CLP-2 Integral Calculus

i
ii
CLP-2 Integral Calculus

Joel Feldman
University of British Columbia

Andrew Rechnitzer
University of British Columbia

Elyse Yeager
University of British Columbia

December 27, 2020

iii
Cover Design: Nick Loewen — licensed under the CC-BY-NC-SA 4.0 License.
Source files: A link to the source files for this document can be found at the CLP
textbook website. The sources are licensed under the CC-BY-NC-SA 4.0 License.

Edition: CLP-2 Integral Calculus: April 2020

Website: CLP-2
© 2016 – 2020 Joel Feldman, Andrew Rechnitzer, Elyse Yeager
This work is licensed under the Creative Commons Attribution-NonCommercial-
ShareAlike 4.0 International License. You can view a copy of the license here.

iv
 P REFACE

This text is a merger of the CLP Integral Calculus textbook and problembook. It is, at
the time that we write this, still a work in progress; some bits and pieces around the
edges still need polish. Consequently we recommend to the student that they still
consult text webpage for links to the errata — especially if they think there might be a
typo or error. We also request that you send us an email at [email protected]
Additionally, if you are not a student at UBC and using these texts please send
us an email (again using the feedback button) — we’d love to hear from you.
Joel Feldman, Andrew Rechnitzer and Elyse Yeager

v
 To our students.

And to the many generations of scholars who have freely shared

all this knowledge with us.

vi
vii
 A CKNOWLEDGEMENTS

Elyse would like to thank her husband Seçkin Demirbaş for his endless patience,
tireless support, and insightful feedback.
Andrew would like to thank Zoe for her patience and support while he hacked
and also Locke for his insightful subediting. He should also thank the various coau-
thors for not being too impatient while other manuscripts were neglected.
Joel would like to thank a range of people for their support and encouragement
over many years.
Thanks to

• Rob Beezer and David Farmer for their help converting this book from LATEX to
this online PreTeXt format.

• Nick Loewen for designing the cover art, help with figures, colours, spelling
and many discussions.

• The many people who have collaborated over the last couple of decades mak-
ing exams and tests for first year calculus courses at UBC Mathematics. A great
many of the exercises in the text come from questions in those tests and exams.

Finally, we’d like to thank those students who reported typos and errors they
found in the text. Many of these students did so through our “bug bounty” program
which was supported by the Department of Mathematics, Skylight and the Loafe
Cafe all at UBC.

viii
 U SING THE EXERCISES IN THIS BOOK

Each problem in this book is split into four parts: Question, Hint, Answer, and So-
lution. As you are working problems, resist the temptation to prematurely peek at
the hint or to click through to the answers and solutions in the appendix! It’s im-
portant to allow yourself to struggle for a time with the material. Even professional
mathematicians don’t always know right away how to solve a problem. The art is in
gathering your thoughts and figuring out a strategy to use what you know to find
out what you don’t.
If you find yourself at a real impasse, go ahead and look at the linked hint. Think
about it for a while, and don’t be afraid to read back in the notes to look for a key
idea that will help you proceed. If you still can’t solve the problem, well, we in-
cluded the Solutions section for a reason! As you’re reading the solutions, try hard
to understand why we took the steps we did, instead of memorizing step-by-step
how to solve that one particular problem.
If you struggled with a question quite a lot, it’s probably a good idea to return
to it in a few days. That might have been enough time for you to internalize the
necessary ideas, and you might find it easily conquerable. Pat yourself on the back
— sometimes math makes you feel good! If you’re still having troubles, read over
the solution again, with an emphasis on understanding why each step makes sense.
One of the reasons so many students are required to study calculus is the hope
that it will improve their problem-solving skills. In this class, you will learn lots of
concepts, and be asked to apply them in a variety of situations. Often, this will in-
volve answering one really big problem by breaking it up into manageable chunks,
solving those chunks, then putting the pieces back together. When you see a partic-
ularly long question, remain calm and look for a way to break it into pieces you can
handle.

• Working with Friends

Study buddies are fantastic! If you don’t already have friends in your class,
you can ask your neighbours in lecture to form a group. Often, a question
that you might bang your head against for an hour can be easily cleared up
by a friend who sees what you’ve missed. Regular study times make sure you

ix
 0.0

don’t procrastinate too much, and friends help you maintain a positive attitude
when you might otherwise succumb to frustration. Struggle in mathematics is
desirable, but suffering is not.
When working in a group, make sure you try out problems on your own be-
fore coming together to discuss with others. Learning is a process, and getting
answers to questions that you haven’t considered on your own can rob you of
the practice you need to master skills and concepts, and the tenacity you need
to develop to become a competent problem-solver.

• Types of Questions
Questions outlined by a blue box make up the representative question set. This set
of questions is intended to cover the most essential ideas in each section. These
questions are usually highly typical of what you’d see on an exam, although
some of them are atypical but carry an important moral. If you find yourself
unconfident with the idea behind one of these, it’s probably a good idea to
practice similar questions.
This representative question set is our suggestion for a minimal selection of
questions to work on. You are highly encouraged to work on more.
In addition to original problems, this book contains problems pulled from
quizzes and exams given at UBC for Math 101 (first-semester calculus) and
Math 121 (honours first-semester calculus). These problems are marked by
“(*)”. The authors would like to acknowledge the contributions of the many
people who collaborated to produce these exams over the years.
Finally, the questions are organized into three types: Stage 1, Stage 2 and
Stage 3.

◦ Exercises — Stage 1
The first category is meant to test and improve your understanding of
basic underlying concepts. These often do not involve much calculation.
They range in difficulty from very basic reviews of definitions to ques-
tions that require you to be thoughtful about the concepts covered in the
section.
◦ Exercises — Stage 2
Questions in this category are for practicing skills. It’s not enough to un-
derstand the philosophical grounding of an idea: you have to be able to
apply it in appropriate situations. This takes practice!
◦ Exercises — Stage 3
The last questions in each section go a little farther than “Stage 2”. Often
they will combine more than one idea, incorporate review material, or ask
you to apply your understanding of a concept to a new situation.

In exams, as in life, you will encounter questions of varying difficulty. A good

skill to practice is recognizing the level of difficulty a problem poses. Exams

x
 0.0

will have some easy question, some standard questions, and some harder ques-
tions.

xi
 F EEDBACK ABOUT THE TEXT

The CLP-2 Integral Calculus text is still undergoing testing and changes. Because of
this we request that if you find a problem or error in the text then:

1. Please check the errata list that can be found at the text webpage.

2. Is the problem in the online version or the PDF version or both?

3. Note the URL of the online version and the page number in the PDF

4. Send an email to [email protected]. Please be sure to include

• a description of the error

• the URL of the page, if found in the online edition
• and if the problem also exists in the PDF, then the page number in the PDF
and the compile date on the front page of PDF.

xii
 C ONTENTS

Preface v

Acknowledgements viii

Using the exercises in this book ix

Feedback about the text xii

1 Integration 1
1.1 Definition of the Integral . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Basic properties of the definite integral . . . . . . . . . . . . . . . . . . . 42
1.3 The Fundamental Theorem of Calculus . . . . . . . . . . . . . . . . . . 60
1.4 Substitution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
1.5 Area between curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
1.6 Volumes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
1.7 Integration by parts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
1.8 Trigonometric Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
1.9 Trigonometric Substitution . . . . . . . . . . . . . . . . . . . . . . . . . . 171
1.10 Partial Fractions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188
1.11 Numerical Integration . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
1.12 Improper Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261
1.13 More Integration Examples . . . . . . . . . . . . . . . . . . . . . . . . . 284

2 Applications of Integration 290

2.1 Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 290
2.2 Averages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 305
2.3 Centre of Mass and Torque . . . . . . . . . . . . . . . . . . . . . . . . . . 319
2.4 Separable Differential Equations . . . . . . . . . . . . . . . . . . . . . . 345

xiii
C ONTENTS

3 Sequence and series 380

3.1 Sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 381
3.2 Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 393
3.3 Convergence Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 410
3.4 Absolute and Conditional Convergence . . . . . . . . . . . . . . . . . . 459
3.5 Power Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 467
3.6 Taylor Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 488
3.7 Optional — Rational and irrational numbers . . . . . . . . . . . . . . . 527

A High School Material 537

A.1 Similar Triangles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 537
A.2 Pythagoras . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 538
A.3 Trigonometry — Definitions . . . . . . . . . . . . . . . . . . . . . . . . . 538
A.4 Radians, Arcs and Sectors . . . . . . . . . . . . . . . . . . . . . . . . . . 539
A.5 Trigonometry — Graphs . . . . . . . . . . . . . . . . . . . . . . . . . . . 539
A.6 Trigonometry — Special Triangles . . . . . . . . . . . . . . . . . . . . . 540
A.7 Trigonometry — Simple Identities . . . . . . . . . . . . . . . . . . . . . 540
A.8 Trigonometry — Add and Subtract Angles . . . . . . . . . . . . . . . . 541
A.9 Inverse Trigonometric Functions . . . . . . . . . . . . . . . . . . . . . . 541
A.10 Areas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 542
A.11 Volumes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 543
A.12 Powers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 544
A.13 Logarithms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 544
A.14 Highschool Material You Should be Able to Derive . . . . . . . . . . . . 545
A.15 Cartesian Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . 546
A.16 Roots of Polynomials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 548

B Hints for Exercises 555

C Answers to Exercises 604

D Solutions to Exercises 672

xiv
 Chapter 1

I NTEGRATION

Calculus is built on two operations — differentiation and integration.

• Differentiation — as we saw last term, differentiation allows us to compute
and study the instantaneous rate of change of quantities. At its most basic it
allows us to compute tangent lines and velocities, but it also led us to quite so-
phisticated applications including approximation of functions through Taylor
polynomials and optimisation of quantities by studying critical and singular
points.
• Integration — at its most basic, allows us to analyse the area under a curve. Of
course, its application and importance extend far beyond areas and it plays a
central role in solving differential equations.

It is not immediately obvious that these two topics are related to each other. How-
ever, as we shall see, they are indeed intimately linked.

1.1q Definition of the Integral

Arguably the easiest way to introduce integration is by considering the area between
the graph of a given function and the x-axis, between two specific vertical lines —

1
 I NTEGRATION 1.1 D EFINITION OF THE I NTEGRAL

such as is shown in the figure above. We’ll follow this route by starting with a moti-
vating example.

1.1.1 tt A Motivating Example

Let us find thearea under the curve y = ex (and above the x-axis) for 0 ≤ x ≤ 1. That
is, the area of (x, y) 0 ≤ y ≤ ex , 0 ≤ x ≤ 1 .

This area is equal to the “definite integral”

Z 1
Area = ex dx
0

Do not worry about this notation or terminology just yet. We discuss it at length
below. In different applications this quantity will have different interpretations —
not just area. For example, if x is time and ex is your velocity at time x, then we’ll see
later (in Example 1.1.18) that the specified area is the net distance travelled between
time 0 and time 1. After we finish with the example, we’ll mimic it to give a general
Rb
definition of the integral a f (x)dx.

Example 1.1.1 Computing an area with vertical strips.


We wish to compute the area of (x, y) 0 ≤ y ≤ ex , 0 ≤ x ≤ 1 . We know, from
our experience with ex in differential calculus, that the curve y = ex is not easily
written in terms of other simpler functions, so it is very unlikely that we would be
able to write the area as a combination of simpler geometric objects such as triangles,
rectangles or circles.
So rather than trying to write down the area exactly, our strategy is to approximate
the area and then make our approximation more and more precise a . We choose b to
approximate the area as a union of a large number of tall thin (vertical) rectangles. As
we take more and more rectangles we get better and better approximations. Taking
the limit as the number of rectangles goes to infinity gives the exact area c .
As a warm up exercise, we’ll now just use four rectangles. In Example 1.1.2, be-
low, we’ll consider an arbitrary number of rectangles and then take the limit as the
number of rectangles goes to infinity. So

2
I NTEGRATION 1.1 D EFINITION OF THE I NTEGRAL

• subdivide the interval 0 ≤ x ≤ 1 into 4 equal subintervals each of width 14 , and

• subdivide the area of interest into four corresponding vertical strips, as in the
figure below.
The area we want is exactly the sum of the areas of all four strips.

Each of these strips is almost, but not quite, a rectangle. While the bottom and sides
are fine (the sides are at right-angles to the base), the top of the strip is not horizontal.
This is where we must start to approximate. We can replace each strip by a rectangle
by just levelling off the top. But now we have to make a choice — at what height do
we level off the top?
Consider, for example, the leftmost strip. On this strip, x runs from 0 to 14 . As x runs
1
from 0 to 14 , the height y runs from e0 to e 4 . It would be reasonable to choose the
1
height of the approximating rectangle to be somewhere between e0 and e 4 . Which

height should we choose? Well, actually it doesn’t matter. When we eventually take
the limit of infinitely many approximating rectangles all of those different choices
give exactly the same final answer. We’ll say more about this later.
In this example we’ll do two sample computations.
• For the first computation we approximate each slice by a rectangle whose
height is the height of the left hand side of the slice.

◦ On the first slice, x runs from 0 to 41 , and the height y runs from e0 , on the
1
left hand side, to e 4 , on the right hand side.

3
I NTEGRATION 1.1 D EFINITION OF THE I NTEGRAL

◦ So we approximate the first slice by the rectangle of height e0 and width

1
4
, and hence of area 14 e0 = 41 .
1 1
◦ On the second slice, x runs from 4
to 12 , and the height y runs from e 4 and
1
e2 .
1
◦ So we approximate the second slice by the rectangle of height e 4 and
1
width 14 , and hence of area 41 e 4 .
◦ And so on.
◦ All together, we approximate the area of interest by the sum of the areas
of the four approximating rectangles, which is
 1 1 3 1
1 + e4 + e2 + e4 = 1.5124
4
◦ This particular approximation is called the “left Riemann sum approxima-
R1
tion to 0 ex dx with 4 subintervals”. We’ll explain this terminology later.
◦ This particular approximation represents the shaded area in the figure on
the left below. Note that, because ex increases as x increases, this approxi-
mation is definitely smaller than the true area.

• For the second computation we approximate each slice by a rectangle whose

height is the height of the right hand side of the slice.

◦ On the first slice, x runs from 0 to 41 , and the height y runs from e0 , on the
1
left hand side, to e 4 , on the right hand side.
1
◦ So we approximate the first slice by the rectangle of height e 4 and width
1 1
4
, and hence of area 14 e 4 .
1 1
◦ On the second slice, x runs from 4
to 12 , and the height y runs from e 4 and
1
e2 .

4
I NTEGRATION 1.1 D EFINITION OF THE I NTEGRAL

1
◦ So we approximate the second slice by the rectangle of height e 2 and
1
width 14 , and hence of area 41 e 2 .
◦ And so on.
◦ All together, we approximate the area of interest by the sum of the areas
of the four approximating rectangles, which is

1 1 3 1
e 4 + e 2 + e 4 + e1

= 1.9420
4

◦ This particular approximation is called the “right Riemann sum approxi-

R1
mation to 0 ex dx with 4 subintervals”.
◦ This particular approximation represents the shaded area in the figure on
the right above. Note that, because ex increases as x increases, this approx-
imation is definitely larger than the true area.

a This should remind the reader of the approach taken to compute the slope of a tangent line way
way back at the start of differential calculus.
b Approximating the area in this way leads to a definition of integration that is called Riemann
integration. This is the most commonly used approach to integration. However we could also
approximate the area by using long thin horizontal strips. This leads to a definition of integration
that is called Lebesgue integration. We will not be covering Lebesgue integration in these notes.
c If we want to be more careful here, we should construct two approximations, one that is always
a little smaller than the desired area and one that is a little larger. We can then take a limit using
the Squeeze Theorem and arrive at the exact area. More on this later.

Example 1.1.1

Now for the full computation that gives the exact area.

Example 1.1.2 Computing an area exactly.

Recall that we wish to compute the area of

(x, y) 0 ≤ y ≤ ex , 0 ≤ x ≤ 1


and that our strategy is to approximate this area by the area of a union of a large
number of very thin rectangles, and then take the limit as the number of rectangles
goes to infinity. In Example 1.1.1, we used just four rectangles. Now we’ll consider a
general number of rectangles, that we’ll call n. Then we’ll take the limit n → ∞. So
• pick a natural number n and

• subdivide the interval 0 ≤ x ≤ 1 into n equal subintervals each of width n1 , and

• subdivide the area of interest into corresponding thin strips, as in the figure
below.

5
I NTEGRATION 1.1 D EFINITION OF THE I NTEGRAL

The area we want is exactly the sum of the areas of all of the thin strips.

Each of these strips is almost, but not quite, a rectangle. As in Example 1.1.1, the
only problem is that the top is not horizontal. So we approximate each strip by a
rectangle, just by levelling off the top. Again, we have to make a choice — at what
height do we level off the top?
Consider, for example, the leftmost strip. On this strip, x runs from 0 to n1 . As x runs
1
from 0 to n1 , the height y runs from e0 to e n . It would be reasonable to choose the
1
height of the approximating rectangle to be somewhere between e0 and e n . Which
height should we choose?
Well, as we said in Example 1.1.1, it doesn’t matter. We shall shortly take the limit
n → ∞ and, in that limit, all of those different choices give exactly the same final an-
swer. We won’t justify that statement in this example, but there will be an (optional)
section shortly that provides the justification. For this example we just, arbitrarily,
choose the height of each rectangle to be the height of the graph y = ex at the small-
est value of x in the corresponding strip a . The figure on the left below shows the
approximating rectangles when n = 4 and the figure on the right shows the approx-
imating rectangles when n = 8.

Now we compute the approximating area when there are n strips.

• We approximate the leftmost strip by a rectangle of height e0 . All of the rectan-

6
I NTEGRATION 1.1 D EFINITION OF THE I NTEGRAL

gles have width n1 . So the leftmost rectangle has area n1 e0 .

• On strip number 2, x runs from n1 to n2 . So the smallest value of x on strip

1
number 2 is n1 , and we approximate strip number 2 by a rectangle of height e n
1
and hence of area n1 e n .

• And so on.
n−1 n
• On the last strip, x runs from n
to n
= 1. So the smallest value of x on the last
(n−1)
n−1
strip is n
, and we approximate the last strip by a rectangle of height e n
(n−1)
and hence of area n1 e n .
The total area of all of the approximating rectangles is
1 0 1 1 1 2 1 3 1 (n−1)
Total approximating area = e + en + en + en + · · · + e n
n n n n n
1 1 2 3 (n−1)

= 1 + e + e + e + ··· + e
n n n n
n
Now the sum in the brackets might look a little intimidating because of all the ex-
ponentials, but it actually has a pretty simple structure that can be easily seen if we
1
rename e n = r. Then
• the first term is 1 = r0 and
1
• the second term is e n = r1 and
2
• the third term is e n = r2 and
3
• the fourth term is e n = r3 and

• and so on and
(n−1)
• the last term is e n = rn−1 .
So
1
1 + r + r2 + · · · + rn−1


Total approximating area =
n
The sum in brackets is known as a geometric sum and satisfies a nice simple formula:

Equation 1.1.3 Geometric sum.

rn − 1
1 + r + r2 + · · · + rn−1 = provided r 6= 1
r−1

7
I NTEGRATION 1.1 D EFINITION OF THE I NTEGRAL

The derivation of the above formula is not too difficult. So let’s derive it in a little
aside.
Geometric sum.
Denote the sum as

S = 1 + r + r2 + · · · + rn−1

Notice that if we multiply the whole sum by r we get back almost the same
thing:

rS
= r 1 + r + · · · + rn−1

= r + r2 + r3 + · · · + rn

This right hand side differs from the original sum S only in that

• the right hand side, which starts with “r+ ”, is missing the “1+ ” that S
starts with, and

• the right hand side has an extra “+rn ” at the end that does not appear in
S.

That is

rS = S − 1 + rn

Moving this around a little gives

(r − 1)S = (rn − 1)
rn − 1
S=
r−1
as required. Notice that the last step in the manipulations only works provid-
ing r 6= 1 (otherwise we are dividing by zero).

Now we can go back to our area approximation armed with the above result about
geometric sums.
1
1 + r + r2 + · · · + rn−1


Total approximating area =
n
1 rn − 1
= remember that r = e1/n
n r−1
1 en/n − 1
=
n e1/n − 1
1 e−1
=
n e1/n − 1

8
I NTEGRATION 1.1 D EFINITION OF THE I NTEGRAL

To get the exact area b all we need to do is make the approximation better and better
by taking the limit n → ∞. The limit will look more familiar if we rename n1 to X.
As n tends to infinity, X tends to 0, so
1 e−1
Area = lim
n→∞ n e1/n − 1
1/n
= (e − 1) lim 1/n
n→∞ e −1
X 1
= (e − 1) lim X (with X = )
X→0 e − 1 n
Examining this limit we see that both numerator and denominator tend to zero as
X → 0, and so we cannot evaluate this limit by computing the limits of the numera-
tor and denominator separately and then dividing the results. Despite this, the limit
is not too hard to evaluate; here we give two ways:
• Perhaps the easiest way to compute the limit is by using l’Hôpital’s rule c . Since
both numerator and denominator go to zero, this is a 00 indeterminate form.
Thus
d
X dX
X 1
lim X
= lim d
= lim =1
X→0 e − 1 X→0 (eX − 1) X→0 eX
dX

• Another way d to evaluate the same limit is to observe that it can be massaged
into the form of the limit definition of the derivative. First notice that
−1
eX − 1

X
lim = lim
X→0 eX − 1 X→0 X
provided this second limit exists and is nonzero e . This second limit should
look a little familiar:
eX − 1 eX − e0
lim = lim
X→0 X X→0 X − 0

which is just the definition of the derivative of ex at x = 0. Hence we have

−1
eX − e0

X
lim = lim
X→0 eX − 1 X→0 X − 0
 −1
d X
= e
dX X=0
 X −1
= e X=0
=1

So, after this short aside into limits, we may now conclude that
X
Area = (e − 1) lim
X→0 eX
−1

9
 I NTEGRATION 1.1 D EFINITION OF THE I NTEGRAL

=e−1

a Notice that since ex is an increasing function, this choice of heights means that each of our rect-
angles is smaller than the strip it came from.
b We haven’t proved that this will give us the exact area, but it should be clear that taking this limit
will give us a lower bound on the area. To complete things rigorously we also need an upper
bound and the squeeze theorem. We do this in the next optional subsection.
c If you do not recall L’Hôpital’s rule and indeterminate forms then we recommend you skim over
your differential calculus notes on the topic.
d Say if you don’t recall l’Hôpital’s rule and have not had time to revise it.
e To hyphenate or not to hypenate: “non-zero” or “nonzero”? The authors took our lead from here
and also here.

Example 1.1.3

1.1.2 tt Optional — A more rigorous area computation


In Example 1.1.1 above we considered the area of the region (x, y) 0 ≤ y ≤ ex ,
0 ≤ x ≤ 1 . We approximated that area by the area of a union of n thin rectangles.
We then claimed that upon taking the number of rectangles to infinity, the approxi-
mation of the area became the exact area. However we did not justify the claim. The
purpose of this optional section is to make that calculation rigorous.
The broad set-up is the same. We divide the region up into n vertical strips, each
of width n1 and we then approximate those strips by rectangles. However rather than
an uncontrolled approximation, we construct two sets of rectangles — one set always
smaller than the original area and one always larger. This then gives us lower and
upper bounds on the area of the region. Finally we make use of the squeeze theorem
1
to establish the result.

• To find our upper and lower bounds we make use of the fact that ex is an
d x
increasing function. We know this because the derivative dx e = ex is always
positive. Consequently, the smallest and largest values of ex on the interval
a ≤ x ≤ b are ea and eb , respectively.
1
• In particular, for 0 ≤ x ≤ n1 , ex takes values only between e0 and e n . As a result,
the first strip
1
(x, y) 0 ≤ x ≤ , 0 ≤ y ≤ ex

n
1
◦ contains the rectangle of 0 ≤ x ≤ n , 0 ≤ y ≤ e0 (the lighter rectangle in the
figure on the left below) and

1 Recall that if we have 3 functions f (x), g(x), h(x) that satisfy f (x) ≤ g(x) ≤ h(x) and we know
that limx→a f (x) = limx→a h(x) = L exists and is finite, then the squeeze theorem tells us that
limx→a g(x) = L.

10
I NTEGRATION 1.1 D EFINITION OF THE I NTEGRAL

1
◦ is contained in the rectangle 0 ≤ x ≤ n1 , 0 ≤ y ≤ e n (the largest rectangle
in the figure on the left below).

Hence
1 0 1 1 1
e ≤ Area (x, y) 0 ≤ x ≤ , 0 ≤ y ≤ ex

≤ en
n n n

• Similarly, for the second, third, . . . , last strips, as in the figure on the right
above,
1 1 1 2 1 2
, 0 ≤ y ≤ ex

e n ≤ Area (x, y) ≤x≤ ≤ en
n n n n
1 2 2 3 1 3
, 0 ≤ y ≤ ex

e n ≤ Area (x, y) ≤x≤ ≤ en
n n n n
..
.
1 (n−1) (n − 1) n 1 n
≤ x ≤ , 0 ≤ y ≤ ex

e n ≤ Area (x, y) ≤ en
n n n n

• Adding these n inequalities together gives

1 1 (n−1)

1 + en + · · · + e n
n
≤ Area (x, y) 0 ≤ x ≤ 1, 0 ≤ y ≤ ex


1 1 2 n

≤ en + en + · · · + en
n

1
 1 n
• We can then recycle equation 1.1.3 with r = e n , so that rn = e n = e. Thus
we have
1 e−1 1 1 e−1
≤ Area (x, y) 0 ≤ x ≤ 1, 0 ≤ y ≤ ex

1 ≤ en
n en − 1 n e n1 − 1

where we have used the fact that the upper bound is a simple multiple of the
lower bound:
 1 2 n
 1
 1 (n−1)

en + en + · · · + en = en 1 + en + · · · + e n .

11
 I NTEGRATION 1.1 D EFINITION OF THE I NTEGRAL

• We now apply the squeeze theorem to the above inequalities. In particular, the
limits of the lower and upper bounds are

1 e−1 X
lim 1 = (e − 1) lim X
=e−1
n→∞ n e n − 1 X= n →0 e − 1
1

(by l’Hôpital’s rule) and

1 1 e−1 XeX
lim en 1 = (e − 1) lim · X
n→∞ n en − 1 1
X= n →0 e − 1

X
= (e − 1) lim eX · lim
X→0 X=→0 eX − 1
= (e − 1) · 1 · 1

Thus, since the exact area is trapped between the lower and upper bounds, the
squeeze theorem then implies that

Exact area = e − 1.

1.1.3 tt Summation notation

As you can see from the above example (and the more careful rigorous computation),
our discussion of integration will involve a fair bit of work with sums of quantities.
To this end, we make a quick aside into summation notation. While one can work
through the material below without this notation, proper summation notation is well
worth learning, so we advise the reader to persevere.
Writing out the summands explicitly can become quite impractical — for exam-
ple, say we need the sum of the first 11 squares:

1 + 22 + 32 + 42 + 52 + 62 + 72 + 82 + 92 + 102 + 112

This becomes tedious. Where the pattern is clear, we will often skip the middle few
terms and instead write

1 + 22 + · · · + 112 .

A far more precise way to write this is using Σ (capital-sigma) notation. For example,
we can write the above sum as
11
X
k2
k=1

This is read as

The sum from k equals 1 to 11 of k 2 .

More generally

12
I NTEGRATION 1.1 D EFINITION OF THE I NTEGRAL

Definition 1.1.4
Let m ≤ n be integers and let f (x) be a function defined on the integers. Then
we write
n
X
f (k)
k=m

to mean the sum of f (k) for k from m to n:

f (m) + f (m + 1) + f (m + 2) + · · · + f (n − 1) + f (n).

Similarly we write
n
X
ai
i=m

to mean

am + am+1 + am+2 + · · · + an−1 + an

for some set of coefficients {am , . . . , an }.

Consider the example

7
X 1 1 1 1 1 1
= + + + +
k=3
k2 32 42 52 62 72
It is important to note that the right hand side of this expression evaluates to a num-
ber 2 ; it does not contain “k”. The summation index k is just a “dummy” variable
and it does not have to be called k. For example
7 7 7 7
X 1 X 1 X 1 X 1
2
= 2
= 2
=
k=3
k i=3
i j=3
j `=3
`2
Also the summation index has no meaning outside the sum. For example
7
X 1
k
k=3
k2
has no mathematical meaning; it is gibberish.
A sum can be represented using summation notation in many different ways. If
you are unsure as to whether or not two summation notations represent the same
sum, just write out the first few terms and the last couple of terms. For example,
m=3 m=4 m=5 m=14 m=15
15
z}|{ z}|{ z}|{ z}|{ z}|{
X 1 1 1 1 1 1
2
= 2 + 2 + 2 +··· + 2 + 2
m=3
m 3 4 5 14 15

46181
2 Some careful addition shows it is 176400 .

13
I NTEGRATION 1.1 D EFINITION OF THE I NTEGRAL

m=4 m=5 m=6 m=15 m=16

16
z}|{ z}|{ z}|{ z}|{ z}|{
X 1 1 1 1 1 1
2
= 2 + 2 + 2 +··· + 2 + 2
m=4
(m − 1) 3 4 5 14 15
are equal.
Here is a theorem that gives a few rules for manipulating summation notation.

Theorem 1.1.5 Arithmetic of Summation Notation.

Let n ≥ m be integers. Then for all real numbers c and ai , bi , m ≤ i ≤ n.
n
 n 
P P
a cai = c ai
i=m i=m

n
 n
  n

P P P
b (ai + bi ) = ai + bi
i=m i=m i=m

n
 n
  n

P P P
c (ai − bi ) = ai − bi
i=m i=m i=m

Proof. We can prove this theorem by just writing out both sides of each
equation, and observing that they are equal, by the usual laws of arithmetic a .
For example, for the first equation, the left and right hand sides are
n
X
cai = cam + cam+1 + · · · + can
i=m
n
X 
and c ai = c(am + am+1 + · · · + an )
i=m

They are equal by the usual distributive law. The “distributive law” is the fancy
name for c(a + b) = ca + cb.

a Since all the sums are finite, this isn’t too hard. More care must be taken when the sums
involve an infinite number of terms. We will examine this in Chapter 3.

Not many sums can be computed exactly 3 . Here are some that can. The first few

3 Of course, any finite sum can be computed exactly — just sum together the terms. What we
mean by “computed exactly” in this context, is that we can rewrite the sum as a simple, and
easily evaluated, formula involving the terminals of the sum. For example
n
X rn+1 − rm
rk = provided r 6= 1
r−1
k=m

14
 I NTEGRATION 1.1 D EFINITION OF THE I NTEGRAL

are used a lot.

Theorem 1.1.6

n n+1
ari = a 1−r
P
a 1−r
, for all real numbers a and r 6= 1 and all integers n ≥ 0.
i=0

n
P
b 1 = n, for all integers n ≥ 1.
i=1

n
i = 12 n(n + 1), for all integers n ≥ 1.
P
c
i=1

n
i2 = 16 n(n + 1)(2n + 1), for all integers n ≥ 1.
P
d
i=1

n h i2
1
i3 =
P
e 2
n(n + 1) , for all integers n ≥ 1.
i=1

ttt Proof of Theorem 1.1.6 (Optional)

1.1.3.1

Proof.
a The first sum is
n
X
ari = ar0 + ar1 + ar2 + · · · + arn
i=0

which is just the left hand side of equation 1.1.3, with n replaced by n + 1
and then multiplied by a.
b The second sum is just n copies of 1 added together, so of course the sum
is n.
c The third and fourth sums are discussed in the appendix of the CLP-1
text. In that discussion certain “tricks” are used to compute the sums

No matter what finite integers we choose for m and n, we can quickly compute the sum in just a
few arithmetic operations. On the other hand, the sums,
n n
X 1 X 1
k k2
k=m k=m

cannot be expressed in such clean formulas (though you can rewrite them quite cleanly using in-
tegrals). To explain more clearly we would need to go into a more detailed and careful discussion
that is beyond the scope of this course.

15
I NTEGRATION 1.1 D EFINITION OF THE I NTEGRAL

with only simple arithmetic. Those tricks do not easily generalise to the
fifth sum.

d Instead of repeating that appendix, we’ll derive the third sum using a
trick that generalises to the fourth and fifth sums (and also to higher pow-
ers). The trick uses the generating function a S(x):

Equation 1.1.7 Finite geometric sum.

2 n xn+1 − 1
S(x) = 1 + x + x + · · · + x =
x−1

Notice that this is just the geometric sum given by equation 1.1.3 with n
replaced by n + 1.
Now, consider the limit

lim S(x) = lim 1 + x + x2 + · · · + xn = n + 1

but also
x→1 x→1
xn+1 − 1
= lim now use l’Hôpital’s rule
x→1 x − 1
(n + 1)xn
= lim = n + 1.
x→1 1
This is not so hard (or useful). But now consider the derivative of S(x):

S 0 (x) = 1 + 2x + 3x2 + · · · + nxn−1

d xn+1 − 1
 
= use the quotient rule
dx x−1
(x − 1) · (n + 1)xn − (xn+1 − 1) · 1
= now clean it up
(x − 1)2
nxn+1 − (n + 1)xn + 1
= .
(x − 1)2
Hence if we take the limit of the above expression as x → 1 we recover

lim S 0 (x) = 1 + 2 + 3 + · · · + n
x→1
nxn+1 − (n + 1)xn + 1
= lim now use l’Hôpital’s rule
x→1 (x − 1)2
n(n + 1)xn − n(n + 1)xn−1
= lim l’Hôpital’s rule again
x→1 2(x − 1)
n2 (n + 1)xn−1 − n(n + 1)(n − 1)xn−2
= lim
x→1 2

16
 I NTEGRATION 1.1 D EFINITION OF THE I NTEGRAL

n2 (n + 1) − n(n − 1)(n + 1) n(n + 1)

= =
2 2
as required. This computation can be done without l’Hôpital’s rule, but
the manipulations required are a fair bit messier.

e The derivation of the fourth and fifth sums is similar to, but even more
tedious than, that of the third sum. One takes two or three derivatives of
the generating functional.

a Generating functions are frequently used in mathematics to analyse sequences and series,
but are beyond the scope of the course. The interested reader should take a look at “Gen-
eratingfunctionology” by Herb Wilf. It is an excellent book and is also free to download.

1.1.4 tt The Definition of the Definite Integral

Z b
In this section we give a definition of the definite integral f (x)dx generalising the
a
machinery we used in Example 1.1.1. But first some terminology and a couple of
remarks to better motivate the definition.
Definition 1.1.8
Z b
The symbol f (x)dx is read “the definite integral of the function f (x) from
a Rb
a to b”. The function f (x) is called the integrand of a f (x)dx and a and b are
called a the limits of integration. The interval a ≤ x ≤ b is called the interval of
integration and is also called the domain of integration.

a a and b are also called the bounds of integration.

Before we explain more precisely what the definite integral actually is, a few
remarks (actually — a few interpretations) are in order.
Z b
• If f (x) ≥ 0 and a ≤ b, one interpretation of the symbol f (x)dx is “the area
 a
of the region (x, y) a ≤ x ≤ b, 0 ≤ y ≤ f (x) ”.

17
I NTEGRATION 1.1 D EFINITION OF THE I NTEGRAL

In this way we can rewrite the area in Example 1.1.1 as the definite integral
R1 x
0
e dx.

• This interpretation breaks down when either a > b or f (x) is not always posi-
tive, but it can be repaired by considering “signed areas”.
Rb
• If a ≤ b, but f (x) is not always positive, one interpretation of a f (x)dx is “the
signed area between y = f (x) and the x-axis for a ≤ x ≤ b”. For “signed area”
(which is also called the “net area”), areas above the x-axis count as positive
while areas below the x-axis count as negative. In the example below, we have
the graph of the function

−1 if 1 ≤ x ≤ 2


f (x) = 2 if 2 < x ≤ 4

0

otherwise

The 2 × 2 shaded square above the x-axis has signed area +2 × 2 = +4. The
1 × 1 shaded square below the x-axis has signed area −1 × 1 = −1. So, for this
f (x),
Z 5
f (x)dx = +4 − 1 = 3
0

• We’ll come back to the case b < a later.

18
I NTEGRATION 1.1 D EFINITION OF THE I NTEGRAL
Rb
We’re now ready to define a f (x)dx. The definition is a little involved, but essen-
tially mimics what we did in Example 1.1.1 (which is why we did the example before
the definition). The main differences are that we replace the function ex by a generic
function f (x) and we replace the interval from 0 to 1 by the generic interval 4 from a
to b.

• We start by selecting any natural number n and subdividing the interval from
a to b into n equal subintervals. Each subinterval has width b−a
n
.

• Just as was the case in Example 1.1.1 we will eventually take the limit as n →
∞, which squeezes the width of each subinterval down to zero.

• For each integer 0 ≤ i ≤ n, define xi = a + i · b−a

n
. Note that this means that
x0 = a and xn = b. It is worth keeping in mind that these numbers xi do depend
on n even though our choice of notation hides this dependence.

• Subinterval number i is xi−1 ≤ x ≤ xi . In particular, on the first subinterval, x

runs from x0 = a to x1 = a + b−a
n
. On the second subinterval, x runs from x1 to
b−a
x2 = a + 2 n .

• On each subinterval we now pick x∗i,n between xi−1 and xi . We then approx-
imate f (x) on the ith subinterval by the constant function y = f (x∗i,n ). We in-
clude n in the subscript to remind ourselves that these numbers depend on
n.
Geometrically, we’re approximating the region

(x, y) x is between xi−1 and xi , and y is between 0 and f (x)

by the rectangle

(x, y) x is between xi−1 and xi , and y is between 0 and f (x∗i,n )



4 We’ll eventually allow a and b to be any two real numbers, not even requiring a < b. But it is
easier to start off assuming a < b, and that’s what we’ll do.

19
I NTEGRATION 1.1 D EFINITION OF THE I NTEGRAL

In Example 1.1.1 we chose x∗i,n = xi−1 and so we approximated the function ex

on each subinterval by the value it took at the leftmost point in that subinterval.
• So, when there are n subintervals our approximation to the signed area be-
tween the curve y = f (x) and the x-axis, with x running from a to b, is
n
X b−a
f (x∗i,n ) ·
i=1
n

We interpret this as the signed area since the summands f (x∗i,n ) · b−a
n
need not
be positive.
• Finally we define the definite integral by taking the limit of this sum as n → ∞.
Oof! This is quite an involved process, but we can now write down the definition
we need.
Definition 1.1.9
Let a and b be two real numbers and let f (x) be a function that is defined for
all x between a and b. Then we define
Z b n
X b−a
f (x)dx = lim f (x∗i,n ) ·
a n→∞
i=1
n

when the limit exists and takes the same value for all choices of the x∗i,n ’s. In
this case, we say that f is integrable on the interval from a to b.

Of course, it is not immediately obvious when this limit should exist. Thankfully
it is easier for a function to be “integrable” than it is for it to be “differentiable”.

Theorem 1.1.10
Let f (x) be a function on the interval [a, b]. If

• f (x) is continuous on [a, b], or

• f (x) has a finite number of jump discontinuities on [a, b] (and is otherwise

20
I NTEGRATION 1.1 D EFINITION OF THE I NTEGRAL

continuous)

then f (x) is integrable on [a, b].

We will not justify this theorem. But a slightly weaker statement is proved in (the
optional) Section 1.1.7. Of course this does not tell us how to actually evaluate any
definite integrals — but we will get to that in time.
Some comments:

• Note that, in Definition 1.1.9, we allow a and b to be any two real numbers. We
Rb
do not require that a < b. That is, even when a > b, the symbol a f (x)dx is
still defined by the formula of Definition 1.1.9. We’ll get an interpretation for
Rb
a
f (x)dx, when a > b, later.
Rb
• It is important to note that the definite integral a f (x)dx represents a number,
not a function of x. The integration Pvariable x is another “dummy” variable,
n
just like the summation index i in i=m ai (see Section 1.1.3). The integration
variable does not have to be called x. For example
Z b Z b Z b
f (x)dx = f (t)dt = f (u)du
a a a

Just as with summation variables, the integration variable x has no meaning

outside of f (x)dx. For example
Z 1 Z x
x
x e dx and ex dx
0 0

are both gibberish.

The sum inside definition 1.1.9 is named after Bernhard Riemann 5 who made
the first rigorous definition of the definite integral and so placed integral calculus on
rigorous footings.
Definition 1.1.11
The sum inside definition 1.1.9
n
X b−a
f (x∗i,n )
i=1
n

5 Bernhard Riemann was a 19th century German mathematician who made extremely important
contributions to many different areas of mathematics — far too many to list here. Arguably two
of the most important (after Riemann sums) are now called Riemann surfaces and the Riemann
hypothesis (he didn’t name them after himself).

21
I NTEGRATION 1.1 D EFINITION OF THE I NTEGRAL

is called a Riemann sum. It is also often written as

n
X
f (x∗i ) ∆x
i=1

b−a
where ∆x = n
.

• If we choose each x∗i,n = xi−1 = a + (i − 1) b−a

n
to be the left hand end point
th
of the i interval, [xi−1 , xi ], we get the approximation
n  
X b−a b−a
f a + (i − 1)
i=1
n n
Rb
which is called the “left Riemann sum approximation to a f (x)dx with
n subintervals”. This is the approximation used in Example 1.1.1.

• In the same way, if we choose x∗i,n = xi = a + i b−a

n
we obtain the approxi-
mation
n  
X b−a b−a
f a+i
i=1
n n
Rb
which is called the “right Riemann sum approximation to a f (x)dx with
n subintervals”. The word “right” signifies that, on each subinterval
[xi−1 , xi ] we approximate f by its value at the right-hand end-point,
xi = a + i b−an
, of the subinterval.

• A third commonly used approximation is

n  
X 1 b−a b−a
f a + (i − )
i=1
2 n n
Rb
which is called the “midpoint Riemann sum approximation to a f (x)dx
with n subintervals”. The word “midpoint” signifies that, on each subin-
terval [xi−1 , xi ] we approximate f by its value at the midpoint, xi−12+xi =
a + (i − 12 ) b−a
n
, of the subinterval.

In order to compute a definite integral using Riemann sums we need to be able

to compute the limit of the sum as the number of summands goes to infinity. This
approach is not always feasible and we will soon arrive at other means of computing
definite integrals based on antiderivatives. However, Riemann sums also provide us
with a good means of approximating definite integrals — if we take n to be a large,
but finite, integer, then the corresponding Riemann sum can be a good approxima-
tion of the definite integral. Under certain circumstances this can be strengthened to
give rigorous bounds on the integral. Let us revisit Example 1.1.1.

22
I NTEGRATION 1.1 D EFINITION OF THE I NTEGRAL

Example 1.1.12 Upper and lower bounds on area.

R1
Let’s say we are again interested in the integral 0 ex dx. We can follow the same pro-
cedure as we used previously to construct Riemann sum approximations. However
since the integrand f (x) = ex is an increasing function, we can make our approxima-
tions into upper and lower bounds without much extra work.
More precisely, we approximate f (x) on each subinterval xi−1 ≤ x ≤ xi

• by its smallest value on the subinterval, namely f (xi−i ), when we compute the
left Riemann sum approximation and

• by its largest value on the subinterval, namely f (xi ), when we compute the
right Riemann sum approximation.

This is illustrated in the two figures below. The shaded region in the left hand figure
is the left Riemann sum approximation and the shaded region in the right hand
figure is the right Riemann sum approximation.

We can see that exactly because f (x) is increasing, the left Riemann sum describes
an area smaller than the definite integral while the right Riemann sum gives an area
larger a than the integral.
R1
When we approximate the integral 0 ex dx using n subintervals, then, on interval
number i,
i−1 i
• x runs from n
to n
and
(i−1)
• y = ex runs from e n , when x is at the left hand end point of the interval, to
i
e n , when x is at the right hand end point of the interval.
R1 (i−1)
Consequently, the left Riemann sum approximation to 0 ex dx is ni=1 e n n1 and
P
i
the right Riemann sum approximation is ni=1 e n · n1 . So
P

n 1 n
1 1
Z
X (i−1) X i
e n ≤ ex dx ≤ en ·
i=1
n 0 i=1
n

23
 I NTEGRATION 1.1 D EFINITION OF THE I NTEGRAL

(i−1)
Thus Ln = ni=1 e n n1 , which for any n can be evaluated by computer, is a lower
P
R1 i
bound on the exact value of 0 ex dx and Rn = ni=1 e n n1 , which for any n can also
P
R1
be evaluated by computer, is an upper bound on the exact value of 0 ex dx. For
example, when n = 1000, Ln = 1.7174 and Rn = 1.7191 (both to four decimal places)
so that, again to four decimal places,
Z 1
1.7174 ≤ ex dx ≤ 1.7191
0

Recall that the exact value is e − 1 = 1.718281828 . . ..

a When a function is decreasing the situation is reversed — the left Riemann sum is always larger
than the integral while the right Riemann sum is smaller than the integral. For more general func-
tions that both increase and decrease it is perhaps easiest to study each increasing (or decreasing)
interval separately.

Example 1.1.12

1.1.5 tt Using Known Areas to Evaluate Integrals

One of the main aims of this course is to build up general machinery for computing
definite integrals (as well as interpreting and applying them). We shall start on this
soon, but not quite yet. We have already seen one concrete, if laborious, method for
computing definite integrals — taking limits of Riemann sums as we did in Exam-
ple 1.1.1. A second method, which will work for some special integrands, works by
interpreting the definite integral as “signed area”. This approach will work nicely
when the area under the curve decomposes into simple geometric shapes like trian-
gles, rectangles and circles. Here are some examples of this second method.

Example 1.1.13 A very simple integral and a very simple area.

Rb Rb
The integral a 1dx (which is also written as just a dx) is the area of the shaded
rectangle (of width b − a and height 1) in the figure on the right below. So
Rb
a
dx = (b−a)×(1) = b−a

24
I NTEGRATION 1.1 D EFINITION OF THE I NTEGRAL

Example 1.1.14 Another simple integral.

Rb
Let b > 0. The integral 0 xdx is the area of the shaded triangle (of base b and of
height b) in the figure on the right below. So
Rb b2
0
xdx = 12 b × b = 2

R0
The integral −b xdx is the signed area of the shaded triangle (again of base b and of
height b) in the figure on the right below. So
R0 2
−b
xdx = − b2

Rb
Notice that it is very easy to extend this example to the integral 0
cxdx for any
real numbers b, c > 0 and find
Z b
c
cxdx = b2 .
0 2

R1
Example 1.1.15 Evaluating −1
(1 − |x|) dx.
R1
In this example, we shall evaluate −1
(1 − |x|) dx. Recall that
(
−x if x ≤ 0
|x| =
x if x ≥ 0
so that
(
1+x if x ≤ 0
1 − |x| =
1 − x if x ≥ 0

25
I NTEGRATION 1.1 D EFINITION OF THE I NTEGRAL

To picture the geometric figure whose area the integral represents observe that

• at the left hand end of the domain of integration x = −1 and the integrand
1 − |x| = 1 − | − 1| = 1 − 1 = 0 and

• as x increases from −1 towards 0, the integrand 1−|x| = 1+x increases linearly,

until

• when x hits 0 the integrand hits 1 − |x| = 1 − |0| = 1 and then

• as x increases from 0, the integrand 1 − |x| = 1 − x decreases linearly, until

• when x hits +1, the right hand end of the domain of integration, the integrand
hits 1 − |x| = 1 − |1| = 0.
R1
So the integral −1 (1 − |x|) dx is the area of the shaded triangle (of base 2 and of
height 1) in the figure on the right below and
R1 1
(1 − |x|) dx =
−1 2
×2×
1=1

Example 1.1.15

R1√
Example 1.1.16 Evaluating 0
1 − x2 dx.
R1√ √
The integral 0 1 − x2 dx has integrand f (x) = 1 − x2 . So it represents the area
√
under y = 1 − x2 with x running from 0 to 1. But we may rewrite
√
y = 1 − x2 as x2 + y 2 = 1, y ≥ 0

But this is the (implicit) equation for a circle — the extra condition that y ≥ 0 makes
it the equation for the semi-circle centred at the origin with radius 1 lying on and
above the x-axis. Thus the integral represents the area of the quarter circle of radius
1, as shown in the figure on the right below. So

26
 I NTEGRATION 1.1 D EFINITION OF THE I NTEGRAL

R1√
0
1 − x2 dx = 41 π(1)2 =
π
4

Example 1.1.16

This next one is a little trickier and relies on us knowing the symmetries of the
sine function.

Example 1.1.17 Integrating sine.

Rπ
The integral −π sin xdx is the signed area of the shaded region in the figure on the
right below. It naturally splits into two regions, one on either side of the y-axis. We
don’t know the formula for the area of either of these regions (yet), however the two
regions are very nearly the same. In fact, the part of the shaded region below the
x-axis is exactly the reflection, in the x-axis, of the part of the shaded region above
the x-axis. So the signed area of part of the shaded region below the x-axis is the
negative of the signed area of part of the shaded region above the x-axis and
Rπ
−π
sin xdx = 0

1.1.6 tt Another Interpretation for Definite Integrals

So far, we have only a single interpretation 6 for definite integrals — namely areas
under graphs. In the following example, we develop a second interpretation.

6 If this were the only interpretation then integrals would be a nice mathematical curiousity and
unlikely to be the core topic of a large first year mathematics course.

27
I NTEGRATION 1.1 D EFINITION OF THE I NTEGRAL

Example 1.1.18 A moving particle.

Suppose that a particle is moving along the x-axis and suppose that at time t its
velocity is v(t) (with v(t) > 0 indicating rightward motion and v(t) < 0 indicating
leftward motion). What is the change in its x-coordinate between time a and time
b > a?
We’ll work this out using a procedure similar to our definition of the integral. First
pick a natural number n and divide the time interval from a to b into n equal subin-
tervals, each of width b−a
n
. We are working our way towards a Riemann sum (as we
have done several times above) and so we will eventually take the limit n → ∞.

• The first time interval runs from a to a + b−a n

. If we think of n as some large
b−a
number, the width of this interval, n is very small and over this time interval,
the velocity does not change very much. Hence we can approximate the veloc-
ity over the first subinterval as being essentially constant at its value at the start
of the time interval — v(a). Over the subinterval the x-coordinate changes by
velocity times time, namely v(a) · b−an
.

• Similarly, the second interval runs from time a + b−a

n
to time a + 2 b−a
n
. Again,
we can assume that the velocity does not change very much and so we can
approximate the velocity as being essentially constant at its value at the start
b−a


of the subinterval — namely v a + n . So during the second
subinterval the
b−a b−a
particle’s x-coordinate changes by approximately v a + n n
.

• In general, time subinterval number i runs from a + (i − 1) b−a

n
to a + i b−a
n
and
during this subinterval the particle’s x-coordinate changes, essentially, by
 
b−a b−a
, v a + (i − 1) .
n n

So the net change in x-coordinate from time a to time b is approximately

b−a  b − a b − a  b − a b − a
v(a) +v a+ + · · · + v a + (i − 1) + ···
n n n n n
 b − a b − a
+ v a + (n − 1)
n n
n
X  b − a b − a
= v a + (i − 1)
i=1
n n

This exactly the left Riemann sum approximation to the integral of v from a to b
Rb
with n subintervals. The limit as n → ∞ is exactly the definite integral a v(t)dt.
Following tradition, we have called the (dummy) integration variable t rather than
x to remind us that it is time that is running from a to b.
The conclusion of the above discussion is that if a particle is moving along the x-axis
and its x-coordinate and velocity at time t are x(t) and v(t), respectively, then, for all

28
 I NTEGRATION 1.1 D EFINITION OF THE I NTEGRAL

b > a,
Z b
x(b) − x(a) = v(t)dt.
a

Example 1.1.18

1.1.7 tt Optional — careful definition of the integral

In this optional section we give a more mathematically rigorous definition of the def-
Z b
inite integral f (x)dx. Some textbooks use a sneakier, but equivalent, definition.
a
The integral will be defined as the limit of a family of approximations to the area
between the graph of y = f (x) and the x-axis, with x running from a to b. We will
then show conditions under which this limit is guaranteed to exist. We should state
up front that these conditions are more restrictive than is strictly necessary — this is
done so as to keep the proof accessible.
The family of approximations needed is slightly more general than that used to
define Riemann sums in the previous sections, though it is quite similar. The main
difference is that we do not require that all the subintervals have the same size.

• We start by selecting a positive integer n. As was the case previously, this will
be the number of subintervals used in the approximation and eventually we
will take the limit as n → ∞.

• Now subdivide the interval from a to b into n subintervals by selecting n + 1

values of x that obey

a = x0 < x1 < x2 < · · · < xn−1 < xn = b.

The subinterval number i runs from xi−1 to xi . This formulation does not re-
quire the subintervals to have the same size. However we will eventually re-
quire that the widths of the subintervals shrink towards zero as n → ∞.

• Then for each subinterval we select a value of x in that interval. That is, for
i = 1, 2, . . . , n, choose x∗i satisfying xi−1 ≤ x∗i ≤ xi . We will use these values of
x to help approximate f (x) on each subinterval.

• The area between the graph of y = f (x) and the x-axis, with x running

29
I NTEGRATION 1.1 D EFINITION OF THE I NTEGRAL

R xi
from xi−1 to xi , i.e. the contribution, xi−1 f (x)dx, from interval number i to the
integral, is approximated by the area of a rectangle. The rectangle has width
xi − xi−1 and height f (x∗i ).

• Thus the approximation to the integral, using all n subintervals, is

Z b
f (x)dx ≈ f (x∗1 )[x1 − x0 ] + f (x∗2 )[x2 − x1 ] + · · · + f (x∗n )[xn − xn−1 ]
a

• Of course every different choice of n and x1 , x2 , · · · , xn−1 and x∗1 , x∗2 , · · · , x∗n
gives a different approximation. So to simplify the discussion that follows,
let us denote a particular choice of all these numbers by P:

P = (n, x1 , x2 , · · · , xn−1 , x∗1 , x∗2 , · · · , x∗n ) .

Similarly let us denote the resulting approximation by I(P):

I(P) = f (x∗1 )[x1 − x0 ] + f (x∗2 )[x2 − x1 ] + · · · + f (x∗n )[xn − xn−1 ]

• We claim that, for any reasonable 7 function f (x), if you take any reasonable 8
sequence of these approximations you always get the exactly the same limiting
Rb
value. We define a f (x)dx to be this limiting value.

7 We’ll be more precise about what “reasonable” means shortly.

8 Again, we’ll explain this “reasonable” shortly

30
I NTEGRATION 1.1 D EFINITION OF THE I NTEGRAL

• Let’s be more precise. We can take the limit of these approximations in two
equivalent ways. Above we did this by taking the number of subintervals n to
infinity. When we did this, the width of all the subintervals went to zero. With
the formulation we are now using, simply taking the number of subintervals
to be very large does not imply that they will all shrink in size. We could have
one very large subinterval and a large number of tiny ones. Thus we take the
limit we need by taking the width of the subintervals to zero. So for any choice
P, we define

M (P) = max x1 − x0 , x2 − x1 , · · · , xn − xn−1

that is the maximum width of the subintervals used in the approximation de-
termined by P. By forcing the maximum width to go to zero, the widths of all
the subintervals go to zero.

• We then define the definite integral as the limit

Z b
f (x)dx = lim I(P).
a M (P)→0

Of course, one is now left with the question of determining when the above limit
exists. A proof of the very general conditions which guarantee existence of this limit
is beyond the scope of this course, so we instead give a weaker result (with stronger
conditions) which is far easier to prove.
For the rest of this section, assume

• that f (x) is continuous for a ≤ x ≤ b,

• that f (x) is differentiable for a < x < b, and

• that f 0 (x) is bounded — ie |f 0 (x)| ≤ F for some constant F .

We will now show that, under these hypotheses, as M (P) approaches zero, I(P)
always approaches the area, A, between the graph of y = f (x) and the x-axis, with x
running from a to b.
These assumptions are chosen to make the argument particularly transparent.
With a little more work one can weaken the hypotheses considerably. We are cheat-
ing a little by implicitly assuming that the area A exists. In fact, one can adjust the
argument below to remove this implicit assumption.

• Consider Aj , the part of the area coming from xj−1 ≤ x ≤ xj .

31
I NTEGRATION 1.1 D EFINITION OF THE I NTEGRAL

We have approximated this area by f (x∗j )[xj − xj−1 ] (see figure left).
• Let f (xj ) and f (xj ) be the largest and smallest values 9 of f (x) for xj−1 ≤ x ≤
xj . Then the true area is bounded by
f (xj )[xj − xj−1 ] ≤ Aj ≤ f (xj )[xj − xj−1 ].
(see figure right).
• Now since f (xj ) ≤ f (x∗j ) ≤ f (xj ), we also know that

f (xj )[xj − xj−1 ] ≤ f (x∗j )[xj−1 − xj ] ≤ f (xj )[xj − xj−1 ].

• So both the true area, Aj , and our approximation of that area f (x∗j )[xj − xj−1 ]
have to lie between f (xj )[xj − xj−1 ] and f (xj )[xj − xj−1 ]. Combining these
bounds we have that the difference between the true area and our approxima-
tion of that area is bounded by
Aj − f (x∗j )[xj − xj−1 ] ≤ [f (xj ) − f (xj )] · [xj − xj−1 ].
(To see this think about the smallest the true area can be and the largest our
approximation can be and vice versa.)
• Now since our function, f (x) is differentiable we can apply one of the main
theorems we learned in CLP-1 — the Mean Value Theorem 10 . The MVT implies
that there exists a c between xj and xj such that

f (xj ) − f (xj ) = f 0 (c) · [xj − xj ]

9 Here we are using the extreme value theorem — its proof is beyond the scope of this course.
The theorem says that any continuous function on a closed interval must attain a minimum and
maximum at least once. In this situation this implies that for any continuous function f (x), there
are xj−1 ≤ xj , xj ≤ xj such that f (xj ) ≤ f (x) ≤ f (xj ) for all xj−1 ≤ x ≤ xj .
10 Recall that the mean value theorem states that for a function continuous on [a, b] and differen-
tiable on (a, b), there exists a number c between a and b so that
f (b) − f (a)
f 0 (c) = .
b−a

32
I NTEGRATION 1.1 D EFINITION OF THE I NTEGRAL

• By the assumption that |f 0 (x)| ≤ F for all x and the fact that xj and xj must
both be between xj−1 and xj

f (xj ) − f (xj ) ≤ F · xj − xj ≤ F · [xj − xj−1 ]

Hence the error in this part of our approximation obeys

Aj − f (x∗j )[xj − xj−1 ] ≤ F · [xj − xj−1 ]2 .

• That was just the error in approximating Aj . Now we bound the total error by
combining the errors from approximating on all the subintervals. This gives
n
X n
X
|A − I(P)| = Aj − f (x∗j )[xj − xj−1 ]
j=1 j=1
n
X
Aj − f (x∗j )[xj − xj−1 ]


= triangle inequality
j=1
n
X
≤ Aj − f (x∗j )[xj − xj−1 ]
j=1
Xn
≤ F · [xj − xj−1 ]2 from above
j=1

Now do something a little sneaky. Replace one of these factors of [xj − xj−1 ]
(which is just the width of the j th subinterval) by the maximum width of the
subintervals:
n
X
≤ F · M (P) · [xj − xj−1 ] F and M (P) are constant
j=1
n
X
≤ F · M (P) · [xj − xj−1 ] sum is total width
j=1

= F · M (P) · (b − a).

• Since a, b and F are fixed, this tends to zero as the maximum rectangle width
M (P) tends to zero.
Thus, we have proven

Theorem 1.1.19
Assume that f (x) is continuous for a ≤ x ≤ b, and is differentiable for all a <
x < b with |f 0 (x)| ≤ F , for some constant F . Then, as the maximum rectangle
width M (P) tends to zero, I(P) always converges to A, the area between the
graph of y = f (x) and the x-axis, with x running from a to b.

33
 I NTEGRATION 1.1 D EFINITION OF THE I NTEGRAL

1.1.8 tt Exercises

Exercises — Stage 1 For Questions 1 through 5, we want you to develop an un-

derstanding of the model we are using to define an integral: we approximate the
area under a curve by bounding it between rectangles. Later, we will learn more so-
phisticated methods of integration, but they are all based on this simple concept.In
Questions 6 through 10, we practice using sigma notation. There are many ways to
write a given sum in sigma notation. You can practice finding several, and deciding
which looks the clearest.Questions 11 through 15 are meant to give you practice in-
terpreting the formulas in Definition 1.1.11. The formulas might look complicated at
first, but if you understand what each piece means, they are easy to learn.
1. Give a range of possible values for the shaded area in the picture below.

1.25
0.75
x
1 3

2. Give a range of possible values for the shaded area in the picture below.
y

2.25
1.75
1.25
0.75
0.25 x
1 2 3 4

3
1
Z
3. Using rectangles, find a lower and upper bound for dx that differ by
1 2x
at most 0.2 square units.

34
I NTEGRATION 1.1 D EFINITION OF THE I NTEGRAL

1
y= 2x

x
1 3

4. Let f (x) be a function that is decreasing from x = 0 to x = 5. Which Riemann

Z 5
sum approximation of f (x)dx is the largest–left, right, or midpoint?
0

5. Give an example of a function f (x), an interval [a, b], and a number n such
that the midpoint Riemann sum of f (x) over [a, b] using n intervals is larger
than both the left and right Riemann sums of f (x) over [a, b] using n inter-
vals.
6. Express the following sums in sigma notation:

a 3+4+5+6+7

b 6 + 8 + 10 + 12 + 14

c 7 + 9 + 11 + 13 + 15

d 1 + 3 + 5 + 7 + 9 + 11 + 13 + 15

7. Express the following sums in sigma notation:

1
a 3
+ 91 + 1
27
+ 1
81
2
b 3
+ 92 + 2
27
+ 2
81

c − 23 + 29 − 2
27
+ 2
81
2
d 3
− 29 + 2
27
− 2
81

8. Express the following sums in sigma notation:

1
a 3
+ 13 + 5
27
+ 7
81
+ 9
243
1 1 1 1 1
b 5
+ 11
+ 29
+ 83
+ 245

c 1000 + 200 + 30 + 4 + 12 + 3
50
+ 7
1000

9. Evaluate the following sums. You might want to use the formulas from
Theorems 5 and 6.

35
I NTEGRATION 1.1 D EFINITION OF THE I NTEGRAL

100  i
X 3
a
i=0
5

100  i
X 3
b
i=50
5

10
X
i2 − 3i + 5


c
i=1

b  n 
X 1 3
d + en , where b is some integer greater than 1.
n=1
e
10. Evaluate the following sums. You might want to use the formulas from
Theorem 1.1.6.
100
X 50
X
a (i − 50) + i
i=50 i=0

100
X
b (i − 5)3
i=10

11
X
c (−1)n
n=1

11
X
d (−1)2n+1
n=2

11. In the picture below, draw in the rectangles whose (signed) area
4
X b−a
is being computed by the midpoint Riemann sum ·
i=1
4
   
1 b−a
f a+ i− .
2 4

y
y = f (x)

x
a b

36
I NTEGRATION 1.1 D EFINITION OF THE I NTEGRAL

4
X
12. *. f (1 + k) · 1 is a left Riemann sum for a function f (x) on the interval
k=1
[a, b] with n subintervals. Find the values of a, b and n.
13. Draw a picture illustrating the area given by the following Riemann sum.
3
X
2 · (5 + 2i)2
i=1

14. Draw a picture illustrating the area given by the following Riemann sum.
5  
X π π(i − 1)
· tan
i=1
20 20

15. *. Fill in the blanks with right, left, or midpoint; an interval; and a value
of n.

3
P
a f (1.5 + k) · 1 is a Riemann sum for f on the interval
k=0
[ , ] with n = .

16. Evaluate the following integral by interpreting it as a signed area, and using
geometry:
Z 5
x dx
0

17. Evaluate the following integral by interpreting it as a signed area, and

using geometry:
Z 5
x dx
−2

Exercises — Stage 2 Questions 26 and 27 use the formula for a geometric sum,
Equation 1.1.3Remember that a definite integral is a signed area between a curve
and the x-axis. We’ll spend a lot of time learning strategies for evaluating definite
integrals, but we already know lots of ways to find area of geometric shapes. In
Questions 28 through 33, use your knowledge of geometry to find the signed areas
described by the integrals given.
8
18. *. Use sigma notation to write the midpoint Riemann sum for f (x) = x on
[5, 15] with n = 50. Do not evaluate the Riemann sum.
Z 5
19. *. Estimate x3 dx using three approximating rectangles and left hand
−1
end points.

37
I NTEGRATION 1.1 D EFINITION OF THE I NTEGRAL

Z 7
20. *. Let f be a function on the whole real line. Express f (x) dx as a
−1
limit of Riemann sums, using the right endpoints.

21. *. The value of the following limit is equal to the area below a graph of
y = f (x), integrated over the interval [0, b]:
n   2
X 4 4i
lim sin 2 +
n→∞
i=1
n n

Find f (x) and b.

22. *. For a certain function f (x), the following equation holds:

n
r Z 1
X k k2
lim 1− 2 = f (x) dx
n→∞
k=1
n2 n 0

Find f (x).

n  
X 3 −i/n 3i
23. lim
*. Express n→∞ e cos as a definite integral.
i=1
n n
n
X iei/n
24. *. Let Rn = n2
. Express lim Rn as a definite integral. Do not evaluate
n→∞
i=1
this integral.

n
X 
−1−2i/n 2
25. lim
*. Express n→∞ e · as an integral in three different ways.
i=1
n

26. Evaluate the sum 1 + r3 + r6 + r9 + · · · + r3n .

27. Evaluate the sum r5 + r6 + r7 + · · · + r100 .

Z 2
28. *. Evaluate |2x| dx.
−1

29. Evaluate the following integral by interpreting it as a signed area, and

using geometry:
Z 5
|t − 1| dt
−3

38
I NTEGRATION 1.1 D EFINITION OF THE I NTEGRAL

30. Evaluate the following integral by interpreting it as a signed area, and using
geometry:
Z b
x dx
a

where 0 ≤ a ≤ b.
31. Evaluate the following integral by interpreting it as a signed area, and using
geometry:
Z b
x dx
a

where a ≤ b ≤ 0.

32. Evaluate the following integral by interpreting it as a signed area, and

using geometry:
Z 4√
16 − x2 dx
0

Z 3
33. *. Use elementary geometry to calculate f (x) dx, where
0
(
x, if x ≤ 1,
f (x) =
1, if x > 1.

34. *. A car’s gas pedal is applied at t = 0 seconds and the car accelerates
continuously until t = 2 seconds. The car’s speed at half-second inter-
vals is given in the table below. Find the best possible upper estimate
for the distance that the car traveled during these two seconds.

t (s) 0 0.5 1.0 1.5 2

v (m/s) 0 14 22 30 40

35. True or false: the answer you gave for Question 34 is definitely greater
than or equal to the distance the car travelled during the two seconds in
question.

36. An airplane’s speed at one-hour intervals is given in the table below. Ap-
proximate the distance travelled by the airplane from noon to 4pm three
ways using a midpoint Riemann sum.

time 12:00 pm 1:00 pm 2:00 pm 3:00 pm 4:00 pm

speed (km/hr) 800 700 850 900 750

39
I NTEGRATION 1.1 D EFINITION OF THE I NTEGRAL

Exercises — Stage 3

37. *. (a) Express s

n  2
X 2 2i
lim 4 − −2 +
n→∞
i=1
n n
as a definite integal.
(b) Evaluate the integral of part (a).

38. *. Consider the integral:

Z 3
(7 + x3 ) dx. (∗)
0

a Approximate this integral using the left Riemann sum with n = 3

intervals.

b Write down the expression for the right Riemann sum with n intervals
and calculate the sum. Now take the limit n → ∞ in your expression
for the Riemann sum, to evaluate the integral (∗) exactly.

You may use the identity

n
X n4 + 2n3 + n2
i3 =
i=1
4
Z 4
39. *. Using a limit of right–endpoint Riemann sums, evaluate x2 dx. You
2
n n
i = n(n+1) i2 = n(n+1)(2n+1)
P P
may use the formulas 2
and 6
.
i=1 i=1
Z 2
40. *. Find (x3 + x) dx using the definition of the definite integral. You may
0
n n
n4 +2n3 +n2 n2 +n
i3 =
P P
use the summation formulas 4
and i= 2
.
i=1 i=1

Z 4
41. *. Using a limit of right-endpoint Riemann sums, evaluate (2x −
1
a
1) dx. Do not use anti-differentiation, except to check your answer.
n
i = n(n+1)
P
You may use the formula 2
.
i=1

a You’ll learn about this method starting in Section 1.3. You can also check this
answer using geometry.

40
I NTEGRATION 1.1 D EFINITION OF THE I NTEGRAL

42. Give a function f (x) that has the following expression as a right Riemann
sum when n = 10, ∆(x) = 10 and a = −5:
10
X
3(7 + 2i)2 sin(4i) .
i=1

43. Using the method of Example 1.1.2, evaluate

Z 1
2x dx
0
44.

a Using the method of Example 1.1.2, evaluate

Z b
10x dx
a

Using your answer from above, make a guess for

Z b
cx dx
a

where c is a positive constant. Does this agree with Question 43?

Z a √
45. Evaluate 1 − x2 dx using geometry, if 0 ≤ a ≤ 1.
0

46. Suppose f (x) is a positive, decreasing function from x = a to x = b. You

give an upper and lower bound on the area under the curve y = f (x) using
n rectangles and a left and right Riemann sum, respectively, as in the picture
below.
y y

y = f (x) y = f (x)
x x
a b a b

a What is the difference between the lower bound and the upper
bound? (That is, if we subtract the smaller estimate from the larger
estimate, what do we get?) Give your answer in terms of f , a, b, and
n.

41
 I NTEGRATION 1.2 B ASIC PROPERTIES OF THE DEFINITE INTEGRAL

b If you want to approximate the area under the curve to within 0.01
square units using this method, how many rectangles should you use?
That is, what should n be?
47. Let f (x) be a linear function, let a < b be integers, and let n be a whole
number. True or false: if we average the left and right Riemann sums for
Z b
f (x)dx using n rectangles, we get the same value as the midpoint Rie-
a
mann sum using n rectangles.

1.2q Basic properties of the definite integral

When we studied limits and derivatives, we developed methods for taking limits
or derivatives of “complicated functions” like f (x) = x2 + sin(x) by understanding
how limits and derivatives interact with basic arithmetic operations like addition
and subtraction. This allowed us to reduce the problem into one of of computing
derivatives of simpler functions like x2 and sin(x). Along the way we established
simple rules such as
d df dg
lim (f (x) + g(x)) = lim f (x) + lim g(x) and (f (x) + g(x)) = +
x→a x→a x→a dx dx dx
Some of these rules have very natural analogues for integrals and we discuss them
below. Unfortunately the analogous rules for integrals of products of functions or
integrals of compositions of functions are more complicated than those for limits or
derivatives. We discuss those rules at length in subsequent sections. For now let us
consider some of the simpler rules of the arithmetic of integrals.

Theorem 1.2.1 Arithmetic of Integration.

Let a, b and A, B, C be real numbers. Let the functions f (x) and g(x) be inte-
grable on an interval that contains a and b. Then
Z b Z b Z b
(a) (f (x) + g(x)) dx = f (x)dx + g(x)dx
a a a
Z b Z b Z b
(b) (f (x) − g(x)) dx = f (x)dx − g(x)dx
a a a
Z b Z b
(c) Cf (x)dx = C · f (x)dx
a a

Combining these three rules we have

Z b Z b Z b
(d) (Af (x) + Bg(x)) dx = A f (x)dx + B g(x)dx
a a a

42
I NTEGRATION 1.2 B ASIC PROPERTIES OF THE DEFINITE INTEGRAL

That is, integrals depend linearly on the integrand.

Z b Z b
(e) dx = 1 · dx = b − a
a a

It is not too hard to prove this result from the definition of the definite integral.
Additionally we only really need to prove (d) and (e) since
• (a) follows from (d) by setting A = B = 1,
• (b) follows from (d) by setting A = 1, B = −1, and
• (c) follows from (d) by setting A = C, B = 0.

Proof. As noted above, it suffices for us to prove (d) and (e). Since (e) is
easier, we will start with that. It is also a good warm-up for (d).
Rb
• The definite integral in (e), a 1dx, can be interpreted geometrically as the
area of the rectangle with height 1 running from x = a to x = b; this area
is clearly b − a. We can also prove this formula from the definition of the
integral (Definition 1.1.9):
Z b n
X b−a
dx = lim f (x∗i,n ) by definition
a n→∞
i=1
n
n
X b−a
= lim 1 since f (x) = 1
n→∞
i=1
n
n
X 1
= lim (b − a) since a, b are constants
n→∞
i=1
n
= lim (b − a)
n→∞
=b−a
as required.
• To prove (d) let us start by defining h(x) = Af (x) + Bg(x) and then we
need to express the integral of h(x) in terms of those of f (x) and g(x). We
use Definition 1.1.9 and some algebraic manipulations a to arrive at the
result.
Z b n
X b−a
h(x)dx = h(x∗i,n ) ·
a i=1
n
by Definition 1.1.9
n
X
b−a
= Af (x∗i,n ) + Bg(x∗i,n ) ·
i=1
n

43
I NTEGRATION 1.2 B ASIC PROPERTIES OF THE DEFINITE INTEGRAL

n  
X
∗ b−a ∗ b−a
= Af (xi,n ) · + Bg(xi,n ) ·
i=1
n n
n
! n
!
X b − a X b − a
= Af (x∗i,n ) · + Bg(x∗i,n ) ·
i=1
n i=1
n
by Theorem 1.1.5(b)
n
! n
!
X b − a X b−a
=A f (x∗i,n ) · +B g(x∗i,n ) ·
i=1
n i=1
n
by Theorem 1.1.5(a)
b Z Z b
=A f (x)dx + B g(x)dx
a a
by Definition 1.1.9

as required.

a Now is a good time to look back at Theorem 1.1.5.

Using this Theorem we can integrate sums, differences and constant multiples of
functions we know how to integrate. For example:

Example 1.2.2 The integral of a sum.

R1
In Example 1.1.1 we saw that 0
ex dx = e − 1. So
Z 1 Z 1 Z 1
x x


e + 7 dx = e dx + 7 1dx
0 0 0
by Theorem 1.2.1(d) with A = 1, f (x) = ex , B = 7, g(x) = 1
= (e − 1) + 7 × (1 − 0)
by Example 1.1.1 and Theorem 1.2.1(e)
=e+6

Rb
When we gave the formal definition of a f (x)dx in Definition 1.1.9 we explained
that the integral could be interpreted as the signed area between the curve y = f (x)
and the x-axis on the interval [a, b]. In order for this interpretation to make sense
we required that a < b, and though we remarked that the integral makes sense
when a > b we did not explain any further. Thankfully there is an easy way to
Rb Ra
express the integral a f (x)dx in terms of b f (x)dx — making it always possible
to write an integral so the lower limit of integration is less than the upper limit of

44
I NTEGRATION 1.2 B ASIC PROPERTIES OF THE DEFINITE INTEGRAL
R3 R7
integration. Theorem 1.2.3, below, tell us that, for example, 7 ex dx = − 3 ex dx. The
same theorem also provides us with two other simple manipulations of the limits of
integration.

Theorem 1.2.3 Arithmetic for the Domain of Integration.

Let a, b, c be real numbers. Let the function f (x) be integrable on an interval

that contains a, b and c. Then
Z a
(a) f (x)dx = 0
a
Z a Z b
(b) f (x)dx = − f (x)dx
b a
Z b Z c Z b
(c) f (x)dx = f (x)dx + f (x)dx
a a c

The proof of this statement is not too difficult.

Proof. Let us prove the statements in order.
• Consider the definition of the definite integral
b n
b−a
Z X
f (x)dx = lim f (x∗i,n ) ·
a n→∞
i=1
n

If we now substitute b = a in this expression we have

a n
a−a
Z X
f (x)dx = lim f (x∗i,n ) ·
a n→∞
i=1
n }
| {z
=0
n
X
= lim f (x∗i,n ) · 0
n→∞
i=1
| {z }
=0
= lim 0
n→∞
=0

as required.
Rb
• Consider now the definite integral a f (x)dx. We will sneak up on the
proof by first examining Riemann sum approximations to both this and
Ra Rb
b
f (x)dx. The midpoint Riemann sum approximation to a f (x)dx with
4 subintervals (so that each subinterval has width b−a 4
) is

f a + 21 b−a + f a + 32 b−a + f a + 52 b−a + f a + 27 b−a





b−a
4 4 4 4
· 4

45
I NTEGRATION 1.2 B ASIC PROPERTIES OF THE DEFINITE INTEGRAL

7
+ 81 b + f 5
+ 83 b + f 3
+ 58 b + f 1
+ 78 b b−a





= f 8
a 8
a 8
a 8
a · 4
Ra
Now we do the same for b f (x)dx with 4 subintervals. Note that b is now
the lower limit on the integral and a is now the upper limit on the integral.
This is likely to cause confusion when we write out the Riemann sum, so
we’ll temporarily rename b to A and a to B. The midpoint Riemann sum
RB
approximation to A f (x)dx with 4 subintervals is
n
f A + 21 B−A + f A + 23 B−A + f A + 52 B−A




4 4 4

o B−A
+ f A + 72 B−A4
· 4
n
= f 78 A + 18 B + f 85 A + 38 B + f 38 A + 85 B

o
+ f 81 A + 87 B · B−A
4

Now recalling that A =Rb and B = a, we have that the midpoint Riemann
a
sum approximation to b f (x)dx with 4 subintervals is
 7
f 8 b + 18 a + f 58 b + 83 a + f 38 b + 58 a + f 81 b + 78 a · a−b





4

Thus we see that the Riemann sums for the two integrals are nearly iden-
tical — the only difference being the factor of b−a
4
versus a−b
4
. Hence the
two Riemann sums are negatives of each other.
The same computation with n subintervals shows that the midpoint Rie-
Ra Rb
mann sum approximations to b f (x)dx and a f (x)dx with n subin-
tervals are negatives of each other. Taking the limit n → ∞ gives
Ra Rb
b
f (x)dx = − a f (x)dx.
• Finally consider (c) — we will not give a formal proof of this, but instead
will interpret it geometrically. Indeed one can also interpret (a) geometri-
cally. In both cases these become statements about areas:
Z a Z b Z c Z b
f (x)dx = 0 and f (x)dx = f (x)dx + f (x)dx
a a a c

are

Area (x, y) a ≤ x ≤ a, 0 ≤ y ≤ f (x) =0

and

Area (x, y) a ≤ x ≤ b, 0 ≤ y ≤ f (x)

= Area (x, y) a ≤ x ≤ c, 0 ≤ y ≤ f (x)

+ Area (x, y) c ≤ x ≤ b, 0 ≤ y ≤ f (x)

respectively. Both of these geometric statements are intuitively obvious.

See the figures below.

46
I NTEGRATION 1.2 B ASIC PROPERTIES OF THE DEFINITE INTEGRAL

y
y = f (x)

x
a c b

Note that we have assumed that a ≤ c ≤ b and that f (x) ≥ 0. One

can remove these restrictions and also make the proof more formal, but it
becomes quite tedious and less intuitive.

Remark 1.2.4 For notational simplicity, let’s assume that a ≤ c ≤ b and f (x) ≥ 0 for
all a ≤ x ≤ b. The geometric interpretations of the identities
Z a Z b Z c Z b
f (x)dx = 0 and f (x)dx = f (x)dx + f (x)dx
a a a c

are

Area (x, y) a ≤ x ≤ a, 0 ≤ y ≤ f (x) =0

and

Area (x, y) a ≤ x ≤ b, 0 ≤ y ≤ f (x)

= Area (x, y) a ≤ x ≤ c, 0 ≤ y ≤ f (x)

+ Area (x, y) c ≤ x ≤ b, 0 ≤ y ≤ f (x)

respectively. Both of these geometric statements are intuitively obvious. See the
figures below. We won’t give a formal proof.

y
y = f (x)

x
a c b
Ra Rb
So we concentrate on the formula b f (x)dx = − a f (x)dx. The midpoint Riemann
Rb
sum approximation to a f (x)dx with 4 subintervals (so that each subinterval has
width b−a
4
) is
n
f a + 12 b−a 3 b−a 5 b−a 7 b−a




b−a
4
+ f a + 2 4
+ f a + 2 4
+ f a + 2 4 4

47
I NTEGRATION 1.2 B ASIC PROPERTIES OF THE DEFINITE INTEGRAL

n  
o
7
+ 18 b + f 5
+ 38 b + f 3
+ 85 b + f 1
+ 87 b b−a


= f 8
a 8
a 8
a 8
a 4
(?)

We’re
Ra now going to write out the midpoint Riemann sum approximation to
b
f (x)dx with 4 subintervals. Note that b is now the lower limit on the integral
and a is now the upper limit on the integral. This is likely to cause confusion when
we write out the Riemann sum, so we’ll temporarily rename b to A and a to B. The
RB
midpoint Riemann sum approximation to A f (x)dx with 4 subintervals is
n
o B−A
f A + 21 B−A 3 B−A 5 B−A 7 B−A




4
+ f A + 2 4
+ f A + 2 4
+ f A + 2 4 4
n o
= f 87 A + 18 B + f 58 A + 38 B + f 83 A + 85 B + f 81 A + 78 B B−A





4

R aA = b and B = a, we have that the midpoint Riemann sum

Now recalling that
approximation to b f (x)dx with 4 subintervals is
n 7 1

o
5 3 3 5 1 7 a−b



f b+ a +f 8
b + 8
a +f 8
b + 8
a +f 8
b + 8
a 4
(??)
8 8
The curly brackets in (?) and (??) are equal to each other — the terms are just in
the reverse order. The factors multiplying the curly brackets in (?) and (??), namely
b−a
4
and a−b
4
, are negatives of each other, so (??)= −(?). The same computation
R a with
n subintervals shows that the midpoint Riemann sum approximations to b f (x)dx
Rb
and a f (x)dx with n subintervals are negatives of each other. Taking the limit n →
Ra Rb
∞ gives b f (x)dx = − a f (x)dx.

Example 1.2.5 Revisiting Example 1.1.14.

Rb 2
Back in Example 1.1.14 we saw that when b > 0 0 xdx = b2 . We’ll now verify that
Rb 2
0
xdx = b2 is still true when b = 0 and also when b < 0.
Rb 2
• First consider b = 0. Then the statement 0 xdx = b2 becomes
Z 0
xdx = 0
0

This is an immediate consequence of Theorem 1.2.3(a).

• Now consider b < 0. Let us write B = −b, so that B > 0. In Example 1.1.14 we
saw that
Z 0
B2
xdx = − .
−B 2
So we have
Z b Z −B Z 0
xdx = xdx = − xdx by Theorem 1.2.3(b)
0 0 −B

48
I NTEGRATION 1.2 B ASIC PROPERTIES OF THE DEFINITE INTEGRAL

B2
 
=− − by Example 1.1.14
2
2
B b2
= =
2 2

We have now shown that

Z b
b2
xdx = for all real numbers b
0 2
Example 1.2.5

Rb
Example 1.2.6 a
xdx.

Applying Theorem 1.2.3 yet again, we have, for all real numbers a and b,
Z b Z 0 Z b
xdx = xdx + xdx by Theorem 1.2.3(c) with c = 0
a a 0
Z b Z a
= xdx − xdx by Theorem 1.2.3(b)
0 0
2 2
b −a
= by Example 1.2.5, twice
2
We can also understand this result geometrically.

• (left) When 0 < a < b, the integral represents the area in green which is the dif-
ference of two right-angle triangles — the larger with area b2 /2 and the smaller
with area a2 /2.
• (centre) When a < 0 < b, the integral represents the signed area of the two
displayed triangles. The one above the axis has area b2 /2 while the one below
has area −a2 /2 (since it is below the axis).
• (right) When a < b < 0, the integral represents the signed area in purple of
the difference between the two triangles — the larger with area −a2 /2 and the
smaller with area −b2 /2.

49
I NTEGRATION 1.2 B ASIC PROPERTIES OF THE DEFINITE INTEGRAL

Theorem 1.2.3(c) shows us how we can split an integral over a larger interval into
one over two (or more) smaller intervals. This is particularly useful for dealing with
piece-wise functions, like |x|.

Example 1.2.7 Integrals involving |x|.

Using Theorem 1.2.3, we can readily evaluate integrals involving |x|. First, recall that
(
x if x ≥ 0
|x| =
−x if x < 0
R3
Now consider (for example) −2 |x|dx. Since the integrand changes at x = 0, it makes
sense to split the interval of integration at that point:
Z 3 Z 0 Z 3
|x|dx = |x|dx + |x|dx by Theorem 1.2.3
−2 −2 0
Z 0 Z 3
= (−x)dx + xdx by definition of |x|
−2 0
Z 0 Z 3
=− xdx + xdx by Theorem 1.2.1(c)
−2 0
= −(−22 /2) + (3 /2) = (4 + 9)/2
2

= 13/2

We can go further still — given a function f (x) we can rewrite the integral of f (|x|)
in terms of the integral of f (x) and f (−x).
Z 1

Z 0

Z 1

f |x| dx = f |x| dx + f |x| dx
−1 −1 0
Z0 Z 1
= f (−x)dx + f (x)dx
−1 0

Here is a more concrete example.

Example 1.2.8 Revisiting Example 1.1.15.

R1

Let us compute −1 1 − |x| dx again. In Example 1.1.15 we evaluated this integral
by interpreting it as the area of a triangle. This time we are going to use only the
properties given in Theorems 1.2.1 and 1.2.3 and the facts that
Z b Z b
b 2 − a2
dx = b − a and xdx =
a a 2
Rb Rb 2 2
That a dx = b − a is part (e) of Theorem 1.2.1. We saw that a xdx = b −a 2
in

50
 I NTEGRATION 1.2 B ASIC PROPERTIES OF THE DEFINITE INTEGRAL

Example 1.2.6.
First we are going to get rid of the absolute value signs by splitting the interval over
which we integrate. Recalling that |x| = x whenever x ≥ 0 and |x| = −x whenever
x ≤ 0, we split the interval by Theorem 1.2.3(c),
Z 1

Z 0

Z 1

1 − |x| dx = 1 − |x| dx + 1 − |x| dx
−1 −1 0
Z 0

Z 1


= 1 − (−x) dx + 1 − x dx
−1 0
Z 0

Z 1


= 1 + x dx + 1 − x dx
−1 0

Now we apply parts (a) and (b) of Theorem 1.2.1, and then
Z 1  
Z 0 Z 0 Z 1 Z 1
1 − |x| dx = 1dx + xdx + 1dx − xdx
−1 −1 −1 0 0
02 − (−1) 2
12 − 02
= [0 − (−1)] + + [1 − 0] −
2 2
=1
Example 1.2.8

1.2.1 tt More properties of integration: even and odd functions

Recall 1 the following definition

Definition 1.2.9
Let f (x) be a function. Then,

• we say that f (x) is even when f (x) = f (−x) for all x, and

• we say that f (x) is odd when f (x) = −f (−x) for all x.

Of course most functions are neither even nor odd, but many of the standard
functions you know are.

1 We haven’t done this in this course, but you should have seen it in your differential calculus
course or perhaps even earlier.

51
I NTEGRATION 1.2 B ASIC PROPERTIES OF THE DEFINITE INTEGRAL

Example 1.2.10 Even functions.

• Three examples of even functions are f (x) = |x|, f (x) = cos x and f (x) = x2 . In
fact, if f (x) is any even power of x, then f (x) is an even function.

• The part of the graph y = f (x) with x ≤ 0, may be constructed by drawing

the part of the graph with x ≥ 0 (as in the figure on the left below) and then
reflecting it in the y-axis (as in the figure on the right below).

• In particular, if f (x) is an even function and a > 0, then the two sets

(x, y) 0 ≤ x ≤ a and y is between 0 and f (x)

(x, y) −a ≤ x ≤ 0 and y is between 0 and f (x)

are reflections of each other in the y-axis and so have the same signed area.
That is
Z a Z 0
f (x)dx = f (x)dx
0 −a

Example 1.2.11 Odd functions.

• Three examples of odd functions are f (x) = sin x, f (x) = tan x and f (x) = x3 .
In fact, if f (x) is any odd power of x, then f (x) is an odd function.

• The part of the graph y = f (x) with x ≤ 0, may be constructed by drawing the
part of the graph with x ≥ 0 (like the solid line in the figure on the left below)
and then reflecting it in the y-axis (like the dashed line in the figure on the left
below) and then reflecting the result in the x-axis (i.e. flipping it upside down,
like in the figure on the right, below).

52
I NTEGRATION 1.2 B ASIC PROPERTIES OF THE DEFINITE INTEGRAL

• In particular, if f (x) is an odd function and a > 0, then the signed areas of the
two sets

(x, y) 0 ≤ x ≤ a and y is between 0 and f (x)

(x, y) −a ≤ x ≤ 0 and y is between 0 and f (x)

are negatives of each other — to get from the first set to the second set, you flip
it upside down, in addition to reflecting it in the y-axis. That is
Z a Z 0
f (x)dx = − f (x)dx
0 −a

Example 1.2.11

We can exploit the symmetries noted in the examples above, namely

Z a Z 0
f (x)dx = f (x)dx for f even
0 −a
Z a Z 0
f (x)dx = − f (x)dx for f odd
0 −a

together with Theorem 1.2.3 Theorem 1.2.3

Z a Z 0 Z a
f (x)dx = f (x)dx + f (x)dx
−a −a 0

in order to simplify the integration of even and odd functions over intervals of the
form [−a, a].

Theorem 1.2.12 Even and Odd.

Let a > 0.
a If f (x) is an even function, then
Z a Z a
f (x)dx = 2 f (x)dx
−a 0

53
 I NTEGRATION 1.2 B ASIC PROPERTIES OF THE DEFINITE INTEGRAL

b If f (x) is an odd function, then

Z a
f (x)dx = 0
−a

Proof. For any function

Z a Z a Z 0
f (x)dx = f (x)dx + f (x)dx
−a 0 −a

When f is even, the two terms on the right hand side are equal. When f is odd,
the two terms on the right hand side are negatives of each other.

1.2.2 tt Optional — More properties of integration: inequalities for in-

tegrals

We are still unable to integrate many functions, however with a little work we can
infer bounds on integrals from bounds on their integrands.

Theorem 1.2.13 Inequalities for Integrals.

Let a ≤ b be real numbers and let the functions f (x) and g(x) be integrable on
the interval a ≤ x ≤ b.

a If f (x) ≥ 0 for all a ≤ x ≤ b, then

Z b
f (x) dx ≥ 0
a

b If f (x) ≤ g(x) for all a ≤ x ≤ b, then

Z b Z b
f (x) dx ≤ g(x) dx
a a

c If there are constants m and M such that m ≤ f (x) ≤ M for all a ≤ x ≤ b,

then
Z b
m(b − a) ≤ f (x) dx ≤ M (b − a)
a

54
I NTEGRATION 1.2 B ASIC PROPERTIES OF THE DEFINITE INTEGRAL

d We have
Z b Z b
f (x) dx ≤ |f (x)| dx
a a

Proof.

a By interpreting the integral as the signed area, this statement simply says
 curve y = f (x) lies above the x-axis and a ≤ b, then the signed
that if the
area of (x, y) a ≤ x ≤ b, 0 ≤ y ≤ f (x) is at least zero. This is
quite clear. Alternatively, we could argue more algebraically from Defini-
Rb
tion 1.1.9. We observe that when we define a f (x)dx via Riemann sums,
every summand, f (x∗i,n ) b−a
n
≥ 0. Thus the whole sum is nonnegative and
consequently, so is the limit, and thus so is the integral.

b We are assuming that g(x) − f (x) ≥ 0, so part (a) gives

Z b  
Z b Z b
g(x) − f (x) dx ≥ 0 =⇒ g(x) dx − f (x) dx ≥ 0
a a a
Z b Z b
=⇒ f (x) dx ≤ g(x) dx
a a

c Applying part (b) with g(x) = M for all a ≤ x ≤ b gives

Z b Z b
f (x) dx ≤ M dx = M (b − a)
a a

Similarly, viewing m as a (constant) function, and applying part (b) gives

=m(b−a)
z }| {
Z b Z b
m ≤ f (x) =⇒ m dx ≤ f (x) dx
a a

d For any x, |f (x)| is either f (x) or −f (x) (depending on whether f (x) is

positive or negative), so we certainly have

f (x) ≤ |f (x)| and −f (x) ≤ |f (x)|

Applying part (c) to each of those inequalities gives

Z b Z b Z b Z b
f (x)dx ≤ |f (x)|dx and − f (x)dx ≤ |f (x)|dx
a a a a

55
 I NTEGRATION 1.2 B ASIC PROPERTIES OF THE DEFINITE INTEGRAL

Rb Rb Rb
Now a f (x)dx is either equal to a f (x)dx or − a f (x)dx (depending
on whether the integral is positive or negative). In either case we can
apply the above two inequalities to get the same result, namely
Z b Z b
f (x)dx ≤ |f (x)|dx.
a a

Rπ√
Example 1.2.14 3
0
cos xdx.

Consider the integral

π
√
Z
3
cos xdx
0

This is not so easy to compute exactly a , but we can bound it quite quickly.
For x between
√ 0 and π3 , the function cos x takes values b between 1 and 12 . Thus the
function cos x takes values between 1 and √12 . That is

1 √ π
√ ≤ cos x ≤ 1 for 0 ≤ x ≤ .
2 3

Consequently, by Theorem 1.2.13(b) with a = 0, b = π3 , m = √1 and M = 1,

2

π
π √ π
Z
3
√ ≤ cos xdx ≤
3 2 0 3

Plugging these expressions into a calculator gives us

π
√
Z
3
0.7404804898 ≤ cos xdx ≤ 1.047197551
0

a It is not too hard to use Riemann sums and a computer to evaluate it numerically: 0.948025319 . . ..
b You know the graphs of sine and cosine, so you should be able to work this out without too much
difficulty.

1.2.3 tt Exercises

Exercises — Stage 1

56
I NTEGRATION 1.2 B ASIC PROPERTIES OF THE DEFINITE INTEGRAL

1. For each of the following properties of definite integrals, draw a picture il-
lustrating the concept, interpreting definite integrals as areas under a curve.
For simplicity, you may assume that a ≤ c ≤ b, and that f (x), g(x) give
positive values.
Z a
a f (x) dx = 0, (Theorem 1.2.3, part (a))
a
Z b Z c Z b
b f (x) dx = f (x) dx + f (x)dx, (Theorem 1.2.3, part (c))
a a c
Z b Z b Z b
c (f (x) + g(x)) dx = f (x) dx + g(x) dx, (Theorem 1.2.1, part
a a a
(a))
Z b Z b
2. If cos xdx = sin b, then what is cos xdx?
0 a

3. *. Decide whether each of the following statements is true or false. If

false, provide a counterexample. If true, provide a brief justification.
(Assume that f (x) and g(x) are continuous functions.)

Z −2 Z 2
a f (x)dx = − f (x)dx.
−3 3
Z −2 Z 3
b If f (x) is an odd function, then f (x) dx = f (x) dx.
−3 2
Z 1 Z 1 Z 1
c f (x) · g(x)dx = f (x)dx · g(x)dx.
0 0 0

4. Suppose we want to make a right Riemann sum with 100 intervals to ap-
R0
proximate f (x)dx, where f (x) is a function that gives only positive values.
5

a What is ∆x?

b Are the heights of our rectangles positive or negative?

c Is our Riemann sum positive or negative?

d Is the signed area under the curve y = f (x) from x = 0 to x = 5

positive or negative?

Exercises — Stage 2

57
I NTEGRATION 1.2 B ASIC PROPERTIES OF THE DEFINITE INTEGRAL
Z 3 Z 3 Z 3
5. *. Suppose f (x) dx = −1 and g(x) dx = 5. Evaluate 6f (x) −

2 2 2
3g(x) dx.
Z 2 Z 2 Z 2


6. *. If f (x) dx = 3 and g(x) dx = −4, calculate 2f (x) + 3g(x) dx.
0 0 0

7. *. The functions f (x) and g(x) obey

Z −1 Z 2
f (x) dx = 1 f (x) dx = 2
0 0
Z 0 Z 2
g(x) dx = 3 g(x) dx = 4
−1 0
R2  
Find −1
3g(x) − f (x) dx.

8. In Question 1.1.8.45, Section 1.1, we found that

Z a√
π 1 1 √
1 − x2 dx = − arccos(a) + a 1 − a2
0 4 2 2

when 0 ≤ a ≤ 1.
Using this fact, evaluate the following:
Z 0√
a 1 − x2 dx, where −1 ≤ a ≤ 0
a
Z 1 √
b 1 − x2 dx, where 0 ≤ a ≤ 1
a

Z 2
9. *. Evaluate |2x|dx.
−1
Rb b2 −a2
You may use the result from Example 1.2.6 that xdx = 2
.
a

Z 5
10. Evaluate x|x|dx .
−5

Z 2
11. Suppose f (x) is an even function and f (x)dx = 10. What is
Z 0 −2

f (x)dx?
−2

Exercises — Stage 3

58
I NTEGRATION 1.2 B ASIC PROPERTIES OF THE DEFINITE INTEGRAL
Z 2  √ 
2
5 + 4 − x dx.
12. *. Evaluate
−2

+2012
sin x
Z
13. *. Evaluate dx.
−2012 log(3 + x2 )
Z +2012
14. *. Evaluate x1/3 cos x dx.
−2012
Z 6
15. Evaluate (x − 3)3 dx .
0

16. We want to compute the area of an ellipse, (ax)2 + (by)2 = 1 for some (let’s
say positive) constants a and b.

a Solve the equation for the upper half of the ellipse. It should have the
form “y = · · ·”

b Write an integral for the area of the upper half of the ellipse. Using
properties of integrals, make the integrand look like the upper half of
a circle.

c Using geometry and your answer to part (b), find the area of the el-
lipse.

17. Fill in the following table: the product of an (even/odd) function with
an (even/odd) function is an (even/odd) function. You may assume
that both functions are defined for all real numbers.

× even odd
even
odd

18. Suppose f (x) is an odd function and g(x) is an even function, both defined
at x = 0. What are the possible values of f (0) and g(0)?
19. Suppose f (x) is a function defined on all real numbers that is both even and
odd. What could f (x) be?

20. Is the derivative of an even function even or odd? Is the derivative of

an odd function even or odd?

59
 I NTEGRATION 1.3 T HE F UNDAMENTAL T HEOREM OF C ALCULUS

1.3q The Fundamental Theorem of Calculus

1.3.1 tt The Fundamental Theorem of Calculus

We have spent quite a few pages (and lectures) talking about definite integrals, what
they are (Definition 1.1.9), when they exist (Theorem 1.1.10), how to compute some
special cases (Section 1.1.5), some ways to manipulate them (Theorem 1.2.1 and 1.2.3)
and how to bound them (Theorem 1.2.13). Conspicuously missing from all of this
has been a discussion of how to compute them in general. It is high time we rectified
that.
The single most important tool used to evaluate integrals is called “the funda-
mental theorem of calculus”. Its grand name is justified — it links the two branches
of calculus by connecting derivatives to integrals. In so doing it also tells us how to
compute integrals. Very roughly speaking the derivative of an integral is the original
function. This fact allows us to compute integrals using antiderivatives 1 . Of course
“very rough” is not enough — let’s be precise.

Theorem 1.3.1 Fundamental Theorem of Calculus.

Let a < b and let f (x) be a function which is defined and continuous on [a, b].
Z x
• Part 1. Let F (x) = f (t)dt for any x ∈ [a, b]. Then the function F (x) is
a
differentiable and further

F 0 (x) = f (x)

• Part 2. Let G(x) be any function which is defined and continuous on

[a, b]. Further let G(x) be differentiable with G0 (x) = f (x) for all a < x <
b. Then
Z b Z b
f (x)dx = G(b) − G(a) or equivalently G0 (x)dx = G(b) − G(a)
a a

Before we prove this theorem and look at a bunch of examples of its application,
it is important that we recall one definition from differential calculus — antideriva-
tives. If F 0 (x) = f (x) on some interval, then F (x) is called an antiderivative of f (x)
on that interval. So Part 2 of the the fundamental theorem of calculus tells us how to
evaluate the definite integral of f (x) in terms of any of its antiderivatives — if G(x)

1 You learned these near the end of your differential calculus course. Now is a good time to revise
— but we’ll go over them here since they are so important in what follows.

60
I NTEGRATION 1.3 T HE F UNDAMENTAL T HEOREM OF C ALCULUS

is any antiderivative of f (x) then

Z b
f (x)dx = G(b) − G(a)
a
Rb
The form a G0 (x) dx = G(b) − G(a) of the fundamental theorem relates the rate of
change of G(x) over the interval a ≤ x ≤ b to the net change of G between x = a and
x = b. For that reason, it is sometimes called the “net change theorem”.
We’ll start with a simple example. Then we’ll see why the fundamental theorem
is true and then we’ll do many more, and more involved, examples.

Example 1.3.2 A first example.

Rb
Consider the integral a
xdx which we have explored previously in Example 1.2.6.

• The integrand is f (x) = x.

x2
• We can readily verify that G(x) = 2
satisfies G0 (x) = f (x) and so is an an-
tiderivative of the integrand.

• Part 2 of Theorem 1.3.1 then tells us that

Z b
f (x)dx = G(b) − G(a)
a
Z b
b 2 a2
xdx = −
a 2 2

which is precisely the result we obtained (with more work) in Example 1.2.6.

We do not give completely rigorous proofs of the two parts of the theorem — that
is not really needed for this course. We just give the main ideas of the proofs so that
you can understand why the theorem is true.
Part 1. We wish to show that if
Z x
F (x) = f (t)dt then F 0 (x) = f (x)
a

• Assume that F is the above integral and then consider F 0 (x). By defini-
tion
F (x + h) − F (x)
F 0 (x) = lim
h→0 h

• To understand this limit, we interpret the terms F (x), F (x + h) as signed

areas. To simplify this further, let’s only consider the case that f is always
nonnegative and that h > 0. These restrictions are not hard to remove,
but the proof ideas are a bit cleaner if we keep them in place. Then we

61
I NTEGRATION 1.3 T HE F UNDAMENTAL T HEOREM OF C ALCULUS

have

F (x + h) = the area of the region (t, y) a ≤ t ≤ x + h, 0 ≤ y ≤ f (t)

F (x) = the area of the region (t, y) a ≤ t ≤ x, 0 ≤ y ≤ f (t)

• Then the numerator


F (x + h) − F (x) = the area of (t, y) x ≤ t ≤ x + h, 0 ≤ y ≤ f (t)

This is just the more darkly shaded region in the figure

• We will be taking the limit h → 0. So suppose that h is very small. Then,

as t runs from x to x = h, f (t) runs only over a very narrow range of
values a , all close to f (x).

• So the darkly shaded region is almost a rectangle of width h and height

f (x) and so has an area which is very close to f (x)h. Thus F (x+h)−F
h
(x)
is
very close to f (x).
F (x+h)−F (x)
• In the limit h → 0, h
becomes exactly f (x), which is precisely
what we want.

a Notice that if f were discontinuous, then this might be false.

We can make the above more rigorous using the Mean Value Theorem 2 .

2 The MVT tells us that there is a number c between x and x + h so that

F (x + h) − F (x) F (x + h) − F (x)
F 0 (c) = =
(x + h) − x h

But since F 0 (x) = f (x), this tells us that

F (x + h) − F (x)
= f (c)
h
where c is trapped between x + h and x. Now when we take the limit as h → 0 we have that this
number c is squeezed to x and the result follows.

62
I NTEGRATION 1.3 T HE F UNDAMENTAL T HEOREM OF C ALCULUS

Rb
Part 2. We want to show that a f (t)dt = G(b) − G(a). To do this we exploit
the fact that the derivative of a constant is zero.

• Let
Z x
H(x) = f (t)dt − G(x) + G(a)
a

Then the result we wish to prove is that H(b) = 0. We will do this by

showing that H(x) = 0 for all x between a and b.

• We first show that H(x) is constant by computing its derivative:

Z x
0 d d d
H (x) = f (t)dt − (G(x)) + (G(a))
dx a dx dx

Since G(a) is a constant, its derivative is 0 and by assumption the deriva-

tive of G(x) is just f (x), so
Z x
d
= f (t)dt − f (x)
dx a

Now Part 1 of the theorem tells us that this derivative is just f (x), so

= f (x) − f (x) = 0

Hence H is constant.

• To determine which constant we just compute H(a):

Z a
H(a) = f (t)dt − G(a) + G(a)
a
Z a
= f (t)dt by Theorem 1.2.3(a)
a
=0

as required.

The simple example we did above (Example 1.3.2), demonstrates the application
of part 2 of the fundamental theorem of calculus. Before we do more examples (and
there will be many more over the coming sections) we should do some examples
illustrating the use of part 1 of the fundamental theorem of calculus. Then we’ll
move on to part 2.

63
I NTEGRATION 1.3 T HE F UNDAMENTAL T HEOREM OF C ALCULUS

d
Rx
Example 1.3.3 dx 0
tdt.
Rx
Consider the integral 0 t dt. We know how to evaluate this — it is just Example 1.3.2
with a = 0, b = x. So we have two ways to compute the derivative. We can evaluate
the integral and then take the derivative, or we can apply Part 1 of the fundamental
theorem. We’ll do both, and check that the two answers are the same.
First, Example 1.3.2 gives
Z x
x2
F (x) = t dt =
0 2

So of course F 0 (x) = x. Second, Part 1 of the fundamental theorem of calculus tells

us that the derivative of F (x) is just the integrand. That is, Part 1 of the fundamental
theorem of calculus also gives F 0 (x) = x.

In the previous example we were able to evaluate the integral explicitly, so we did
not need the fundamental theorem to determine its derivative. Here is an example
that really does require the use of the fundamental theorem.

d
Rx 2
Example 1.3.4 dx 0
e−t dt.
d
R x −t2
We would like to find dx 0
e dt. In the previous example, we were able to compute
the corresponding derivative in two ways — we could explicitly compute the inte-
gral and then differentiate the result, or we could apply part 1 of the fundamental
theorem of calculus. In this example weRdo not know the integral explicitly. Indeed
x 2
it is not possible to express a the integral 0 e−t dt as a finite combination of standard
functions such as polynomials, exponentials, trigonometric functions and so on.
Despite this, we can find its derivative by just applying the first part of the funda-
2
mental theorem of calculus with f (t) = e−t and a = 0. That gives
Z x Z x
d −t2 d
e dt = f (t)dt
dx 0 dx 0
2
= f (x) = e−x

Rx 2
a The integral 0 e−t dt is closely related to the “error function” which is an extremely important
function in mathematics. While we cannot express this integral (or the error function) as a finite
combination of polynomials, exponentials etc, we can express it as an infinite series
Z x
2 x3 x5 x7 x9
e−t dt = x − + − + + ···
0 3 · 1 5 · 2 7 · 3! 9 · 4!
x2k+1
+ (−1)k + ···
(2k + 1) · k!

But more on this in Chapter 3.

64
I NTEGRATION 1.3 T HE F UNDAMENTAL T HEOREM OF C ALCULUS

Let us ratchet up the complexity of the previous example — we can make the
limits of the integral more complicated functions. So consider the previous example
with the upper limit x replaced by x2 :

d
R x2 2
Example 1.3.5 dx 0
e−t dt.
R x2 2
Consider the integral 0 e−t dt. We would like to compute its derivative with re-
spect to x using part 1 of the fundamental theorem of calculus.
The fundamental
Rx theorem tells us how to compute the derivative of functions of
the form a f (t)dt but the integral at hand is not of the specified form because the
upper limit we have is x2 , rather than x, — so more care is required. Thankfully we
can deal with this obstacle with only a little extra work. The trick is to define an
auxiliary function by simply changing the upper limit to x. That is, define
Z x
2
E(x) = e−t dt
0

Then the integral we want to work with is

Z x2
2 2
E(x ) = e−t dt
0

The derivative E 0 (x) can be found via part 1 of the fundamental theorem of calculus
2
(as we did in Example 1.3.4) and is E 0 (x) = e−x . We can then use this fact with the
chain rule to compute the derivative we need:
x2
d d
Z
2
e−t dt = E(x2 ) use the chain rule
dx 0 dx
= 2xE 0 (x2 )
4
= 2xe−x

What if both limits of integration are functions of x? We can still make this work,
but we have to split the integral using Theorem 1.2.3.

d
R x2 2
Example 1.3.6 dx x
e−t dt.

Consider the integral

Z x2
2
e−t dt
x

As was the case in the previous example, we have to do a little pre-processing before
we can apply the fundamental theorem.
This time (by design), not only is the upper limit of integration x2 rather than x,

65
I NTEGRATION 1.3 T HE F UNDAMENTAL T HEOREM OF C ALCULUS

but the lower

Rx limit of integration also depends on x — this is different from the
integral a f (t)dt in the fundamental theorem where the lower limit of integration is
a constant.
Fortunately we can use the basic properties of integrals (Theorem 1.2.3(b) and (c)) to
R x2 2
split x e−t dt into pieces whose derivatives we already know.
Z x2 Z 0 Z x2
−t2 −t2 2
e dt = e dt + e−t dt by Theorem 1.2.3(c)
x x 0
Z x Z x2
−t2 2
=− e dt + e−t dt by Theorem 1.2.3(b)
0 0

With this pre-processing, both integrals are of the right form. Using what we have
learned in the the previous two examples,
Z x2 Z x Z x2 !
d 2 d 2 2
e−t dt = − e−t dt + e−t dt
dx x dx 0 0
Z x Z x2
d 2 d 2
=− e−t dt + e−t dt
dx 0 dx 0
2 4
= −e−x + 2xe−x
Example 1.3.6

Before we start to work with part 2 of the fundamental theorem, we need a lit-
tle terminology and notation. First some terminology — you may have seen this
definition in your differential calculus course.
Definition 1.3.7 Antiderivatives.
Let f (x) and F (x) be functions. If F 0 (x) = f (x) on an interval, then we say that
F (x) is an antiderivative of f (x) on that interval.

As we saw above, an antiderivative of f (x) = x is F (x) = x2 /2 — we can easily

verify this by differentiation. Notice that x2 /2 + 3 is also an antiderivative of x, as is
x2 /2 + C for any constant C. This observation gives us the following simple lemma.

Lemma 1.3.8
Let f (x) be a function and let F (x) be an antiderivative of f (x). Then F (x) + C
is also an antiderivative for any constant C. Further, every antiderivative of
f (x) must be of this form.

66
I NTEGRATION 1.3 T HE F UNDAMENTAL T HEOREM OF C ALCULUS

Proof. There are two parts to the lemma and we prove each in turn.

• Let F (x) be an antiderivative of f (x) and let C be some constant. Then

d d d
(F (x) + C) = (F (x)) + (C)
dx dx dx
= f (x) + 0

since the derivative of a constant is zero, and by definition the derivative

of F (x) is just f (x). Thus F (x) + C is also an antiderivative of f (x).

• Now let F (x) and G(x) both be antiderivatives of f (x) — we will show
that G(x) = F (x) + C for some constant C. To do this let H(x) = G(x) −
F (x). Then

d d d d
H(x) = (G(x) − F (x)) = G(x) − F (x)
dx dx dx dx
= f (x) − f (x) = 0

Since the derivative of H(x) is zero, H(x) must be a constant function a .

Thus H(x) = G(x)−F (x) = C for some constant C and the result follows.

a This follows from the Mean Value Theorem. Say H(x) were not constant, then there would
be two numbers a < b so that H(a) 6= H(b). Then the MVT tells us that there is a number
c between a and b so that
H(b) − H(a)
H 0 (c) = .
b−a
Since both numerator and denominator are nonzero, we know the derivative at c is
nonzero. But this would contradict the assumption that derivative of H is zero. Hence
we cannot have a < b with H(a) 6= H(b) and so H(x) must be constant.

Based on the above lemma we have the following definition.

Definition 1.3.9
R
The “indefinite integral of f (x)” is denoted by f (x)dx and should be re-
garded as the general antiderivative of f (x). In particular, if F (x) is an an-
tiderivative of f (x) then
Z
f (x)dx = F (x) + C

where the C is an arbitrary constant. In this context, the constant C is also

often called a “constant of integration”.

Now we just need a tiny bit more notation.

67
I NTEGRATION 1.3 T HE F UNDAMENTAL T HEOREM OF C ALCULUS

Definition 1.3.10
The symbol
Z b
f (x)dx
a

denotes the change in an antiderivative of f (x) from x = a to x = b. More

precisely, let F (x) be any antiderivative of f (x). Then
Z b
f (x)dx = F (x)|ba = F (b) − F (a)
a

Notice that this notation allows us to write part 2 of the fundamental theorem as
Z b Z b
f (x)dx = f (x)dx
a a
= F (x)|ba = F (b) − F (a)

Some texts also use an equivalent notation using square brackets:

Z b h ib
f (x)dx = F (x) = F (b) − F (a).
a a

You should be familiar with both notations.

We’ll soon develop some strategies for computing more complicated integrals.
But for now, we’ll try a few integrals that are simple enough that we can just guess
the answer. Of course, any antiderivative that we can guess we can also check —
simply differentiate the guess and verify you get back to the original function:

d
Z
f (x)dx = f (x).
dx
We do these examples in some detail to help us become comfortable finding indefi-
nite integrals.

R2
Example 1.3.11 Compute the definite integral 1
xdx.
R2
Compute the definite integral 1 xdx.
R2 2 2
Solution: We have already seen, in Example 1.2.6, that 1 xdx = 2 −1
2
= 32 . We shall
now rederive that result using the fundamental theorem of calculus.

• The main difficulty in this approach is finding the indefinite integral (an an-
tiderivative) of x. That is, we need to find a function F (x) whose derivative
is x. So think back to all the derivatives you computed last term a and try to
remember a function whose derivative was something like x.

68
I NTEGRATION 1.3 T HE F UNDAMENTAL T HEOREM OF C ALCULUS

• This shouldn’t be too hard — we recall that the derivatives of polynomials are
polynomials. More precisely, we know that

d n
x = nxn−1
dx
So if we want to end up with just x = x1 , we need to take n = 2. However this
gives us

d 2
x = 2x
dx

• This is pretty close to what we want except for the factor of 2. Since this is a
constant we can just divide both sides by 2 to obtain:

1 d 2 1
· x = · 2x which becomes
2 dx 2
d x2
· =x
dx 2
which is exactly what we need. It tells us that x2 /2 is an antiderivative of x.

• Once one has an antiderivative, it is easy to compute the indefinite integral

1
Z
xdx = x2 + C
2
as well as the definite integral:
2 2
1
Z
xdx = x2 since x2 /2 is an antiderivative of x
1 2 1
1 1 3
= 22 − 12 =
2 2 2

a Of course, this assumes that you did your differential calculus course last term. If you did that
course at a different time then please think back to that point in time. If it is long enough ago that
you don’t quite remember when it was, then you should probably do some revision of derivatives
of simple functions before proceeding further.

Example 1.3.11

While the previous example could be computed using signed areas, the following
example would be very difficult to compute without using the fundamental theorem
of calculus.

69
I NTEGRATION 1.3 T HE F UNDAMENTAL T HEOREM OF C ALCULUS

Rπ
Example 1.3.12 Compute 0
2
sin xdx.
Rπ
Compute 0
2
sin xdx.
Solution:

• Once again, the crux of the solution is guessing the antiderivative of sin x —
that is finding a function whose derivative is sin x.

• The standard derivative that comes closest to sin x is

d
cos x = − sin x
dx
which is the derivative we want, multiplied by a factor of −1.

• Just as we did in the previous example, we multiply this equation by a constant

to remove this unwanted factor:
d
(−1) · cos x = (−1) · (− sin x) giving us
dx
d

− cos x = sin x
dx
This tells us that − cos x is an antiderivative of sin x.

• Now it is straightforward to compute the integral:

Z π
2 π
sin xdx = − cos x|02 since − cos x is an antiderivative of sin x
0
π
= − cos
+ cos 0
2
=0+1=1

R2 1
Example 1.3.13 Compute 1 x
dx.
R2
Find 1 x1 dx.
Solution:

• Once again, the crux of the solution is guessing a function whose derivative is
1
x
. Our standard way to differentiate powers of x, namely

d n
x = nxn−1 ,
dx
doesn’t work in this case — since it would require us to pick n = 0 and this

70
I NTEGRATION 1.3 T HE F UNDAMENTAL T HEOREM OF C ALCULUS

would give

d 0 d
x = 1 = 0.
dx dx

• Fortunately, we also know a that

d 1
log x =
dx x
which is exactly the derivative we want.

• We’re now ready to compute the prescribed integral.

2
1
Z
dx = log x|21 since log x is an antiderivative of 1/x
1 x
= log 2 − log 1 since log 1 = 0
= log 2

a Recall that in most mathematics courses (especially this one) we use log x without any indicated
base to denote the natural logarithm — the logarithm base e. Many widely used computer lan-
guages, like Java, C, Python, MATLAB, · · ·, use log(x) to denote the logarithm base e too. But
many texts also use ln x to denote the natural logarithm

log x = loge x = ln x.

The reader should be comfortable with all three notations for this function. They should also be
aware that in different contexts — such as in chemistry or physics — it is common to use log x to
denote the logarithm base 10, while in computer science often log x denotes the logarithm base
2. Context is key.

Example 1.3.13

R −11
Example 1.3.14 −2 x
dx.
R −1
Find −2 x1 dx.
Solution:
• As we saw in the last example,
d 1
log x =
dx x
and if we naively use this here, then we will obtain
Z −1
1
dx = log(−1) − log(−2)
−2 x

71
I NTEGRATION 1.3 T HE F UNDAMENTAL T HEOREM OF C ALCULUS

which makes no sense since the logarithm is only defined for positive numbers
a
.

• We can work around this problem using a slight variation of the logarithm —
log |x|.

◦ When x > 0, we know that |x| = x and so we have

log |x| = log x differentiating gives us

d d 1
log |x| = log x = .
dx dx x

◦ When x < 0 we have that |x| = −x and so

log |x| = log(−x) differentiating with the chain rule gives

d d
log |x| = log(−x)
dx dx
1 1
= · (−1) =
(−x) x

◦ Indeed, more generally we should write the indefinite integral of 1/x as

1
Z
dx = log |x| + C
x
which is valid for all positive and negative x. It is, however, undefined at
x = 0.

• We’re now ready to compute the prescribed integral.

−1 −1
1
Z
dx = log |x| since log |x| is an antiderivative of 1/x
−2 x −2
= log | − 1| − log | − 2| = log 1 − log 2
1
= − log 2 = log .
2

a This is not entirely true — one can extend the definition of the logarithm to negative numbers,
but to do so one needs to understand complex numbers which is a topic beyond the scope of this
course.

Example 1.3.14

This next example raises a nasty issue that requires a little care. We know that
the function 1/x is not defined at x = 0 — so can we integrate over an interval that
contains x = 0 and still obtain an answer that makes sense? More generally can we
integrate a function over an interval on which that function has discontinuities?

72
I NTEGRATION 1.3 T HE F UNDAMENTAL T HEOREM OF C ALCULUS

R1 1
Example 1.3.15 −1 x2
dx.
R1
Find −1 x12 dx.
Solution: Beware that this is a particularly nasty example, which illustrates a booby
trap hidden in the fundamental theorem of calculus. The booby trap explodes when
the theorem is applied sloppily.

• The sloppy solution starts, as our previous examples have, by finding an an-
tiderivative of the integrand. In this case we know that

d 1 1
=− 2
dx x x
which means that −x−1 is an antiderivative of x−2 .

• This suggests (if we proceed naively) that

1 1
1
Z
−2
x dx = − since −1/x is an antiderivative of 1/x2
−1 x −1
1  1 
=− − −
1 −1
= −2

Unfortunately,

• At this point we should really start to be concerned. This answer cannot be

correct. Our integrand, being a square, is positive everywhere. So our integral
represents the area of a region above the x-axis and must be positive.

• So what has gone wrong? The flaw in the computation is that the fundamental
theorem of calculus, which says that
Z b
0
if F (x) = f (x) then f (x)dx = F (b) − F (a),
a

is only applicable when F 0 (x) exists and equals f (x) for all x between a and b.

• In this case F 0 (x) = x12 does not exist for x = 0. So we cannot apply the funda-
mental theorem of calculus as we tried to above.
R1
An integral, like −1 x12 dx, whose integrand is undefined somewhere in the domain
of integration is called improper. We’ll give a more thorough treatment of improper
integrals later in the text. For now, we’ll just say that the correct way to define (and
evaluate) improper integrals is as a limit of well-defined approximating integrals.
R1
We shall later see that, not only is −1 x12 dx not negative, it is infinite.

73
I NTEGRATION 1.3 T HE F UNDAMENTAL T HEOREM OF C ALCULUS

Remark 1.3.16 For completeness we’ll show how to evaluate this integral by sneak-
ing up on the point of discontinuity in the interval of integration. As noted above,
we will give a fuller explanation of such integrals later in the text.

• Rather than evaluating the integral directly, we will approximate the integral
using definite integrals on intervals that avoid the discontinuity. In the current
example, the original domain of integration is −1 ≤ x ≤ 1. The domains of
integration of the approximating integrals exclude from [−1, 1] small intervals
around x = 0.

• The shaded area in the figure below illustrates a typical approximating integral,
whose domain of integration consists of the original domain of integration,
[−1, 1], but with the interval [−t, T ] excluded.

The full domain of integration is only recovered in the limit t, T → 0.

• For this example, the correct computation is

1 Z −t Z 1
1 1 1
Z
2
dx = lim+ 2
dx + lim+ 2
dx
−1 x t→0 −1 x T →0 T x
 −t  1
1 1
= lim+ − + lim −
t→0 x −1 T →0+ x T
h 1   1 i h 1   1 i
= lim+ − − − + lim+ − − −
t→0 −t −1 T →0 1 T
1 1
= lim+ + lim+ − 2
t→0 t T →0 T
= +∞

• We can interpret this to mean that the signed area under the curve x−2 between
x = −1 and x = 1 is infinite.

Example 1.3.16

74
I NTEGRATION 1.3 T HE F UNDAMENTAL T HEOREM OF C ALCULUS

The above examples have illustrated how we can use the fundamental theorem
of calculus to convert knowledge of derivatives into knowledge of integrals. We are
now in a position to easily built a table of integrals. Here is a short table of the most
important derivatives that we know.

F (x) 1 xn sin x cos x tan x ex loge |x| arcsin x arctan x

1 √ 1 1
f (x) = F 0 (x) 0 nxn−1 cos x − sin x sec2 x ex x 1−x2 1+x2

Of course we know other derivatives, such as those of sec x and cot x, however
the ones listed above are arguably the most important ones. From this table (with a
very little massaging) we can write down a short table of indefinite integrals.

Theorem 1.3.17 Important indefinite integrals.

R
f (x) F (x) = f (x)dx
1 x+C
1
xn n+1
xn+1 + C provided that n 6= −1
1
loge |x| + C
x
ex ex + C
sin x − cos x + C
cos x sin x + C
sec2 x tan x + C
1
√ arcsin x + C
1 − x2
1
arctan x + C
1 + x2

Example 1.3.18 Using Theorem 1.3.17 to compute some integrals.

Find the following integrals

R7
i 2 ex dx
R2 1
ii −2 1+x 2 dx

R3
iii 0 (2x3 + 7x − 2)dx

Solution: We can proceed with each of these as before — find the antiderivative and
then apply the fundamental theorem. The third integral is a little more complicated,
but we can split it up into monomials using Theorem 1.2.1 and do each separately.

75
I NTEGRATION 1.3 T HE F UNDAMENTAL T HEOREM OF C ALCULUS

i An antiderivative of ex is just ex , so
Z 7 7
ex dx = ex
2 2
= e − e2 = e2 (e5 − 1).
7

1
ii An antiderivative of 1+x2
is arctan(x), so

2 2
1
Z
dx = arctan(x)
−2 1 + x2 −2
= arctan(2) − arctan(−2)

We can simplify this a little further by noting that arctan(x) is an odd function,
so arctan(−2) = − arctan(2) and thus our integral is

= 2 arctan(2)

iii We can proceed by splitting the integral using Theorem 1.2.1(d)

Z 3 Z 3 Z 3 Z 3
3 3
(2x + 7x − 2)dx = 2x dx + 7xdx − 2dx
0 0 0 0
Z 3 Z 3 Z 3
3
=2 x dx + 7 xdx − 2 dx
0 0 0

and because we know that x4 /4, x2 /2, x are antiderivatives of x3 , x, 1 respec-

tively, this becomes
3  2 3
x4

7x
= + − [2x]30
2 0 2 0
81 7 · 9
= + −6
2 2
81 + 63 − 12 132
= = = 66.
2 2
We can also just find the antiderivative of the whole polynomial by finding
the antiderivatives of each term of the polynomial and then recombining them.
This is equivalent to what we have done above, but perhaps a little neater:
3 3
x4 7x2
Z 
3
(2x + 7x − 2)dx = + − 2x
0 2 2 0
81 7 · 9
= + − 6 = 66.
2 2
Example 1.3.18

76
 I NTEGRATION 1.3 T HE F UNDAMENTAL T HEOREM OF C ALCULUS

1.3.2 tt Exercises

Exercises — Stage 1 Questions 11 through 14 are meant to help reinforce key ideas
in the Fundamental Theorem of Calculus and its proof.So far, we have been able
to guess many antiderivatives. Often, however, antiderivatives are very difficult to
guess. In Questions 16 through 19, we will find some antiderivatives that might
appear in a table of integrals. Coming up with the antiderivative might be quite
difficult (strategies to do just that will form a large part of this semester), but verifying
that your antiderivative is correct is as simple as differentiating.
2
1. *. Suppose that f (x) is a function and F (x) = e(x −3) + 1 is an antiderivative
Z √5
of f (x). Evaluate the definite integral f (x) dx.
1
3
2. *. For the function f (x) = x − sin 2x, find its antiderivative F (x) that satis-
fies F (0) = 1.

3. *. Decide whether each of the following statements is true or false. Pro-

vide a brief justification.

If f (x) is continuous on [1, π] and differentiable on (1, π), then

a Z
π
f 0 (x) dx = f (π) − f (1).
1

1
1
Z
b dx = 0.
−1 x2
Z b Z b
c If f is continuous on [a, b] then xf (x) dx = x f (x) dx.
a a

1
4. True or false: an antiderivative of 2 is log(x2 ) (where by log x we mean
x
logarithm base e).

sin(ex )
5. True or false: an antiderivative of cos(ex ) is ex
.
Z x
6. Suppose F (x) = sin(t2 )dt. What is the instantaneous rate of change of
7
F (x) with respect to x?

77
I NTEGRATION 1.3 T HE F UNDAMENTAL T HEOREM OF C ALCULUS

Z x
7. Suppose F (x) = e1/t dt. What is the slope of the tangent line to y =
2
F (x) when x = 3?

8. Suppose F 0 (x) = f (x). Give two different antiderivatives of f (x).

9. In Question 1.1.8.45, Section 1.1, we found that
Z a√
π 1 1 √
1 − x2 dx = − arccos(a) + a 1 − a2 .
0 4 2 2

1 √ √
 
d π 1
a Verify that − arccos(a) + a 1 − a = 1 − a2 .
2
da 4 2 2
√
b Find a function F (x) that satisfies F 0 (x) = 1 − x2 and F (0) = π.
10. Evaluate the following integrals using the Fundamental Theorem of Calcu-
lus Part 2, or explain why it does not apply.
Z π
a cos xdx.
−π
Z π
b sec2 xdx.
−π

0
1
Z
c dx.
−2 x+1
11. RAs in the proof of the Fundamental Theorem of Calculus, let F (x) =
x
a
f (t)dt. In the diagram below, shade the area corresponding to F (x +
h) − F (x).

y = f (t)

t
a x x+h

Z x
12. Let F (x) = f (t)dt, where f (t) is shown in the graph below, and
0
0 ≤ x ≤ 4.
a Is F (0) positive, negative, or zero?

78
I NTEGRATION 1.3 T HE F UNDAMENTAL T HEOREM OF C ALCULUS

b Where is F (x) increasing and where is it decreasing?

y = f (t)

t
1 2 3 4

Z 0
13. Let G(x) = f (t)dt, where f (t) is shown in the graph below, and 0 ≤ x ≤
x
4.

a Is G(0) positive, negative, or zero?

b Where is G(x) increasing and where is it decreasing?

y = f (t)

t
1 2 3 4

Z x
14. Let F (x) = tdt. Using the definition of the derivative, find F 0 (x).
a
Z x
15. Give a continuous function f (x) so that F (x) = f (t)dt is a constant.
0
d
16. Evaluate and simplify dx {x log(ax)−x}, where a is some constant and log(x)
is the logarithm base e. What antiderivative does this tell you?

79
I NTEGRATION 1.3 T HE F UNDAMENTAL T HEOREM OF C ALCULUS

d
17. Evaluate and simplify dx
{ex (x3 − 3x2 + 6x − 6)}. What antiderivative does
this tell you?
d
 √
18. Evaluate and simplify dx log x + x2 + a2 , where a is some constant.
What antiderivative does this tell you?
d np √ √
o
19. Evaluate and simplify x(a + x) − a log x + a + x , where a is
dx
some constant. What antiderivative does this tell you?

Exercises — Stage 2
Z 2
20. *. Evaluate x3 + sin x) dx.
0

2
x2 + 2
Z
21. *. Evaluate dx.
1 x2
1
Z
22. Evaluate dx.
1 + 25x2
1
Z
23. Evaluate √ dx.
2 − x2
Z
24. Evaluate tan2 xdx.
Z
25. Evaluate 3 sin x cos xdx.
Z
26. Evaluate cos2 xdx.

27. *. If
Z x Z 0
F (x) = log(2 + sin t) dt and G(y) = log(2 + sin t) dt
0 y

find F 0 π
and G0 π



2 2
.
Z x
2
28. *. Let f (x) = 100(t2 − 3t + 2)e−t dt. Find the interval(s) on which f
1
is increasing.

cos x
1
Z
29. *. If F (x) = dt, find F 0 (x).
0 t3 +6

Z 1+x4
2
30. 0
*. Compute f (x) where f (x) = et dt.
0

80
I NTEGRATION 1.3 T HE F UNDAMENTAL T HEOREM OF C ALCULUS
Z sin x 
d 6
31. *. Evaluate (t + 8)dt .
dx 0
Z x3  
−t πt
32. *. Let F (x) = e sin dt. Calculate F 0 (1).
0 2
Z 0 
d dt
33. *. Find 3
.
du cos u 1 + t
Z x
2
34. *. Find f (x) if x = 1 + f (t)dt.
1
Z x
35. *. If x sin(πx) = f (t) dt where f is a continuous function, find f (4).
0
Z x2 Z 0
−t 2
36. *. Consider the function F (x) = e dt + e−t dt.
0 −x

a Find F 0 (x).

b Find the value of x for which F (x) takes its minimum value.
Z x
37. *. If F (x) is defined by F (x) = esin t dt, find F 0 (x).
x4 −x3

Z −x2 
d t


38. *. Evaluate dx cos e dt .
x5

Z ex √
39. *. Differentiate sin t dt for 0 < x < log π.
x
(
5
3 if x ≤ 3
Z
40. *. Evaluate f (x) dx, where f (x) = .
1 x if x ≥ 3

Exercises — Stage 3
Z 2
41. 0
*. If f (1) = 2 and f (2) = 3, find
0
f 0 (x)f 00 (x) dx.
1

42. *. A car traveling at 30 m/s applies its brakes at time t = 0, its velocity
(in m/s) decreasing according to the formula v(t) = 30 − 10t. How far
does the car go before it stops?

Z 2x−x2
0
log 1 + et dt. Does f (x) have an


43. *. Compute f (x) where f (x) =
0
absolute maximum? Explain.

81
I NTEGRATION 1.3 T HE F UNDAMENTAL T HEOREM OF C ALCULUS

x2 −2x
dt
Z
44. *. Find the minimum value of 1 + t4
. Express your answer as an
0
integral.
Z x2 √
45. *. Define the function F (x) = sin( t) dt on the interval 0 < x < 4. On
0
this interval, where does F (x) have a maximum?

n  
πX jπ
46. *. Evaluate lim sin by interpreting it as a limit of Riemann
n→∞ n n
j=1
sums.
n
1X 1
47. lim
*. Use Riemann sums to evaluate the limit n→∞ j .
n j=1 1 + n

Z x
48. Below is the graph of y = f (t), −5 ≤ t ≤ 5. Define F (x) = f (t)dt for
0
any x in [−5, 5]. Sketch F (x).
y

y = f (x)
x
−5 −3 −1 1 3 5

Z x3 +1
3
49. *. Define f (x) = x
3
et dt.
0

a Find a formula for the derivative f 0 (x). (Your formula may include in
integral sign.)

b Find the equation of the tangent line to the graph of y = f (x) at x =

−1.
R
50. Two students calculate f (x)dx for some function f (x).
• Student A calculates f (x)dx = tan2 x + x + C
R

• Student B calculates f (x)dx = sec2 x + x + C

R

d
• It is a fact that dx
{tan2 x} = f (x) − 1

82
 I NTEGRATION 1.4 S UBSTITUTION

Who ended up with the correct answer?

Z x
51. Let F (x) = x3 sin(t)dt.
0

a Evaluate F (3).

b What is F 0 (x)?

52. Let f (x) be an even function, defined everywhere, and let F (x) be an an-
tiderivative of f (x). Is F (x) even, odd, or not necessarily either one? (You
may use your answer from Section 1.2, Question 1.2.3.20. )

1.4q Substitution

1.4.1 tt Substitution

In the previous section we explored the fundamental theorem of calculus and the
link it provides between definite integrals and antiderivatives. Indeed, integrals
with simple integrands are usually evaluated via this link. In this section we start to
explore methods for integrating more complicated integrals. We have already seen
— via Theorem 1.2.1 — that integrals interact very nicely with addition, subtraction
and multiplication by constants:
Z b Z b Z b
(Af (x) + Bg(x)) dx = A f (x)dx + B g(x)dx
a a a

for A, B constants. By combining this with the list of indefinite integrals in Theo-
rem 1.3.17, we can compute integrals of linear combinations of simple functions. For
example
Z 4 Z 4 Z 4 Z 4
2
x x
x2 dx


e − 2 sin x + 3x dx = e dx − 2 sin xdx + 3
1 1 1 1
 3
 4
x
= ex + (−2) · (− cos x) + 3 and so on
3 1

Of course there are a great many functions that can be approached in this way, how-
ever there are some very simple examples that cannot.
x
Z Z Z
x
sin(πx)dx xe dx 2
dx
x − 5x + 6

83
I NTEGRATION 1.4 S UBSTITUTION

In each case the integrands are not linear combinations of simpler functions; in order
to compute them we need to understand how integrals (and antiderivatives) interact
with compositions, products and quotients. We reached a very similar point in our
differential calculus course where we understood the linearity of the derivative,
d df dg
(Af (x) + Bg(x)) = A +B ,
dx dx dx
but had not yet seen the chain, product and quotient rules 1 . While we will develop
tools to find the second and third integrals in later sections, we should really start
with how to integrate compositions of functions.
It is important to state up front, that in general one cannot write down the integral
of the composition of two functions — even if those functions are simple. This is
not because the integral does not exist. Rather it is because the integral cannot be
written down as a finite combination of the standard functions we know. A very
good example of this, which we encountered in Example 1.3.4, is the composition of
ex and −x2 . Even though we know
1
Z Z
x x
e dx = e + C and −x2 dx = − x3 + C
3
there is no simple function that is equal to the indefinite integral
Z
2
e−x dx.

even though the indefinite integral exists. In this way integration is very different
from differentiation.
With that caveat out of the way, we can introduce the substitution rule. The
substitution rule is obtained by antidifferentiating the chain rule. In some sense it is
the chain rule in reverse. For completeness, let us restate the chain rule:

Theorem 1.4.1 The chain rule.

Let F (u) and u(x) be differentiable functions and form their composition
F (u(x)). Then

d
F u(x) = F 0 u(x) · u0 (x)



dx
Equivalently, if y(x) = F (u(x)), then

dy dF du
= · .
dx du dx

1 If your memory of these rules is a little hazy then you really should go back and revise them
before proceeding. You will definitely need a good grasp of the chain rule for what follows in
this section.

84
I NTEGRATION 1.4 S UBSTITUTION

Consider a function f (u), which has antiderivative F (u). Then we know that
Z Z
f (u)du = F 0 (u)du = F (u) + C

Now take the above equation and substitute into it u = u(x) — i.e. replace the
variable u with any (differentiable) function of x to get
Z
f (u)du = F (u(x)) + C
u=u(x)

But now the right-hand side is a function of x, so we can differentiate it with respect
to x to get
d
F (u(x)) = F 0 (u(x)) · u0 (x)
dx
This tells us that F (u(x)) is an antiderivative of the function F 0 (u(x))·u0 (x) = f (u(x))u0 (x).
Thus we know
Z Z

0

f u(x) · u (x) dx = F u(x) + C = f (u) du
u=u(x)

This is the substitution rule for indefinite integrals.

Theorem 1.4.2 The substitution rule — indefinite integral version.

For any differentiable function u(x):

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

In order to apply the substitution rule successfully we will have to write the in-
tegrand in the form f (u(x)) · u0 (x). To do this we need to make a good choice of the
function u(x); after that it is not hard to then find f (u) and u0 (x). Unfortunately there
is no one strategy for choosing u(x). This can make applying the substitution rule
more art than science 2 . Here we suggest two possible strategies for picking u(x):
1 Factor the integrand and choose one of the factors to be u0 (x). For this to work,
you must be able to easily find the antiderivative of the chosen factor. The
antiderivative will be u(x).
2 Look for a factor in the integrand that is a function with an argument
that is
more complicated than just “x”. That factor will play the role of f u(x) Choose
u(x) to be the complicated argument.
Here are two examples which illustrate each of those strategies in turn.

2 Thankfully this does become easier with experience and we recommend that the reader read
some examples and then practice a LOT.

85
I NTEGRATION 1.4 S UBSTITUTION

9 sin8 (x) cos(x)dx.

R
Example 1.4.3

Consider the integral

Z
9 sin8 (x) cos(x)dx

We want to massage this into the form of the integrand in the substitution rule —
namely f (u(x)) · u0 (x). Our integrand can be written as the product of the two factors

9 sin8 (x) · cos(x)

| {z } | {z }
first factor second factor

and we start by determining (or guessing) which factor plays the role of u0 (x). We
can choose u0 (x) = 9 sin8 (x) or u0 (x) = cos(x).

• If we choose u0 (x) = 9 sin8 (x), then antidifferentiating this to find u(x) is really
not very easy. So it is perhaps better to investigate the other choice before
proceeding further with this one.

• If we choose u0 (x) = cos(x), then we know (Theorem 1.3.17) that u(x) = sin(x).
This also works nicely because it makes the other factor simplify quite a bit
9 sin8 (x) = 9u8 . This looks like the right way to go.

So we go with the second choice. Set u0 (x) = cos(x), u(x) = sin(x), then
Z Z
9 sin (x) cos(x)dx = 9u(x)8 · u0 (x)dx
8

Z
= 9u8 du by the substitution rule
u=sin(x)

We are now left with the problem of antidifferentiating a monomial; this we can do
with Theorem 1.3.17.

= u9 + C


u=sin(x)
9
= sin (x) + C

Note that 9 sin8 (x) cos(x) is a function of x. So our answer, which is the indefinite
integral of 9 sin8 (x) cos(x), must also be a function of x. This is why we have substi-
tuted u = sin(x) in the last step of our solution — it makes our solution a function of
x.

86
I NTEGRATION 1.4 S UBSTITUTION

3x2 cos(x3 )dx.

R
Example 1.4.4

Evaluate the integral

Z
3x2 cos(x3 )dx

Solution: Again we are going to use the substitution rule and helpfully our integrand
is a product of two factors

3x2 · cos(x3 )
|{z} | {z }
first factor second factor

The second factor, cos x3 is a function, namely cos, with a complicated argument,

namely x3 . So we try u(x) = x3 . Then u0 (x) = 3x2 , which is the other factor in the
integrand. So the integral becomes
Z Z
3x cos(x )dx = u0 (x) cos u(x) dx
2 3


just swap order of factors
Z
= cos u(x) u0 (x)dx


by the substitution rule
Z
= cos(u)du
u=x3

= (sin(u) + C) using Theorem 1.3.17)

u=x3
= sin(x3 ) + C

One more — we’ll use this to show how to use the substitution rule with definite
integrals.

R1
Example 1.4.5 0
ex sin(ex )dx.

Compute
Z 1
ex sin ex dx.


0

Solution: Again we use the substitution rule.

• The integrand is again the product of two factors and we can choose u0 (x) = ex
or u0 (x) = sin(ex ).
• If we choose u0 (x) = ex then u(x) = ex and the other factor becomes sin(u) —
this looks promising. Notice that if we applied the other strategy of looking for
a complicated argument then we would arrive at the same choice.

87
I NTEGRATION 1.4 S UBSTITUTION

• So we try u0 (x) = ex and u(x) = ex . This gives (if we ignore the limits of
integration for a moment)
Z Z
e sin e dx = sin u(x) u0 (x)dx
x x



apply the substitution rule
Z
= sin(u)du
u=ex

= (− cos(u) + C)
u=ex
= − cos ex + C

• But what happened to the limits of integration? We can incorporate them now.
We have just shown that the indefinite integral is − cos(ex ), so by the funda-
mental theorem of calculus
Z 1

1
ex sin ex dx = − cos ex 0


0
= − cos(e1 ) − (− cos(e0 ))
= − cos(e) + cos(1)
Example 1.4.5

0if F (u)
Theorem 1.4.2, the substitution rule for
indefinite integrals, tells us that
is any antiderivative for f (u), then F u(x) is an antiderivative for f u(x) u (x). So
the fundamental theorem of calculus gives us

x=b
Z b
f u(x) u0 (x) dx = F u(x)


a

x=a

= F u(b) − F u(a)
Z u(b)
= f (u) du since F (u) is an antiderivative for f (u)
u(a)

and we have just found

Theorem 1.4.6 The substitution rule — definite integral version.

For any differentiable function u(x):

Z b Z u(b)
0
f (u(x))u (x)dx = f (u)du
a u(a)

Notice that to get from the integral on the left hand side to the integral on the
right hand side you

88
I NTEGRATION 1.4 S UBSTITUTION

• substitute 3 u(x) → u and u0 (x)dx → du,

• set the lower limit for the u integral to the value of u (namely u(a)) that corre-
sponds to the lower limit of the x integral (namely x = a), and
• set the upper limit for the u integral to the value of u (namely u(b)) that corre-
sponds to the upper limit of the x integral (namely x = b).
Also note that we now have two ways to evaluate definite integrals of the form
Rb
0
a
f u(x) u (x) dx.
• We can find the indefinite integral f u(x) u0 (x) dx, using Theorem 1.4.2, and
R

then evaluate the result between x = a and x = b. This is what was done in
Example 1.4.5.
• ROr we can apply Theorem 1.4.2. This entails finding the indefinite integral
f (u) du and evaluating the result between u = u(a) and u = u(b). This is
what we will do in the following example.

R1
Example 1.4.7 0
x2 sin(x3 + 1)dx.

Compute
Z 1
x2 sin x3 + 1 dx


0

Solution:
• In this example the integrand is already neatly factored into two pieces. While
we could deploy either of our two strategies, it is perhaps easier in this case to
choose u(x) by looking for a complicated argument.
• The second factor of the integrand is sin x3 + 1 , which is the function sin

evaluated at x3 + 1. So set u(x) = x3 + 1, giving u0 (x) = 3x2 and f (u) = sin(u)

• The first factor of the integrand is x2 which is not quite u0 (x), however we
can easily massage the integrand into the required form by multiplying and
dividing by 3:

1
x2 sin x3 + 1 = · 3x2 · sin x3 + 1 .


3
• We want this in the form of the substitution rule, so we do a little massaging:
Z 1 Z 1
2 3 1
· 3x2 · sin x3 + 1 dx



x sin x + 1 dx =
0 0 3

3 A good way to remember this last step is that we replace dudx dx by just du — which looks like
du 
we cancelled out the dx terms: dx
dx = du. While using “cancel the dx” is a good mnemonic

(memory aid), you should not think of the derivative du
dx as a fraction — you are not dividing du
by dx.

89
I NTEGRATION 1.4 S UBSTITUTION

1
1
Z
sin x3 + 1 · 3x2 dx


=
3 0
by Theorem 1.2.1(c)

• Now we are ready for the substitution rule:

1 1
1 1
Z Z
3 2
sin x3 + 1 · |{z}
3x2 dx



sin x + 1 · 3x dx =
3 0 3 0 | {z } 0 =u (x)
=f (u(x))

Now set u(x) = x3 + 1 and f (u) = sin(u)

1 1
Z
= f (u(x))u0 (x)dx
3 0
1 u(1)
Z
= f (u)du by the substitution rule
3 u(0)
1 2
Z
= sin(u)du since u(0) = 1 and u(1) = 2
3 1
1 2
= − cos(u) 1
3
1

= − cos(2) − (− cos(1))
3
cos(1) − cos(2)
= .
3
Example 1.4.7

There is another, and perhaps easier, way to view the manipulations in the previous
example. Once you have chosen u(x) you

• make the substitution u(x) → u,

1
• replace dx → du.
u0 (x)
In so doing, we take the integral
b u(b)
1
Z Z
0
f (u(x)) · u (x)dx = f (u) · u0 (x) · du
a u(a) u0 (x)
Z u(b)
= f (u)du exactly the substitution rule
u(a)

but we do not have to manipulate the integrand so as to make u0 (x) explicit. Let us
redo the previous example by this approach.

90
I NTEGRATION 1.4 S UBSTITUTION

Example 1.4.8 Example 1.4.7 revisited.

Compute the integral

Z 1
x2 sin x3 + 1 dx


0

Solution:

• We have already observed that one factor of the integrand is sin x3 + 1 , which

is sin evaluated at x3 + 1. Thus we try setting u(x) = x3 + 1.

1
• This makes u0 (x) = 3x2 , and we replace u(x) = x3 + 1 → u and dx → u0 (x)
du =
1
3x2
du:
Z 1 Z u(1)
1
2 3
x2 sin x3 + 1


x sin x + 1 dx = du
0 u(0) | {z } 3x2
=sin(u)
2
x2
Z
= sin(u) du
3x2
Z1 2
1
= sin(u)du
1 3
2
1
Z
= sin(u)du
3 1

which is precisely the integral we found in Example 1.4.7.

Example 1.4.9 Some more substitutions.

Compute the indefinite integrals

√
Z Z
2x + 1dx and e3x−2 dx

Solution:
• Starting with the first integral, we see that it is not too hard to spot√the compli-
cated argument. If we set u(x) = 2x + 1 then the integrand is just u.

• Hence we substitute 2x + 1 → u and dx → u01(x) du = 12 du:

√ √ 1
Z Z
2x + 1dx = u du
2
1
Z
= u1/2 du
2

91
I NTEGRATION 1.4 S UBSTITUTION

 
2 3/2 1
= u · +C
3 2 u=2x+1
1
= (2x + 1)3/2 + C
3

• We can evaluate the second integral in much the same way. Set u(x) = 3x − 2
and replace dx by u01(x) du = 13 du:

1
Z Z
3x−2
e dx = eu du
3
 
1 u
= e +C
3 u=3x−2
1
= e3x−2 + C
3
Example 1.4.9

This last example illustrates that substitution can be used to easily deal with argu-
ments of the form ax+b, i.e. that are linear functions of x, and suggests the following
theorem.

Theorem 1.4.10
Let F (u) be an antiderivative of f (u) and let a, b be constants. Then

1
Z
f (ax + b)dx = F (ax + b) + C
a

Proof. We can show this using the substitution rule. Let u(x) = ax + b so
0
u (x) = a, then

1
Z Z
f (ax + b)dx = f (u) · 0 du
u (x)
1
Z
= f (u)du
a
1
Z
= f (u)du since a is a constant
a
1
= F (u) + C since F (u) is an antiderivative of f (u)
a u=ax+b
1
= F (ax + b) + C.
a


92
I NTEGRATION 1.4 S UBSTITUTION

Now we can do the following example using the substitution rule or the above
theorem:
Rπ
Example 1.4.11 0
2
cos(3x)dx.
Rπ
Compute 0
2
cos(3x)dx.

• In this example we should set u = 3x, and substitute dx → u01(x) du = 13 du.

When we do this we also have to convert the limits of the integral: u(0) = 0
and u(π/2) = 3π/2. This gives
π 3π
1
Z Z
2 2
cos(3x)dx = cos(u) du
0 0 3
  3π2
1
= sin(u)
3 0
sin(3π/2) − sin(0)
=
3
−1 − 0 1
= =− .
3 3

• We can also do this example more directly using the above theorem. Since
sin(x) is an antiderivative of cos(x), Theorem 1.4.10 tells us that sin(3x)
3
is an
antiderivative of cos(3x). Hence
π  π
sin(3x) 2
Z
2
cos(3x)dx =
0 3 0
sin(3π/2) − sin(0)
=
3
1
=− .
3

The rest of this section is just more examples of the substitution rule. We rec-
ommend that you after reading these that you practice many examples by yourself
under exam conditions.
R1
Example 1.4.12 0
x2 sin(1 − x3 )dx.

This integral looks a lot like that of Example 1.4.7. It makes sense to try u(x) = 1 − x3
since it is the argument of sin(1 − x3 ). We

• substitute u = 1 − x3 and
1 1
• replace dx with u0 (x)
du = −3x2
du,

93
I NTEGRATION 1.4 S UBSTITUTION

• when x = 0, we have u = 1 − 03 = 1 and

• when x = 1, we have u = 1 − 13 = 0.

So
1 0
1
Z Z
2 3
x2 sin(u) ·


x sin 1 − x · dx = du
0 −3x2
Z1 0
1
= − sin(u)du.
1 3

Note that the lower limit of the u-integral, namely 1, is larger than the upper limit,
which is 0. There is absolutely nothing wrong with that. We can simply evaluate the
u-integral in the normal way. Since − cos(u) is an antiderivative of sin(u):
0
cos(u)
=
3 1
cos(0) − cos(1)
=
3
1 − cos(1)
= .
3
Example 1.4.12

R1 1
Example 1.4.13 0 (2x+1)3
dx.
R1 1
Compute 0 (2x+1) 3 dx.

We could do this one using Theorem 1.4.10, but its not too hard to do without. We
can think of the integrand as the function “one over a cube” with the argument 2x+1.
So it makes sense to substitute u = 2x + 1. That is

• set u = 2x + 1 and
1
• replace dx → u0 (x)
du = 21 du.

• When x = 0, we have u = 2 × 0 + 1 = 1 and

• when x = 1, we have u = 2 × 1 + 1 = 3.

So
1 3
1 1 1
Z Z
dx = · du
0 (2x + 1)3 1 u3 2
3
1
Z
= u−3 du
2 1

94
I NTEGRATION 1.4 S UBSTITUTION

3
1 u−2

=
2 −2 1
 
1 1 1 1 1
= · − ·
2 −2 9 −2 1
 
1 1 1 1 8
= − = ·
2 2 18 2 18
2
=
9
Example 1.4.13

R1 x
Example 1.4.14 0 1+x2
dx.
R1 x
Evaluate 0 1+x2
dx.
Solution:
1
• The integrand can be rewritten as x · 1+x 2 . This second factor suggests that we

should try setting u = 1 + x2 — and so we interpret the second factor as the

function “one over” evaluated at argument 1 + x2 .

• With this choice we

◦ set u = 1 + x2 ,
1
◦ substitute dx → 2x
du, and
◦ translate the limits of integration: when x = 0, we have u = 1 + 02 = 1 and
when x = 1, we have u = 1 + 12 = 2.

• The integral then becomes

1 2
x x 1
Z Z
dx = du
0 1 + x2 1 u 2x
Z 2
1
= du
1 2u
1 2
= log |u| 1
2
log 2 − log 1 log 2
= = .
2 2

Remember that we are using the notation “log” for the natural logarithm, i.e. the
logarithm with base e. You might also see it written as “ln x”, or with the base made
explicit as “loge x”.

95
I NTEGRATION 1.4 S UBSTITUTION

x3 cos x4 + 2 dx.
R

Example 1.4.15

Compute the integral x3 cos x4 + 2 dx.

R

Solution:

• The integrand is the product of cos evaluated at the argument x4 + 2 times x3 ,

which aside from a factor of 4, is the derivative of the argument x4 + 2.
1 1
• Hence we set u = x4 + 2 and then substitute dx → u0 (x)
du = 4x3
du.

• Before proceeding further, we should note that this is an indefinite integral so

we don’t have to worry about the limits of integration. However we do need
to make sure our answer is a function of x — we cannot leave it as a function
of u.

• With this choice of u, the integral then becomes

1
Z Z
x cos x + 2 dx = x3 cos(u) 3 du
3 4


4x u=x4 +2
1
Z
= cos(u)du
4 u=x4 +2
 
1
= sin(u) + C
4 u=x4 +2
1
= sin(x4 + 2) + C.
4

The next two examples are more involved and require more careful thinking.

R√
Example 1.4.16 1 + x2 x3 dx.
R√
Compute 1 + x2 x3 dx.

• An obvious choice of u is the argument inside the square root. So substitute

1
u = 1 + x2 and dx → 2x du.

• When we do this we obtain

Z √
√ 1
Z
3
2
1 + x · x dx = u · x3 · du
2x
1√
Z
= u · x2 du
2
Unlike all our previous examples, we have not cancelled out all of the x’s from
the integrand. However before we do the integral with respect to u, the inte-

96
I NTEGRATION 1.4 S UBSTITUTION

grand must be expressed solely in terms of u — no x’s are allowed. (Look that
integrand on the right hand side of Theorem 1.4.2.)

• But all is not lost. We can rewrite the factor x2 in terms of the variable u. We
know that u = 1 + x2 , so this means x2 = u − 1. Substituting this into our
integral gives
Z √
1√
Z
3
2
1 + x · x dx = u · x2 du
2
1√
Z
= u · (u − 1)du
2
1
Z
u3/2 − u1/2 du


=
2
 
1 2 5/2 2 3/2
= u − u +C
2 5 3 u=x2 +1
 
1 5/2 1 3/2
= u − u +C
5 3 u=x2 +1
1 1
= (x2 + 1)5/2 − (x2 + 1)3/2 + C.
5 3
Oof!

• Don’t forget that you can always check the answer by differentiating:
 
d 1 2 5/2 1 2 3/2
(x + 1) − (x + 1) + C
dx 5 3
   
d 1 2 5/2 d 1 2 3/2
= (x + 1) − (x + 1)
dx 5 dx 3
1 5 1 3
= · 2x · · (x2 + 1)3/2 − · 2x · · (x2 + 1)1/2
5 2 3 2
= x(x2 + 1)3/2 − x(x2 + 1)1/2
 √
= x (x2 + 1) − 1 · x2 + 1

√
= x3 x2 + 1.

which is the original integrand X.

Example 1.4.16

R
Example 1.4.17 tan xdx.
R
Evaluate the indefinite integral tan(x)dx.
Solution:
• At first glance there is nothing to manipulate here and so very little to go on.

97
 I NTEGRATION 1.4 S UBSTITUTION

sin x sin x
R
However we can rewrite tan x as cos x
, making the integral cos x
dx. This gives
us more to work with.

• Now think of the integrand as being the product cos1 x · sin x. This suggests that
we set u = cos x and that we interpret the first factor as the function “one over”
evaluated at u = cos x.
1
• Substitute u = cos x and dx → − sin x
du to give:

sin x sin x 1
Z Z
dx = du
cos x u − sin x u=cos x
1
Z
= − du
u u=cos x
= − log | cos x| + C and if we want to go further
1
= log +C
cos x
= log | sec x| + C.
Example 1.4.17

1.4.2 tt Exercises

Recall that we are using log x to denote the logarithm of x with base e. In other
courses it is often denoted ln x.
Exercises — Stage 1

1.
Z Z
x x
a True or False: sin(e ) · e dx = sin(u)du = − cos(ex ) + C
u=ex
Z 1 Z 1
x x
b True or False: sin(e ) · e dx = sin(u)du = 1 − cos(1)
0 0

2. Is the following reasoning sound? If not, fix it.

Z
Problem: Evaluate (2x + 1)2 dx.

Work: We use the substitution u = 2x + 1. Then:

Z Z
(2x + 1) dx = u2 du
2

98
I NTEGRATION 1.4 S UBSTITUTION

1
= u3 + C
3
1
= (2x + 1)3 + C
3

3. Is the following reasoning sound? If not, fix it.

Z π
cos(log t)
Problem: Evaluate dt.
1 t
Work: We use the substitution u = log t, so du = 1t dt. Then:
Z π Z π
cos(log t)
dt = cos(u)du
1 t 1
= sin(π) − sin(1) = − sin(1) .
4. Is the following reasoning sound? If not, fix it.
Z π/4
Problem: Evaluate x tan(x2 )dx.
0
Work: We begin with the substitution u = x2 , du = 2xdx:
π/4 π/4
1
Z Z
2
x tan(x )dx = tan(x2 ) · 2xdx
0 0 2
π 2 /16
1
Z
= tan udu
0 2
π 2 /16
1 sin u
Z
= du
2 0 cos u

Now we use the substitution v = cos u, dv = − sin udu:

cos(π 2 /16)
1 1
Z
= − dv
2 cos 0 v
Z cos(π2 /16)
1 1
=− dv
2 1 v
1 2 /16)
= − [log |v|]cos(π
1
2
1
= − log cos(π 2 /16) − log(1)



2
1
= − log cos(π 2 /16)


2

5. *. What is the integralZ that results when the substitution u = sin x is

π/2
applied to the integral f (sin x) dx?
0

99
I NTEGRATION 1.4 S UBSTITUTION

6. Let f and g be functions that are continuous and differentiable everywhere.

Simplify Z
f 0 (g(x))g 0 (x)dx − f (g(x)).

Exercises — Stage 2
Z 1
2 2
7. *. Use substitution to evaluate xex cos(ex ) dx.
0
Z 8
8. *. Let f (t) be any function for which f (t) dt = 1. Calculate the integral
Z 2 1

x2 f (x3 ) dx.
1

x2
Z
9. *. Evaluate dx.
(x3 + 1)101

e4
dx
Z
10. *. Evaluate x log x
.
e

π/2
cos x
Z
11. *. Evaluate dx.
0 1 + sin x
Z π/2
12. *. Evaluate cos x · (1 + sin2 x) dx.
0
Z 3
2 −x
13. *. Evaluate (2x − 1)ex dx.
1

(x2 − 4)x
Z
14. *. Evaluate √ dx.
4 − x2
√
e log x
Z
15. Evaluate √ dx .
2x log x

Exercises — Stage 3 Questions 18 through 22 can be solved by substitution, but

it may not be obvious which substitution will work. In general, when evaluating
integrals, it is not always immediately clear which methods are appropriate. If this
happens to you, don’t despair, and definitely don’t give up! Just guess a method
and try it. Even if it fails, you’ll probably learn something that you can use to make
a better guess. 4

4 This is also pretty decent life advice.

100
 I NTEGRATION 1.5 A REA BETWEEN CURVES
Z 2
2
16. *. Calculate xex dx.
−2
n
j2
 
X j
17. lim
*. Calculate n→∞ 2
sin 1 + 2 .
j=1
n n

1
u3
Z
18. Evaluate du.
0 u2 + 1
Z
19. Evaluate tan3 θ dθ.

1
Z
20. Evaluate dx
ex + e−x
Z 1 √
21. Evaluate (1 − 2x) 1 − x2 dx
0

Z
22. Evaluate tan x · log (cos x) dx

n  2
X j j
23. lim
*. Evaluate n→∞ 2
cos .
j=1
n n2
n
r
X j j2
24. *. Calculate lim 2
1 + 2
.
n→∞
j=1
n n

25. Using Riemann sums, prove that

Z b Z 2b
2f (2x)dx = f (x)dx
a 2a

1.5q Area between curves

1.5.1 tt Area between curves

Before we continue our exploration of different methods for integrating functions,

we have now have sufficient tools to examine some simple applications of definite

101
I NTEGRATION 1.5 A REA BETWEEN CURVES

integrals. One of the motivations for our definition of “integral” was the problem
of finding the area between some curve and the x-axis for x running between two
specified values. More precisely
Z b
f (x)dx
a

is equal to the signed area between the the curve y = f (x), the x-axis, and the vertical
lines x = a and x = b.
We found the area of this region by approximating it by the union of tall thin
rectangles, and then found the exact area by taking the limit as the width of the
approximating rectangles went to zero. We can use the same strategy to find areas
of more complicated regions in the xy-plane.
As a preview of the material to come, let f (x) > g(x) > 0 and a < b and suppose
that we are interested in the area of the region

S1 = (x, y) a ≤ x ≤ b , g(x) ≤ y ≤ f (x)

that is sketched in the left hand figure below.

Rb
We already know that a
f (x) dx is the area of the region

S2 = (x, y) a ≤ x ≤ b , 0 ≤ y ≤ f (x)
Rb
sketched in the middle figure above and that a g(x) dx is the area of the region

S3 = (x, y) a ≤ x ≤ b , 0 ≤ y ≤ g(x)

sketched in the right hand figure above. Now the region S1 of the left hand figure
can be constructed by taking the region S2 of center figure and removing from it the
region S3 of the right hand figure. So the area of S1 is exactly
Z b Z b Z b

f (x) dx − g(x) dx = f (x) − g(x) dx
a a a

This computation depended on the assumption that f (x) > g(x) and, in particular,
that the curves y = g(x) and y = f (x) did not cross. If they do cross, as in this figure

102
I NTEGRATION 1.5 A REA BETWEEN CURVES

then we have to be a lot more careful. The idea is to separate the domain of
integration depending on where f (x) − g(x) changes sign — i.e. where the curves
intersect. We will illustrate this in Example 1.5.5 below.
Let us start with an example that makes the link to Riemann sums and definite
integrals quite explicit.

Example 1.5.1 The area between y = 4 − x2 and y = x.

Find the area bounded by the curves y = 4 − x2 , y = x, x = −1 and x = 1.

Solution:

• Before we do any calculus, it is a very good idea to make a sketch of the area
in question. The curves y = x, x = −1 and x = 1 are all straight lines, while
the curve y = 4 − x2 is a parabola whose apex is at (0, 4) and then curves down
(because of the minus sign in −x2 ) with x-intercepts at (±2, 0). Putting these
together gives

2 2
√ y
= 4 − x and y = x intersect when 4 − x =2 x, namely
Notice that the curves
1
when x = 2 −1 ± 17 ≈ 1.56, −2.56. Hence the curve y = 4 − x lies above
the line y = x for all −1 ≤ x ≤ 1.

103
I NTEGRATION 1.5 A REA BETWEEN CURVES

• We are to find the area of the shaded region. Each point (x, y) in this shaded
region has −1 ≤ x ≤ 1 and x ≤ y ≤ 4 − x2 . When we were defining the
integral (way back in Definition 1.1.9) we used a and b to denote the smallest
and largest allowed values of x; let’s do that here too. Let’s also use B(x) to
denote the bottom curve (i.e. to denote the smallest allowed value of y for a
given x) and use T (x) to denote the top curve (i.e. to denote the largest allowed
value of y for a given x). So in this example

a = −1 b=1 B(x) = x T (x) = 4 − x2

and the shaded region is


(x, y) a ≤ x ≤ b, B(x) ≤ y ≤ T (x)

• We use the same strategy as we used when defining the integral in Section 1.1.4:

◦ Pick a natural number n (that we will later send to infinity), then

b−a
◦ subdivide the region into n narrow slices, each of width ∆x = n
.
◦ For each i = 1, 2, · · · , n, slice number i runs from x = xi−1 to x = xi , and
we approximate its area by the area of a rectangle. We pick a number x∗i
between xi−1 and xi and approximate the slice by a rectangle whose top is
at y = T (x∗i ) and whose bottom is at y = B(x∗i ).
◦ Thus the area of slice i is approximately T (x∗i ) − B(x∗i ) ∆x (as shown in
 

the figure below).

• So the Riemann sum approximation of the area is

n
X
T (x∗i ) − B(x∗i ) ∆x
 
Area ≈
i=1

104
I NTEGRATION 1.5 A REA BETWEEN CURVES

• By taking the limit as n → ∞ (i.e. taking the limit as the width of the rectangles
goes to zero), we convert the Riemann sum into a definite integral (see Defi-
nition 1.1.9) and at the same time our approximation of the area becomes the
exact area:
Xn Z b
∗ ∗
   
lim T (xi ) − B(xi ) ∆x = T (x) − B(x) dx
n→∞ a
i=1
Riemann sum → integral
Z 1
(4 − x2 ) − x dx
 
=
−1
Z 1
4 − x − x2 dx
 
=
−1
1
x2 x3

= 4x − −
2 3
  −1 
1 1 1 1
= 4− − − −4 − +
2 3 2 3
24 − 3 − 2 −24 − 3 + 2
= −
6 6
19 25
= +
6 6
44 22
= = .
6 3
Example 1.5.1

Oof! Thankfully we generally do not need to go through the Riemann sum steps
to get to the answer. Usually, provided we are careful to check where curves intersect
and which curve lies above which, we can just jump straight to the integral
Z b  
Area = T (x) − B(x) dx.
a

So let us redo the above example.

Example 1.5.2 Example 1.5.1 revisited.

Find the area bounded by the curves y = 4 − x2 , y = x, x = −1 and x = 1.

Solution:

• We first sketch the region

105
I NTEGRATION 1.5 A REA BETWEEN CURVES

and verify a that y = T (x) = 4 − x2 lies above the curve y = B(x) = x on the
region −1 ≤ x ≤ 1.

• The area between the curves is then

Z b
 
Area = T (x) − B(x) dx
Za 1
4 − x − x2 dx
 
=
−1
1
x2 x3

= 4x − −
2 3 −1
19 25 44 22
= + = = .
6 6 6 3

a We should do this by checking where the curves intersect; that is by solving T (x) = B(x) and
seeing if any of the solutions lie in the range −1 ≤ x ≤ 1.

Example 1.5.2

Example 1.5.3 The area between y = x2 and y = 6x − 2x2 .

Find the area of the finite region bounded by y = x2 and y = 6x − 2x2 .

Solution: This is a little different from the previous question, since we are not given
bounding lines x = a and x = b — instead we have to determine the minimum and
maximum allowed values of x by determining where the curves intersect. Hence
our very first task is to get a good idea of what the region looks like by sketching it.
• Start by sketching the region:

106
I NTEGRATION 1.5 A REA BETWEEN CURVES

◦ The curve y = x2 is a parabola. The point on this parabola with the small-
est y-coordinate is (0, 0). As |x| increases, y increases so the parabola opens
upward.
◦ The curve y = 6x − 2x2 = −2(x2 − 3x) = −2(x − 32 )2 + 92 is also a parabola.
The point on this parabola with the largest value of y has x = 23 (so that
the negative term in −2(x − 32 )2 + 92 is zero). So the point with the largest
value of y is is ( 32 , 92 ). As x moves away from 32 , either to the right or to the
left, y decreases. So the parabola opens downward. The parabola crosses
the x-axis when 0 = 6x − 2x2 = 2x(3 − x). That is, when x = 0 and x = 3.
◦ The two parabolas intersect when x2 = 6x − 2x2 , or
3x2 − 6x = 0
3x(x − 2) = 0
So there are two points of intersection, one being x = 0, y = 02 = 0 and
the other being x = 2, y = 22 = 4.
◦ The finite region between the curves lies between these two points of in-
tersection.
This leads us to the sketch

• So on this region we have 0 ≤ x ≤ 2, the top curve is T (x) = 6x − x2 and the

bottom curve is B(x) = x2 . Hence the area is given by
Z b
 
Area = T (x) − B(x) dx
Za 2
(6x − 2x2 ) − (x2 ) dx
 
=
Z0 2
6x − 3x2 dx
 
=
0
 2 2
x x3
= 6 −3
2 3 0
2 3
= 3(2) − 2 = 4
Example 1.5.3
107
I NTEGRATION 1.5 A REA BETWEEN CURVES

Example 1.5.4 The area between y 2 = 2x + 6 and y = x − 1.

Find the area of the finite region bounded by y 2 = 2x + 6 and y = x − 1.

Solution: We show two different solutions to this problem. The first takes the ap-
proach we have in Example 1.5.3 but leads to messy algebra. The second requires a
little bit of thinking at the beginning but then is quite straightforward. Before we get
to that we should start by by sketching the region.

• The curve y 2 = 2x + 6, or equivalently x = 12 y 2 − 3 is a parabola. The point on

this parabola with the smallest x-coordinate has y = 0 (so that the positive term
in 12 y 2 − 3 is zero). So the point on this parabola with the smallest x-coordinate
is (−3, 0). As |y| increases, x increases so the parabola opens to the right.

• The curve y = x − 1 is a straight line of slope 1 that passes through x = 1, y = 0.

y2
• The two curves intersect when 2
− 3 = y + 1, or

y 2 − 6 = 2y + 2
y 2 − 2y − 8 = 0
(y + 2)(y − 4) = 0

So there are two points of intersection, one being y = 4, x = 4 + 1 = 5 and the

other being y = −2, x = −2 + 1 = −1.

• Putting this all together gives us the sketch

As noted above, we can find the area of this region by approximating it by a union of
narrow vertical rectangles, as we did in Example 1.5.3 — though it is a little harder.
The easy way is to approximate it by a union of narrow horizontal rectangles. Just
for practice, here is the hard solution. The easy solution is after it.
Harder solution:

• As we have done previously, we approximate the region by a union of narrow

vertical rectangles, each of width ∆x. Two of those rectangles are illustrated in
the sketch

108
I NTEGRATION 1.5 A REA BETWEEN CURVES

• In this region, x runs from a = −3 to b = 5. The curve at the top of the region is
√
y = T x) = 2x + 6

The curve at the bottom of the region is more complicated. To the left of
(−1, −2) the lower half of the parabola gives the bottom of the region while
to the right of (−1, −2) the straight line gives the bottom of the region. So
( √
− 2x + 6 if − 3 ≤ x ≤ −1
B(x) =
x−1 if − 1 ≤ x ≤ 5
Rb 
• Just as before, the area is still given by the formula a T (x) − B(x) dx, but to
accommodate our B(x), we have to split up the domain of integration when
we evaluate the integral.
Z b
 
T (x) − B(x) dx
a
Z −1 Z 5
   
= T (x) − B(x) dx + T (x) − B(x) dx
−3 −1
Z −1 Z 5
√ √  √ 
= 2x + 6 − (− 2x + 6) dx + 2x + 6 − (x − 1) dx
−3 −1
Z −1 Z 5 Z 5
√ √
=2 2x + 6dx + 2x + 6 − (x − 1)dx
−3 −1 −1

• The third integral is straightforward, while we evaluate the first two via the
substitution rule. In particular, set u = 2x + 6 and replace dx → 21 du. Also
u(−3) = 0, u(−1) = 4, u(5) = 16. Hence
Z 4 Z 16 Z 5
√ du √ du
Area = 2 u + u − (x − 1)dx
0 2 4 2 −1
 3 4  3 16  2 5
u2 1 u2 1 x
=2 3 + 3 − −x
2
2 0 2
2 4 2 −1

109
I NTEGRATION 1.5 A REA BETWEEN CURVES

2 1 h 25  1 i
= 8 − 0] + [64 − 8] − −5 − +1
3 3 2 2
72 24
= − +6
3 2
= 18

Oof!

Easier solution: The easy way to determine the area of our region is to approximate
by narrow horizontal rectangles, rather than narrow vertical rectangles. (Really we
are just swapping the roles of x and y in this problem)

• Look at our sketch of the region again — each point (x, y) in our region has
−2 ≤ y ≤ 4 and 21 (y 2 − 6) ≤ x ≤ y + 1.

• Let’s use

◦ c to denote the smallest allowed value of y,

◦ d to denote the largest allowed value of y
◦ L(y) (“L” stands for “left”) to denote the smallest allowed value of x,
when the y-coordinate is y, and
◦ R(y) (“R” stands for “right”) to denote the largest allowed value of x,
when the y-coordinate is y.

So, in this example,

1
c = −2 d=4 L(y) = (y 2 − 6) R(y) = y + 1
2
and the shaded region is

(x, y) c ≤ y ≤ d, L(y) ≤ x ≤ R(y)

• Our strategy is now nearly the same as that used in Example 1.5.1:

◦ Pick a natural number n (that we will later send to infinity), then

◦ subdivide the interval c ≤ y ≤ d into n narrow subintervals, each of width
∆y = d−cn
. Each subinterval cuts a thin horizontal slice from the region
(see the figure below).
◦ We approximate the area of slice number i by the area of a thin horizontal
rectangle (indicated by the dark rectangle in the figure below). On this
slice, the y-coordinate runs over a very narrow range. We pick a number
yi∗ , somewhere in that range. We approximate slice i by a rectangle whose
left side is at x = L(yi∗ ) and whose right side is at x = R(yi∗ ).
◦ Thus the area of slice i is approximately R(x∗i ) − L(x∗i ) ∆y.
 

110
I NTEGRATION 1.5 A REA BETWEEN CURVES

• The desired area is

n
X Z d
R(yi∗ ) L(yi∗ )
   
lim − ∆y = R(y) − L(y) dy
n→∞ c
i=1
Riemann sum → integral
Z 4
(y + 1) − 21 y 2 − 6 dy


=
−2
Z 4
− 21 y 2 + y + 4 dy
 
=
−2
h i4
= − 16 y 3 + 12 y 2 + 4y
1

1 −2
= − 6 64 − (−8) + 2 (16 − 4) + 4(4 + 2)
= −12 + 6 + 24
= 18
Example 1.5.4

One last example.

Example 1.5.5 Another area.

1
Find the area between the curves y = √ and y = sin(x) with x running from 0 to π2 .
2
Solution: This one is a little trickier since (as we shall see) the region is split into two
pieces and we need to treat them separately.

• Again we start by sketching the region.

111
I NTEGRATION 1.5 A REA BETWEEN CURVES

We want the shaded area.

• Unlike our previous examples, the bounding curves y = √12 and y = sin(x)
cross in the middle of the region of interest. They cross when y = √12 and
sin(x) = y = √12 , i.e. when x = π4 . So

◦ to the left of x = π4 , the top boundary is part of the straight line y = √1

2
and the bottom boundary is part of the curve y = sin(x)
◦ while to the right of x = π4 , the top boundary is part of the curve y = sin(x)
and the bottom boundary is part of the straight line y = √12 .

• Thus the formulae for the top and bottom boundaries are
( )
√1 if 0 ≤ x ≤ π
2 4
T (x) =
sin(x) if π4 ≤ x ≤ π2
( )
sin(x) if 0 ≤ x ≤ π4
B(x) =
√1 if π4 ≤ x ≤ π2
2

We may compute the area of interest using our canned formula

Z b
 
Area = T (x) − B(x) dx
a

but since the formulas for T (x) and B(x) change at the point x = π4 , we must
split the domain of the integral in two at that point a .
π π
• Our integral over the domain 0 ≤ x ≤ 2
is split into an integral over 0 ≤ x ≤ 4
and one over π4 ≤ x ≤ π2 :
Z π
2  
Area = T (x) − B(x) dx
0
Z π Z π
4   2  
= T (x) − B(x) dx + T (x) − B(x) dx
π
0 4

112
 I NTEGRATION 1.5 A REA BETWEEN CURVES

π Z πh
Z h 1
4
i 2 1 i
= √ − sin(x) dx + sin(x) − √ dx
0 2 π
4
2
h x i π4 h x 2
i π

= √ + cos(x) + − cos(x) − √ π
2 0 2 4
h 1 π 1 i h 1 1 πi
= √ + √ −1 + √ − √
24 2 2 24
2
= √ −1
2
√
= 2−1

a We are effectively computing the area of the region by computing the area of the two disjoint
pieces separately. Alternatively, if we set f (x) = sin(x) and g(x) = √12 , we can rewrite the
Rb  Rb
integral a T (x) − B(x) dx as a f (x) − g(x) dx. To see that the two integrals are the same,
split the domain of integration where f (x) − g(x) changes sign.

Example 1.5.5

1.5.2 tt Exercises

Exercises — Stage 1
1. We want to approximate the area between the graphs of y = cos x and y =
sin x from x = 0 to x = π using a left Riemann sum with n = 4 rectangles.

a On the graph below, sketch the four rectangles.

b Calculate the Riemann approximation.

x
π π 3π π
4 2 4
y = sin x
y = cos x

2. We want  to approximate
r the bounded area between the curves y =
2x πx
arcsin and y = using n = 5 rectangles.
π 2

113
I NTEGRATION 1.5 A REA BETWEEN CURVES

a Draw the five (vertical) rectangles on the picture below corresponding

to a right Riemann sum.

b Draw five rectangles on the picture below we might use if we were

using horizontal rectangles.

p xπ
y= 2

2x


y = arcsin π

x
π
2

3. *. Write down a definite integral that represents the finite area bounded
by the curves y = x3 − x and y = x for x ≥ 0. Do not evaluate the integral
explicitly.

4. *. Write down a definite integral that represents the area of the region
x 5x
bounded by the line y = − and the parabola y 2 = 6 − . Do not eval-
2 4
uate the integral explicitly.
5. *. Write down a definite integral that represents the area of the finite plane
region bounded by y 2 = 4ax and x2 = 4ay, where a > 0 is a constant. Do
not evaluate the integral explicitly.

6. *. Write down a definite integral that represents the area of the region
bounded between the line x + 12y + 5 = 0 and the curve x = 4y 2 . Do not
evaluate the integral explicitly.

Exercises — Stage 2

1
7. *. Find the area of the region bounded by the graph of f (x) = (2x − 4)2
and the x–axis between x = 0 and x = 1.

114
I NTEGRATION 1.5 A REA BETWEEN CURVES

2
8. *. Find the area between the curves y = x and y = 3x − x , by first identify-
ing the points of intersection and then integrating.
x
√
9. *. Calculate the area of the region enclosed by y = 2 and y = x + 1.
10. *. Find the area of the finite region bounded between the two curves y =
√
2 cos(πx/4) and y = |x|.

11. *. Find the√ area of the finite region that is bounded by the graphs of
2 3 2
f (x) = x x + 1 and g(x) = 3x .

12. *. Find the area to the left of the y–axis and to the right of the curve
x = y 2 + y.

√
13. Find the area of√the finite region below y = 9 − x2 and above both
y = |x| and y = 1 − x2 .

Exercises — Stage 3
14. *. The graph below shows the region between y = 4 + π sin x and y =
4 + 2π − 2x.

Find the area of this region.

15. *. Compute the area of the finite region bounded by the curves x = 0, x = 3,
y = x + 2 and y = x2 .
√
16. *. Find the total area between the curves y = x 25 − x2 and y = 3x, on the
interval 0 ≤ x ≤ 4.
√
17. Findp the area of the finite region below y = 9 − x2 and y = x, and above
y = 1 − (x − 1)2 .

18. Find the area of the finite region bounded by the curve y = x(x2 − 4)
and the line y = x − 2.

115
 I NTEGRATION 1.6 V OLUMES

1.6q Volumes

Another simple 1 application of integration is computing volumes. We use the same

strategy as we used to express areas of regions in two dimensions as integrals —
approximate the region by a union of small, simple pieces whose volume we can
compute and then then take the limit as the “piece size” tends to zero.
In many cases this will lead to “multivariable integrals” that are beyond our
present scope 2 . But there are some special cases in which this leads to integrals
that we can handle. Here are some examples.

Example 1.6.1 Cone.

Find the volume of the circular cone of height h and radius r.

Solution: Here is a sketch of the cone.

We have called the vertical axis x, just so that we end up with a “dx” integral.

• In what follows we will slice the cone into thin horizontal “pancakes”. In order
to approximate the volume of those slices, we need to know the radius of the
cone at a height x above its point. Consider the cross sections shown in the
following figure.

1 Well — arguably the idea isn’t too complicated and is a continuation of the idea used to compute
areas in the previous section. In practice this can be quite tricky as we shall see.
2 Typically such integrals (and more) are covered in a third calculus course.

116
I NTEGRATION 1.6 V OLUMES

At full height h, the cone has radius r. If we cut the cone at height x, then by
similar triangles (see the figure on the right) the radius will be hx · r.

• Now think of cutting the cone into n thin horizontal “pancakes”. Each such
pancake is approximately a squat cylinder of height ∆x = nh . This is very sim-
ilar to how we approximated the area under a curve by n tall thin rectangles.
Just as we approximated the area under the curve by summing these rectan-
gles, we can approximate the volume of the cone by summing the volumes of
these cylinders. Here is a side view of the cone and one of the cylinders.

• We follow the method we used in Example 1.5.1, except that our slices are now
pancakes instead of rectangles.

◦ Pick a natural number n (that we will later send to infinity), then

◦ subdivide the cone into n thin pancakes, each of width ∆x = nh .
◦ For each i = 1, 2, · · · , n, pancake number i runs from x = xi−1 = (i−1)·∆x
to x = xi = i · ∆x, and we approximate its volume by the volume of a
squat cone. We pick a number x∗i between xi−1 and xi and approximate
x∗
the pancake by a cylinder of height ∆x and radius hi r.
 ∗ 2
x
◦ Thus the volume of pancake i is approximately π hi r ∆x (as shown in
the figure above).

117
I NTEGRATION 1.6 V OLUMES

• So the Riemann sum approximation of the volume is

n  ∗ 2
X xi
Area ≈ π r ∆x
i=1
h

• By taking the limit as n → ∞ (i.e. taking the limit as the thickness of the
pancakes goes to zero), we convert the Riemann sum into a definite integral
(see Definition 1.1.9) and at the same time our approximation of the volume
becomes the exact volume:
Z h  2
x
π r dx
0 h

Our life a would be easier if we could avoid all this formal work with Riemann sums
every time we encounter a new volume. So before we compute the above integral,
let us redo the above calculation in a less formal manner.
• Start again from the picture of the cone

and think of slicing it into thin pancakes, each of width dx.

x
• The pancake at height x above the point of the cone (which is the fraction h
of
the total height of the cone) has

118
I NTEGRATION 1.6 V OLUMES

x x
◦ radius h
· r (the fraction h
of the full radius, r) and so
x

2
◦ cross-sectional area π h
r ,
◦ thickness dx — we have done something a little sneaky here, see the dis-
cussion below.

2
◦ volume π hx r dx

As x runs from 0 to h, the total volume is

Z h  2
x πr2 h 2
Z
π r dx = 2 x dx
0 h h 0
 h
πr2 x3
= 2
h 3 0
1 2
= πr h
3

In this second computation we are using a time-saving trick. As we saw in the formal
computation above, what we really need to do is pick a natural number n, slice the
cone into n pancakes each of thickness ∆x = nh and then take the limit as n → ∞.
This led to the Riemann sum
n  ∗ 2 Z h  2
X xi x
π r ∆x which becomes π r dx
i=1
h 0 h

So knowing that we will replace

n
X Z h
−→
i=1 0

x∗i
−→ x
∆x −→ dx

when we take the limit, we have just skipped the intermediate steps. While this is
not entirely rigorous, it can be made so, and does save us a lot of algebra.

a At least the bits of it involving integrals.

Example 1.6.1

119
I NTEGRATION 1.6 V OLUMES

Example 1.6.2 Sphere.

Find the volume of the sphere of radius r.

Solution: We’ll find the volume of the part of the sphere in the first octant a , sketched
below. Then we’ll multiply by 8.

• To compute the volume,

we slice it up into thin vertical “pancakes” (just as we did in the previous ex-
ample).

• Each pancake is one quarter of a thin circular disk. The pancake a distance x
from the yz-plane is shown in the sketch above. The radius of that pancake is
the distance from the dot shown in the figure to the x-axis, i.e. the y-coordinate
of the dot. To get the coordinates of the dot, observe that

◦ it lies the xy-plane, and so has z-coordinate zero, and that

2 2 2 2
◦ it also lies on the sphere, so
√ that its coordinates obey x + y + z = r .
Since z = 0 and y > 0, y = r2 − x2 .

• So the pancake at distance x from the yz-plane has

◦ thickness b dx and
√
◦ radius r2 − x2
√
2
◦ cross-sectional area 14 π r 2 − x2 and hence

120
I NTEGRATION 1.6 V OLUMES

π
r2 − x2 dx


◦ volume 4

• As x runs from 0 to r, the total volume of the part of the sphere in the first
octant is
Z r r
x3

π 2 2 π 2 1
= πr3


r − x dx = r x−
0 4 4 3 0 6

and the total volume of the whole sphere is eight times that, which is 34 πr3 , as
expected.

a The first octant is the set of all points (x, y, z) with x ≥ 0, y ≥ 0 and z ≥ 0.
b Yet again what we really do is pick a natural number n, slice the octant of the sphere into n
pancakes each of thickness ∆x = nr and then take the limit n → ∞. In the integral ∆x is replaced
by dx. Knowing that this is what is going to happen, we again just skip a few steps.

Example 1.6.2

Example 1.6.3 Revolving a region.

The region between the lines y = 3, y = 5, x = 0 and x = 4 is rotated around the line
y = 2. Find the volume of the region swept out.
Solution: As with most of these problems, we should start by sketching the problem.

• Consider the region and slice it into thin vertical strips of width dx.

• Now we are to rotate this region about the line y = 2. Imagine looking straight
down the axis of rotation, y = 2, end on. The symbol in the figure above just to
the right of the end the line y = 2 is supposed to represent your eye a . Here is
what you see as the rotation takes place.

121
I NTEGRATION 1.6 V OLUMES

• Upon rotation about the line y = 2 our strip sweeps out a “washer”

◦ whose cross-section is a disk of radius 5 − 2 = 3 from which a disk of

radius 3 − 2 = 1 has been removed so that it has a
◦ cross-sectional area of π32 − π12 = 8π and a
◦ thickness dx and hence a
◦ volume 8π dx.

• As our leftmost strip is at x = 0 and our rightmost strip is at x = 4, the total

Z 4
Volume = 8π dx = (8π)(4) = 32π
0

Notice that we could also reach this answer by writing the volume as the difference
of two cylinders.

• The outer cylinder has radius (5 − 2) and length 4. This has volume

Vouter = πr2 ` = π · 32 · 4 = 36π.

• The inner cylinder has radius (3 − 2) and length 4. This has volume

Vinner = πr2 ` = π · 12 · 4 = 4π.

• The volume we want is the difference of these two, namely

V = Vouter − Vinner = 32π.

a Okay okay. . . We missed the pupil. I’m sure there is a pun in there somewhere.

Example 1.6.3
122
I NTEGRATION 1.6 V OLUMES

Let us turn up the difficulty a little on this last example.

Example 1.6.4 Revolving again.

√
The region between the curve y = x, and the lines y = 0, x = 0 and x = 4 is rotated
around the line y = 0. Find the volume of the region swept out.
Solution: We can approach this in much the same way as the previous example.

• Consider the region and cut it into thin vertical strips of width dx.

• When we rotate the region about the line y = 0, each strip sweeps out a thin
pancake
√
◦ whose cross-section is a disk of radius x with a
√
◦ cross-sectional area of π( x)2 = πx and a
◦ thickness dx and hence a
◦ volume πxdx.

• As our leftmost strip is at x = 0 and our rightmost strip is at x = 4, the total

Z 4 hπ i4
Volume = πxdx = x2 = 8π
0 2 0

In the last example we considered rotating a region around the x-axis. Let us do
the same but rotating around the y-axis.

123
I NTEGRATION 1.6 V OLUMES

Example 1.6.5 Revolving yet again.

√
The region between the curve y = x, and the lines y = 0, x = 0 and x = 4 is rotated
around the line x = 0. Find the volume of the region swept out.
Solution:

• We will cut the region into horizontal slices, so we should write x as a function
of y. That is, the region is bounded by x = y 2 , x = 4, y = 0 and y = 2.

• Now slice the region into thin horizontal strips of width dy.

• When we rotate the region about the line x = 0, each strip sweeps out a thin
washer

◦ whose inner radius is y 2 and outer radius is 4, and

◦ thickness is dy and hence
2 2
◦ has volume π(rout − rin )dy = π(16 − y 4 )dy.

• As our bottommost strip is at y = 0 and our topmost strip is at y = 2, the total

2
π i2 32π 128π
Z h
Volume = π(16 − y 4 )dy = 16πy − y 5 = 32π − = .
0 5 0 5 5

There is another way 3 to do this one which we show at the end of this section.

Example 1.6.6 Pyramid.

Find the volume of the pyramid which has height h and whose base is a square of
side b.
Solution: Here is a sketch of the part of the pyramid that is in the first octant; we
display only this portion to make the diagrams simpler.

3 The method is not a core part of the course and should be considered optional.

124
I NTEGRATION 1.6 V OLUMES

Note that this diagram shows only 1 quarter of the whole pyramid.

• To compute its volume, we slice it up into thin horizontal “square pancakes”.

A typical pancake also appears in the sketch above.
h−z
◦ The pancake at height z is the fraction h
of the distance from the peak
of the pyramid to its base.
◦ So the full pancake a at height z is a square of side h−z
h
b. As a check, note
that when z = h the pancake has side h−h h
b = 0, and when z = 0 the
h−0
pancake has side h b = b.

2
◦ So the pancake has cross-sectional area h−z h
b and thickness b dz and
hence

2
◦ volume h−z h
b dz.
• The volume of the whole pyramid (not just the part of the pyramid in the first
octant) is
Z h
h − z 2 b2 h
Z
b dz = 2 (h − z)2 dz
0 h h 0

Now use the substitution rule with t = (h − z), dz → −dt

b2 0 2
Z
= 2 −t dt
h h
 0
b2 t3
=− 2
h 3 h
b2 h3
 
=− 2 −
h 3
1 2
= bh
3

125
I NTEGRATION 1.6 V OLUMES

a Note that this is the full pancake, not just the part in the first octant.
b We are again using our Riemann sum avoiding trick.

Example 1.6.6

Let’s ramp up the difficulty a little.

Example 1.6.7 Napkin Ring.

Suppose you make two napkin rings a by drilling holes with different diameters
through two wooden balls. One ball has radius r and the other radius R with r < R.
You choose the diameter of the holes so that both napkin rings have the same height,
2h. See the figure below.

Which b ring has more wood in it?

Solution: We’ll compute the volume of the napkin ring with radius R. We can then
obtain the volume of the napkin ring of radius r, by just replacing R 7→ r in the
result.

• To compute the volume of the napkin ring of radius R, we slice it up into thin
horizontal “pancakes”. Here is a sketch of the part of the napkin ring in the
first octant showing a typical pancake.

126
I NTEGRATION 1.6 V OLUMES

• The coordinates of the two points marked in the yz-plane of that figure are
found by remembering that
◦ the equation of the sphere is x2 + y 2 + z 2 = R2 .
◦ The two√points have y > 0 and are in the yz-plane, so that x = 0 for them.
So y = R2 − z 2 .
√
◦ In particular, at the top of the napkin ring z = h, so that y = R2 − h2 .
• The pancake at height z, shown in the sketch, is a “washer” — a circular disk
with a circular hole cut in its center.
√
◦ The outer radius of the washer is R2 − z 2 and
√
◦ the inner radius of the washer is R2 − h2 . So the
◦ cross-sectional area of the washer is
√
2 √
2
π R2 − z 2 − π R2 − h2 = π(h2 − z 2 )

• The pancake at height z

◦ has thickness dz and
◦ cross-sectional area π(h2 − z 2 ) and hence
◦ volume π(h2 − z 2 )dz.
• Since z runs from −h to +h, the total volume of wood in the napkin ring of
radius R is
Z h h z 3 ih
π(h2 − z 2 )dz = π h2 z −
−h 3 −h

127
I NTEGRATION 1.6 V OLUMES

h h3  
3 3 (−h)3 i
=π h − − (−h) −
3 3
h2 2
3 i
3
=π h − −h
3 3
4π 3
= h
3

This volume is independent of R. Hence the napkin ring of radius r contains pre-
cisely the same volume of wood as the napkin ring of radius R!

a Handy things to have (when combined with cloth napkins) if your parents are coming to dinner
and you want to convince them that you are “taking care of yourself”.
b A good question to ask to distract your parents from the fact you are serving frozen burritos.

Example 1.6.7

Example 1.6.8 Notch.

A 45◦ notch is cut to the centre of a cylindrical log having radius 20cm. One plane
face of the notch is perpendicular to the axis of the log. See the sketch below. What
volume of wood was removed?

Solution: We show two solutions to this problem which are of comparable difficulty.
The difference lies in the shape of the pancakes we use to slice up the volume. In
solution 1 we cut rectangular pancakes parallel to the yz-plane and in solution 2 we
slice triangular pancakes parallel to the xz-plane.
Solution 1:

• Concentrate on the notch. Rotate it around so that the plane face lies in the
xy-plane.

• Then slice the notch into vertical rectangles (parallel to the yz-plane) as in the
figure on the left below.

128
I NTEGRATION 1.6 V OLUMES

• The cylindrical log had radius 20cm. So the circular part of the boundary of the
base of the notch has equation x2 + y 2 = 202 . (We’re putting the origin of the
xy-plane at the centre of the circle.) If our coordinate system is such that x is
constant on each slice, then
◦ the base
√ of the slice is the line segment from (x, −y, 0) to (x, +y, 0) where
y = 202 − x2 so that
√
◦ the slice has width 2y = 2 202 − x2 and
◦ height x (since the upper face of the notch is at 45◦ to the base — see the
side view sketched in the figure on the right above).
√
◦ So the slice has cross-sectional area 2x 202 − x2 .
• On the base of the notch x runs from 0 to 20 so the volume of the notch is
Z 20 √
V = 2x 202 − x2 dx
0
1
Make the change of variables u = 202 −x2 (don’t forget to change dx → − 2x du):
Z 0
√
V = − u du
202
 3/2 0
u
= −
3/2 202
2 16, 000
= 203 =
3 3
Solution 2:
• Concentrate of the notch. Rotate it around so that its base lies in the xy-plane
with the skinny edge along the y-axis.

129
I NTEGRATION 1.6 V OLUMES

• Slice the notch into triangles parallel to the xz-plane as in the figure on the left
below. In the figure below, the triangle happens to lie in a plane where y is
negative.

• The cylindrical log had radius 20cm. So the circular part of the boundary of the
base of the notch has equation x2 + y 2 = 202 . Our coordinate system is such
that y is constant on each slice, so that

◦ the base
p of the triangle is the line segment from (0, y, 0) to (x, y, 0) where
x = 202 − y 2 so that
p
◦ the triangle has base x = 202 − y 2 and
p
◦ height x = 202 − y 2 (since the upper face of the notch is at 45◦ to the
base — see the side view sketched in the figure on the right above).
p
2
◦ So the slice has cross-sectional area 21 202 − y 2 .

• On the base of the notch y runs from −20 to 20, so the volume of the notch is

1 20
Z
V = (202 − y 2 )dy
2 −20
Z 20
= (202 − y 2 )dy
0
h y 3 i20
2
= 20 y −
3 0
2 3 16, 000
= 20 =
3 3
Example 1.6.8

130
 I NTEGRATION 1.6 V OLUMES

1.6.1 tt Optional — Cylindrical shells

Let us return to Example 1.6.5 in which we rotate a region around the y-axis. Here
we show another solution to this problem which is obtained by slicing the region
into vertical strips. When rotated about the y-axis, each such strip sweeps out a thin
cylindrical shell. Hence the name of this approach (and this subsection).

Example 1.6.9 Revolving yet again.

√
The region between the curve y = x, and the lines y = 0, x = 0 and x = 4 is rotated
around the line x = 0. Find the volume of the region swept out.
Solution:

• Consider the region and cut it into thin vertical strips of width dx.

• When we rotate the region about the line y = 0, each strip sweeps out a thin
cylindrical shell

◦ whose radius is x,
√
◦ height is x, and
◦ thickness is dx and hence
◦ has volume 2π × radius × height × thickness = 2πx3/2 dx.

• As our leftmost strip is at x = 0 and our rightmost strip is at x = 4, the total

4  4
4π 5/2 4π 128π
Z
3/2
Volume = 2πx dx = x = · 32 =
0 5 0 5 5

which (thankfully) agrees with our previous computation.

131
 I NTEGRATION 1.6 V OLUMES

1.6.2 tt Exercises

Exercises — Stage 1
1. Consider a right circular cone.

What shape are horizontal cross-sections? Are the vertical cross-sections

the same?

2. Two potters start with a block of clay h units tall, and identical square
cookie cutters. They form columns by pushing the square cookie cutter
straight down over the clay, so that its cross-section is the same square as
the cookie cutter. Potter A pushes their cookie cutter down while their
clay block is sitting motionless on a table; Potter B pushes their cookie
cutter down while their clay block is rotating on a potter’s wheel, so
their column looks twisted. Which column has greater volume?

Column A Column B

3. Let R be the region bounded above by the graph of y = f (x) shown below
and bounded below by the x-axis, from x = 0 to x = 6. Sketch the washers
that are formed by rotating R about the y-axis. In your sketch, label the all
radii in terms of y, and label the thickness.

132
I NTEGRATION 1.6 V OLUMES

1 y = f (x)

x
1 2 4 6

4. *. Write down definite integrals that represent the following quantities. Do

not evaluate the integrals explicitly.

a The volume of the solid obtained by

√ rotating around the x–axis the
x2
region between the x–axis and y = x e for 0 ≤ x ≤ 3.

b The volume of the solid obtained by revolving the region bounded by

the curves y = x2 and y = x + 2 about the line x = 3.

5. *. Write down definite integrals that represent the following quantities.

Do not evaluate the integrals explicitly.

a The volume of the solid obtained by rotating the finite plane re-
gion bounded by the curves y = 1 − x2 and y = 4 − 4x2 about the
line y = −1.

b The volume of the solid obtained by rotating the finite plane re-
gion bounded by the curve y = x2 − 1 and the line y = 0 about the
line x = 5.

6. *. Write down a definite integral that represents the volume of the solid
obtained by rotating around the line y = −1 the region between the curves
y = x2 and y = 8 − x2 . Do not evaluate the integrals explicitly.
7. A tetrahedron is a three-dimensional shape with four faces, each of which
is an equilateral triangle. (You might have seen this shape as a 4-sided die;
think of a pyramid with a triangular base.) Using the methods from this
section, calculate the volume of a tetrahedron with side-length `. You may
assume
q without proof that the height of a tetrahedron with side-length ` is
2
3
`.

133
I NTEGRATION 1.6 V OLUMES

` q
2
3
`

Exercises — Stage 2

8. *. Let a > 0 be a√constant. Let R be the finite region bounded by the

x2
graph of y = 1 + xe , the line y = 1, and the line x = a. Using vertical
slices, find the volume generated when R is rotated about the line y = 1.

9. *. Find the volume of the solid generated by rotating the finite region
bounded by y = 1/x and 3x + 3y = 10 about the x–axis.
2 2
10. *. Let R be the region inside the circle x + (y − 2) = 1. Let S be the solid
obtained by rotating R about the x-axis.

a Write down an integral representing the volume of S.

b Evaluate the integral you wrote down in part (a).

11. *. The region R is the portion of the first quadrant which is below the
parabola y 2 = 8x and above the hyperbola y 2 − x2 = 15.

a Sketch the region R.

b Find the volume of the solid obtained by revolving R about the x axis.

12. *. The region R is bounded by y = log x, y = 0, x = 1 and x = 2. (Recall

that we are using log x to denote the logarithm of x with base e. In other
courses it is often denoted log x.)

a Sketch the region R.

b Find the volume of the solid obtained by revolving this region

about the y axis.

x 2 2
13. *. The finite region between the curves y = cos( 2 ) and y = x − π is rotated
about the line y = −π 2 . Using vertical slices (disks and/or washers), find
the volume of the resulting solid.

14. *. The solid V is 2 meters high and has square horizontal cross sections.
The length of the side of the square cross section at height x meters
2
above the base is 1+x m. Find the volume of this solid.

134
I NTEGRATION 1.6 V OLUMES

15. *. Consider a solid whose base is the finite portion of the xy–plane bounded
by the curves y = x2 and y = 8 − x2 . The cross–sections perpendicular to
the x–axis are squares with one side in the xy–plane. Compute the volume
of this solid.

16. *. A frustrum of a right circular cone (as shown below) has height h.
Its base is a circular disc with radius 4 and its top is a circular disc with
radius 2. Calculate the volume of the frustrum.

Exercises — Stage 3
17. The shape of the earth is often approximated by an oblate spheroid, rather
than a sphere. An oblate spheroid is formed by rotating an ellipse about its
minor axis (its shortest diameter).

a Find the volume of the oblate spheroid obtained by rotating the upper
(positive) half of the ellipse (ax)2 + (by)2 = 1 about the x-axis, where
a and b are positive constants with a ≥ b.

b Suppose a the earth has radius at the equator of 6378.137 km, and
radius at the poles of 6356.752 km. If we model the earth as an
oblate spheroid formed by rotating the upper half of the ellipse
(ax)2 + (by)2 = 1 about the x-axis, what are a and b?

c What is the volume of this model of the earth? (Use a calculator.)

d Suppose we had calculated the volume of the earth by modelling it

as a sphere with radius 6378.137 km. What would our absolute and
relative errors be, compared to our oblate spheroid calculation?

a Earth Fact Sheet, NASA, accessed 2 July 2017

2
18. *. Let R be the bounded region that lies between the curve y = 4 − (x − 1)
and the line y = x + 1.

a Sketch R and find its area.

b Write down a definite integral giving the volume of the region ob-
tained by rotating R about the line y = 5. Do not evaluate this integral.

135
I NTEGRATION 1.6 V OLUMES

2 2 2 2

19. *. Let R = (x, y) : (x − 1) + y ≤ 1 and x + (y − 1) ≤ 1 .
a Sketch R and find its area.

b If R rotates around the y–axis, what volume is generated?

20. √ R be the plane region bounded by x = 0, x = 1, y = 0 and

*. Let
y = c 1 + x2 , where c is a positive constant.

a Find the volume V1 of the solid obtained by revolving R about the

x–axis.

b Find the volume V2 of the solid obtained by revolving R about the

y–axis.

c If V1 = V2 , what is the value of c?

21. *. The graph below shows the region between y = 4 + π sin x and y =
4 + 2π − 2x.

The region is rotated about the line y = −1. Express in terms of definite
integrals the volume of the resulting solid. Do not evaluate the integrals.

22. On a particular, highly homogeneous a planet, we observe that the den-

sity of the atmosphere h kilometres above the surface is given by the
equation ρ(h) = c2−h/6 mkg3 , where c is the density on the planet’s sur-
face.

a What is the mass of the atmosphere contained in a vertical column

with radius one metre, sixty kilometres high?
3000cπ
b What height should a column be to contain kilograms of
log 2
air?

a This is clearly a simplified model: air density changes all the time, and depends
on lots of complicated factors aside from altitude. However, the equation we’re

136
 I NTEGRATION 1.7 I NTEGRATION BY PARTS

using is not so far off from an idealized model of the earth’s atmosphere, taken
from Pressure and the Gas Laws by H.P. Schmid, accessed 3 July 2017.

1.7q Integration by parts

1.7.1 tt Integration by parts

The fundamental theorem of calculus tells us that it is very easy to integrate a deriva-
tive. In particular, we know that
d
Z
(F (x)) dx = F (x) + C
dx
We can exploit this in order to develop another rule for integration — in particular a
rule to help us integrate products of simpler function such as
Z
xex dx

In so doing we will arrive at a method called “integration by parts”.

To do this we start with the product rule and integrate. Recall that the product
rule says
d
u(x)v(x) = u0 (x) v(x) + u(x) v 0 (x)
dx
Integrating this gives
Z
 0
u (x) v(x) + u(x) v 0 (x) dx = a function whose derivative is u0 v + uv 0 + C
  

= u(x)v(x) + C

Now this, by itself, is not terribly useful. In order to apply it we need to have a
function whose integrand is a sum of products that is in exactly this form u0 (x)v(x) +
u(x)v 0 (x). This is far too specialised.
However if we tease this apart a little:
Z Z Z
 0
u (x) v(x) + u(x) v (x) dx = u (x) v(x) dx + u(x) v 0 (x) dx
0 0


Bring one of the integrals to the left-hand side

Z Z
u(x)v(x) − u (x) v(x)dx = u(x) v 0 (x)dx
0

137
I NTEGRATION 1.7 I NTEGRATION BY PARTS

Swap left and right sides

Z Z
0
u(x) v (x)dx = u(x)v(x) − u0 (x) v(x)dx

In this form we take the integral of one product and express it in terms of the integral
of a different product. If we express it like that, it doesn’t seem too useful. However,
if the second integral is easier, then this process helps us.
Let us do a simple example before explaining this more generally.
R
Example 1.7.1 xex dx.
Z
Compute the integral xex dx.
Solution:

• We start by taking the equation above

Z Z
u(x) v (x)dx = u(x)v(x) − u0 (x) v(x)dx
0

• Now set u(x) = x and v 0 (x) = ex . How did we know how to make this choice?
We will explain some strategies later. For now, let us just accept this choice and
keep going.

• In order to use the formula we need to know u0 (x) and v(x). In this case it is
quite straightforward: u0 (x) = 1 and v(x) = ex .

• Plug everything into the formula:

Z Z
xe dx = xe − ex dx
x x

So our original more difficult integral has been turned into a question of com-
puting an easy one.

= xex − ex + C

• We can check our answer by differentiating:

d
(xex − ex + C) = xe x
+ 1 · e}x −ex + 0
dx | {z
by product rule

= xex as required.

The process we have used in the above example is called “integration by parts”.
When our integrand is a product we try to write it as u(x)v 0 (x) — we need to choose

138
I NTEGRATION 1.7 I NTEGRATION BY PARTS

one factor to be u(x) and the other to be v 0 (x). We then compute u0 (x) and v(x) and
then apply the following theorem:

Theorem 1.7.2 Integration by parts.

Let u(x) and v(x) be continuously differentiable. Then

Z Z
u(x) v (x)dx = u(x) v(x) − v(x) u0 (x)dx
0

If we write dv for v 0 (x)dx and du for u0 (x)dx (as the substitution rule suggests),
then the formula becomes
Z Z
udv = u v − vdu

The application of this formula is known as integration by parts.

The corresponding statement for definite integrals is
Z b Z b
0
u(x) v (x)dx = u(b) v(b) − u(a) v(a) − v(x) u0 (x)dx
a a

Integration by parts is not as easy to apply as the product rule for derivatives.
This is because it relies on us

1 judiciously choosing u(x) and v 0 (x), then

2 computing u0 (x) and v(x) — which requires us to antidifferentiate v 0 (x), and

finally

3 that the integral u0 (x)v(x)dx is easier than the integral we started with.
R

Notice that any antiderivative of v 0 (x) will do. All antiderivatives of v 0 (x) are of
the form v(x)+A with A a constant. Putting this into the integration by parts formula
gives
Z Z
u(x)v (x)dx = u(x) (v(x) + A) − u0 (x) (v(x) + A) dx
0

Z Z
= u(x)v(x) + Au(x) − u (x)v(x)dx − A u0 (x)dx
0

| {z }
=Au(x)+C
Z
= u(x)v(x) − u0 (x)v(x)dx + C

So that constant A will always cancel out.

In most applications (but not all) our integrand will be a product of two factors
so we have two choices for u(x) and v 0 (x). Typically one of these choices will be

139
I NTEGRATION 1.7 I NTEGRATION BY PARTS

“good” (in that it results in a simpler integral) while the other will be “bad” (we
cannot antidifferentiate our choice of v 0 (x) or the resulting integral is harder). Let us
illustrate what we mean by returning to our previous example.
R
Example 1.7.3 xex dx — again.

Our integrand is the product of two factors

x and ex
This gives us two obvious choices of u and v 0 :
u(x) = x v 0 (x) = ex
or
u(x) = ex v 0 (x) = x
We should explore both choices:
1. If take u(x) = x and v 0 (x) = ex . We then quickly compute
u0 (x) = 1 and v(x) = ex
which means we will need to integrate (in the right-hand side of the integration
by parts formula)
Z Z
0
u (x)v(x)dx = 1 · ex dx

which looks straightforward. This is a good indication that this is the right
choice of u(x) and v 0 (x).
2. But before we do that, we should also explore the other choice, namely u(x) =
ex and v 0 (x) = x. This implies that
1
u0 (x) = ex and v(x) = x2
2
which means we need to integrate
1 2 x
Z Z
0
u (x)v(x)dx = x · e dx.
2
This is at least as hard as the integral we started with. Hence we should try the
first choice.
With our choice made, we integrate by parts to get
Z Z
xe dx = xe − ex dx
x x

= xex − ex + C.
The above reasoning is a very typical workflow when using integration by parts.

140
I NTEGRATION 1.7 I NTEGRATION BY PARTS

Integration by parts is often used

d
• to eliminate factors of x from an integrand like xex by using that dx
x = 1 and
d 1
• to eliminate a log x from an integrand by using that dx
log x = x
and
• to eliminate inverse trig functions, like arctan x, from an integrand by using
d 1
that, for example, dx arctan x = 1+x 2.

R
Example 1.7.4 x sin xdx.

Solution:
• Again we have a product of two factors giving us two possible choices.
1 If we choose u(x) = x and v 0 (x) = sin x, then we get
u0 (x) = 1 and v(x) = − cos x
which is looking promising.
2 On the other hand if we choose u(x) = sin x and v 0 (x) = x, then we have
1
u0 (x) = cos x and v(x) = x2
2
1 2
R
which is looking worse — we’d need to integrate 2
x cos xdx.
• So we stick with the first choice. Plugging u(x) = x, v(x) = − cos x into inte-
gration by parts gives us
Z Z
x sin xdx = −x cos x − 1 · (− cos x)dx

= −x cos x + sin x + C

• Again we can check our answer by differentiating:

d
(−x cos x + sin x + C) = − cos x + x sin x + cos x + 0
dx
= x sin xX

Once we have practised this a bit we do not really need to write as much. Let us
solve it again, but showing only what we need to.
Solution:
• We use integration by parts to solve the integral.
• Set u(x) = x and v 0 (x) = sin x. Then u0 (x) = 1 and v(x) = − cos x, and
Z Z
x sin xdx = −x cos x + cos xdx

= −x cos x + sin x + C.

141
I NTEGRATION 1.7 I NTEGRATION BY PARTS

It is pretty standard practice to reduce the notation even further in these problems.
As noted above, many people write the integration by parts formula as
Z Z
udv = uv − vdu

where du, dv are shorthand for u0 (x)dx, v 0 (x)dx. Let us write up the previous example
using this notation.
R
Example 1.7.5 x sin xdx yet again.

Solution: Using integration by parts, we set u = x and dv = sin xdx. This makes
du = 1dx and v = − cos x. Consequently
Z Z
x sin xdx = udv
Z
= uv − vdu
Z
= −x cos x + cos xdx

= −x cos x + sin x + C

You can see that this is a very neat way to write up these problems and we will
continue using this shorthand in the examples that follow below.

We can also use integration by parts to eliminate higher powers of x. We just

need to apply the method more than once.

x2 ex dx.
R
Example 1.7.6

Solution:

• Let u = x2 and dv = ex dx. This then gives du = 2xdx and v = ex , and

Z Z
2 x 2 x
x e dx = x e − 2xex dx

• So we have reduced the problem of computing the original integral to one of

integrating 2xex . We know how to do this — just integrate by parts again:
Z Z
2 x 2 x
x e dx = x e − 2xex dx set u = 2x, dv = ex dx
 Z 
2 x x x
= x e − 2xe − 2e dx since du = 2dx, v = ex

= x2 ex − 2xex + 2ex + C

142
I NTEGRATION 1.7 I NTEGRATION BY PARTS

• We can, if needed, check our answer by differentiating:

d
x2 ex − 2xex + 2ex + C


dx
= x2 ex + 2xex − (2xex + 2ex ) + 2ex + 0

= x2 e x X

A similar iterated application of integration by parts will work for integrals

Z
P (x) (Aeax + B sin(bx) + C cos(cx)) dx

where P (x) is a polynomial and A, B, C, a, b, c are constants.

Example 1.7.6

Now let us look at integrands containing logarithms. We don’t know the an-
tiderivative of log x, but we can eliminate log x from an integrand by using integra-
tion by parts with u = log x. Remember log x = loge x = ln x.
R
Example 1.7.7 x log xdx.

Solution:
• We have two choices for u and dv.

1 Set u = x and dv = log xdx. This gives du = dx but v is hard to compute —

we haven’t done it yet a . Before we go further along this path, we should
look to see what happens with the other choice.
2 Set u = log x and dv = xdx. This gives du = x1 dx and v = 12 x2 , and we
have to integrate
1 1 2
Z Z
v du = · x dx
x 2
which is easy.

• So we proceed with the second choice.

1 2 1
Z Z
x log xdx = x log x − xdx
2 2
1 1
= x2 log x − x2 + C
2 4

• We can check our answer quickly:

d  x2 x2  x2 1 x
ln x − + C = x ln x + − + 0 = x ln x
dx 2 4 2 x 2

143
I NTEGRATION 1.7 I NTEGRATION BY PARTS

a We will soon.

Example 1.7.7

R
Example 1.7.8 log xdx.

It is not immediately obvious that one should use integration by parts to compute
the integral
Z
log xdx

since the integrand is not a product. But we should persevere — indeed this is a
situation where our shorter notation helps to clarify how to proceed.
Solution:

• In the previous example we saw that we could remove the factor log x by set-
ting u = log x and using integration by parts. Let us try repeating this. When
we make this choice, we are then forced to take dv = dx — that is we choose
v 0 (x) = 1. Once we have made this sneaky move everything follows quite
directly.

• We then have du = x1 dx and v = x, and the integration by parts formula gives

us
1
Z Z
log xdx = x log x − · xdx
x
Z
= x log x − 1dx

= x log x − x + C

• As always, it is a good idea to check our result by verifying that the derivative
of the answer really is the integrand.

d
1
x ln x − x + C = ln x + x − 1 + 0 = ln x
dx x

The same method works almost exactly to compute the antiderivatives of arcsin(x)
and arctan(x):

144
I NTEGRATION 1.7 I NTEGRATION BY PARTS

R R
Example 1.7.9 arctan(x)dx and arcsin(x)dx.

Compute the antiderivatives of the inverse sine and inverse tangent functions.
Solution:

• Again neither of these integrands are products, but that is no impediment. In

both cases we set dv = dx (ie v 0 (x) = 1) and choose v(x) = x.
1
• For inverse tan we choose u = arctan(x), so du = 1+x 2 dx:

1
Z Z
arctan(x)dx = x arctan(x) − x · dx
1 + x2

now use substitution rule with w(x) = 1 + x2 , w0 (x) = 2x

Z 0
w (x) 1
= x arctan(x) − · dx
2 w
1 1
Z
= x arctan(x) − dw
2 w
1
= x arctan(x) − log |w| + C
2
1
= x arctan(x) − log |1 + x2 | + C but 1 + x2 > 0, so
2
1
= x arctan(x) − log(1 + x2 ) + C
2

• Similarly for inverse sine we choose u = arcsin(x) so du = √ 1 dx:

1−x2

x
Z Z
arcsin(x)dx = x arcsin(x) − √ dx
1 − x2

Now use substitution rule with w(x) = 1 − x2 , w0 (x) = −2x

−w0 (x)
Z
= x arcsin(x) − · w−1/2 dx
2
1
Z
= x arcsin(x) + w−1/2 dw
2
1
= x arcsin(x) + · 2w1/2 + C
2
√
= x arcsin(x) + 1 − x2 + C

• Both can be checked quite quickly by differentiating — but we leave that as an

exercise for the reader.

There are many other examples we could do, but we’ll finish with a tricky one.

145
I NTEGRATION 1.7 I NTEGRATION BY PARTS

R
Example 1.7.10 ex sin xdx.

Solution: Let us attempt this one a little naively and then we’ll come back and do it
more carefully (and successfully).

• We can choose either u = ex , dv = sin xdx or the other way around.

1. Let u = ex , dv = sin xdx. Then du = ex dx and v = − cos x. This gives

Z Z
e sin x = −e cos x + ex cos xdx
x x

So we are left with an integrand that is very similar to the one we started
with. What about the other choice?
2. Let u = sin x, dv = ex dx. Then du = cos xdx and v = ex . This gives
Z Z
e sin x = e sin x − ex cos xdx
x x

So we are again left with an integrand that is very similar to the one we
started with.
R
• RHow do we proceed? — It turns out to be easier if you do both ex sin xdx and
ex cos xdx simultaneously. We do so in the next example.

Rb Rb
Example 1.7.11 a
ex sin xdx and a
ex cos xdx.

This time we’re going to do the two integrals

Z b Z b
I1 = ex sin xdx I2 = ex cos xdx
a a

at more or less the same time.

• First
Z b Z b
x
I1 = e sin xdx = udv
a a

Choose u = ex , dv = sin xdx, so v = − cos x, du = ex dx

h ib Z b
x
= − e cos x + ex cos xdx
a a

We have not found I1 but we have related it to I2 .

h ib
I1 = − ex cos x + I2
a

146
I NTEGRATION 1.7 I NTEGRATION BY PARTS

• Now start over with I2 .

Z b Z b
x
I2 = e cos xdx = udv
a a

Choose u = ex , dv = cos xdx, so v = sin x, du = ex dx

h ib Z b
x
= e sin x − ex sin xdx
a a

Once again, we have not found I2 but we have related it back to I1 .

h ib
I2 = ex sin x − I1
a

• So summarising, we have
h ib h ib
x x
I1 = − e cos x + I2 I2 = e sin x − I1
a a

• So now, substitute the expression for I2 from the second equation into the first
equation to get
h ib
I1 = − ex cos x + ex sin x − I1
a
1h x
ib
which implies I1 = e sin x − cos x
2 a

If we substitute the other way around we get

h ib
x x
I2 = e sin x + e cos x − I2
a
1h x
ib
which implies I2 = e sin x + cos x
2 a

That is,
b
1h x
Z
x

ib
e sin xdx = e sin x − cos x
a 2 a
Z b
1 h i b
ex cos xdx = ex sin x + cos x


a 2 a

• This also says, for example, that 21 ex sin x−cos x is an antiderivative of ex sin x

so that
1
Z
ex sin xdx = ex sin x − cos x + C


2

147
 I NTEGRATION 1.7 I NTEGRATION BY PARTS

• Note that we can always check whether or not this is correct. It is correct if
and only if the derivative of the right hand side is ex sin x. Here goes. By the
product rule

d h1 x
i
= e sin x − cos x + C
dx 2
1h
i
= ex sin x − cos x + ex cos x + sin x = ex sin x


2
which is the desired derivative.
R R
• There is another way to find ex sin xdx and ex cos xdx that, in contrast to the
above computations, doesn’t involve any trickery. But it does require the use
of complex numbers and so is beyond the scope of this course. The secret is to
ix −ix ix −ix
use that sin x = e −e2i
and cos x = e +e2
, where i is the square root of −1 of
the complex number system.
Example 1.7.11

1.7.2 tt Exercises

Exercises — Stage 1

1. The method of integration by substitution comes from the rule for dif-
ferentiation.
The method of integration by parts comes from the rule for differentia-
tion.

2. Suppose you want to evaluate an integral using integration by parts. You

choose part of your integrand to be u, and part to be dv. The part chosen as
u will be: (differentiated, antidifferentiated). The part chosen as dv will be:
(differentiated, antidifferentiated).
3. Let f (x) and g(x) be differentiable functions. Using
Z 0 the quotient rule for
f (x)
differentiation, give an equivalent expression to dx.
g(x)
Z
4. Suppose we want to use integration by parts to evaluate u(x) · v 0 (x)dx
for some differentiable functions u and v. We need to find an antideriva-
tive of v 0 (x), but there are infinitely many choices. Show that every an-
tiderivative of v 0 (x) gives an equivalent final answer.

148
I NTEGRATION 1.7 I NTEGRATION BY PARTS
Z
5. Suppose you want to evaluate f (x)dx using integration by parts. Explain
why dv = f (x)dx, u = 1 is generally a bad choice.
Note: compare this to Example 1.7.8, where we chose u = f (x), dv = 1dx.

Exercises — Stage 2
Z
6. *. Evaluate x log x dx.

log x
Z
7. *. Evaluate dx.
x7
Z π
8. *. Evaluate x sin x dx.
0

Z π
2
9. *. Evaluate x cos x dx.
0
Z
10. Evaluate x3 ex dx.
Z
11. Evaluate x log3 xdx.

Z
12. Evaluate x2 sin xdx.

Z
13. Evaluate (3t2 − 5t + 6) log tdt.

√
Z √
14. Evaluate se s ds.
Z
15. Evaluate log2 xdx.
Z
2 +1
16. Evaluate 2xex dx.

Z
17. *. Evaluate arccos y dy.

Exercises — Stage 3
Z
18. *. Evaluate 4y arctan(2y) dy.

149
I NTEGRATION 1.7 I NTEGRATION BY PARTS

Z
19. Evaluate x2 arctan xdx.

Z
20. Evaluate ex/2 cos(2x)dx.
Z
21. Evaluate sin(log x)dx.
Z
22. Evaluate 2x+log2 x dx.

Z
23. Evaluate ecos x sin(2x)dx.

xe−x
Z
24. Evaluate dx.
(1 − x)2
25. *. A reduction formula.
a Derive the reduction formula
sinn−1 (x) cos(x) n − 1
Z Z
n
sin (x) dx = − + sinn−2 (x) dx.
n n
Z π/2
b Calculate sin8 (x) dx.
0

26. *. Let R be the part of the first quadrant that lies below the curve y =
arctan x and between the lines x = 0 and x = 1.

a Sketch the region R and determine its area.

b Find the volume of the solid obtained by rotating R about the y–axis.
√ 3x √
27. *. Let R be the region between the curves T (x) = xe and B(x) = x(1 +
2x) on the interval 0 ≤ x ≤ 3. (It is true that T (x) ≥ B(x) for all 0 ≤ x ≤ 3.)
Compute the volume of the solid formed by rotating R about the x-axis.
Z 4 √

28. 0
*. Let f (0) = 1, f (2) = 3 and f (2) = 4. Calculate f 00 x dx.
0

n  
X 2 2 2
29. Evaluate lim i − 1 e n i−1 .
n→∞
i=1
n n

150
 I NTEGRATION 1.8 T RIGONOMETRIC I NTEGRALS

1.8q Trigonometric Integrals

Integrals of polynomials of the trigonometric functions sin x, cos x, tan x and so on,
are generally evaluated by using a combination of simple substitutions and trigono-
metric identities. There are of course a very large number 1 of trigonometric identi-
ties, but usually we use only a handful of them. The most important three are:

Equation 1.8.1

sin2 x + cos2 x = 1

Equation 1.8.2

sin(2x) = 2 sin x cos x

Equation 1.8.3

cos(2x) = cos2 x − sin2 x

= 2 cos2 x − 1
= 1 − 2 sin2 x

Notice that the last two lines of Equation 1.8.3 follow from the first line by replac-
ing either sin2 x or cos2 x using Equation 1.8.1. It is also useful to rewrite these last
two lines:

Equation 1.8.4

1 − cos(2x)
sin2 x =
2

1 The more pedantic reader could construct an infinite list of them.

151
 I NTEGRATION 1.8 T RIGONOMETRIC I NTEGRALS

Equation 1.8.5

1 + cos(2x)
cos2 x =
2

These last two are particularly useful since they allow us to rewrite higher powers
of sine and cosine in terms of lower powers. For example:
 2
4 1 − cos(2x)
sin (x) = by Equation 1.8.4
2
1 1 1
= − cos(2x) + cos2 (2x) use Equation 1.8.5
4 2 4 | {z }
do it again
1 1 1
= − cos(2x) + (1 + cos(4x))
4 2 8
3 1 1
= − cos(2x) + cos(4x)
8 2 8
So while it was hard to integrate sin4 (x) directly, the final expression is quite straight-
forward (with a little substitution rule).
There are many such tricks for integrating powers of trigonometric functions.
Here we concentrate on two families
Z Z
m n
sin x cos xdx and tanm x secn xdx

for integer n, m. The details of the technique depend on the parity of n and m — that
is, whether n and m are even or odd numbers.

1.8.1 tt Integrating sinm x cosn xdx

R

ttt One of n and m is odd

1.8.1.1

Consider the integral sin2 x cos xdx. We can integrate this by substituting u = sin x
R

and du = cos xdx. This gives

Z Z
sin x cos xdx = u2 du
2

1 1
= u3 + C = sin3 x + C
3 3
This method can be used whenever n is an odd integer.

152
 I NTEGRATION 1.8 T RIGONOMETRIC I NTEGRALS

• Substitute u = sin x and du = cos xdx.

• This leaves an even power of cosines — convert them using cos2 x = 1−sin2 x =
1 − u2 .
Here is an example.

sin2 x cos3 xdx.

R
Example 1.8.6

Start by factoring off one power of cos x to combine with dx to get cos xdx = du.
Z Z
2 3 2 2
| {z x} cos
sin x cos xdx = sin | {z x} |cos{z
xdx} set u = sin x
=u2 =1−u2 =du
Z
= u2 (1 − u2 )du
u3 u5
= − +C
3 5
sin3 x sin5 x
= − +C
3 5

Of course if m is an odd integer we can use the same strategy with the roles
of sin x and cos x exchanged. That is, we substitute u = cos x, du = − sin xdx and
sin2 x = 1 − cos2 x = 1 − u2 .

ttt Both n and m are even

1.8.1.2

If m and n are both even, the strategy is to use the trig identities 1.8.4 and 1.8.5 to get
back to the m or n odd case. This is typically more laborious than the previous case
we studied. Here are a couple of examples that arise quite commonly in applications.

cos2 xdx.
R
Example 1.8.7

By 1.8.5
1 1h 1
Z Z i
2
 
cos xdx = 1 + cos(2x) dx = x + sin(2x) + C
2 2 2

cos4 xdx.
R
Example 1.8.8

First we’ll prepare the integrand cos4 x for easy integration by applying 1.8.5 a couple
times. We have already used 1.8.5 once to get

1
cos2 x =

1 + cos(2x)
2

153
I NTEGRATION 1.8 T RIGONOMETRIC I NTEGRALS

Squaring it gives

1 2 1 1 1
cos4 x = 1 + cos(2x) = + cos(2x) + cos2 (2x)
4 4 2 4
Now by 1.8.5 a second time

1 1 1 1 + cos(4x)
cos4 x = + cos(2x) +
4 2 4 2
3 1 1
= + cos(2x) + cos(4x)
8 2 8
Now it’s easy to integrate

3 1 1
Z Z Z Z
4
cos xdx = dx + cos(2x)dx + cos(4x)dx
8 2 8
3 1 1
= x + sin(2x) + sin(4x) + C
8 4 32
Example 1.8.8

cos2 x sin2 xdx.

R
Example 1.8.9

Here we apply both 1.8.4 and 1.8.5.

1 
Z Z
2 2
 
cos x sin xdx = 1 + cos(2x) 1 − cos(2x) dx
4
1 
Z
1 − cos2 (2x) dx

=
4
We can then apply 1.8.5 again

1  1
Z

= 1 − (1 + cos(4x)) dx
4 2
1 
Z

= 1 − cos(4x) dx
8
1 1
= x− sin(4x) + C
8 32
Oof! We could also have done this one using 1.8.2 to write the integrand as sin2 (2x)
and then used 1.8.4 to write it in terms of cos(4x).

154
 I NTEGRATION 1.8 T RIGONOMETRIC I NTEGRALS

Rπ Rπ
Example 1.8.10 0
cos2 xdx and 0
sin2 xdx.
Rπ
Of course we can compute the definite integral 0 cos2 xdx by using the antideriva-
tive for cos2 x that we found in Example
R π 2 1.8.7. But here is a trickier way to evaluate
that integral, and also the integral 0 sin xdx at the same time, very quickly without
needing the antiderivative of Example 1.8.7.
Solution:
Rπ Rπ
• Observe that 0 cos2 xdx and 0 sin2 xdx are equal because they represent the
same area — look at the graphs below — the darkly shaded regions in the two
graphs have the same area and the lightly shaded regions in the two graphs
have the same area.

• Consequently,
Z π π Z π Z π 
1
Z
2 2 2 2
cos xdx = sin xdx = sin xdx + cos xdx
0 0 2 0 0
1 π 2
Z
sin x + cos2 x dx

=
2 0
1 π
Z
= dx
2 0
π
=
2

1.8.2 tt Integrating
R
tanm x secn xdx

The strategy for dealing with theseR integrals is similar to the strategy that we used
m n
to evaluate integrals of the form sin x cos xdx and again depends on the parity of
the exponents n and m. It uses 2
d d
tan x = sec2 x sec x = sec x tan x 1 + tan2 x = sec2 x
dx dx

2 You will need to memorise the derivatives of tangent and secant. However there is no need to
memorise 1 + tan2 x = sec2 x. To derive it very quickly just divide sin2 x + cos2 x = 1 by cos2 x.

155
 I NTEGRATION 1.8 T RIGONOMETRIC I NTEGRALS
R
We split the methods for integrating tanm x secn xdx into 5 cases which we list be-
low. These will become much more clear after an example (or two).

1 When m is odd and any n — rewrite the integrand in terms of sin x and cos x:
 m  n
m n sin x 1
tan x sec xdx = dx
cos x cos x
sinm−1 x
= sin xdx
cosn+m x
and then substitute u = cos x, du = − sin xdx, sin2 x = 1 − cos2 x = 1 − u2 . See
Examples 1.8.11 and 1.8.12.

2 Alternatively, if m is odd and n ≥ 1 move one factor of sec x tan x to the side
so that you can see sec x tan xdx in the integral, and substitute u = sec x, du =
sec x tan x dx and tan2 x = sec2 x − 1 = u2 − 1. See Example 1.8.13.

3 If n is even with n ≥ 2, move one factor of sec2 x to the side so that you can see
sec2 xdx in the integral, and substitute u = tan x, du = sec2 x dx and sec2 x =
1 + tan2 x = 1 + u2 . See Example 1.8.14.

4 When m is even and n = 0 — that is the integrand is just an even power of

tangent — we can still use the u = tan x substitution, after using tan2 x =
sec2 x − 1 (possibly more than once) to create a sec2 x. See Example 1.8.16.

5 This leaves the case n odd and m even. There are strategies like those above
for treating this case. But they are more complicated and also involve more
tricks (that basically have to be memorized). Examples using them are pro-
vided in the optional section entitled “Integrating sec x, csc x, sec3 x and csc3 x”,
below. A more straight forward strategy uses another technique called “partial
fractions”. We shall return to this strategy after we have learned about partial
fractions. See Example 1.10.5 and 1.10.6 in Section 1.10.

ttt m is odd — odd power of tangent

1.8.2.1

In this case we rewrite the integrand in terms of sine and cosine and then substitute
u = cos x, du = − sin xdx.
R
Example 1.8.11 tan xdx.

Solution:
1
• Write the integrand tan x = cos x
sin x.

• Now substitute u = cos x, du = − sin x dx just as we did in treating integrands

156
 I NTEGRATION 1.8 T RIGONOMETRIC I NTEGRALS

of the form sinm x cosn x with m odd.

1
Z Z
tan x dx = sin x dx substitute u = cos x
cos x
1
Z
= · (−1)du
u
= − log |u| + C
= − log |cos x| + C can also write in terms of secant
−1
= log |cos x| + C = log |sec x| + C
Example 1.8.11

tan3 xdx.
R
Example 1.8.12

Solution:
sin2 x
• Write the integrand tan3 x = cos3 x
sin x.

• Again substitute u = cos x, du = − sin x dx. We rewrite the remaining even

powers of sin x using sin2 x = 1 − cos2 x = 1 − u2 .

• Hence
sin2 x
Z Z
3
tan x dx = sin x dx substitute u = cos x
cos3 x
1 − u2
Z
= (−1)du
u3
u−2
= + log |u| + C
2
1
= + log |cos x| + C rewrite in terms of secant
2 cos2 x
1
= sec2 x − log |sec x| + C
2

ttt m is odd and n ≥ 1 — odd power of tangent and at least one secant
1.8.2.2

Here we collect a factor of tan x sec x and then substitute u = sec x and du = sec x tan xdx.
We can then rewrite any remaining even powers of tanx in terms of sec x using
tan2 x = sec2 x − 1 = u2 − 1.

157
 I NTEGRATION 1.8 T RIGONOMETRIC I NTEGRALS

tan3 x sec4 xdx.

R
Example 1.8.13

Solution:

• Start by factoring off one copy of sec x tan x and combine it with dx to form
sec x tan xdx, which will be du.

• Now substitute u = sec x, du = sec x tan xdx and tan2 x = sec2 x − 1 = u2 − 1.

• This gives
Z Z
3 4 2 3
tan x sec xdx = | {z x} sec
tan | {z x} sec
| x tan
{z xdx}
u2 −1 u3 du
Z
 2
= u − 1]u3 du
u6 u4
= − +C
6 4
1 1
= sec6 x − sec4 x + C
6 4

ttt n ≥ 2 is even — a positive even power of secant

1.8.2.3

In the previous case we substituted u = sec x, while in this case we substitute u =

tan x. When we do this we write du = sec2 xdx and then rewrite any remaining even
powers of sec x as powers of tan x using sec2 x = 1 + tan2 x = 1 + u2 .

sec4 xdx.
R
Example 1.8.14

Solution:
• Factor off one copy of sec2 x and combine it with dx to form sec2 xdx, which
will be du.
• Then substitute u = tan x, du = sec2 xdx and rewrite any remaining even pow-
ers of sec x as powers of tan x = u using sec2 x = 1 + tan2 x = 1 + u2 .
• This gives
Z Z
4 2 2
sec xdx = sec
| {z x} sec
| {zxdx}
1+u2 du
Z
1 + u2 ]du

=
u3
=u+ +C
3

158
I NTEGRATION 1.8 T RIGONOMETRIC I NTEGRALS

1
= tan x + tan3 x + C
3
Example 1.8.14

tan3 x sec4 xdx — redux.

R
Example 1.8.15

Solution: Let us revisit this example using this slightly different approach.

• Factor off one copy of sec2 x and combine it with dx to form sec2 xdx, which
will be du.

• Then substitute u = tan x, du = sec2 xdx and rewrite any remaining even pow-
ers of sec x as powers of tan x = u using sec2 x = 1 + tan2 x = 1 + u2 .

• This gives
Z Z
3 4 3 2 2
tan x sec xdx = | {z x} sec
tan | {z x} sec
| {zxdx}
u3 1+u2 du
Z
 3
= u + u5 ]du
u4 u6
= + +C
4 6
1 1
= tan4 x + tan6 x + C
4 6

• This is not quite the same as the answer we got above in Example 1.8.13. How-
ever we can show they are (nearly) equivalent. To do so we substitute v = sec x
and tan2 x = sec2 x − 1 = v 2 − 1:
1 1 1 1
tan6 x + tan4 x = (v 2 − 1)3 + (v 2 − 1)2
6 4 6 4
1 6 1
= (v − 3v 4 + 3v 2 − 1) + (v 4 − 2v 2 + 1)
6 4
6 4 2 4
v v v 1 v v2 1
= − + − + − +
6 2 2 6  4 2 4
v6 v4

1 1
= − + 0 · v2 + −
6 4 4 6
1 1 1
= sec6 x − sec4 x + .
6 4 12
So while 61 tan6 x + 14 tan4 x 6= 16 sec6 x − 41 sec4 x, they only differ by a constant.
Hence both are valid antiderivatives of tan3 x sec4 x.

159
 I NTEGRATION 1.8 T RIGONOMETRIC I NTEGRALS

ttt m is even and n = 0 — even powers of tangent

1.8.2.4

We integrate this by setting u = tan x. For this to work we need to pull one factor
of sec2 x to one side to form du = sec2 xdx. To find this factor of sec2 x we (perhaps
repeatedly) apply the identity tan2 x = sec2 x − 1.

tan4 xdx.
R
Example 1.8.16

Solution:

• There is no sec2 x term present, so we try to create it from tan4 x by using

tan2 x = sec2 x − 1.

tan4 x = tan2 x · tan2 x

= tan2 x sec2 x − 1
 

= tan2 x sec2 x − tan 2

| {z x}
sec2 x−1
2 2 2
= tan x sec x − sec x + 1

• Now we can substitute u = tan x, du = sec2 xdx.

Z Z Z Z
4 2 2 2
tan xdx = tan | {z x} sec
| {zxdx} − sec
| {zxdx} + dx
Z du du
2
Z u Z
= u2 du − du + dx
u3
= −u+x+C
3
tan3 x
= − tan x + x + C
3

tan8 xdx.
R
Example 1.8.17

Solution: Let us try the same approach.

• First pull out a factor of tan2 x to create a sec2 x factor:
tan8 x = tan6 x · tan2 x
= tan6 x · sec2 x − 1
 

= tan6 x sec2 x − tan6 x

The first term is now ready to be integrated, but we need to reapply the method
to the second term:
= tan6 x sec2 x − tan4 x · sec2 x − 1
 

160
I NTEGRATION 1.8 T RIGONOMETRIC I NTEGRALS

= tan6 x sec2 x − tan4 x sec2 x + tan4 x do it again

= tan x sec x − tan x sec x + tan x · sec2 x − 1
6 2 4 2 2
 

= tan6 x sec2 x − tan4 x sec2 x + tan2 x sec2 x − tan2 x and again

6 2 4 2 2 2
 2 
= tan x sec x − tan x sec x + tan x sec x − sec x − 1

• Hence
Z
tan8 xdx
Z
 6
tan x sec2 x − tan4 x sec2 x + tan2 x sec2 x − sec2 x + 1 dx

=
Z Z
 6 4 2
 2
= tan x − tan x + tan x − 1 sec xdx + dx
Z
 6
u − u4 + u2 − 1 du + x + C

=
u7 u5 u3
= − + −u+x+C
7 5 3
1 1 1
= tan7 x − tan5 x + tan3 x − tan x + x + C
7 5 3

Indeed this example suggests that for integer k ≥ 0:

1 1
Z
tan2k xdx = tan2k−1 (x) − tan2k−3 x + · · ·
2k − 1 2k − 3
− (−1) tan x + (−1)k x + C
k

Example 1.8.17

This last example also shows how we might integrate an odd power of tangent:

tan7 x.
R
Example 1.8.18

Solution: We follow the same steps

• Pull out a factor of tan2 x to create a factor of sec2 x:

tan7 x = tan5 x · tan2 x

= tan5 x · sec2 x − 1
 

= tan5 x sec2 x − tan5 x do it again

= tan x sec x − tan x · sec2 x − 1
5 2 3
 

= tan5 x sec2 x − tan3 x sec2 x + tan3 x and again

5 2 3 2
 2 
= tan x sec x − tan x sec x + tan x sec x − 1
= tan5 x sec2 x − tan3 x sec2 x + tan x sec2 x − tan x

161
 I NTEGRATION 1.8 T RIGONOMETRIC I NTEGRALS

• Now we can substitute u = tan x and du = sec2 xdx and also use the result from
Example 1.8.11 to take care of the last term:
Z Z
7
tan5 x sec2 x − tan3 x sec2 x + tan x sec2 x dx
 
tan xdx =
Z
− tan xdx

Now factor out the common sec2 x term and integrate tan x via Example 1.8.11
Z
tan5 x − tan3 x + tan x sec xdx − log | sec x| + C
 
=
Z
 5
u − u3 + u du − log | sec x| + C

=
u6 u4 u2
= − + − log | sec x| + C
6 4 2
1 1 1
= tan6 x − tan4 x + tan2 x − log | sec x| + C
6 4 2

This example suggests that for integer k ≥ 0:

1 1
Z
tan2k+1 xdx = tan2k (x) − tan2k−2 x + · · ·
2k 2k − 2
1
− (−1)k tan2 x + (−1)k log | sec x| + C
2
Example 1.8.18

Of course we have not considered integrals involving powers of cot x and csc x.
But they can be treated in much the same way as tan x and sec x were.

1.8.3 tt Optional — Integrating sec x, csc x, sec3 x and csc3 x

As noted above, when n is odd and m is even, one can use similar strategies as to
the previous cases. However the computations are often more involved and more
tricks need to be deployed. For this reason we make this section optional — the
computations are definitely non-trivial. Rather than trying to construct a coherent
“method” for this case, we instead give some examples to give the idea of what to
expect.

162
I NTEGRATION 1.8 T RIGONOMETRIC I NTEGRALS

R
Example 1.8.19 sec xdx — by trickery.

Solution: There is a very sneaky trick to compute this integral.

sec x+tan x
• The standard trick for this integral is to multiply the integrand by 1 = sec x+tan x

sec x + tan x sec2 x + sec x tan x

sec x = sec x =
sec x + tan x sec x + tan x

• Notice now that the numerator of this expression is exactly the derivative its
denominator. Hence we can substitute u = sec x + tan x and du = (sec x tan x +
sec2 x) dx.

• Hence
sec x + tan x sec2 x + sec x tan x
Z Z Z
sec xdx = sec x dx = dx
sec x + tan x sec x + tan x
1
Z
= du
u
= log |u| + C
= log | sec x + tan x| + C

• The above trick appears both totally unguessable and very hard to remember.
Fortunately, there is a simple wayWe thank Serban Raianu for bringing this to
our attention. to recover the trick. Here it is.

◦ The goal is to guess a function whose derivative is sec x.

◦ So get out a table of derivatives and look for functions whose derivatives
at least contain sec x. There are two:
d
tan x = sec2 x
dx
d
sec x = tan x sec x
dx

◦ Notice that if we add these together we get

d

sec x + tan x = (sec x + tan x) sec x =⇒
dx
d


dx
sec x + tan x
= sec x
sec x + tan x

◦ We’ve done it! The right hand side is sec x and the left hand side is the
derivative of log | sec x + tan x|.

R
There is another method for integrating sec xdx, that is more tedious, but more

163
I NTEGRATION 1.8 T RIGONOMETRIC I NTEGRALS

straight forward. In particular, it does not involve a memorized trick. We first use
the substitution u = sin x, du = cos x dx, together with cos2 x = 1 − sin2 x = 1 − u2 .
This converts the integral into
1 cos x dx
Z Z Z
sec xdx = dx =
cos x cos2 x
du
Z
=
1 − u2 u=sin x
1
The integrand 1−u 2 is a rational function, i.e. a ratio of two polynomials. There is

a procedure, called the method of partial fractions, that may be used to integrate
any rational function. We shall learn aboutR it in Section
R du 1.10 “Partial Fractions”.
The detailed evaluation of the integral sec x dx = 1−u2 by the method of partial
fractions is presented in Example 1.10.5 below. R du
In addition, there is a standard trick for evaluating 1−u 2 that allows us to avoid

going through the whole partial fractions algorithm.

R
Example 1.8.20 sec xdx — by more trickery.

Solution: We have already seen that

du
Z Z
sec xdx =
1 − u2 u=sin x

The trick uses the obervations that

1 1+u−u 1 u
• 1−u2
= 1−u2
= 1−u
− 1−u2
1
• 1−u
has antiderivative − log(1 − u) (for u < 1)
d
• The derivative du (1 − u2 ) = −2u of the denominator of 1−u u
2 is the same, up to
u
a factor of −2, as the numerator of 1−u 2 . So we can easily evaluate the integral
u 2
of 1−u2 by substituting v = 1 − u , dv = −2u du.
Z dv
u du 1
Z
−2
2
= = − log(1 − u2 ) + C
1−u v v=1−u2 2

Combining these observations gives

Z  Z 
du 1 u
Z Z
sec xdx = = du − du
1 − u2 u=sin x 1−u 1 − u2 u=sin x
h 1 2
i
= − log(1 − u) + log(1 − u ) + C
2 u=sin x
1 2
= − log(1 − sin x) + log(1 − sin x) + C
2
1 1
= − log(1 − sin x) + log(1 − sin x) + log(1 + sin x) + C
2 2
1 1 + sin x
= log +C
2 1 − sin x

164
I NTEGRATION 1.8 T RIGONOMETRIC I NTEGRALS

Example 1.8.20 has given the answer

1 1 + sin x
Z
sec xdx = log +C
2 1 − sin x
which appears to be different than the answer in Example 1.8.19. But they are really
the same since
1 + sin x (1 + sin x)2 (1 + sin x)2
= =
1 − sin x 1 − sin2 x cos2 x
1 1 + sin x 1 (1 + sin x)2 sin x + 1
=⇒ log = log 2
= log = log | tan x + sec x|
2 1 − sin x 2 cos x cos x
Oof!

csc xdx — by the u = tan x2 substitution.

R
Example 1.8.21
R
Solution: The integral csc xdx may also be evaluated by both the methods above.
That is either

• by multiplying the integrand by a cleverly chosen 1 = cot x−csc x

cot x−csc x
and then sub-
stituting u = cot x − csc x, du = (− csc2 x + csc x cot x) dx, or
R R du
• by substituting u = cos x, du = − sin x dx to give csc xdx = − 1−u 2 and then

using the method of partial fractions.

These two methods give the answers

1 1 + cos x
Z
csc xdx = log | cot x − csc x| + C = − log +C (?)
2 1 − cos x
R
In this example, we shall evaluate csc xdx by yet a third method, which can be used
to integrate rational functions a of sin x and cos x.

• This method uses the substitution

x 2
x = 2 arctan u i.e. u = tan and dx = du
2 1 + u2
— a half-angle substitution.

• To express sin x and cos x in terms of u, we first use the double angle trig iden-
tities (Equations 1.8.2 and 1.8.3 with x 7→ x2 ) to express sin x and cos x in terms
of sin x2 and cos x2 :
x x
sin x = 2 sin cos
2 2
2 x x
cos x = cos − sin2
2 2

• We then use the triangle

165
I NTEGRATION 1.8 T RIGONOMETRIC I NTEGRALS

to express sin x2 and cos x2 in terms of u. The bottom and right hand sides of the
triangle have been chosen so that tan x2 = u. This tells us that

x u x 1
sin =√ cos =√
2 1 + u2 2 1 + u2

• This in turn implies that:

x x u 1 2u
sin x = 2 sin cos = 2 √ √ =
2 2 2
1+u 1+u 2 1 + u2
x x 1 u2 1 − u2
cos x = cos2 − sin2 = − =
2 2 1 + u2 1 + u2 1 + u2
Oof!
R
• Let’s use this substitution to evaluate csc x dx.

1 1 + u2 2 1
Z Z Z Z
csc xdx = dx = 2
du = du
sin x 2u 1 + u u
x
= log |u| + C = log tan + C
2
To see that this answer is really the same as that in (?), note that

cos x − 1 −2 sin2 (x/2) x

cot x − csc x = = = − tan
sin x 2 sin(x/2) cos(x/2) 2

a A rational function of sin x and cos x is a ratio with both the numerator and denominator being
finite sums of terms of the form a sinm x cosn x, where a is a constant and m and n are positive
integers.

Example 1.8.21

sec3 xdx — by trickery.

R
Example 1.8.22

sec3 xdx is integration by parts.

R
Solution: The standard trick used to evaluate

166
I NTEGRATION 1.8 T RIGONOMETRIC I NTEGRALS

• Set u = sec x, dv = sec2 xdx. Hence du = sec x tan xdx, v = tan x and
Z Z
3 2
sec xdx = sec x sec
|{z} | {zxdx}
u dv
Z
= |{z}
sec x tan
| {z x} − tan
| {z x} |sec x tan
{z xdx}
u v v du

• Since tan2 x + 1 = sec2 x, we have tan2 x = sec2 x − 1 and

Z Z
sec xdx = sec x tan x − [sec3 x − sec x]dx
3

Z
= sec x tan x + log | sec x + tan x| + C − sec3 xdx
R
where we used sec xdx = log | sec x + tan x| + C, which we saw in Example
1.8.19.

• Now moving the sec3 xdx from the right hand side to the left hand side
R

Z
2 sec3 xdx = sec x tan x + log | sec x + tan x| + C and so
1 1
Z
sec3 xdx = sec x tan x + log | sec x + tan x| + C
2 2
for a new arbitrary constant C (which is just one half the old one).
Example 1.8.22

sec3 dx can also be evaluated by two other methods.

R
The integral
R du
• Substitute u = sin x, du = cos xdx to convert sec3 xdx into [1−u
R
2 ]2 and eval-

uate the latter using the method of partial fractions. This is done in Exam-
ple 1.10.6 in Section 1.10.

• Use the u = tan x2 substitution. We use this method to evaluate csc3 xdx in
R

Example 1.8.23, below.

csc3 xdx – by the u = tan x2 substitution.

R
Example 1.8.23

Solution: Let us use the half-angle substitution that we introduced in Example 1.8.21.

• In this method we set

x 2 2u 1 − u2
u = tan dx = du sin x = cos x =
2 1 + u2 1 + u2 1 + u2

167
I NTEGRATION 1.8 T RIGONOMETRIC I NTEGRALS

• The integral then becomes

1
Z Z
3
csc xdx = dx
sin3 x
3
1 + u2  2
Z 
= du
2u 1 + u2
1 1 + 2u2 + u4
Z
= du
4 u3
1 n u−2 u2 o
= + 2 log |u| + +C
4 −2 2
1n x x xo
= − cot2 + 4 log tan + tan2 +C
8 2 2 2
Oof!

• This is a perfectly acceptable answer. But if you don’t like the x2 ’s, they may be
eliminated by using

x x sin2 x2 cos2 x2
tan2 − cot2 = −
2 2 cos2 x2 sin2 x2
sin4 x2 − cos4 x2
=
sin2 x2 cos2 x2
sin2 x
− cos2 x2 sin2 x2 + cos2 x2



2
=
sin2 x2 cos2 x2
sin2 x2 − cos2 x2 x x
= since sin2 + cos2 = 1
sin2 x2 cos2 x2 2 2
− cos x
= 1 2 by 1.8.2 and 1.8.3
4
sin x

and
x sin x2 sin2 x2
tan = =
2 cos x2 sin x2 cos x2
1
2
[1 − cos x]
= 1 by 1.8.2 and 1.8.3
2
sin x

So we may also write

1 1
Z
csc3 xdx = − cot x csc x + log | csc x − cot x| + C
2 2
Example 1.8.23

That last optional section was a little scary — let’s get back to something a little

168
 I NTEGRATION 1.8 T RIGONOMETRIC I NTEGRALS

easier.

1.8.4 tt Exercises

Recall that we are using log x to denote the logarithm of x with base e. In other
courses it is often denoted ln x.

Exercises — Stage 1

Z π/4
1. Suppose you want to evaluate sin x cosn xdx using the substitution
0
u = cos x. Which of the following need to be true for your substitution
to work?

a n must be even

b n must be odd

c n must be an integer

d n must be positive

e n can be any real number

Z
2. Evaluate secn x tan xdx, where n is a strictly positive integer.

3. Derive the identity tan2 x + 1 = sec2 x from the easier-to-remember identity

sin2 x + cos2 x = 1.

Exercises — Stage 2 Questions 4 through 10 deal with powers of sines and cosines.
Review Section 1.8.1 in the notes for integration strategies.Questions 12 through 21
deal with powers of tangents and secants. Review Section 1.8.2 in the notes for strate-
gies.
Z
3
4. *. Evaluate cos x dx.

Z π
5. *. Evaluate cos2 x dx.
0

Z
6. *. Evaluate sin36 t cos3 t dt.

sin3 x
Z
7. Evaluate dx.
cos4 x

169
I NTEGRATION 1.8 T RIGONOMETRIC I NTEGRALS
Z π/3
8. Evaluate sin4 xdx.
0
Z
9. Evaluate sin5 xdx.

Z
10. Evaluate sin1.2 x cos xdx.

Z
11. Evaluate tan x sec2 xdx.

Z
12. *. Evaluate tan3 x sec5 x dx.

Z
13. *. Evaluate sec4 x tan46 x dx.

Z
14. Evaluate tan3 x sec1.5 xdx.
Z
15. Evaluate tan3 x sec2 xdx.
Z
16. Evaluate tan4 x sec2 xdx.
Z
17. Evaluate tan3 x sec−0.7 xdx.
Z
18. Evaluate tan5 xdx.

Z π/6
19. Evaluate tan6 xdx.
0

Z π/4
20. Evaluate tan8 x sec4 xdx.
0
√
Z
21. Evaluate tan x sec xdx.

Z
22. Evaluate sec8 θ tane θdθ.

Exercises — Stage 3

170
 I NTEGRATION 1.9 T RIGONOMETRIC S UBSTITUTION

23. *. A reduction formula.

a Let n be a positive integer with n ≥ 2. Derive the reduction for-
mula
tann−1 (x)
Z Z
n
tan (x) dx = − tann−2 (x) dx.
n−1
Z π/4
b Calculate tan6 (x) dx.
0

Z
24. Evaluate tan5 x cos2 xdx.

1
Z
25. Evaluate dθ.
cos2 θ
Z
26. Evaluate cot xdx.

Z
27. Evaluate ex sin(ex ) cos(ex )dx.
Z
28. Evaluate sin(cos x) sin3 xdx.

Z
29. Evaluate x sin x cos xdx.

1.9q Trigonometric Substitution

1.9.1 tt Trigonometric Substitution

In this section we discuss substitutions that simplify integrals containing square

roots of the form
√ √ √
a2 − x 2 a2 + x 2 x 2 − a2 .

When the integrand contains one of these square roots, then we can use trigonomet-
ric substitutions to eliminate them. That is, we substitute

x = a sin u or x = a tan u or x = a sec u

171
I NTEGRATION 1.9 T RIGONOMETRIC S UBSTITUTION

and then use trigonometric identities

sin2 θ + cos2 θ = 1 and 1 + tan2 θ = sec2 θ

to simplify the result. To be more precise, we can

√
• eliminate a2 − x2 from an integrand by substituting x = a sin u to give
√ p √
a2 − x2 = a2 − a2 sin2 u = a2 cos2 u = |a cos u|
√
• eliminate a2 + x2 from an integrand by substituting x = a tan u to give
√ p √
a2 + x2 = a2 + a2 tan2 u = a2 sec2 u = |a sec u|
√
• eliminate x2 − a2 from an integrand by substituting x = a sec u to give
√ √ √
x2 − a2 = a2 sec2 u − a2 = a2 tan2 u = |a tan u|

Be very careful with signs and absolute values when using this substitution.
See Example 1.9.6.

When we have used substitutions before, we usually gave the new integration
variable, u, as a function of the old integration variable x. Here we are doing the
reverse — we are giving the old integration variable, x, in terms of the new integra-
tion variable u. We may do so, as long as we may invert to get u as a function of x.
For example, with x = a sin u, we may take u = arcsin xa . This is a good time for you
to review the definitions of arcsin θ, arctan θ and arcsec θ. See Section 2.12, “Inverse
Functions”, of the CLP-1 text.
As a warm-up, consider the area of a quarter of the unit circle.

Example 1.9.1 Quarter of the unit circle.

Compute the area of the unit circle lying in the first quadrant.
Solution: We know that the answer is π4 , but we can also compute this as an integral
— we saw this way back in Example 1.1.16:
Z 1 √
area = 1 − x2 dx
0

dx
• To simplify the integrand we substitute x = sin u. With this choice du
= cos u
and so dx = cos udu.

• We also need to translate the limits of integration and it is perhaps easiest to do

this by writing u as a function of x — namely u(x) = arcsin x. Hence u(0) = 0
and u(1) = π2 .

172
I NTEGRATION 1.9 T RIGONOMETRIC S UBSTITUTION

• Hence the integral becomes

π
Z 1 √ Z
2 p
1− x2 dx = 1 − sin2 u · cos udu
0 0
π
Z
2 √
= cos2 u · cos udu
0
Z π
2
= cos2 udu
0
√
Notice that here we have used that the positive square root cos2 u = | cos u| =
cos u because cos(u) ≥ 0 for 0 ≤ u ≤ π2 .

• To go further we use the techniques of Section 1.8.

π
Z 1 √ Z
2 1 + cos 2u
1 − x2 dx = cos2 udu and since cos2 u =
0 0 2
π
1
Z
2
= (1 + cos(2u))du
2 0
  π2
1 1
= u + sin(2u)
2 2 0
 
1 π sin π sin 0
= −0+ −
2 2 2 2
π
= X
4
Example 1.9.1

2
√x
R
Example 1.9.2 1−x2
dx.

Solution: We proceed much as we did in the previous example.

dx
• To simplify the integrand we substitute x = sin u. With this choice du
= cos u
and so dx = cos udu. Also note that u = arcsin x.

• The integral becomes

x2 sin2 u
Z Z
√ dx = p · cos udu
1 − x2 1 − sin2 u
sin2 u
Z
= √ · cos udu
cos2 u

• To proceed further we need to get rid of the square-root. Since u = arcsin x has
domain −1 ≤ x ≤ 1 and range − π2 ≤ u ≤ π2 , it follows that cos u ≥ 0 (since

173
I NTEGRATION 1.9 T RIGONOMETRIC S UBSTITUTION

cosine is non-negative on these inputs). Hence

√ π π
cos2 u = cos u when − ≤u≤
2 2

• So our integral now becomes

x2 sin2 u
Z Z
√ dx = √ · cos udu
1 − x2 cos2 u
sin2 u
Z
= · cos udu
cos u
Z
= sin2 udu
1
Z
= (1 − cos 2u)du by Equation 1.8.4
2
u 1
= − sin 2u + C
2 4
1 1
= arcsin x − sin(2 arcsin x) + C
2 4

• We can simplify this further using a double-angle identity. Recall that u =

arcsin x and that x = sin u. Then

sin 2u = 2 sin u cos u

2 2
We can replace cos u using
p cos u = 1 − sin u. Taking a square-root of this
formula gives cos u = ± 1 − sin2 u. We need the positive branch here since
cos u ≥ 0 when − π2 ≤ u ≤ π2 (which is exactly the range of arcsin x). Continuing
along:
p
sin 2u = 2 sin u · 1 − sin2 u
√
= 2x 1 − x2

Thus our solution is

x2 1 1
Z
√ dx = arcsin x − sin(2 arcsin x) + C
1 − x2 2 4
1 1 √
= arcsin x − x 1 − x2 + C
2 2
Example 1.9.2

The above two example illustrate the main steps of the approach. The next ex-
ample is similar, but with more complicated limits of integration.

174
I NTEGRATION 1.9 T RIGONOMETRIC S UBSTITUTION

Rr√
Example 1.9.3 a
r2 − x2 dx.

Let’s find the area of the shaded region in the sketch below.

We’ll set up√the integral using vertical strips. The strip in the figure has width dx
and height r2 − x2 . So the area is given by the integral
Z r√
area = r2 − x2 dx
a

Which is very similar to the previous example.

Solution:
• To evaluate the integral we substitute
dx
x = x(u) = r sin u dx = du = r cos udu
du
It is also helpful to write u as a function of x — namely u = arcsin xr .

• The integral runs from x = a to x = r. These correspond to

r π
u(r) = arcsin = arcsin 1 =
r 2
a
u(a) = arcsin which does not simplify further
r

• The integral then becomes

Z r√ Z π
2 p
2 2
r − x dx = r2 − r2 sin2 u · r cos udu
a arcsin(a/r)
Z π
2 p
= r2 1 − sin2 u · cos udu
arcsin(a/r)
π
Z
2 √
2
=r cos2 u · cos udu
arcsin(a/r)

To proceed further (as we did in Examples 1.9.1 and 1.9.2) we need to think
about whether cos u is positive or negative.

175
I NTEGRATION 1.9 T RIGONOMETRIC S UBSTITUTION

• Since a (as shown in the diagram) satisfies 0 ≤ a ≤ r, we know that u(a) lies
between arcsin(0) = 0 and arcsin(1) = π2 . Hence the variable u lies between 0
and π2 , and on this range cos u ≥ 0. This allows us get rid of the square-root:
√
cos2 u = | cos u| = cos u

• Putting this fact into our integral we get

π
Z r√ Z
2 √
2
r2 − x2 dx =r cos2 u · cos udu
a arcsin(a/r)
Z π
2
2
=r cos2 udu
arcsin(a/r)

1+cos 2u
Recall the identity cos2 u = 2
from Section 1.8
π
r2
Z
2
= (1 + cos 2u)du
2 arcsin(a/r)
 π2
r2

1
= u + sin(2u)
2 2 arcsin(a/r)
2
 
r π 1 1
= + sin π − arcsin(a/r) − sin(2 arcsin(a/r))
2 2 2 2
2
 
r π 1
= − arcsin(a/r) − sin(2 arcsin(a/r))
2 2 2

Oof! But there is a little further to go before we are done.

• We can again simplify the term sin(2 arcsin(a/r)) using a double angle identity.
Set θ = arcsin(a/r). Then θ is the angle in the triangle on the right below. By
the double angle formula for sin(2θ) (Equation 1.8.2)

sin(2θ) = 2 sin θ cos θ

√
a r 2 − a2
=2 .
r r

176
I NTEGRATION 1.9 T RIGONOMETRIC S UBSTITUTION

• So finally the area is

Z r√
area = r2 − x2 dx
a
r2 π
 
1
= − arcsin(a/r) − sin(2 arcsin(a/r))
2 2 2
πr 2
r 2
a√ 2
= − arcsin(a/r) − r − a2
4 2 2

• This is a relatively complicated formula, but we can make some “reasonable-

ness” checks, by looking at special values of a.

◦ If a = 0 the shaded region, in the figure at the beginning of this example, is

exactly one quarter of a disk of radius r and so has area 14 πr2 . Substituting
a = 0 into our answer does indeed give 14 πr2 .
◦ At the other extreme, if a = r, the shaded region disappears completely
and so has area 0. Subbing a = r into our answer does indeed give 0, since
arcsin 1 = π2 .
Example 1.9.3

Rr √
Example 1.9.4 a
x r2 − x2 dx.
Rr √
The integral a x r2 − x2 dx looks a lot like the integral we just did in the previous
3 examples. It can also be evaluated using the trigonometric substitution x = r sin u
— but that is unnecessarily complicated. Just because you have now learned how to
use trigonometric substitution a doesn’t mean that you should forget everything you
learned before.
Solution: This integral is much more easily evaluated using the simple substitution
u = r 2 − x2 .

• Set u = r2 − x2 . Then du = −2xdx, and so

Z r √ 0 √ du
Z
2 2
x r − x dx = u
a r2 −a2 −2
 3/2 0
1 u
=−
2 3/2 r2 −a2
1 3/2
= r 2 − a2
3

a To paraphrase the Law of the Instrument, possibly Mark Twain and definitely some psycholo-
gists, when you have a shiny new hammer, everything looks like a nail.

177
I NTEGRATION 1.9 T RIGONOMETRIC S UBSTITUTION

Enough sines and cosines — let us try a tangent substitution.

√dx
R
Example 1.9.5 x2 9+x2
.

Solution: √
As per our guidelines at the start of this section, the presence of the square
root term 32 + x2 tells us to substitute x = 3 tan u.

• Substitute

x = 3 tan u dx = 3 sec2 udu

This allows us to remove the square root:

√ p p √
9 + x2 = 9 + 9 tan2 u = 3 1 + tan2 u = 3 sec2 u = 3| sec u|

• Hence our integral becomes

dx 3 sec2 u
Z Z
√ = du
x2 9 + x2 9 tan2 u · 3| sec u|

• To remove the absolute value we must consider the range of values of u in the
integral. Since x = 3 tan u we have u = arctan(x/3). The range a of arctangent is
− π2 ≤ arctan ≤ π2 and so u = arctan(x/3) will always like between − π2 and + π2 .
Hence cos u will always be positive, which in turn implies that | sec u| = sec u.

• Using this fact our integral becomes:

dx 3 sec2 u
Z Z
√ = du
x2 9 + x2 27 tan2 u| sec u|
1 sec u
Z
= du since sec u > 0
9 tan2 u

• Rewrite this in terms of sine and cosine

dx 1 sec u
Z Z
√ = du
x2 9 + x2 9 tan2 u
1 1 cos2 u 1 cos u
Z Z
= · 2 du = du
9 cos u sin u 9 sin2 u
Now we can use the substitution rule with y = sin u and dy = cos udu
1 dy
Z
=
9 y2
1
=− +C
9y
1
=− +C
9 sin u

178
I NTEGRATION 1.9 T RIGONOMETRIC S UBSTITUTION

• The original integral was a function of x, so we still have to rewrite sin u in

terms of x. Remember that x = 3 tan u or u = arctan(x/3). So u is the angle
shown in the triangle below and we can read off the triangle that
x
sin u = √
9 + x2
√
dx 9 + x2
Z
=⇒ √ =− +C
x2 9 + x2 9x

a To be pedantic, we mean the range of the “standard” arctangent function or its “principle value”.
One can define other arctangent functions with different ranges.

Example 1.9.5

2
√x
R
Example 1.9.6 x2 −1
dx.

Solution: This one requires a secant substitution, but otherwise is very similar to
those above.
• Set x = sec u and dx = sec u tan udu. Then
x2 sec2 u
Z Z
√ dx = √ sec u tan udu
x2 − 1 sec 2u−1

tan u
Z
= sec3 u · √ du since tan2 u = sec2 u − 1
2
tan u
tan u
Z
= sec3 u · du
| tan u|
• As before we need to consider the range of u values in order to determine the
sign of tan u. Notice that the integrand is only defined when either x < −1
or x > 1; thus we should treat the cases x < −1 and x > 1 separately. Let us
assume that x > 1 and we will come back to the case x < −1 at the end of the
example.
When x > 1, our u = arcsec x takes values in (0, π2 ). This follows since when
0 < u < π2 , we have 0 < cos u < 1 and so sec u > 1. Further, when 0 < u < π2 ,
we have tan u > 0. Thus | tan u| = tan u.

179
I NTEGRATION 1.9 T RIGONOMETRIC S UBSTITUTION

• Back to our integral, when x > 1:

x2 tan u
Z Z
√ dx = sec3 u · du
x2 − 1 | tan u|
Z
= sec3 udu since tan u ≥ 0

This is exactly Example 1.8.22

1 1
= sec u tan u + log | sec u + tan u| + C
2 2

• Since we started with a function of x we need to finish with one. We know that
sec u = x and then we can use trig identities
√
tan2 u = sec2 u − 1 = x2 − 1 so tan u = ± x2 − 1

but we know
√
tan u ≥ 0 so tan u = x2 − 1

Thus
x2 1 √ 1 √
Z
√ dx = x x2 − 1 + log |x + x2 − 1| + C
x2 − 1 2 2

• The above holds when x > 1. We can confirm that it is also true when x < −1
by showing the right-hand side is a valid antiderivative of the integrand. To
do so we must differentiate
√ our answer. Notice that we do not need to consider
2
the sign of x + x − 1 when we differentiate since we have already seen that

d 1
log |x| =
dx x
when either x < 0 or x > 0. So the following computation applies to both x > 1
and x < −1. The expressions become quite long so we differentiate each term
separately:

d h √ 2 i √ 2

x
x x −1 = x2 − 1 + √
dx x2 − 1
1  2
(x − 1) + x2

=√
x2 − 1
√
 
d 1 x
log x + x2 − 1 = √ · 1+ √
dx x + x2 − 1 x2 − 1
√
1 x + x2 − 1
= √ · √
x + x2 − 1 x2 − 1

180
I NTEGRATION 1.9 T RIGONOMETRIC S UBSTITUTION

1
=√
x2 −1
Putting things together then gives us

d 1 √ 2 √
 
1
x x − 1 + log |x + x2 − 1| + C
dx 2 2
1  2
(x − 1) + x2 + 1 + 0

= √
2 x2 − 1
x2
=√
x2 − 1
This tells us that our answer for x > 1 is also valid when x < −1 and so
x2 1 √ 1 √
Z
√ dx = x x2 − 1 + log |x + x2 − 1| + C
x2 − 1 2 2

when x < −1 and when x > 1.

In this example, we were lucky. The answer that we derived for x > 1 happened
to be also valid for x < −1. This does not always happen with the x = a sec u
substitution. When it doesn’t, we have to apply separate x > a and x < −a analyses
that are very similar to our x > 1 analysis above. Of course that doubles the tedium.
So in the CLP-2 problem book, we will not pose questions that require separate x > a
and x < −a computations.
Example 1.9.6

The method, as we have demonstrated it above, works when our integrand con-
tains the square root of very specific families of quadratic polynomials. In fact, the
same method works for more general quadratic polynomials — all we need to do is
complete the square 1 .

R5 √
x2 −2x−3
Example 1.9.7 3 x−1
dx.

This time we have an integral with a square root in the integrand, but the argument
of the square
√ √ a quadratic function of x, is not in one of the standard forms
√ root, while
a2 − x2 , a2 + x2 , x2 − a2 . The reason that it is not in one of those forms is that
the argument, x2 − 2x − 3, contains a term , namely −2x that is of degree one in x. So

1 If you have not heard of “completing the square” don’t worry. It is not a difficult method and it
will only take you a few moments to learn. It refers to rewriting a quadratic polynomial

P (x) = ax2 + bx + c as P (x) = a(x + d)2 + e

for new constants d, e.

181
I NTEGRATION 1.9 T RIGONOMETRIC S UBSTITUTION

we try to manipulate it into one of the standard forms by completing the square.
Solution:

• We first rewrite the quadratic polynomial x2 − 2x − 3 in the form (x − a)2 + b for

some constants a, b. The easiest way to do this is to expand both expressions
and compare coefficients of x:

x2 − 2x − 3 = (x − a)2 + b = (x2 − 2ax + a2 ) + b

So — if we choose −2a = −2 (so the coefficients of x1 match) and a2 + b = −3

(so the coefficients of x0 match), then both expressions are equal. Hence we set
a = 1 and b = −4. That is

x2 − 2x − 3 = (x − 1)2 − 4

Many of you may have seen this method when learning to sketch parabolas.

• Once this is done we can convert the square root of the integrand into a stan-
dard form by making the simple substitution y = x − 1. Here goes
Z 5√ 2
x − 2x − 3
dx
3 x−1
Z 5p
(x − 1)2 − 4
= dx
3 x−1
Z 4p 2
y −4
= dy with y = x − 1, dy = dx
2 y
Z π/3 √
4 sec2 u − 4
= 2 sec u tan udu with y = 2 sec u
0 2 sec u
and dy = 2 sec u tan udu

Notice that we could also do this in fewer steps by setting (x−1) = 2 sec u, dx =
2 sec u tan udu.

• To get the limits of integration we used that

2
◦ the value of u that corresponds to y = 2 obeys 2 = y = 2 sec u = cos u
or
cos u = 1, so that u = 0 works and
2
◦ the value of u that corresponds to y = 4 obeys 4 = y = 2 sec u = cos u
or
cos u = 12 , so that u = π3 works.

• Now returning to the evaluation of the integral, we simplify and continue.

Z 5√ 2 Z π/3 √
x − 2x − 3
dx = 2 sec2 u − 1 tan udu
3 x − 1 0

182
 I NTEGRATION 1.9 T RIGONOMETRIC S UBSTITUTION

Z π/3
=2 tan2 udu since sec2 u = 1 + tan2 u
0

In taking the square root of sec2 u − 1 = tan2 u we used that tan u ≥ 0 on the
range 0 ≤ u ≤ π3 .
Z π/3
sec2 u − 1 du since sec2 u = 1 + tan2 u, again
 
=2
0
h iπ/3
= 2 tan u − u
0
√ π
=2 3−
3
Example 1.9.7

1.9.2 tt Exercises

Recall that we are using log x to denote the logarithm of x with base e. In other
courses it is often denoted ln x.
Exercises — Stage 1

1. *. For each of the following integrals, choose the substitution that is

most beneficial for evaluating the integral.
2x2
Z
a √ dx
9x2 − 16
x4 − 3
Z
b √ dx
1 − 4x2
Z
−5/2
c (25 + x2 ) dx

2. For each of the following integrals, choose a trigonometric substitution

that will eliminate the roots.
1
Z
a √ dx
x2 − 4x + 1
(x − 1)6
Z
b dx
(−x2 + 2x + 4)3/2
1
Z
c √ dx
4x2 + 6x + 10

183
I NTEGRATION 1.9 T RIGONOMETRIC S UBSTITUTION

Z √
d x2 − xdx

3. In each part of this question, assume θ is an angle in the interval [0, π/2].
1
a If sin θ = , what is cos θ ?
20
b If tan θ = 7, what is csc θ ?
√
x−1
c If sec θ = , what is tan θ ?
2

4. Simplify the following expressions.

a sin arccos x2

  
b sin arctan √13
√
c sec (arcsin ( x))

Exercises — Stage 2
1
Z
5. *. Evaluate dx.
(x2 + 4)3/2
4
1
Z
6. *. Evaluate dx. Your answer may not contain inverse
(4 + x2 )3/2
0
trigonometric functions.

5/2
dx
Z
7. *. Evaluate √ .
0 25 − x2

dx
Z Z
8. *. Evaluate √ . You may use that sec dx = log sec x +
x2 + 25
tan x + C.

x+1
Z
9. Evaluate √ dx.
2x2 + 4x
dx
Z
10. *. Evaluate √ .
x2 x2 + 16
dx
Z
11. *. Evaluate √ for x ≥ 3. Do not include any inverse trigonomet-
x2 x2 − 9
ric functions in your answer.

184
I NTEGRATION 1.9 T RIGONOMETRIC S UBSTITUTION

Z π/4
12. *. (a) Show that cos4 θdθ = (8 + 3π)/32.
Z 1 0
dx
(b) Evaluate .
−1 (x2 + 1)3

π/12
15x3
Z
13. Evaluate dx.
−π/12 (x2 + 1)(9 − x2 )5/2
√
Z
14. *. Evaluate 4 − x2 dx.
Z √
25x2 − 4
15. *. Evaluate dx for x > 25 .
x
√
17
x3
Z
16. Evaluate √
√ dx.
10 x2 − 1

dx
Z
17. *. Evaluate √ .
3 − 2x − x2

1
Z
18. Evaluate √ dx for x > 2.
(2x − 3)3 4x2 − 12x + 8

1
x2
Z
19. Evaluate 2 3/2
dx.
0 (x +R1)
You may use that sec xdx = log | sec x + tan x| + C.

1
Z
20. Evaluate dx.
(x2 + 1)2

Exercises — Stage 3
x2
Z
21. Evaluate √ dx.
x2 − 2x + 2
1 1
Z
You may assume without proof that sec3 θdθ = sec θ tan θ+ log | sec θ+
2 2
tan θ| + C.

1
Z
22. Evaluate √ dx.
3x2 +
R 5x
You may use that sec xdx = log | sec x + tan x| + C.

(1 + x2 )3/2
Z Z
23. Evaluate dx. You may use the fact that csc θdθ = log | cot θ −
x

185
I NTEGRATION 1.9 T RIGONOMETRIC S UBSTITUTION

csc θ| + C.

2
2
24. Below is the graph of the ellipse x4 + y2 = 1. Find the area of the
shaded region using the ideas from this section.
y

−1

|x|
25. Let f (x) = √
4
, and let R be the region between f (x) and the x-axis
1 − x2
over the interval [− 21 , 12 ].

a Find the area of R.

b Find the volume of the solid formed by rotating R about the x-axis.
√
Z Z
26. Evaluate x
1 + e dx. You may use the antiderivative csc θdθ =
log | cot θ − csc θ| + C.

27. Consider the following work.

1 1
Z Z
dx = 2 cos θdθ using x = sin θ, dx = cos θdθ
1 − x2 1 − sin θ
cos θ
Z
= dθ
cos2 θ
Z
= sec θdθ

= log | sec θ + tan θ| + C Example 1.8.19

1 x
= log √ +√ +C using the triangle below
1−x 2 1 − x2
1+x
= log √ +C
1 − x2

186
I NTEGRATION 1.9 T RIGONOMETRIC S UBSTITUTION

1
x
θ
√
1 − x2

1+x
a Differentiate log √ .
1 − x2
Z 3  x=3
1 1+x
b True or false: 2
dx = log √
2 1−x 1 − x2 x=2
c Was the work in the question correct? Explain.

28.

a √
Suppose we are evaluating an integral that contains the term
a2 − x2 , where a is a positive constant, and we use the substitution
x = a sin u (with inverse u = arcsin(x/a)), so that
√ √
a2 − x2 = a2 cos2 u = |a cos u|

Under what circumstances is |a cos u| =

6 a cos u?
√
b Suppose we are evaluating an integral that contains the term a2 + x2 ,
where a is a positive constant, and we use the substitution x = a tan u
(with inverse u = arctan(x/a)), so that
√ √
a2 + x2 = a2 sec2 u = |a sec u|

Under what circumstances is |a sec u| =

6 a sec u?

c √
Suppose we are evaluating an integral that contains the term
x2 − a2 , where a is a positive constant, and we use the substitution
x = a sec u (with inverse u = arcsec(x/a) = arccos(a/x)), so that
√ √
x2 − a2 = a2 tan2 u = |a tan u|

Under what circumstances is |a tan u| =

6 a tan u?

187
 I NTEGRATION 1.10 PARTIAL F RACTIONS

1.10q Partial Fractions

Partial fractions is the name given to a technique of integration that may be used to
integrate any rational function 1 . We already know how to integrate some simple
rational functions
1 1
Z Z
dx = log |x| + C dx = arctan(x) + C
x 1 + x2
Combining these with the substitution rule, we can integrate similar but more com-
plicated rational functions:
 
1 1 1 1 2x
Z Z
dx = log |2x + 3| + C dx = √ arctan √ +C
2x + 3 2 3 + 4x2 2 3 3
By summing such terms together we can integrate yet more complicated forms

x2
Z  
1 1
x+ + dx = + log |x + 1| + log |x − 1| + C
x+1 x−1 2
However we are not (typically) presented with a rational function nicely decom-
posed into neat little pieces. It is far more likely that the rational function will be
written as the ratio of two polynomials. For example:
Z 3
x +x
dx
x2 − 1
In this specific example it is not hard to confirm that

1 1 x(x + 1)(x − 1) + (x − 1) + (x + 1) x3 + x
x+ + = = 2
x+1 x−1 (x + 1)(x − 1) x −1
and hence
x3 + x
Z  
1 1
Z
dx = x+ + dx
x2 − 1 x+1 x−1
x2
= + log |x + 1| + log |x − 1| + C
2
Of course going in this direction (from a sum of terms to a single rational function)
is straightforward. To be useful we need to understand how to do this in reverse:
decompose a given rational function into a sum of simpler pieces that we can inte-
grate.
(x)
Suppose that N (x) and D(x) are polynomials. The basic strategy is to write ND(x)
as a sum of very simple, easy to integrate rational functions, namely

1 Recall that a rational function is the ratio of two polynomials.

188
 I NTEGRATION 1.10 PARTIAL F RACTIONS

1 polynomials — we shall see below that these are needed when the degree 2 of
N (x) is equal to or strictly bigger than the degree of D(x), and
A
2 rational functions of the particularly simple form (ax+b)n
and
Ax+B
3 rational functions of the form (ax2 +bx+c)m
.

We already know how to integrate the first two forms, and we’ll see how to integrate
the third form in the near future.
To begin to explore this method of decomposition, let us go back to the example
we just saw

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

The technique that we will use is based on two observations:

1 The denominators on the left-hand side of are the factors of the denominator
x2 − 1 = (x − 1)(x + 1) on the right-hand side.

2 Use P (x) to denote the polynomial on the left hand side, and then use N (x) and
D(x) to denote the numerator and denominator of the right hand side. That is

P (x) = x N (x) = x3 + x D(x) = x2 − 1.

Then the degree of N (x) is the sum of the degrees of P (x) and D(x). This is
because the highest degree term in N (x) is x3 , which comes from multiplying
P (x) by D(x), as we see in

P (x) D(x)
z}|{ z }| {
1 1 x (x + 1)(x − 1) +(x − 1) + (x + 1) x3 + x
x+ + = = 2
x+1 x−1 (x + 1)(x − 1) x −1

More generally, the presence of a polynomial on the left hand side is signalled
on the right hand side by the fact that the degree of the numerator is at least as
large as the degree of the denominator.

1.10.1 tt Partial fraction decomposition examples

Rather than writing up the technique — known as the partial fraction decomposition
— in full generality, we will instead illustrate it through a sequence of examples.

2 The degree of a polynomial is the largest power of x. For example, the degree of 2x3 +4x2 +6x+8
is three.

189
I NTEGRATION 1.10 PARTIAL F RACTIONS

x−3
R
Example 1.10.1 x2 −3x+2
dx.

N (x) x−3
In this example, we integrate D(x)
= x2 −3x+2
.
Solution:

• Step 1. We first check to see if a polynomial P (x) is needed. To do so, we check

to see if the degree of the numerator, N (x), is strictly smaller than the degree
of the denominator D(x). In this example, the numerator, x − 3, has degree one
and that is indeed strictly smaller than the degree of the denominator, x2 − 3x +
2, which is two. In this case a we do not need to extract a polynomial P (x) and
we move on to step 2.

• Step 2. The second step is to factor the denominator

x2 − 3x + 2 = (x − 1)(x − 2)

In this example it is quite easy, but in future examples (and quite possibly in
your homework, quizzes and exam) you will have to work harder to factor the
denominator. In Appendix A.16 we have written up some simple tricks for
factoring polynomials. We will illustrate them in Example 1.10.3 below.
x−3
• Step 3. The third step is to write x2 −3x+2
in the form

x−3 A B
= +
x2 − 3x + 2 x−1 x−2
for some constants A and B. More generally, if the denominator consists of n
different linear factors, then we decompose the ratio as

A1 A2 An
rational function = + + ··· +
linear factor 1 linear factor 2 linear factor n

To proceed we need to determine the values of the constants A, B and there

are several different methods to do so. Here are two methods

• Step 3 — Algebra Method. This approach has the benefit of being conceptually
clearer and easier, but the downside is that it is more tedious.
b
To determine the values of the constants A, B, we put the right-hand side
back over the common denominator (x − 1)(x − 2).

x−3 A B A(x − 2) + B(x − 1)

= + =
x2 − 3x + 2 x−1 x−2 (x − 1)(x − 2)

The fraction on the far left is the same as the fraction on the far right if and only
if their numerators are the same.

x − 3 = A(x − 2) + B(x − 1)

190
I NTEGRATION 1.10 PARTIAL F RACTIONS

Write the right hand side as a polynomial in standard form (i.e. collect up all x
terms and all constant terms)

x − 3 = (A + B)x + (−2A − B)

For these two polynomials to be the same, the coefficient of x on the left hand
side and the coefficient of x on the right hand side must be the same. Similarly
the coefficients of x0 (i.e. the constant terms) must match. This gives us a
system of two equations.

A+B =1 −2A − B = −3

in the two unknowns A, B. We can solve this system by

◦ using the first equation, namely A + B = 1, to determine A in terms of B:

A=1−B

◦ Substituting this into the remaining equation eliminates the A from sec-
ond equation, leaving one equation in the one unknown B, which can
then be solved for B:

−2A − B = −3 substitute A = 1 − B
−2(1 − B) − B = −3 clean up
−2 + B = −3 so B = −1

◦ Once we know B, we can substitute it back into A = 1 − B to get A.

A = 1 − B = 1 − (−1) = 2

Hence
x−3 2 1
= −
x2 − 3x + 2 x−1 x−2

• Step 3 — Sneaky Method. This takes a little more work to understand, but it is
more efficient than the algebra method.
We wish to find A and B for which
x−3 A B
= +
(x − 1)(x − 2) x−1 x−2

Note that the denominator on the left hand side has been written in factored
form.

191
I NTEGRATION 1.10 PARTIAL F RACTIONS

◦ To determine A, we multiply both sides of the equation by A’s denomina-

tor, which is x − 1,
x−3 (x − 1)B
=A+
x−2 x−2
and then we completely eliminate B from the equation by evaluating at
x = 1. This value of x is chosen to make x − 1 = 0.
x−3 (x − 1)B 1−3
=A+ =⇒ A = =2
x−2 x=1 x−2 x=1 1−2

◦ To determine B, we multiply both sides of the equation by B’s denomina-

tor, which is x − 2,
x−3 (x − 2)A
= +B
x−1 x−1
and then we completely eliminate A from the equation by evaluating at
x = 2. This value of x is chosen to make x − 2 = 0.
x−3 (x − 2)A 2−3
= + B =⇒ B = = −1
x−1 x=2 x−1 x=2 2−1

Hence we have (the thankfully consistent answer)

x−3 2 1
= −
x2 − 3x + 2 x−1 x−2
Notice that no matter which method we use to find the constants we can easily
check our answer by summing the terms back together:

2 1 2(x − 2) − (x − 1)
− =
x−1 x−2 (x − 2)(x − 1)
2x − 4 − x + 1 x−3
= = X
x2 − 3x + 2 x2 − 3x + 2

• Step 4. The final step is to integrate.

x−3 2 −1
Z Z Z
2
dx = dx + dx
x − 3x + 2 x−1 x−2
= 2 log |x − 1| − log |x − 2| + C

a We will soon get to an example (Example 1.10.2 in fact) in which the numerator degree is at least
as large as the denominator degree — in that situation we have to extract a polynomial P (x)
before we can move on to step 2.
b That is, we take the decomposed form and sum it back together.

Example 1.10.1
192
I NTEGRATION 1.10 PARTIAL F RACTIONS

Perhaps the first thing that you notice is that this process takes quite a few steps
3
. However no single step is all that complicated; it only takes practice. With that
said, let’s do another, slightly more complicated, one.

3x3 −8x2 +4x−1

R
Example 1.10.2 x2 −3x+2
dx.

N (x) 3x3 −8x2 +4x−1

In this example, we integrate D(x)
= x2 −3x+2
.
Solution:

• Step 1. We first check to see if the degree of the numerator N (x) is strictly
smaller than the degree of the denominator D(x). In this example, the numer-
ator, 3x3 − 8x2 + 4x − 1, has degree three and the denominator, x2 − 3x + 2, has
degree two. As 3 ≥ 2, we have to implement the first step.
N (x)
The goal of the first step is to write D(x)
in the form

N (x) R(x)
= P (x) +
D(x) D(x)

with P (x) being a polynomial and R(x) being a polynomial of degree strictly
smaller than the degree of D(x). The right hand side is P (x)D(x)+R(x)
D(x)
, so we have
to express the numerator in the form N (x) = P (x)D(x) + R(x), with P (x) and
R(x) being polynomials and with the degree of R being strictly smaller than
the degree of D. P (x)D(x) is a sum of expressions of the form axn D(x). We
want to pull as many expressions of this form as possible out of the numerator
N (x), leaving only a low degree remainder R(x).
We do this using long division — the same long division you learned in school,
but with the base 10 replaced by x.

◦ We start by observing that to get from the highest degree term in the de-
nominator (x2 ) to the highest degree term in the numerator (3x3 ), we have
to multiply it by 3x. So we write,

In the above expression, the denominator is on the left, the numerator is

on the right and 3x is written above the highest order term of the numer-
ator. Always put lower powers of x to the right of higher powers of x —
this mirrors how you do long division with numbers; lower powers of ten
sit to the right of lower powers of ten.

3 Though, in fairness, we did step 3 twice — and that is the most tedious bit. . . Actually — some-
times factoring the denominator can be quite challenging. We’ll consider this issue in more detail
shortly.

193
I NTEGRATION 1.10 PARTIAL F RACTIONS

◦ Now we subtract 3x times the denominator, x2 − 3x + 2, which is 3x3 −

9x2 + 6x, from the numerator.

◦ This has left a remainder of x2 − 2x − 1. To get from the highest degree

term in the denominator (x2 ) to the highest degree term in the remainder
(x2 ), we have to multiply by 1. So we write,

◦ Now we subtract 1 times the denominator, x2 − 3x + 2, which is x2 − 3x + 2,

from the remainder.

◦ This leaves a remainder of x − 3. Because the remainder has degree 1,

which is smaller than the degree of the denominator (being degree 2), we
stop.
◦ In this example, when we subtracted 3x(x2 − 3x + 2) and 1(x2 − 3x + 2)
from 3x3 − 8x2 + 4x − 1 we ended up with x − 3. That is,
3x3 − 8x2 + 4x − 1 − 3x(x2 − 3x + 2) − 1(x2 − 3x + 2)
=x−3
or, collecting the two terms proportional to (x2 − 3x + 2)
3x3 − 8x2 + 4x − 1 − (3x + 1)(x2 − 3x + 2) = x − 3
Moving the (3x + 1)(x2 − 3x + 2) to the right hand side and dividing the
whole equation by x2 − 3x + 2 gives
3x3 − 8x2 + 4x − 1 x−3
2
= 3x + 1 + 2
x − 3x + 2 x − 3x + 2
And we can easily check this expression just by summing the two terms
on the right-hand side.

194
I NTEGRATION 1.10 PARTIAL F RACTIONS

(x)
We have written the integrand in the form N D(x)
R(x)
= P (x) + D(x) , with the degree
of R(x) strictly smaller than the degree of D(x), which is what we wanted.
Observe that R(x) is the final remainder of the long division procedure and
P (x) is at the top of the long division computation. This is the end of Step 1.
Oof! You should definitely practice this step.

• Step 2. The second step is to factor the denominator

x2 − 3x + 2 = (x − 1)(x − 2)

We already did this in Example 1.10.1.

x−3
• Step 3. The third step is to write x2 −3x+2
in the form

x−3 A B
= +
x2 − 3x + 2 x−1 x−2
for some constants A and B. We already did this in Example 1.10.1. We found
A = 2 and B = −1.

• Step 4. The final step is to integrate.

3x3 − 8x2 + 4x − 1
Z
dx
x2 − 3x + 2
2 −1
Z Z Z
 
= 3x + 1 dx + dx + dx
x−1 x−2
3
= x2 + x + 2 log |x − 1| − log |x − 2| + C
2

You can see that the integration step is quite quick — almost all the work is in prepar-
ing the integrand.
Example 1.10.2

Here is a very solid example. It is quite long and the steps are involved. However
please persist. No single step is too difficult.

195
I NTEGRATION 1.10 PARTIAL F RACTIONS

x4 +5x3 +16x2 +26x+22

R
Example 1.10.3 x3 +3x2 +7x+5
dx.

N (x) x4 +5x3 +16x2 +26x+22

In this example, we integrate D(x)
= x3 +3x2 +7x+5
.
Solution:

• Step 1. Again, we start by comparing the degrees of the numerator and denom-
inator. In this example, the numerator, x4 + 5x3 + 16x2 + 26x + 22, has degree
four and the denominator, x3 + 3x2 + 7x + 5, has degree three. As 4 ≥ 3, we
(x)
must execute the first step, which is to write N
D(x)
in the form

N (x) R(x)
= P (x) +
D(x) D(x)

with P (x) being a polynomial and R(x) being a polynomial of degree strictly
smaller than the degree of D(x). This step is accomplished by long division,
just as we did in Example 1.10.3. We’ll go through the whole process in detail
again.
Actually — before you read on ahead, please have a go at the long division. It
is good practice.

◦ We start by observing that to get from the highest degree term in the de-
nominator (x3 ) to the highest degree term in the numerator (x4 ), we have
to multiply by x. So we write,

◦ Now we subtract x times the denominator x3 + 3x2 + 7x + 5, which is

x4 + 3x3 + 7x2 + 5x, from the numerator.

◦ The remainder was 2x3 + 9x2 + 21x + 22. To get from the highest degree
term in the denominator (x3 ) to the highest degree term in the remainder
(2x3 ), we have to multiply by 2. So we write,

196
I NTEGRATION 1.10 PARTIAL F RACTIONS

◦ Now we subtract 2 times the denominator x3 + 3x2 + 7x + 5, which is

2x3 + 6x2 + 14x + 10, from the remainder.

◦ This leaves a remainder of 3x2 +7x+12. Because the remainder has degree
2, which is smaller than the degree of the denominator, which is 3, we stop.
◦ In this example, when we subtracted x(x3 + 3x2 + 7x + 5) and 2(x3 + 3x2 +
7x + 5) from x4 + 5x3 + 16x2 + 26x + 22 we ended up with 3x2 + 7x + 12.
That is,

x4 + 5x3 + 16x2 + 26x + 22 − x(x3 + 3x2 + 7x + 5)

− 2(x3 + 3x2 + 7x + 5)
= 3x2 + 7x + 12

or, collecting the two terms proportional to (x3 + 3x2 + 7x + 5) we get

x4 + 5x3 + 16x2 + 26x + 22 − (x + 2)(x3 + 3x2 + 7x + 5)

= 3x2 + 7x + 12

Moving the (x + 2)(x3 + 3x2 + 7x + 5) to the right hand side and dividing
the whole equation by x3 + 3x2 + 7x + 5 gives

x4 + 5x3 + 16x2 + 26x + 22 3x2 + 7x + 12

= x + 2 +
x3 + 3x2 + 7x + 5 x3 + 3x2 + 7x + 5
(x)
This is of the form ND(x)
R(x)
= P (x) + D(x) , with the degree of R(x) strictly smaller
than the degree of D(x), which is what we wanted. Observe, once again, that
R(x) is the final remainder of the long division procedure and P (x) is at the top
of the long division computation.

197
I NTEGRATION 1.10 PARTIAL F RACTIONS

• Step 2. The second step is to factor the denominator D(x) = x3 + 3x2 + 7x + 5. In

the “real world” factorisation of polynomials is often very hard. Fortunately a ,
this is not the “real world” and there is a trick available to help us find this fac-
torisation. The reader should take some time to look at Appendix A.16 before
proceeding.
◦ The trick exploits the fact that most polynomials that appear in homework
assignments and on tests have integer coefficients and some integer roots.
Any integer root of a polynomial that has integer coefficients, like D(x) =
x3 + 3x2 + 7x + 5, must divide the constant term of the polynomial exactly.
Why this is true is explained b in Appendix A.16.
◦ So any integer root of x3 + 3x2 + 7x + 5 must divide 5 exactly. Thus the
only integers which can be roots of D(x) are ±1 and ±5. Of course, not all
of these give roots of the polynomial — in fact there is no guarantee that
any of them will be. We have to test each one.
◦ To test if +1 is a root, we sub x = 1 into D(x):
D(1) = 13 + 3(1)2 + 7(1) + 5 = 16
As D(1) 6= 0, 1 is not a root of D(x).
◦ To test if −1 is a root, we sub it into D(x):
D(−1) = (−1)3 + 3(−1)2 + 7(−1) + 5 = −1 + 3 − 7 + 5 = 0


As D(−1) = 0, −1 is a root of D(x). As −1 is a root of D(x), x − (−1) =
(x + 1) must factor D(x) exactly. We can factor the (x + 1) out of D(x) =
x3 + 3x2 + 7x + 5 by long division once again.
◦ Dividing D(x) by (x + 1) gives:

This time, when we subtracted x2 (x + 1) and 2x(x + 1) and 5(x + 1) from

x3 +3x2 +7x+5 we ended up with 0 — as we knew would happen, because
we knew that x + 1 divides x3 + 3x2 + 7x + 5 exactly. Hence
x3 + 3x2 + 7x + 5 − x2 (x + 1) − 2x(x + 1) − 5(x + 1) = 0
or
x3 + 3x2 + 7x + 5 = x2 (x + 1) + 2x(x + 1) + 5(x + 1)

198
I NTEGRATION 1.10 PARTIAL F RACTIONS

or

x3 + 3x2 + 7x + 5 = (x2 + 2x + 5)(x + 1)

◦ It isn’t quite time to stop yet; we should attempt to factor the quadratic
factor, x2 + 2x + 5. We can use the quadratic formula c to find the roots of
x2 + 2x + 5:
√ √ √
−b ± b2 − 4ac −2 ± 4 − 20 −2 ± −16
= =
2a 2 2
Since this expression contains the square root of a negative number the
equation x2 + 2x + 5 = 0 has no real solutions; without the use of complex
numbers, x2 + 2x + 5 cannot be factored.

We have reached the end of step 2. At this point we have

x4 + 5x3 + 16x2 + 26x + 22 3x2 + 7x + 12

=x+2+
x3 + 3x2 + 7x + 5 (x + 1)(x2 + 2x + 5)

3x2 +7x+12
• Step 3. The third step is to write (x+1)(x2 +2x+5)
in the form

3x2 + 7x + 12 A Bx + C
2
= + 2
(x + 1)(x + 2x + 5) x + 1 x + 2x + 5
for some constants A, B and C.
Note that the numerator, Bx + C of the second term on the right hand side is
not just a constant. It is of degree one, which is exactly one smaller than the
degree of the denominator, x2 + 2x + 5. More generally, if the denominator
consists of n different linear factors and m different quadratic factors, then we
decompose the ratio as
A1 A2 An
rational function = + + ··· +
linear factor 1 linear factor 2 linear factor n
B1 x + C1 B2 x + C2
+ + + ···
quadratic factor 1 quadratic factor 2
Bm x + Cm
+
quadratic factor m

To determine the values of the constants A, B, C, we put the right hand side
back over the common denominator (x + 1)(x2 + 2x + 5).

3x2 + 7x + 12 A Bx + C
2
= + 2
(x + 1)(x + 2x + 5) x + 1 x + 2x + 5
A(x2 + 2x + 5) + (Bx + C)(x + 1)
=
(x + 1)(x2 + 2x + 5)

199
I NTEGRATION 1.10 PARTIAL F RACTIONS

The fraction on the far left is the same as the fraction on the far right if and only
if their numerators are the same.

3x2 + 7x + 12 = A(x2 + 2x + 5) + (Bx + C)(x + 1)

Again, as in Example 1.10.1, there are a couple of different ways to determine

the values of A, B and C from this equation.

• Step 3 — Algebra Method. The conceptually clearest procedure is to write the

right hand side as a polynomial in standard form (i.e. collect up all x2 terms,
all x terms and all constant terms)

3x2 + 7x + 12 = (A + B)x2 + (2A + B + C)x + (5A + C)

For these two polynomials to be the same, the coefficient of x2 on the left hand
side and the coefficient of x2 on the right hand side must be the same. Similarly
the coefficients of x1 must match and the coefficients of x0 must match.
This gives us a system of three equations

A+B =3 2A + B + C = 7 5A + C = 12

in the three unknowns A, B, C. We can solve this system by

◦ using the first equation, namely A + B = 3, to determine A in terms of B:

A = 3 − B.
◦ Substituting A = 3 − B into the remaining two equations eliminates the
A’s from these two equations, leaving two equations in the two unknowns
B and C.

A=3−B 2A + B + C = 7 5A + C = 12
⇒ 2(3 − B) + B + C = 7 5(3 − B) + C = 12
⇒ −B + C = 1 −5B + C = −3

◦ Now we can use the equation −B + C = 1, to determine B in terms of C:

B = C − 1.
◦ Substituting this into the remaining equation eliminates the B’s leaving
an equation in the one unknown C, which is easy to solve.

B =C −1 −5B + C = −3
⇒ −5(C − 1) + C = −3
⇒ −4C = −8

◦ So C = 2, and then B = C − 1 = 1, and then A = 3 − B = 2. Hence

3x2 + 7x + 12 2 x+2
2
= + 2
(x + 1)(x + 2x + 5) x + 1 x + 2x + 5

200
I NTEGRATION 1.10 PARTIAL F RACTIONS

• Step 3 — Sneaky Method. While the above method is transparent, it is rather

tedious. It is arguably better to use the second, sneakier and more efficient,
procedure. In order for

3x2 + 7x + 12 = A(x2 + 2x + 5) + (Bx + C)(x + 1)

the equation must hold for all values of x.

◦ In particular, it must be true for x = −1. When x = −1, the factor (x +
1) multiplying Bx + C is exactly zero. So B and C disappear from the
equation, leaving us with an easy equation to solve for A:
h i
3x2 + 7x + 12 = A(x2 + 2x + 5) + (Bx + C)(x + 1)
x=−1 x=−1
=⇒ 8 = 4A =⇒ A = 2

◦ Sub this value of A back in and simplify.

3x2 + 7x + 12 = 2(x2 + 2x + 5) + (Bx + C)(x + 1)

x2 + 3x + 2 = (Bx + C)(x + 1)

Since (x + 1) is a factor on the right hand side, it must also be a factor on

the left hand side.

(x + 2)(x + 1) = (Bx + C)(x + 1)

⇒ (x + 2) = (Bx + C) ⇒ B = 1, C = 2

So again we find that

3x2 + 7x + 12 2 x+2
= + X
(x + 1)(x2 + 2x + 5) x + 1 x2 + 2x + 5

Thus our integrand can be written as

x4 + 5x3 + 16x2 + 26x + 22 2 x+2

3 2
=x+2+ + 2 .
x + 3x + 7x + 5 x + 1 x + 2x + 5

• Step 4. Now we can finally integrate! The first two pieces are easy.
1 2 2
Z Z
(x + 2)dx = x + 2x dx = 2 log |x + 1|
2 x+1
(We’re leaving the arbitrary constant to the end of the computation.)
d
The final piece is a little harder. The idea is to complete the square in the
denominator
x+2 x+2
=
x2 + 2x + 5 (x + 1)2 + 4

201
I NTEGRATION 1.10 PARTIAL F RACTIONS

ay+b
and then make a change of variables to make the fraction look like y 2 +1
. In this
case
x+2 1 x+2
=
2
(x + 1) + 4 4 ( x+1
2
)2 + 1

so we make the change of variables y = x+1 2

, dy = dx 2
, x = 2y − 1, dx = 2 dy

x+2 1 x+2
Z Z
dx = 2 dx
(x + 1)2 + 4 4 ( x+1 ) + 1
2
1 (2y − 1) + 2 1 2y + 1
Z Z
= 2 dy = dy
4 y2 + 1 2 y2 + 1
y 1 1
Z Z
= 2
dy + 2
dy
y +1 2 y +1

Both integrals are easily evaluated, using the substitution u = y 2 +1, du = 2y dy

for the first.
y 1 du 1 1
Z Z
2
dy = = log |u| = log(y 2 + 1)
y +1 u 2 2 2
1 h x+1  2 i
= log +1
2 2
1 1 1 1 x + 1
Z
dy = arctan y = arctan
2 y2 + 1 2 2 2

That’s finally it. Putting all of the pieces together

Z 4
x +5x3 +16x2 +26x + 22 1 2
dx = x + 2x + 2 log |x + 1|
x3 + 3x2 + 7x + 5 2
1 h x+1 2 i 1  x+1 
+ log + 1 + arctan +C
2 2 2 2

a One does not typically think of mathematics assignments or exams as nice kind places. . . The
polynomials that appear in the “real world” are not so forgiving. Nature, red in tooth and claw —
to quote Tennyson inappropriately (especially when this author doesn’t know any other words
from the poem).
b Appendix A.16 contains several simple tricks for factoring polynomials. We recommend that
you have a look at them.
c To be precise, the quadratic equation ax2 + bx + c = 0 has solutions
√
−b ± b2 − 4ac
x= .
2a
The term b2 − 4ac is called the discriminant and it tells us about the number of solutions. If
the discriminant is positive then there are two real solutions. When it is zero, there is a single
solution. And if it is negative, there is no real solutions (you need complex numbers to say more
than this).

202
I NTEGRATION 1.10 PARTIAL F RACTIONS

d This same idea arose in Section 1.9. Given a quadratic written as

Q(x) = ax2 + bx + c

rewrite it as

Q(x) = a(x + d)2 + e.

We can determine d and e by expanding and comparing coefficients of x:

ax2 + bx + c = a(x2 + 2dx + d2 ) + e = ax2 + 2dax + (e + ad2 )

Hence d = b/2a and e = c − ad2 .

Example 1.10.3

The best thing after working through a few a nice long examples is to do another
nice long example — it is excellent practice 4 . We recommend that the reader attempt
the problem before reading through our solution.

4x3 +23x2 +45x+27

R
Example 1.10.4 x3 +5x2 +8x+4
dx.

N (x) 4x3 +23x2 +45x+27

In this example, we integrate D(x)
= x3 +5x2 +8x+4
.

• Step 1. The degree of the numerator N (x) is equal to the degree of the denomi-
(x)
nator D(x), so the first step to write N
D(x)
in the form
N (x) R(x)
= P (x) +
D(x) D(x)
with P (x) being a polynomial (which should be of degree 0, i.e. just a constant)
and R(x) being a polynomial of degree strictly smaller than the degree of D(x).
By long division

so
4x3 + 23x2 + 45x + 27 3x2 + 13x + 11
= 4 +
x3 + 5x2 + 8x + 4 x3 + 5x2 + 8x + 4

4 At the risk of quoting Nietzsche, “That which does not kill us makes us stronger.” Though
this author always preferred the logically equivalent contrapositive — “That which does not
make us stronger will kill us.” However no one is likely to be injured by practicing partial
fractions or looking up quotes on Wikipedia. Its also a good excuse to remind yourself of what
a contrapositive is — though we will likely look at them again when we get to sequences and
series.

203
I NTEGRATION 1.10 PARTIAL F RACTIONS

• Step 2. The second step is to factorise D(x) = x3 + 5x2 + 8x + 4.

◦ To start, we’ll try and guess an integer root. Any integer root of D(x) must
divide the constant term, 4, exactly. Only ±1, ±2, ±4 can be integer roots
of x3 + 5x2 + 8x + 4.
◦ We test to see if ±1 are roots.

D(1) = (1)3 + 5(1)2 + 8(1) + 4 6= 0 ⇒ x = 1 is not a root

D(−1) = (−1)3 +5(−1)2 +8(−1)+4 = 0 ⇒ x = −1 is a root

So (x + 1) must divide x3 + 5x2 + 8x + 4 exactly.

◦ By long division

so

x3 + 5x2 + 8x + 4 = (x + 1)(x2 + 4x + 4)
= (x + 1)(x + 2)(x + 2)

◦ Notice that we could have instead checked whether or not ±2 are roots

D(2) = (2)3 + 5(2)2 + 8(2) + 4 6= 0 ⇒ x = 2 is not a root

D(−2) = (−2)3 +5(−2)2 +8(−2)+4 = 0 ⇒ x = −2 is a root

We now know that both −1 and −2 are roots of x3 + 5x2 + 8x + 4 and

hence both (x + 1) and (x + 2) are factors of x3 + 5x2 + 8x + 4. Because
x3 + 5x2 + 8x + 4 is of degree three and the coefficient of x3 is 1, we must
have x3 +5x2 +8x+4 = (x+1)(x+2)(x+a) for some constant a. Multiplying
out the right hand side shows that the constant term is 2a. So 2a = 4 and
a = 2.

This is the end of step 2. We now know that

4x3 + 23x2 + 45x + 27 3x2 + 13x + 11

= 4 +
x3 + 5x2 + 8x + 4 (x + 1)(x + 2)2

204
I NTEGRATION 1.10 PARTIAL F RACTIONS

3x2 +13x+11
• Step 3. The third step is to write (x+1)(x+2)2
in the form

3x2 + 13x + 11 A B C
2
= + +
(x + 1)(x + 2) x + 1 x + 2 (x + 2)2

for some constants A, B and C.

Note that there are two terms on the right hand arising from the factor (x + 2)2 .
One has denominator (x+2) and one has denominator (x+2)2 . More generally,
for each factor (x + a)n in the denominator of the rational function on the left
hand side, we include
A1 A2 An
+ 2
+ ··· +
x + a (x + a) (x + a)n

in the partial fraction decomposition on the right hand side a .

To determine the values of the constants A, B, C, we put the right hand side
back over the common denominator (x + 1)(x + 2)2 .

3x2 + 13x + 11 A B C
2
= + +
(x + 1)(x + 2) x + 1 x + 2 (x + 2)2
A(x + 2)2 + B(x + 1)(x + 2) + C(x + 1)
=
(x + 1)(x + 2)2

The fraction on the far left is the same as the fraction on the far right if and only
if their numerators are the same.

3x2 + 13x + 11 = A(x + 2)2 + B(x + 1)(x + 2) + C(x + 1)

As in the previous examples, there are a couple of different ways to determine

the values of A, B and C from this equation.

• Step 3 — Algebra Method. The conceptually clearest procedure is to write the

right hand side as a polynomial in standard form (i.e. collect up all x2 terms,
all x terms and all constant terms)

3x2 + 13x + 11 = (A + B)x2 + (4A + 3B + C)x + (4A + 2B + C)

For these two polynomials to be the same, the coefficient of x2 on the left hand
side and the coefficient of x2 on the right hand side must be the same. Similarly
the coefficients of x1 and the coefficients of x0 (i.e. the constant terms) must
match. This gives us a system of three equations,

A+B =3 4A + 3B + C = 13 4A + 2B + C = 11

in the three unknowns A, B, C. We can solve this system by

205
I NTEGRATION 1.10 PARTIAL F RACTIONS

◦ using the first equation, namely A + B = 3, to determine A in terms of B:

A = 3 − B.
◦ Substituting this into the remaining equations eliminates the A, leaving
two equations in the two unknown B, C.

4(3 − B) + 3B + C = 13 4(3 − B) + 2B + C = 11

or

−B + C = 1 − 2B + C = −1

◦ We can now solve the first of these equations, namely −B + C = 1, for B

in terms of C, giving B = C − 1.
◦ Substituting this into the last equation, namely −2B + C = −1, gives
−2(C − 1) + C = −1 which is easily solved to give
◦ C = 3, and then B = C − 1 = 2 and then A = 3 − B = 1.

Hence
4x3 + 23x2 + 45x + 27 3x2 + 13x + 11
= 4 +
x3 + 5x2 + 8x + 4 (x + 1)(x + 2)2
1 2 3
=4+ + +
x + 1 x + 2 (x + 2)2

• Step 3 — Sneaky Method. The second, sneakier, method for finding A, B and C
exploits the fact that 3x2 + 13x + 11 = A(x + 2)2 + B(x + 1)(x + 2) + C(x + 1)
must be true for all values of x. In particular, it must be true for x = −1. When
x = −1, the factor (x + 1) multiplying B and C is exactly zero. So B and C
disappear from the equation, leaving us with an easy equation to solve for A:
h i
2 2
3x + 13x + 11 = A(x+2) + B(x+1)(x+2) + C(x+1)
x=−1 x=−1
=⇒ 1 = A

Sub this value of A back in and simplify.

3x2 + 13x + 11 = (1)(x + 2)2 + B(x + 1)(x + 2) + C(x + 1)

2x2 + 9x + 7 = B(x + 1)(x + 2) + C(x + 1)
= (xB + 2B + C)(x + 1)

Since (x + 1) is a factor on the right hand side, it must also be a factor on the
left hand side.

(2x + 7)(x + 1) = (xB + 2B + C)(x + 1)

⇒ (2x + 7) = (xB + 2B + C)

206
I NTEGRATION 1.10 PARTIAL F RACTIONS

For the coefficients of x to match, B must be 2. For the constant terms to match,
2B + C must be 7, so C must be 3. Hence we again have

4x3 + 23x2 + 45x + 27 3x2 + 13x + 11

= 4 +
x3 + 5x2 + 8x + 4 (x + 1)(x + 2)2
1 2 3
=4+ + +
x + 1 x + 2 (x + 2)2

• Step 4. The final step is to integrate

4x3 + 23x2 + 45x + 27

Z
dx
x3 + 5x2 + 8x + 4
1 2 3
Z Z Z Z
= 4dx + dx + dx + dx
x+1 x+2 (x + 2)2
3
= 4x + log |x + 1| + 2 log |x + 2| − +C
x+2

a This is justified in the (optional) subsection “Justification of the Partial Fraction Decompositions”
below.

Example 1.10.4

The method of partial fractions is not just confined toRthe problem ofR integrating
rational functions. There are other integrals — such as sec xdx and sec3 xdx —
that can be transformed (via substitutions) into integrals of rational functions. We
encountered both of these integrals in Sections 1.8 and 1.9 on trigonometric integrals
and substitutions.
R
Example 1.10.5 sec xdx.

Solution: In this example, we integrate sec x. It is not yet clear what this integral has
to do with partial fractions. To get to a partial fractions computation, we first make
one of our old substitutions.
1
Z Z
sec xdx = dx massage the expression a little
cos x
cos x
Z
= dx substitute u = sin x, du = cos xdx
cos2 x
du
Z
=− 2
and use cos2 x = 1 − sin2 x = 1 − u2
u −1
1
So we now have to integrate u2 −1
, which is a rational function of u, and so is perfect
for partial fractions.

• Step 1. The degree of the numerator, 1, is zero, which is strictly smaller than the

207
I NTEGRATION 1.10 PARTIAL F RACTIONS

degree of the denominator, u2 − 1, which is two. So the first step is skipped.

• Step 2. The second step is to factor the denominator:

u2 − 1 = (u − 1)(u + 1)

1
• Step 3. The third step is to write u2 −1
in the form

1 1 A B
= = +
u2 −1 (u − 1)(u + 1) u−1 u+1

for some constants A and B.

• Step 3 — Sneaky Method.

◦ Multiply through by the denominator to get

1 = A(u + 1) + B(u − 1)

This equation must be true for all u.

◦ If we now set u = 1 then we eliminate B from the equation leaving us
with
1
1 = 2A so A = .
2

◦ Similarly, if we set u = −1 then we eliminate A, leaving

1
1 = −2B which implies B = − .
2

We have now found that A = 12 , B = − 12 , so

1 1h 1 1 i
= − .
u2 − 1 2 u−1 u+1

• It is always a good idea to check our work.

1
2
− 21 1
(u + 1) − 12 (u − 1) 1
+ = 2 = X
u−1 u+1 (u − 1)(u + 1) (u − 1)(u + 1)

• Step 4. The final step is to integrate.

du
Z Z
sec xdx = − 2
after substitution
u −1
1 du 1 du
Z Z
=− + partial fractions
2 u−1 2 u+1

208
I NTEGRATION 1.10 PARTIAL F RACTIONS

1 1
= − log |u − 1| + log |u + 1| + C
2 2
1 1
= − log | sin(x) − 1| + log | sin(x) + 1| + C rearrange a little
2 2
1 1 + sin x
= log +C
2 1 − sin x

Notice that since −1 ≤ sin x ≤ 1, we are free to drop the absolute values in the
last line if we wish.
Example 1.10.5

Another example in the same spirit, though a touch harder. Again, we saw this
problem in Section 1.8 and 1.9.

sec3 xdx.
R
Example 1.10.6

Solution:
• We’ll start by converting it into the integral of a rational function using the
substitution u = sin x, du = cos xdx.
1
Z Z
3
sec xdx = dx massage this a little
cos3 x
cos x
Z
= dx replace cos2 x = 1 − sin2 x = 1 − u2
cos4 x
cos xdx
Z
= 2
[1 − sin2 x]
du
Z
=
[1 − u2 ]2
1
• We could now find the partial fraction decomposition of the integrand [1−u2 ]2
by executing the usual four steps. But it is easier to use
1 1h 1 1 i
= −
u2 − 1 2 u−1 u+1
which we worked out in Example 1.10.5 above.
• Squaring this gives
1 1h 1 1 i2
= −
[1 − u2 ]2 4 u−1 u+1
1h 1 2 1 i
= − +
4 (u − 1)2 (u − 1)(u + 1) (u + 1)2
1h 1 1 1 1 i
= − + +
4 (u − 1)2 u − 1 u + 1 (u + 1)2

209
 I NTEGRATION 1.10 PARTIAL F RACTIONS

1 1 1 1
 
where we have again used u2 −1
= 2 u−1
− u+1
in the last step.

• It only remains to do the integrals and simplify.

1 h 1 1 1 1 i
Z Z
3
sec xdx = − + + du
4 (u−1)2 u−1 u+1 (u+1)2
1h 1 1 i
= − − log |u − 1| + log |u + 1| − +C
4 u−1 u+1
group carefully
−1 h 1 1 i 1 h i
= + + log |u + 1| − log |u − 1| + C
4 u−1 u+1 4
sum carefully
1 2u 1 u+1
=− 2 + log +C
4u −1 4 u−1
clean up
1 u 1 u+1
= 2
+ log +C
21−u 4 u−1
put u = sin x
1 sin x 1 sin x + 1
= + log +C
2 cos2 x 4 sin x − 1
Example 1.10.6

1.10.2 tt The form of partial fraction decompositions

In the examples above we used the partial fractions method to decompose rational
functions into easily integrated pieces. Each of those examples was quite involved
and we had to spend quite a bit of time factoring and doing long division. The
key step in each of the computations was Step 3 — in that step we decomposed
(x)
the rational function N D(x)
R(x)
(or D(x) ), for which the degree of the numerator is strictly
smaller than the degree of the denominator, into a sum of particularly simple rational
A
functions, like x−a . We did not, however, give a systematic description of those
decompositions.
In this subsection we fill that gap by describing the general 5 form of partial frac-
tion decompositions. The justification of these forms is not part of the course, but
the interested reader is invited to read the next (optional) subsection where such
justification is given. In the following it is assumed that
• N (x) and D(x) are polynomials with the degree of N (x) strictly smaller than
the degree of D(x).

5 Well — not the completely general form, in the sense that we are not allowing the use of complex
numbers. As a result we have to use both linear and quadratic factors in the denominator. If we
could use complex numbers we would be able to restrict ourselves to linear factors.
210
 I NTEGRATION 1.10 PARTIAL F RACTIONS

• K is a constant.
• a1 , a2 , · · ·, aj are all different numbers.
• m1 , m2 , · · ·, mj , and n1 , n2 , · · ·, nk are all strictly positive integers.
• x2 + b1 x + c1 , x2 + b2 x + c2 , · · · , x2 + bk x + ck are all different.

ttt Simple linear factor case

1.10.2.1

If the denominator D(x) = K(x − a1 )(x − a2 ) · · · (x − aj ) is a product of j different

linear factors, then

Equation 1.10.7

N (x) A1 A2 Aj
= + + ··· +
D(x) x − a1 x − a2 x − aj

We can then integrate each term

A
Z
dx = A log |x − a| + C.
x−a

ttt General linear factor case

1.10.2.2

If the denominator D(x) = K(x − a1 )m1 (x − a2 )m2 · · · (x − aj )mj then

Claim 1.10.8

N (x) A1,1 A1,2 A1,m1

= + + · · · +
D(x) x − a1 (x − a1 )2 (x − a1 )m1
A2,1 A2,2 A2,m2
+ + + · · · + + ···
x − a2 (x − a2 )2 (x − a2 )m2
Aj,1 Aj,2 Aj,mj
+ + + · · · +
x − aj (x − aj )2 (x − aj )mj

Notice that we could rewrite each line as

A1 A2 Am A1 (x − a)m−1 + A2 (x − a)m−2 + · · · + Am
+ + · · · + =
x − a (x − a)2 (x − a)m (x − a)m

211
 I NTEGRATION 1.10 PARTIAL F RACTIONS

B1 xm−1 + B2 xm−2 + · · · + Bm
=
(x − a)m
which is a polynomial whose degree, m − 1, is strictly smaller than that of the de-
nominator (x − a)m . But the form of Equation 1.10.8 is preferable because it is easier
to integrate.
A
Z
dx = A log |x − a| + C
x−a
A 1 A
Z
k
dx = − · provided k > 1.
(x − a) k − 1 (x − a)k−1

ttt Simple linear and quadratic factor case

1.10.2.3

If D(x) = K(x − a1 ) · · · (x − aj )(x2 + b1 x + c1 ) · · · (x2 + bk x + ck ) then

Claim 1.10.9

N (x) A1 Aj B1 x + C1 Bk x + Ck
= + ··· + + 2 + ··· + 2
D(x) x − a1 x − aj x + b 1 x + c 1 x + bk x + c k

Note that the numerator of each term on the right hand side has degree one
smaller than the degree of the denominator.
The quadratic terms xBx+C
2 +bx+c are integrated in a two-step process that is best illus-

trated with a simple example (see also Example 1.10.3 above).

2x+7
R
Example 1.10.10 x2 +4x+13
dx.

Solution:

• Start by completing the square in the denominator:

x2 + 4x + 13 = (x + 2)2 + 9 and thus

2x + 7 2x + 7
2
=
x + 4x + 13 (x + 2)2 + 32

• Now set y = (x + 2)/3, dy = 13 dx, or equivalently x = 3y − 2, dx = 3dy:

2x + 7 2x + 7
Z Z
dx = dx
x2 + 4x + 13 (x + 2)2 + 32
6y − 4 + 7
Z
= · 3dy
32 y 2 + 32

212
 I NTEGRATION 1.10 PARTIAL F RACTIONS

6y + 3
Z
= dy
3(y 2 + 1)
2y + 1
Z
= dy
y2 + 1

Notice that we chose 3 in y = (x + 2)/3 precisely to transform the denominator

into the form y 2 + 1.

• Now almost always the numerator will be a linear polynomial of y and we

decompose as follows

2x + 7 2y + 1
Z Z
2
dx = dy
x + 4x + 13 y2 + 1
2y 1
Z Z
= 2
dy + 2
dy
y +1 y +1
= log |y 2 + 1| + arctan y + C
 2  
x+2 x+2
= log + 1 + arctan +C
3 3

Example 1.10.10

ttt Optional — General linear and quadratic factor case

1.10.2.4

If D(x) = K(x − a1 )m1 · · · (x − aj )mj (x2 + b1 x + c1 )n1 · · · (x2 + bk x + ck )nk

Claim 1.10.11

N (x) A1,1 A1,2 A1,m1

= + 2
+ ··· + + ···
D(x) x − a1 (x − a1 ) (x − a1 )m1
Aj,1 Aj,2 Aj,mj
+ + + · · · +
x − aj (x − aj )2 (x − aj )mj
B1,1 x + C1,1 B1,2 x + C1,2 B1,n x + C1,n1
+ 2 + 2 2
+· · ·+ 2 1 +· · ·
x + b1 x + c1 (x + b1 x + c1 ) (x + b1 x + c1 )n1
Bk,1 x + Ck,1 Bk,2 x + Ck,2 Bk,nk x + C1,nk
+ 2 + 2 +· · ·+
x + bk x + ck (x + bk x + ck )2 (x2 + bk x + ck )nk

We have already seen how to integrate the simple and general linear terms, and
the simple quadratic terms. Integrating general quadratic terms is not so straightfor-
ward.

213
I NTEGRATION 1.10 PARTIAL F RACTIONS

dx
R
Example 1.10.12 (x2 +1)n
.

This example is not so easy, so it should definitely be considered optional.

Solution: In what follows write
dx
Z
In = .
(x2 + 1)n
• When n = 1 we know that
dx
Z
= arctan x + C
x2+1

• Now assume that n > 1, then

1 (x2 + 1 − x2 )
Z Z
dx = dx sneaky
(x2 + 1)n (x2 + 1)n
1 x2
Z Z
= dx − dx
(x2 + 1)n−1 (x2 + 1)n
x2
Z
= In−1 − dx
(x2 + 1)n
So we can write In in terms of In−1 and this second integral.

• We can use integration by parts to compute the second integral:

x2 x 2x
Z Z
2 n
dx = · 2 dx sneaky
(x + 1) 2 (x + 1)n

We set u = x/2 and dv = (x22x +1)n

dx, which gives du = 12 dx and v = − n−1
1
·
1
(x2 +1)n−1
. You can check v by differentiating. Integration by parts gives

x 2x
Z
· 2 dx
2 (x + 1)n
x dx
Z
=− 2
+
2(n − 1)(x + 1) n−1 2(n − 1)(x2 + 1)n−1
x 1
=− 2 n−1
+ · In−1
2(n − 1)(x + 1) 2(n − 1)

• Now put everything together:

1
Z
In = 2
dx
(x + 1)n
x 1
= In−1 + 2 n−1
− · In−1
2(n − 1)(x + 1) 2(n − 1)
2n − 3 x
= In−1 +
2(n − 1) 2(n − 1)(x2 + 1)n−1

214
 I NTEGRATION 1.10 PARTIAL F RACTIONS

• We can then use this recurrence to write down In for the first few n:
1 x
I2 = I1 + 2
+C
2 2(x + 1)
1 x
= arctan x +
2 2(x2 + 1)
3 x
I3 = I2 +
4 4(x + 1)2
2

3 3x x
= arctan x + + +C
8 8(x + 1) 4(x + 1)2
2 2

5 x
I4 = I3 +
6 6(x + 1)3
2

5 5x 5x x
= arctan x + + + +C
16 16(x2 + 1) 24(x2 + 1)2 6(x2 + 1)3

and so forth. You can see why partial fraction questions involving denomina-
tors with repeated quadratic factors do not often appear on exams.
Example 1.10.12

1.10.3 tt Optional — Justification of the partial fraction decompositions

We will now see the justification for the form of the partial fraction decompositions.
We will only consider the case in which the denominator has only linear factors. The
arguments when there are quadratic factors too are similar 6 .

ttt Simple linear factor case

1.10.3.1

In the most common partial fraction decomposition, we split up

N (x)
(x − a1 ) × · · · × (x − ad )
into a sum of the form
A1 Ad
+ ··· +
x − a1 x − ad
We now show that this decomposition can always be achieved, under the assump-
tions that the ai ’s are all different and N (x) is a polynomial of degree at most d − 1.
To do so, we shall repeatedly apply the following Lemma.

6 In fact, quadratic factors are completely avoidable because, if we use complex numbers, then
every polynomial can be written as a product of linear factors. This is the fundamental theorem
of algebra.

215
I NTEGRATION 1.10 PARTIAL F RACTIONS

Lemma 1.10.13
Let N (x) and D(x) be polynomials of degree n and d respectively, with n ≤ d.
Suppose that a is NOT a zero of D(x). Then there is a polynomial P (x) of
degree p < d and a number A such that

N (x) P (x) A
= +
D(x) (x − a) D(x) x − a

Proof.

• To save writing, let z = x − a. We then write Ñ (z) = N (z + a) and D̃(z) =

D(z + a), which are again polynomials of degree n and d respectively. We
also know that D̃(0) = D(a) 6= 0.

• In order to complete the proof we need to find a polynomial P̃ (z) of de-

gree p < d and a number A such that

Ñ (z) P̃ (z) A P̃ (z)z + AD̃(z)

= + =
D̃(z) z D̃(z) z D̃(z) z

or equivalently, such that

P̃ (z)z + AD̃(z) = Ñ (z).

• Now look at the polynomial on the left hand side. Every term in P̃ (z)z,
has at least one power of z. So the constant term on the left hand side is
exactly the constant term in AD̃(z), which is equal to AD̃(0). The constant
term on the right hand side is equal to Ñ (0). So the constant terms on the
(0)
left and right hand sides are the same if we choose A = Ñ D̃(0)
. Recall that
D̃(0) cannot be zero, so A is well defined.

• Now move AD̃(z) to the right hand side.

P̃ (z)z = Ñ (z) − AD̃(z)

The constant terms in Ñ (z) and AD̃(z) are the same, so the right hand side
contains no constant term and the right hand side is of the form Ñ1 (z)z
for some polynomial Ñ1 (z).

• Since Ñ (z) is of degree at most d and AD̃(z) is of degree exactly d, Ñ1 is a

polynomial of degree d − 1. It now suffices to choose P̃ (z) = Ñ1 (z).

216
 I NTEGRATION 1.10 PARTIAL F RACTIONS

Now back to
N (x)
(x − a1 ) × · · · × (x − ad )

Apply Lemma 1.10.13, with D(x) = (x − a2 ) × · · · × (x − ad ) and a = a1 . It says

N (x) A1 P (x)
= +
(x − a1 ) × · · · × (x − ad ) x − a1 (x − a2 ) × · · · × (x − ad )

for some polynomial P of degree at most d − 2 and some number A1 .

Apply Lemma 1.10.13 a second time, with D(x) = (x − a3 ) × · · · × (x − ad ),
N (x) = P (x) and a = a2 . It says

P (x) A2 Q(x)
= +
(x − a2 ) × · · · × (x − ad ) x − a2 (x − a3 ) × · · · × (x − ad )

for some polynomial Q of degree at most d − 3 and some number A2 .

At this stage, we know that

N (x) A1 A2 Q(x)
= + +
(x − a1 ) × · · · × (x − ad ) x − a1 x − a2 (x − a3 ) × · · · × (x − ad )

If we just keep going, repeatedly applying Lemma 1, we eventually end up with

N (x) A1 Ad
= + ··· +
(x − a1 ) × · · · × (x − ad ) x − a1 x − ad

as required.

ttt The general case with linear factors

1.10.3.2

Now consider splitting

N (x)
(x − a1 )n1 × · · · × (x − ad )nd

into a sum of the form 7

h A A1,n1 i h A Ad,nd i
1,1 d,1
+ ··· + + · · · + + · · · +
x − a1 (x − a1 )n1 x − ad (x − ad )nd

We now show that this decomposition can always be achieved, under the assump-
tions that the ai ’s are all different and N (x) is a polynomial of degree at most n1 +
· · · + nd − 1. To do so, we shall repeatedly apply the following Lemma.

7 If we allow ourselves to use complex numbers as roots, this is the general case. We don’t need to
consider quadratic (or higher) factors since all polynomials can be written as products of linear
factors with complex coefficients.

217
I NTEGRATION 1.10 PARTIAL F RACTIONS

Lemma 1.10.14
Let N (x) and D(x) be polynomials of degree n and d respectively, with n <
d + m. Suppose that a is NOT a zero of D(x). Then there is a polynomial P (x)
of degree p < d and numbers A1 , · · · , Am such that

N (x) P (x) A1 A2 Am
m
= + + 2
+ ··· +
D(x) (x − a) D(x) x − a (x − a) (x − a)m

Proof.

• As we did in the proof of the previous lemma, we write z = x − a. Then

Ñ (z) = N (z + a) and D̃(z) = D(z + a) are polynomials of degree n and d
respectively, D̃(0) = D(a) 6= 0.

• In order to complete the proof we have to find a polynomial P̃ (z) of de-

gree p < d and numbers A1 , · · · , Am such that

Ñ (z) P̃ (z) A1 A2 Am
= + + 2 + ··· + m
D̃(z) z m D̃(z) z z z
m m−1
P̃ (z)z + A1 z D̃(z) + A2 z m−2 D̃(z) + · · · + Am D̃(z)
=
D̃(z) z m

or equivalently, such that

P̃ (z)z m + A1 z m−1 D̃(z) + A2 z m−2 D̃(z) + · · · + Am−1 z D̃(z) + Am D̃(z)

= Ñ (z)

• Now look at the polynomial on the left hand side. Every single term
on the left hand side, except for the very last one, Am D̃(z), has at least
one power of z. So the constant term on the left hand side is exactly the
constant term in Am D̃(z), which is equal to Am D̃(0). The constant term
on the right hand side is equal to Ñ (0). So the constant terms on the left
(0)
and right hand sides are the same if we choose Am = Ñ D̃(0)
. Recall that
D̃(0) 6= 0 so Am is well defined.

• Now move Am D̃(z) to the right hand side.

P̃ (z)z m + A1 z m−1 D̃(z) + A2 z m−2 D̃(z) + · · · + Am−1 z D̃(z)

= Ñ (z) − Am D̃(z)

The constant terms in Ñ (z) and Am D̃(z) are the same, so the right hand
side contains no constant term and the right hand side is of the form

218
I NTEGRATION 1.10 PARTIAL F RACTIONS

Ñ1 (z)z with Ñ1 a polynomial of degree at most d + m − 2. (Recall that

Ñ is of degree at most d + m − 1 and D̃ is of degree at most d.) Divide the
whole equation by z to get

P̃ (z)z m−1 + A1 z m−2 D̃(z) + A2 z m−3 D̃(z) + · · · + Am−1 D̃(z) = Ñ1 (z).

• Now, we can repeat the previous argument. The constant term on the left
hand side, which is exactly equal to Am−1 D̃(0) matches the constant term
1 (0)
on the right hand side, which is equal to Ñ1 (0) if we choose Am−1 = ÑD̃(0) .
With this choice of Am−1

P̃ (z)z m−1 + A1 z m−2 D̃(z) + A2 z m−3 D̃(z) + · · · + Am−2 z D̃(z)

= Ñ1 (z) − Am−1 D̃(z) = Ñ2 (z)z

with Ñ2 a polynomial of degree at most d + m − 3. Divide by z and

continue.

• After m steps like this, we end up with

P̃ (z)z = Ñm−1 (z) − A1 D̃(z)

Ñm−1 (0)
after having chosen A1 = D̃(0)
.

• There is no constant term on the right side so that Ñm−1 (z) − A1 D̃(z) is
of the form Ñm (z)z with Ñm a polynomial of degree d − 1. Choosing
P̃ (z) = Ñm (z) completes the proof.

Now back to
N (x)
(x − a1 )n1 × · · · × (x − ad )nd
Apply Lemma 1.10.14, with D(x) = (x − a2 )n2 × · · · × (x − ad )nd , m = n1 and a = a1 .
It says
N (x)
(x − a1 )n1 × · · · × (x − ad )nd
A1,1 A1,2 A1,n1 P (x)
= + + · · · + +
x − a1 (x − a1 )2 (x − a)n1 (x − a2 )n2 × · · · × (x − ad )nd
Apply Lemma 1.10.14 a second time, with D(x) = (x − a3 )n3 × · · · × (x − ad )nd ,
N (x) = P (x), m = n2 and a = a2 . And so on. Eventually, we end up with
h A A1,n1 i h A Ad,nd i
1,1 d,1
+ ··· + + · · · + + · · · +
x − a1 (x − a1 )n1 x − ad (x − ad )nd
which is exactly what we were trying to show.

219
 I NTEGRATION 1.10 PARTIAL F RACTIONS

ttt Really Optional — The Fully General Case

1.10.3.3

We are now going to see that, in general, if N (x) and D(x) are polynomials with the
degree of N being strictly smaller than the degree of D (which we’ll denote deg(N ) <
deg(D)) and if

D(x) = K(x − a1 )m1 · · · (x − aj )mj (x2 + b1 x + c1 )n1 · · · (x2 + bk x + ck )nk (?)

(with b2` − 4ac` < 0 for all 1 ≤ ` ≤ k so that no quadratic factor can be written as a
product of linear factors with real coefficients) then there are real numbers Ai,j , Bi,j ,
Ci,j such that

N (x) A1,1 A1,2 A1,m1

= + + · · · + + ···
D(x) x − a1 (x − a1 )2 (x − a1 )m1
Aj,1 Aj,2 Aj,mj
+ + + · · · +
x − aj (x − aj )2 (x − aj )mj
B1,1 x + C1,1 B1,2 x + C1,2 B1,n1 x + C1,n1
+ 2 + 2 +· · ·+ +· · ·
x + b1 x + c1 (x + b1 x + c1 )2 (x2 + b1 x + c1 )n1
Bk,1 x + Ck,1 Bk,2 x + Ck,2 Bk,n x + C1,nk
+ 2 + 2 2
+· · ·+ 2 k
x + bk x + ck (x + bk x + ck ) (x + bk x + ck )nk

This was Equation 1.10.11. We start with two simpler results, that we’ll use repeat-
edly to get Equation 1.10.11. In the first simpler result, we consider the fraction
P (x)
Q1 (x) Q2 (x)
with P (x), Q1 (x) and Q2 (x) being polynomials with real coefficients and
we are going to assume that when P (x), Q1 (x) and Q2 (x) are factored as in (?), no
two of them have a common linear or quadratic factor. As an example, no two of

P (x) = 2(x − 3)(x − 4)(x2 + 3x + 3)

Q1 (x) = 2(x − 1)(x2 + 2x + 2)
Q2 (x) = 2(x − 2)(x2 + 2x + 3)

have such a common factor. But, for

P (x) = 2(x − 3)(x − 4)(x2 + x + 1)

Q1 (x) = 2(x − 1)(x2 + 2x + 2)
Q2 (x) = 2(x − 2)(x2 + x + 1)

P (x) and Q2 (x) have the common factor x2 + x + 1.

Lemma 1.10.15
Let P (x), Q1 (x) and Q2 (x) be polynomials with real coefficients and with
deg(P ) < deg(Q1 Q2 ). Assume that no two of P (x), Q1 (x) and Q2 (x) have a
common linear or quadratic factor. Then there are polynomials P1 , P2 with

220
I NTEGRATION 1.10 PARTIAL F RACTIONS

deg(P1 ) < deg(Q1 ), deg(P2 ) < deg(Q2 ), and

P (x) P1 (x) P2 (x)

= +
Q1 (x) Q2 (x) Q1 (x) Q2 (x)

Proof. We are to find polynomials P1 and P2 that obey

P (x) = P1 (x) Q2 (x) + P2 (x) Q1 (x)

Actually, we are going to find polynomials p1 and p2 that obey

p1 (x) Q1 (x) + p2 (x) Q2 (x) = C (??)

for some nonzero constant C, and then just multiply (??) by PC(x) . To find p1 , p2
and C we are going to use something called the Euclidean algorithm. It is an
algorithma that is used to efficiently find the greatest common divisors of two
numbers. Because Q1 (x) and Q2 (x) have no common factors of degree 1 or 2,
their ``greatest common divisor” has degree 0, i.e. is a constant.
Q1 (x)
• The first step is to apply long division to Q2 (x)
to find polynomials n0 (x)
and r0 (x) such that

Q1 (x) r0 (x)
= n0 (x) + with deg(r0 ) < deg(Q2 )
Q2 (x) Q2 (x)

or, equivalently,

Q1 (x) = n0 (x) Q2 (x) + r0 (x) with deg(r0 ) < deg(Q2 )

Q2 (x)
• The second step is to apply long division to r0 (x)
to find polynomials
n1 (x) and r1 (x) such that

Q2 (x) = n1 (x) r0 (x) + r1 (x) with deg(r1 ) < deg(r0 ) or r1 (x) = 0

• The third step (assuming that r1 (x) was not zero) is to apply long division
to rr01 (x)
(x)
to find polynomials n2 (x) and r2 (x) such that

r0 (x) = n2 (x) r1 (x) + r2 (x) with deg(r2 ) < deg(r1 ) or r2 (x) = 0

• And so on.

As the degree of the remainder ri (x) decreases by at least one each time i is
increased by one, the above iteration has to terminate with some r`+1 (x) =

221
I NTEGRATION 1.10 PARTIAL F RACTIONS

0. That is, we choose ` to be index of the last nonzero remainder. Here is a

summary of all of the long division steps.

Q1 (x) = n0 (x) Q2 (x) + r0 (x) with deg(r0 ) < deg(Q2 )

Q2 (x) = n1 (x) r0 (x) + r1 (x) with deg(r1 ) < deg(r0 )
r0 (x) = n2 (x) r1 (x) + r2 (x) with deg(r2 ) < deg(r1 )
r1 (x) = n3 (x) r2 (x) + r3 (x) with deg(r3 ) < deg(r2 )
..
.
r`−2 (x) = n` (x) r`−1 (x) + r` (x) with deg(r` ) < deg(r`−1 )
r`−1 (x) = n`+1 (x) r` (x) + r`+1 (x) with r`+1 = 0

Now we are going to take a closer look at all of the different remainders that
we have generated.

• From first long division step, namely Q1 (x) = n0 (x) Q2 (x)+r0 (x) we have
that the remainder

r0 (x) = Q1 (x) − n0 (x) Q2 (x)

• From the second long division step, namely Q2 (x) = n1 (x) r0 (x) + r1 (x)
we have that the remainder
 
r1 (x) = Q2 (x) − n1 (x) r0 (x) = Q2 (x) − n1 (x) Q1 (x) − n0 (x) Q2 (x)
= A1 (x) Q1 (x) + B1 (x) Q2 (x)

with A1 (x) = −n1 (x) and B1 (x) = 1 + n0 (x) n1 (x).

• From the third long division step (assuming that r1 (x) was not zero),
namely r0 (x) = n2 (x) r1 (x) + r2 (x), we have that the remainder

r2 (x) = r0 (x) − n2 (x) r1 (x)

   
= Q1 (x) − n0 (x) Q2 (x) − n2 (x) A1 (x) Q1 (x) + B1 (x) Q2 (x)
= A2 (x) Q1 (x) + B2 (x) Q2 (x)

with A2 (x) = 1 − n2 (x) A1 (x) and B2 (x) = −n0 (x) − n2 (x) B1 (x).

• And so on. Continuing in this way, we conclude that the final nonzero
remainder r` (x) = A` (x) Q1 (x) + B` (x) Q2 (x) for some polynomials A`
and B` .

Now the last nonzero remainder r` (x) has to be a nonzero constant C because

• it is nonzero by the definition of r` (x) and

• if r` (x) were a polynomial of degree at least one, then

222
I NTEGRATION 1.10 PARTIAL F RACTIONS

◦ r` (x) would be a factor of r`−1 (x) because r`−1 (x) = n`+1 (x) r` (x) and
◦ r` (x) would be a factor of r`−2 (x) because r`−2 (x) = n` (x) r`−1 (x) +
r` (x) and
◦ r` (x) would be a factor of r`−3 (x) because r`−3 (x) = n`−1 (x) r`−2 (x) +
r`−1 (x) and
◦ · · · and · · ·
◦ r` (x) would be a factor of r1 (x) because r1 (x) = n3 (x) r2 (x) + r3 (x)
and
◦ r` (x) would be a factor of r0 (x) because r0 (x) = n2 (x) r1 (x) + r2 (x)
and
◦ r` (x) would be a factor of Q2 (x) because Q2 (x) = n1 (x) r0 (x) + r1 (x)
and
◦ r` (x) would be a factor of Q1 (x) because Q1 (x) = n0 (x) Q2 (x) + r0 (x)
• so that r` (x) would be a common factor for Q1 (x) and Q2 (x), in contra-
diction to the hypothesis that no two of P (x), Q1 (x) and Q2 (x) have a
common linear or quadratic factor.
P (x)
We now have that A` (x) Q1 (x) + B` (x) Q2 (x) = r` (x) = C. Multiplying by C
gives
P̃1 (x) P̃2 (x) P (x)
P̃2 (x) Q1 (x) + P̃1 (x) Q2 (x) = P (x) or + =
Q1 (x) Q2 (x) Q1 (x) Q2 (x)
with P̃2 (x) = P (x)CA` (x) and P̃1 (x) = P (x)CB` (x) . We’re not quite done, because
there is still the danger that deg(P̃1 ) ≥ deg(Q1 ) or deg(P̃2 ) ≥ deg(Q2 ). To deal
P̃1 (x)
with that possibility, we long divide Q 1 (x)
and call the remainder P1 (x).

P̃1 (x) P1 (x)

= N (x) + with deg(P1 ) < deg(Q1 )
Q1 (x) Q1 (x)
Therefore we have that
P (x) P1 (x) P̃2 (x)
= + N (x) +
Q1 (x) Q2 (x) Q1 (x) Q2 (x)
P1 (x) P̃2 (x) + N (x)Q2 (x)
= +
Q1 (x) Q2 (x)
P1 P2
Denoting P2 (x) = P̃2 (x) + N (x)Q2 (x) gives Q1PQ2 = Q 1
+Q 2
and since deg(P1 ) <
deg(Q1 ), the only thing left to prove is that deg(P2 ) < deg(Q2 ). We assume that
deg(P2 ) ≥ deg(Q2 ) and look for a contradiction. We have
deg(P2 Q1 ) ≥ deg(Q1 Q2 ) > deg(P1 Q2 )
=⇒ deg(P ) = deg(P1 Q2 + P2 Q1 ) = deg(P2 Q1 ) ≥ deg(Q1 Q2 )
which contradicts the hypothesis that deg(P ) < deg(Q1 Q2 ) and the proof is
complete.

223
I NTEGRATION 1.10 PARTIAL F RACTIONS

a It appears in Euclid’s Elements, which was written about 300 BC, and it was probably
known even before that.

For the second of the two simpler results, that we’ll shortly use repeatedly to get
P (x) P (x)
Equation 1.10.11, we consider (x−a) m and (x2 +bx+c)m .

Lemma 1.10.16
Let m ≥ 2 be an integer, and let Q(x) be either x − a or x2 + bx + c, with a, b and
c being real numbers. Let P (x) be a polynomial with real coefficients, which
does not contain Q(x) as a factor, and with deg(P ) < deg(Qm ) = m deg(Q).
Then, for each 1 ≤ i ≤ m, there is a polynomial Pi with deg(Pi ) < deg(Q) or
Pi = 0, such that
P (x) P1 (x) P2 (x) P3 (x) Pm−1 (x) Pm (x)
m
= + 2
+ 3
+ ··· + m−1
+ .
Q(x) Q(x) Q(x) Q(x) Q(x) Q(x)m

In particular, if Q(x) = x − a, then each Pi (x) is just a constant Ai , and if

Q(x) = x2 + bx + c, then each Pi (x) is a polynomial Bi x + Ci of degree at most
one.

Proof. We simply repeatedly use long divison to get

 
P (x) P (x) 1 r1 (x) 1
m
= m−1
= n1 (x) +
Q(x) Q(x) Q(x) Q(x) Q(x)m−1
r1 (x) n1 (x) 1
= m
+
Q(x) Q(x) Q(x)m−2
 
r1 (x) r2 (x) 1
= m
+ n2 (x) +
Q(x) Q(x) Q(x)m−2
r1 (x) r2 (x) n2 (x) 1
= m
+ m−1
+
Q(x) Q(x) Q(x) Q(x)m−3
..
.
r1 (x) r2 (x) rm−2 (x) nm−2 (x) 1
= + + ··· + +
Q(x) m Q(x) m−1 Q(x)3 Q(x) Q(x)
r1 (x) r2 (x) rm−2 (x)
= + + ··· + +
Q(x) m Q(x) m−1 Q(x)3
 
rm−1 (x) 1
nm−1 (x) +
Q(x) Q(x)

224
 I NTEGRATION 1.10 PARTIAL F RACTIONS

r1 (x) r2 (x) rm−2 (x) rm−1 (x) nm−1 (x)

= + + ··· + + +
Q(x) m Q(x) m−1 Q(x)3 Q(x)2 Q(x)

By the rules of long division every deg(ri ) < deg(Q). It is also true that the final
numerator, nm−1 , has deg(nm−1 ) < deg(Q) — that is, we kept dividing by Q
until the degree of the quotient was less than the degree of Q. To see this, note
that deg(P ) < m deg(Q) and

deg(n1 ) = deg(P ) − deg(Q)

deg(n2 ) = deg(n1 ) − deg(Q) = deg(P ) − 2 deg(Q)
..
.
deg(nm−1 ) = deg(nm−2 ) − deg(Q) = deg(P ) − (m − 1) deg(Q)
< m deg(Q) − (m − 1) deg(Q)
= deg(Q)

So, if deg(Q) = 1, then r1 , r2 , . . . , rm−1 , nm−1 are all real numbers, and if
deg(Q) = 2, then r1 , r2 , . . . , rm−1 , nm−1 all have degree at most one.

We are now in a position to get Equation 1.10.11. We use (?) to factor8 D(x) =
(x − a1 )m1 Q2 (x) and use Lemma 1.10.15 to get

N (x) N (x) P1 (x) P2 (x)

= = +
D(x) (x − a1 )m1 Q2 (x) (x − a1 )m1 Q2 (x)

where deg(P1 ) < m1 , and deg(P2 ) < deg(Q2 ). Then we use Lemma Lemma 1.10.16 to
get

N (x) P1 (x) P2 (x) A1,1 A1,2 A1,m1 P2 (x)

= + = + + · · · + +
D(x) (x − a1 )m1 Q2 (x) x − a1 (x − a1 )2 (x − a1 )m1 Q2 (x)
P2 (x)
We continue working on Q 2 (x)
in this way, pulling off of the denominator one (x −
mi 2 ni
ai ) or one (x + bi x + ci ) at a time, until we exhaust all of the factors in the de-
nominator D(x).

1.10.4 tt Exercises

Recall that we are using log x to denote the logarithm of x with base e. In other
courses it is often denoted ln x.

Exercises — Stage 1

8 This is assuming that there is at least one linear factor. If not, we factor D(x) = (x2 + b1 x +
c1 )n1 Q2 (x) instead.

225
I NTEGRATION 1.10 PARTIAL F RACTIONS

1. Below are the graphs of four different quadratic functions. For each
quadratic function, decide whether it is: (i) irreducible, (ii) the product
of two distinct linear factors, or (iii) the product of a repeated linear
factor (and possibly a constant).
y y y y

x x x x

(a) (b) (c) (d)

2. *. Write out the general form of the partial-fractions decomposition of

x3 + 3
. You need not determine the values of any of the coef-
(x2 − 1)2 (x2 + 1)
ficients.

1
3. *. Find the coefficient of x − 1 in the partial fraction decomposition of
3x3 − 2x2 + 11
.
x2 (x − 1)(x2 + 3)

4. Re-write the following rational functions as the sum of a polynomial

and a rational function whose numerator has a strictly smaller degree
than its denominator. (Remember our method of partial fraction de-
composition of a rational function only works when the degree of the
numerator is strictly smaller than the degree of the denominator.)
x3 + 2x + 2
a
x2 + 1
15x4 + 6x3 + 34x2 + 4x + 20
b
5x2 + 2x + 8
2x5 + 9x3 + 12x2 + 10x + 30
c
2x2 + 5

5. Factor the following polynomials into linear and irreducible factors.

a 5x3 − 3x2 − 10x + 6

b x4 − 3x2 − 5

c x4 − 4x3 − 10x2 − 11x − 6

d 2x4 + 12x3 − x2 − 52x + 15

226
I NTEGRATION 1.10 PARTIAL F RACTIONS

6. Here is a fact:

Suppose we have a rational function with a repeated linear fac-

tor (ax+b)n in the denominator, and the degree of the numerator
is strictly less than the degree of the denominator. In the partial
fraction decomposition, we can replace the terms

A1 A2 A3 An
+ 2
+ 3
+ ··· + (1)
ax + b (ax + b) (ax + b) (ax + b)n

with the single term

B1 + B2 x + B3 x2 + · · · + Bn xn−1
(2)
(ax + b)n

and still be guaranteed to find a solution.

Why do we use the sum in (1), rather than the single term in (2), in partial
fraction decomposition?

Exercises — Stage 2
2
dx
Z
7. *. Evaluate x + x2
.
1

1
Z
8. *. Calculate dx.
x4 + x2

12x + 4
Z
9. *. Calculate dx.
(x − 3)(x2 + 1)

10. *. Evaluate the following indefinite integral using partial fraction:

3x2 − 4
Z
F (x) = dx.
(x − 2)(x2 + 4)
x − 13
Z
11. *. Evaluate 2
dx.
x −x−6
5x + 1
Z
12. *. Evaluate dx.
x2 + 5x + 6

5x2 − 3x − 1
Z
13. Evaluate dx.
x2 − 1

4x4 + 14x2 + 2
Z
14. Evaluate dx.
4x4 + x2

227
I NTEGRATION 1.10 PARTIAL F RACTIONS

x2 + 2x − 1
Z
15. Evaluate dx.
x4 − 2x3 + x2

3x2 − 4x − 10
Z
16. Evaluate dx.
2x3 − x2 − 8x + 4

1
10x2 + 24x + 8
Z
17. Evaluate dx.
0 2x3 + 11x2 + 6x + 5

Exercises — Stage 3 In Questions 18 and 19, we use partial fraction to find the an-
tiderivatives of two important functions: cosecant, and cosecant cubed.The purpose
of performing a partial fraction decomposition is to manipulate an integrand into a
form that is easily integrable. These “easily integrable” forms are rational functions
whose denominator is a power of a linear function, or of an irreducible quadratic
function. In Questions 20 through 23, we explore the integration of rational func-
tions whose denominators involve irreducible quadratics.In Questions 24 through
26, we use substitution to turn a non-rational integrand into a rational integrand,
then evaluate the resulting integral using partial fraction. Till now, the partial frac-
tion problems you’ve seen have all looked largely the same, but keep in mind that a
partial fraction decomposition can be a small step in a larger problem.
Z
18. Using the method of Example 1.10.5, integrate csc xdx.

Z
19. Using the method of Example 1.10.6, integrate csc3 xdx.

2
3x3 + 15x2 + 35x + 10
Z
20. Evaluate dx.
1 x4 + 5x3 + 10x2
Z  
3 x−3
21. Evaluate + 2 dx.
x + 2 (x + 2)2
2

1
Z
22. Evaluate dx.
(1 + x2 )3
Z  
3x + 1 3x
23. Evaluate 3x + 2 + dx.
x + 5 (x2 + 5)2

cos θ
Z
24. Evaluate dθ.
3 sin θ + cos2 θ − 3

1
Z
25. Evaluate dt.
e2t + et + 1

228
 I NTEGRATION 1.11 N UMERICAL I NTEGRATION

√
Z
26. Evaluate 1 + ex dx using partial fraction.

27. *. The region R is the portion of the first quadrant where 3 ≤ x ≤ 4 and
10
0≤y≤ √ .
25 − x2
a Sketch the region R.

b Determine the volume of the solid obtained by revolving R around

the x-axis.

c Determine the volume of the solid obtained by revolving R around

the y-axis.
4
28. Find the area of the finite region bounded by the curves y = ,y =
3 + x2
2 1
, x = , and x = 3.
x(x + 1) 4

x
1
Z
29. Let F (x) = dt.
1 t2 − 9
a Give a formula for F (x) that does not involve an integral.

b Find F 0 (x).

1.11q Numerical Integration

By now the reader will have come to appreciate that integration is generally quite a
bit more difficult
R than differentiation. There are a great many simple-looking inte-
2
grals, such as e−x dx, that are either very difficult or even impossible to express in
terms of standard functions 1 . Such integrals are not merely mathematical curiosities,
but arise very naturally in many contexts. For example, the error function
Z x
2 2
erf(x) = √ e−t dt
π 0
is extremely important in many areas of mathematics, and also in many practical
applications of statistics.
In such applications we need to be able to evaluate this integral (and many oth-
ers) at a given numerical value of x. In this section we turn to the problem of how to

1 We apologise for being a little sloppy here — but we just want to say that it can be very hard or
even impossible to write some integrals as some finite sized expression involving polynomials,
exponentials, logarithms and trigonometric functions. We don’t want to get into a discussion of
computability, though that is a very interesting topic.

229
I NTEGRATION 1.11 N UMERICAL I NTEGRATION

find (approximate) numerical values for integrals, without having to evaluate them
algebraically. To develop these methods we return to Riemann sums and our geo-
metric interpretation of the definite integral as the signed area.
We start by describing (and applying) three simple algorithms for generating, nu-
Rb
merically, approximate values for the definite integral a f (x) dx. In each algorithm,
we begin in much the same way as we approached Riemann sums.
• We first select an integer n > 0, called the “number of steps”.
• We then divide the interval of integration, a ≤ x ≤ b, into n equal subintervals,
each of length ∆x = b−a
n
. The first subinterval runs from x0 = a to x1 = a + ∆x.
The second runs from x1 to x2 = a + 2∆x, and so on. The last runs from
xn−1 = b − ∆x to xn = b.

This splits the original integral into n pieces:

Z b Z x1 Z x2 Z xn
f (x) dx = f (x) dx + f (x) dx + · · · + f (x) dx
a x0 x1 xn−1
R xj
Each subintegral xj−1 f (x) dx is approximated by the area of a simple geometric fig-
ure. The three algorithms we consider approximate the area by rectangles, trape-
zoids and parabolas (respectively).

We will explain these rules in detail below, but we give a brief overview here:
1 The midpoint rule approximates each subintegral by the area of a rectangle of
height given by the value of the function at the midpoint of the subinterval
Z xj  
xj−1 + xj
f (x)dx ≈ f ∆x
xj−1 2

230
 I NTEGRATION 1.11 N UMERICAL I NTEGRATION

This is illustrated in the leftmost figure above.

2 The trapezoidal rule approximates each subintegral by the area of a trapezoid

with vertices at (xj−1 , 0), (xj−1 , f (xj−1 )), (xj , f (xj )), (xj , 0):
Z xj
1
f (x)dx ≈ [f (xj−1 ) + f (xj )] ∆x
xj−1 2

The trapezoid is illustrated in the middle figure above. We shall derive the
formula for the area shortly.

3 Simpson’s rule approximates two adjacent subintegrals by the area under a

parabola that passes through the points (xj−1 , f (xj−1 )), (xj , f (xj )) and (xj+1 , f (xj+1 )):
Z xj+1
1
f (x)dx ≈ [f (xj−1 ) + 4f (xj ) + f (xj+1 )] ∆x
xj−1 3

The parabola is illustrated in the right hand figure above. We shall derive the
formula for the area shortly.

Definition 1.11.1 Midpoints.

In what follows we need to refer to the midpoint between xj−1 and xj very
frequently. To save on writing (and typing) we introduce the notation

1
x̄j = (xj−1 + xj ) .
2

1.11.1 tt The midpoint rule

R xj
The integral xj−1 f (x) dx represents the area between the curve y = f (x) and the
x-axis with x running from xj−1 to xj . The width of this region is xj − xj−1 = ∆x.
The height varies over the different values that f (x) takes as x runs from xj−1 to xj .
The midpoint rule approximates this area by the area of a rectangle of width
xj − xj−1 = ∆x and height f (x̄j ) which is the exact height at the midpoint of the
range covered by x.

231
I NTEGRATION 1.11 N UMERICAL I NTEGRATION

The area of the approximating rectangle is f (x̄j )∆x, and the midpoint rule ap-
proximates each subintegral by
Z xj
f (x) dx ≈ f (x̄j )∆x.
xj−1

Applying this approximation to each subinterval and summing gives us the fol-
lowing approximation of the full integral:
Z b Z x1 Z x2 Z xn
f (x) dx = f (x) dx + f (x) dx + · · · + f (x) dx
a x0 x1 xn−1

≈ f (x̄1 )∆x + f (x̄2 )∆x + · · · + f (x̄n )∆x

So notice that the approximation is the sum of the function evaluated at the midpoint
of each interval and then multiplied by ∆x. Our other approximations will have
similar forms.
In summary:

Equation 1.11.2 The midpoint rule.

The midpoint rule approximation is

Z b h i
f (x) dx ≈ f (x̄1 ) + f (x̄2 ) + · · · + f (x̄n ) ∆x
a

b−a
where ∆x = n
and

x0 = a x1 = a + ∆x x2 = a + 2∆x · · · xn−1 = b − ∆x xn = b
x0 +x1 x1 +x2 xn−2 +xn−1 xn−1 +xn
x̄1 = 2
x̄2 = 2
· · · x̄n−1 = 2
x̄n = 2

R1 4
Example 1.11.3 0 1+x2
dx.

We approximate the above integral using the midpoint rule with n = 8 step.
Solution:
1
• First we set up all the x-values that we will need. Note that a = 0, b = 1, ∆x = 8
and
1 2 7 8
x0 = 0 x1 = 8
x2 = 8
··· x7 = 8
x8 = 8
=1

Consequently
1 3 5 15
x̄1 = 16
x̄2 = 16
x̄3 = 16
··· x̄8 = 16

232
I NTEGRATION 1.11 N UMERICAL I NTEGRATION

4
• We now apply Equation 1.11.2 to the integrand f (x) = 1+x2
:

f (x̄1 ) f (x̄2 ) f (x̄n−1 ) f (x̄n )

1  z }| { z }| { z }| { z }| { 
4 4 4 4 4
Z
dx ≈ 2
+ 2
+· · ·+ + ∆x
0 1+x
2 1 + x̄1 1 + x̄2 1 + x̄7 1 + x̄28
2

4 4 4 4 4
= 1 + 32 + 52 + 72 + 92
1 + 162 1 + 16 2 1 + 16 2 1 + 16 2 1 + 16 2

4 4 4 1
+ 2 + 2 + 2
1 + 11 13
1 + 16 1 + 15 8
162 2 162

= 3.98444 + 3.86415 + 3.64413 + 3.35738 + 3.03858+
1
2.71618 + 2.40941 + 2.12890
8
= 3.1429

where we have rounded to four decimal places.

• In this case we can compute the integral exactly (which is one of the reasons it
was chosen as a first example):
1
4
Z 1
dx = 4 arctan x =π
0 1 + x2 0

• So the error in the approximation generated by eight steps of the midpoint rule
is

|3.1429 − π| = 0.0013

• The relative error is then

|approximate − exact| |3.1429 − π|
= = 0.0004
exact π
That is the error is 0.0004 times the actual value of the integral.

• We can write this as a percentage error by multiplying it by 100

|approximate − exact|
percentage error = 100 × = 0.04%
exact
That is, the error is about 0.04% of the exact value.
Example 1.11.3

The midpoint rule gives us quite good estimates of the integral without too much

233
I NTEGRATION 1.11 N UMERICAL I NTEGRATION

work — though it is perhaps a little tedious to do by hand 2 . Of course, it would be

very helpful to quantify what we mean by “good” in this context and that requires
us to discuss errors.
Definition 1.11.4
Suppose that α is an approximation to A. This approximation has

• absolute error |A − α| and

|A−α|
• relative error A
and

• percentage error 100 |A−α|

A

We will discuss errors further in Section 1.11.4 below.

Rπ
Example 1.11.5 0
sin x dx.

As a second example, we apply the midpoint rule with n = 8 steps to the above
integral.

• We again start by setting up all the x-values that we will need. So a = 0, b = π,

∆x = π8 and
π 2π 7π 8π
x0 = 0 x1 = 8
x2 = 8
··· x7 = 8
x8 = 8
=π
Consequently,
π 3π 13π 15π
x̄1 = 16
x̄2 = 16
··· x̄7 = 16
x̄8 = 16

• Now apply Equation 1.11.2 to the integrand f (x) = sin x:

Z π h i
sin x dx ≈ sin(x̄1 ) + sin(x̄2 ) + · · · + sin(x̄8 ) ∆x
0
h
= sin( 16 π
) + sin( 3π
16
) + sin( 5π
16
) + sin( 7π
16
) + sin( 9π
16
)+
i
sin( 11π
16
) + sin( 13π
16
) + sin( 15π
16
) π
8
h
= 0.1951 + 0.5556 + 0.8315 + 0.9808 + 0.9808+
i
0.8315 + 0.5556 + 0.1951 × 0.3927
= 5.1260 × 0.3927 = 2.013

• Again, we have chosen this example so that we can compare it against the exact
value:
Z π
 π
sin xdx = − cos x 0 = − cos π + cos 0 = 2.
0

2 Thankfully it is very easy to write a program to apply the midpoint rule.

234
 I NTEGRATION 1.11 N UMERICAL I NTEGRATION

• So with eight steps of the midpoint rule we achieved

absolute error = |2.013 − 2| = 0.013

|2.013 − 2|
relative error = = 0.0065
2
|2.013 − 2|
percentage error = 100 × = 0.65%
2
With little work we have managed to estimate the integral to within 1% of its
true value.
Example 1.11.5

1.11.2 tt The trapezoidal rule

R xj
Consider again the area represented by the integral xj−1 f (x) dx. The trapezoidal
rule (unsurprisingly) approximates this area by a trapezoid 4 whose vertices lie at
3

(xj−1 , 0), (xj−1 , f (xj−1 )), (xj , f (xj )) and (xj , 0).

R xj
The trapezoidal approximation of the integral xj−1 f (x) dx is the shaded region
in the figure on the right above. It has width xj − xj−1 = ∆x. Its left hand side has
height f (xj−1 ) and its right hand side has height f (xj ).
As the figure below shows, the area of a trapezoid is its width times its average
height.

3 This method is also called the “trapezoid rule” and “trapezium rule”.
4 A trapezoid is a four sided polygon, like a rectangle. But, unlike a rectangle, the top and bottom
of a trapezoid need not be parallel.

235
I NTEGRATION 1.11 N UMERICAL I NTEGRATION

So the trapezoidal rule approximates each subintegral by

Z xj
f (x )+f (xj )
f (x) dx ≈ j−12 ∆x
xj−1

Applying this approximation to each subinterval and then summing the result gives
us the following approximation of the full integral
Z b Z x1 Z x2 Z xn
f (x) dx = f (x) dx + f (x) dx + · · · + f (x) dx
a x0 x1 xn−1
f (x0 )+f (x1 ) f (x1 )+f (x2 )
≈ 2
∆x + 2
∆x + ··· + f (xn−12)+f (xn ) ∆x
h1 1 i
= f (x0 ) + f (x1 ) + f (x2 ) + · · · + f (xn−1 ) + f (xn ) ∆x
2 2
So notice that the approximation has a very similar form to the midpoint rule, ex-
cepting that

• we evaluate the function at the xj ’s rather than at the midpoints, and

• we multiply the value of the function at the endpoints x0 , xn by 21 .

In summary:

Equation 1.11.6 The trapezoidal rule.

The trapezoidal rule approximation is

b h1 1
Z i
f (x) dx ≈ f (x0 ) + f (x1 ) + f (x2 ) + · · · + f (xn−1 ) + f (xn ) ∆x
a 2 2

where
b−a
∆x = n
, x0 = a, x1 = a + ∆x, x2 = a + 2∆x, ··· , xn−1 = b − ∆x, xn = b

To compare and contrast we apply the trapezoidal rule to the examples we did
above with the midpoint rule.

236
I NTEGRATION 1.11 N UMERICAL I NTEGRATION

R1 4
Example 1.11.7 0 1+x2
dx — using the trapezoidal rule.

Solution: We proceed very similarly to Example 1.11.3 and again use n = 8 steps.
4 1
• We again have f (x) = 1+x2
, a = 0, b = 1, ∆x = 8
and
1 2 7 8
x0 = 0 x1 = 8
x2 = 8
··· x7 = 8
x8 = 8
=1

• Applying the trapezoidal rule, Equation 1.11.6, gives

f (x0 ) f (x1 ) f (xn−1 ) f (xn )
1  z }| { z }| { z }| { z }| { 
4 1 4 4 4 1 4
Z
dx ≈ + +· · ·+ + ∆x
0 1 + x2 2 1+x20 1+x21 1+x27 2 1+x28

1 4 4 4 4
= + 1 + 2 + 2
2 1 + 02 1 + 82 1 + 282 1 + 832

4 4 4 4 1 4 1
+ 2 + 2 + 2 + 2 + 2
1 + 482 1 + 582 1 + 682 1 + 782 2 1 + 82 8
8
h1
= × 4 + 3.939 + 3.765 + 3.507
2
1 i1
+ 3.2 + 2.876 + 2.56 + 2.266 + × 2
2 8
= 3.139

to three decimal places.

• The exact value of the integral is still π. So the error in the approximation
generated by eight steps of the trapezoidal rule is |3.139 − π| = 0.0026, which is
100 |3.139−π|
π
% = 0.08% of the exact answer. Notice that this is roughly twice the
error that we achieved using the midpoint rule in Example 1.11.3.

Let us also redo Example 1.11.5 using the trapezoidal rule.

Rπ
Example 1.11.8 0
sin x dx — using the trapezoidal rule.

Solution: We proceed very similarly to Example 1.11.5 and again use n = 8 steps.
π
• We again have a = 0, b = π, ∆x = 8
and
π 2π 7π 8π
x0 = 0 x1 = 8
x2 = 8
··· x7 = 8
x8 = 8
=π

• Applying the trapezoidal rule, Equation 1.11.6, gives

Z π h1 1 i
sin x dx ≈ sin(x0 ) + sin(x1 ) + · · · + sin(x7 ) + sin(x8 ) ∆x
0 2 2

237
 I NTEGRATION 1.11 N UMERICAL I NTEGRATION

h1
= sin 0 + sin π8 + sin 2π
8
+ sin 3π
8
+ sin 4π
8
+ sin 5π
8
2
1 i
+ sin 6π
8
+ sin 7π8
+ sin 8π 8 8
π
2
h1
= ×0 + 0.3827 + 0.7071 + 0.9239 + 1.0000 + 0.9239+
2
1 i
0.7071 + 0.3827 + ×0 × 0.3927
2
= 5.0274 × 0.3927 = 1.974

Rπ π
• The exact answer is 0
sin x dx = − cos x = 2. So with eight steps of the
0
trapezoidal rule we achieved 100 |1.974−2|
2
= 1.3% accuracy. Again this is ap-
proximately twice the error we achieved in Example 1.11.5 using the midpoint
rule.
Example 1.11.8

These two examples suggest that the midpoint rule is more accurate than the
trapezoidal rule. Indeed, this observation is born out by a rigorous analysis of the
error — see Section 1.11.4.

1.11.3 tt Simpson’s Rule

R xj
When we use the trapezoidal rule we approximate the area xj−1 f (x)dx by the area
between the x-axis and a straight line that runs from (xj−1 , f (xj−1 )) to (xj , f (xj )) —
that is, we approximate the function f (x) on this interval by a linear function that
agrees with the function at each endpoint. An obvious way to extend this — just as
we did when extending linear approximations to quadratic approximations in our
differential calculus course — is to approximate the function with a quadratic. This
is precisely what Simpson’s 5 rule does.
Simpson’s rule approximates the integral over two neighbouring subintervals by
the area between a parabola and the x-axis. In order to describe this parabola we
need 3 distinct points (which is why we approximate two subintegrals at a time).
That is, we approximate
Z x1 Z x2 Z x2
f (x) dx + f (x) dx = f (x) dx
x0 x1 x0


by the area
bounded by the
parabola that passes through the three points x0 , f (x0 ) ,
x1 , f (x1 ) and x2 , f (x2 ) , the x-axis and the vertical lines x = x0 and x = x2 .

5 Simpson’s rule is named after the 18th century English mathematician Thomas Simpson, despite
its use a century earlier by the German mathematician and astronomer Johannes Kepler. In many
German texts the rule is often called Kepler’s rule.

238
I NTEGRATION 1.11 N UMERICAL I NTEGRATION

Rx
We repeat this on the next pair of subintervals and approximate x24 f (x) dx by the
area between the x-axis and the part of a parabola

with
x2 ≤ x ≤ x4 .
This parabola
passes through the three points x2 , f (x2 ) , x3 , f (x3 ) and x4 , f (x4 ) . And so on.
Because Simpson’s rule does the approximation two slices at a time, n must be even.
To derive Simpson’s rule formula, we first
find the equation

of the parabola


that passes through the three points x0 , f (x0 ) , x1 , f (x1 ) and x2 , f (x2 ) . Then
we find the area between the x-axis and the part of that parabola with x0 ≤ x ≤
x2 . To simplify this computation consider a parabola passing through the points
(−h, y−1 ), (0, y0 ) and (h, y1 ).
Write the equation of the parabola as
y = Ax2 + Bx + C
Then the area between it and the x-axis with x running from −h to h is
Z h  h
 2  A 3 B 2
Ax + Bx + C dx = x + x + Cx
−h 3 2 −h
2A 3
= h + 2Ch it is helpful to write it as
3
h
2Ah2 + 6C


=
3
Now, the the three points (−h, y−1 ), (0, y0 ) and (h, y1 ) lie on this parabola if and
only if
Ah2 − Bh + C = y−1 at (−h, y−1 )
C = y0 at (0, y0 )
2
Ah + Bh + C = y1 at (h, y1 )
Adding the first and third equations together gives us
2Ah2 + (B − B)h + 2C = y−1 + y1
To this we add four times the middle equation
2Ah2 + 6C = y−1 + 4y0 + y1 .
This means that
h
h
Z
Ax2 + Bx + C dx = 2Ah2 + 6C
 

area =
−h 3

239
I NTEGRATION 1.11 N UMERICAL I NTEGRATION

h
= (y−1 + 4y0 + y1 )
3
Note that here

• h is one half of the length of the x-interval under consideration

• y−1 is the height of the parabola at the left hand end of the interval under con-
sideration

• y0 is the height of the parabola at the middle point of the interval under con-
sideration

• y1 is the height of the parabola at the right hand end of the interval under
consideration

So Simpson’s rule approximates

Z x2
f (x) dx ≈ 13 ∆x f (x0 ) + 4f (x1 ) + f (x2 )
 
x0

and Z x4
f (x) dx ≈ 13 ∆x f (x2 ) + 4f (x3 ) + f (x4 )
 
x2

and so on. Summing these all together gives:

Z b Z x2 Z x4 Z x6 Z xn
f (x) dx = f (x) dx + f (x) dx + f (x) dx + · · · + f (x) dx
a x0 x2 x4 xn−2

≈ ∆x
  ∆x  
3
f (x 0 ) + 4f (x 1 ) + f (x 2 ) + f (x 2 ) + 4f (x 3 ) + f (x 4 )
 3
+ ∆x + · · · + ∆x
  
3
f (x 4 ) + 4f (x 5 ) + f (x 6 ) 3
f (xn−2 ) + 4f (xn−1 ) + f (xn )
h i
= f (x0 )+ 4f (x1 )+ 2f (x2 )+ 4f (x3 )+ 2f (x4 )+ · · · + 2f (xn−2 )+ 4f (xn−1 )+ f (xn ) ∆x 3

In summary

Equation 1.11.9 Simpson’s rule.

The Simpson’s rule approximation is

Z b h
f (x) dx ≈ f (x0 )+ 4f (x1 )+ 2f (x2 )+ 4f (x3 )+ 2f (x4 )+ · · ·
a
i
· · · + 2f (xn−2 )+ 4f (xn−1 )+ f (xn ) ∆x
3

where n is even and

b−a
∆x = n
, x0 = a, x1 = a + ∆x, x2 = a + 2∆x, ··· , xn−1 = b − ∆x, xn = b

240
I NTEGRATION 1.11 N UMERICAL I NTEGRATION

Notice that Simpson’s rule requires essentially no more work than the trapezoidal
rule. In both rules we must evaluate f (x) at x = x0 , x1 , · · · , xn , but we add those
terms multiplied by different constants 6 .
Let’s put it to work on our two running examples.

R1 4
Example 1.11.10 0 1+x2
dx — using Simpson’s rule.

Solution: We proceed almost identically to Example 1.11.7 and again use n = 8 steps.

• We have the same ∆, a, b, x0 , · · · , xn as Example 1.11.7.

• Applying Equation 1.11.9 gives

1
4
Z
2
dx
0 1+x

4 4 4 4 4
≈ + 4 1 + 2 2 + 4 2 + 2 2
1 + 02 1 + 82 1 + 282 1 + 382 1 + 842

4 4 4 4 1
+4 2 + 2 2 + 4 2 + 2
1 + 582 1 + 682 1 + 782 1 + 882 8 × 3
h
= 4+4 × 3.938461538+2 × 3.764705882+4 × 3.506849315+2 × 3.2
i 1
+ 4 × 2.876404494 + 2 × 2.56 + 4 × 2.265486726 + 2
8×3
= 3.14159250

to eight decimal places.

• This agrees with π (the exact value of the integral) to six decimal places. So
the error in the approximation generated by eight steps of Simpson’s rule is
|3.14159250 − π| = 1.5 × 10−7 , which is 100 |3.14159250−π|
π
% = 5 × 10−6 % of the
exact answer.

It is striking that the absolute error approximating with Simpson’s rule is so much
smaller than the error from the midpoint and trapezoidal rules.

midpoint error = 0.0013

trapezoid error = 0.0026
Simpson error = 0.00000015

Buoyed by this success, we will also redo Example 1.11.8 using Simpson’s rule.

6 There is an easy generalisation of Simpson’s rule that uses cubics instead of parabolas. It is
known as Simpson’s second rule and Simpson’s 83 rule. While one can push this approach further
(using quartics, quintics etc), it can sometimes lead to larger errors — the interested reader should
look up Runge’s phenomenon.

241
 I NTEGRATION 1.11 N UMERICAL I NTEGRATION

Rπ
Example 1.11.11 0
sin x dx — Simpson’s rule.

Solution: We proceed almost identically to Example 1.11.8 and again use n = 8 steps.

• We have the same ∆, a, b, x0 , · · · , xn as Example 1.11.7.

• Applying Equation 1.11.9 gives

Z π
sin x dx
0
h i
≈ sin(x0 ) + 4 sin(x1 ) + 2 sin(x2 ) + · · · + 4 sin(x7 ) + sin(x8 ) ∆x 3
h
= sin(0) + 4 sin( π8 ) + 2 sin( 2π 8
) + 4 sin( 3π
8
) + 2 sin( 4π8
)
i
+ 4 sin( 5π
8
) + 2 sin( 6π
8
) + 4 sin( 7π
8
) + sin( 8π
8
π
) 8×3
h
= 0 + 4 × 0.382683 + 2 × 0.707107 + 4 × 0.923880 + 2 × 1.0
i
π
+ 4 × 0.923880 + 2 × 0.707107 + 4 × 0.382683 + 0 8×3
= 15.280932 × 0.130900
= 2.00027

• With only eight steps of Simpson’s rule we achieved 100 2.00027−2

2
= 0.014% ac-
curacy.

Again we contrast the error we achieved with the other two rules:
midpoint error = 0.013
trapezoid error = 0.026
Simpson error = 0.00027
This completes our derivation of the midpoint, trapezoidal and Simpson’s rules
for approximating the values of definite integrals. So far we have not attempted to
see how efficient and how accurate the algorithms are in general. That’s our next
task.

1.11.4 tt Three Simple Numerical Integrators — Error Behaviour

Now we are armed with our three (relatively simple) method for numerical integra-
tion we should give thought to how practical they might be in the real world 7 . Two

7 Indeed, even beyond the “real world” of many applications in first year calculus texts, some of
the methods we have described are used by actual people (such as ship builders, engineers and
surveyors) to estimate areas and volumes of actual objects!

242
I NTEGRATION 1.11 N UMERICAL I NTEGRATION

obvious considerations when deciding whether or not a given algorithm is of any

practical value are

a the amount of computational effort required to execute the algorithm and

b the accuracy that this computational effort yields.

For algorithms like our simple integrators, the bulk of the computational effort usu-
ally goes into evaluating the function f (x). The number of evaluations of f (x) re-
quired for n steps of the midpoint rule is n, while the number required for n steps
of the trapezoidal and Simpson’s rules is n + 1. So all three of our rules require
essentially the same amount of effort — one evaluation of f (x) per step.
To get a first impression of the error behaviour of these methods, we apply them
to a problem whose answer we know exactly:
Z π
π
sin x dx = − cos x 0 = 2.
0

To be a little more precise, we would like to understand how the errors of the three
methods change as we increase the effort we put in (as measured by the number of
steps n). The following table lists the error in the approximate value for this number
generated by our three rules applied with three different choices of n. It also lists the
number of evaluations of f required to compute the approximation.

Midpoint Trapezoidal Simpson’s

n error # evals error # evals error # evals
10 4.1 × 10−1 10 8.2 × 10−1
11 5.5 × 10−3 11
100 4.1 × 10−3 100 8.2 × 10−3
101 5.4 × 10−7 101
1000 4.1 × 10−5 1000 8.2 × 10−5
1001 5.5 × 10−11 1001

Observe that

• Using 101 evaluations of f worth of Simpson’s rule gives an error 80 times

smaller than 1000 evaluations of f worth of the midpoint rule.

• The trapezoidal rule error with n steps is about twice the midpoint rule error
with n steps.

• With the midpoint rule, increasing the number of steps by a factor of 10 appears
to reduce the error by about a factor of 100 = 102 = n2 .

• With the trapezoidal rule, increasing the number of steps by a factor of 10 ap-
pears to reduce the error by about a factor of 102 = n2 .

• With Simpson’s rule, increasing the number of steps by a factor of 10 appears

to reduce the error by about a factor of 104 = n4 .

243
I NTEGRATION 1.11 N UMERICAL I NTEGRATION

So it looks like
b b
1
Z Z
approx value of f (x) dx given by n midpoint steps ≈ f (x) dx + KM ·
a a n2
Z b Z b
1
approx value of f (x) dx given by n trapezoidal steps≈ f (x) dx + KT · 2
a a n
Z b Z b
1
approx value of f (x) dx given by n Simpson’s steps≈ f (x) dx + KM · 4
a a n

with some constants KM , KT and KS . It also seems that KT ≈ 2KM .

Figure
Z π 1.11.12: A log-log plot of the error in the n step approximation to
sin x dx.
0

To test these conjectures for the behaviour of the errors we apply our three rules
with about ten different choices of n of the form n = 2m with m integer. Figure
1.11.12 contains two graphs of the results. The left-hand plot shows the results for

244
I NTEGRATION 1.11 N UMERICAL I NTEGRATION

the midpoint and trapezoidal rules and the right-hand plot shows the results for
Simpson’s rule.
For each rule we are expecting (based on our conjectures above) that the error
en = |exact value − approximate value|
with n steps is (roughly) of the form
1
en = K
nk
for some constants K and k. We would like to test if this is really the case, by graph-
ing Y = en against X = n and seeing if the graph “looks right”. But it is not easy to
tell whether or not a given curve really is Y = XKk , for some specific k, by just looking
at it. However, your eye is pretty good at determining whether or not a graph is a
straight line. Fortunately, there is a little trick that turns the curve Y = XKk into a
straight line — no matter what k is.
Instead of plotting Y against X, we plot log Y against log X. This transformation
8
works because when Y = XKk
log Y = log K − k log X
So plotting y = log Y against x = log X gives the straight line y = log K − kx, which
has slope −k and y-intercept log K.
The three graphs in Figure 1.11.12 plot y = log2 en against x = log2 n for our three
rules. Note that we have chosen to use logarithms 9 with this “unusual base” because
it makes it very clear how much the error is improved if we double the number of
steps used. To be more precise — one unit step along the x-axis represents changing
n 7→ 2n. For example, applying Simpson’s rule with n = 24 steps results in an error
of 0000166, so the point (x = log2 24 = 4, y = log2 0000166 = log 0000166
log 2
= −15.8)
has been included on the graph. Doubling the effort used — that is, doubling the
number of steps to n = 25 — results in an error of 0.00000103. So, the data point
(x = log2 25 = 5 , y = log2 0.00000103 = ln 0.00000103
ln 2
= −19.9) lies on the graph. Note
that the x-coordinates of these points differ by 1 unit.
For each of the three sets of data points, a straight line has also been plotted
“through” the data points. A procedure called linear regression 10 has been used to

8 There is a variant of this trick that works even when you don’t know the answer to the integral
ahead of time. Suppose that you suspect that the approximation satisfies
Mn = A + K n1k
where A is the exact value of the integral and suppose that you don’t know the values of A, K
and k. Then
Mn − M2n = K n1k − K (2n) 1 1

1
k = K 1 − 2k nk

so plotting y = log(Mn − M2n ) against x = log n gives the straight line y = log K 1 − 21k − kx.



9 Now is a good time for a quick revision of logarithms — see “Whirlwind review of logarithms”
in Section 2.7 of the CLP-1 text.
10 Linear regression is not part of this course as its derivation requires some multivariable calculus.
It is a very standard technique in statistics.

245
I NTEGRATION 1.11 N UMERICAL I NTEGRATION

decide precisely which straight line to plot. It provides a formula for the slope and
y-intercept of the straight line which “best fits” any given set of data points. From
the three lines, it sure looks like k = 2 for the midpoint and trapezoidal rules and
k = 4 for Simpson’s rule. It also looks like the ratio between the value of K for
the trapezoidal rule, namely K = 20.7253 , and the value of K for the midpoint rule,
namely K = 2−0.2706 , is pretty close to 2: 20.7253 /2−0.2706 = 20.9959 .
The intuition, about the error behaviour, that we have just developed is in fact
correct — provided the integrand f (x) is reasonably smooth. To be more precise

Theorem 1.11.13 Numerical integration errors.

Assume that |f 00 (x)| ≤ M for all a ≤ x ≤ b. Then

M (b − a)3
the total error introduced by the midpoint rule is bounded by
24 n2
and
M (b − a)3
the total error introduced by the trapezoidal rule is bounded by
12 n2
Z b
when approximating f (x)dx. Further, if |f (4) (x)| ≤ L for all a ≤ x ≤ b, then
a

L (b − a)5
the total error introduced by Simpson’s rule is bounded by .
180 n4

The first of these error bounds in proven in the following (optional) section. Here
are some examples which illustrate how they are used. First let us check that the
above result is consistent with our data in Figure 1.11.12
Rπ
Example 1.11.14 Midpoint rule error approximating 0
sin x dx.

Rπ
• The integral 0
sin x dx has b − a = π.

• The second derivative of the integrand satisfies

d2
sin x = | − sin x| ≤ 1
dx2

So we take M = 1.

• So the error, en , introduced when n steps are used is bounded by

M (b − a)3
|en | ≤
24 n2

246
I NTEGRATION 1.11 N UMERICAL I NTEGRATION

π3 1
=
24 n2
1
≈ 1.29 2
n

• The data in the graph in Figure 1.11.12 gives

1 1
|en | ≈ 2−.2706 2
= 0.83 2
n n
π3 1
which is consistent with the bound |en | ≤ 24 n2
.
Example 1.11.14

In a typical application we would be asked to evaluate a given integral to some

specified accuracy. For example, if you are manufacturer and your machinery can
1 th
only cut materials to an accuracy of 10 of a millimeter, there is no point in making
1 th
design specifications more accurate than 10 of a millimeter.

Example 1.11.15 How many steps for a given accuracy?

Suppose, for example, that we wish to use the midpoint rule to evaluate a
Z 1
2
e−x dx
0

to within an accuracy of 10−6 .

Solution:
• The integral has a = 0 and b = 1.
• The first two derivatives of the integrand are
d −x2 2
e = −2xe−x and
dx
d2 −x2 d −x2 2 2 2
= −2e−x + 4x2 e−x = 2(2x2 − 1)e−x


2
e = − 2xe
dx dx
• As x runs from 0 to 1, 2x2 − 1 increases from −1 to 1, so that
2 2
0 ≤ x ≤ 1 =⇒ |2x2 − 1| ≤ 1, e−x ≤ 1 =⇒ 2(2x2 − 1)e−x ≤ 2
So we take M = 2.
• The error introduced by the n step midpoint rule is at most
M (b − a)3
en ≤
24 n2
2 (1 − 0)3 1
≤ 2
=
24 n 12n2

247
I NTEGRATION 1.11 N UMERICAL I NTEGRATION

• We need this error to be smaller than 10−6 so

1
en ≤ ≤ 10−6 and so
12n2
12n2 ≥ 106 clean up
106
n2 ≥ = 83333.3 square root both sides
12
n ≥ 288.7

So 289 steps of the midpoint rule will do the job.

• In fact n = 289 results in an error of about 3.7 × 10−7 .

a This is our favourite running example of an integral that cannot be evaluated algebraically — we
need to use numerical methods.

Example 1.11.15

That seems like far too much work, and the trapezoidal rule will have twice the
error. So we should look at Simpson’s rule.

Example 1.11.16 How many steps using Simpson’s rule?

R1 2
Suppose now that we wish evaluate 0 e−x dx to within an accuracy of 10−6 — but
now using Simpson’s rule. How many steps should we use?
Solution:
• Again we have a = 0, b = 1.
d4 −x2
• We then need to bound dx4
e on the domain of integration, 0 ≤ x ≤ 1.

d3 −x2 d 2 2 2

3
e = 2(2x2 − 1)e−x = 8xe−x − 4x(2x2 − 1)e−x
dx dx
2
= 4(−2x3 + 3x)e−x
d4 −x2 d 2

4
e = 4(−2x3 + 3x)e−x
dx dx
2 2
= 4(−6x2 + 3)e−x − 8x(−2x3 + 3x)e−x
2
= 4(4x4 − 12x2 + 3)e−x
2
• Now, for any x, e−x ≤ 1. Also, for 0 ≤ x ≤ 1,

0 ≤ x2 , x4 ≤ 1 so
3 ≤ 4x4 + 3 ≤ 7 and
−12 ≤ −12x2 ≤ 0 adding these together gives

248
 I NTEGRATION 1.11 N UMERICAL I NTEGRATION

−9 ≤ 4x4 − 12x2 + 3 ≤ 7

Consequently, |4x4 − 12x2 + 3| is bounded by 9 and so

d4 −x2
e ≤ 4 × 9 = 36
dx4

So take L = 36.

• The error introduced by the n step Simpson’s rule is at most

L (b − a)5
en ≤
180 n4
36 (1 − 0)5 1
≤ 4
= 4
180 n 5n

• In order for this error to be no more than 10−6 we require n to satisfy

1
en ≤ ≤ 10−6 and so
5n4
5n4 ≥ 106
n4 ≥ 200000 take fourth root
n ≥ 21.15

So 22 steps of Simpson’s rule will do the job.

• n = 22 steps actually results in an error of 3.5 × 10−8 . The reason that we get an
error so much smaller than we need is that we have overestimated the number
of steps required. This, in turn, occurred because we made quite a rough bound
d4
of dx4
f (x) ≤ 36. If we are more careful then we will get a slightly smaller n.
It actually turns out a that you only need n = 10 to approximate within 10−6 .

a The authors tested this empirically.

Example 1.11.16

1.11.5 tt Optional — An error bound for the midpoint rule

We now try develop some understanding as to why we got the above experimental
results. We start with the error generated by a single step of the midpoint rule. That
is, the error introduced by the approximation
Z x1
f (x) dx ≈ f (x̄1 )∆x where ∆x = x1 − x0 , x̄1 = x0 +x
2
1

x0

249
I NTEGRATION 1.11 N UMERICAL I NTEGRATION

To do this we are going to need to apply integration by parts in a sneaky way. Let
us start by considering 11 a subinterval α ≤ x ≤ β and let’s call the width of the
subinterval 2q so that β = α + 2q. If we were to now apply the midpoint rule to this
subinterval, then we would write
Z β
f (x)dx ≈ 2q · f (α + q) = qf (α + q) + qf (β − q)
α

since the interval has width 2q and the midpoint is α + q = β − q.

The sneaky trick we will employ is to write
Z β Z α+q Z β
f (x)dx = f (x)dx + f (x)dx
α α β−q

and then examine each of the integrals on the right-hand side (using integration by
parts) and show that they are each of the form
Z α+q
f (x)dx ≈ qf (α + q) + small error term
α
Z β
f (x)dx ≈ qf (β − q) + small error term
β−q
R α+q
Let us apply integration by parts to α f (x)dx — with u = f (x), dv = dx so
du = f 0 (x)dx and we will make the slightly non-standard choice of v = x − α:
Z α+q Z α+q
α+q
(x − α)f 0 (x)dx

f (x)dx = (x − α)f (x) α −
α
Z α+q α
= qf (α + q) − (x − α)f 0 (x)dx
α

Notice that the first term on the right-hand side is the term we need, and that our
non-standard choice of v allowed us to avoid introducing an f (α) term.
Now integrate by parts again using u = f 0 (x), dv = (x − α)dx, so du = f 00 (x), v =
(x−α)2
2
:
Z α+q Z α+q
f (x)dx = qf (α + q) − (x − α)f 0 (x)dx
α α
α+q Z α+q
(x − α)2 0 (x − α)2 00

= qf (α + q) − f (x) + f (x)dx
2 α α 2
Z α+q
q2 0 (x − α)2 00
= qf (α + q) − f (α + q) + f (x)dx
2 α 2
To obtain a similar expression for the other integral, we repeat the above steps and
obtain:
Z β Z β
q2 0 (x − β)2 00
f (x)dx = qf (β − q) + f (β − q) + f (x)dx
β−q 2 β−q 2

11 We chose this interval so that we didn’t have lots of subscripts floating around in the algebra.

250
I NTEGRATION 1.11 N UMERICAL I NTEGRATION

Now add together these two expressions

Z α+q Z β
q2
f (x)dx + f (x)dx = qf (α + q) + qf (β − q) + (f 0 (β − q) − f 0 (α + q))
α β−q 2
Z α+q β
(x − α)2 00 (x − β)2 00
Z
+ f (x)dx + f (x)dx
α 2 β−q 2
Then since α + q = β − q we can combine the integrals on the left-hand side and
eliminate some terms from the right-hand side:
Z β Z α+q Z β
(x − α)2 00 (x − β)2 00
f (x)dx = 2qf (α + q) + f (x)dx + f (x)dx
α α 2 β−q 2
Rearrange this expression a little and take absolute values
Z β Z α+q Z β
(x − α)2 00 (x − β)2 00
f (x)dx − 2qf (α + q) ≤ f (x)dx + f (x)dx
α α 2 β−q 2
where we have also made use of the triangle inequality 12 . By assumption |f 00 (x)| ≤
M on the interval α ≤ x ≤ β, so
Z β Z α+q Z β
(x − α)2 (x − β)2
f (x)dx − 2qf (α + q) ≤ M dx + M dx
α α 2 β−q 2
M q3 M (β − α)3
= =
3 24
where we have used q = β−α2
in the last step.
Thus on any interval xi ≤ x ≤ xi+1 = xi + ∆x
Z xi+1  
xi + xi+1 M
f (x)dx − ∆xf ≤ (∆x)3
xi 2 24
Putting everything together we see that the error using the midpoint rule is
bounded by
Z b
f (x)dx − [f (x̄1 ) + f (x̄2 ) + · · · + f (x̄n )] ∆x
a
Z x1 Z xn
≤ f (x)dx − ∆xf (x̄1 ) + · · · + f (x)dx − ∆xf (x̄n )
x0 xn−1
3
M (b − a)3

M M b−a
≤n× (∆x)3 = n × =
24 24 n 24n2
as required.
A very similar analysis shows that, as was stated in Theorem 1.11.13 above,

12 The triangle inequality says that for any real numbers x, y

|x + y| ≤ |x| + |y|.

251
 I NTEGRATION 1.11 N UMERICAL I NTEGRATION

M (b − a)3
• the total error introduced by the trapezoidal rule is bounded by ,
12 n2
M (b − a)5
• the total error introduced by Simpson’s rule is bounded by
180 n4

1.11.6 tt Exercises

Recall that we are using log x to denote the logarithm of x with base e. In other
courses it is often denoted ln x.
Exercises — Stage 1

1. Suppose we approximate an object to have volume 1.5m3 , when its exact

volume is 1.387m3 . Give the relative error, absolute error, and percent
error of our approximation.
Z 10
2. Consider approximating f (x)dx, where f (x) is the function in the
2
graph below.

x
2 10

a Draw the rectangles associated with the midpoint rule approximation

and n = 4.

b Draw the trapezoids associated with the trapezoidal rule approxima-

tion and n = 4.

You don’t have to give an approximation.

1 4 7 3
3. Let f (x) = − x + x − 3x2 .
12 6
a Find a reasonable value M such that |f 00 (x)| ≤ M for all 1 ≤ x ≤ 6.

b Find a reasonable value L such that |f (4) (x)| ≤ L for all 1 ≤ x ≤ 6.

252
I NTEGRATION 1.11 N UMERICAL I NTEGRATION

4. Let f (x) = x sin x+2 cos x. Find a reasonable value M such that |f 00 (x)| ≤ M
for all −3 ≤ x ≤ 2.
Z π
5. Consider the quantity A = cos xdx.
−π

a Find the upper bound on the error using Simpson’s rule with n = 4
to approximate A using Theorem 1.11.13 in the text.

b Find the Simpson’s rule approximation of A using n = 4.

c What is the (actual) absolute error in the Simpson’s rule approxi-

mation of A with n = 4?

6. Give a function f (x) such that:

• f 00 (x) ≤ 3 for every x in [0, 1], and

Z 1
• the error using the trapezoidal rule approximating f (x)dx with
0
1
n = 2 intervals is exactly
.
16
a
7. Suppose my mother is under 100 years old, and I am under 200 years old.
Who is older?

a We’re going somewhere with this.

8.

a True or False: for fixed positive constants M , n, a, and b, with b > a,

M (b − a)3 M (b − a)3
≤
24 n2 12 n2

b True or False: for a function f (x) and fixed constants n, a, and b, with
Z b
b > a, the n-interval midpoint approximation of f (x)dx is more
a
accurate than the n-interval trapezoidal approximation.

9. *. Decide whether the following statement is true or false. If false, pro-

vide a counterexample. If true, provide a brief justification.
When f (x) is positive and concave up, any trapezoidal rule
Z b
approximation for f (x) dx will be an upper estimate for
Z b a

f (x) dx.
a

253
I NTEGRATION 1.11 N UMERICAL I NTEGRATION

10. Give a polynomial f (x) with the property that the Simpson’s rule approxi-
Z b
mation of f (x)dx is exact for all a, b, and n.
a

Exercises — Stage 2 Questions 11 and 12 ask you to approximate a given integral

using the formulas in Equations 1.11.2, 1.11.6, and 1.11.9 in the text.Questions 13
though 17 ask you to approximate a quantity based on observed data.In Questions 18
through 24, we practice finding error bounds for our approximations.

30
1
Z
11. Write out all three approximations of dx with n = 6. (That
0 +1 x3
is: midpoint, trapezoidal, and Simpson’s.) You do not need to simplify
your answers.
Z π
12. *. Find the midpoint rule approximation to sin xdx with n = 3.
0

13. *. The solid V is 40 cm high and the horizontal cross sections are circular
disks. The table below gives the diameters of the cross sections in centime-
ters at 10 cm intervals. Use the trapezoidal rule to estimate the volume of
V.

height 0 10 20 30 40
diameter 24 16 10 6 4

14. *. A 6 metre long cedar log has cross sections that are approximately cir-
cular. The diameters of the log, measured at one metre intervals, are given
below:

metres from left end of log 0 1 2 3 4 5 6

diameter in metres 1.2 1 0.8 0.8 1 1 1.2

Use Simpson’s Rule to estimate the volume of the log.

15. *. The circumference of an 8 metre high tree at different heights above
the ground is given in the table below. Assume that all horizontal cross–
sections of the tree are circular disks.

height (metres) 0 2 4 6 8
circumference (metres) 1.2 1.1 1.3 0.9 0.2

Use Simpson’s rule to approximate the volume of the tree.

16. *. By measuring the areas enclosed by contours on a topographic map, a

geologist determines the cross sectional areas A in m2 of a 60 m high hill.
The table below gives the cross sectional area A(h) at various heights h.

254
I NTEGRATION 1.11 N UMERICAL I NTEGRATION

R 60
The volume of the hill is V = 0
A(h) dh.

h 0 10 20 30 40 50 60
A 10,200 9,200 8,000 7,100 4,500 2,400 100

a If the geologist uses the Trapezoidal Rule to estimate the volume

of the hill, what will be their estimate, to the nearest 1,000m3 ?

b What will be the geologist’s estimate of the volume of the hill if

they use Simpson’s Rule instead of the Trapezoidal Rule?

17. *. The graph below applies to both parts (a) and (b).

a Use the Trapezoidal Rule, with n = 4, to estimate the area under

the graph between x = 2 and x = 6. Simplify your answer com-
pletely.

b Use Simpson’s Rule, with n = 4, to estimate the area under the

graph between x = 2 and x = 6.
Z 1
18. *. The integral sin(x2 ) dx is estimated using the Midpoint Rule with
−1
1000 intervals. Show that the absolute error in this approximation is at most
2 · 10−6 . Rb
You may use the fact that when approximating a f (x) dx with the Midpoint
Rule using n points, the absolute value of the error is at most M (b−a)3 /24n2
when |f 00 (x)| ≤ M for all x ∈ [a, b].

255
I NTEGRATION 1.11 N UMERICAL I NTEGRATION

19. *. The total error using the midpoint rule with n subintervals to ap-
M (b − a)3
proximate the integral of f (x) over [a, b] is bounded by , if
(24n2 )
|f 00 (x)| ≤ M for all a ≤ x ≤ b. Z 1
Using this bound, if the integral 2x4 dx is approximated using the
−2
midpoint rule with 60 subintervals, what is the largest possible error
between the approximation M60 and the true value of the integral?
Z 2
20. *. Both parts of this question concern the integral I = (x − 3)5 dx.
0

a Write down the Simpson’s Rule approximation to I with n = 6. Leave

your answer in calculator-ready form.

b Which method of approximating I results in a smaller error bound:

the Midpoint Rule with n = 100 intervals, or Simpson’s Rule with
n = 10 intervals? You may use the formulas

M (b − a)3 L(b − a)5

|EM | ≤ and |ES | ≤ ,
24n2 180n4
where M is an upper bound for |f 00 (x)| and L is an upper bound for
|f (4) (x)|, and EM and ES are the absolute errors arising from the mid-
point rule and Simpson’s rule, respectively.
Z 5
1
21. *. Find a bound for the error in approximating dx using Simpson’s
1 x
rule with n = 4. Do not write down the Simpson’s rule approximation S4 .
Rb
In general the error in approximating a f (x)dx using Simpson’s rule with
L(b − a) b−a
n steps is bounded by (∆x)4 where ∆x = and L ≥ |f (4) (x)|
180 n
for all a ≤ x ≤ b.

22. *. Find a bound for the error in approximating

Z 1
e−2x + 3x3 dx


0

using Simpson’s rule with n = 6. Do not write down the Simpson’s rule
approximation Sn .
Rb
In general, the error in approximating a f (x)dx using Simpson’s rule
L(b − a) b−a
with n steps is bounded by (∆x)4 where ∆x = and L ≥
180 n
|f (4) (x)| for all a ≤ x ≤ b.

256
I NTEGRATION 1.11 N UMERICAL I NTEGRATION
Z 2
23. *. Let I = (1/x) dx.
1

a Write down the trapezoidal approximation T4 for I. You do not need to

simplify your answer.

b Write down the Simpson’s approximation S4 for I. You do not need to

simplify your answer.

c Without computing I, find an upper bound for |I − S4 |. You may use

the fact that if f (4) (x) ≤ L on the interval [a, b], then the error in using
Rb
Sn to approximate a f (x) dx has absolute value less than or equal to
L(b − a)5 /180n4 .
24. *. A function s(x) satisfies s(0) = 1.00664, s(2) = 1.00543, s(4) = 1.00435,
k
s(6) = 1.00331, s(8) = 1.00233. Also, it is known to satisfy s(k) (x) ≤
1000
for 0 ≤ x ≤ 8 and all positive integers k.

a Find the best Trapezoidal Rule and Simpson’s Rule approximations

Z 8
that you can for I = s(x)dx.
0

b Determine the maximum possible sizes of errors in the approxima-

tions you gave in part (a). Recall that if a function f (x) satisfies
f (k) (x) ≤ Kk on [a, b], then
b b
K2 (b − a)3 K4 (b − a)5
Z Z
f (x)dx−Tn ≤ and f (x)dx−Sn ≤
a 12n2 a 180n4

25. *. Consider
Z b
the trapezoidal rule for making numerical approxima-
tions to f (x)dx. The error for the trapezoidal rule satisfies |ET | ≤
a
M (b − a)3
, where |f 00 (x)| ≤ M for a ≤ x ≤ b. If −2 < f 00 (x) < 0
12n2
for 1 ≤ x ≤ 4, find a value of n to guarantee the trapezoidal rule will
Z 4
give an approximation for f (x)dx with absolute error, |ET |, less than
1
0.001.

Exercises — Stage 3
26. *. A swimming pool has the shape shown in the figure below. The verti-
cal cross–sections of the pool are semi–circular disks. The distances in feet
across the pool are given in the figure at 2–foot intervals along the sixteen–
foot length of the pool. Use Simpson’s Rule to estimate the volume of the
pool.

257
I NTEGRATION 1.11 N UMERICAL I NTEGRATION

27. *. A piece of wire 1m long with radius 1mm is made in such a way that
the density varies in its cross–section, but is radially symmetric (that is,
the local density g(r) in kg/m3 depends only on the distance r in mm
from the centre of the wire). Take as given that the total mass W of the
wire in kg is given by
Z 1
−6
W = 2π10 rg(r) dr
0

Data from the manufacturer is given below:

r 0 1/4 1/2 3/4 1

g(r) 8051 8100 8144 8170 8190

a Find the best Trapezoidal Rule approximation that you can for W
based on the data in the table.

b Suppose that it is known that |g 0 (r)| < 200 and |g 00 (r)| < 150 for
all values of r. Determine the maximum possible size of the error
in the approximation you gave in part (a). Recall that if a function
f (x) satisfies |f 00 (x)| ≤ M on [a, b], then

M (b − a)3
|I − Tn | ≤
12n2
Rb
where I = a f (x) dx and Tn is the Trapezoidal Rule approxima-
tion to I using n subintervals.

258
I NTEGRATION 1.11 N UMERICAL I NTEGRATION

28. *. Simpson’s rule can be used to approximate log 2, since log 2 =

2
1
Z
dx.
1 x
a Use Simpson’s rule with 6 subintervals to approximate log 2.

b How many subintervals are required in order to guarantee that the

absolute error is less than 0.00001?
Note that if En is the error using n subintervals, then |En | ≤
L(b − a)5
where L is the maximum absolute value of the fourth
180n4
derivative of the function being integrated and a and b are the end
points of the interval.
Z 2
29. *. Let I = cos(x2 )dx and let Sn be the Simpson’s rule approximation to
0
I using n subintervals.

a Estimate the maximum absolute error in using S8 to approximate I.

b How large should n be in order to ensure that |I − Sn | ≤ 0.0001?

Note: The graph of f 0000 (x), where f (x) = cos(x2 ), is shown below. The
L(b − a)5
absolute error in the Simpson’s rule approximation is bounded by
0000
180n4
when |f (x)| ≤ L on the interval [a, b].

30. *. Define a function f (x) and an integral I by

Z x2 √ Z 1
f (x) = sin( t) dt, I= f (t) dt
0 0

Estimate how many subdivisions are needed to calculate I to five decimal

places of accuracy using the trapezoidal rule.
M (b − a)3
Note that if En is the error using n subintervals, then |En | ≤ ,
12n2

259
I NTEGRATION 1.11 N UMERICAL I NTEGRATION

where M is the maximum absolute value of the second derivative of the

function being integrated and a and b are the limits of integration.
x2
31. Let f (x) be a function a with f 00 (x) = .
x+1

a Show that |f 00 (x)| ≤ 1 whenever x is in the interval [0, 1].

b Find the maximum value of |f 00 (x)| over the interval [0, 1].

c Assuming M = 1, how many intervals should you use to approximate

Z 1
f (x)dx to within 10−5 ?
0

d Using the value of M you found in (b), how many intervals should
Z 1
you use to approximate f (x)dx to within 10−5 ?
0

a For example, f (x) = 16 x3 − 12 x2 + (1 + x) log |x + 1| will do, but you don’t need to know
what f (x) is for this problem.

32. Approximate the function log x with a rational function by approximat-

Z x
1
ing the integral dt using Simpson’s rule. Your rational function
1 t
f (x) should approximate log x with an error of not more than 0.1 for
any x in the interval [1, 3].

1
33. Using an approximation of the area under the curve 2 , show that the
hπ x i+ 1
π
constant arctan 2 is in the interval + 0.321, + 0.323 .
4 4
d4
 
1
You may assume use without proof that =
dx4 1 + x2
24(5x4 − 10x2 + 1)
. You may use a calculator, but only to add, sub-
(x2 + 1)5
tract, multiply, and divide.

260
 I NTEGRATION 1.12 I MPROPER I NTEGRALS

1.12q Improper Integrals

1.12.1 tt Definitions

Rb
To this point we have only considered nicely behaved integrals a f (x)dx. Though
the algebra involved in some of our examples was quite difficult, all the integrals
had
• finite limits of integration a and b, and
• a bounded integrand f (x) (and in fact continuous except possibly for finitely
many jump discontinuities).
Not all integrals we need to study are quite so nice.
Definition 1.12.1
An integral having either an infinite limit of integration or an unbounded in-
tegrand is called an improper integral.

Two examples are

Z ∞ 1
dx dx
Z
and
0 1 + x2 0 x
The first has an infinite domain of integration and the integrand of the second tends
to ∞ as x approaches the left end of the domain of integration. We’ll start with an
example that illustrates the traps that you can fall into if you treat such integrals
sloppily. Then we’ll see how to treat them carefully.

R1 1
Example 1.12.2 −1 x2
dx.

Consider the integral

1
1
Z
dx
−1 x2

If we “do” this integral completely naively then we get

1 1
1 x−1
Z
dx =
−1 x2 −1 −1
1 −1
= −
−1 −1
= −2

261
I NTEGRATION 1.12 I MPROPER I NTEGRALS

which is wrong a . In fact, the answer is ridiculous. The integrand x12 > 0, so the
integral has to be positive.
The flaw in the argument is that the fundamental theorem of calculus, which says
that
Rb
if F 0 (x) = f (x) then a f (x) dx = F (b) − F (a)

is applicable only when F 0 (x) exists and equals f (x) for all a ≤ x ≤ b. In this case
F 0 (x) = x12 does not exist for x = 0. The given integral is improper. We’ll see later
that the correct answer is +∞.

a Very wrong. But it is not an example of “not even wrong” — which is a phrase attributed to
the physicist Wolfgang Pauli who was known for his harsh critiques of sloppy arguments. The
phrase is typically used to describe arguments that are so incoherent that not only can one not
prove they are true, but they lack enough coherence to be able to show they are false. The inter-
ested reader should do a little searchengineing and look at the concept of falisfyability.

Example 1.12.2
R ∞ dx
Let us put this example to one side for a moment and turn to the integral a 1+x 2.

In this case, the integrand is bounded but the domain of integration extends to +∞.
We can evaluate this integral by sneaking up on it. We compute it on a bounded
R R dx
domain of integration, like a 1+x 2 , and then take the limit R → ∞.

Let us put this into practice:

R∞ dx
Example 1.12.3 a 1+x2
.

Solution:

• Since the domain extends to +∞ we first integrate on a finite domain

R R
dx
Z
= arctan x
a 1 + x2 a
= arctan R − arctan a

262
I NTEGRATION 1.12 I MPROPER I NTEGRALS

• We then take the limit as R → +∞:

Z ∞ Z R
dx dx
2
= lim
a 1+x R→∞ a 1 + x2
 
= lim arctan R − arctan a
R→∞
π
= − arctan a.
2
Example 1.12.3

To be more precise, we actually formally define an integral with an infinite domain

as the limit of the integral with a finite domain as we take one or more of the limits
of integration to infinity.
Definition 1.12.4 Improper integral with infinite domain of integration.

RR
a If the integral a
f (x)dx exists for all R > a, then
Z ∞ Z R
f (x)dx = lim f (x)dx
a R→∞ a

when the limit exists (and is finite).

Rb
b If the integral r f (x)dx exists for all r < b, then
Z b Z b
f (x)dx = lim f (x)dx
−∞ r→−∞ r

when the limit exists (and is finite).

RR
c If the integral r f (x)dx exists for all r < R, then
Z ∞ Z c Z R
f (x)dx = lim f (x)dx + lim f (x)dx
−∞ r→−∞ r R→∞ c

when both limits exist (and are finite). Any c can be used.

When the limit(s) exist, the integral is said to be convergent. Otherwise it is

said to be divergent.
R1
We must also be able to treat an integral like 0 dx
x
that has a finite domain of
integration but whose integrand is unbounded near one limit of integration 1 Our
approach is similar — we sneak up on the problem. We compute the integral on a
R1
smaller domain, such as t dx
x
, with t > 0, and then take the limit t → 0+.

1 This will, in turn, allow us to deal with integrals whose integrand is unbounded somewhere
inside the domain of integration.

263
I NTEGRATION 1.12 I MPROPER I NTEGRALS

R1 1
Example 1.12.5 0 x
dx.

Solution:

• Since the integrand is unbounded near x = 0, we integrate on the smaller do-

main t ≤ x ≤ 1 with t > 0:
Z 1 1
1
dx = log |x| = − log |t|
t x t

• We then take the limit as t → 0+ to obtain

Z 1 Z 1
1 1
dx = lim+ dx = lim+ − log |t| = +∞
0 x t=0 t x t=0

Thus this integral diverges to +∞.

Indeed, we define integrals with unbounded integrands via this process:

Definition 1.12.6 Improper integral with unbounded integrand.

Rb
a If the integral t
f (x)dx exists for all a < t < b, then
Z b Z b
f (x)dx = lim f (x)dx
a t→a+ t

when the limit exists (and is finite).

RT
b If the integral a f (x)dx exists for all a < T < b, then
Z b Z T
f (x)dx = lim f (x)dx
a T →b− a

when the limit exists (and is finite).

264
I NTEGRATION 1.12 I MPROPER I NTEGRALS

RT Rb
c Let a < c < b. If the integrals a f (x)dx and t f (x)dx exist for all a <
T < c and c < t < b, then
Z b Z T Z b
f (x)dx = lim f (x)dx + lim f (x)dx
a T →c− a t→c+ t

when both limit exist (and are finite).

When the limit(s) exist, the integral is said to be convergent. Otherwise it is

said to be divergent.

Notice that (c) is used when the integrand is unbounded at some point in the
middle of the domain of integration, such as was the case in our original example
Z 1
1
2
dx
−1 x

A quick computation shows that this integral diverges to +∞

Z 1 Z a Z 1
1 1 1
2
dx = lim− 2
dx + lim+ 2
dx
−1 x a→0 −1 x b→0 b x
   
1 1
= lim− 1 − + lim+ −1
a→0 a b→0 b
= +∞
More generally, if an integral has more than one “source of impropriety” (for
example an infinite domain of integration and an integrand with an unbounded in-
tegrand or multiple infinite discontinuities) then you split it up into a sum of inte-
grals with a single “source of impropriety” in each. For the integral, as a whole, to
converge every term in that sum has to converge.
For example
R∞ dx
Example 1.12.7 −∞ (x−2)x2
.

Consider the integral

∞
dx
Z

−∞ (x − 2)x2
• The domain of integration that extends to both +∞ and −∞.
• The integrand is singular (i.e. becomes infinite) at x = 2 and at x = 0.
• So we would write the integral as
Z ∞ Z a Z 0 Z b
dx dx dx dx
2
= 2
+ 2
+ 2
−∞ (x − 2)x −∞ (x − 2)x a (x − 2)x 0 (x − 2)x

265
 I NTEGRATION 1.12 I MPROPER I NTEGRALS

2 c ∞
dx dx dx
Z Z Z
+ + +
b (x − 2)x2 2 (x − 2)x2 c (x − 2)x2

where

◦ a is any number strictly less than 0,

◦ b is any number strictly between 0 and 2, and
◦ c is any number strictly bigger than 2.

So, for example, take a = −1, b = 1, c = 3.

• When we examine the right-hand side we see that

◦ the first integral has domain of integration extending to −∞

◦ the second integral has an integrand that becomes unbounded as x → 0−,
◦ the third integral has an integrand that becomes unbounded as x → 0+,
◦ the fourth integral has an integrand that becomes unbounded as x → 2−,
◦ the fifth integral has an integrand that becomes unbounded as x → 2+,
and
◦ the last integral has domain of integration extending to +∞.

• Each of these integrals can then be expressed as a limit of an integral on a small

domain.
Example 1.12.7

1.12.2 tt Examples

With the more formal definitions out of the way, we are now ready for some (impor-
tant) examples.
R∞ dx
Example 1.12.8 1 xp
with p > 0.

Solution:
• Fix any p > 0.
R∞
• The domain of the integral 1 dx
xp
extends to +∞ and the integrand 1
xp
is con-
tinuous and bounded on the whole domain.
• So we write this integral as the limit
Z ∞ Z R
dx dx
p
= lim
1 x R→∞ 1 xp

266
I NTEGRATION 1.12 I MPROPER I NTEGRALS

• The antiderivative of 1/xp changes when p = 1, so we will split the problem

into three cases, p > 1, p = 1 and p < 1.

• When p > 1,
R R
dx 1
Z
p
= x1−p
1 x 1−p 1
R1−p − 1
=
1−p

Taking the limit as R → ∞ gives

∞ R
dx dx
Z Z
= lim
1 xp R→∞ 1 xp
1−p
R −1
= lim
R→∞ 1−p
−1 1
= =
1−p p−1
since 1 − p < 0.

• Similarly when p < 1 we have

∞ R
dx dx R1−p − 1
Z Z
= lim = lim
1 xp R→∞ 1 xp R→∞ 1−p
= +∞

because 1 − p > 0 and the term R1−p diverges to +∞.

• Finally when p = 1
R
dx
Z
= log |R| − log 1 = log R
1 x

Then taking the limit as R → ∞ gives us

Z ∞
dx
= lim log |R| = +∞.
1 xp R→∞

• So summarising, we have
(
∞
dx divergent if p ≤ 1
Z
=
1 xp 1
if p > 1
p−1

Example 1.12.8

267
I NTEGRATION 1.12 I MPROPER I NTEGRALS

R1dx
Example 1.12.9 0 xp
with p > 0.

Solution:
• Again fix any p > 0.
R1
• The domain of integration of the integral 0 dx
xp
is finite, but the integrand x1p
becomes unbounded as x approaches the left end, 0, of the domain of integra-
tion.
• So we write this integral as
1 1
dx dx
Z Z
= lim
0 xp t→0+ t xp
• Again, the antiderivative changes at p = 1, so we split the problem into three
cases.
• When p > 1 we have
1 1
dx 1
Z
p
= x1−p
t x 1−p t
1 − t1−p
=
1−p
Since 1 − p < 0 when we take the limit as t → 0 the term t1−p diverges to +∞
and we obtain
Z 1
dx 1 − t1−p
p
= lim+ = +∞
0 x t→0 1−p
• When p = 1 we similarly obtain
Z 1 Z 1
dx dx
= lim
0 x t→0+ t x


= lim − log |t|
t→0+

= +∞
• Finally, when p < 1 we have
Z 1 Z 1
dx dx
p
= lim+ p
0 x t→0 t x
1 − t1−p 1
= lim+ =
t→0 1−p 1−p
since 1 − p > 0.
• In summary
(
1 1
dx if p < 1
Z
1−p
=
0 xp divergent if p ≥ 1

268
I NTEGRATION 1.12 I MPROPER I NTEGRALS

R∞ dx
Example 1.12.10 0 xp
with p > 0.

Solution:

• Yet again fix p > 0.

R∞
• This time the domain of integration of the integral 0 dx
xp
extends to +∞, and
1
in addition the integrand xp becomes unbounded as x approaches the left end,
0, of the domain of integration.

• So we split the domain in two — given our last two examples, the obvious
place to cut is at x = 1:
∞ 1 ∞
dx dx dx
Z Z Z
= +
0 xp 0 xp 1 xp

• We saw, in Example 1.12.9, that the first integral diverged whenever p ≥ 1, and
we also saw, in Example 1.12.8, that the second integral diverged whenever
p ≤ 1.
R∞
• So the integral 0 dx
xp
diverges for all values of p.

R1 dx
Example 1.12.11 −1 x
.

This is a pretty subtle example. Look at the sketch below:

269
I NTEGRATION 1.12 I MPROPER I NTEGRALS

This suggests that the signed area to the left of the y-axis should exactly cancel the
R1
area to the right of the y-axis making the value of the integral −1 dx
x
exactly zero.
But both of the integrals
1 Z 1
dx dx 1
Z h i1
= lim = lim log x = lim log = +∞
0 x t→0+ t x t→0+ t t→0+ t
0 Z T
dx dx
Z h i T
= lim = lim log |x| = lim log |T | = −∞
−1 x T →0− −1 x T →0− −1 T →0−

R1
diverge so −1 dx
x
diverges. Don’t make the mistake of thinking that ∞ − ∞ = 0. It
is undefined. And it is undefined for good reason.
For example, we have just seen that the area to the right of the y-axis is
1
dx
Z
lim = +∞
t→0+ t x

and that the area to the left of the y-axis is (substitute −7t for T above)
Z −7t
dx
lim = −∞
t→0+ −1 x

If ∞ − ∞ = 0, the following limit should be 0.

Z 1 Z −7t 
dx dx h 1 i
lim + = lim log + log | − 7t|
t→0+ t x −1 x t→0+ t
h 1 i
= lim log + log(7t)
t→0+
h t i
= lim − log t + log 7 + log t = lim log 7
t→0+ t→0+

= log 7

This appears to give ∞ − ∞ = log 7. Of course the number 7 was picked at random.
You can make ∞ − ∞ be any number at all, by making a suitable replacement for 7.
Example 1.12.11

Example 1.12.12 Example 1.12.2 revisited.

The careful computation of the integral of Example 1.12.2 is

1 T Z 1
1 1 1
Z Z
dx = lim dx + lim dx
−1 x2 T →0− −1 x
2 t→0+ t x2
h 1 iT h 1 i1
= lim − + lim −
T →0− x −1 t→0+ x t

270
I NTEGRATION 1.12 I MPROPER I NTEGRALS

=∞+∞

Hence the integral diverges to +∞.

Example 1.12.12

R∞ dx
Example 1.12.13 −∞ 1+x2
.

Since
R
dx π
Z h iR
lim 2
= lim arctan x = lim arctan R =
R→∞ 1+x R→∞ 0 R→∞ 2
Z0 0
dx h i 0 π
lim 2
= lim arctan x = lim − arctan r =
r→−∞ r 1+x r→−∞ r r→−∞ 2
R∞ dx
The integral −∞ 1+x2
converges and takes the value π.

R∞ dx
Example 1.12.14 When does e x(log x)p
converge?
R∞ dx
For what values of p does e x(log x)p
converge?
Solution:

• For x ≥ e, the denominator x(log x)p is never zero. So the integrand is bounded
on the entire domain of integration and this integral is improper only because
the domain of integration extends to +∞ and we proceed as usual.

• We have
∞ R
dx dx
Z Z
= lim use substitution
e x(log x)p R→∞ e x(log x)
p
Z log R
du dx
= lim with u = log x, du =
R→∞ 1 up x
 h i
 1 (log R)1−p − 1 if p 6= 1
= lim 1−p
R→∞ 
log(log R) if p = 1
(
divergent if p ≤ 1
= 1
p−1
if p > 1

In this last step we have used similar logic that that used in Example 1.12.8, but
with R replaced by log R.

271
I NTEGRATION 1.12 I MPROPER I NTEGRALS

Example 1.12.15 The gamma function.

The gamma function Γ(x) is defined by the improper integral

Z ∞
Γ(t) = xt−1 e−x dx
0

We shall now compute Γ(n) for all natural numbers n.

• To get started, we’ll compute

Z ∞ Z R h iR
−x
Γ(1) = e dx = lim e−x dx = lim − e−x =1
0 R→∞ 0 R→∞ 0

• Then compute
Z ∞
Γ(2) = xe−x dx
0
Z R
= lim xe−x dx
R→∞ 0

Use integration by parts with u = x, dv = e−x dx, so v = −e−x , du = dx


R
Z R 
−x −x
= lim − xe + e dx
R→∞ 0 0
h iR
−x −x
= lim − xe −e
R→∞ 0
=1
For the last equality, we used that lim xe−x = 0.
x→∞

• Now we move on to general n, using the same type of computation as we just

used to evaluate Γ(2). For any natural number n,
Z ∞
Γ(n + 1) = xn e−x dx
0
Z R
= lim xn e−x dx
R→∞ 0

Again integrate by parts with u = xn , dv = e−x dx, so v = −e−x , du = nxn−1 dx


R
Z R 
n −x n−1 −x
= lim − x e + nx e dx
R→∞ 0 0
Z R
= lim n xn−1 e−x dx
R→∞ 0
= nΓ(n)
To get to the third row, we used that lim xn e−x = 0.
x→∞

272
 I NTEGRATION 1.12 I MPROPER I NTEGRALS

• Now that we know Γ(2) = 1 and Γ(n + 1) = nΓ(n), for all n ∈ N, we can
compute all of the Γ(n)’s.

Γ(2) = 1
Γ(3) = Γ(2 + 1) = 2Γ(2) = 2 · 1
Γ(4) = Γ(3 + 1) = 3Γ(3) = 3 · 2 · 1
Γ(5) = Γ(4 + 1) = 4Γ(4) = 4 · 3 · 2 · 1
..
.
Γ(n) = (n − 1) · (n − 2) · · · 4 · 3 · 2 · 1 = (n − 1)!

That is, the factorial is just a the Gamma function shifted by one.

a The Gamma function is far more important than just a generalisation of the factorial. It appears
all over mathematics, physics, statistics and beyond. It has all sorts of interesting properties and
its definition can be extended from natural numbers n to all numbers excluding 0, −1, −2, −3, · · ·.
For example, one can show that
π
Γ(1 − z)Γ(z) = .
sin πz

Example 1.12.15

1.12.3 tt Convergence Tests for Improper Integrals

It is very common to encounter integrals that are too complicated to evaluate ex-
plicitly. Numerical approximation schemes, evaluated by computer, are often used
instead (see Section 1.11). You want to be sure that at least the integral converges
before feeding it into a computer 2 . Fortunately it is usually possible to determine
whether or not an improper integral converges even when you cannot evaluate it
explicitly.

Remark 1.12.16 For pedagogical purposes, we are R ∞ going to concentrate on the

problem of determining whether or not an integral a f (x)dx converges, when f (x)
has no singularities for x ≥ a. Recall that the first step in analyzing any improper
integral is to write it as a sum of integrals each of has only a single “source of im-
propriety” — either a domain of integration that extends to +∞, or a domain of
integration that extends to −∞, or an integrand which is singular at one end of the
domain of integration. So we are now going to consider only the first of these three

2 Applying numerical integration methods to a divergent integral may result in perfectly reason-
ably looking but very wrong answers.

273
I NTEGRATION 1.12 I MPROPER I NTEGRALS

possibilities. But the techniques that we are about to see have obvious analogues for
the other two possibilities.
R∞
Now let’s start. Imagine that we have an improper integral a f (x)dx, that f (x)
has no singularities for x ≥ a and that f (x) is complicated enough that we can-
Rnot evaluate the integral explicitly 3 . The idea is find another improper integral
∞
a
g(x)dx
R∞
• with g(x) simple enough that we can evaluate R ∞the integral a g(x)dx explicitly,
or at least determine easily whether or not a g(x)dx converges, and
R∞
• with g(x) behaving enough R∞ like f (x) for large x that the integral a
f (x)dx
converges if and only if a g(x)dx converges.
So far, this is a pretty vague strategy. Here is a theorem which starts to make it more
precise.

Theorem 1.12.17 Comparison.

Let a be a real number. Let f and g be functions that are defined and continu-
ous for all x ≥ a and assume that g(x) ≥ 0 for all x ≥ a.
R∞ R∞
a If |f (x)| ≤ g(x) for all x ≥ a and if a g(x)dx converges then a f (x)dx
also converges.
R∞ R∞
b If f (x) ≥ g(x) for all x ≥ a and if a g(x)dx diverges then a f (x)dx also
diverges.

We will not prove this theorem, but, hopefully, the following supporting argu-
ments should at least appear reasonable to you. Consider the figure below:

R∞
• If a
g(x)dx converges, then the area of

(x, y) x ≥ a, 0 ≤ y ≤ g(x) is finite.

R∞ 2
3 You could, for example, think of something like our running example a
e−t dt.

274
I NTEGRATION 1.12 I MPROPER I NTEGRALS

When |f (x)| ≤ g(x), the region

 
(x, y) x ≥ a, 0 ≤ y ≤ |f (x)| is contained inside (x, y) x ≥ a, 0 ≤ y ≤ g(x)

and so must also have finite area. Consequently the areas of both the regions
 
(x, y) x ≥ a, 0 ≤ y ≤ f (x) and (x, y) x ≥ a, f (x) ≤ y ≤ 0

are finite too 4 .

R∞
• If a g(x)dx diverges, then the area of

(x, y) x ≥ a, 0 ≤ y ≤ g(x) is infinite.

When f (x) ≥ g(x), the region

 
(x, y) x ≥ a, 0 ≤ y ≤ f (x) contains the region (x, y) x ≥ a, 0 ≤ y ≤ g(x)

and so also has infinite area.

R∞ 2
Example 1.12.18 1
e−x dx.
R∞ 2
We cannot evaluate the integral 1 e−x dx explicitly a , however we would still like
to understand if it is finite or not — does it converge or diverge?
Solution: We will use Theorem 1.12.17 to answer the question.
• So we want
R ∞ −xto find another integral that we can compute and that we can com-
2 2
pare to 1 e dx. To do so we pick an integrand that looks like e−x , but whose
indefinite integral we know — such as e−x .
2
• When x ≥ 1, we have x2 ≥ x and hence e−x ≤ e−x . Thus we can use Theo-
rem 1.12.17 to compare
Z ∞ Z ∞
−x2
e dx with e−x dx
1 1

• The integral
Z ∞ Z R
−x
e dx = lim e−x dx
1 R→∞ 1
h iR
= lim − e−x
R→∞ 1
h i
−1 −R
= lim e − e = e−1
R→∞

converges.

4 We
R ∞ have separated the regions in which f (x) is positive
 and negative, because the integral
a f (x)dx represents the signed area of the union of (x, y) x ≥ a, 0 ≤ y ≤ f (x) and
(x, y) x ≥ a, f (x) ≤ y ≤ 0 .

275
I NTEGRATION 1.12 I MPROPER I NTEGRALS

2
• RSo, by Theorem 1.12.17, with a = 1, f (x) = e−x and g(x) = e−x , the integral
∞ −x2
1
e dx converges too (it is approximately equal to 0.1394).

a It has been the subject of many remarks and footnotes.

Example 1.12.18

R∞ 2
Example 1.12.19 1/2
e−x dx.

Solution:
R∞ 2 R∞ 2
• The integral 1/2 e−x dx is quite similar to the integral 1 e−x dx of Example
1.12.18. But we cannot just repeat the argument of Example 1.12.18 because it
2
is not true that e−x ≤ e−x when 0 < x < 1.
2
• In fact, for 0 < x < 1, x2 < x so that e−x > e−x .

• However the difference between the current example and Example 1.12.18 is
Z ∞ Z ∞ Z 1
−x2 −x2 2
e dx − e dx = e−x dx
1/2 1 1/2

which is clearly a well defined finite number (its actually about 0.286). It is
important to note that we are being a little sloppy by taking the difference of
two integrals like this — we are assuming that both integrals converge. More
on this below.
R∞ 2
• So we would expect that 1/2 e−x dx should be the sum of the proper integral
R1 2 R∞ 2
integral 1/2 e−x dx and the convergent integral 1 e−x dx and so should be a
convergent integral. This is indeed the case. The Theorem below provides the
justification.

Theorem 1.12.20
Let a and c be real numbers with a < c andRlet the function f (x) be continuous
∞
for all x ≥ a. Then theR improper integral a f (x) dx converges if and only if
∞
the improper integral c f (x) dx converges.

276
I NTEGRATION 1.12 I MPROPER I NTEGRALS

R∞
Proof. By definition the improper integral a f (x)dx converges if and only
if the limit
Z R Z c Z R 
lim f (x)dx = lim f (x)dx + f (x)dx
R→∞ a R→∞ a c
Z c Z R
= f (x)dx + lim f (x)dx
a R→∞ c
Rc
exists and is finite. (Remember that, in computing the limit, a f (x)dx is a finite
constant independent of R and so can be pulled out of the limit.) But that is the
RR
case if and only if the limit limR→∞ Rc f (x)dx exists and is finite, which in turn
∞
is the case if and only if the integral c f (x)dx converges.

R∞ √
x
Example 1.12.21 Does 1 x2 +x
dx converge?
R∞ √
x
Does the integral 1 x2 +x
dx converge or diverge?
Solution:
• Our first task is to identify the potential sources of impropriety for this integral.
• The domain of integration extends to +∞, but we must also check to see if the
integrand contains any singularities. On the domain of integration x ≥ 1 so
the denominator is never zero and the integrand is continuous. So the only
problem is at +∞.
• Our second task is to develop some intuition a . As the only problem is that the
domain of integration extends to infinity, whether or not the integral converges
will be determined by the behavior of the integrand for very large x.
• When x is very large, x2 is much much larger than x (which we can write as
x2  x) so that the denominator x2 + x ≈ x2 and the integrand
√ √
x x 1
2
≈ 2 = 3/2
x +x x x
R∞
• By Example 1.12.8, with p = 32 , the integral 1 xdx
3/2 converges. So we would
R ∞ √x
expect that 1 x2 +x dx converges too.
• Our final task is to verify that our intuition is correct.
√
To do so, we want to
x 1
apply part (a) of Theorem 1.12.17 with f (x) = x2 +x and g(x) being x3/2 , or
1
possibly some constant times x3/2 . That is, we need to show that for all x ≥ 1
(i.e. on the domain of integration)
√
x A
2
≤ 3/2
x +x x

277
I NTEGRATION 1.12 I MPROPER I NTEGRALS

for some constant A. Let’s try this.

• Since x ≥ 1 we know that

x2 + x > x2

Now take the reciprocal of both sides:

1 1
<
x2 + x x2
√
Multiply both sides by x (which is always positive, so the sign of the inequal-
ity does not change)
√ √
x x 1
< =
x2 + x x2 x3/2

3
• So Theorem 1.12.17(a) and Example 1.12.8, with p = 2
do indeed show that the
R ∞ √x
integral 1 x2 +x dx converges.

a This takes practice, practice and more practice. At the risk of alliteration — please perform plenty
of practice problems .

Example 1.12.21

Notice that in this last example we managed to show that the integral exists by find-
ing an integrand that behaved the same way for large x. Our intuition then had to be
bolstered with some careful inequalities to apply the comparison Theorem 1.12.17.
It would be nice to avoid this last step and be able jump from the intuition to the
conclusion without messing around with inequalities. Thankfully there is a variant
of Theorem 1.12.17 that is often easier to apply and that also fits well with the sort of
intuition that we developed to solve Example 1.12.21.
A key phrase in the previous paragraph is “behaves the same way for large x”.
A good way to formalise this expression — “f (x) behaves like g(x) for large x” — is
to require that the limit
f (x)
lim exists and is a finite nonzero number.
x→∞ g(x)

Suppose that this is the case and call the limit L 6= 0. Then
f (x)
• the ratio g(x)
must approach L as x tends to +∞.

• So when x is very large — say x > B, for some big number B — we must have
that
1 f (x)
L≤ ≤ 2L for all x > B
2 g(x)

278
I NTEGRATION 1.12 I MPROPER I NTEGRALS

Equivalently, f (x) lies between L2 g(x) and 2Lg(x), for all x ≥ B.

• Consequently, the integral of f (x) converges if and only if the integral of g(x)
converges, by Theorems 1.12.17 and 1.12.20.
These considerations lead to the following variant of Theorem 1.12.17.

Theorem 1.12.22 Limiting comparison.

Let −∞ < a < ∞. Let f and g be functions that are defined and continuous for
all x ≥ a and assume that g(x) ≥ 0 for all x ≥ a.
R∞
a If a g(x)dx converges and the limit

f (x)
lim
x→∞ g(x)

R∞
exists, then a f (x)dx converges.
R∞
b If a g(x)dx diverges and the limit

f (x)
lim
x→∞ g(x)

R∞
exists and is nonzero, then a
f (x) diverges.

Note that in (b) the limit must exist and be nonzero, while in (a) we only re-
quire that the limit exists (it can be zero).

Here is an example of how Theorem 1.12.22 is used.

R∞ x+sin x
Example 1.12.23 1 e−x +x2
dx.
∞
x + sin x
Z
Does the integral dx converge or diverge?
1 e−x + x2
Solution:
• Our first task is to identify the potential sources of impropriety for this integral.
• The domain of integration extends to +∞. On the domain of integration the
denominator is never zero so the integrand is continuous. Thus the only prob-
lem is at +∞.
• Our second task is to develop some intuition about the behavior of the inte-
grand for very large x. A good way to start is to think about the size of each
term when x becomes big.
• When x is very large:

279
 I NTEGRATION 1.12 I MPROPER I NTEGRALS

◦ e−x  x2 , so that the denominator e−x + x2 ≈ x2 , and

◦ | sin x| ≤ 1  x, so that the numerator x + sin x ≈ x, and
◦ the integrand x+sin x
e−x +x2
≈ x
x2
= x1 .

Notice that we are using A  B to mean that “A is much much smaller than
B”. Similarly A  B means “A is much much bigger than B”. We don’t really
need to be too precise about its meaning beyond this in the present context.
R∞ R∞
• Now, since 1 dx x
diverges, we would expect 1 ex+sin x
−x +x2 dx to diverge too.

• Our final task is to verify that our intuition is correct. To do so, we set

x + sin x 1
f (x) = g(x) =
e−x + x2 x
and compute

f (x) x + sin x 1
lim = lim −x ÷
x→∞ g(x) x→∞ e + x2 x
(1 + sin x/x)x
= lim −x 2 ×x
x→∞ (e /x + 1)x2

1 + sin x/x
= lim −x 2
x→∞ e /x + 1

=1
R∞ R∞
• Since 1 g(x)dx = 1 dx R ∞ x+sin x1.12.8 with p = 1, Theorem
diverges, by Example
x R∞
1.12.22(b) now tells us that 1 f (x)dx = 1 e−x +x2 dx diverges too.

Example 1.12.23

1.12.4 tt Exercises

Exercises — Stage 1

b
1
Z
1. For which values of b is the integral dx improper?
0 x2 − 1

b
1
Z
2. For which values of b is the integral dx improper?
0 x2 +1

280
I NTEGRATION 1.12 I MPROPER I NTEGRALS

Z ∞
3. Below are the graphs y = f (x) and y = g(x). Suppose f (x)dx con-
Z ∞ 0

verges, and g(x)dx diverges. Assuming the graphs continue on as

0
shown as x → ∞, which graph is f (x), and which is g(x)?

4. *. Decide whether the following statement is true or false. If false, provide

a counterexample. If true, provide a brief justification. (Assume that f (x)
andZ g(x) are continuous functions.)
∞ Z ∞
If f (x) dx converges and g(x) ≥ f (x) ≥ 0 for all x, then g(x) dx
1 1
converges.
1 R∞
5. Let f (x) = e−x and g(x) = . Note 0 f (x)dx converges while
R∞ x+1
0
g(x)dx diverges. R∞
For each of the functions h(x) described below, decide whether 0 h(x)dx
converges or diverges, or whether there isn’t enough information to decide.
Justify your decision.

a h(x), continuous and defined for all x ≥ 0, h(x) ≤ f (x).

b h(x), continuous and defined for all x ≥ 0, f (x) ≤ h(x) ≤ g(x).

c h(x), continuous and defined for all x ≥ 0, −2f (x) ≤ h(x) ≤ f (x).

Exercises — Stage 2

1
x4
Z
6. *. Evaluate the integral dx or state that it diverges.
0 x5 − 1

2
1
Z
7. *. Determine whether the integral dx is convergent or diver-
−2 (x + 1)4/3
gent. If it is convergent, find its value.

281
I NTEGRATION 1.12 I MPROPER I NTEGRALS
∞
1
Z
8. *. Does the improper integral √ dx converge? Justify your an-
1 4x2 − x
swer.
∞
dx
Z
9. *. Does the integral √ converge or diverge? Justify your
0 x2 + x
claim.
Z ∞
10. Does the integral cos xdx converge or diverge? If it converges, evaluate
−∞
it.
Z ∞
11. Does the integral sin xdx converge or diverge? If it converges, eval-
−∞
uate it.
∞
x4 − 5x3 + 2x − 7
Z
12. Evaluate dx, or state that it diverges.
10 x5 + 3x + 8

10
x−1
Z
13. Evaluate dx, or state that it diverges.
0 x2 − 11x + 10

14. *. Determine
Z +∞
(with justification!) which of the following applies to the inte-
x
gral 2
dx:
−∞ x + 1
Z +∞
x
i 2
dx diverges
−∞ x + 1
Z +∞ Z +∞
x x
ii 2
dx converges but dx diverges
−∞ x + 1 −∞ x2 + 1
Z +∞ Z +∞
x x
iii 2
dx converges, as does 2
dx
−∞ x + 1 −∞ x +1
Remark: these options, respectively, are that the integral diverges, con-
verges conditionally, and converges absolutely. You’ll see this terminology
used for series in Section 3.4.1.
Z ∞
| sin x|
15. *. Decide whether I = x + x1/2
3/2
dx converges or diverges. Justify.
0

∞
x+1
Z
16. *. Does the integral dx converge or diverge?
0 x1/3 (x2
+ x + 1)

Exercises — Stage 3

282
I NTEGRATION 1.12 I MPROPER I NTEGRALS

1
17. We craft a tall, vuvuzela-shaped solid by rotating the line y = from
x
x = a to x = 1 about the y-axis, where a is some constant between 0 and
1.
y
1
a

1
y= x

x
−1 a 1

True or false: No matter how large a constant M is, there is some value
of a that makes a solid with volume larger than M .

∞
1
Z
18. *. What is the largest value of q for which the integral x5q
dx diverges?
1

∞
x
Z
19. For which values of p does the integral dx converge?
0 (x2 + 1)p

∞
1
Z
20. Evaluate dt, or state that it diverges.
2 t4 − 1
Z 5 !
1 1 1
21. Does the integral p +p +p dx converge or di-
−5 |x| |x − 1| |x − 2|
verge?
Z ∞
22. Evaluate e−x sin xdx, or state that it diverges.
0

∞
sin4 x
Z
23. *. Is the integral dx convergent or divergent? Explain why.
0 x2

∞
x
Z
24. Does the integral √ dx converge or diverge?
0 ex + x

t
e−x
Z
25. *. Let Mn,t be the Midpoint Rule approximation for 1+x
dx with n equal
0

283
 I NTEGRATION 1.13 M ORE I NTEGRATION E XAMPLES

subintervals. Find a value of t and a value of n such that Mn,t differs from
R ∞ e−x
0 1+x
dx by at most 10−4 . Recall that the error En introduced when the
Midpoint Rule is used with n subintervals obeys

M (b − a)3
|En | ≤
24n2
where M is the maximum absolute value of the second derivative of the
integrand and a and b are the end points of the interval of integration.
Z ∞
26. Suppose f (x) is continuous for all real numbers, and f (x)dx con-
1
verges.
Z −1
a If f (x) is odd, does f (x)dx converge or diverge, or is there not
−∞
enough information to decide?
Z ∞
b If f (x) is even, does f (x)dx converge or diverge, or is there
−∞
not enough information to decide?

27. True or false: x

1
Z
There is some real number x, with x ≥ 1, such that dt = 1.
0 et

1.13q More Integration Examples

tt Exercises
Recall that we are using log x to denote the logarithm of x with base e. In other
courses it is often denoted ln x.

Exercises — Stage 1
1. Match the integration method to a common kind of integrand it’s used to
antidifferentiate.

(A) u-substitution (I) a function multiplied by its derivative

(B) trigonometric substitution (II) a polynomial times an exponential
(C) integration by parts (III) a rational function
(D) partial fractions (IV) the square root of a quadratic function

284
I NTEGRATION 1.13 M ORE I NTEGRATION E XAMPLES

Exercises — Stage 2
Z π/2
2. Evaluate sin4 x cos5 xdx.
0
Z √
3. Evaluate 3 − 5x2 dx.
∞
x−1
Z
4. Evaluate dx.
0 ex
−2
Z
5. Evaluate dx.
3x2 + 4x + 1
Z 2
6. Evaluate x2 log xdx.
1

x
Z
7. *. Evaluate dx.
x2 −3
8. *. Evaluate the following integrals.
Z 4
x
a √ dx
0 9 + x2
Z π/2
b cos3 x sin2 x dx
0
Z e
c x3 log x dx
1

9. *. Evaluate the following integrals.

Z π/2
a x sin x dx
0
Z π/2
b cos5 x dx
0

10. *. Evaluate the following integrals.

Z 2
a xex dx
0

1
1
Z
b √ dx
0 1 + x2
5
4x
Z
c dx
3 (x2 − 1)(x2 + 1)
11. *. Calculate the following integrals.
Z 3√
a 9 − x2 dx
0

285
I NTEGRATION 1.13 M ORE I NTEGRATION E XAMPLES
Z 1
b log(1 + x2 ) dx
0
∞
x
Z
c dx
3 (x − 1)2 (x − 2)
sin4 θ − 5 sin3 θ + 4 sin2 θ + 10 sin θ
Z
12. Evaluate cos θdθ.
sin2 θ − 5 sin θ + 6
13. *. Evaluate the following integrals. Show your work.
Z π
4
a sin2 (2x) cos3 (2x) dx
0
Z

− 32
b 9 + x2 dx

dx
Z
c
(x − 1)(x2 + 1)
Z
d x arctan x dx

14. *. Evaluate the following integrals.

Z π/4
a sin5 (2x) cos(2x) dx
0
Z √
b 4 − x2 dx

x+1
Z
c dx
x2 (x − 1)
15. *. Calculate the following integrals.
Z ∞
a e−x sin(2x) dx
0
√
2
1
Z
b dx
0 (2 + x2 )3/2
Z 1
c x log(1 + x2 ) dx
0
∞
1
Z
d dx
3 (x − 1)2 (x − 2)
16. *. Evaluate the following integrals.
Z
a x log x dx

286
I NTEGRATION 1.13 M ORE I NTEGRATION E XAMPLES

(x − 1) dx
Z
b
x2 + 4x + 5
dx
Z
c
x2 − 4x + 3
x2 dx
Z
d
1 + x6
17. *. Evaluate the following integrals.
Z 1
a arctan x dx.
0

2x − 1
Z
b dx.
x2 − 2x + 5
18. *.
x2
Z
a Evaluate dx.
(x3 + 1)101
Z
b Evaluate cos3 x sin4 x dx.
π
cos x
Z
19. Evaluate √ dx.
π/2 sin x
20. *. Evaluate the following integrals.
ex
Z
a dx
(ex + 1)(ex − 3)
Z 4
x2 − 4x + 4
b √ dx
2 12 + 4x − x2
21. *. Evaluate these integrals.

sin3 x
Z
a dx
cos3 x
2
x4
Z
b 10
dx
−2 x + 16
√
Z
22. Evaluate x x − 1dx.
Z √
x2 − 2
23. Evaluate dx.
x2 R
You may use that sec xdx = log | sec x + tan x| + C.
Z π/4
24. Evaluate sec4 x tan5 x dx.
0

287
I NTEGRATION 1.13 M ORE I NTEGRATION E XAMPLES

3x2 + 4x + 6
Z
25. Evaluate dx.
(x + 1)3
1
Z
26. Evaluate dx.
x2 + x + 1
Z
27. Evaluate sin x cos x tan xdx.

1
Z
28. Evaluate dx.
x3 +1
Z
29. Evaluate (3x)2 arcsin xdx.

Exercises — Stage 3
√
Z π/2
30. Evaluate cos t + 1 dt.
0
e √
log x
Z
31. Evaluate dx.
1 x
0.2
tan x
Z
32. Evaluate dx.
0.1 log(cos x)
33. *. Evaluate these integrals.
Z
a sin(log x) dx

1
1
Z
b dx
0 x2 − 5x + 6
34. *. Evaluate (with justification).

Z 3 √
a (x + 1) 9 − x2 dx
0

4x + 8
Z
b dx
(x − 2)(x2 + 4)
+∞
1
Z
c dx
−∞ + e−xex
Z r
x
35. Evaluate dx.
1−x
Z 1
x
36. Evaluate e2x ee dx.
0

288
I NTEGRATION 1.13 M ORE I NTEGRATION E XAMPLES

xex
Z
37. Evaluate dx.
(x + 1)2
x sin x
Z
38. Evaluate dx.
cos2 xR
You may use that sec xdx = log | sec x + tan x| + C.
Z
39. Evaluate x(x + a)n dx, where a and n are constants.
Z
40. Evaluate arctan(x2 )dx.

289
 Chapter 2

A PPLICATIONS OF I NTEGRATION

In the previous chapter we defined the definite integral, based on its interpretation
as the area of a region in the xy-plane. We also developed a bunch of theory to help
us work with integrals. This abstract definition, and the associated theory, turns out
to be extremely useful simply because "areas of regions in the xy-plane" appear in a
huge number of different settings, many of which seem superficially not to involve
"areas of regions in the xy-plane". Here are some examples.

• The work involved in moving a particle or in pumping a fluid out of a reservoir.

See section 2.1.

• The average value of a function. See section 2.2.

• The center of mass of an object. See section 2.3.

• The time dependence of temperature. See section 2.4.

• Radiocarbon dating. See section 2.4.

Let us start with the first of these examples.

2.1q Work

2.1.1 tt Work

While computing areas and volumes are nice mathematical applications of integra-
tion we can also use integration to compute quantities of importance in physics and

290
A PPLICATIONS OF I NTEGRATION 2.1 W ORK

statistics. One such quantity is work. Work is a way of quantifying the amount of en-
ergy that is required to act against a force 1 . In SI 2 metric units the force F has units
newtons (which are kilogram-metres per second squared), x has units metres and
the work W has units joules (which are newton-metres or kilogram-metres squared
per second squared).
Definition 2.1.1
The work done by a force F (x) in moving an object from x = a to x = b is
Z b
W = F (x)dx
a

In particular, if the force is a constant, F , independent of x, the work is F ·(b−a).

Here is some motivation for this definition. Consider a particle of mass m moving
along the x-axis. Let the position of the particle at time t be x(t). The particle starts
at position a at time α, moves to the right, finishing at position b > a at time β. While
the particle moves, it is subject to a position-dependent force F (x). Then Newton’s
law of motion 3 says 4 that force is mass times acceleration

d2 x

m (t) = F x(t)
dt2
Now consider our definition of work above. It tells us that the work done in moving
the particle from x = a to x = b is
Z b
W = F (x)dx
a

However, we know the position as a function of time, so we can substitute x = x(t),

dx = dx
dt
dt (using Theorem 1.4.6) and rewrite the above integral:
Z b Z t=β
dx
W = F (x)dx = F (x(t)) dt
a t=α dt

1 For example — if your expensive closed-source textbook has fallen on the floor, work quantifies
the amount of energy required to lift the object from the floor acting against the force of gravity.
2 SI is short for “le système international d’unités” which is French for “the international system
of units”. It is the most recent internationally sanctioned version of the metric system, published
in 1960. It aims to establish sensible units of measurement (no cubic furlongs per hogshead-
Fahrenheit). It defines seven base units — metre (length), kilogram (mass), second (time), kelvin
(temperature), ampere (electric current), mole (quantity of substance) and candela (luminous
intensity). From these one can then establish derived units — such as metres per second for
velocity and speed.
3 Specifically, the second of Newton’s three law of motion. These were first published in 1687 in
his “Philosophiæ Naturalis Principia Mathematica”.
4 It actually says something more graceful in Latin - Mutationem motus proportionalem esse vi
motrici impressae, et fieri secundum lineam rectam qua vis illa imprimitur. Or — The alteration
of motion is ever proportional to the motive force impressed; and is made in the line in which
that force is impressed. It is amazing what you can find on the internet.

291
A PPLICATIONS OF I NTEGRATION 2.1 W ORK

Using Newton’s second law we can rewrite our integrand:

Z β 2
d x dx
=m 2
dt
α dt dt
Z β
dv dx
=m v(t)dt since v(t) =
α dt dt
Z β  
d 1
=m v(t)2 dt
α dt 2
What happened here? By the chain rule, for any function f (t):
 
d 1
f (t) = f (t)f 0 (t).
2
dt 2
In the above computation we have used this fact with f (t) = v(t). Now using the
fundamental theorem of calculus (Theorem 1.3.1 part 2), we have
Z β  
d 1 2
W =m v(t) dt
α dt 2
1 1
= mv(β)2 − mv(α)2 .
2 2
By definition, the function 21 mv(t)2 is the kinetic energy 5 of the particle at time t.
So the work W of Definition 2.1.1 is the change in kinetic energy from the time the
particle was at x = a to the time it was at x = b.

Example 2.1.2 Hooke’s Law.

Imagine that a spring lies along the x-axis. The left hand end is fixed to a wall, but
the right hand end lies freely at x = 0. So the spring is at its “natural length”.

• Now suppose that we wish to stretch out the spring so that its right hand end
is at x = L.
• Hooke’s Law a says that when a (linear) spring is stretched (or compressed) by
x units beyond its natural length, it exerts a force of magnitude kx, where the

5 This is not a physics text so we will not be too precise. Roughly speaking, kinetic energy is the
energy an object possesses due to it being in motion, as opposed to potential energy, which is
the energy of the object due to its position in a force field. Leibniz and Bernoulli determined
that kinetic energy is proportional to the square of the velocity, while the modern term “kinetic
energy” was first used by Lord Kelvin (back while he was still William Thompson).

292
A PPLICATIONS OF I NTEGRATION 2.1 W ORK

constant k is the spring constant of that spring.

• In our case, once we have stretched the spring by x units to the right, the spring
will be trying to pull back the right hand end by applying a force of magnitude
kx directed to the left.

• For us to continue stretching the spring we will have to apply a compensating

force of magnitude kx directed to the right. That is, we have to apply the force
F (x) = +kx.

• So to stretch a spring by L units from its natural length we have to supply the
work
Z L
1
W = kxdx = kL2
0 2

a Robert Hooke (1635–1703) was an English contemporary of Isaac Newton (1643–1727). It was in
a 1676 letter to Hooke that Newton wrote “If I have seen further it is by standing on the shoulders
of Giants.” There is some thought that this was sarcasm and Newton was actually making fun of
Hooke, who had a spinal deformity. However at that time Hooke and Newton were still friends.
Several years later they did have a somewhat public falling-out over some of Newton’s work on
optics.

Example 2.1.2

Example 2.1.3 Spring.

A spring has a natural length of 0.1m. If a 12N force is needed to keep it stretched to
a length of 0.12m, how much work is required to stretch it from 0.12m to 0.15m?
Solution: In order to answer this question we will need to determine the spring
constant and then integrate the appropriate function.
• Our first task is to determine the spring constant k. We are told that when the
spring is stretched to a length of 0.12m, i.e. to a length of 0.12 − 0.1 = 0.02m
beyond its natural length, then the spring generates a force of magnitude 12N.

• Hooke’s law states that the force exerted by the spring, when it is stretched by
x units, has magnitude kx, so
2
12 = k · 0.02 = k · thus
100
k = 600.

• So to stretch the spring

◦ from a length of 0.12m, i.e. a length of x = 0.12 − 0.1 = 0.02m beyond its
natural length,

293
A PPLICATIONS OF I NTEGRATION 2.1 W ORK

◦ to a length of 0.15m, i.e. a length of x = 0.15 − 0.1 = 0.05m beyond its

natural length,

takes work
0.05  0.05
1
Z
W = kxdx = kx2
0.02 2 0.02
= 300 0.052 − 0.022

= 0.63J
Example 2.1.3

Example 2.1.4 Pumping Out a Reservoir.

A cylindrical reservoir a of height h and radius r is filled with a fluid of density ρ.

We would like to know how much work is required to pump all of the fluid out the
top of the reservoir.

Solution: We are going to tackle this problem by applying the standard integral cal-
culus “slice into small pieces” strategy. This is how we computed areas and volumes
— slice the problem into small pieces, work out how much each piece contributes,
and then add up the contributions using an integral.

• Start by slicing the reservoir (or rather the fluid inside it) into thin, horizontal,
cylindrical pancakes, as in the figure above. We proceed by determining how
much work is required to pump out this pancake volume of fluid b .

• Each pancake is a squat cylinder with thickness dx and circular cross section of
radius r and area πr2 . Hence it has volume πr2 dx and mass ρ × πr2 dx.

• Near the surface of the Earth gravity exerts a downward force of mg on a body
of mass m. The constant g = 9.8m/sec2 is called the standard acceleration due to
gravity c . For us to raise the pancake we have to apply a compensating upward

294
A PPLICATIONS OF I NTEGRATION 2.1 W ORK

force of mg, which, for our pancake, is

F = gρ × πr2 dx

• To remove the pancake at height x from the reservoir we need to raise it to

height h. So we have to lift it a distance h − x using the force F = πρgr2 dx,
which takes work πρgr2 (h − x) dx.

• The total work to empty the whole reservoir is

Z h Z h
2 2
W = π ρg r (h − x)dx = π ρg r (h − x)dx
0 0
h x2 i h
= π ρg r2 hx −
2 0
π 2 2
= ρg r h
2

• If we measure lengths in metres and mass in kilograms, then this quantity has
units of Joules. If we instead used feet and pounds d then this would have units
of “foot-pounds”. One foot-pound is equal to 1.355817. . . Joules.

a We could assign units to these measurements — such as metres for the lengths h and r, and
kilograms per cubic metre for the density ρ.
b Potential for a bad “work out how much work out” pun here.
c This quantity is not actually constant — it varies slightly across the surface of earth depending
on local density, height above sea-level and centrifugal force from the earth’s rotation. It is,
for example, slightly higher in Oslo and slightly lower in Singapore. It is actually defined to be
9.80665 m/sec2 by the International Organisation for Standardization.
d It is extremely mysterious to both authors why a person would do science or engineering in
imperial units. One of the authors still has nightmares about having had to do so as a student.
The other author is thankful that he escaped such tortures.

Example 2.1.4

Example 2.1.5 Escape Velocity.

Suppose that you shoot a probe straight up from the surface of the Earth — at what
initial speed must the probe move in order to escape Earth’s gravity?
Solution: We determine this by computing how much work must be done in order
to escape Earth’s gravity. If we assume that all of this work comes from the probe’s
initial kinetic energy, then we can establish the minimum initial velocity required.

• The work done by gravity when a mass moves from the surface of the Earth to

295
A PPLICATIONS OF I NTEGRATION 2.1 W ORK

a height h above the surface is

Z h
W = F (x)dx
0

where F (x) is the gravitational force acting on the mass at height x above the
Earth’s surface.

• The gravitational force a of the Earth acting on a particle of mass m at a height

x above the surface of the Earth is
GM m
F =− ,
(R + x)2

where G is the gravitational constant, M is the mass of the Earth and R is the
radius of the Earth. Note that R + x is the distance from the object to the centre
of the Earth. Additionally, note that this force is negative because gravity acts
downward.

• So the work done by gravity on the probe, as it travels from the surface of the
Earth to a height h, is
h
GM m
Z
W =− 2
dx
0 (R + x)
Z h
1
= −GM m 2
dx
0 (R + x)

A quick application of the substitution rule with u = R + x gives

u(h)
1
Z
= −GM m du
u(0) u2
u=R+h

1
= −GM m −
u u=R
GM m GM m
= −
R+h R

• So if the probe completely escapes the Earth and travels all the way to h = ∞,
gravity does work
h GM m GM m i GM m
lim − =−
h→∞ R+h R R
GM m
The minus sign means that gravity has removed energy R
from the probe.

• To finish the problem we need one more assumption. Let us assume that all
of this energy comes from the probe’s initial kinetic energy and that the probe

296
A PPLICATIONS OF I NTEGRATION 2.1 W ORK

is not fitted with any sort of rocket engine. Hence the initial kinetic energy
1
2
mv 2 (coming from an initial velocity v) must be at least as large as the work
computed above. That is we need

1 2 GM m
mv ≥ which rearranges to give
2 rR
2GM
v≥
R
q
2GM
• The right hand side of this inequality, R
, is called the escape velocity.

a Newton published his inverse square law of universal gravitation in his Principia in 1687. His
law states that the gravitational force between two masses m1 and m2 is
m1 m2
F = −G
r2
where r is the distance separating the (centres of the) masses and G = 6.674 × 10−11 Nm2 /kg2
is the gravitational constant. Notice that r measures the separation between the centres of the
masses not the distance between the surfaces of the objects.

Also, do not confuse G with g — standard acceleration due to gravity. The first measurement
of G was performed by Henry Cavendish in 1798 — the interested reader should look up the
“Cavendish experiment” for details of this very impressive work.

Example 2.1.5

Example 2.1.6 Lifting a Cable.

A 10-metre-long cable of mass 5kg is used to lift a bucket of water, with mass 8kg,
out of a well. Find the work done.
Solution: Denote by y the height of the bucket above the top of the water in the well.
So the bucket is raised from y = 0 to y = 10. The cable has mass density 0.5kg/m. So
when the bucket is at height y,
• the cable that remains to be lifted has mass 0.5(10 − y) kg and
• the remaining
 cable and water
 is subject
 to a downward gravitational force of
magnitude 0.5(10 − y) + 8 g = 13 − y2 g, where g = 9.8 m/sec2 .
So to raise the bucket from height y to height y +dy we need to apply
 a compensating
upward force of 13 − 2 g through distance dy. This takes work 13 − y2 gdy. So the
y
 

total work required is

Z 10 h 10
y2

yi  
13 − gdy = g 13y − = 130 − 25 g = 105g J
0 2 4 0

297
 A PPLICATIONS OF I NTEGRATION 2.1 W ORK

2.1.2 tt Exercises

Exercises — Stage 1
1. Find the work (in joules) required to lift a 3-gram block of matter a height
of 10 centimetres against the force of gravity (with g = 9.8 m/sec2 ).

2. A rock exerts a force of 1 N on the ground where it sits due to gravity.

Use g = 9.8 m/sec2 .
What is the mass of the rock?
How much work (in joules) does it take to lift that rock one metre in the
air?

3. Consider the equation

Z b
W = F (x) dx
a

where x is measured in metres and F (x) is measured in kilogram-metres

per second squared (newtons).
For some large n, we might approximate
n
X
W ≈ F (xi )∆x
i=1

where ∆x = b−a n
and xi is some number in the interval [a+(i−1)∆x, a+i∆x].
(This is just the general form of a Riemann sum).

a What are the units of ∆x?

b What are the units of F (xi )?

c Using your answers above, what are the units of W ?

Remark: we already know the units of W from the text, but the Riemann
sum illustrates why they make sense arising from this particular integral.

smoot
4. Suppose f (x) has units , and x is measured in barns a . What
megaFonzie
R1
are the units of the quantity 0 f (x) dx?

a For this problem, it doesn’t matter what the units measure, but a smoot is a
silly measure of length; a megaFonzie is an apocryphal measure of coolness;
and a barn is a humorous (but actually used) measure of area. For explanations
(and entertainment) see https://en.wikipedia.org/wiki/List_of_
humorous_units_of_measurement and https://en.wikipedia.
org/wiki/List_of_unusual_units_of_measurement (accessed 27

298
A PPLICATIONS OF I NTEGRATION 2.1 W ORK

July 2017).

5. You want to weigh your luggage before a flight. You don’t have a scale or
balance, but you do have a heavy-duty spring from your local engineering-
supply store. You nail it to your wall, marking where the bottom hangs. You
hang a one-litre bag of water (with mass one kilogram) from the spring, and
observe that the spring stretches 1 cm. Where on the wall should you mark
the bottom of the spring corresponding to a hanging mass of 10kg?

bottom of unloaded spring

bottom of spring with 10kg mass

You may assume that the spring obeys Hooke’s law.

6. The work done by a force in moving an object from position x = 1 to

x = b is W (b) = −b3 + 6b2 − 9b + 4 for any b in [1, 3]. At what position x
in [1, 3] is the force the strongest?

Exercises — Stage 2 Questions 9 through 16 offer practice on two broad types of

calculations covered in the text: lifting things against gravity, and stretching springs.
You may make the same physical assumptions as in the text: that is, springs follow
Hooke’s law, and the acceleration due to gravity is a constant −9.8 metres per second
squared.For Questions 18 and 19, use the principle (introduced after Definition 2.1.1
and utilized in Example 2.1.5) that the work done on a particle by a force over a
distance is equal to the change in kinetic energy of that particle.
a
7. *. A variable force F (x) = √x Newtons moves an object along a straight
line when it is a distance of x meters from the origin. If the work done in
moving the object from x = 1 meters to x = 16 meters is 18 joules, what is
the value of a? Don’t worry about the units of a.

8. A tube of air is fitted with a plunger that compresses the air as it is

pushed in. If the natural length of the tube of air is `, when the plunger
has been pushed x metres past its natural position, the force exerted
c
by the air is `−x N, where c is a positive constant (depending on the
particulars of the tube of air) and x < `.

299
A PPLICATIONS OF I NTEGRATION 2.1 W ORK

a What are the units of c?

b How much work does it take to push the plunger from 1 metre
past its natural position to 1.5 metres past its natural position?
(You may assume ` > 1.5.)

9. *. Find the work (in joules) required to stretch a string 10 cm beyond equi-
librium, if its spring constant is k = 50 N/m.

10. *. A force of 10 N (newtons) is required to hold a spring stretched 5

cm beyond its natural length. How much work, in joules (J), is done in
stretching the spring from its natural length to 50 cm beyond its natural
length?

11. *. A 5-metre-long cable of mass 8 kg is used to lift a bucket off the

ground. How much work is needed to raise the entire cable to height 5
m? Ignore the mass of the bucket and its contents.

12. A tank 1 metre high has pentagonal cross sections of area 3 m2 and is filled
with water. How much work does it take to pump out all the water?
You may assume the density of water is 1 kg per 1000 cm3 .

13. *. A sculpture, shaped like a pyramid 3m high sitting on the ground, has
been made by stacking smaller and smaller (very thin) iron plates on top
of one another. The iron plate at height z m above ground level is a square

300
A PPLICATIONS OF I NTEGRATION 2.1 W ORK

whose side length is (3 − z) m. All of the iron plates started on the floor of
a basement 2 m below ground level.
Write down an integral that represents the work, in joules, it took to move
all of the iron from its starting position to its present position. Do not eval-
uate the integral. (You can use 9.8 m/s2 for the acceleration due to gravity
and 8000 kg/m3 for the density of iron.)
14. Suppose a spring extends 5 cm past its natural length when one kilogram
is hung from its end. How much work is done to extend the spring from 5
cm past its natural length to 7 cm past its natural length?

15. Ten kilograms of firewood are hoisted on a rope up a height of 4 metres

to a second-floor deck. If the total work done is 400 joules, what is the
mass of the 4 metres of rope?
You may assume that the rope has the same density all the way along.

16. A 5 kg weight is attached to the middle of a 10-metre long rope, which

dangles out a window. The rope alone has mass 1 kg. How much work
does it take to pull the entire rope in through the window, together with the
weight?
17. A box is dragged along the floor. Friction exerts a force in the opposite
direction of motion from the box, and that force is equal to µ × m × g, where
µ is a constant, m is the mass of the box and g is the acceleration due to
gravity. You may assume g = 9.8 m/sec2 .

a How much work is done dragging a box of mass 10 kg along the floor
for three metres if µ = 0.4?

b Suppose the box contains a volatile substance that rapidly evaporates.

You pull the box at a constant rate of 1 m/sec for three
√ seconds, and
the mass of the box at t seconds (0 ≤ t ≤ 3) is (10 − t) kilograms. If
µ = 0.4, how much work is done pulling the box for three seconds?
18. A ball of mass 1 kg is attached to a spring, and the spring is attached to
a table. The ball moves with some initial velocity, and the spring slows it
down. At its farthest, the spring stretches 10 cm past its natural length. If
the spring constant is 5 N/m, what was the initial velocity of the ball?

You may assume that the ball starts moving with initial velocity v0 , and that
the only force slowing it down is the spring. You may also assume that the
spring started out at its natural length, it follows Hooke’s law, and when it
is stretched its farthest, the velocity of the ball is 0 m/sec.

301
A PPLICATIONS OF I NTEGRATION 2.1 W ORK

19. A mild-mannered university professor who is definitely not a spy no-

tices that when their car is on the ground, it is 2 cm shorter than when it
is on a jack. (That is: when the car is on a jack, its struts are at their nat-
ural length; when on the ground, the weight of the car causes the struts
to compress 2 cm.) The university professor calculates that if they were
to jump a local neighborhood drawbridge, their car would fall to the
ground with a speed of 4 m/sec. If the car can sag 20 cm before impor-
tant parts scrape the ground, and the car has mass 2000 kg unoccupied
(2100 kg with the professor inside), can the professor, who is certainly
not involved in international intrigue, safely jump the bridge?
Assume the car falls vertically, the struts obey Hooke’s law, and the
work done by the struts is equal to the change in kinetic energy of the
car + professor. Use 9.8 m/sec2 for the acceleration due to gravity.

Exercises — Stage 3
20. A disposable paper cup has the shape of a right circular cone with radius 5
cm and height 15 cm, and is completely filled with water. How much work
is done sucking all the water out of the cone with a straw?

You may assume that 1 m3 of water has mass 1000 kilograms, the accelera-
tion due to gravity is −9.8 m/sec2 , and that the water moves as high up as
the very top of the cup and no higher.

21. *. A spherical tank of radius 3 metres is half-full of water. It has a

spout of length 1 metre sticking up from the top of the tank. Find the
work required to pump all of the water in the tank out the spout. The
density of water is 1000 kilograms per cubic metre. The acceleration due
to gravity is 9.8 metres per second squared.

302
A PPLICATIONS OF I NTEGRATION 2.1 W ORK

22. A 5-metre cable is pulled out of a deep hole, where it was dangling
straight down. The cable has density ρ(x) = (10 − x) kg/m, where x
is the distance from the bottom end of the rope. (So, the bottom of the
cable is denser than the top.) How much work is done pulling the cable
out of the hole?

23. A rectangular tank is fitted with a plunger that can raise and lower the
water level by decreasing and increasing the length of its base, as in the
diagrams below. The tank has base width 1 m (which does not change) and
contains 3 m3 of water.
m
1

3m
1

The force of the water acting on any tiny piece of the plunger is P A, where
P is the pressure of the water, and dA is the area of the tiny piece. The
pressure varies with the depth of the piece (below the surface of the wa-
ter). Specifically, P = cD, where D is the depth of the tiny piece and c is a
constant, in this case c = 9800 N/m3 .

a If the length of the base is 3 m, give the force of the water on the entire
plunger. (You can do this with an integral: it’s the sum of the force on
all the tiny pieces of the plunger.)

b If the length of the base is x m, give the force of the water on the entire
plunger.

c Give the work required to move the plunger in so that the base length

303
A PPLICATIONS OF I NTEGRATION 2.1 W ORK

changes from 3 m to 1 m.

24. A leaky bucket picks up 5 L of water from a well, but drips out 1 L every
ten seconds. If the bucket was hauled up 5 metres at a constant speed of
1 metre every two seconds, how much work was done?
Assume the rope and bucket have negligible mass and one litre of water
has 1 kg mass, and use 9.8 m/sec2 for the acceleration due to gravity.

25. The force of gravity between two objects, one of mass m1 and another of
m1 m2
mass m2 , is F = G 2 , where r is the distance between them and G is the
r
gravitational constant.
How much work is required to separate the earth and the moon far enough
apart that the gravitational attraction between them is negligible?
Assume the mass of the earth is 6 × 1024 kg and the mass of the moon is
7 × 1022 kg, and that they are currently 400 000 km away from each other.
m3
Also, assume G = 6.7 × 10−11 kg·sec 2 , and the only force acting on the earth

and moon is the gravity between them.

26. True or false: the work done pulling up a dangling cable of length ` and
mass m (with uniform density) is the same as the work done lifting up
a ball of mass m a height of `/2.

`/2

27. A tank one metre high is filled with watery mud that has settled to be denser
at the bottom than at the top.
At height h metres above the bottom of the tank, the cross-section of the
tank has the shape of the finite region bounded by the two curves y = x2
and y = 2 − h − 3x2 . At height
√ h metres above the bottom of the tank, the
density of the liquid is 1000 2 − h kilograms per cubic metre.
How much work is done to pump all the liquid out of the tank?
You may assume the acceleration due to gravity is 9.8 m/sec2 .
28. An hourglass is 0.2 m tall and shaped such that that y metres above or below

304
 A PPLICATIONS OF I NTEGRATION 2.2 AVERAGES

its vertical centre it has a radius of y 2 + 0.01 m.

It is exactly half-full of sand, which has mass M = 17 kilograms.
How much work is done on the sand by quickly flipping the hourglass
over?

Assume that the work done is only moving against gravity, with g = 9.8
m/sec2 , and the sand has uniform density. Also assume that at the instant
the hourglass is flipped over, the sand has not yet begun to fall, as in the
picture above.
√
29. Suppose at position x a particle experiences a force of F (x) = 1 − x4 N.
Approximate the work done moving the particle from x = 0 to x = 1/2,
accurate to within 0.01 J.

2.2q Averages

Another frequent 1 application of integration is computing averages and other sta-

tistical quantities. We will not spend too much time on this topic — that is best left
to a proper course in statistics — however, we will demonstrate the application of
integration to the problem of computing averages.
Let us start with the definition 2 of the average of a finite set of numbers.

1 Awful pun. The two main approaches to statistics are frequentism and Bayesianism; the latter
named after Bayes’ Theorem which is, in turn, named for Reverend Thomas Bayes. While this
(both the approaches to statistics and their history and naming) is a very interesting and quite
philosophical topic, it is beyond the scope of this course. The interested reader has plenty of
interesting reading here to interest them.
2 We are being a little loose here with the distinction between mean and average. To be much more
pedantic — the average is the arithmetic mean. Other interesting “means” are the geometric and
harmonic means:
1
arithmetic mean = (y1 + y2 + · · · + yn )
n
1
geometric mean = (y1 · y2 · · · yn ) n

305
A PPLICATIONS OF I NTEGRATION 2.2 AVERAGES

Definition 2.2.1
The average (mean) of a set of n numbers y1 , y2 , · · ·, yn is
y1 + y2 + · · · + yn
yave = ȳ = hyi =
n
The notations yave , ȳ and hyi are all commonly used to represent the average.

Now suppose that we want to take the average of a function f (x) with x running
continuously from a to b. How do we even define what that means? A natural
approach is to

• select, for each natural number n, a sample of n, more or less uniformly dis-
tributed, values of x between a and b,

• take the average of the values of f at the selected points,

• and then take the limit as n tends to infinity.

Unsurprisingly, this process looks very much like how we computed areas and vol-
umes previously. So let’s get to it.

• First fix any natural number n.

• Subdivide the interval a ≤ x ≤ b into n equal subintervals, each of width

∆x = b−a
n
.

• The subinterval number i runs from xi−1 to xi with xi = a + i b−a

n
.

• Select, for each 1 ≤ i ≤ n, one value of x from subinterval number i and call it
x∗i . So xi−1 ≤ x∗i ≤ xi .

• The average value of f at the selected points is

n n
1X 1 X b−a
f (x∗i ) = f (x∗i )∆x since ∆x =
n i=1 b − a i=1 n

giving us a Riemann sum.

1
Rb
Now when we take the limit n → ∞ we get exactly b−a a
f (x)dx. That’s why we
define

  −1
1 1 1 1
harmonic mean = + + ···
n y1 y2 yn

All of these quantities, along with the median and mode, are ways to measure the typical value
of a set of numbers. They all have advantages and disadvantages — another interesting topic
beyond the scope of this course, but plenty of fodder for the interested reader and their favourite
search engine. But let us put pedantry (and beyond-the-scope-of-the-course-reading) aside and
just use the terms average and mean interchangeably for our purposes here.

306
A PPLICATIONS OF I NTEGRATION 2.2 AVERAGES

Definition 2.2.2
Let f (x) be an integrable function defined on the interval a ≤ x ≤ b. The
average value of f on that interval is
b
1
Z
fave = f¯ = hf i = f (x)dx
b−a a

Consider the case when f (x) is positive. Then rewriting Definition 2.2.2 as
Rb
fave (b − a) = a
f (x)dx

gives us a link between the average value and the area under the curve. The
right-hand side is the area of the region

(x, y) a ≤ x ≤ b, 0 ≤ y ≤ f (x)
while the left-hand side can be seen as the area of a rectangle of width b − a and
height fave . Since these areas must be the same, we interpretfave as the height of the
rectangle which has the same width and the same area as (x, y) a ≤ x ≤ b, 0 ≤
y ≤ f (x) .
Let us start with a couple of simple examples and then work our way up to harder
ones.

Example 2.2.3 An easy warm-up.

Let f (x) = x and g(x) = x2 and compute their average values over 1 ≤ x ≤ 5.
Solution: We can just plug things into the definition.
Z 5
1
fave = xdx
5−1 1
 5
1 x2
=
4 2 1
1 24
= (25 − 1) =
8 8
=3

as we might expect. And then

5
1
Z
gave = x2 dx
5−1 1

307
A PPLICATIONS OF I NTEGRATION 2.2 AVERAGES

 5
1 x3
=
4 3 1
1 124
= (125 − 1) =
12 12
31
=
3
Example 2.2.3

Something a little more trigonometric

Example 2.2.4 Average of sine.

Find the average value of sin(x) over 0 ≤ x ≤ π2 .

Solution: Again, we just need the definition.
Z π
1 2
average = π sin(x)dx
2
−0 0
  π2
2
= · − cos(x)
π 0
2 π
= (− cos( ) + cos(0))
π 2
2
= .
π

We could keep going. . . But better to do some more substantial examples.

Example 2.2.5 Average velocity.

Let x(t) be the position at time t of a car moving along the x-axis. The velocity of the
car at time t is the derivative v(t) = x0 (t). The average velocity of the car over the
time interval a ≤ t ≤ b is
Z b
1
vave = v(t)dt
b−a a
Z b
1
= x0 (t)dt
b−a a
x(b) − x(a)
= by the fundamental theorem of calculus.
b−a
The numerator in this formula is just the displacement (net distance travelled — if
x0 (t) ≥ 0, it’s the distance travelled) between time a and time b and the denominator
is just the time it took.
Notice that this is exactly the formula we used way back at the start of your differ-

308
A PPLICATIONS OF I NTEGRATION 2.2 AVERAGES

ential calculus class to help introduce the idea of the derivative. Of course this is a
very circuitous way to get to this formula — but it is reassuring that we get the same
answer.
Example 2.2.5

A very physics example.

Example 2.2.6 Peak vs RMS voltage.

When you plug a light bulb into a socket a and turn it on, it is subjected to a voltage
V (t) = V0 sin(ωt − δ)
where
• V0 = 170 volts,
• ω = 2π × 60 (which corresponds to 60 cycles per second b ) and
• the constant δ is an (unimportant) phase. It just shifts the time at which the
voltage is zero
The voltage V0 is the “peak voltage” — the maximum value the voltage takes over
time. More typically we quote the “root mean square” voltage c (or RMS-voltage).
In this example we explain the difference, but to simplify the calculations, let us
simplify the voltage function and just use
V (t) = V0 sin(t)
Since the voltage is a sine-function, it takes both positive and negative values. If we
take its simple average over 1 period then we get
Z 2π
1
Vave = V0 sin(t)dt
2π − 0 0
 2π
V0
= − cos(t)
2π 0
V0 V0
= (− cos(2π) + cos 0) = (−1 + 1)
2π 2π
=0
This is clearly not a good indication of the typical voltage.
What we actually want here is a measure of how far the voltage is from zero. Now
we could do this by taking the average of |V (t)|, but this is a little harder to work
with. Instead we take the average of the square d of the voltage (so it is always
positive) and then take the square root at the end. That is
s
Z 2π
1
Vrms = V (t)2 dt
2π − 0 0

309
A PPLICATIONS OF I NTEGRATION 2.2 AVERAGES

s
2π
1
Z
= V02 sin2 (t)dt
2π 0
s
2π
V02
Z
= sin2 (t)dt
2π 0

This is called the “root mean square” voltage.

Though we do know how to integrate sine and cosine, we don’t (yet) know how to
integrate their squares. A quick look at double-angle formulas e gives us a way to
eliminate the square:

1 − cos(2θ)
cos(2θ) = 1 − 2 sin2 θ =⇒ sin2 θ =
2
Using this we manipulate our integrand a little more:
s
V02 2π 1
Z
Vrms = (1 − cos(2t))dt
2π 0 2
s  2π
V02 1
= t − sin(2t)
4π 2 0
s
V02
 
1 1
= 2π − sin(4π) − 0 + sin(0)
4π 2 2
r
V02
= · 2π
4π
V0
=√
2
170
So if the peak voltage is 170 volts then the RMS voltage is √
2
≈ 120.2.

a A normal household socket delivers alternating current, rather than the direct current USB sup-
plies. At the risk of yet another “the interested reader” suggestion — the how and why house-
hold plugs supply AC current is another worthwhile and interesting digression from studying
integration. The interested reader should look up the “War of Currents”. The diligent and inter-
ested reader should bookmark this, finish the section and come back to it later.
b Some countries supply power at 50 cycles per second. Japan actually supplies both — 50 cycles
in the east of the country and 60 in the west.
c This example was written in North America where the standard voltage supplied to homes is
120 volts. Most of the rest of the world supplies homes with 240 volts. The main reason for this
difference is the development of the light bulb. The USA electrified earlier when the best voltage
for bulb technology was 110 volts. As time went on, bulb technology improved and countries
that electrified later took advantage of this (and the cheaper transmission costs that come with
higher voltage) and standardised at 240 volts. So many digressions in this section!

310
A PPLICATIONS OF I NTEGRATION 2.2 AVERAGES

d For a finite set of numbers one can compute the “quadratic mean” which is another way to
generalise the notion of the average:
r
1 2
quadratic mean = (y + y22 + · · · + yn2 )
n 1

e A quick glance at Appendix A.14 will refresh your memory.

Example 2.2.6

Continuing this very physics example:

Example 2.2.7 Peak vs RMS voltage — continued.

Let us take our same light bulb with voltage (after it is plugged in) given by

V (t) = V0 sin(ωt − δ)

where

• V0 is the peak voltage,

• ω = 2π × 60, and

• the constant δ is an (unimportant) phase.

If the light bulb is “100 watts”, then what is its resistance?

To answer this question we need the following facts from physics.

• If the light bulb has resistance R ohms, this causes, by Ohm’s law, a current of

1
I(t) = V (t)
R
(amps) to flow through the light bulb.

• The current I is the number of units of charge moving through the bulb per
unit time.

• The voltage is the energy required to move one unit of charge through the bulb.

• The power is the energy used by the bulb per unit time and is measured in
watts.

So the power is the product of the current times the voltage and, so

V (t)2 V02
P (t) = I(t)V (t) = = sin2 (ωt − δ)
R R

311
A PPLICATIONS OF I NTEGRATION 2.2 AVERAGES

The average power used over the time interval a ≤ t ≤ b is

Z b Z b
1 V02
Pave = P (t)dt = sin2 (ωt − δ)dt
b−a a R(b − a) a
Notice that this is almost exactly the form we had in the previous example when
computing the root mean square voltage.
Again we simplify the integrand using the identity
1 − cos(2θ)
cos(2θ) = 1 − 2 sin2 θ =⇒ sin2 θ =
2
So
b Z b
1 V02
Z
 
Pave = P (t)dt = 1 − cos(2ωt − 2δ) dt
b−a a 2R(b − a) a
b
V02

sin(2ωt − 2δ)
= t−
2R(b − a) 2ω a
2
 
V0 sin(2ωb − 2δ) sin(2ωa − 2δ)
= b−a− +
2R(b − a) 2ω 2ω
2 2
V V0
= 0 −
 
sin(2ωb − 2δ) − sin(2ωa − 2δ)
2R 4ωR(b − a)
In the limit as the length of the time interval b − a tends to infinity, this converges to
V02
2R
. The resistance R of a “100 watt bulb” obeys

V02 V02
= 100 so that R= .
2R 200
We finish this example off with two side remarks.
• If we translate the peak voltage to the root mean square voltage using
√
V0 = Vrms · 2
then we have
2
Vrms
P =
R
• If we were using direct voltage rather than alternating current then the compu-
tation is much simpler. The voltage and current are constants, so
P =V ·I but I = V /R by Ohm’s law
V2
=
R
So if we have a direct current giving voltage equal to the root mean square
voltage, then we would expend the same power.
Example 2.2.7
312
 A PPLICATIONS OF I NTEGRATION 2.2 AVERAGES

2.2.1 tt Optional — Return to the mean value theorem

One last application of Definition 2.2.2. The following theorem can be thought of
as an analogue of the mean-value theorem (which was covered in your differential
calculus class) but for integrals. The theorem says that a continuous function has
to take its average value exactly somewhere. For example, if you went for a drive
along the x-axis and you were at x(a) at time a and at x(b) at time b, then your
velocity x0 (t) had to be exactly your average velocity x(b)−x(a)
b−a
at some time t between
a and b. In particular, if your average velocity was greater than the speed limit, you
were definitely speeding at some point during the trip. This is, of course, no great
surprise.

Theorem 2.2.8 Mean Value Theorem for Integrals.

Let f (x) be a continuous function on the interval a ≤ x ≤ b. Then there is some

c obeying a ≤ c ≤ b such that
b b
1
Z Z
f (x)dx = f (c) or f (x)dx = f (c) (b − a)
b−a a a

Here is why the theorem is true. Let M and m be the largest and smallest values,
respectively, that f (x) takes for x between a and b. Then m ≤ f (x) ≤ M for all
a ≤ x ≤ b, so that 3
Z b Z b
1
m(b − a) ≤ f (x)dx ≤ M (b − a) ⇐⇒ m≤ f (x)dx ≤ M
a b−a a
As x runs from a to b the continuous function f (x) takes all values between m and
1
Rb
M including, in particular, b−a a
f (x).

2.2.2 tt Exercises

Recall that we are using log x to denote the logarithm of x with base e. In other
courses it is often denoted ln x.
Exercises — Stage 1

1. Below is the graph of a function y = f (x). Its average value on the

R5
interval [0, 5] is A. Draw a rectangle on the graph with area 0 f (x) dx.

3 The symbol ⇐⇒ is read “if and only if”. This is used in mathematics to express the logical
equivalence of two statements. To be more precise, the statement P ⇐⇒ Q tells us that P is true
whenever Q is true and Q is true whenever P is true.

313
A PPLICATIONS OF I NTEGRATION 2.2 AVERAGES

x
5

2. Suppose a car travels for 5 hours in a straight line, with an average velocity
of 100 kph. How far did the car travel?
3. A force F (x) acts on an object from position x = a metres to position x = b
metres, for a total of W joules of work. What was the average force on the
object?
4. Suppose we want to approximate the average value of the function f (x) on
the interval [a, b]. To do this, we cut the interval [a, b] into n pieces, then
take n samples by finding the function’s output at the left endpoint of each
piece, starting with a. Then, we average those n samples. (In the example
below, n = 4.)

y average these y-values

x
a b

a Using n samples, what is the distance between two consecutive sam-

ple points xi and xi+1 ?

b Assuming n ≥ 4, what is the x-coordinate of the fourth sample?

c Assuming n ≥ 4, what is the y-value of the fourth sample?

d Write the approximation of the average value of f (x) over the interval
[a, b] using sigma notation.

314
A PPLICATIONS OF I NTEGRATION 2.2 AVERAGES

5. Suppose f (x) and g(x) are functions that are defined for all numbers in
the interval [0, 10].

a If f (x) ≤ g(x) for all x in [0, 10], then is the average value of f (x)
is less than or equal to the average value of g(x) on the interval
[0, 10], or is there not enough information to tell?

b Suppose f (x) ≤ g(x) for all x in [0.01, 10]. Is the average value
of f (x) less than or equal to the average value of g(x) over the
interval [0, 10], or is there not enough information to tell?

6. Suppose f is an odd function, defined for all real numbers. What is the
average of f on the interval [−10, 10]?

Exercises — sStage 2 For Questions 16 through 18, let the root mean square of f (x)
Z b
1
on [a, b] be f 2 (x) dx. This is the formula used in Example 2.2.6 in the text.
b−a a

7. *. Find the average value of f (x) = sin(5x) + 1 over the interval −π/2 ≤
x ≤ π/2.

2
8. *. Find the average value of the function y = x log x on the interval 1 ≤ x ≤
e.
3 2
9. *. Find the average value of the function f (x) = 3 cos x + 2 cos x on the
π
interval 0 ≤ x ≤ 2 .
10. *. Let k be a positive constant. Find the average value of the function f (x) =
sin(kx) on the interval 0 ≤ x ≤ π/k.

11. *. The temperature in Celsius in a 3 m long rod at a point x metres from

80
the left end of the rod is given by the function T (x) = 16−x 2 . Determine

the average temperature in the rod.

log x
12. *. What is the average value of the function f (x) = x
on the interval
[1, e]?
2
13. *. Find the average value of f (x) = cos (x) over 0 ≤ x ≤ 2π.

14. The carbon dioxide concentration in the air at a particular location over

one year is approximated by C(t) = 400 + 50 cos 12t π + 200 cos 4380 t


π
parts per million, where t is measured in hours.

a What is the average carbon dioxide concentration for that location

for that year?

315
A PPLICATIONS OF I NTEGRATION 2.2 AVERAGES

b What is the average over the first day?

c Suppose measurements were only made at noon every day: that is,
when t = 12 + 24n, where n is any whole number between
 0 and
364. Then the daily variation would cease: 50 cos (12+24n)
12
π =
50 cos (π + 2πn) = 50 cos π = −50. So, the approximation for the
concentration of carbon dioxide in the atmosphere might be given
as  
t
N (t) = 350 + 200 cos π ppm
4380

What is the relative error in the yearly average concentration of

carbon dioxide involved in using N (t), instead of C(t)?

You may assume a day has exactly 24 hours, and a year has exactly 8760
hours.

15. Let S be the solid formed by rotating the parabola y = x2 from x = 0 to

x = 2 about the x-axis.

a What is the average area of the circular cross-sections of S? Call this

value A.

b What is the volume of S?

c What is the volume of a cylinder with circular cross-sectional area A

and length 2?

16. Let f (x) = x.

a Calculate the average of f (x) over [−3, 3].

b Calculate the root mean square of f (x) over [−3, 3].

Calculate the root mean square of f (x) = tan x over − π4 , π4 .

 
17.
18. A force acts on a spring, and the spring stretches and contracts. The distance
beyond its natural length at time t is f (t) = sin (tπ) cm, where t is measured
in seconds. The spring constant is 3 N/cm.

a What is the force exerted by the spring at time t, if it obeys Hooke’s

law?

b Find the average of the force exerted by the spring from t = 0 to t = 6.

c Find the root mean square of the force exerted by the spring from t = 0
to t = 6.

Exercises — Stage 3

316
A PPLICATIONS OF I NTEGRATION 2.2 AVERAGES

19. *. A car travels two hours without stopping. The driver records the
car’s speed every 20 minutes, as indicated in the table below:

time in hours 0 1/3 2/3 1 4/3 5/3 2

speed in km/hr 50 70 80 55 60 80 40

a Use the trapezoidal rule to estimate the total distance traveled in

the two hours.

b Use the answer to part (a) to estimate the average speed of the car
during this period.

20. Let s(t) = et .

a Find the average of s(t) on the interval [0, 1]. Call this quantity A.

b For any point t, the difference between s(t) and A is s(t) − A. Find the
average value of s(t) − A on the interval [0, 1].

c For any point t, the absolute difference between s(t) and A is |s(t)−A|.
Find the average value of |s(t) − A| on the interval [0, 1].

21. Consider the two functions f (x) and g(x) below, both of which have
average A on [0, 4].
y

y = f (x)

y = g(x)
A
x
4

a Which function has a larger average on [0, 4]: f (x) − A or g(x) − A?

b Which function has a larger average on [0, 4]: |f (x) − A| or |g(x) −

A|?

317
A PPLICATIONS OF I NTEGRATION 2.2 AVERAGES

22. Suppose the root mean square of a function f (x) on the interval [a, b] is R.
What is the volume of the solid formed by rotating the portion of f (x) from
a to b about the x-axis?
y

b
x
a

y = f (x)

As in Example 2.2.6, let the root mean square of f (x) on [a, b] be

s
Z b
1
f 2 (x) dx.
b−a a
23. Suppose f (x) = ax2 + bx + c, and the average value of f (x) on the interval
[0, 1] is the same as the average of f (0) and f (1). What is a?
24. Suppose f (x) = ax2 + bx + c, and the average value of f (x) on the interval
[s, t] is the same as the average of f (s) and f (t). Is it possible that a 6= 0?
That is — does the result of Question 23 generalize?
25. Let f (x) be a function defined for all numbers in the interval [a, b], with
average value A over that interval. What is the average of f (a + b − x) over
the interval [a, b]?
26. Suppose f (t) is a continuous function, and A(x) is the average of f (t) on the
interval from 0 to x.

a What is the average of f (t) on [a, b], where a < b? Give your answer
in terms of A.

b What is f (t)? Again, give your answer in terms of A.

27.

a Find a function f (x) with average 0 over [−1, 1] but f (x) 6= 0 for
all x in [−1, 1], or show that no such function exists.

b Find a continuous function f (x) with average 0 over [−1, 1] but

f (x) 6= 0 for all x in [−1, 1], or show that no such function exists.

28. Suppose f (x) is a positive, continuous function with lim f (x) = 0, and let
x→∞
A(x) be the average of f (x) on [0, x].
True or false: lim A(x) = 0.
x→∞

318
 A PPLICATIONS OF I NTEGRATION 2.3 C ENTRE OF M ASS AND T ORQUE
2
29. Let A(x) be the average of the function f (t) = e−t on the interval [0, x].
What is lim A(x)?
x→∞

2.3q Centre of Mass and Torque

2.3.1 tt Centre of Mass

If you support a body at its center of mass (in a uniform gravitational field) it bal-
ances perfectly. That’s the definition of the center of mass of the body.

If the body consists of a finite number of masses m1 , · · ·, mn attached to an in-

finitely strong, weightless (idealized) rod with mass number i attached at position
xi , then the center of mass is at the (weighted) average value of x:

Equation 2.3.1 Centre of mass (discrete masses).

Pn
mi xi
x̄ = Pi=1
n
i=1 mi

The denominator m = ni=1 mi is the total mass of the body.

P
This formula for the center of mass is derived in the following (optional) section.
See equation (2.3.14).
For many (but certainly not all) purposes an (extended rigid) body acts like a
point particle located at its center of mass. For example it is very common to treat
the Earth as a point particle. Here is a more detailed example in which we think of a
body as being made up of a number of component parts and compute the center of
mass of the body as a whole by using the center of masses of the component parts.
Suppose that we have a dumbbell which consists of

• a left end made up of particles of masses ml,1 , · · ·, ml,3 located at xl,1 , · · ·, xl,3
and

• a right end made up of particles of masses mr,1 , · · ·, mr,4 located at xr,1 , · · ·, xr,4
and

• an infinitely strong, weightless (idealized) rod joining all of the particles.

319
A PPLICATIONS OF I NTEGRATION 2.3 C ENTRE OF M ASS AND T ORQUE

Then the mass and center of mass of the left end are
ml,1 xl,1 + · · · + ml,3 xl,3
Ml = ml,1 + · · · + ml,3 X̄l =
Ml
and the mass and center of mass of the right end are
mr,1 xr,1 + · · · + mr,4 xr,4
Mr = mr,1 + · · · + mr,4 X̄r =
Mr
The mass and center of mass of the entire dumbbell are

M = ml,1 + · · · + ml,3 + mr,1 + · · · + mr,4

= Ml + Mr
ml,1 xl,1 + · · · + ml,3 xl,3 + mr,1 xr,1 + · · · + mr,4 xr,4
x̄ =
M
Ml X̄l + Mr X̄r
=
Mr + Ml
So we can compute the center of mass of the entire dumbbell by treating it as being
made up of two point particles, one of mass Ml located at the centre of mass of the
left end, and one of mass Mr located at the center of mass of the right end.

Example 2.3.2 Work and Centre of Mass.

Here is another example in which an extended body acts like a point particle located
at its centre of mass. Imagine that there are a finite number of masses m1 , · · · , mn
arrayed along a (vertical) z-axis with
P mass number i attached at height zi . Note that
the total mass of the array is M = ni=1 mi and that the centre of mass of the array is
at height
Pn n
i=1 mi zi 1 X
z̄ = Pn = mi zi
i=1 mi M i=1

Now suppose that we lift all of the masses, against gravity, to height Z. So after the
lift there is a total mass M located at height Z. The ith mass is subject to a downward
gravitational force of mi g. So to lift the ith mass we need to apply a compensating
upward force of mi g through a distance of Z − zi . This takes work mi g(Z − zi ). So
the total work required to lift all n masses is
n
X
Work = mi g(Z − zi )
i=1
n
X n
X
= gZ mi − g mi zi
i=1 i=1
= gZM − gM z̄
= M g(Z − z̄)

320
A PPLICATIONS OF I NTEGRATION 2.3 C ENTRE OF M ASS AND T ORQUE

So the work required to lift the array of n particles is identical to the work required
to lift a single particle, whose mass, M , is the total mass of the array, from height z̄,
the centre of mass of the array, to height Z.
Example 2.3.2

Example 2.3.3 Example 2.3.2, continued.

Imagine, as in Example 2.3.2, that there are a finite number of masses m1 , · · · , mn

arrayed along a (vertical) z-axis with mass number i attached at height zi . Again, the
total mass and centre of mass of the array are
n Pn n
X
i=1 mi zi 1 X
M= mi z̄ = Pn = mi zi
i=1 i=1 mi M i=1

Now suppose that we lift, for each 1 ≤ i ≤ n, mass number i, against gravity, from
its initial height zi to a final height Zi . So after the lift we have a new array of masses
with total mass and centre of mass
n Pn n
X mi Zi 1 X
M= mi Z̄ = Pi=1 n = mi Zi
i=1 i=1 mi M i=1

To lift the ith mass took work mi g(Zi − zi ). So the total work required to lift all n
masses was
n
X
Work = mi g(Zi − zi )
i=1
Xn n
X
=g mi Zi − g mi zi
i=1 i=1
= gM Z̄ − gM z̄ = M g(Z̄ − z̄)

321
A PPLICATIONS OF I NTEGRATION 2.3 C ENTRE OF M ASS AND T ORQUE

So the work required to lift the array of n particles is identical to the work required
to lift a single particle, whose mass, M , is the total mass of the array, from height z̄,
the initial centre of mass of the array, to height Z̄, the final centre of mass of the array.
Example 2.3.3

Now we’ll extend the above ideas to cover more general classes of bodies. If the
body consists of mass distributed continuously along a straight line, say with mass
density ρ(x)kg/m and with x running from a to b, rather than consisting of a finite
number of point masses, the formula for the center of mass becomes

Equation 2.3.4 Centre of mass (continuous mass).

Rb
x ρ(x) dx
x̄ = Ra b
a
ρ(x) dx

Think of ρ(x) dx as the mass of the “almost point particle” between x and x + dx.
If the body is a two dimensional object, like a metal plate, lying in the xy-plane,
its center of mass is a point (x̄, ȳ) with x̄ being the (weighted) average value of
the x-coordinate over the body and ȳ being the (weighted) average value of the y-
coordinate over the body. To be concrete, suppose the body fills the region

(x, y) a ≤ x ≤ b, B(x) ≤ y ≤ T (x)

in the xy-plane. For simplicity, we will assume that the density of the body is a
constant, say ρ. When the density is constant, the center of mass is also called the
centroid and is thought of as the geometric center of the body.
To find the centroid of the body, we use the use our standard “slicing” strategy.
We slice the body into thin vertical strips, as illustrated in the figure below.

Here is a detailed description of a generic strip.

322
A PPLICATIONS OF I NTEGRATION 2.3 C ENTRE OF M ASS AND T ORQUE

• The strip has width dx.

• Each point of the strip has essentially the same x-coordinate. Call it x.

• The top of the strip is at y = T (x) and the bottom of the strip is at y = B(x).

• So the strip has

◦ height T (x) − B(x)

◦ area [T (x) − B(x)] dx
◦ mass ρ[T (x) − B(x)] dx
B(x)+T (x)

◦ centroid, i.e. middle point, x , 2
.

In computing the centroid of the entire body, we may treat each strip as a single
(x)
particle of mass ρ[T (x) − B(x)] dx located at x , B(x)+T


2
. So:

Equation 2.3.5 Centroid of object with constant density.

The mass of the entire body bounded by curves T (x) above and B(x) below is
Z b
M =ρ [T (x) − B(x)] dx = ρA (a)
a
Rb
where A = a [T (x) − B(x)] dx is the area of the region. The coordinates of the
centroid are
mass of slice
Rb z }| { Rb
x ρ[T (x) − B(x)] dx x[T (x) − B(x)] dx
x̄ = a = a
(b)
M A
average y on slice
}| { z mass}|
R b zB(x)+T
of slice
{ Rb
(x)
2
ρ[T (x) − B(x)] dx [T (x)2 − B(x)2 ] dx
ȳ = a = a
(c)
M 2A

We can of course also slice up the body using horizontal slices.

323
A PPLICATIONS OF I NTEGRATION 2.3 C ENTRE OF M ASS AND T ORQUE

If the body has constant density ρ and fills the region


(x, y) L(y) ≤ x ≤ R(y), c ≤ y ≤ d

then the same computation as above gives:

Equation 2.3.6 Centroid of object with constant density.

The mass of the entire body bounded by curves L(y) to the left and R(y) to the
right is
Z d
M =ρ [R(y) − L(y)] dy = ρA (a)
c
Rd
where A = c [R(y)−L(y)] dy is the area of the region, and gives the coordinates
of the centroid to be
average x on slice
}| { z mass}|
R d zR(y)+L(y)
of slice
{ Rd
2
ρ[R(y) − L(y)] dy [R(y)2 − L(y)2 ] dy
x̄ = c = c
(b)
M 2A
mass of slice
Rd z }| { Rd
y ρ[R(y) − L(y)] dy y[R(y) − L(y)] dy
ȳ = c = c
(c)
M A

Example 2.3.7 Centroid of a quarter ellipse.

Find the x-coordinate of the centroid (centre of gravity) of the plane region R that
lies in the first quadrant x ≥ 0, y ≥ 0 and inside the ellipse 4x2 + 9y 2 = 36. (The area
2 2
bounded by the ellipse xa2 + yb2 = 1 is πab square units.)

324
A PPLICATIONS OF I NTEGRATION 2.3 C ENTRE OF M ASS AND T ORQUE

2 2
Solution: In standard form 4x2 + 9y 2 = 36 is x9 + y4 = 1. So, on R, x runs from 0 to 3
and R has area A = 14 π × 3 × 2 = 32 π. For each fixed x, between 0 and 3, y runs from 0
q q
x2 2
to 2 1 − 9 . So, applying (2.3.5.b) with a = 0, b = 3, T (x) = 2 1 − x9 and B(x) = 0,
r r
3 3 3
1 1 x2 4 x2
Z Z Z
x̄ = x T (x) dx = x2 1 − dx = x 1− dx
A 0 A 0 9 3π 0 9
x2
Sub in u = 1 − 9
, du = − 29 x dx.
0
9 4 √ 9 4 h u3/2 i0 9 4 h 2i 4
Z
x̄ = − u du = − =− − =
2 3π 1 2 3π 3/2 1 2 3π 3 π
Example 2.3.7

Example 2.3.8 Centroid of a quarter disk.

Find the centroid of the quarter circular disk x ≥ 0, y ≥ 0, x2 + y 2 ≤ r2 .

Solution: By symmetry, x̄ = ȳ. The area of the quarter disk is A = 14 πr2 . By (2.3.5.b)
√
with a = 0, b = r, T (x) = r2 − x2 and B(x) = 0,

1 r √ 2
Z
x̄ = x r − x2 dx
A 0

To evaluate the integral, sub in u = r2 − x2 , du = −2x dx.

Z r √ Z 0 √ du 1 h u3/2 i0 r3
x r2 − x2 dx = u =− = (∗)
0 r2 −2 2 3/2 r2 3

325
A PPLICATIONS OF I NTEGRATION 2.3 C ENTRE OF M ASS AND T ORQUE

So
4 h r3 i 4r
x̄ = =
πr2 3 3π
As we observed above, we should have x̄ = ȳ. But, just for practice, let’s
√ compute
ȳ by the integral formula (2.3.5.c), again with a = 0, b = r, T (x) = r2 − x2 and
B(x) = 0,
Z r √ Z r
1
2 2
r2 − x2 dx


ȳ = r2 − x2 dx = 2
2A 0 πr 0
3 ir
2 h
2 x 2 2r3
= 2 r x− = 2
πr 3 0 πr 3
4r
=
3π
as expected.
Example 2.3.8

Example 2.3.9 Centroid of a half disc.

Find the centroid of the half circular disk y ≥ 0, x2 + y 2 ≤ r2 .

Solution: Once again, we have a symmetry —- namely the half disk is symmetric
about the y-axis. So the centroid lies on the y-axis and x̄√ = 0. The area of the half
disk is A = 12 πr2 . By (2.3.5.c), with a = −r, b = r, T (x) = r2 − x2 and B(x) = 0,
Z r √ Z r
1
2 1
r2 − x2 dx


ȳ = 2
r − x dx 2 = 2
2A −r πr −r
Z r
2
r2 − x2 dx


= 2 since the integrand is even
πr 0
2 h x3 i r
= 2 r2 x −
πr 3 0
4r
=
3π

326
A PPLICATIONS OF I NTEGRATION 2.3 C ENTRE OF M ASS AND T ORQUE

Example 2.3.10 Another centroid.

Find the centroid of the region R in the diagram.

Solution: By symmetry, x̄ = ȳ. The region R is a 2 × 2 square with one quarter of

a circle of radius 1 removed and so has area 2 × 2 − 41 π = 16−π . The top of R is
√ 4
2
y = T (x) = 2. The bottom is y = B(x) with B(x) = 1 − x when 0 ≤ x ≤ 1 and
B(x) = 0 when 1 ≤ x ≤ 2. So
√
Z 1 Z 2 
1
ȳ = x̄ = x[2 − 1 − x2 ] dx + x[2 − 0] dx
A 0 1
 Z 1 √ 
4 2 1 2 2 2
= x 0+x 1− x 1 − x dx
16 − π 0

Now we can make use of the starred equation in Example 2.3.8 with r = 1 to obtain

4 h 1i
= 4−
16 − π 3
44
=
48 − 3π

Example 2.3.11 Centroid of a triangle and its medians.

Prove that the centroid of any triangle is located at the point of intersection of the
medians. A median of a triangle is a line segment joining a vertex to the midpoint of
the opposite side.

327
A PPLICATIONS OF I NTEGRATION 2.3 C ENTRE OF M ASS AND T ORQUE

Solution: Choose a coordinate system so that the vertices of the triangle are located
at (a, 0), (0, b) and (c, 0). (In the figure below, a is negative.)

The line joining (a, 0) and (0, b) has equation bx+ay = ab. (Check that (a, 0) and (0, b)
both really are on this line.) The line joining (c, 0) and (0, b) has equation bx + cy =
bc. (Check that (c, 0) and (0, b) both really are on this line.) Hence for each fixed y
between 0 and b, x runs from a − ab y to c − cb y.
We’ll use horizontal strips to compute x̄ and ȳ. We could just apply equation (2.3.6)
with c = 0, d = b, R(y) = cb (b − y) (which is gotten by solving bx + cy = bc for x) and
L(y) = ab (b − y) (which is gotten by solving bx + ay = ab for x).
But rather than memorizing or looking up those formulae, we’ll derive them for this
example. So consider a thin strip at height y as illustrated in the figure above.

• The strip has length

hc a i c−a
`(y) = (b − y) − (b − y) = (b − y)
b b b

• The strip has width dy.

• On this strip, y has average value y.

• On this strip, x has average value 12 ab (b − y) + cb (b − y) = a+c

 
2b
(b − y).

As the area of the triangle is A = 12 (c − a)b,

b b
1 2 c−a
Z Z
ȳ = y `(y)dy = y (b − y)dy
A 0 (c − a)b 0 b

328
A PPLICATIONS OF I NTEGRATION 2.3 C ENTRE OF M ASS AND T ORQUE

2 b 2  b2 b3 
Z
2
= (by − y )dy = b −
b2 0 b2 2 3
3
2b b
= 2
=
b 6 3
Z b
1 a+c
x̄ = (b − y) `(y)dy
A 0 2b
Z b
2 a+c c−a
= (b − y) (b − y)dy
(c − a)b 0 2b b
a+c b
Z
= 3
(y − b)2 dy
b 0
a+c 1 h
3
ib a + c b3 a+c
= 3
(y − b) = 3
=
b 3 0 b 3 3
a+c b


We have found that the centroid of the triangle is at (x̄, ȳ) = 3
,3 . We shall now
show that this point lies on all three medians.

• One vertex is at (a, 0). The opposite side runs from (0, b) and (c, 0) and so has
b/2
midpoint 21 (c, b). The line from (a, 0) to 21 (c, b) has slope c/2−a b
= c−2a and so has
b b b a+c 1 b


equation y = c−2a (x − a). As c−2a (x̄ − a) = c−2a 3 − a = 3 c−2a (c + a − 3a) =
b
3
= ȳ, the centroid does indeed lie on this median. In this computation we
have implicitly assumed that c 6= 2a so that the denominator c − 2a 6= 0. In
b


the event that c = 2a, the median runs from (a, 0) to a, 2 and so has equation
x = a. When c = 2a we also have x̄ = a+c 3
= a, so that the centroid still lies on
the median.

• Another vertex is at (c, 0). The opposite side runs from (a, 0) and (0, b) and so
b/2
has midpoint 21 (a, b). The line from (c, 0) to 21 (a, b) has slope a/2−c b
= a−2c and so
b b b a+c 1 b


has equation y = a−2c (x−c). As a−2c (x̄−c) = a−2c 3 −c = 3 a−2c (a+c−3c) =
b
3
= ȳ, the centroid does indeed lie on this median. In this computation we have
implicitly assumed that a 6= 2c so that the denominator a − 2c 6= 0. In the event
b


that a = 2c, the median runs from (c, 0) to c, 2 and so has equation x = c.
When a = 2c we also have x̄ = a+c 3
= c, so that the centroid still lies on the
median.

• The third vertex is at
(0, b). The opposite side runs from
(a, 0) and −b (c, 0) and so
a+c a+c 2b
has midpoint 2 , 0 . The line from (0, b) to 2 , 0 has slope (a+c)/2 = − a+c
2b 2b 2b a+c
and so has equation y = b − a+c x. As b − a+c x̄ = b − a+c 3
= 3b = ȳ, the centroid
does indeed lie on this median. This time, we have implicitly assumed that
a + c 6= 0. In the event that a + c = 0, the median runs from (0, b) to (0, 0) and
so has equation x = 0. When a + c = 0 we also have x̄ = a+c 3
= 0, so that the
centroid still lies on the median.
Example 2.3.11

329
 A PPLICATIONS OF I NTEGRATION 2.3 C ENTRE OF M ASS AND T ORQUE

2.3.2 tt Optional — Torque

Newton’s law of motion says that the position x(t) of a single particle moving under
the influence of a force F obeys mx00 (t) = F . Similarly, the positions xi (t), 1 ≤ i ≤ n,
of a set of particles moving under the influence of forces Fi obey mx00i (t) = Fi , 1 ≤ i ≤
n. Often systems of interest consist of some small number of rigid bodies. Suppose
that we are interested in the motion of a single rigid body, say a piece of wood. The
piece of wood is made up of a huge number of atoms. So the system of equations
determining the motion of all of the individual atoms in the piece of wood is huge.
On the other hand, because the piece of wood is rigid, its configuration is completely
determined by the position of, for example, its centre of mass and its orientation.
(Rather than get into what is precisely meant by “orientation”, let’s just say that
it is certainly determined by, for example, the positions of a few of the corners of
the piece of wood). It is possible to extract from the huge system of equations that
determine the motion of all of the individual atoms, a small system of equations that
determine the motion of the centre of mass and the orientation. We can avoid some
vector analysis, that is beyond the scope of this course, by assuming that our rigid
body is moving in two rather than three dimensions.
So, imagine a piece of wood moving in the xy-plane.

Furthermore, imagine that the piece of wood consists of a huge number of par-
ticles joined by a huge number of weightless but very strong steel rods. The steel
rod joining particle number one to particle number two just represents a force acting
between particles number one and two. Suppose that
• there are n particles, with particle number i having mass mi

• at time t, particle number i has x-coordinate xi (t) and y-coordinate yi (t)

• at time t, the external force (gravity and the like) acting on particle number
i has x-coordinate Hi (t) and y-coordinate Vi (t). Here H stands for horizontal
and V stands for vertical.

• at time t, the force acting on particle number i, due to the steel rod joining
particle number i to particle number j has x-coordinate Hi,j (t) and y-coordinate
Vi,j (t). If there is no steel rod joining particles number i and j, just set Hi,j (t) =
Vi,j (t) = 0. In particular, Hi,i (t) = Vi,i (t) = 0.
The only assumptions that we shall make about the steel rod forces are

330
A PPLICATIONS OF I NTEGRATION 2.3 C ENTRE OF M ASS AND T ORQUE

(A1) for each i 6= j, Hi,j (t) = −Hj,i (t) and Vi,j (t) = −Vj,i (t). In words, the steel rod
joining particles i and j applies equal and opposite forces to particles i and j.
 
(A2) for each i 6= j, there is
 a function M
 i,j (t) such that Hi,j (t) = Mi,j (t) x i (t) − x j (t)
and Vi,j (t) = Mi,j (t) yi (t) − yj (t) . In words, the force due to the rod joining
particles i and j acts parallel to the line joining particles i and j. For (A1) to be
true, we need Mi,j (t) = Mj,i (t).
Newton’s law of motion, applied to particle number i, now tells us that
n
X
mi x00i (t) = Hi (t) + Hi,j (t) (Xi )
j=1
n
X
mi yi00 (t) = Vi (t) + Vi,j (t) (Yi )
j=1

Adding up all of the equations (Xi ), for i = 1, 2, 3, · · · , n and adding up all of the
equations (Yi ), for i = 1, 2, 3, · · · , n gives
n
X n
X X
mi x00i (t) = Hi (t) + Hi,j (t) (Σi Xi )
i=1 i=1 1≤i,j≤n
n
X n
X X
mi yi00 (t) = Vi (t) + Vi,j (t) (Σi Yi )
i=1 i=1 1≤i,j≤n
P
The sum 1≤i,j≤n Hi,j (t) contains H1,2 (t) exactly once and it also contains H2,1 (t)
exactly once
P and these two terms cancel exactly, by assumption (A1). In this way, all
terms in 1≤i,j≤n P Hi,j (t) with i 6= j exactly cancel.
P All terms with i = j are assumed
to be zero. So 1≤i,j≤n Hi,j (t) = 0. Similarly, 1≤i,j≤n Vi,j (t) = 0, so the equations
(Σi Xi ) and (Σi Yi ) simplify to
n
X n
X
mi x00i (t) = Hi (t) (Σi Xi )
i=1 i=1
n
X n
X
mi yi00 (t) = Vi (t) (Σi Yi )
i=1 i=1

Denote by
n
X
M= mi
i=1
the total mass of the system, by
n n
1 X 1 X
X(t) = mi xi (t) and Y (t) = mi yi (t)
M i=1 M i=1

the x- and y-coordinates of the centre of mass of the system at time t and by
n
X n
X
H(t) = Hi (t) and V (t) = Vi (t)
i=1 i=1

331
A PPLICATIONS OF I NTEGRATION 2.3 C ENTRE OF M ASS AND T ORQUE

the x- and y-coordinates of the total external force acting on the system at time t. In
this notation, the equations (Σi Xi ) and (Σi Yi ) are

Equation 2.3.12 Rectilinear motion of centre of mass.

M X 00 (t) = H(t) M Y 00 (t) = V (t)

So the centre of mass of the system moves just like a single particle of mass M
subject to the total external force.
Now multiply equation (Yi ) by xi (t), subtract
P from
 it equation (Xi ) multiplied by
yi (t), and sum over i. This gives the equation i xi (t) (Yi ) − yi (t) (Xi ) :
n
X n
X
mi xi (t)yi00 (t) yi (t)x00i (t)
   
− = xi (t)Vi (t) − yi (t)Hi (t)
i=1 i=1
X  
+ xi (t)Vi,j (t) − yi (t)Hi,j (t)
1≤i,j≤n

By the assumption (A2)

   
x1 (t)V1,2 (t) − y1 (t)H1,2 (t) = x1 (t)M1,2 (t) y1 (t) − y2 (t) − y1 (t)M1,2 (t) x1 (t) − x2 (t)
 
= M1,2 (t) y1 (t)x2 (t) − x1 (t)y2 (t)
   
x2 (t)V2,1 (t) − y2 (t)H2,1 (t) = x2 (t)M2,1 (t) y2 (t) − y1 (t) − y2 (t)M2,1 (t) x2 (t) − x1 (t)
 
= M2,1 (t) − y1 (t)x2 (t) + x1 (t)y2 (t)
 
= M1,2 (t) − y1 (t)x2 (t) + x1 (t)y2 (t)
P  
So the i = 1, j = 2 term in 1≤i,j≤n xi (t)Vi,j (t)−yi (t)Hi,j (t) exactly cancels the i = 2,
P  
j = 1 term. In this way all of the terms in 1≤i,j≤n xi (t)Vi,j (t)−yi (t)Hi,j (t) with i 6= j
P  
cancel. Each term with i = j is exactly zero. So 1≤i,j≤n xi (t)Vi,j (t)−yi (t)Hi,j (t) = 0
and n n
X  X
00 00
  
mi xi (t)yi (t) − yi (t)xi (t) = xi (t)Vi (t) − yi (t)Hi (t)
i=1 i=1

Define
n
X
mi xi (t)yi0 (t) − yi (t)x0i (t)
 
L(t) =
i=1
Xn
 
T (t) = xi (t)Vi (t) − yi (t)Hi (t)
i=1

In this notation

332
A PPLICATIONS OF I NTEGRATION 2.3 C ENTRE OF M ASS AND T ORQUE

Equation 2.3.13 Rotational motion of centre of mass.

d
L(t) = T (t)
dt

• Equation (2.3.13) plays the role of Newton’s law of motion for rotational mo-
tion.

• T (t) is called the torque and plays the role of “rotational force”.

• L(t) is called the angular momentum (about the origin) and is a measure of the
rate at which the piece of wood is rotating.

◦ For example, if a particle of mass m is traveling in a circle of radius r,

centred on the origin, at ω radians per unit time, then x(t) = r cos(ωt),
y(t) = r sin(ωt) and

m x(t)y 0 (t) − y(t)x0 (t) = m r cos(ωt) rω cos(ωt) − r sin(ωt) − rω sin(ωt)

  


= mr2 ω

is proportional to ω, which is the rate of rotation about the origin.

In any event, in order for the piece of wood to remain stationary, that is to have xi (t)
and yi (t) be constant for all 1 ≤ i ≤ n, we need to have

X 00 (y) = Y 00 (t) = L(t) = 0

and then equations (2.3.12) and (2.3.13) force

H(t) = V (t) = T (t) = 0

Now suppose that the piece of wood is a seesaw that is long and thin and is lying
on the x-axis, supported
P on a fulcrum at x = p. Then every yi = 0 and the torque
simplifies to T (t) = ni=1 xi (t)Vi (t). The forces consist of

333
 A PPLICATIONS OF I NTEGRATION 2.3 C ENTRE OF M ASS AND T ORQUE

• gravity, mi g, acting downwards on particle number i, for each 1 ≤ i ≤ n and

the

• force F imposed by the fulcrum that is pushing straight up on the particle at

x = p.

So
n
P
• The net vertical force is V (t) = F − mi g = F − M g. If the seesaw is to remain
i=1
stationary, this must be zero so that F = M g.

• The total torque (about the origin) is

n
X n
X
T = Fp − mi gxi = M gp − mi gxi
i=1 i=1

If the seesaw is to remain stationary, this must also be zero and the fulcrum
must be placed at

Equation 2.3.14 Placement of fulcrum.

n
1 X
p= m i xi
M i=1

which is the centre of mass of the piece of wood.

2.3.3 tt Exercises

Exercises — Stage 1 In Questions 8 through 10, you will derive the formulas for the
centre of mass of a rod of variable density, and the centroid of a two-dimensional
region using vertical slices (Equations 2.3.4 and 2.3.5 in the text). Knowing the equa-
tions by heart will allow you to answer many questions in this section; understand-
ing where they came from will you allow to generalize their ideas to answer even
more questions.

334
A PPLICATIONS OF I NTEGRATION 2.3 C ENTRE OF M ASS AND T ORQUE

1. Using symmetry, find the centroid of the finite region between the
curves y = (x − 1)2 and y = −x2 + 2x + 1.

y = (x − 1)2

x
y = −x2 + 2x + 1

2. Using symmetry, find the centroid of the region inside the unit circle, and
outside a rectangle centred at the origin with width 1 and height 0.5.
y

3. A long, straight, thin rod has a number of weights attached along it.
True or false: if it balances at position x, then the mass to the right of x
is the same as the mass to the left of x.

4. A straight rod with negligible mass has the following weights attached to
it:
• A weight of mass 1 kg, 1m from the left end,

• a weight of mass 2 kg, 3m from the left end,

• a weight of mass 2 kg, 4m from the left end, and

• a weight of mass 1 kg, 6m from the left end.

Where is the centre of mass of the weighted rod?

335
A PPLICATIONS OF I NTEGRATION 2.3 C ENTRE OF M ASS AND T ORQUE

5. For each picture below, determine whether the centroid is to the left of, to
the right of, or along the line x = a, or whether there is not enough infor-
mation to tell. The shading of a region indicates density: darker shading
corresponds to a denser area.

y y y

x x x
a a a

(a) (b) (c)

y B y
(d) (e)
A x
x
a a
A A/2
A
2B
B B

6. Tank A is spherical, of radius 1 metre, and filled completely with water. The
bottom of tank A is three metres above the ground, where Tank B sits. Tank
B is tall and rectangular, with base dimensions 1 metre by 12 metres, and
empty. Calculate the work done by gravity to drain all the water from Tank
A to Tank B by modelling the situation as a point mass, of the same mass
as the water, being moved from the height of the centre of mass of A to the
height of the centre of mass of B.

A
1m

3m
B

m
2m 0.5

You may use 1000 kg/m3 for the density of water, and g = 9.8 m/sec2 for
the acceleration due to gravity.

336
A PPLICATIONS OF I NTEGRATION 2.3 C ENTRE OF M ASS AND T ORQUE

7. Let S be the region bounded above by y = x1 and and below by the x-axis,
1 ≤ x ≤ 3. Let R be a rod with density ρ(x) = x1 at position x, 1 ≤ x ≤ 3.

a What is the area of a thin slice of S at position x with width dx?

b What is the mass of a small piece of R at position x with length dx?

c What is the total area of S?

d What is the total mass of R?

e What is the x-coordinate of the centroid of S?

f What is the centre of mass of R?

8. Suppose R is a straight, thin rod with density ρ(x) at a position x. Let

the left endpoint of R lie at x = a, and the right endpoint lie at x = b.

a To approximate the centre of mass of R, imagine chopping it into n

pieces of equal length, and approximating the mass of each piece
using the density at its midpoint. Give your approximation for the
centre of mass in sigma notation.

a m1 m2 mn b

b Take the limit as n goes to infinity of your approximation in part

(a), and express the result using a definite integral.

9. Suppose S is a two-dimensional object and at (horizontal) position x its

height is T (x)−B(x). Its leftmost point is at position x = a, and its rightmost
point is at position x = b.
To approximate the x-coordinate of the centroid of S, we imagine it as a
straight, thin rod R, where the mass of R from a ≤ x ≤ b is equal to the area
of S from a ≤ x ≤ b.

a If S is the sheet shown below, sketch R as a rod with the same hori-
zontal length, shaded darker when R is denser, and lighter when R is
less dense.

337
A PPLICATIONS OF I NTEGRATION 2.3 C ENTRE OF M ASS AND T ORQUE

y T (x)

B(x)
x
a b

b If we cut S into strips of very small width dx, what is the area of the
strip at position x?

c Using your answer from (b), what is the density ρ(x) of R at position
x?

d Using your result from Question 8(b), give the x-coordinate of the cen-
troid of S. Your answer will be in terms of a, b, T (X), and B(x).
10. Suppose S is flat sheet with uniform density, and at (horizontal) position
x its height is T (x) − B(x). Its leftmost point is at position x = a, and its
rightmost point is at position x = b.
To approximate the y-coordinate of the centroid of S, we imagine it as a
straight, thin, vertical rod R. We slice S into thin, vertical strips, and model
these as weights on R with:
• position y on R, where y is the centre of mass of the strip, and
• mass in R equal to the area of the strip in S.

a If S is the sheet shown below, slice it into a number of vertical pieces

of equal length, approximated by rectangles. For each rectangle, mark
its centre of mass. Sketch R as a rod with the same vertical height, with
weights corresponding to the slices you made of S.

y T (x)

B(x)
x
a a 0
b 0 b

b Imagine a thin strip of S at position x, with thickness dx. What is the

area of the strip? What is the y-value of its centre of mass?

338
A PPLICATIONS OF I NTEGRATION 2.3 C ENTRE OF M ASS AND T ORQUE

c Recall the centre of mass of a rod with n weights of mass Mi at position

yi is given by
Pn
(Mi × yi )
i=1
n
P
Mi
i=1

Considering the limit of this formula as n goes to infinity, give the

y-coordinate of the centre of mass of S.
11. *. Express the x-coordinate of the centroid of the triangle with vertices
(−1, −3), (−1, 3), and (0, 0) in terms of a definite integral. Do not evalu-
ate the integral.

Exercises — Stage 2 Use Equations 2.3.4 and 2.3.5 to find centroids and centres of
mass in Questions 12 through 23.
12. A long, thin rod extends from x = 0 to x = 7 metres, and its density at
position x is given by ρ(x) = x kg/m. Where is the centre of mass of the
rod?

13. A long, thin rod extends from x = −3 to x = 10 metres, and its density
1
at position x is given by ρ(x) = 1+x 2 kg/m. Where is the centre of mass

of the rod?

14. *. Find the y-coordinate of the centroid of the region bounded by the curves
y = 1, y = −ex , x = 0 and x = 1. You may use the fact that the area of this
region equals e.
√ 1
15. *. Consider the region bounded by y = 16−x2
, y = 0, x = 0 and x = 2.

a Sketch this region.

b Find the y-coordinate of the centroid of this region.

16. *. Find the centroid of the finite region bounded by y = sin(x), y = cos(x),
x = 0, and x = π/4.

17. *. Let A denote the area of the plane region bounded by x = 0, x = 1,

k
y = 0 and y = √ , where k is a positive constant.
1 + x2
a Find the coordinates of the centroid of this region in terms of k and
A.

b For what value of k is the centroid on the line y = x?

18. *. The region R is the portion of the plane which is above the curve
y = x2 − 3x and below the curve y = x − x2 .

339
A PPLICATIONS OF I NTEGRATION 2.3 C ENTRE OF M ASS AND T ORQUE

a Sketch the region R

b Find the area of R.

c Find the x coordinate of the centroid of R.

1
19. *. Let R be the region where 0 ≤ x ≤ 1 and 0 ≤ y ≤ 1+x2 . Find the
x-coordinate of the centroid of R.
20. *. Find the centroid of the region below, which consists of a semicircle of
radius 3 on top of a rectangle of width 6 and height 2.

√
21. *. Let D be the region below the graph of the curve y = 9 − 4x2 and
above the x-axis.

a Using an appropriate integral, find the area of the region D; sim-

plify your answer completely.

b Find the centre of mass of the region D; simplify your answer com-
pletely. (Assume it has constant density ρ.)

22. The finite region S is bounded by the lines y = arcsin x, y = arcsin(2 − x),
and y = − π2 . Find the centroid of S.
23. Calculate the centroid of the figure bounded by the curves y = ex , y =
3(x − 1), y = 0, x = 0, and x = 2.

Exercises — Stage 3

24. *. Find the y-coordinate of the centre of mass of the (infinite) region lying
to the right of the line x = 1, above the x-axis, and below the graph of
y = 8/x3 .

340
A PPLICATIONS OF I NTEGRATION 2.3 C ENTRE OF M ASS AND T ORQUE

25. *. Let A be the region to the right of the y-axis that is bounded by the graphs
of y = x2 and y = 6 − x.

a Find the centroid of A, assuming it has constant density ρ = 1. The

22
area of A is (you don’t have to show this).
3
b Write down an expression, using horizontal slices (disks), for the vol-
ume obtained when the region A is rotated around the y-axis. Do not
evaluate any integrals; simply write down an expression for the vol-
ume.
x
26. *. (a) Find the y-coordinate of the centroid of the region bounded by y = e ,
x = 0, x = 1, and y = −1.
(b) Calculate the volume of the solid generated by rotating the region from
part (a) about the line y = −1.

27. Suppose a rectangle has width 4 m, height 3 m, and its density x metres
from its left edge is x2 kg/m2 . Find the centre of mass of the rectangle.
y

x
4

28. Suppose a circle of radius 3 m has density (2 + y) kg/m2 at any point y

metres above its bottom. Find the centre of mass of the circle.
y

x
−3 3

29. A right circular cone of uniform density has base radius r m and height
h m. We want to find its centre of mass. By symmetry, we know that
the centre of mass will occur somewhere along the straight vertical line
through the tip of the cone and the centre of its base. The only question
is the height of the centre of mass.

341
A PPLICATIONS OF I NTEGRATION 2.3 C ENTRE OF M ASS AND T ORQUE

x
r

We will model the cone as a rod R with height h, such that the mass
of the section of the rod from position a to position b is the same as the
volume of the cone from height a to height b. (You can imagine that the
cone is an umbrella, and we’ve closed it up to look like a cane. a )

h h

x 0
r

a Using this model, calculate how high above the base of the cone
its centre of mass is.
b If we cut off the top h − k metres of the cone (leaving an object of
height k), how high above the base is the new centre of mass?

a This analogy isn’t exact: if the cone were an umbrella, closing it would move
the outside fabric vertically. A more accurate, but less familiar, image might be
vacuum-wrapping an umbrella, watching it shrivel towards the middle but not
move vertically.

342
A PPLICATIONS OF I NTEGRATION 2.3 C ENTRE OF M ASS AND T ORQUE

30. An hourglass is shaped like two identical truncated cones attached together.
Their base radius is 5 cm, the height of the entire hourglass is 18 cm, and
the radius at the thinnest point is .5 cm. The hourglass contains sand that
fills up the bottom 6 cm when it’s settled, with mass 600 grams and uniform
density. We want to know the work done flipping the hourglass smoothly,
so the sand settles into a truncated, inverted-cone shape before it starts to
fall down.
5 cm

8.8 cm

0.5
18 cm

6 cm

Using the methods of Section 2.1 to calculate the work done would be quite
tedious. Instead, we will model the sand as a point of mass 0.6 kg, being
lifted from the centre of mass of its original position to the centre of mass
of its upturned position. Using the results of Question 29, how much work
was done on the sand?
To simplify your calculation, you may assume that the height of the up-
turned sand (that is, the distance from the skinniest
√ part of the hourglass to
the top of the sand) is 8.8 cm. (Actually, it’s 3 937 − 1 ≈ 8.7854 cm.) So, the
top 0.2 cm of the hourglass is empty.

31. Tank A is in the shape of half a sphere of radius 1 metre, with its flat
face resting on the ground, and is completely filled with water. Tank B
is empty and rectangular, with a square base of side length 1 m and a
height of 3 m.

343
A PPLICATIONS OF I NTEGRATION 2.3 C ENTRE OF M ASS AND T ORQUE

3m

1m 1m

a To pump the water from Tank A to Tank B, we need to pump all

the water from Tank A to a height of 3 m. How much work is done
to pump all the water from Tank A to a height of 3 m? You may
model the water as a point mass, originally situated at the centre
of mass of the full Tank A.

b Suppose we could move the water from Tank A directly to its final
position in Tank B without going over the top of Tank B. (For
example, maybe tank A is elastic, and Tank B is just Tank A after
being smooshed into a different form.) How much work is done
pumping the water? (That is, how much work is done moving a
point mass from the centre of mass of Tank A to the centre of mass
of Tank B?)

c What percentage of work from part (a) was “wasted” by pumping

the water over the top of Tank B, instead of moving it directly to
its final position?

You may assume that the only work done is against the acceleration due
to gravity, g = 9.8 m/sec2 , and that the density of water is 1000 kg/m3 .
Remark: the answer from (b) is what you might think of as the net work
involved in pumping the water from Tank A to Tank B. When work gets
“wasted,” the pump does some work pumping water up, then gravity
does equal and opposite work bringing the water back down.

2
32. Let R be the region
p π bounded above by y = 2x sin(x ) and below by the x-
axis, 0 ≤ x ≤ 2 . Give an approximation of the x-value of the centroid of
1
R with error no more than 100 .
d4
You may assume without proof that dx4
{2x2 sin(x2 )} ≤ 415 over the inter-
val 0, π2 .
 p 

344
 A PPLICATIONS OF I NTEGRATION 2.4 S EPARABLE D IFFERENTIAL E QUATIONS

2.4q Separable Differential Equations

A differential equation is an equation for an unknown function that involves the

derivative of the unknown function. Differential equations play a central role in
modelling a huge number of different phenomena. Here is a table giving a bunch of
named differential equations and what they are used for. It is far from complete.

Newton’s Law of Motion describes motion of particles

Maxwell’s equations describes electromagnetic radiation
Navier-Stokes equations describes fluid motion
Heat equation describes heat flow
Wave equation describes wave motion
Schrödinger equation describes atoms, molecules and crystals
Stress-strain equations describes elastic materials
Black-Scholes models used for pricing financial options
Predator-prey equations describes ecosystem populations
Einstein’s equations connects gravity and geometry
Ludwig-Jones-Holling’s equation models spruce budworm/Balsam fir ecosystem
Zeeman’s model models heart beats and nerve impulses
Sherman-Rinzel-Keizer model for electrical activity in Pancreatic β-cells
Hodgkin-Huxley equations models nerve action potentials

We are just going to scratch the surface of the study of differential equations.
Most universities offer half a dozen different undergraduate courses on various as-
pects of differential equations. We will just look at one special, but important, type
of equation.

2.4.1 tt Separate and integrate

Definition 2.4.1
A separable differential equation is an equation for a function y(x) of the form

dy

(x) = f (x) g y(x)
dx

We’ll start by developing a recipe for solving separable differential equations.

Then we’ll look at many examples. Usually one suppresses the argument of y(x)

345
A PPLICATIONS OF I NTEGRATION 2.4 S EPARABLE D IFFERENTIAL E QUATIONS

and writes the equation 1

dy
= f (x) g(y)
dx
and solves such an equation by cross multiplying/dividing to get all of the y’s, in-
cluding the dy on one side of the equation and all of the x’s, including the dx, on the
other side of the equation.
dy
= f (x) dx
g(y)
(We are of course assuming that g(y) is nonzero.) Then you integrate both sides
dy
Z Z
= f (x) dx (?)
g(y)
dy
This looks illegal, and indeed is illegal — dx is not a fraction. But we’ll now see that
the answer is still correct. This procedure is simply a mnemonic device to help you
remember that answer (?).
dy


• Our goal is to find all functions y(x) that obey dx (x) = f (x) g y(x) .
• Assuming that g is nonzero,
y 0 (x) y 0 (x)
Z Z
0
y (x) = f (x) g(y(x)) ⇐⇒ = f (x) ⇐⇒ dx = f (x) dx
g(y(x)) g(y(x))
dy
Z Z
⇐⇒ = f (x) dx
g(y) y=y(x)
with the substitution y = y(x), dy = y 0 (x) dx

• That’s our answer (?) again.

1 1
Let G(y) be an antiderivative of g(y) (i.e. G0 (y) = g(y) ) and F (x) be an antideriva-
0
tive of f (x) (i.e. F (x) = f (x)). If we reinstate the argument of y, (?) is


G y(x) = F (x) + C (2.4.1)
Observe that the solution equation (2.4.1) contains an arbitrary constant, C. The
value of this arbitrary constant can not be determined by the differential equation.
You need additional data to determine it. Often this data consists of the value of the
unknown function for one value of x. That is, often the problem you have to solve is
of the form
dy

(x) = f (x) g y(x) y(x0 ) = y0
dx
where f (x) and g(y) are given functions and x0 and y0 are given numbers. This type
of problem is called an “initial value problem”. It is solved by first using the method
above to find the general solution to the differential equation, including the arbitrary
constant C, and then using the “initial condition” y(x0 ) = y0 to determine the value
of C. We’ll see examples of this shortly.

1 Look at the right hand side of the equation. The x-dependence is separated from the y-
dependence. That’s the reason for the name “separable”.

346
A PPLICATIONS OF I NTEGRATION 2.4 S EPARABLE D IFFERENTIAL E QUATIONS

Example 2.4.2 A separable warm-up.

The differential equation

dy
= xe−y
dx
is separable, and we now find all of its solutions by using our mnemonic device. We
start by cross-multiplying so as to move all y’s to the left hand side and all x’s to the
right hand side.
ey dy = x dx
Then we integrate both sides

x2
Z Z
e dy = xdx ⇐⇒ ey =
y
+C
2
The C onR the right hand side contains both the arbitrary constant forR the indefinite
integral ey dy and the arbitrary constant for the indefinite integral xdx. Finally,
we solve for y, which is really a function of x.
 x2 
y(x) = log +C
2
Recall that we are using log to refer to the natural (base e) logarithm.
Note that C is an arbitrary constant. It can take any value. It cannot be determined
by the differential equation itself. In applications C is usually determined by a re-
quirement that y take some prescribed value (determined by the application) when
x is some prescribed value. For example, suppose that we wish to find a function
y(x) that obeys both
dy
= xe−y and y(0) = 1
dx
dy 2
= xe−y satisfied, we must have y(x) = log x2 + C , for


We know that, to have dx
some constant C. To also have y(0) = 1, we must have
 x2 
1 = y(0) = log +C = log C ⇐⇒ log C = 1 ⇐⇒ C = e
2 x=0

x2


So our final solution is y(x) = log 2
+e .

Example 2.4.3 A little more warm-up.

Let a and b be any two constants. We’ll now solve the family of differential equations

dy
= a(y − b)
dx

347
A PPLICATIONS OF I NTEGRATION 2.4 S EPARABLE D IFFERENTIAL E QUATIONS

using our mnemonic device.

dy dy
Z Z
= a dx =⇒ = a dx
y−b y−b
=⇒ log |y − b| = ax + c =⇒ |y − b| = eax+c = ec eax
=⇒ y − b = Ceax

where C is either +ec or −ec . Note that as c runs over all real numbers, +ec runs over
all strictly positive real numbers and −ec runs over all strictly negative real numbers.
So, so far, C can be any real number except 0. But we were a bit sloppy here. We
implicitly assumed that y − b was nonzero, so that we could divide it across. None-
the-less, the constant function y = b, which corresponds to C = 0, is a perfectly good
dy
solution — when y is the constant function y = b, both dx and a(y −b) are zero. So the
dy ax
general solution to dx = a(y − b) is y(x) = Ce + b, where the constant C can be any
real number. Note that when y(x) = Ceax + b we have y(0) = C + b. So C = y(0) − b
and the general solution is

y(x) = {y(0) − b} eax + b

Example 2.4.3

This is worth stating as a theorem.

Theorem 2.4.4
Let a and b be constants. The differentiable function y(x) obeys the differential
equation
dy
= a(y − b)
dx
if and only if
y(x) = {y(0) − b} eax + b

dy
Example 2.4.5 Solve dx
= y2.
dy
Solve dx = y2
Solution: When y 6= 0,

dy 2 dy y −1 1
= y =⇒ 2 = dx =⇒ = x + C =⇒ y = −
dx y −1 x+C

When y = 0, this computation breaks down because dy y2

contains a division by 0. We
can check if the function y(x) = 0 satisfies the differential equation by just subbing it

348
A PPLICATIONS OF I NTEGRATION 2.4 S EPARABLE D IFFERENTIAL E QUATIONS

in:
y(x) = 0 =⇒ y 0 (x) = 0, y(x)2 = 0 =⇒ y 0 (x) = y(x)2
So y(x) = 0 is a solution and the full solution is
1
y(x) = 0 or y(x) = − , for any constant C
x+C
Example 2.4.5

Example 2.4.6 A falling raindrop.

When a raindrop falls it increases in size so that its mass m(t), is a function of time t.
The rate of growth of mass, i.e. dm
dt
, is km(t) for some positive constant k. According
d
to Newton’s law of motion, dt (mv) = gm, where v is the velocity of the raindrop
(with v being positive for downward motion) and g is the acceleration due to gravity.
Find the terminal velocity, lim v(t), of a raindrop.
t→∞
Solution: In this problem we have two unknown functions, m(t) and v(t), and two
d
differential equations, dmdt
= km and dt (mv) = gm. The first differential equation,
dm
dt
= km, involves only m(t), not v(t), so we use it to determine m(t). By Theorem
2.4.4, with b = 0, a = k, y replaced by m and x replaced by t,
dm
= km =⇒ m(t) = m(0)ekt
dt
Now that we know m(t) (except for the value of the constant m(0)), we can sub-
stitute it into the second differential equation, which we can then use to deter-
mine the remaining unknown function v(t). Observe that the second equation,
d
dt
(mv) = gm(t) = gm(0)ekt tells that the derivative of the function y(t) = m(t)v(t) is
gm(0)ekt . So y(t) is just an antiderivative of gm(0)ekt .

dy ekt
Z
kt
= gm(t) = gm(0)e =⇒ y(t) = gm(0)ekt dt = gm(0) +C
dt k
Now that we know y(t) = m(t)v(t) = m(0)ekt v(t), we can get v(t) just by dividing
out the m(0)ekt .

ekt kt ekt
y(t) = gm(0) + C =⇒ m(0)e v(t) = gm(0) +C
k k
g C
=⇒ v(t) = +
k m(0)ekt
Our solution, v(t), contains two arbitrary constants, namely C and m(0). They will
be determined by, for example, the mass and velocity at time t = 0. But since we are
only interested in the terminal velocity lim v(t), we don’t need to know C and m(0).
t→∞
Since k > 0, lim C
kt = 0 and the terminal velocity lim v(t) = kg .
t→∞ e t→∞

349
 A PPLICATIONS OF I NTEGRATION 2.4 S EPARABLE D IFFERENTIAL E QUATIONS

Example 2.4.7 Intravenous glucose.

A glucose solution is administered intravenously into the bloodstream at a constant

rate r. As the glucose is added, it is converted into other substances at a rate that is
proportional to the concentration at that time. The concentration, C(t), of the glucose
in the bloodstream at time t obeys the differential equation

dC
= r − kC
dt
where k is a positive constant of proportionality.

a Express C(t) in terms of k and C(0).

b Find lim C(t).

t→∞

r


Solution: (a) Since r − kC = −k C − k
the given equation is

dC r

= −k C −
dt k
which is of the form solved in Theorem 2.4.4 with a = −k and b = kr . So the solution
is
r  r  −kt
C(t) = + C(0) − e
k k
For any k > 0, lim e−kt = 0. Consequently, for any C(0) and any k > 0, lim C(t) = kr ,.
t→∞ t→∞
We could have predicted this limit without solving for C(t). If we assume that C(t)
approaches some equilibrium value Ce as t approaches infinity, then taking the limits
of both sides of dC
dt
= r − kC as t → ∞ gives
r
0 = r − kCe =⇒ Ce =
k

2.4.2 tt Optional — Carbon Dating

Scientists can determine the age of objects containing organic material by a method
called carbon dating or radiocarbon dating 2 . The bombardment of the upper atmo-
sphere by cosmic rays converts nitrogen to a radioactive isotope of carbon, 14 C, with
a half-life of about 5730 years. Vegetation absorbs carbon dioxide from the atmo-
sphere through photosynthesis and animals acquire 14 C by eating plants. When a
plant or animal dies, it stops replacing its carbon and the amount of 14 C begins to de-
crease through radioactive decay. Therefore the level of radioactivity also decreases.

2 Willard Libby, of Chicago University was awarded the Nobel Prize in Chemistry in 1960, for
developing radiocarbon dating.

350
A PPLICATIONS OF I NTEGRATION 2.4 S EPARABLE D IFFERENTIAL E QUATIONS

More precisely, let Q(t) denote the amount of 14 C in the plant or animal t years after
it dies. The number of radioactive decays per unit time, at time t, is proportional to
the amount of 14 C present at time t, which is Q(t). Thus

Equation 2.4.8 Radioactive decay.

dQ
(t) = −kQ(t)
dt

Here k is a constant of proportionality that is determined by the half-life. We shall

explain what half-life is, and also determine the value of k, in Example 2.4.9, below.
Before we do so, let’s think about the sign in equation (2.4.8).
• Recall that Q(t) denotes a quantity, namely the amount of 14 C present at time
t. There cannot be a negative amount of 14 C. Nor can this quantity be zero.
(We would not use carbon dating when there is no 14 C present.) Consequently,
Q(t) > 0.
• As the time t increases, Q(t) decreases, because 14 C is being continuously con-
verted into 14 N by radioactive decay 3 . Thus dQ
dt
(t) < 0.

• The signs Q(t) > 0 and dQdt

(t) < 0 are consistent with (2.4.8) provided the con-
stant of proportionality k > 0.
• In (2.4.8), we chose to call the constant of proportionality “−k”. We did so in
order to make k > 0. We could just as well have chosen to call the constant
of proportionality “K”. That is, we could have replaced equation (2.4.8) by
dQ
dt
(t) = KQ(t). The constant of proportionality K would have to be negative,
(and K and k would be related by K = −k).

Example 2.4.9 Half life and the constant k.

In this example, we determine the value of the constant of proportionality k in (2.4.8)

that corresponds to the half-life of 14 C, which is 5730 years.
• Imagine that some plant or animal contains a quantity Q0 of 14 C at its time of
death. Let’s choose the zero point of time t = 0 to be the instant that the plant
or animal died.
• Denote by Q(t) the amount of 14 C in the plant or animal t years after it died.
Then Q(t) must obey both (2.4.8) and Q(0) = Q0 .
• Theorem 2.4.4, with b = 0 and a = −k, then tells us that Q(t) = Q0 e−kt for all
t ≥ 0.

3 The precise transition is 14

C → 14
N + e− + ν̄e where e− is an electron and ν̄e is an electron
neutrino.

351
A PPLICATIONS OF I NTEGRATION 2.4 S EPARABLE D IFFERENTIAL E QUATIONS

• By definition, the half-life of 14 C is the length of time that it takes for half of the
14
C to decay. That is, the half-life t1/2 is determined by

1 1
Q(t1/2 ) = Q(0) = Q0 but we know that Q(t) = Q0 e−kt
2 2
1
Q0 e−kt1/2 = Q0 now cancel Q0
2
1
e−kt1/2 =
2
Taking the logarithm of both sides gives

1 log 2
−kt1/2 = log = − log 2 =⇒ k =
2 t1/2

Recall that, in this text, we use log x to indicate the natural logarithm. That is,

log x = loge x = log x

We are told that, for 14 C, the half-life t1/2 = 5730, so

log 2
k= = 0.000121 to 6 decimal places
5730
Example 2.4.9

From the work in the above example we have accumulated enough new facts to
make a corollary to Theorem 2.4.4.

Corollary 2.4.10

The function Q(t) satisfies the equation

dQ
= −kQ(t)
dt
if and only if

Q(t) = Q(0) e−kt

The half-life is defined to be the time t1/2 which obeys

1
Q t1/2 = Q(0)
2
The half-life is related to the constant k by
log 2
t1/2 =
k

352
A PPLICATIONS OF I NTEGRATION 2.4 S EPARABLE D IFFERENTIAL E QUATIONS

Now here is a typical problem that is solved using Corollary 2.4.10.

Example 2.4.11 The age of a piece of parchment.

A particular piece of parchment contains about 64% as much 14 C as plants do today.

Estimate the age of the parchment.
Solution: Let Q(t) denote the amount of 14 C in the parchment t years after it was
first created. By equation (2.4.8) and Example 2.4.9

dQ log 2
(t) = −kQ(t) with k = = 0.000121
dt 5730
By Corollary 2.4.10

Q(t) = Q(0) e−kt

The time at which Q(t) reaches 0.64 Q(0) is determined by

Q(t) = 0.64 Q(0) but Q(t) = Q(0) e−kt

Q(0) e−kt = 0.64 Q(0) cancel Q(0)
−kt
e = 0.64 take logarithms
−kt = log 0.64
log 0.64 log 0.64
t= = = 3700 to 2 significant digits
−k −0.000121
That is, the parchment a is about 37 centuries old.

a The British Museum has an Egyptian mathematical text from the seventeenth century B.C.

We have stated that the half-life of 14 C is 5730 years. How can this be determined?
We can explain this using the following example.

Example 2.4.12 Half life of implausium.

A scientist in a B-grade science fiction film is studying a sample of the rare and
fictitious element, implausium. With great effort he has produced a sample of pure
implausium. The next day — 17 hours later — he comes back to his lab and discovers
that his sample is now only 37% pure. What is the half-life of the element?
Solution: We can again set up our problem using Corollary 2.4.10. Let Q(t) denote
the quantity of implausium at time t, measured in hours. Then we know

Q(t) = Q(0) · e−kt

353
 A PPLICATIONS OF I NTEGRATION 2.4 S EPARABLE D IFFERENTIAL E QUATIONS

We also know that

Q(17) = 0.37Q(0).

That enables us to determine k via

Q(17) = 0.37Q(0) = Q(0)e−17k divide both sides by Q(0)

0.37 = e−17k

and so
log 0.37
k=− = 0.05849
17
We can then convert this to the half life using Corollary 2.4.10:

log 2
t1/2 = ≈ 11.85 hours
k
While this example is entirely fictitious, one really can use this approach to measure
the half-life of materials.
Example 2.4.12

2.4.3 tt Optional — Newton’s Law of Cooling

Newton’s law of cooling says:

• The rate of change of temperature of an object is proportional to the difference

in temperature between the object and its surroundings. The temperature of
the surroundings is sometimes called the ambient temperature.

If we denote by T (t) the temperature of the object at time t and by A the temperature
of its surroundings, Newton’s law of cooling says that there is some constant of
proportionality, K, such that

Equation 2.4.13 Newton’s law of cooling.

dT  
(t) = K T (t) − A
dt

This mathematical model of temperature change works well when studying a

small object in a large, fixed temperature, environment. For example, a hot cup of

354
A PPLICATIONS OF I NTEGRATION 2.4 S EPARABLE D IFFERENTIAL E QUATIONS

coffee in a large room 4 . Let’s start by thinking a little about the sign of the constant
of proportionality. At any time t, there are three possibilities.
• If T (t) > A, that is, if the body is warmer than its surroundings, we would ex-
pect heat to flow from the body into its surroundings and so we would expect
the body to cool off so that dT dt
(t) < 0. For this expectation to be consistent with
(2.4.13), we need K < 0.
• If T (t) < A, that is the body is cooler than its surroundings, we would expect
heat to flow from the surroundings into the body and so we would expect the
body to warm up so that dT dt
(t) > 0. For this expectation to be consistent with
(2.4.13), we again need K < 0.
• Finally if T (t) = A, that is the body and its environment have the same temper-
ature, we would not expect any heat to flow between the two and so we would
expect that dTdt
(t) = 0. This does not impose any condition on K.
In conclusion, we would expect K < 0. Of course, we could have chosen to call the
constant of proportionality −k, rather than K. Then the differential equation would
be dT


dt
= −k T − A and we would expect k > 0.

Example 2.4.14 Warming iced tea.

The temperature of a glass of iced tea is initially 5◦ . After 5 minutes, the tea has
heated to 10◦ in a room where the air temperature is 30◦ .
a Determine the temperature as a function of time.
b What is the temperature after 10 minutes?
c Determine when the tea will reach a temperature of 20◦ .
Solution: (a)
• Denote by T (t) the temperature of the tea t minutes after it was removed from
the fridge, and let A = 30 be the ambient temperature.
• By Newton’s law of cooling,
dT
= K(T − A) = K(T − 30)
dt
for some, as yet unknown, constant of proportionality K.
• By Theorem 2.4.4 with a = K and b = 30,
T (t) = [T (0) − 30] eKt + 30 = 30 − 25eKt
since the initial temperature T (0) = 5.

4 It does not work so well when the object is of a similar size to its surroundings since the tempera-
ture of the surroundings will rise as the object cools. It also fails when there are phase transitions
involved — for example, an ice-cube melting in a warm room does not obey Newton’s law of
cooling.

355
A PPLICATIONS OF I NTEGRATION 2.4 S EPARABLE D IFFERENTIAL E QUATIONS

• This solution is not complete because it still contains an unknown constant,

namely K. We have not yet used the given data that T (5) = 10. We can use it
to determine K. At t = 5,
20 20
T (5) = 30 − 25e5K = 10 =⇒ e5K = =⇒ 5K = log
25 25
1 4
=⇒ K = log = −0.044629
5 5
to six decimal places.

(b) To find the temperature at 10 minutes we can just use the solution we have deter-
mined above.

T (10) = 30 − 25e10K
1 4
= 30 − 25e10× 5 log 5
4 16
= 30 − 25e2 log 5 = 30 − 25elog 25
= 30 − 16 = 14◦

(c) The temperature is 20◦ when

10 10
30 − 25eKt = 20 =⇒ eKt = =⇒ Kt = log
25 25
1 2
=⇒ t = log = 20.5 min
K 5
to one decimal place.
Example 2.4.14

Example 2.4.15 Temperature back in time.

A dead body is discovered at 3:45pm in a room where the temperature is 20◦ C. At

that time the temperature of the body 1s 27◦ C. Two hours later, at 5:45pm, the tem-
perature of the body is 25.3◦ C. What was the time of death? Note that the normal
(adult human) body temperature is 37◦ C.
Solution: We will assume that the body’s temperature obeys Newton’s law of cool-
ing.

• Denote by T (t) the temperature of the body at time t, with t = 0 corresponding

to 3:45pm. We wish to find the time of death — call it td .

• There is a lot of data in the statement of the problem. We are told

1 the ambient temperature: A = 20

2 the temperature of the body when discovered: T (0) = 27

356
A PPLICATIONS OF I NTEGRATION 2.4 S EPARABLE D IFFERENTIAL E QUATIONS

3 the temperature of the body 2 hours later: T (2) = 25.3

4 assuming the person was a healthy adult right up until he died, the tem-
perature at the time of death: T (td ) = 37.

• Theorem 2.4.4 with a = K and b = A = 20

T (t) = [T (0) − A] eKt + A = 20 + 7eKt

Two unknowns remain, K and td .

• We can find the first, K, by using the condition (3), which says T (2) = 25.3.

25.3 = T (2) = 20 + 7e2K =⇒ 7e2K = 5.3 =⇒ 2K = log 5.3

7
=⇒ K = 12 log 5.3


7
= −0.139

• Finally, td is determined by the condition (4).

37 = T (td ) = 20 + 7e−0.139td =⇒ e−0.139td = 17

7
17


=⇒ −0.139td = log 7
1 17


=⇒ td = − 0.139 log 7
= −6.38

to two decimal places. Now 6.38 hours is 6 hours and 0.38×60 = 23 minutes. So
the time of death was 6 hours and 23 minutes before 3:45pm, which is 9:22am.
Example 2.4.15

A slightly tricky example — we need to determine the ambient temperature from

three measurements at different times.

Example 2.4.16 Finding the ambient temperature.

A glass of room-temperature water is carried out onto a balcony from an apartment

where the temperature is 22◦ C. After one minute the water has temperature 26◦ C
and after two minutes it has temperature 28◦ C. What is the outdoor temperature?
Solution: We will assume that the temperature of the thermometer obeys Newton’s
law of cooling.

• Let A be the outdoor temperature and T (t) be the temperature of the water t
minutes after it is taken outside.

• By Newton’s law of cooling,

T (t) = A + T (0) − A eKt

Theorem 2.4.4 with a = K and b = A. Notice there are 3 unknowns here — A,

T (0) and K — so we need three pieces of information to find them all.

357
 A PPLICATIONS OF I NTEGRATION 2.4 S EPARABLE D IFFERENTIAL E QUATIONS

• We are told T (0) = 22, so

T (t) = A + 22 − A eKt .

• We are also told T (1) = 26, which gives

26 = A + 22 − A eK


rearrange things
26 − A
eK =
22 − A

• Finally, T (2) = 28, so

28 = A + 22 − A e2K


rearrange
28 − A 26 − A
e2K = but eK = , so
22 − A 22 − A
 2
26 − A 28 − A
= multiply through by (22 − A)2
22 − A 22 − A
2
(26 − A) = (28 − A)(22 − A)

We can expand out both sides and collect up terms to get

262 −52A + A2 = 28 × 22 −50A + A2

|{z} | {z }
=676 =616
60 = 2A
30 = A

So the temperature outside is 30◦ .

Example 2.4.16

2.4.4 tt Optional — Population Growth

Suppose that we wish to predict the size P (t) of a population as a function of the time
t. In the most naive model of population growth, each couple produces β offspring
(for some constant β) and then dies. Thus over the course of one generation β P 2(t)
children are produced and P (t) parents die so that the size of the population grows
from P (t) to

P (t) β
P (t + tg ) = P (t) + β − P (t) = P (t)
| {z 2 } parents
|{z}
die
2
parents+offspring

358
A PPLICATIONS OF I NTEGRATION 2.4 S EPARABLE D IFFERENTIAL E QUATIONS

where tg denotes the lifespan of one generation. The rate of change of the size of the
population per unit time is
P (t + tg ) − P (t) 1 hβ i
= P (t) − P (t) = bP (t)
tg tg 2

where b = β−2
2tg
is the net birthrate per member of the population per unit time. If we
approximate
P (t + tg ) − P (t) dP
≈ (t)
tg dt
we get the differential equation

Equation 2.4.17 Population growth.

dP
= bP (t)
dt

By Corollary 2.4.10, with −k replaced by b,

Equation 2.4.18 Malthusian growth model.

P (t) = P (0) · ebt

This is called the Malthusian 5 growth model. It is, of course, very simplistic. One
of its main characteristics is that, since P (t + T ) = P (0) · eb(t+T ) = P (t) · ebT , every
time you add T to the time, the population size is multiplied by ebT . In particular, the
population size doubles every logb 2 units of time. The Malthusian growth model can
be a reasonably good model only when the population size is very small compared
to its environment 6 . A more sophisticated model of population growth, that takes
into account the “carrying capacity of the environment” is considered below.

Example 2.4.19 A rough estimate of the earth’s population.

In 1927 the population of the world was about 2 billion. In 1974 it was about 4 billion.
Estimate when it reached 6 billion. What will the population of the world be in 2100,
assuming the Malthusian growth model?
Solution: We follow our usual pattern for dealing with such problems.

5 This is named after Rev. Thomas Robert Malthus. He described this model in a 1798 paper called
“An essay on the principle of population”.
6 That is, the population has plenty of food and space to grow.

359
A PPLICATIONS OF I NTEGRATION 2.4 S EPARABLE D IFFERENTIAL E QUATIONS

• Let P (t) be the world’s population, in billions, t years after 1927. Note that
1974 corresponds to t = 1974 − 1927 = 47.

• We are assuming that P (t) obeys equation (2.4.17). So, by (2.4.18)

P (t) = P (0) · ebt

Notice that there are 2 unknowns here — b and P (0) — so we need two pieces
of information to find them.

• We are told P (0) = 2, so

P (t) = 2 · ebt

• We are also told P (47) = 4, which gives

4 = 2 · e47b clean up
e47b = 2 take the log and clean up
log 2
b= = 0.0147 to 3 decimal places
47

• We now know P (t) completely, so we can easily determine the predicted pop-
ulation a in 2100, i.e. at t = 2100 − 1927 = 173.

P (173) = 2e173b = 2e173×0.0147 = 12.7 billion

• Finally, our crude model predicts that the population is 6 billion at the time t
that obeys

P (t) = 2ebt = 6 clean up

ebt = 3 take the log and clean up
log 3 log 3
t= = 47 = 74.5
b log 2

which corresponds b to the middle of 2001.

a The 2015 Revision of World Population, a publication of the United Nations, predicts that the
world’s population in 2100 will be about 11 billion. But “about” covers a pretty large range.
They give an 80% confidence interval running from 10 billion to 12.5 billion.
b The world population really reached 6 billion in about 1999.

Example 2.4.19

Logistic growth adds one more wrinkle to the simple population model. It as-
sumes that the population only has access to limited resources. As the size of the

360
A PPLICATIONS OF I NTEGRATION 2.4 S EPARABLE D IFFERENTIAL E QUATIONS

population grows the amount of food available to each member decreases. This
in turn causes the net birth rate b to decrease. In the logistic growth model b =
P


b0 1 − K , where K is called the carrying capacity of the environment, so that
 
0 P (t)
P (t) = b0 1 − P (t)
K

This is a separable differential equation and we can solve it explicitly. We shall do

so shortly. See Example 2.4.20, below. But, before doing that, we’ll see what we can
learn about the behaviour of solutions to differential equations like this without find-
ing formulae for the solutions. It turns out that we can learn a lot just by watching
the sign of P 0 (t). For concreteness, we’ll look at solutions of the differential equation

dP

(t) = 6000 − 3P (t) P (t)
dt
We’ll sketch the graphs of four functions P (t) that obey this equation.

• For the first function, P (0) = 0.

• For the second function, P (0) = 1000.

• For the third function, P (0) = 2000.

• For the fourth function, P (0) = 3000.

The sketches will be based on the observation that (6000 − 3P ) P = 3(2000 − P ) P

• is zero for P = 0, 2000,

• is strictly positive for 0 < P < 2000 and

• is strictly negative for P > 2000.

Consequently

= 0

 if P (t) = 0

> 0
dP if 0 < P (t) < 2000
(t)
dt 
 =0 if P (t) = 2000


<0 if P (t) > 2000


Thus if P (t) is some function that obeys dP

dt (t) = 6000−3P (t) P (t), then as the graph
of P (t) passes through the point t, P (t)



 slope zero, i.e. is horizontal, if P (t) = 0

positive slope, i.e. is increasing, if 0 < P (t) < 2000
the graph has


 slope zero, i.e. is horizontal, if P (t) = 2000

negative slope, i.e. is decreasing, if 0 < P (t) < 2000


361
A PPLICATIONS OF I NTEGRATION 2.4 S EPARABLE D IFFERENTIAL E QUATIONS

as illustrated in the figure

As a result,

• if P (0) = 0, the graph starts out horizontally. In other words, as t starts to

increase, P (t) remains at zero, so the slope of the graph remains at zero. The
population size remains zero for all time. As a check, observe that the function
dP


P (t) = 0 obeys dt (t) = 6000 − 3P (t) P (t) for all t.

• Similarly, if P (0) = 2000, the graph again starts out horizontally. So P (t) re-
mains at 2000 and the slope remains at zero. The population size remains
2000
for all time. Again, the function P (t) = 2000 obeys dP
dt
(t) = 6000 − 3P (t) P (t)
for all t.

• If P (0) = 1000, the graph starts out with positive slope. So P (t)
increases
with t. As P (t) increases towards 2000, the slope (6000 − 3P (t) P (t), while
remaining positive, gets closer and closer to zero. As the graph approaches
height 2000, it becomes more and more horizontal. The graph cannot actually
cross from below 2000 to above 2000, because to do so it would have to have
strictly positive slope for some value of P above 2000, which is not allowed.

• If P (0) = 3000, the graph starts out with negative slope. So P (t)

decreases
with t. As P (t) decreases towards 2000, the slope (6000 − 3P (t) P (t), while
remaining negative, gets closer and closer to zero. As the graph approaches
height 2000, it becomes more and more horizontal. The graph cannot actually
cross from above 2000 to below 2000, because to do so it would have to have
negative slope for some value of P below 2000, which is not allowed.

These curves are sketched in the figure below. We conclude that for any initial
population size P (0), except P (0) = 0, the population size approaches 2000 as t → ∞.

362
A PPLICATIONS OF I NTEGRATION 2.4 S EPARABLE D IFFERENTIAL E QUATIONS

Now we’ll do an example in which we explicitly solve the logistic growth equa-
tion.

Example 2.4.20 Population predictions using logistic growth.

In 1986, the population of the world was 5 billion and was increasing at a rate of 2%
per year. Using the logistic growth model with an assumed maximum population of
100 billion, predict the population of the world in the years 2000, 2100 and 2500.
Solution: Let y(t) be the population of the world, in billions of people, at time 1986 +
t. The logistic growth model assumes
y 0 = ay(K − y)
where K is the carrying capacity and a = bK0 .
First we’ll determine the values of the constants a and K from the given data.
• We know that, if at time zero the population is below K, then as time increases
the population increases, approaching the limit K as t tends to infinity. So in
this problem K is the maximum population. That is, K = 100.
• We are also told that, at time zero, the percentage rate of change of population,
0 0
100 yy , is 2, so that, at time zero, yy = 0.02. But, from the differential equation,
y0 2
y
= a(K − y). Hence at time zero, 0.02 = a(100 − 5), so that a = 9500
.
We now know a and K and can solve the (separable) differential equation
dy dy
= ay(K − y) =⇒ = a dt
dt y(K − y)
1 h1 1 i
Z Z
=⇒ − dy = a dt
K y y−K
1
=⇒ [log |y| − log |y − K|] = at + C
K
|y| y
=⇒ log = aKt + CK =⇒ = DeaKt
|y − K| y−K
y y
with D = eCK . We know that y remains between 0 and K, so that y−K = K−y and
our solution obeys
y
= DeaKt
K −y
At this stage, we know the values of the constants a and K, but not the value of the
constant D. We are given that at t = 0, y = 5. Subbing in this, and the values of K
and a,
5 5
= De0 =⇒ D =
100 − 5 95
So the solution obeys the algebraic equation
y 5
= e2t/95
100 − y 95

363
 A PPLICATIONS OF I NTEGRATION 2.4 S EPARABLE D IFFERENTIAL E QUATIONS

which we can solve to get y as a function of t.

5 2t/95
y = (100 − y) e =⇒ 95y = (500 − 5y)e2t/95
95
=⇒ 95 + 5e2t/95 y = 500e2t/95

500e2t/95 100e2t/95 100

=⇒ y = 2t/95
= 2t/95
=
95 + 5e 19 + e 1 + 19e−2t/95
Finally,
100
• In the year 2000, t = 14 and y = 1+19e−28/95
≈ 6.6 billion.
100
• In the year 2100, t = 114 and y = 1+19e−228/95
≈ 36.7 billion.
100
• In the year 2200, t = 514 and y = 1+19e−1028/95
≈ 100 billion.

Example 2.4.20

2.4.5 tt Optional — Mixing Problems

Example 2.4.21 Dissolving salt.

At time t = 0, where t is measured in minutes, a tank with a 5-litre capacity contains

3 litres of water in which 1 kg of salt is dissolved. Fresh water enters the tank at a
rate of 2 litres per minute and the fully mixed solution leaks out of the tank at the
varying rate of 2t litres per minute.

a Determine the volume of solution V (t) in the tank at time t.

b Determine the amount of salt Q(t) in solution when the amount of water in the
tank is at maximum.

Solutiona : (a) The rate of change of the volume in the tank, at time t, is 2−2t, because
water is entering at a rate 2 and solution is leaking out at a rate 2t. Thus

dV
Z
= 2 − 2t =⇒ dV = (2 − 2t) dt =⇒ V = (2 − 2t) dt = 2t − t2 + C
dt

at least until V (t) reaches either the capacity of the tank or zero. When t = 0, V = 3 so
C = 3 and V (t) = 3+2t−t2 . Observe that V (t) is at a maximum when dV dt
= 2−2t = 0,
or t = 1.
(b) In the very short time interval from time t to time t + dt, 2t dt litres of brine
leaves the tank. That is, the fraction V2t(t)
dt
of the total salt in the tank, namely Q(t) V2t(t)
dt

364
A PPLICATIONS OF I NTEGRATION 2.4 S EPARABLE D IFFERENTIAL E QUATIONS

kilograms, leaves. Thus salt is leaving the tank at the rate

Q(t) V2t(t)
dt
2tQ(t) 2tQ(t)
= = kilograms per minute
dt V (t) 3 + 2t − t2
so
dQ 2tQ(t) dQ 2t 2t
=− 2
=⇒ =− 2
dt = − dt
dt 3 + 2t − t Q 3 + 2t − t (3 − t)(1 + t)
h 3/2 1/2 i
= + dt
t−3 t+1
3 1
=⇒ log Q = log |t − 3| + log |t + 1| + C
2 2
We are interested in the time interval 0 ≤ t ≤ 1. In this time interval |t − 3| = 3 − t
and |t + 1| = t + 1 so
3 1
log Q = log(3 − t) + log(t + 1) + C
2 2
At t = 0, Q is 1 so
3 1
log 1 = log(3 − 0) + log(0 + 1) + C
2 2
3 1 3
=⇒ C = log 1 − log 3 − log 1 = − log 3
2 2 2
At t = 1
3 1 3
log Q = log(3 − 1) + log(1 + 1) − log 3
2 2 2
3 3
= 2 log 2 − log 3 = log 4 − log 3 2
2
4
so Q = 3 .
32

a No pun intended (sorry).

Example 2.4.21

Example 2.4.22 Mixing brines.

A tank contains 1500 liters of brine with a concentration of 0.3 kg of salt per liter.
Another brine solution, this with a concentration of 0.1 kg of salt per liter is poured
into the tank at a rate of 20 li/min. At the same time, 20 li/min of the solution in the
tank, which is stirred continuously, is drained from the tank.

365
 A PPLICATIONS OF I NTEGRATION 2.4 S EPARABLE D IFFERENTIAL E QUATIONS

a How many kilograms of salt will remain in the tank after half an hour?

b How long will it take to reduce the concentration to 0.2 kg/li?

Solution: Denote by Q(t) the amount of salt in the tank at time t. In a very short time
interval dt, the incoming solution adds 20 dt liters of a solution carrying 0.1 kg/li. So
the incoming solution adds 0.1 × 20 dt = 2 dt kg of salt. In the same time interval
Q(t)
20 dt liters is drained from the tank. The concentration of the drained brine is 1500 .
Q(t)
So 1500 20 dt kg were removed. All together, the change in the salt content of the tank
during the short time interval is

Q(t)  Q(t) 
dQ = 2 dt − 20 dt = 2 − dt
1500 75
The rate of change of salt content per unit time is

dQ Q(t) 1

=2− =− Q(t) − 150
dt 75 75
The solution of this equation is

Q(t) = Q(0) − 150 e−t/75 + 150



1
by Theorem 2.4.4, with a = − 75 and b = 150. At time 0, Q(0) = 1500 × 0.3 = 450. So

Q(t) = 150 + 300e−t/75

(a) At t = 30
Q(30) = 150 + 300e−30/75 = 351.1 kg
(b) Q(t) = 0.2 × 1500 = 300 kg is achieved when

150 + 300e−t/75 = 300 =⇒ 300e−t/75 = 150 =⇒ e−t/75 = 0.5

t
=⇒ − = log(0.5) =⇒ t = −75 log(0.5) = 51.99 min
75
Example 2.4.22

2.4.6 tt Optional — Interest on Investments

Suppose that you deposit $P in a bank account at time t = 0. The account pays r%
interest per year compounded n times per year.

• The first interest payment is made at time t = n1 . Because the balance in the
account during the time interval 0 < t < n1 is $P and interest is being paid for

366
A PPLICATIONS OF I NTEGRATION 2.4 S EPARABLE D IFFERENTIAL E QUATIONS

1 th
of a year, that first interest payment is n1 × 100
r


n
× P . After the first

interest
1 r r
payment, the balance in the account is P + n × 100 × P = 1 + 100n P .
• The second interest payment is made at time t = n2 . Because the balance in the
account during the time interval n1 < t < n2 is 1 + 100n
r


P and interest is being

th
paid for n1 of a year, the second interest payment is n1 × 100
r r


× 1 + 100n
P.
r
After the second interest payment, the balance in the account is 1 + 100n P+
1 r r

r

2
n
× 100 × 1 + 100n P = 1 + 100n P.
• And so on.
In general, at time t = m
n
(just after the mth interest payment), the balance in the
account is

Equation 2.4.23 Discrete compounding interest.

 r m  r nt
B(t) = 1 + P = 1+ P
100n 100n

Three common values of n are 1 (interest is paid once a year), 12 (i.e. interest is
paid once a month) and 365 (i.e. interest is paid daily). The limit n → ∞ is called
continuous compounding 7 . Under continuous compounding, the balance at time t
is
 r nt
B(t) = lim 1 + P
n→∞ 100n
You may have already seen the limit

Equation 2.4.24 A useful limit.

lim (1 + x)a/x = ea
x→0

r rt
If so, you can evaluate B(t) by applying 2.4.24 with x = 100n
and a = 100
(so that
a
x
= nt). As n → ∞, x → 0 so that

Equation 2.4.25 Continuous compound interest.

 r nt
B(t) = lim 1+ P = lim (1 + x)a/x P = ea P = ert/100 P
n→∞ 100n x→0

7 There are banks that advertise continuous compounding. You can find some by googling “inter-
est is compounded continuously and paid”

367
A PPLICATIONS OF I NTEGRATION 2.4 S EPARABLE D IFFERENTIAL E QUATIONS

If you haven’t seen (2.4.24) before, that’s OK. In the following example, we red-
erive (2.4.25) using a differential equation instead of (2.4.24).

Example 2.4.26 Computing my future bank balance.

Suppose, again, that you deposit $P in a bank account at time t = 0, and that the
account pays r% interest per year compounded n times per year, and denote by
B(t) the balance at time t. Suppose that you have just received an interest payment
at time t. Then the next interest payment will be made at time t + n1 and will be
1
n
r
× 100 r
× B(t) = 100n B(t). So, calling n1 = h,

r B(t + h) − B(t) r
B(t + h) = B(t) + B(t)h or = B(t)
100 h 100
To get continuous compounding we take the limit n → ∞ or, equivalently, h → 0.
This gives

B(t + h) − B(t) r dB r
lim = B(t) or (t) = B(t)
h→0 h 100 dt 100
r r
By Theorem 2.4.4, with a = 100
and b = 0, (or Corollary 2.4.10 with k = − 100 ),

B(t) = ert/100 B(0) = ert/100 P

once again.

Example 2.4.27 Double your money.

a A bank advertises that it compounds interest continuously and that it will dou-
ble your money in ten years. What is the annual interest rate?

b A bank advertises that it compounds monthly and that it will double your
money in ten years. What is the annual interest rate?

Solution: (a) Let the interest rate be r% per year. If you start with $P , then af-
ter t years, you have P ert/100 , under continuous compounding. This was equa-
tion (2.4.25). After 10 years you have P er/10 . This is supposed to be 2P , so

P er/10 = 2P =⇒ er/10 = 2
r
=⇒ = log 2 =⇒ r = 10 log 2 = 6.93%
10
(b) Let the interest rate be r% per year. If you start with $P , then after t years, you
r

12t
have P 1+ 100×12 , under monthly compounding. This was (2.4.23). After 10 years

368
A PPLICATIONS OF I NTEGRATION 2.4 S EPARABLE D IFFERENTIAL E QUATIONS

r

120
you have P 1 + 100×12
. This is supposed to be 2P , so
r
120 r
120
P 1+ = 2P =⇒ 1+ =2
100 × 12 1200
r
=⇒ 1+ = 21/120
1200
r
=⇒ = 21/120 − 1
1200
r = 1200 21/120 − 1 = 6.95%


=⇒
Example 2.4.27

Example 2.4.28 Pension planning.

A 25 year old graduate of UBC is given $50,000 which is invested at 5% per year com-
pounded continuously. The graduate also intends to deposit money continuously at
the rate of $2000 per year.
a Find a differential equation that A(t) obeys, assuming that the interest rate re-
mains 5%.
b Determine the amount of money in the account when the graduate is 65.
c At age 65, the graduate will start withdrawing money continuously at the rate
of W dollars per year. If the money must last until the person is 85, what is the
largest possible value of W ?
Solution: (a) Let’s consider what happens to A over a very short time interval from
time t to time t + ∆t. At time t the account balance is A(t). During the (really short)
specified time interval the balance remains very close to A(t) and so earns interest
5
of 100 × ∆t × A(t). During the same time interval, the graduate also deposits an
additional $2000∆t. So
A(t + ∆t) ≈ A(t) + 0.05A(t)∆t + 2000∆t
A(t + ∆t) − A(t)
=⇒ ≈ 0.05A(t) + 2000
∆t
In the limit ∆t → 0, the approximation becomes exact and we get
dA
= 0.05A + 2000
dt
(b) The amount of money at time t obeys
dA

= 0.05A(t) + 2,000 = 0.05 A(t) + 40,000
dt
So by Theorem 2.4.4 (with a = 0.05 and b = −40,000),
A(t) = A(0) + 40,000 e0.05t − 40,000

369
 A PPLICATIONS OF I NTEGRATION 2.4 S EPARABLE D IFFERENTIAL E QUATIONS

At time 0 (when the graduate is 25), A(0) = 50,000, so the amount of money at time
t is
A(t) = 90,000 e0.05t − 40, 000
In particular, when the graduate is 65 years old, t = 40 and

A(40) = 90,000 e0.05×40 − 40, 000 = $625,015.05

(c) When the graduate stops depositing money and instead starts withdrawing
money at a rate W , the equation for A becomes

dA
= 0.05A − W = 0.05(A − 20W )
dt
assuming that the interest rate remains 5%. This time, Theorem 2.4.4 (with a = 0.05
and b = 20W ) gives
A(t) = A(0) − 20W e0.05t + 20W

If we now reset our clock so that t = 0 when the graduate is 65, A(0) = 625, 015.05.
So the amount of money at time t is

A(t) = 20W + e0.05t (625, 015.05 − 20W )

We want the account to be depleted when the graduate is 85. So, we want A(20) = 0.
This is the case if

20W + e0.05×20 (625, 015.05 − 20W ) = 0

=⇒ 20W + e(625, 015.05 − 20W ) = 0
=⇒ 20(e − 1)W = 625, 015.05e
625, 015.05e
=⇒ W = = $49, 437.96
20(e − 1)
Example 2.4.28

2.4.7 tt Exercises

Recall that we are using log x to denote the logarithm of x with base e. In other
courses it is often denoted ln x.

Exercises — Stage 1

1. Below are pairs of functions y = f (x) and differential equations. For

each pair, decide whether the function is a solution of the differential
equation.

370
A PPLICATIONS OF I NTEGRATION 2.4 S EPARABLE D IFFERENTIAL E QUATIONS

function differential equation

dy
(a) y = 5(ex − 3x2 − 6x − 6) = y + 15x2
dx
−2
(b) y= 2 y 0 (x) = xy 2
x +1  2
dy dy
(c) y = x3/2 + x + =y
dx dx

2. Following Definition 2.4.1, a separable differential equation has the

form
dy

(x) = f (x) g y(x) .
dx
Show that each of the following equations can be written in this form,
identifying f (x) and g(y).
dy
a 3y dx = x sin y
dy
b dx
= ex+y
dy
c dx
+1=x
dy 2 dy
+ x2 = 0


d dx
− 2x dx

3. Suppose we have the following functions:

• y is a differentiable function of x
R
• f is a function of x, with f (x) dx = F (x)
R 1
• g is a nonzero function of y, with g(y) dy = G(y) = G(y(x)).

In the work below, we set up a solution to the separable differential equa-

tion
dy
= f (x)g(y) = f (x)g(y(x)))
dx
without using the mnemonic of Equation (?).
By deleting some portion of our work, we can create the solution as it would
look using the mnemonic. What portion can be deleted?
Remark: the purpose of this exercise is to illuminate what, exactly, the
mnemonic is a shortcut for. Despite its peculiar look, it agrees with what
we already know about integration.

dy
= f (x)g(y(x))
dx

371
A PPLICATIONS OF I NTEGRATION 2.4 S EPARABLE D IFFERENTIAL E QUATIONS

Since g(y(x)) is a nonzero function, we can divide both sides by

it.
1 dy
· = f (x)
g(y(x)) dx

If these functions of x are the same, then they have the same
antiderivative with respect to x.

1 dy
Z Z
· dx = f (x) dx
g(y(x)) dx

The left integral is in the correct form for a change of variables to

y. To make this easier to see, we’ll use a u-substitution, since it’s
dy
a little more familiar than a y-substitution. If u = y, then du
dx
= dx ,
dy
so du = dx dx.

1
Z Z
du = f (x) dx
g(u)

Since u was just the same as y, again for cosmetic reasons, we

can swap it back. (Formally, you could have skipped the step
above–we just included it to be extra clear that we’re not using
any integration techniques we haven’t seen before.)

1
Z Z
dy = f (x) dx
g(y)

We’re given the antiderivatives in question.

G(y) + C1 = F (x) + C2
G(y) = F (x) + (C2 − C1 )

where C1 and C2 are arbitrary constants. Then also C2 − C1 is an

arbitrary constant, so we might as well call it C.

G(y) = F (x) + C

dy
4. Suppose y = f (x) is a solution to the differential equation dx = xy.
True or false: f (x) + C is also a solution, for any constant C.

5. Suppose a function y = f (x) satisfies |y| = Cx, for some constant C > 0.

a What is the largest possible domain of f (x), given the information at

hand?

b Give an example of function y = f (x) with the following properties,

or show that none exists:

372
A PPLICATIONS OF I NTEGRATION 2.4 S EPARABLE D IFFERENTIAL E QUATIONS

• |y| = Cx,
dy
• dx
exists for all x > 0, and
• y > 0 for some values of x, and y < 0 for others.
6. Express the following sentence a as a differential equation. You don’t have
to solve the equation.

About 0.3 percent of the total quantity of morphine in the blood-

stream is eliminated every minute.

a The sentence is paraphrased from the Pharmakokinetics website of Université de Lau-

sanne, Elimination Kinetics. The half-life of morphine is given on the same website.
Accessed 12 August 2017.

7. Suppose a particular change is occurring in a language, from an old

form to a new form. a Let p(t) be the proportion (measured as a number
between 0, meaning none, and 1, meaning all) of the time that speakers
use the new form. Piotrowski’s law b predicts the following.

Use of the new form over time spreads at a rate that is pro-
portional to the product of the proportion of the new form
and the proportion of the old form.

Express this as a differential equation. You do not need to solve the

differential equation.

a An example is the change in German from “wollt” to “wollst” for the second-
person conjugation of the verb “wollen.” This example is provided by the site
Laws in Quantitative Linguistics, Change in Language, accessed 18 August 2017.
b Piotrowski’s law is paraphrased from the page Piotrowski-Gesetz on Glottopedia,
accessed 18 August 2017. According to this source, the law was based on work by
the married couple R. G. Piotrowski and A. A. Piotrowskaja, later generalized by
G. Altmann.

y
8. Consider the differential equation y 0 = 2
− 1.

a When y = 0, what is y 0 ?

b When y = 2, what is y 0 ?

c When y = 3, what is y 0 ?

d On the axes below, interpret the marks we have made, and use
them to sketch a possible solution to the differential equation.

373
A PPLICATIONS OF I NTEGRATION 2.4 S EPARABLE D IFFERENTIAL E QUATIONS

x
1

9. Consider the differential equation y 0 = y − x2 .

a If y(1) = 0, what is y 0 (1)?

b If y(1) = 2, what is y 0 (1)?

c If y(1) = −2, what is y 0 (1)?

d Draw a sketch similar to that of Question 8(d) showing the derivatives

of y at the points with integer values for x in [0, 6] and y in [−3, 3].

e Sketch a possible graph of y.

Exercises — Stage 2

10. *. Find the solution to the separable initial value problem:

dy 2x
= y, y(0) = log 2
dx e
Express your solution explicitly as y = y(x).

dy xy
11. *. Find the solution y(x) of dx = x2 + 1 , y(0) = 3.
y
0
12. *. Solve the differential equation y (t) = e 3 cos t. You should express the
solution y(t) in terms of t explicitly.

13. *. Solve the differential equation

dy 2 2
= xex −log(y )
dx

374
A PPLICATIONS OF I NTEGRATION 2.4 S EPARABLE D IFFERENTIAL E QUATIONS

0
14. *. Let y = y(x). Find the general solution of the differential equation y =
xey .
yy 0 1
15. *. Find the solution to the differential equation x
= that satisfies
e − 2x y
y(0) = 3. Solve completely for y as a function of x.
16. *. Find the function y = f (x) that satisfies
dy 1
= −xy 3 and f (0) = −
dx 4

17. *. Find the function y = y(x) that satisfies y(1) = 4 and

dy 15x2 + 4x + 3
=
dx y

0 3
18. *. Find the solution y(x) of y = x y with y(0) = 1.

19. *. Find the solution of the initial value problem

dy
x + y = y2 y(1) = −1
dx

0
20. *. A function f (x) is always positive, has f (0) = e and satisfies f (x) =
x f (x) for all x. Find this function.

21. *. Solve the following initial value problem:

dy 1
= 2 y(1) = 2
dx (x + x)y
p
1+y2 − 4 0 sec x
22. *. Find the solution of the differential equation y = that
tan x y
satisfies y(0) = 2. You don’t have to solve for y in terms of x.

23. *. The fish population in a lake is attacked by a disease at time t = 0,

with the result that the size P (t) of the population at time t ≥ 0 satisfies

dP √
= −k P
dt
where k is a positive constant. If there were initially 90,000 fish in the
lake and 40,000 were left after 6 weeks, when will the fish population be
reduced to 10,000?

24. *. An object of mass m is projected straight upward at time t = 0 with initial

speed v0 . While it is going up, the only forces acting on it are gravity (as-

375
A PPLICATIONS OF I NTEGRATION 2.4 S EPARABLE D IFFERENTIAL E QUATIONS

sumed constant) and a drag force proportional to the square of the object’s
speed v(t). It follows that the differential equation of motion is

dv
m = −(mg + kv 2 )
dt
where g and k are positive constants. At what time does the object reach its
highest point?

25. *. A motor boat is traveling with a velocity of 40 ft/sec when its motor
shuts off at time t = 0. Thereafter, its deceleration due to water resis-
tance is given by

dv
= −k v 2
dt
where k is a positive constant. After 10 seconds, the boat’s velocity is 20
ft/sec.

a What is the value of k?

b When will the boat’s velocity be 5 ft/sec?

dx
26. *. Consider the initial value problem dt = k(3−x)(2−x), x(0) = 1, where
k is a positive constant. (This kind of problem occurs in the analysis of
certain chemical reactions.)

a Solve the initial value problem. That is, find x as a function of t.

b What value will x(t) approach as t approaches +∞.

27. *. The quantity P = P (t), which is a function of time t, satisfies the differ-
ential equation

dP
= 4P − P 2
dt
and the initial condition P (0) = 2.

a Solve this equation for P (t).

b What is P when t = 0.5? What is the limiting value of P as t becomes

large?
28. *. An object moving in a fluid has an initial velocity v of 400 m/min. The
velocity is decreasing at a rate proportional to the square of the velocity.

376
A PPLICATIONS OF I NTEGRATION 2.4 S EPARABLE D IFFERENTIAL E QUATIONS

After 1 minute the velocity is 200 m/min.

a Give a differential equation for the velocity v = v(t) where t is time.

b Solve this differential equation.

c When will the object be moving at 50 m/min?

Exercises — Stage 3
29. *. An investor places some money in a mutual fund where the interest is
compounded continuously and where the interest rate fluctuates between
4% and 8%. Assume that the amount of money B = B(t) in the account in
dollars after t years satisfies the differential equation

dB

= 0.06 + 0.02 sin t B
dt

a Solve this differential equation for B as a function of t.

b If the initial investment is $1000, what will the balance be at the end
of two years?

30. *. An endowment is an investment account in which the balance ideally

remains constant and withdrawals are made on the interest earned by
the account. Such an account may be modeled by the initial value prob-
lem B 0 (t) = aB − m for t ≥ 0, with B(0) = B0 . The constant a reflects
the annual interest rate, m is the annual rate of withdrawal, and B0 is
the initial balance in the account.

a Solve the initial value problem with a = 0.02 and B(0) = B0 =

$30, 000. Note that your answer depends on the constant m.

b If a = 0.02 and B(0) = B0 = $30, 000, what is the annual with-

drawal rate m that ensures a constant balance in the account?

31. *. A certain continuous function y = y(x) satisfies the integral equation

Z x
y(t)2 − 3y(t) + 2 sin tdt


y(x) = 3 + (∗)
0

for all x in some open interval containing 0. Find y(x) and the largest
interval for which (∗) holds.

377
A PPLICATIONS OF I NTEGRATION 2.4 S EPARABLE D IFFERENTIAL E QUATIONS

32. *. A cylindrical water tank, of radius 3 meters and height 6 meters, is

full of water when its bottom is punctured. Water drains out through a
hole of radius 1 centimeter. If

• h(t) is the height of the water in the tank at time t (in meters) and

• v(t) is the velocity of the escaping water at time t (in meters per
second) then
p
• Torricelli’s law states that v(t) = 2gh(t) where g = 9.8 m/sec2 .
Determine how long it takes for the tank to empty.

33. *. A spherical tank of radius 6 feet is full of mercury when a circular hole
of radius 1 inch is opened in the bottom. How long will it take for all of the
mercury to drain from the tank?
Use the value g = 32 feet/sec2 . Also use Torricelli’s law, which states when
the height of mercury √ in the tank is h, the speed of the mercury escaping
from the tank is v = 2gh.
34. *. Consider the equation
Z x


f (x) = 3 + f (t) − 1 f (t) − 2 dt
0

a What is f (0)?

b Find the differential equation satisfied by f (x).

c Solve the initial value problem determined in (a) and (b).

35. *. A tank 2 m tall is to be made with circular cross–sections with radius
r = y p . Here y measures the vertical distance from the bottom of the tank
and p is a positive constant to be determined. You may assume that when
the tank drains, it obeys Torricelli’s law, that is

dy √
A(y) = −c y
dt
for some constant c where A(y) is the cross–sectional area of the tank at
height y. It is desired that the tank be constructed so that the top half (y = 2
to y = 1) takes exactly the same amount of time to drain as the bottom half
(y = 1 to y = 0). Determine the value of p so that the tank has this property.
Note: it is not possible or necessary to find c for this question.
36. Suppose f (t) is a continuous, differentiable function and the root mean
square of f (t) on [a, x] is equal to the average of f (t) on [a, x] for all x. That
is, s
Z x Z x
1 1
f (t) d(t) = f 2 (t) dt (∗)
x−a a x−a a

378
A PPLICATIONS OF I NTEGRATION 2.4 S EPARABLE D IFFERENTIAL E QUATIONS

You may assume x > a.

a Guess a function f (t) for which the average of f (t) is the same as the
root mean square of f (t) on any interval.

b Differentiate both sides of the given equation.

c Simplify your answer

R x from (b) by using EquationR x(∗) to replace all
terms containing a f 2 (t) dt with terms containing a f (t) dt.
Rx
d Let Y (x) = a f (t) dt, so the equation from (c) becomes a differential
equation. Find all functions that satisfy it.

e What is f (t)?

37. Find the function y(x) such that

d2 y 2 dy
2
= 3·
dx y dx
1 dy
and if x = − 16 log 3, then y = 1 and dx = 3.
You do not need to solve for y explicitly.

379
 Chapter 3

S EQUENCE AND SERIES

You have probably learned about Taylor polynomials 1 and, in particular, that
x2 x3 xn
ex = 1 + x + + + ··· + + En (x)
2! 3! n!
where En (x) is the error introduced when you approximate ex by its Taylor polyno-
mial of degree n. You may have even seen a formula for En (x). We are now going
to ask what happens as n goes to infinity? Does the error go zero, giving an exact
formula for ex ? We shall later see that it does and that
∞
x x2 x3 X xn
e =1+x+ + + ··· =
2! 3! n=0
n!

At this point we haven’t defined, or developed any understanding of, this infinite
sum. How do we compute the sum of an infinite number of terms? Indeed, when
does a sum of an infinite number of terms even make sense? Clearly we need to
build up foundations to deal with these ideas. Along the way we shall also see other
functions for which the corresponding error obeys lim En (x) = 0 for some values of
n→∞
x and not for other values of x.
To motivate the next section, consider using the above formula with x = 1 to
compute the number e:
∞
1 1 X 1
e = 1 + 1 + + + ··· =
2! 3! n=0
n!

As we stated above, we don’t yet understand what to make of this infinite number
of terms, but we might try to sneak up on it by thinking about what happens as we
take more and more terms.
1 term 1=1

1 Now would be an excellent time to quickly read over your notes on the topic.

380
 S EQUENCE AND SERIES 3.1 S EQUENCES

2 terms 1+1=2
1
3 terms 1 + 1 + = 2.5
2
1 1
4 terms 1 + 1 + + = 2.666666 . . .
2 6
1 1 1
5 terms 1+1+ + + = 2.708333 . . .
2 6 24
1 1 1 1
6 terms 1+1+ + + + = 2.716666 . . .
2 6 24 120
By looking at the infinite sum in this way, we naturally obtain a sequence of numbers

{ 1 , 2 , 2.5 , 2.666666 , · · · , 2.708333 , · · · , 2.716666 , · · · , · · · }.

The key to understanding the original infinite sum is to understand the behaviour
of this sequence of numbers — in particularly, what do the numbers do as we go
further and further? Does it settle down 2 to a given limit?

3.1q Sequences

3.1.1 tt Sequences

In the discussion above we used the term “sequence” without giving it a precise
mathematical meaning. Let us rectify this now.
Definition 3.1.1
A sequence is a list of infinitely a many numbers with a specified order. It is
denoted    ∞
a1 , a2 , a3 , · · · , an , · · · or an or an n=1

a For the more pedantic reader, here we mean a list of countably infinitely many numbers.
The interested (pedantic or otherwise) reader should look up countable and uncountable
sets.

We will often specify a sequence by writing it more explicitly, like

n o∞
an = f (n)
n=1

where f (n) is some function from the natural numbers to the real numbers.

2 You will notice a great deal of similarity between the results of the next section and “limits at
infinity” which was covered last term.

381
S EQUENCE AND SERIES 3.1 S EQUENCES

Example 3.1.2 Three sequences and another one.

Here are three sequences.

n 1 1 1 o n 1 o∞
1, , , ··· , , ··· or an =
2 3 n no n=1
n o n ∞
1, 2, 3, · · · , n, · · · or an = n
n=1
n o n o∞
1, −1, 1, −1, · · · , (−1)n−1 , · · · or an = (−1)n−1
n=1

It is not necessary that there be a simple explicit formula for the nth term of a se-
quence. For example the decimal digits of π is a perfectly good sequence

3, 1, 4, 1, 5, 9, 2, 6, 5, 3, 5, 8, 9, 7, 9, 3, 2, 3, 8, 4, 6, 2, 6, 4, · · ·

but there is no simple formula a for the nth digit.

a There is, however, a remarkable result due to Bailey, Borwein and Plouffe that can be used to
compute the nth binary digit of π (i.e. writing π in base 2 rather than base 10) without having to
work out the preceding digits.

Our primary concern with sequences will be the behaviour of an as n tends to

infinity and, in particular, whether or not an “settles down” to some value as n tends
to infinity.
Definition 3.1.3
 ∞
A sequence an n=1 is said to converge to the limit A if an approaches A as n
tends to infinity. If so, we write

lim an = A or an → A as n → ∞
n→∞

A sequence is said to converge if it converges to some limit. Otherwise it is

said to diverge.

The reader should immediately recognise the similarity with limits at infinity
lim f (x) = L if f (x) → L as x → ∞
x→∞

Example 3.1.4 Convergence in Example 3.1.2.

Three of the four sequences in Example 3.1.2 diverge:

 ∞
• The sequence an = n n=1 diverges because an grows without bound, rather
than approaching some finite value, as n tends to infinity.

382
S EQUENCE AND SERIES 3.1 S EQUENCES

 ∞
• The sequence an = (−1)n−1 n=1 diverges because an oscillates between +1
and −1 rather than approaching a singe value as n tends to infinity.

• The sequence of the decimal digits of π also diverges, though the proof that
this is the case is a bit beyond us right now a .

The other sequence in Example 3.1.2 has an = n1 . As n tends to infinity, 1

n
tends to
zero. So
1
lim = 0
n→∞ n

a If the digits of π were to converge, then π would have to be a rational number. The irrationality
of π (that it cannot be written as a fraction) was first proved by Lambert in 1761. Niven’s 1947
proof is more accessible and we invite the interested reader to use their favourite search engine
to find step-by-step guides to that proof.

Example 3.1.4

n
Example 3.1.5 lim .
n→∞ 2n+1

n
Here is a little less trivial example. To study the behaviour of 2n+1
as n → ∞, it is a
good idea to write it as
n 1
=
2n + 1 2 + n1
As n → ∞, the n1 in the denominator tends to zero, so that the denominator 2 + 1
n
tends to 2 and 2+1 1 tends to 12 . So
n

n 1 1
lim = lim 1 =
n→∞ 2n + 1 n→∞ 2 + 2
n

Notice that in this last example, we are really using techniques that we used be-
fore to study infinite limits like lim f (x). This experience can be easily transferred
x→∞
to dealing with lim an limits by using the following result.
n→∞

Theorem 3.1.6
If
lim f (x) = L
x→∞

and if an = f (n) for all positive integers n, then

lim an = L
n→∞

383
S EQUENCE AND SERIES 3.1 S EQUENCES

Example 3.1.7 lim e−n .

n→∞

Set f (x) = e−x . Then e−n = f (n) and

since lim e−x = 0 we know that lim e−n = 0

x→∞ n→∞

The bulk of the rules for the arithmetic of limits of functions that you already
know also apply to the limits of sequences. That is, the rules you learned to work
with limits such as lim f (x) also apply to limits like lim an .
x→∞ n→∞

Theorem 3.1.8 Arithmetic of limits.

 ∞  ∞
Let A, B and C be real numbers and let the two sequences an n=1
and bn n=1
converge to A and B respectively. That is, assume that

lim an = A lim bn = B
n→∞ n→∞

Then the following limits hold.

 
a lim an + bn = A + B
n→∞
(The limit of the sum is the sum of the limits.)
 
b lim an − bn = A − B
n→∞
(The limit of the difference is the difference of the limits.)

c lim Can = CA.

n→∞

d lim an bn = A B
n→∞
(The limit of the product is the product of the limits.)
an A
e If B 6= 0 then lim =
n→∞ bn B
(The limit of the quotient is the quotient of the limits provided the limit of
the denominator is not zero.)

We use these rules to evaluate limits of more complicated sequences in terms of

the limits of simpler sequences — just as we did for limits of functions.

384
S EQUENCE AND SERIES 3.1 S EQUENCES

Example 3.1.9 Arithmetic of limits.

Combining Examples 3.1.5 and 3.1.7,

h n i n
lim + 7e−n = lim + lim 7e−n by Theorem 3.1.8(a)
n→∞ 2n + 1 n→∞ 2n + 1 n→∞
n
= lim + 7 lim e−n by Theorem 3.1.8(c)
n→∞ 2n + 1 n→∞

and then using Examples 3.1.5 and 3.1.7

1
= +7·0
2
1
=
2

There is also a squeeze theorem for sequences.

Theorem 3.1.10 Squeeze theorem.

If an ≤ cn ≤ bn for all natural numbers n, and if

lim an = lim bn = L
n→∞ n→∞

then
lim cn = L
n→∞

Example 3.1.11 A simple squeeze.

In this example we use the squeeze theorem to evaluate

h πn i
lim 1 +
n→∞ n
where πn is the nth decimal digit of π. That is,
π1 = 3 π2 = 1 π3 = 4 π4 = 1 π5 = 5 π 6 = 9 ···
We do not have a simple formula for πn . But we do know that
πn 9 πn 9
0 ≤ πn ≤ 9 =⇒ 0 ≤ ≤ =⇒ 1 ≤ 1 + ≤1+
n n n n
and we also know that
h 9i
lim 1 = 1 lim 1 + =1
n→∞ n→∞ n

385
 S EQUENCE AND SERIES 3.1 S EQUENCES

9
So the squeeze theorem with an = 1, bn = 1 + πnn , and cn = 1 + n
gives
h πn i
lim 1 + =1
n→∞ n
Example 3.1.11

Finally, recall that we can compute the limit of the composition of two functions
using continuity. In the same way, we have the following result:

Theorem 3.1.12 Continuous functions of limits.

If lim an = L and if the function g(x) is continuous at L, then
n→∞

lim g(an ) = g(L)

n→∞

πn
Example 3.1.13 lim sin 2n+1 .
n→∞

πn n


Write sin 2n+1 =g 2n+1
with g(x) = sin(πx). We saw, in Example 3.1.5 that

n 1
lim =
n→∞ 2n + 1 2
Since g(x) = sin(πx) is continuous at x = 12 , which is the limit of n
2n+1
, we have

πn  n  1 π
lim sin = lim g =g = sin = 1
n→∞ 2n + 1 n→∞ 2n + 1 2 2

With this introduction to sequences and some tools to determine their limits, we
can now return to the problem of understanding infinite sums.

3.1.2 tt Exercises

Exercises — Stage 1

1. Assuming the sequences continue as shown, estimate the limit of each

sequence from its graph.

386
S EQUENCE AND SERIES 3.1 S EQUENCES

1
(a) x

1
(b) x

1
(c) x

2. Suppose an and bn are sequences, and an = bn for all n ≥ 100, but an 6= bn

for n < 100.
True or false: lim an = lim bn .
n→∞ n→∞

3. Let {an }∞ ∞ ∞
n=1 , {bn }n=1 , and {cn }n=1 , be sequences with lim an = A, lim bn =
n→∞ n→∞
B, and lim cn = C. Assume A, B, and C are nonzero real numbers.
n→∞
Evaluate the limits of the following sequences.
an − b n
a
cn
cn
b
n
a2n+5
c
bn
4. Give an example of a sequence {an }∞
n=1 with the following properties:

• an > 1000 for all n ≤ 1000,

• an+1 < an for all n, and

• lim an = −2
n→∞

387
S EQUENCE AND SERIES 3.1 S EQUENCES

5. Give an example of a sequence {an }∞

n=1 with the following properties:

• an > 0 for all even n,

• an < 0 for all odd n,

• lim an does not exist.

n→∞

6. Give an example of a sequence {an }∞

n=1 with the following properties:

• an > 0 for all even n,

• an < 0 for all odd n,

• lim an exists.
n→∞

7. The limits of the sequences below can be evaluated using the squeeze
theorem. For each sequence, choose an upper bounding sequence and
lower bounding sequence that will work with the squeeze theorem. You
do not have to evaluate the limits.
sin n
a an =
n
n2
b bn =
en (7 + sin n − 5 cos n)
c cn = (−n)−n

8. Below is a list of sequences, and a list of functions.

a Match each sequence an to any and all functions f (x) such that
f (n) = an for all whole numbers n.

b Match each sequence an to any and all functions f (x) such that
lim an = lim f (x).
n→∞ x→∞

1
an = 1 + f (x) = cos(πx)
n
1 cos(πx)
bn = 1 + g(x) =
|n| x
(
x+1
x is a whole number
cn = e−n h(x) = x
1 else

388
S EQUENCE AND SERIES 3.1 S EQUENCES

(
x+1
x is a whole number
dn = (−1)n i(x) = x
0 else
n
(−1) 1
en = j(x) =
n ex

9. Let {an }∞
n=1 be a sequence defined by an = cos n.

a Give three different whole numbers n that are within 0.1 of an odd
integer multiple of π, and find the corresponding values of an .

b Give three different whole numbers n such that an is close to 0. Justify

your answers.

Remark: this demonstrates intuitively, though not rigorously, why

lim cos n is undefined. We consistently find terms in the series that are
n→∞
close to −1, and also consistently
 find terms in the series that are close to
1. Contrast this to a series like cos(2πn) , whose terms are always 1, and
whose limit therefore is 1. It is possible to turn the ideas of this question
into a rigorous proof that lim cos n is undefined. See the solution.
n→∞

Exercises — Stage 2

10. Determine the limits of the following sequences.

3n2 − 2n + 5
a an =
4n + 3
3n2 − 2n + 5
b bn =
4n2 + 3
3n2 − 2n + 5
c cn =
4n3 + 3

4n3 − 21
11. Determine the limit of the sequence an = .
ne + n1
√4
n+1
12. Determine the limit of the sequence bn = √ .
9n + 3

sin n
13. Determine the limit of the sequence cn = .
n

nsin n
14. Determine the limit of the sequence an = .
n2

389
S EQUENCE AND SERIES 3.1 S EQUENCES

15. Determine the limit of the sequence dn = e−1/n .

1 + 3 sin(n2 ) − 2 sin n
16. Determine the limit of the sequence an = .
n

en
17. Determine the limit of the sequence bn = n .
2 + n2

18. *. Find the limit, if it exists, of the sequence ak , where

k! sin3 k
ak =
(k + 1)!
n
o
n 1
19. *. Consider the sequence (−1) sin n . State whether this sequence con-
verges or diverges, and if it converges give its limit.

6n2 + 5n
 
2
20. lim
*. Evaluate n→∞ + 3 cos(1/n ) .
n2 + 1

Exercises — Stage 3
   
1
21. *. Find the limit of the sequence log sin n + log(2n) .

h√ √ i
22. Evaluate lim 2 2
n + 5n − n − 5n .
n→∞

h√ √ i
23. Evaluate lim n2 + 5n − 2n2 − 5 .
n→∞

n h
100 io∞
24. Evaluate the limit of the sequence n 2 + n1 − 2100 .
n=1

25. Write a sequence {an }∞ 0

n=1 whose limit is f (a) for a function f (x) that is
differentiable at the point a.
Your answer will depend on f and a.

26. Let {An }∞

n=3 be the area of a regular polygon with n sides, with the distance
from the centroid of the polygon to each corner equal to 1.

390
S EQUENCE AND SERIES 3.1 S EQUENCES

1 1
1

√
A(3) = 3 3 A(4) = 2 A(5) = 2.5 sin(0.4π)
4

a By dividing the polygon into n triangles, give a formula for An .

b What is lim An ?
n→∞

27. Suppose we define a sequence {fn }, which depends on some constant x, as

the following: (
1 n≤x<n+1
fn (x) =
0 else
For a fixed constant x ≥ 1, {fn } is the sequence
{0, 0, 0, . . . , 0, 1, 0, . . . , 0, 0, 0, . . .}. The sole nonzero element comes in
position k, where k is what we get when we round x down to a whole
number. If x < 1, then the sequence consists of all zeroes.
Since we can plug in different values of x, we can think of fn (x) as a function
of sequences: a different x gives you a different sequence. On the other
hand, if we imagine fixing n, then fn (x) is just a function, where fn (x) gives
the nth term in the sequence corresponding to x.

a Sketch the curve y = f2 (x).

b Sketch the curve y = f3 (x).

R∞
c Define An = 0 fn (x) dx. Give a simple description of the sequence
{An }∞
n=1 .

d Evaluate lim An .
n→∞

e Evaluate lim fn (x) for a constant x, and call the result g(x).
n→∞
Z ∞
f Evaluate g(x) dx.
0
 n
3 5
28. Determine the limit of the sequence bn = 1+ + 2 .
n n
 ∞
29. A sequence an n=1
of real numbers satisfies the recursion relation an+1 =

391
S EQUENCE AND SERIES 3.1 S EQUENCES

an + 8
for n ≥ 2.
3
a Suppose a1 = 4. What is lim an ?
n→∞

x+8
b Find x if x = .
3
c Suppose a1 = 1. Show that lim an = L, where L is the solution to
n→∞
equation above.
30. Zipf’s Law applied to word frequency can be phrased as follows:

The most-used word in a language is used n times as frequently

as the n-th most word used in a language.

a Suppose the sequence {w1 , w2 , w3 , . . .} is a list of all words in a lan-

guage, where wn is the word that is the nth most frequently used. Let
fn be the frequency of word wn . Is {f1 , f2 , f3 , . . .} an increasing se-
quence or a decreasing sequence?

b Give a general formula for fn , treating f1 as a constant.

c Suppose in a language, w1 (the most frequently used word) has fre-

quency 6%. If the language follows Zipf’s Law, then what frequency
does w3 have?

d Suppose f6 = 0.3% for a language following Zipf’s law. What is f10 ?

e The word “the” is the most-used word in contemporary American

English. In a collection of about 450 million words, “the” appeared
22,038,615 times. The second-most used word is “be,” followed by
“and.” About how many usages of these words do you expect in the
same collection of 450 million words?

Sources:

• Zipf’s word frequency law in natural language: A critical review and

future directions, Steven T. Piantadosi. Psychon Bull Rev. 2014 Oct;
21(5): 1112–1130. Accessed online 11 October 2017.

• Word Frequency Data. Accessed online 11 October 2017.

392
 S EQUENCE AND SERIES 3.2 S ERIES

3.2q Series

3.2.1 tt Series

A series is a sum

a1 + a2 + a3 + · · · + an + · · ·

of infinitely many terms. In summation notation, it is written

∞
X
an
n=1

You already have a lot of experience with series, though you might not realise it.
When you write a number using its decimal expansion you are really expressing it
as a series. Perhaps the simplest example of this is the decimal expansion of 13 :

1
= 0.3333 · · ·
3
Recall that the expansion written in this way actually means
∞
3 3 3 3 X 3
0.333333 · · · = + + + + ··· =
10 100 1000 10000 n=1
10n

The summation index n is of course a dummy index. You can use any symbol you
like (within reason) for the summation index.
∞ ∞ ∞ ∞
X 3 X 3 X 3 X 3
n
= i
= j
=
n=1
10 i=1
10 j=1
10 `=1
10`

A series can be expressed using summation notation in many different ways. For
example the following expressions all represent the same series:
n=1 n=2 n=3
∞
z}|{ z}|{ z }| {
X 3 3 3 3
n
= + + +···
n=1
10 10 100 1000
j=2 j=3 j=4
∞
z}|{ z}|{ z }| {
X 3 3 3 3
= + + +···
j=2
10j−1 10 100 1000
`=0 `=1 `=3
∞
z}|{ z}|{ z }| {
X 3 3 3 3
= + + +···
`=0
10`+1 10 100 1000

393
S EQUENCE AND SERIES 3.2 S ERIES

n=2 n=3
∞
z}|{ z }| {
3 X 3 3 3 3
+ n
= + + +···
10 n=2 10 10 100 1000
We can get from the first line to the second line by substituting n = j − 1 — don’t
forget to also change the limits of summation (so that n = 1 becomes j − 1 = 1 which
is rewritten as j = 2). To get from the first line to the third line, substitute n = ` + 1
everywhere, including in the limits of summation (so that n = 1 becomes ` + 1 = 1
which is rewritten as ` = 0).
Whenever you are in doubt as to what series a summation notation expression
represents, it is a good habit to write out the first few terms, just as we did above.
Of course, at this point, it is not clear whether the sum of infinitely many terms
adds up to a finite number or not. In order to make sense of this we will recast
the problem in terms of the convergence of sequences (hence the discussion of the
previous section). Before we proceed more formally let us illustrate the basic idea
with a few simple examples.

∞
X 3
Example 3.2.1 .
n=1
10n
P∞ 3
As we have just seen above the series n=1 10n is
n=1 n=2 n=3
∞
z}|{ z}|{ z }| {
X 3 3 3 3
n
= + + +···
n=1
10 10 100 1000

Notice that the nth term in that sum is

n−1 zeroes
z }| {
−n
3 × 10 = 0. 00 · · · 0 3

So the sum of the first 5, 10, 15 and 20 terms in that series are
5 10
X 3 X 3
= 0.33333 = 0.3333333333
n=1
10n n=1
10n
15 20
X 3 X 3
n
= 0.333333333333333 n
= 0.33333333333333333333
n=1
10 n=1
10

It sure looks like that, as we add more andPmore terms, we get closer and closer to
0.3̇ = 31 . So it is very reasonable a to define ∞ 3 1
n=1 10n to be 3 .

a Of course we are free to define the series to be whatever we want. The hard part is defining it to
be something that makes sense and doesn’t lead to contradictions. We’ll get to a more systematic
definition shortly.

394
S EQUENCE AND SERIES 3.2 S ERIES

∞
X P∞ n
Example 3.2.2 1 and n=1 (−1) .
n=1

Every term in the series ∞

P
n=1 1 is exactly 1. So the sum of the first N terms is exactly
N . As we add more andP∞more terms this grows unboundedly. So it is very reasonable
to say that the series n=1 1 diverges.
The series
n=1 n=2 n=3 n=4 n=5
X∞ z }| { z}|{ z }| { z}|{ z }| {
(−1)n = (−1) + 1 + (−1) + 1 + (−1) + · · ·
n=1

So the sum of the first N terms is 0 if N is even and −1 if N is odd. As we add more
and more terms from the series, the sum alternates between 0 and −1 for ever and
ever. So the sum of all infinitely many terms does not make any sense and it is again
reasonable to say that the series ∞ n
P
n=1 (−1) diverges.

In the above examples we have tried to understand the series by examining the
sum of the first few terms and then extrapolating as we add in more and more terms.
That is, we tried to sneak up on the infinite sum by looking at the limit of (partial)
sums of the first few terms. This approach can be made into a more formal
P rigorous
definition. More precisely, to define what is meant by the infinite sum ∞ n=1 an , we
approximate it by the sum of its first N terms and then take the limit as N tends to
infinity.
Definition 3.2.3
P∞
The N th partial sum of the series n=1 an is the sum of its first N terms
N
X
SN = an .
n=1
 ∞
The partial sums form a sequence SN N =1 . If this sequence
P of partial sums
converges SN → S as N → ∞ then we say that the series ∞ n=1 an converges
to S and we write ∞
X
an = S
n=1

If the sequence of partial sums diverges, we say that the series diverges.

395
S EQUENCE AND SERIES 3.2 S ERIES

Example 3.2.4 Geometric Series.

Let a and r be any two fixed real numbers with a 6= 0. The series
∞
X
2 n
a + ar + ar + · · · + ar + · · · = arn
n=0

is called the geometric series with first term a and ratio r.

Notice that we have chosen to start the summation index at n = 0. That’s fine. The
first a term is the n = 0 term, which is ar0 = a. The second term is the P n = 1 term,
which is ar1 = ar. And so on. We could have also written the series ∞ n=1 ar
n−1
.
n−1 1−1
That’s exactly the same series — the first term is ar n=1
= ar = a, the sec-
n−1 2−1 b
ond term is ar n=2
= ar = ar, and so on . Regardless of how we write the
geometric series, a is the first term and r is the ratio between successive terms.
Geometric series have the extremely useful property that there is a very simple for-
mula for their partial sums. Denote the partial sum by
N
X
SN = arn = a + ar + ar2 + · · · + arN .
n=0

The secret to evaluating this sum is to see what happens when we multiply it r:
rSN = r a + ar + ar2 + · · · + arN

= ar + ar2 + ar3 + · · · + arN +1

Notice that this is almost the same c as SN . The only differences are that the first
term, a, is missing and one additional term, arN +1 , has been tacked on the end. So
SN = a + ar + ar2 + · · · + arN
rSN = ar + ar2 + · · · + arN + arN +1
Hence taking the difference of these expressions cancels almost all the terms:
(1 − r)SN = a − arN +1 = a(1 − rN +1 )
Provided r 6= 1 we can divide both side by 1 − r to isolate SN :
1 − rN +1
SN = a · .
1−r
On the other hand, if r = 1, then
SN = a
|+a+
{z· · · + a} = a(N + 1)
N +1 terms

So in summary: ( N +1
a 1−r
1−r
if r 6= 1
SN =
a(N + 1) if r = 1

396
S EQUENCE AND SERIES 3.2 S ERIES

Now that we have this expression we can determine whether or not the series con-
verges. If |r| < 1, then rN +1 tends to zero as N → ∞, so that SN converges to 1−r
a
as
N → ∞ and ∞
X a
arn = provided |r| < 1.
n=0
1−r
On the other hand if |r| ≥ 1, SN diverges. To understand this divergence, consider
the following 4 cases:

• If r > 1, then rN grows to ∞ as N → ∞.

• If r < −1, then the magnitude of rN grows to ∞, and the sign of rN oscillates
between + and −, as N → ∞.

• If r = +1, then N + 1 grows to ∞ as N → ∞.

• If r = −1, then rN just oscillates between +1 and −1 as N → ∞.

In each case the sequence of partial sums does not converge and so the series does
not converge.

a It is actually quite common in computer science to think of 0 as the first integer. In that context,
the set of natural numbers is defined to contain 0:

N = {0, 1, 2, · · · }

while the notation

Z+ = {1, 2, 3, · · · }

is used to denote the (strictly) positive integers. Remember that in this text, as is more standard
in mathematics, we define the set of natural numbers to be the set of (strictly) positive integers.
b This reminds the authors of the paradox of Hilbert’s hotel. The hotel with an infinite number of
rooms is completely full, but can always accommodate one more guest. The interested reader
should use their favourite search engine to find more information on this.
c One can find similar properties of other special series, that allow us, with some work, to cancel
many terms in the partial sums. We will shortly see a good example of this. The interested reader
should look up “creative telescoping” to see how this idea might be used more generally, though
it is somewhat beyond this course.

Example 3.2.4

We should summarise the results in the previous example in a lemma.

397
S EQUENCE AND SERIES 3.2 S ERIES

Lemma 3.2.5 Geometric series.

Let a and r be real numbers and let N ≥ 0 be an integer then
N +1

XN a 1 − r if r 6= 1
n
ar = 1−r .
a(N + 1) if r = 1

n=0

Further, if |r| < 1 then

∞
X a
arn = .
n=0
1−r

Now that we know how to handle geometric series let’s return to Example 3.2.1.

Example 3.2.6 Decimal Expansions.

The decimal expansion

∞
3 3 3 3 X 3
0.3333 · · · = + + + + ··· =
10 100 1000 10000 n=1
10n

3 1
is a geometric series with the first term a = 10 and the ratio r = 10
. So, by
Lemma 3.2.5,
∞ 3 3
X 3 10 10 1
0.3333 · · · = n
= 1 = 9 =
n=1
10 1 − 10 10
3
just as we would have expected.
We can push this idea further. Consider the repeating decimal expansion:

16 16 16
0.16161616 · · · = + + + ···
100 10000 1000000
16 1
This is another geometric series with the first term a = 100
and the ratio r = 100
. So,
by Lemma 3.2.5,
∞ 16 16
X 16 100 100 16
0.16161616 · · · = n
= 1 = 99 =
n=1
100 1 − 100 100
99

again, as expected. In this way any periodic decimal expansion converges to a ratio
of two integers — that is, to a rational number a .
Here is another more complicated example.

12 34 34
0.1234343434 · · · = + + + ···
100 10000 1000000

398
S EQUENCE AND SERIES 3.2 S ERIES

∞
12 X 34
= +
100 n=2 100n

34 1
Now apply Lemma 3.2.5 with a = 1002
and r = 100

12 34 1
= + 1
100 10000 1 − 100
12 34 100
= +
100 10000 99
1222
=
9900

a We have included a (more) formal proof of this fact in the optional §3.7 at the end of this chapter.
Proving that a repeating decimal expansion gives a rational number isn’t too hard. Proving the
converse — that every rational number has a repeating decimal expansion is a little trickier, but
we also do that in the same optional section.

Example 3.2.6

Typically, it is quite difficult to write down a neat closed form expression for the
partial sums of a series. Geometric series are very notable exceptions to this. Another
family of series for which we can write down partial sums is called “telescoping
series”. These series have the desirable property that many of the terms in the sum
cancel each other out rendering the partial sums quite simple.

Example 3.2.7 Telescoping Series.

In this example, we are going to study the series ∞ 1

P
n=1 n(n+1) . This is a rather artificial
series a that has been rigged to illustrate a phenomenon call “telescoping”. Notice
that the nth term can be rewritten as
1 1 1
= −
n(n + 1) n n+1

and so we have
1
an = bn − bn+1 where bn = .
n
Because of this we get big cancellations when we add terms together. This allows us
to get a simple formula for the partial sums of this series.

1 1 1 1
SN = + + + ··· +
1·2 2·3 3·4 N · (N + 1)
1 1 1 1 1 1 1 1 
= − + − + − + ··· + −
1 2 2 3 3 4 N N +1

399
S EQUENCE AND SERIES 3.2 S ERIES

The second term of each bracket exactly cancels the first term of the following
bracket. So the sum “telescopes” leaving just

1
SN = 1 −
N +1
and we can now easily compute
∞
X 1  1 
= lim SN = lim 1 − =1
n=1
n(n + 1) N →∞ N →∞ N +1

a Well. . . this sort of series does show up when you start to look at the Maclaurin polynomial of
functions like (1 − x) log(1 − x). So it is not totally artificial. At any rate, it illustrates the basic
idea of telescoping very nicely, and the idea of “creative telescoping” turns out to be extremely
useful in the study of series — though it is well beyond the scope of this course.

Example 3.2.7

More generally, if we can write

an = bn − bn+1

for some other known sequence bn , then the series telescopes and we can compute
partial sums using
N
X N
X
an = (bn − bn+1 )
n=1 n=1
XN N
X
= bn − bn+1
n=1 n=1
= b1 − bN +1 .

and hence
∞
X
an = b1 − lim bN +1
N →∞
n=1

∞
P
provided this limit exists.Often lim bN +1 = 0 and then an = b1 . But this does not
N →∞ n=1
always happen. Here is an example.

Example 3.2.8 A Divergent Telescoping Series.

∞
1
P

In this example, we are going to study the series log 1 + n
. Let’s start by just
n=1

400
S EQUENCE AND SERIES 3.2 S ERIES

writing out the first few terms.

n=1 n=2 n=3
∞
z }| { z }|
{ { z }|
X  1  1  1 1 
log 1 + = log 1 + + log 1 + + log 1 +
n=1
n 1 2 3
n=4
z { }|
1

+ log 1 + +···
4
3 4
= log(2) + log + log
2 3
5
+ log + ···
4
This is pretty suggestive since
3 4 5
3 4 5 
log(2) + log + log + log = log 2 × × × = log(5)
2 3 4 2 3 4
So let’s try using this idea to compute the partial sum SN :
N
X  1
SN = log 1 +
n=1
n
n=1 n=2 n=3
z }| { z }| { z }| {
 1  1  1
= log 1 + + log 1 + + log 1 + +···
1 2 3
n=N −1 n=N
z }| { z }| {
 1   1
+ log 1 + + log 1 +
N −1 N
3 4  N  N + 1
= log(2) + log + log + · · · + log + log
2 3 N −1 N
 3 4 N N + 1
= log 2 × × × · · · × ×
2 3 N −1 N
= log(N + 1)

Uh oh!

lim SN = lim log(N + 1) = +∞

N →∞ N →∞

This telescoping series diverges! There is an important lesson here. Telescoping

series can diverge. They do not always converge to b1 .
Example 3.2.8

As was the case for limits, differentiation and antidifferentiation, we can com-
pute more complicated series in terms of simpler ones by understanding how series

401
S EQUENCE AND SERIES 3.2 S ERIES

interact with the usual operations of arithmetic. It is, perhaps, not so surprising that
there are simple rules for addition and subtraction of series and for multiplication of
a series by a constant. Unfortunately there are no simple general rules for computing
products or ratios of series.

Theorem 3.2.9 Arithmetic of series.

P∞ P∞
Let A, B and C be real numbers and let the two series n=1 an and n=1 bn
converge to S and T respectively. That is, assume that
∞
X ∞
X
an = S bn = T
n=1 n=1

Then the following hold.

∞
X ∞
X
   
a an + b n = S + T and an − b n = S − T
n=1 n=1

∞
X
b Can = CS.
n=1

P∞  1 2

Example 3.2.10 n=1 7n
+ n(n+1)
.

As a simple example of how we use the arithmetic of series Theorem 3.2.9, consider
∞ h
X 1 2 i
+
n=1
7n n(n + 1)
We recognize that we know how to compute parts of this sum. We know that
∞ 1
X 1 1
n
= 7 1 =
n=1
7 1− 7
6
1
because it is a geometric series (Example 3.2.4 and Lemma 3.2.5) with first term a = 7
and ratio r = 17 . And we know that
∞
X 1
=1
n=1
n(n + 1)
by Example 3.2.7. We can now use Theorem 3.2.9 to build the specified “compli-
cated” series out of these two “simple” pieces.
∞ h ∞ ∞
X 1 2 i X 1 X 2
n
+ = n
+ by Theorem 3.2.9(a)
n=1
7 n(n + 1) n=1
7 n=1
n(n + 1)

402
 S EQUENCE AND SERIES 3.2 S ERIES

∞ ∞
X 1 X 1
= + 2 by Theorem 3.2.9(b)
n=1
7n n=1
n(n + 1)
1 13
= +2·1=
6 6
Example 3.2.10

3.2.2 tt Exercises

Exercises — Stage 1
∞
X 1
1. Write out the first five partial sums corresponding to the series .
n=1
n
You don’t need to simplify the terms.

2. Every student who comes to class brings their instructor cookies, and
leaves them on the instructor’s desk. Let Ck be the total number of cook-
ies on the instructor’s desk after the kth student comes.
If C11 = 20, and C10 = 17, how many cookies did the 11th student bring
to class?

∞
X
3. Suppose the sequence of partial sums of the series an is {SN } =
  n=1
N
.
N +1
a What is {an }?

b What is lim an ?
n→∞

∞
X
c Evaluate an .
n=1

∞
X
4. Suppose the sequence of partial sums of the series an is {SN } =
  n=1
N 1
(−1) + .
N
What is {an }?

403
S EQUENCE AND SERIES 3.2 S ERIES

∞
X
5. Let f (N ) be a formula for the N th partial sum of an . (That is, f (N ) =
n=1
SN .) If f 0 (N ) < 0 for all N > 1, what does that say about an ?

Questions 6 through 8 invite you to explore geometric sums in a geometric way. This
is complementary to than the algebraic method discussed in the text.
6. Suppose the triangle outlined in red in the picture below has area one.

a Express the combined area of the black triangles as a series, assuming

the pattern continues forever.

b Evaluate the series using the picture (not the formula from your book).

7. Suppose the square outlined in red in the picture below has area one.

a Express the combined area of the black square as a series, assum-

404
S EQUENCE AND SERIES 3.2 S ERIES

ing the pattern continues forever.

b Evaluate the series using the picture (not the formula from your
book).

∞
X 1
8. In the style of Questions 6 and 7, draw a picture that represents
n=1
3n
as an area.

100
X 1
9. Evaluate n
.
n=0
5

10. Every student who comes to class brings their instructor cookies, and
leaves them on the instructor’s desk. Let Ck be the total number of cook-
ies on the instructor’s desk after the kth student comes.
If C20 = 53, and C10 = 17, what does C20 − C10 = 36 represent?

100
X 1
11. Evaluate . (Note the starting index.)
n=50
5n

12.
1
a Starting on day d = 1, every day you give your friend $ d+1 , and they
1
give $ d back to you. After a long time, how much money have you
gained by this arrangement?
∞  
X 1 1
b Evaluate − .
d=1
d (d + 1)

c Starting on day d = 1, every day your friend gives you $(d + 1), and
they take $(d + 2) from you. After a long time, how much money have
you gained by this arrangement?
∞
X
d Evaluate ((d + 1) − (d + 2)).
d=1
∞
X ∞
X ∞
X
13. Suppose an = A, bn = B, and cn = C.
n=1 n=1 n=1
X∞
Evaluate (an + bn + cn+1 ).
n=1

405
S EQUENCE AND SERIES 3.2 S ERIES

∞
X ∞
X ∞
X
14. Suppose an = A, bn = B 6= 0, and cn = C.
n=1 n=1 n=1
∞  
X an A
True or false: + cn = +C .
n=1
bn B

Exercises — Stage 2
1 1 1 1 1
15. *. To what value does the series 1 + 3 + 9 + 27 + 81 + 243 + · · · converge?

∞
X 1
16. *. Evaluate 8k
k=7

∞  
X 6 6
17. *. Show that the series − converges and find its limit.
k=1
k 2 (k + 1)2

X∞  π   π 
18. *. Find the sum of the convergent series cos − cos .
n=3
n n+1

∞
X
th
19. *. The n partial sum of a series an is known to have the formula sn =
n=1
1 + 3n
.
5 + 4n
a (a) Find an expression for an , valid for n ≥ 2.
∞
X
b (b) Show that the series an converges and find its value.
n=1

∞
X 3 · 4n+1
20. *. Find the sum of the series 8 · 5n
. Simplify your answer com-
n=2
pletely.

21. *. Relate the number 0.23̄ = 0.233333 . . . to the sum of a geometric series,
and use that to represent it as a rational number (a fraction or combination
of fractions, with no decimals).

22. *. Express 2.656565 . . . as a rational number, i.e. in the form p/q where p
and q are integers.

23. *. Express the decimal 0.321 = 0.321321321 . . . as a fraction.

406
S EQUENCE AND SERIES 3.2 S ERIES

24. *. Find the value of the convergent series

∞  n+1 
X 2 1 1
+ −
n=2
3n 2n − 1 2n + 1

Simplify your answer completely.

25. *. Evaluate
∞  n
2 n−1

X 1 
+ −
n=1
3 5
∞
X 1 + 3n+1
26. *. Find the sum of the series 4n
.
n=0

∞  
X n−3
27. Evaluate log .
n=5
n

∞  
X 2 1 1
28. Evaluate − − .
n=2
n n+1 n−1

Exercises — Stage 3
29. An infinitely long, flat cliff has stones hanging off it, attached to thin wire
of negligible mass. Starting at position x = 1, every metre (at position x,
1
where x is some whole number) the stone has mass x kg and is hanging 2x
4
metres below the top of the cliff.

407
S EQUENCE AND SERIES 3.2 S ERIES

1 2 3 4 5 6

How much work in joules does it take

to pull up all the stones to the top of the cliff?

You may use g = 9.8 m/sec2 .

30. Find the combined volume of an infinite collection of spheres, where for
1
each whole number n = 1, 2, 3, . . . there is exactly one sphere of radius n .
π
∞
X sin2 n cos2 (n + 1)
 
31. Evaluate + .
n=3
2n 2n+1

∞
X
32. Suppose a series an has sequence of partial sums {SN }, and the series
n=1
∞
(M )
X X
SN has sequence of partial sums {SM } = SN .
N =1 N =1
M +1
If SM = , what is an ?
M

33. Create a bullseye using the following method:

Starting with a red circle of area 1, divide the radius into thirds, creating
two rings and a circle. Colour the middle ring blue.
Continue the pattern with the inside circle: divide its radius into thirds, and
colour the middle ring blue.

408
S EQUENCE AND SERIES 3.2 S ERIES

Step 1

Step 2

Continue in this way indefinitely: dividing the radius of the innermost cir-
cle into thirds, creating two rings and another circle, and colouring the mid-
dle ring blue.

409
 S EQUENCE AND SERIES 3.3 C ONVERGENCE T ESTS

What is the area of the red portion?

3.3q Convergence Tests

It is very common to encounter series for which it is difficult, or even virtually impos-
sible, to determine the sum exactly. Often you try to evaluate the sum approximately
by truncating it, i.e. having the index run only up to some finite N , rather than in-
finity. But there is no point in doing so if the series diverges. So you like to at least
know if the series converges or diverges. FurthermoreP∞ you would also like to know
what error is introduced when you approximate n=1 an by the “truncated series”
PN
n=1 an . That’s called the truncation error. There are a number of “convergence
tests” to help you with this.

3.3.1 tt The Divergence Test

Our first test is very easy to apply, but it is also rarely useful. It just allows us to
quickly reject some “trivially divergent” series. It is based on the observation that

• by definition, a series ∞
P
PN n=1 an converges to S when the partial sums SN =
n=1 an converge to S.

• Then, as N → ∞, we have SN → S and, because N − 1 → ∞ too, we also have

SN −1 → S.

• So aN = SN − SN −1 → S − S = 0.
P
This tells us that, if we already know that a given series an is convergent, then the
nth term of the series, an , must converge to 0 as n tends to infinity. In this form, the

410
S EQUENCE AND SERIES 3.3 C ONVERGENCE T ESTS

test is not so useful. However the contrapositive 1 of the statement is a useful test for
divergence.

Theorem 3.3.1 Divergence Test.

 ∞
If
Pthe sequence an n=1 fails to converge to zero as n → ∞, then the series
∞
n=1 an diverges.

Example 3.3.2 A simple divergence.

n
Let an = n+1
. Then

n 1
lim an = lim = lim 1 = 1 6= 0
n→∞ n→∞ n+1 n→∞ 1+ n
P∞ n
So the series n=1 n+1 diverges.

Warning 3.3.3

The divergence test is a “one way P∞test”. It tells us that if limn→∞ an is nonzero,
or fails to exist, then the series n=1 an diverges. But it tells us absolutely noth-
P∞when limn→∞ an = 0. In particular, it is perfectly possible
ing for a series
P∞ 1
n=1 na to diverge even though limn→∞ na = 0. An example is n=1 n . We’ll
show in Example 3.3.6, below, that it diverges.

Now while convergence or divergence of series like ∞ 1

P
n=1 n can be determined
using some clever tricks — see the optional §3.3.9 —, it would be much better of
have methods that are more systematic and rely less on being sneaky. Over the next
subsections we will discuss several methods for testing series for convergence.
Note that while these tests will tell us whether or not a series converges, they do
not (except in rare cases) tell us what the series adds up to. For example, the test we

1 We have discussed the contrapositive a few times in the CLP notes, but it doesn’t hurt to discuss
it again here (or for the reader to quickly look up the relevant footnote in Section 1.3 of the CLP-1
text). At any rate, given a statement of the form “If A is true, then B is true” the contrapositive is
“If B is not true, then A is not true”. The two statements in quotation marks areP logically equiv-
alent — if one is true, then so is the other. In the present context we have “If ( an converges)
then (an converges
P to 0).” The contrapositive of this statement is then “If (an does not converge
to 0) then ( an does not converge).”

411
 S EQUENCE AND SERIES 3.3 C ONVERGENCE T ESTS

will see in the next subsection tells us quite immediately that the series
∞
X 1
n=1
n3

converges. However it does not tell us its value 2 .

3.3.2 tt The Integral Test

In the integral test, we think of a series ∞

P
n=1 an , that we cannot evaluate explicitly,
as the area of a union of rectangles, with an representing the area of a rectangle of
width one and height an . Then we compare that area with the area represented
by an integral, that we can evaluate explicitly, much as we did in Theorem 1.12.17,
the comparison test for improper integrals. We’ll start with a simple example, to
illustrate the idea. Then we’ll move on to a formulation of the test in general.

Example 3.3.4 Convergence of the harmonic series.

Visualise the terms of the harmonic series ∞ 1

P
n=1 n as a bar graph — each term is a
rectangle of height n1 and width 1. The limit of the series is then the limiting area of
this union of rectangles. Consider the sketch on the left below.

P4 1
It shows that the area of the shaded columns, n=1 n , is bigger than the area under
the curve y = x1 with 1 ≤ x ≤ 5. That is
4 Z 5
X 1 1
≥ dx
n=1
n 1 x

2 This series converges to Apéry’s constant 1.2020569031 . . .. The constant is named for Roger
Apéry (1916–1994) who proved that this number must be irrational. This number appears in
many contexts including the following cute fact — the reciprocal of Apéry’s constant gives the
probability that three positive integers, chosen at random, do not share a common prime factor.

412
S EQUENCE AND SERIES 3.3 C ONVERGENCE T ESTS

If we were to continue drawing the columns all the way out to infinity, then we
would have
∞ Z ∞
X 1 1
≥ dx
n=1
n 1 x

We are able to compute this improper integral exactly:

Z ∞
1 h iR
dx = lim log |x| = +∞
1 x R→∞ 1

That is the area under the curve diverges to +∞ and so the area represented by the
columns must also diverge to +∞.
It should be clear that the above argument can be quite easily generalised. For ex-
ample the same argument holds mutatis mutandis a for the series
∞
X 1
n=1
n2

Indeed we see from the sketch on the right above that

N Z N
X 1 1
2
≤ dx
n=2
n 1 x2

and hence
∞ Z ∞
X 1 1
2
≤ 2
dx
n=2
n 1 x

This last improper integral is easy to evaluate:

Z ∞  R
1 1
dx = lim −
2 x2 R→∞ x 2
 
1 1 1
= lim − =
R→∞ 2 R 2
Thus we know that
∞ ∞
X 1 X 1 3
= 1 + ≤ .
n=1
n2 n=2
n2 2

and so the series must converge.

a Latin for “Once the necessary changes are made”. This phrase still gets used a little, but these
days mathematicians tend to write something equivalent in English. Indeed, English is pretty
much the lingua franca for mathematical publishing. Quidquid erit.

Example 3.3.4
413
S EQUENCE AND SERIES 3.3 C ONVERGENCE T ESTS

The above arguments are formalised in the following theorem.

Theorem 3.3.5 The Integral Test.

Let N0 be any natural number. If f (x) is a function which is defined and con-
tinuous for all x ≥ N0 and which obeys

i f (x) ≥ 0 for all x ≥ N0 and

ii f (x) decreases as x increases and

iii f (n) = an for all n ≥ N0 .

Then ∞
X Z ∞
an converges ⇐⇒ f (x) dx converges
n=1 N0

Furthermore, when the series converges, the truncation error

∞
X N
X Z ∞
an − an ≤ f (x) dx for all N ≥ N0
n=1 n=1 N

Proof. Let I be any fixed integer with I > N0 . Then

P∞ P∞
• n=1 an converges if and only if n=I an converges — removing a fixed
finite number of terms from a series cannot impact whether or not it con-
verges.

• Since an ≥ 0 for all n ≥ I > N0 , the sequence of partial sums s` = `n=I an

P
obeys s`+1 = s` + an+1 ≥ s` . That is, s` increases as ` increases.

• So s` must either
P∞ converge to some finite number or increase to infinity.
That is, either n=I an converges to a finite number or it is +∞.

414
S EQUENCE AND SERIES 3.3 C ONVERGENCE T ESTS

P∞
Look at the figure above. The shaded area in the figure is n=I an because

• the first shaded rectangle has height aI and width 1, and hence area aI
and

• the second shaded rectangle has height aI+1 and width 1, and hence area
aI+1 , and so on

This shaded area is smaller than the area under the curve y = f (x) for I − 1 ≤
x < ∞. So ∞ Z ∞
X
an ≤ f (x) dx
n=I I−1
P∞
and, if the integral is finite, the sum n=I an is finite too. Furthermore, the
desired bound on the truncation error is just the special case of this inequality
with I = N + 1:
∞
X N
X ∞
X Z ∞
an − an = an ≤ f (x) dx
n=1 n=1 n=N +1 N

For the “divergencePcase” look at the figure above. The (new) shaded area in
∞
the figure is again n=I an because

• the first shaded rectangle has height aI and width 1, and hence area aI
and

• the second shaded rectangle has height aI+1 and width 1, and hence area
aI+1 , and so on

415
S EQUENCE AND SERIES 3.3 C ONVERGENCE T ESTS

This time the shaded area is larger than the area under the curve y = f (x) for
I ≤ x < ∞. So
X∞ Z ∞
an ≥ f (x) dx
n=I I
P∞
and, if the integral is infinite, the sum n=I an is infinite too.

Now that we have the integral test, it is straightforward to determine for which
values of p the series 3
∞
X 1
n=1
np
converges.

∞
1
P
Example 3.3.6 The p test: np
.
n=1

P∞ p >
Let
1
0. We’ll now use the integral test to determine whether or not the series
n=1 np (which is sometimes called the p-series) converges.

• To do so, we need a function f (x) that obeys f (n) = an = n1p for all n bigger
than some N0 . Certainly f (x) = x1p obeys f (n) = n1p for all n ≥ 1. So let’s pick
this f and try N0 = 1. (We can always increase N0 later if we need to.)

• This function also obeys the other two conditions of Theorem 3.3.5:

i f (x) > 0 for all x ≥ N0 = 1 and

1
ii f (x) decreases as x increases because f 0 (x) = −p xp+1 < 0 for all x ≥ N0 =
1.
P∞ 1
• So the integral
R ∞ dx test tells us that the series n=1 np converges if and only if the
integral 1 xp converges.
R∞
• We have already seen, in Example 1.12.8, that the integral 1 dx xp
converges if
and only if p > 1.

3 This series, viewed as a function of p, is called the Riemann zeta function, ζ(p), or the Euler-
Riemann zeta function. It is extremely important because of its connections to prime numbers
(among many other things). Indeed Euler proved that
∞
X 1 Y
−1
ζ(p) = = 1 − P−p
n=1
np
P prime

Riemann showed the connections between the zeros of this function (over complex numbers p)
and the distribution of prime numbers. Arguably the most famous unsolved problem in mathe-
matics, the Riemann hypothesis, concerns the locations of zeros of this function.

416
S EQUENCE AND SERIES 3.3 C ONVERGENCE T ESTS

P∞ 1
So we conclude that n=1 np converges if and only if p > 1. This is sometimes called
the p-test.

• In particular, the series ∞ 1

P
n=1 n , which is called the harmonic series, has p = 1
and so diverges. As we add more and more terms of this series together, the
terms we add, namely n1 , get smaller and smaller and tend to zero, but they
tend to zero so slowly that the full sum is still infinite.

• On the other hand, the series ∞ 1

P
n=1 n1.000001 has p = 1.000001 > 1 and so con-
verges. This time as we add more and more terms of this series together, the
1
terms we add, namely n1.000001 , tend to zero (just) fast enough that the full sum is
finite. Mind you, for this example, the convergence takes place very slowly —
you have to take a huge number of terms to get a decent approximation to the
full sum. If we approximate ∞
P 1
PN 1
n=1 n1.000001 by the truncated series n=1 n1.000001 ,
we make an error of at most
Z ∞ Z R
dx dx
1.000001
= lim 1.000001
N x R→∞ N x
1 h 1 1 i
= lim − −
R→∞ 0.000001 R0.000001 N 0.000001
6
10
= 0.000001
N
This does tend to zero as N → ∞, but really slowly.
Example 3.3.6

P∞We 1now know that the dividing line between convergence and divergence of
n=1 np occurs at p = 1. We can dig a little deeper and ask ourselves how much
more quickly than n1 the nth term needs to shrink in order for the series to converge.
We know that for large x, the function log x is smaller than xa for any positive a —
you can convince yourself of this with a quick application of L’Hôpital’s rule. So it
is not unreasonable to ask whether the series
∞
X 1
n=2
n log n

converges. Notice that we sum from n = 2 because when n = 1, n log n = 0. And

we don’t need to stop there 4 . We can analyse the convergence of this sum with any
power of log n.

4 We could go even further and see what happens if we include powers of log(log(n)) and other
more exotic slow growing functions.

417
 S EQUENCE AND SERIES 3.3 C ONVERGENCE T ESTS

∞
1
P
Example 3.3.7 n(log n)p
.
n=2

Let p > 0. We’ll now use the integral test to determine whether or not the series
∞
1
P
n(log n)p
converges.
n=2

• As in the last example, we start by choosing a function that obeys f (n) = an =

1
n(log n)p
for all n bigger than some N0 . Certainly f (x) = x(log1 x)p obeys f (n) =
1
n(log n)p
for all n ≥ 2. So let’s use that f and try N0 = 2.

• Now let’s check the other two conditions of Theorem 3.3.5:

i Both x and log x are positive for all x > 1, so f (x) > 0 for all x ≥ N0 = 2.
ii As x increases both x and log x increase and so x(log x)p increases and f (x)
decreases.
∞
1
P
• So the integral test tells us that the series n(log n)p
converges if and only if the
R ∞ dx n=2
integral 2 x(log x)p converges.

• To test the convergence of the integral, we make the substitution u = log x,

du = dx
x
.
R log R
dx du
Z Z
=
2 x(log x)p log 2 up
R ∞ du
We already know that the integral the integral 1 up
, and hence the integral
R R dx
2 x(log x)p
, converges if and only if p > 1.
∞
1
P
So we conclude that n(log n)p
converges if and only if p > 1.
n=2

3.3.3 tt The Comparison Test

Our next convergence test is the comparison test. It is much like the comparison test
for improper integrals (see Theorem 1.12.17) and is true for much the same reasons.
The rough idea is quite simple. A sum of larger terms must be bigger than a sum
of smaller terms. So if we know the big sum converges, then the small sum must
converge too. On the other hand, if we know the small sum diverges, then the big
sum must also diverge. Formalising this idea gives the following theorem.

418
S EQUENCE AND SERIES 3.3 C ONVERGENCE T ESTS

Theorem 3.3.8 The Comparison Test.

Let N0 be a natural number and let K > 0.

∞
P ∞
P
a If |an | ≤ Kcn for all n ≥ N0 and cn converges, then an converges.
n=0 n=0

∞
P ∞
P
b If an ≥ Kdn ≥ 0 for all n ≥ N0 and dn diverges, then an diverges.
n=0 n=0

“Proof”. We will not prove this theorem here. We’ll just observe that it is
very reasonable. That’s why there are quotation marks around “Proof”. For an
actual proof see the optional section 3.3.10.
∞
P ∞
P
a If cn converges to a finite number and if the terms in an are smaller
n=0 n=0
∞
P ∞
P
than the terms in cn , then it is no surprise that an converges too.
n=0 n=0

∞
P ∞
P
b If dn diverges (i.e. adds up to ∞) and if the terms in an are larger
n=0 n=0
∞
P ∞
P
than the terms in dn , then of course an adds up to ∞, and so di-
n=0 n=0
verges, too.

The comparison test for series is also used in much the same way as is the com-
parison test for improper integrals.
P −p Of course, one needs a good series to compare
against, and often the series n (from Example 3.3.6), for some p > 0, turns out
to be just what is needed.

P∞ 1
Example 3.3.9 n=1 n2 +2n+3 .

We could determine whether or not the series ∞ 1

P
n=1 n2 +2n+3 converges by applying
the integral test. But it is not worth the effort a . Whether or not any series converges
is determined by the behaviour of the summand b for very large n. So the first step
in tackling such a problem is to develop some intuition about the behaviour of an
when n is very large.
2
• Step 1: Develop intuition. In this case, when n is very large c nP  2n  3 so that
1
n2 +2n+3
≈ n2 . We already know, from Example 3.3.6, that ∞
1 1
n=1 np converges
P∞ 1
if p > 1. So n=1 n2 , which has p = 2, converges, and we would
if and only P
expect that ∞ 1
n=1 n2 +2n+3 converges too.

419
S EQUENCE AND SERIES 3.3 C ONVERGENCE T ESTS

• Step 2: Verify intuition. We can use the comparison test to confirm that this is
1
indeed the case. For any n ≥ 1, n2 + 2n + 3 > n2 , so that n2 +2n+3 ≤ n12 . So the
1
comparison test, Theorem 3.3.8, with an = n2 +2n+3 and cn = n12 , tells us that
P∞ 1
n=1 n2 +2n+3 converges.

1
a Go back and quickly scan Theorem 3.3.5; to apply it we need to show that n2 +2n+3 is positive
1
R
and decreasing (it is), and then we need to integrate x2 +2x+3 dx. To do that we reread the notes
on partial fractions, then rewrite x2 + 2x + 3 = (x + 1)2 + 2 and so
Z ∞ Z ∞
1 1
2 + 2x + 3
dx = 2+2
dx · · ·
1 x 1 (x + 1)

and then arctangent appears, etc etc. Urgh. Okay — let’s go back to the text now and see how to
avoid this. P∞
b To understand this consider any series n=1 an . We can always cut such a series into two parts
— pick some huge number like 106 . Then
6
∞
X 10
X ∞
X
an = an + an
n=1 n=1 n=106 +1
P∞
The first sum, though it could be humongous,
P∞ is finite. So the left hand side, n=1 an , is a
well-defined finite number if and only if n=106 +1 an , is a well-defined finite number. The con-
vergence or divergence of the series is determined by the second sum, which only contains an
for “large” n.
c The symbol “” means “much larger than”. Similarly, the symbol “” means “much less than”.
Good shorthand symbols can be quite expressive.

Example 3.3.9

Of course the previous example was “rigged” to give an easy application of the
comparison test. It is often relatively
P∞ easy, using arguments like those in Example
3.3.9, to find a “simple” series n=1 bn with bn almost the same as an when n is
large. However it is pretty rare that an ≤ bn for all n. It is much more common that
an ≤ Kbn for some constant K. This is enough to allow application of the comparison
test. Here is an example.
P∞ n+cos n
Example 3.3.10 n=1 n3 −1/3 .

As in the previous example, the first step is to develop some intuition about the
behaviour of an when n is very large.

• Step 1: Develop intuition. When n is very large,

◦ n  | cos n| so that the numerator n + cos n ≈ n and

1 1
◦ n3  3
so that the denominator n3 − 3
≈ n3 .

420
S EQUENCE AND SERIES 3.3 C ONVERGENCE T ESTS

So when n is very large

n + cos n n 1
an = 3 1 ≈ 3 = 2
n −3 n n
P∞ 1
We already know fromPExample 3.3.6, with p = 2, that n=1 n2 converges, so
we would expect that ∞ n+cos n
n=1 n3 − 1 converges too.
3

• Step 2: Verify intuition. We can use the comparison test to confirm that this
is indeed the case. To do so we need to find a constant K such that |an | =
|n+cos n|
n3 −1/3
= n+cos n
n3 −1/3
is smaller than nK2 for all n. A good way a to do that is to
factor the dominant term (in this case n) out of the numerator and also factor
the dominant term (in this case n3 ) out of the denominator.
n + cos n n 1 + cosn n 1 1 + cosn n
an = = =
n3 − 31 n3 1 − 3n1 3 n2 1 − 3n1 3
(cos n)
1+
So now we need to find a constant K such that 1−
n
1 is smaller than K for all
3n3
n ≥ 1.
◦ First consider the numerator 1 + (cos n) n1 . For all n ≥ 1
 n1 ≤ 1 and
 | cos n| ≤ 1
So the numerator 1 + (cos n) n1 is always smaller than 1 + (1) 11 = 2.
1
◦ Next consider the denominator 1 − 3n3
.
 When n ≥ 1, 3n1 3 lies between 31 and 0 so that
 1 − 3n1 3 is between 32 and 1 and consequently
 1−1 1 is between 23 and 1.
3n3

◦ As the numerator 1 + (cos n) n1 is always smaller than 2 and 1

1− 13
is always
3n
3
smaller than 2
, the fraction
1 + cosn n 3
≤ 2 =3
1 − 3n1 3 2

We now know that

1 1 + n2 3
|an | = 2 1 ≤ 2
n 1 − 3n3 n
P∞ −2
and, since
P∞ n+cos n we know n=1 n converges, the comparison test tells us that
n=1 n3 −1/3 converges.

a This is very similar to how we computed limits at infinity way way back near the beginning of
CLP-1.

Example 3.3.10
421
S EQUENCE AND SERIES 3.3 C ONVERGENCE T ESTS

The last example was actually a relatively simple application of the comparison
theorem — finding a suitable constant K can be really tedious 5 . Fortunately, there
is a variant of the comparison test that completely eliminates the need to explicitly
find K.
The idea behind this isn’t too complicated. We have already seen that the conver-
gence or divergence of a series depends not on its first few terms, but just on what
happens when n is really large. Consequently, if we can work out how the series
terms behave for really big n then we can work out if the series converges. So in-
stead of comparing the terms of our series for all n, just compare them when n is
big.

Theorem 3.3.11 Limit Comparison Theorem.

Let ∞
P P∞
n=1 an and n=1 bn be two series with bn > 0 for all n. Assume that

an
lim =L
n→∞ bn

exists.
P∞ P∞
a If n=1 b n converges, then n=1 an converges too.

b If L 6= 0 and ∞
P P∞
n=1 bn diverges, then n=1 an diverges too.

In particular, if L 6= 0, then ∞
P P∞
n=1 an converges if and only if n=1 bn converges.

an
Proof. (a) Because we are told that limn→∞ bn
= L, we know that,

an an
• when n is large, bn
is very close to L, so that bn
is very close to |L|.

an
• In particular, there is some natural number N0 so that bn
≤ |L| + 1, for
all n ≥ N0 , and hence

• |an | ≤ Kbn with K = |L| + 1, for all n ≥ N0 .

P∞
• The comparison Theorem 3.3.8 now implies that n=1 an converges.

(b) Let’s suppose that L > 0. (If L < 0, just replace an with −an .) Because we
are told that limn→∞ abnn = L, we know that,
an
• when n is large, bn
is very close to L.
an
• In particular, there is some natural number N so that bn
≥ L2 , and hence

5 Really, really tedious. And you thought some of those partial fractions computations were bad
...

422
S EQUENCE AND SERIES 3.3 C ONVERGENCE T ESTS

L
• an ≥ Kbn with K = 2
> 0, for all n ≥ N .
P∞
• The comparison Theorem 3.3.8 now implies that n=1 an diverges.

The next two examples illustrate how much of an improvement the above theo-
rem is over the straight comparison test (though of course, we needed the compari-
son test to develop the limit comparison test).

P∞ √
n+1
Example 3.3.12 n=1 n2 −2n+3 .
√
n+1
Set an = n2 −2n+3 . We first try to develop some intuition about the behaviour of an for
large n and then we confirm that our intuition was correct.
√ √
• Step 1: Develop intuition. When n  1, the√numerator n + 1 ≈ n, and the
denominator n2 − 2n + 3 ≈ n2 so that an ≈ n2n = n3/2 1
and it looks like our series
3
should converge by Example 3.3.6 with p = 2 .
1
• Step 2: Verify intuition. To confirm our intuition we set bn = n3/2
and compute
the limit √ √
n+1 3/2
an 2 n n+1
lim = lim n −2n+3
1 = lim 2
n→∞ bn n→∞ n→∞ n − 2n + 3
n3/2

Again it is a good idea to factor the dominant term out of the numerator and
the dominant term out of the denominator.
q q
1
an
2
n 1+ n 1 + n1
lim = lim 2
= lim =1
n→∞ bn n→∞ n 1 − 2 + 32 n→∞ 1 − 2 + 32
n n n n

We already know that the series ∞

P P∞ 1
n=1 b n = n=1 n3/2 converges by Example
3.3.6 with p = 23 . So our series converges by the limit comparison test, Theorem
3.3.11.

P∞ √
n+1
Example 3.3.13 n=1 n2 −2n+3
, again.

We can also try to deal with the series of Example 3.3.12, using the comparison test
directly. But that requires us to find K so that
√
n+1 K
2
≤ 3/2
n − 2n + 3 n
We might do this by examining the numerator and denominator separately:

423
 S EQUENCE AND SERIES 3.3 C ONVERGENCE T ESTS

• The numerator isn’t too bad since for all n ≥ 1:

n + 1 ≤ 2n and so
√ √
n + 1 ≤ 2n

• The denominator is quite a bit more tricky, since we need a lower bound, rather
than an upper bound, and we cannot just write |n2 − 2n + 3| ≥ n2 , which is
false. Instead we have to make a more careful argument. In particular, we’d
1
like to find N0 and K 0 so that n2 − 2n + 3 ≥ K 0 n2 , i.e. n2 −2n+3 ≤ K 01n2 for all
n ≥ N0 . For n ≥ 4, we have 2n = 12 4n ≤ 12 n · n = 12 n2 . So for n ≥ 4,

1 1
n2 − 2n + 3 ≥ n2 − n2 + 3 ≥ n2
2 2

Putting the numerator and denominator back together we have

√ √
n+1 2n √ 1
≤ = 2 2 3/2 for all n ≥ 4
n2 − 2n + 3 n2 /2 n

and the comparison test then tells us that our series converges. It is pretty clear that
the approach of Example 3.3.12 was much more straightforward.
Example 3.3.13

3.3.4 tt The Alternating Series Test

When the signs of successive terms in a series alternate between + and −, like for
example in 1 − 12 + 13 − 14 + · · · , the series is called an alternating series. More generally,
the series ∞
X
A1 − A2 + A3 − A4 + · · · = (−1)n−1 An
n=1

is alternating if every An ≥ 0. Often (but not always) the terms in alternating series
get successively smaller. That is, then A1 ≥ A2 ≥ A3 ≥ · · ·. In this case:
• The first partial sum is S1 = A1 .
• The second partial sum, S2 = A1 − A2 , is smaller than S1 by A2 .
• The third partial sum, S3 = S2 + A3 , is bigger than S2 by A3 , but because A3 ≤
A2 , S3 remains smaller than S1 . See the figure below.
• The fourth partial sum, S4 = S3 − A4 , is smaller than S3 by A4 , but because
A4 ≤ A3 , S4 remains bigger than S2 . Again, see the figure below.
• And so on.

424
S EQUENCE AND SERIES 3.3 C ONVERGENCE T ESTS

So the successive partial sums oscillate, but with ever decreasing amplitude. If, in
addition, An tends to 0 as n tends to ∞, the amplitude of oscillation tends to zero
and the sequence S1 , S2 , S3 , · · · converges to some limit S.
This is illustrated in the figure

Here is a convergence test for alternating series that exploits this structure, and
that is really easy to apply.

Theorem 3.3.14 Alternating Series Test.

 ∞
Let An n=1 be a sequence of real numbers that obeys

i An ≥ 0 for all n ≥ 1 and

ii An+1 ≤ An for all n ≥ 1 (i.e. the sequence is monotone decreasing) and

iii limn→∞ An = 0.

Then ∞
X
A1 − A2 + A3 − A4 + · · · = (−1)n−1 An = S
n=1

converges and, for each natural number N , S − SN is between 0 and (the first
dropped term) (−1)N AN +1 . Here SN is, as previously, the N th partial sum
N
(−1)n−1 An .
P
n=1

“Proof”. We shall only give part of the proof here. For the rest of the proof
see the optional section 3.3.10. We shall fix any natural number N and concen-
trate on the last statement, which gives a bound on the truncation error (which
is the error introduced when you approximate the full series by the partial sum

425
S EQUENCE AND SERIES 3.3 C ONVERGENCE T ESTS

SN )
∞
X
EN = S − SN = (−1)n−1 An
n=N +1
h i
= (−1)N AN +1 − AN +2 + AN +3 − AN +4 + · · ·
This is of course another series. We’re going to study the partial sums
`
X `−N
X
n−1 N
SN,` = (−1) An = (−1) (−1)m−1 AN +m
n=N +1 m=1

for that series.

• If `0 > N + 1, with `0 − N even,
≥0 ≥0
z }| { z }| {
N
(−1) SN,`0 = (AN +1 − AN +2 ) + (AN +3 − AN +4 ) + · · ·
≥0
z }| {
+ (A`0 −1 − A`0 )
≥0

and
≥0
z }| { z ≥0 }| {
(−1)N SN,`0 +1 = (−1)N SN,`0 + A`0 +1 ≥ 0
This tells us that (−1)N SN,` ≥ 0 for all ` > N + 1, both even and odd.
• Similarly, if `0 > N + 1, with `0 − N odd,
≥0 ≥0
z }| { z }| {
(−1)N SN,`0 = AN +1 − (AN +2 − AN +3 ) − (AN +4 − AN +5 ) − · · ·
≥0
z }| {
− (A` −1 − A` )
0 0

≤ AN +1
≤AN +1
z }| { z ≥0 }| {
(−1)N SN,`0 +1 = (−1)N SN,`0 − A`0 +1 ≤ AN +1
This tells us that (−1)N SN,` ≤ AN +1 for all for all ` > N + 1, both even and
odd.
So we now know that SN,` lies between its first term, (−1)N AN +1 , and 0 for all
` > N + 1. While we are not going to prove it here (see the optional section
3.3.10), this implies that, since AN +1 → 0 as N → ∞, the series converges and
that
S − SN = lim SN,`
`→∞
lies between (−1)N AN +1 and 0.


426
S EQUENCE AND SERIES 3.3 C ONVERGENCE T ESTS

Example 3.3.15 Convergence of the alternating harmonic series.

that the harmonic series ∞ 1

P
We have already seen, in Example
P∞ 3.3.6,n−1 n=1 n diverges.
1
On the other hand, the series n=1 (−1) n converges by the alternating series test
with An = n1 . Note that

i An = n1 ≥ 0 for all n ≥ 1, so that ∞ n−1 1

P
n=1 (−1) n
really is an alternating series,
and
1
ii An = n
decreases as n increases, and
1
iii lim An = lim = 0.
n→∞ n→∞ n

so that all of the hypotheses of the alternating series test, i.e. of Theorem 3.3.14, are
satisfied. We shall see, in Example 3.5.20, that
∞
X (−1)n−1
= log 2.
n=1
n

Example 3.3.16 e.
P∞ xn
You may already know that ex = n=0 n! . In any event, we shall prove this in
Example 3.6.5, below. In particular
∞
X (−1)n
1 1 1 1 1 1
= e−1 = = 1 − + − + − + ···
e n=0
n! 1! 2! 3! 4! 5!

is an alternating series and satisfies all of the conditions of the alternating series test,
Theorem 3.3.14a:
i The terms in the series alternate in sign.
ii The magnitude of the nth term in the series decreases monotonically as n in-
creases.
iii The nth term in the series converges to zero as n → ∞.
So the alternating series test guarantees that, if we approximate, for example,
1 1 1 1 1 1 1 1 1
≈ − + − + − + −
e 2! 3! 4! 5! 6! 7! 8! 9!
then the error in this approximation lies between 0 and the next term in the series,
1
which is 10! . That is
1 1 1 1 1 1 1 1 1
− + − + − + − ≤
2! 3! 4! 5! 6! 7! 8! 9! e

427
S EQUENCE AND SERIES 3.3 C ONVERGENCE T ESTS

1 1 1 1 1 1 1 1 1
≤ − + − + − + − +
2! 3! 4! 5! 6! 7! 8! 9! 10!
so that
1
1 1 1 1 1 1 1 1 1 ≤e
2!
− 3!
+ 4!
− 5!
+ 6!
− 7!
+ 8!
− 9!
+ 10!
1
≤ 1 1 1 1 1 1 1 1
2!
− 3!
+ 4!
− 5!
+ 6!
− 7!
+ 8!
− 9!

which, to seven decimal places says

2.7182816 ≤ e ≤2.7182837

(To seven decimal places e = 2.7182818.)

The alternating series test tells us that, for any natural number N , the error that we
(−1)n
make when we approximate 1e by the partial sum SN = N
P
n=0 n! has magnitude
1
no larger than (N +1)! . This tends to zero spectacularly quickly as N increases, simply
because (N + 1)! increases spectacularly quickly as N increases a . For example 20! ≈
2.4 × 1027 .

a The interested
√
n reader may wish to check out “Stirling’s approximation”, which says that n! ≈
2πn ne .

Example 3.3.16

11
Example 3.3.17 Computing log 10 .

We will shortly see, in Example 3.5.20, that if −1 < x ≤ 1, then

∞
x2 x3 x4 X xn
log(1 + x) = x − + − + ··· = (−1)n−1
2 3 4 n=1
n
11 11
Suppose that we have to compute log 10 to within an accuracy of 10−12 . Since 10
=
1
1 + 10 , we can get log 11
10
1
by evaluating log(1 + x) at x = 10 , so that
11  1 1 1 1 1
log = log 1 + = − 2
+ 3
− + ···
10 10 10 2 × 10 3 × 10 4 × 104
∞
X 1
= (−1)n−1
n=1
n × 10n
By the alternating series test, this series converges. Also by the alternating series test,
approximating log 1110
by throwing away all but the first N terms
11 1 1 1 1 1
log ≈ − 2
+ 3
− 4
+ · · · + (−1)N −1
10 10 2 × 10 3 × 10 4 × 10 N × 10N

428
S EQUENCE AND SERIES 3.3 C ONVERGENCE T ESTS

N
X 1
= (−1)n−1
n=1
n × 10n

introduces an error whose magnitude is no more than the magnitude of the first term
that we threw away.
1
error ≤
(N + 1) × 10N +1
To achieve an error that is no more than 10−12 , we have to choose N so that
1
≤ 10−12
(N + 1) × 10N +1

The best way to do so is simply to guess — we are not going to be able to manipulate
1 1
the inequality (N +1)×10 N +1 ≤ 1012 into the form N ≤ · · ·, and even if we could, it

would not be worth the effort. We need to choose N so that the denominator (N +
1) × 10N +1 is at least 1012 . That is easy, because the denominator contains the factor
10N +1 which is at least 1012 whenever N + 1 ≥ 12, i.e. whenever N ≥ 11. So we will
achieve an error of less than 10−12 if we choose N = 11.
1 1 1
= < 12
(N + 1) × 10N +1 N =11 12 × 1012 10

This is not the smallest possible choice of N , but in practice that just doesn’t matter
— your computer is not going to care whether or not you ask it to compute a few
1 1
extra terms. If you really need the smallest N that obeys (N +1)×10 N +1 ≤ 1012 , you can

next just try N = 10, then N = 9, and so on.

1 1 1
N +1
= 12
< 12
(N + 1) × 10 N =11 12 × 10 10
1 1 1 1
N +1
= 11
< 11
= 12
(N + 1) × 10 N =10 11 × 10 10 × 10 10
1 1 1 1
N +1
= 10
= 11 > 12
(N + 1) × 10 N =9 10 × 10 10 10

So in this problem, the smallest acceptable N = 10.

Example 3.3.17

429
 S EQUENCE AND SERIES 3.3 C ONVERGENCE T ESTS

3.3.5 tt The Ratio Test

The idea behind the ratio test comes from a reexamination of the geometric series.
Recall that the geometric series
∞
X ∞
X
an = arn
n=0 n=0

converges when |r| < 1 and diverges otherwise. So the convergence of this series is
completely determined by the number r. This number is just the ratio of successive
terms — that is r = an+1 /an .
In general the ratio of successive terms of a series, an+1
an
, is not constant, but
P de-
pends on n. However, as we have noted above, the convergence of a series an is
determined by the behaviour of its terms when n is large. In this way, the behaviour
of this ratio when n is small tells us nothing about the convergence of the series, but
the limit of the ratio as n → ∞ does. This is the basis of the ratio test.

Theorem 3.3.18 Ratio Test.

Let N be any positive integer and assume that an 6= 0 for all n ≥ N .
∞
an+1 P
a If lim = L < 1, then an converges.
n→∞ an n=1

∞
an+1 an+1 P
b If lim = L > 1, or lim = +∞, then an diverges.
n→∞ an n→∞ an n=1

Warning 3.3.19

Beware that the ratio test provides absolutely no conclusion about the conver-
∞
an if lim an+1
P
gence or divergence of the series an
= 1. See Example 3.3.22,
n=1 n→∞
below.

Proof. (a) Pick any number R obeying L < R < 1. We are assuming that
an+1
an
approaches L as n → ∞. In particular there must be some natural num-
an+1
ber M so that an
≤ R for all n ≥ M . So |an+1 | ≤ R|an | for all n ≥ M . In
particular

|aM +1 | ≤ R |aM |
|aM +2 | ≤ R |aM +1 | ≤ R2 |aM |

430
S EQUENCE AND SERIES 3.3 C ONVERGENCE T ESTS

|aM +3 | ≤ R |aM +2 | ≤ R3 |aM |

..
.
|aM +` | ≤ R` |aM |

for all ` ≥ 0. The series ∞ `

P
`=0 R |aM | is a geometric series with ratio R smaller
than one in magnitude and so converges. Consequently, by the comparison
test with an replaced by A` = an+` and cn replaced by C` = R` |aM |, the series
P∞ P∞ ∞
P
aM +` = an converges. So the series an converges too.
`=1 n=M +1 n=1
an+1
(b) We are assuming that an
approaches L > 1 as n → ∞. In particular
there must be some natural number M > N so that an+1 an
≥ 1 for all n ≥ M .
So |an+1 | ≥ |an | for all n ≥ M . That is, |an | increases as n increases as long as
n ≥ M . So |an | ≥ |aM | for all n ≥ M and an cannot converge to zero as n → ∞.
So the series diverges by the divergence test.

P∞
Example 3.3.20 n=0 anxn−1 .

Fix any two nonzero real numbers a and x. We have already seen in Example 3.2.4
and
P∞ Lemma 3.2.5 — we have just renamed r to x — that the geometric series
n
n=0 ax converges when |x| < 1 and diverges when |x| ≥ 1. We are now going
to consider
P∞ a new series, constructed by differentiating a each term in the geometric
series n=0 axn . This new series is
∞
X
an with an = a n xn−1
n=0

Let’s apply the ratio test.

an+1 a (n + 1) xn n+1  1
= = |x| = 1 + |x| → L = |x| as n → ∞
an a n xn−1 n n
The ratio test now tells us that the series ∞ n−1
P
n=0 a n x converges if |x| < 1 and
diverges if |x| > 1. It says nothing about the cases x = ±1. But in both of those cases
an = a n (±1)n does not converge to zero as n → ∞ and the series diverges by the
divergence test.

P∞
a We shall see later,
P∞in Theorem 3.5.13, that the function n=0 anxn−1 is indeed the derivative
of the function n=0 axn . Of course, such a statement only makes sense where these series
converge — how can you differentiate a divergent series? (This is not an allusion to a popular
series of dystopian novels.) Actually, there is quite a bit of interesting and useful mathematics
involving divergent series, but it is well beyond the scope of this course.

431
S EQUENCE AND SERIES 3.3 C ONVERGENCE T ESTS

Notice that in the above example, we had to apply another convergence test in
addition to the ratio test. This will be commonplace when we reach power series
and Taylor series — the ratio test will tell us something like

The series converges for |x| < R and diverges for |x| > R.

Of course, we will still have to to determine what happens when x = +R, −R. To
determine convergence or divergence in those cases we will need to use one of the
other tests we have seen.
P∞ a n+1
Example 3.3.21 n=0 n+1 X .

Once again, fix anyP∞two nnonzero real numbers a and X. We again start with thea
geometric series n=0 ax but this time we construct a new series by integrating
a
each term, axn , from x = 0 to x = X giving n+1 X n+1 . The resulting new series is
∞
X a
an with an = X n+1
n=0
n+1

To apply the ratio test we need to compute

a
an+1 n+2
X n+2 n+1 1 + n1
= = |X| = |X| → L = |X| as n → ∞
an a
n+1
X n+1 n+2 1 + n2

The ratio test now tells us that the series ∞ a n+1

P
n=0 n+1 X converges if |X| < 1 and
diverges if |X| > 1. It says nothing about the cases X = ±1.
If X = 1, the series reduces to
∞ ∞ ∞
X a X a X 1
X n+1 = =a with m = n + 1
n=0
n+1 X=1 n=0
n+1 m=1
m

which is just a times the harmonic series, which we know diverges, by Example 3.3.6.
If X = −1, the series reduces to
∞ ∞
X a X a
X n+1 = (−1)n+1
n=0
n+1 X=−1 n=0
n+1

which converges by the alternating series test. See Example 3.3.15.

In conclusion, the series ∞ a n+1
P
n=0 n+1 X converges if and only if −1 ≤ X < 1.

P∞ a n+1
a We shall also see later, in Theorem 3.5.13, that the function n=0 n+1 x is indeed an an-
P∞
tiderivative of the function n=0 axn .

The ratio test is often quite easy to apply, but one must always be careful when
the limit of the ratio is 1. The next example illustrates this.

432
S EQUENCE AND SERIES 3.3 C ONVERGENCE T ESTS

Example 3.3.22 L = 1.

In this example, we are going to see three different series that all have
an+1
limn→∞ an
= 1. One is going to diverge and the other two are going to converge.

• The first series is the harmonic series

∞
X 1
an with an =
n=1
n

We have already seen, in Example 3.3.6, that this series diverges. It has
1
an+1 n+1 n 1
= 1 = = 1 → L = 1 as n → ∞
an n
n+1 1+ n

• The second series is the alternating harmonic series

∞
X 1
an with an = (−1)n−1
n=1
n

We have already seen, in Example 3.3.15, that this series converges. But it also
has
1
an+1 (−1)n n+1 n 1
= 1 = = → L = 1 as n → ∞
an n−1
(−1) n n+1 1 + n1

• The third series is ∞

X 1
an with an =
n=1
n2
We have already seen, in Example 3.3.6 with p = 2, that this series converges.
But it also has
1
an+1 (n+1)2 n2 1
= 1 = = → L = 1 as n → ∞
an n2
(n + 1)2 (1 + n1 )2

Let’s do a somewhat artificial example that forces us to combine a few of the

techniques we have seen.

√
P∞ (−3)n n+1 n
Example 3.3.23 n=1 2n+3
X .

Again, the convergence of this series will depend on X.

433
S EQUENCE AND SERIES 3.3 C ONVERGENCE T ESTS

• Let us start with the ratio test — so we compute

√
an+1 (−3)n+1 n + 2(2n + 5)X n+1
= √
an (−3)n n + 1(2n + 3)X n
√
n + 2 2n + 5
= | − 3| · √ · · |X|
n + 1 2n + 3
So in the limit as n → ∞ we are left with
an+1
lim = 3|X|
n→∞ an

• The ratio test then tells us that if 3|X| > 1 the series diverges, while when
3|X| < 1 the series converges.

• This leaves us with the cases X = + 31 and − 31 .

1
• Setting X = 3
gives the series
∞ √
X (−1)n n + 1
n=1
2n + 3

The fact that the terms alternate here suggests that we use the alternating series
test. That will show that this series converges provided
√
x+1
f (x) =
2x + 3
is a decreasing function. To prove that, it suffices to show its derivative is
negative when x ≥ 1:
1 −1/2
√
(2x + 3) · · (x + 1) − 2 x+1
f 0 (x) = 2
2
(2x + 3)
(2x + 3) − 4(x + 1)
= √
2 x + 1(2x + 3)2
−2x − 1
= √
2 x + 1(2x + 3)2
So when x ≥ 1 this is negative and so f (x) is a decreasing function. Thus we
can apply the alternating series test to show that the series converges when
x = 13 .

• When X = − 13 the series becomes

∞ √
X n+1
.
n=1
2n + 3

434
 S EQUENCE AND SERIES 3.3 C ONVERGENCE T ESTS

√
n
Notice that when n is large, the summand is approximately
P −1/2 2n
which suggests
that the series will diverge by comparison with n . To formalise this, we
can use the limit comparison theorem:
√ √ p
n+1 1 n · 1 + 1/n 1/2
lim −1/2
= lim ·n
n→∞ 2n + 3 n n→∞ n(2 + 3/n)
p
n · 1 + 1/n
= lim
n→∞ n(2 + 3/n)
1
=
2
P −1/2
So since this ratio has a finite limit and the series n diverges, we know
that our series also diverges.

So in summary the series converges when − 13 < X ≤ 1

3
and diverges otherwise.
Example 3.3.23

3.3.6 tt Convergence Test List

We now have half a dozen convergence tests:

• Divergence Test

◦ works well when the nth term in the series fails to converge to zero as n
tends to infinity

• Alternating Series Test

◦ works well when successive terms in the series alternate in sign

◦ don’t forget to check that successive terms decrease in magnitude and
tend to zero as n tends to infinity

• Integral Test

◦ works well when, if you substitute x for n in the nth term you get a func-
tion, f (x), that you can integrate
◦ don’t forget to check that f (x) ≥ 0 and that f (x) decreases as x increases

• Ratio Test

◦ works well when an+1

an
simplifies enough that you can easily compute lim an+1
an
=
n→∞
L
◦ this often happens when an contains powers, like 7n , or factorials, like n!

435
 S EQUENCE AND SERIES 3.3 C ONVERGENCE T ESTS

◦ don’t forget that L = 1 tells you nothing about the convergence/divergence

of the series

• Comparison Test and Limit Comparison Test

◦ works well when, for very large n, the nth term an is approximately the
P term bn (see Example 3.3.10) and it is easy to determine
same as a simpler
whether or not ∞ n=1 bn converges
◦ don’t forget to check that bn ≥ 0
◦ usually the Limit Comparison Test is easier to apply than the Comparison
Test

3.3.7 tt Optional — The Leaning Tower of Books

Imagine that you are about to stack a bunch of identical books on a table. But you
don’t want to just stack them exactly vertically. You want to built a “leaning tower
of books” that overhangs the edge of the table as much as possible.

How big an overhang can you get? The answer to that question, which we’ll now
derive, uses a series!

• Let’s start by just putting book #1 on the table. It’s the red book labelled “B1 ”
in the figure below.

L
B1

x1 x
0

Use a horizontal x-axis with x = 0 corresponding to the right hand edge of

the table. Imagine that we have placed book #1 so that its right hand edge
overhangs the end of the table by a distance x1 .

436
S EQUENCE AND SERIES 3.3 C ONVERGENCE T ESTS

◦ In order for the book to not topple off of the table, we need its centre of
mass to lie above the table. That is, we need the x-coordinate of the centre
mass of B1 , which we shall denote X̄(B1 ), to obey

X̄(B1 ) ≤ 0

Assuming that our books have uniform density and are of length L, X̄(B1 )
will be exactly half way between the right hand end of the book, which is
at x = x1 , and the left hand end of the book, which is at x = x1 − L. So
1 1 L
X̄(B1 ) = x1 + (x1 − L) = x1 −
2 2 2
Thus book #1 does not topple off of the table provided
L
x1 ≤
2

• Now let’s put books #1 and #2 on the table, with the right hand edge of book #1
at x = x1 and the right hand edge of book #2 at x = x2 , as in the figure below.

B2
B1

x1 x2 x
0

◦ In order for book #2 to not topple off of book #1, we need the centre of
mass of book #2 to lie above book #1. That is, we need the x-coordinate of
the centre mass of B2 , which is X̄(B2 ) = x2 − L2 , to obey

L L
X̄(B2 ) ≤ x1 ⇐⇒ x2 − ≤ x1 ⇐⇒ x2 ≤ x1 +
2 2
◦ Assuming that book #2 does not topple off of book #1, we still need to
arrange that the pair of books does not topple off of the table. Think of the
pair of books as the combined red object in the figure

B2
B1

x1 x2 x
0

437
S EQUENCE AND SERIES 3.3 C ONVERGENCE T ESTS

In order for the combined red object to not topple off of the table, we need
the centre of mass of the combined red object to lie above the table. That
is, we need the x-coordinate of the centre mass of the combined red object,
which we shall denote X̄(B1 ∪ B2 ), to obey

X̄(B1 ∪ B2 ) ≤ 0

The centre of mass of the combined red object is the weighted average 6 of
the centres of mass of B1 and B2 . As B1 and B2 have the same weight,

1 1 1 L 1 L
X̄(B1 ∪ B2 ) = X̄(B1 ) + X̄(B2 ) = x1 − + x2 −
2 2 2 2 2 2
1 L
= (x1 + x2 ) −
2 2
and the combined red object does not topple off of the table if

1 L
X̄(B1 ∪ B2 ) = (x1 + x2 ) − ≤ 0 ⇐⇒ x1 + x2 ≤ L
2 2

In conclusion, our two-book tower survives if

L
x2 ≤ x1 + and x1 + x2 ≤ L
2
L
In particular we may choose x1 and x2 to satisfy x2 = x1 + 2
and x1 + x2 = L.
Then, substituting x2 = x1 + L2 into x1 + x2 = L gives
 L L L1 L 1
x1 + x1 + = L ⇐⇒ 2x1 = ⇐⇒ x1 = , x2 = 1+
2 2 2 2 2 2

• Before considering the general “n-book tower”, let’s now put books #1, #2 and
#3 on the table, with the right hand edge of book #1 at x = x1 , the right hand
edge of book #2 at x = x2 , and the right hand edge of book #3 at x = x3 , as in
the figure below.

B3
B2
B1

x
0 x1 x2 x3

6 It might be a good idea to review the beginning of §2.3 at this point.

438
S EQUENCE AND SERIES 3.3 C ONVERGENCE T ESTS

◦ In order for book #3 to not topple off of book #2, we need the centre of
mass of book #3 to lie above book #2. That is, we need the x-coordinate of
the centre mass of B3 , which is X̄(B3 ) = x3 − L2 , to obey
L L
X̄(B3 ) ≤ x2 ⇐⇒ x3 − ≤ x2 ⇐⇒ x3 ≤ x2 +
2 2
◦ Assuming that book #3 does not topple off of book #2, we still need to
arrange that the pair of books, book #2 plus book #3 (the red object in the
figure below), does not topple off of book #1.

B3
B2
B1

x
0 x1 x2 x3

In order for this combined red object to not topple off of book #1, we need
the x-coordinate of its centre mass, which we denote X̄(B2 ∪ B3 ), to obey
X̄(B2 ∪ B3 ) ≤ x1
The centre of mass of the combined red object is the weighted average of
the centre of masses of B2 and B3 . As B2 and B3 have the same weight,
1 1 1 L 1 L
X̄(B2 ∪ B3 ) = X̄(B2 ) + X̄(B3 ) = x2 − + x3 −
2 2 2 2 2 2
1 L
= (x2 + x3 ) −
2 2
and the combined red object does not topple off of book #1 if
1 L
(x2 + x3 ) − ≤ x1 ⇐⇒ x2 + x3 ≤ 2x1 + L
2 2
◦ Assuming that book #3 does not topple off of book #2, and also that the
combined book #2 plus book #3 does not topple off of book #1, we still
need to arrange that the whole tower of books, book #1 plus book #2 plus
book #3 (the red object in the figure below), does not topple off of the table.

B3
B2
B1

x
0 x1 x2 x3

439
S EQUENCE AND SERIES 3.3 C ONVERGENCE T ESTS

In order for this combined red object to not topple off of the table, we need
the x-coordinate of its centre mass, which we denote X̄(B1 ∪ B2 ∪ B3 ), to
obey
X̄(B1 ∪ B2 ∪ B3 ) ≤ 0
The centre of mass of the combined red object is the weighted average
of the centre of masses of B1 and B2 and B3 . As they all have the same
weight,

1 1 1
X̄(B1 ∪ B2 ∪ B3 ) = X̄(B1 ) + X̄(B2 ) + X̄(B3 )
3 3 3
1 L 1 L 1 L
= x1 − + x2 − + x3 −
3 2 3 2 3 2
1 L
= (x1 + x2 + x3 ) −
3 2
and the combined red object does not topple off of the table if

1 L 3L
(x1 + x2 + x3 ) − ≤ 0 ⇐⇒ x1 + x2 + x3 ≤
3 2 2

In conclusion, our three-book tower survives if

L 3L
x3 ≤ x2 + and x2 + x3 ≤ 2x1 + L and x1 + x2 + x3 ≤
2 2
In particular, we may choose x1 , x2 and x3 to satisfy

3L
x1 + x2 + x 3 = and
2
x2 + x3 = 2x1 + L and
L
x3 = + x2
2
Substituting the second equation into the first gives

3L L1
3x1 + L = =⇒ x1 =
2 2 3
Next substituting the third equation into the second, and then using the for-
mula above for x1 , gives

L L L1 1
2x2 + = 2x1 + L = + L =⇒ x2 = +
2 3 2 2 3
and finally

L L 1 1
x3 = + x2 = 1+ +
2 2 2 3

440
S EQUENCE AND SERIES 3.3 C ONVERGENCE T ESTS

• We are finally ready for the general “n-book tower”. Stack n books on the table,
with book B1 on the bottom and book Bn at the top, and with the right hand
edge of book #j at x = xj . The same centre of mass considerations as above
show that the tower survives if
L
X̄(Bn ) ≤ xn−1 xn − ≤ xn−1
2
1 L
X̄(Bn−1 ∪ Bn ) ≤ xn−2 (xn−1 + xn ) − ≤ xn−2
2 2
.. ..
. .
1 L
X̄(B3 ∪ · · · ∪ Bn ) ≤ x2 (x3 + · · · + xn ) − ≤ x2
n−2 2
1 L
X̄(B2 ∪ B3 ∪ · · · ∪ Bn ) ≤ x1 (x2 + x3 + · · · + xn ) − ≤ x1
n−1 2
1 L
X̄(B1 ∪ B2 ∪ B3 ∪ · · · ∪ Bn ) ≤ 0 (x1 + x2 + x3 + · · · + xn ) − ≤ 0
n 2
In particular, we may choose the xj ’s to obey

1 L
(x1 + x2 + x3 + · · · + xn ) =
n 2
1 L
(x2 + x3 + · · · + xn ) = + x1
n−1 2
1 L
(x3 + · · · + xn ) = + x2
n−2 2
.. ..
. .
1 L
(xn−1 + xn ) = + xn−2
2 2
L
xn = + xn−1
2
Substituting x2 + x3 + · · · + xn = (n − 1)x1 + L2 (n − 1) from the second equation
into the first equation gives
nx1
1nz }| { L o L L 1  L1
x1 + (n − 1)x1 + (n − 1) = =⇒ x1 + 1− =
n 2 2 2 n 2 2
L 1 
=⇒ x1 =
2 n
Substituting x3 + · · · + xn = (n − 2)x2 + L2 (n − 2) from the third equation into
the second equation gives
(n−1)x
2 (n−1)−1
1 nz }| { L z }| { o L L 1
x2 + (n − 2)x2 + ( n − 2 ) = + x1 = 1+
n−1 2 2 2 n
L 1  L 1
=⇒ x2 + 1− = 1+
2 n−1 2 n

441
 S EQUENCE AND SERIES 3.3 C ONVERGENCE T ESTS

L 1 1
=⇒ x2 = +
2 n−1 n
Just keep going. We end up with

L 1 
x1 =
2 n
L 1 1
x2 = +
2 n−1 n
L 1 1 1
x3 = + +
2 n−2 n−1 n
..
.
L1 1
xn−2 = + ··· +
2 3 n
L1 1 1
xn−1 = + + ··· +
2 2 3 n
L 1 1 1
xn = 1 + + + ··· +
2 2 3 n
1 1 1 th
Our overhang is xn = L2 1 + L


P2∞+ 31 + · · · + n . This is 2 times the n partial
sum of the harmonic series m=1 m . As we saw in Example 3.3.6 (the p test),
the harmonic series diverges. So, as n goes to infinity 1 + 21 + 13 + · · · + n1 also
goes to infinity. We may make the overhang as large 7 as we like!

3.3.8 tt Optional — The Root Test

There is another test that is very similar in spirit to the ratio test. It also comes from
a reexamination of the geometric series
∞
X ∞
X
an = arn
n=0 n=0

The ratio test was based on the observation that r, which largely determines whether
or not the series converges, could be found by computing the ratio r = an+1 /an . The
root test is based on the observation that |r| can also be determined by looking that
the nth root of the nth term with n very large:
q q
n
lim ar = |r| lim n a = |r|
n if a 6= 0
n→∞ n→∞

Of course, in general, the nth term is not exactly arn . However, if for very large n, the
nth term is approximately proportional to rn , with |r| given by the above limit, we
would expect the series to converge when |r| < 1 and diverge when |r| > 1. That is
indeed the case.

7 At least if our table is strong enough.

442
S EQUENCE AND SERIES 3.3 C ONVERGENCE T ESTS

Theorem 3.3.24 Root Test.

Assume that q
n
L = lim an
n→∞

exists or is +∞.
∞
P
a If L < 1, then an converges.
n=1

∞
P
b If L > 1, or L = +∞, then an diverges.
n=1

Warning 3.3.25

Beware that the root test provides absolutely no conclusion about the conver-
∞
P q
gence or divergence of the series an if lim n an = 1.
n=1 n→∞

Proof.
p (a) Pick any number R obeying L < R < 1. We are assuming that
n
|an | approaches
p L as n → ∞. In particular there must be some natural num-
ber M so that |an | ≤ R for all n ≥ M . So |an | ≤ Rn for all n ≥ M and the
n

∞ ∞
Rn
P P
series an converges by comparison to the geometric series
n=1 p n=1
(b) We are assuming that n |an | approaches L > 1 (or grows unboundedly)
p as
n → ∞. In particular there must be some natural number M so that |an | ≥
n

1 for all n ≥ M . So |an | ≥ 1 for all n ≥ M and the series diverges by the
divergence test.

√
P∞ (−3)n n+1 n
Example 3.3.26 n=1 2n+3
X .

We have already used the ratio test, in Example 3.3.23, to show that this series con-
verges when |X| < 13 and diverges when |X| > 13 . We’ll now use the root test to draw
the same conclusions.
√
(−3)n n+1 n
• Write an = 2n+3
X .

443
S EQUENCE AND SERIES 3.3 C ONVERGENCE T ESTS

• We compute
s √
p
n (−3)n n + 1 n
n
|an | = X
2n + 3

1 1
= 3|X| n + 1 2n 2n + 3)− n

1
• We’ll now show that the limit of n + 1 2n as n → ∞ is exactly 1. To do, so we
first compute the limit of the logarithm.



2n1 log n + 1
lim log n + 1 = lim now apply Theorem 3.1.6
n→∞ n→∞ 2n

log x + 1
= lim
x→∞ 2x
1
x+1
= lim by l’Hôpital
x→∞ 2
=0

So

2n1
1
= e0 = 1

lim n + 1 = lim exp log n + 1 2n
n→∞ n→∞

1
An essentially identical computation also gives that limn→∞ 2n + 3)− n = e0 =
1.

• So
p
n
lim |an | = 3|X|
n→∞

and the root test also tells us that if 3|X| > 1 the series diverges, while when 3|X| < 1
the series converges.
Example 3.3.26

We have done the last example once, in Example 3.3.23, using the ratio test and
once, in Example 3.3.26, using the root test. It was clearly much easier to use the
ratio test. Here is an example that is most easily handled by the root test.

444
 S EQUENCE AND SERIES 3.3 C ONVERGENCE T ESTS

P∞ n

n2
Example 3.3.27 n=1 n+1
.

n

n2
Write an = n+1
. Then
r
p
n n n n2  n n  1 −n
|an | = = = 1+
n+1 n+1 n
Now we take the limit,
 1 −n  1 −X
lim 1 + = lim 1 + by Theorem 3.1.6
n→∞ n X→∞ X

−1/x 1
= lim 1 + x where x =
x→0 X
−1
=e

by Example 3.7.20 in the CLP-1 text with a = −1. As the limit is strictly smaller than

n2
1, the series ∞ n
P
n=1 n+1 converges.
To draw the same conclusion using the ratio test, one would have to show that the
limit of
an+1  n + 1 (n+1)2  n + 1 n2
=
an n+2 n
as n → ∞ is strictly smaller than 1. It’s clearly better to stick with the root test.

3.3.9 tt Optional — Harmonic and Basel Series

ttt The Harmonic Series

3.3.9.1

The series
∞
X 1
n=1
n

that appeared in Warning 3.3.3, is called the Harmonic series 8 , and its partial sums
N
X 1
HN =
n=1
n

8 The interested reader should use their favourite search engine to read more on the link between
this series and musical harmonics. You can also find interesting links between the Harmonic
series and the so-called “jeep problem” and also the problem of stacking a tower of dominoes to
create an overhang that does not topple over.

445
 S EQUENCE AND SERIES 3.3 C ONVERGENCE T ESTS

are called the Harmonic numbers. Though these numbers have been studied at
least as far back as Pythagoras, the divergence of the series was first proved in
around 1350 by Nicholas Oresme (1320-5 – 1382), though the proof was lost for many
years and rediscovered by Mengoli (1626–1686) and the Bernoulli brothers (Johann
1667–1748 and Jacob 1655–1705).
Oresme’s proof is beautiful and all the more remarkable that it was produced
more than 300 years before calculus was developed by Newton and Leibnitz. It
starts by grouping the terms of the harmonic series carefully:
∞
X 1 1 1 1 1 1 1 1
= 1 + + + + + + + + ···
n=1
n 2 3 4 5 6 7 8
     
1 1 1 1 1 1 1 1 1 1 1
=1+ + + + + + + + + + ··· + + + ···
2 3 4 5 6 7 8 9 10 15 16
     
1 1 1 1 1 1 1 1 1 1 1
>1+ + + + + + + + + + ··· + + + ···
2 4 4 8 8 8 8 16 16 16 16
     
1 2 4 8
=1+ + + + + ···
2 4 8 16
So one can see that this is 1 + 12 + 21 + 12 + 12 + · · · and so must diverge 9 .
There are many variations on Oresme’s proof — for example, using groups of
two or three. A rather different proof relies on the inequality

ex > 1 + x for x > 0

which follows immediately from the Taylor series for ex given in Theorem 3.6.7.
From this we can bound the exponential of the Harmonic numbers:
1 1 1 1
eHn = e1+ 2 + 3 + 4 +···+ n
= e1 · e1/2 · e1/3 · e1/4 · · · e1/n
> (1 + 1) · (1 + 1/2) · (1 + 1/3) · (1 + 1/4) · · · (1 + 1/n)
2 3 4 5 n+1
= · · · ···
1 2 3 4 n
=n+1

Since eHn grows unboundedly with n, the harmonic series diverges.

ttt The Basel Problem

3.3.9.2

The problem of determining the exact value of the sum of the series
∞
X 1
n=1
n2

9 The grouping argument can be generalised further and the interested reader should look up
Cauchy’s condensation test.

446
S EQUENCE AND SERIES 3.3 C ONVERGENCE T ESTS

is called the Basel problem. The problem is named after the home town of Leonhard
Euler, who solved it. One can use telescoping series to show that this series must
converge. Notice that
1 1 1 1
2
< = −
n n(n − 1) n−1 n
Hence we can bound the partial sum:
k k
X 1 X 1
Sk = 2
<1+ avoid dividing by 0
n=1
n n=2
n(n − 1)
k  
X 1 1
=1+ − which telescopes to
n=2
n − 1 n
1
=1+1−
k
Thus, as k increases, the partial sum Sk increases (the series is a sum of positive
terms), but is always smaller than 2. So the sequence of partial sums converges.
Mengoli posed the problem of evaluating the series exactly in 1644 and it was
solved — not entirely rigorously — by Euler in 1734. A rigorous proof had to wait
another 7 years. Euler used some extremely cunning observations and manipula-
tions of the sine function to show that
∞
X 1 π2
= .
n=1
n2 6
He used the Maclaurin series
x3 x5
sin x = 1 − + − ···
6 24
and a product formula for sine
 x  x  x  x  x  x
sin x = x · 1 − · 1+ · 1− · 1+ · 1− · 1+ ···
π   π   2π 2π 3π 3π

x2 x2 x2
 (?)
=x· 1− · 1− · 1− ···
π 4π 9π
Extracting the coefficient of x3 from both expansions gives the desired result. The
proof of the product formula is well beyond the scope of this course. But notice that
at least the values of x which make the left hand side of (?) zero, namely x = nπ with
n integer, are exactly the same as the values of x which make the right hand side of
(?) zero 10 .
This approach can also be used to compute ∞ −2p
P
n=1 n for p = 1, 2, 3, · · · and show
11 2p
that they are rational multiples of π . The corresponding series of odd powers are
significantly nastier and getting closed form expressions for them remains a famous
open problem.

10 Knowing that the left and right hand sides of (?) are zero for the same values of x is far from
the end of the story. Two functions f (x) and g(x) having the same zeros, need not be equal. It
is certainly possible that f (x) = g(x) ∗ A(x) where A(x) is a function that is nowhere zero. The
interested reader should look up the Weierstrass factorisation theorem.
11 Search-engine your way to “Riemann zeta function”.

447
 S EQUENCE AND SERIES 3.3 C ONVERGENCE T ESTS

3.3.10 tt Optional — Some Proofs

In this optional section we provide proofs of two convergence tests. We shall repeat-
edly use the fact that any sequence a1 , a2 , a3 , · · ·, of real numbers which is increasing
(i.e. an+1 ≥ an for all n) and bounded (i.e. there is a constant M such that an ≤ M for
all n) converges. We shall not prove this fact 12 .
We start with the comparison test, and then move on to the alternating series test.

Theorem 3.3.28 The Comparison Test (stated again).

Let N0 be a natural number and let K > 0.

∞
P ∞
P
a If |an | ≤ Kcn for all n ≥ N0 and cn converges, then an converges.
n=0 n=0

∞
P ∞
P
b If an ≥ Kdn ≥ 0 for all n ≥ N0 and dn diverges, then an diverges.
n=0 n=0

P∞
Proof.
P∞ (a) By hypothesis n=0 cn converges. So it suffices to prove that
n=0 [Kcn − an ] converges, because then, by our Arithmetic of series Theorem
3.2.9,
X∞ X∞ X∞
an = Kcn − [Kcn − an ]
n=0 n=0 n=0

will converge too. But for all n ≥ N0 , Kcn − an ≥ 0 so that, for all N ≥ N0 , the
partial sums
XN
SN = [Kcn − an ]
n=0
N0
P
increase with N , but never gets bigger than the finite number [Kcn − an ] +
n=0
∞
P
K cn . So the partial sums SN converge as N → ∞.
n=N0 +1
(b) For all N > N0 , the partial sum
N
X N0
X N
X
SN = an ≥ an + K dn
n=0 n=0 n=N0 +1

By hypothesis, N
P
n=N0 +1 dn , and hence SN , grows without bound as N → ∞.
So SN → ∞ as N → ∞.


12 It is one way to state a property of the real number system called “completeness”. The interested
reader should use their favourite search engine to look up “completeness of the real numbers”.

448
S EQUENCE AND SERIES 3.3 C ONVERGENCE T ESTS

Theorem 3.3.29 Alternating Series Test (stated again).

 ∞
Let an n=1 be a sequence of real numbers that obeys

i an ≥ 0 for all n ≥ 1 and

ii an+1 ≤ an for all n ≥ 1 (i.e. the sequence is monotone decreasing) and

iii limn→∞ an = 0.

Then ∞
X
a1 − a2 + a3 − a4 + · · · = (−1)n−1 an = S
n=1

converges and, for each natural number N , S − SN is between 0 and (the first
dropped term) (−1)N aN +1 . Here SN is, as previously, the N th partial sum
N
(−1)n−1 an .
P
n=1

Proof. Let 2n be an even natural number. Then the 2nth partial sum obeys
≥0 ≥0 ≥0
z }| { z }| { z }| {
S2n = (a1 − a2 ) + (a3 − a4 ) + · · · + (a2n−1 − a2n )
≥0 ≥0 ≥0 ≥0
z }| { z }| { z }| { z }| {
≤ (a1 − a2 ) + (a3 − a4 ) + · · · + (a2n−1 − a2n ) + (a2n+1 − a2n+2 )
= S2(n+1)

and
≥0 ≥0 ≥0 ≥0
z }| { z }| { z }| { z}|{
S2n = a1 − (a2 − a3 ) − (a4 − a5 ) − · · · − (a2n−2 − a2n−1 ) − a2n
≤ a1

So the sequence S2 , S4 , S6 , · · · of even partial sums is a bounded, increasing

sequence and hence converges to some real number S. Since S2n+1 = S2n +a2n+1
and a2n+1 converges zero as n → ∞, the odd partial sums S2n+1 also converge
to S. That S − SN is between 0 and (the first dropped term) (−1)N aN +1 was
already proved in §3.3.4.

449
 S EQUENCE AND SERIES 3.3 C ONVERGENCE T ESTS

3.3.11 tt Exercises

Exercises — Stage 1

1. Select the series below that diverge by the divergence test.

∞
X 1
(A)
n=1
n
∞
X n2
(B)
n=1
n+1
∞
X
(C) sin n
n=1
∞
X
(D) sin(πn)
n=1

2. Select the series below whose terms satisfy the conditions to apply the
integral test.
∞
X 1
(A)
n=1
n
∞
X n2
(B)
n=1
n+1
X∞
(C) sin n
n=1
∞
X sin n + 1
(D)
n=1
n2

3. Suppose there is some threshold after which a person is considered old, and
before which they are young.
Let Olaf be an old person, and let Yuan be a young person.

a Suppose I am older than Olaf. Am I old?

b Suppose I am younger than Olaf. Am I old?

c Suppose I am older than Yuan. Am I young?

d Suppose I am younger than Yuan. Am I young?

4. Below are graphs of two sequences with positive terms. Assume the se-
quences continue as shown. Fill in the table with conclusions that can be
made from the direct comparison test, if any.

450
S EQUENCE AND SERIES 3.3 C ONVERGENCE T ESTS

nx

P P
if an converges if an diverges
P P
and if {an } is the red series then b then b
P n P n
and if {an } is the blue series then bn then bn

5. For each pair of series below, decide whether the second series is a valid
comparison series to determine the convergence of the first series, using
the direct comparison test and/or the limit comparison test.
∞ ∞
X 1 X 1
a , compared to the divergent series .
n=10
n−1 n=10
n
∞ ∞
X sin n X 1
b 2+1
, compared to the convergent series 2
.
n=1
n n=1
n
∞ ∞
X n3 + 5n + 1 X 1
c , compared to the convergent series .
n=5
n6 − 2 n=5
n3
∞ ∞
X 1 X 1
d √ , compared to the divergent series √ .
n=5
n n=5
4
n

∞
1 X
6. Suppose an is a sequence with lim an = . Does an converge or diverge,
n→∞ 2 n=7
or is it not possible to determine this from the information given? Why?

7. What flaw renders the following reasoning invalid?

∞
X sin n
Q: Determine whether converges or diverges.
n=1
n
sin n
A: First, we will evaluate lim .
n→∞ n

451
S EQUENCE AND SERIES 3.3 C ONVERGENCE T ESTS

−1 sin n 1
• Note ≤ ≤ for n ≥ 1.
n n n
−1 1
• Note also that lim = lim = 0.
n→∞ n n→∞ n
sin n
• Therefore, by the Squeeze Theorem, lim = 0 as
n→∞ n
well.
∞
X sin n
So, by the divergence test, converges.
n=1
n

8. What flaw renders the following reasoning invalid?

∞
X
Q: Determine whether (sin(πn) + 2) converges or diverges.
n=1

A: We use the integral test. Let f (x) = sin(πx) + 2. Note f (x)

is always positive, since sin(x) + 2 ≥ −1 + 2 = 1. Also, f (x) is
continuous.
Z ∞ Z b
[sin(πx) + 2]dx = lim [sin(πx) + 2]dx
1 b→∞ 1
" #
b
1
= lim − cos(πx) + 2x
b→∞ π 1
 
1 1
= lim − cos(πb) + 2b + (−1) − 2
b→∞ π π
=∞

By the integral test, since the integral diverges, also

X∞
(sin(πn) + 2) diverges.
n=1

9. What flaw renders the following reasoning invalid?

∞
X 2n+1 n2
Q: Determine whether the series n + 2n
converges or di-
n=1
e
verges.
∞
X 2n+1
A: We want to compare this series to the series . Note
n=1
en
both this series and the series in the question have positive
terms.
2n+1 n2 2n+1
First, we find that n > n when n is sufficiently large.
e + 2n e
The justification for this claim is as follows:
• We note that en (n2 −1) > n2 −1 > 2n for n sufficiently large.

452
S EQUENCE AND SERIES 3.3 C ONVERGENCE T ESTS

• Therefore, en · n2 > en + 2n
• Therefore, 2n+1 · en · n2 > 2n+1 (en + 2n)
• Since en + 2n and en are both expressions that work out to
be positive for the values of n under consideration, we can
divide both sides of the inequality by these terms without
2n+1 n2 2n+1
having to flip the inequality. So, n > n .
e + 2n e
∞
X 2n+1
Now, we claim converges.
n=1
en
∞ ∞ ∞  n
X 2n+1 X 2n X 2
Note =2 =2 . This is a geometric se-
n=1
en n=1
en n=1
e
ries with r = 2e . Since 2/e < 1, the series converges.
Now, by the Direct Comparison Test, we conclude that
∞
X 2n+1 n2
n + 2n
converges.
n=1
e
10. Which of the series below are alternating?
∞
X
(A) sin n
n=1
∞
X cos(πn)
(B)
n=1
n3
∞
X 7
(C)
n=1
(−n)2n
∞
X (−2)n
(D)
n=1
3n+1

11. Give an example of a convergent series for which the ratio test is incon-
clusive.

12. Imagine you’re taking an exam, and you momentarily forget exactly how
the inequality in the ratio test works. You remember there’s a ratio, but you
don’t remember which term goes on top; you remember there’s something
about the limit being greater than or less than one, but you don’t remember
which way implies convergence.
Explain why
an+1
lim >1
n→∞ an
or, equivalently,
an
lim <1
n→∞ an+1

453
S EQUENCE AND SERIES 3.3 C ONVERGENCE T ESTS

∞
P
should mean that the sum an diverges (rather than converging).
n=1
∞
X
13. Give an example of a series an , with a function f (x) such that f (n) = an
n=a
for all whole numbers n, such that:
Z ∞
• f (x) dx diverges, while
a
∞
X
• an converges.
n=a

14. *. Suppose that you want to use the Limit Comparison Test on the series
∞
X 2n + n
an where an = n .
n=0
3 +1
an
Write down a sequence {bn } such that lim exists and is nonzero. (You
n→∞ bn
don’t have to carry out the Limit Comparison Test)

15. *. Decide whether each of the following statements is true or false. If

false, provide a counterexample. If true provide a brief justification.

∞
P
a If lim an = 0, then an converges.
n→∞ n=1

∞
(−1)n an converges.
P
b If lim an = 0, then
n→∞ n=1

∞
P ∞
P
c If 0 ≤ an ≤ bn and bn diverges, then an diverges.
n=1 n=1

Exercises — Stage 2
∞
X n2
16. *. Does the series √ converge?
n=2
3n2 + n
∞
X 5k
17. *. Determine, with explanation, whether the series 4k + 3k
converges or
n=1
diverges.

∞
X 1
18. *. Determine whether the series 1 is convergent or divergent. If
n=0
n+ 2
it is convergent, find its value.

454
S EQUENCE AND SERIES 3.3 C ONVERGENCE T ESTS

∞
X 1
19. Does the following series converge or diverge? √ √
k=1
k k+1

∞
X
20. Evaluate the following series, or show that it diverges: 3(1.001)k .
k=30

∞  n
X −1
21. Evaluate the following series, or show that it diverges: .
n=3
5

∞
X
22. Does the following series converge or diverge? sin(πn)
n=7
∞
X
23. Does the following series converge or diverge? cos(πn)
n=7

∞
X ek
24. Does the following series converge or diverge? .
k=1
k!

∞
X 2k
25. Evaluate the following series, or show that it diverges: k+2
.
k=0
3
∞
X n!n!
26. Does the following series converge or diverge? .
n=1
(2n)!

∞
X n2 + 1
27. Does the following series converge or diverge? .
n=1
2n4 + n

∞
X 5
28. *. Show that the series n(log n)3/2
converges.
n=3

∞
X 1
29. *. Find the values of p for which the series n(log n)p
converges.
n=2

∞ √
X e− n
30. *. Does √ converge or diverge?
n=1
n

31. *. Use the comparison test (not the limit comparison test) to show whether

455
S EQUENCE AND SERIES 3.3 C ONVERGENCE T ESTS

the series ∞ √
X 3n2 − 7
n=2
n3
converges or diverges.
∞ √
3
X k4 + 1
32. *. Determine whether the series √ converges.
k=1
k5 + 9

∞
X n4 2n/3
33. *. Does 4
converge or diverge?
n=1
(2n + 7)

34. *. Determine, with explanation, whether each of the following series con-
verge or diverge.

∞
X 1
a √
n=1
n2 + 1
∞
X n cos(nπ)
b
n=1
2n
35. *. Determine whether the series
∞
X k 4 − 2k 3 + 2
k=1
k5 + k2 + k

converges or diverges.
36. *. Determine whether each of the following series converge or diverge.

∞
X n2 + n + 1
a
n=2
n5 − n
∞ √
X 3m + sin m
b
m=1
m2
∞
X 1
37. Evaluate the following series, or show that it diverges: n
.
n=5
e
∞
X 6
38. *. Determine whether the series 7n
is convergent or divergent. If it is
n=2
convergent, find its value.
39. *. Determine, with explanation, whether each of the following series con-
verge or diverge.

456
S EQUENCE AND SERIES 3.3 C ONVERGENCE T ESTS

a 1 + 13 + 15 + 17 + 91 + · · ·.
∞
X 2n + 1
b
n=1
22n+1
40. *. Determine, with explanation, whether each of the following series con-
verges or diverges.

∞ √
3
X k
a .
k=2
k2 −k
∞
X k 10 10k (k!)2
b .
k=1
(2k)!
∞
X 1
c .
k=3
k(log k)(log log k)
∞
X n3 − 4
41. *. Determine whether the series 2n5 − 6n
is convergent or divergent.
n=1
N
X (−1)n
42. *. What is the smallest value of N such that the partial sum n · 10n
n=1
∞
X (−1)n
approximates n
within an accuracy of 10−6 ?
n=1
n · 10

∞
X (−1)n−1 π2
43. *. It is known that = (you don’t have to show this). Find
n=1
n2 12
2
th
N so that SN , the N partial sum of the series, satisfies | π12 − SN | ≤ 10−6 .
Be sure to say why your method can be applied to this particular series.

∞
X (−1)n+1
44. *. The series 2
converges to some number S (you don’t have to
n=1
(2n + 1)
prove this). According to the Alternating Series Estimation Theorem, what
is the smallest value of N for which the N th partial sum of the series is at
1
most 100 away from S? For this value of N , write out the N th partial sum of
the series.

Exercises — Stage 3 A number of phenomena roughly follow a distribution called

Zipf’s law. We discuss some of these in Questions 52 and 53.
45. *. Determine, with explanation, whether the following series converge or
diverge.

457
S EQUENCE AND SERIES 3.3 C ONVERGENCE T ESTS

∞
X nn
a
n=1
9n n!
∞
X 1
b
n=1
nlog n

∞
x + sin x
Z
46. *. (a) Prove that dx diverges.
2 1 + x2
∞
X n + sin n
(b) Explain why you cannot conclude that diverges from
n=1
1 + n2
part (a) and the Integral Test.
∞
X n + sin n
(c) Determine, with explanation, whether converges or di-
n=1
1 + n2
verges.

∞ √
X e− n
47. *. Show that √ converges and find an interval of length 0.05 or less
n=1
n
that contains its exact value.
∞
X
48. *. Suppose that the series an converges and that 1 > an ≥ 0 for all n.
n=1
∞
X an
Prove that the series also converges.
n=1
1 − an

∞
P
49. *. Suppose that the series (1 − an ) converges, where an > 0 for
n=0
∞
2n an converges or
P
n = 0, 1, 2, 3, · · ·. Determine whether the series
n=0
diverges.

∞
X nan − 2n + 1
50. *. Assume that the series converges, where an > 0 for
n=1
n+1
n = 1, 2, · · ·. Is the following series
∞
X  a 
n
− log a1 + log
n=1
an+1

convergent? If your answer is NO, justify your answer. If your answer is

∞
an
P

YES, evaluate the sum of the series − log a1 + log an+1 .
n=1

458
 S EQUENCE AND SERIES 3.4 A BSOLUTE AND C ONDITIONAL C ONVERGENCE

∞
X
51. *. Prove that if an ≥ 0 for all n and if the series an converges, then
n=1
∞
X
the series a2n also converges.
n=1

52. Suppose the frequency of word use in a language has the following pattern:
α
The n-th most frequently used word accounts for percent of
n
the total words used.

So, in a text of 100 words, we expect the most frequently used word to ap-
pear α times, while the second-most-frequently used word should appear
about α2 times, and so on.
If books written in this language use 20, 000 distinct words, then the most
commonly used word accounts for roughly what percentage of total words
used?
53.
Suppose the sizes of cities in a country adhere to the following pattern: if
the largest city has population α, then the n-th largest city has population
α
n
.
If the largest city in this country has 2 million people and the smallest city
6
has 1 person, then the population of the entire country is 2×10 2×106
P
n=1 n
. (For
2×106
many n’s in this sum n is not an integer. Ignore that.) Evaluate this sum
approximately, with an error of no more than 1 million people.

3.4q Absolute and Conditional Convergence

We have now seen examples of series that converge and of series that diverge. But
we haven’t really discussed how robust the convergence of series is — that is, can
we tweak the coefficients in some way while leaving the convergence unchanged. A
good example of this is the series
∞  n
X 1
n=1
3

This is a simple geometric series and we know it converges. We have also seen, as
examples 3.3.20 and 3.3.21 showed us, that we can multiply or divide the nth term
by n and it will still converge. We can even multiply the nth term by (−1)n (making
it an alternating series), and it will still converge. Pretty robust.
On the other hand, we have explored the Harmonic series and its relatives quite

459
 S EQUENCE AND SERIES 3.4 A BSOLUTE AND C ONDITIONAL C ONVERGENCE

a lot and we know it is much more delicate. While

∞
X 1
n=1
n

diverges, we also know the following two series converge:

∞ ∞
X 1 X 1
(−1)n .
n=1
n1.00000001 n=1
n

This suggests that the divergence of the Harmonic series is much more delicate. In
this section, we discuss one way to characterise this sort of delicate convergence —
especially in the presence of changes of sign.

3.4.1 tt Definitions

Definition 3.4.1 Absolute and conditional convergence.

∞
P ∞
P
a A series an is said to converge absolutely if the series |an | con-
n=1 n=1
verges.
∞
P ∞
P ∞
P
b If an converges but |an | diverges we say that an is conditionally
n=1 n=1 n=1
convergent.

If you consider these definitions for a moment, it should be clear that absolute
convergence
P is a stronger condition than just simple convergence. AllP the terms in
n |a n | are forced
P to be positive (by the absolute
P value signs), so that n |an | must
be bigger than n an — making it easier for n |an | to diverge. This is formalised by
the following theorem, which is an immediate consequence of the comparison test,
Theorem 3.3.8.a, with cn = |an |.

Theorem 3.4.2 Absolute convergence implies convergence.

∞
P ∞
P
If the series |an | converges then the series an also converges. That is,
n=1 n=1
absolute convergence implies convergence.

Recall that some of our convergence tests (for example, the integral test) may only
be applied to series with positive terms. Theorem 3.4.2 opens up the possibility of
applying “positive only” convergence tests to series whose terms are not all positive,
by checking for “absolute convergence” rather than for plain “convergence”.

460
 S EQUENCE AND SERIES 3.4 A BSOLUTE AND C ONDITIONAL C ONVERGENCE

P∞ n−1 1
Example 3.4.3 n=1 (−1) n
.
∞
(−1)n−1 n1 of Example 3.3.15 converges (by the al-
P
The alternating harmonic series
n=1
∞
1
P
ternating series test). But the harmonic series n
of Example 3.3.6 diverges (by the
n=1
∞
(−1)n−1 n1 converges condition-
P
integral test). So the alternating harmonic series
n=1
ally.

P∞ n−1 1
Example 3.4.4 n=1 (−1) n2
.

P∞ ∞
(−1)n−1 n12 1
P
Because the series n=1 = n2
of Example 3.3.6 converges (by the
n=1
∞
(−1)n−1 n12 converges absolutely, and hence converges.
P
integral test), the series
n=1

Example 3.4.5 Random signs.

Imagine flipping a coin infinitely many times. Set σn = +1 ifPthe nth flip comes up
∞
heads and σn = −1 if the n flip comes up tails. The series n=1 (−1)σn n12 is not in
th
∞
general an alternating series. But we know that the series ∞ σn 1 1
P P
n=1 (−1) n2 = n2
n=1
converges. So ∞ σn 1
P
n=1 (−1) n2
converges absolutely, and hence converges.

3.4.2 tt Optional — The delicacy of conditionally convergent series

Conditionally convergent series have to be treated with great care. For example,
switching the order of the terms in a finite sum does not change its value.

1+2+3+4+5+6=6+3+5+2+4+1

The same is true for absolutely convergent series. But it is not true for conditionally
convergent series. In fact by reordering any conditionally convergent series, you
can make it add up to any number you like, including +∞ and −∞. This very
strange result is known as Riemann’s rearrangement theorem, named after Bernhard
Riemann (1826–1866). The following example illustrates the phenomenon.

461
S EQUENCE AND SERIES 3.4 A BSOLUTE AND C ONDITIONAL C ONVERGENCE

Example 3.4.6 The alternating Harmonic series.

The alternating Harmonic series

∞
X 1
(−1)n−1
n=1
n

is a very good example of conditional convergence. We can show, quite explicitly,

how we can rearrange the terms to make it add up to two different numbers. Later, in
Example 3.5.20, we’ll show that this series is equal to log 2. However, by rearranging
the terms we can make it sum to 12 log 2. The usual order is

1 1 1 1 1 1
− + − + − + ···
1 2 3 4 5 6
For the moment think of the terms being paired as follows:
     
1 1 1 1 1 1
− + − + − + ···
1 2 3 4 5 6

so the denominators go odd-even odd-even. Now rearrange the terms so the de-
nominators are odd-even-even odd-even-even:
     
1 1 1 1 1 1 1 1
1− − + − − + − − + ···
2 4 3 6 8 5 10 12

Now notice that the first term in each triple is exactly twice the second term. If we
now combine those terms we get
     

1 − 1 − 1  +  1 − 1 − 1  +  1 − 1 − 1  + · · ·
     
| {z 2} 4
   3 6 8   5 10 12 
| {z } | {z }
=1/2 =1/6 =1/10
     
1 1 1 1 1 1
= − + − + − + ···
2 4 6 8 10 12

We can now extract a factor of 21 from each term, so

     
1 1 1 1 1 1 1 1 1
= − + − + − + ···
2 1 2 2 3 4 2 5 6
      
1 1 1 1 1 1 1
= − + − + − + ···
2 1 2 3 4 5 6

So by rearranging the terms, the sum of the series is now exactly half the original
sum!

462
S EQUENCE AND SERIES 3.4 A BSOLUTE AND C ONDITIONAL C ONVERGENCE

In fact, we can go even further, and show how we can rearrange the terms of the
alternating harmonic series to add up to any given number 1 . For the purposes of
the example we have chosen 1.234, but it could really be any number. The example
below can actually be formalised to give a proof of the rearrangement theorem.

Example 3.4.7 Reorder summands to get 1.234.

∞
(−1)n−1 n1 so that it
P
We’ll show how to reorder the conditionally convergent series
n=1
adds up to exactly 1.234 (but the reader should keep in mind that any fixed number
will work).
• First create two lists of numbers — the first list consisting of the positive terms
of the series, in order, and the second consisting of the negative numbers of the
series, in order.
1 1 1 1 1 1
1, , , , ··· and − , − , − , ···
3 5 7 2 4 6
• Notice that that if we add together the numbers in the second list,we get
1h 1 1 i
− 1 + + + ···
2 2 3
which is just − 12 times the harmonic series. So the numbers in the second list
add up to −∞.
Also, if we add together the numbers in the first list, we get
1 1 1 1 1 1 1
1+ + + ··· which is greater than + + + + ···
3 5 7 2 4 6 8
That is, the sum of the first set of numbers must be bigger than the sum of the
second set of numbers (which is just −1 times the second list). So the numbers
in the first list add up to +∞.
• Now we build up our reordered series. Start by moving just enough numbers
from the beginning of the first list into the reordered series to get a sum bigger
than 1.234.
1
1 + = 1.3333
3
We know that we can do this, because the sum of the terms in the first list
diverges to +∞.
• Next move just enough numbers from the beginning of the second list into the
reordered series to get a number less than 1.234.
1 1
1+ − = 0.8333
3 2

1 This is reminiscent of the accounting trick of pushing all the company’s debts off to next year so
that this year’s accounts look really good and you can collect your bonus.

463
 S EQUENCE AND SERIES 3.4 A BSOLUTE AND C ONDITIONAL C ONVERGENCE

Again, we know that we can do this because the sum of the numbers in the
second list diverges to −∞.

• Next move just enough numbers from the beginning of the remaining part of
the first list into the reordered series to get a number bigger than 1.234.

1 1 1 1 1
1+ − + + + = 1.2873
3 2 5 7 9
Again, this is possible because the sum of the numbers in the first list diverges.
Even though we have already used the first few numbers, the sum of the rest
of the list will still diverge.

• Next move just enough numbers from the beginning of the remaining part of
the second list into the reordered series to get a number less than 1.234.

1 1 1 1 1 1
1+ − + + + − = 1.0373
3 2 5 7 9 4

• At this point the idea is clear, just keep going like this. At the end of each step,
the difference between the sum and 1.234 is smaller than the magnitude of the
first unused number in the lists. Since the numbers in both lists tend to zero
as you go farther and farther up the list, this procedure will generate a series
whose sum is exactly 1.234. Since in each step we remove at least one number
from a list and we alternate between the two lists, the reordered series will
∞
(−1)n−1 n1 , with each term appearing exactly
P
contain all of the terms from
n=1
once.
Example 3.4.7

3.4.3 tt Exercises

Exercises — Stage 1
1. *. Decide whether the following statement is true or false. If false, provide
a counterexample. If true provide a brief justification.
∞
X ∞
X
n+1
• If (−1) bn converges, then bn also converges.
n=1 n=1

∞
X ∞
X ∞
X
2. Describe the sequence an based on whether an and |an | con-
n=1 n=1 n=1
verge or diverge, using vocabulary from this section where possible.

464
S EQUENCE AND SERIES 3.4 A BSOLUTE AND C ONDITIONAL C ONVERGENCE

P P
an converges an diverges
P
|a | converges
P n
|an | diverges

Exercises — Stage 2

∞
X (−1)n
3. *. Determine whether the series is absolutely convergent,
n=1
9n + 5
conditionally convergent, or divergent; justify your answer.

∞
X (−1)2n+1
4. *. Determine whether the series 1+n
is absolutely convergent,
n=1
conditionally convergent, or divergent.

∞
X
n−1 1 + 4n
5. *. The series (−1) either:
n=1
3 + 22n

• converges absolutely;

• converges conditionally;

• diverges;

• or none of the above.

Determine which is correct.

∞ √
X n cos n
6. *. Does the series n2 − 1
converge conditionally, converge abso-
n=5
lutely, or diverge?

∞
X n2 − sin n
7. *. Determine (with justification!) whether the series n6 + n2
con-
n=1
verges absolutely, converges but not absolutely, or diverges.

∞
X (−1)n (2n)!
8. *. Determine (with justification!) whether the series
n=0
(n2 + 1)(n!)2
converges absolutely, converges but not absolutely, or diverges.

465
S EQUENCE AND SERIES 3.4 A BSOLUTE AND C ONDITIONAL C ONVERGENCE

∞
(−1)n X
9. *. Determine (with justification!) whether the series n(log n)101
con-
n=2
verges absolutely, converges but not absolutely, or diverges.

∞
X sin n
10. Show that the series converges.
n2
n=1
∞  n
X sin n 1
11. Show that the series − converges.
n=1
4 8
∞
X sin2 n − cos2 n + 1
2
12. Show that the series converges.
n=1
2n

Exercises — Stage 3
∞
3
X
13. *. Both parts of this question concern the series S = (−1)n−1 24n2 e−n .
n=1

a Show that the series S converges absolutely.

b Suppose that you approximate the series S by its fifth partial sum S5 .
Give an upper bound for the error resulting from this approximation.

14. You may assume without proof the following:

∞
X (−1)n
= cos(1)
n=0
(2n)!

Using this fact, approximate cos 1 as a rational number, accurate to

1
within 1000 .
Check your answer against a calculator’s approximation of cos(1): what
was your actual error?

15. Let an be defined as

(
−en/2 if n is prime
an =
n2 if n is not prime
∞
X an
Show that the series converges.
n=1
en

466
 S EQUENCE AND SERIES 3.5 P OWER S ERIES

3.5q Power Series

Let’s return to the simple geometric series

∞
X
xn
n=0

where x is some real number. As we have seen (back in Example 3.2.4 and Lemma 3.2.5),
for |x| < 1 this series converges to a limit, that varies with x, while for |x| ≥ 1 the
series diverges. Consequently we can consider this series to be a function of x
∞
X
f (x) = xn on the domain |x| < 1.
n=0

Furthermore (also from Example 3.2.4 and Lemma 3.2.5) we know what the function
is.
∞
X 1
f (x) = xn = .
n=0
1−x

Hence we can consider the series ∞ n

P
n=0 x as a new way of representing the function
1
1−x
when |x| < 1. This series is an example of a power series.
1
Of course, representing a function as simple as 1−x by a series doesn’t seem like it
is going to make life easier. However the idea of representing a function by a series
turns out to be extremely helpful. Power series turn out to be very robust mathemat-
ical objects and interact very nicely with not only standard arithmetic operations,
but also with differentiation and integration (see Theorem 3.5.13). This means, for
example, that
  ∞
d 1 d X n
= x provided |x| < 1
dx 1−x dx n=0
∞
X d n
= x just differentiate term by term
n=0
dx
∞
X
= nxn−1
n=0

and in a very similar way

∞
1
Z Z X
dx = xn dx provided |x| < 1
1−x n=0
∞ Z
X
= xn dx just integrate term by term
n=0

467
 S EQUENCE AND SERIES 3.5 P OWER S ERIES

∞
X 1
=C+ xn+1
n=0
n+1

We are hiding some mathematics under the word “just” in the above, but you can
see that once we have a power series representation of a function, differentiation and
integration become very straightforward.
So we should set as our goal for this section, the development of machinery to
define and understand power series. This will allow us to answer questions 1 like
∞
x
X xn
Is e = ?
n=0
n!

Our starting point (now that we have equipped ourselves with basic ideas about
series), is the definition of power series.

3.5.1 tt Definitions

Definition 3.5.1
A series of the form
∞
X
2 3
A0 + A1 (x − c) + A2 (x − c) + A3 (x − c) + · · · = An (x − c)n
n=0

is called a power series in (x − c) or a power series centered on c. The numbers An

are called the coefficients of the power series.
One often considers power series centered on c = 0 and then the series reduces
to ∞
X
2 3
A0 + A1 x + A2 x + A3 x + · · · = A n xn
n=0

For example ∞ xn 1
P
n=0 n! is the power series with c = 0 and An = n! . Typically, as in
that case, the coefficients An are given fixed numbers, but the “x” is to be thought of
as a variable. Thus each power series is really a whole family of series — a different
series for each value of x.
One possible value of x is x = c and then the series reduces 2 to
∞
X ∞
X
n
An (x − c) = An (c − c)n
x=c
n=0 n=0
= A0 + |{z}
0 + |{z}
0 + |{z}
0 +···
|{z}
n=0 n=1 n=2 n=3

1 Recall that n! = 1 × 2 × 3 × · · · × n is called “n factorial”. By convention 0! = 1.

2 By convention, when the term (x − c)0 appears in a power series, it has value 1 for all values of x,
even x = c.

468
S EQUENCE AND SERIES 3.5 P OWER S ERIES

and so simply converges to A0 .

We now know that a power series converges when x = c. We can now use our
convergence tests to determine for what other values of x the series P converges. Per-
haps most straightforward is the ratio test. The n term in the series ∞
th
n=0 An (x − c)
n

is an = An (x − c)n . To apply the ratio test we need to compute the limit

an+1 An+1 (x − c)n+1
lim = lim
n→∞ an n→∞ An (x − c)n
An+1
= lim · |x − c|
n→∞ An
An+1
= |x − c| · lim .
n→∞ An
When we do so there are several possible outcomes.
• If the limit of ratios exists and is nonzero
An+1
lim = A 6= 0,
n→∞ An
then the ratio test says that the series ∞ n
P
n=0 An (x − c)
1
◦ converges when A · |x − c| < 1, i.e. when |x − c| < A
, and
1
◦ diverges when A · |x − c| > 1, i.e. when |x − c| > A
.
Because of this, when the limit exists, the quantity

Equation 3.5.2 Radius of convergence.

 −1
1 An+1
R= = lim
A n→∞ An

is called the radius of convergence of the series 3 .

• If the limit of ratios exists and is zero
An+1
lim =0
n→∞ An
then limn→∞ AAn+1n
|x − c| = 0 for every x and the ratio test tells us that the
P∞
series n=0 An (x − c)n converges for every number x. In this case we say that
the series has an infinite radius of convergence.

3 The use of the word “radius” might seem a little odd here, since we are really describing the
interval in the real line where the series converges. However, when one starts to consider power
series over complex numbers, the radius of convergence does describe a circle inside the complex
plane and so “radius” is a more natural descriptor.

469
S EQUENCE AND SERIES 3.5 P OWER S ERIES

• If the limit of ratios diverges to +∞

An+1
lim = +∞
n→∞ An

then limn→∞ AAn+1 |x − c| = +∞ for every x 6= c. The ratio test then tells us that
P∞ n
the series n=0 An (x − c)n diverges for every number x 6= c. As we have seen
above, when x = c, the series reduces to A0 + 0 + 0 + 0 + 0 + · · ·, which of course
converges. In this case we say that the series has radius of convergence zero.

• If AAn+1
n
does not approach a limit as n → ∞, then we learn nothing from the
ratio test and we must use other tools to understand the convergence of the
series.

All of these possibilities do happen. We give an example of each below. But first,
the concept of “radius of convergence” is important enough to warrant a formal
definition.
Definition 3.5.3

a Let 0 < R < ∞. If ∞ n

P
n=0 An (x−c) converges for |x−c| < R, and diverges
for |x − c| > R, then we say that the series has radius of convergence R.

b If ∞ n
P
n=0 An (x − c) converges for every number x, we say that the series
has an infinite radius of convergence.

c If ∞ n
P
n=0 An (x − c) diverges for every x 6= c, we say that the series has
radius of convergence zero.

Example 3.5.4 Finite nonzero radius of convergence.

∞
axn converges when |x| < 1
P
We already know that, if a 6= 0, the geometric series
n=0
and diverges when |x| ≥ 1. So, in the terminology of Definition 3.5.3, the geometric
series has radius of convergence R = 1. As a consistency check, we can also compute
∞
axn has An = a. So
P
R using 3.5.2. The series
n=0

 −1
An+1 h i−1
R = lim = lim 1 =1
n→∞ An n→∞

as expected.

470
S EQUENCE AND SERIES 3.5 P OWER S ERIES

Example 3.5.5 Radius of convergence = +∞.

∞
xn 1
P
The series n!
has An = n!
. So
n=0

1
An+1 (n+1)!
lim = lim 1
n→∞ An n→∞
n!
n!
= lim
n→∞ (n + 1)!
1 × 2 × 3 × ··· × n
= lim
n→∞ 1 × 2 × 3 × · · · × n × (n + 1)
1
= lim
n→∞ n + 1

=0
∞
xn
P
and n!
has radius of convergence ∞. It converges for every x.
n=0

Example 3.5.6 Radius of convergence = 0.

∞
n!xn has An = n!. So
P
The series
n=0

An+1 (n + 1)! 1 × 2 × 3 × 4 × · · · × n × (n + 1)
lim = lim = lim
n→∞ An n→∞ n! n→∞ 1 × 2 × 3 × 4 × ··· × n
= lim (n + 1)
n→∞
= +∞
∞
n!xn has radius of convergence zero a . It converges only for x = 0, where it
P
and
n=0
takes the value 0! = 1.

a Because of this, it might seem that such a series is fairly pointless. However there are all sorts of
mathematical games that can be played with them without worrying about their convergence.
Such “formal” power series can still impart useful information and the interested reader is in-
vited to look up “generating functions” with their preferred search engine.

471
S EQUENCE AND SERIES 3.5 P OWER S ERIES

Example 3.5.7 An awkward series to test.

Comparing the series

1 + 2x +x2 +2x3 +x4 +2x5 + · · ·

to
∞
X
An xn =A0 +A1 x+A2 x2 +A3 x3 +A4 x4 +A5 x5 + · · ·
n=1

we see that

A0 = 1 A1 = 2 A2 = 1 A3 = 2 A4 = 1 A5 = 2 ···

so that
A1 A2 1 A3 A4 1 A5
=2 = =2 = =2 ···
A0 A1 2 A2 A3 2 A4

and AAn+1
n
does not converge as n → ∞. Since the limit of the ratios does not exist,
we cannot tell anything from the ratio test. Nonetheless, we can still figure out for
which x’s our power series converges.

• Because every coefficient An is either 1 or 2, the nth term in our series obeys

An xn ≤ 2|x|n

and so is smaller than the nth term in the geometric series ∞ n

P
n=0 2|x| . This
geometric series converges if |x| < 1. So, by the comparison test, our series
converges for |x| < 1 too.

• Since every An is at least one, the nth term in our series obeys

An xn ≥ |x|n

If |x| ≥ 1, this an = An xn cannot converge to zero as n → ∞, and our series

diverges by the divergence test.

In conclusion, our series converges if and only if |x| < 1, and so has radius of con-
vergence 1.

472
S EQUENCE AND SERIES 3.5 P OWER S ERIES

Example 3.5.8 A series from π.

Lets construct a series from the digits of π. Now to avoid dividing by zero, let us set

An = 1 + the nth digit of π

Since π = 3.141591 . . .

A0 = 4 A1 = 2 A2 = 5 A3 = 2 A4 = 6 A5 = 10 A6 = 2 ···

Consequently every An is an integer between 1 and 10 and gives us the series

∞
X
An xn = 4 + 2x + 5x2 + 2x3 + 6x4 + 10x5 + · · ·
n=0

The number π is irrational a and consequently the ratio AAn+1 n

cannot have a limit as
n → ∞. If you do not understand why this is the case then don’t worry too much
about it b . As in the last example, the limit of the ratios does not exist and we cannot
tell anything from the ratio test. But we can still figure out for which x’s it converges.
• Because every coefficient An is no bigger (in magnitude) than 10, the nth term
in our series obeys
An xn ≤ 10|x|n
and so is smaller than the nth term in the geometric series ∞ n
P
n=0 10|x| . This
geometric series converges if |x| < 1. So, by the comparison test, our series
converges for |x| < 1 too.

• Since every An is at least one, the nth term in our series obeys

An xn ≥ |x|n

If |x| ≥ 1, this an = An xn cannot converge to zero as n → ∞, and our series

diverges by the divergence test.
In conclusion, our series converges if and only if |x| < 1, and so has radius of con-
vergence 1.

a We give a proof of this in the optional §3.7 at the end of this chapter.
b This is a little beyond the scope of the course. Roughly speaking, think about what would happen
if the limit of the ratios did exist. If the limit were smaller than 1, then it would tell you that
the terms of our series must be getting smaller and smaller and smaller — which is impossible
because they are all integers between 1 and 10. Similarly if the limit existed and were bigger than
1 then the terms of the series would have to get bigger and bigger and bigger — also impossible.
Hence if the ratio exists then it must be equal to 1 — but in that case because the terms are
integers, they would have to be all equal when n became big enough. But that means that the
expansion of π would be eventually periodic — something that only rational numbers do (a proof
is given in the optional §3.7 at the end of this chapter).

473
S EQUENCE AND SERIES 3.5 P OWER S ERIES

Though we won’t prove it, it is true that every power series has a radius of con-
An+1
vergence, whether or not the limit lim exists.
n→∞ An

Theorem 3.5.9
∞
An (x − c)n be a power series. Then one of the following alternatives
P
Let
n=0
must hold.

a The power series converges for every number x. In this case we say that
the radius of convergence is ∞.

b There is a number 0 < R < ∞ such that the series converges for |x − c| <
R and diverges for |x−c| > R. Then R is called the radius of convergence.

c The series converges for x = c and diverges for all x 6= c. In this case, we
say that the radius of convergence is 0.

Definition 3.5.10
Consider the power series
∞
X
An (x − c)n .
n=0

The set of real x-values for which it converges is called the interval of conver-
gence of the series.
∞
An (x − c)n has radius of convergence R. Then
P
Suppose that the power series
n=0
from Theorem 3.5.9, we have that

• if R = ∞, then its interval of convergence is −∞ < x < ∞, which is also

denoted (−∞, ∞), and

• if R = 0, then its interval of convergence is just the point x = c, and

• if 0 < R < ∞, then we know that the series converges for any x which obeys

|x − c| < R or equivalently − R < x − c < R

or equivalently c − R < x < c + R

But we do not (yet) know whether or not the series converges at the two end
points of that interval. We do know, however, that its interval of convergence
must be one of

◦ c − R < x < c + R, which is also denoted (c − R , c + R), or

474
S EQUENCE AND SERIES 3.5 P OWER S ERIES

◦ c − R ≤ x < c + R, which is also denoted [c − R , c + R), or

◦ c − R < x ≤ c + R, which is also denoted (c − R , c + R], or
◦ c − R ≤ x ≤ c + R, which is also denoted [c − R , c + R].

To reiterate — while the radius convergence, R with 0 < R < ∞, tells us that the
series converges for |x − c| < R and diverges for |x − c| > R, it does not (by itself) tell
us whether or not the series converges when |x − c| = R, i.e. when x = c ± R. The
following example shows that all four possibilities can occur.

P∞ xn
Example 3.5.11 The series n=0 np .

Let p be any real number and consider the series ∞ xn 1

P
n=0 np . This series has An = np
.
Since
An+1 np 1
lim = lim p = lim p = 1
n→∞ An n→∞ (n + 1) n→∞ (1 + 1 )
n

the series has radius of convergence 1. So it certainly converges for |x| < 1 and
diverges for |x| > 1. That just leaves x = ±1.

• When x = 1, the series reduces to ∞ 1

P
n=0 np . We know, from Example 3.3.6, that
this series converges if and only if p > 1.
(−1)n
• When x = −1, the series reduces to ∞
P
n=0 np . By the alternating series test,
Theorem 3.3.14, this series converges whenever p > 0 (so that n1p tends to zero
as n tends to infinity). When p ≤ 0 (so that n1p does not tend to zero as n tends
to infinity), it diverges by the divergence test, Theorem 3.3.1.

So
P∞
• The power series n=0 xn (i.e. p = 0) has interval of convergence −1 < x < 1.
P∞ xn
• The power series n=0 n (i.e. p = 1) has interval of convergence −1 ≤ x < 1.
P∞ (−1)n n
• The power series n=0 n
x (i.e. p = 1) has interval of convergence −1 <
x ≤ 1.
P∞ xn
• The power series n=0 n2 (i.e. p = 2) has interval of convergence −1 ≤ x ≤ 1.

Example 3.5.12 Playing with intervals of convergence.

We are told that a certain power series with centre c = 3, converges at x = 4 and
diverges at x = 1. What else can we say about the convergence or divergence of the
series for other values of x?
We are told that the series is centred at 3, so its terms are all powers of (x − 3) and it

475
S EQUENCE AND SERIES 3.5 P OWER S ERIES

is of the form
X
An (x − 3)n .
n≥0

A good way to summarise the convergence data we are given is with a figure like the
one below. Green dots mark the values of x where the series is known to converge.
(Recall that every power series converges at its centre.) The red dot marks the value
of x where the series is known to diverge. The bull’s eye marks the centre.

Can we say more about the convergence and/or divergence of the series for other
values of x? Yes!
Let us think about the radius of convergence, R, of the series. We know that it must
exist and the information we have been given allows us to bound R. Recall that
• the series converges at x provided that |x − 3| < R and
• the series diverges at x if |x − 3| > R.
We have been told that
• the series converges when x = 4, which tells us that

◦ x = 4 cannot obey |x − 3| > R so

◦ x = 4 must obey |x − 3| ≤ R, i.e. |4 − 3| ≤ R, i.e. R ≥ 1
• the series diverges when x = 1 so we also know that
◦ x = 1 cannot obey |x − 3| < R so
◦ x = 1 must obey |x − 3| ≥ R, i.e. |1 − 3| ≥ R, i.e. R ≤ 2
We still don’t know R exactly. But we do know that 1 ≤ R ≤ 2. Consequently,
• since 1 is the smallest that R could be, the series certainly converges at x if
|x − 3| < 1, i.e. if 2 < x < 4 and

• since 2 is the largest that R could be, the series certainly diverges at x if |x−3| >
2, i.e. if x > 5 or if x < 1.
The following figure provides a resume of all of this convergence data — there is
convergence at green x’s and divergence at red x’s.

476
 S EQUENCE AND SERIES 3.5 P OWER S ERIES

Notice that from the data given we cannot say anything about the convergence or
divergence of the series on the intervals (1, 2] and (4, 5].
One lesson that we can derive from this example is that,

• if a series has centre c and converges at a,

• then it also converges at all points between c and a, as well as at all points of
distance strictly less than |a − c| from c on the other side of c from a.
Example 3.5.12

3.5.2 tt Working With Power Series

Just as we have done previously with limits, differentiation and integration, we can
construct power series representations of more complicated functions by using those
of simpler functions. Here is a theorem that helps us to do so.

Theorem 3.5.13 Operations on Power Series.

Assume that the functions f (x) and g(x) are given by the power series
∞
X ∞
X
f (x) = An (x − c)n g(x) = Bn (x − c)n
n=0 n=0

for all x obeying |x − c| < R. In particular, we are assuming that both power
series have radius of convergence at least R. Also let K be a constant. Then
∞
X
f (x) + g(x) = [An + Bn ] (x − c)n
n=0
∞
X
Kf (x) = K An (x − c)n
n=0
X∞
(x − c)N f (x) = An (x − c)n+N for any integer N ≥ 1
n=0
X∞
= Ak−N (x − c)k where k = n + N
k=N
∞
X ∞
X
0 n−1
f (x) = An n (x − c) = An n (x − c)n−1
n=0 n=1
x ∞ n+1
(x − c)
Z X
f (t)dt = An
c n=0
n+1

477
S EQUENCE AND SERIES 3.5 P OWER S ERIES

∞
(x − c)n+1
Z X 
f (x) dx = An +C with C an arbitrary constant
n=0
n+1

for all x obeying |x − c| < R.

In particular the radius of convergence of each of the six power series on the
right hand sides is at least R. In fact, if R is the radius of convergence of
∞
An (x − c)n , then R is also the radius of convergence of all of the above
P
n=0
∞
[An + Bn ] (x − c)n and
P
right hand sides, with the possible exceptions of
n=0
∞
KAn (x − c)n when K = 0.
P
n=0

Example 3.5.14 More on the last part of Theorem 3.5.13.

The last statement of Theorem 3.5.13 might seem a little odd, but consider the fol-
lowing two power series centred at 0:
∞
X ∞
X
n n
2 x and (1 − 2n )xn .
n=0 n=0

The ratio test tells us that they both have radius of convergence R = 21 . However
their sum is
∞
X ∞
X ∞
X
n n n n
2 x + (1 − 2 )x = xn
n=0 n=0 n=0

which has the larger radius of convergence 1.

A more extreme example of the same phenomenon is supplied by the two series
∞
X ∞
X
n n
2 x and (−2n )xn .
n=0 n=0

They are both geometric series with radius of convergence R = 21 . But their sum is
∞
X ∞
X ∞
X
n n n n
2 x + (−2 )x = (0)xn
n=0 n=0 n=0

which has radius of convergence +∞.

We’ll now use this theorem to build power series representations for a bunch of
functions out of the one simple power series representation that we know — the

478
S EQUENCE AND SERIES 3.5 P OWER S ERIES

geometric series
∞
1 X
= xn for all |x| < 1
1 − x n=0

1
Example 3.5.15 1−x2
.
1
Find a power series representation for 1−x 2.

Solution: The secret to finding power series representations for a good many func-
1
tions is to manipulate them into a form in which 1−y appears and use the geometric
1
P ∞ n
series representation 1−y = n=0 y . We have deliberately renamed the variable to
y here — it does not have to be x. We can use that strategy to find a power series
1 1 1
expansion for 1−x 2 — we just have to recognize that 1−x2 is the same as 1−y if we set

y to x2 .
∞
X 
1 1 n
2
= = y if |y| < 1, i.e. |x| < 1
1−x 1−y y=x2 n=0 y=x2
∞ ∞
2 n
X X
x2n


= x =
n=0 n=0
= 1 + x2 + x4 + x6 + · · ·

This is a perfectly good power series. There is nothing wrong with the power of
x being 2n. (This just means that the coefficients of all odd powers of x are zero.)
In fact, you should try to always write power series in forms that are as easy to
understand as possible. The geometric series that we used at the end of the first line
converges for

|y| < 1 ⇐⇒ x2 < 1 ⇐⇒ |x| < 1

So our power series has radius of convergence 1 and interval of convergence −1 <
x < 1.

x
Example 3.5.16 2+x2
.
x
Find a power series representation for 2+x 2.

Solution: This example is just a more algebraically involved variant of the last one.
x 1
Again, the strategy is to manipulate 2+x2 into a form in which 1−y appears.

x x 1 x 1 x2
= = set − =y
2 + x2 2 1 + x22 2 1 − − x22 2


 ∞ 
x 1 x X n
= = y if |y| < 1
2 1 − y y=− x2 2 n=0 y=− x
2
2 2

479
S EQUENCE AND SERIES 3.5 P OWER S ERIES

Now use Theorem 3.5.13 twice

∞ n ∞ ∞
x X  x2  x X (−1)n 2n X (−1)n 2n+1
= − = x = x
2 n=0 2 2 n=0 2n n=0
2n+1
x x3 x5 x7
= − + − + ···
2 4 8 16
The geometric series that we used in the second line converges when

x2
|y| < 1 ⇐⇒ − <1
2 √
⇐⇒ |x|2 < 2 ⇐⇒ |x| < 2
√
So√the given√power series has radius of convergence 2 and interval of convergence
− 2 < x < 2.
Example 3.5.16

Example 3.5.17 Nonzero centre.

1
Find a power series representation for 5−x with centre 3.
Solution: The new wrinkle in this example is the requirement that the centre be 3.
That the centre is to be 3 means that we need a power series P in powers ofnx − c, with
c = 3. So we are looking for a power series of the form ∞ n=0 An (x − 3) . The easy
way to find such a series is to force an x − 3 to appear by adding and subtracting a 3.
1 1 1
= =
5−x 5 − (x − 3) − 3 2 − (x − 3)
1
Now we continue, as in the last example, by manipulating 2−(x−3)
into a form in
1
which 1−y appears.
1 1 1 1 x−3
= = set =y
5−x 2 − (x − 3) 2 1 − x−3
2
2
 ∞ 
1 1 1 X n
= = y if |y| < 1
2 1 − y y= x−3 2 n=0 y= x−3
2 2
∞  n ∞ n
1 X x−3  X (x − 3)
= =
2 n=0
2 n=0
2n+1
x − 3 (x − 3)2 (x − 3)3
= + + + ···
2 4 8
The geometric series that we used in the second line converges when
x−3
|y| < 1 ⇐⇒ <1
2

480
S EQUENCE AND SERIES 3.5 P OWER S ERIES

⇐⇒ |x − 3| < 2
⇐⇒ −2 < x − 3 < 2
⇐⇒ 1 < x < 5

So the power series has radius of convergence 2 and interval of convergence 1 < x <
5.
Example 3.5.17

In the previous two examples, to construct a new series from an existing series,
we replaced x by a simple function. The following theorem gives us some more (but
certainly not all) commonly used substitutions.

Theorem 3.5.18 Substituting in a Power Series.

Assume that the function f (x) is given by the power series

∞
X
f (x) = A n xn
n=0

for all x in the interval I. Also let K and k be real constants. Then
∞
X
k
An K n xkn


f Kx =
n=0

whenever Kxk is in I. In particular, if ∞ n

P
n=0 An xP has radius of convergence
R, K is nonzero
p and k is a natural number, then ∞ n kn
n=0 An K x has radius of
convergence R/|K|.
k

1
Example 3.5.19 (1−x)2
.

1
Find a power series representation for (1−x)2.
1 1
Solution: Once again the trick is to express (1−x)2 in terms of 1−x
. Notice that

1 d 1
=
(1 − x)2 dx 1−x
(∞ )
d X n
= x
dx n=0
∞
X
= nxn−1 by Theorem 3.5.13
n=1

481
S EQUENCE AND SERIES 3.5 P OWER S ERIES

Note that the n = 0 term has disappeared because, for n = 0,

d n d 0 d
x = x = 1=0
dx dx dx
Also note that the radius of convergence of this series is one. We can see this via The-
orem 3.5.13. That theorem tells us that the radius
P∞ ofnconvergence of a power series is
not changed by differentiation — and since n=0 x has radius of convergence one,
so too does its derivative.
Without much more work we can determine the interval of convergence by testing
at x = ±1. When x = ±1 the terms of the series do not go to zero as n → ∞ and
so, by the divergence test, the series does not converge there. Hence the interval of
convergence for the series is −1 < x < 1.
Example 3.5.19

Notice that, in this last example, we differentiated a known series to get to our
answer. As per Theorem 3.5.13, the radius of convergence didn’t change. In addition,
in this particular example, the interval of convergence didn’t change. This is not
always the case. Differentiation of some series causes the interval of convergence to
shrink. In particular the differentiated series may no longer be convergent at the end
points of the interval 4 . Similarly, when we integrate a power series the radius of
convergence is unchanged, but the interval of convergence may expand to include
one or both ends, as illustrated by the next example.

Example 3.5.20 log(1 + x).

Find a power series representation for log(1 + x).

d 1 1
Solution: Recall that dx log(1 + x) = 1+x so that log(1 + t) is an antiderivative of 1+t
and
Z x Z xhX ∞
dt i
log(1 + x) = = (−t)n dt
0 1+t 0 n=0
∞
X x Z
= (−t)n dt by Theorem 3.5.13
n=0 0
∞
X xn+1
n
= (−1)
n=0
n+1
x2 x3 x4
= x− + − + ···
2 3 4
Theorem 3.5.13 guarantees that the radius of convergence is exactly one (the radius

P∞ n
4 Consider the power series n=1 xn . We know that its interval of convergence is −1 ≤ x < 1.
(Indeed
P∞ nsee the next example.) When we differentiate the series we get the geometric series
n=0 x which has interval of convergence −1 < x < 1.

482
S EQUENCE AND SERIES 3.5 P OWER S ERIES

P∞ n
of convergence of the geometric series n=0 (−t) ) and that
∞
X xn+1
log(1 + x) = (−1)n for all −1<x<1
n=0
n+1

When x = −1 our series reduces to ∞ −1

P
n=0 n+1 , which is (minus) the harmonic series
and so diverges. That’s no surprise — log(1 + (−1)) = log 0 = −∞. When x = 1, the
series converges by the alternating series test. It is possible to prove, by continuity,
though we won’t do so here, that the sum is log 2. So the interval of convergence is
−1 < x ≤ 1.
Example 3.5.20

Example 3.5.21 arctan x.

Find a power series representation for arctan x.

d 1 1
Solution: Recall that dx arctan x = 1+x 2 so that arctan t is an antiderivative of 1+t2
and
Z x Z xhX ∞ ∞ Z x
dt 2 n
i X
arctan x = 2
= (−t ) dt = (−1)n t2n dt
0 1+t 0 n=0 n=0 0
∞
X x2n+1
= (−1)n
n=0
2n + 1
x3 x5
=x− + − ···
3 5
Theorem 3.5.13 guarantees that the radius
P of convergence is exactly one (the radius
of convergence of the geometric series ∞n=0 (−t2 n
) ) and that
∞
X x2n+1
arctan x = (−1)n for all −1 < x < 1
n=0
2n + 1

When x = ±1, the series converges by the alternating series test. So the interval of
convergence is −1 ≤ x ≤ 1. It is possible to prove, though once again we won’t do
so here, that when x = ±1, the series ∞ n x2n+1
P
n=0 (−1) 2n+1 converges to the value of the
left hand side, arctan x, at x = ±1. That is, to arctan(±1) = ± π4 .

The operations on power series dealt with in Theorem 3.5.13 are fairly easy to
apply. Unfortunately taking the product, ratio or composition of two power series
is more involved and is beyond the scope of this course 5 . Unfortunately Theo-
rem 3.5.13 alone will not get us power series representations of many of our stan-

5 As always, a quick visit to your favourite search engine will direct the interested reader to more
information.

483
 S EQUENCE AND SERIES 3.5 P OWER S ERIES

dard functions (like ex and sin x). Fortunately we can find such representations by
extending Taylor polynomials 6 to Taylor series.

3.5.3 tt Exercises

Exercises — Stage 1

∞  n
X 3−x
1. Suppose f (x) = . What is f (1)?
n=0
4

∞
X (x − 5)n
2. Suppose f (x) = . Give a power series representation of
n=1
n! + 2
f 0 (x).

∞
X
3. Let f (x) = An (x − c)n for some positive constants a and c, and some
n=a
sequence of constants {An }. For which values of x does f (x) definitely con-
verge?

4. Let f (x) be a power series centred at c = 5. If f (x) converges at x = −1,

and diverges at x = 11, what is the radius of convergence of f (x)?

Exercises — Stage 2

5. *. (a) Find the radius of convergence of the series

∞
X
(−1)k 2k+1 xk
k=0

(b) You are given the formula for the sum of a geometric series, namely:

1
1 + r + r2 + · · · = , |r| < 1
1−r
Use this fact to evaluate the series in part (a).

6 Now is a good time to review your notes from last term, though we’ll give you a whirlwind
review over the next page or two.

484
S EQUENCE AND SERIES 3.5 P OWER S ERIES

∞
X xk
6. *. Find the radius of convergence for the power series 10k+1 (k + 1)!
k=0

∞
X (x − 2)n
7. *. Find the radius of convergence for the power series n2 + 1
.
n=0
∞
X (−1)n (x + 2)n
8. *. Consider the power series √ , where x is a real number.
n=1
n
Find the interval of convergence of this series.

9. *. Find the radius of convergence and interval of convergence of the

series
∞ n
(−1)n x + 1
X 

n=0
n+1 3

10. *. Find the interval of convergence for the power series

∞
X (x − 2)n
.
n=1
n4/5 (5n − 4)

∞
X (x + 2)n
11. *. Find all values x for which the series n2
converges.
n=1
∞
X 4n
12. *. Find the interval of convergence for (x − 1)n .
n=1
n

13. *. Find, with explanation, the radius of convergence and the interval of
convergence of the power series
∞ n
n (x − 1)
X
(−1) n
n=0
2 (n + 2)

∞
X
14. *. Find the interval of convergence for the series (−1)n n2 (x − a)2n where a
n=1
is a constant.

15. *. Find the interval of convergence of the following series:

∞
X (x + 1)k
a .
k=1
k 2 9k

485
S EQUENCE AND SERIES 3.5 P OWER S ERIES

∞
X
b ak (x − 1)k , where ak > 0 for k = 1, 2, · · · and
k=1
∞ 
X ak ak+1  a1
− = .
k=1
ak+1 ak+2 a2

x3
16. *. Find a power series representation for 1 − x .

∞ x ∞
(x − 1)n (x − 1)n+1
X Z X
0
17. Suppose f (x) = , and f (t)dt = 3x + 2
.
n=0
n + 2 5 n=1
n(n + 1)
Give a power series representation of f (x).

Exercises — Stage 3

18. *. Determine the values of x for which the series

∞
X xn
n=2
32n log n

converges absolutely, converges conditionally, or diverges.

1
Z
19. *. (a) Find the power–series representation for dx centred at 0
1 + x3
(i.e. in powers of x).
1/4
1
Z
(b) The power series above is used to approximate dx. How
0 1 + x3
many terms are required to guarantee that the resulting approximation
is within 10−5 of the exact value? Justify your answer.

∞
X x
20. *. (a) Show that nxn = for −1 < x < 1.
n=0
(1 − x)2
∞
X
(b) Express n2 xn as a ratio of polynomials. For which x does this series
n=0
converge?
P∞
21. *. Suppose that you have a sequence {bn } such that the series n=0 (1 −
bn ) converges. Using the tests we’ve learned in class, prove that the
X∞
radius of convergence of the power series bn xn is equal to 1.
n=0

486
S EQUENCE AND SERIES 3.5 P OWER S ERIES

22. *. Assume an is a sequence such that nan decreases to C as n → ∞ for
some real number C > 0 ∞
X
(a) Find the radius of convergence of an xn . Justify your answer care-
n=1
fully.
(b) Find the interval of convergence of the above power series, that is, find
all x for which the power series in (a) converges. Justify your answer care-
fully.

23. An infinitely long, straight rod of negligible mass has the following
weights:
1
• At every whole number n, a mass of weight at position n, and
2n
1
• a mass of weight at position −n.
3n
At what position is the centre of mass of the rod?

−3 −2 −1 0 1 2 3

1
33

1 1
32 23

1 1
31 22

1
21

∞
X
24. Let f (x) = An (x − c)n , for some constant c and a sequence of constants
n=0
{An }. Further, let f (x) have a positive radius of covergence.
If A1 = 0, show that y = f (x) has a critical point at x = c. What is the
relationship between the behaviour of the graph at that point and the value
of A2 ?

∞
X n
25. Evaluate n−1
.
n=3
5

26.
−5 1

f (x) = log(1+x) to within an error
Find a polynomial that approximates
of 10 for all values of x in 0, 10 .
Then, use your polynomial to approximate log(1.05) as a rational num-
ber.

487
 S EQUENCE AND SERIES 3.6 TAYLOR S ERIES

27. Find a polynomial that approximates f (x) = arctan x to within an error of

−5 1 1


10 for all values of x in − 4 , 4 .

3.6q Taylor Series

3.6.1 tt Extending Taylor Polynomials

Recall 1 that Taylor polynomials provide a hierarchy of approximations to a given

function f (x) near a given point a. Typically, the quality of these approximations
improves as we move up the hierarchy.

• The crudest approximation is the constant approximation f (x) ≈ f (a).

• Then comes the linear, or tangent line, approximation f (x) ≈ f (a)+f 0 (a) (x−a).

• Then comes the quadratic approximation

1
f (x) ≈ f (a) + f 0 (a) (x − a) + f 00 (a) (x − a)2
2

• In general, the Taylor polynomial of degree n, for the function f (x), about the
expansion point a, is the polynomial, Tn (x), determined by the requirements
(k)
that f (k) (a) = Tn (a) for all 0 ≤ k ≤ n. That is, f and Tn have the same
derivatives at a, up to order n. Explicitly,

1 1
f (x) ≈ Tn (x) = f (a) + f 0 (a) (x − a) + f 00 (a) (x − a)2 + · · · + f (n) (a) (x − a)n
2 n!
n
X 1 (k)
= f (a) (x − a)k
k=0
k!

These are, of course, approximations — often very good approximations near x = a

— but still just approximations. One might hope that if we let the degree, n, of
the approximation go to infinity then the error in the approximation might go to
zero. If that is the case then the “infinite” Taylor polynomial would be an exact
representation of the function. Let’s see how this might work.
Fix a real number a and suppose that all derivatives of the function f (x) exist.
Then, for any natural number n,

1 Please review your notes from last term if this material is feeling a little unfamiliar.

488
S EQUENCE AND SERIES 3.6 TAYLOR S ERIES

Equation 3.6.1

f (x) = Tn (x) + En (x)

where Tn (x) is the Taylor polynomial of degree n for the function f (x) expanded
about a, and En (x) = f (x) − Tn (x) is the error in our approximation. The Taylor
polynomial 2 is given by the formula

Equation 3.6.2

Tn (x) = f (a) + f 0 (a) (x − a) + · · · + 1 (n)

n!
f (a) (x − a)n

while the error satisfies

Equation 3.6.3

En (x) = 1
(n+1)!
f (n+1) (c) (x − a)n+1

for some c strictly between a and x.

Note that we typically do not know the value of c in the formula for the error.
Instead we use the bounds on c to find bounds on f (n+1) (c) and so bound the error 3 .
In order for our Taylor polynomial to be an exact representation of the function
f (x) we need the error En (x) to be zero. This will not happen when n is finite unless
f (x) is a polynomial. However it can happen in the limit as n → ∞, and in that case
we can write f (x) as the limit
n
X
1 (k)
f (x) = lim Tn (x) = lim k!
f (a) (x − a)k
n→∞ n→∞
k=0

This is really a limit of partial sums, and so we can write

∞
X
1 (k)
f (x) = k!
f (a) (x − a)k
k=0

2 Did you take a quick look at your notes?

3 The discussion here is only supposed to jog your memory. If it is feeling insufficiently jogged,
then please look at your notes from last term.

489
S EQUENCE AND SERIES 3.6 TAYLOR S ERIES

which is a power series representation of the function. Let us formalise this in a

definition.
Definition 3.6.4 Taylor series.

The Taylor series for the function f (x) expanded around a is the power series
∞
X
1 (n)
f (x) = n!
f (a) (x − a)n
n=0

provided the series converges. When a = 0 it is also called the Maclaurin series
of f (x).

This definition hides the discussion of whether or not En (x) → 0 as n → ∞ within

the caveat “provided the series converges”. Demonstrating that for a given function
can be difficult, but for many of the standard functions you are used to dealing with,
it turns out to be pretty easy. Let’s compute a few Taylor series and see how we do
it.

Example 3.6.5 Exponential Series.

Find the Maclaurin series for f (x) = ex .

Solution: Just as was the case for computing Taylor polynomials, we need to com-
pute the derivatives of the function at the particular choice of a. Since we are asked
for a Maclaurin series, a = 0. So now we just need to find f (k) (0) for all integers
k ≥ 0.
d x
We know that dx e = ex and so

ex = f (x) = f 0 (x) = f 00 (x) = · · · = f (k) (x) = · · · which gives

0 00 (k)
1 = f (0) = f (0) = f (0) = · · · = f (0) = · · · .

Equations 3.6.1 and 3.6.2 then give us

x2 xn
ex = f (x) = 1 + x + + ··· + + En (x)
2! n!
We shall see, in the optional Example 3.6.8 below, that, for any fixed x, lim En (x) =
n→∞
0. Consequently, for all x,
∞
x
h 1 2 1 3 1 ni X 1 n
e = lim 1 + x + x + x + · · · + x = x
n→∞ 2 3! n! n=0
n!

We have now seen power series representations for the functions

1 1
log(1 + x) arctan(x) ex .
1−x (1 − x)2

490
S EQUENCE AND SERIES 3.6 TAYLOR S ERIES

We do not think that you, the reader, will be terribly surprised to see that we develop
series for sine and cosine next.

Example 3.6.6 Sine and Cosine Series.

The trigonometric functions sin x and cos x also have widely used Maclaurin series
expansions (i.e. Taylor series expansions about a = 0). To find them, we first com-
pute all derivatives at general x.

f (x) = sin x f 0 (x) = cos x f 00 (x) = − sin x f (3) (x) = − cos x f (4) (x) = sin x · · ·
g(x) = cos x g 0 (x) = − sin x g 00 (x) = − cos x g (3) (x) = sin x g (4) (x) = cos x · · ·

Now set x = a = 0.

f (x) = sin x f (0) = 0 f 0 (0) = 1 f 00 (0) = 0 f (3) (0) = −1 f (4) (0) = 0 ···
0 00 (3) (4)
g(x) = cos x g(0) = 1 g (0) = 0 g (0) = −1 g (0) = 0 g (0) = 1 ···

For sin x, all even numbered derivatives (at x = 0) are zero, while the odd numbered
derivatives alternate between 1 and −1. Very similarly, for cos x, all odd numbered
derivatives (at x = 0) are zero, while the even numbered derivatives alternate be-
tween 1 and −1. So, the Taylor polynomials that best approximate sin x and cos x
near x = a = 0 are

sin x ≈ x − 3!1 x3 + 5!1 x5 − · · ·

cos x ≈ 1 − 2!1 x2 + 4!1 x4 − · · ·

We shall see, in the optional Example 3.6.10 below, that, for both sin x and cos x, we
have lim En (x) = 0 so that
n→∞
h i
f (x) = lim f (0) + f 0 (0) x + · · · + n!1 f (n) (0) xn
n→∞
h i
0 1 (n) n
g(x) = lim g(0) + g (0) x + · · · + n! g (0) x
n→∞

Reviewing the patterns we found in the derivatives, we conclude that, for all x,
∞
X
sin x = x − 3!1 x3 + 5!1 x5 − · · · = 1
(−1)n (2n+1)! x2n+1
n=0
∞
X
cos x = 1 − 2!1 x2 + 4!1 x4 − · · · = 1
(−1)n (2n)! x2n
n=0

and, in particular, both of the series on the right hand sides converge for all x.
We could also test for convergence of the series using the ratio test. Computing the
ratios of successive terms in these two series gives us

An+1 |x|2n+3 /(2n + 3)! |x|2

= 2n+1 =
An |x| /(2n + 1)! (2n + 3)(2n + 2)

491
S EQUENCE AND SERIES 3.6 TAYLOR S ERIES

An+1 |x|2n+2 /(2n + 2)! |x|2

= =
An |x|2n /(2n)! (2n + 2)(2n + 1)

for sine and cosine respectively. Hence as n → ∞ these ratios go to zero and con-
sequently both series are convergent for all x. (This is very similar to what was
observed in Example 3.5.5.)
Example 3.6.6

We have developed power series representations for a number of important func-

tions 4 . Here is a theorem that summarizes them.

Theorem 3.6.7

∞
x
X xn 1 2 1 3
e = =1+x+ x + x + · · · for all −∞ < x < ∞
n=0
n! 2! 3!
∞
X x2n+1 1 1
sin(x) = (−1)n = x − x3 + x5 − · · · for all −∞ < x < ∞
n=0
(2n + 1)! 3! 5!
∞
X x2n 1 2 1 4
cos(x) = (−1)n =1− x + x − ··· for all −∞ < x < ∞
n=0
(2n)! 2! 4!
∞
1 X
= xn = 1 + x + x2 + x3 + · · · for all −1 < x < 1
1−x n=0
∞
X xn+1 x2 x3 x 4
log(1 + x) = (−1)n =x− + − + ··· for all −1 < x ≤ 1
n=0
n+1 2 3 4
∞
X x2n+1 x3 x5
arctan x = (−1)n =x− + − ··· for all −1 ≤ x ≤ 1
n=0
2n + 1 3 5

Notice that the series for sine and cosine sum to something that looks very similar

4 The reader might ask whether or not we will give the series for other trigonometric functions or
their inverses. While the tangent function has a perfectly well defined series, its coefficients are
not as simple as those of the series we have seen — they form a sequence of numbers known (per-
haps unsurprisingly) as the “tangent numbers”. They, and the related Bernoulli numbers, have
many interesting properties, links to which the interested reader can find with their favourite
search engine. The Maclaurin series for inverse sine is
∞
X 4−n (2n)! 2n+1
arcsin(x) = x
n=0
2n + 1 (n!)2

which is quite tidy, but proving it is beyond the scope of the course.

492
S EQUENCE AND SERIES 3.6 TAYLOR S ERIES

to the series for ex :

   
1 3 1 5 1 2 1 4
sin(x) + cos(x) = x − x + x − · · · + 1 − x + x − · · ·
3! 5! 2! 4!
1 2 1 3 1 4 1 5
= 1 + x − x − x + x + x − ···
2! 3! 4! 5!
1 1 1 1
ex = 1 + x + x2 + x3 + x4 + x5 + · · ·
2! 3! 4! 5!
So both series have coefficients with the same absolute value (namely n!1 ), but there
are differences in sign 5 . This is not a coincidence and we direct the interested reader
to
√ the optional Section 3.6.3 where will show how these series are linked through
−1.
P∞ 1 n
Example 3.6.8 Optional — Why n=0 n! x is ex ..

We have already seen, in Example 3.6.5, that

x2 xn
ex = 1 + x + + ··· + + En (x)
2! n!
By (3.6.3)
1
En (x) = ec xn+1
(n + 1)!
for some (unknown) c between 0 and x. Fix any real number x. We’ll now show that
En (x) converges to zero as n → ∞.
To do this we need get bound the size of ec , and to do this, consider what happens if
x is positive or negative.

• If x < 0 then x ≤ c ≤ 0 and hence ex ≤ ec ≤ e0 = 1.

• On the other hand, if x ≥ 0 then 0 ≤ c ≤ x and so 1 = e0 ≤ ec ≤ ex .

In either case we have that 0 ≤ ec ≤ 1 + ex . Because of this the error term

ec |x|n+1
|En (x)| = xn+1 ≤ [ex + 1]
(n + 1)! (n + 1)!

We claim that this upper bound, and hence the error En (x), quickly shrinks to zero
as n → ∞.
Call the upper bound (except for the factor ex + 1, which is independent of n) en (x) =
|x|n+1
(n+1)!
. To show that this shrinks to zero as n → ∞, let’s write it as follows.

n + 1 factors
z }| {
|x|n+1 |x| |x| |x| |x| |x|
en (x) = = · · ··· ·
(n + 1)! 1 2 3 n |n + 1|

5 Warning: antique sign–sine pun. No doubt the reader first saw it many years syne.

493
S EQUENCE AND SERIES 3.6 TAYLOR S ERIES

Now let k be an integer bigger than |x|. We can split the product
k factors
z }| {  
|x| |x| |x| |x| |x| |x|
en (x) = · · ··· · ···
1 2 3 k k+1 |n + 1|
   n+1−k
|x| |x| |x| |x| |x|
≤ · · ··· ·
1 2 3 k k+1
| {z }
=Q(x)
 n+1−k
|x|
= Q(x) ·
k+1

Since k does not depend not n (though it does depend on x), the function Q(x) does
|x|
not change as we increase n. Additionally, we know that |x| < k + 1 and so k+1 < 1.
Hence as we let n → ∞ the above bound must go to zero.
Alternatively, compare en (x) and en+1 (x).

|x|n+2
en+1 (x) (n+2)! |x|
= =
en (x) |x|n+1 n+2
(n+1)!

(x)
When n is bigger than, for example 2|x|, we have en+1en (x)
< 12 . That is, increasing the
index on en (x) by one decreases the size of en (x) by a factor of at least two. As a
result en (x) must tend to zero as n → ∞.
Consequently, for all x, lim En (x) = 0, as claimed, and we really have
n→∞

∞
x
h 1 2 1 3 1 ni X 1 n
e = lim 1 + x + x + x + · · · + x = x
n→∞ 2 3! n! n=0
n!

Example 3.6.8

There is another way to prove that the series ∞ xn

P
n=0 n! converges to the function
ex . Rather than looking at how the error term En (x) behaves as n → ∞, we can show
that the series satisfies the same simple differential equation 6 and the same initial
condition as the function.
P∞ 1 n
Example 3.6.9 Optional — Another approach to showing that n=0 n! x is ex ..

We already know from Example 3.5.5, that the series ∞ 1 n

P
n=0 n! x converges to some
function f (x) for all values of x . All that remains to do is to show that f (x) is really

6 Recall, you studied that differential equation in the section on separable differential equations
(Theorem 2.4.4 in Section 2.4) as well as wayyyy back in the section on exponential growth and
decay in differential calculus.

494
S EQUENCE AND SERIES 3.6 TAYLOR S ERIES

ex . We will do this by showing that f (x) and ex satisfy the same differential equation
with the same initial conditions a . We know that y = ex satisfies
dy
=y and y(0) = 1
dx
and by Theorem 2.4.4 (with aP= 1, b = 0 and y(0) = 1), this is the only solution. So it
suffices to show that f (x) = ∞ xn
n=0 n! satisfies

df
= f (x) and f (0) = 1.
dx

• By Theorem 3.5.13,
(∞ ) ∞ ∞
df d X 1
n
X n n−1 X 1
= x = x = xn−1
dx dx n=0
n! n=1
n! n=1
(n − 1)!
n=3 n=4
n=1 z}|{ z}|{
n=2
z}|{ z}|{ x2 x3
= 1 + x + + +···
2! 3!
= f (x)

• When we substitute x = 0 into the series we get (see the discussion after Defi-
nition 3.5.1)
0 0
f (0) = 1 + + + · · · = 1.
1! 2!

Hence f (x) solves the same initial value problem and we must have f (x) = ex .

a Recall that when we solve of a separable differential equation our general solution will have an
arbitrary constant in it. That constant cannot be determined from the differential equation alone
and we need some extra data to find it. This extra information is often information about the
system at its beginning (for example when position or time is zero) — hence “initial conditions”.
Of course the reader is already familiar with this because it was covered back in Section 2.4.

Example 3.6.9

We can show that the error terms in Maclaurin polynomials for sine and cosine
go to zero as n → ∞ using very much the same approach as in Example 3.6.8.

495
 S EQUENCE AND SERIES 3.6 TAYLOR S ERIES

P∞ (−1)n 2n+1 P∞ (−1)n 2n

Example 3.6.10 Optional — Why n=0 (2n+1)! x = sin x and n=0 (2n)! x = cos x.

Let f (x) be either sin x or cos x. We know that every derivative of f (x) will be one
of ± sin(x) or ± cos(x). Consequently, when we compute the error term using equa-
tion 3.6.3 we always have f (n+1) (c) ≤ 1 and hence

|x|n+1
|En (x)| ≤ .
(n + 1)!
|x| n+1
In Example 3.6.5, we showed that (n+1)! → 0 as n → ∞ — so all the hard work is
already done. Since the error term shrinks to zero for both f (x) = sin x and f (x) =
cos x, and
h i
f (x) = lim f (0) + f 0 (0) x + · · · + n!1 f (n) (0) xn
n→∞

as required.

3.6.2 tt Computing with Taylor Series

Taylor series have a great many applications. (Hence their place in this course.) One
of the most immediate of these is that they give us an alternate way of computing
many functions. For example, the first definition we see for the sine and cosine
functions is in terms of triangles. Those definitions, however, do not lend themselves
to computing sine and cosine except at very special angles. Armed with power series
representations, however, we can compute them to very high precision at any angle.
To illustrate this, consider the computation of π — a problem that dates back to the
Babylonians.

Example 3.6.11 Computing the number π.

There are numerous methods for computing π to any desired degree of accuracy a .
Many of them use the Maclaurin expansion
∞
X x2n+1
arctan x = (−1)n
n=0
2n + 1
π
of Theorem 3.6.7. Since arctan(1) = 4
, the series gives us a very pretty formula for π:
X (−1)n∞
π
= arctan 1 =
4 n=0
2n + 1
 
1 1 1
π = 4 1 − + − + ···
3 5 7

496
S EQUENCE AND SERIES 3.6 TAYLOR S ERIES

Unfortunately, this series is not very useful for computing π because it converges
so slowly. If we approximate the series by its N th partial sum, then the alternating
series test (Theorem 3.3.14) tells us that the error is bounded by the first term we
drop. To guarantee that we have 2 decimal digits of π correct, we need to sum about
the first 200 terms!
A much better way to compute π using this series is to take advantage of the fact
that tan π6 = √13 :

 1  ∞
X 1 1
π = 6 arctan √ = 6 (−1)n √ 2n+1
3 n=0
2n + 1 ( 3)
∞
√ X 1 1
=2 3 (−1)n
n=0
2n + 1 3n
√  1 1 1 1 1 
=2 3 1− + − + − + ···
3 × 3 5 × 9 7 × 27 9 × 81 11 × 243
Again, this is an alternating series and so (via Theorem 3.3.14) the error we introduce
by truncating it is bounded by the first term dropped. For example, if we keep ten
terms, stopping at n = 9, we get π = 3.141591 (to 6 decimal places) with an error
between zero and √
2 3
10
< 3 × 10−6
21 × 3
In 1699, the English astronomer/mathematician Abraham Sharp (1653–1742) used
150 terms of this series to compute 72 digits of π — by hand!
This is just one of very many ways to compute π. Another one, which still uses the
Maclaurin expansion of arctan x, but is much more efficient, is

1 1
π = 16 arctan − 4 arctan
5 239
This formula was used by John Machin in 1706 to compute π to 100 decimal digits
— again, by hand.

a The computation of π has a very, very long history and your favourite search engine will turn up
many sites that explore the topic. For a more comprehensive history one can turn to books such
as “A history of Pi” by Petr Beckmann and “The joy of π” by David Blatner.

Example 3.6.11

Power series also give us access to new functions which might not be easily ex-
pressed in terms of the functions we have been introduced to so far. The following
is a good example of this.

497
S EQUENCE AND SERIES 3.6 TAYLOR S ERIES

Example 3.6.12 Error function.

The error function

x
2
Z
2
erf(x) = √ e−t dt
π 0

is used in computing “bell curve” probabilities. The indefinite integral of the in-
2
tegrand e−t cannot be expressed in terms of standard functions. But we can still
evaluate the integral to within any desired degree of accuracy by using the Taylor
expansion of the exponential. Start with the Maclaurin series for ex :
∞
X 1 n
ex = x
n=0
n!

and then substitute x = −t2 into this:

∞
−t2
X (−1)n
e = t2n
n=0
n!

We can then apply Theorem 3.5.13 to integrate term-by-term:

Z x "X∞ n
#
2 (−t2 )
erf(x) = √ dt
π 0 n=0 n!
∞
2 X x2n+1
=√ (−1)n
π n=0 (2n + 1)n!

For example, for the bell curve, the probability of being within one standard devia-
tion of the mean a , is
2n+1
 1  2 X
∞
n
( √12 ) 2 X
∞
1
erf √ = √ (−1) =√ (−1)n
2 π n=0 (2n + 1)n! 2π n=0 (2n + 1)2n n!
r 
2 1 1 1 1 
= 1− + − + − · · ·
π 3 × 2 5 × 22 × 2 7 × 23 × 3! 9 × 24 × 4!
This is yet another alternating series. If we keep five terms, stopping at n = 4, we get
0.68271 (to 5 decimal places) with, by Theorem 3.3.14 again, an error between zero
and the first dropped term, which is minus
r
2 1
< 2 × 10−5
π 11 × 25 × 5!

a If you don’t know what this means (forgive the pun) don’t worry, because it is not part of the
course. Standard deviation a way of quantifying variation within a population.

498
S EQUENCE AND SERIES 3.6 TAYLOR S ERIES

Example 3.6.13 Two nice series.

Evaluate ∞ ∞
X (−1)n−1 X 1
and
n=1
n3n n=1
n3n
Solution. There are not very many series that can be easily evaluated exactly. But
occasionally one encounters a series that can be evaluated simply by realizing that it
is exactly one of the series in Theorem 3.6.7, just with a specific value of x. The left
hand given series is
∞
X (−1)n−1 1 1 1 1 1 1 1 1
n
= − 2
+ 3
− 4
+ ···
n=1
n 3 3 2 3 3 3 4 3

The series in Theorem 3.6.7 that this most closely resembles is

x2 x3 x 4
log(1 + x) = x − + − − ···
2 3 4
Indeed
∞
X (−1)n−1 1 1 1 1 1 1 1 1
= − 2
+ 3
− + ···
n=1
n 3n 3 2 3 3 3 4 34
x2 x3 x4
 
= x− + − − ···
2 3 4 x= 31
h i
= log(1 + x) 1
x= 3
4
= log
3
The right hand series above differs from the left hand series above only that the signs
of the left hand series alternate while those of the right hand series do not. We can
flip every second sign in a power series just by using a negative x.

x 2 x3 x4
h i  
log(1 + x) = x− + − − ···
x=− 13 2 3 4 x=− 13
1 1 1 1 1 1 1
=− − 2
− 3
− + ···
3 2 3 3 3 4 34
which is exactly minus the desired right hand series. So
∞
X 1 h i 2 3
n
= − log(1 + x) = − log = log
n=1
n3 1
x=− 3 3 2

499
S EQUENCE AND SERIES 3.6 TAYLOR S ERIES

Example 3.6.14 Finding a derivative from a series.

Let f (x) = sin(2x3 ). Find f (15) (0), the fifteenth derivative of f at x = 0.

Solution: This is a bit of a trick question. We could of course use the product and
chain rules to directly apply fifteen derivatives and then set x = 0, but that would
be extremely tedious a . There is a much more efficient approach that exploits two
pieces of knowledge that we have.

• From equation 3.6.2, we see that the coefficient of (x − a)n in the Taylor series
of f (x) with expansion point a is exactly n!1 f (n) (a). So f (n) (a) is exactly n! times
the coefficient of (x − a)n in the Taylor series of f (x) with expansion point a.

• We know, or at least can easily find, the Taylor series for sin(2x3 ).

Let’s apply that strategy.

• First, we know that, for all y,

1 3 1 5
sin y = y − y + y − ···
3! 5!

• Just substituting y = 2x3 , we have

1 3 1 5
sin(2x3 ) = 2x3 − (2x3 ) + (2x3 ) − · · ·
3! 5!
5
8 2
= 2x3 − x9 + x15 − · · ·
3! 5!

• So the coefficient of x15 in the Taylor series of f (x) = sin(2x3 ) with expansion
5
point a = 0 is 25!

and we have
25
f (15) (0) = 15! × = 348,713,164,800
5!

a We could get a computer algebra system to do it for us without much difficulty — but we
wouldn’t learn much in the process. The point of this example is to illustrate that one can do
more than just represent a function with Taylor series. More on this in the next section.

Example 3.6.15 Optional — Computing the number e.

Back in Example 3.6.8, we saw that

x2 xn 1
ex = 1 + x + 2!
+ ··· + n!
+ (n+1)!
ec xn+1
for some (unknown) c between 0 and x. This can be used to approximate the number

500
S EQUENCE AND SERIES 3.6 TAYLOR S ERIES

e, with any desired degree of accuracy. Setting x = 1 in this equation gives

1 1 1
e=1+1+ 2!
+ ··· + n!
+ (n+1)!
ec

for some c between 0 and 1. Even though we don’t know c exactly, we can bound
that term quite readily. We do know that ec in an increasing function a of c, and so
1 = e0 ≤ ec ≤ e1 = e. Thus we know that
1 1 1

e
≤e− 1+1+ 2!
+ ··· + n!
≤
(n + 1)! (n + 1)!

So we have a lower bound on the error, but our upper bound involves the e — pre-
cisely the quantity we are trying to get a handle on.
But all is not lost. Let’s look a little more closely at the right-hand inequality when
n = 1:
e
e − (1 + 1) ≤ move the e’s to one side
2
e
≤2 and clean it up
2
e ≤ 4.

Now this is a pretty crude bound b but it isn’t hard to improve. Try this again with
n = 1:
1 e
e − (1 + 1 + ) ≤ move e’s to one side
2 6
5e 5
≤
6 2
e ≤ 3.

Better. Now we can rewrite our bound:

1 1 1

e 3
≤e− 1+1+ 2!
+ ··· + n!
≤ ≤
(n + 1)! (n + 1)! (n + 1)!

If we set n = 4 in this we get

 
1 1 1 1 1 3
= ≤e− 1+1+ + + ≤
120 5! 2 6 24 120
1 3 1
So the error is between 120 and 120 = 40 — this approximation isn’t guaranteed to
give us the first 2 decimal places. If we ramp n up to 9 however, we get
 
1 1 1 3
≤ e − 1 + 1 + + ··· + ≤
10! 2 9! 10!

501
 S EQUENCE AND SERIES 3.6 TAYLOR S ERIES

3 3
Since 10! = 3628800, the upper bound on the error is 3628800
< 3000000
= 10−6 , and we
can approximate e by
1 1 1 1 1 1
1+1+ 2!
+ 3!
+ 4!
+ 5!
+ 6!
+ 7!
1 1
+ 8!
+ 9!
=1 + 1 + 0.5 + 0.16̇ + 0.0416̇ + 0.0083̇ + 0.00138̇ + 0.0001984
+ 0.0000248 + 0.0000028
=2.718282

and it is correct to six decimal places.

a Check the derivative!

b The authors hope that by now we all “know” that e is between 2 and 3, but maybe we don’t
know how to prove it.

Example 3.6.15

3.6.3 tt Optional — Linking ex with trigonometric functions

Let us return to the observation that we made earlier about the Maclaurin series for
sine, cosine and the exponential functions:
1 2 1 3 1 4 1 5
cos x + sin x = 1 + x − x − x + x + x − ···
2! 3! 4! 5!
1 1 1 1
e x = 1 + x + x2 + x3 + x4 + x5 + · · ·
2! 3! 4! 5!
We see that these series are identical except for the differences in the signs of the
coefficients. Let us try to make them look even more alike by introducing extra
constants A, B and q into the equations. Consider
A 2 B 3 A 4 B 5
A cos x + B sin x = A + Bx − x − x + x + x − ···
2! 3! 4! 5!
2 3 4
q q q q5
eqx = 1 + qx + x2 + x3 + x4 + x5 + · · ·
2! 3! 4! 5!
Let’s try to choose A, B and q so that these to expressions are equal. To do so we
must make sure that the coefficients of the various powers of x agree. Looking just
at the coefficients of x0 and x1 , we see that we need
A=1 and B=q
Substituting this into our expansions gives
1 2 q 1 q
cos x + q sin x = 1 + qx − x − x3 + x4 + x5 − · · ·
2! 3! 4! 5!
502
S EQUENCE AND SERIES 3.6 TAYLOR S ERIES

q2 2 q3 3 q4 4 q5 5
eqx = 1 + qx + x + x + x + x + ···
2! 3! 4! 5!
Now the coefficients of x0 and x1 agree, but the coefficient of x2 tells us that we need
q to be a number so that q 2 = −1, or
√
q = −1
We know that no such real number q exists. But for the moment let us see what
happens if we just assume 7 that we can find q so that q 2 = −1. Then we will have
that
q 3 = −q q4 = 1 q5 = q ···
so that the series for cos x + q sin x and eqx are identical. That is
eqx = cos x + q sin x
√
If we now write this with the more usual notation q = −1 = i we arrive at what is
now known as Euler’s formula

Equation 3.6.16

eix = cos x + i sin x

Euler’s proof of this formula (in 1740) was based on Maclaurin expansions (much
like our explanation above). Euler’s formula 8 is widely regarded as one of the most
important and beautiful in all of mathematics.
Of course having established Euler’s formula one can find slicker demonstra-
tions. For example, let
f (x) = e−ix (cos x + i sin x)
Differentiating (with product and chain rules and the fact that i2 = −1) gives us
f 0 (x) = −ie−ix (cos x + i sin x) + e−ix (− sin x + i cos x)

7 We do not wish to give a primer on imaginary and complex numbers here. The interested reader
can find many many resources with their favourite search engine.
8 It is worth mentioning here that history of this topic is perhaps a little rough on Roger Cotes
(1682–1716) who was one of the strongest mathematicians of his time and a collaborator of New-
ton. Cotes published a paper on logarithms in 1714 in which he states
ix = log(cos x + i sin x).
(after translating his results into more modern notation). He proved this result by computing in
two different ways the surface area of an ellipse rotated about one axis and equating the results.
Unfortunately Cotes died only 2 years later at the age of 33. Upon hearing of his death Newton
is supposed to have said “If he had lived, we might have known something.” The reader might
think this a rather weak statement, however coming from Newton it was high praise.

503
 S EQUENCE AND SERIES 3.6 TAYLOR S ERIES

=0

Since the derivative is zero, the function f (x) must be a constant. Setting x = 0 tells
us that

f (0) = e0 (cos 0 + i sin 0) = 1.

Hence f (x) = 1 for all x. Rearranging then arrives at

eix = cos x + i sin x

as required.
Substituting x = π into Euler’s formula we get Euler’s identity

eiπ = −1

which is more often stated

Equation 3.6.17 Euler’s identity.

eiπ + 1 = 0

√
which links the 5 most important constants in mathematics, 1, 0, π, e and −1.

3.6.4 tt Evaluating Limits using Taylor Expansions

Taylor polynomials provide a good way to understand the behaviour of a function

near a specified point and so are useful for evaluating complicated limits. Here are
some examples.

Example 3.6.18 A simple limit from a Taylor expansion.

In this example, we’ll start with a relatively simple limit, namely

sin x
lim
x→0 x

The first thing to notice about this limit is that, as x tends to zero, both the numerator,
sin x, and the denominator, x, tend to 0. So we may not evaluate the limit of the ratio
by simply dividing the limits of the numerator and denominator. To find the limit, or
show that it does not exist, we are going to have to exhibit a cancellation between the
numerator and the denominator. Let’s start by taking a closer look at the numerator.
By Example 3.6.6,
1 1
sin x = x − x3 + x5 − · · ·
3! 5!

504
S EQUENCE AND SERIES 3.6 TAYLOR S ERIES

Consequently a
sin x 1 1
= 1 − x2 + x4 − · · ·
x 3! 5!
Every term in this series, except for the very first term, is proportional to a strictly
positive power of x. Consequently, as x tends to zero, all terms in this series, except
for the very first term, tend to zero. In fact the sum of all terms, starting with the
second term, also tends to zero. That is,
h 1 1 i
lim − x2 + x4 − · · · = 0
x→0 3! 5!
We won’t justify that statement here, but it will be justified in the following (optional)
subsection. So
sin x h 1 2 1 4 i
lim = lim 1 − x + x − · · ·
x→0 x x→0 3! 5!
h 1 1 4 i
2
= 1 + lim − x + x − · · ·
x→0 3! 5!
=1

a We are hiding some mathematics behind this “consequently”. What we are really using our
knowledge of Taylor polynomials to write

1 3 1
f (x) = sin(x) = x − x + x5 + E5 (x)
3! 5!
(6)
where E5 (x) = f 6!(c) x6 and c is between 0 and x. We are effectively hiding “E5 (x)” inside the
“· · ·”. Now we can divide both sides by x (assuming x 6= 0):

sin(x) 1 1 E5 (x)
= 1 − x2 + x4 + .
x 3! 5! x
E5 (x)
and everything is fine provided the term x stays well behaved.

Example 3.6.18

The limit in the previous example can also be evaluated relatively easily using
l’Hôpital’s rule 9 . While the following limit can also, in principal, be evaluated using
l’Hôpital’s rule, it is much more efficient to use Taylor series 10 .

9 Many of you learned about l’Hôptial’s rule in school and all of you should have seen it last term
in your differential calculus course.
10 It takes 3 applications of l’Hôpital’s rule and some careful cleaning up of the intermediate ex-
pressions. Oof!

505
S EQUENCE AND SERIES 3.6 TAYLOR S ERIES

Example 3.6.19 A not so easy limit made easier.

In this example we evaluate

arctan x − x
lim
x→0 sin x − x
Once again, the first thing to notice about this limit is that, as x tends to zero, the
numerator tends to arctan 0 − 0, which is 0, and the denominator tends to sin 0 − 0,
which is also 0. So we may not evaluate the limit of the ratio by simply dividing the
limits of the numerator and denominator. Again, to find the limit, or show that it
does not exist, we are going to have to exhibit a cancellation between the numerator
and the denominator. To get a more detailed understanding of the behaviour of
the numerator and denominator near x = 0, we find their Taylor expansions. By
Example 3.5.21,
x3 x 5
arctan x = x − + − ···
3 5
so the numerator
x3 x5
arctan x − x = − + − ···
3 5
By Example 3.6.6,
1 1
sin x = x − x3 + x5 − · · ·
3! 5!
so the denominator
1 1
sin x − x = − x3 + x5 − · · ·
3! 5!
and the ratio 3 5
arctan x − x − x3 + x5 − · · ·
= 1 3 1 5
sin x − x − 3! x + 5! x − · · ·
Notice that every term in both the numerator and the denominator contains a com-
mon factor of x3 , which we can cancel out.
2
arctan x − x −1 + x − · · ·
= 13 15 2
sin x − x − 3! + 5! x − · · ·

As x tends to zero,

• the numerator tends to − 13 , which is not 0, and

• the denominator tends to − 3!1 = − 61 , which is also not 0.

so we may now legitimately evaluate the limit of the ratio by simply dividing the
limits of the numerator and denominator.
2
arctan x − x − 13 + x5 − · · ·
lim = lim 1
x→0 sin x − x x→0 − + 1 x2 − · · ·
3! 5!

506
 S EQUENCE AND SERIES 3.6 TAYLOR S ERIES

2
limx→0 − 31 + x5 − · · ·
 
=
limx→0 − 3!1 + 5!1 x2 − · · ·
 

− 31
=
− 3!1
=2
Example 3.6.19

3.6.5 tt Optional — The Big O Notation

In Example 3.6.18 we used, without justification 11 , that, as x tends to zero, not only
does every term in
∞
sin x 1 1 X 1
− 1 = − x2 + x4 − · · · = (−1)n x2n
x 3! 5! n=1
(2n + 1)!

converge to zero, but in fact the sum of all infinitely many terms also converges to
zero. We did something similar twice in Example 3.6.19; once in computing the limit
of the numerator and once in computing the limit of the denominator.
We’ll now develop some machinery that provides the justification. We start by
recalling, from equation 3.6.1, that if, for some natural number n, the function f (x)
has n + 1 derivatives near the point a, then

f (x) = Tn (x) + En (x)

where
Tn (x) = f (a) + f 0 (a) (x − a) + · · · + 1 (n)
n!
f (a) (x − a)n
is the Taylor polynomial of degree n for the function f (x) and expansion point a and

En (x) = f (x) − Tn (x) = 1

(n+1)!
f (n+1) (c) (x − a)n+1

is the error introduced when we approximate f (x) by the polynomial Tn (x). Here
c is some unknown number between a and x. As c is not known, we do not know
exactly what the error En (x) is. But that is usually not a problem.
In the present context 12 we are interested in taking the limit as x → a. So we are
only interested in x-values that are very close to a, and because c lies between x and
a, c is also very close to a. Now, as long as f (n+1) (x) is continuous at a, as x → a,
f (n+1) (c) must approach f (n+1) (a) which is some finite value. This, in turn, means

11 Though there were a few comments in a footnote.

12 It is worth pointing out that our Taylor series must be expanded about the point to which we are
limiting — i.e. a to work out a limit as x → a we need Taylor series expanded about a and not
some other point.

507
S EQUENCE AND SERIES 3.6 TAYLOR S ERIES

that there must be constants M, D > 0 such that f (n+1) (c) ≤ M for all c’s within a
M
distance D of a. If so, there is another constant C (namely (n+1)! ) such that

En (x) ≤ C|x − a|n+1 whenever |x − a| ≤ D

There is some notation for this behaviour.

Definition 3.6.20 Big O.

Let a and m be real numbers. We say that

the function “g(x) is of order |x − a|m
near a” and we write g(x) = O |x − a|m if there exist constants a C, D > 0 such
that

g(x) ≤ C|x − a|m whenever |x − a| ≤ D (?)

Whenever O |x − a|m appears in an algebraic expression, it just stands for
some (unknown) function g(x) that obeys (?). This is called “big O” notation.

a To be precise, C and D do not depend on x, though they may, and usually do, depend on
m.

How should we parse the big O notation when we see it? Consider the following

g(x) = O(|x − 3|2 )

First of all, we know from the definition that the notation only tells us something
about g(x) for x near the point a. The equation above contains “O(|x − 3|2 )” which
tells us something about what the function looks like when x is close to 3. Further,
because it is “|x − 3|” squared, it says that the graph of the function lies below a
parabola y = C(x − 3)2 and above a parabola y = −C(x − 3)2 near x = 3. The
notation doesn’t tell us anything more than this — we don’t know, for example,
that the graph of g(x) is concave up or concave down. It also tells us that Taylor
expansion of g(x) around x = 3 does not contain any constant or linear term — the
first nonzero term in the expansion is of degree at least two. For example, all of the
following functions are O(|x − 3|2 ).
5
5(x − 3)2 + 6(x − 3)3 , −7(x − 3)2 − 8(x − 3)4 , (x − 3)3 , (x − 3) 2

In the next few examples we will rewrite a few of the Taylor polynomials that we
know using this big O notation.

Example 3.6.21 Sine and the big O.

Let f (x) = sin x and a = 0. Then

f (x) = sin x f 0 (x) = cos x f 00 (x) = − sin x f (3) (x) = − cos x

f (0) = 0 f 0 (0) = 1 f 00 (0) = 0 f (3) (0) = −1

508
S EQUENCE AND SERIES 3.6 TAYLOR S ERIES

f (4) (x) = sin x ···

f (4) (0) = 0 ···

and the pattern repeats. So every derivative is plus or minus either sine or cosine
and, as we saw in previous examples, this makes analysing the error term for the
sine and cosine series quite straightforward. In particular, f (n+1) (c) ≤ 1 for all real
numbers c and all natural numbers n. So the Taylor polynomial of, for example,
degree 3 and its error term are

sin x = x − 3!1 x3 + cos

5!
c 5
x
= x − 3!1 x3 + O(|x|5 )
1
under Definition 3.6.20, with C = 5!
and any D > 0. Similarly, for any natural
number n,

Equation 3.6.22

sin x = x − 3!1 x3 + 5!1 x5 − · · · + (−1)n (2n+1)!

1
x2n+1 + O |x|2n+3

cos x = 1 − 2!1 x2 + 4!1 x4 − · · · + (−1)n (2n)!

1
x2n + O |x|2n+2

Example 3.6.22

When we studied the error in the expansion of the exponential function (way
back in optional Example 3.6.8), we had to go to some length to understand the
behaviour of the error term well enough to prove convergence for all numbers x.
However, in the big O notation, we are free to assume that x is close to 0. Further-
more we do not need to derive an explicit bound on the size of the coefficient C. This
makes it quite a bit easier to verify that the big O notation is correct.

Example 3.6.23 Exponential and the big O.

d x dk
Let n be any natural number. Since dx
e = ex , we know that dxk
{ex } = ex for every
integer k ≥ 0. Thus
x2 x3 xn ec
ex = 1 + x + 2!
+ 3!
+ ··· + n!
+ (n+1)!
xn+1

for some c between 0 and x. If, for example, |x| ≤ 1, then |ec | ≤ e, so that the error
term
ec
(n+1)!
xn+1 ≤ C|x|n+1 e
with C = (n+1)! whenever |x| ≤ 1
e
So, under Definition 3.6.20, with C = (n+1)!
and D = 1,

509
S EQUENCE AND SERIES 3.6 TAYLOR S ERIES

Equation 3.6.24

x2 x3 xn
ex = 1 + x + + O |x|n+1


2!
+ 3!
+ ··· + n!

You can see that, because we only have to consider x’s that are close to the expansion
point (in this example, 0) it is relatively easy to derive the bounds that are required
to justify the use of the big O notation.
Example 3.6.24

Example 3.6.25 Logarithms and the big O.

Let f (x) = log(1 + x) and a = 0. Then

f 0 (x) = 1
1+x
f 00 (x) = − (1+x)
1
2 f (3) (x) = 2
(1+x)3

f 0 (0) = 1 f 00 (0) = −1 f (3) (0) = 2

f (4) (x) = − (1+x)
2×3
4 f (5) (x) = 2×3×4
(1+x)5

f (4) (0) = −3! f (5) (0) = 4!

We can see a pattern for f (n) (x) forming here — f (n) (x) is a sign times a ratio with
• the sign being + when n is odd and being − when n is even. So the sign is
(−1)n−1 .

• The denominator is (1 + x)n .

• The numerator a is the product 2 × 3 × 4 × · · · × (n − 1) = (n − 1)!.

Thus b , for any natural number n,
(n−1)!
f (n) (x) = (−1)n−1 (1+x) n which means that
1 (n)
n!
f (0) xn = (−1)n−1 (n−1)!
n!
xn = (−1) n−1 xn
n

so
x2 x3 n
log(1 + x) = x − 2
+ 3
− · · · + (−1)n−1 xn + En (x)
with
(−1)n
En (x) = 1
(n+1)!
f (n+1) (c) (x − a)n+1 = 1
n+1
· (1+c)n+1
· xn+1
If we choose, for example D = 12 , then c for any x obeying |x| ≤ D = 21 , we have
|c| ≤ 12 and |1 + c| ≥ 12 so that
1 n+1
= O |x|n+1


|En (x)| ≤ (n+1)(1/2)n+1 |x|

510
S EQUENCE AND SERIES 3.6 TAYLOR S ERIES

2n+1
under Definition 3.6.20, with C = n+1
and D = 12 . Thus we may write

Equation 3.6.26

x2 x3 n
− · · · + (−1)n−1 xn + O |x|n+1


log(1 + x) = x − 2
+ 3

a remember that n! = 1 × 2 × 3 × · · · × n, and that we use the convention 0! = 1.

b It is not too hard to make this rigorous using the principle of mathematical induction. The in-
terested reader should do a little search-engine-ing. Induction is a very standard technique for
proving statements of the form “For every natural number n,. . . ”. For example
n
X n(n + 1)
For every natural number n, k= or
2
k=1
n
d (n − 1)!
For every natural number n, {log(1 + x)} = (−1)n−1
dxn (1 + x)n

It was also used by Polya (1887–1985) to give a very convincing (but subtly (and deliberately)
flawed) proof that all horses have the same colour.
c Since |c| ≤ 21 , − 12 ≤ c ≤ 21 . If we now add 1 to every term we get 12 ≤ 1 + c ≤ 23 and so |1 + c| ≥ 12 .
You can also do this with the triangle inequality which tells us that for any x, y we know that
|x + y| ≤ |x| + |y|. Actually, you want the reverse triangle inequality (which is a simple corollary
of the triangle inequality) which says that for any x, y we have |x + y| ≥ |x| − |y| .

Example 3.6.26

Remark 3.6.27 The big O notation has a few properties that are useful in computa-
tions and taking limits. All follow immediately from Definition 3.6.20.
a If p > 0, then
lim O(|x|p ) = 0
x→0

b For any real numbers p and q,

O(|x|p ) O(|x|q ) = O(|x|p+q )

(This is just because C|x|p × C 0 |x|q = (CC 0 )|x|p+q .) In particular,

axm O(|x|p ) = O(|x|p+m )

for any constant a and any integer m.

c For any real numbers p and q,

O(|x|p ) + O(|x|q ) = O(|x|min{p,q} )

511
 S EQUENCE AND SERIES 3.6 TAYLOR S ERIES

(For example, if p = 2 and q = 5, then C|x|2 + C 0 |x|5 = C + C 0 |x|3 |x|2 ≤

(C + C 0 )|x|2 whenever |x| ≤ 1.)

d For any real numbers p and q with p > q, any function which is O(|x|p ) is also
O(|x|q ) because C|x|p = C|x|p−q |x|q ≤ C|x|q whenever |x| ≤ 1.

e All of the above observations also hold for more general expressions with |x|
replaced by |x − a|, i.e. for O(|x − a|p ). The only difference being in (a) where
we must take the limit as x → a instead of x → 0.

3.6.6 tt Optional — Evaluating Limits Using Taylor Expansions — More

Examples

512
S EQUENCE AND SERIES 3.6 TAYLOR S ERIES

Example 3.6.28 Example 3.6.18 revisited.

In this example, we’ll return to the limit

sin x
lim
x→0 x

of Example 3.6.18 and treat it more carefully. By Example 3.6.21,

1 3
sin x = x − x + O(|x|5 )
3!
That is, for small x, sin x is the same as x − 3!1 x3 , up to an error that is bounded by
some constant times |x|5 . So, dividing by x, sinx x is the same as 1 − 3!1 x2 , up to an error
that is bounded by some constant times x4 — see Remark 3.6.27(b). That is
sin x 1
= 1 − x2 + O(x4 )
x 3!
But any function that is bounded by some constant times x4 (for all x smaller than
some constant D > 0) necessarily tends to 0 as x → 0 — see Remark 3.6.27(a). . Thus
sin x h 1 i h 1 i
lim = lim 1 − x2 + O(x4 ) = lim 1 − x2 = 1
x→0 x x→0 3! x→0 3!
Reviewing the above computation, we see that we did a little more work than we had
to. It wasn’t necessary to keep track of the − 3!1 x3 contribution to sin x so carefully.
We could have just said that
sin x = x + O(|x|3 )
so that
sin x x + O(|x|3 )
= lim 1 + O(x2 ) = 1
 
lim = lim
x→0 x x→0 x x→0
We’ll spend a little time in the later, more complicated, examples learning how to
choose the number of terms we keep in our Taylor expansions so as to make our
computations as efficient as possible.

Example 3.6.29 Practicing using Taylor polynomials for limits.

In this example, we’ll use the Taylor polynomial of Example 3.6.25 to evaluate
lim log(1+x)
x
and lim (1 + x)a/x . The Taylor expansion of equation 3.6.26 with n = 1
x→0 x→0
tells us that
log(1 + x) = x + O(|x|2 )
That is, for small x, log(1+x) is the same as x, up to an error that is bounded by some
constant times x2 . So, dividing by x, x1 log(1 + x) is the same as 1, up to an error that
is bounded by some constant times |x|. That is
1
log(1 + x) = 1 + O(|x|)
x
But any function that is bounded by some constant times |x|, for all x smaller than

513
S EQUENCE AND SERIES 3.6 TAYLOR S ERIES

some constant D > 0, necessarily tends to 0 as x → 0. Thus

log(1 + x) x + O(|x|2 )  
lim = lim = lim 1 + O(|x|) = 1
x→0 x x→0 x x→0

We can now use this limit to evaluate

lim (1 + x)a/x .
x→0

Now, we could either evaluate the limit of the logarithm of this expression, or we
can carefully rewrite the expression as e(something) . Let us do the latter.
a
lim (1 + x)a/x = lim e x log(1+x)
x→0 x→0
a 2 )]
= lim e x [x+O(|x|
x→0

= lim ea+O(|x|) = ea
x→0

Here we have used that if F (x) = O(|x|2 ) then xa F (x) = O(x) — see Remark 3.6.27(b).
We have also used that the exponential is continuous — as x tends to zero, the expo-
nent of ea+O(|x|) tends to a so that ea+O(|x|) tends to ea — see Remark 3.6.27(a).
Example 3.6.29

Example 3.6.30 A difficult limit.

In this example, we’ll evaluate a the harder limit

cos x − 1 + 21 x sin x
lim
x→0 [log(1 + x)]4
The first thing to notice about this limit is that, as x tends to zero, the numerator
1 1
cos x − 1 + x sin x → cos 0 − 1 + · 0 · sin 0 = 0
2 2
and the denominator
[log(1 + x)]4 → [log(1 + 0)]4 = 0
too. So both the numerator and denominator tend to zero and we may not simply
evaluate the limit of the ratio by taking the limits of the numerator and denominator
and dividing.
To find the limit, or show that it does not exist, we are going to have to exhibit a
cancellation between the numerator and the denominator. To develop a strategy for
evaluating this limit, let’s do a “little scratch work”, starting by taking a closer look
at the denominator. By Example 3.6.25,
log(1 + x) = x + O(x2 )

514
S EQUENCE AND SERIES 3.6 TAYLOR S ERIES

This tells us that log(1 + x) looks a lot like x for very small x. So the denominator
[x + O(x2 )]4 looks a lot like x4 for very small x. Now, what about the numerator?
• If the numerator looks like some constant times xp with p > 4, for very small x,
p
then the ratio will look like the constant times xx4 = xp−4 and, as p − 4 > 0, will
tend to 0 as x tends to zero.
• If the numerator looks like some constant times xp with p < 4, for very small x,
p
then the ratio will look like the constant times xx4 = xp−4 and will, as p − 4 < 0,
tend to infinity, and in particular diverge, as x tends to zero.
• If the numerator looks like Cx4 , for very small x, then the ratio will look like
Cx4
x4
= C and will tend to C as x tends to zero.
The moral of the above “scratch work” is that we need to know the behaviour of the
numerator, for small x, up to order x4 . Any contributions of order xp with p > 4 may
be put into error terms O(|x|p ).
Now we are ready to evaluate the limit. Because the expressions are a little involved,
we will simplify the numerator and denominator separately and then put things
together. Using the expansions we developed in Example 3.6.21, the numerator,
 
1 1 2 1 4 6
cos x − 1 + x sin x = 1 − x + x + O(|x| )
2 2! 4!
 
x 1 3 5
−1+ x − x + O(|x| ) expand
2 3!
 
1 1 x
= − x4 + O(|x|6 ) + O(|x|5 )
24 12 2
Then by Remark 3.6.27(b)
1 4
=− x + O(|x|6 ) + O(|x|6 )
24
and now by Remark3.6.27(c)
1 4
=− x + O(|x|6 )
24
Similarly, using the expansion that we developed in Example 3.6.25,
4
[log(1 + x)]4 = x + O(|x|2 )

 4
= x + xO(|x|) by Remark 3.6.27(b)
4 4
= x [1 + O(|x|)]

Now put these together and take the limit as x → 0:

cos x − 1 + 21 x sin x 1 4
− 24 x + O(|x|6 )
lim = lim
x→0 [log(1 + x)]4 x→0 x4 [1 + O(|x|)]4

515
S EQUENCE AND SERIES 3.6 TAYLOR S ERIES

1 4
− 24 x + x4 O(|x|2 )
= lim by Remark 3.6.27(b)
x→0 x4 [1 + O(|x|)]4
− 1 + O(|x|2 )
= lim 24
x→0 [1 + O(|x|)]4
1
=− by Remark 3.6.27(a).
24

a Use of l’Hôpital’s rule here could be characterised as a “courageous decision”. The interested
reader should search-engine their way to Sir Humphrey Appleby and “Yes Minister” to better
understand this reference (and the workings of government in the Westminster system). Discre-
tion being the better part of valour, we’ll stop and think a little before limiting (ha) our choices.

Example 3.6.30

The next two limits have much the same flavour as those above — expand the
numerator and denominator to high enough order, do some cancellations and then
take the limit. We have increased the difficulty a little by introducing “expansions of
expansions”.

Example 3.6.31 Another difficult limit.

In this example we’ll evaluate another harder limit, namely

log sinx x


lim
x→0 x2
The first thing to notice about this limit is that, as x tends to zero, the denominator
x2 tends to 0. So, yet again, to find the limit, we are going to have to show that the
numerator also tends to 0 and we are going to have to exhibit a cancellation between
the numerator and the denominator.
Because the denominator is x2 any terms in the numerator, log sinx x that are of order


log( sin x )
x3 or higher will contribute terms in the ratio x2x that are of order x or higher.
Those terms in the ratio will converge to zero as x → 0. The moral of this discussion
is that we need to compute log sinx x to order x2 with errors of order x3 . Now we saw,
in Example 3.6.28, that
sin x 1
= 1 − x2 + O(x4 )
x 3!
We also saw, in equation 3.6.26 with n = 1, that

log(1 + X) = X + O(X 2 )

Substituting a X = − 3!1 x2 +O(x4 ), and using that X 2 = O(x4 ) (by Remark 3.6.27(b,c)),
we have that the numerator
 sin x  1
log = log(1 + X) = X + O(X 2 ) = − x2 + O(x4 )
x 3!

516
S EQUENCE AND SERIES 3.6 TAYLOR S ERIES

and the limit

sin x
− 3!1 x2 + O(x4 )


log x
h 1
2
i 1
lim 2
= lim 2
= lim − + O(x ) = −
x→0 x x→0 x x→0 3! 3!
1
=−
6

a In our derivation of log(1+X) = X +O(X 2 ) in Example 3.6.25, we required only that |X| ≤ 12 . So
1 2 1 2
we are free to substitute X = − 3! x +O(x4 ) for any x that is small enough that − 3! x +O(x4 ) <
1
2.

Example 3.6.31

Example 3.6.32 Yet another difficult limit.

Evaluate 2
ex − cos x
lim
x→0 log(1 + x) − sin x

Solution: Step 1: Find the limit of the denominator.

 
lim log(1 + x) − sin x = log(1 + 0) − sin 0 = 0
x→0

This tells us that we can’t evaluate the limit just by finding the limits of the numera-
tor and denominator separately and then dividing.
Step 2: Determine the leading order behaviour of the denominator near x = 0. By
equations 3.6.26 and 3.6.22,
log(1 + x) = x − 21 x2 + 13 x3 − · · ·
sin x = x − 3!1 x3 + 5!1 x5 − · · ·
Taking the difference of these expansions gives
log(1 + x) − sin x = − 21 x2 + 1 1

3
3
+ 3!
x + ···
2
This tells us that, for x near zero, the denominator is − x2 (that’s the leading order
term) plus contributions that are of order x3 and smaller. That is
2
log(1 + x) − sin x = − x2 + O(|x|3 )
Step 3: Determine the behaviour of the numerator near x = 0 to order x2 with errors
of order x3 and smaller (just like the denominator). By equation 3.6.24
eX = 1 + X + O X 2

Substituting X = x2
2
ex = 1 + x2 + O x4

517
 S EQUENCE AND SERIES 3.6 TAYLOR S ERIES

cos x = 1 − 21 x2 + O x4

by equation 3.6.22. Subtracting, the numerator

2
ex − cos x = 23 x2 + O x4

Step 4: Evaluate the limit.

2 3 2
ex − cos x x + O(x4 )
lim = lim 2 x2
x→0 log(1 + x) − sin x x→0 − + O(|x|3 )
2
3
2
+ O(x2 )
= lim
x→0 − 1 + O(|x|)
2
3
2
= = −3
− 12
Example 3.6.32

3.6.7 tt Exercises

Exercises — Stage 1

1. Below is a graph of y = f (x), along with the constant approximation,

linear approximation, and quadratic approximation centred at a = 2.
Which is which?
y
C

y = f (x)
B

x
2

2. Suppose T (x) is the Taylor series for f (x) = arctan3 (ex + 7) centred at a = 5.
What is T (5)?

3. Below are a list of common functions, and their Taylor series represen-
tations. Match the function to the Taylor series.

518
S EQUENCE AND SERIES 3.6 TAYLOR S ERIES

function series
∞
1 X xn+1
A. I. (−1)n
1−x n=0
n+1
∞
X x2n+1
B. log(1 + x) II. (−1)n
n=0
(2n + 1)!
∞
X x2n
C. arctan x III. (−1)n
n=0
(2n)!
∞
X x2n+1
D. ex IV. (−1)n
n=0
2n + 1
X ∞
E. sin x V. xn
n=0
∞
X xn
F. cos x VI.
n=0
n!

4.
∞
X n2
a Suppose f (x) = (x − 3)n for all real x. What is f (20) (3)
n=0
(n! + 1)
(the twentieth derivative of f (x) at x = 3)?
∞
X n2
b Suppose g(x) = (x − 3)2n for all real x. What is g (20) (3)?
n=0
(n! + 1)

arctan(5x2 )
c If h(x) = , what is h(20) (0)? What is h(22) (0)?
x4

Exercises — Stage 2 In Questions 5 through 8, you will create Taylor series from
scratch. In practice, it is often preferable to modify an existing series, rather than
creating a new one, but you should understand both ways.
5. Using the definition of a Taylor series, find the Taylor series for f (x) =
log(x) centred at x = 1.
6. Find the Taylor series for f (x) = sin x centred at a = π.

1
7. Using the definition of a Taylor series, find the Taylor series for g(x) =
x
centred at x = 10. What is the interval of convergence of the resulting
series?

519
S EQUENCE AND SERIES 3.6 TAYLOR S ERIES

8. Using the definition of a Taylor series, find the Taylor series for h(x) =
e3x centred at x = a, where a is some constant. What is the radius of
convergence of the resulting series?

In Questions 9 through 16, practice creating new Taylor series by modifying known
Taylor series, rather than creating your series from scratch.

1
9. *. Find the Maclaurin series for f (x) = 2x − 1 .

∞
X 3 1
10. *. Let bn xn be the Maclaurin series for f (x) = − ,
n=0
x + 1 2x − 1
∞
X 3 1
i.e. bn x n = − .
n=0
x + 1 2x − 1
Find bn .
11. *. Find the coefficient c5 of the fifth degree term in the Maclaurin series
X∞
cn xn for e3x .
n=0

12. *. Express the Taylor series of the function

f (x) = log(1 + 2x)

about x = 0 in summation notation.

2 3 5 11
13. *. The first two terms in the Maclaurin series for x sin(x ) are ax + bx ,
where a and b are constants. Find the values of a and b.

14. *. Give
2
the first two nonzero terms in the Maclaurin series for
e−x − 1
Z
dx.
x
Z
15. *. Find the Maclaurin series for x4 arctan(2x) dx.

df x
16. *. Suppose that dx = 1 + 3x3 and f (0) = 1. Find the Maclaurin series
for f (x).

In past chapters, we were only able to exactly evaluate very specific types of series:
geometric and telescoping. In Questions 17 through 25, we expand our range by
relating given series to Taylor series.

520
S EQUENCE AND SERIES 3.6 TAYLOR S ERIES

17. *. The Maclaurin series for arctan x is given by

∞
X x2n+1
arctan x = (−1)n
n=0
2n + 1

which has radius of convergence equal to 1. Use this fact to compute the
exact value of the series below:
∞
X (−1)n
n=0
(2n + 1)3n

∞
X (−1)n
18. *. Evaluate n!
.
n=0

∞
X 1
19. *. Evaluate e k k!
.
k=0

∞
X 1
20. *. Evaluate the sum of the convergent series π k k!
.
k=1

∞
X (−1)n−1
21. *. Evaluate n 2n
.
n=1

∞
X n+2
22. *. Evaluate en .
n=1
n!

∞
X 2n
23. Evaluate , or show that it diverges.
n=1
n
24. Evaluate ∞
X (−1)n  π 2n+1
1 + 22n+1


n=0
(2n + 1)! 4
or show that it diverges.

∞
X x2n
25. *. (a) Show that the power series (2n)!
converges absolutely for all
n=0
x.
real numbers ∞
X 1
(b) Evaluate .
n=0
(2n)!

521
S EQUENCE AND SERIES 3.6 TAYLOR S ERIES

26.
π
a Using the fact that arctan(1) = , how many terms of the Taylor
4
series for arctangent would you have to add up to approximate π
with an error of at most 4 × 10−5 ?

b Example 3.6.11 mentions the formula

1 1
π = 16 arctan − 4 arctan
5 239
Using the Taylor series for arctangent, how many terms would
you have to add up to approximate π with an error of at most
4 × 10−5 ?

c Assume without proof the following:

 
1 1 3+2
arctan + arctan = arctan
2 3 2·3−1

Using the Taylor series for arctangent, how many terms would
you have to add up to approximate π with an error of at most
4 × 10−5 ?

27. Suppose you wanted to approximate the number log(1.5) as a rational num-
ber using the Taylor expansion of log(1 + x). How many terms would you
need to add to get 10 decimal places of accuracy? (That is, an absolute error
less than 5 × 10−11 .)
28. Suppose you wanted to approximate the number e as a rational number
using the Maclaurin expansion of ex . How many terms would you need to
add to get 10 decimal places of accuracy? (That is, an absolute error less
than 5 × 10−11 .)
You may assume without proof that 2 < e < 3.

29. Suppose you wanted to approximate the number log(0.9) as a rational

number using the Taylor expansion of log(1 − x). Which partial sum
should you use to get 10 decimal places of accuracy? (That is, an abso-
lute error less than 5 × 10−11 .)

30. Define the hyperbolic sine function as

ex − e−x
sinh x = .
2
Suppose you wanted to approximate the number sinh(b) using the Maclau-
rin series of sinh x, where b is some number in (−2, 1). Which partial sum
should you use to guarantee 10 decimal places of accuracy? (That is, an

522
S EQUENCE AND SERIES 3.6 TAYLOR S ERIES

absolute error less than 5 × 10−11 .)

You may assume without proof that 2 < e < 3.
31. Let f (x) be a function with

(n − 1)! 
f (n) (x) = (1 − x)−n + (−1)n−1 (1 + x)−n

2
for all n ≥ 1.
Give reasonable bounds
(both upper and lower) on the error involved in
1
approximating f − 3 using the partial sum S6 of the Taylor series for f (x)
centred at a = 21 .
Remark: One function with this quality is the inverse hyperbolic tangent
functiona .

ex − e−x
a Of course it is! Actually, hyperbolic tangent is tanh(x) = , and inverse hy-
ex + e−x
perbolic tangent is its functional inverse.

Exercises — Stage 3

1 − cos x
32. lim
*. Use series to evaluate x→0 .
1 + x − ex

x3
sin x − x + 6
33. lim
*. Evaluate x→0 .
x5

2/x
34. Evaluate lim 1 + x + x2 using a Taylor series for the natural logarithm.
x→0

35. Use series to evaluate  x

1
lim 1 +
x→∞ 2x

∞
X (n + 1)(n + 2)
36. Evaluate the series or show that it diverges.
n=0
7n

∞
X (−1)n x2n+4
37. Write the series f (x) = as a combination of familiar
n=0
(2n + 1)(2n + 2)
functions.

38.

a Find the Maclaurin series for f (x) = (1 − x)−1/2 . What is its radius
of convergence?

523
S EQUENCE AND SERIES 3.6 TAYLOR S ERIES

b Manipulate the series you just found to find the Maclaurin series
for g(x) = arcsin x. What is its radius of convergence?

39. *. Find the Taylor series for f (x) = log(x) centred at a = 2. Find the interval
of convergence for this series.
Z x
1
40. *. Let I(x) = 4
dt.
0 1+t

a Find the Maclaurin series for I(x).

b Approximate I(1/2) to within ±0.0001.

c Is your approximation in (b) larger or smaller than the true value of

I(1/2)? Explain.
41. *. Using a Maclaurin series, the number a = 1/5 − 1/7 + 1/18 is found to be
Z 1
2
an approximation for I = x4 e−x dx. Give the best upper bound you can
0
for |I − a|.
42. *. Find an interval of length 0.0002 or less that contains the number
Z 1
2 2
I= x2 e−x dx
0
x −t
e −1
Z
43. *. Let I(x) = t
dt.
0

a Find the Maclaurin series for I(x).

b Approximate I(1) to within ±0.01.

c Explain why your answer to part (b) has the desired accuracy.

x
sin t
Z
44. *. The function Σ(x) is defined by Σ(x) = t
dt.
0

a Find the Maclaurin series for Σ(x).

b It can be shown that Σ(x) has an absolute maximum which occurs

at its smallest positive critical point (see the graph of Σ(x) below).
Find this critical point.

c Use the previous information to find the maximum value of Σ(x)

to within ±0.01.

524
S EQUENCE AND SERIES 3.6 TAYLOR S ERIES

x
cos t − 1
Z
45. *. Let I(x) = t2
dt.
0

a Find the Maclaurin series for I(x).

b Use this series to approximate I(1) to within ±0.01

c Is your estimate in (b) greater than I(1)? Explain.

x
cos t + t sin t − 1
Z
46. *. Let I(x) = t2
dt
0

a Find the Maclaurin series for I(x).

b Use this series to approximate I(1) to within ±0.001

c Is your estimate in (b) greater than or less than I(1)?

Z x
1 − e−t
47. *. Define f (x) = dt.
0 t
∞
X (−1)n−1
a Show that the Maclaurin series for f (x) is xn .
n=1
n · n!

b Use the∞ratio test to determine the values of x for which the Maclaurin
X (−1)n−1
series xn converges.
n=1
n · n!
Z 1
x3 1
48. *. Show that x
dx ≤ .
0 e −1 3
ex + e−x
49. *. Let cosh(x) = 2
.

a Find the power series expansion of cosh(x) about x0 = 0 and deter-

mine its interval of convergence.
b Show that 3 23 ≤ cosh(2) ≤ 3 23 + 0.1.
1 2
c Show that cosh(t) ≤ e 2 t for all t.

525
S EQUENCE AND SERIES 3.6 TAYLOR S ERIES

50. The law of the instrument says “If you have a hammer then everything
looks like a nail” — it is really a description of the “tendency of jobs to
be adapted to tools rather than adapting tools to jobs”a . Anyway, this is
a long way of saying that just because we know how to compute things
using Taylor series doesn’t mean we should neglect other techniques.
√
a Using Newton’s method, approximate the constant 3 2 as a root of
the function g(x) = x3 − 2. Using a calculator, make your estima-
tion accurate to within 0.01.

b You may assume without proof that

∞
√ 1 X (2)(5)(8) · · · (3n − 4)
3
x = 1 + (x − 1) + (−1)n−1 n
(x − 1)n .
6 n=2
3 n!

for all real numbers x. Using the fact that this is an alternating
series, how many√ terms would you have to add for the partial
3
sum to estimate 2 with an error less than 0.01?

a Quote from Silvan Tomkins’s Computer Simulation of Personality: Frontier of Psycho-

logical Theory. See also Birmingham screwdrivers.

Let f (x) = arctan(x3 ). Write f (10) 15 as a sum of rational numbers with an

51.
error less than 10−6 using the Maclaurin series for arctangent.
52. Consider the following function:
( 2
e−1/x x 6= 0
f (x) =
0 x=0

a Sketch y = f (x).

b Assume (without proof) that f (n) (0) = 0 for all whole numbers n. Find
the Maclaurin series for f (x).

c Where does the Maclaurin series for f (x) converge?

d For which values of x is f (x) equal to its Maclaurin series?

∞
X f (n) (0)
53. Suppose f (x) is an odd function, and f (x) = xn . Simplify
n=0
n!
∞ (2n)
X f (0) 2n
x .
n=0
(2n)!

526
 S EQUENCE AND SERIES 3.7 O PTIONAL — R ATIONAL AND IRRATIONAL NUMBERS

3.7q Optional — Rational and irrational numbers

In this optional section we shall use series techniques to look a little at rationality
and irrationality of real numbers. We shall see the following results.
• A real number is rational (i.e. a ratio of two integers) if and only if its decimal
expansion is eventually periodic. “Eventually periodic” means that, if we de-
note the nth decimal place by dn , then there are two positive integers k and p
such that dn+p = dn whenever n > k. So the part of the decimal expansion after
the decimal point looks like

. a1 a2 a3 · · · ak b1 b2 · · · bp b1 b2 · · · bp b1 b2 · · · bp · · ·
| {z } | {z } | {z } | {z }
It is possible that a finite number of decimal places right after the decimal point
do not participate in the periodicity. It is also possible that p = 1 and b1 = 0, so
that the decimal expansion ends with an infinite string of zeros.
• e is irrational.
• π is irrational.

3.7.1 tt Decimal expansions of rational numbers

We start by showing that a real number is rational if and only if its decimal expansion
is eventually periodic. We need only consider the expansions of numbers 0 < x <
1. If a number is negative then we can just multiply it by −1 and not change the
expansion. Similarly if the number is larger than 1 then we can just subtract off the
integer part of the number and leave the expansion unchanged.

3.7.2 tt Eventually periodic implies rational

Let us assume that a number 0 < x < 1 has a decimal expansion that is eventually
periodic. Hence we can write

x = 0. a1 a2 a3 · · · ak b1 b2 · · · bp b1 b2 · · · bp b1 b2 · · · bp · · ·
| {z } | {z } | {z } | {z }
Let α = a1 a2 a3 · · · ak and β = b1 b2 · · · bp . In particular, α has at most k digits and β
has at most p digits. Then we can (carefully) write
α β β β
x= k
+ k+p + k+2p + k+3p + · · ·
10 10 10 10
527
 S EQUENCE AND SERIES 3.7 O PTIONAL — R ATIONAL AND IRRATIONAL NUMBERS

∞
α β X −p
= k + k+p 10
10 10 j=0

This sum is just a geometric series (see Lemma 3.2.5) and we can evaluate it

α β 1 α β 1
= k
+ k+p · −p
= k+ k· p
10  10 1 −
10 10 10 10 − 1
1 β α(10p − 1) + β
= k α+ p =
10 10 − 1 10k (10p − 1)

This is a ratio of integers, so x is a rational number.

3.7.3 tt Rational implies eventually periodic

Let 0 < x < 1 be rational with x = ab , where a and b are positive integers. We wish to
show that x’s decimal expansion is eventually periodic. Start by looking at the last
formula we derived in the “eventually periodic implies rational” subsection. If we
k p
can express the denominator b in the form 10 (10q −1) with k, p and q integers, we will
aq
be in business because ab = 10k (10p −1) . From this we can generate the desired decimal

expansion by running the argument of the last subsection backwards. So we want to

find integers k, p, q such that 10k+p − 10k = b · q. To do so consider the powers of 10
up to 10b :

1, 101 , 102 , 103 , · · · , 10b

For each j = 0, 1, 2, · · · , b, find integers cj and 0 ≤ rj < b so that

10j = b · cj + rj

To do so, start with 10j and repeatedly subtract b from it until the remainder drops
strictly below b. The rj ’s can take at most b different values, namely 0, 1, 2, · · ·, b − 1,
and we now have b + 1 rj ’s, namely r0 , r1 , · · ·, rb . So we must be able to find two
powers of 10 which give the same remainder 1 . That is there must be 0 ≤ k < l ≤ b
so that rk = rl . Hence

10l − 10k = (bcl + rl ) − (bck + rk )

= b(cl − ck ) since rk = rl .

and we have
10k (10p − 1)
b=
q

1 This is an application of the pigeon hole principle — the very simple but surprisingly useful idea
that if you have n items which you have to put in m boxes, and if n > m, then at least one box
must contain more than one item.

528
 S EQUENCE AND SERIES 3.7 O PTIONAL — R ATIONAL AND IRRATIONAL NUMBERS

where p = l − k and q = cl − ck are both strictly positive integers, since l > k so that
10l − 10k > 0. Thus we can write
a aq
= k p
b 10 (10 − 1)

Next divide the numerator aq by 10p − 1 and compute the remainder. That is, write
aq = α(10p − 1) + β with 0 ≤ β < 10p − 1. Notice that 0 ≤ α < 10k , as otherwise
x = ab ≥ 1. That is, α has at most k digits and β has at most p digits. This, finally,
gives us

a α(10p − 1) + β
x= =
b 10k (10p − 1)
α β
= k+ k p
10 10 (10 − 1)
α β
= k + k+p
10 10 (1 − 10−p )
∞
α β X −pj
= k + k+p 10
10 10 j=0

which gives the required eventually periodic expansion.

3.7.4 tt Irrationality of e

We will give 2 proofs that the number e is irrational, the first due to Fourier (1768–1830)
and the second due to Pennisi (1918–2010). Both are proofs by contradiction 2 — we
first assume that e is rational and then show that this implies a contradiction. In
both cases we reach the contradiction by showing that a given quantity (related to
the series expression for e) must be both a positive integer and also strictly less than
1.

ttt Proof 1
3.7.4.1

This proof is due to Fourier. Let us assume that the number e is rational so we can
write it as
a
e=
b

2 Proof by contradiction is a standard and very powerful method of proof in mathematics. It

relies on the law of the excluded middle which states that any given mathematical statement P
is either true or false. Because of this, if we can show that the statement P being false implies
something contradictory — like 1 = 0 or a > a — then we can conclude that P must be true. The
interested reader can certainly find many examples (and a far more detailed explanation) using
their favourite search engine.

529
S EQUENCE AND SERIES 3.7 O PTIONAL — R ATIONAL AND IRRATIONAL NUMBERS

where a, b are positive integers. Using the Maclaurin series for ex we have
∞
a X 1
= e1 =
b n=0
n!

Now multiply both sides by b! to get

∞
b! X b!
a =
b n=0
n!

The left-hand side of this expression is an integer. We complete the proof by showing
that the right-hand side cannot be an integer (and hence that we have a contradic-
tion).
First split the series on the right-hand side into two piece as follows
∞ b ∞
X b! X b! X b!
= +
n=0
n! n=0 n! n=b+1 n!
| {z } | {z }
=A =B

The first sum, A, is finite sum of integers:

b b
X b! X
A= = (n + 1)(n + 2) · · · (b − 1)b.
n=0
n! n=0

Consequently A must be an integer. Notice that we simplified the ratio of factorials

using the fact that when b ≥ n we have
b! 1 · 2 · · · n(n + 1)(n + 2) · · · (b − 1)b
= = (n + 1)(n + 2) · · · (b − 1)b.
n! 1 · 2···n
Now we turn to the second sum. Since it is a sum of strictly positive terms we
must have

B>0

We complete the proof by showing that B < 1. To do this we bound each term from
above:
b! 1
=
n! (b + 1)(b + 2) · · · (n − 1)n
| {z }
n−b factors
1 1
≤ =
(b + 1)(b + 1) · · · (b + 1)(b + 1) (b + 1)n−b
| {z }
n−b factors

Indeed the inequality is strict except when n = b + 1. Hence we have that

∞
X 1
B<
n=b+1
(b + 1)n−b

530
 S EQUENCE AND SERIES 3.7 O PTIONAL — R ATIONAL AND IRRATIONAL NUMBERS

1 1 1
= + 2
+ + ···
(b + 1) (b + 1) (b + 1)3

This is just a geometric series (see Lemma 3.2.5) and equals

1 1
= 1
(b + 1) 1 − b+1
1 1
= =
b+1−1 b
And since b is a positive integer, we have shown that

0<B<1

and thus B cannot be an integer.

Thus we have that
b!
a = |{z}
A + |{z} B
b
|{z} integer not integer
integer

which gives a contradiction. Thus e cannot be rational.

ttt Proof 2
3.7.4.2

This proof is due to Pennisi (1953). Let us (again) assume that the number e is ratio-
nal. Hence it can be written as
a
e= ,
b
where a, b are positive integers. This means that we can write

b
e−1 = .
a
Using the Maclaurin series for ex we have
∞
X (−1)n
b
= e−1 =
a n=0
n!

Before we do anything else, we multiply both sides by (−1)a+1 a! — this might seem
a little strange at this point, but the reason will become clear as we proceed through
the proof. The expression is now
∞
a! X (−1)n+a+1 a!
(−1)a+1 b =
a n=0
n!

531
 S EQUENCE AND SERIES 3.7 O PTIONAL — R ATIONAL AND IRRATIONAL NUMBERS

The left-hand side of the expression is an integer. We again complete the proof by
showing that the right-hand side cannot be an integer.
We split the series on the right-hand side into two pieces:
∞ a ∞
X (−1)n+a+1 a! X (−1)n+a+1 a! X (−1)n+a+1 a!
= +
n=0
n! n=0
n! n=a+1
n!
| {z } | {z }
=A =B

We will show that A is an integer while 0 < B < 1; this gives the required contradic-
tion.
Every term in the sum A is an integer. To see this we simplify the ratio of factorials
as we did in the previous proof:
a a
X (−1)n+a+1 a! X
A= = (−1)n+a+1 (n + 1)(n + 2) · · · (a − 1)a
n=0
n! n=0

Let us now examine the series B. Again clean up the ratio of factorials:
∞ ∞
X (−1)n+a+1 a! X (−1)n+a+1
B= =
n=a+1
n! n=a+1
(a + 1) · (a + 2) · · · (n − 1) · n
(−1)2a+2 (−1)2a+3 (−1)2a+4
= + + + ···
a+1 (a + 1)(a + 2) (a + 1)(a + 2)(a + 3)
1 1 1
= − + − ···
a + 1 (a + 1)(a + 2) (a + 1)(a + 2)(a + 3)
Hence B is an alternating series of decreasing terms and by the alternating series
test (Theorem 3.3.14) it converges. Further, it must converge to a number between
its first and second partial sums (see the discussion before Theorem 3.3.14). Hence
the right-hand side lies between
1 1 1 1
and − =
a+1 a + 1 (a + 1)(a + 2) a+2
Since a is a positive integer the above tells us that B converges to a real number
strictly greater than 0 and strictly less than 1. Hence it cannot be an integer.
This gives us a contradiction and hence e cannot be rational.

3.7.5 tt Irrationality of π

This proof is due to Niven (1946) and doesn’t require any mathematics beyond the
level of this course. Much like the proofs above we will start by assuming that π is
rational and then reach a contradiction. Again this contradiction will be that a given
quantity must be an integer but at the same time must lie strictly between 0 and 1.
Assume that π is a rational number and so can be written as π = ab with a, b
positive integers. Now let n be a positive integer and define the polynomial
xn (a − bx)n
f (x) = .
n!
532
 S EQUENCE AND SERIES 3.7 O PTIONAL — R ATIONAL AND IRRATIONAL NUMBERS

It is certainly not immediately obvious why and how Niven chose this polynomial,
but you will see that it has been very carefully crafted to make the proof work. In
particular we will show — under our assumption that π is rational — that, if n is
really big, then
Z π
In = f (x) sin(x)dx
0

is an integer and it also lies strictly between 0 and 1, giving the required contradic-
tion.

ttt Bounding the integral

3.7.5.1

Consider again the polynomial

xn (a − bx)n
f (x) = .
n!
Notice that

f (0) = 0
f (π) = f (a/b) = 0.
a
Furthermore, for 0 ≤ x ≤ π = a/b, we have x ≤ b
and a − bx ≤ a so that

0 ≤ x(a − bx) ≤ a2 /b.

3
We could work out a more precise upper bound, but this one is sufficient for the
analysis that follows. Hence
n
a2

1
0 ≤ f (x) ≤
b n!

We also know that for 0 ≤ x ≤ π = a/b, 0 ≤ sin(x) ≤ 1. Thus

 2 n
a 1
0 ≤ f (x) sin(x) ≤
b n!

for all 0 ≤ x ≤ 1. Using this inequality we bound

Z π  2 n
a 1
0 < In = f (x) sin(x)dx < .
0 b n!
2
n
We will later show that, if n is really big, then ab n!1 < 1. We’ll first show, starting
now, that In is an integer.

3 You got lots of practice finding the maximum and minimum values of continuous functions on
closed intervals when you took calculus last term.

533
 S EQUENCE AND SERIES 3.7 O PTIONAL — R ATIONAL AND IRRATIONAL NUMBERS

ttt Integration by parts

3.7.5.2

In order to show that the value of this integral is an integer we will use integration by
parts. You have already practiced using integration by parts to integrate quantities
like
Z
x2 sin(x) dx

and this integral isn’t much different. For the moment let us just use the fact that f (x)
is a polynomial of degree 2n. Using integration by parts with u = f (x), dv = sin(x)
and v = − cos(x) gives us
Z Z
f (x) sin(x) dx = −f (x) cos(x) + f 0 (x) cos(x) dx

Use integration by parts again with u = f 0 (x), dv = cos(x) and v = sin(x).

Z
= −f (x) cos(x) + f (x) sin(x) − f 00 (x) sin(x) dx
0

Use integration by parts yet again, with u = f 00 (x), dv = sin(x) and v = − cos(x).
Z
= −f (x) cos(x) + f (x) sin(x) + f (x) cos(x) − f 000 (x) cos(x) dx
0 00

And now we can see the pattern; we get alternating signs, and then derivatives mul-
tiplied by sines and cosines:
Z
f (x) sin(x)dx = cos(x) −f (x) + f 00 (x) − f (4) (x) + f (6) (x) − · · ·

+ sin(x) f 0 (x) − f 000 (x) + f (5) (x) − f (7) (x) + · · ·

This terminates at the 2nth derivative since f (x) is a polynomial of degree 2n. We can
check this computation by differentiating the terms on the right-hand side:

d
cos(x) −f (x) + f 00 (x) − f (4) (x) + f (6) (x) − · · ·



dx
= − sin(x) −f (x) + f 00 (x) − f (4) (x) + f (6) (x) − · · ·

+ cos(x) −f 0 (x) + f 000 (x) − f (5) (x) + f (7) (x) − · · ·

and similarly

d
sin(x) f 0 (x) − f 000 (x) + f (5) (x) − f (7) (x) + · · ·



dx
= cos(x) f 0 (x) − f 000 (x) + f (5) (x) − f (7) (x) + · · ·

+ sin(x) f 00 (x) − f (4) (x) + f (6) (x) − · · ·

534
 S EQUENCE AND SERIES 3.7 O PTIONAL — R ATIONAL AND IRRATIONAL NUMBERS

When we add these two expressions together all the terms cancel except f (x) sin(x),
as required.
Now when we take the definite integral from 0 to π , all the sine terms give 0
because sin(0) = sin(π) = 0. Since cos(π) = −1 and cos(0) = +1, we are just left with:
Z π
f (x) sin(x)dx = f (0) − f 00 (0) + f (4) (0) − f (6) (0) + · · · + (−1)n f (2n) (0)


0
+ f (π) − f 00 (π) + f (4) (π) − f (6) (π) + · · · + (−1)n f (2n) (π)

So to show that In is an integer, it now suffices to show that f (j) (0) and f (j) (π) are
integers.

ttt The derivatives are integers

3.7.5.3

Recall that
xn (a − bx)n
f (x) =
n!
and expand it:
c0 0 c1 1 cn c2n 2n
f (x) = x + x + · · · + xn + · · · + x
n! n! n! n!
All the cj are integers, and clearly cj = 0 for all j = 0, 1, · · · , n − 1, because of the
factor xn in f (x).
dk
Now take the k th derivative and set x = 0. Note that, if j < k, then dx kx
j
= 0
dk
for all x and, if j > k, then dxk xj is some number times xj−k which evaluates to zero
when we set x = 0. So
dk  ck k  k!ck
f (k) (0) = k x =
dx k! n!
If k < n, then this is zero since ck = 0. If k > n, this is an integer because ck is an
integer and k!/n! = (n + 1)(n + 2) · · · (k − 1)k is an integer. If k = n, then f (k) (0) = cn
is again an integer. Thus all the derivatives of f (x) evaluated at x = 0 are integers.
But what about the derivatives at π = a/b? To see this, we can make use of a
handy symmetry. Notice that

f (x) = f (π − x) = f (a/b − x)

You can confirm this by just grinding through the algebra:

xn (a − bx)n
f (x) = now replace x with a/b − x
n!
(a/b − x)n (a − b(a/b − x))n
f (a/b − x) = start cleaning this up:
n!
a−bx n


b
(a − a + bx)n
=
n!
535
 S EQUENCE AND SERIES 3.7 O PTIONAL — R ATIONAL AND IRRATIONAL NUMBERS

a−bx n


b
(bx)n
=
n!
(a − bx)n xn
= = f (x)
n!
Using this symmetry (and the chain rule) we see that

f 0 (x) = −f 0 (π − x)

and if we keep differentiating

f (k) (x) = (−1)k f (k) (π − x)

Setting x = 0 in this tells us that

f (k) (0) = (−1)k f (k) (π)

So because all the derivatives at x = 0 are integers, we know that all the derivatives
at x = π are also integers.
Hence the integral we are interested in
Z π
f (x) sin(x)dx
0

must be an integer.

ttt Putting it together

3.7.5.4

Based on our assumption that π = a/b is rational, we have shown that the integral
Z π n
x (a − bx)
In = sin(x)dx
0 n!

satisfies
n
a2

1
0 < In <
b n!

and also that In is an integer.

We are, however, free to choose n to be any positive integer we want. If we take
n to be very large — in particular much much larger than a — then n! will be much
much larger than a2n (we showed this in Example 3.6.8), and consequently
 2 n
a 1
0 < In < <1
b n!

Which means that the integral cannot be an integer. This gives the required contra-
diction, showing that π is irrational.

536
 Appendix A

H IGH S CHOOL M ATERIAL

This chapter is really split into three parts.

• Sections A.1 to A.11 contains results that we expect you to understand and
know.
• Then Section A.14 contains results that we don’t expect you to memorise, but
that we think you should be able to quickly derive from other results you know.
• The remaining sections contain some material (that may be new to you) that is
related to topics covered in the main body of these notes.

A.1q Similar Triangles

Two triangles T1 , T2 are similar when

• (AAA — angle angle angle) The angles of T1 are the same as the angles of T2 .
• (SSS — side side side) The ratios of the side lengths are the same. That is
A B C
= =
a b c
537
 H IGH S CHOOL M ATERIAL A.3 T RIGONOMETRY — D EFINITIONS

• (SAS — side angle side) Two sides have lengths in the same ratio and the angle
between them is the same. For example
A C
= and angle β is same
a c

A.2q Pythagoras

For a right-angled triangle the length of the hypotenuse is related to the lengths of
the other two sides by

(adjacent)2 + (opposite)2 = (hypotenuse)2

A.3q Trigonometry — Definitions

opposite 1
sin θ = csc θ =
hypotenuse sin θ
adjacent 1
cos θ = sec θ =
hypotenuse cos θ
opposite 1
tan θ = cot θ =
adjacent tan θ

538
 H IGH S CHOOL M ATERIAL A.5 T RIGONOMETRY — G RAPHS

A.4q Radians, Arcs and Sectors

For a circle of radius r and angle of θ radians:

• Arc length L(θ) = rθ.

• Area of sector A(θ) = 2θ r2 .

A.5q Trigonometry — Graphs

sin θ cos θ tan θ

1 1

−π − π2 π π 3π 2π −π − π2 π π 3π 2π −π − π2 π π 3π 2π
2 2 2 2 2 2

−1 −1

539
 H IGH S CHOOL M ATERIAL A.7 T RIGONOMETRY — S IMPLE I DENTITIES

A.6q Trigonometry — Special Triangles

From the above pair of special triangles we have

√
π 1 π 1 π 3
sin = √ sin = sin =
4 2 6 2 3 2
√
π 1 π 3 π 1
cos = √ cos = cos =
4 2 6 2 3 2
π π 1 π √
tan = 1 tan = √ tan = 3
4 6 3 3

A.7q Trigonometry — Simple Identities

• Periodicity

sin(θ + 2π) = sin(θ) cos(θ + 2π) = cos(θ)

• Reflection

sin(−θ) = − sin(θ) cos(−θ) = cos(θ)

• Reflection around π/4

π π



sin 2
− θ = cos θ cos 2
− θ = sin θ

• Reflection around π/2

sin (π − θ) = sin θ cos (π − θ) = − cos θ

540
 H IGH S CHOOL M ATERIAL A.9 I NVERSE T RIGONOMETRIC F UNCTIONS

• Rotation by π

sin (θ + π) = − sin θ cos (θ + π) = − cos θ

• Pythagoras

sin2 θ + cos2 θ = 1

A.8q Trigonometry — Add and Subtract Angles

• Sine

sin(α ± β) = sin(α) cos(β) ± cos(α) sin(β)

• Cosine

cos(α ± β) = cos(α) cos(β) ∓ sin(α) sin(β)

A.9q Inverse Trigonometric Functions

Some of you may not have studied inverse trigonometric functions in highschool,
however we still expect you to know them by the end of the course.

arcsin x arccos x arctan x

Domain: −1 ≤ x ≤ 1 Domain: −1 ≤ x ≤ 1 Domain: all real numbers
Range: − π2 ≤ arcsin x ≤ π
2
Range: 0 ≤ arccos x ≤ π Range: − π2 < arctan x < π2
π π π
2 2

π
2
−1 1

− π2 − π2
−1 1

Since these functions are inverses of each other we have

π π
arcsin(sin θ) = θ − ≤θ≤
2 2
arccos(cos θ) = θ 0≤θ≤π

541
 H IGH S CHOOL M ATERIAL A.10 A REAS

π π
arctan(tan θ) = θ − ≤θ≤
2 2
and also
sin(arcsin x) = x −1 ≤ x ≤ 1
cos(arccos x) = x −1 ≤ x ≤ 1
tan(arctan x) = x any real x

arccsc x arcsec x arccot x

Domain: |x| ≥ 1 Domain: |x| ≥ 1 Domain: all real numbers
Range: − π2 ≤ arccsc x ≤ π
2
Range: 0 ≤ arcsec x ≤ π Range: 0 < arccot x < π
π
arccsc x 6= 0 arcsec x 6=
2
π π
2 π

π
π 2
2
−1 1

− π2

−1 1

Again
π π
arccsc(csc θ) = θ − ≤ θ ≤ , θ 6= 0
2 2
π
arcsec(sec θ) = θ 0 ≤ θ ≤ π, θ 6=
2
arccot(cot θ) = θ 0<θ<π
and
csc(arccsc x) = x |x| ≥ 1
sec(arcsec x) = x |x| ≥ 1
cot(arccot x) = x any real x

A.10q Areas

542
 H IGH S CHOOL M ATERIAL A.11 V OLUMES

• Area of a rectangle
A = bh

• Area of a triangle
1 1
A = bh = ab sin θ
2 2
• Area of a circle
A = πr2

• Area of an ellipse
A = πab

A.11q Volumes

• Volume of a rectangular prism

V = lwh

• Volume of a cylinder
V = πr2 h

• Volume of a cone
1
V = πr2 h
3
• Volume of a sphere
4
V = πr3
3

543
 H IGH S CHOOL M ATERIAL A.13 L OGARITHMS

A.12q Powers

In the following, x and y are arbitrary real numbers, and q is an arbitrary constant
that is strictly bigger than zero.

• q0 = 1
qx
• q x+y = q x q y , q x−y = qy

• q −x = q1x

y
• q x = q xy

• lim q x = ∞, lim q x = 0 if q > 1

x→∞ x→−∞

• lim q x = 0, lim q x = ∞ if 0 < q < 1

x→∞ x→−∞

• The graph of 2x is given below. The graph of q x , for any q > 1, is similar.

A.13q Logarithms

In the following, x and y are arbitrary real numbers that are strictly bigger than 0,
and p and q are arbitrary constants that are strictly bigger than one.

• q logq x = x,


logq q x = x
logp x
• logq x = logp q

• logq 1 = 0, logq q = 1

• logq (xy) = logq x + logq y

544
 H IGH S CHOOL M ATERIAL A.14 H IGHSCHOOL M ATERIAL Y OU S HOULD BE A BLE TO D ERIVE

x


• logq y
= logq x − logq y
1


• logq y
= − logq y,

• logq (xy ) = y logq x

• lim logq x = ∞, lim logq x = −∞
x→∞ x→0

• The graph of log10 x is given below. The graph of logq x, for any q > 1, is similar.

A.14q Highschool Material You Should be Able to Derive

• Graphs of csc θ, sec θ and cot θ:

csc θ sec θ cot θ

1 1

−π − π2−1 π π 3π 2π −π − π2−1 π π 3π 2π −π − π2 π π 3π 2π
2 2 2 2 2 2

• More Pythagoras
divide by cos2 θ
sin2 θ + cos2 θ = 1 7−−−−−−−−→ tan2 θ + 1 = sec2 θ
divide by sin2 θ
sin2 θ + cos2 θ = 1 7−−−−−−−−→ 1 + cot2 θ = csc2 θ

• Sine — double angle (set β = α in sine angle addition formula)

sin(2α) = 2 sin(α) cos(α)

545
 H IGH S CHOOL M ATERIAL A.15 C ARTESIAN C OORDINATES

• Cosine — double angle (set β = α in cosine angle addition formula)

cos(2α) = cos2 (α) − sin2 (α)

= 2 cos2 (α) − 1 (use sin2 (α) = 1 − cos2 (α))
= 1 − 2 sin2 (α) (use cos2 (α) = 1 − sin2 (α))

• Composition of trigonometric and inverse trigonometric functions:

√ √
cos(arcsin x) = 1 − x2 sec(arctan x) = 1 + x2

and similar expressions.

A.15q Cartesian Coordinates

Each point in two dimensions may be labeled by two coordinates (x, y) which specify
the position of the point in some units with respect to some axes as in the figure
below.

The set of all points in two dimensions is denoted R2 . Observe that

• the distance from the point (x, y) to the x-axis is |y|

• the distance from the point (x, y) to the y-axis is |x|

p
• the distance from the point (x, y) to the origin (0, 0) is x2 + y 2

Similarly, each point in three dimensions may be labeled by three coordinates

(x, y, z), as in the two figures below.

546
H IGH S CHOOL M ATERIAL A.15 C ARTESIAN C OORDINATES

The set of all points in three dimensions is denoted R3 . The plane that contains,
for example, the x- and y-axes is called the xy-plane.
• The xy-plane is the set of all points (x, y, z) that obey z = 0.
• The xz-plane is the set of all points (x, y, z) that obey y = 0.
• The yz-plane is the set of all points (x, y, z) that obey x = 0.
More generally,
• The set of all points (x, y, z) that obey z = c is a plane that is parallel to the
xy-plane and is a distance |c| from it. If c > 0, the plane z = c is above the
xy-plane. If c < 0, the plane z = c is below the xy-plane. We say that the plane
z = c is a signed distance c from the xy-plane.
• The set of all points (x, y, z) that obey y = b is a plane that is parallel to the
xz-plane and is a signed distance b from it.
• The set of all points (x, y, z) that obey x = a is a plane that is parallel to the
yz-plane and is a signed distance a from it.

Observe that

547
 H IGH S CHOOL M ATERIAL A.16 R OOTS OF P OLYNOMIALS

• the distance from the point (x, y, z) to the xy-plane is |z|

• the distance from the point (x, y, z) to the xz-plane is |y|
• the distance from the point (x, y, z) to the yz-plane is |x|
p
• the distance from the point (x, y, z) to the origin (0, 0, 0) is x2 + y 2 + z 2
The distance from the point (x, y, z) to the point (x0 , y 0 , z 0 ) is
p
(x − x0 )2 + (y − y 0 )2 + (z − z 0 )2
so that the equation of the sphere centered on (1, 2, 3) with radius 4, that is, the set of
all points (x, y, z) whose distance from (1, 2, 3) is 4, is
(x − 1)2 + (y − 2)2 + (z − 3)2 = 16

A.16q Roots of Polynomials

Being able to factor polynomials is a very important part of many of the compu-
tations in this course. Related to this is the process of finding roots (or zeros) of
polynomials. That is, given a polynomial P (x), find all numbers r so that P (r) = 0.
In the case of a quadratic P (x) = ax2 + bx + c, we can use the formula
√
−b ± b2 − 4ac
x=
2a
The corresponding formulas for cubics and quartics 1 are extremely cumbersome,
and no such formula exists for polynomials of degree 5 and higher 2 .
Despite this there are many tricks 3 for finding roots of polynomials that work
well in some situations but not all. Here we describe approaches that will help you
find integer and rational roots of polynomials that will work well on exams, quizzes
and homework assignments.
Consider the quadratic equation x2 − 5x + 6 = 0. We could 4 solve this using the
quadratic formula
√
5 ± 25 − 4 × 1 × 6 5±1
x= = = 2, 3.
2 2

1 The method for cubics was developed in the 15th century by del Ferro, Cardano and Ferrari
(Cardano’s student). Ferrari then went on to discover a formula for the roots of a quartic. His
formula requires the solution of an associated cubic polynomial.
2 This is the famous Abel-Ruffini theorem.
3 There is actually a large body of mathematics devoted to developing methods for factoring poly-
nomials. Polynomial factorisation is a fundamental problem for most computer algebra systems.
The interested reader should make use of their favourite search engine to find out more.
4 We probably shouldn’t do it this way for such a simple polynomial, but for pedagogical purposes
we do here.

548
H IGH S CHOOL M ATERIAL A.16 R OOTS OF P OLYNOMIALS

Hence x2 − 5x + 6 has roots x = 2, 3 and so it factors as (x − 3)(x − 2). Notice 5 that

the numbers 2 and 3 divide the constant term of the polynomial, 6. This happens in
general and forms the basis of our first trick.

Lemma A.16.1 A very useful trick.

If r or −r is an integer root of a polynomial P (x) = an xn + · · · + a1 x + a0 with

integer coefficients, then r is a factor of the constant term a0 .

Proof. If r is a root of the polynomial we know that P (r) = 0. Hence

an · rn + · · · + a1 · r + a0 = 0

If we isolate a0 in this expression we get

a0 = − an r n + · · · + a1 r
 

We can see that r divides every term on the right-hand side. This means that
the right-hand side is an integer times r. Thus the left-hand side, being a0 , is
an integer times r, as required. The argument for when −r is a root is almost
identical.

Let us put this observation to work.

Example A.16.2 Integer roots of x3 − x2 + 2.

Find the integer roots of P (x) = x3 − x2 + 2.

Solution:

• The constant term in this polynomial is 2.

• The only divisors of 2 are 1, 2. So the only candidates for integer roots are
±1, ±2.

• Trying each in turn

P (1) = 2 P (−1) = 0
P (2) = 6 P (−2) = −10

• Thus the only integer root is −1.

5 Many of you may have been taught this approach in highschool.

549
H IGH S CHOOL M ATERIAL A.16 R OOTS OF P OLYNOMIALS

Example A.16.3 Integer roots of 3x3 + 8x2 − 5x − 6.

Find the integer roots of P (x) = 3x3 + 8x2 − 5x − 6.

Solution:

• The constant term is −6.

• The divisors of 6 are 1, 2, 3, 6. So the only candidates for integer roots are
±1, ±2, ±3, ±6.

• We try each in turn (it is tedious but not difficult):

P (1) = 0 P (−1) = 4
P (2) = 40 P (−2) = 12
P (3) = 132 P (−3) = 0
P (6) = 900 P (−6) = −336

• Thus the only integer roots are 1 and −3.

We can generalise this approach in order to find rational roots. Consider the
polynomial 6x2 − x − 2. We can find its zeros using the quadratic formula:
√
1 ± 1 + 48 1±7 1 2
x= = =− , .
12 12 2 3
Notice now that the numerators, 1 and 2, both divide the constant term of the poly-
nomial (being 2). Similarly, the denominators, 2 and 3, both divide the coefficient of
the highest power of x (being 6). This is quite general.

Lemma A.16.4 Another nice trick.

If b/d or −b/d is a rational root in lowest terms (i.e. b and d are integers with
no common factors) of a polynomial Q(x) = an xn + · · · + a1 x + a0 with integer
coefficients, then the numerator b is a factor of the constant term a0 and the
denominator d is a factor of an .

b
Proof. Since d
is a root of P (x) we know that

an (b/d)n + · · · + a1 (b/d) + a0 = 0

Multiply this equation through by dn to get

an bn + · · · + a1 bdn−1 + a0 dn = 0

550
H IGH S CHOOL M ATERIAL A.16 R OOTS OF P OLYNOMIALS

Move terms around to isolate a0 dn :

a0 dn = − an bn + · · · + a1 bdn−1
 

Now every term on the right-hand side is some integer times b. Thus the left-
hand side must also be an integer times b. We know that d does not contain any
factors of b, hence a0 must be some integer times b (as required).
Similarly we can isolate the term an bn :

an bn = − an−1 bn−1 d + · · · + a1 bdn−1 + a0 dn

 

Now every term on the right-hand side is some integer times d. Thus the left-
hand side must also be an integer times d. We know that b does not contain any
factors of d, hence an must be some integer times d (as required).
The argument when − db is a root is nearly identical.

We should put this to work:

Example A.16.5 Rational roots of 2x2 − x − 3.

P (x) = 2x2 − x − 3.
Solution:

• The constant term in this polynomial is 3 = 1 × 3 and the coefficient of the

highest power of x is 2 = 1 × 2.

• Thus the only candidates for integer roots are ±1, ±3.

• By our newest trick, the only candidates for fractional roots are ± 21 , ± 32 .

• We try each in turn a

P (1) = −2 P (−1) = 0
P (3) = 12 P (−3) = 18
P 12 = −3 P − 21 = −2

P 32 = 0 P − 23 = 3

so the roots are −1 and 32 .

a Again, this is a little tedious, but not difficult. Its actually pretty easy to code up for a computer
to do. Modern polynomial factoring algorithms do more sophisticated things, but these are a
pretty good way to start.

The tricks above help us to find integer and rational roots of polynomials. With

551
H IGH S CHOOL M ATERIAL A.16 R OOTS OF P OLYNOMIALS

a little extra work we can extend those methods to help us factor polynomials. Say
we have a polynomial P (x) of degree p and have established that r is one of its roots.
That is, we know P (r) = 0. Then we can factor (x − r) out from P (x) — it is always
possible to find a polynomial Q(x) of degree p − 1 so that
P (x) = (x − r)Q(x)
In sufficiently simple cases, you can probably do this factoring by inspection. For
example, P (x) = x2 − 4 has r = 2 as a root because P (2) = 22 − 4 = 0. In this case,
P (x) = (x − 2)(x + 2) so that Q(x) = (x + 2). As another example, P (x) = x2 − 2x − 3
has r = −1 as a root because P (−1) = (−1)2 − 2(−1) − 3 = 1 + 2 − 3 = 0. In this case,
P (x) = (x + 1)(x − 3) so that Q(x) = (x − 3).
For higher degree polynomials we need to use something more systematic —
long divison.

Lemma A.16.6 Long Division.

Once you have found a root r of a polynomial, even if you cannot factor (x − r)
out of the polynomial by inspection, you can find Q(x) by dividing P (x) by
x − r, using the long division algorithm you learned a in school, but with 10
replaced by x.

a This is a standard part of most highschool mathematics curricula, but perhaps not all. You
should revise this carefully.

Example A.16.7 Roots of x3 − x2 + 2.

Factor P (x) = x3 − x2 + 2.
Solution:

• We can go hunting for integer roots of the polynomial by looking at the divisors
of the constant term. This tells us to try x = ±1, ±2.

• A quick computation shows that P (−1) = 0 while P (1), P (−2), P (2) 6= 0.

Hence x = −1 is a root of the polynomial and so x + 1 must be a factor.
3 2
• So we divide x −x
x+1
+2
. The first term, x2 , in the quotient is chosen so that when
you multiply it by the denominator, x2 (x + 1) = x3 + x2 , the leading term, x3 ,
matches the leading term in the numerator, x3 − x2 + 2, exactly.

552
H IGH S CHOOL M ATERIAL A.16 R OOTS OF P OLYNOMIALS

• When you subtract x2 (x + 1) = x3 + x2 from the numerator x3 − x2 + 2 you

get the remainder −2x2 + 2. Just like in public school, the 2 is not normally
“brought down” until it is actually needed.

• The next term, −2x, in the quotient is chosen so that when you multiply it by
the denominator, −2x(x + 1) = −2x2 − 2x, the leading term −2x2 matches the
leading term in the remainder exactly.

And so on.

• Note that we finally end up with a remainder 0. A nonzero remainder would

have signalled a computational error, since we know that the denominator x −
(−1) must divide the numerator x3 − x2 + 2 exactly.

• We conclude that

(x + 1)(x2 − 2x + 2) = x3 − x2 + 2

To check this, just multiply out the left hand side explicitly.

553
H IGH S CHOOL M ATERIAL A.16 R OOTS OF P OLYNOMIALS

√
2
• Applying the high school quadratic root formula −b± 2ab −4ac to x2 − 2x + 2 tells
us that it has no real roots and that we cannot factor it further a .

a Because we are not permitted to use complex numbers.

Example A.16.7

We finish by describing an alternative to long division. The approach is roughly

equivalent, but is perhaps more straightforward at the expense of requiring more
algebra.

Example A.16.8 Roots of x3 − x2 + 2 again.

Factor P (x) = x3 − x2 + 2, again.

Solution: Let us do this again but avoid long division.
3 2
• From the previous example, we know that x −x x+1
+2
must be a polynomial (since
−1 is a root of the numerator) of degree 2. So write

x3 − x2 + 2
= ax2 + bx + c
x+1
for some, as yet unknown, coefficients a, b and c.

• Cross multiplying and simplifying gives us

x3 − x2 + 2 = (ax2 + bx + c)(x + 1)
= ax3 + (a + b)x2 + (b + c)x + c

• Now matching coefficients of the various powers of x on the left and right hand
sides

coefficient of x3 : a=1
coefficient of x2 : a + b = −1
coefficient of x1 : b+c=0
coefficient of x0 : c=2

• This gives us a system of equations that we can solve quite directly. Indeed it
tells us immediately that that a = 1 and c = 2. Subbing a = 1 into a + b = −1
tells us that 1 + b = −1 and hence b = −2.

• Thus

x3 − x2 + 2 = (x + 1)(x2 − 2x + 2).

554
 Appendix B

H INTS FOR E XERCISES

1 · Integration
1.1 · Definition of the Integral
1.1.8 · Exercises

Exercises — Stage 1
1.1.8.1. Hint. Draw a rectangle that encompasses the entire shaded area, and one
that is encompassed by the shaded area. The shaded area is no more than the area
of the bigger rectangle, and no less than the area of the smaller rectangle.
1.1.8.2. Hint. We can improve on the method of Question 1 by using three rect-
angles that together encompass the shaded region, and three rectangles that to-
gether are encompassed by the shaded region.
1.1.8.3. Hint. Four rectangles suffice.
1.1.8.4. Hint. Try drawing a picture.
1.1.8.5. Hint. Try an oscillating function.
1.1.8.6. Hint. The ordering of the parts is intentional: each sum can be written
by changing some small part of the sum before it.
1.1.8.7. Hint. If we raise −1 to an even power, we get +1, and if we raise it to
an odd power, we get −1.
1.1.8.8. Hint. Sometimes a little anti-simplification can make the pattern more
clear.
1
a Re-write as 3
+ 39 + 5
27
+ 7
81
+ 9
243
.

b Compare to the sum in the hint for (a).

555
H INTS FOR E XERCISES

c Re-write as 1 · 1000 + 2 · 100 + 3 · 10 + 14 + 5

10
+ 6
100
+ 7
1000
.
1.1.8.9. Hint.

• (a), (b) These are geometric sums.

• (c) You can write this as three separate sums.

• (d) You can write this as two separate sums. Remember that e is a constant.
Don’t be thrown off by the index being n instead of i.
1.1.8.10. Hint.

a Write out the terms of the two sums.

b A change of index is an easier option than expanding the cubic.

c Which terms cancel?

d Remember 2n + 1 is odd for every integer n. The index starts at n = 2, not

n = 1.
1.1.8.11. Hint. Since the sum adds four pieces, there will be four rectangles.
However, one might be extremely small.
1.1.8.12. *. Hint. Write out the general formula for the left Riemann sum from
Definition 1.1.11 and choose a, b and n to make it match the given sum.
1.1.8.13. Hint. Since the sum runs from 1 to 3, there are three intervals. Suppose
2 = ∆x = b−a n
. You may assume the sum given is a right Riemann sum (as
opposed to left or midpoint).
π
1.1.8.14. Hint. Let ∆x = . Then what is b − a?
20
1.1.8.15. *. Hint. Notice that the index starts at k = 0, instead of k = 1. Write out
the given sum explicitly without using summation notation, and sketch where
the rectangles would fall on a graph of y = f (x).
Then try to identify b − a, and n, followed by “right”, “left”, or “midpoint”, and
finally a.
1.1.8.16. Hint. The area is a triangle.
1.1.8.17. Hint. There is one triangle of positive area, and one of negative area.

Exercises — Stage 2
1.1.8.18. *. Hint. Review Definition 1.1.11.
1.1.8.20. *. Hint. You’ll want the limit as n goes to infinity of a sum with n
terms. If you’re having a hard time coming up with the sum in terms of n, try
writing a sum with a finite number of terms of your choosing. Then, think about
how that sum would change if it had n terms.

556
H INTS FOR E XERCISES

1.1.8.21. *. Hint. The main step is to express the given sum as the right Riemann
sum,
n
X
f (a + i∆x)∆x.
i=1

Don’t be afraid to guess ∆x and f (x) (review Definition 1.1.11). Then write out
n
P
explicitly f (a+i∆x)∆x with your guess substituted in, and compare the result
i=1
with the given sum. Adjust your guess if they don’t match.
1.1.8.22. *. Hint. The main step is to express the given sum as the right Riemann
Pn
sum f (a + k∆x)∆x. Don’t be afraid to guess ∆x and f (x) (review Definition
k=1
n
P
1.1.11). Then write out explicitly f (a + k∆x)∆x with your guess substituted
k=1
in, and compare the result with the given sum. Adjust your guess if they don’t
match.
1.1.8.23.
Pn *. Hint. The main step is to express the given sum in the form
∗
i=1 f (x i )∆x. Don’t be afraid to guess ∆x, x∗i (for either a left or a right or a
midpoint sum — review Definition 1.1.11) and f (x). Then write out explicitly
P n ∗
i=1 f (xi )∆x with your guess substituted in, and compare the result with the
given sum. Adjust your guess if they don’t match.
1.1.8.24. *. Hint. The main step is to express the given sum in the form
n
f (x∗i )∆x. Don’t be afraid to guess ∆x, x∗i (probably, based on the symbol Rn ,
P
i=1
assuming we have a right Riemann sum — review Definition 1.1.11) and f (x).
n
f (x∗i )∆x with your guess substituted in, and com-
P
Then write out explicitly
i=1
pare the result with the given sum. Adjust your guess if they don’t match.
1.1.8.25. *. Hint. Try several different choices of ∆x and x∗i .
1.1.8.26. Hint. Let x = r3 , and re–write the sum in terms of x.
1.1.8.27. Hint. Note the sum does not start at r0 = 1.
1.1.8.28. *. Hint. Draw a picture. See Example 1.1.15.

x x≥0
1.1.8.29. Hint. Draw a picture. Remember |x| = .
−x x < 0
1.1.8.30. Hint. Draw a picture: the area we want is a trapezoid. If you don’t
remember a formula for the area of a trapezoid, think of it as the difference of
two triangles.
1.1.8.31. Hint. You can draw a very similar picture to Question 30, but remem-
ber the areas are negative.
√
1.1.8.32. Hint. If y = 16 − x2 , then y is nonnegative, and y 2 + x2 = 16.

557
H INTS FOR E XERCISES

1.1.8.33. *. Hint. Sketch the graph of f (x).

1.1.8.34. *. Hint. At which time in the interval, for example, 0 ≤ t ≤ 0.5, is the
car moving the fastest?
1.1.8.35. Hint. What are the possible speeds the car could have reached at time
t = 0.25?
1.1.8.36. Hint. You need to know the speed of the plane at the midpoints of
your intervals, so (for example) noon to 1pm is not one of your intervals.

Exercises — Stage 3
1.1.8.37. *. Hint. Sure looks like a Riemann sum.
1.1.8.38. *. Hint. For part (b): don’t panic! Just take it one step at a time. The
first step is to write down the Riemann sum. The second step is to evaluate the
sum, using the given identity. The third step is to evaluate the limit n → ∞.
1.1.8.39. *. Hint. The first step is to write down the Riemann sum. The second
step is to evaluate the sum, using the given formulas. The third step is to evaluate
the limit as n → ∞.
1.1.8.40. *. Hint. The first step is to write down the Riemann sum. The second
step is to evaluate the sum, using the given formulas. The third step is to evaluate
the limit n → ∞.
1.1.8.41. *. Hint. You’ve probably seen this hint before. It is worth repeating.
Don’t panic! Just take it one step at a time. The first step is to write down the
Riemann sum. The second step is to evaluate the sum, using the given formula.
The third step is to evaluate the limit n → ∞.
1.1.8.42. Hint. Using the definition of a right Riemann sum, we can come up
with an expression for f (−5 + 10i). In order to find f (x), set x = −5 + 10i.
d
1.1.8.43. Hint. Recall that for a positive constant a, dx
{ax } = ax log a, where
log a is the natural logarithm (base e) of a.
1.1.8.44. Hint. Part (a) follows the same pattern as Question 43–there’s just a
little more algebra involved, since our lower limit of integration is not 0.
1.1.8.45. Hint. Your area can be divided into a section of a circle and a triangle.
Then you can use geometry to find the area of each piece.
1.1.8.46. Hint.

a The difference between the upper and lower bounds is the area that is out-
side of the smaller rectangles but inside the larger rectangles. Drawing both
sets of rectangles on one picture might make things clearer. Look for an
easy way to compute the area you want.

b Use your answer from Part (a). Your answer will depend on f , a, and b.

558
H INTS FOR E XERCISES

1.1.8.47. Hint. Since f (x) is linear, there exist real numbers m and c such that
f (x) = mx + c. It’s a little easier to first look at a single triangle from each sum,
rather than the sums in their entirety.

1.2 · Basic properties of the definite integral

1.2.3 · Exercises

Exercises — Stage 1
1.2.3.1. Hint.

a What is the length of this figure?

b Think about cutting the area into two pieces vertically.

c Think about cutting the area into two pieces another way.
Rb Rc Rb
1.2.3.2. Hint. Use the identity f (x)dx = f (x)dx + f (x)dx.
a a c

1.2.3.4. Hint. Note that the limits of the integral given are in the opposite order
from what we might expect: the smaller number is the top limit of integration.
Recall ∆x = b−a
n
.

Exercises — Stage 2
1.2.3.5. *. Hint. Split the “target integral” up into pieces that can be evaluated
using the given integrals.
1.2.3.6. *. Hint. Split the “target integral” up into pieces that can be evaluated
using the given integrals.
1.2.3.7. *. Hint. Split the “target integral” up into pieces that can be evaluated
using the given integrals.
1.2.3.8. Hint. For part (a), use the symmetry of the integrand. For part (b), the
R1 √
area 1 − x2 dx is easy to find–how is this useful to you?
0

1.2.3.9. *. Hint. The evaluation of this integral was also the subject of Question
1.2.3.9 in Section 1.1. This time try using the method of Example 1.2.7.
1.2.3.10. Hint. Use symmetry.
1.2.3.11. Hint. Check Theorem 1.2.12.

Exercises — Stage 3
1.2.3.12. *. Hint. Split the integral into a sum of two integrals. Interpret each
geometrically.

559
H INTS FOR E XERCISES

1.2.3.13. *. Hint. Hmmmm. Looks like a complicated integral. It’s probably a

trick question. Check for symmetries.
1.2.3.14. *. Hint. Check for symmetries again.
1.2.3.15. Hint. What does the integrand look like to the left and right of x = 3?
1.2.3.16. Hint. In part (b), you’ll have to factor a constant
√ out through a square
root. Remember the upper half of a circle looks like r − x2 .
2

1.2.3.17. Hint. For two functions f (x) and g(x), define h(x) = f (x) · g(x). If
h(−x) = h(x), then the product is even; if h(−x) = −h(x), then the product is
odd.
The table will not be the same as if we were multiplying even and odd numbers.
1.2.3.18. Hint. Note f (0) = f (−0).
1.2.3.19. Hint. If f (x) is even and odd, then f (x) = −f (x) for every x.
1.2.3.20. Hint. Think about mirroring a function across an axis. What does this
do to the slope?

1.3 · The Fundamental Theorem of Calculus

1.3.2 · Exercises

Exercises — Stage 1
1.3.2.2. *. Hint. First find the general antiderivative by guessing and checking.
1.3.2.3. *. Hint. Be careful. Two of these make no sense at all.
1.3.2.4. Hint. Check by differentiating.
1.3.2.5. Hint. Check by differentiating.
1.3.2.6. Hint. Use the Fundamental Theorem of Calculus Part 1.
1.3.2.7. Hint. Use the Fundamental Theorem of Calculus, Part 1.
1.3.2.8. Hint. You already know that F (x) is an antiderivative of f (x).
d −1
1.3.2.9. Hint. (a) Recall dx {arccos x} = √1−x2.
√
(b) All antiderivatives of 1 − x2 differ from one another by a constant. You
already know one antiderivative.
1.3.2.10. Hint. In order to apply the Fundamental Theorem of Calculus Part 2,
the integrand must be continuous over the interval of integration.
1.3.2.11. Hint. Use the definition of F (x) as an area.
1.3.2.12. Hint. F (x) represents net signed area.
1.3.2.13. Hint. Note G(x) = −F (x), when F (x) is defined as in Question 12.

560
H INTS FOR E XERCISES

1.3.2.14. Hint. Using the definition of the derivative, F 0 (x) =

F (x + h) − F (x)
lim .
h→0 h
The area of a trapezoid with base b and heights h1 and h2 is 12 b(h1 + h2 ).
1.3.2.15. Hint. There is only one!
d
R
1.3.2.16. Hint. If dx
{F (x)} = f (x), that tells us f (x)dx = F (x) + C.
1.3.2.17. Hint. When you’re differentiating, you can leave the ex factored out.
1.3.2.18. Hint. After differentiation, you can simplify pretty far. Keep at it!
1.3.2.19. Hint. This derivative also simplifies considerably. You might need to
add fractions by finding a common denominator.

Exercises — Stage 2
1.3.2.20. *. Hint. Guess a function whose derivative is the integrand, then use
the Fundamental Theorem of Calculus Part 2.
1.3.2.21. *. Hint. Split the given integral up into two integrals.
1
1.3.2.22. Hint. The integrand is similar to , so something with arctangent
1 + x2
seems in order.
1 √
The integrand is similar to √
1.3.2.23. Hint. , so factoring out 2 from the
1 − x2
denominator will make it look like some flavour of arcsine.
1.3.2.24. Hint. We know how to antidifferentiate sec2 x, and there is an identity
linking sec2 x with tan2 x.
1.3.2.25. Hint. Recall 2 sin x cos x = sin(2x).
1 + cos(2x)
1.3.2.26. Hint. cos2 x =
2
1.3.2.28. *. Hint. There is a good way to test where a function is increasing,
decreasing, or constant, that also has something to do with topic of this section.
1.3.2.29. *. Hint. See Example 1.3.5.
1.3.2.30. *. Hint. See Example 1.3.5.
1.3.2.31. *. Hint. See Example 1.3.5.
1.3.2.32. *. Hint. See Example 1.3.5.
1.3.2.33. *. Hint. See Example 1.3.6.
d
1.3.2.34. *. Hint. Apply dx
to both sides.
1.3.2.35. *. Hint. What is the title of this section?

561
H INTS FOR E XERCISES

1.3.2.36. *. Hint. See Example 1.3.6.

1.3.2.37. *. Hint. See Example 1.3.6.
1.3.2.38. *. Hint. See Example 1.3.6.
1.3.2.39. *. Hint. See Example 1.3.6.
1.3.2.40. *. Hint. Split up the domain of integration.

Exercises — Stage 3
1.3.2.41. *. Hint. It is possible to guess an antiderivative for f 0 (x)f 00 (x) that is
expressed in terms of f 0 (x).
1.3.2.42. *. Hint. When does the car stop? What is the relation between velocity
and distance travelled?
1.3.2.43. *. Hint. See Example 1.3.5. For the absolute maximum part of the
question, study the sign of f 0 (x).
1.3.2.44. *. Hint. See Example 1.3.5. For the “minimum value” part of the
question, study the sign of f 0 (x).
1.3.2.45. *. Hint. See Example 1.3.5. For the “maximum” part of the question,
study the sign of F 0 (x).
1.3.2.46. *. Hint. Review the definition of the definite integral and in particular
Definitions 1.1.9 and 1.1.11.
1.3.2.47. *. Hint. Review the definition of the definite integral and in particular
Definitions 1.1.9 and 1.1.11.
1.3.2.48. Hint. Carefully check the Fundamental Theorem of Calculus: as writ-
ten, it only applies directly to F (x) when x ≥ 0.
Is F (x) even or odd?
1.3.2.49. *. Hint. In general, the equation of the tangent line to the graph of
y = f (x) at x = a is y = f (a) + f 0 (a) (x − a).
1.3.2.50. Hint. Recall tan2 x + 1 = sec2 x.
1.3.2.51. Hint. Since the integration is with respect to t, the x3 term can be
moved outside the integral.
1.3.2.52. Hint. Remember that antiderivatives may have a constant term.

1.4 · Substitution
1.4.2 · Exercises

Exercises — Stage 1
1.4.2.1. Hint. One is true, the other false.

562
H INTS FOR E XERCISES

1.4.2.2. Hint. You can check whether the final answer is correct by differentiat-
ing.
1.4.2.3. Hint. Check the limits.
1.4.2.4. Hint. Check every step. Do they all make sense?
d
1.4.2.6. Hint. What is dx
{f (g(x))}?

Exercises — Stage 2
1.4.2.7. *. Hint. What is the derivative of the argument of the cosine?
1.4.2.8. *. Hint. What is the title of the current section?
1.4.2.9. *. Hint. What is the derivative of x3 + 1?
1.4.2.10. *. Hint. What is the derivative of log x?
1.4.2.11. *. Hint. What is the derivative of 1 + sin x?
1.4.2.12. *. Hint. cos x is the derivative of what?
1.4.2.13. *. Hint. What is the derivative of the exponent?
1.4.2.14. *. Hint. What is the derivative of the argument of the square root?
d
√
1.4.2.15. Hint. What is dx log x ?

Exercises — Stage 3
1.4.2.16. *. Hint. There is a short, slightly sneaky method — guess an an-
tiderivative — and a really short, still-more-sneaky method.
1.4.2.17. *. Hint. Review the definition of the definite integral and in particular
Definitions 1.1.9 and 1.1.11.
1.4.2.18. Hint. If w = u2 + 1, then u2 = w − 1.

R Hint. Using a trigonometric identity, this is similar (though not identi-

1.4.2.19.
cal) to tan θ · sec2 θdθ.
1.4.2.20. Hint. If you multiply the top and the bottom by ex , what does this look
like the antiderivative of?
1.4.2.21. Hint. You know methods other than substitution to evaluate definite
integrals.
sin x
1.4.2.22. Hint. tan x =
cos x
1.4.2.23. *. Hint. Review the definition of the definite integral and in particular
Definitions 1.1.9 and 1.1.11.
1.4.2.24. *. Hint. Review the definition of the definite integral and in particular
Definitions 1.1.9 and 1.1.11.

563
H INTS FOR E XERCISES

1.4.2.25. Hint. Find the right Riemann sum for both definite integrals.

1.5 · Area between curves

1.5.2 · Exercises

Exercises — Stage 1
1.5.2.1. Hint. When we say “area between,” we want positive area, not signed
area.
1.5.2.2. Hint. We’re taking rectangles that reach from one function to the other.
1.5.2.3. *. Hint. Draw a sketch first.
1.5.2.4. *. Hint. Draw a sketch first.
1.5.2.5. *. Hint. You can probably find the intersections by inspection.
1.5.2.6. *. Hint. To find the intersection, plug x = 4y 2 into the equation x+12y +
5 = 0.

Exercises — Stage 2
1.5.2.7. *. Hint. If the bottom function is the x-axis, this is a familiar question.
1.5.2.8. *. Hint. Part of the job is to determine whether y = x lies above or
below y = 3x − x2 .
1.5.2.9. *. Hint. Guess the intersection points by trying small integers.
1.5.2.10. *. Hint. Draw a sketch first. You can also exploit a symmetry of the
region to simplify your solution.
1.5.2.11. *. Hint. Figure out where the two curves cross. To determine which
curve is above the other, try evaluating f (x) and g(x) for some simple value of x.
Alternatively, consider x very close to zero.
1.5.2.12. *. Hint. Think about whether it will easier to use vertical strips or
horizontal strips.
1.5.2.13. Hint. Writing an integral for this is nasty. How can you avoid it?

Exercises — Stage 3
1.5.2.14. *. Hint. You are asked for the area, not the signed area. Be very careful
about signs.
1.5.2.15. *. Hint. You are asked for the area, not the signed area. Draw a sketch
of the region and be very careful about signs.
1.5.2.16. *. Hint. You have to determine whether
√
• the curve y = f (x) = x 25 − x2 lies above the line y = g(x) = 3x for all
0 ≤ x ≤ 4 or

564
H INTS FOR E XERCISES

• the curve y = f (x) lies below the line y = g(x) for all 0 ≤ x ≤ 4 or

• y = f (x) and y = g(x) cross somewhere between x = 0 and x = 4.

√

One way to do so is to study the sign of f (x) − g(x) = x 25 − x2 − 3 .
1.5.2.17. Hint. Flex those geometry muscles.
1.5.2.18. Hint. These two functions have three points of intersection. This ques-
tion is slightly messy, but uses the same concepts we’ve been practicing so far.

1.6 · Volumes
1.6.2 · Exercises

Exercises — Stage 1
1.6.2.1. Hint. The horizontal cross-sections were discussed in Example 1.6.1.
1.6.2.2. Hint. What are the dimensions of the cross-sections?
1.6.2.3. Hint. There are two different kinds of washers.
1.6.2.4. *. Hint. Draw sketches. The mechanically easiest way to answer part
(b) uses the method of cylindrical shells, which is in the optional section 1.6. The
method of washers also works, but requires you to have more patience and also
to have a good idea what the specified region looks like. Look at your sketch
very careful when identifying the ends of your horizontal strips.
1.6.2.5. *. Hint. Draw sketchs.
1.6.2.6. *. Hint. Draw a sketch.
1.6.2.7. Hint. If you take horizontal slices (parallel to one face), they will all be
equilateral triangles.
Be careful not to confuse the height of a triangle with the height of the tetrahe-
dron.

Exercises — Stage 2
1.6.2.8. *. Hint. Sketch the region.
1.6.2.9. *. Hint. Sketch the region first.
1.6.2.10. *. Hint. You can save yourself quite a bit of work by interpreting the
integral as the area of a known geometric figure.
1.6.2.11. *. Hint. See Example 1.6.3.
1.6.2.12. *. Hint. See Example 1.6.5.
1.6.2.13. *. Hint. Sketch the region. To find where the curves intersect, look at
where cos( x2 ) and x2 − π 2 both have roots.
1.6.2.14. *. Hint. See Example 1.6.6.

565
H INTS FOR E XERCISES

1.6.2.15. *. Hint. See Example 1.6.6. Imagine cross-sections with shadow paral-
lel to the y-axis, sticking straight out of the xy-plane.
1.6.2.16. *. Hint. See Example 1.6.1.

Exercises — Stage 3
1.6.2.17. Hint. (a) Don’t be put off by phrases like “rotating an ellipse about its
minor axis.” This is the same kind of volume you’ve been calculating all section.
(b) Hopefully, you sketched the ellipse in part (a). What was its smallest radius?
Its largest? These correspond to the polar and equitorial radii, respectively.
(c) Combine your answers from (a) and (b).
(d) Remember that the absolute error is the absolute difference of your two
results–that is, you subtract them and take the absolute value. The relative er-
ror is the absolute error divided by the actual value (which we’re taking, for our
purposes, to be your answer from (c)). When you take the relative error, lots of
terms will cancel, so it’s easiest to not use a calculator till the end.
1.6.2.18. *. Hint. To find the points of intersection, set 4 − (x − 1)2 = x + 1.
1.6.2.19. *. Hint. You can somewhat simplify your calculations in part (a) (but
not part (b)) by using the fact that R is symmetric about the line y = x.
When you’re solving an equation for x, be careful about your signs: x − 1 is
negative.
1.6.2.20. *. Hint. The mechanically easiest way to answer part (b) uses the
method of cylindrical shells, which we have not covered. The method of washers
also works, but requires you have enough patience and also to have a good idea
what R looks like. So it is crucial to first sketch R. Then be very careful in
identifying the left end of your horizontal strips.
1.6.2.21. *. Hint. Note that the curves cross. The area of this region was found
in Problem 1.5.2.14 of Section 1.5. It would be useful to review that problem.
1.6.2.22. Hint. You can use ideas from this section to answer the question. If
you take a very thin slice of the column, the density is almost constant, so you
can find the mass. Then you can add up all your little slices. It’s the same idea as
volume, only applied to mass.
Do be careful about units: in the problem statement, some are given in metres,
others in kilometres.
If you’re having a hard time with the antiderivative, try writing the exponential
function with base e. Remember 2 = elog 2 .

1.7 · Integration by parts

1.7.2 · Exercises

566
H INTS FOR E XERCISES

Exercises — Stage 1
1.7.2.1. Hint. Read back over Sections 1.4 and 1.7. When these methods are
introduced, they are justified using the corresponding differentiation rules.
R R
1.7.2.2. Hint. Remember our rule: udv = uv − vdu. So, we take u and use it
to make du, and we take dv and use it to make v.
1.7.2.3. Hint. According to the quotient rule,

g(x)f 0 (x) − f (x)g 0 (x)

 
d f (x)
= .
dx g(x) g 2 (x)

Antidifferentiate both sides of the equation, then solve for the expression in the
question.
1.7.2.4. Hint. Remember all the antiderivatives differ only by a constant, so you
can write them all as v(x) + C for some C.
1.7.2.5. Hint. What integral do you have to evaluate, after you plug in your
choices to the integration by parts formula?

Exercises — Stage 2
1.7.2.6. *. Hint. You’ll probably want to use integration by parts. (It’s the title
of the section, after all). You’ll break the integrand into two parts, integrate one,
and differentiate the other. Would you rather integrate log x, or differentiate it?
1.7.2.7. *. Hint. This problem is similar to Question 6.
1.7.2.8. *. Hint. Example 1.7.5 shows you how to find the antiderivative. Then
the Fundamental Theorem of Calculus Part 2 gives you the definite integral.
1.7.2.9. *. Hint. Compare to Question 8. Try to do this one all the way through
without peeking at another solution!
1.7.2.10. Hint. If at first you don’t succeed, try using integration by parts a few
times in a row. Eventually, one part will go away.
1.7.2.11. Hint. Similarly to Question 10, look for a way to use integration by
parts a few times to simplify the integrand until it is antidifferentiatable.
1.7.2.12. Hint. Use integration by parts twice to get an integral with only a
trigonometric function in it.
1.7.2.13. Hint. If you let u = log t in the integration by parts, then du works
quite nicely with the rest of the integrand.
1.7.2.14. Hint. Those square roots are a little disconcerting– get rid of them with
a substitution.
1.7.2.15. Hint. This can be solved using the same ideas as Example 1.7.8 in your
text.

567
H INTS FOR E XERCISES

1.7.2.16. Hint. Not every integral should be evaluated using integration by

parts.
1.7.2.17. *. Hint. You know, or can easily look up, the derivative of arccosine.
You can use a similar trick as the book did when antidifferentiating other inverse
trigonometric functions in Example 1.7.9.

Exercises — Stage 3
1.7.2.18. *. Hint. After integrating by parts, do some algebraic manipulation to
the integral until it’s clear how to evaluate it.
1.7.2.19. Hint. After integration by parts, use a substitution.
1.7.2.20. Hint. This example is similar to Example 1.7.10 in the text. The func-
tions ex/2 and cos(2x) both do not substantially alter when we differentiate or
antidifferentiate them. If we use integration by parts twice, we’ll end up with
an expression that includes our original integral. Then we can just solve for the
original integral in the equation, without actually integrating.
1.7.2.21. Hint. This looks a bit like a substitution problem, because we have an
“inside function.”
It might help to review Example 1.7.11.
1.7.2.22. Hint. Start by simplifying.
1.7.2.23. Hint. sin(2x) = 2 sin x cos x
1.7.2.24. Hint. What is the derivative of xe−x ?
1.7.2.25. *. Hint. You’ll want to do an integration by parts for (a)–check the end
result to get a guess as to what your parts should be. A trig identity and some
amount of algebraic manipulation will be necessary to get the final form.
1.7.2.26. *. Hint. See Examples 1.7.9 and 1.6.5 for refreshers on integrating
arctangent, and using washers.
Remember tan2 x + 1 = sec2 x, and sec2 x is easy to integrate.
1.7.2.27. *. Hint. Your integral can be broken into two integrals, which yield to
two different integration methods.
1.7.2.28. *. Hint. Think, first, about how to get rid of the square root in the
argument of f 00 , and, second, how to convert f 00 into f 0 . Note that you are told
that f 0 (2) = 4 and f (0) = 1, f (2) = 3.
1.7.2.29. Hint. Interpret the limit as a right Riemann sum.

1.8 · Trigonometric Integrals

1.8.4 · Exercises

Exercises — Stage 1

568
H INTS FOR E XERCISES

1.8.4.1. Hint. Go ahead and try it!

1.8.4.2. Hint. Use the substitution u = sec x.
1.8.4.3. Hint. Divide both sides of the second identity by cos2 x.

Exercises — Stage 2
1.8.4.4. *. Hint. See Example 1.8.6. Note that the power of cosine is odd, and
the power of sine is even (it’s zero).
1.8.4.5. *. Hint. See Example 1.8.7. All you need is a helpful trig identity.
1.8.4.6. *. Hint. The power of cosine is odd, so we can reserve one cosine for
du, and turn the rest into sines using the identity sin2 x + cos2 x = 1.
1.8.4.7. Hint. Since the power of sine is odd (and positive), we can reserve one
sine for du, and turn the rest into cosines using the identity sin2 + cos2 x = 1.
1.8.4.8. Hint. When we have even powers of sine and cosine both, we use the
identities in the last two lines of Equation 1.8.3.
1.8.4.9. Hint. Since the power of sine is odd, you can use the substitution u =
cos x.
1.8.4.10. Hint. Which substitution will work better: u = sin x, or u = cos x?
1.8.4.11. Hint. Try a substitution.
1.8.4.12. *. Hint. For practice, try doing this in two ways, with different substi-
tutions.
1.8.4.13. *. Hint. A substitution will work. See Example 1.8.14 for a template
for integrands with even powers of secant.
1.8.4.14. Hint. Try the substitution u = sec x.
1.8.4.15. Hint. Compare to Question 14.
1.8.4.16. Hint. What is the derivative of tangent?
1.8.4.17. Hint. Don’t be scared off by the non-integer power of secant. You can
still use the strategies in the notes for an odd power of tangent.
1.8.4.18. Hint. Since there are no secants in the problem, it’s difficult to use the
substitution u = sec x that we’ve enjoyed in the past. Example 1.8.12 in the text
provides a template for antidifferentiating an odd power of tangent.
1.8.4.19. Hint. Integrating even powers of tangent is surprisingly different from
integrating odd powers of tangent. You’ll want to use the identity tan2 x =
sec2 x − 1, then use the substitution u = tan x, du = sec2 xdx on (perhaps only
a part of) the resulting integral. Example 1.8.16 show you how this can be ac-
complished.

569
H INTS FOR E XERCISES

1.8.4.20. Hint. Since there is an even power of secant in the integrand, we can
use the substitution u = tan x.
1.8.4.21. Hint. How have we handled integration in the past that involved an
odd power of tangent?
1.8.4.22. Hint. Remember e is some constant. What are our strategies when
the power of secant is even and positive? We’ve seen one such substitution in
Example 1.8.15.

Exercises — Stage 3
1.8.4.23. *. Hint. See Example 1.8.16 for a strategy for integrating powers of
tangent.
sin x
1.8.4.24. Hint. Write tan x = .
cos x
1
1.8.4.25. Hint. = sec θ
cos θ
cos x
1.8.4.26. Hint. cot x =
sin x
1.8.4.27. Hint. Try substituting.
1.8.4.28. Hint. To deal with the “inside function,” start with a substitution.
1.8.4.29. Hint. Try an integration by parts.

1.9 · Trigonometric Substitution

1.9.2 · Exercises

Exercises — Stage 1
1.9.2.1. *. Hint. The beginning of this section has a template for choosing a
substitution. Your goal is to use a trig identity
pto turn the argument of the square
root into a perfect square, so you can cancel (something)2 = |something|.
1.9.2.2. Hint. You want to do the same thing you did in Question 1, but you’ll
have to complete the square first.
1.9.2.3. Hint. Since θ is acute, you can draw it as an angle of a right trian-
gle. The given information will let you label two sides of the triangle, and the
Pythagorean Theorem will lead you to the third.
1.9.2.4. Hint. You can draw a right triangle with angle θ, and use the given
information to label two of the sides. The Pythagorean Theorem gives you the
third side.

Exercises — Stage 2

570
H INTS FOR E XERCISES

1.9.2.5. *. Hint. As in Question 1, choose an appropriate substitution. Your

answer should be in terms of your original variable, x, which can be achieved
using the methods of Question 3.
1.9.2.6. *. Hint. As in Question 1, choose an appropriate substitution. Your
answer will be a number, so as long as you change your limits of integration
when you substitute, you don’t need to bother changing the antiderivative back
into the original variable x. However, you might want to use the techniques of
Question 4 to simplify your final answer.
1.9.2.7. *. Hint. Question 1 guides the way to finding the appropriate substi-
tution. Since the integral is definite, your final answer will be a number. Your
limits of integration should be common reference angles.
1.9.2.8. *. Hint. Question 1 guides the way to finding the appropriate substitu-
tion. Since you have in indefinite integral, make sure to get your answer back in
terms of the original variable, x. Question 3 gives a reliable method for this.
1.9.2.9. Hint. A trig substitution is not the easiest path.
1.9.2.10. *. Hint. To antidifferentiate, change your trig functions into sines and
cosines.
1.9.2.11. *. Hint. The integrand should simplify quite far after your substitu-
tion.
1.9.2.12. *. Hint. In part (a) you are asked to integrate an even power of cos x.
For part (b) you can use a trigonometric substitution to reduce the integral of part
(b) almost to the integral of part (a).
1.9.2.13. Hint. What is the symmetry of the integrand?
1.9.2.14. *. Hint. See Example 1.9.3.
1.9.2.15. *. Hint. To integrate an even power of tangent, use the identity
tan2 x = sec2 x − 1.
1.9.2.16. Hint. A trig substitution is not the easiest path.
1.9.2.17. *. Hint. Complete the square. Your final answer will have an inverse
trig function in it.
1.9.2.18. Hint. To antidifferentiate even powers of cosine, use the formula
cos2 θ = 21 (1 + cos(2θ)). Then, remember sin(2θ) = 2 sin θ cos θ.

1.9.2.19. Hint. After substituting, use the identity tan2 x = sec2 x − 1 more than
once. Z
Remember sec xdx = log sec x + tan x + C.

1.9.2.20. Hint. There’s no square root, but we can still make use of the substitu-
tion x = tan θ.

571
H INTS FOR E XERCISES

Exercises — Stage 3
1.9.2.21. Hint. You’ll probably want to use the identity tan2 θ + 1 = sec2 θ more
than once.
1.9.2.22. Hint. Complete the square — refer to Question 2 if you want a re-
fresher. The constants aren’t pretty, but don’t let them scare you.
1.9.2.23. Hint. After substituting, use the identity sec2 u = tan2 u + 1. It might
help to break the integral into a few pieces.
1.9.2.24. Hint. Make use of symmetry, and integrate with respect to y (rather
than x). The limits of integration should be reference angles.
1.9.2.25. Hint. Use the symmetry of the function to re-write your integrals with-
out an absolute value.

2
1.9.2.26. Hint. Think of ex as ex/2 , and use a trig substitution. Then, use the
identity sec2 θ = tan2 θ + 1.
1.9.2.27. Hint.

a Use logarithm rules to simplify first.

b Think about domains.

c What went wrong in part (b)? At what point in the work was that problem
introduced?
There is a subtle but important point mentioned in the introductory text to
Section 1.9 that may help you make sense of things.
1.9.2.28. Hint. Consider the ranges
√ of the inverse trigonometric functions. For
(c), also consider the domain of x2 − a2 .

1.10 · Partial Fractions

1.10.4 · Exercises

Exercises — Stage 1
1.10.4.1. Hint. If a quadratic function can be factored as (ax+b)(cx+d) for some
constants a, b, c, d, then it has roots − ab and − dc .
1.10.4.2. *. Hint. Review Equations 1.10.7 through 1.10.11. Be careful to fully
factor the denominator.
1.10.4.3. *. Hint. Review Example 1.10.1. Is the “Algebraic Method” or the
“Sneaky Method” going to be easier?
1.10.4.4. Hint. For each part, use long division as in Example 1.10.4.
1.10.4.5. Hint. (a) Look for a pattern you can exploit to factor out
√ a linear√term.
(b) If you set y = x2 , this is quadratic. Remember (x2 − a) = (x + a)(x − a) as

572
H INTS FOR E XERCISES

long as a is positive
(c),(d) Look for integer roots, then use long division.
1.10.4.6. Hint. Why do we do partial fraction decomposition at all?

Exercises — Stage 2
1.10.4.7. *. Hint. What is the title of this section?
1.10.4.8. *. Hint. You can save yourself some work in developing your par-
tial fraction decomposition by renaming x2 to y and comparing the result with
Question 7.
1.10.4.9. *. Hint. Review Steps 3 (particularly the “Sneaky Method”) and 4 of
Example 1.10.3.
1.10.4.10. *. Hint. Review Steps 3 (particularly the “Sneaky Method”) and 4 of
d 1
Example 1.10.3. Remember dx {arctan x} = 1+x 2.

1.10.4.11. *. Hint. Fill in the blank: the integrand is a function.

1.10.4.12. *. Hint. The integrand is yet another function.
1.10.4.13. Hint. Since the degree of the numerator is the same as the degree
of the denominator, we can’t do our partial fraction decomposition before we
simplify the integrand.
1.10.4.14. Hint. The degree of the numerator is not smaller than the degree of
the denominator.
Your final answer will have an arctangent in it.
1.10.4.15. Hint. In the partial fraction decomposition, several constants turn out
to be 0.
1.10.4.16. Hint. Factor (2x − 1) out of the denominator to get started. You don’t
need long division for this step.
1.10.4.17. Hint. When it comes time to integrate, look for a convenient substi-
tution.

Exercises — Stage 3
1 sin x
1.10.4.18. Hint. csc x = =
sin x sin2 x
1.10.4.19. Hint. Use the partial fraction decomposition from Queston 18 to save
yourself some time.
1.10.4.20. Hint. In the final integration, complete the square to make a piece of
the integrand look more like the derivative of arctangent.
1.10.4.21. Hint. Review Question 1.9.2.20 in Section 1.9 for antidifferentiation
tips.

573
H INTS FOR E XERCISES

1.10.4.22. Hint. Partial fraction decomposition won’t simplify this any more.
Use a trig substitution.
1.10.4.23. Hint. To evaluate the antiderivative, break one of the fractions into
two fractions.
1.10.4.24. Hint. cos2 θ = 1 − sin2 θ
1.10.4.25. Hint. If you’re having a hard time making the substitution, multiply
the numerator and the denominator by ex .
√
1.10.4.26. Hint. Try the substitution u = 1 + ex . You’ll need to do long divi-
sion before you can use partial fraction decomposition.
1.10.4.27. *. Hint. The mechanically easiest way to answer part (c) uses the
method of cylindrical shells, which we have not covered. The method of washers
also works, but requires you have enough patience and also to have a good idea
what R looks like. So look at the sketch in part (a) very carefully when identifying
the left endpoints of your horizontal strips.
1.10.4.28. Hint. You’ll need to use two regions, because the curves cross.
1.10.4.29. Hint. For (b), use the Fundamental Theorem of Calculus Part 1.

1.11 · Numerical Integration

1.11.6 · Exercises

Exercises — Stage 1
1.11.6.1. Hint. The absolute error is the difference of the two values; the relative
error is the absolute error divided by the exact value; the percent error is one
hundred times the relative error.
1.11.6.2. Hint. You should have four rectangles in one drawing, and four trape-
zoids in another.
1.11.6.3. Hint. Sketch the second derivative–it’s quadratic.
1.11.6.4. Hint. You don’t have to find the actual, exact maximum the second
derivative achieves–you only have to give a reasonable “ceiling” that it never
breaks through.
1.11.6.5. Hint. To compute the upper bound on the error, find an upper bound
on the fourth derivative of cosine, then use Theorem 1.11.13 in the text.
To find the actual error, you need to find the actual value of A.
1.11.6.6. Hint. Find a function with f 00 (x) = 3 for all x in [0, 1].
1.11.6.7. Hint. You’re allowed to use common sense for this one.
1.11.6.8. Hint. For part (b), consider Question 7.
1.11.6.9. *. Hint. Draw a sketch.

574
H INTS FOR E XERCISES

1.11.6.10. Hint. The error bound for the approximation is given in Theo-
rem 1.11.13 in the text. You want this bound to be zero.

Exercises — Stage 2
1.11.6.11. Hint. Follow the formulas in Equations 1.11.2, 1.11.6, and 1.11.9 in
the text.
1.11.6.12. *. Hint. See Section 1.11.1. You should be able to simplify your an-
swer to an exact value (in terms of π).
1.11.6.13. *. Hint. See Section 1.11.2. To set up the volume integral, see Example
1.6.6. Note the dimensions given for the cross sections are diameters, not radii.
1.11.6.14. *. Hint. See Section 1.11.3 and compare to Question 1.11.6.13. Note
the table gives diameters, not radii.
1.11.6.15. *. Hint. See §1.11.3. To set up the volume integral, see Example 1.6.6,
or Question 14.
Note that the table gives the circumference, not radius, of the tree at a given
height.
1.11.6.18. *. Hint. The main step is to find an appropriate value of M . It is not
necessary to find the smallest possible M .
1.11.6.19. *. Hint. The main step is to find M . This question is unusual in that
its wording requires you to find the smallest possible allowed M .
1.11.6.20. *. Hint. The main steps in part (b) are to find the smallest possible
values of M and L.
1.11.6.21. *. Hint. As usual, the biggest part of this problem is finding L. Don’t
be thrown off by the error bound being given slightly differently from Theo-
rem 1.11.13 in the text: these expressions are equivalent, since ∆x = b−a
n
.
1
1.11.6.22. *. Hint. The function e−2x = 2x is positive and decreasing, so its
e
maximum occurs when x is as small as possible.
1
1.11.6.23. *. Hint. Since 5 is a decreasing function when x > 0, look for its
x
maximum value when x is as small as possible.
1.11.6.24. *. Hint. The “best ... approximations that you can” means using the
maximum number of intervals, given the information available.
The final sentence in part (b) is just a re-statement of the error bounds we’re
k
familiar with from Theorem 1.11.13 in the text. The information s(k) (x) ≤
1000
gives you values of M and L when you set k = 2 and k = 4, respectively.
1.11.6.25. *. Hint. Set the error bound to be less than 0.001, then solve for n.

Exercises — Stage 3

575
H INTS FOR E XERCISES

1.11.6.26. *. Hint. See Section 1.11.3. To set up the volume integral, see Example
1.6.2.
Since the cross-sections of the pool are semi-circular disks, a section that is d

2
metres across will have area 21 π d2 square feet. Based on the drawing, you may
assume the very ends of the pool have distance 0 feet across.
1.11.6.27. *. Hint. See Example 1.11.15.
Don’t get caught up in the interpretation of the integral. It’s nice to see how
integrals can be used, but for this problem, you’re still just approximating the
integral given, and bounding the error.
When you find the second derivative to bound your error, pay attention to the
difference between the integrand and g(r).
1.11.6.28. *. Hint. See Example 1.11.16. You’ll want to use a calculator for the
approximation in (a), and for finding the appropriate number of intervals in (b).
Remember that Simpson’s rule requires an even number of intervals.
1.11.6.29. *. Hint. See Example 1.11.16.
Rather than calculating the fourth derivative of the integrand, use the graph to
find the largest absolute value it attains over our interval.
1.11.6.30. *. Hint. See Example 1.11.15.
You’ll have to differentiate f (x). To that end, you may also want to review the
fundamental theorem of calculus and, in particular, Example 1.3.5.
You don’t have to find the best possible value for M . A reasonable upper bound
on |f 00 (x)| will do.
To have five decimal places of accuracy, your error must be less than 0.000005.
This ensures that, if you round your approximation to five decimal places, they
will all be correct.
1.11.6.31. Hint. To find the maximum value of |f 00 (x)|, check its critical points
and endpoints.
x
1
Z
1.11.6.32. Hint. In using Simpson’s rule to approximate dt with n inter-
1 t
x−1
vals, a = 1, b = x, and ∆x = .
n
1.11.6.33. Hint.
R2 1 π π
R2 1
• 1 1+x 2 dx = arctan(2) − 4 , so arctan(2) = 4 + 1 1+x2 dx

R2 1
• If an approximation A of the integral 1 1+x 2 dx has error at most ε, then
R2 1
A − ε ≤ 1 1+x2 dx ≤ A + ε.

• Looking at our target interval will tell you how small ε needs to be, which
in turn will tell you how many intervals you need to use.

• You can show, by considering the numerator and denominator separately,

that |f (4) (x)| ≤ 30.75 for every x in [1, 2].

576
H INTS FOR E XERCISES
R2 1
• If you use Simpson’s rule to approximate 1 1+x2
dx, you won’t need very
many intervals to get the requisite accuracy.

1.12 · Improper Integrals

1.12.4 · Exercises

Exercises — Stage 1
1.12.4.1. Hint. There are two kinds of impropreity in an integral: an infinite
discontinuity in the integrand, and an infinite limit of integration.
1.12.4.2. Hint. The integrand is continuous for all x.
1.12.4.3. Hint. What matters is which function is bigger for large values of x,
not near the origin.
1.12.4.4. *. Hint. Read both the question and Theorem 1.12.17 very carefully.
1.12.4.5. Hint. (a) What if h(x) is negative? What if it’s not?
(b) What if h(x) is very close to f (x) or g(x), rather than right in the middle?
(c) Note |h(x)| ≤ 2f (x).

Exercises — Stage 2
1.12.4.6. *. Hint. First: is the integrand unbounded, and if so, where?
Second: when evaluating integrals, always check to see if you can use a simple
substitution before trying a complicated procedure like partial fractions.
1.12.4.7. *. Hint. Is the integrand bounded?
1.12.4.8. *. Hint. See Example 1.12.21. Rather than antidifferentiating, you can
find a nice comparison.
1.12.4.9. *. Hint. Which of the two terms in the denominator is more important
when x ≈ 0? Which one is more important when x is very large?
1.12.4.10. Hint. Remember to break the integral into two pieces.
1.12.4.11. Hint. Remember to break the integral into two pieces.
1.12.4.12. Hint. The easiest test in this case is limiting comparison, Theo-
rem 1.12.22.
1.12.4.13. Hint. Not all discontinuities cause an integral to be improper–only
infinite discontinuities.
1.12.4.14. *. Hint. Which of the two terms in the denominator is more important
when x is very large?
1.12.4.15. *. Hint. Which of the two terms in the denominator is more important
when x ≈ 0? Which one is more important when x is very large?

577
H INTS FOR E XERCISES

1.12.4.16. *. Hint. What are the “problem x’s” for this integral? Get a simple
approximation to the integrand near each.

Exercises — Stage 3
1.12.4.17. Hint. To find the volume of the solid, cut it into horizontal slices,
which are thin circular disks.
The true/false statement is equivalent to saying that the improper integral giving
the volume of the solid when a = 0 diverges to infinity.
1
1.12.4.18. *. Hint. Review Example 1.12.8. Remember the antiderivative of x
looks very different from the antiderivative of other powers of x.
1.12.4.19. Hint. Compare to Example 1.12.14 in the text. You can antidifferenti-
ate with a u-substitution.
1.12.4.20. Hint. To evaluate the integral, you can factor the denominator.
π
Recall lim arctan x = . For the other limits, use logarithm rules, and beware of
x→∞ 2
indeterminate forms.
1.12.4.21. Hint. Break up the integral. The absolute values give you a nice even
function, so you can replace |x − a| with x − a if you’re careful about the limits of
integration.
1.12.4.22. Hint. Use integration by parts twice to find the antiderivative of
e−x sin x, as in Example 1.7.10. Be careful with your signs–it’s easy to make a
mistake with all those negatives.
If you’re having a hard time taking the limit at the end, review the Squeeze The-
orem (see the CLP-1 text).
1.12.4.23. *. Hint. What is the limit of the integrand when x → 0?
1.12.4.24. Hint. The only “source of impropriety” is the infinite domain of inte-
gration. Don’t be afraid to be a little creative to make a comparison work.
1.12.4.25. *. Hint. There are two things that contribute to your error: using t as
the upper bound instead of infinity, and using n intervals for the approximation.
R ∞ e−x R t e−x
First, find a t so that the error introduced by approximating 0 1+x dx by 0 1+x dx
1 −4
is at most 2 10 . Then, find your n.
1.12.4.26. Hint. Look for a place to use Theorem 1.12.20.
Examples 1.2.10 and 1.2.11 have nice results about the area under an even/odd
curve.
1.12.4.27. Hint. x should be a real number

1.13 · More Integration Examples

· Exercises

Exercises — Stage 1

578
H INTS FOR E XERCISES

1.13.1. Hint. Each option in each column should be used exactly once.

Exercises — Stage 2
1.13.2. Hint. The integrand is the product of sines and cosines. See how this
was handled with a substitution in Section 1.8.1.
After your substitution, you should have a polynomial expression in u—but it
might take some simplification to get it into a form you can easily integrate.
1.13.3. Hint. We notice that the integrand has a quadratic polynomial under
the square root. If that polynomial were a perfect square, we could get rid of the
square root: try a trig substitution, as in Section 1.9.
The identity sin(2θ) = 2 sin θ cos θ might come in handy.
1.13.4. Hint. Notice the integral is improper. When you compute the limit,
l’Hôpital’s rule might help.
x−1
If you’re struggling to think of how to antidifferentiate, try writing x = (x −
e
1)e−x .
1.13.5. Hint. Which method usually works for rational functions (the quotient
of two polynomials)?
1
1.13.6. Hint. It would be nice to replace logarithm with its derivative, .
x
1.13.7. *. Hint. The integrand is a rational function, so it is possible to use
partial fractions. But there is a much easier way!
1.13.8. *. Hint. You should prepare your own personal internal list of integra-
tion techniques ordered from easiest to hardest. You should have associated to
each technique your own personal list of signals that you use to decide when the
technique is likely to be useful.
1.13.9. *. Hint. Despite both containing a trig function, the two integrals are
easiest to evaluate using different methods.
1.13.10. *. Hint. For the integral of secant, see See Section 1.8.3 or Example
1.10.5.
In (c), notice the denominator is not yet entirely factored.
1.13.11. *. Hint. Part (a) can be done by inspection — use a little highschool
geometry! Part (b) is reminiscent of the antiderivative of logarithm—how did
we find that one out? Part (c) is an improper integral.
1.13.12. Hint. Use the substitution u = sin θ.
1.13.13. *. Hint. For (c), try a little algebra to split the integral into pieces that
are easy to antidifferentiate.
1.13.14. *. Hint. If you’re stumped, review Sections 1.8, 1.9, and 1.10.

579
H INTS FOR E XERCISES

1.13.15. *. Hint. For part (a), see Example 1.7.11. For part (d), see Example
1.10.4.
1.13.16. *. Hint. For part (b), first complete the square in the denominator. You
can save some work by first comparing the derivative of the denominator with
the numerator. For part (d) use a simple substitution.
1.13.17. *. Hint. For part (b), complete the square in the denominator. You can
save some work by first comparing the derivative of the denominator with the
numerator.
1.13.18. *. Hint. For part (a), the numerator is the derivative of a function that
appears in the denominator.
1.13.19. Hint. The integral is improper.
1.13.20. *. Hint. For part (a), can you convert this into a partial fractions inte-
gral? For part (b), start by completing the square inside the square root.
1.13.21. *. Hint. For part (b), the numerator is the derivative of a function that
is embedded in the denominator.
1.13.22. Hint. Try a substitution.
1.13.23. Hint. Note the quadratic function under the square root: you can solve
this with trigonometric substitution, as in Section 1.9.
1.13.24. Hint. Try a u-substitution, as in Section 1.8.2.
1.13.25. Hint. What’s the usual trick for evaluating a rational function (quotient
of polynomials)?
1.13.26. Hint. If the denominator were x2 + 1, the antiderivative would be arct-
angent.
1.13.27. Hint. Simplify first.
1.13.28. Hint. x3 + 1 = (x + 1)(x2 − x + 1)
1.13.29. Hint. You have the product of two quite dissimilar functions in the
integrand—try integration by parts.

Exercises — Stage 3
1.13.30. Hint. Use the identity cos(2x) = 2 cos2 x − 1.
1.13.31. Hint. Using logarithm rules can make the integrand simpler.
1.13.32. Hint. What is the derivative of the function in the denominator? How
could that be useful to you?
1.13.33. *. Hint. For part (a), the substitution u = log x gives an integral that
you have seen before.

580
H INTS FOR E XERCISES

1.13.34. *. Hint. For part (a), split the integral in two. One part may be eval-
uated by interpreting it geometrically, without doing any integration at all. For
part (c), multiply both the numerator and denominator by ex and then make a
substitution.
√
1.13.35. Hint. Let u = 1 − x.
1.13.36. Hint. Use the substitution u = ex .
1.13.37. Hint. Use integration by parts. If you choose your parts well, the re-
sulting integration will be very simple.
sin x
1.13.38. Hint. cos2 x
= tan x sec x
1.13.39. Hint. The cases n = −1 and n = −2 are different from all other values
of n.
√ √
1.13.40. Hint. x4 + 1 = (x2 + 2x + 1)(x2 − 2x + 1)

2 · Applications of Integration
2.1 · Work
2.1.2 · Exercises

Exercises — Stage 1
2
2.1.2.1. Hint. Watch your units: 1 J = 1 kg·m
sec2
, but your mass is not given in
kilograms, and your height is not given in metres.
2.1.2.2. Hint. The force of the rock on the ground is the product of its mass and
the acceleration due to gravity.
2.1.2.3. Hint. Adding or subtracting two quantities of the same units doesn’t
change the units. For example, if I have one metre of rope, and I tie on two more
metres of rope, I have 1 + 2 = 3 metres of rope — not 3 centimetres of rope, or 3
kilograms of rope.
Multiplying or dividing quantities of some units gives rise to a quantity with the
product or quotient of those units. For example, if I buy ten pounds of salmon
50 dollars 50 dollars
for $50, the price of my salmon is = = 5 dollars . (Not 5 pound-
10 pounds 10 pound pound

dollars, or 5 pounds.)
2.1.2.4. Hint. See Question 3.
2.1.2.5. Hint. Hooke’s law says that the force required to stretch a spring x
units past its natural length is proportional to x; that is, there is some constant k
associated with the individual spring such that the force required to stretch it x
m past its natural length is kx.
2.1.2.6. Hint. Definition 2.1.1 tells us the work done by the force from x = 1 to
Rb
x = b is W (b) = 1 F (x)dx, where F (x) is the force on the object at position x. To
recover the equation for F (x), use the Fundamental Theorem of Calculus.

581
H INTS FOR E XERCISES

Exercises — Stage 2
2.1.2.7. *. Hint. Review Definition 2.1.1 for calculating the work done by a force
over a distance.
c
2.1.2.8. Hint. For (a), `−x is meausured in Newtons, while ` and x are in metres.
For (b), notice the similarities and differences between the tube of air and a spring
obeying Hooke’s law.
2.1.2.9. *. Hint. See Example 2.1.2. Be careful about your units.
2.1.2.10. *. Hint. Be careful about the units.
2.1.2.11. *. Hint. Suppose that the bucket is a distance y above the ground.
How much work is required to raise it an additional height dy?
2.1.2.12. Hint. Since you’re given the area of the cross-section, it doesn’t matter
what shape it has. However, the density of water is given in cubic centimetres,
while the measurements of the tank are given in metres.
2.1.2.13. *. Hint. Consider the work done to lift a horizontal plate from 2 m
below the ground to a height z. You’ll need to know the mass of the plate, which
you can calculate from its volume, since its density is given to you.

dz

2.1.2.14. Hint. You can find the spring constant k from the information about
the hanging kilogram.
2.1.2.15. Hint. Follow the method of Example 2.1.6 and Question 11 in this
section.
2.1.2.16. Hint. Calculating the work done on the rope and the weight separately
makes the computation somewhat easier.
2.1.2.17. Hint. When you pull the box, the force you’re exerting is exactly the
same as the frictional force, but in the opposite direction. In (a), that force is
constant. In (b), it changes. Check Definition 2.1.1 for how to turn force into
work.
2.1.2.18. Hint. Remember that the work done on an object is equal to the change
in its kinetic energy, which is 12 mv 2 , where m is the mass of the object and v is its
velocity. Hooke’s law will tell you how much work was done stretching the
spring.

582
H INTS FOR E XERCISES

2.1.2.19. Hint. As in Question 18 in this section, the change in kinetic energy of

the car is equal to the work done by the compressing struts. The only added step
is to calculate the spring constant, given that a car with mass 2000 kg compresses
the spring 2 cm in Earth’s gravity. You’re not calculating work to find the spring
constant: you’re using the fact that when the car is sitting still, the force exerted
upward by the struts is equal to the force exerted downward by the mass of the
car under gravity.

Exercises — Stage 3
2.1.2.20. Hint. To find the radius of a horizontal layer of water, use similar
triangles. Be careful with centimetres versus metres.
2.1.2.21. *. Hint. See Example 2.1.4 for a basic method for calculating the work
done pumping water.
To find the area of a horizontal layer of water, use some geometry. A horizontal
cross-section of a sphere is a circle, and its radius will depend on the height of
the layer in the tank.
2.1.2.22. Hint. The basic ideas you’ve used already with “cable problems” still
work, you only need to take care that the density of the cable is no longer con-
stant. The mass of a tiny piece of cable, say of length dx, is (density)×(length) =
(10 − x)dx, where x is the distance of our piece from the bottom of the cable.
If you want more work to reference, Question 1.6.2.22 in Section 1.6 finds the
mass of an object of variable density.
2.1.2.23. Hint. To calculate the force on the entire plunger, first find the force on
a horizontal rectangle with height dy at depth y.
Checking units can be a good way to make sure your calculation makes sense.
2.1.2.24. Hint. When y metres of rope have been hauled up, what is the mass of
the water?
2.1.2.25. Hint. The work you’re asked for is an improper integral, moving the
earth and moon infinitely far apart.
2.1.2.26. Hint. You can formulate a guess by considering the work done on the
ball versus the work done on the rope in Question 16, Solution 1. But be careful
— the ball in that problem did not have the same mass as the rope.
2.1.2.27. Hint. There are two things that vary with height: the density of the
liquid, and the area of the cross-section of the tank. Make a formula M (h) for
the mass of a thin layer of liquid h metres below the top of the tank, using
mass=volume×density. The rest of the problem is similar to other tank-pumping
problems in this section.
2.1.2.28. Hint. You can model the motion, instead of a rotation, as dividing the
sand into thin horizontal slices and lifting each of them to their new position.

• In order to calculate the work involved lifting a layer of sand, you need to
know the mass of the layer of sand.

583
H INTS FOR E XERCISES

• To find the mass of a layer of sand, you need its volume and the density of
the sand.

• To find the density of the sand, you need to the volume of the sand: that is,
the volume of half the hourglass.

• The hourglass is a solid of rotation: you can find its volume using an inte-
gral, as in Section 1.6.
2.1.2.29. Hint. Theorem 1.11.13 gives error bounds for the standard types of
numerical approximations. You won’t need very many intervals to achieve the
desired accuracy.

2.2 · Averages
2.2.2 · Exercises

Exercises — Stage 1
2.2.2.1. Hint. See Definition 2.2.2 and the discussion following it for the link
between area under the curve and averages.
2.2.2.2. Hint. Average velocity is discussed in Example 2.2.5. You don’t need
an integral for this.
2.2.2.3. Hint. Much like Problem 2, you don’t need to do any integration here.
2.2.2.4. Hint. Part (a) is asking the length of the pieces we’ve cut our interval
into. Part (c) should be given in terms of f . Our final answer in (d) will resemble
a Riemann sum, but without some extra manipulation it won’t be in exactly the
form of a Riemann sum we’re used to.
2.2.2.5. Hint. For (b), the value of f (0) could be much, much larger than g(0).
2.2.2.6. Hint. The answer is something very simple.

Exercises — Stage 2
2.2.2.7. *. Hint. Apply the definition of “average value” in Section 2.2.
2.2.2.8. *. Hint. You can antidifferentiate x2 log x using integration by parts.
2.2.2.9. *. Hint. You can antidifferentiate an odd power of cosine with a substi-
tution; for an even power of cosine, use the identity cos2 x = 12 1 + cos(2x) .

2.2.2.10. *. Hint. If you’re not sure how to antidifferentiate, try the substitution
u = kx, du = kdx, keeping in mind that k is a constant. Interestingly, your final
answer won’t depend on k.
2.2.2.11. *. Hint. The method of partial fractions can help you antidifferentiate.
1
2.2.2.12. *. Hint. Try the substitution u = log x, du = x
dx.

584
H INTS FOR E XERCISES

1
Remember cos2 x =


2.2.2.13. *. Hint. 1 + cos(2x) .
2

the term 50 cos 12t π has a period of 24 hours, while the

2.2.2.14. Hint. Notice
t


term 200 cos 4380 π has a period of one year.
If n is an approximation of c, then the relative error of n is |n−c|
c
.

2.2.2.15. Hint. A cross section of S at location x is a circle with radius x2 , so area

πx4 . Part (a) is asking for the average of this function on [0, 2].
2.2.2.16. Hint. (a) can be done without calculation
2.2.2.17. Hint. tan2 x = sec2 x − 1
2.2.2.18. Hint. Remember force is the product of the spring constant with the
distance it’s stretched past its natural length. The units given in the question are
not exactly standard, but they are compatible with each other.
You can
find part (b) without any calculation. For (c), remember sin2 x = 12 1 −
cos(2x) .

Exercises — Stage 3
2.2.2.19. *. Hint. The trapezoidal rule is found in Section 1.11.2.
2.2.2.20. Hint. To find a definite integral of the absolute value of a function,
break up the interval of integration into regions where the function is positive,
and intervals where it’s negative.
2.2.2.21. Hint. This is an application of the ideas in Question 20.
2.2.2.22. Hint. Slice the solid into circular disks of radius |f (x)| and thickness
dx.
1
R1 f (0)+f (1)
2.2.2.23. Hint. The question tells you 1−0 0
f (x) dx = 2
.

2.2.2.24. Hint. Set up this question just like Question 23, but with variables for
your limits of integration.
Note (s − t)2 = s2 − 2st + t2 .
2.2.2.25. Hint. What are the graphs of f (x) and f (a + b − x) like?
2.2.2.26. Hint. For (b), express A(x) as an integral, then differentiate.
2.2.2.27. Hint. For (b), consider the cases that f (x) is always bigger or always
smaller than 0. Then, use the intermediate value theorem (see the CLP-1 text).
2.2.2.28. Hint. Try l’Hôpital’s rule.
2.2.2.29. Hint. Use the result of Question 28.

2.3 · Centre of Mass and Torque

2.3.3 · Exercises

585
H INTS FOR E XERCISES

Exercises — Stage 1
2.3.3.1. Hint. It might help to know that −x2 + 2x + 1 = 2 − (x − 1)2 .
2.3.3.2. Hint. The centroid of a region doesn’t have to be a point in the region.
2.3.3.3. Hint. Read over the very beginning of Section 2.3, specifically Equa-
tion 2.3.1.
2.3.3.4. Hint. Use Equation 2.3.1.
2.3.3.5. Hint. Imagine cutting out the shape and setting it on top of a pencil, so
that the pencil lines up with the vertical line x = a. Will the figure balance, or fall
to one side? Which side?
2.3.3.6. Hint. You can find the heights of the centres of mass using symmetry.
2.3.3.7. Hint. Think about whether your answers should have repetition.
2.3.3.8. Hint. The definition of a definite integral (Definition 1.1.9) will tell you
how to convert your limits of sums into integrals.
2.3.3.9. Hint. In (a), the slices all have the same width, so the area of the slices
is larger (and hence the density of R is higher) where T (x) − B(x) is larger.
2.3.3.10. Hint. Part (a) is a significantly different model from the last question.
2.3.3.11. *. Hint. Which method involves more work: horizontal strips or ver-
tical strips?

Exercises — Stage 2
2.3.3.12. Hint. This is a straightforward application of Equation 2.3.4.
1
2.3.3.13. Hint. Remember the derivative of arctangent is 1+x2

2.3.3.14. *. Hint. This is a straightforward application of Equations 2.3.5 and

2.3.6. Note that you’re only asked for the y-coordinate of the centroid.
2.3.3.15. *. Hint. You can use a trigonometric substitution to find the area,
then a partial fraction decomposition to find the y-coordinate of the centroid.
Remember sin(1/2) = π/6.
2.3.3.16. *. Hint. Vertical slices will be easier than horizontal. An integration by
parts might be helpful to find x̄, while trigonometric identities are important to
finding ȳ.
2.3.3.17. *. Hint. No trigonometric substitution is necessary if you’re clever
with your u-substitutions, and remember the derivative of arctangent.
2.3.3.18. *. Hint. In R, the top function is x − x2 , and the bottom function is
x2 − 3x.
d 1
2.3.3.19. *. Hint. Remember dx
{arctan x} = 1+x2
.

586
H INTS FOR E XERCISES

2.3.3.20. *. Hint. You can save quite a bit of work by, firstly, exploiting symme-
try and, secondly, thinking about whether it is more efficient to use vertical strips
or horizontal strips.
√
2.3.3.21. *. Hint. Sketch the region, being careful the domain of 9 − 4x2 . You
can save quite a bit of work by exploiting symmetry.
2.3.3.22. Hint. Horizontal slices will be easier than vertical.
2.3.3.23. Hint. Start with a picture: whether you use vertical slices or horizon-
tal, you’ll need to break your integral into multiple pieces.

Exercises — Stage 3
2.3.3.24. *. Hint. For practice, do the computation twice — once with horizontal
strips and once with vertical strips. Watch for improper integrals.
2.3.3.25. *. Hint. Draw a sketch. In part (b) be careful about the equation of the
right hand boundary of A.
2.3.3.26. *. Hint. Draw a sketch. Rotating about a horizontal line is similar
to rotating about the x-axis, but for the radius of a slice, you’ll need to know
|y−(−1)|: the distance from the outer edge of the region (the boundary function’s
y-value) to y = −1.
2.3.3.27. Hint. Go back to the derivation of Equation 2.3.5 (centroid for a region)
to figure out what to do when your surface does not have uniform density. We
will consider a rod R that reaches from x = 0 to x = 4, and the mass of the section
of the rod along [a, b] is equal to the mass of the strip of our rectangle along [a, b].
2.3.3.28. Hint. Horizontal slices will help you, where symmetry doesn’t, to set
up a rod R whose centre of mass is the same as one coordinate of the centre
of mass of the circle. When you’re integrating, trigonometric substitutions are
sometimes the easiest way, and sometimes not.
The equation of a circle of radius 3, centred at (0, 3), is x2 + (y − 3)2 = 9.
2.3.3.29. Hint. The model in the question gives you the setup to solve this prob-
lem. You know how to find the centre of mass of a rod — that’s Equation 2.3.4
— so all you need to find is ρ(y), the density of the rod at position y. To find this,
consider a thin slice of the cone at position y with thickness dy. Its volume V (y)
is the same as the mass of the small section of the rod at position y with thickness
(y)
dy. So, the density of the rod at position y is ρ(y) = Vdy .

2.3.3.30. Hint. Use similar triangles to show that the shape of the lower (also
upper) half of the hourglass is a truncated cone, where the untruncated cone
would have had a height 10 cm.
To calculate the centre of mass of the upturned sand using the result of Ques-
tion 29, you should find h = 9.8 (not h = 10 — think carefully about our model
from Question 29) and k = 8.8. For the centre of mass of the sand before turning,
h = 10 and k = 6.

587
H INTS FOR E XERCISES

2.3.3.31. Hint. The techniques of Section 2.1 get pretty complicated here, so
it’s easiest to use the techniques we developed in Questions 6, 29, and 30 in this
section. That is, (1) find the height of the centre of mass of the water in its starting
and ending positions, and then (2) model the work done as the work moving
a point mass with the weight of the water from the first centre of mass to the
second.
The height change of the centre of mass is all that matters to calculate the work
done against gravity, so you only have to worry about the height of the centres of
mass.
2.3.3.32. Hint. The area of R is precisely√one, so the error in your approximation
R π/2 2
is the error involved in approximating 0 2x sin(x2 ) dx.

2.4 · Separable Differential Equations

2.4.7 · Exercises

Exercises — Stage 1
2.4.7.1. Hint. You don’t need to solve the differential equation from scratch,
dy
only verify whether the given function y = f (x) makes it true. Find dx and plug
it into the differential equation.
dy
2.4.7.2. Hint. For (d), note the equation given is quadratic in the variable dx .
1
Z Z
2.4.7.3. Hint. The step dy = f (x) dx shows up whether we’re using
g(y)
our mnemonic or not.
d d
2.4.7.4. Hint. Note dx {f (x)} = dx {f (x) + C}. Plug in y = f (x) + C to the
dy
equation dx = xy to see whether it makes the equation is true.
2.4.7.5. Hint. If a function is differentiable at a point, it is also continuous at
that point.
2.4.7.6. Hint. Let Q(t) be the quantity of morphine in a patient’s bloodstream at
time t, where t is measured in minutes.
Using the definition of a derivative,

dQ Q(t + h) − Q(t) Q(t + 1) − Q(t)

= lim ≈
dt h→0 h 1

So, dQ
dt
is roughly the change in the amount of morphine in one minute, from t to
t + 1.
2.4.7.7. Hint. If p(t) is the proportion of the new form, then 1 − p(t) is the pro-
portion of the old form.
When we say two quantities are proportional, we mean that one is a constant
multiple of the other.

588
H INTS FOR E XERCISES

2.4.7.8. Hint. The red marks show the slope y(x) would have at a point if it
crosses that point. So, pick a value of y(0); based on the red marks, you can see
how fast y(x) is increasing or decreasing at that point, which leads you roughly
to a value of y(1); again, the red marks tell you how fast y(x) is increasing or
decreasing, which leads you to a value of y(2), etc (unless you’re already off the
graph).
2.4.7.9. Hint. To draw the sketch similar to Question 8(d), don’t actually calcu-
late every single slope; find a few (for instance, where the slope is zero, or where
it’s negative), and use a pattern (for instance, the slope increases as y increases)
to approximate most of the points.

Exercises — Stage 2
2.4.7.10. *. Hint. Start by multiplying both sides of the equation by ey and dx,
dy
pretending that dx is a fraction, according to our mnemonic.
2.4.7.11. *. Hint. You need to solve for your function y(x) explicitly. Be careful
with absolute values: if |y| = F , then y = F or y = −F . However, y = ±F is not
a function. You have to choose one: y = F or y = −F .
2.4.7.12. *. Hint. If your answer doesn’t quite look like the answer given, try
manipulating it with logarithm rules: log a + log b = log(ab), and a log b = log(ba ).
2.4.7.13. *. Hint. Simplify the equation.
2.4.7.14. *. Hint. Be careful with the arbitrary constant.
2.4.7.15. *. Hint. Start by cross-multiplying.
√
√ Be careful about signs.
2.4.7.16. *. Hint. √ If y 2 = F , then possibly y = F , and
possibly y = − F . However, y = ± F is not a function.
2.4.7.17. *. Hint. Be careful about signs.

2.4.7.18. *. Hint. Be careful about signs. If log |y| = F , then |y| = eF . Since you
should give your answer as an explicit function y(x), you need to decide whether
y = eF or y = −eF .
2.4.7.19. *. Hint. Move the y from the left hand side to the right hand side, then
use partial fractions to integrate.
Be careful about the signs. Remember that we need y = −1 when x = 1. This
suggests how to deal with absolute values.
2.4.7.20. *. Hint. The unknown function f (x) satisfies an equation that involves
the derivative of f .
1
2.4.7.21. *. Hint. Try guessing the partial fractions expansion of .
x(x + 1)
Since x = 1 is in the domain and x = 0 is not, you may assume x > 0 for all x in
the domain.

589
H INTS FOR E XERCISES

d
2.4.7.22. *. Hint. {sec x} = sec x tan x
dx
2.4.7.23. *. Hint. The general solution to the differential equation will contain
the constant k and one other constant. They are determined by the data given in
the question.
2.4.7.24. *. Hint.

• When you’re solving the differential equation, you should have an integral
that you can massage to look something like arctangent.

• What is the velocity of the object at its highest point?

• Your final answer will depend on the (unspecified) constants v0 , m, g and

k.
2.4.7.25. *. Hint. The general solution to the differential equation will contain
the constant k and one other constant. They are determined by the data given in
the question.
2.4.7.26. *. Hint. The method of partial fractions will help you integrate.
To solve x−a
x−b
= Y for x, move the terms containing x out of the denominator, then
gather them on one side of the equals sign and factor out the x.
x−a
=Y
x−b
x − a = Y (x − b) = Y x − Y b
x−Yx=a−Yb
x(1 − Y ) = a − Y b
a−Yb
x=
1−Y
To find the limit, you can avoid l’Hôpital’s rule using some clever algebra–but
you can also just use l’Hôpital’s rule.
2.4.7.27. *. Hint. Be careful about signs.
Part (a) has some algebraic similarities to Question 26.
2.4.7.28. *. Hint. The general solution to the differential equation will contain a
constant of proportionality and one other constant. They are determined by the
data given in the question.

Exercises — Stage 3
2.4.7.29. *. Hint. You do not need to know anything about investing or contin-
uous compounding to do this problem. You are given the differential equation
explicitly. The whole first sentence is just window dressing.

590
H INTS FOR E XERCISES

2.4.7.30. *. Hint. Again, you do not need to know anything about investing to
do this problem. You are given the differential equation explicitly.
2.4.7.31. *. Hint. Differentiate the given integral equation. Plugging in x = 0
gives you y(0).
2.4.7.32. *. Hint. Suppose that in a very short time interval dt, the height of
water in the tank changes by dh (which is negative). Express in two different
ways the volume of water that has escaped during this time interval. Equating
the two gives the needed differential equation.
As the water escapes, it forms a cylinder of radius 1 cm.
2.4.7.33. *. Hint. Sketch the mercury in the tank at time t, when it has height h,
and also at time t + dt, when it has height h + dh (with dh < 0). The difference
between those two volumes is the volume of (essentially) a disk of thickness −dh.
Figure out the radius and then the volume of that disk. This volume has to be
the same as the volume of mercury that left through the hole in the bottom of the
sphere, which runs out in the shape of a cylinder. Toricelli’s law tells you what
the length of that cylinder is, and from there you can find its volume. Setting the
two volumes equal to each other gives the differential equation that determines
h(t).
2.4.7.34. *. Hint. The fundamental theorem of calculus will be useful in part
(b).
2.4.7.35. *. Hint. For any p > 0, determine first y(t) (in terms of p and c) and
then the times (also depending on p and c) at which y = 2, y = 1 and y = 0. The
condition that “the top half takes exactly the same amount of time to drain as the
bottom half” then gives an equation that determines p.
2.4.7.36. Hint. For (a), think of a very simple function.
The equation in the question statement is equivalent to the equation
Z x sZ
x
1
√ f (t) d(t) = f 2 (t) dt
x−a a a

which is, in some cases, easier to use.

Rx
For (d), you’ll want to let Y (x) = a f (t) dt, and use the quadratic equation.
2.4.7.37. Hint. Start by antidifferentiating both sides of the equation with re-
spect to x.

3 · Sequence and series

3.1 · Sequences
3.1.2 · Exercises

591
H INTS FOR E XERCISES

Exercises — Stage 1
3.1.2.1. Hint. Not every limit exists.
3.1.2.2. Hint. 100 isn’t all that big when you’re contemplating infinity. (Neither
is any other number.)
3.1.2.3. Hint. lim a2n+5 = lim an
n→∞ n→∞

3.1.2.4. Hint. The sequence might be defined by different functions when n is

large than when n is small.
3.1.2.5. Hint. Recall (−1)n is positive when n is even, and negative when n is
odd.
3.1.2.6. Hint. Modify your answer from Question 5, but make the terms ap-
proach zero.
(−1)n
3.1.2.7. Hint. (−n)−n =
nn
3.1.2.8. Hint. What might cause your answers in (a) and (b) to differ? Care-
fully read Theorem 3.1.6 about convergent functions and their corresponding
sequences.
22
3.1.2.9. Hint. You can use the fact that π is somewhat close to , or you can
7
use trial and error.

Exercises — Stage 2
3.1.2.10. Hint. You can compare the leading terms, or factor a high power of n
from the numerator and denominator.
3.1.2.11. Hint. This isn’t a rational expression, but you can treat it in a similar
way. Recall e < 3.
3.1.2.12. Hint. The techniques of evaluating limits of rational sequences are
again useful here.
3.1.2.13. Hint. Use the squeeze theorem.
1
3.1.2.14. Hint. ≤ nsin n ≤ n
n
1 1
3.1.2.15. Hint. e−1/n = ; what happens to as n grows?
e1/n n
3.1.2.16. Hint. Use the squeeze theorem.
3.1.2.17. Hint. L’Hôpital’s rule might help you decide what happens if you are
unsure.
3.1.2.18. *. Hint. Simplify ak .

592
H INTS FOR E XERCISES

1
3.1.2.19. *. Hint. What happens to as n gets very big?
n
3.1.2.20. *. Hint. cos 0 = 1

Exercises — Stage 3
1
3.1.2.21. *. Hint. This is trickier than it looks. Write = x and look at the limit
n
as x → 0.
3.1.2.22. Hint. Multiply and divide by the conjugate.
3.1.2.23. Hint. Compared to Question 22, there’s an easier path.
3.1.2.24. Hint. Consider f 0 (x), when f (x) = x100 .
3.1.2.25. Hint. Look to Question 24 for inspiration.
3.1.2.26. Hint. The area of an isosceles triangle with two sides of length 1, meet-
ing at an angle θ, is 12 sin θ.

1
sin θ

θ
1

3.1.2.27. Hint. Every term of An is the same, and g(x) is a constant function.
3.1.2.28. Hint. You’ll need to use a logarithm before you can apply l’Hôpital’s
rule.
3.1.2.29. Hint. (a) Write out the first few terms of the sequence.
(c) Consider how an+1 −L relates to an −L. What should happen to these numbers
if an converges to L?
3.1.2.30. Hint. Your answer from (b) will help you a lot with the subsequent
parts.

3.2 · Series
3.2.2 · Exercises

Exercises — Stage 1
3.2.2.1. Hint. SN is the sum of the terms corresponding to n = 1 through n = N .
3.2.2.2. Hint. Note Ck is the cumulative number of cookies.
3.2.2.3. Hint. How is (a) related to Question 2?

593
H INTS FOR E XERCISES

3.2.2.4. Hint. You’ll have to calculate a1 separately from the other terms.
3.2.2.5. Hint. When does adding a number decrease the total sum?

3.2.2.6. Hint. For (b), imagine cutting up the triangle into its black and white
parts, then sharing it equally among a certain number of friends. What is the
easiest number of friends to share with, making sure each has the same area in
their pile?
3.2.2.7. Hint. Compare to Question 6.
3.2.2.8. Hint. Iteratively divide a shape into thirds.
N
X 1 − rN +1
3.2.2.9. Hint. Lemma 3.2.5 tells us arn = a , for r 6= 1.
n=0
1−r

3.2.2.10. Hint. Note Ck is the cumulative number of cookies.

3.2.2.11. Hint. To adjust the starting index, either factor out the first term in the
series, or subtract two series. For the subtraction option, consider Question 10.
3.2.2.12. Hint. Express your gains in (a) and (c) as series.
∞
X ∞
X
3.2.2.13. Hint. To find the difference between cn and cn+1 , try writing
n=1 n=1
out the first few terms.
∞
X an ?
3.2.2.14. Hint. You might want to first consider a simpler true or false: =
n=1
bn
A
.
B

Exercises — Stage 2
3.2.2.15. *. Hint. What kind of a series is this?
3.2.2.16. *. Hint.This is a special kind of series, that you should recognize.
X 
3.2.2.17. *. Hint. When you see · · · k · · · − · · · k + 1 · · · , you should
k
think “telescoping series.”
X 
3.2.2.18. *. Hint. When you see · · · n · · · − · · · n + 1 · · · , you should
n
immediately think “telescoping series”. But be careful not to jump to conclusions
— evaluate the nth partial sum explicitly.
3.2.2.19. *. Hint. Review Definition 3.2.3.
3.2.2.20. *. Hint. This is a special case of a general series whose sum we know.

594
H INTS FOR E XERCISES

3.2.2.21. *. Hint. Review Example 3.2.6. To write the number as a geometric

series, the first few terms might not fit the pattern of the rest of the terms.
3.2.2.22. *. Hint. Start by writing it as a geometric series.
3.2.2.23. *. Hint. Review Example 3.2.6. Since the pattern repeats every three
1
decimals, your common ratio r will be 3 .
10
3.2.2.24. *. Hint. Split the series into two parts.
3.2.2.25. *. Hint. Split the series into two parts.
3.2.2.26. *. Hint. Split the series into two parts.
3.2.2.27. Hint. Use logarithm rules to turn this into a more obvious telescoping
series.
3.2.2.28. Hint. This is a telescoping series.

Exercises — Stage 3
1
3.2.2.29. Hint. The stone at position x has mass x kg, and we have to pull it a
4
distance of 2x metres. From this, you can find the work involved in pulling up a
single stone. Then, add up the work involved in pulling up all the stones.
4
3.2.2.30. Hint. The volume of a sphere of radius r is πr3 .
3
3.2.2.31. Hint. Use the properties of a telescoping series to simplify the terms.
Recall sin2 θ + cos2 θ = 1.
3.2.2.32. Hint. Review Question 3 for using the sequence of partial sums.
3.2.2.33. Hint. What is the ratio of areas between the outermost (red) ring and
the next (blue) ring?

3.3 · Convergence Tests

3.3.11 · Exercises

Exercises — Stage 1
3.3.11.1. Hint. That is, which series have terms whose limit is not zero?
3.3.11.2. Hint. That is, if f (x) is a function with f (n) = an for all whole numbers
n, is f (x) nonnegative and decreasing?
3.3.11.3. Hint. This isn’t a trick. It’s meant to give you intuition to the direct
comparison test.
3.3.11.4. Hint. The comparison test is Theorem 3.3.8. However, rather than
trying to memorize which way the inequalities go in all cases, you can use the
same reasoning as Question 3.

595
H INTS FOR E XERCISES

3.3.11.5. Hint. Think about Question 4 to remind yourself which way the in-
equalities have to go for direct comparison.
Note that all the comparison series have positive terms, so we don’t need to
worry about that part of the limit comparison test.
3.3.11.6. Hint. The divergence test is Theorem 3.3.1.
3.3.11.7. Hint. The limit is calculated correctly.
3.3.11.8. Hint. It is true that f (x) is positive. What else has to be true of f (x) for
the integral test to apply?
3.3.11.9. Hint. Refer to Question 4.
3.3.11.10. Hint. The definition of an alternating series is given in the start of
Section 3.3.4.
an+1
3.3.11.11. Hint. For the ratio test to be inconclusive, lim should be 1 or
n→∞ an
nonexistent.
P
3.3.11.12. Hint. By the divergence test, for a series an to converge, we need
lim an = 0. That is, the magnitude (absolute value) of the terms needs to be
n→∞
getting smaller.
3.3.11.13. Hint. If f (x) is positive and decreasing, then the integral test tells you
that the integral and the series either both increase or both decrease. So, in order
to find an example with the properties required in the question, you need f (x) to
not be both positive and decreasing.
3.3.11.14. *. Hint. Review Theorem 3.3.11 and Example 3.3.12.
3.3.11.15. *. Hint. Don’t jump to conclusions about properties of the an ’s.

Exercises — Stage 2
3.3.11.16. *. Hint. Always try the divergence test first (in your head).
3.3.11.17. *. Hint. Which test should you always try first (in your head)?
3.3.11.18. *. Hint. Review the integral test, which is Theorem 3.3.5.
3.3.11.19. Hint. A comparison might be helpful — try some algebraic manipu-
lation to find a likely series to compare it to.
3.3.11.20. Hint. This is a geometric series.
3.3.11.21. Hint. Notice that the series is geometric, but it doesn’t start at n = 0.
3.3.11.22. Hint. Note n only takes integer values: what’s sin(πn) when n is an
integer?
3.3.11.23. Hint. Note n only takes integer values: what’s cos(πn) when n is an
integer?

596
H INTS FOR E XERCISES

3.3.11.24. Hint. What’s the test that you should always think of when you see a
factorial?
3.3.11.25. Hint. This is a geometric series, but you’ll need to do a little algebra
to figure out r.
3.3.11.26. Hint. Which test fits most often with factorials?
3.3.11.27. Hint. Try finding a nice comparison.
1
3.3.11.28. *. Hint. With the substitution u = log x, the function is
x(log x)3/2
easily integrable.
3.3.11.29. *. Hint. Combine the integral test with the results about p-series,
Example 3.3.6.
√
3.3.11.30. *. Hint. Try the substitution u = x.
3.3.11.31. *. Hint. Review Example 3.3.9 for developing intuition about com-
parisons, and Example 3.3.10 for an example where finding an appropriate com-
parison series calls for some creativity.
3.3.11.32. *. Hint. What does the summand look like when k is very large?
3.3.11.33. *. Hint. What does the summand look like when n is very large?
3.3.11.34. *. Hint. cos(nπ) is a sneaky way to write (−1)n .
3.3.11.35. *. Hint. What is the behaviour for large k?
√
3.3.11.36. *. Hint. When m is large, 3m + sin m ≈ 3m.
3.3.11.37. Hint. This is a geometric series, but it doesn’t start at n = 0.
3.3.11.38. *. Hint. The series is geometric.
∞
X 1
3.3.11.39. *. Hint. The first series can be written as .
n=1
2n − 1

3.3.11.41. *. Hint. What does the summand look like when n is very large?
3.3.11.42. *. Hint. Review the alternating series test, which is given in Theorem
3.3.14.
3.3.11.43. *. Hint. Review the alternating series test, which is given in Theorem
3.3.14.
3.3.11.44. *. Hint. Review the alternating series test, which is given in Theorem
3.3.14.

Exercises — Stage 3

597
H INTS FOR E XERCISES

3.3.11.46. *. Hint. For part (a), see Example 1.12.23.

For part (b), review Theorem 3.3.5.
For part (c), see Example 3.3.12.
3.3.11.47. *. Hint. The truncation error arising from the approximation
∞ √ N √ ∞ √
X e− n X e− n X e− n
√ ≈ √ is precisely EN = √ . You’ll want to find a bound
n=1
n n=1
n n=N +1
n
on this sum using the integral test. √
e− x
A key observation is that, since f (x) = √ is decreasing, we can show that
x
√ √
n
e− n e− x
Z
√ ≤ √ dx
n n−1 x

for every n ≥ 1.
∞
P
3.3.11.48. *. Hint. What does the fact that the series an converges guarantee
n=0
about the behavior of an for large n?
∞
P
3.3.11.49. *. Hint. What does the fact that the series (1 − an ) converges guar-
n=0
antee about the behavior of an for large n?
∞
X nan − 2n + 1
3.3.11.50. *. Hint. What does the fact that the series converges
n=1
n+1
guarantee about the behavior of an for large n?
3.3.11.51. *. Hint. What does the fact that the series ∞
P
n=1 an converges guaran-
2
tee about the behavior of an for large n? When is x ≤ x?
3.3.11.52. Hint. If we add together the frequencies of all the words, they should
amount to 100%. We can approximate this sum using ideas from Example 3.3.4.
3.3.11.53. Hint. We are approximating a finite sum — not an infinite series. To
get greater accuracy, use exact values for the first several terms in the sum, and
use an integral to approximate the rest.

3.4 · Absolute and Conditional Convergence

3.4.3 · Exercises

Exercises — Stage 1
3.4.3.1. *. Hint. What is conditional convergence?
P P
3.4.3.2. Hint. If |an | converges, then an is guaranteed to converge as well.
(That’s Theorem 3.4.2.) So, one of the blank spaces describes an impossible se-
quence.

598
H INTS FOR E XERCISES

Exercises — Stage 2
3.4.3.4. *. Hint. Be careful about the signs.
3.4.3.5. *. Hint. Does the alternating series test really apply?
3.4.3.6. *. Hint. What does the summand look like when n is very large?
3.4.3.7. *. Hint. What does the summand look like when n is very large?
3.4.3.8. *. Hint. This is a trick question. Be sure to verify all of the hypotheses
of any convergence test you apply.
3.4.3.9. *. Hint. Try the substitution u = log x.
3.4.3.10. Hint. Show that it converges absolutely.
3.4.3.11. Hint. Use a similar method to Queston 10.
3.4.3.12. Hint. Show it converges absolutely using a direct comparison test.

Exercises — Stage 3
3.4.3.13. *. Hint. For part (a), replace n by x in the absolute value of the sum-
mand. Can you integrate the resulting function?
3.4.3.14. Hint. You don’t need to add up very many terms for this level of ac-
curacy.
3.4.3.15. Hint. Use the direct comparison test to show that the series converges
absolutely.

3.5 · Power Series

3.5.3 · Exercises

Exercises — Stage 1
3.5.3.1. Hint. f (1) is the sum of a geometric series.
d (x − 5)n
 
3.5.3.2. Hint. Calculate when n is a constant.
dx n! + 2
3.5.3.3. Hint. There is only one.
3.5.3.4. Hint. Use Theorem 3.5.9.

Exercises — Stage 2
3.5.3.5. *. Hint. Review the discussion immediately following Definition 3.5.1.
3.5.3.6. *. Hint. Review the discussion immediately following Definition 3.5.1.
3.5.3.7. *. Hint. Review the discussion immediately following Definition 3.5.1.
3.5.3.8. *. Hint. See Example 3.5.11.

599
H INTS FOR E XERCISES

3.5.3.9. *. Hint. See Example 3.5.11.

3.5.3.15. *. Hint. Start part (b) by computing the partial sums of
∞ 
X ak ak+1 
−
k=1
ak+1 ak+2

1
3.5.3.16. *. Hint. You should know a power series representation for . Use
1−x
it.
3.5.3.17. Hint. You can safely ignore one of the given equations, but not the
other.

Exercises — Stage 3
3.5.3.18. *. Hint. n ≥ log n for all n ≥ 1.
3.5.3.19. *. Hint. See Example 3.5.21. For part (b), review §3.3.4.
1
3.5.3.20. *. Hint. You know the geometric series expansion of 1−x . What (cal-
culus) operation(s) can you apply to that geometric series to convert it into the
given series?
3.5.3.21. *. Hint. First show that the fact that the series ∞
P
n=0 (1 − bn ) converges
guarantees that limn→∞ bn = 1.
3.5.3.22. *. Hint. What does an look like for large n?
3.5.3.23. Hint. Equation 2.3.1Ptells us the centre of mass of a rod with weights
m n xn
{mn } at positions {xn } is x̄ = P .
mn
3.5.3.24. Hint. Use the second derivative test.
∞
X
3.5.3.25. Hint. What function has nxn−1 as its power series representation?
n=1

3.5.3.26. Hint. The power series representation in Example 3.5.20 is an alternat-

ing series when x is positive.
3.5.3.27. Hint. The power series representation in Example 3.5.21 is an alternat-
ing series when x is nonzero.

3.6 · Taylor Series

3.6.7 · Exercises

Exercises — Stage 1
3.6.7.1. Hint. Which of the functions are constant, linear, and quadratic?
3.6.7.2. Hint. You don’t have to actually calculate the entire series T (x) to an-
swer the question.

600
H INTS FOR E XERCISES

3.6.7.3. Hint. If you don’t have these memorized, it’s good to be able to derive
1
them. For instance, log(1 + x) is the antiderivative of , whose Taylor series
P n 1+x
can be found by modifying the geometric series x .
3.6.7.4. Hint. See Example 3.6.14.

Exercises — Stage 2
3.6.7.5. Hint. The series will bear some resemblance to the Maclaurin series for
log(1 + x).

3.6.7.6. Hint. The terms f (n) (π) are going to be similar to the terms f (n) (0) that
we used in the Maclaurin series for sine.
3.6.7.7. Hint. The Taylor series will look similar to a geometric series.
3.6.7.8. Hint. Your answer will depend on a.

1
3.6.7.9. *. Hint. You should know the Maclaurin series for . Use it.
1−x
1
3.6.7.10. *. Hint. You should know the Maclaurin series for . Use it.
1−x
3.6.7.11. *. Hint. You should know the Maclaurin series for ex . Use it.
3.6.7.12. *. Hint. Review Example 3.5.20.
3.6.7.13. *. Hint. You should know the Maclaurin series for sin x. Use it.
3.6.7.14. *. Hint. You should know the Maclaurin series for ex . Use it.
3.6.7.15. *. Hint. You should know the Maclaurin series for arctan(x). Use it.
1
3.6.7.16. *. Hint. You should know the Maclaurin series for . Use it.
1−x

x2n+1 (−1)n
3.6.7.17. *. Hint. Set (−1)n =C , for some constant C. What
2n + 1 (2n + 1)3n
are x and C?
3.6.7.18. *. Hint. There is an important Taylor series, one of the series in Theo-
rem 3.6.7, that looks a lot like the given series.
3.6.7.19. *. Hint. There is an important Taylor series, one of the series in Theo-
rem 3.6.7, that looks a lot like the given series.
3.6.7.20. *. Hint. There is an important Taylor series, one of the series in The-
orem 3.6.7, that looks a lot like the given series. Be careful about the limits of
summation.

601
H INTS FOR E XERCISES

3.6.7.21. *. Hint. There is an important Taylor series, one of the series in Theo-
rem 3.6.7, that looks a lot like the given series.
3.6.7.22. *. Hint. Split the series into a sum of two series. There is an important
Taylor series, one of the series in Theorem 3.6.7, that looks a lot like each of the
two series.
3.6.7.23. Hint. Try the ratio test.
3.6.7.24. Hint. Write it as the sum of two Taylor series.
3.6.7.25. *. Hint. Can you think of a way to eliminate the odd terms from ex =
∞
X xn
?
n=0
n!

3.6.7.26. Hint. The series you’re adding up are alternating, so it’s simple to
bound the error using a partial sum.
3.6.7.27. Hint. The Taylor Series is alternating, so bounding the error in a
partial-sum approximation is straightforward.
3.6.7.28. Hint. The Taylor Series is not alternating, so use Theorem 3.6.3 to
bound the error in a partial-sum approximation.
3.6.7.29. Hint. The Taylor Series is not alternating, so use Theorem 3.6.3 to
bound the error in a partial-sum approximation.
3.6.7.30. Hint. Use Theorem 3.6.3 to bound the error in a partial-sum approxi-
mation. This theorem requires you to consider values of c between x and x = 0;
since x could be anything from −2 to 1, you should think about values of c be-
tween −2 and 1.
3.6.7.31. Hint. Use Theorem 3.6.3 to bound the error in a partial-sum approxi-
mation.
To bound the derivative over the appropriate range, remember how to find ab-
solute extrema.

Exercises — Stage 3
3.6.7.32. *. Hint. See Example 3.6.19.
3.6.7.33. *. Hint. See Example 3.6.19.
2/x
3.6.7.34. Hint. Set f (x) = (1 + x + x2 ) , and find lim log (f (x)).
x→0

1
3.6.7.35. Hint. Use the substitution y = , and compare to Question 34.
x
∞
X
3.6.7.36. Hint. Start by differentiating xn .
n=0

3.6.7.37. Hint. The series bears a resemblance to the Taylor series for arctangent.

602
H INTS FOR E XERCISES

(2n)!
3.6.7.38. Hint. For simplification purposes, note (1)(3)(5)(7) · · · (2n−1) = .
2n n!
3.6.7.39. *. Hint. You know the Maclaurin series for log(1 + y). Use it! Remem-
ber that you are asked for a series expansion in powers of x − 2. So you want y
to be some constant times x − 2.
3.6.7.40. *. Hint. See Example 3.5.21. For parts (b) and (c), review § 3.3.4.
3.6.7.41. *. Hint. Look at the signs of successive terms in the series.
3.6.7.42. *. Hint. The magic word is “series”.
3.6.7.43. *. Hint. See Example 3.6.12. For parts (b) and (c), review § 3.3.4.
3.6.7.44. *. Hint. See Example 3.6.12. For part (b), review the fundamental
theorem of calculus in § 1.3. For part (c), review § 3.3.4.
3.6.7.45. *. Hint. See Example 3.6.12. For parts (b) and (c), review § 3.3.4.
3.6.7.46. *. Hint. See Example 3.6.12. For parts (b) and (c), review § 3.3.4.
3.6.7.48. *. Hint. Use the Maclaurin series for ex .
3.6.7.49. *. Hint. For part (c), compare two power series term-by-term.
3.6.7.50. Hint. For Newton’s method, recall we approximate a root of the func-
tion g(x) in iterations: given an approximation xn , our next approximation is
g(xn )
xn+1 = xn − 0 .
g (xn )
To gauge your error, note that from approximation to approximation, the first
digits stabilize. Keep refining your approximation until the first two digits stop
changing.
3.6.7.51. Hint. First, modify your known Maclaurin series for arctangent into a
Maclaurin series for f (x). This series is not hard to repeatedly differentiate, so
use it to find a power series for f (10) (x).
3.6.7.52. Hint. Remember ex is never negative for any real number x.
3.6.7.53. Hint. Since f (x) is odd, f (−x) = −f (x) for all x in its domain. Con-
sider the even-indexed terms and odd-indexed terms of the Taylor series.

603
 Appendix C

A NSWERS TO E XERCISES

1 · Integration
1.1 · Definition of the Integral
1.1.8 · Exercises

Exercises — Stage 1
1.1.8.1. Answer. The area is between 1.5 and 2.5 square units.
1.1.8.2. Answer. The shaded area is between 2.75 and 4.25 square units. (Other
estimates are possible, but this is a reasonable estimate, using methods from this
chapter.)
1.1.8.3.
 h Answer.
i h Thei area under the curve is a number in the interval
3 1 √1 3 √1
8 2
+ 2
, 8
1 + 2
.

1.1.8.4. Answer. left

1.1.8.5. Answer. Many answers are possible. One example is f (x) = sin x,
[a, b] = [0, π], n = 1. Another example is f (x) = sin x, [a, b] = [0, 5π], n = 5.
1.1.8.6. Answer. Some of the possible answers are given, but more exist.
7
X 5
X
a i ; (i + 2)
i=3 i=1

7
X 5
X
b 2i ; (2i + 4)
i=3 i=1

7
X 5
X
c (2i + 1) ; (2i + 5)
i=3 i=1

604
A NSWERS TO E XERCISES

8
X 7
X
d (2i − 1) ; (2i + 1)
i=1 i=0

1.1.8.7. Answer. Some answers are below, but others are possible.
4 4  i
X 1 X 1
a i
and
i=1
3 i=1
3

4 4  i
X 2 X 1
b i
and 2
i=1
3 i=1
3

4 4
X 2 X 2
c (−1)i i and
i=1
3 i=1
(−3)i

4 4
i+1 2 2
X X
d (−1) i
and −
i=1
3 i=1
(−3)i

1.1.8.8. Answer.
5
X 2i − 1
a
i=1
3i

5
X 1
b
i=1
3i + 2

7 7
X
4−i
X i
c i · 10 and
i=1 i=1
10i−4

1.1.8.9. Answer.
"  101 #
5 3
a 1−
2 5
 50 "  51 #
5 3 3
b 1−
2 5 5

c 270

b
1 − 1e e
d + [b(b + 1)]2
e−1 4
1.1.8.10. Answer.
a 50 · 51 = 2550
2  2
b 21 (95)(96) − 21 (4)(5)


c −1

605
A NSWERS TO E XERCISES

d −10
1.1.8.11. Answer.

y
y = f (x)

x
a b

1.1.8.12. *. Answer. n = 4, a = 2, and b = 6

1.1.8.13. Answer. One answer is below, but other interpretations exist.
y

y = x2

121

81

49

x
5 7 9 11

1.1.8.14. Answer. Many interpretations are possible–see the solution to Ques-

tion 13 for a more thorough discussion–but the most obvious is given below.

606
A NSWERS TO E XERCISES

y y = tan x

x
π 2π 3π 4π 5π
20 20 20 20 20

1.1.8.15. *. Answer. Three answers are possible. It is a midpoint Riemann sum

for f on the interval [1, 5] with n = 4. It is also a left Riemann sum for f on the
interval [1.5, 5.5] with n = 4. It is also a right Riemann sum for f on the interval
[0.5, 4.5] with n = 4.
25
1.1.8.16. Answer.
2
21
1.1.8.17. Answer.
2

Exercises — Stage 2
50 
8
1 1 1
P
1.1.8.18. *. Answer. 5+ i− 2 5 5
i=1

1.1.8.19. *. Answer. 54
Z 7 n  
X 8i 8
1.1.8.20. *. Answer. f (x) dx = lim f −1 +
−1 n→∞
i=1
n n

1.1.8.21. *. Answer. f (x) = sin2 (2 + x) and b = 4

√
1.1.8.22. *. Answer. f (x) = x 1 − x2
R 3 −x/3
1.1.8.23. *. Answer. 0
e cos(x) dx
Z 1
1.1.8.24. *. Answer. xex dx
0

1.1.8.25. *. Answer. Possible answers include:

Z2
e−1−x dx
0

607
A NSWERS TO E XERCISES

Z3
e−x dx
1
Z 3/2
2 e−2x dx and
1/2
Z1
2 e−1−2x dx.
0

r3n+3 − 1
1.1.8.26. Answer.
r−1
 96 
5 r −1
1.1.8.27. Answer. r
r−1
1.1.8.28. *. Answer. 5
1.1.8.29. Answer. 16
b 2 − a2
1.1.8.30. Answer.
2
b 2 − a2
1.1.8.31. Answer.
2
1.1.8.32. Answer. 4π
Z 3
1.1.8.33. *. Answer. f (x) dx = 2.5
0

1.1.8.34. *. Answer. 53 m
1.1.8.35. Answer. true
1.1.8.36. Answer. 3200 km

Exercises — Stage 3
1.1.8.37. *. Answer. (a) There are many possible answers. Two are
R0 √ R2p
−2
4 − x dx and 0 4 − (−2 + x)2 dx.
2

(b) π
1.1.8.38. *. Answer. (a) 30
(b) 41 14
56
1.1.8.39. *. Answer.
3
1.1.8.40. *. Answer. 6
1.1.8.41. *. Answer. 12
 
3 x 2 2x
1.1.8.42. Answer. f (x) = + 8 sin +2
10 5 5

608
A NSWERS TO E XERCISES

1
1.1.8.43. Answer.
log 2
1

1.1.8.44. Answer. (a) 10b − 10a
log 10
1

(b) cb − ca ; yes, it agrees.
log c
√
1.1.8.45. Answer. π4 − 21 arccos(a) + 12 a 1 − a2
1.1.8.46. Answer.
b−a
a [f (b) − f (a)] ·
n
b Choose n to be an integer that is greater than or equal to
100 [f (b) − f (a)] (b − a).
1.1.8.47. Answer. true (but note, for a non-linear function, it is possible that the
midpoint Riemann sum is not the average of the other two)

1.2 · Basic properties of the definite integral

1.2.3 · Exercises

Exercises — Stage 1
1.2.3.1. Answer. Possible drawings:

y y y
y = f (x) + g(x)
y = f (x) y = f (x)
y = f (x)

x
a b
x x
a a c b

1.2.3.2. Answer. sin b − sin a

1.2.3.3. *. Answer. (a) False. For example, the function
(
0 for x < 0
f (x) =
1 for x ≥ 0
provides a counterexample.
(b) False. For example, the function f (x) = x provides a counterexample.
(c) False. For example, the functions
( (
0 for x < 12 0 for x ≥ 12
f (x) = and g(x) =
1 for x ≥ 12 1 for x < 12
provide a counterexample.

609
A NSWERS TO E XERCISES

1
1.2.3.4. Answer. (a) − , (b) positive, (c) negative, (d) positive.
20

Exercises — Stage 2
1.2.3.5. *. Answer. −21
1.2.3.6. *. Answer. −6
1.2.3.7. *. Answer. 20
1.2.3.8. Answer.
√ √
a π
4
− 12
arccos(−a) − 12 a 1−a2 = − π4 + 12 arccos(a) − 12 a 1−a2
√
b 12 arccos(a) − 12 a 1 − a2
1.2.3.9. *. Answer. 5
1.2.3.10. Answer. 0
1.2.3.11. Answer. 5

Exercises — Stage 3
1.2.3.12. *. Answer. 20 + 2π
1.2.3.13. *. Answer. 0
1.2.3.14. *. Answer. 0
1.2.3.15. Answer. 0
1p
1.2.3.16. Answer. (a) y = 1 − (ax)2
Z 1 r b
a a 1
(b) − x2 dx
b − a1 a2
π
(c)
ab
1.2.3.17. Answer.

× even odd
even even odd
odd odd even

1.2.3.18. Answer. f (0) = 0; g(0) can be any real number

1.2.3.19. Answer. f (x) = 0 for every x
1.2.3.20. Answer. The derivative of an even function is odd, and the derivative
of an odd function is even.

1.3 · The Fundamental Theorem of Calculus

1.3.2 · Exercises
610
A NSWERS TO E XERCISES

Exercises — Stage 1
1.3.2.1. *. Answer. e2 − e−2
x4 1 1
1.3.2.2. *. Answer. F (x) = + cos 2x + .
4 2 2
1.3.2.3. *. Answer. (a) True
(b) False Rb Rb
(c) False, unless a f (x) dx = a xf (x) dx = 0.
1.3.2.4. Answer. false
1.3.2.5. Answer. false
1.3.2.6. Answer. sin(x2 )
√
1.3.2.7. Answer. 3
e
1.3.2.8. Answer. For any constant C, F (x) + C is an antiderivative of f (x). So,
for example, F (x) and F (x) + 1 are both antiderivatives of f (x).
1.3.2.9. Answer.
d −1
a We differentiate with respect to a. Recall dx {arccos x} = √1−x 2 . To differen-
1
√
2
tiate 2 a 1 − a , we use the product and chain rules.

1 √
 
d π 1 2
− arccos(a) + a 1 − a
da 4 2 2
1√
 
1 −1 1 −2a
=0− · √ + a · √ + 1 − a2
2 1 − a2 2 2 1 − a2 2
1 a2 1 − a2
= √ − √ + √
2 1 − a2 2 1 − a2 2 1 − a2
1 − a2 + 1 − a2
= √
2 1 − a2
2(1 − a2 )
= √
2 1 − a2
√
= 1 − a2

5π 1 1 √
b F (x) = − arccos(x) + x 1 − x2
4 2 2
1.3.2.10. Answer. (a) 0
(b),(c) The FTC does not apply, because the integrand is not continuous over the
interval of integration.
1.3.2.11. Answer.

611
A NSWERS TO E XERCISES

y = f (t)

t
a x x+h

1.3.2.12. Answer. (a) zero

(b) increasing when 0 < x < 1 and 3 < x < 4; decreasing when 1 < x < 3
1.3.2.13. Answer. (a) zero
(b) G(x) is increasing when 1 < x < 3, and it is decreasing when 0 < x < 1 and
when 3 < x < 4.
1.3.2.14. Answer. Using the definition of the derivative,
F (x + h) − F (x)
F 0 (x) = lim
h→0 h
R x+h Rx
a
tdt − a tdt
= lim
h→0 h
R x+h
tdt
= lim x
h→0 h
The numerator describes the area of a trapezoid with base h and heights x and
x + h.
1
2
h(x
+ x + h)
= lim
h→0
 h
1
= lim x + h
h→0 2
=x

y
y=t
x+h
R x+h
x x
tdt

t
x x+h

So, F 0 (x) = x.

612
A NSWERS TO E XERCISES

1.3.2.15. Answer. f (t) = 0

R
1.3.2.16. Answer. log(ax)dx = x log(ax) − x + C, where a is a given constant,
and C is any constant.
R 3 x
1.3.2.17. Answer. x e dx = ex (x3 − 3x2 + 6x − 6) + C
Z
1 √
1.3.2.18. Answer. √ dx = log x + x2 + a2 + C when a is a given
x 2 + a2
constant. As usual, C is an arbitrary constant.
x √ √
Z p

1.3.2.19. Answer. p dx = x(a + x) − a log x + a + x + C
x(a + x)

Exercises — Stage 2
1.3.2.20. *. Answer. 5 − cos 2
1.3.2.21. *. Answer. 2
1
1.3.2.22. Answer. arctan(5x) + C
5
 
x
1.3.2.23. Answer. arcsin √ +C
2
1.3.2.24. Answer. tan x − x + C
3 3
1.3.2.25. Answer. − cos(2x) + C, or equivalently, sin2 x + C
4 2
1 1
1.3.2.26. Answer. x + sin(2x) + C
2 4
0 π


1.3.2.27. *. Answer. F 2
= log(3)
0 π


G 2 = − log(3)

1.3.2.28. *. Answer. f (x) is increasing when −∞ < x < 1 and when 2 < x < ∞.
sin x
1.3.2.29. *. Answer. F 0 (x) = −
cos3 x + 6
4 )2
1.3.2.30. *. Answer. 4x3 e(1+x
1.3.2.31. *. Answer. sin6 x + 8) cos x

1.3.2.32. *. Answer. F 0 (1) = 3e−1

sin u
1.3.2.33. *. Answer.
1 + cos3 u
1.3.2.34. *. Answer. f (x) = 2x
1.3.2.35. *. Answer. f (4) = 4π

613
A NSWERS TO E XERCISES
2
1.3.2.36. *. Answer. (a) (2x + 1)e−x
(b) x = −1/2
4 −x3 )
esin x − esin(x 4x3 − 3x2


1.3.2.37. *. Answer.
2
5

1.3.2.38. *. Answer. −2x cos e−x − 5x4 cos ex
p p
1.3.2.39. *. Answer. ex sin(ex ) − sin(x)
1.3.2.40. *. Answer. 14

Exercises — Stage 3
5
1.3.2.41. *. Answer.
2
1.3.2.42. *. Answer. 45 m
2

1.3.2.43. *. Answer. f 0 (x) = (2 − 2x) log 1 + e2x−x and f (x) achieves its abso-
lute maximum at x = 1, because f (x) is increasing for x < 1 and decreasing for
x > 1.
R −1 dt
1.3.2.44. *. Answer. The minimum is 0 1+t 4 . As x runs from −∞ to ∞, the
R x2 −2x dt
function f (x) = 0 1+t4
decreases until x reaches 1 and then increases all
x > 1. So the minimum is achieved for x = 1. At x = 1, x2 − 2x = −1.
1.3.2.45. *. Answer. F achieves its maximum value at x = π.
1.3.2.46. *. Answer. 2
1.3.2.47. *. Answer. log 2
1.3.2.48. Answer. In the sketch below, open dots denote inflection points, and
closed dots denote extrema.
y

y = F (x)

y = f (x)
x
−5 −3 −1 1 3 5

614
A NSWERS TO E XERCISES
Z x3 +1
3 3 +1)3
1.3.2.49. *. Answer. (a) 3x 2
et dt + 3x5 e(x
0
(b) y = −3(x + 1)
1.3.2.50. Answer. Both students.
1.3.2.51. Answer. (a) 27(1 − cos 3)
(b) x3 sin(x) + 3x2 [1 − cos(x)]
1.3.2.52. Answer. If f (x) = 0 for all x, then F (x) is even and possibly also odd.
If f (x) 6= 0 for some x, then F (x) is not even. It might be odd, and it might be
neither even nor odd.
(Perhaps surprisingly, every antiderivative of an odd function is even.)

1.4 · Substitution
1.4.2 · Exercises

Exercises — Stage 1
1.4.2.1. Answer. (a) true
(b) false
1.4.2.2. Answer. The reasoning is not sound: when we do a substitution, we
need to take care of the differential (dx). Remember the method of substitution
comes from the chain rule: there should be a function and its derivative. Here’s
the way to do it:
Z
Problem: Evaluate (2x + 1)2 dx.

Work: We use the substitution u = 2x+1. Then du = 2dx, so dx = 12 du:

1
Z Z
(2x + 1) dx = u2 · du
2
2
1 3
= u +C
6
1
= (2x + 1)3 + C
6
1.4.2.3. Answer. The problem is with the limits of integration, as in Question 1.
Here’s how it ought to go:
Z π
cos(log t)
Problem: Evaluate dt.
1 t
Work: We use the substitution u = log t, so du = 1t dt. When t = 1, we
have u = log 1 = 0 and when t = π, we have u = log(π). Then:
Z π Z log(π)
cos(log t)
dt = cos(u)du
1 t log 1
Z log(π)
= cos(u)du
0

615
A NSWERS TO E XERCISES

= sin(log(π)) − sin(0) = sin(log(π)).

1.4.2.4. Answer. This one is OK.
Z 1
f (u) √
1.4.2.5. *. Answer. √ du. Because the denominator 1 − u2 vanishes
0 1 − u2
when u = 1, this is what is known as an improper integral. Improper integrals
will be discussed in Section 1.12.
1.4.2.6. Answer. some constant C

Exercises — Stage 2
1

1.4.2.7. *. Answer. sin(e) − sin(1)
2
1
1.4.2.8. *. Answer.
3
1
1.4.2.9. *. Answer. − +C
300(x3 + 1)100
1.4.2.10. *. Answer. log 4
1.4.2.11. *. Answer. log 2
4
1.4.2.12. *. Answer.
3
1.4.2.13. *. Answer. e6 − 1
1
1.4.2.14. *. Answer. (4 − x2 )3/2 + C
3
√
log x
1.4.2.15. Answer. e +C

Exercises — Stage 3
1.4.2.16. *. Answer. 0
1
1.4.2.17. *. Answer. [cos 1 − cos 2] ≈ 0.478
2
1 1
1.4.2.18. Answer. − log 2
2 2
1
1.4.2.19. Answer. 2
tan2 θ − log | sec θ| + C
1.4.2.20. Answer. arctan(ex ) + C
π 2
1.4.2.21. Answer. −
4 3
1.4.2.22. Answer. − 12 (log(cos x))2 + C
1
1.4.2.23. *. Answer. 2
sin(1)

616
A NSWERS TO E XERCISES

1 √
1.4.2.24. *. Answer. [2 2 − 1] ≈ 0.609
3
1.4.2.25. Answer. Using the definition of a definite integral with right Riemann
sums:
Z b n
X b−a
2f (2x)dx = lim ∆x · 2f (2(a + i∆x)) ∆x =
a n→∞
i=1
n
n      
X b−a b−a
= lim · 2f 2 a + i
n→∞
i=1
n n
n 
X 2b − 2a    
2b − 2a
= lim · f 2a + i
n→∞
i=1
n n
Z 2b n
X 2b − 2a
f (x)dx = lim ∆x · f (2a + i∆x) ∆x =
2a n→∞
i=1
n
n     
X 2b − 2a 2b − 2a
= lim · f 2a + i
n→∞
i=1
n n

Since the Riemann sums are exactly the same,

Z b Z 2b
2f (2x)dx = f (x)dx
a 2a

1.5 · Area between curves

1.5.2 · Exercises

Exercises — Stage 1 √

π
1.5.2.1. Answer. Area between curves ≈ 4
2+ 2

x
π π 3π π
4 2 4
y = sin x
y = cos x

1.5.2.2. Answer. (a) Vertical rectangles:

617
A NSWERS TO E XERCISES

y
p xπ
y= 2

2x


y = arcsin π

x
π π 3π 2π π
10 5 10 5 2

(b) One possible answer:

π
2

2π x = π2 y 2
5

3π
10

π
5

π
10 π
x= 2
sin y

√
Z 2
2x − x3 dx
 
1.5.2.3. *. Answer.
0
4  
4
Z
2
1.5.2.4. *. Answer. (6 − y ) + 2y dy
−3/2 5
Z 4a √
x2

1.5.2.5. *. Answer. 4ax − dx
0 4a

618
A NSWERS TO E XERCISES

25
1√
 
1
Z
1.5.2.6. *. Answer. − (x + 5) + x dx
1 12 2

Exercises — Stage 2
1
1.5.2.7. *. Answer.
8
4
1.5.2.8. *. Answer.
3
5 1
1.5.2.9. *. Answer. −
3 log 2
8
1.5.2.10. *. Answer. −1
π
20
1.5.2.11. *. Answer.
9
1
1.5.2.12. *. Answer.
6
1.5.2.13. Answer. 2π

Exercises — Stage 3 h i
1.5.2.14. *. Answer. 2 π − 14 π 2

31
1.5.2.15. *. Answer.
6
26
1.5.2.16. *. Answer.
3
7π 1
1.5.2.17. Answer. −
8 2
√ 13
1.5.2.18. Answer. 12 2 −
4

1.6 · Volumes
1.6.2 · Exercises

Exercises — Stage 1
1.6.2.1. Answer. The horizontal cross-sections are circles, but the vertical cross-
sections are not.
1.6.2.2. Answer. The columns have the same volume.
1.6.2.3. Answer.

• Washers when 1 < y ≤ 6: If y > 1, then our washer has inner radius 2 + 23 y,

619
A NSWERS TO E XERCISES

outer radius 6 − 32 y, and height dy.

r = 2 + 23 y

thickness: dy

R = 6 − 32 y

• Washers when 0 ≤ y < 1: When 0 ≤ y < 1, we have a “double washer,”

two concentric rings. The inner washer has inner radius r1 = y and outer
radius R1 = 2 − y. The outer washer has inner radius r2 = 2 + 23 y and outer
radius R2 = 6 − 23 y. The thickness of the washers is dy.

r2 = 2 + 23 y

r1 = y

thickness: dy

R1 = 2 − y

R2 = 6 − 23 y

620
A NSWERS TO E XERCISES
Z 3
2
1.6.2.4. *. Answer. (a) π xe2x dx
Z 1 0
4
√
2 √
2  √
2 
Z
 
2
(b) π 3 + y − 3 − y dy + π 5 − y − 3 − y dy
0 1
1  Z
2 2
1.6.2.5. *. Answer. (a) π (5 − 4x2 ) − (2 − x2 ) dx
Z 0 −1
 p
2 p
2 
(b) π 5 + y + 1 − 5 − y + 1 dy
−1
Z 2
2 2
(9 − x2 ) − (x2 + 1) dx

1.6.2.6. *. Answer. π
−2
√
2 3
1.6.2.7. Answer. `
12

Exercises — Stage 2
π  2a2 
1.6.2.8. *. Answer. e −1
4
 
38 514 512
1.6.2.9. *. Answer. π − 4 =π
3 3 81
R1 √
1.6.2.10. *. Answer. (a) 8π −1 1 − x2 dx
(b) 4π 2
1.6.2.11. *. Answer. (a) The region R is the region between the blue and red
curves, with 3 ≤ x ≤ 5, in the figures below.

(b) 43 π ≈ 4.19
1.6.2.12. *. Answer. (a) The region R is sketched below.

621
A NSWERS TO E XERCISES

h i
(b) π 4 log 2 − 32 ≈ 3.998

8π 6
1.6.2.13. *. Answer. π 2 + 8π 3 + 5

8
1.6.2.14. *. Answer.
3
256 × 8
1.6.2.15. *. Answer. = 136.53̇
15
28
1.6.2.16. *. Answer. πh
3

Exercises — Stage 3
1.6.2.17. Answer.
4π
• (a) cubic units,
3b2 a
1 1
• (b) a = and b = ,
6356.752 6378.137
• (c) Approximately 1.08321 × 1012 km3 , or 1.08321 × 1021 m3 ,

• (d) Absolute error is about 3.64×109 km3 , and relative error is about 0.00336,
or 0.336%.
Z 2
9
2
2 
4 − x − 1 + (x − 1)2 dx

1.6.2.18. *. Answer. (a) (b) π
2 −1

π π2
1.6.2.19. *. Answer. (a) − 1 (b) − π ≈ 1.793
2 2
4 π c √ √ 1
(a) V1 = πc2 (b) V2 =

1.6.2.20. *. Answer. 4 2 − 2 (c) c = 0 or c = 2 − 2
3 3
1.6.2.21. *. Answer.
Z π
π (5 + π sin x)2 − (5 + 2π − 2x)2 dx
 
π/2
Z 3π/2
π (5 + 2π − 2x)2 − (5 + π sin x)2 dx
 
+
π
 
6000cπ 1 6000cπ
1.6.2.22. Answer. (a) 1 − 10 , which is close to .
log 2 2 log 2
(b) 6km: that is, there is roughly the same mass of air in the lowest 6 km of the

622
A NSWERS TO E XERCISES

column as there is in the remaining 54 km.

1.7 · Integration by parts

1.7.2 · Exercises

Exercises — Stage 1
1.7.2.1. Answer. chain; product
1.7.2.2. Answer. The part chosen as u will be differentiated. The part chosen as
dv will be antidifferentiated.
Z 0
f (x) f (x) f (x)g 0 (x)
Z
1.7.2.3. Answer. dx = + dx + C
g(x) g(x) g 2 (x)
1.7.2.4. Answer. All the antiderivatives differ only by a constant, so we can
write them all as v(x) + C for some C. Then, using the formula for integration by
parts,
Z Z
0
v(x) + C u0 (x)dx
   
u(x) · v (x)dx = u(x) v(x) + C −
|{z} | {z } | {z } | {z }
u v v du
Z Z
= u(x)v(x) + Cu(x) − v(x)u (x)dx − Cu0 (x)dx
0

Z
= u(x)v(x) + Cu(x) − v(x)u0 (x)dx − Cu(x) + D
Z
= u(x)v(x) − v(x)u0 (x)dx + D

where D is any constant.

Since the terms with C cancel out, it didn’t matter what we chose for C–all
choices end up the same.
Z
1.7.2.5. Answer. Suppose we choose dv = f (x)dx, u = 1. Then v = f (x)dx,
and du = dx. So, our integral becomes:
Z Z Z Z 
(1) f (x)dx = (1) f (x)dx − f (x)dx |{z}
dx
|{z} | {z } |{z}
u dv u | {z } | {z } du
v v

In order to figure out the first product (and the second integrand), you need to
know the antiderivative of f (x)–but that’s exactly what you’re trying to figure
out!

Exercises — Stage 2
x2 log x x2
1.7.2.6. *. Answer. − +C
2 4

623
A NSWERS TO E XERCISES

log x 1
1.7.2.7. *. Answer. − 6 − +C
6x 36x6
1.7.2.8. *. Answer. π
π
1.7.2.9. *. Answer. −1
2
1.7.2.10. Answer. ex (x3 − 3x2 + 6x − 6) + C
x2 3x2 3x2 3x2
1.7.2.11. Answer. log3 x − log2 x + log x − +C
2 4 4 8
1.7.2.12. Answer. (2 − x2 ) cos x + 2x sin x + C

t3 − 25 t2 + 6t log t − 13 t3 + 54 t2 − 6t + C


1.7.2.13. Answer.
√ √
1.7.2.14. Answer. e s (2s − 4 s + 4) + C

1.7.2.15. Answer. x log2 x − 2x log x + 2x + C

2 +1
1.7.2.16. Answer. ex +C
p
1.7.2.17. *. Answer. y arccos y − 1 − y2 + C

Exercises — Stage 3
1.7.2.18. *. Answer. 2y 2 arctan(2y) − y + 12 arctan(2y) + C

x3 1 1
1.7.2.19. Answer. arctan x − (1 + x2 ) + log(1 + x2 ) + C
3 6 6
2 x/2 8
1.7.2.20. Answer. e cos(2x) + ex/2 sin(2x) + C
17 17
x  
1.7.2.21. Answer. sin(log x) − cos(log x) + C
2
2x
 
1
1.7.2.22. Answer. x− +C
log 2 log 2
1.7.2.23. Answer. 2ecos x [1 − cos x] + C
xe−x −x e−x
1.7.2.24. Answer. +e +C = +C
1−x 1−x
1.7.2.25. *. Answer. (a) We integrate by parts with u = sinn−1 x and dv =
sin x dx, so that du = (n − 1) sinn−2 x cos x and v = − cos x.
Z Z
n
sin x dx = − sin n−1
x cos x + (n − 1) cos2 x sinn−2 x dx
| {z }
uv | {z }
R
− vdu

624
A NSWERS TO E XERCISES

Using the identity sin2 x + cos2 x = 1,

Z
= − sin n−1
(1 − sin2 x) sinn−2 x dx
x cos x + (n − 1)
Z Z
n−1 n−2
= − sin x cos x + (n − 1) sin x dx − (n − 1) sinn x dx

Moving the last term on the right hand side to the left hand side gives
Z Z
n n−1
n sin x dx = − sin x cos x + (n − 1) sinn−2 x dx

Dividing across by n gives the desired reduction formula.

35
(b) π ≈ 0.4295
256
π log 2
1.7.2.26. *. Answer. (a) Area: −
4 2

π2
(b) Volume: −π
2
17e18 − 4373
 
1.7.2.27. *. Answer. π
36
1.7.2.28. *. Answer. 12
2
1.7.2.29. Answer.
e
1.8 · Trigonometric Integrals
1.8.4 · Exercises

Exercises — Stage 1
1.8.4.1. Answer. (e)
1
1.8.4.2. Answer. secn x + C
n
1.8.4.3. Answer. We divide both sides by cos2 x, and simplify.
sin2 x + cos2 x = 1
sin2 x + cos2 x 1
=
cos2 x cos2 x

625
A NSWERS TO E XERCISES

sin2 x
2
+ 1 = sec2 x
cos x
tan2 x + 1 = sec2 x

Exercises — Stage 2
sin3 x
1.8.4.4. *. Answer. sin x − +C
3
π
1.8.4.5. *. Answer.
2
sin37 t sin39 t
1.8.4.6. *. Answer. − +C
37 39
1 1
1.8.4.7. Answer. 3
− +C
3 cos x cos x
√
π 9 3
1.8.4.8. Answer. −
8 64
2 1
1.8.4.9. Answer. − cos x + cos3 x − cos5 x + C
3 5
1
1.8.4.10. Answer. sin2.2 x + C
2.2
1 1
1.8.4.11. Answer. tan2 x + C, or equivalently, sec2 +C
2 2
1 1
1.8.4.12. *. Answer. sec7 x − sec5 x + C
7 5
tan49 x tan47 x
1.8.4.13. *. Answer. + +C
49 47
1 1
1.8.4.14. Answer. sec3.5 x − sec1.5 x + C
3.5 1.5
1 1
1.8.4.15. Answer. sec4 x − sec2 x + C
4 2
1
1.8.4.16. Answer. tan5 x + C
5
1 1
1.8.4.17. Answer. sec1.3 x + cos0.7 x + C
1.3 0.7
1
1.8.4.18. Answer. = sec4 x − sec2 x + log | sec x| + C
4
41 π
1.8.4.19. Answer. √ −
45 3 6

626
A NSWERS TO E XERCISES

1 1
1.8.4.20. Answer. +
11 9
√
1.8.4.21. Answer. 2 sec x + C
tan6 θ 3 tan4 θ 3 tan2 θ
 
e+1 1
1.8.4.22. Answer. tan θ + + + +C
7+e 5+e 3+e 1+e

Exercises — Stage 3
1.8.4.23. *. Answer. (a) Using the trig identity tan2 x = sec2 x − 1 and the sub-
stitution y = tan x, dy = sec2 x dx,
Z Z
tan xdx = tann−2 x tan2 xdx
n

Z Z
n−2 2
= tan x sec xdx − tann−2 xdx
Z Z
n−2
= y dy − tann−2 xdx
y n−1
Z
= − tann−2 xdx
n−1
tann−1 x
Z
= − tann−2 xdx
n−1
13 π
(b) − ≈ 0.0813
15 4
1 1
1.8.4.24. Answer. 2
+ 2 log | cos x| − cos2 x + C
2 cos x 2
1.8.4.25. Answer. tan θ + C
1.8.4.26. Answer. log | sin x| + C
1 2 x
1.8.4.27. Answer. sin (e ) + C
2
1.8.4.28. Answer. (sin2 x + 2) cos(cos x) + 2 cos x sin(cos x) + C
x 2 x 1
1.8.4.29. Answer. sin x − + sin x cos x + C
2 4 4

1.9 · Trigonometric Substitution

1.9.2 · Exercises

Exercises — Stage 1

627
A NSWERS TO E XERCISES

4
1.9.2.1. *. Answer. (a) x = sec θ
3
1
(b) x = sin θ
2
(c) x = 5 tan θ
√
1.9.2.2. Answer.
√ (a) x − 2 = 3 sec u
x − 1 =  5 sin
(b)  √ u
3 31
(c) 2x + = tan u
2 2
1 1
(d) x − = sec u
2 2
√
399
1.9.2.3. Answer. (a)
√ 20
5 2
(b)
√7
x−5
(c)
2
√
4 − x2
1.9.2.4. Answer. (a)
2
1
(b)
2
1
(c) √
1−x

Exercises — Stage 2
1 x
1.9.2.5. *. Answer. ·√ +C
4 x2 + 4
1
1.9.2.6. *. Answer. √
2 5
π
1.9.2.7. *. Answer.
6
r
x2 x
1.9.2.8. *. Answer. log 1+ + +C
25 5

1√ 2
1.9.2.9. Answer. 2x + 4x + C
2
√
1 x2 + 16
1.9.2.10. *. Answer. − +C
16 x
√
x2 − 9
1.9.2.11. *. Answer. +C
9x

628
A NSWERS TO E XERCISES

1.9.2.12. *. Answer. (a) We’ll use the trig identity cos 2θ = 2 cos2 θ − 1. It implies
that
cos 2θ + 1 1 2
cos2 θ = =⇒ cos4 θ =

cos 2θ + 2 cos 2θ + 1
2 4
1 h cos 4θ + 1 i
= + 2 cos 2θ + 1
4 2
cos 4θ cos 2θ 3
= + +
8 2 8
So,
π/4 π/4  cos 4θ
cos 2θ 3 
Z Z
4
cos θdθ = + + dθ
0 0 8 2 8
 π/4
sin 4θ sin 2θ 3
= + + θ
32 4 8 0
1 3 π
= + ·
4 8 4
8 + 3π
=
32
as required.
8 + 3π
(b)
16
1.9.2.13. Answer. 0
x x√
1.9.2.14. *. Answer. 2 arcsin + 4 − x2 + C
2 2
√
1.9.2.15. *. Answer. 25x2 − 4 − 2 arcsec 5x
2
+C
40
1.9.2.16. Answer.
3
x+1
1.9.2.17. *. Answer. arcsin +C
2
   √ 2 
1 1 4x − 12x + 8
1.9.2.18. Answer. arccos + + C, or equivalently,
4√ 2x − 3 (2x − 3)2
4x2 − 12x + 8
 
1
arcsec (2x − 3) + +C
4 (2x − 3)2
√ 1
1.9.2.19. Answer. log(1 + 2) − √
2
 
1 x
1.9.2.20. Answer. arctan x + 2
+C
2 x +1

Exercises — Stage 3

629
A NSWERS TO E XERCISES

3 + x√ 2 1 √
1.9.2.21. Answer. x − 2x + 2 + log x2 − 2x + 2 + x − 1 + C
2 2
2√ 2
 
1 6
1.9.2.22. Answer. √ log x+1 + 9x + 15x + C
3 5 5
√
1√ 1 − 1 + x2
1.9.2.23. Answer. 1 + x2 (4 + x2 ) + log +C
3 x

8π √
1.9.2.24. Answer. +4 3
3
r
4 4 4
1.9.2.25. Answer. Area: −
√ 3 3
π2 3π
Volume: −
6 4
√ √
1.9.2.26. Answer. 2 1 + ex + 2 log 1 − 1 + ex − x + C
1.9.2.27. Answer.
1
a
1 − x2
b False

c The work in the question is not correct. The most salient problem is that
when we make the substitution x = sin θ, we restrict the possible values of
x to [−1, 1], since this is the range of the sine function. However, the original
integral had no such restriction.
How can we be sure we avoid this problem in the future? In the intro-
ductory text to Section 1.9 (before Example 1.9.1), the notes tell us that we
are allowed to write our old variable as a function of a new variable (say
x = s(u)) as long as that function is invertible to recover our original variable
x. There is one very obvious reason why invertibility is necessary: after we
antidifferentiate using our new variable u, we need to get it back in terms of
our original variable, so we need to be able to recover x. Moreover, invert-
ibility reconciles potential problems with domains: if an inverse function
u = s−1 (x) exists, then for any x, there exists a u with s(u) = x. (This
was not the case in the work for the question, because we chose x = sin θ,
but if x = 2, there is no corresponding θ. Note, however, that x = sin θ is
invertible over [−1, 1], so the work is correct if we restrict x to those values.)
1.9.2.28. Answer. (a), (b): None.
(c): x < −a

1.10 · Partial Fractions

1.10.4 · Exercises

630
A NSWERS TO E XERCISES

Exercises — Stage 1
1.10.4.1. Answer. (a) (iii)
(b) (ii)
(c) (ii)
(d) (i)
A B C D Ex + F
1.10.4.2. *. Answer. + 2
+ + 2
+ 2
x − 1 (x − 1) x + 1 (x + 1) x +1
1.10.4.3. *. Answer. 3
x3 + 2x + 2 x+2
1.10.4.4. Answer. (a) 2
=x+ 2
x +1 x +1
15x4 + 6x3 + 34x2 + 4x + 20 2 4
(b) 2
= 3x + 2 + 2
5x + 2x + 8 5x + 2x + 8
2x5 + 9x3 + 12x2 + 10x + 30 3
(c) = x + 2x + 6
2x2 + 5
√ √
1.10.4.5. Answer. (a) 5x3 − 3x2 − 10x + 6 = (x + 2)(x − 2)(5x − 3)
(b)

x4 − 3x2 − 5
√ √ √
 s  s  !
3+ 29   3+ 29  29 − 3
= x + x− x2 +
2 2 2

(c) x4 − 4x3 − 10x2 − 11x − 6 = (x + 1)(x − 6)(x2 + x + 1)

(d)

2x4 + 12x3 − x2 − 52x + 15

 √  √ 
= (x + 3)(x + 5) x − (2 + 2) x − (2 − 2)

1.10.4.6. Answer. The goal of partial fraction decomposition is to write our

integrand in a form that is easy to integrate. The antiderivative of (1) can be
easily determined with the substitution u = (ax + b). It’s less clear how to find
the antiderivative of (2).

Exercises — Stage 2
4
1.10.4.7. *. Answer. log
3
1
1.10.4.8. *. Answer. − − arctan x + C
x
1.10.4.9. *. Answer. 4 log |x − 3| − 2 log(x2 + 1) + C
1.10.4.10. *. Answer. F (x) = log |x − 2| + log |x2 + 4| + 2 arctan(x/2) + D
1.10.4.11. *. Answer. −2 log |x − 3| + 3 log |x + 2| + C

631
A NSWERS TO E XERCISES

1.10.4.12. *. Answer. −9 log |x + 2| + 14 log |x + 3| + C

1 7
1.10.4.13. Answer. 5x + log |x − 1| − log |x + 1| + C
2 2
2 5
1.10.4.14. Answer. x− + arctan(2x) + C
x 2
1 2
1.10.4.15. Answer. − +C
x x−1
1 1 3
1.10.4.16. Answer. − log |x − 2| + log |x + 2| + log |2x − 1| + C
2 2 2
4 · 63
 
1.10.4.17. Answer. log
53

Exercises — Stage 3
1 1 − cos x
1.10.4.18. Answer. log +C
2 1 + cos x
− cos x 1 1 − cos x
1.10.4.19. Answer. 2 + log +C
2 sin x 4 1 + cos x
    
1 2 7 9
1.10.4.20. Answer. 3 log 2 + + √ arctan √ − arctan √
2 15 15 15
 
9 x 2 + 3x
1.10.4.21. Answer. = √ arctan √ − +C
4 2 2 4(x2 + 2)
3 3x3 + 5x
1.10.4.22. Answer. arctan x + +C
8 8(1 + x2 )2
 
3 2 1 x 3 3
1.10.4.23. Answer. x + √ arctan √ + log |x2 + 5| − 2 +C
2 5 5 2 2x + 10
sin θ − 1
1.10.4.24. Answer. log +C
sin θ − 2
2et + 1
 
1 1
1.10.4.25. Answer. t − log |e2t + et + 1| − √ arctan √ +C
2 3 3
√
√ 1 + ex − 1
1.10.4.26. Answer. 2 1 + e + log √
x +C
1 + ex + 1
1.10.4.27. *. Answer. (a) The region R is

632
A NSWERS TO E XERCISES

9 3
(b) 10π log = 20π log
4 2
(c) 20π
5 4 1
1.10.4.28. Answer. 2 log + √ arctan √
3 3 4 3
 
1 x−3
1.10.4.29. Answer. (a) log 2 ·
6 x+3
1
(b) F 0 (x) = x2 −9

1.11 · Numerical Integration

1.11.6 · Exercises

Exercises — Stage 1
1.11.6.1. Answer. Relative error: ≈ 0.08147; absolute error: 0.113; percent error:
≈ 8.147%.
1.11.6.2. Answer. Midpoint rule:

x
2 10

Trapezoidal rule:

633
A NSWERS TO E XERCISES

x
2 10

1.11.6.3. Answer. M = 6.25, L = 2

1.11.6.4. Answer. One reasonable answer is M = 3.
π5
1.11.6.5. Answer. (a)
180 · 8
(b) 0
(c) 0
3
1.11.6.6. Answer. Possible answers: f (x) = x2 + Cx + D for any constants C,
2
D.
1.11.6.7. Answer. my mother
1.11.6.8. Answer. (a) true
(b) false
1.11.6.9. *. Answer. True. Because f (x) is positive and concave up, the graph
of f (x) is always below the top edges of the trapezoids used in the trapezoidal
rule.
y
y = f (x)

1.11.6.10. Answer. Any polynomial of degree at most 3 will do. For example,
f (x) = 5x3 − 27, or f (x) = x2 .

Exercises — Stage 2

634
A NSWERS TO E XERCISES

1.11.6.11. Answer. Midpoint:

30 
1 1 1 1
Z
dx ≈ (2.5)13 +1 + + +
0 x3 +1 (7.5) + 1 (12.5) + 1 (17.5)3 + 1
3 3

1 1
+ + 3 5
(22.5)3 + 1 (27.5) +1

Trapezoidal:
30 
1 1/2 1 1 1 1
Z
3
dx ≈ 3 + 3 + 3 + 3 + 3
0 x +1 0 + 1 5 + 1 10 + 1 15 + 1 20 + 1

1 1/2
+ 3 + 5
25 + 1 303 + 1

Simpson’s:
30 
1 1 4 2 4 2
Z
3
dx ≈ 3 + 3 + 3 + 3 + 3
0 x +1 0 + 1 5 + 1 10 + 1 15 + 1 20 + 1

4 1 5
+ 3 + 3
25 + 1 30 + 1 3
2π
1.11.6.12. *. Answer.
3
1.11.6.13. *. Answer. 1720π ≈ 5403.5 cm3
π
1.11.6.14. *. Answer. (16.72) ≈ 4.377 m3
12
12.94
1.11.6.15. *. Answer. ≈ 0.6865 m3
6π
1.11.6.16. *. Answer. (a) 363,500
(b) 367,000
49
1.11.6.17. *. Answer. (a)
2
77
(b)
3
1.11.6.18. *. Answer. Let f (x) = sin(x2 ). Then f 0 (x) = 2x cos(x2 ) and
f 00 (x) = 2 cos(x2 ) − 4x2 sin(x2 ).
Since |x2 | ≤ 1 when |x| ≤ 1, and |sin θ| ≤ 1 and |cos θ| ≤ 1 for all θ, we have
2 cos(x2 ) − 4x2 sin(x2 ) ≤ 2| cos(x2 )| + 4x2 | sin(x2 )|
≤2×1+4×1×1=2+4=6
We can therefore choose M = 6, and it follows that the error is at most
M [b − a]3 6 · [1 − (−1)]3 2
≤ = = 2 · 10−6
24n2 24 · 10002 106

635
A NSWERS TO E XERCISES

3
1.11.6.19. *. Answer.
100
1.11.6.20. *. Answer. (a)

1/3  1 5 2 5 4 5
(−3)5 + 4 −3 +2 − 3 + 4(−2)5 + 2 −3
3 3 3 3
5 5 
+4 − 3 + (−1)5
3
(b) Simpson’s Rule results in a smaller error bound.
8
1.11.6.21. *. Answer.
15
1 1
1.11.6.22. *. Answer. 4
=
180 × 3 14580
   
1 1 4 2 4 1 1
1.11.6.23. *. Answer. (a) T4 = ×1 + + + + × ,
    4 2  5 3  7 2 2
1 4 2 4 1
(b) S4 = 1+ 4× + 2× + 4× +
12 5 3 7 2
24 3
(c) I − S4 ≤ =
180 × 44 5760
1.11.6.24. *. Answer. (a) T4 = 8.03515, S4 ≈ 8.03509
(b) Z b
2 83
f (x) dx − Tn ≤ ≤ 0.00533
a 1000 12(4)2
Z b
4 85
f (x) dx − Sn ≤ ≤ 0.00284
a 1000 180(4)4
1.11.6.25. *. Answer. Any n ≥ 68 works.

Exercises — Stage 3
472
1.11.6.26. *. Answer. ≈ 494 ft3
3
1.11.6.27. *. Answer. (a) 0.025635
(b) 1.8 × 10−5
1.11.6.28. *. Answer. (a) ≈ 0.6931698
(b) n ≥ 12 with n even
1.11.6.29. *. Answer. (a) 0.01345
(b) n ≥ 28 with n even
1.11.6.30. *. Answer. n ≥ 259

636
A NSWERS TO E XERCISES

1.11.6.31. Answer. (a) When 0 ≤ x ≤ 1, then x2 ≤ 1 and x + 1 ≥ 1, so

x2 1
|f 00 (x)| = ≤ =1
|x + 1| 1
1
(b)
2
(c) n ≥ 65
(d) n ≥ 46
 
x−1 16 4 16 1
1.11.6.32. Answer. 1+ + + +
12 x + 3 x + 1 3x + 1 x
1.11.6.33. Answer. Note: for more detail, see the solutions.
R2 1
First, we use Simpson’s rule with n = 4 to approximate 1 1+x 2 dx. The choice

of this method (what we’re approximating, why n = 4, etc.) is explained in the

solutions–here, we only show that it works.
Z 2  
1 1 1 64 8 64 1
2
dx ≈ + + + + ≈ 0.321748
1 1+x 12 2 41 13 65 5
For ease of notation, define A = 0.321748.
Now, we bound the error associated with this approximation. Define N (x) =
24(5x4 − 10x2 + 1) and D(x) = (x2 + 1)5 , so N (x)/D(x) gives the fourth derivative
1
of 1+x 2 . When 1 ≤ x ≤ 2, |N (x)| ≤ N (2) = 984 (because N (x) is increasing over

that interval) and |D(x)| ≥ D(1) = 25 (because D(x) is also increasing over that
d4 (x)
= N
 1
interval), so dx 4 1+x2 D(x)
≤ 984
25
= 30.75. Now we find the error bound
for Simpson’s rule with L = 30.75, b = 2, a = 1, and n = 4.
Z 2
1 L(b − a)5 30.75
2
dx − A = |error| ≤ 4
= < 0.00067
1 1+x 180 · n 180 · 44
So,
2
1
Z
−0.00067 < 2
dx − A < 0.00067
1 1+x
Z 2
1
A − 0.0067 < 2
dx < A + 0.00067
1 1+x
A − 0.00067 < arctan(2) − arctan(1) < A + 0.00067
π
A − 0.00067 < arctan(2) − < A + 0.00067
4
π π
+ A − 0.00067 < arctan(2) < + A + 0.00067
4 4
π π
+ 0.321748 − 0.00067 < arctan(2) < + 0.321748 + 0.00067
4 4
π π
+ 0.321078 < arctan(2) < + 0.322418
4 4
π π
+ 0.321 < arctan(2) < + 0.323
4 4
This was the desired bound.

637
A NSWERS TO E XERCISES

1.12 · Improper Integrals

1.12.4 · Exercises

Exercises — Stage 1
1.12.4.1. Answer. Any real number in [1, ∞) or (−∞, −1], and b = ±∞.
1.12.4.2. Answer. b = ±∞
1.12.4.3. Answer. The red function is f (x), and the blue function is g(x).
1.12.4.4. *. Answer. False. For example, the functions f (x) = e−x and g(x) = 1
provide a counterexample.
1.12.4.5. Answer.

a Not enough information to decide. For example, consider h(x) = 0 versus

h(x) = −1.

b Not enough information to decide. For example, consider h(x) = f (x) ver-
sus h(x) = g(x).
Z ∞
c h(x)dx converges by the comparison test, since |h(x)| ≤ 2f (x) and
0
Z ∞
2f (x)dx converges.
0

Exercises — Stage 2
1.12.4.6. *. Answer. The integral diverges.
1.12.4.7. *. Answer. The integral diverges.
1.12.4.8. *. Answer. The integral does not converge.
1.12.4.9. *. Answer. The integral converges.
1.12.4.10. Answer. The integral diverges.
1.12.4.11. Answer. The integral diverges.
1.12.4.12. Answer. The integral diverges.
1.12.4.13. Answer. The integral diverges.
1.12.4.14. *. Answer. The integral diverges.
1.12.4.15. *. Answer. The integral converges.
1.12.4.16. *. Answer. The integral converges.

Exercises — Stage 3
1.12.4.17. Answer. false

638
A NSWERS TO E XERCISES

1
1.12.4.18. *. Answer. q= 5

1.12.4.19. Answer. p>1

log 3 − π 1
1.12.4.20. Answer. + arctan 2
4 2
1.12.4.21. Answer. The integral converges.
1
1.12.4.22. Answer.
2
1.12.4.23. *. Answer. The integral converges.
1.12.4.24. Answer. The integral converges.
1.12.4.25. *. Answer. t = 10 and n = 2042 will do the job. There are many other
correct answers.
1.12.4.26. Answer. (a) The integral converges.
(b) The interval converges.
1.12.4.27. Answer. false

1.13 · More Integration Examples

· Exercises

Exercises — Stage 1
1.13.1. Answer. (A)–(I), (B)–(IV), (C)–(II), (D)–(III)

Exercises — Stage 2
1 2 1 8
1.13.2. Answer. − + =
5 7 9 315
r !
3 5 x√
1.13.3. Answer. √ arcsin x + 3 − 5x2 + C
2 5 3 2

1.13.4. Answer. 0
x+1
1.13.5. Answer. log +C
3x + 1
8 7
1.13.6. Answer. log 2 −
3 9
1
1.13.7. *. Answer. log x2 − 3 + C
2

639
A NSWERS TO E XERCISES

1.13.8. *. Answer. (a) 2

2
(b)
15
3e4 1
(c) +
16 16
1.13.9. *. Answer. (a) 1
8
(b)
15
√
1.13.10. *. Answer. (a) e2 + 1 (b) log( 2 + 1) (c) log 15
13
≈ 0.1431
9
1.13.11. *. Answer. (a) π
4
π
(b) log 2 − 2 + ≈ 0.264
2
(c) 2 log 2 − 21 ≈ 0.886
1 3 sin θ − 3
1.13.12. Answer. sin θ − 2 sin θ + 12 log +C
3 sin θ − 2
1
1.13.13. *. Answer. (a)
15
1 x
(b) · √ +C
9 x2 + 9
1 1 1
(c) log |x − 1| − log(x2 + 1) − arctan x + C
2 4 2
1 2 
(d) x arctan x − x + arctan x + C
2
1
1.13.14. *. Answer. (a)
r 12
x x 2
(b) 2 sin−1 + x 1 − +C
2 4
1
(c) −2 log |x| + + 2 log |x − 1| + C
x
2
1.13.15. *. Answer. (a)
5
1
(b) √
2 2
1
(c) log 2 − ≈ 0.193
2
1
(d) log 2 − ≈ 0.193
2
1 1
1.13.16. *. Answer. (a) x2 log x − x2 + C
2 4
1 2
(b) log[x + 4x + 5] − 3 arctan(x + 2) + C
2
1 1
(c) log |x − 3| − log |x − 1| + C
2 2
1 3
(d) arctan x + C
3

640
A NSWERS TO E XERCISES

π 1 1 x−1
1.13.17. *. Answer. (a) − log 2 (b) log |x2 − 2x + 5| + arctan +C
4 2 2 2
1
1.13.18. *. Answer. (a) − +C
300(x3 + 1)100
sin5 x sin7 x
(b) − +C
5 7
1.13.19. Answer. -2
1 1 4π √
1.13.20. *. Answer. (a) − log |ex + 1| + log |ex − 3| + C (b) −2 3
4 4 3
1
1.13.21. *. Answer. (a) sec2 x + log | cos x| + C
2
1
(b) arctan 8 ≈ 0.1446
10
2 2
1.13.22. Answer. (x − 1)5/2 + (x − 1)3/2 + C
5 3
√
√ x2 − 2
1.13.23. Answer. log x + x2 − 2 − +C
x
7
1.13.24. Answer.
24
2 5
1.13.25. Answer. 3 log |x + 1| + − +C
x + 1 2(x + 1)2
 
2 2 1
1.13.26. Answer. √ arctan √ x + √ +C
3 3 3
1
1.13.27. Answer. (x − sin x cos x) + C
2
 
1 1 1 2x − 1
1.13.28. Answer. log |x + 1| − log |x2 + x + 1| + √ arctan √ +C
3 6 3 3
√
1.13.29. Answer. 3x3 arcsin x + 3 1 − x2 − (1 − x2 )3/2 + C

Exercises — Stage 3
1.13.30. Answer. 2
1
1.13.31. Answer.
4
 
log(cos(0.1))
1.13.32. Answer. log
log(cos(0.2))
1  
1.13.33. *. Answer. (a) x sin(log x) − cos(log x) + C
2
(b) 2 log 2 − log 3 = log 34

641
A NSWERS TO E XERCISES

9
1.13.34. *. Answer. (a) π + 9
4
2 π
(b) 2 log |x − 2| − log(x + 4) + C (c)
2
√ √ √
1.13.35. Answer. − arcsin( 1 − x) − 1 − x x + C
1.13.36. Answer. ee (e − 1)
ex
1.13.37. Answer. +C
x+1
1.13.38. Answer. x sec x − log | sec x + tan x| + C

(x+a)(n+2) (x+a)n+1
 n+2 − a n+1 + C if n 6= −1, −2
Z 

1.13.39. Answer. x(x + a)n dx = (x + a) − a log |x + a| + C if n = −1

log |x + a| + a + C

if n = −2
x+a

1.13.40. Answer.

1 1 √ √ 
x2 −√2x+1
x arctan(x2 ) − √ log 2
x + 2x+1
+ arctan 2x + 1
2 2
!
√ 
+ arctan 2x − 1 +C

2 · Applications of Integration
2.1 · Work
2.1.2 · Exercises

Exercises — Stage 1
2.1.2.1. Answer. 0.00294 J
1
2.1.2.2. Answer. The rock has mass kg (about 102 grams); lifting it one
9.8
metre takes 1 J of work.
2.1.2.3. Answer. (a) metres
(b) newtons
(c) joules
smoot · barn
2.1.2.4. Answer. (smoot-barns per megaFonzie)
megaFonzie
2.1.2.5. Answer. 10 cm below the bottom of the unloaded spring
2.1.2.6. Answer. x=2

Exercises — Stage 2

642
A NSWERS TO E XERCISES

2.1.2.7. *. Answer. a = 3
2.1.2.8. Answer.
  (a) joules
`−1
(b) c log J
` − 1.5
1
2.1.2.9. *. Answer. J
4
2.1.2.10. *. Answer. 25 J
2.1.2.11. *. Answer. 196 J
2.1.2.12. Answer. 14700 J
Z 3
2.1.2.13. *. Answer. (9.8)(8000)(2 + z)(3 − z)2 dz joules
0

2.1.2.14. Answer. 0.2352 J

20
2.1.2.15. Answer. kg, or about 408 grams
49
2.1.2.16. Answer. 294 J

√  (a) 117.6 J
2.1.2.17.Answer.
(b) 3.92 30 − 2 3 ≈ 104 J
1
2.1.2.18. Answer. √ m/sec, or about 22.36 cm/sec
2 5
2.1.2.19. Answer. yes (at least, the car won’t scrape the ground)

Exercises — Stage 3
2.1.2.20. Answer. ≈ 0.144 J
2.1.2.21. *. Answer. 904,050π J
2.1.2.22. Answer. 1020 65 J
2.1.2.23. Answer. (a) 4900 N
(b) 44100
x2
N
(c) 29 400 J
2.1.2.24. Answer. 220.5 J
2.1.2.25. Answer. About 7 × 1028 J
2.1.2.26. Answer. true
2.1.2.27. Answer. 9255 95 J
7
2.1.2.28. Answer. = 0.175 J
40

643
A NSWERS TO E XERCISES
s 
 4 s  4
1 1 3 
2.1.2.29. Answer. One possible answer: 1− + 1−
4 8 8

2.2 · Averages
2.2.2 · Exercises

Exercises — Stage 1
2.2.2.1. Answer. The most straightforward of many possible answers is shown.

x
5

2.2.2.2. Answer. 500 km

W
2.2.2.3. Answer. N
b−a
b−a
2.2.2.4. Answer. (a)
n
b−a
(b) a + 3
 n 
b−a
(c) f a + 3
n
1P n
f a + (i − 1) b−a


(d) n
n i=1
2.2.2.5. Answer. (a) yes
(b) not enough information
2.2.2.6. Answer. 0

Exercises — Stage 2
2.2.2.7. *. Answer. 1
1 h2 3 1i
2.2.2.8. *. Answer. e +
e−1 9 9

644
A NSWERS TO E XERCISES

4
2.2.2.9. *. Answer. +1
π
2
2.2.2.10. *. Answer.
π
10
2.2.2.11. *. Answer. log 7 degrees Celsius
3
1
2.2.2.12. *. Answer.
2(e − 1)
1
2.2.2.13. *. Answer.
2
2.2.2.14. Answer. (a) 400 ppm
(b) ≈ 599.99 ppm
(c) 0.125, or 12.5%
16π
2.2.2.15. Answer. (a)
5
32π
(b)
5
32π
(c)
5
2.2.2.16.
√ Answer. (a) 0
(b) 3
r
4
2.2.2.17. Answer. − 1 ≈ 0.52
π
2.2.2.18. Answer. (a) F (t) = 3f (t) = 3 sin (tπ) N
(b) 0
3
(c) √ ≈ 2.12
2

Exercises — Stage 3
2.2.2.19. *. Answer. (a) 130 km
(b) 65 km/hr
2.2.2.20. Answer. (a) A = e − 1
(b) 0
(c) 4 − 2e + 2(e − 1) log(e − 1) ≈ 0.42
2.2.2.21. Answer. (a) neither — both are zero
(b) |f (x) − A| has the larger average on [0, 4]
2.2.2.22. Answer. (b − a)πR2
2.2.2.23. Answer. 0
2.2.2.24. Answer. Yes, but if a 6= 0, then s = t.

645
A NSWERS TO E XERCISES

2.2.2.25. Answer. A
bA(b) − aA(a)
2.2.2.26. Answer. (a)
b−a
(b) f (t) = A(t) + tA0 (t)
2.2.2.27. Answer.
(
−1 if x ≤ 0
a One of many possible answers: f (x) = .
1 if x > 0

b No such function exists.

R1
• Note 1: Suppose f (x) > 0 for all x in [−1, 1]. Then 21 −1 f (x) dx >
1 1
R
2 −1
0 dx = 0. That is, the average value of f (x) on the interval [−1, 1]
is not zero — it’s something greater than zero.
R1
• Note 2: Suppose f (x) < 0 for all x in [−1, 1]. Then 21 −1 f (x) dx <
1 1
R
2 −1
0 dx = 0. That is, the average value of f (x) on the interval [−1, 1]
is not zero — it’s something less than zero.

So, if the average value of f (x) is zero, then f (x) ≥ 0 for some x in [−1, 1],
and f (y) ≤ 0 for some y ∈ [−1, 1]. Since f is a continuous function, and 0
is between f (x) and f (y), by the intermediate value theorem (see the CLP-
1 text) there is some value c between x and y such that f (c) = 0. Since x
and y are both in [−1, 1], then c is as well. Therefore, no function exists as
described in the question.
2.2.2.28. Answer. true
2.2.2.29. Answer. 0

2.3 · Centre of Mass and Torque

2.3.3 · Exercises

Exercises — Stage 1
2.3.3.1. Answer. (1, 1)
2.3.3.2. Answer. (0, 0)
2.3.3.3. Answer. In general, false.
2.3.3.4. Answer. 3.5 metres from the left end
2.3.3.5. Answer. (a) to the left
(b) to the left
(c) not enough information
(d) along the line x = a
(e) to the right

646
A NSWERS TO E XERCISES

78400π
2.3.3.6. Answer. (6 − π) ≈ 78, 225J
9
1
2.3.3.7. Answer. (a), (b) dx
x
(c), (d) log 3
2
(e), (f)
log 3
2.3.3.8. Answer. (a)
n 
1
b−a
( b−a ) × a + (i − 21 ) b−a
P




n
ρ a+ i− 2 n n
i=1
n
b−a
a + (i − 12 ) b−a
P


n
ρ n
i=1

Rb
xρ(x) dx
(b) x̄ = Ra b
a
ρ(x) dx
2.3.3.9. Answer. (a)

y T (x)

B(x)
x
a a 0
b 0 b

(b) (T (x) − B(x)) dx

(c) T (x) − B(x)
Rb
x(T (x) − B(x)) dx
(d) x̄ = Ra b
a
(T (x) − B(x)) dx
2.3.3.10. Answer. (a) The strips between x = a and x = a0 at the left end of the
figure all have the same centre of mass, which is the y-value where T (x) = B(x),
x < 0. So, there should be multiple weights of different mass piled up at that
y-value.
Similarly, the strips between x = b0 and x = b at the right end of the figure all
have the same centre of mass, which is the y-value where T (x) = B(x), x > 0.
So, there should be a second pile of weights of different mass, at that (higher)

647
A NSWERS TO E XERCISES

y-value.
Between these two piles, there are a collection of weights with identical mass
distributed fairly evenly. The top and bottom ends of R (above the uppermost
pile, and below the lowermost pile) have no weights.
One possible answer (using twelve slices):

y T (x) R

B(x)

x
a b

(b) The area of the strip is (T (x) − B(x)) dx, and its centre of mass is at height
T (x) + B(x)
.
2R
b
T (x)2 − B(x)2 dx


a
(c) ȳ = R b

2 a T (x) − B(x) dx

1 0 2
Z
2.3.3.11. *. Answer. x̄ = − 6x dx
3 −1

Exercises — Stage 2
14
2.3.3.12. Answer. x̄ =
3
log 10.1
2.3.3.13. Answer. x̄ = ≈ 0.43
2(arctan 10 + arctan(3))
3 e
2.3.3.14. *. Answer. ȳ = −
4e 4
2.3.3.15. *. Answer. (a)

648
A NSWERS TO E XERCISES

3 log 3
(b)
8π
π
√
4
2−1 1
2.3.3.16. *. Answer. x̄ = √ and ȳ = √
2−1 4( 2 − 1)
k √  k2π
2.3.3.17. *. Answer. (a) x̄ = 2−1 , ȳ =
A 8A
8 √ 
(b) k = 2−1
π
2.3.3.18. *. Answer. (a)

8
(b)
3
(c) 1
2
2.3.3.19. *. Answer. log 2 ≈ 0.44127
π
12
2.3.3.20. *. Answer. x̄ = 0 and ȳ =
24 + 9π
9
2.3.3.21. *. Answer. (a) π
4
4
(b) x̄ = 0 and ȳ =
π
 
2
2.3.3.22. Answer. (x̄, ȳ) = 1, −
π

649
A NSWERS TO E XERCISES

e2 − 3/2 e4 − 7
 
2.3.3.23. Answer. , ≈ (1.2, 2.4)
e2 − 5/2 4e2 − 10

Exercises — Stage 3
8
2.3.3.24. *. Answer. ȳ =
5
8 166
2.3.3.25. *. Answer. (a) x̄ = , ȳ =
Z 4 Z 6 11 55
2
(b) π y dy + π (6 − y) dy
0 4

e 3
2.3.3.26. *. Answer. (a) ȳ = −
 2  4 4e
e 3
(b) π + 2e −
2 2
2.3.3.27. Answer. (3, 1.5)
2.3.3.28. Answer. (0, 3.45)
h
2.3.3.29. Answer. (a)
4
1 2
2
h k − 23 hk 2 + 14 k 3
(b)
h2 − hk + 13 k 2
2.3.3.30. Answer. about 0.833 N
2.3.3.31. Answer. (a) 17,150π J
2450
(b) π (8π − 9) ≈ 13, 797 J
9
(c) about 74%
r   
π π π π π  2π
2.3.3.32. Answer. x̄ = sin + 2 sin + 9 sin + 8 sin +
   162 2 72 18 8 9
25π
25 sin + 9 ≈ 0.976
72

2.4 · Separable Differential Equations

2.4.7 · Exercises

Exercises — Stage 1
2.4.7.1. Answer. (a) yes
(b) yes
(c) no
2.4.7.2. Answer.
sin y
a One possible answer: f (x) = x, g(y) = .
3y

650
A NSWERS TO E XERCISES

b One possible answer: f (x) = ex , g(y) = ey .

c One possible answer: f (x) = x − 1, g(y) = 1.

dy
d The given equation is equivalent to the equation dx
= x, which fits the form
of a separable equation with f (x) = x, g(y) = 1.
2.4.7.3. Answer. The mnemonic allows us to skip from the separable differen-
tial equation we want to solve (very first line) to the equation

1
Z Z
dy = f (x) dx
g(y)
2.4.7.4. Answer. false
2.4.7.5. Answer. (a) [0, ∞)
(b) No such function exists. If |f (x)| = Cx and f (x) switches from f (x) = Cx to
f (x) = −Cx at some point, then that point is a jump discontinuity. Where f (x)
dy
contains a discontinuity, dx does not exist.
dQ
2.4.7.6. Answer. = −0.003Q(t)
dt
dp


2.4.7.7. Answer. dt
= αp(t) 1 − p(t) , for some constant α.
2.4.7.8. Answer. (a) −1
(b) 0
(c) 0.5
(d) Two possible answers are shown below:
y

x
1

651
A NSWERS TO E XERCISES

x
1

Another possible answer is the constant function y = 2.

1
2.4.7.9. Answer. (a) −
2
3
(b)
2
5
(c) −
2
(d) Your sketch should look something like this:
y

x
1

(e) There are lots of possible answers. Several are shown below.

652
A NSWERS TO E XERCISES

x
1

x
1

x
1

653
A NSWERS TO E XERCISES

x
1

Exercises — Stage 2
2.4.7.10. *. Answer. y = log(x2 + 2)
√
2.4.7.11. *. Answer. y(x) = 3 1 + x2
 
−3
2.4.7.12. *. Answer. y(t) = 3 log
C + sin t
q
2.4.7.13. *. Answer. y = 3 32 ex2 + C.

x2
 
2.4.7.14. *. Answer. y = − log C −
2
x2
The solution only exists for C − 2 > 0, i.e. C > 0 and the function has domain
n √ o
x : |x| < 2C .

2.4.7.15. *. Answer. y = (3ex − 3x2 + 24)1/3

1
2.4.7.16. *. Answer. y = f (x) = − √
x2 + 16
√
2.4.7.17. *. Answer. y = 10x3 + 4x2 + 6x − 4
4 /4
2.4.7.18. *. Answer. y(x) = ex
1
2.4.7.19. *. Answer. y= 1−2x
2
2.4.7.20. *. Answer. f (x) = e · ex /2
q
2x
2.4.7.21. *. Answer. y(x) = 4 + 2 log x+1 . Note that, to satisfy y(1) = 2, we
need the positive square root.

654
A NSWERS TO E XERCISES

2
2.4.7.22. *. Answer. y 2 + (y 2 − 4)3/2 = 2 sec x + 2
3
2.4.7.23. *. Answer. 12 weeks
s !
m k
r
2.4.7.24. *. Answer. t= arctan v0
kg mg
1
2.4.7.25. *. Answer. (a) k = 400
(b) t = 70sec
3 − 4ekt
2.4.7.26. *. Answer. (a) x(t) =
1 − 2ekt
(b) As t → ∞, x → 2.
4
2.4.7.27. *. Answer. (a) P =
1 + e−4t
1
(b) At t = , P ≈ 3.523. As t → ∞, P → 4.
2
dv
2.4.7.28. *. Answer. (a) = −kv 2
dt
400
(b) v =
t+1
(c) t = 7

Exercises — Stage 3
2.4.7.29. *. Answer. (a) B(t) = C e0.06t−0.02 cos t with the arbitrary constant C ≥ 0.
(b) $1159.89
2.4.7.30. *. Answer. (a) B(t) = {30000 − 50m} et/50 + 50m
(b) $600
4 − e1−cos x
2.4.7.31. *. Answer. y(x) = . The largest allowed interval is
2 − e1−cos x
− arccos(1 − log 2) < x < arccos(1 − log 2)

or, roughly, −1.259 < x < 1.259.

q
2.4.7.32. *. Answer. 180, 000 g3 ≈ 99, 591 sec ≈ 27.66 hr
s
4 × 144 125
2.4.7.33. *. Answer. t = ≈ 2, 394 sec ≈ 0.665 hr
15 2g

2.4.7.34. *. Answer. (a) 3

(b) y 0 = (y − 1)(y − 2)
4 − ex
(c) f (x) =
2 − ex

655
A NSWERS TO E XERCISES

1
2.4.7.35. *. Answer. p= 4

2.4.7.36. Answer.

a One possible answer: f (t) = 0

Z x
f 2 (x)
 
1 1
b √ f (x) − f (t) dt = q
x−a 2(x − a) a Rx
2 a f 2 (t) dt
x  x 
2 1
Z Z
c f (t) dt f (x) − f (t) dt = f 2 (x)
x−a a 2(x − a) a

d Y (x) = D(x − a), where D is any constant

e f (t) = D, for any nonnegative constant D

 
1 1 2y − 1
2.4.7.37. Answer. x = y − 1 + log
4 4 2y + 1

3 · Sequence and series

3.1 · Sequences
3.1.2 · Exercises

Exercises — Stage 1
3.1.2.1. Answer. (a) −2 (b) 0 (c) the limit does not exist
3.1.2.2. Answer. true
A−B A
3.1.2.3. Answer. (a) (b) 0 (c)
C B
3.1.2.4. Answer. Two possible answers, of many:
(
3000 − n if n ≤ 1000
• an =
−2 + n1 if n > 1000
1, 002, 001
• an = −2
n
3.1.2.5. Answer. One possible answer is an = (−1)n =
{−1, 1, −1, 1, −1, 1, −1, . . .}.
Another is an = n(−1)n = {−1, 2, −3, 4, −5, 6, −7, . . .}.
(−1)n
3.1.2.6. Answer. One sequence of many possible is an = =
  n
1 1 1 1 1
−1, , − , , − , , . . . .
2 3 4 5 6
3.1.2.7. Answer. Some possible answers:

656
A NSWERS TO E XERCISES

−1 sin n 1
a ≤ ≤
n n n
n2 n2 n2
b ≤ ≤
13en en (7 + sin n − 5 cos n) en
−1 1
c n
≤ (−n)−n ≤ n
n n
3.1.2.8. Answer. (a) an = bn = h(n) = i(n), cn = j(n), dn = f (n), en = g(n)
(b) lim an = lim bn = lim h(x) = 1, lim cn = lim en = lim g(x) = lim j(x) =
n→∞ n→∞ x→∞ n→∞ n→∞ x→∞ x→∞
0, lim dn , lim f (x) and lim i(x) do not exist.
n→∞ x→∞ x→∞

3.1.2.9. Answer. (a) Some possible answers: a22 ≈ −0.99996, a66 ≈ −0.99965,
and a110 ≈ −0.99902.
(b) Some possible answers: a11 ≈ 0.0044, a33 ≈ −0.0133, and a55 ≈ 0.0221.
22
The integers 11, 33, and 55 were found by approximating π by and finding
7
11
when an odd multiple of (which is the corresponding approximation of π/2)
7
is an integer.

Exercises — Stage 2
3
3.1.2.10. Answer. (a) ∞ (b) (c) 0
4
3.1.2.11. Answer. ∞
3.1.2.12. Answer. 0
3.1.2.13. Answer. 0
3.1.2.14. Answer. 0
3.1.2.15. Answer. 1
3.1.2.16. Answer. 0
3.1.2.17. Answer. ∞
3.1.2.18. *. Answer. lim ak = 0.
k→∞

3.1.2.19. *. Answer. The sequence converges to 0.

3.1.2.20. *. Answer. 9

Exercises — Stage 3
3.1.2.21. *. Answer. log 2
3.1.2.22. Answer. 5
3.1.2.23. Answer. −∞
3.1.2.24. Answer. 100 · 299 .

657
A NSWERS TO E XERCISES
    
1
3.1.2.25. Answer. Possible answers are {an } = n f a + − f (a) or
    n
1
{an } = n f (a) − f a − .
n
 
n 2π
3.1.2.26. Answer. (a) An = sin (b) π
2 n
3.1.2.27. Answer.

x
2 3

x
3 4

c An = 1 for all n

d lim An = 1.
n→∞

e g(x) = 0
Z ∞
f g(x) dx = 0.
0

3.1.2.28. Answer. e3
3.1.2.29. Answer. (a) 4 (b) x = 4 (c) see solution
3.1.2.30. Answer. (a) decreasing (b)fn = n1 f1 (c) 2% (d) 0.18%
(e) “be”: 11,019,308; “and”: 7,346,205

658
A NSWERS TO E XERCISES

3.2 · Series
3.2.2 · Exercises

Exercises — Stage 1
3.2.2.1. Answer.
N SN
1 1
2 1 + 21
3 1 + 21 + 13
4 1 + 21 + 13 + 41
5 1 + 21 + 13 + 41 + 1
5

3.2.2.2. Answer. 3

 12 if n = 1
3.2.2.3. Answer. (a) an = 1 (b) 0 (c) 1
 else
n(n + 1)
(
0 if n = 1
3.2.2.4. Answer. an = n 1
2(−1) − n(n−1)
else

3.2.2.5. Answer. an < 0 for all n ≥ 2

∞
X 2 2
3.2.2.6. Answer. (a) n
(b)
n=1
4 3
∞
X 1 1
3.2.2.7. Answer. (a) (b)
n=1
9n 8

3.2.2.8. Answer. Two possible pictures:

659
A NSWERS TO E XERCISES

5101 − 1
3.2.2.9. Answer.
4 · 5100
3.2.2.10. Answer. All together, there were 36 cookies brought by Student 11
through Student 20.
551 − 1
3.2.2.11. Answer.
4 · 5100
3.2.2.12. Answer. (a) As time passes, your gains increase, approaching $1.
(b) 1
(c) As time passes, you lose more and more money, without bound. (d) −∞
3.2.2.13. Answer. A + B + C − c1
3.2.2.14. Answer. in general, false

Exercises — Stage 2
3
3.2.2.15. *. Answer.
2
1
3.2.2.16. *. Answer.
7 × 86
3.2.2.17. *. Answer. 6
π  1
3.2.2.18. *. Answer. cos − cos(0) = −
3 2
11 3
3.2.2.19. *. Answer. (a) an = (b)
16n2 + 24n + 5 4

660
A NSWERS TO E XERCISES

24
3.2.2.20. *. Answer.
5
7
3.2.2.21. *. Answer.
30
263
3.2.2.22. *. Answer.
99
321 107
3.2.2.23. *. Answer. =
999 333
3.2.2.24. *. Answer. 3
1 5 17
3.2.2.25. *. Answer. + =
2 7 14
40
3.2.2.26. *. Answer.
3
3.2.2.27. Answer. The series diverges to −∞.
1
3.2.2.28. Answer. −
2

Exercises — Stage 3
3.2.2.29. Answer. 9.8 J
4π
3.2.2.30. Answer.
3 (π 3 − 1)
sin2 3
3.2.2.31. Answer. + 32 ≈ 32.0025
8

2
 n(n−1)(n−2) if n ≥ 3,


3.2.2.32. Answer. an = − 52 if n = 2,


2 if n = 1
5
3.2.2.33. Answer.
8

3.3 · Convergence Tests

3.3.11 · Exercises

Exercises — Stage 1
3.3.11.1. Answer. (B), (C)
3.3.11.2. Answer. (A)
3.3.11.3. Answer. (a) I am old (b) not enough information to tell
(c) not enough information to tell (d) I am young

661
A NSWERS TO E XERCISES

3.3.11.4. Answer.
P P
if an converges if an diverges
P
and if {an } is the red series then bn CONVERGES inconclusive
P
and if {an } is the blue series inconclusive then bn DIVERGES

3.3.11.5. Answer. (a) both direct comparison and limit comparison

(b) direct comparison
(c) limit comparison (d) neither
3.3.11.6. Answer. It diverges by the divergence test, because lim an 6= 0.
n→∞

3.3.11.7. Answer. We cannot use the divergence test to show that a series con-
verges. It is inconclusive in this case.
3.3.11.8. Answer. The integral test does not apply because f (x) is not decreas-
ing.
3.3.11.9. Answer. The inequality goes the wrong way, so the direct comparison
test (with this comparison series) is inconclusive.
3.3.11.10. Answer. (B), (D)
∞
X 1
3.3.11.11. Answer. One possible answer: 2
.
n=1
n
P
3.3.11.12. Answer. By the divergence test, for a series an to converge, we need
lim an = 0. That is, the magnitude (absolute value) of the terms needs to be
n→∞
an an+1
getting smaller. If lim < 1 or (equivalently) lim > 1, then |an+1 | >
n→∞ an+1 n→∞ an
|an | for sufficiently large n, so the terms are actually growing in magnitude. That
means the series diverges, by the divergence test.
3.3.11.13. Answer. One possible answer: f (x) = sin(πx), an = 0 for every n.
By the integral test, any answer will use a function f (x) that is not both positive
and decreasing.
2n
3.3.11.14. *. Answer. One possible answer: bn =
3n
∞
1
P
3.3.11.15. *. Answer. (a) In general false. The harmonic series n
provides a
n=1
counterexample.
∞
(b) In general false. If an = (−1)n n1 , then (−1)n an is again the harmonic series
P
n=1
∞
1
P
n
, which diverges.
n=1
(c) In general false. Take, for example, an = 0 and bn = 1.

662
A NSWERS TO E XERCISES

Exercises — Stage 2
3.3.11.16. *. Answer. No. It diverges.
3.3.11.17. *. Answer. It diverges.
3.3.11.18. *. Answer. The series diverges.
3.3.11.19. Answer. It diverges.
3.3.11.20. Answer. This is a geometric series with r = 1.001. Since |r| > 1, it is
divergent.
1
3.3.11.21. Answer. The series converges to − .
50
3.3.11.22. Answer. The series converges.
3.3.11.23. Answer. It diverges.
3.3.11.24. Answer. The series converges.
1
3.3.11.25. Answer. The series converges to .
3
3.3.11.26. Answer. The series converges.
3.3.11.27. Answer. It converges.
5
3.3.11.28. *. Answer. Let f (x) = . Then f (x) is positive and de-
x(log x)3/2
∞
X Z ∞
creases as x increases. So the sum f (n) and the integral f (x) dx either
3 3
both converge or both diverge, by the integral test, which is Theorem 3.3.5. For
the integral, we use the substitution u = log x, du = dx x
to get
Z ∞ Z ∞
5 dx 5 du
3/2
= 3/2
3 x(log x) log 3 u

3
which converges by the p-test (which is Example 1.12.8) with p = 2
> 1.
3.3.11.29. *. Answer. p>1
3.3.11.30. *. Answer. It converges.
∞ √
X 3
3.3.11.31. *. Answer. The series 2
converges by the p-test with p = 2.
n=2
n
Note that
√ √ √
3n2 − 7 3n2 3
0 < an = 3
< 3
= 2
n n n
∞ √
3
P
for all n ≥ 2. As the series n2
converges, the comparison test says that
n=2

663
A NSWERS TO E XERCISES

∞ √
3n2 −7
P
n3
converges too.
n=2

3.3.11.32. *. Answer. The series converges.

3.3.11.33. *. Answer. It diverges.
3.3.11.34. *. Answer. (a) diverges (b) converges
3.3.11.35. *. Answer. The series diverges.
3.3.11.36. *. Answer. (a) converges (b) diverges
1
3.3.11.37. Answer.
e5 − e4
1
3.3.11.38. *. Answer. 7

3.3.11.39. *. Answer. (a) diverges by limit comparison with the harmonic series
(b) converges by the ratio test
1
3.3.11.40. *. Answer. (a) Converges by the limit comparison test with b = k5/3
.
(b) Diverges by the ratio test.
(c) Diverges by the integral test.
3.3.11.41. *. Answer. It converges.
3.3.11.42. *. Answer. N =5
3.3.11.43. *. Answer. N ≥ 999
1 1 1 1
3.3.11.44. *. Answer. We need N = 4 and then S4 = 32
− 52
+ 72
− 92

Exercises — Stage 3
3.3.11.45. *. Answer. (a) converges (b) converges
3.3.11.46. *. Answer. (a) See the solution.
x + sin x
(b) f (x) = is not a decreasing function.
1 + x2
(c) See the solution.
3.3.11.47. *. Answer. The sum is between 0.9035 and 0.9535.
3.3.11.48. *. Answer. Since lim an = 0, there must be some integer N such that
n→∞
1
2
> an ≥ 0 for all n > N . Then, for n > N ,
an an
≤ = 2an
1 − an 1 − 1/2

From the information in the problem statement, we know

∞
X ∞
X
2an = 2 an converges.
n=N +1 n=N +1

664
A NSWERS TO E XERCISES

So, by the direct comparison test,

∞
X an
converges as well.
n=N +1
1 − an

Since the convergence of a series is not affected by its first N terms, as long as N
is finite, we conclude
∞
X an
converges.
n=1
1 − an

3.3.11.49. *. Answer. It diverges.

3.3.11.50. *. Answer. It converges to − log 2 = log 21 ,
3.3.11.51. *. Answer. See the solution.
3.3.11.52. Answer. About 9% to 10%
3.3.11.53. Answer. The total population is between 29,820,091 and 30,631,021
people.

3.4 · Absolute and Conditional Convergence

3.4.3 · Exercises

Exercises — Stage 1
1
3.4.3.1. *. Answer. False. For example, bn = n
provides a counterexample.
3.4.3.2. Answer.
P P
an converges an diverges
P
|a | converges converges absolutely not possible
P n
|an | diverges converges conditionally diverges

Exercises — Stage 2
3.4.3.3. *. Answer. Conditionally convergent
3.4.3.4. *. Answer. The series diverges.
3.4.3.5. *. Answer. It diverges.
3.4.3.6. *. Answer. It converges absolutely.
3.4.3.7. *. Answer. It converges absolutely.
3.4.3.8. *. Answer. It diverges.
3.4.3.9. *. Answer. It converges absolutely.
3.4.3.10. Answer. See solution.

665
A NSWERS TO E XERCISES

3.4.3.11. Answer. See solution.

3.4.3.12. Answer. See solution.

Exercises — Stage 3
3.4.3.13. *. Answer. (a) See the solution.
3
(b) |S − S5 | ≤ 24 × 36e−6
389
3.4.3.14. Answer. cos 1 ≈ 720
; the actual associated error (using a calculator) is
about 0.000025.
3.4.3.15. Answer. See solution.

3.5 · Power Series

3.5.3 · Exercises

Exercises — Stage 1
3.5.3.1. Answer. 2
∞
X n(x − 5)n−1
3.5.3.2. Answer. f (x) =
n=1
n! + 2

3.5.3.3. Answer. Only x = c

3.5.3.4. Answer. R=6

Exercises — Stage 2
1
3.5.3.5. *. Answer. (a) R =
2
2 1
(b) for all |x| <
1 + 2x 2
3.5.3.6. *. Answer. R=∞
3.5.3.7. *. Answer. 1
3.5.3.8. *. Answer. The interval of convergence is −1 < x + 2 ≤ 1 or (−3, −1].
3.5.3.9. *. Answer. The interval of convergence is −4 < x ≤ 2, or simply (−4, 2].
3.5.3.10. *. Answer. −3 ≤ x < 7 or [−3, 7)
3.5.3.11. *. Answer. The given series converges if and only if −3 ≤ x ≤ −1.
Equivalently, the series has interval of convergence [−3, −1].

3.5.3.12. *. Answer. The interval of convergence is 34 ≤ x < 45 , or 34 , 54 .



3.5.3.13. *. Answer. The radius of convergence is 2. The interval of convergence

is −1 < x ≤ 3, or − 1, 3 .

666
A NSWERS TO E XERCISES

3.5.3.14. *.
Answer. The interval of convergence is a − 1 < x < a + 1, or

a − 1, a + 1 .

3.5.3.15. *. Answer. (a) |x + 1| ≤ 9 or −10 ≤ x ≤ 8 or [−10, 8]

(b) This series converges only for x = 1.
∞
X ∞
X
n+3
3.5.3.16. *. Answer. x = xn
n=0 n=3

∞
X (x − 1)n
3.5.3.17. Answer. f (x) = 3 +
n=1
n(n + 1)

Exercises — Stage 3
3.5.3.18. *. Answer. The series converges absolutely for |x| < 9, converges
conditionally for x = −9 and diverges otherwise.
∞ 3n+1
n x
X
3.5.3.19. *. Answer. (a) (−1) +C
n=0
3n + 1
(b) We need to keep two terms (the n = 0 and n = 1 terms).
3.5.3.20. *. Answer. (a) See the solution.
∞
X x(1 + x)
(b) n 2 xn = 3
. The series converges for −1 < x < 1.
n=0
(1 − x)

3.5.3.21. *. Answer. See the solution.

3.5.3.22. *. Answer. (a) 1.
(b) The series converges for −1 ≤ x < 1, i.e. for the interval [−1, 1)
4
3.5.3.23. Answer.
9
3.5.3.24. Answer. The point x = c corresponds to a local maximum if A2 < 0
and a local minimum if A2 > 0.
13
3.5.3.25. Answer.
80
x2 x3 x 4
3.5.3.26. Answer. x− + −
2 3 4
x3 x5
3.5.3.27. Answer. x− +
3 5

3.6 · Taylor Series

3.6.7 · Exercises

Exercises — Stage 1

667
A NSWERS TO E XERCISES

3.6.7.1. Answer. A: linear

B: constant
C: quadratic
3.6.7.2. Answer. T (5) = arctan3 (e5 + 7)
3.6.7.3. Answer. A-V
B-I
C - IV
D - VI
E - II
F - III
 
(20) 2 20!
3.6.7.4. Answer. (a) f (3) = 20
  20! + 1
(20) 2 20!
(b) g (3) = 10
10! + 1
22! · 513
(c) h(20) (0) = 0; h(22) (0) =
13

Exercises — Stage 2
∞
X (−1)n−1
3.6.7.5. Answer. (x − 1)n
n=1
n
∞
X (−1)n+1
3.6.7.6. Answer. (x − π)2n+1
n=0
(2n + 1)!
∞  n
1 X 10 − x
3.6.7.7. Answer. with interval of convergence (0, 20).
10 n=0 10
∞
X 3n e3a
3.6.7.8. Answer. (x − a)n , with infinite radius of convergence
n=0
n!

∞
X
3.6.7.9. *. Answer. − 2n xn
n=0

3.6.7.10. *. Answer. bn = 3(−1)n + 2n

35
3.6.7.11. *. Answer. c5 =
5!
∞
X 2n+1 xn+1 1
3.6.7.12. *. Answer. (−1)n for all |x| < 2
n=0
n+1

668
A NSWERS TO E XERCISES

1 1
3.6.7.13. *. Answer. a = 1, b = − =− .
3! 6
2
e−x − 1 x2 x4
Z
3.6.7.14. *. Answer. dx = C − + + · · ·.
x 2 8
It is not clear from the wording of the question whether or not the arbitrary con-
stant C is to be counted as one of the “first two nonzero terms”.
∞ ∞
X 22n+1 x2n+6 X 22n x2n+6
3.6.7.15. *. Answer. (−1)n +C = (−1)n +
n=0
(2n + 1)(2n + 6) n=0
(2n + 1)(n + 3)
C
∞
X 3n
3.6.7.16. *. Answer. f (x) = 1 + (−1)n x3n+2
n=0
3n + 2

π
3.6.7.17. *. Answer. √
2 3
1
3.6.7.18. *. Answer.
e
3.6.7.19. *. Answer. e1/e
3.6.7.20. *. Answer. e1/π − 1
3.6.7.21. *. Answer. log(3/2)
3.6.7.22. *. Answer. (e + 2)ee − 2
3.6.7.23. Answer. The sum diverges — see the solution.
√
1+ 2
3.6.7.24. Answer. √
2
3.6.7.25.
 *. Answer.
 (a) See the solution.
1 1
(b) e+
2 e
3.6.7.26. Answer. (a) 50,000
(b) three terms (n = 0 to n = 2)
(c) six terms (n = 0 to n = 5)
3.6.7.27. Answer. 29
3.6.7.28. Answer. S13 or higher
3.6.7.29. Answer. S9 or higher
3.6.7.30. Answer. S18 or higher
−57 −57
   
1
3.6.7.31. Answer. The error is in the interval 1+ 7 , ≈
14 · 37 3 7 · 67
(−0.199, −0.040)

669
A NSWERS TO E XERCISES

Exercises — Stage 3
3.6.7.32. *. Answer. −1
1 1
3.6.7.33. *. Answer. =
5! 120
3.6.7.34. Answer. e2
√
3.6.7.35. Answer. e
2 343
3.6.7.36. Answer. =
(6/7)3 108
∞
X x2n+4 x2
3.6.7.37. Answer. (−1)n = x3 arctan x − log(1 + x2 )
n=0
(2n + 1)(2n + 2) 2
∞
X (2n)! n
3.6.7.38. Answer. (a) the Maclaurin series for f (x) is 2n 2
x , and its
n=0
2 (n!)
radius of convergence is R = 1.
∞
X (2n)!
(b) the Maclaurin series for arcsin x is 2n (n!)2 (2n + 1)
x2n+1 , and its radius of
n=0
2
convergence is R = 1.
∞
X (−1)n−1
3.6.7.39. *. Answer. log(x) = log 2 + (x − 2)n . It converges when
n=1
n 2n
0 < x ≤ 4.
∞
X x4n+1
3.6.7.40. *. Answer. (a) (−1)n
n=0
4n + 1
(b) 0.493967
(c) The approximate value of part (b) is larger than the true value of I(1/2)
1
3.6.7.41. *. Answer.
66
3.6.7.42. *. Answer. Any interval of length 0.0002 that contains 0.03592 and
0.03600 is fine.
∞
X xn
3.6.7.43. *. Answer. (a) (−1)n
n=1
n n!
(b)−0.80
(c) See the solution.
∞
X x2n+1
3.6.7.44. *. Answer. (a) Σ(x) = (−1)n
n=0
(2n + 1)(2n + 1)!
(b) x = π
(c) 1.8525

670
A NSWERS TO E XERCISES

∞
X x2n−1
3.6.7.45. *. Answer. (a) I(x) = (−1)n
n=1
(2n)!(2n − 1)
1 1 1
(b) I(1) = − + ± = −0.486 ± 0.001
2 4!3 6!5
1 1
(c) I(1) < − +
2 4!3
∞
X x2n−1 1 1 1 1
3.6.7.46. *. Answer. (a) I(x) = (−1)n+1 = x − x3 + x5 − x8 + · · ·
n=1
(2n)! 2! 4! 6! 8!
(b) 0.460
1 1 1
(c) I(1) < − + < 0.460
2! 4! 6!
3.6.7.47. *. Answer. (a) See the solution.
(b) The series converges for all x.
3.6.7.48. *. Answer. See the solution.
∞ ∞
X xn X x2n
3.6.7.49. *. Answer. (a) cosh(x) = = for all x.
n=0
n! n=0
(2n)!
n even
(b), (c) See the solution.
√
3.6.7.50. Answer. (a) 3 3 ≈ 1.26
(b) 12 terms (S11 )
15! 21! 27! 33!
3.6.7.51. Answer. − + −
5! · 56 7! · 11! · 511 9! · 17! · 517 11! · 23! · 523
3.6.7.52. Answer. (a)
y

1 y = f (x)

x
p p
− 2/3 2/3

(b) the constant function 0

(c) everywhere
(d) only at x = 0
3.6.7.53. Answer. 0

671
 Appendix D

S OLUTIONS TO E XERCISES

1 · Integration
1.1 · Definition of the Integral
1.1.8 · Exercises

Exercises — Stage 1
1.1.8.1. Solution.
y

1.25
0.75
x
1 3

1.25
0.75
x
1 3

The diagram on the left shows a rectangle with area 2 × 1.25 = 2.5 square units.
Since the blue-shaded region is entirely inside this rectangle, the area of the blue-

672
S OLUTIONS TO E XERCISES

shaded region is no more than 2.5 square units.

The diagram on the right shows a rectangle with area 2 × 0.75 = 1.5 square units.
Since the blue-shaded region contains this entire rectangle, the area of the blue
region is no less than 1.5 square units.
So, the area of the blue-shaded region is between 1.5 and 2.5 square units.
Remark: we could also give an obvious range, like “the shaded area is between
zero and one million square units.” This would be true, but not very useful or
interesting.
1.1.8.2. Solution.

• Solution 1: One naive way to solve this is to simply use the same method
as Question 1.

2.25
1.75
1.25
0.75
0.25 x
1 2 3 4

2.25
1.75
1.25
0.75
0.25 x
1 2 3 4

The rectangle on the left has area 3 × 2.25 = 6.75 square units, and en-
compasses the entire shaded region. The rectangle on the right has area
3 × 0.25 = 0.75 square units, and is entirely contained inside the blue-
shaded region. So, the area of the blue-shaded region is between 0.75 and
6.75 square units.
This is a legitimate approximation, but we can easily do much better. The
shape of this graph suggests that using the areas of three rectangles would
be a natural way to improve our estimate.

• Solution 2: Let’s use these rectangles instead:

673
S OLUTIONS TO E XERCISES

y y

2.25 2.25
1.75 1.75
1.25 1.25
0.75 0.75
0.25 x 0.25 x
1 2 3 4 1 2 3 4

In the left picture, the red area is (1 × 1.25) + (1 × 2.25) + (1 × 0.75) = 4.25
square units. In the right picture, the red area is (1 × 0.75) + (1 × 1.75) + (1 ×
0.25) = 2.75 square units. So, the blue shaded area is between 2.75 and 4.25
square units.
1.1.8.3. Solution. Remark: in the solution below, we find the appropriate ap-
proximation using trial and error. In Question 46, we take a more systematic
approach.

• Try 1: First, we can try by using a single rectangle as an overestimate, and

a single rectangle as an underestimate.

y y

1/2 1/2

1 1
1/8 y= 2x 1/8 y= 2x

x x
1 3 1 3

The area under the curve is less than the area of the rectangle on the left
(2 × 12 = 1) and greater than the area of the rectangle on the right (2 × 81 = 14 ).
So, the area is in the range 14 , 1 . Unfortunately, this range is too big–we

need our range to have length at most 0.2. So, we refine our approximation
by using more rectangles.

• Try 2: Let’s try using two rectangles each for the upper and lower bounds.

674
S OLUTIONS TO E XERCISES

y y

1/2 1/2

1/4 1/4
1 1
1/8 y= 2x 1/8 y= 2x

x x
1 2 3 1 2 3

1
+ 1
× 14 = 34 , and the



The rectangles in the left picture have area 1 × 2
1 × 14 + 1 × 18 = 83 . So, the area


rectangles in the right picture have area
under the curve is in the interval 38 , 34 . The length of this interval is 38 , and


3 3
8
> 15 = 15 = 0.2. (Indeed, 38 = 0.375 > 0.2.) Since the length of our interval
is still bigger than 0.2, we need even more rectangles.

• Try 3: Let’s go ahead and try four rectangles each for the upper and lower
estimates.

y y

1/2 1/2
√ √
1/(2 2) 1/(2 2)
√ 1/4 √ 1/4
1/(4 2) 1 1/(4 2) 1
1/8 y= 2x 1/8 y= 2x

x x
1 2 3 1 2 3

The area of the rectangles on the left is:

       
1 1 1 1 1 1 1 1
× + × √ + × + × √
2 2 2 2 2 2 4 2 4 2
 
3 1
= 1+ √ ,
8 2
and the area of the rectangles on the right is:
       
1 1 1 1 1 1 1 1
× √ + × + × √ + ×
2 2 2 2 4 2 4 2 2 8
 
3 1 1
= +√ .
8 2 2
 h i h i
So, the area under the curve is in the interval 38 21 + √12 , 38 1 + √1
2
. The

675
S OLUTIONS TO E XERCISES

3 3 3 1
length of this interval is 16
, and 16
< 15
= 5
= 0.2, as desired. (Indeed,
3
16
= 0.1875 < 0.2.)
 h i h i
Note, if we choose any value in the interval 83 12 + √12 , 38 1 + √12 as an
approximation for the area under the curve, our error is no more than 0.2.
1.1.8.4. Solution. Since f (x) is decreasing, it is larger on the left endpoint of an
interval than on the right endpoint of an interval. So, a left Riemann sum gives a
larger approximation. Notice this does not depend on n.
Z 5
Furthermore, the actual area f (x)dx is larger than its right Riemann sum, and
0
smaller than its left Riemann sum.
y y

x x
left Riemann sum right Riemann sum

1.1.8.5. Solution. If f (x) is always increasing or always decreasing, then the

midpoint Riemann sum will be between the left and right Riemann sums. So, we
need a function that goes up and down. Many examples are possible, but let’s
work with a familiar one: sin x.
If our intervals have endpoints that are integer multiples of π, then the left and
right Riemann sums will be 0, since sin(0) = sin(π) = sin(2π) = · · · = 0. The
midpoints of these intervals will give y-values of 1 and -1. So, for example, we
can let f (x) = sin x, [a, b] = [0, π], and n = 1. Then the right and left Riemann
sums are 0, while the midpoint Riemann sum is π.
We can extend the example of f (x) = sin x to have more intervals. As long as
we have more positive terms than negative, the midpoint approximation will be
a positive number, and so it will be larger than both the left and right Riemann
sums. So, for example, we can let f (x) = sin x, [a, b] = [0, 5π], and n = 5. Then
the midpoint Riemann sum is π − π + π − π + π = π, which is strictly larger than
0 and so it is larger than both the left and right Riemann sums.

676
S OLUTIONS TO E XERCISES

x
5π

1.1.8.6. Solution.
7
X 5
X
a Two possible answers are i and (i + 2). The first has simpler terms (i
i=3 i=1
versus i + 2), while the second has simpler indices (we often like to start at
i = 1). Neither is objectively better than the other, but depending on your
purposes you might find one more useful.

b The terms of this sum are each double the terms of the sum from part (a),
X7 X5
so two possible answers are 2i and (2i + 4). We often want to write a
i=3 i=1
sum that involves even numbers: it will be useful for you to remember that
the term 2i (with index i) generates evens.

c The terms of this sum are each one more than the terms of the sum from
7
X 5
X
part (b), so two possible answers are (2i + 1) and (2i + 5).
i=3 i=1
In the last part, we used the expression 2i to generate even numbers; 2i + 1
will generate odds. So will the index 2i + 5, and indeed, 2i + k for any odd
number k. The choice of what you add will depend on the bounds of i.

d This sum adds up the odd numbers from 1 to 15. From Part (c), we know
that the formula 2i+1 is a simple way of generating odd numbers.PSince our
first term should be 1 and our last term should be 15, if we use (2i + 1),
then i should run from 0 to 7. So, one way of expressing our sum in sigma
X7
notation is (2i + 1).
i=0
Sometimes we like our sum to start at i = 1 instead of i = 0. If this is our
desire, we can use 2i − 1 as our terms, and let i run from 1 to 8. This gives
X8
us another way of expressing our sum: (2i − 1).
i=1

1.1.8.7. Solution.

a The denominators are successive powers of three, so one way of writing

677
S OLUTIONS TO E XERCISES

4
X 1
this is i
. Equivalently, the terms we’re adding are powers of 1/3, so we
i=1
3
4  i
X 1
can also write .
i=1
3

b This sum is obtained from the sum in (a) by multiplying each term by two,
4 4  i
X 2 X 1
so we can write i
or 2 .
i=1
3 i=1
3

c The difference between this sum and the previous sum is its alternating
sign, minus-plus-minus-plus. This behaviour appears when we raise a neg-
ative number to successive powers. We can multiply each term by (−1)i , or
we can slip a negative into the number that is already raised to the power
4 4
i 2 2
X X
i: (−1) i , or .
i=1
3 i=1
(−3)i

d This sum is the negative of the sum in part (c), so we can simply multiply
4 4
i+1 2 2
X X
each term by negative one: (−1) i
or − .
i=1
3 i=1
(−3)i
Be careful with the second form: a common mistake is to think that
2 2
− = , but these are not the same.
(−3)i 3i
1.1.8.8. Solution.
3
a If we re-write the second term as 9
instead of 13 , our sum becomes:

1 3 5 7 9
+ + + +
3 9 27 81 243

The numerators are the first five odd numbers, and the denominators are
the first five positive powers of 3. We learned how to generate odd num-
bers in Question 6, and we learned how to generate powers of three in
5
X 2i − 1
Question 7. Combining these, we can write our sum as .
i=1
3i

b The denominators of these terms differ from the denominators of part (a)
by precisely two, while the numerators are simply 1. So, we can modify
5
X 1
our previous answer: i+2
.
i=1
3

c Let’s re-write the sum to make the pattern clearer.

678
S OLUTIONS TO E XERCISES

1000 + 200 + 30 + 4
1 3 7
+ 2
+ 50
+ 1000
4
= 1 · 1000 + 2 · 100 + 3 · 10 + 1
5 6 7
+ 10
+ 100
+ 1000
= 1 · 103 + 2 · 102 + 3 · 101 + 4 · 100
+ 5 · 10−1 + 6 · 10−2 + 7 · 10−3
= 1 · 104−1 + 2 · 104−2 + 3 · 104−3 + 4 · 104−4
+ 5 · 104−5 + 6 · 104−6 + 7 · 104−7
7
X
If we let the red numbers be our index i, this gives us the expression i · 104−i .
i=1
7
X i
Equivalently, we can write .
i=1
10i−4

1.1.8.9. Solution.
a Using Theorem 1.1.6, part (a) with a = 1, r = 35 and n = 100:
100  i
101 "  101 #
X 3 1 − 35 5 3
= 3 = 1−
i=0
5 1− 5 2 5

b We want to use Theorem 1.1.6, part (a) again, but our sum doesn’t start at
3 0


5
= 1. We have two options: factor out the leading term, or use the
difference of two sums that start where we want them to.
• Solution 1: In this solution, we’ll make our sum start at 1 by factoring
out the leading term. We wrote our work out the long way (expanding
the sigma into “dot-dot-dot” notation) for clarity, but it’s faster to do
the algebra in sigma notation all the way through.
100  i  50  51  52  100
X 3 3 3 3 3
= + + + ··· +
i=50
5 5 5 5 5
 50 "    2  50 #
3 3 3 3
= 1+ + + ··· +
5 5 5 5
 50
51
3 1 − 35
=
5 1 − 35
 50 "  51 #
5 3 3
= 1− .
2 5 5

• Solution 2: In this solution, we write our given expression as the dif-

ference of two sums, both starting at i = 0.
100  i 100  i 49  i
X 3 X 3 X 3
= −
i=50
5 i=0
5 i=0
5

679
S OLUTIONS TO E XERCISES

101
50
1 − 35 1 − 53
= −
1 − 35 1 − 35
"   101 #
50
5 3 3
= −
2 5 5
 50 "  51 #
5 3 3
= 1− .
2 5 5

c Before we can use the equations in Theorem 1.1.6, we’ll need to do a little
simplification.
10
X 10
X 10
X 10
X
2 2


i − 3i + 5 = i + −3i + 5
i=1 i=1 i=1 i=1
X10 10
X X10
= i2 − 3 i+5 1
i=1 i=1 i=1
 
1 1
= (10)(11)(21) − 3 (10 · 11) + 5 · 10
6 2
= 270

d As in part (c), we’ll simplify first. The first part (shown here in red) is a

0
geometric sum, but it does not start at 1 = 1e .
b  n  b  n b
X 1 3
X 1 X
+ en = + en3
n=1
e n=1
e n=1
b  n b
X 1 X
= −1+e n3
n=0
e n=1
b+1 2
1 − 1e


1
= − 1 + e b(b + 1)
1 − 1e 2
1 b+1 2
− 1e


e 1
= + e b(b + 1)
1 − 1e 2
b
1 − 1e


e
= + [b(b + 1)]2
e−1 4
1.1.8.10. Solution.
a The two pieces are very similar, which we can see by changing the index,
or expanding them out:
100
X 50
X
(i − 50) + i
i=50 i=0
= (0 + 1 + 2 + · · · + 50) + (0 + 1 + 2 + · · · + 50)

680
S OLUTIONS TO E XERCISES

= (1 + 2 + · · · + 50) + (1 + 2 + · · · + 50)
= 2 (1 + 2 + · · · + 50)
50
X
=2 i
i=1
 
50 · 51
=2 = 50 · 51 = 2550
2

b If we expand (i − 5)3 = i3 − 15i2 + 75i − 225, we can break the sum into four
parts, and evaluate each separately. However, it is much simpler to change
the index and make the term (i − 5)3 into i3 .
100
X
(i − 5)3 = 53 + 63 + 73 + · · · + 953
i=10

We have a formula to evaluate the sum of cubes if they start at 1, so we turn

our expression into the difference of two sums starting at 1:

= 13 + 23 + 33 + 43 + 53 + 63 + 73 + · · · + 953
 

− 13 + 23 + 33 + 43
 

95
X 4
X
3
= i − i3
i=1 i=1
 2  2
1 1
= (95)(96) − (4)(5) .
2 2

c Notice every two terms cancel with each other, since the sum is (−1)+(+1),
etc. Then the terms n = 1 through n = 10 cancel, and we’re left only with
the final term, (−1)11 = −1.
Written out more explicitly:
11
X
(−1)n = −1 + 1 − 1 + 1 − 1 + 1 − 1 + 1 − 1 + 1 − 1
n=1
= [−1 + 1] + [−1 + 1] + [−1 + 1] + [−1 + 1] + [−1 + 1] − 1
= 0 + 0 + 0 + 0 + 0 − 1 = −1.

11
X
d For every integer n, 2n + 1 is odd, so (−1)2n+1 = −1. Then (−1)2n+1 =
n=2
11
X
−1 = −10.
n=2

681
S OLUTIONS TO E XERCISES

1.1.8.11. Solution. The index of the sum runs from 1 to 4: the first, second,
third, and fourth rectangles. So, we have four rectangles in our Riemann sum.
Let’s start by drawing in the intervals along the x-axis taken up by these four
b−a
rectangles. Note each has the same width: .
4

y
y = f (x)

x
a b

Since this is a midpoint Riemann sum, the height of each rectangle is given by
the y-value of the function in the midpoint of the interval. So, now let’s find the
height of the function at the midpoints of each of the four intervals.

y
y = f (x)

x
a b

The left-most interval has a height of about 0, so it gives a “trivial” rectangle

with no height and no area. The middle two intervals have rectangles of about
the same height, and the right-most interval has the highest rectangle.

682
S OLUTIONS TO E XERCISES

y
y = f (x)

x
a b

1.1.8.12. *. Solution. In general, the {left} Riemann sum for the integral
Rb
a
f (x) dx is of the form
n  
X b−a b−a
f a + (k − 1)
k=1
n n

• To get the limits of summation to match the given sum, we need n = 4.

• Then to get the factor multiplying f to match that in the given sum, we
need b−a
n
= 1, so b − a = 4.

• Finally, to get the argument of f to match that in the given sum, we need

b−a b−a b−a

a + (k − 1) =a− +k =1+k
n n n
Subbing in n = 4 and b − a = 4 gives a − 1 + k = 1 + k, so a = 2 and b = 6.
n
X
1.1.8.13. Solution. The general form of a Riemann sum is ∆x · f (x∗i ), where
i=1
∆x = b−an
is the width of each rectangle, and f (x∗i ) is the height.
There are different ways to interpret the given sum as a Riemann sum. The most
obvious is given in Solution 1. You may notice that we make some convenient
assumptions in this solution about values for ∆x and a, and we assume the sum
is a right Riemann sum. Other visualizations of the sum arise from making more
exotic choices. Some of these are explored in Solutions 2-4.
All cases have three rectangles, and the three rectangles will have the same areas:
98, 162, and 242 square units, respectively. This is because the terms of the given
sum simplify to 98 + 162 + 242.

• Solution 1:

◦ Because the index runs from 1 to 3, there are three intervals: n = 3.

683
S OLUTIONS TO E XERCISES

◦ Looking at our sum, it seems reasonable to interpret ∆x = 2. Then,

since n = 3, we conclude b−a
3
= 2, hence b − a = 6.
◦ If ∆x = 2, then f (x∗i ) = (5 + 2i)2 . Recall that x∗i is the x-coordinate we
use to decide the height of the ith rectangle. In a right Riemann sum,
x∗i = a+i·∆x. So, using 2 = ∆x, we can let f (x∗i ) = f (a+2i) = (5 + 2i)2 .
This fits with the function f (x) = x2 , and a = 5.
◦ Since b − a = 6, and a = 5, this tells us b = 11
To sum up, we can interpret the Riemann sum as a right Riemann sum,
with three intervals, of the function f (x) = x2 from x = 5 to x = 11.

y = x2

121

81

49

x
5 7 9 11

• Solution 2: We could have chosen a different value for ∆x.

◦ The index of the sum runs from 1 to 3, so we have n = 3.
◦ We didn’t have to interpret ∆x as 2–that was just the path of least
resistance. We could have chosen it to be any other number–for the
sake of argument, let’s say ∆x = 10. (Positive numbers are easiest to
interpret, but negatives are technically allowed as well.)
b−a b−a
◦ Then 10 = n
= 3
, so b − a = 30.
◦ Let’s use the paradigm of a right Riemann sum, and match up the
terms of the sum given in the problem to the terms in the definition:

∆x · f (a + i · ∆x) = 2 · (5 + 2i)2
10 · f (a + 10i) = 2 · (5 + 2i)2
1
f (a + 10i) = · (5 + 2i)2
5
 2
1 1
f (a + 10i) = · 5 + · 10i
5 5
1
◦ The easiest value of a in this case is a = 0. Then f (10i) = 5
·

2
2
5 + 15 · 10i , so f (x) = 15 · 5 + 15 · x .

684
S OLUTIONS TO E XERCISES

◦ If a = 0 and b − a = 30, then b = 30.

1 x 2


◦ To sum up: n = 3, a = 0, b = 30, ∆x = 10, and f (x) = 5
· 5+ 5
.

y 1 x 2


y= 5
5+ 5
121
5
81
5
49
5

x
10 20 30

By changing ∆x, we changed the widths of the rectangles. The rectan-

gles in this picture are wider and shorter than the rectangles in Solu-
tion 1. Their areas are the same: 98, 162, and 242.

• Solution 3: We could have chosen a different value of a.

◦ Suppose ∆x = 2, and we interpret our sum as a right Riemann sum,

but we didn’t assume a = 5. We could have chosen a to be any
number–say, a = 1.
◦ Let’s match up what we’re given in the problem to what we’re given
as a definition:

∆x · f (a + i · ∆x) = 2 · (5 + 2i)2
2 · f (1 + 2i) = 2 · (5 + 2i)2
f (1 + 2i) = (5 + 2i)2
f (1 + 2i) = (4 + 1 + 2i)2

◦ Since f (1 + 2i) = (4 + 1 + 2i)2 , we have f (x) = (4 + x)2

b−a
◦ Since a = 1 and 3
= 2, in this case b = 7.
◦ To sum up: n = 3, a = 1, b = 7, ∆x = 2, and f (x) = (4 + x)2 .

685
S OLUTIONS TO E XERCISES

y = (4 + x)2

121

81

49

x
1 3 5 7

This picture is a lot like the picture in Solution 1, but shifted to the left.
By changing a, we changed the left endpoint of our region.

• Solution 4: We could have chosen a different kind of Riemann sum.

◦ We didn’t have to assume that we were dealing with a right Riemann

sum. Suppose ∆x = 2, and we have a midpoint Riemann sum.
◦ Let’s match up what we’re given in the problem with what we’re given
in the definition:

∆x · f a + i − 21 ∆x = 2 · (5 + 2i)2

2 · f a + i − 12 2 = 2 · (5 + 2i)2

f a + i − 12 2 = (5 + 2i)2

f (a + 2i − 1) = (5 + 2i)2
f ((a − 1) + 2i) = (5 + 2i)2

◦ It is now convenient to set a − 1 = 5, hence a = 6.

◦ Then f (5 + 2i) = (5 + 2i)2 , so f (x) = x2
b−a
◦ Since 2 = 3
and a = 6, we see b = 12.
◦ To sum up: n = 3, a = 6, b = 12, ∆x = 2, and f (x) = x2 .

686
S OLUTIONS TO E XERCISES

y = x2
121

81

49

x
6 8 10 12

By choosing to interpret our sum as a midpoint Riemann sum instead

of a right Riemann sum, we changed where our rectangles intersect
the graph y = f (x): instead of the graph hitting the right corner of the
rectangle, it hits in the middle.
1.1.8.14. Solution. Many interpretations are possible–see the solution to Ques-
tion 13 for a more thorough discussion–but the most obvious is given below.
Recall the definition of a left Riemann sum:
n
X
∆x · f (a + (i − 1)∆x)
i=1

We chose a left Riemann sum instead of right or midpoint because our given sum
has (i − 1) in it, rather than (i − 21 ) or simply i.

• Since the sum has five terms (i runs from 1 to 5), there are 5 rectangles. That
is, n = 5.

• In the definition of the Riemann sum, note that the term ∆x appears twice:
once multiplied by the entire term, and once multiplied by i−1. So, a conve-
π
nient choice for ∆x is 20 , because this is the constant that is both multiplied
at the start of the term, and multiplied by i − 1.
π b−a b−a 5π π
• Since = ∆x = = , we see b − a = = .
20 n 5 20 4
• We match the terms in the definition with the terms in the problem:
 
π(i − 1)
f (a + (i − 1)∆x) = tan
20
 π  π
f a + (i − 1) = tan (i − 1)
20 20
So, we choose a = 0 and f (x) = tan x.

• Since a = 0 and b − a = π4 , we see b = π4 .

687
S OLUTIONS TO E XERCISES

y y = tan x

x
π 2π 3π 4π 5π
20 20 20 20 20

We note that the first rectangle of the five is a “trivial” rectangle, with height (and
area) 0.
1.1.8.15. *. Solution. Since there are four terms in the sum, n = 4. (Note the
sum starts at k = 0, instead of k = 1.) Since the function is multiplied by 1,
b−a b−a
1 = ∆x = = , hence b − a = 4.
n 4
We can choose to view the given sum as a left, right, or midpoint Riemann sum.
The choice we make determines the interval. Note that the heights of the rectan-
gles are determined when x = 1.5, 2.5, 3.5, and 4.5.

f (4.5)

f (3.5)

f (2.5)

f (1.5)

• Option 1: right Riemann sum. If our sum is a right Riemann sum, then we
take the heights of the rectangles from the right endpoint of each interval.

688
S OLUTIONS TO E XERCISES

f (4.5)

f (3.5)

f (2.5)

f (1.5)

x
1.5 2.5 3.5 4.5

3
P
Then a = 0.5 and b = 4.5. Therefore: f (1.5 + k) · 1 is a right Riemann
k=0
sum on the interval [0.5, 4.5] with n = 4.

• Option 2: left Riemann sum. If our sum is a left Riemann sum, then we take
the heights of the rectangles from the left endpoint of each interval.

f (4.5)

f (3.5)

f (2.5)

f (1.5)

x
1.5 2.5 3.5 4.5

3
P
Then a = 1.5 and b = 5.5. Therefore: f (1.5 + k) · 1 is a left Riemann sum
k=0
on the interval [1.5, 5.5] with n = 4.

• Option 3: midpoint Riemann sum. If our sum is a midpoint Riemann sum,

then we take the heights of the rectangles from the midpoint of each inter-
val.

689
S OLUTIONS TO E XERCISES

f (4.5)

f (3.5)

f (2.5)

f (1.5)

x
1.5 2.5 3.5 4.5

3
P
Then a = 1 and b = 5. Therefore: f (1.5 + k) · 1 is a midpoint Riemann
k=0
sum on the interval [1, 5] with n = 4.
1.1.8.16. Solution. The area in question is a triangle with base 5 and height 5,
25
so its area is .
2
y
y=x
5

x
5

1.1.8.17. Solution. There is a positive and a negative portion of this area. The
25
positive area is a triangle with base 5 and height 5, so area square units. The
2
4
negative area is a triangle with base 2 and height 2, so negative area = 2 square
2
25 4 21
units. So, the net area is − = square units.
2 2 2

690
S OLUTIONS TO E XERCISES

y
y=x
5

2
x
2 5

Exercises — Stage 2
1.1.8.18. *. Solution. In general, the midpoint Riemann sum is given by
n
X  1
 b−a
f a + i − ∆x ∆x , where ∆x = .
i=1
2 n

In this problem we are told that f (x) = x8 , a = 5, b = 15 and n = 50, so that

∆x = b−an
= 15 and the desired Riemann sum is:
50 
X 1
1 8 1
5+ i−
i=1
2 5 5

1.1.8.19. *. Solution. The given integral has interval of integration going from
a = −1 to b = 5. So when we use three approximating rectangles, all of the same
width, the common width is ∆x = b−a n
= 2. The first rectangle has left endpoint
x0 = a = −1, the second has left hand endpoint x1 = a + ∆x = 1, and the third
has left hand end point x2 = a + 2∆x = 3. So
Z 5
x3 dx ≈ f (x0 ) + f (x1 ) + f (x2 ) ∆x = (−1)3 + 13 + 33 × 2 = 54
   
−1

1.1.8.20. *. Solution. In the given integral, the domain of integration runs from
a = −1 to b = 7. So, we have ∆x = (b−a)n
= (7−(−1))
n
= n8 . The left-hand end of the
first subinterval is at x0 = a = −1. So, the right-hand end of the ith interval is at
x∗i = −1 + 8in
. So:
7 n  
8i 8
Z X
f (x) dx = lim f −1 +
−1 n→∞
i=1
n n

1.1.8.21. *. Solution. We identify the given sum as the right Riemann sum
Pn
f (a + i∆x)∆x, with a = 0 (that’s specified in the statement of the question).
i=1
4 4
Since n
is multiplied in every term, and is also multiplied by i, we let ∆x = n
.

691
S OLUTIONS TO E XERCISES

Then x∗i = a + i∆x = 4i

n
and f (x) = sin2 (2 + x). So, b = a + n∆x = 0 + n · 4
n
= 4.
1.1.8.22. *. Solution. The given sum is of the form
s
n n    2
r
X k k2 X 1 k k
lim 1 − 2 = lim 1−
n→∞
k=1
n2 n n→∞
k=1
n n n
n
X
= lim ∆xf (x∗k )
n→∞
k=1
√
with ∆x = n1 , a = 0, x∗k = nk = a + k∆x and f (x) = x 1 − x2 . Since x∗0 = 0 and
x∗n = 1, the right hand side is the definition (using the right Riemann sum) of
R1
0
f (x) dx.
1.1.8.23. *. Solution. As i ranges from 1 to n, 3i/n range from 3/n to 3 with
jumps of ∆x = 3/n, so this is
n n Z b
X 3 −i/n X
∗
lim e cos(3i/n) = lim f (xi )∆x = f (x) dx
n→∞
i=1
n n→∞
i=1 a

where x∗i = 3i/n, f (x) = e−x/3 cos(x), a = x0 = 0 and b = xn = 3. Thus

n Z 3
X 3 −i/n
lim e cos(3i/n) = e−x/3 cos(x) dx
n→∞
i=1
n 0

i 1
1.1.8.24. *. Solution. As i ranges from 1 to n, the exponent n
ranges from n
to 1
with jumps of ∆x = n1 . So let’s try x∗i = ni , ∆x = n1 . Then:
n n n n
X iei/n X i i/n 1 X ∗ x∗i X
Rn = = e = xi e ∆x = f (x∗i )∆x
i=1
n2 i=1
n n i=1 i=1

with f (x) = xex , and the limit

n
X Z b
lim Rn = lim f (x∗i )∆x = f (x) dx
n→∞ n→∞ a
i=1

1
Since we chose x∗i = i
n
= 0 + i∆x, we let a = 0. Then n
= ∆x = b−a
n
= b
n
tells us
b = 1. Thus,
Z 1
lim Rn = xex dx .
n→∞ 0

1.1.8.25. *. Solution.
• Choice #1: If we set ∆x = n2 and x∗i = 2in
, i.e. x∗i = a + i∆x with a = 0, then
X n  X n 
−1−2i/n 2 −1−x∗i
lim e · = lim e ∆x
n→∞
i=1
n n→∞
i=1

692
S OLUTIONS TO E XERCISES

n
X 
= lim f (x∗i )∆x
n→∞
i=1
with f (x) = e−1−x
Z b
= f (x) dx
a
with a = x0 = 0 and b = xn = 2
Z 2
= e−1−x dx
0

• Choice #2: If we set ∆x = n2 and x∗i = 1 + 2i

n
, i.e. x∗i = a + i∆x with a = 1,
then
Xn  X n 
−1−2i/n 2 −x∗i
lim e · = lim e ∆x
n→∞
i=1
n n→∞
i=1
X n 
∗
= lim f (xi )∆x
n→∞
i=1
with f (x) = e−x
Z b
= f (x) dx
a
with a = x0 = 1 and b = xn = 3
Z 3
= e−x dx
1

• Choice #3: If we set ∆x = n1 and x∗i = ni , i.e. x∗i = a + i∆x with a = 0, then
X n  X n 
−1−2i/n 2 −1−2x∗i
lim e · = lim e 2∆x
n→∞
i=1
n n→∞
i=1
X n 
∗
= lim f (xi )∆x
n→∞
i=1
with f (x) = 2e−1−2x
Z b
= f (x) dx
a
with a = x0 = 0 and b = xn = 1
Z 1
=2 e−1−2x dx
0

• Choice #4: If we set ∆x = n1 and x∗i = 12 + ni , i.e. xi = a + i∆x with a = 12 ,

then
X n  X n 
−1−2i/n 2 −2xi
lim e · = lim e 2∆x
n→∞
i=1
n n→∞
i=1

693
S OLUTIONS TO E XERCISES

n
X 
= lim f (x∗i )∆x
n→∞
i=1
with f (x) = 2e−2x
Z b
= f (x) dx
a
1 3
with a = x0 = and b = xn =
2 2
Z 3/2
=2 e−2x dx
1/2

1.1.8.26. Solution. This is similar to the familiar form of a geometric sum, but
the powers go up by threes. So, we make a subsitution. If x = r3 , then:

1 + r3 + r6 + r9 + · · · + r3n = 1 + x + x2 + x3 + · · · + xn

Now, using Equation 1.1.3,

xn+1 − 1
1 + x + x2 + x3 + · · · + xn =
x−1
(r3 )n+1 − 1 r3n+3 − 1
Substituting back in x = r3 , we find our sum is equal to , or .
r3 − 1 r3 − 1
1.1.8.27. Solution. The sum does not start at 1, so we need to do some algebra.
We can either factor out the first term, or subtract off the initial terms that are
missing.

• Solution 1: If we factor out r5 , then what’s left fits the form of Equation 1.1.3:

r5 + r6 + r7 + · · · + r100 = r5 1 + r + r2 + · · · + r95
 
 96 
5 r −1
=r .
r−1

• Solution 2: We know how to evaluate sums of this form if they start at 1, so

we re-write our sum as follows:

r5 + r6 + r7 + · · · + r100 = 1 + r + r2 + r3 + r4 + r5 + · · · + r100

− 1 + r + r2 + r3 + r4

r101 − 1 r5 − 1
= −
r−1 r−1
r − 1 − r5 + 1
101
r101 − r5
= =
 96r −1 r−1
r − 1
= r5 .
r−1

694
S OLUTIONS TO E XERCISES

1.1.8.28. *. Solution. Recall that

(
−x if x ≤ 0
|x| =
x if x ≥ 0

so that
(
−2x if x ≤ 0
|2x| =
2x if x ≥ 0

To picture the geometric figure whose area the integral represents observe that

• at the left hand end of the domain of integration x = −1 and the integrand
|2x| = | − 2| = 2 and

• as x increases from −1 towards 0, the integrand |2x| = −2x decreases lin-

early, until

• when x hits 0 the integrand hits |2x| = |0| = 0 and then

• as x increases from 0, the integrand |2x| = 2x increases linearly, until

• when x hits +2, the right hand end of the domain of integration, the inte-
grand hits |2x| = |4| = 4.
R2
So the integral −1 |2x| dx is the area of the union of the two shaded triangles
(one of base 1 and of height 2 and the other of base 2 and height 4) in the figure
on the right below and
2
1 1
Z
|2x| dx = ×1×2+ ×2×4=5
−1 2 2

1.1.8.29. Solution. The area we want is two triangles, both  the x-axis.
 above
4·4
Each triangle has base 4 and height 4, so the total area is 2 · = 16.
2

695
S OLUTIONS TO E XERCISES

y
y = |t − 1|
4

x
−3 1 5

If you had a hard time sketching the function, recall that the absolute value of
a number leaves it unchanged if it is positive or zero, and flips the sign if it
is negative. So, when t − 1 ≥ 0 (that is, when t ≥ 1), our function is simply
f (t) = |t − 1| = t − 1. On the other hand, when t = 1 is negative (that is, when
t < 1), the absolute value changes the sign, so f (t) = |t − 1| = −(t − 1) = −t + 1.
1.1.8.30. Solution. The area we want is a trapezoid with base (b−a) and heights
(b − a)(b + a) b 2 − a2
a and b, so its area is = .
2 2
y
y=x
b

x
a b

Instead of using a formula for the area of a trapezoid, you can find the blue area
as the area of a triangle with base and height b, minus the area of a triangle with
base and height a.
1.1.8.31. Solution. The area is negative. The shape is a trapezoid with base
length (b − a) and heights 0 − a = −a and 0 − b = −b (note: those are nonnegative
(b − a)(−b − a) −b2 + a2
numbers), so its area is = . Since the shape is below the
2 2
b 2 − a2
x-axis, we change its sign. Thus, the integral evaluates to .
2

696
S OLUTIONS TO E XERCISES

a b y=x
x

The signs can be a little hard to keep track of. The base of our trapezoid is |a − b|;
since b > a, this is b − a. The heights of the trapezoid are |a| and |b|; since these
are both negative, |a| = −a and |b| = −b.
We note that this is the same result as in Question 30.
√
1.1.8.32. Solution.
√ If y = 16 − x2 , then y is nonnegative, and y 2 + x2 = 16. So,
the graph y = 16 − x2 is the upper half of a circle of radius 4. Since x only runs
from 0 to 4, we have a quarter of a circle of radius 4. Then the area under the
1
curve is [π · 42 ] = 4π.
4
y

√
y= 16 − x2
x

1.1.8.33. *. Solution. Here is a sketch the graph of f (x).

There is a linear increase from x = 0 to x = 1, followed by a constant. Using the

R3
interpretation of 0 f (x) dx as the area between y = f (x) and the x–axis with x
between 0 and 3, we can break this area into:

697
S OLUTIONS TO E XERCISES
R1
• f (x) dx: a right-angled triangle of height 1 and base 1 and hence area
0
0.5.
R3
• 1 f (x) dx: a rectangle of height 1 and base 2 and hence area 2.
R3
Summing up: 0 f (x) dx = 2.5.
1.1.8.34. *. Solution. The car’s speed increases with time. So its highest speed
on any time interval occurs at the right hand end of the interval and the best
possible upper estimate for the distance traveled is given by the right Riemann
sum with ∆x = 0.5, which is
   
v(0.5) + v(1.0) + v(1.5) + v(2.0) × 0.5 = 14 + 22 + 30 + 40 × 0.5
= 53 m
1.1.8.35. Solution. There is a key detail in the statement of Question 34: namely,
that the car is continuously accelerating. So, although we don’t know exactly
what’s going on in between our brief snippets of information, we know that the
car is not going any faster during an interval than at the end of that interval.
Therefore, the car certainly travelled no farther than our estimation.
We ask this question in order to point out an important detail. If we did not have
the information that the car was continuously accelerating, we would not be able
to give a certain upper bound on its distance travelled. It would be possible that,
when the car is not being observed (for example, when t = 0.25), it is going much
faster than when it is being observed.
1.1.8.36. Solution. First, note that the distance travelled by the plane is equal to
the area under the curve of its speed.
We need to know the speed of the plane at the midpoints of our intervals. So (for
example) noon to 1pm is not one of your intervals–we don’t know the speed at
12:30. (A common idea is to average the two end values, 700 and 800. This is a
fine approximation, but it is not a Riemann sum.) So, we use the two intervals
12:00 to 2:00, and 2:00 to 4:00. Then our intervals have length 2 hours, and at
the midpoints of the intervals the speed of the plane is 700 kph and 900 kph,
respectively. So, our midpoint Riemann sum gives us:

700(2) + 900(2) = 3200

an approximation of 3200 km travelled by the plane from noon to 4:00 pm.

Remark: if we had been asked to approximate the distance travelled from 11:30
am to 4:30 pm, then we could have used the midpoint rule with five intervals and
made use of every entry in the data table. With the question as stated, however,
we ignore three out of five entries in the table because they are not the midpoints
of our intervals.

Exercises — Stage 3

698
S OLUTIONS TO E XERCISES

1.1.8.37. *. Solution.
• Solution #1: Set x∗i = −2 + 2i
n
. Then a = x0 = −2 and b = xn = 0 and
2
∆x = n . So
s
n  2
X 2 2i
lim 4 − −2 +
n→∞
i=1
n n
n
X √ 2
= lim f (x∗i )∆x with f (x) = 4 − x2 and ∆x =
n→∞
i=1
n
Z 0 √
= 4 − x2 dx
−2

R0 √ √
For the integral −2 4 − x2 dx, y = 4 − x2 is equivalent to x2 + y 2 = 4,
y ≥ 0. So the integral represents the area between the upper half of the
circle x2 + y 2 = 4 (which has radius 2) and the x-axis with −2 ≤ x ≤ 0,
which is a quarter circle with area 14 · π 22 = π.

y
√
y= 4 − x2

x
−2

• Solution #2: Set x∗i = 2i

n
. Then a = x0 = 0 and b = xn = 2 and ∆x = n2 . So
s
n  2
X 2 2i
lim 4 − −2 +
n→∞
i=1
n n
n
X p 2
= lim f (x∗i )∆x with f (x) = 4 − (−2 + x)2 , ∆x =
n→∞
i=1
n
Z 2 p
= 4 − (−2 + x)2 dx
0

R2p p
For the integral 0 4 − (−2 + x)2 dx , y = 4 − (x − 2)2 is equivalent to
(x − 2)2 + y 2 = 4, y ≥ 0. So the integral represents the area between the
upper half of the circle (x − 2)2 + y 2 = 4 (which is centered at (2, 0) and has
radius 2) and the x-axis with 0 ≤ x ≤ 2, which is a quarter circle with area
1
4
· π 22 = π.

699
S OLUTIONS TO E XERCISES

y
p
y= 4 − (x − 2)2

x
2

1.1.8.38. *. Solution. (a) The left Riemann sum is defined as

n
X
Ln = f (xi−1 )∆x with xi = a + i∆x
i=1

We subdivide into n = 3 intervals, so that ∆x = b−a n

= 3−0
3
= 1, x0 = 0, x1 = 1
and x2 = 2. The function f (x) = 7 + x has the values f (x0 ) = 7 + 03 = 7,
3

f (x1 ) = 7 + 13 = 8, and f (x2 ) = 7 + 23 = 15, from which we evaluate

   
L3 = f (x0 ) + f (x1 ) + f (x2 ) ∆x = 7 + 8 + 15 × 1 = 30
b−a 3 3i
(b) We divide into n intervals so that ∆x = n
= n
and xi = a + i∆x = n
. The
right Riemann sum is therefore:
n n  n 
(3i)3 3 X 21 81 i3
X X  
Rn = f (xi )∆x = 7+ 3 = + 4
i=1 i=1
n n i=1
n n

To calculate the sum:

n
! n
!
21 X 81 X 3
Rn = 1 + i
n i=1 n4 i=1
81 n4 + 2n3 + n2
   
21
= ×n + ×
n n4 4
81
= 21 + (1 + 2/n + 1/n2 )
4
To evaluate the limit exactly, we take n → ∞. The expressions involving 1/n
vanish leaving:
3
81 1
Z
(7 + x3 ) dx = lim Rn = 21 + = 41
0 n→∞ 4 4
1.1.8.39. *. Solution. In general, the right-endpoint Riemann sum approxima-

700
S OLUTIONS TO E XERCISES
Rb
tion to the integral a
f (x) dx using n rectangles is
n
X
f (a + i∆x)∆x
i=1

2
where ∆x = b−a
n
. In this problem, a = 2, b = 4, and f (x) = x2 , so that ∆x = n
and
the right–endpoint Riemann sum approximation becomes
n n
X  2i  2 X  2i 2 2
f 2+ = 2+
i=1
n n i=1
n n
n
8i 4i2 2
X  
= 4+ + 2
i=1
n n n
n 
8 16i 8i2
X 
= + 2 + 3
i=1
n n n
n n n
X 8 X 16i X 8i2
= + +
i=1
n i=1 n2 i=1
n3
n n n
8X 16 X 8 X 2
= 1+ 2 i+ 3 i
n i=1 n i=1 n i=1
8 16 n(n + 1) 8 n(n + 1)(2n + 1)
= n+ 2 · + 3·
n n 2 n 6
 1  4 1  1
=8+8 1+ + 1+ 2+
n 3 n n

So
4
1  4 1 1 i
Z h 
x2 dx = lim 8 + 8 1 + + 1+ 2+
2 n→∞ n 3 n n
4 56
=8+8+ ×2=
3 3
1.1.8.40. *. Solution. We’ll use right Riemann sums with a = 0 and b = 2. When
there are n rectangles, ∆x = b−a
n
= n2 and xi = a + i∆x = 2i/n. So we need to
evaluate
n
X n
X
(xi )3 + xi ∆x


lim f (xi )∆x = lim
n→∞ n→∞
i=1 i=1
n  3 !
X 2i 2i 2
= lim +
n→∞
i=1
n n n
n
2 X 8i3 2i
 
= lim +
n→∞ n
i=1
n3 n

701
S OLUTIONS TO E XERCISES

n n
!
16 X 3 4 X
= lim i + i
n→∞ n4 i=1 n2 i=1
16(n4 + 2n3 + n2 ) 4(n2 + n)
 
= lim +
n→∞ n4 · 4 n2 · 2
    
16 2 1 4 1
= lim 1+ + 2 + 1+
n→∞ 4 n n 2 n
16 4
= + = 6.
4 2
1.1.8.41. *. Solution. We’ll use right Riemann sums with a = 1, b = 4 and
f (x) = 2x − 1. When there are n rectangles, ∆x = b−a
n
= n3 and xi = a + i∆x =
1 + 3i/n. So we need to evaluate
n
X n
X
lim f (xi )∆x = lim (2xi − 1) ∆x
n→∞ n→∞
i=1 i=1
n  
X 6i 3
= lim 2+ −1
n→∞
i=1
n n
n  
3 X 6i
= lim +1
n→∞ n n
i=1
n n
!
18 X 3X
= lim i+ 1
n→∞ n2 i=1 n i=1
 
18 · n(n + 1) 3
= lim + n
n→∞ n2 · 2 n
   
1
= lim 9 1 + +3
n→∞ n
= 9 + 3 = 12.
1.1.8.42. Solution. Using the definition of a right Riemann sum,
10
X 10
X
2
3(7 + 2i) sin(4i) = ∆xf (a + i∆x)
i=1 i=1

Since ∆x = 10 and a = −5,

10
X 10
X
2
3(7 + 2i) sin(4i) = 10f (−5 + 10i)
i=1 i=1

Dividing both expressions by 10,

10 10
X 3 2
X
(7 + 2i) sin(4i) = f (−5 + 10i)
i=1
10 i=1

So, we have an expression for f (−5 + 10i):

3
f (−5 + 10i) = (7 + 2i)2 sin(4i)
10

702
S OLUTIONS TO E XERCISES

In order to find f (x), let x = −5 + 10i. Then i = x

10
+ 21 .
  2   
3 x 1 x 1
f (x) = 7+2 + sin 4 +
10 10 2 10 2
 
3 x 2 2x
= + 8 sin +2 .
10 5 5
1.1.8.43. Solution. As in the text, we’ll set up a Riemann sum for the given
integral. Right Riemann sums have the simplest form, so we use a right Riemann
sum, but we could equally well use left or midpoint.
Z 1 n
X
x
2 dx = lim ∆xf (a + i∆x)
0 n→∞
i=1
n  
X 1 i
= lim f
n→∞
i=1
n n
n
X 1 i/n
= lim ·2
n→∞
i=1
n
1 1/n
2 + 22/n + 23/n + · · · + 2n/n


= lim
n→∞ n
21/n  n−1

= lim 1 + 21/n + 22/n + · · · + 2 n
n→∞ n
21/n 
2
n−1 
= lim 1 + 21/n + 21/n + · · · + 21/n
n→∞ n

The sum in parenthesis has the form of a geometric sum, with r = 21/n :

n !
21/n 21/n − 1
= lim
n→∞ n 21/n − 1
21/n
 
2−1
= lim
n→∞ n 21/n − 1
21/n
= lim
n→∞ n(21/n − 1)

Note as n → ∞, 1/n → 0, so the numerator has limit 1, while the denominator has
indeterminate form ∞ · 0. So, we’ll do a little algebra to get this into a l’Hôpital-
style indeterminate form:
1
n
· 21/n
= lim
n→∞ 21/n − 1
1
n
= lim
n→∞ 1 − 2−1/n
| {z }
num→0
den→0

703
S OLUTIONS TO E XERCISES

d
Now we can use l’Hôpital’s rule. Recall dx {2x } = 2x log x, where log x is the
natural logarithm of x, also sometimes written ln x. We’ll need to use the chain
rule when we differentiate the denominator.
−1
n2
= lim −1/n 1
n→∞ −2 log 2 · n2
21/n
= lim
n→∞ log 2
1
=
log 2

Using a calculator, we see this is about 1.44 square units.

1.1.8.44. Solution. As in the text, we’ll set up a Riemann sum for the given
integral. Right Riemann sums have the simplest form:
Z b n
X
x
10 dx = lim ∆xf (a + i∆x)
a n→∞
i=1
n  
X b−a b−a
= lim f a+i
n→∞
i=1
n n
n
X b−a b−a
= lim · 10a+i n
n→∞
i=1
n
n
b−a
X
a
 b−a i
= lim · 10 · 10 n
n→∞
i=1
n
  b−a n 
b−a a b−a 1
  b−a 2  b−a 3
= lim · 10 10 n + 10 n + 10 n + · · · + 10 n
n→∞ n
  b−a n−1 
b−a a b−a
 b−a   b−a 2
= lim · 10 · 10 n 1 + 10 n + 10 n + · · · + 10 n
n→∞ n
b−a
Now the sum in parentheses has the form of a geometric sum, with r = 10 n :
  b−a n 
b−a b−a
10 n − 1
= lim · 10a · 10 n  b−a

n→∞ n 10 n − 1
 b−a 
b−a a b−a 10 − 1
= lim · 10 · 10 n
b−a
n→∞ n 10 n − 1
The coloured parts do not depend on n, so for simplicity we can move them
outside the limit.
b−a
!
1 10 n
= (b − a) · 10a 10b−a − 1 lim ·


b−a
n→∞ n 10 n − 1
 
b a

1/n
= (b − a) · 10 − 10 lim b−a
n→∞ 1 − 10− n
| {z }
num→0
den→0

704
S OLUTIONS TO E XERCISES

d
Now we can use l’Hôpital’s rule. Recall dx {10x } = 10x log x, where log x is the
natural logarithm of x, also sometimes written ln x. For the denominator, we will
have to use the chain rule.
!
2
−1/n
= (b − a) · 10b − 10a lim


b−a
n→∞ −10− n · log 10 · b−a n2
!
1
= (b − a) · 10b − 10a lim


b−a
n→∞ 10− n · log 10 · (b − a)
 
b a

1
= (b − a) · 10 − 10
log 10 · (b − a)
1
10b − 10a


=
log 10

For part (b), we can guess that if 10 were changed to c, our answer would be
b
1
Z
cx dx = cb − ca


a log c

In Question 43, we had a = 0, b = 1, and c = 2. In this case, the formula we

guessed above gives
1
1 1
Z
2x dx = 21 − 20 =


0 log 2 log 2

This does indeed match the answer

Z b we calculated.
1
cx dx = cb − ca using the method of this


(In fact, we can directly show
a log c
problem.)
√
1.1.8.45. Solution. First, we note y = 1 − x2 is the upper half of a circle of
radius 1, centred at the origin. We’re taking the area under the curve from 0 to a,
so the area in question is as shown in the picture below.

x
−1 a 1

In order to use geometry to find this area, we break it up into two pieces: a sector
of a circle, and a triangle, shown below.

705
S OLUTIONS TO E XERCISES

θ
x
−1 a 1

• Area of sector: The sector is a portion of a circle with radius 1, with inner
θ θ
angle θ. So, its area is 2π (area of circle) = 2π (π) = 2θ .
Our job now is to find θ in terms of a. Note π2 − θ is the inner angle of
π


the red triangle, which lies in the unit circle. So, cos 2 − θ = a. Then
π
2
− θ = arccos(a), and so θ = π2 − arccos(a).
Then the area of the sector is π
4
− 12 arccos(a) square units.

√ a. Its height is the y-value of the

• Area of triangle: The triangle has base
2
√ when x = a, so its height is 1 − a . Then the area of the triangle
function
1
is 2 a 1 − a2 .
Z a√
π 1 1 √
We conclude 1 − x2 dx = − arccos(a) + a 1 − a2 .
0 4 2 2
1.1.8.46. Solution.

a The difference between our upper and lower bounds is the difference in
areas between the larger set of rectangles and the smaller set of rectangles.
Drawing them on a single picture makes this a little clearer.

y = f (x)
x
a b

b−a
Each of the rectangles has width n
, since we took a segment of the x-

706
S OLUTIONS TO E XERCISES

axis with length b − a and chopped it into n pieces. We could calculate the
height of each rectangle, but it would be a little complicated, since it differs
for each of them. An easier method is to notice that the area we want to
calculate can be imagined as a single rectangle:

f (a)

f (b) y = f (x)
x
a b

The rectangle has base b−a n

. Its highest coordinate is f (a), and its lowest is
f (b), so its height is f (b) − f (a). Therefore, the difference in area between
our lower bound and our upper bound is:

b−a
[f (b) − f (a)] ·
n

b We want to give a range with length at most 0.01, and guarantee that the
area under the curve y = f (x) is inside that range. In the previous part,
we figured out that when we use n rectangles, the length of our range is
[f (b) − f (a)] · b−a
n
. So, all we have to do is set this to be less than or equal to
0.01, and solve for n:
b−a
[f (b) − f (a)] · ≤ 0.01
n
100 [f (b) − f (a)] · (b − a) ≤ n

We can choose n to be an integer that is greater than or equal to

100 [f (b) − f (a)] · (b − a). Using that many rectangles, we find an upper
and lower bound for the area under the curve. If we choose any number
between our upper and lower bound as an approximation for the area un-
der the curve, our error is no more than 0.01.

Remark: this question depends on the fact that f is decreasing and positive from
a to b. In general, bounding errors on approximations like this is not so straight-
forward.
1.1.8.47. Solution. Since f (x) is linear, there exist real numbers m and c such
that f (x) = mx + c. Now we can do some calculations. Suppose we have a
rectangle in our Riemann sum that takes up the interval [x, x + w].

707
S OLUTIONS TO E XERCISES

• If we are using a left Riemann sum, our rectangle has height f (x) = mx + c.
Then it has area w(mx + c).

• If we are using a right Riemann sum, our rectangle has height f (x + w) =

m(x + w) + c = mx + c + mw. Then it has area w(mx + c + mw).

• If we are using a midpoint Riemann sum, our rectangle has height f (x
+
1
2
w) = m(x + 12 w) + c = mx + c + 21 mw. Then it has area w mx + c + 12 w .

So, for each rectangle in our sums, the midpoint rectangle has the same area as
the average of the left and right rectangles:
 
1 w(mx + c) + w(mx + c + mw)
w mx + c + mw =
2 2

It follows that the midpoint Riemann sum has a value equal to the average of the
values of the left and right Riemann sums. To see this, let the rectangles in the
midpoint Riemann sum have areas M1 , M2 , . . . , Mn , let the rectangles in the left
Riemann sum have areas L1 , L2 , . . . , Ln , and let the rectangles in the right Rie-
mann sum have areas R1 , R2 , . . . , Rn . Then the midpoint Riemann sum evaluates
to M1 + M2 + · · · + Mn , and:
[L1 + L2 + . . . + Ln ] + [R1 + R2 + . . . + Rn ]
2
L1 + R1 L2 + R2 Ln + Rn
= + + ··· +
2 2 2
= M1 + M2 + · · · + Mn

So, the statement is true.

(Note, however, it is false for many non-linear functions f (x).)

1.2 · Basic properties of the definite integral

1.2.3 · Exercises

Exercises — Stage 1
1.2.3.1. Solution.
Z a
a f (x) dx = 0
a

y y = f (x)

x
a

The area under the curve is zero, because it’s a region with no width.

708
S OLUTIONS TO E XERCISES
Z b Z c Z b
b f (x) dx = f (x) dx + f (x)dx
a a c

y y = f (x)

x
a c b

If we assume a ≤ c ≤ b, then this identity simply tells us that if we add up

the area under the curve from a to c, and from c to b, then we get the whole
area under the curve from a to b.
(The situation is slightly more complicated when c is not between a and b,
but it still works out.)
Z b Z b Z b
c (f (x) + g(x)) dx = f (x) dx + g(x) dx
a a a

y
y = f (x) + g(x)

y = f (x)

x
a b

Z b
The blue-shaded area in the picture above is f (x)dx. The area under the
Z b a

curve f (x) + g(x) but above the curve f (x) (shown in red) is g(x)dx.
a

1.2.3.2. Solution. Using the identity

Zb Zc Zb
f (x)dx = f (x)dx + f (x)dx ,
a a c

we see
Zb Z0 Zb
cos xdx = cos xdx + cos xdx
a a 0
Za Zb
=− cos xdx + cos xdx
0 0

709
S OLUTIONS TO E XERCISES

= − sin a + sin b
= sin b − sin a
1.2.3.3. *. Solution. (a) False. For example if
(
0 for x < 0
f (x) =
1 for x ≥ 0
R −2 R2
then −3
f (x)dx = 0 and − 3
f (x)dx = −1.

x
−3 −2 2 3
R −2 R3
(b) False. For example, if f (x) = x, then −3
f (x) dx is negative while 2
f (x) dx
is positive, so they cannot be the same.

−3 −2
x
2 3

(c) False. For example, consider the functions

( (
0 for x < 12 0 for x ≥ 1
2
f (x) = and g(x) =
1 for x ≥ 12 1 for x < 1
2

R1 R1 1
Then f (x) · g(x) = 0 for all x, so 0 f (x) · g(x)dx = 0. However, 0 f (x)dx = 2
and
R1 R1 R1
0
g(x)dx = 12 , so 0 f (x)dx · 0 g(x)dx = 14 .

y y
1 f (x) 1
g(x)
x x
1 1 1 1
2
2

710
S OLUTIONS TO E XERCISES

1.2.3.4. Solution.
b−a 0−5 1
a ∆x = = =−
n 100 20
Note: if we were to use the Riemann-sum definition of a definite integral,
Rb Ra
this is how we would justify the identity f (x)dx = − f (x)dx.
a b

b The heights of the rectangles are given by f (xi ), where xi = a + i∆x =

5 − 20i . Since f (x) only gives positive values, f (xi ) > 0, so the heights of the
rectangles are positive.

c Our Riemann sum is the sum of the signed areas of individual rectangles.
Each rectangle has a negative base (∆x) and a positive height (f (xi )). So,
each term of our sum is negative. If we add up negative numbers, the sum
is negative. So, the Riemann sum is negative.
R5
d Since f (x) is always above the x-axis, f (x)dx is positive.
0

Exercises — Stage 2
1.2.3.5. *. Solution. The operation of integration is linear (that’s part (d) of the
“arithmetic of integration” Theorem 1.2.1), so that:
Z 3 Z 3 Z 3
[6f (x) − 3g(x)] dx = 6f (x) dx − 3g(x) dx
2 2 2
Z 3 Z 3
=6 f (x) dx − 3 g(x) dx
2 2
= (6 × (−1)) − (3 × 5) = −21
1.2.3.6. *. Solution. The operation of integration is linear (that’s part (d) of the
“arithmetic of integration” Theorem 1.2.1), so that:
Z 2 Z 2 Z 2
[2f (x) + 3g(x)] dx = 2f (x) dx + 3g(x) dx
0 0 0
Z 2 Z 2
=2 f (x) dx + 3 g(x) dx
0 0
= (2 × 3) + (3 × (−4)) = −6
1.2.3.7. *. Solution. Using part (d) of the “arithmetic of integration” Theorem
1.2.1, followed by parts (c) and (b) of the “arithmetic for the domain of integra-
tion” Theorem 1.2.3,
Z 2 Z 2 Z 2
 
3g(x) − f (x) dx = 3 g(x) dx − f (x) dx
−1 −1 −1

711
S OLUTIONS TO E XERCISES
Z 0 Z 2 Z 0 Z 2
=3 g(x) dx + 3 g(x) dx − f (x) dx − f (x) dx
−1 0 −1 0
Z 0 Z 2 Z −1 Z 2
=3 g(x) dx + 3 g(x) dx + f (x) dx − f (x) dx
−1 0 0 0
= 3 × 3 + 3 × 4 + 1 − 2 = 20
1.2.3.8. Solution.
√
a Since 1 − x2 is an even function,
Z 0√ Z |a| √
2
1 − x dx = 1 − x2 dx
a 0
π 1 1 p
= − arccos(|a|) + |a| 1 − |a|2
4 2 2
π 1 1 √
= − arccos(−a) − a 1 − a2
4 2 2
Alternatively, since arccos(−a) = π − arccos(a) we also have
Z 0 √ π 1 1 √
1 − x2 dx = − + arccos(a) − a 1 − a2
a 4 2 2
Z 1 √ π
b Note 1 − x2 dx =
, since the area under the curve represents one-
0 4
quarter of the unit circle. Then,
Z 1 √ Z√ 1 Z a√
1− x2 dx = 2
1 − x dx − 1 − x2 dx
a 0 0
1 √
 
π π 1
= − − arccos(a) + a 1 − a2
4 4 2 2
1 1 √
= arccos(a) − a 1 − a2
2 2
1.2.3.9. *. Solution. Recall that
(
−x if x ≤ 0
|x| =
x if x ≥ 0

so that
(
−2x if x ≤ 0
|2x| =
2x if x ≥ 0

Also recall, from Example 1.2.6, that

b
b 2 − a2
Z
xdx =
a 2

712
S OLUTIONS TO E XERCISES

So
Z 2 Z 0 Z 2 Z 0 Z 2
|2x|dx = |2x|dx + |2x|dx = (−2x)dx + 2xdx
−1 −1 0 −1 0
Z 0 Z 2
02 − (−1)2 22 − 02
= −2 xdx + 2 xdx = −2 · +2·
−1 0 2 2
=1+4=5
1.2.3.10. Solution. We note that the integrand f (x) = x|x| is an odd function,
because f (−x) = −x| − x| = −x|x| = −f (x). Then by Theorem 1.2.12 part (b),
Z 5
x|x|dx = 0.
−5

1.2.3.11. Solution. Using Theorem 1.2.12 part (a),

Z 2 Z 2
10 = f (x)dx = 2 f (x)dx
−2 0
Z 2
5= f (x)dx
0

Also,
Z 2 Z 0 Z 2
f (x)dx = f (x)dx + f (x)dx
−2 −2 0

So,
Z 0 Z 2 Z 2
f (x)dx = f (x)dx − f (x)dx
−2 −2 0
= 10 − 5 = 5

R0 Ra
Indeed, for any even function f (x), f (x)dx = f (x)dx.
−a 0

Exercises — Stage 3
1.2.3.12. *. Solution. We first use additivity:
Z 2 √  Z 2 Z 2 √
2
5 + 4 − x dx = 5 dx + 4 − x2 dx
−2 −2 −2

The first integral represents the area of a rectangle of height 5 and width 4 and
so equals 20. The√ second integral represents the area above the x–axis and below
2 2 2
the curve y = 4 − x or x + y = 4. That is a semicircle of radius 2, which has
area 12 π22 . So
Z 2 √ 
5 + 4 − x2 dx = 20 + 2π
−2

713
S OLUTIONS TO E XERCISES

y y

y=5

√
y= 4 − x2

x
−2 2
x R2 √
4 − x2 dx = 2π
−2 2 −2

R2
5dx = 5
−2

sin x
1.2.3.13. *. Solution. Note that the integrand f (x) = log(3+x2 )
is an odd function,
because:
sin(−x) − sin x
f (−x) = 2
= = −f (x)
log(3 + (−x) ) log(3 + x2 )
The domain of integration −2012 ≤ x ≤ 2012 is symmetric about x = 0. So, by
Theorem 1.2.12, Z +2012
sin x
2
dx = 0
−2012 log(3 + x )

1.2.3.14. *. Solution. Note that the integrand f (x) = x1/3 cos x is an odd func-
tion, because:

f (−x) = (−x)1/3 cos(−x) = −x1/3 cos x = −f (x)

The domain of integration −2012 ≤ x ≤ 2012 is symmetric about x = 0. So, by

Theorem 1.2.12, Z +2012
x1/3 cos x dx = 0
−2012

1.2.3.15. Solution. Our integrand f (x) = (x − 3)3 is neither even nor odd. How-
ever, it does have a similar symmetry. Namely, f (3 + x) = −f (3 − x). So, f is
“negatively symmetric” across the line x = 3. This suggests that the integral
should be 0: the positive area to the right of x = 3 will be the same as the nega-
tive area to the left of x = 3.
Another way to see this is to notice that the graph of f (x) = (x − 3)3 is equivalent
to the graph of g(x) = x3 shifted three units to the right, and g(x) is an odd
function. So, Z 6 Z 3
3
(x − 3) dx = x3 dx = 0
0 −3

714
S OLUTIONS TO E XERCISES

y
y = x3 y = (x − 3)3

x
−3 3 6

1.2.3.16. Solution.
a

(ax)2 + (by)2 = 1
p
by = 1 − (ax)2
1p
y= 1 − (ax)2
b

b The values of x in the domain of the function above are those that satisfy
1 − (ax)2 ≥ 0. That is, − a1 ≤ x ≤ a1 . Therefore, the upper half of the ellipse
has area
Z 1
1 a p
1 − (ax)2 dx
b − a1
√
The upper half of a circle has equation y = r2 − x2 .
Z 1 s  
1 a 2
1 2
= a − x dx
b − a1 a2
Z 1 r
1 a 1
= a − x2 dx
b − a1 a2
Z 1 r
a a 1
= − x2 dx
b − a1 a2
r
1
c The function y = − x2 is the upper-half of the circle centred at the
a2
1 a π
origin with radius . So, the expression from (b) evaluates to =
a b 2a2
π
.
2ab
The expression from (b) was half of the ellipse, so the area of the ellipse is
π
.
ab

715
S OLUTIONS TO E XERCISES

Remark: this was a slightly long-winded way of getting the result. The reasoning
is basically this:

• The area of the unit circle x2 + y 2 = 1 is π.

• The ellipse (ax)2 + y 2 = 1 is obtained by shrinking the unit circle horizon-

π
tally by a factor of a. So, its area is .
a
• Further, the ellipse (ax)2 + (by)2 = 1 is obtained from the previous ellipse
π
by shrinking it vertically by a factor of b. So, its area is .
ab
1.2.3.17. Solution. Let’s recall the definitions of even and odd functions: f (x) is
even if f (−x) = f (x) for every x in its domain, and f (x) is odd if f (−x) = −f (x)
for every x in its domain.
Let h(x) = f (x) · g(x).

• even × even: If f and g are both even, then h(−x) = f (−x) · g(−x) =
f (x) · g(x) = h(x), so their product is even.

• odd × odd: If f and g are both odd, then h(−x) = f (−x) · g(−x) = [−f (x)] ·
[−g(x)] = f (x) · g(x) = h(x), so their product is even.

• even × odd: If f is even and g is odd, then h(−x) = f (−x) · g(−x) =

f (x)·[−g(x)] = −[f (x)·g(x)] = −h(x), so their product is odd. Because mul-
tiplication is commutative, the order we multiply the functions in doesn’t
matter.

We note that the table would be the same as if we were adding (not multiplying)
even and odd numbers (not functions).
1.2.3.18. Solution. Since f (x) is odd, f (0) = −f (−0) = −f (0). So, f (0) = 0.
However, this restriction does not apply to g(x). For example, for any constant c,
let g(x) = c. Then g(x) is even and g(0) = c. So, g(0) can be any real number.
1.2.3.19. Solution. Let x be any real number.

• f (x) = f (−x) (since f (x) is even), and

• f (x) = −f (−x) (since f (x) is odd).

• So, f (x) = −f (x).

• Then (adding f (x) to both sides) we see 2f (x) = 0, so f (x) = 0.

So, f (x) = 0 for every x.

1.2.3.20. Solution.
• Solution 1: Suppose f (x) is an odd function. We investigate f 0 (x) using the
chain rule:
f (−x) = −f (x) (odd function)

716
S OLUTIONS TO E XERCISES

d d
{f (−x)} = {−f (x)}
dx dx
−f 0 (−x) = −f 0 (x) (chain rule)
f 0 (−x) = f 0 (x)
So, when f (x) is odd, f 0 (x) is even.
Similarly, suppose f (x) is even.
f (−x) = f (x) (even function)
d d
{f (−x)} = {f (x)}
dx dx
−f 0 (−x) = f 0 (x) (chain rule)
f 0 (−x) = −f 0 (x)
So, when f (x) is even, f 0 (x) is odd.
• Solution 2: Another way to think about this problem is to notice that “mir-
roring” a function changes the sign of its derivative. Then since an even
function is “mirrored once” (across the y-axis), it should have f 0 (x) =
−f 0 (−x), and so the derivative of an even function should be an odd func-
tion. Since an odd function is “mirrored twice” (across the y-axis and across
the x-axis), it should have f 0 (x) = −(−f 0 (−x)) = f 0 (−x). So the derivative
of an odd function should be even. These ideas are presented in more detail
below.
First, we consider the case where f (x) is even, and investigate f 0 (x).

x
−a3 −a2 −a1 a1 a2 a3

y = f (x)

The whole function has a mirror-like symmetry across the y-axis. So, at x
and −x, the function will have the same “steepness,” but if one is increasing
then the other is decreasing. That is, f 0 (−x) = −f 0 (x). (In the picture above,
compare the slope at some point ai with its corresponding point −ai .) So,
f 0 (x) is odd when f (x) is even.
Second, let’s consider the case where f (x) is odd, and investigate f 0 (x).
Suppose the blue graph below is y = f (x). If f (x) were even, then to the left
of the y-axis, it would look like the orange graph, which we’ll call y = g(x).

717
S OLUTIONS TO E XERCISES

y = g(x) y = f (x)

From our work above, we know that, for every x > 0, −f 0 (x) = g 0 (−x).
When x < 0, f (x) = −g(x). So, if x > 0, then −f 0 (x) = g 0 (−x) = −f 0 (−x).
In other words, f 0 (x) = f 0 (−x). Similarly, if x < 0, then f 0 (x) = −g 0 (x) =
f 0 (−x). Therefore f 0 (x) is even. (In the graph below, you can anecdotally
verify that f 0 (ai ) = f 0 (−ai ).)

y = g(x) y = f (x)

x
−a3 −a2 −a1 a1 a2 a3

1.3 · The Fundamental Theorem of Calculus

1.3.2 · Exercises

Exercises — Stage 1
1.3.2.1. *. Solution. The Fundamental Theorem of Calculus Part 2 (Theorem
1.3.1) tells us that
√
Z 5 √
f (x) dx = F ( 5) − F (1)
1
√ 2 2
= e( 5 −3) + 1 − e(1 −3) + 1

= e5−3 − e1−3 = e2 − e−2

1.3.2.2. *. Solution. First, let’s find a general antiderivative of x3 − sin(2x).
x4
• One function with derivative x3 is .
4
• To find an antiderivative of sin(2x), we might first guess cos(2x); checking,

718
S OLUTIONS TO E XERCISES

d 1
we see dx {cos(2x)} = −2 sin(2x). So, we only need to multiply by − :
  2
d 1
− cos 2x = sin(2x).
dx 2
x4 1
So, the general antiderivative of f (x) is + cos 2x + C. To satisfy F (0) = 1, we
a
4 2
need

h x4 1 i 1 1
+ cos 2x + C = 1 ⇐⇒ + C = 1 ⇐⇒ C =
4 2 x=0 2 2
x4 1 1
So F (x) = + cos 2x + .
4 2 2

a The symbol ⇐⇒ is read “if and only if”. This is used in mathematics to express the logical
equivalence of two statements. To be more precise, the statement P ⇐⇒ Q tells us that P
is true whenever Q is true and Q is true whenever P is true.

1.3.2.3. *. Solution. (a) This is true, by part 2 of the Fundamental Theorem of

Calculus, Thereom 1.3.1, with G(x) = f (x) and f (x) replaced by f 0 (x).
(b) This is not only false, but it makes no sense at all. The integrand is strictly pos-
itive so the integral has to be strictly positive. In fact it’s +∞. The Fundamental
Theorem of Calculus does not apply because the integrand has an infinite dis-
continuity at x = 0.

1
y= x2

x
−1 1

Rb
(c) This is not only false, but it makes no sense at all, unless a f (x) dx =
Rb
a
xf (x) dx = 0. The left hand side is a number. The right hand side is a number
times x.
Z b Z b
xf (x)dx vs x ·
|{z} f (x)dx
a a
| {z } variable | {z }
area area
R1 1
For example, if a = 0, b = 1 and f (x) = 1, then the left hand side is 0
x dx = 2
R1
and the right hand side is x 0 dx = x.

719
S OLUTIONS TO E XERCISES

1.3.2.4. Solution. This is a tempting thought:

1
Z
dx = log |x| + C
x
so perhaps similarly

1
Z
?
2
dx = log |x2 | + C = log(x2 ) + C
x
We check by differentiating:

d d 2 1
{log(x2 )} = {2 log x} = 6= 2
dx dx x x
So, it wasn’t so easy: false.
When we’re guessing antiderivatives, we often need to adjust our original
guesses a little. Changing constants works well; changing functions usually does
not.
1.3.2.5. Solution. This is tempting:

d
{sin(ex )} = ex cos(ex )
dx
so perhaps

sin(ex )
 
d ?
= cos(ex )
dx ex

We check by differentiating:

d sin(ex ) ex (cos(ex ) · ex ) − sin(ex )ex

 
= (quotient rule)
dx ex e2x
sin(ex )
= cos(ex ) −
ex
x
6= cos(e )

So, the statement is false.

When we’re guessing antiderivatives, we often need to adjust our original
guesses a little. Dividing by constants works well; dividing by functions usu-
ally does not.
1.3.2.6. Solution. “The instantaneous rate of change of F (x) with respect to x”
is another way of saying “F 0 (x)”. From the Fundamental Theorem of Calculus
Part 1, we know this is sin(x2 ).
1.3.2.7. Solution. The slope of the tangent line to y = F (x) when x = 3 is
exactly F 0 (3). By the
√ Fundamental Theorem of Calculus Part 1, F 0 (x) = e1/x .
Then F 0 (3) = e1/3 = 3 e.

720
S OLUTIONS TO E XERCISES

1.3.2.8. Solution. For any constant C, F (x) + C is an antiderivative of f (x),

d d
because dx {F (x) + C} = dx {F (x)} = f (x). So, for example, F (x) and F (x) + 1
are both antiderivatives of f (x).
1.3.2.9. Solution.
d −1
a We differentiate with respect to a. Recall dx {arccos x} = √1−x 2 . To differen-
√
tiate 12 a 1 − a2 , we use the product and chain rules.

1 √
 
d π 1
− arccos(a) + a 1 − a2
da 4 2 2
1√
 
1 −1 1 −2a
=0− · √ + a · √ + 1 − a2
2 1 − a2 2 2 1 − a2 2
1 a2 1 − a2
= √ − √ + √
2 1 − a2 2 1 − a2 2 1 − a2
1 − a2 + 1 − a2
= √
2 1 − a2
2(1 − a2 )
= √
2 1 − a2
√
= 1 − a2
√
b Let G(x) = π4 − 12 arccos(x) + 12 x 1 − x2 . We showed in part (a) that G(x) is
√ √
an antiderivative of 1 − x2 . Since F (x) is also an antiderivative of 1 − x2 ,
F (x) = G(x) + C for some constant C (this is Lemma 1.3.8).
Z 0√
Note G(0) = 1 − x2 dx = 0, so if F (0) = π, then F (x) = G(x) + π. That
0
is,
5π 1 1 √
F (x) = − arccos(x) + x 1 − x2 .
4 2 2
1.3.2.10. Solution.

The antiderivative of cos x is sin x, and cos x is continuous everywhere, so

a Z
π
cos xdx = sin(π) − sin(−π) = 0.
−π

b Since sec2 x is discontinuous atZx = ± π2 , the Fundamental Theorem of Cal-

π
culus Part 2 does not apply to sec2 xdx.
−π

1
c Since x+1
is discontinuous at x = −1, the Fundamental Theorem of Calcu-
Z 0
1
lus Part 2 does not apply to dx.
−2 x + 1

1.3.2.11. Solution. Using the definition of F , F (x) is the area under the curve
from a to x, and F (x + h) is the area under the curve from a to x + h. These are

721
S OLUTIONS TO E XERCISES

shown on the same diagram, below.

y = f (t)

t
a x x+h

Then the area represented by F (x + h) − F (x) is the area that is outside the red,
x+h
R
but inside the blue. Equivalently, it is f (t)dt.
x

y = f (t)

t
a x x+h

R0
1.3.2.12. Solution. We evaluate F (0) using the definition: F (0) = 0 f (t)dt = 0.
Although f (0) > 0, the area from t = 0 to t = 0 is zero.
As x moves along, F (x) adds bits of signed area. If it’s adding positive area, it’s
increasing, and if it’s adding negative area, it’s decreasing. So, F (x) is increasing
when 0 < x < 1 and 3 < x < 4, and F (x) is decreasing when 1 < x < 3.
1.3.2.13. Solution. This question is nearly identical to Question 12, with
Z 0 Z x
G(x) = f (t)dt = − f (t)dt = −F (x).
x 0

So, G(x) increases when F (x) decreases, and vice-versa. Therefore: G(0) = 0,
G(x) is increasing when 1 < x < 3, and G(x) is decreasing when 0 < x < 1 and
when 3 < x < 4.
1.3.2.14. Solution. Using the definition of the derivative,

F (x + h) − F (x)
F 0 (x) = lim
h→0 h

722
S OLUTIONS TO E XERCISES
R x+h Rx
a
tdt − a
tdt
= lim
h→0 h
R x+h
tdt
= lim x
h→0 h
The numerator describes the area of a trapezoid with base h and heights x and
x + h.
1
2
h(x
+ x + h)
= lim
h→0
 h
1
= lim x + h
h→0 2
=x

y
y=t
x+h
R x+h
x x
tdt

t
x x+h

So, F 0 (x) = x.
1.3.2.15. Solution. If F (x) is constant, then F 0 (x) = 0. By the Fundamental The-
orem of Calculus Part 1, F 0 (x) = f (x). So, the only possible continuous function
fitting the question is f (x) = 0.
This makes intuitive sense: if moving x doesn’t add or subtract area under the
curve, then there must not be any area under the curve–the curve should be the
same as the x-axis.
As an aside, we mention that there are other, non-continuous
 functions f (t) such
Rx 0 x 6= 0
that 0 f (t)dt = 0 for all x. For example, f (t) = . These kinds of
1 x=0
removable discontinuities will not factor heavily in our discussion of integrals.
1.3.2.16. Solution.
d a
{x log(ax) − x} = x + log(ax) − 1 (product rule, chain rule)
dx ax
= log(ax)

So, we know
Z
log(ax)dx = x log(ax) − x + C

723
S OLUTIONS TO E XERCISES

where a isR a given constant, and C is any constant.

Remark: log(ax)dx can be calculated using the method of Integration by Parts,
which you will learn in Section 1.7.
1.3.2.17. Solution.
d  x 3
e x − 3x2 + 6x − 6


dx
= ex 3x2 − 6x + 6 + ex x3 − 3x2 + 6x − 6



(product rule)
= ex 3x2 − 6x + 6 + x3 − 3x2 + 6x − 6

= x3 ex

So,
Z
x3 ex dx = ex x3 − 3x2 + 6x − 6 + C

Remark: x3 ex dx can be calculated using the method of Integration by Parts,

R

which you will learn in Section 1.7.

1.3.2.18. Solution.
√
 
d n o 1 1
log x + x2 + a2 = √ · 1+ √ · 2x
dx x + x 2 + a2 2 x 2 + a2
(chain rule)
√
2 2
x +a +x
1 + √x2x+a2 √
2 2
= √ = √x +a
x+ x +a2 2 x + x + a2
2

1
=√
x + a2
2

So,
Z
1 √
√ dx = log x + x2 + a2 + C
x 2 + a2

Remark: √x21+a2 dx can be calculated using the method of Trigonometric Substi-

R

tution, which you will learn in Section 1.9.

1.3.2.19. Solution. Using the chain rule:

d np √ √
o
x(a + x) − a log x + a + x
dx   
x + (a + x) 1 1 1
= p −a √ √ · √ + √
2 x(a + x) x+ a+x 2 x 2 a+x
√ √ !!
2x + a 1 a+x+ x
= p −a √ √ · p
2 x(a + x) x+ a+x 2 x(a + x)

724
S OLUTIONS TO E XERCISES
!
2x + a 1
= p −a p
2 x(a + x) 2 x(a + x)
2x x
= p =p
2 x(a + x) x(a + x)

So,
x √ √
Z p

p dx = x(a + x) − a log x + a + x + C
x(a + x)

Remark: √ x dx can be calculated using the method of Trigonometric Sub-

R
x(a+x)
stitution, which you will learn in Section 1.9.

Exercises — Stage 2
1.3.2.20. *. Solution. By the Fundamental Theorem of Calculus,
2 2
x4
Z 
3
x + sin x) dx = − cos x
0 4
 4 0
2
= − cos 2 − (0 − cos 0)
4
= 4 − cos 2 + 1 = 5 − cos 2.
1.3.2.21. *. Solution. By part (d) of our “Arithmetic of Integration” theorem,
Theorem 1.2.1,
Z 2 2 Z 2h Z 2 Z 2
x +2 2i 1
2
dx = 1 + 2 dx = dx + 2 2
dx
1 x 1 x 1 1 x

Then by the Fundamental Theorem of Calculus Part 2,

2 2
1 h 1 i2  h 1
Z Z h i2 i

dx + 2 dx = x + 2 − = 2−1 +2 − +1
1 1 x2 1 x 1 2
=2
1
1.3.2.22. Solution. The integrand is similar to , which is the derivative of
1 + x2
arctangent. Indeed, we have
1 1
Z Z
2
dx = dx.
1 + 25x 1 + (5x)2
So, a reasonable first guess for the antiderivative might be
?
F (x)= arctan(5x).
However, because of the chain rule,
5
F 0 (x)= .
1 + (5x)2

725
S OLUTIONS TO E XERCISES

In order to “fix” the numerator, we make a second guess:

1
F (x)= arctan(5x)
5 
0 1 5 1
F (x)= 2
=
5 1 + (5x) 1 + 25x2
1 1
Z
So, 2
dx = arctan(5x) + C.
1 + 25x 5
1
1.3.2.23. Solution. The integrand is similar to √ . In order to formulate a
√ 1 − x2
guess for the antiderivative, let’s factor out 2 from the denominator:

1 1
Z Z
√ dx = q dx
2−x 2 2

2 1 − x2
1
Z
= √ q dx
2
2 1 − x2
1 1
Z
= √ ·s 2 dx
2 
x
1− √
2

 we might guess that our antiderivative is something like F (x) =

At thispoint,
x
arcsin √ . To explore this possibility, we can differentiate, and see what we
2
get.
  
d x 1 1
arcsin √ =√ ·s
dx 2 2 
x
2
1− √
2

This is exactly what we want! So,

 
1 x
Z
√ dx = arcsin √ +C
2 − x2 2
1.3.2.24. Solution. We know that sec2 xdx = tan x + C, and sec2 x = tan2 x + 1,
R

so
Z Z
tan xdx = sec2 x − 1dx
2

Z Z
2
= sec xdx − 1dx

= tan x − x + C

726
S OLUTIONS TO E XERCISES

1.3.2.25. Solution.

• Solution 1: This might not obviously look like the derivative of anything
familiar, but it does look like half of a familiar trig identity: 2 sin x cos x =
sin(2x).

3
Z Z
3 sin x cos xdx = · 2 sin x cos xdx
2
3
Z
= sin(2x)dx
2

So, we might guess that the antiderivative is something like − cos(2x). We

only need to figure out the constants.

d
{− cos(2x)} = 2 sin(2x)
dx
 
d 3 3
So, − cos(2x) = sin(2x)
dx 4 2
3
Z
Therefore, 3 sin x cos xdx = − cos(2x) + C
4

• Solution 2: You might notice that the integrand looks like it came from the
chain rule, since cos x is the derivative of sin x. Using this observation, we
can work out the antideriative:
d  2
sin x = 2 sin x cos x
dx
 
d 3 2
sin x = 3 sin x cos x
dx 2
3
Z
So, 3 sin x cos xdx = sin2 x + C
2

These two answers look different. Using the identity cos(2x) = 1 − 2 sin2 (x), we
reconcile them:
3 3
− cos(2x) + C = − 1 − 2 sin2 x + C


4 4  
3 2 3
= sin x + C −
2 4

The 34 here is not significant. Remember that C is used to designate a constant

that can take any value between −∞ and +∞. So C − 34 is also just a constant that
can take any value between −∞ and +∞. As the two answers we found differ
by a constant, they are equivalent.
1.3.2.26. Solution. It’s not immediately obvious which function has cos2 x as
its derivative, but we can make the situation a little clearer by using the identity

727
S OLUTIONS TO E XERCISES

1 + cos(2x)
cos2 x = :
2
1
Z Z
2
cos xdx = · (1 + cos(2x)) dx
2
1 1
Z Z
= dx + cos(2x)dx
2 2
1 1
Z
= x+C + cos(2x)dx
2 2

For the remaining integral, we might guess something like F (x) = sin(2x). Let’s
figure out the appropriate constant:

d
{sin(2x)} = 2 cos(2x)
dx
 
d 1 1
sin(2x) = cos(2x)
dx 4 2
1 1
Z
So, cos(2x)dx = sin(2x) + C
2 4
1 1
Z
Therefore, cos2 xdx = x + sin(2x) + C
2 4
1.3.2.27. *. Solution. By the Fundamental Theorem of Calculus Part 1,
Z x
0 d
F (x) = log(2 + sin t) dt = log(2 + sin x)
dx 0
 Z y 
0 d
G (y) = − log(2 + sin t) dt = − log(2 + sin y)
dy 0

So, π  π 
0 0
F = log 3 G = − log(3)
2 2
1.3.2.28. *. Solution. By the Fundamental Theorem of Calculus Part 1,
2 2
f 0 (x) = 100(x2 − 3x + 2)e−x = 100(x − 1)(x − 2)e−x
2
As f (x) is increasing whenever f 0 (x) > 0 and 100e−x is always strictly bigger
than 0, we have f (x) increasing if and only if (x − 1)(x − 2) > 0, which is the case
if and only if (x − 1) and (x − 2) are of the same sign. Both are positive when
x > 2 and both are negative when x < 1. So f (x) is increasing when −∞ < x < 1
and when 2 < x < ∞.
Remark: even without the Fundamental Theorem of Calculus, since f (x) is the
area under a curve from 1 to x, f (x) is increasing when the curve is above the
x-axis (because we’re adding positive area), and it’s decreasing when the curve
is below the x-axis (because we’re adding negative area).

728
S OLUTIONS TO E XERCISES
x
1
Z
1.3.2.29. *. Solution. Write G(x) = dt. By the Fundamental Theorem
0 t3 +6
0 1
of Calculus Part 1, G (x) = 3 . Since F (x) = G(cos x), the chain rule gives us
x +6
sin x
F 0 (x) = G0 (cos x) · (− sin x) = −
cos3 x + 6
Z x
2
1.3.2.30. *. Solution. Define g(x) = et dt. By the Fundamental Theorem of
0
2
Calculus Part 1, g 0 (x) = ex . As f (x) = g(1 + x4 ) the chain rule gives us
4 2
f 0 (x) = 4x3 g 0 (1 + x4 ) = 4x3 e(1+x )
Rx
1.3.2.31. *. Solution. Define g(x) = 0 (t6 + 8)dt. By the fundamental theorem
of calculus, g 0 (x) = x6 + 8. We are to compute the derivative of f (x) = g(sin x).
The chain rule gives
Z sin x 
d
(t + 8)dt = g 0 (sin x) · cos x = sin6 x + 8 cos x
6


dx 0
Z x  
−t πt
1.3.2.32. *. Solution. Let G(x) = e sin dt. By the Fundamen-
0 2

tal Theorem of Calculus Part 1, G0 (x) = e−x sin πx 2
and, since F (x) = G(x3 ),
3 3
F 0 (x) = 3x2 G0 (x3 ) = 3x2 e−x sin πx2 . Then F 0 (1) = 3e−1 sin π2 = 3e−1 .

Z 0 Z x
dt 1
1.3.2.33. *. Solution. Define G(x) = 3
=− 3
dt, so that G0 (x) =
x 1+t 0 1+t
1
− by the Fundamental Theorem of Calculus Part 1. Then by the chain rule,
1 + x3
Z 0 
d dt d d
3
= G(cos u) = G0 (cos u) · cos u
du cos u 1 + t du du
1
=− · (− sin u).
1 + cos3 u
d
Rx
1.3.2.34. *. Solution. Applying dx to both sides of x2 = 1 + 1 f (t)dt gives, by
the Fundamental Theorem of Calculus Part 1, 2x = f (x).
d
Rx
1.3.2.35. *. Solution. Apply dx to both sides of x sin(πx) = 0 f (t) dt. Then, by
the Fundamental Theorem of Calculus Part 1,
Z x
d d
f (x) = f (t) dt = x sin(πx)
dx 0 dx
d
=⇒ f (x) = x sin(πx) = sin(πx) + πx cos(πx)
dx
=⇒ f (4) = sin(4π) + 4π cos(4π) = 4π

729
S OLUTIONS TO E XERCISES

1.3.2.36. *. Solution. (a) Write

Z y Z y
2 −t 2
F (x) = G(x ) − H(−x) with G(y) = e dt, H(y) = e−t dt
0 0

By the Fundamental Theorem of Calculus Part 1,

2
G0 (y) = e−y , H 0 (y) = e−y

Hence, by the chain rule,

2 2 2
F 0 (x) = 2xG0 (x2 ) − (−1)H 0 (−x) = 2xe−(x ) + e−(−x) = (2x + 1)e−x

(b) Observe that F 0 (x) < 0 for x < −1/2 and F 0 (x) > 0 for x > −1/2. Hence F (x)
is decreasing for x < −1/2 and increasing for x > −1/2, and F (x) must take its
minimum value when x = −1/2.
Z y
1.3.2.37. *. Solution. Define G(y) = esin t dt. Then:
0

Z x Z 0 Z x Z x4 −x3
sin t sin t sin t
F (x) = e dt + e dt = e dt − esin t dt
0 x4 −x3 0 0
4 3
= G(x) − G(x − x )

By the Fundamental Theorem of Calculus Part 1,

G0 (y) = esin y

Hence, by the chain rule,

d 4
F 0 (x) = G0 (x) − G0 (x4 − x3 ) x − x3
dx
= G0 (x) − G0 (x4 − x3 ) (4x3 − 3x2 )
4 3
= esin x − esin(x −x ) 4x3 − 3x2

Z y
cos et dt. Then:


1.3.2.38. *. Solution. Define with G(y) =
0

Z −x2 Z −x2 Z 0
t t
cos et dt




F (x) = cos e dt = cos e dt +
x5 0 x5
Z −x2 Z x5
cos et dt − cos et dt



=
0 0
2 5
= G(−x ) − G(x )

By the Fundamental Theorem of Calculus,

G0 (y) = cos ey

730
S OLUTIONS TO E XERCISES

Hence, by the chain rule,

d d 5
F 0 (x) = G0 (−x2 ) − x2 − G0 (x5 ) x
dx dx
= G0 (−x2 ) (−2x) − G0 (x5 ) (5x4 )
2
5

= −2x cos e−x − 5x4 cos ex
Z y√
1.3.2.39. *. Solution. Define with G(y) = sin t dt. Then:
0
Z ex √
F (x) = sin t dt
x
Z ex √ Z 0 √ Z ex √ Z x √
= sin t dt + sin t dt = sin t dt − sin t dt
0 x 0 0
x
= G(e ) − G(x)

By the Fundamental Theorem of Calculus Part 1,

p
G0 (y) = sin y

Hence, by the chain rule,

d x
F 0 (x) = G0 (ex ) e − G0 (x)
dx
= ex G0 (ex ) − G0 (x)
p p
= ex sin(ex ) − sin(x)
1.3.2.40. *. Solution. Splitting up the domain of integration,
Z 5 Z 3 Z 5
f (x)dx = f (x) dx + f (x) dx
1 1 3
Z 3 Z 5
= 3 dx + x dx
1 3
x=3 2 x=5
x
= 3x +
x=1 2 x=3
= 14

y
y = f (x)

x
1 3 5

731
S OLUTIONS TO E XERCISES

Exercises — Stage 3
1.3.2.41. *. Solution. By the chain rule,

d 0 2
(f (x)) = 2f 0 (x) f 00 (x)
dx
so 12 f 0 (x)2 is an antiderivative for f 0 (x) f 00 (x) and, by the Fundamental Theorem
of Calculus Part 2,
Z 2  x=2
0 00 1 0 2 1 1 5
f (x)f (x) dx = (f (x)) = f 0 (2)2 − f 0 (1)2 =
1 2 x=1 2 2 2

Remark: evaluating antiderivatives of this type will occupy the next section, Sec-
tion 1.4.
1.3.2.42. *. Solution. The car stops when v(t) = 30 − 10t = 0, which occurs at
time t = 3. The distance covered up to that time is
Z 3 3
v(t) dt = (30t − 5t2 ) = (90 − 45) − 0 = 45 m.
0 0
Z x
log 1 + et dt. By the Fundamental


1.3.2.43. *. Solution. Define g(x) =
0
Theorem of Calculus Part 1, g 0 (x) = log 1 + ex . But f (x) = g(2x − x2 ), so by the

chain rule,
d 2

f 0 (x) = g 0 (2x − x2 ) · {2x − x2 } = (2 − 2x) · log 1 + e2x−x
dx
2 2 2

Observe that e2x−x > 0 for all x so that 1+e2x−x > 1 for all x and log 1+e2x−x >
0 for all x. Since 2 − 2x is positive for x < 1 and negative for x > 1, f 0 (x) is also
positive for x < 1 and negative for x > 1. That is, f (x) is increasing for x < 1 and
decreasing for x > 1. So f (x) achieves its absolute maximum at x = 1.
R x2 −2x dt R x dt
1.3.2.44. *. Solution. Let f (x) = 0 1+t4 and g(x) = 0 1+t4
. Then g 0 (x) =
1
1+x4
and, since f (x) = g(x2 − 2x), f 0 (x) = (2x − 2)g 0 (x2 − 2x) = 2 1+(xx−1
2 −2x)4 . This

is zero for x = 1, negative for x < 1 and positive for x > 1. Thus as x runs
from −∞ to ∞, f (x) decreases until x reaches 1 and then increases all x > 1.
So the Rminimum of f (x) is achieved for x = 1. At x = 1, x2 − 2x = −1 and
−1 dt
f (1) = 0 1+t 4.

Z x √
1.3.2.45. *. Solution. Define G(x) = sin( t) dt. By the Fundamental Theo-
√ 0
rem of Calculus Part 1, G0 (x) = sin( x). Since F (x) = G(x2 ), and since x > 0, we
have
F 0 (x) = 2xG0 (x2 ) = 2x sin |x| = 2x sin x.
Thus F increases as x runs from to 0 to π (since F 0 (x) > 0 there) and decreases as
x runs from π to 4 (since F 0 (x) < 0 there). Thus F achieves its maximum value at
x = π.

732
S OLUTIONS TO E XERCISES

1.3.2.46. *. Solution. The given sum is of the form

n n
X π  jπ  X
lim sin = lim f (x∗j )∆x
n→∞
j=1
n n n→∞
j=1

with ∆x = πn , x∗j = jπ
n
and f (x) = sin(x). Since x∗0 = 0 and x∗n = π, the right hand
side is the definition (using the right Riemann sum) of
Z π Z π
f (x) dx = sin(x) dx = [− cos(x)]π0 = 2
0 0

where we evaluate the definite integral using the Fundamental Theorem of Cal-
culus Part 2.
1.3.2.47. *. Solution. The given sum is of the form
n n
1X 1 X
lim j = lim f (xj )∆x
n→∞ n 1+ n→∞
j=1 n j=1

with ∆x = n1 , xj = nj and f (x) = 1

1+x
. The right hand side is the definition (using
the right Riemann sum) of
1 1
1
Z Z 1
f (x) dx = dx = log |1 + x| = log 2
0 0 1+x 0

1.3.2.48. Solution.

• F(x), x ≥ 0: We learned quite a lot last semester about curve sketching.

We can use those techniques here. We have to be quite careful about the
sign of x, though. We can only directly apply the Fundamental Theorem of
Calculus Part 1 (as it’s written in your text) when x ≥ 0. So first, let’s graph
the right-hand portion. Notice f (x) has even symmetry–so, if we know one
half of F (x), we should be able to figure out the other half with relative
ease.
Z 0
◦ F (0) = f (t)dt = 0 (so, F (x) passes through the origin)
0
◦ Using the Fundamental Theorem of Calculus Part 1, F 0 (x) > 0 when
0 < x < 1 and when 3 < x < 5; F 0 (x) < 0 when 1 < x < 3. So, F (x) is
decreasing from 1 to 3, and increasing from 0 to 1 and also from 3 to 5.
That gives us a skeleton to work with.

733
S OLUTIONS TO E XERCISES

x
1 3 5

We get the relative sizes of the maxes and mins by eyeballing the area
under y = f (t). The first lobe (from x = 0 to x = 1 has a small positive
area, so F (1) is a small positive number. The next lobe (from x = 1
to x = 3) has a larger absolute area than the first, so F (3) is negative.
Indeed, the second lobe seems to have more than twice the area of the
first, so |F (3)| should be larger than F (1). The third lobe is larger still,
and even after subtracting the area of the second lobe it looks much
larger than the first or second lobe, so |F (3)| < F (5).
◦ We can use F 00 (x) to get the concavity of F (x). Note F 00 (x) = f 0 (x).
We observe f (x) is decreasing on (roughly) (0, 2.5) and (4, 5), so F (x)
is concave down on those intervals. Further, f (x) is increasing on
(roughly) (2.5, 4), so F (x) is concave up there, and has inflection points
at about x = 2.5 and x = 4.
y

y = F (x)

y = f (x)
x
−5 −3 −1 1 3 5

In the sketch above, closed dots are extrema, and open dots are inflec-
tion points.

• F(x), x < 0: Now we can consider the left half of the graph. If you stare at
it long enough, you might convince yourself that F (x) is an odd function.
We can also show this with the following calculation:
Z −x Z 0
F (−x) = f (t)dt = f (t)dt
0 x
as in Example 1.2.10, since f (t) is even,

734
S OLUTIONS TO E XERCISES
Z x
=− f (t)dt
0
= −F (x)

Knowing that F (x) is odd allows us to finish our sketch.

y = F (x)

y = f (x)
x
−5 −3 −1 1 3 5

1.3.2.49. *. Solution. (a) Using the product rule, followed by the chain rule,
followed by the Fundamental Theorem of Calculus Part 1,
x3 +1 Z x3 +1
d
Z
0 2 t3 3 3
f (x) = 3x e dt + x et dt
0 dx 0
Z x3 +1 " Z #
y
3 d 3
= 3x2 et dt + x3 3x2 et dt
 
0 dy 0
y=x3 +1
Z x3 +1
3  h 3 i
= 3x2 et dt + x3 3x2 ey
0 y=x3 +1
Z x3 +1
3   3 3
= 3x2 et dt + x3 3x2 e(x +1)
0
Z x3 +1
2 3 3 +1)3
= 3x et dt + 3x5 e(x
0

(b) In general, the equation of the tangent line to the graph of y = f (x) at x = a is

y = f (a) + f 0 (a) (x − a)

Substituting in the given f (x) and a = −1:

Z 0
3 3
f (a) = f (−1) = (−1) et dt = 0
Z0 0
3
f 0 (a) = f 0 (−1) = 3(−1)2 et dt + 3(−1)5 e0
0
= 0 − 3 = −3

735
S OLUTIONS TO E XERCISES

(x − a) = x − (−1) = x + 1

So, the equation of the tangent line is

y = −3(x + 1) .
1.3.2.50. Solution. Recall that “+C” means that we can add any constant to the
function. Since tan2 x = sec2 x − 1, Students A and B have equivalent answers:
they only differ by a constant.
So, if one is right, both are right; if one is wrong, both are wrong. We check
Student A’s work:
d d
{tan2 x + x + C} = {tan2 x} + 1 + 0 = f (x) − 1 + 1 = f (x)
dx dx
So, Student A’s answer is indeed an anditerivative of f (x). Therefore, both stu-
dents ended up with the correct answer.
Remark: it is a frequent occurrence that equivalent answers might look quite
different. As you are comparing your work to others’, this is a good thing to
keep in mind!
1.3.2.51. Solution.
a When x = 3,
Z 3 Z 3
3
F (3) = 3 sin(t)dt = 27 sin tdt
0 0

Using the Fundamental Theorem of Calculus Part 2,

= 27 [− cos t]t=3
t=0 = 27 [− cos 3 − (− cos 0)]
= 27(1 − cos 3)

b Since the integration is with respect to t, the x3 term can be moved out-
side the integral. That is: for the purposes of the integral, x3 is a constant
(although for the purposes of the derivative, it certainly is not).
Z x Z x
3 3
F (x) = x sin(t)dt = x sin(t)dt
0 0

Using the product rule and the Fundamental Theorem of Calculus Part 1,
Z x
0 3 2
F (x) = x · sin(x) + 3x sin(t)dt
0
= x sin(x) + 3x [− cos(t)]t=x
3 2
t=0
= x3 sin(x) + 3x2 [− cos(x) − (− cos(0))]
= x3 sin(x) + 3x2 [1 − cos(x)]

Remark: Since x and t play different roles in our problem, it’s crucial that they
have different names.
R x This is one reason why R x we should avoid the common
mistake of writing a f (x)dx when we mean a f (t)dt.

736
S OLUTIONS TO E XERCISES

1.3.2.52. Solution. If F (x) is even, then f (x) is odd (by the result of Ques-
tion 1.2.3.20 in Section 1.2). So, F (x) can only be even if f (x) is both even and
odd. By the result in Question 1.2.3.19, Section 1.2, this means F (x) is only even
if f (x) = 0 for all x. Note if f (x) = 0, then F (x) is a constant function. So, it is
certainly even, and it might be odd as well if F (x) = f (x) = 0.
Therefore, if f (x) 6= 0 for some x, then F (x) is not even. It could be odd, or it
could be neither even nor odd. We can come up with examples of both types:
if f (x) = 1, then F (x) = x is an odd antiderivative, and F (x) = x + 1 is an
antiderivative that is neither even nor odd.
Interestingly, the antiderivative of an odd function is always even. The proof
is a little beyond what we might ask you, but is given below for completeness.
The proof goes like this: First, we’ll show that if g(x) is odd, then there is some
antiderivative of g(x) that is even. Then, we’ll show that every antiderivative of
g(x) is even. Z x
So, suppose g(x) is odd and define G(x) = g(t)dt. By the Fundamental The-
0
orem of Calculus Part 1, G0 (x) = g(x), so G(x) is an antiderivative of g(x). Since
g(x) is odd, for any x ≥ 0, the net signed area under the curve along [0, x] is the
negative of the net signed area under the curve along [−x, 0]. So,
Z x Z 0
g(t)dt = − g(t)dt (See Example 1.2.11)
0 −x
Z −x
= g(t)dt
0

By the definition of G(x),

G(x) = G(−x)

That is, G(x) is even. We’ve shown that there exists some antiderivative of g(x)
that is even; it remains to show that all of them are even.
Recall that every antiderivative of g(x) differs from G(x) by some constant. So,
any antiderivative of g(x) can be written as G(x) + C, and G(−x) + C = G(x) + C.
So, every antiderivative of an odd function is even.

1.4 · Substitution
1.4.2 · Exercises

Exercises — Stage 1
1.4.2.1. Solution. (a) This is true: it is an application of Theorem 1.4.2 with
f (x) = sin x and u(x) = ex .
(b) This is false: the upper limit of integration is incorrect. Using Theorem 1.4.6,
the correct form is
Z 1 Z e
x x
sin(e ) · e dx = sin(u)du = − cos(e) + cos(1)
0 1

737
S OLUTIONS TO E XERCISES

= cos(1) − cos(e).

Alternately, we can use the Fundamental Theorem of Calculus Part 2, and our
answer from (a):
Z 1
sin(ex ) · ex dx = [− cos(ex ) + C]10 = cos(1) − cos(e) .
0

1.4.2.2. Solution. The reasoning is not sound: when we do a substitution, we

need to take care of the differential (dx). Remember the method of substitution
comes from the chain rule: there should be a function and its derivative. Here’s
the way to do it:
Z
Problem: Evaluate (2x + 1)2 dx.

Work: We use the substitution u = 2x+1. Then du = 2dx, so dx = 12 du:

1
Z Z
(2x + 1) dx = u2 · du
2
2
1
= u3 + C
6
1
= (2x + 1)3 + C
6
1.4.2.3. Solution. The problem is with the limits of integration, as in Question 1.
Here’s how it ought to go:
Z π
log(log t)
Problem: Evaluate dt.
1 t
Work: We use the substitution u = log t, so du = 1t dt. When t = 1, we
have u = log 1 = 0 and when t = π, we have u = log(π). Then:
π log(π)
cos(log t)
Z Z
dt = cos(u)du
1 t log 1
Z log(π)
= cos(u)du
0
= sin(log(π)) − sin(0) = sin(log(π)).
1.4.2.4. Solution.Perhaps shorter ways exist, but the reasoning here is valid.
Z π/4
Problem: Evaluate x tan(x2 )dx.
0
Work: We begin with the substitution u = x2 , du = 2xdx: If u = x2 ,
then du
dx
= 2x, so indeed du = 2xdx.
π/4 π/4
1
Z Z
2
x tan(x )dx = tan(x2 ) · 2xdx algebra
0 0 2

738
S OLUTIONS TO E XERCISES

π 2 /16
1
Z
= tan udu
0 2

Note that every piece was changed from x to u: integrand, differential,

limits. So
Z π/4 Z 2
2 1 π /16 sin u
x tan(x )dx = du
0 2 0 cos u
sin u
since tan u = cos u
. Now we use the substitution v = cos u, dv =
− sin udu:
π 2 /16 cos(π 2 /16)
1 sin u 1 1
Z Z
du = − dv
2 0 cos u 2 cos 0 v

Note that every piece was changed from u to v: integrand, differential,

limits. So
2
1 cos(π /16) 1
Z π/4 Z
2
x tan(x )dx = − dv
0 2 1 v
since cos(0) = 1
 cos(π2 /16)
1
= − log |v|
2 1
FTC Part 2
1
= − log cos(π 2 /16) − log(1)



2
1
= − log cos(π 2 /16)


2
since log(1) = 0
1.4.2.5. *. Solution. We substitute:

u = sin x,
du = cos x dx,
p √
cos x = 1 − sin2 x = 1 − u2 ,
du du
dx = =√
cos x 1 − u2
u(0) = sin 0 = 0
π  π 
u = sin =1
2 2
So,
x=π/2 u=1
du
Z Z
f (sin x) dx = f (u) √
x=0 u=0 1 − u2

739
S OLUTIONS TO E XERCISES
√
Because the denominator 1 − u2 vanishes when u = 1, this is what is known as
an improper integral. Improper integrals will be discussed in Section 1.12.
1.4.2.6. Solution. Using the chain rule, we see that

d
{f (g(x))} = f 0 (g(x))g 0 (x)
dx
So, f (g(x)) is an antiderivative of f 0 (g(x))g 0 (x). All antiderivatives of
f 0 (g(x))g 0 (x) differ by only a constant, so:
Z
f 0 (g(x))g 0 (x)dx − f (g(x)) = f (g(x)) + C − f (g(x))

=C

That is, our expression simplifies to some constant C.

Remark: since Z
f 0 (g(x))g 0 (x)dt − f (g(x)) = C

we conclude Z
f 0 (g(x))g 0 (x)dt = f (g(x)) + C

which is precisely how we perform substitution on integrals.

Exercises — Stage 2
2 2
1.4.2.7. *. Solution. We write u(x) = ex and find du = u0 (x) dx = 2xex dx. Note
2 2
that u(1) = e1 = e when x = 1, and u(0) = e0 = 1 when x = 0. Therefore:
Z 1
1 x=1
Z
x2 x2
xe cos(e ) dx = cos(u(x))u0 (x) dx
0 2
Zx=0
1 u=e
= cos(u) du
2 u=1
 e
1 1

= sin(u) = sin(e) − sin(1) .
2 1 2

1.4.2.8. *. Solution. Substituting y = x3 , dy = 3x2 dx:

2 8
1 1
Z Z
2 3
x f (x ) dx = f (y) dy =
1 3 1 3
1.4.2.9. *. Solution. Setting u = x3 + 1, we have du = 3x2 dx and so

x2 dx du/3
Z Z
101 =
3
(x + 1) u101
1
Z
= u−101 du
3

740
S OLUTIONS TO E XERCISES

1 u−100
= ·
3 −100
1
=− +C
3 × 100u100
1
=− +C
300(x3 + 1)100
1
1.4.2.10. *. Solution. Setting u = log x, we have du = x
dx and so
e4 x=e4 u=4
dx 1 1 1
Z Z Z
= · dx = du,
e x · log x x=e log x x u=1 u

since u = log(e) = 1 when x = e and u = log(e4 ) = 4 when x = e4 . Then, by the

Fundamental Theorem of Calculus Part 2,
Z 4
1 h i4
du = log |u| = log 4 − log 1 = log 4.
1 u 1

1.4.2.11. *. Solution. Setting u = 1 + sin x, we have du = cos xdx and so

Z π/2 Z x=π/2 Z u=2
cos x 1 du
dx = cos xdx =
0 1 + sin x x=0 1 + sin x u=1 u

since u = 1 + sin 0 = 1 when x = 0 and u = 1 + sin(π/2) = 2 when x = π/2. Then,

by the Fundamental Theorem of Calculus Part 2,
u=2
du h
Z i2
= log |u| = log 2
u=1 u 1

1.4.2.12. *. Solution. Setting u = sin x, we have du = cos xdx and so

Z π/2 Z x=π/2
2
cos x · (1 + sin x)dx = (1 + sin2 x) · cos xdx
0
Zx=0
u=1
= (1 + u2 ) du,
u=0

since u = sin 0 = 0 when x = 0 and u = sin(π/2) = 1 when x = π/2. Then, by the

Fundamental Theorem of Calculus Part 2,
Z 1 1 
u3
 
2 1 4
(1 + u ) du = u + = 1+ −0= .
0 3 0 3 3

1.4.2.13. *. Solution. Substituting t = x2 − x, dt = (2x − 1) dx and noting that

t = 0 when x = 1 and t = 6 when x = 3,
Z 3 Z 6
x2 −x
 6
(2x − 1)e dx = et dt = et 0 = e6 − 1
1 0

741
S OLUTIONS TO E XERCISES

1.4.2.14. *. Solution. We use the substitution u = 4 − x2 , for which

du = −2x dx:
x2 − 4 1 4 − x2
Z Z
√ x dx = ·√ (−2x) dx
4 − x2 2 4 − x 2

1 u
Z
= √ du
2 u
1 √
Z
= u du
2
1 u3/2
= +C
2 3/2
1
= (4 − x2 )3/2 + C
3
1.4.2.15. Solution.
√ 1
• Solution 1: If we let u = log x, then du = √ dx, and:
2x log x
√
e log x √
Z Z
log x
√ dx = eu du = eu + C = e +C
2x log x

• Solution 2: In Solution 1, we made a pretty slick choice. We might have

tried to work with something a little less convenient. For example, it’s not
1
unnatural to think that u = log x, du = dx would be a good choice. In that
x
case:
√ Z √u
e log x e
Z
√ dx = √ du
2x log x 2 u
√ 1
Now, we should be able to see that w = u, dw = √ du is a good choice:
2 u
√
e u
Z Z
√ du = ew dw
2 u
√
u
=e +C
√
log x
=e +C

Exercises — Stage 3
1.4.2.16. *. Solution.
d n x2 o 2 1 2
• The slightly sneaky method: We note that e = 2x ex , so that ex is
2
dx 2
a antiderivative for the integrand xex . So
Z 2  2
x2 1 x2 1 1
xe dx = e = e4 − e4 = 0
−2 2 −2 2 2

742
S OLUTIONS TO E XERCISES
2
• The really sneaky method: The integrand f (x) = xex is an odd function
(meaning
R a that f (−x) = −f (x)). So by Theorem 1.2.12 every integral of the
2
form −a xex dx is zero.

1.4.2.17. *. Solution. The given sum is of the form

n n
X j  j2  X
lim 2
sin 1 + 2 = lim f (x∗j )∆x
n→∞
j=1
n n n→∞
j=1

with ∆x = n1 , x∗j = nj and f (x) = x sin(1 + x2 ). Since x∗0 = 0 and x∗n = 1, the right
hand side is the definition (using the right Riemann sum) of
Z 1 Z 1
f (x) dx = x sin(1 + x2 ) dx
0 0
1 2
Z
= sin(y) dy with y = 1 + x2 , dy = 2x dx
2 1
1h iy=2
= − cos(y)
2 y=1
1
= [cos 1 − cos 2]
2
Using a calculator, we see this is close to 0.478.
1.4.2.18. Solution. Often, the denominator of a function is a good guess for the
substitution. So, let’s try setting w = u2 + 1. Then dw = 2udu:
1 1
u3 1 u2
Z Z
2
du = 2udu
0 u +1 2 0 u2 + 1

The numerator now is u2 , and looking at our substitution, we see u2 = w − 1:

1 2 w−1
Z
= dw
2 1 w
1 2
Z  
1
= 1− dw
2 1 w
1
= [w − log |w|]w=2
w=1
2
1 1 1
= (2 − log 2 − 1) = − log 2
2 2 2
1.4.2.19. Solution. The only thing we really have to work with is a tangent,
so it’s worth considering what would happen if we substituted u = tan θ. Then
du = sec2 θdθ. This doesn’t show up in the integrand as it’s written, but we can
try and bring it out by using the identity tan2 = sec2 θ − 1:
Z Z
tan θ dθ = tan θ · tan2 θdθ
3

743
S OLUTIONS TO E XERCISES
Z
tan θ · sec2 θ − 1 dθ


=
Z Z
2
= tan θ · sec θdθ − tan θdθ
R
In Example 1.4.17, we learned tan θdθ = log | sec θ| + C
Z
= u du − log | sec θ| + C
1
= u2 − log | sec θ| + C
2
1
= tan2 θ − log | sec θ| + C
2
1.4.2.20. Solution. At first glance, it’s not clear what substitution to use. If we
try the denominator, u = ex + e−x , then du = (ex − e−x )dx, but it’s not clear how
to make this work with our integral. So, we can try something else.
If we want to tidy things up, we might think to take u = ex as a substitution.
Then du = ex dx, so we need an ex in the numerator. That can be arranged.
 x
1 e ex
Z Z
· dx = dx
ex + e−x ex (ex )2 + 1
1
Z
= 2
du
u +1
= arctan(u) + C
= arctan(ex ) + C
1.4.2.21. Solution. We often like to take the “inside” function as our substitu-
tion, in this case u = 1 − x2 , so du = −2xdx. This takes care of part of the integral:
Z 1 √ Z 1 √ Z 1 √
(1 − 2x) 1 − x2 dx = 1 − x2 dx + (−2x) 1 − x2 dx
0 0 0

The left integral is tough to solve with substitution, but luckily we don’t have
to–it’s the area of a quarter of a circle of radius 1.
Z 0
π √
= + u du
4 1
 u=0
π 2 3/2
= + u
4 3 u=1
π 2 π 2
= +0− = −
4 3 4 3
1.4.2.22. Solution.

• Solution 1: We often find it useful to take “inside” functions as our substi-

tutions, so let’s try u = cos x, du = − sin xdx. In order to dig up a sine, we

744
S OLUTIONS TO E XERCISES

sin x
use the identity tan x = :
cos x
− sin x
Z Z
tan x · log (cos x) dx = − · log (cos x) dx
cos x
1
Z
=− log(u)du
u

Now, it is convenient to let w = log u, dw = u1 du:

1
Z Z
− log(u)du = − w dw
u
1
= − w2 + C
2
1
= − (log u)2 + C
2
1
= − (log(cos x))2 + C
2

• Solution 2: We might guess that it’s useful to have u = log(cos x),

− sin x
du = dx = − tan xdx:
cos x
Z Z
tan x · log (cos x)dx = − − tan x · log (cos x)dx
Z
= − u du
1
= − u2 + C
2
1
= − (log(cos x))2 + C
2
1.4.2.23. *. Solution. The given sum is of the form
n  j2  n
X j X
lim 2
cos 2 = lim f (x∗j )∆x
n→∞
j=1
n n n→∞
j=1

with ∆x = n1 , x∗j = nj and f (x) = x cos(x2 ). Since x∗0 = 0 and x∗n = 1, the right
hand side is the definition (using the right Riemann sum) of
Z 1 Z 1
f (x) dx = x cos(x2 ) dx
0 0
1 1
Z
= cos(y) dy with y = x2 , dy = 2x dx
2 0
1h i1
= sin(y)
2 0
1
= sin 1
2

745
S OLUTIONS TO E XERCISES

1.4.2.24. *. Solution. The given sum is of the form

n n
r
X j j2 X
lim 2
1 + 2
= lim f (x∗j )∆x
n→∞
j=1
n n n→∞
j=1

√
with ∆x = n1 , x∗j = nj and f (x) = x 1 + x2 . Since x∗0 = 0 and x∗n = 1, the right
hand side is the definition (using the right Riemann sum) of
Z 1 Z 1 √
f (x) dx = x 1 + x2 dx
0 0
1 2√
Z
= y dy with y = 1 + x2 , dy = 2x dx
2 1
 y=2
1 2 3/2
= y
2 3 y=1
1 √
= [2 2 − 1]
3
Using a calculator, we see this is approximately 0.609.
1.4.2.25. Solution. Using the definition of a definite integral with right Riemann
sums:
Z b n
X b−a
2f (2x)dx = lim ∆x · 2f (2(a + i∆x)) ∆x =
a n→∞
i=1
n
n      
X b−a b−a
= lim · 2f 2 a + i
n→∞
i=1
n n
n 
X 2b − 2a    
2b − 2a
= lim · f 2a + i
n→∞
i=1
n n
Z 2b n
X 2b − 2a
f (x)dx = lim ∆x · f (2a + i∆x) ∆x =
2a n→∞
i=1
n
n     
X 2b − 2a 2b − 2a
= lim · f 2a + i
n→∞
i=1
n n

Since the Riemann sums are exactly the same,

Z b Z 2b
2f (2x)dx = f (x)dx
a 2a

Looking at the Riemann sum in this way is instructive, because it is very clear
why the two integrals should be equal (without using substitution). The rectan-
gles in the first Riemann sum are half as wide, but twice as tall, as the rectangles
in the second Riemann sum. So, the two Riemann sums have rectangles of the
same area.

746
S OLUTIONS TO E XERCISES

2f (x∗i )

f (x∗i )

b−a
n
2 b−a
n

(Not every substitution corresponds to such a simple picture.)

1.5 · Area between curves

1.5.2 · Exercises

Exercises — Stage 1
1.5.2.1. Solution.

x
π π 3π π
4 2 4
y = sin x
y = cos x

The intervals of our rectangles are [0, π4 ], [ π4 , π2 ], [ π2 , 3π

4
], and [ 3π
4
, π]. Since we’re tak-
ing a left Riemann sum, we find the height of the rectangles at the left endpoints
of the intervals.
• x = 0: The distance from cos 0 to sin 0 is 1, so our first rectangle has height
1.
• x = π4 : The distance from cos π4 to sin π4 is 0, so our second rectangle has
height 0.
• x = π2 : The distance from cos π2 to sin π2 is 1, so our third rectangle has height
1.
3π
• x= 4
: The distance from cos 3π
4
to sin 3π
4
is sin(3π/4) − cos(3π/4) = √1
2
−

747
S OLUTIONS TO E XERCISES
  √ √
− √12 = 2, so our fourth rectangle has height 2.

So, our approximation for the area between the two curves is
π √  π √ 
1+0+1+ 2 = 2+ 2
4 4
1.5.2.2. Solution.

a We are finding the area in the interval from x = 0 to x = π2 . Since we’re

taking n = 5 rectangles, our rectangles cover the following intervals:
h πi h π πi      
π 3π 3π 2π 2π π
0, , , , , , , , , .
10 10 5 5 10 10 5 5 2

y
p xπ
y= 2

2x


y = arcsin π

x
π π 3π 2π π
10 5 10 5 2

b We are finding the area in the interval from y = 0 to y = π2 . (In general,

when we switch from horizontal rectangles to vertical, the limits of inte-
gration will change–it’s only coincidence that they are the same in this ex-
ample.) Since we’re taking n = 5 rectangles, these rectangles cover the
following intervals of the y-axis:
h πi h π πi      
π 3π 3π 2π 2π π
0, , , , , , , , , .
10 10 5 5 10 10 5 5 2

The question doesn’t specify which endpoints we’re using. Let’s use upper
endpoints, to match part (a).

748
S OLUTIONS TO E XERCISES

π
2

2π x = π2 y 2
5

3π
10

π
5

π
10 π
x= 2
sin y

1.5.2.3. *. Solution. The curves intersect when y = x and y = x3 − x. To find

these points, we set:

x = x3 − x
0 = x3 − 2x
0 = x(x2 − 2)
0 = x or 0 = x2 − 2
√ √
For x ≥ 0, the curves intersect at (0, 0) and ( 2, 2).
A handy observation is that, since both curves √ are continuous and they do not
meet each other between x = 0 and x = 2, we don’t have to worry about
dividing our area into two regions: one of the functions is always on the top, and
the other is always on the bottom.
Using vertical strips:

The top and bottom boundaries of the specified region are y = T (x) = x and

749
S OLUTIONS TO E XERCISES

y = B(x) = x3 − x, respectively. So,

√ √
Z 2 Z 2
x − (x3 − x) dx
   
Area = T (x) − B(x) dx =
0 0
√
Z 2
2x − x3 dx
 
=
0

1.5.2.4. *. Solution. We need to find where the curves intersect.

x2 5x
= y2 = 6 −
4 4
1 2 5
x + x−6=0
4 4
x2 + 5x − 24 = 0
(x + 8)(x − 3) = 0
x = −8, x = 3

The curves intersect at (−8, 4) and (3, − 23 ). Using horizontal strips:

we have
Z 4 h4 i
Area = (6 − y 2 ) + 2y dy
−3/2 5

1.5.2.5. *. Solution. If the curves intersect at (x, y), then

2
x2 = (4a)2 y 2 = (4a)2 4ax
x4 = (4a)3 x
x4 − (4a)3 x = 0
x(x3 − (4a)3 ) = 0
x = 0 or x3 = (4a)3

The curves intersect at (0, 0) and (4a, 4a). (It is also possible to find these points
by inspection.) Using vertical strips:

750
S OLUTIONS TO E XERCISES

√
We want the y-values of the functions. We write the top function as y = 4ax
(we care about the positive square root, not the negative one) and we write the
2
bottom function as y = x4a . Then we have
4a √ x2
Z  
Area = 4ax − dx
0 4a
1.5.2.6. *. Solution. The curves intersect when x = 4y 2 and 0 = 4y 2 + 12y + 5 =
(2y + 5)(2y + 1). So, the curves intersect at (1, − 12 ) and (25, − 25 ). Using vertical
strips:

we have
25
1√
 
1
Z
Area = − (x + 5) + x dx
1 12 2

Exercises — Stage 2
1.5.2.7. *. Solution.
y
1
y= (2x−4)2

x
1

751
S OLUTIONS TO E XERCISES

1
The area between the curve y = (2x−4) 2 and the x-axis, with x running from a = 0
1
to b = 1, is exactly the definite integral of (2x+4)2 with limits 0 and 1.

1
dx
Z
Area = u = 2x − 4, du = 2dx
0 (2x − 4)2
 u=−2
1 −2 1 1 −1
Z
= du =
2 −4 u2 2 u u=−4
1 1 1
h i 1
= − =
2 2 4 8
1.5.2.8. *. Solution. If the curves y = f (x) = x and y = g(x) = 3x − x2 intersect
at (x, y), then

3x − x2 = y = x
x2 − 2x = 0
x(x − 2) = 0
x = 0 or x = 2

Furthermore, g(x) − f (x) = 2x − x2 = x(2 − x) is positive for all 0 ≤ x ≤ 2. That

is, the curve y = 3x − x2 lies above the line y = x for all 0 ≤ x ≤ 2.
y=x
y

y = 3x − x2

x
2

We therefore evaluate the integral:

2 2 2
x3
Z Z 
(3x − x2 ) − x dx = 2 2
 
[2x − x ] dx = x −
0 3
0  0
8 4
= 4− −0=
3 3
1.5.2.9. *. Solution. By inspection, the two curves cross at (0, 1) and (1, 2).

752
S OLUTIONS TO E XERCISES

y √
y= x+1

y = 2x

x
1

x
To antidifferentiate 2x , we write 2x = (elog 2 ) = ex log 2 .
1 1
√

2 3/2 1 x
Z
x log 2

Area = ( x + 1) − e dx = x + x − 2
0 3 log 2 0
2 1 5 1
= +1− [2 − 1] = −
3 log 2 3 log 2
1.5.2.10. *. Solution. Here is a sketch of the specified region.

Both functions are even, so the region is symmetric about the y–axis. So, we
compute the area of the part with x ≥ 0 and
will √ √ multiply by 2. The curves
y = 2 cos(πx/4) and y = x intersect when x = 2 cos(πx/4) or cos(πx/4) = √x2 ,
which is the case a when x = 1. So, using vertical strips as in the figure above,
the area (including the multiplication by 2) is
1
1 √ √ 4 x2
Z 

2 2 cos(πx/4) − x dx = 2 2 sin(πx/4) −
0 π 2 0
 
4 1 8
=2 − = −1
π 2 π

a The solution x = 1 was found by guessing. To guess a solution to cos(πx/4) = √x2 just
√
ask yourself what simple angle has a cosine that involves 2. This guessing strategy is
essentially useless in the real world, but works great on problem sets and exams.

753
S OLUTIONS TO E XERCISES

1.5.2.11.
√ *. Solution. For our computation, we will need an antiderivative of
2
x x + 1, which can be found using the substitution u = x3 + 1, du = 3x2 dx:
3

Z
2
√ Z
√ 1 1
Z
3
x x + 1 dx = u · du = u1/2 du
3 3
1 u3/2 2
= · + C = (x3 + 1)3/2 + C.
3 3/2 9

The two functions f (x) and g(x) are clearly equal at x = 0. If x 6= 0, then the
functions are equal when
√
3x2 = x2 x3 + 1
√
3 = x3 + 1
9 = x3 + 1
8 = x3
2 = x.

The function g(x) = 3x2 is the larger of the two on the interval [0, 2], as can
be seen by
2
√ plugging 2in x = 1, say, or
2
by observing that when x is very small
3
f (x) = x x + 1 ≈ x and g(x) = 3x .

The area in question is therefore:

2
2 √

2
Z
2 2 3 3 3/2


3x − x x3 + 1 dx = x − (x + 1)
0 9 0
   
3 2 3 3/2 3 2 3 3/2
= 2 − (2 + 1) − 0 − (0 + 1)
9 9
   
2 20
= 8−6 − 0− = .
9 9
1.5.2.12. *. Solution. First, let’s figure out what our curve x = y 2 + y = y(y + 1)
looks like.

• The curve intercepts the y-axis when y = 0 and y = −1.

• The x-values of the curve are negative when −1 < y < 0, and positive
elsewhere.

754
S OLUTIONS TO E XERCISES

This leads to the figure below. We’re evaluating the area from y = −1 to y = 0.
Since y 2 + y is negative there, the length of our (horizontal) slices are 0 − (y 2 + y).

0  3 0
y y2 1 1 1
Z
2


Area = 0 − (y + y) dy = − + =− + =
−1 3 2 −1 3 2 6

√
1.5.2.13. Solution.
√ Let’s begin by sketching our region. Note that y = 1 − x2
and y = 9 − x2 are the top halves of circles centred at the origin with radii 1
and 3, respectively.

y = |x|

√
y= 9 − x2
x

Our region is the difference of two quarter-circles, so we find its area using ge-
ometry:
1
1
π · 32 − π · 12 = 2π


Area =
4 4

Exercises — Stage 3
1.5.2.14. *. Solution. We will compute the area by using thin vertical strips, as
in the sketch below:

755
S OLUTIONS TO E XERCISES

By looking at the sketch above, we guess the line y = 4 + 2π − 2x intersects the

curve y = 4 + π sin x when x = π2 , x = π, and x = 3π
2
. Let’s make sure these are
correct by plugging them into the two equations, and making sure the y-values
match:

x 4 + 2π − 2x 4 + π sin(x) match?
π
2
4+π 4+π X
π 4 4 X
3π
2
4−π 4−π X

Also from the sketch, we see that:

• When π2 ≤ x ≤ π, the top of the strip is at y = 4 + π sin x and the bottom of

the stripis at y = 4 + 2π − 2x. So the strip
 has height (4 + π sin x) −  (4 +
2π − 2x) and width dx, and hence area (4 + π sin x) − (4 + 2π − 2x) dx.

• When π ≤ x ≤ 3π 2
, the top of the strip is at y = 4 + 2π − 2x and the bottom
of the strip
 is at y = 4 + π sin x. So the strip
 has height (4 + 2π − 2x) − (4 +
π sin x) and width dx, and hence area (4 + 2π − 2x) − (4 + π sin x) dx.

Now we can calculate:

Z π
 
Area = (4 + π sin x) − (4 + 2π − 2x) dx
π/2
Z 3π/2
 
+ (4 + 2π − 2x) − (4 + π sin x) dx
π
Z π Z 3π/2
   
= π sin x − 2π + 2x dx + 2π − 2x − π sin x dx
π/2 π
h iπ h i3π/2
2 2
= − π cos x − 2πx + x + 2πx − x + π cos x
π/2 π
   
2 3 2 2 5 2
= π−π + π + π − π +π
4 4
h 1 2i
=2 π− π
4

756
S OLUTIONS TO E XERCISES

1.5.2.15. *. Solution. First, here is a sketch of the region. We are not asked for
it, but it is crucial for understanding the question.

The two curves y = x + 2 and y = x2 cross at (2, 4). The area of the part between
them with 0 ≤ x ≤ 2 is:
Z 2 h1 1 i2 8 10
x + 2 − x2 dx = x2 + 2x − x3 = 2 + 4 − =
 
0 2 3 0 3 3

The area of the part between the two curves with 2 ≤ x ≤ 3 is:
3 h1 1
Z i3
x2 − (x + 2) dx = x3 − x2 − 2x
 
2 3 2 2
9 8 11
=9− −6− +2+4=
2 3 6
10 11 31
The total area is + = .
3 6 6
1.5.2.16. *. Solution. We√need to figure out which curve is on top, when. To
do this, set h(x) = 3x 2
√ − x 25 − x . If h(x) > 0, then y = 3x is the top curve; if
h(x) < 0, then y = x 25 − x2 is the top curve.
√ h √ i
h(x) = 3x − x 25 − x2 = x 3 − 25 − x2

We only care about values of x in [0, 4], so x is nonnegative. Then h(x) is positive
when:
√
3 > 25 − x2
9 > 25 − x2
x2 > 16
x>4
√
That is, h(x) is never positive over the interval [0, 4]. So, y = x 25 − x2 lies above
y = 3x for all 0 ≤ x ≤ 4.
The area we need to calculate is therefore:
Z 4h √ i
A= 2
x 25 − x − 3x dx
0

757
S OLUTIONS TO E XERCISES
Z 4 √ Z 4
= x 25 − x2 dx − 3x dx
0 0
= A1 − A2 .

To evaluate A1 , we use the substitution u(x) = 25 − x2 , for which du = u0 (x) dx =

−2x dx; and u(4) = 25 − 42 = 9 when x = 4, while u(0) = 25 − 02 = 25 when
x = 0. Therefore
Z x=4 √
1 u=9 √
Z
A1 = 2
x 25 − x dx = − u du
x=0 2 u=25
 9
1 3/2 125 − 27 98
= − u = =
3 25 3 3

For A2 we use the antiderivative directly:

4 4
3x2
Z 
A2 = 3x dx = = 24
0 2 0

Therefore the total area is:

98 26
A= − 24 =
3 3
√
1.5.2.17. Solution. Let’s begin by sketching our region. Note that yp= 9 − x2 is
the top half of a circle centred at the origin with radius 3, while y = 1 − (x − 1)2
is the top half of a circle of radius 1 centred at (1, 0).
y

p
Note y = x intersects y = 1 − (x − 1)2 at (1, 1), the highest part of the smaller
half-circle.
We can easily take the area of triangles and sectors of circles. With that in mind,
we cut up our region the following way:

758
S OLUTIONS TO E XERCISES

A3

x
A1 A2

• The desired area is A3 − (A1 + A2 ).

• A1 is the area of right a triangle with base 1 and height 1, so A1 = 12 .

• A2 is the area of a quarter circle of radius 1, so A2 = π4 .

9π
• A3 is the area of an eighth of a circle of radius 3, so A2 = 8

9π 1 π 7π 1
So, the area of our region is − − = − .
8 2 4 8 2
1.5.2.18. Solution. The first function is a cubic, with intercepts at x = 0, ±2. The
second is a straight line with a positive slope.
We need to figure out what these functions look like in relation to one another,
so let’s find their points of intersection.

x(x2 − 4) = x − 2
x(x + 2)(x − 2) = x − 2
x−2=0 or x(x + 2) = 1
x2 + 2x − 1 = 0
p
−2 ±4 − 4(1)(−1)
x=
2
√
x = −1 ± 2
√
So, our three points of intersection are when x = 2 and when x = −1 ± 2. We
note √ √ √
−1 − 2 < −1 + 2 < −1 + 4 < 2 .
So,
 we
√ need to
√ see which
 function
√  is on top over the two intervals
−1 − 2, −1 + 2 and −1 + 2, 2 . It suffices to check points in these inter-
vals.

x x(x2 − 4) x−2 top function:

0 0 −2 x(x2 − 4)
1 -3 −1 x−2

759
S OLUTIONS TO E XERCISES
√ √ 
Since 0 is in the interval −1 − 2, −1 + 2 , x(x2 − 4) is the top function in that

 √ 
interval. Since 1 is in the interval −1 + 2, 2 , x − 2 is the top function in that
interval. Now we can set up the integral to evaluate the area:
√
Z −1+ 2
x(x2 − 4) − (x − 2) dx
 
Area = √
−1− 2
Z 2
(x − 2) − x(x2 − 4) dx
 
+ √
−1+ 2
√
Z −1+ 2
x3 − 5x + 2 dx
 
= √
−1− 2
Z 2  3 
+ √
−x + 5x − 2 dx
−1+ 2
 −1+√2
1 4 5 2
= x − x + 2x
4 2 √
−1− 2
 2
1 4 5 2
+ − x + x − 2x
4 2 √
−1+ 2

After some taxing but rudimentary algebra:

 √   √ 13

√ 13
= 8 2 + 4 2− = 12 2 −
4 4

1.6 · Volumes
1.6.2 · Exercises

Exercises — Stage 1
1.6.2.1. Solution. If we take a horizontal slice of a cone, we get a circle. If
we take a vertical cross-section, the base is flat (it’s a chord on the circular base
of the cone), so we know right away it isn’t a circle. Indeed, if we slice down
through the very centre, we get a triangle. (Other vertical slices have a curvy top,
corresponding to a class of curves known as hyperbolas.)
1.6.2.2. Solution. The columns have the same volume. We can see this by chop-
ping up the columns into horizontal cross-sections. Each cross-section has the
same area as the cookie cutter, A, and height dy. Then in both cases, the volume
of the column is Z h
Ady = hA cubic units
0

1.6.2.3. Solution. Notice f (x) is a piecewise linear function, so we can find ex-
plicit equations for each of its pieces from the graph. The radii will be determined
by the x-values, so below we give the x-values as functions of y.

760
S OLUTIONS TO E XERCISES

1 y = f (x)

x
1 2 4 6
x=y

x=2−y

x = 2 + 23 y

x = 6 − 23 y
If we imagine rotating the region from the picture about the y-axis, there will
be two kinds of washers formed: when y < 1, we have a “double washer,” two
concentric rings. When y > 1, we have a single ring.

• Washers when 1 < y ≤ 6: If y > 1, then our washer has inner radius 2 + 23 y,
outer radius 6 − 23 y, and height dy.

r = 2 + 23 y

thickness: dy

R = 6 − 32 y

• Washers when 0 ≤ y < 1: When 0 ≤ y < 1, we have a “double washer,”

two concentric rings corresponding to the two “humps” in the function.
The inner washer has inner radius r1 = y and outer radius R1 = 2 − y. The
outer washer has inner radius r2 = 2 + 23 y and outer radius R2 = 6 − 23 y.
The thickness of the washers is dy.

761
S OLUTIONS TO E XERCISES

r2 = 2 + 23 y

r1 = y

thickness: dy

R1 = 2 − y

R2 = 6 − 23 y

1.6.2.4. *. Solution. (a) When the strip shown in the figure

√ 2
is rotated about the x–axis, it forms a thin disk of radius xex and thickness dx
2 2
and hence of cross sectional area πxe2x and volume πxe2x dx So the volume of
the solid is
Z 3
2
π xe2x dx
0

(b) The curves intersect at (−1, 1) and (2, 4).

762
S OLUTIONS TO E XERCISES

We’ll use horizontal washers as in Example 1.6.5.

• We use thin horizontal strips of width dy as in the figure above.

• When we rotate about the line x = 3, each strip sweeps out a thin washer
√
◦ whose inner radius is rin = 3 − y, and
◦ whose outer radius is rout = 3 − (y − 2) = 5 − y when y ≥ 1 (see the
red strip in the figure on the right above), and whose outer radius is
√ √
rout = 3 − (− y) = 3 + y when y ≤ 1 (see the blue strip in the figure
on the right above) and
◦ whose thickness is dy and hence
2 2

2 √
2 
◦ whose volume is π(rout −rin )dy = π 5−y − 3− y dy when y ≥ 1
2 2
 √
2 √
2 
and whose volume is π(rout −rin )dy = π 3+ y − 3− y dy when
y ≤ 1 and

• As our bottommost strip is at y = 0 and our topmost strip is at y = 4, the

total volume is
Z 1
 √
2 √
2 
π 3 + y − 3 − y dy
0
Z 4

2 √
2 
+ π 5 − y − 3 − y dy
1

1.6.2.5. *. Solution. (a) The curves intersect at (1, 0) and (−1, 0). When the strip
shown in the figure

763
S OLUTIONS TO E XERCISES

is rotated about the line y = −1, it forms a thin washer with:

• inner radius (1 − x2 ) − (−1) = 2 − x2 ,

• outer radius (4 − 4x2 ) − (−1) = 5 − 4x2 and

• thickness dx ; so, it has

2 2
• cross sectional area π (5 − 4x2 ) − (2 − x2 ) and


2 2
• volume π (5 − 4x2 ) − (2 − x2 ) dx.


So the volume of the solid is

Z 1
2 2
π (5 − 4x2 ) − (2 − x2 ) dx

−1

(b) The curve y = x2 − 1 intersects y = 0 at (1, 0) and (−1, 0).

We’ll use horizontal washers.

• We use thin horizontal strips of height dy as in the figure above.

• When we rotate about the line x = 5, each strip sweeps out a thin washer
√
◦ whose inner radius is rin = 5 − y + 1, and
√ √
◦ whose outer radius is rout = 5 − (− y + 1) = 5 + y + 1 and
◦ whose thickness is dy and hence
2 2
 √
2 √
2 
◦ whose volume is π(rout − rin ) dy = π 5 + y + 1 − 5 − y + 1 dy

• As our topmost strip is at y = 0 and our bottommost strip is at y = −1

(when x = 0), the total volume is
Z 0 h p
2 p
2 i
π 5+ y+1 − 5− y+1 dy
−1

1.6.2.6. *. Solution. The curves intersect at (−2, 4) and (2, 4). When the strip
shown in the figure

764
S OLUTIONS TO E XERCISES

is rotated about the line y = −1, it forms a thin washer (punctured disc) of

• inner radius x2 + 1,

• outer radius 9 − x2 and

• thickness dx and hence of

2 2
• cross sectional area π (9 − x2 ) − (x2 + 1) and


2 2
• volume π (9 − x2 ) − (x2 + 1) dx.


So the volume of the solid is

Z 2
2 2
(9 − x2 ) − (x2 + 1) dx

π
−2

1.6.2.7. Solution. We’ll make horizontal slices, parallel to one of the faces of the
tetrahedron. Then our slices will be equilateral triangles, of varying sizes.

For the sake of ease, as in Example 1.6.1, we picture the tetrahedron perched on
a tip, one base horizontal on top.

765
S OLUTIONS TO E XERCISES

y
q
2
3
`

x
− 2` `
2

Notice our slice forms the horizontal top of a smaller tetrahedron.

q The horizontal
top of the full tetrahedron has side length `, which is 32 times the height of the
q slice is the horizontal top of a tetrahedron of height y and
full tetrahedron. Our
3
so has side length 2
y.
An equilateral triangle with side length L has base L and
√ √ q
height 23 L, and hence area 43 L2 . So, the area of our slice with side length 32 y
is
√ r !2 √
3 3 3 3 2
A= y = y
4 2 8
So, the volume of a tetrahedron with side length ` is:
Z √2` √
3 3 3 2
Volume = y dy
0 8
√ r !3 √
3 2 2 3
= · ` = `
8 3 12

You were given the height of a tetrahedron, but for completeness we calculate it here.
Draw a line starting at one tip, and dropping straight down to the middle of the
opposite face. It forms a right triangle with one edge of the tetrahedron, and a
line from the middle of the face to the corner.

766
S OLUTIONS TO E XERCISES

A C
c

We know the length of the hypotenuse of this right triangle (it’s `), so if we know
the length of its base (labeled Ac in the diagram), we can figure out its third side,
the height of our tetrahedron. Note by using the Pythagorean qtheorem, we see
3
that the height of an equilateral triangle with edge length ` is 2
`.
Here is a sketch of the base of the pyramid:

b
A C

c
B

The triangles ABC and Abc are similar (since b and B are right angles, and also
A has the same angle in both). Therefore,

Ac AC
=
Ab AB
Ac `
=√
`/2 3`/2
1
Ac = √ `
3
With
r this in our pocket, we can find the height of the tetrahedron:
 2 q
`2 − √13 ` = 23 `.

Exercises — Stage 2

767
S OLUTIONS TO E XERCISES
√ 2
1.6.2.8. *. Solution. Let f (x) = 1 + xex . On the vertical slice a distance x from
the y-axis, sketched in the figure below, y runs from 1 to f (x). Upon rotation
about the line y = 1, this thin slice sweeps out a thin disk of thickness dx and
radius f (x) − 1 and hence of volume π[f (x) − 1]2 dx. The full volume generated
(for any fixed a > 0) is
Z a Z a
2 2
π[f (x) − 1] dx = π xe2x dx.
0 0

Using the substitution u = 2x2 , so that du = 4x dx:

2a2 2a2
du π π  2a2
Z 
Volume = π eu = eu = e −1
0 4 4 0 4

Remark: we spent a good deal of time last semester developing highly accurate
but time-consuming methods for sketching common functions. For the purposes
of questions like this, we√don’t need a detailed picture of a function–broad out-
x2
√ x2suffice. Notice that x > 0 whenever x > 0, and e > 0 for all x. Therefore,
lines √ 2
xe is nonnegative over its entire domain, and so the graph y = 1 + xex
is always the top function, above the bottom function y = 1. That is the only
information we needed to perform our calculation.
10
1.6.2.9. *. Solution. The curves y = 1/x and 3x+3y = 10, i.e. y = 3
−x intersect
when
1 10
= − x ⇐⇒ 3 = 10x − 3x2 ⇐⇒ 3x2 − 10x + 3 = 0
x 3
⇐⇒ (3x − 1)(x − 3) = 0
1
⇐⇒ x = 3 ,
3

768
S OLUTIONS TO E XERCISES

When the region is rotated about the x–axis, the vertical strip in the figure above
sweeps out a washer with thickness dx, outer radius T (x) = 10 3
− x and inner
radius B(x) = x1 . This washer has volume
 10020 1
2 2 2


π T (x) − B(x) dx = π − x + x − 2 dx
9 3 x
Hence the volume of the solid is
Z 3 
100 20 1
π − x + x2 − 2 dx
1/3 9 3 x
h 100x 10 1 1 i3
=π − x2 + x3 +
9 3 3 x 1/3
h 38 514 i 512
=π − 4 =π
3 3 81
1.6.2.10. *. Solution.
√ (a) The top and the bottom
√ of the circle have equations
2 2
y = T (x) = 2 + 1 − x and y = B(x) = 2 − 1 − x , respectively.

When R is rotated about the x–axis, the vertical strip of R in the figure above
sweeps out a washer with thickness dx, outer radius T (x) and inner radius B(x).
This washer has volume
π T (x)2 − B(x)2 dx = π T (x) + B(x) T (x) − B(x) dx

769
S OLUTIONS TO E XERCISES
√
= π × 4 × 2 1 − x2 dx

Hence the volume of the solid is

Z 1 √
8π 1 − x2 dx
−1
√
(b) Since y = 1 − x2 is equivalent to x2 + y 2 = 1, y ≥ 0, the integral is 8π times
the area of the upper half of the circle x2 + y 2 = 1 and hence is 8π × 12 π12 = 4π 2 .
1.6.2.11. *. Solution. (a) The two curves intersect when x obeys 8x = x2 + 15 or
x2 − 8x√+ 15 = (x − 5)(x
√ − 3) = 0. The points of intersection, in the first quadrant,
are (3, 24) and (5, 40). The region R is the region between the blue and red
curves, with 3 ≤ x ≤ 5, in the figures below.

(b) The part of the solid with x coordinate

√ between x and√x + dx is a “washer”
shaped region with inner radius √ x2 + 15, outer
√ radius2 8x and thickness dx.
2 2 2
The surface area of the washer is π( 8x) − π( x + 15) = π(8x − x − 15) and
its volume is π(8x − x2 − 15) dx. The total volume is
Z 5 h 1 i5
π(8x − x2 − 15) dx = π 4x2 − x3 − 15x
3 3 3
h 125 i
= π 100 − − 75 − 36 + 9 + 45
3
4
= π ≈ 4.19
3
1.6.2.12. *. Solution. (a) The region R is sketched in the figure on the left below.
(The bound y = 0 renders the bound x = 1 unnecessary, since the graph y = log x
hits the x-axis when x = 1.)

(b) We’ll use horizontal washers as in Example 1.6.5.

• We cut R into thin horizontal strips of height dy as in the figure on the right

770
S OLUTIONS TO E XERCISES

above.

• When we rotate R about the y–axis, i.e. about the line x = 0, each strip
sweeps out a thin washer

◦ whose inner radius is rin = ey and outer radius is rout = 2, and

◦ whose thickness is dy and hence
2 2
)dy = π 4 − e2y dy.


◦ whose volume π(rout − rin

• As our bottommost strip is at y = 0 and our topmost strip is at y = log 2

(since at the top x = 2 and x = ey ), the total
Z log 2 log 2
π 4 − e2y dy = π 4y − e2y /2 0


Volume =
0
h 1i h 3i
= π 4 log 2 − 2 + = π 4 log 2 −
2 2
Using a calculator, we see this is approximately 3.998.
1.6.2.13. *. Solution. Here is a sketch of the curves y = cos( x2 ) and y = x2 − π 2 .

By inspection, the curves meet at x = ±π where both cos( x2 ) and x2 − π 2 take the
value zero. We’ll use vertical washers as specified in the question.
• We cut the specified region into thin vertical strips of width dx as in the
figure above.
• When we rotate about the line y = −π 2 , each strip sweeps out a thin washer
◦ whose inner radius is rin = (x2 − π 2 ) − (−π 2 ) = x2 and outer radius is
rout = cos( x2 ) − (−π 2 ) = cos( x2 ) + π 2 , and
◦ whose thickness is dx and hence
2 2 2 2

◦ whose volume π(rout − rin )dx = π (cos( x2 ) + π 2 ) − (x2 ) dx.
• As our leftmost strip is at x = −π and our rightmost strip is at x = π,
the total volume is
Z π
cos2 ( x2 ) + 2π 2 cos( x2 ) + π 4 − x4 dx


π
−π

771
S OLUTIONS TO E XERCISES
π  
1 + cos(x)
Z
=π + 2π 2 cos( x2 ) + π 4 − x4 dx
−π 2

Because the integrand is even,

Z π 
1 + cos(x) 2 x 4 4
= 2π + 2π cos( 2 ) + π − x dx
0 2
 π
1 1 2 x 4 1 5
= 2π x + sin(x) + 4π sin( 2 ) + π x − x
2 2 5 0
5
 
π π
= 2π + 0 + 4π 2 + π 5 −
2 5
6
8π
= π 2 + 8π 3 +
5
We used the fact that the integrand is an even function and the interval of inte-
gration [−π, π] is symmetric, but one can also compute directly.
1.6.2.14. *. Solution. As in Example 1.6.6, we slice V into thin horizontal
“square pancakes”.

2
• We are told that the pancake at height x is a square of side 1+x
and so
2

2
• has cross-sectional area 1+x
and thickness dx and hence
2

2
• has volume 1+x
dx.

Hence the volume of V is

Z 2h Z 3 3
2 i2 4 u−1 h1 i 8
dx = 2
du = 4 = −4 − 1 =
0 1+x 1 u −1 1 3 3

We made the change of variables u = 1 + x, du = dx.

1.6.2.15. *. Solution. Here is a sketch of the base region.

Consider the thin vertical cross–section resting on the heavy red line in the figure
above. It has thickness dx. Its face is a square whose side runs from y = x2 to
2
y = 8 − x2 , a distance of 8 − 2x2 . So the face has area (8 − 2x2 ) and the slice has

772
S OLUTIONS TO E XERCISES

2
volume (8 − 2x2 ) dx. The two curves cross when x2 = 8 − x2 , i.e. when x2 = 4 or
x = ±2. So x runs from −2 to 2 and the total volume is
Z 2 Z 2 Z 2
2 2 2 2
16 − 8x2 + x4 dx
 
(8 − 2x ) dx = 2 4(4 − x ) dx = 8
−2 0 0
h 8 3 1 5 i 256 × 8
= 8 16 × 2 − 2 + 2 = = 136.53̇
3 5 15
In the first simplification step, we used the fact that our integrand was even, but
we also could have finished our computation without this step.
1.6.2.16. *. Solution. Slice the frustrum into horizontal discs. When the disc is
a distance t from the top of the frustrum it has radius 2 + 2t/h. Note that as t runs
from 0 (the top of the frustrum) to t = h (the bottom of the frustrum) the radius
2 + 2t/h increases linearly from 2 to 4.

2
Thus the disk has volume π 2 + 2t/h dt. The total volume of the frustrum is

h h h
(1 + t/h)3
Z Z 

2
2
π 2 + 2t/h dt = 4π 1 + t/h dt = 4π
0 0 3/h 0
4 28
= πh × 7 = πh
3 3
Remark: we could also solve this problem using the formula for the volume of
a cone. Using similar triangles, the frustrum in question is shaped like a right
1
circular cone of height 2h and base radius 4 (and hence of volume π(42 )(2h)),
3
but missing its top, which is a right circular cone of height h and base radius 2
1
(and hence volume π(22 )h). So, the volume of the frustrum is
3
1 1 28
π(42 )(2h) − π(22 )h = πh.
3 3 3

Exercises — Stage 3
1.6.2.17. Solution. (a)
We’ll want to start by graphing the upper half of the ellipse (ax)2 + (by)2 = 1. Its
intercepts will be enough to get us an idea: (0, 1b ) and (± a1 , 0):

773
S OLUTIONS TO E XERCISES

y
1
b

1
p
y= b
1 − (ax)2

x
− a1 1
a

We note a few things at the outset: first, since a ≥ b, then a1 ≤ 1b , so indeed the
x-axis is the minor axis. That is, we’re rotating about the proper axis to create an
oblate spheroid. p
Second, if we solve our equation for y, we get y = 1b 1 − (ax)2 . (Since we only
want the upper half of the ellipse, we only need to consider the positive square
root.)
Now, we have a standard volume-of-revolution
p problem. We make vertical
1 2
slices, of width dx and height y = b 1 − (ax) . When we rotate these slices
h p i2
about the x-axis, they form thin disks of volume π 1b 1 − (ax)2 dx. Since x
runs from − a1 to a1 , the volume of our oblate spheroid is:
Z 1  p 2
a 1 2
Volume = π 1 − (ax) dx
− a1 b
Z 1
π a
= 2 1 − (ax)2 dx
b −a 1

Z 1
2π a
= 2 1 − (ax)2 dx (even function)
b 0
1
a2 x 3 a

2π
= 2 x−
b 3 0
 
2π 1 1 4π
= 2 − = 2
b a 3a 3b a
(b) As we saw in the sketch from part (a), the shortest radius of the ellipse is a1 ,
1
while the largest is 1b . So, a1 = 6356.752, and 1b = 6378.137. That is, a =
6356.752
1
and b = .
6378.137
Note a ≥ b, as specified in part (a).
(c) Combining our answers from (a) and (b), the volume of an oblate spheroid
with approximately the same dimensions as the earth is:
 2  
4π 4π 1 1
2
=
3b a 3 b a

774
S OLUTIONS TO E XERCISES

4π
= (6378.137)2 (6356.752)
3
≈ 1.08321 × 1012 km3
≈ 1.08321 × 1021 m3

(d) A sphere of radius 6378.137 has volume

4
π (6378.137)3
3
So, our absolute error is:

4π 4
(6378.137)2 (6356.752) − π (6378.137)3
3 3
4π
= (6378.137)2 6356.752 − 6378.137
3
4π
= (6378.137)2 (21.385)
3
≈3.64 × 109 km3

And our relative error is:

abs error 4π
3
(6378.137)2 6356.752 − 6378.137
=
actual value 4π
3
(6378.137)2 (6356.752)
6356.752 − 6378.137
=
6356.752
6378.137
= −1
6356.752
≈ 0.00336

That is, about 0.336%, or about one-third of one percent.

1.6.2.18. *. Solution. (a) The curve y = 4−(x−1)2 is an “upside down parabola”
and line y = x + 1 has slope 1. They intersect at points (x, y) which satisfy both
y = x + 1 and y = 4 − (x − 1)2 . That is, when x obeys

x + 1 = 4 − (x − 1)2
x + 1 = 4 − x2 + 2x − 1
x2 − x − 2 = 0
(x − 2)(x + 1) = 0
x = −1 or x = 2

Thus the intersection points are (−1, 0) and (2, 3). Here is a sketch of R:

775
S OLUTIONS TO E XERCISES

The red strip in the sketch above runs from y = x + 1 to y = 4 − (x − 1)2 and so
has area [4 − (x − 1)2 − (x + 1)] dx = [2 + x − x2 ] dx. All together R has
Z 2
2 + x − x2 dx
 
Area =
−1
2
x2 x3

= 2x + −
2 3 −1
3 9 9
=6+ − =
2 3 2
(b) We’ll use vertical washers as in Example 1.6.3. Note that the highest point
achieved by y = 4 − (x − 1)2 is y = 4, so rotating around the line y = 5 causes no
unexpected problems.

• We cut R into thin vertical strips of width dx like the red strip in the figure
above.

• When we rotate R about the horizontal line y = 5, each strip sweeps out a
thin washer

◦ whose inner radius is rin = 5 − [4 − (x − 1)2 ] = 1 + (x − 1)2 , and

◦ whose outer radius is rout = 5 − [x + 1] = 4 − x and
◦ whose thickness is dx and hence

776
S OLUTIONS TO E XERCISES
 2 2

2
2 
dx = π 4 − x − 1 + (x − 1)2 dx
 
◦ whose volume is π rout − rin

• As our leftmost strip is at x = −1 and our rightmost strip is at x = 2, the

total
Z 2

2
2 
4 − x − 1 + (x − 1)2 dx

Volume = π
−1

1.6.2.19. *. Solution. (a) The curves (x − 1)2 + y 2 = 1 and x2 + (y − 1)2 = 1

are circles of radius 1 centered on (1, 0) and (0, 1) respectively. Both circles pass
through (0, 0) and (1, 1). They are sketched below.

The region R is symmetric about the line y = x, so the area of R is twice the area
of the part of R to the left of the line y = x. The red strip in the sketch above runs
from the edge of the lower circle to x = y. So, given a value of y in [0, 1], we need
to find the corresponding value of x along the circle. We solve (x − 1)2 + y 2 = 1
for x, keeping in mind that 0 ≤ x ≤ 1:

(x − 1)2 + y 2 = 1
(x − 1)2 = 1 − y 2
p
|x − 1| = 1 − y 2
p
1 − x = 1 − y2
p
x = 1 − 1 − y2

Now, we calculate:
Z
 1 p

Area = 2 y − 1 − 1 − y 2 dy
0Z 1 Z 1p 
=2 y − 1dy + 2
1 − y dy
0 0
nh y 2 i1 Z 1 p o
=2 −y + 2
1 − y dy
2 0 0

777
S OLUTIONS TO E XERCISES

π
= −1
2
R1p
Here the integral 0 1 − y 2 dy was evaluated simply as the area of one quarter
of a cicular disk of radius 1. It can also be evaluated by substituting y = sin θ, a
technique we’ll learn more about in Section 1.9.
(b) We’ll use horizontal washers as in Example 1.6.5.

• We cut R into thin horizontal strips of width dy like the blue strip in the
figure above.

• When we rotate R about the y–axis, each strip sweeps out a thin washer
p
◦ whose inner radius is rin = 1 − 1 − y 2 , and
p
◦ whose outer radius is rout = 1 − (y − 1)2 and
◦ whose thickness is dy and hence
◦ whose volume is
 p
2 p
2 
π 1 − (y − 1)2 − 1 − 1 − y 2 dy
p
=π 1 − (y − 1)2 − 1 + 2 1 − y 2 − (1 − y 2 )
 
p 
=2π 1 − y 2 + y − 1 dy

• As our bottommost strip is at y = 0 and our topmost strip is at y = 1, the

total
Z 1 p hπ 1 i
 
Volume = 2π 1 − y 2 + y − 1 dy = 2π + −1
0 4 2
2
π
= − π ≈ 1.793
2
R1p
Here, we again used that 0 1 − y 2 dy is the area of a quarter circle of
radius one, and we used a calculator to approximate the final answer.
1.6.2.20. *. Solution. Before√ we start, it will be useful to have a reasonable
sketch of√the graph y = c 1 + x2 over the interval [0, 1]. Its endpoints are (0, c)
and (1, c 2). The function is entirely above the x-axis, which we need to know
for part (a). For part (b), we need to know whether it is always increasing or not:
when we’re drawing horizontal strips, we need to know their endpoints, and if
the function has “humps,” the right endpoint will not be simply the line x = 1.
If you’re comfortable noticing that 1+x2 increases as x increases because√ we only
consider nonnegative values of x, then you can also be confident that 1 + x2 is
simply increasing. Alternately, we can consider the derivative:

d n √ o 1 cx
c 1 + x2 = c · √ · 2x = √
dx 2 1 + x2 1 + x2
Since we only consider positive values of x, this derivative is never negative, so

778
S OLUTIONS TO E XERCISES

the function is never decreasing. This gives us the following basic sketch:

x
1

The figures in the solution below use a slightly more detailed rendering of our
function, but so much accuracy is not necessary.
(a) Let V1 be the solid obtained by revolving R about the x–axis. The portion of V1
with x–coordinate between x and x + dx is obtained by rotating the red vertical
the figure on the left below about the x–axis. That portion
strip in √ √ is a disk of
radius c 1 + x2 and thickness dx. The volume of this disk is π(c 1 + x2 )2 dx =
πc2 (1 + x2 ) dx. So the total volume of V1 is
Z 1
2 2 2
h x3 i 1 4 2
V1 = πc (1 + x ) dx = πc x + = πc
0 3 0 3

(b) We’ll use horizontal washers as in Example 1.6.5.

• We cut R into thin horizontal strips of width dy as in the figure on the right
above.

• When we rotate R about the y–axis, i.e. about the line x = 0, each strip
sweeps out a thin washer

◦ whose outer radius is rout = 1, and

q
2 √
◦ whose inner radius is rin = yc2 − 1 when y ≥ c 1 + 02 = c (see the
red strip in the figure on the right above), and whose inner radius is
rin = 0 when y ≤ c (see the blue strip in the figure on the right above)
and
◦ whose thickness is dy and hence

779
S OLUTIONS TO E XERCISES
2

◦ whose volume is π(rout 2 2
− rin )dy = π 2 − yc2 dy when y ≥ c and whose
2 2
volume is π(rout − rin )dy = π dy when y ≤ c and
√
• As our bottommost strip is at y√= 0 and our topmost strip is at y = 2 c
(since at the top x = 1 and y = c 1 + x2 ), the total
√
2c c
y2 
Z  Z
V2 = π 2 − 2 dy + π dy
c c 0
√
2c
h y3 i
= π 2y − 2 + πc
√ 3c c
h4 2 5i
= πc − + πc
3 3
π c √ 
= 4 2−2
3

(c) We have V1 = V2 if and only if

4 2 π c √ 
πc = 4 2−2
3 3
2
√ 
4c = c 4 2 − 2
 √ 
2
4c − c 4 2 − 2 = 0
√
  
1
4c c − 2− =0
2
√ 1
c = 0 or c = 2 −
2
1.6.2.21. *. Solution. We will compute the volume by rotating thin vertical
strips as in the sketch

about the line y = −1 to generate thin washers. We need to know when the line
y = 4 + 2π − 2x intersects the curve y = 4 + π sin x. Looking at the graph, it
appears to be at π2 , π, and 3π
2
. By plugging in these values of x to both functions,
we see they are indeed the points of intersection.

π
• When 2
≤ x ≤ π, the top of the strip is at y = 4 + π sin x and the bottom

780
S OLUTIONS TO E XERCISES

of the strip is at y = 4 + 2π − 2x. When the strip is rotated, we get a thin

washer with outer radius R1 (x) = 1 + 4 + π sin x = 5 + π sin x and inner
radius r1 (x) = 1 + 4 + 2π − 2x = 5 + 2π − 2x.

• When π ≤ x ≤ 3π 2
, the top of the strip is at y = 4 + 2π − 2x and the bottom of
the strip is at y = 4 + π sin x. When the strip is rotated, we get a thin washer
with outer radius R2 (x) = 1 + 4 + 2π − 2x = 5 + 2π − 2x and inner radius
r2 (x) = 1 + 4 + π sin x = 5 + π sin x.

So, the total

Z π Z 3π/2
π R1 (x)2 − r1 (x)2 dx + π R2 (x)2 − r2 (x)2 dx
   
Volume =
π/2 π
Z π
π (5 + π sin x)2 − (5 + 2π − 2x)2 dx
 
=
π/2
Z 3π/2
π (5 + 2π − 2x)2 − (5 + π sin x)2 dx
 
+
π

1.6.2.22. Solution. (a)

We use the same ideas for volume, and apply them to mass. We want to take
slices of the column, approximate their mass, then add them up. To reconcile our
units, let k = 1000h, so k is the height in metres. Then the density of air at height
k is c2−k/6000 mkg3 .
A horizontal slice of the column is a circular disk with height dk and radius 1 m.
So, its volume is πdkm3 . What we’re interested in, though, is its mass. At height
k, its mass is
 
3 −k/6000 kg


(volume) × (density) = πdkm × c2
m3
= cπ2−k/6000 dk kg

Since k runs from 0 to 60, 000, the total mass is given by

Z 60000 Z 60000
−k/6000
cπ2 dk = cπ 2−k/6000 dk
0 0

To facilitate integration, we can write our exponential function in terms of e, then

k 1
use the substitution u = − 6000 log 2, du = − 6000 log 2dk.
Z 60000

−k/6000
= cπ elog 2 dk
0
Z 60000
k
= cπ e− 6000 log 2 dk
0
6000cπ −10 log 2 u
Z
=− e du
log 2 0
6000cπ 0
Z
= eu du
log 2 −10 log 2

781
S OLUTIONS TO E XERCISES
 
6000cπ 1
= 1 − 10
log 2 2
6000cπ
We note this is fairly close to .
log 2
We also remark that this is a demonstration of the usefulness of integrals. We
wanted to know how much of something there was, but the amount of that some-
thing was different everywhere: more in some places, less in others. Integration
allowed us to account for this gradient. You’ve seen this behaviour exploited to
find distances travelled, areas, volumes, and now mass. In your studies, you will
doubtless learn to use it to find still more quantities, and we will discuss other
applications in Chapter 2.
3000cπ
(b) We want to find the value of k that gives a mass of . By following our
log 2
reasoning above, the mass of air in the column from the ground to height k is
 
6000cπ 1
1 − k/6000
log 2 2

So, we set this equal to the mass we want, and solve for k.
 
6000cπ 1 3000cπ
1 − k/6000 =
log 2 2 log 2
 
1
2 1 − k/6000 = 1
2
2
1 = k/6000
2
2k/6000
= 21
k = 6000
h=6

This means that there is roughly the same mass of air in the lowest 6 km of the
column as there is in the remaining 54 km.

1.7 · Integration by parts

1.7.2 · Exercises

Exercises — Stage 1
1.7.2.1. Solution. Integration by substitution is just using the chain rule, back-
wards:
d
{f (g(x))} = f 0 (g(x))g 0 (x)
dx
d
Z Z
⇔ {f (g(x))}dx + C = f 0 (g(x))g 0 (x)dx
dx

782
S OLUTIONS TO E XERCISES
Z
⇔ f (g(x)) +C = f 0 (g(x)) g 0 (x)dx
| {z } | {z } | {z }
f (u) f 0 (u) du

Similarly, integration by parts comes from the product rule:

d
{f (x)g(x)} = f 0 (x)g(x) + f (x)g 0 (x)
dx
d
Z Z
⇔ {f (x)g(x)}dx + C = f 0 (x)g(x) + f (x)g 0 (x)dx
dx
Z Z
⇔ f (x)g(x) + C = f (x)g(x)dx + f (x)g 0 (x)dx
0

Z Z
⇔ f (x) g (x)dx = f (x) g(x) − g(x) f 0 (x)dx
0
|{z} | {z } |{z} |{z} |{z} | {z }
u dv u v v du

R R
1.7.2.2. Solution. Remember our rule: udv = uv − vdu. So, we take u and
use it to make du–that is, we differentiate it. We take dv and use it to make v–that
is, we antidifferentiate it.
1.7.2.3. Solution. We’ll use the same ideas that lead to the methods of substi-
tution and integration by parts. (You can review these in your text, or see the
solution to Question 1 in this section.) According to the quotient rule,

g(x)f 0 (x) − f (x)g 0 (x)

 
d f (x)
= .
dx g(x) g 2 (x)

Antidifferentiating both sides gives us:

g(x)f 0 (x) − f (x)g 0 (x)

 
d f (x)
Z Z
dx + C = dx
dx g(x) g 2 (x)
Z 0
f (x) f (x) f (x)g 0 (x)
Z
+C = dx − dx
g(x) g(x) g 2 (x)
Z 0
f (x) f (x) f (x)g 0 (x)
Z
dx = + dx + C
g(x) g(x) g 2 (x)

This isn’t quite at catchy as integration by parts–which is probably why it hasn’t

caught on as a rule with its own name.
1.7.2.4. Solution. All the antiderivatives differ only by a constant, so we can
write them all as v(x) + C for some C. Then, using the formula for integration by
parts,
Z Z
0
v(x) + C u0 (x)dx
   
u(x) · v (x)dx = u(x) v(x) + C −
|{z} | {z } | {z } | {z }
u v v du
Z Z
= u(x)v(x) + Cu(x) − v(x)u (x)dx − Cu0 (x)dx
0

783
S OLUTIONS TO E XERCISES
Z
= u(x)v(x) + Cu(x) − v(x)u0 (x)dx − Cu(x) + D
Z
= u(x)v(x) − v(x)u0 (x)dx + D

where D is any constant.

Since the terms with C cancel out, it didn’t matter what we chose for C–all
choices end up the same.
Z
1.7.2.5. Solution. Suppose we choose dv = f (x)dx, u = 1. Then v = f (x)dx,
and du = dx. So, our integral becomes:
Z Z Z Z 
(1) f (x)dx = (1) f (x)dx − f (x)dx |{z}
dx
|{z} | {z } |{z}
u dv u | {z } | {z } du
v v

In order to figure out the first product (and the second integrand), you need to
know the antiderivative of f (x)–but that’s exactly what you’re trying to figure
out! So, using integration by parts has not Reased your pain.
We note here that in certain cases, such as log xdx (Example 1.7.8 in your text),
it is useful to choose dv = 1dx. This is similar to, but crucially different from, the
do-nothing method in this problem.

Exercises — Stage 2
1.7.2.6. *. Solution. For integration by parts, we want to break the integrand
into two pieces, multiplied together. There is an obvious choice for how to do
this: one piece is x, and the other is log x. Remember that one piece will be
integrated, while the other is differentiated. The question is, which choice will be
more helpful. After some practice, you’ll get the hang of making the choice. For
now, we’ll present both choices–but when you’re writing a solution, you don’t
have to write this part down. It’s enough to present your choice, and then a
successful computation is justification enough.

Option 1: u = x du = 1 dx
dv = log x dx v =??
1
Option 2: u = log x du = dx
x
1 2
dv = xdx v= x
2
In Example 1.7.8, we found the antiderivative of logarithm, but it wasn’t trivial.
We might reasonably want to avoid using this complicated antiderivative. In-
deed, Option 2 (differentiating logarithm, antidifferentiating x) looks promising–
when we multiply the blue equations, we get something easily integrable– so
let’s not even bother going deeper into Option 1.
That is, we perform integration by parts with u = log x and dv = x dx, so that

784
S OLUTIONS TO E XERCISES

dx x2
du = x
and v = 2
.

x2 log x x2 dx x2 log x 1
Z Z Z
x log x dx = − = − xdx
|{z} | {z } 2 } 2 |{z}
x 2 2
u dv
| {z |{z}
uv v du
2 2
x log x x
= − +C
2 4
1.7.2.7. *. Solution. Our integrand is the product of two functions, and there’s
no clear substitution. So, we might reasonably want to try integration by parts.
Again, we have two obvious pieces: log x, and x−7 . We’ll consider our options
for assigning these to u and dv:

1
Option 1: u = log x dx
du =
x
1 −6
dv = x−7 dx v = x
−6
Option 2: u = x−7 du = −7x−8 dx
dv = log xdx v =??

Again, we remember that logarithm has some antiderivative we found in Ex-

ample 1.7.8, but it was something complicated. Luckily, we don’t need to bother
with it: when we multiply the red equations in Option 1, we get a perfectly work-
able integral.
We perform integration by parts with u = log x and dv = x−7 dx, so that du = dxx
−6
and v = − x 6 .

log x x−6 x−6 dx log x 1

Z Z Z
dx = − log x + =− 6 + x−7 dx
x7 | {z 6 } 6 |{z}
|{z} x 6x 6
uv −v du
log x 1
=− 6 − +C
6x 36x6
1.7.2.8. *. Solution. To integrate by parts, we need to decide what to use as
u, and what to use as dv. The salient parts of this integrand are x and sin x, so
we only need to decide which is u and which dv. Again, this process will soon
become familiar, but to help you along we show you both options below.

Option 1: u = x du = 1 dx
dv = sin x dx v = − cos x
Option 2: u = sin x du = cos x dx
1
dv = xdx v = x2
2
The derivative and antiderivative of sine are almost the same, but x turns into

785
S OLUTIONS TO E XERCISES

something simpler when we differentiate it. So, we choose Option 1.

We integrate by parts, using u = x, dv = sin x dx so that v = − cos x and du = dx:
Z π π
Z π
x sin x dx = −x cos x − (− cos x) |{z}
dx
0
| {z } 0 0 | {z }
uv du
h iπ v
= − x cos x + sin x = −π(−1) = π
0

1.7.2.9. *. Solution. When we have two functions multiplied like this, and
there’s no obvious substitution, our minds turn to integration by parts. We hope
that our integral will be improved by differentiating one part and antidifferenti-
ating the other. Let’s see what our choices are:

Option 1: u = x du = 1 dx
dv = cos x dx v = sin x
Option 2: u = cos x du = − sin x dx
1
dv = xdx v = x2
2
Option 1 seems preferable. We integrate by parts, using u = x, dv = cos x dx so
that v = sin x and du = dx:
Z π π Z π
2 2 2
x cos x dx = |{z}
x sin x − sin x |{z}
dx
0
|{z} 0
|{z} 0
u v v du
h i π2 π
= x sin x + cos x = − 1
0 2
1.7.2.10. Solution. This integrand is the product of two functions, with no obvi-
ous substitution. So, let’s try integration by parts, with one part ex and one part
x3 .
Option 1: u = ex du = ex dx
dv = x3 dx v = 14 x4
Option 2: u = x3 du = 3x2 dx
dv = ex dx v = ex
At first glance, multiplying the red functions and multiplying the blue functions
give largely equivalent integrands to what we started with–none of them with
obvious antiderivatives. In previous questions, we were able to choose u = x,
and then du = dx, so the “x” in the integrand effectively went away. Here, we
see that choosing u = x3 will lead to du = 3x2 dx, which has a lower power. If
we repeatedly perform integration by parts, choosing u to be the power of x each
time, then after a few iterations it should go away, because the third derivative
of x3 is a constant.
So, we start with Option 2: u = x3 , dv = ex dx, du = 3x2 dx, and v = ex .

Z Z
3 x 3 x 2
x e dx = |{z}
x |{z}
e − ex · 3x
|{z} | {zdx}
u v v du

786
S OLUTIONS TO E XERCISES
Z
3 x
=x e −3 ex · x2 dx

Now, we take u = x2 and dv = ex dx, so du = 2xdx and v = ex . We’re only using

integration by parts on the actual integral–the rest of the function stays the way
it is.
" Z #
= x3 ex − 3 x 2 x
e − |{z}
|{z} ex · 2xdx
| {z }
uv v du
Z
= x3 ex − 3x2 ex + 6 xex dx

Continuing, we take u = x and dv = ex dx, so du = dx and v = ex . This is the step

where the polynomial part of the integrand finally disappears.
" Z #
= x3 ex − 3x2 ex + 6 |{z}
xex − ex |{z}
|{z} dx
uv v du
3 x 2 x x x
= x e − 3x e + 6xe − 6e + C
= ex x3 − 3x2 + 6x − 6 + C

Let’s check that this makes sense: the derivative of ex (x3 − 3x2 + 6x − 6) + C
should be x3 ex . We differentiate using the product rule.

d  x 3
e x − 3x2 + 6x − 6 + C


dx
= ex x3 − 3x2 + 6x − 6 + ex 3x2 − 6x + 6

= ex x3 − 3x2 + 3x2 + 6x − 6x − 6 + 6 = x3 ex

Remark: In order to be technically correct in our antidifferentiation, we should

add the +C as soon as we do the first integration by parts. However, when we
are using integration by parts, we usually end up evaluating an integral at the
end, and we add the +C at that point. Since the +C comes up eventually, it is
common practice to not clutter our calculations with it until the end.
1.7.2.11. Solution. Since our integrand is two functions multiplied together,
and there isn’t an obvious substitution, let’s try integration by parts. Here are
our salient options.

Option 1: u = x du = 1 dx
3
dv = log x dx v =??
Option 2: u = log3 x du = 3 log2 x · 1
x
dx
1
dv = xdx v = x2
2
This calls for some strategizing. Using the template of Example 1.7.8, we could
probably figure out the antiderivative of log3 x. Option 1 is tempting, because
our x-term goes away. So, there might be a benefit there, but on the other hand,

787
S OLUTIONS TO E XERCISES

the antiderivative of log3 x is probably pretty complicated.

Now let’s consider Option 2. When we multiply the blue functions together, we
get something similar to our original integrand, but the power of logarithm is
smaller. If we were to iterate this method (using integration by parts a few times,
always choosing u to be the part with a logarithm) then eventually we would
end up differentiating logarithm. This seems like a safer plan: let’s do Option 2.
We use integration by parts with u = log3 x, dv = xdx, du = x3 log2 xdx, and
v = 12 x2 .

1 2 3 3
Z Z
3
x log xdx = x log x − x log2 xdx
2
| {z } |2 {z }
uv vdu
1 3
Z
= x2 log3 x − x log2 xdx
2 2
Continuing our quest to differentiate away the logarithm, we use integration by
2 1
parts with u = log2 x, dv = xdx, du = log xdx, and v = x2 .
x 2
 
1 3 1
Z
= x2 log3 x −  x2 log2 x − x log xdx

2 2 |2 {z } | {z }
vdu
uv
1 3 3
Z
= x2 log3 x − x2 log2 x + x log xdx
2 4 2
1 1
One last integration by parts: u = log x, dv = xdx, du = dx, and v = x2 .
x 2
 
1 3 3 1 1
Z
= x2 log3 x − x2 log2 x +  x2 log x − xdx

2 4 2 |2 {z } |2 {z }
uv vdu
1 2 3 3 2 2 3 2 3
Z
= x log x − x log x + x log x − xdx
2 4 4 4
1 3 3 3
= x2 log3 x − x2 log2 x + x2 log x − x2 + C
2 4 4 8
Once again, technically there is a +C in the work after the first integration by
parts, but we follow convention by conveniently suppressing it until the final
integration.
1.7.2.12. Solution. The integrand is the product of two functions, without an

788
S OLUTIONS TO E XERCISES

obvious substitution, so let’s see what integration by parts can do for us.

Option 1: u = x2 du = 2x dx
dv = sin x dx v = − cos x
Option 2: u = sin x du = cos x dx
dv = x2 dx v = 13 x3

Neither option gives us something immediately integrable, but Option 1 replaces

our x2 term with a lower power of x. If we repeatedly apply integration by parts,
we can reduce this power to zero. So, we start by choosing u = x2 and dv =
sin xdx, so du = 2xdx and v = − cos x.

Z Z
2 2
x sin xdx = −x cos x + 2x cos xdx
| {z }
uv | {z }
−vdu
Z
2
= −x cos x + 2 x cos xdx

Using integration by parts again, we want to be differentiating (not antidifferen-

tiating) x, so we choose u = x, dv = cos xdx, and then du = dx (x went away!),
v = sin x.
" Z #
= −x2 cos x + 2 x {z x} − sin
| sin | {zxdx}
uv vdu
2
= −x cos x + 2x sin x + 2 cos x + C
= (2 − x2 ) cos x + 2x sin x + C
1.7.2.13. Solution. This problem is similar to Questions 6 and 7: integrating
a polynomial multiplied by a logarithm. Just as in these questions, if we use
1
integration by parts with u = log t, then du = dt, and our new integrand will
t
consist of powers of t–which are easy to antidifferentiate.
So, we use u = log t, dv = 3t2 − 5t + 6, du = 1t dt, and v = t3 − 25 t2 + 6t.
Z
(3t2 − 5t + 6) log tdt
 
Z  
3 5 2 1 3 5 2
= log t t − t + 6t − t − t + 6t dt

|{z} | 2{z t 2
u } | {z }
v vdu
  Z  
3 5 2 2 5
= t − t + 6t log t − t − t + 6 dt
2 2
 
5 1 5
= t3 − t2 + 6t log t − t3 + t2 − 6t + C
2 3 4

789
S OLUTIONS TO E XERCISES

1.7.2.14. Solution. Before we jump to integration by parts, we

√ notice that the
square roots lend themselves to substitution. Let’s take w = s. Then dw =
1
√ ds, so 2wdw = ds.
2 s
√ √s
Z Z Z
se ds = w · e · 2wdw = 2 w2 ew dw
w

Now we have nearly the situation of Question 10. We can repeatedly use inte-
gration by parts with u as the power of w to get rid of the polynomial part. We’ll
start with u = w2 dv = ew dw, du = 2wdw, and v = ew .
" Z #
2 w w
=2 w
| {ze } − 2we
| {z dw}
uv vdu
Z
= 2w2 ew − 4 wew dw

We use integration by parts again, this time with u = w, dv = ew dw, du = dw,

and v = ew .
" Z #
= 2w2 ew − 4 |{z}
wew − e|w{z
dw}
uv vdu
2 w w w
= 2w e − 4we + 4e + C
= ew 2w2 − 4w + 4 + C


√ √
= e s 2s − 4 s + 4 + C

1.7.2.15. Solution. Let’s use integration by parts. What are our parts? We have
a few options.

• Solution 1: Following Example 1.7.8, we choose u = log2 x and dv = dx, so

that du = x2 log xdx and v = x.
Z Z
2 2
log xdx = x log x − 2 log xdx
| {z } | {z }
uv vdu

Here we can either use the antiderivative of logarithm from memory, or

re-derive it. We do the latter, using integration by parts with u = log x,
dv = 2dx, du = x1 dx, and v = 2x.
 
Z
= x log2 x − 2x log x − |{z}
2dx 
| {z }
uv vdu
2
= x log x − 2x log x + 2x + C

• Solution 2: Our integrand is two functions multiplied together: log x and

log x. So, we will use integration by parts with u = log x, dv = log x, du =

790
S OLUTIONS TO E XERCISES

1
x
dx,and (using the antiderivative of logarithm, found in Example 1.7.8 in
the text) v = x log x − x.
1
Z Z
2
log xdx = (log x)(x log x − x) − (x log x − x) dx
|{z} | {z } | {z } |{z} x
u v v
du
Z
= x log2 x − x log x − (log x − 1) dx

= x log2 x − x log x − [(x log x − x) − x] + C

= x log2 x − 2x log x + 2x + C
1.7.2.16. Solution. This is your friendly reminder that to a person with a ham-
mer, everything looks like a nail. The integral in the problem is a classic ex-
ample of an integral to solve using substitution. We have an “inside function,”
x2 + 1, whose derivative shows up multiplied to the rest of the integrand. We
take u = x2 + 1, then du = 2xdx, so
Z Z
x2 +1 2
2xe dx = eu du = eu + C = ex +1 + C

1.7.2.17. *. Solution. In Example 1.7.9, we saw that integration by parts was

useful when the integrand has a derivative that works nicely when multiplied
by x. We use the same idea here. Let u = arccos y and dv = dy, so that v = y and
du = − √dy 2 .
1−y

y
Z Z
arccos y dy = y arccos y + p dy
| {z } 1 − y2
uv | {z }
−vdu

Using the substitution u = 1 − y 2 , du = −2ydy,

1
Z
= y arccos y − u−1/2 du
2
= y arccos y − u1/2 + C
p
= y arccos y − 1 − y 2 + C

Exercises — Stage 3
1.7.2.18. *. Solution. We integrate by parts, using u = arctan(2y), dv = 4y dy,
2 dy
so that v = 2y 2 and du = 1+(2y) 2:

4y 2
Z Z
2
4y arctan(2y) dy = 2y arctan(2y) − dy
| {z } (2y)2 + 1
uv | {z }
vdu
4y2
The integrand (2y) 2 +1 is a rational function. So the remaining integral can be

evaluated using the method of partial fractions, starting with long division. But

791
S OLUTIONS TO E XERCISES

4y 2 4y 2 +1 1
it is easier to just notice that 4y 2 +1
= 4y 2 +1
− 4y 2 +1
. We therefore have:

4y 2
Z  
1 1
Z
dy = 1− dy = y − arctan(2y) + C
4y 2 + 1 2
4y + 1 2

The final answer is then

1
Z
4y arctan(2y) dy = 2y 2 arctan(2y) − y + arctan(2y) + C
2
1.7.2.19. Solution. We’ve got an integrand that consists of two functions mul-
tiplied together, and no obvious substitution. So, we think about integration by
parts. Let’s consider our options. Note in Example 1.7.9, we found that the an-
tiderivative of arctangent is x arctan x − 12 log(1 + x2 ) + C.
1
Option 1: u = arctan x du = dx
1 + x2
1
dv = x2 dx v = x3
3
Option 2: u = x2 du = 2x dx
1
dv = arctan xdx v = x arctan x − log(1 + x2 )
2

• Option 1: Option 1 seems likelier. Let’s see how it plays out. We use inte-
dx 1 3
gration by parts with u = arctan x, dv = x2 dx, du = 1+x 2 , and v = 3 x .

x3 x3
Z Z
2
x arctan xdx = arctan x − dx
|3 {z } 3(1 + x2 )
| {z }
uv vdu
x3 1 x3
Z
= arctan x − dx
3 3 1 + x2
This is starting to look like a candidate for a substitution! Let’s try the
denominator, s = 1 + x2 . Then ds = 2xdx, and x2 = s − 1.
x3 1 x2
Z
= arctan x − · 2xdx
3 6 1 + x2
x3 1 s−1
Z
= arctan x − ds
3 6 s
x3 1 1
Z
= arctan x − 1 − ds
3 6 s
3
x 1 1
= arctan x − s + log |s| + C
3 6 6
3
x 1 1
= arctan x − (1 + x2 ) + log(1 + x2 ) + C
3 6 6

• Option 2: What if we had tried the other option? That is, u = x2 , du = 2xdx,
dv = arctan x, and v = x arctan x − 12 log(1 + x2 ). It’s not always the case

792
S OLUTIONS TO E XERCISES

that both options work, but sometimes they do. (They are almost never of
equal difficulty.) This solution takes advantage of two previously hard-won
results: the antiderivatives of logarithm and arctangent.

 
1
Z
2 2 2
x arctan xdx = |{z}
x x arctan x − log(1 + x )
2
u | {z }
v
 
1
Z
2
− x arctan x − log(1 + x ) · 2xdx
2 | {z }
| {z } du
v
2
x
Z
3
= x arctan x − log(1 + x ) − 2 x2 arctan xdx
2
2
Z
+ x log(1 + x2 )dx
Z
Adding 2 x2 arctan xdx to both sides:

x2
Z
2 3
3 x arctan xdx = x arctan x − log(1 + x2 )
2
Z
+ x log(1 + x2 )dx
x3 x2
Z
x2 arctan xdx = arctan x − log(1 + x2 )
3 6
1
Z
+ x log(1 + x2 )dx
3

Using the substitution s = 1 + x2 , ds = 2xdx:

x3 x2 1
Z
= arctan x − log(1 + x2 ) + log sds
3 6 6
Using the antiderivative of logarithm found in Example 1.7.8,

x3 x2 1
= arctan x − log(1 + x2 ) +
(s log s − s) + C
3 6 6
x3 x2 1
= arctan x − log(1 + x2 ) +(1 + x2 ) log(1 + x2 )
3 6 6
− (1 + x2 ) + C

x3
 2
1 + x2

x 1
= arctan x + − + log(1 + x2 ) − (1 + x2 ) + C
3 6 6 6
x3 1 1
= arctan x + log(1 + x2 ) − (1 + x2 ) + C
3 6 6

793
S OLUTIONS TO E XERCISES

1.7.2.20. Solution. This example is similar to Example 1.7.10 in the text. The
functions ex/2 and cos(2x) both do not substantially alter when we differentiate
or antidifferentiate them. If we use integration by parts twice, we’ll end up with
an expression that includes our original integral. Then we can just solve for the
original integral in the equation, without actually antidifferentiating.
Let’s use u = cos(2x) and dv = ex/2 dx, so du = −2 sin(2x)dx and v = 2ex/2 .
Z Z
e cos(2x)dx = 2e cos(2x) − −4ex/2 sin(2x)dx
x/2 x/2
| {z } | {z }
uv vdu
Z
= 2e cos(2x) + 4 ex/2 sin(2x)dx
x/2

Similarly to our first integration by parts, we use u = sin(2x), dv = ex/2 dx, du =

2 cos(2x)dx, and v = 2ex/2 .
 
Z
= 2ex/2 cos(2x) + 4 2ex/2 sin(2x) − 4ex/2 cos(2x)dx
| {z } | {z }
uv vdu

So, we’ve found the equation

Z
ex/2 cos(2x)dx = 2ex/2 cos(2x) + 8ex/2 sin(2x)
Z
− 16 ex/2 cos(2x)dx + C
Z
We add 16 ex/2 cos(2x)dx to both sides.
Z
17 ex/2 cos(2x)dx = 2ex/2 cos(2x) + 8ex/2 sin(2x) + C
2 x/2 8
Z
ex/2 cos(2x)dx = e cos(2x) + ex/2 sin(2x) + C
17 17
Remark: remember that C is a stand-in for “we can add any real constant”. Since
C
C can be any number in (−∞, ∞), also 17 can be any number in (−∞, ∞). So,
C C
rather than write 17 in the last line, we re-named 17 to C.
1.7.2.21. Solution.
• Solution 1: This question looks like a substitution, since we have an “inside
1
function.” So, let’s see where that leads: let u = log x. Then du = dx. We
x
don’t see this right away in our function, but we can bring it into the func-
tion by multiplying and dividing by x, and noting from our substitution
that eu = x.
x sin(log x)
Z Z
sin(log x)dx = dx
x

794
S OLUTIONS TO E XERCISES
Z
= eu sin udu

Using the result of Example 1.7.11:

1
= eu (sin u − cos u) + C
2
1
= elog x (sin(log x) − cos(log x)) + C
2
1
= x (sin(log x) − cos(log x)) + C
2

• Solution 2: It’s not clear how to antidifferentiate the integrand, but we can
certainly differentiate it. So, keeping in mind the method of Example 1.7.11
in the text, we take u = sin(log x) and dv = dx, so du = x1 cos(log x)dx and
v = x.
Z Z
sin(log x)dx = x sin(log x) − cos(log x)dx
| {z } | {z }
uv vdu

Continuing on, we again use integration by parts, with u = cos(log x), dv =

dx, du = − x1 sin(log x)dx, and v = x.
 Z 
= x sin(log x) − x cos(log x) + sin(log x) dx
| {z } | {z }
uv −vdu

That is, we have

Z Z
sin(log x)dx = x [sin(log x) − cos(log x)] − sin(log x)dx + C
R
Adding sin(log x)dx to both sides,
Z
2 sin(log x)dx = x [sin(log x) − cos(log x)] + C
x
Z
sin(log x)dx = [sin(log x) − cos(log x)] + C
2

Remark: remember that C is a stand-in for “we can add any real constant”.
Since C can be any number in (−∞, ∞), also C2 can be any number in
(−∞, ∞). So, rather than write C2 in the last line, we re-named C2 to C.
1.7.2.22. Solution. We begin by simplifying the integrand.
Z Z Z
x+log2 x x log2 x
2 dx = 2 · 2 dx = 2x · xdx

795
S OLUTIONS TO E XERCISES
Z
This is similar to the integral xex dx, which we saw in Example 1.7.1. Let’s
write 2 = elog 2 to take advantage of the easy integrability of ex .
Z
= x · ex log 2 dx

We use integration by parts with u = x, dv = ex log 2 dx; du = dx, v = log1 2 ex log 2 .

(Remember log 2 is a constant. If you’d prefer, you can do a substitution with
s = x log 2 first, to have a simpler exponent of e.)

x x log 2 1 x log 2
Z
= e − e dx
log 2 log 2
| {z } | {z }
uv vdu
x x log 2 1
= e − ex log 2 + C
log 2 (log 2)2
x x 1
= 2 − 2
2x + C
log 2 (log 2)
1.7.2.23. Solution. It’s not obvious where to start, but in general it’s nice to have
the arguments of our trig functions the same. So, we use the identity sin(2x) =
2 sin x cos x.
Z Z
cos x
e sin(2x)dx = 2 ecos x cos x sin x dx

Now we can use the substitution w = cos x, dw = − sin xdx.

Z
= −2 wew dw

From here the integral should look more familiar. We can use integration by parts
with u = w, dv = ew dw, du = dw, and v = ew .
" Z #
wew −
= −2 |{z} e|w{z
dw}
uv vdu
w
= 2e [1 − w] + C
= 2ecos x [1 − cos x] + C
1.7.2.24. Solution. We’ve got an integrand that consists of several functions
multiplied together, and no obvious substitution. So, we think about integration
1 d 1 1
by parts. We know an antiderivative for (1−x) 2 , because we know dx 1−x = (1−x)2 .
dx −x 1
So let’s try dv = (1−x) 2 and u = xe . Then v = 1−x and du = (1 − x)e−x dx. So,
by integration by parts,
dx xe−x 1
Z Z
−x
xe
| {z } (1 − x)2 = − (1 − x)e−x dx
1−x 1−x| {z }
u | {z } | {z } | {z } du
dv uv v

796
S OLUTIONS TO E XERCISES

xe−x
Z
= − e−x dx
1−x
xe−x e−x
= + e−x + C = +C
1−x 1−x
1.7.2.25. *. Solution. (a) The “parts” in the integrand are powers of sine. Look-
ing at the right hand side of the reduction formula, we see that it looks a little like
the derivative of sinn−1 x, although not exactly. So, let’s integrate by parts with
u = sinn−1 x and dv = sin x dx, so that du = (n − 1) sinn−2 x cos x and v = − cos x.
Z Z
n
sin x dx = − sin n−1
x cos x + (n − 1) cos2 x sinn−2 x dx
| {z }
uv | {z }
R
− vdu

Using the identity sin2 x + cos2 x = 1,

Z
= − sin n−1
(1 − sin2 x) sinn−2 x dx
x cos x + (n − 1)
Z Z
n−1 n−2
= − sin x cos x + (n − 1) sin x dx − (n − 1) sinn x dx

Moving the last term on the right hand side to the left hand side gives
Z Z
n n−1
n sin x dx = − sin x cos x + (n − 1) sinn−2 x dx

Dividing across by n gives the desired reduction formula.

(b) By the reduction formula of part (a), if n ≥ 2,
π/2 π/2
n−1
Z Z
n
sin (x) dx = sinn−2 (x) dx
0 n 0

since sin 0 = cos π2 = 0. Applying this reduction formula, with n = 8, 6, 4, 2:

π/2
7 π/2 6 7 5 π/2 4
Z Z Z
8
sin (x) dx = sin (x) dx = · sin (x) dx
0 8 0 8 6 0
7 5 3 π/2 2
Z
= · · sin (x) dx
8 6 4 0
7 5 3 1 π/2
Z
= · · · dx
8 6 4 2 0
7 5 3 1 π
= · · · ·
8 6 4 2 2
35
= π
256
Using a calculator, we see this is approximately 0.4295.

797
S OLUTIONS TO E XERCISES

1.7.2.26. *. Solution. (a) The sketch is the figure on the left below. By integration
1
by parts with u = arctan x, dv = dx, v = x and du = 1+x 2 dx, and then the
2
substitution s = 1 + x ,
Z 1 Z 1
1 x
A= arctan xdx = x arctan x} − dx
0
| {z 0 0 | x2 }
1 + {z
uv
vdu
1 1
= arctan 1 − log(1 + x2 )
2 0
π log 2
= −
4 2

(b) We’ll use horizontal washers as in Example 1.6.5.

• We cut R into thin horizontal strips of width dy as in the figure on the right
above.

• When we rotate R about the y–axis, each strip sweeps out a thin washer

◦ whose inner radius is rin = tan y and outer radius is rout = 1, and
◦ whose thickness is dy and hence
2 2
◦ whose volume π(rout − rin )dy = π(1 − tan2 y)dy.
π
• As our bottommost strip is at y = 0 and our topmost strip is at y = 4
(since
at the top x = 1 and x = tan y), the total
Z π Z π
4 4
2
Volume = π(1 − tan y)dy = π(2 − sec2 y)dy
0 0
 π
= π 2y − tan y 04
π2
= −π
2
1.7.2.27. *. Solution. For a fixed value of x, if we rotate about the x-axis, we
form a washer of inner radius B(x) and outer radius T (x) and hence of area
π[T (x)2 − B(x)2 ]. We integrate this function from x = 0 to x = 3 to find the total
volume V :
Z 3
V = π[T (x)2 − B(x)2 ] dx
0

798
S OLUTIONS TO E XERCISES
Z 3 √ √
=π ( xe3x )2 − ( x(1 + 2x))2 dx
Z0 3
xe6x − (x + 4x2 + 4x3 ) dx


=π
Z0 3 h x2
4x3 i3
=π xe6x dx − π + x4 +
2 3 0
Z0 3 h 32 4 · 33 i
=π xe6x dx − π + + 34
0 2 3

For the first integral, we use integration by parts with u(x) = x, dv = e6x dx, so
that du = dx and v(x) = 16 e6x :
3 3 3
xe6x 1 6x
Z Z
6x
xe dx = − e dx
0 6}
| {z 0 0 6
| {z }
uv vdu
18 3
e18 e18
 
3e 1 6x 1
= −0− e = − − .
6 36 0 2 36 36

Therefore, the total volume is

 18  18  2
4 · 33
 
e e 1 3 4
V =π − − −π + +3
2 36 36 2 3
17e18 − 4373
 
=π .
36
1.7.2.28. *. Solution. To get rid of the square root in the argument of f 00 , we
make the change of variables (also called “substitution”) x = t2 , dx = 2t dt.
Z 4 √

Z 2
00
f x dx = 2 tf 00 (t) dt
0 0

Then, to convert f 00 into f 0 , we integrate by parts with u = t, dv = f 00 (t) dt,

v = f 0 (t).
Z 4 i2 Z 2
√

h 
00 0 0
f x dx = 2 tf (t) − f (t) dt
0 | {z } 0 0 | {z }
uv vdu
h i2
= 2 tf 0 (t) − f (t)
0
 0   
= 2 2f (2) − f (2) + f (0) = 2 2 × 4 − 3 + 1
= 12
1.7.2.29. Solution. As we saw in Section 1.1, there are many different ways to
interpret a limit as a Riemann sum. In the absence of instructions that restrain
our choices, we go with the most convenient interpretations.
With that in mind, we choose:

799
S OLUTIONS TO E XERCISES

• that our Riemann sum is a right Riemann sum (because we see i, not i − 1
or i − 21 )

• ∆x = n2 (because it is multiplied by the rest of the integrand, and also shows

up multiplied by i),

• then xi = a + i∆x = n2 i − 1, which leads us to a = −1 and

• f (x) = xex .
b−a 2
• Finally, since ∆x = n
= n
and a = −1, we have b = 1.

So, the limit is equal to the definite integral

n   Z 1
X 2 2 2
i−1
lim i − 1 en = xex dx
n→∞
i=1
n n −1

which we evaluate using integration by parts with u = x, dv = ex dx, du = dx,

and v = ex .
h i1 Z 1
x x
= |{z}
xe − e|{z}
dx
−1 −1
uv vdu
   
1 1 2
= e+ − e− =
e e e

1.8 · Trigonometric Integrals

1.8.4 · Exercises

Exercises — Stage 1
1.8.4.1. Solution. If u = cos x, then −du = dx. If n 6= −1, then
π/4
√
1/ 2  1/√2
1
Z Z
sin x cosn xdx = − un du = − un+1
0 1 n+1 1
!
1 1
= 1 − √ n+1
n+1 2

If n = −1, then
√ √
π/4 1/ 2 1/ 2
1
Z Z Z
sin x cosn xdx = − un du = − du
0 1 1 u
 1/√2
= − log |u|
1
 
1 1
= − log √ = log 2
2 2

So, (e) n can be any real number.

800
S OLUTIONS TO E XERCISES

1.8.4.2. Solution. We use the substitution u = sec x, du = sec x tan xdx.

Z Z Z
n n−1
sec x tan xdx = sec x · sec x tan xdx = un−1 du

Since n is positive, n − 1 6= −1, so we antidifferentiate using the power rule.

un 1
= + C = secn x + C
n n
1.8.4.3. Solution. We divide both sides by cos2 x, and simplify.

sin2 x + cos2 x = 1
sin2 x + cos2 x 1
=
cos2 x cos2 x
2
sin x
+ 1 = sec2 x
cos2 x
tan2 x + 1 = sec2 x

Exercises — Stage 2
1.8.4.4. *. Solution. The power of cosine is odd, and the power of sine is even
(zero). Following the strategy in the text, we make the substitution u = sin x, so
that du = cos x dx and cos2 x = 1 − sin2 x = 1 − u2 :
Z Z Z
cos x dx = (1 − sin x) cos x dx = (1 − u2 ) du
3 2

u3 sin3 x
=u− + C = sin x − +C
3 3
1 + cos(2x)
1.8.4.5. *. Solution. Using the trig identity cos2 x = , we have
2
1 π 1h 1 iπ π
Z Z
2

cos xdx = 1 + cos(2x) dx= x + sin(2x) =
2 0 2 2 0 2
1.8.4.6. *. Solution. Since the power of cosine is odd, following the strategies
in the text, we make the substitution u = sin t, so that du = cos t dt and cos2 t =
1 − sin2 t = 1 − u2 .
Z Z Z
sin t cos t dt = sin t (1 − sin t) cos t dt = u36 (1 − u2 ) du
36 3 36 2

u37 u39 sin37 t sin39 t

= − +C = − +C
37 39 37 39
1.8.4.7. Solution. Since the power of sine is odd (and positive), we can reserve
one sine for du, and turn the rest into cosines using the identity sin2 + cos2 x = 1.
This allows us to use the substitution u = cos x, du = − sin xdx, and sin2 x =

801
S OLUTIONS TO E XERCISES

1 − cos2 x = 1 − u2 .
sin3 x sin2 x 1 − u2
Z Z Z
dx = sin xdx = − du
cos4 x cos4 x u4
Z  
1 1 1 1
= − 4 + 2 du = 3 − + C
u u 3u u
1 1
= 3
− +C
3 cos x cos x
1.8.4.8. Solution. Both sine and cosine have even powers (four and zero, re-
spectively), so we don’t have the option of using a substitution like u = sin x or
1 − cos(2θ)
u = cos x. Instead, we use the identity sin2 θ = .
2
Z π/3 Z π/3 Z π/3  2
4 2

2 1 − cos(2x)
sin xdx = sin x dx = dx
0 0 0 2
1 π/3
Z
1 − 2 cos(2x) + cos2 (2x) dx


=
4 0
1 π/3 1 π/3
Z Z
= (1 − 2 cos(2x)) dx + cos2 (2x)dx
4 0 4 0

We can antidifferentiate the first integral right away. For the second integral, we
1 + cos(2θ)
use the identity cos2 θ = , with θ = 2x.
2

1h iπ/3 1 Z π/3
= x − sin(2x) + (1 + cos(4x))dx
4 0 8 0
" √ #
1 π 3 1h 1 iπ/3
= − + x + sin(4x)
4 3 2 8 4 0
" √ # " √ #
1 π 3 1 π 3
= − + −
4 3 2 8 3 8
√
π 9 3
= −
8 64
1.8.4.9. Solution. Since the power of sine is odd, we can reserve one sine for du,
and change the remaining four into cosines. This sets us up to use the substitu-
tion u = cos x, du = − sin xdx.
Z Z Z
sin xdx = sin x · sin xdx = (1 − cos2 x)2 sin xdx
5 4

Z Z
= − (1 − u ) du = − (1 − 2u2 + u4 )du
2 2

2 1
= −u + u3 − u5 + C
3 5
2 1
= − cos x + cos3 x − cos5 x + C
3 5

802
S OLUTIONS TO E XERCISES

1.8.4.10. Solution. If we use the substitution u = sin x, then du = cos xdx, which
very conveniently shows up in the integrand.

u2.2 1
Z Z
sin x cos xdx = u1.2 du =
1.2
+C = sin2.2 x + C
2.2 2.2
Note this is exactly the strategy described in the text when the power of cosine is
odd. The non-integer power of sine doesn’t cause a problem.
1.8.4.11. Solution.

• Solution 1: Let’s use the substitution u = tan x, du = sec2 xdx:

1 1
Z Z
tan x sec xdx = udu = u2 + C = tan2 x + C
2
2 2

• Solution 2: We can also use the substitution u = sec x, du = sec x tan xdx:
1 1
Z Z
2
tan x sec xdx = udu = u2 + C = sec2 x + C
2 2

We note that because tan2 x and sec2 x only differ by a constant, the two answers
are equivalent.
1.8.4.12. *. Solution.

• Solution 1: Substituting u = cos x, du = − sin x dx, sin2 x = 1 − cos2 x =

1 − u2 , gives

sin3 x (1 − cos2 x) sin x

Z Z Z
3 5
tan x sec x dx = dx = dx
cos8 x cos8 x
1 − u2 h u−7 u−5 i
Z
=− du = − − +C
u8 −7 −5
1 1
= sec7 x − sec5 x + C
7 5

• Solution 2: Alternatively, substituting u = sec x, du = sec x tan x dx, tan2 x =

sec2 x − 1 = u2 − 1, gives
Z Z
tan x sec x dx = tan2 x sec4 x (tan x sec x) dx
3 5

Z h u7 u5 i
= (u2 − 1)u4 du = − +C
7 5
1 1
= sec7 x − sec5 x + C
7 5
1.8.4.13. *. Solution. Use the substitution u = tan x, so that du = sec2 x dx:
Z Z
sec x tan x dx = (tan2 x + 1) tan46 x sec2 x dx
4 46

803
S OLUTIONS TO E XERCISES

u49 u47
Z
2 46
= (u + 1)u du = + +C
49 47
tan49 x tan47 x
= + +C
49 47
1.8.4.14. Solution. We use the substitution u = sec x, du = sec x tan xdx. Then
tan2 x = sec2 x − 1 = u2 − 1.
Z Z
tan x sec xdx = tan2 x · sec0.5 x · sec x tan xdx
3 1.5

Z Z
2 0.5
u2.5 − u0.5 du


= (u − 1)u du =
u3.5 u1.5
= − +C
3.5 1.5
1 1
= sec3.5 x − sec1.5 x + C
3.5 1.5
Note this solution used the same method as Example 1.8.13 for the case that the
power of tangent is odd and there is at least one secant.
1.8.4.15. Solution. As in Question 14, we have an odd power of tangent and at
least one secant. So, we can use the substitution u = sec x, du = sec x tan xdx, and
tan2 x = sec2 x − 1 = u2 − 1.
Z Z
tan x sec xdx = tan2 x sec x · sec x tan xdx
3 2

Z Z
2
u3 − u du


= (u − 1)udu =
1 1
= u4 − u2 + C
4 2
1 1
= sec4 x − sec2 x + C
4 2
1.8.4.16. Solution. In contrast to Questions 14 and 15, we do not have an odd
power of tangent, so we should consider a different substitution. Luckily, if we
choose u = tan x, then du = sec2 xdx, and this fits our integrand nicely.

1 1
Z Z
tan x sec xdx = u4 du = u5 + C = tan5 x + C
4 2
5 5
1.8.4.17. Solution.

• Solution 1: Since the power of tangent is odd, let’s try to use the substitution
u = sec x, du = sec x tan xdx, and tan2 x = sec2 x − 1 = u2 − 1, as in Ques-
tions 14 and 15. In order to make this work, we need to see sec x tan xdx in
the integrand, so we do a little algebraic manipulation.

tan3 x tan3 x
Z Z Z
3 −0.7
tan x sec xdx = dx = sec xdx
sec0.7x sec1.7x

804
S OLUTIONS TO E XERCISES

tan2 x
Z
= · sec x tan xdx
sec1.7 x
Z 2
u −1
Z
0.3 −1.7


= du = u − u du
u1.7
u1.3 1
= + +C
1.3 0.7u0.7
1 1
= sec1.3 x + +C
1.3 0.7 sec0.7 x
1 1
= sec1.3 x + cos0.7 x + C
1.3 0.7

• Solution 2: Let’s convert the secants and tangents to sines and cosines.

sin3 x
Z Z
3 −0.7
tan x sec xdx = 3
· cos0.7 xdx
cos x
sin3 x sin2 x
Z Z
= dx = · sin xdx
cos2.3 x cos2.3 x

Using the substitution u = cos x, du = − sin dx, and sin2 x = 1 − cos2 x =

1 − u2 :
1 − u2
Z Z
−2.3 −0.3


=− du = −u + u du
u2.3
1 −1.3 1 0.7
= u + u +C
1.3 0.7
1 1
= sec1.3 x + cos0.7 x + C
1.3 0.7
sin x
1.8.4.18. Solution. We replace tan x with.
cos x
5
sin4 x
Z 
sin x
Z Z
5
tan xdx = dx = · sin xdx
cos x cos5 x

Now we use the substitution u = cos x, du = − sin xdx, and sin2 x = 1 − cos2 x =
1 − u2 .
(1 − u2 )2
Z Z
−u−5 + 2u−3 − u−1 du


=− 5
du =
u
1 −4
= u − u−2 − log |u| + C
4
1
= sec4 x − sec2 x − log | cos x| + C
4
1
= sec4 x − sec2 x + log | sec x| + C
4
where in the last line, we used the logarithm rule log(ba ) = a log b, with ba =
cos x = (sec x)−1 .

805
S OLUTIONS TO E XERCISES

1.8.4.19. Solution. Integrating even powers of tangent is surprisingly different

from integrating odd powers of tangent. For even powers, we use the identity
tan2 x = sec2 x − 1, then use the substitution u = tan x, du = sec2 xdx on (perhaps
only a part of) the resulting integral.
Z π/6 Z π/6
6
tan xdx = tan4 x(sec2 x − 1)dx
0 0
Z π/6  
4 2 4
= tan
| x{zsec x} − tan x dx
0
u4 du
Z π/6  
4 2 2 2
= tan x sec x − tan x(sec x − 1) dx
0
Z π/6  
4 2 2 2 2
= tan x sec x − tan
| x{zsec x} + tan x dx
0
u2 du
Z π/6  
4 2 2 2 2
= tan x sec x − tan x sec x + (sec
| {z x} −1) dx
0
du
Z π/6 Z π/6
tan4 x − tan2 x + 1 sec2 xdx −


= 1dx
0 0
√
Note tan(0) = 0, and tan(π/6) = 1/ 3.
√
Z 1/ 3  π/6
= (u4 − u2 + 1)du − x 0
0
 1/√3
1 5 1 3 π
= u − u +u −
5 3 0 6
1 1 1 π
= √ 5− √ 3+√ −
5 3 3 3 3 6
41 π
= √ −
45 3 6
1.8.4.20. Solution. Since there is an even power of secant in the integrand, we
can reserve two secants for du and change the rest to tangents. That sets us up
nicely to use the substitution u = tan x, du = sec2 xdx. Note tan(0) = 0 and
tan(π/4) = 1.
Z π/4 Z π/4
8 4
tan x sec xdx = tan8 x(tan2 x + 1) sec2 xdx
0
Z0 1
= u8 (u2 + 1)du
Z0 1
= u10 + u8 du
0
1 1
= +
11 9

806
S OLUTIONS TO E XERCISES

1.8.4.21. Solution.

• Solution 1: Let’s use the substitution u = sec x, du = sec x tan xdx. In order
to make this work, we need to see sec x tan x in the integrand, so we start
with some algebraic manipulation.

√
√ 
sec x 1
Z Z
tan x sec x √ dx = √ sec x tan xdx
sec x sec x
1 √
Z
= √ du = 2 u + C
u
√
= 2 sec x + C

• Solution 2: Let’s turn our secants and tangents into sines and cosines.
√ sin x sin x
Z Z Z
tan x sec xdx = √ dx = dx
cos x · cos x cos1.5 x

We use the substitution u = cos x, du = − sin xdx.

2
Z
= −u−1.5 du = √ + C
u
√
= 2 sec x + C
1.8.4.22. Solution. Since the power of secant is even and positive, we can re-
serve two secants for du, and change the rest into tangents, setting the stage for
the substitution u = tan θ, du = sec2 θdθ.
Z Z
sec θ tan θdθ = sec6 θ tane θ sec2 θdθ
8 e

Z
= (tan2 θ + 1)3 tane θ sec2 θdθ
Z
= (u2 + 1)3 · ue du
Z
= (u6 + 3u4 + 3u2 + 1) · ue du
Z
= (u6+e + 3u4+e + 3u2+e + ue )du
1 7+e 3 5+e 3 3+e 1 1+e
= u + u + u + u +C
7+e 5+e 3+e 1+e
1 3 3
= tan7+e θ + tan5+e θ + tan3+e θ
7+e 5+e 3+e
1
+ tan1+e θ + C
1 + e
tan6 θ 3 tan4 θ 3 tan2 θ
 
1+e 1
= tan θ + + + +C
7+e 5+e 3+e 1+e

807
S OLUTIONS TO E XERCISES

Exercises — Stage 3
1.8.4.23. *. Solution. (a) Using the trig identity tan2 x = sec2 x − 1 and the
substitution y = tan x, dy = sec2 x dx,
Z Z
tan xdx = tann−2 x tan2 xdx
n

Z Z
n−2 2
= tan x sec xdx − tann−2 xdx
Z Z
n−2
= y dy − tann−2 xdx
y n−1
Z
= − tann−2 xdx
n−1
tann−1 x
Z
= − tann−2 xdx
n−1
(b) By the reduction formula of part (a),
π/4 π/4 Z π/4
tann−1 x
Z 
n
tan (x) dx = − tann−2 (x) dx
0 n−1 0 0
Z π/4
1
= − tann−2 (x) dx
n−1 0

for all integers n ≥ 2, since tan 0 = 0 and tan π4 = 1. We apply this reduction
formula, with n = 6, 4, 2.
π/4 π/4
1
Z Z
6
tan (x) dx = − tan4 (x) dx
0 5 0
Z π/4
1 1
= − + tan2 (x) dx
5 3 0
Z π/4
1 1
= − +1− dx
5 3 0
1 1 π 13 π
= − +1− = −
5 3 4 15 4
Using a calculator, we see this is approximately 0.0813.
Notice how much faster this was than the method of Question 19.
sin x
1.8.4.24. Solution. Recall tan x = .
cos x
sin5 x sin5 x
Z Z Z
5 2
tan x cos xdx = 5
cos2 xdx = dx
cos x cos3 x

Substitute u = cos x, so du = − sin xdx and sin2 x = 1 − cos2 x = 1 − u2 .

sin4 x (1 − u2 )2
Z Z
= sin xdx = − du
cos3 x u3

808
S OLUTIONS TO E XERCISES

1 − 2u2 + u4
Z  
1 2
Z
=− du = − 3 + − u du
u3 u u
1 1 2
= 2 + 2 log |u| − u + C
2u 2
1 1
= 2
+ 2 log | cos x| − cos2 x + C
2 cos x 2
1.8.4.25. Solution. We can use the definition of secant to make this integral look
more familiar.
1
Z Z
dθ = sec2 θdθ = tan θ + C
cos2 θ
cos x
1.8.4.26. Solution. We re-write cot x = , and use the substitution u = sin x,
sin x
du = cos xdx.
cos x 1
Z Z Z
cot xdx = dx = du
sin x u
= log |u| + C = log | sin x| + C
1.8.4.27. Solution.

• Solution 1: We begin with the obvious substitution, w = ex , dw = ex dw.

Z Z
x x x
e sin(e ) cos(e )dx = sin w cos wdw

Now we see another substitution, u = sin w, du = cos wdw.

1 1
Z
= udu = u2 + C = sin2 w + C
2 2
1 2 x
= sin (e ) + C
2
d
• Solution 2: Notice that dx {sin(ex )} = ex cos(ex ). This suggests to us the
substitution u = sin(ex ), du = ex cos(ex )dx.

1 1
Z Z
e sin(e ) cos(e )dx = udu = u2 + C = sin2 (ex ) + C
x x x
2 2
1.8.4.28. Solution. Since we have an “inside function,” we start with the sub-
stitution s = cos x, so −ds = sin xdx and sin2 x = 1 − cos2 x = 1 − s2 .
Z Z
sin(cos x) sin xdx = sin(cos x) · sin2 x · sin xdx
3

Z
= − sin(s) · (1 − s2 )ds

809
S OLUTIONS TO E XERCISES

We use integration by parts with u = (1 − s2 ), dv = sin sds; du = −2sds, and

v = − cos s.
 Z 
2
= − −(1 − s ) cos s − 2s cos sds
Z
2
= (1 − s ) cos s + 2s cos sds

We integrate by parts again, with u = 2s, dv = cos sds; du = 2ds, and v = sin s.
Z
2
= (1 − s ) cos s + 2s sin s − 2 sin sds

= (1 − s2 ) cos s + 2s sin s + 2 cos s + C

= sin2 x · cos(cos x) + 2 cos x · sin(cos x) + 2 cos(cos x) + C
= (sin2 x + 2) cos(cos x) + 2 cos x · sin(cos x) + C
1.8.4.29. Solution. Since the integrand is the product of polynomial and
trigonometric functions, we suspect it might yield to integration by parts. There
are a number of ways this can be accomplished.

• Solution 1: Before we choose parts, let’s use the identity sin(2x) =

2 sin x cos x.
1
Z Z
x sin x cos xdx = x sin(2x)dx
2
Now let u = x, dv = sin(2x)dx; du = dx, and v = − 12 cos(2x). Using
integration by parts:
 
1 x 1
Z
= − cos(2x) + cos(2x)dx
2 2 2
x 1
= − cos(2x) + sin(2x) + C
4 8
x 1
= − (1 − 2 sin2 x) + sin x cos x + C
4 4
x x 2 1
= − + sin x + sin x cos x + C
4 2 4
• Solution 2: If we let u = x, then du = dx, and this seems desirable for
integration by parts. If u = x, then dv = sin x cos xdx. To find v we can use
the substitution u = sin x, du = cos xdx.
1 1
Z Z
v = sin x cos xdx = udu = u2 + C = sin2 x + C
2 2
So, we take v = 21 sin2 x. Now we can apply integration by parts to our
original integral.
x 2 1 2
Z Z
x sin x cos xdx = sin x − sin xdx
2 2

810
S OLUTIONS TO E XERCISES

1 − cos(2x)
Apply the identity sin2 x = .
2
x 1
Z
= sin2 x − 1 − cos(2x)dx
2 4
x 2 x 1
= sin x − + sin(2x) + C
2 4 8
x 2 x 1
= sin x − + sin x cos x + C
2 4 4

• Solution 3: Let u = x sin x and dv = cos xdx; then du = (x cos x + sin x)dx
and v = sin x.
Z Z
2
x sin x cos xdx = x sin x − sin x(x cos x + sin x)dx
Z Z
= x sin x − x sin x cos xdx − sin2 xdx
2

1 − cos(2x)
Apply the identity sin2 x = to the second integral.
2
1 − cos(2x)
Z Z
2
= x sin x − x sin x cos xdx − dx
2
x 1
Z
2
= x sin x − x sin x cos xdx − + sin(2x) + C
2 4
So, we have the equation
Z Z
2
x sin x cos xdx = x sin x − x sin x cos xdx
x 1
−
+ sin(2x) + C
2 4
x 1
Z
2 x sin x cos xdx = x sin2 x − + sin(2x) + C
2 4
x 2 x 1 C
Z
x sin x cos xdx = sin x − + sin(2x) +
2 4 8 2
x x 1 C
= sin2 x − + sin x cos x +
2 4 4 2
Since C is an arbitrary constant that can take any number in (−∞, ∞), also
C
2
is an arbitrary constant that can take any number in (−∞, ∞), so we’re
free to rename C2 to C.

1.9 · Trigonometric Substitution

1.9.2 · Exercises

Exercises — Stage 1

811
S OLUTIONS TO E XERCISES

1.9.2.1. *. Solution. In the text, there is a template for choosing an appropriate

substitution, but for this problem we will explain the logic of the choices.
The trig identities that we can use are:

1 − sin2 θ = cos2 θ tan2 θ + 1 = sec2 θ sec2 θ − 1 = tan2 θ

They have the following forms:

constant − function function + constant function − constant

In order to cancel out the square root, we should choose a substitution that will
match the argument under the square root with the trig identity of the corre-
sponding form.
(a) There’s not an obvious non-trig substitution for evaluating this problem, so
we want a trigonometric substitution to get rid of the square root in the denom-
inator. Under the square root is the function 9x2 − 16, which has the form (func-
tion) − (constant). This form matches the trig identity sec2 θ − 1 = tan2 θ. We can
set x to be whatever we need it to be, but we don’t have the same control over
the constant, 16. So, to make the substitution work, we use a different form of
the trig identity: multiplying both sides by 16, we get

16 sec2 θ − 16 = 16 tan2 θ

What we want is a substitution that gives us

9x2 − 16 = 16 sec2 θ − 16
So, 9x2 = 16 sec2 θ
4
x = sec θ
3
Using this substitution,
√ √
9x2 − 16 = 16 sec2 θ − 16
√
= 16 tan2 θ
= 4| tan θ|

So, we eliminated the square root.

(b) There’s not an obvious non-trig substitution for evaluating this problem, so
we want a trigonometric substitution to get rid of the square root in the denomi-
nator. Under the square root is the function 1−4x2 , which has the form (constant)
− (function). This form matches the trig identity 1−sin2 θ = cos2 θ. What we want
is a substitution that gives us

1 − 4x2 = 1 − sin2 θ
So, 4x2 = sin2 θ
1
x = sin θ
2

812
S OLUTIONS TO E XERCISES

Using this substitution,

√ p
1 − 4x2 = 1 − sin2 θ
√
= cos2 θ
= | cos θ|

So, we eliminated the square root.

(c) There’s not an obvious non-trig substitution for evaluating this problem, so
we want a trigonometric substitution to get rid of the fractional power. (That is,
we want to eliminate the square root.) The function under the power is 25 + x2 ,
which has the form (constant) + (function). This form matches the trig identity
tan2 θ + 1 = sec2 θ. We can set x to be whatever we need it to be, but we don’t
have the same control over the constant, 25. So, to make the substitution work,
we use a different form of the trig identity: multiplying both sides by 25, we get

25 tan2 θ + 25 = 25 sec2 θ

What we want is a substitution that gives us

25 + x2 = 25 tan2 θ + 25
So, x2 = 25 tan2 θ
x = 5 tan θ

Using this substitution,

(25 + x2 )−5/2 = (25 + 25 tan2 θ)−5/2

= (25 sec2 θ)−5/2
= (5| sec θ|)−5

So, we eliminated the square root.

1.9.2.2. Solution. Just as in Question 1, we want to choose a trigonometric sub-
stitution that will allow us to eliminate the square roots. Before we can make
that choice, though, we need to complete the square. In subsequent problems,
we won’t show the algebra behind completing the square, but for this problem
we’ll work it out explicitly. After some practice, you’ll be able to do this step in
your head for many cases.
After the squares are completed, the choice of trig substitution follows the logic
outlined in the solutions to Question 1, or (equivalently) the template in the text.

a The quadratic function under the square root is x2 − 4x + 1. To complete the

square, we match the non-constant terms to those of a perfect square.

(ax + b)2 = a2 x2 + 2abx + b2

x2 − 4x + 1 = a2 x2 + 2abx + b2 + c for some constant c

813
S OLUTIONS TO E XERCISES

• Looking at the leading term tells us a = 1.

• Then the second term tells us −4 = 2ab = 2b, so b = −2.
• Finally, the constant terms give us 1 = b2 + c = 4 + c, so c = −3.

1 1
Z Z
√ dx = p dx
2
x − 4x + 1 (x − 2)2 − 3
1
Z
= q √ 2 dx
2
(x − 2) − 3

√
So we use the substitution (x − 2) = 3 sec u, which eliminates the square
root: q √ √ √
(x − 2)2 − 3 = 3 sec2 u − 3 = 3 tan2 u = 3| tan u|

b The quadratic function under the square root is −x2 + 2x + 4 = −[x2 − 2x −

4]. To complete the square, we match the non-constant terms to those of a
perfect square. We factored out the negative to make things a little easier
— don’t forget to put it back in before choosing a substitution!

(ax + b)2 = a2 x2 + 2abx + b2

x2 − 2x − 4 = a2 x2 + 2abx + b2 + c for some constant c

• Looking at the leading term tells us a = 1.

• Then the second term tells us −2 = 2ab = 2b, so b = −1.
• Finally, the constant terms give us −4 = b2 + c = 1 + c, so c = −5.
• Then

−x2 + 2x + 4 = −[x2 − 2x − 4] = −[(x − 1)2 − 5]

= 5 − (x − 1)2

(x − 1)6 (x − 1)6
Z Z
dx = dx
(−x2 + 2x + 4)3/2 (5 − (x − 1)2 )3/2
(x − 1)6
Z
= √ 2 3/2 dx
5 − (x − 1) 2

√
So we use the substitution (x − 1) = 5 sin u, which eliminates the square
root (fractional power):
3/2 3/2 3/2
(5 − (x − 1)2 ) = (5 − 5 sin2 u) = (5 cos2 u)
√
= 5 5| cos3 u|

814
S OLUTIONS TO E XERCISES

c The quadratic function under the square root is 4x2 + 6x + 10. To complete
the square, we match the non-constant terms to those of a perfect square.
(ax + b)2 = a2 x2 + 2abx + b2
4x2 + 6x + 10 = a2 x2 + 2abx + b2 + c for some constant c
• Looking at the leading term tells us a = 2.
• Then the second term tells us 6 = 2ab = 4b, so b = 23 .
9 31
• Finally, the constant terms give us 10 = b2 + c = 4
+ c, so c = 4
.

1 1
Z Z
√ dx = q dx
2
4x + 6x + 10 3 2

31
2x + 2
+ 4
1
Z
= r  √ 2 dx
3 2 31


2x + 2
+ 2
√
So we use the substitution 2x + 23 = 231 tan u, which eliminates the

square root:
s 2 r r
3 31 31 31 31
2x + + = tan2 u + = sec2 u
2 4 4 4 4
√
31
= | sec u|
2
d The quadratic function under the square root is x2 − x. To complete the
square, we match the non-constant terms to those of a perfect square.
(ax + b)2 = a2 x2 + 2abx + b2
x2 − x = a2 x2 + 2abx + b2 + c for some constant c
• Looking at the leading term tells us a = 1.
• Then the second term tells us −1 = 2ab = 2b, so b = − 12 .
1
• Finally, the constant terms give us 0 = b2 + c = 4
+ c, so c = − 14 .
s 2
Z √
1 1
Z
x2 − xdx = x− − dx
2 4
s 2  2
1 1
Z
= x− − dx
2 2

So we use the substitution (x − 1/2) = 21 sec u, which eliminates the square

root:
s
1.9.2.3. Solution. 2 r r
1 1 1 1 1 1
x− − = sec 2 u − = tan2 u = | tan u|
2 4 4 4 4 2

815
S OLUTIONS TO E XERCISES

1
a If sin θ = and θ is between 0 and π/2, then we can draw a right triangle
20
with angle θ that has opposite side length 1, and hypotenuse
√ length 20.
√ By
the Pythagorean Theorem, the adjacent side has length 202 − 12 = 399.
√
adj 399
So, cos θ = = .
hyp 20

20
1
θ
√
399

1
We can do a quick “reasonableness” check here: 20 is pretty close to 0, so
we might expect θ√to be pretty
√ close to 0, and so cos θ should be pretty close
399 400 20
to 1. Indeed it is: ≈ = = 1.
20 20 20
Alternatively, we can solve this problem using identities.
sin2 θ + cos2 θ = 1
 2
1
+ cos2 θ = 1
20
r √
1 399
cos θ = ± 1− =±
400 20
Since 0 ≤ θ ≤ π2 , cos θ ≥ 0, so
√
399
cos θ =
20
b If tan θ = 7 and θ is between 0 and π/2, then we can draw a right triangle
with angle θ that has opposite side length 7 and adjacent
√ side length√1. By
2 2
the Pythagorean Theorem, √ the hypotenuse has length 7 + 1 = 50 =
√ hyp 5 2
5 2. So, csc θ = = .
opp 7

√ 2
5 7
θ
1

Again, we can do a quick reasonableness check. Since 7 is much larger

than 1, the triangle we’re thinking of doesn’t look much like the triangle
in our standardized picture above: it’s really quite tall, with a small base.
So,
√ the opposite side and hypotenuse are pretty close in length. Indeed,
5 2
≈ 7.071, so this dimension seems reasonable.
7

816
S OLUTIONS TO E XERCISES
√
x−1
c If sec θ = and θ is between 0 and π/2, then we can draw a right
2 √
triangle with angle θ that has hypotenuse length x − 1 and adjacent
side length 2. By the Pythagorean Theorem, the opposite √side has length
√ √ √ opp x−5
q
2
x − 1 − 22 = x − 1 − 4 = x − 5. So, tan θ = = .
adj 2

1
√ x− √
x−5
θ
2

We can also solve this using identities. Note that since sec θ exists, θ 6= π2 .

tan2 θ + 1 = sec2 θ
√ 2
2 x−1 x−1
tan θ + 1 = =
2 4
r √
x−1 x−5
tan θ = ± −1=±
4 2
Since 0 ≤ θ < π2 , tan θ ≥ 0, so
√
x−5
tan θ =
2
1.9.2.4. Solution.

a Let θ = arccos x2 . That is, cos(θ) = x2 , and 0 ≤ θ ≤ π. Then we can draw

the corresponding right triangle with angle θ with adjacent side of signed
length x (we note that if θ > π2 , then x is negative) and hypotenuse of length
2.
√ By the Pythagorean Theorem, the opposite side of the triangle has length
4 − x2 .

2 √
4 − x2
θ
x

So, √
  x  opp 4 − x2
sin arccos = sin θ = =
2 hyp 2
 
b Let θ = arctan √1 . That is, tan(θ) = √1 , and − π2 ≤ θ ≤ π2 .
3 3

817
S OLUTIONS TO E XERCISES

π 1
• Solution 1: Then θ = , so sin θ = .
6 2
• Solution 2: Then we can draw the corresponding right triangle with
angle θ with opposite side of length 1 and adjacent side of length
√
3. Byqthe Pythagorean Theorem, the hypotenuse of the triangle has
√ 2
length 3 + 12 = 2.

2
1
θ
√
3

So,   
1 opp 1
sin arctan √ = sin θ = =
3 hyp 2
√ √
c Let θ = arcsin ( x). That is, sin(θ) = x, and − π2 ≤ θ ≤ π2 . Then we can
√ corresponding right triangle with angle θ with opposite side of
draw the
length x and hypotenuse of length 1.√By the Pythagorean Theorem, the
adjacent side of the triangle has length 1 − x.

1 √
x
θ
√
1−x

So,
√

hyp 1
sec arcsin x = sec θ = =√
adj 1−x

Exercises — Stage 2
1.9.2.5. *. Solution. Let x = 2 tan θ, so that x2 + 4 = 4 tan2 θ + 4 = 4 sec2 θ and
dx = 2 sec2 θ dθ. Then
1 1
Z Z
2 3/2
dx = 2 3/2
· 2 sec2 θ dθ
(x + 4) (4 sec θ)
2 sec2 θ
Z
= dθ
8 sec3 θ
1
Z
= cos θ dθ
4
1 1 x
= sin θ + C = √ +C
4 4 x2 + 4

818
S OLUTIONS TO E XERCISES

4
√ x2 +
x
θ
2
x
To find sin θ in terms of x, we construct the right triangle above. Since tan θ = =
2
opp
, we label the opposite side x and the adjacent side 2. By the Pythagorean
adj
√ opp x
Theorem, the hypotenuse has length x2 + 4. Then sin θ = =√ .
hyp x2 + 4
To see why we could write (sec2 θ)3/2 = sec3 θ, as opposed to (sec2 θ)3/2 = sec3 θ ,
in the second line above, see Example 1.9.5.
As a check, we observe that the derivative of the answer
 
d 1 x 1 1 1 x(2x)
√ +C = √ −
dx 4 x2 + 4 4 x2 + 4 2 × 4 x2 + 4 3/2

x2 2
+ 1 − x4
4 1
=
3/2 =
3/2
x2 + 4 x2 + 4

is exactly the integrand.

1.9.2.6. *. Solution.
• Solution 1: As in Question 5, substitute x = 2 tan u, dx = 2 sec2 u du. Note
that when x = 4 we have 4 = 2 tan u, so that tan u = 2.
Z 4 Z arctan 2
1 1
3/2
dx = 3/2
2 sec2 u du
2 2
0 (4 + x ) 0 (4 + 4 tan u)
Z arctan 2
2 sec2 u
= du
0 (2 sec u)3
1 arctan 2 sec2 u
Z
= du
4 0 sec3 u
1 arctan 2
Z
= cos u du
4 0
 arctan 2
1
= sin u
4 0
1
1
= sin(arctan 2) − 0 = √
4 2 5

√ 5
2
u
1

819
S OLUTIONS TO E XERCISES

To find sin(arctan 2), we use the right triangle above, with angle u =
opp
arctan 2. Since tan u = 2 = , we label the opposite side as 2, and the
adj
adjacent side as 1. The Pythagorean Theorem tells us the hypotenuse has
√ opp 2
length 5, so sin u = =√ .
hyp 5
• Solution 2: Using our result from Question 5,
4  4
1 1 x
Z
dx = √
0 (4 + x2 )3/2 4 x2 + 4 0
1 4 1
= ·√ = √
4 42 + 4 2 5
1.9.2.7. *. Solution. Make the change of variables x = 5 sin θ, dx = 5 cos θ dθ.
Since x = 0 corresponds to θ = 0 and x = 52 correponds to sin θ = 21 or θ = π6 ,
5/2 π/6 π/6
dx 5 cos θ dθ π
Z Z Z
√ = p = dθ =
0 25 − x2 0 25 − 25 sin2 θ 0 6

1.9.2.8. *. Solution. Substitute x = 5 tan u, so that dx = 5 sec2 u du.

1 1
Z Z
√ dx = √ 5 sec2 u du
2
x + 25 25 tan2 u + 25
5 sec2 u
Z Z
= du = sec u du
5 sec u
= log sec u + tan u + C
r
x2 x
= log 1+ + +C
25 5

25
√ x2 + x
u
5

To find sec u and tan u, we have two options. One is to set up a right triangle with
angle u and tan u = x5 . Then we can label the opposite side x and the adjacent
√
side 5, and use Pythagorus to find that the hypotenuse is x2 + 25.
Another option is to look back at our work a little more closely — in fact, we’ve
already found what we’re looking for. Since we used the substitution x = 5 tan u,
this gives us tan u = x5 . In the denominator of the integrand, we simplified
√ √ q
2
x2 + 25 = 5 sec u, so sec u = 51 x2 + 25 = 1 + x25 .
√ √
To see why we could write x2 + 25 = 5 sec u, as opposed to x2 + 25 = 5| sec u|,
see Example 1.9.5.

820
S OLUTIONS TO E XERCISES

1.9.2.9. Solution. The quadratic formula underneath the square root makes us
think of a trig substitution, but in the interest of developing good habits, let’s
check for an easier way first. If we let u = 2x2 + 4x, then du = (4x + 4)dx, so
1
4
du = (x + 1) dx. This substitution looks easier than a trig substitution (which
would start with completing the square).

x+1 1 1 1√ 1√ 2
Z Z
√ dx = √ du = u+C = 2x + 4x + C
2x2 + 4x 4 u 2 2
1.9.2.10. *. Solution. Substitute x = 4 tan u, dx = 4 sec2 u du.
1 1
Z Z
√ dx = 2
√
2
4 sec2 u du
2 2
x x + 16 16 tan u 16 tan u + 16
sec2 u 1 sec u
Z Z
= 2
du = du
16 tan u sec u 16 tan2 u
1 cos u
Z
= du
16 sin2 u
To finish off the integral, we’ll substitute v = sin u, dv = cos u du.

1 1 cos u 1 dv 1
Z Z Z
√ dx = 2 du = = − +C
x2 x2 + 16 16 sin u 16 v2 16v
√
1 1 x2 + 16
=− +C =− +C
16 sin u 16 x

16
√ x2 + x
u
4

To find sin u, we draw a right triangle with angle u and tan u = x4 . We label the
opposite side x and the adjacent side 4, and then√from Pythagorus we find that
√ x2 + 16
the hypotenuse has length x2 + 16. So, sin u = .
x
As a check, we observe that the derivative of the answer
√ ! √
d 1 x2 + 16 1 x2 + 16 1 x
− +C = 2
− √
dx 16 x 16 x 16 x x2 + 16
1 (x2 + 16) − x2
= √
16 x2 x2 + 16
1
= √
x2 x2 + 16
is exactly the integrand.

821
S OLUTIONS TO E XERCISES

1.9.2.11. *. Solution. Substitute x = 3 sec u with 0 ≤ u < π2 . Then dx =

√ √ √
3 sec u tan u du and x2 − 9 = 9 sec2 u − 9 = 9 tan2 u = 3 tan u, so that
dx 3 sec u tan udu
Z Z
√ = √
x2 x2 − 9 9 sec 2 u 9 tan2 u

1 du
Z
=
9 sec u
1 1
Z
= cos udu = sin u + C.
9 9
x hyp
To evaluate sin u, we make a right triangle with angle u. Since sec u = = ,
3 adj
we label the hypotenuse x and the adjacent side 3.

x √
x2 − 9
u
3
√
Using the
√ Pythagorean Theorem, the opposite side has length x2 − 9. So,
x2 − 9
sin u = and
x
√
dx x2 − 9
Z
√ = + C.
x2 x2 − 9 9x

As a check, we observe that the derivative of the answer

√ 2 √
x2 − 9

d x −9 x
+C =− 2
+ √
dx 9x 9x 9x x2 − 9
1 −(x2 − 9) + x2 1
= √ = √
9 x x −9
2 2 x x2 − 9
2

is exactly the integrand. (We remark that this is the case even for x ≤ −3.)
1.9.2.12. *. Solution. (a) We’ll use the trig identity cos 2θ = 2 cos2 θ−1. It implies
that
cos 2θ + 1 1 2
cos2 θ = =⇒ cos4 θ =

cos 2θ + 2 cos 2θ + 1
2 4
1 h cos 4θ + 1 i
= + 2 cos 2θ + 1
4 2
cos 4θ cos 2θ 3
= + +
8 2 8
So,
Z π/4 Z π/4 
4 cos 4θ cos 2θ 3 
cos θdθ = + + dθ
0 0 8 2 8

822
S OLUTIONS TO E XERCISES
 π/4
sin 4θ sin 2θ 3
= + + θ
32 4 8 0
1 3 π
= + ·
4 8 4
8 + 3π
=
32
as required.
(b) We’ll use the trig substitution x = tan θ, dx = sec2 θdθ. Note that when θ =
± π4 , we have x = ±1. Also note that dividing the trig identity sin2 θ + cos2 θ = 1
by cos2 θ gives the trig identity tan2 θ + 1 = sec2 θ. So
1 1
dx dx
Z Z
=2 (even integrand)
−1 (x2 + 1)3 0 (x2 + 1)3
π/4
sec2 θdθ
Z
=2 3
0 (tan2 θ + 1)
π/4
sec2 θdθ
Z
=2
0 (sec2 θ)3
Z π/4
=2 cos4 θdθ
0
8 + 3π
=
16
by part (a).
1.9.2.13. Solution. The integrand is an odd function, and the limits of integra-
Z π/12
15x3
tion are symmetric, so √ 5 dx = 0.
−π/12 (x2 + 1) 9 − x2

1.9.2.14. *. Solution. Substitute x = 2 sin u, so that dx = 2 cos u du.

Z √ Z p
2
4 − x dx = 4 − 4 sin2 u 2 cos u du
Z √
= 4 cos2 u 2 cos u du
Z Z
2


= 4 cos u du = 2 1 + cos(2u) du

= 2u + sin(2u) + C
= 2u + 2 sin u cos u + C
x x√
= 2 arcsin + 4 − x2 + C
2 2

823
S OLUTIONS TO E XERCISES

2
x
u
√
4 − x2
√ √
To see why we could write 4 cos2 u = 2 cos u, as opposed to 4 cos2 u = 2| cos u|,
in the third line above, see Example 1.9.2.
We used the substitution x = 2 sin u, so we know sin u = x2 and u = arcsin( x2 ). We
have three options for finding cos u.
First, we can draw a right triangle with angle u. Since sin u = x2 , we label the
opposite side x and the hypotenuse 2, then by the√Pythagorean Theorem the
√ adj 4 − x2
adjacent side has length 4 − x2 . So, cos u = = .
hyp 2 √
Second, we√can look back carefully at our work. We simplified 4 − x2 = 2 cos u,
4 − x2
so cos u = .
2 p
2
Third,
q we could use the identity sin u + cos 2
u = 1. Then cos u = ± 1 − sin2 u =
x2
± 1− 4
Since u = arcsin(x/2), u is in the range of arcsine, which means − π2 ≤
.
q √
2 2
u ≤ 2 . Therefore, cos u ≥ 0, so cos u = 1 − x4 = 4−x
π
2
.
So,
Z √
4 − x2 dx = 2u + 2 sin u cos u + C
√
x 4 − x2
= 2 arcsin + x · +C
2 2
1.9.2.15. *. Solution. Substitute x = 25 sec u with 0 < u < π2 , so that dx =
√ p √
2
5
sec u tan u du and 25x2 − 4 = 4(sec2 u − 1) = 4 tan2 u = 2 tan u. Then
Z √
25x2 − 4 2 tan u 2
Z
dx = 2 · sec u tan u du
x 5
sec u 5
Z Z
2
sec2 u − 1 du


= 2 tan u du = 2

= 2 tan u − 2u + C
√
= 25x2 − 4 − 2 arcsec 5x
2
+C

5x √
25x2 − 4
u
2

5x
To find tan u, we draw a right triangle with angle u. Since sec u = , we label
2

824
S OLUTIONS TO E XERCISES

the hypotenuse 5x and the adjacent side 2. Then the Pythagorean√ Theorem gives
√ opp 25x2 − 4
us the opposite side as length 25x2 − 4. Then tan u = = .
adj 2
Alternately,
√ we can notice
√ that in our work, we already showed 2 tan u =
2 1 2
25x − 4, so tan u = 2 25x − 4.
As a check, we observe that the derivative of the answer

d √
 
5x
25x2 − 4 − 2 arcsec +C
dx 2
5
25x
=√ −2 q2
25x2 − 4 5x 25x2
−1
2 4
25x 4
=√ − √
25x2 − 4 x 25x2 − 4
since x > 0
25x2 − 4
= √
x 25x2 − 4
√
25x2 − 4
=
x
is exactly the integrand (provided x > 25 ).
1.9.2.16. Solution. The integrand has a quadratic polynomial under a square
root, which makes us think of trig substitutions. However, it’s good practice to
look for simpler methods before we jump into more complicated ones, and in this
case we find something nicer than a trig substitution: the substitution
√ u = x2 − 1,
1
du = 2x dx. Then xdx = 2 du, and x2 = u + 1. When x = 10, u = 9, and when
√
x = 17, u = 16.

√ √
17 17
x3 x2
Z Z
√
√ dx = √
√ · xdx
10 x2 − 1 10 x2 − 1
1 16 u + 1
Z
= √ du
2 9 u
Z 16
1
u1/2 + u−1/2 du


=
2 9
 16
1 2 3/2 1/2
= u + 2u
2 3 9
 
1 2 3 2 3
= ·4 +2·4− ·3 −2·3
2 3 3
40
=
3
1.9.2.17. *. Solution. This integrand looks very different from those above. But

825
S OLUTIONS TO E XERCISES

it is only slightly disguised. If we complete the square

dx dx
Z Z
√ = p
3 − 2x − x2 4 − (x + 1)2

and make the substitution y = x + 1, dy = dx

dx dx dy
Z Z Z
√ = p = p
3 − 2x − x2 4 − (x + 1)2 4 − y2

we get a typical trig substitution integral. So, we substitute y = 2 sin θ, dy =

2 cos θ dθ to get

dx dy 2 cos θ dθ 2 cos θ dθ
Z Z Z Z
√ = p = p = √
3 − 2x − x 2 4−y 2 2 4 cos2 θ
4 − 4 sin θ
y
Z
= dθ = θ + C = arcsin + C
2
x+1
= arcsin +C
2
An experienced integrator would probably substitute x + 1 = 2 sin θ directly,
without going through y.
1.9.2.18. Solution. Completing the square, we see 4x2 − 12x + 8 = (2x − 3)2 − 1.

1 1
Z Z
√ dx = p dx
3 2
(2x − 3) 4x − 12x + 8 (2x − 3) (2x − 3)2 − 1
3

As x > 2, we have 2x − 3 > 1. Wep use the substitution

√ 2x − 3 =√sec θ with 0 ≤ θ <
π
2
. So 2 dx = sec θ tan θ dθ and (2x − 3) − 1 = sec θ − 1 = tan2 θ = tan θ.
2 2

1 1
Z
= √ sec θ tan θdθ
2 sec 3 θ sec2 θ − 1

1 1
Z
= 3
sec θ tan θdθ
2 sec θ tan θ
1 1
Z
= dθ
2 sec2 θ
1
Z
= cos2 θdθ
2
1
Z
= (1 + cos(2θ))dθ
4
 
1 1
= θ + sin(2θ) + C
4 2
1
= (θ + sin θ cos θ) + C
4
   √ 2 
1 1 4x − 12x + 8
= arccos + +C
4 2x − 3 (2x − 3)2

826
S OLUTIONS TO E XERCISES

− 3
√
2x 4x2 − 12x + 8
θ
1

1 1


Since 2x − 3 = sec θ, we know cos θ = 2x−3 and θ = arccos 2x−3 . (Equivalently,
θ = arcsec(2x − 3).) To find sin θ, we draw a right triangle with adjacent side
√ of length 2x − 3. By the Pythagorean Theorem, the
of length 1, and hypotenuse
opposite side has length 4x2 − 12x + 8.
1.9.2.19. Solution. We use the substitution x = tan u, dx = sec2 udu. Note
tan 0 = 0 and tan π4 = 1.
1 π/4
x2 tan2 u
Z Z
√ 3 dx = √ 3 sec2 udu
x2 +1 2
0 0 tan u + 1
π/4
tan2 u
Z
2
= √ 3 sec udu
0 2
sec u
π/4
tan2 u
Z
= du
0 sec u
π/4
sec2 u − 1
Z
= du
0 sec u
Z π/4

= sec u − cos u du
0
h iπ/4
= log |sec u + tan u| − sin u
0
√
 
1
= log 2 + 1 − √ − (log |1 + 0| − 0)
2
√ 1
= log(1 + 2) − √
2
1.9.2.20. Solution. There’s no square root, but we can still make use of the sub-
stitution x = tan θ, dx = sec2 θdθ.
1 1
Z Z
2 2
dx = 2
sec2 θdθ
(x + 1) (tan θ + 1)2
1
Z Z
= sec θdθ = cos2 θdθ
2
sec4 θ
1
Z


= 1 + cos(2θ) dθ
2
 
1 1
= θ + sin(2θ) + C
2 2
1
= (θ + sin θ cos θ) + C
2

827
S OLUTIONS TO E XERCISES
 
1 x
= arctan x + +C
2 x2 + 1

1
√ x2 +
x
θ
1

Since x = tan θ, we can draw a right triangle with angle θ, opposite side x, and
adjacent side 1. Then by the Pythagorean Theorem, its hypotenuse has length
√
x2 + 1, which allows us to find sin θ and cos θ.

Exercises — Stage 3
1.9.2.21. Solution. We complete the square to find x2 − 2x + 2 = (x − 1)2 + 1.

x2 x2
Z Z
√ dx = p dx
x2 − 2x + 2 (x − 1)2 + 1

We use the substitution x−1 = tan θ, which implies dx = sec2 θdθ and x = tan θ+1

(tan θ + 1)2
Z
= p sec2 θdθ
2
(tan θ) + 1
tan2 θ + 2 tan θ + 1
Z
= sec2 θdθ
sec θ
Z
= (sec2 θ + 2 tan θ) sec θdθ
Z
sec3 θ + 2 tan θ sec θ dθ


=
1 1
= sec θ tan θ + log | sec θ + tan θ| + 2 sec θ + C
2 2
1√ 2 1 √
= x − 2x + 2(x − 1) + log x2 − 2x + 2 + x − 1
2 2 √
+ 2 x2 − 2x + 2 + C
3 + x√ 2 1 √
= x − 2x + 2 + log x2 − 2x + 2 + x − 1 + C
2 2
p
To
p see why we could write (tan θ)2 + 25 = sec θ, as opposed to
(tan θ)2 + 25 = | sec θ|, see Example 1.9.5.
From our substitution, we know√ tan θ = x − 1. To find sec θ, we can notice that in
our work we already simplified x2 − 2x + 1 = sec θ.

828
S OLUTIONS TO E XERCISES

2
x+
√ x2 −2 x−1
θ
1

Alternately, we can draw a right triangle with angle θ, opposite side x − 1, adja-
cent side 1, and use the Pythagorean Theorem to find the hypotenuse.
1.9.2.22. Solution. First, we complete the square. The constants aren’t integers,
but we can still use the same method as in Question 2. The quadratic function
under the square root is 3x2 + 5x. We match the non-constant terms to those of a
perfect square.

(ax + b)2 = a2 x2 + 2abx + b2

3x2 + 5x = a2 x2 + 2abx + b2 + c for some constant c
√
• Looking at the leading term tells us a = 3.
√ 5
• Then the second term tells us 5 = 2ab = 2 3b, so b = √
2 3
.
25
• Finally, the constant terms give us 0 = b2 + c = 12
+ c, so c = − 25
12
.
√ 2
So, 3x2 + 5x = 3x + 2√5 3 − 25
12
.

1 1
Z Z
√ dx = r dx
3x2 + 5x √ 5
2
25
3x + √
2 3
− 12

√ √
We use the substitution 3x + 2√5 3 = 5
√
2 3
sec θ, which leads to 3dx =
5
√
2 3
sec θ tan θdθ, i.e. dx = 65 sec θ tan θdθ.

1 5
Z
= r · sec θ tan θdθ
5
2
25
6
√
2 3
sec θ − 12

1 5
Z
= q · sec θ tan θdθ
25
sec2 θ − 25 6
12 12
1 5
Z
= q ·
sec θ tan θdθ
25 2
tan θ 6
12
1 5
Z
= 5 · sec θ tan θdθ
√
2 3
tan θ 6
1
Z
=√ sec θdθ
3

829
S OLUTIONS TO E XERCISES

1
= √ log |sec θ + tan θ| + C
3
2√ 2
 
1 6
= √ log x+1 + 9x + 15x + C
3 5 5

+ 5
√
6x 2 9x2 + 15x
θ
5
√
Since we used the substitution 3x + 2√5 3 = 2√5 3 sec θ, we have sec θ = 65 x +
1 = 6x+55
. To find tan θ in terms of x, we have two options. We can make a
right triangle with angle θ, hypotenuse 6x + 5, and adjacent side 5, then use the
Pythagorean Theorem to find the opposite side. Or, we can look through our
√ √ √
2 3
√
work and see that 3x + 5 = 2 3 tan θ, so tan θ = 5 3x2 + 5 = 52 9x2 + 15.
2 5
√

As a check, we observe that the derivative of the answer

2√ 2
   
d 1 6
√ log x+1 + 9x + 15x + C
dx 3 5 5
6
1 5
+ 51 √18x+15
2
=√ 6
29x √ +15x
3 5 x + 1 + 5 9x2 + 15x
1 6 + 3 √9x6x+5
2 √ 2 + √9x6x+5
2
=√ √ +15x = 3 √+15x
3 (6x + 5) + 2 9x2 + 15x (6x + 5) + 2 9x2 + 15x
√
3 1
=√ =√
9x2 + 15x 3x2 + 5x
is exactly the integrand.
Remark: in applications, often the numbers involved are messier than they are in
textbooks. The ideas of this problem are similar to other problems in this section,
but it’s good practice to apply them in a slightly messy context.
1.9.2.23. Solution. We use the substitution x = tan u, dx = sec2 udu.
Z √ 3 Z √ 3
1 + x2 1 + tan2 u
dx = sec2 udu
x tan u
sec3 u
Z
= sec2 udu
tan u
(sec2 u)2
Z
= sec udu
tan u
(tan2 u + 1)2
Z
= sec udu
tan u
tan4 u + 2 tan2 u + 1
Z
= sec udu
tan u

830
S OLUTIONS TO E XERCISES

sec u
Z Z Z
3
= tan u sec udu + 2 sec u tan udu + du
tan u
For the first integral, we use the substitution w = sec u. The second is the an-
sec u
tiderivative of 2 sec u. The third we simplify as tan u
= cos1 u · cos u
sin u
= csc u . This
brings us to
Z √ 3
1 + x2
dx
x
Z  
2
= (sec u − 1) sec u tan u du + 2 sec u + log | cot u − csc u| + C
Z
= (w2 − 1)dw + 2 sec u + log | cot u − csc u| + C
1
= w3 − w + 2 sec u + log | cot u − csc u| + C
3
1
= sec3 u − sec u + 2 sec u + log | cot u − csc u| + C
3
1
= sec3 u + sec u + log | cot u − csc u| + C
3
We began with the substitution x = tan u. Then cot u = x1 . To find csc u and sec u,
we draw a right triangle with angle u, opposite side x, and adjacent side 1. The
Pythagorean Theorem gives us the hypotenuse.

2
x
√ 1+
x
u
1

So
Z √ 3
1 + x2
dx
x
√
1√ 3 √ 1 1 + x2
= 1 + x2 + 1 + x2 + log − +C
3 x x
√
1√ 1 − 1 + x2
= 1 + x2 (4 + x2 ) + log +C
3 x

1.9.2.24. Solution. The half of the ellipse to the right of the y-axis is given by
the equation
r  y 2
x = f (y) = 4 1 −
2

831
S OLUTIONS TO E XERCISES

The area we want is the twice the area between the right-hand side of the curve
and the y-axis, from y = −1 to y = 1. In other words,
Z 1 r  y 2
Area = 2 4 1− dy
−1 2

Making use of symmetry,

Z 1
r  y 2
= 16 1− dy
0 2
y
We use the substitution 2
= sin θ, 12 dy = cos θdθ. Notice sin π6 = 1
2
and sin 0 = 0.
Z π/6 q
= 16 1 − (sin θ)2 · 2 cos θdθ
0
Z π/6 √
= 32 cos2 θ cos θdθ
0
Z π/6
= 32 cos2 θdθ
0
Z π/6
= 16 (1 + cos(2θ)) dθ
0
 π/6
1
= 16 θ + sin(2θ)
2 0
√ !
π 1 3
= 16 + ·
6 2 2
8π √
= +4 3
3
Remark: we also investigated areas of ellipses in Question 1.2.3.16, Section 1.2.
1.9.2.25. Solution. Note that f (x) is an even function, nonnegative over its en-
tire domain.
(a) To find the area of R, we evaluate
Z 1/2 Z 1/2
|x| x
Area = √
4
dx = 2 √
4
dx
1−x 2 1 − x2
−1/2 0

We use the substitution u = 1 − x2 , du = −2xdx.

Z 3/4
1
=− du
1 u1/4
 3/4  3/4 !
4 3/4 4 3
=− u =− −1
3 1 3 4
r
4 4 4
= −
3 3

832
S OLUTIONS TO E XERCISES

(b) We slice the solid of rotation into circular disks of width dx and radius
|x|
√4
.
1 − x2
Z 1/2  2
|x|
Volume = π √ 4
dx
−1/2 1 − x2
Z 1/2
x2
= 2π √ dx
0 1 − x2
√ p
We use the substitution x = sin θ, dx = cos θdθ, so 1 − x2 = 1 − sin2 θ = cos θ.
Note sin 0 = 0 and sin π6 = 12 .
π/6
sin2 θ
Z
= 2π cos θdθ
0 cos θ
Z π/6
= 2π sin2 θdθ
0
Z π/6

=π 1 − cos(2θ) dθ
0
 π/6
1
= π θ − sin(2θ)
2 0
√ !
π 1 3
=π − ·
6 2 2
√
π2 3π
= −
6 4

2
1.9.2.26. Solution. If we think of ex as ex/2 , the function under the square
root suggests the substitution ex/2 = tan θ. Then 12 ex/2 dx = sec2 θdθ, so dx =
2
ex/2
sec2 θdθ = tan2 θ sec θdθ.
√
√ 2 1 + tan2 θ
Z Z
1 + ex dx = sec2 θdθ
tan θ
sec3 θ
Z
=2 dθ
tan θ
sec θ(tan2 θ + 1)
Z
=2 dθ
tan θ
Z  
sec θ
=2 sec θ tan θ + dθ
tan θ
Z


=2 sec θ tan θ + csc θ dθ

= 2 sec θ + 2 log | cot θ − csc θ| + C

√
√ 1 1 + ex
= 2 1 + ex + 2 log x/2 − +C
e ex/2

833
S OLUTIONS TO E XERCISES
√ √
= 2 1 + ex + 2 log 1 − 1 + ex − 2 log(ex/2 ) + C
√ √
= 2 1 + ex + 2 log 1 − 1 + ex − x + C

ex
√ 1+
ex/2
θ
1

1
We used the substitution ex/2 = tan θ, so cot θ = ex/2 . To find sec θ and csc θ, we
x/2
draw a right triangle with opposite side e and √ adjacent side 1. They by the
Pythagorean Theorem, the hypotenuse has √ length 1 + ex .
Remark: x
if we use the substitution u = 1 + e , then we can change the integral
2u2
Z
to du. We can integrate this using the method of partial fractions, which
u2 − 1
we’ll learn in the next section. You can explore this option in Question 1.10.4.26,
Section 1.10.
1.9.2.27. Solution.

a We can save ourselves some trouble by applying logarithm rules before we

differentiate.
1+x √
log √ = log |1 + x| − log | 1 − x2 |
1 − x2
1
= log |1 + x| − log |1 − x2 |
2
1
= log |1 + x| − log |(1 + x)(1 − x)|
2
1 1
= log |1 + x| − log |1 + x| − log |1 − x|
2 2

 
d 1+x
log √
dx 1 − x2
 
d 1 1
= log |1 + x| − log |1 + x| − log |1 − x|
dx 2 2
1 1/2 1/2
= − +
1+x 1+x 1−x
1/2 1/2
= +
1+x 1−x
1
=
1 − x2
Notice this is the integrand from our work in blue.

834
S OLUTIONS TO E XERCISES

3
1
Z
b False: 2
dx is a number, because it is the area under a finite por-
2 1−x
tion of a continuous curve. (We note that the integrand is continuous over
the interval [2, 3], although it is not continuous everywhere.) However,
 x=3
1+x
log √ is not defined, since the denominator takes the square
1 − x2 x=2
root of a negative number. So, these two expressions are not the same.

c The work in the question is not correct. The most salient problem is that
when we make the substitution x = sin θ, we restrict the possible values of
x to [−1, 1], since this is the range of the sine function. However, the original
integral had no such restriction.
How can we be sure we avoid this problem in the future? In the intro-
ductory text to Section 1.9 (before Example 1.9.1), the notes tell us that we
are allowed to write our old variable as a function of a new variable (say
x = s(u)) as long as that function is invertible to recover our original variable
x. There is one very obvious reason why invertibility is necessary: after we
antidifferentiate using our new variable u, we need to get it back in terms of
our original variable, so we need to be able to recover x. Moreover, invert-
ibility reconciles potential problems with domains: if an inverse function
u = s−1 (x) exists, then for any x, there exists a u with s(u) = x. (This
was not the case in the work for the question, because we chose x = sin θ,
but if x = 2, there is no corresponding θ. Note, however, that x = sin θ is
invertible over [−1, 1], so the work is correct if we restrict x to those values.)
Remark: in the next section, you will learn to use partial fractions to find
1 1
Z
2
dx = log |1 + x| − log |1 − x|. When −1 < x < 1, this is equivalent
1−x 2
1+x
to log √ .
1 − x2
1.9.2.28. Solution. Remember that for any value X,

X if X ≥ 0
|X| =
−X if X ≤ 0

So, |X| =6 X precisely when X < 0.

π π
(a) The
π π
range of arcsine is − 2 , 2 . So, since u = arcsin(x/a), u is in the range
− 2 , 2 . Therefore cos u ≥ 0. Since a is positive, a cos u ≥ 0, so a cos u = |a cos u|.
That is, √
a2 − x2 = |a cos u| = a cos u
all the time.
(b) The range
of arctangent is − π2 , π2 . So, since u = arctan(x/a), u is in the

range − π2 , π2 . Therefore sec u = cos1 u > 0. Since a is positive, a sec u > 0, so

a sec u = |a sec u|.That is,
√
a2 + x2 = |a sec u| = a sec u

835
S OLUTIONS TO E XERCISES

all the time.  

(c) The range of arccosine is 0, π . So, since u = arcsec(x/a) = arccos(a/x), u is in
the range 0, π . (Actually, it’s in the range [0, π2 )∪( π2 , π], since secant is undefined
 

at π/2.) If |a tan u| =
6 a tan u, then tan u < 0, which happens when u is in the range
π
2
, π). This is the same range over which −1 < cos u < 0, and so −1 < xa < 0.
Since xa < 0, a and x have different signs, so x < 0. Then since −1 < xa , also
x < −a.
So, √
x2 − a2 = |a tan u| = −a tan u 6= a tan u
happens precisely when when x < −a.

1.10 · Partial Fractions

1.10.4 · Exercises

Exercises — Stage 1
1.10.4.1. Solution. If a quadratic function can be factored as (ax + b)(cx + d) for
some constants a, b, c, d, then it has roots − ab and − dc . So, if a quadratic function
has no roots, it is irreducible: this is the case for the function in graph (d).
If a quadratic function has two different roots, then (ax + b) 6= α(cx + d) for any
constant α. That is, the quadratic function is the product of distinct linear factors.
This is the case for the functions graphed in (b) and (c), since these each have two
distinct places where they cross the x-axis.
Finally, if a quadratic function has precisely one root, then ab = dc , so:

(ax + b)(cx + d) = a(x + ab )(cx + d) = a(x + dc )(cx + d)

= ac (cx + d)(cx + d)

That is, the quadratic function is the product of a repeated linear factor, and a
constant ac (which might simply be ac = 1).
1.10.4.2. *. Solution. Our first step is to fully factor the denominator:
(x2 − 1)2 (x2 + 1) = (x − 1)2 (x + 1)2 (x2 + 1)
Once a term is linear, it can’t be factored further; for quadratic terms, we should
check that they are irreducible. Since x2 + 1 has no real roots (we are familiar
with its graph, which is entirely above the x-axis), it is irreducible, so now our
denominator is fully factored.

x3 + 3 x3 + 3
=
(x2 − 1)2 (x2 + 1) (x − 1)2 (x + 1)2 (x2 + 1)
A B C D Ex + F
= + 2
+ + 2
+ 2
x − 1 (x − 1) x + 1 (x + 1) x +1
Notice (x − 1) and (x + 1) are (repeated) linear factors, while (x2 + 1) is an ir-
reducible quadratic factor. This accounts for the difference in the numerators of
their corresponding terms.

836
S OLUTIONS TO E XERCISES

1.10.4.3. *. Solution. The partial fraction decomposition has the form

3x3 − 2x2 + 11 A
2 2
= + various terms
x (x − 1)(x + 3) x−1

When we multiply through by the original denominator, this becomes

3x3 − 2x2 + 11 = x2 (x2 + 3)A + (x − 1)(other terms).

Evaluating both sides at x = 1 yields 3 · 13 − 2 · 12 + 11 = 12 (12 + 3)A + 0, or A = 3.

1.10.4.4. Solution.

a We start by dividing. The leading term of the numerator is x times the

leading term of the denominator. The remainder is x + 2.

x
x2 + 1 x3 + 2x + 2

− x3 − x
x+2

That is, x3 + 2x + 2 = x(x2 + 1) + (x + 2). So,

x3 + 2x + 2 x+2
2
=x+ 2
x +1 x +1

b We start by dividing. The leading term of the numerator is 3x2 times the
leading term of the denominator.

3x2
5x2 + 2x + 8 15x4 + 6x3 + 34x2 + 4x + 20

− 15x4 − 6x3 − 24x2

10x2 + 4x + 20

Then 5x2 goes into 10x2 twice, so

3x2 +2
2 4 3 2


5x + 2x + 8 15x + 6x + 34x + 4x + 20
− 15x4 − 6x3 − 24x2
10x2 + 4x + 20
− 10x2 − 4x − 16
4

Our remainder is 4. That is,

15x4 + 6x3 + 34x2 + 4x + 20 4

2
= 3x2 + 2 + 2 .
5x + 2x + 8 5x + 2x + 8

837
S OLUTIONS TO E XERCISES

c We start by dividing. The leading term of the numerator is x3 times the

leading term of the denominator.

x3
2
2x + 9x3 + 12x2 + 10x + 30
5


2x + 5
− 2x5 − 5x3
4x3 + 12x2 + 10x

Then 2x2 (2x) gives us 4x3 .

x3 + 2x
2x2 + 5 2x5 + 9x3 + 12x2 + 10x + 30

− 2x5 − 5x3
4x3 + 12x2 + 10x
− 4x3 − 10x
2
12x + 30

Finally, 2x2 goes into 12x2 six times.

x3 + 2x + 6
2
2x + 9x3 + 12x2 + 10x + 30
5


2x + 5
− 2x5 − 5x3
4x3 + 12x2 + 10x
− 4x3 − 10x
2
12x + 30
− 12x2 − 30
0

Since there is no remainder,

2x5 + 9x3 + 12x2 + 10x + 30

= x3 + 2x + 6
2x2 + 5

Remark: if we wanted to be pedantic about the question statement, we

could write our final answer as x3 + 2x + 6 + x0 , so that we are indeed
adding a polynomial to a rational function whose numerator has degree
strictly smaller than its denominator.
1.10.4.5. Solution.
a The polynomial 5x3 − 3x2 − 10x + 6 has a repeated pattern: the ratio of the
first two coefficients is the same as the ratio of the last two coefficients. We
can use this to factor.

5x3 − 3x2 − 10x + 6 = x2 (5x − 3) − 2(5x − 3)

= (x2 − 2)(5x − 3)
√ √
= (x + 2)(x − 2)(5x − 3)

838
S OLUTIONS TO E XERCISES

b The polynomial x4 − 3x2 − 5 has only even powers of x, so we can (tem-

porarily) replace them with x2 = y to turn our quartic polynomial into a
quadratic.

x4 − 3x2 − 5 = y 2 − 3y − 5

There’s no obvious factoring here, but we can find its roots, if any, using the
quadratic equation.
p
3 ± 32 − 4(1)(−5)
y=
√ 2
3 ± 29
=
2
√ ! √ !
3 + 29 3 − 29
So, y 2 − 3y − 5 = y − y−
2 2
√ ! √ !
3 + 29 3 − 29
Therefore, x4 − 3x2 − 5 = x2 − x2 −
2 2

We’d like to use the difference of two squares to factor these quadratic ex-
pressions. For this to work,
√ the constants must be positive (so their square
roots are real). Since 29 > 3, only the first quadratic is factorable. The
other is irreducible — it’s always positive, so it had no roots.
√ √
 s  s 
3 + 29   3 + 29 
x4 − 3x2 − 5 = x + x−
2 2
√ !
29 − 3
x2 +
2

c Without seeing any obvious patterns, we start hunting for roots. Since we
have all integer coefficients, if there are any integer roots, they will divide
our constant term, −6. So, our candidates for roots are ±1, ±2, ±3, and ±6.
To save time, we don’t need to know exactly the value of our polynomial at
these points: only whether or not it is 0. Write f (x) = x4 −4x3 −10x2 −11x−6.

f (−1)= 0 f (−2) 6= 0 f (−3) 6= 0 f (−6) 6= 0

f (1) 6= 0 f (2) 6= 0 f (3) 6= 0 f (6)= 0

Since x = −1 and x = 6 are roots of our polynomial, it has factors (x + 1)

and (x − 6). Note (x + 1)(x − 6) = x2 − 5x − 6. We use long division to figure
out what else is lurking in our polynomial.

839
S OLUTIONS TO E XERCISES

x2 + x + 1
2
x − 4x − 10x2 − 11x − 6
4 3


x − 5x − 6
− x4 + 5x3 + 6x2
x3 − 4x2 − 11x
− x3 + 5x2 + 6x
x2 − 5x − 6
− x2 + 5x + 6
0

So, x4 − 4x3 − 10x2 − 11x − 6 = (x + 1)(x − 6)(x2 + x + 1).

We should check whether x2 + x + 1 is reducible or √ not. If we try to find
−1 ± −3
its roots with the quadratic equation, we get , which are not real
2
numbers. So, we’re at the end of our factoring.
d Without seeing any obvious patterns, we start hunting for roots. Since we
have all integer coefficients, if there are any integer roots, they will divide
our constant term, −15. So, our candidates for roots are ±1, ±3, ±5, and
±15. Write f (x) = 2x4 + 12x3 − x2 − 52x + 15.

f (−1) 6= 0 f (−3)= 0 f (−5)= 0 f (−5) 6= 0

f (1) 6= 0 f (3) 6= 0 f (5) 6= 0 f (15) 6= 0

Since x = −3 and x = −5 are roots of our polynomial, it has factors (x + 3)

and (x + 5). Note (x + 3)(x + 5) = x2 + 8x + 15. We use long division to
move forward.

2x2 − 4x + 1
x2 + 8x + 15 2x4 + 12x3 − x2 − 52x + 15

− 2x4 − 16x3 − 30x2

− 4x3 − 31x2 − 52x
4x3 + 32x2 + 60x
x2 + 8x + 15
− x2 − 8x − 15
0

So, 2x4 + 12x3 − x2 − 52x + 15 = (x + 3)(x + 5)(2x2 − 4x + 1).

We should check whether 2x2 − 4x + 1 is reducible or not. There’s not an
obvious way√to factor it, but we can use the quadratic equation. This gives
4 ± 16 − 8 √
us roots = 2 ± 2. So, we have two more linear factors.
2
Specifically:
 √ 
2x4 + 12x3 − x2 − 52x + 15 = (x + 3)(x + 5) x − (2 + 2)
 √ 
x − (2 − 2)

840
S OLUTIONS TO E XERCISES

1.10.4.6. Solution. The goal of partial fraction decomposition is to write our

integrand in a form that is easy to integrate. The antiderivative of (1) can be
easily determined with the substitution u = (ax + b). It’s less clear how to find
the antiderivative of (2).

Exercises — Stage 2
1.10.4.7. *. Solution. The integrand is a rational function, so it’s a candidate for
partial fraction. We quickly rule out any obvious substitution or integration by
parts, so we go ahead with the decomposition.
1 1
We start by expressing the integrand, i.e. the fraction x+x 2 = x(1+x) , as a linear

combination of the simpler fractions x1 and x+1

1
(which we already know how to
integrate). We will have

1 1 a b a(x + 1) + bx
2
= = + =
x+x x(1 + x) x x+1 x(1 + x)

The fraction on the left hand side is the same as the fraction on the right hand side
if and only if the numerator on the left hand side, which is 1 = 0x + 1, is equal to
the numerator on the right hand side, which is a(x + 1) + bx = (a + b)x + a. This
in turn is the case if and only of a = 1 (i.e. the constant terms are the same in the
two numerators) and a + b = 0 (i.e. the coefficients of x are the same in the two
numerators). So a = 1 and b = −1. Now we can easily evaluate the integral
2 2 2
dx dx 1  1
Z Z Z
= = dx −
1 x + x2 1 x(x + 1)
1 x x+1
h i2 3 4
= log x − log(x + 1) = log 2 − log = log
1 2 3
1.10.4.8. *. Solution. We’ll first do a partial fraction decomposition. The sneaky
way is to temporarily rename x2 to y. Then x4 + x2 = y 2 + y and

1 1 1 1
= = −
x 4 + x2 y(y + 1) y y+1

as we found in Question 7. Now we restore y to x2 .

1 1 1  1
Z Z 
4 2
dx = 2
− 2 dx = − − arctan x + C
x +x x x +1 x
1.10.4.9. *. Solution. The integrand is of the form N (x)/D(x) with D(x) already
factored and N (x) of lower degree. We immediately look for a partial fraction
decomposition:
12x + 4 A Bx + C
2
= + 2 .
(x − 3)(x + 1) x−3 x +1
Multiplying through by the denominator yields

12x + 4 = A(x2 + 1) + (Bx + C)(x − 3) (∗)

841
S OLUTIONS TO E XERCISES

Setting x = 3 we find:

36 + 4 = A(9 + 1) + 0 =⇒ 40 = 10A =⇒ A = 4

Substituting A = 4 in (∗) gives

12x + 4 = 4(x2 + 1) + (Bx + C)(x − 3)

=⇒ −4x2 + 12x = (x − 3)(Bx + C)
=⇒ (−4x)(x − 3) = (Bx + C)(x − 3)
=⇒ B = −4, C = 0

So we have found that A = 4, B = −4, and C = 0. Therefore

Z  
12x + 4 4 4x
Z
dx = − dx
(x − 3)(x2 + 1) x − 3 x2 + 1
= 4 log |x − 3| − 2 log(x2 + 1) + C

The second integral was found just by guessing an antiderivative. Alternatively,

one could use the substitution u = x2 + 1, du = 2x dx.
1.10.4.10. *. Solution. The integrand is of the form N (x)/D(x) with D(x) al-
ready factored and N (x) of lower degree. With no obvious substitution available,
we look for a partial fraction decomposition.
3x2 − 4 A Bx + C
= +
(x − 2)(x2 + 4) x−2 x2 + 4
Multiplying through by the denominator gives
3x2 − 4 = A(x2 + 4) + (Bx + C)(x − 2) (∗)
Setting x = 2 we find:
12 − 4 = A(4 + 4) + 0 =⇒ 8 = 8A =⇒ A = 1
Substituting A = 1 in (∗) gives
3x2 − 4 = (x2 + 4) + (x − 2)(Bx + C)
=⇒ 2x2 − 8 = (x − 2)(Bx + C)
=⇒ (x − 2)(2x + 4) = (x − 2)(Bx + C)
=⇒ B = 2, C = 4
Thus, we have:
3x2 − 4 1 2x + 4 1 2x 4
2
= + 2 = + 2 + 2
(x − 2)(x + 4) x−2 x +4 x−2 x +4 x +4
The first two of these are directly integrable:
4
Z
2
F (x) = log |x − 2| + log |x + 4| + dx
x2 +4

842
S OLUTIONS TO E XERCISES

(The second integral was found just by guessing an antiderivative. Alternatively,

one could use the substitution u = x2 + 4, du = 2x dx.) For the final integral, we
substitute: x = 2y, dx = 2dy, and see that:
4 1
Z Z
dx = 2 dy = 2 arctan y + D = 2 arctan(x/2) + D
x2 + 4 y2 + 1

for any constant D. All together we have:

F (x) = log |x − 2| + log |x2 + 4| + 2 arctan(x/2) + D

1.10.4.11. *. Solution. This integrand is a rational function, with no obvious
substitution. This sure looks like a partial fraction problem. Let’s go through our
protocol.
• The degree of the numerator x − 13 is one, which is strictly smaller than the
dergee of the denominator x2 − x − 6, which is two. So we don’t need long
division to pull out a polynomial.
• Next we factor the denominator.

x2 − x − 6 = (x − 3)(x + 2)

• Next we find the partial fraction decomposition of the integrand. It is of the

form
x − 13 A B
= +
(x − 3)(x + 2) x−3 x+2
To find A and B, using the sneaky method, we cross multiply by the de-
nominator.

x − 13 = A(x + 2) + B(x − 3)

Now we can find A by evaluating at x = 3

3 − 13 = A(3 + 2) + B(3 − 3) =⇒ A = −2

and find B by evaluating at x = −2.

−2 − 13 = A(−2 + 2) + B(−2 − 3) =⇒ B = 3

(Hmmm. A and B are nice round numbers. Sure looks like a rigged exam
or homework question.) Our partial fraction decomposition is
x − 13 −2 3
= +
(x − 3)(x + 2) x−3 x+2
As a check, we recombine the right hand side and make sure that it matches
the left hand side.
−2 3 −2(x + 2) + 3(x − 3) x − 13
+ = =
x−3 x+2 (x − 3)(x + 2) (x − 3)(x + 2)

843
S OLUTIONS TO E XERCISES

• Finally, we evaluate the integral.

Z  
x − 13 −2 3
Z
dx = + dx
x2 − x − 6 x−3 x+2
= −2 log |x − 3| + 3 log |x + 2| + C
1.10.4.12. *. Solution. Again, this sure looks like a partial fraction problem. So
let’s go through our protocol.
• The degree of the numerator 5x + 1 is one, which is strictly smaller than the
dergee of the denominator x2 + 5x + 6, which is two. So we do not long
divide to pull out a polynomial.
• Next we factor the denominator.

x2 + 5x + 6 = (x + 2)(x + 3)

• Next we find the partial fraction decomposition of the integrand. It is of the

form
5x + 1 A B
= +
(x + 2)(x + 3) x+2 x+3
To find A and B, using the sneaky method, we cross multiply by the de-
nominator.

5x + 1 = A(x + 3) + B(x + 2)

Now we can find A by evaluating at x = −2

−10 + 1 = A(−2 + 3) + B(−2 + 2) =⇒ A = −9

and find B by evaluating at x = −3.

−15 + 1 = A(−3 + 3) + B(−3 + 2) =⇒ B = 14

So our partial fraction decomposition is

5x + 1 −9 14
= +
(x + 2)(x + 3) x+2 x+3
As a check, we recombine the right hand side and make sure that it matches
the left hand side
−9 14 −9(x + 3) + 14(x + 2) 5x + 1
+ = =
x+2 x+3 (x + 2)(x + 3) (x + 2)(x + 3)

• Finally, we evaluate the integral

Z  
5x + 1 −9 14
Z
dx = + dx
x2 + 5x + 6 x+2 x+3
= −9 log |x + 2| + 14 log |x + 3| + C

844
S OLUTIONS TO E XERCISES

1.10.4.13. Solution. We have a rational function with no obvious substitution,

so let’s use partial fraction decomposition.

• Since the degree of the numerator is the same as the degree of the denomi-
nator, we need to pull out a polynomial.

5
2 2


x −1 5x − 3x − 1
− 5x2 +5
− 3x + 4

That is,

5x2 − 3x − 1
Z  
−3x + 4
Z
dx = 5+ 2 dx
x2 − 1 x −1
−3x + 4
Z
= 5x + dx
x2 − 1

• Again, there’s no obvious substitution for the new integrand, so we want

to use partial fraction. The denominator factors as (x − 1)(x + 1), so our
decomposition has this form:

−3x + 4 −3x + 4 A B
= = +
x2 − 1 (x − 1)(x + 1) x−1 x+1
(A + B)x + (A − B)
=
(x − 1)(x + 1)

So, (1) A + B = −3 and (2) A − B = 4.

• We solve (2) for A in terms of B, namely A = 4 + B. Plugging this into (1),

we see (4 + B) + B = −3. So, B = − 27 , and A = 12 .

• Now we can write our integral in a friendlier form and evaluate.

5x2 − 3x − 1 −3x + 4
Z Z
2
dx = = 5x + dx
x −1 x2 − 1
1/2 7/2
Z
= 5x + − dx
x−1 x+1
1 7
= 5x + log |x − 1| − log |x + 1| + C
2 2
1.10.4.14. Solution. The integrand is a rational function with no obvious sub-
stitution, so we use partial fraction decomposition.

• The degree of the numerator is the same as the degree of the denominator.
Since it’s not smaller, we need to re-write our integrand. We could do this

845
S OLUTIONS TO E XERCISES

using long division, but this case is simple enough to do more informally.

4x4 + 14x2 + 2 4x4 + x2 + 13x2 + 2

=
4x4 + x2 4x4 + x2
4x + x2 13x2 + 2
4
= 4 +
4x + x2 4x4 + x2
13x2 + 2
=1+ 4
4x + x2

• The denominator factors as x2 (4x2 + 1).

• We want to find the partial fraction decomposition of the fractional part of

our simplified integrand.

13x2 + 2 13x2 + 2 A B Cx + D
= = + +
4x4 + x2 x2 (4x2 + 1) x x2 4x2 + 1

Multiply through by the original denominator.

13x2 + 2 = Ax(4x2 + 1) + B(4x2 + 1) + (Cx + D)x2 (1)

Setting x = 0 gives us:

2= B

We use B = 2 to simplify Equation (1).

13x2 + 2 = Ax(4x2 + 1) + 2(4x2 + 1) + (Cx + D)x2

5x2 = Ax(4x2 + 1) + (Cx + D)x2
5x = A(4x2 + 1) + (Cx + D)x (2)

Again, let x = 0.

0= A

Using A = 0, simplify Equation (2).

5x = (Cx + D)x
5 = Cx + D
C= 0, D=5

• Now we can write our integral in pieces.

4x4 + 14x2 + 2 13x2 + 2

Z Z  
dx = 1+ 4 dx
4x4 + x2 4x + x2
Z  
2 5
= 1+ 2 + 2 dx
x 4x + 1

846
S OLUTIONS TO E XERCISES

2 5
Z
=x− + dx
x (2x)2 + 1

Substitute u = 2x, du = 2dx.

2 5/2
Z
=x− + du
x u2 + 1
2 5
= x − + arctan u + C
x 2
2 5
= x − + arctan(2x) + C
x 2
1.10.4.15. Solution. The integrand is a rational function with no obvious sub-
stitution, so we’ll use a partial fraction decomposition.
• Since the numerator has strictly smaller degree than the denominator, we
don’t need to start off with a long division.
• We do, however, need to factor the denominator. We can immediately pull
out x2 ; the remaining part is x2 − 2x + 1 = (x − 1)2 .
• Now we can perform our partial fraction decomposition.
x2 + 2x − 1 x2 + 2x − 1 A B C D
4 3 2
= 2 2
= + 2+ +
x − 2x + x x (x − 1) x x x − 1 (x − 1)2
Multiply both sides by the original denominator.
x2 + 2x − 1 = Ax(x−1)2 + B(x−1)2 + Cx2 (x−1) + Dx2 (1)
To be sneaky, we set x = 0, and find:
−1= B
We also set x = 1, and find:
2= D
We use B and D to simplify Equation (1).
x2 + 2x − 1 = Ax(x − 1)2 −1(x − 1)2 + Cx2 (x − 1) + 2x2
0 = Ax(x − 1)2 + Cx2 (x − 1)
= x(x − 1)[(A + C)x − A]
So, 0 = (A + C)x − A
That is, A = C = 0.
• Now we can evaluate our integral.
x2 + 2x − 1
Z  
−1 2
Z
dx = + dx
x4 − 2x3 + x2 x2 (x − 1)2
1 2
= − +C
x x−1

847
S OLUTIONS TO E XERCISES

1.10.4.16. Solution. Our integrand is a rational function with no obvious sub-

stitution, so we’ll use the method of partial fractions.
• The degree of the numerator is less than the degree of the denominator.
• We need to factor the denominator. The first two terms have the same ratio
as the last two terms.
2x3 − x2 − 8x + 4 = x2 (2x − 1) − 4(2x − 1)
= (x2 − 4)(2x − 1)
= (x − 2)(x + 2)(2x − 1)

• Now we find our partial fraction decomposition.

3x2 − 4x − 10 3x2 − 4x − 10
=
2x3 − x2 − 8x + 4 (x − 2)(x + 2)(2x − 1)
A B C
= + +
x − 2 x + 2 2x − 1
Multiply both sides by the original denominator.
3x2 − 4x − 10 = A(x + 2)(2x − 1) + B(x − 2)(2x − 1)
+ C(x − 2)(x + 2)
Distinct linear factors is the best possible scenario for the sneaky method.
First, let’s set x = 2.
3(4) − 4(2) − 10 = A(4)(3) + B(0) + C(0)
1
A= −
2
Now, let x = −2.
3(4) − 4(−2) − 10 = A(0) + B(−4)(−5) + C(0)
1
B=
2
Finally, let x = 12 .
  
3 3 5
− 2 − 10 = A(0) + B(0) + C −
4 2 2
C= 3

• Now we can evaluate our integral in its new form.

3x2 − 4x − 10
Z  
−1/2 1/2 3
Z
dx = + + dx
2x3 − x2 − 8x + 4 x − 2 x + 2 2x − 1
1 1 3
= − log |x − 2| + log |x + 2| + log |2x − 1| + C
2 2 2
1 x+2 3
= log + log |2x − 1| + C
2 x−2 2

848
S OLUTIONS TO E XERCISES

1.10.4.17. Solution. The integrand is a rational function with no obvious sub-

stitution, so we use the method of partial fractions.
• The numerator has smaller degree than the denominator.
• We need to factor the denominator. In the absence of any clues, we look for
an integer root. The constant term is 5, so the possible integer roots are ±1
and ±5. Name f (x) = 2x3 + 11x2 + 6x + 5.
f (−1) 6= 0 f (−5) = 0 f (1) 6= 0 f (5) 6= 0

So, (x + 5) is a factor of the denominator.

• We use long division to pull out the factor of (x + 5).

2x2 + x + 1
2x + 11x2 + 6x + 5
3


x+5
− 2x3 − 10x2
x2 + 6x
− x2 − 5x
x+5
−x−5
0

That is, our denominator is (x + 5)(2x2 + x + 1).

• The quadratic function 2x2 + x + 1 is irreducible: we can see this by using
the quadratic equation, and finding no real roots. So, we are ready to find
our partial fraction decomposition.
10x2 + 24x + 8 10x2 + 24x + 8
=
2x3 + 11x2 + 6x + 5 (x + 5)(2x2 + x + 1)
A Bx + C
= + 2
x + 5 2x + x + 1
Multiply through by the original denominator.
10x2 + 24x + 8 = A(2x2 + x + 1) + (Bx + C)(x + 5) (1)
Set x = −5.
10(25) − 24(5) + 8 = A(2(25) − 5 + 1) + (B(−5) + C)(0)
A= 3
Using our value of A, we simplify Equation (1).
10x2 + 24x + 8 = 3(2x2 + x + 1) + (Bx + C)(x + 5)
4x2 + 21x + 5 = (Bx + C)(x + 5)
We factor the left side. We know (x + 5) must be one of its factors.
(4x + 1)(x + 5) = (Bx + C)(x + 5)
4x + 1 = Bx + C
So, B = 4 and C = 1.

849
S OLUTIONS TO E XERCISES

• Now we can write our integral in smaller pieces.

Z 1 Z 1
10x2 + 24x + 8

3 4x + 1
3 2
dx = + dx
0 2x + 11x + 6x + 5 0 x + 5 2x2 + x + 1

The antiderivative of the left fraction is 3 log |x + 5|. For the right fraction,
we use the substitution u = 2x2 + x + 1, du = (4x + 1)dx to antidifferentiate.
1
= 3 log |x + 5| + log |2x2 + x + 1| 0


= 3 log 6 + log 4 − 3 log 5 − log 1

4 · 63
 
= log
53

Exercises — Stage 3
1.10.4.18. Solution. We follow the example in the text.

1 sin x sin x
Z Z Z Z
csc xdx = dx = 2 dx = dx
sin x sin x 1 − cos2 x
Let u = cos x, du = − sin xdx.
−1 −1
Z Z
= 2
du = du
1−u (1 + u)(1 − u)

We see an opportunity for partial fraction.

−1 A B
= +
(1 + u)(1 − u) 1+u 1−u

Multiply both sides by the original denominator.

−1 = A(1 − u) + B(1 + u)

Let u = 1.
1
−1 = 2B ⇒B=−
2
Let u = −1.
1
−1 = 2A ⇒A=−
2
We can now re-write our integral.
Z  
−1 −1/2 −1/2
Z Z
csc xdx = du = + du
(1 + u)(1 − u) 1+u 1−u
1 1
= − log |1 + u| + log |1 − u| + C
2 2

850
S OLUTIONS TO E XERCISES

1 1−u
= log +C
2 1+u
1 1 − cos x
= log +C
2 1 + cos x

Remark: Elsewhere in the text, and in many tables of integrals, the antiderivative
of cosecant is given as log | csc x − cot x|. We show that this is equivalent to our
result.
1
log | csc x − cot x| = log (csc x − cot x)2
2
1
= log csc2 x − 2 csc x cot x + cot2 x
2
1 1 2 cos x cos2 x
= log − +
2 sin2 x sin2 x sin2 x
1 1 − 2 cos x + cos2 x 1 (1 − cos x)2
= log = log
2 sin2 x 2 1 − cos2 x
1 (1 − cos x)2 1 1 − cos x
= log = log
2 (1 − cos x)(1 + cos x) 2 1 + cos x

1.10.4.19. Solution. We follow the example in the text.

1 sin x sin x
Z Z Z Z
3
csc xdx = 3 dx = 4 dx = dx
sin x sin x (1 − cos2 x)2
Let u = cos x, du = − sin xdx.
−1
Z
= du
(1 − u2 )2
1 1/2 1/2
In Question 18 we saw 1−u 2 = 1+u + 1−u , so

Z  2 Z  2
−1 1 1/2 1/2
Z
du = − du = − + du
(1 − u2 )2 1 − u2 1+u 1−u
Z  
1 1 2 1
=− + + du
4 (1 + u)2 1 − u2 (1 − u)2
Z  
1 1 1 1 1
=− + + + du
4 (1 + u)2 1 + u 1 − u (1 − u)2
 
1 1 1
=− − + log |1 + u| − log |1 − u| + +C
4 1+u 1−u
 
1 2u 1+u
=− + log +C
4 1 − u2 1−u
−u 1 1−u
= + log +C
2(1 − u2 ) 4 1+u
− cos x 1 1 − cos x
= 2 + log +C
2 sin x 4 1 + cos x

851
S OLUTIONS TO E XERCISES

Remark: In Example 1.8.23, and in many tables of integrals, the antiderivative

of csc3 x is given as − 21 cot x csc x + 12 log | csc x − cot x| + C. This is equivalent
to our result. Recall in the remark after the solution to Question 18, we saw
1
2
log 1−cos x
1+cos x
= log | csc x − cot x|.

1 1
− cot x csc x + log | csc x − cot x|
2 2
1 1 1 − cos x
= − cot x csc x + log
2 4 1 + cos x
 
1 cos x
  1 1 1 − cos x
=− + log
2 sin x sin x 4 1 + cos x
− cos x 1 1 − cos x
= 2 + log
2 sin x 4 1 + cos x
1.10.4.20. Solution. This is a rational function, and there’s no obvious substitu-
tion, so we’ll use partial fraction decomposition.
• First, we check that the numerator has strictly smaller degree than the de-
nominator, so we don’t have to use long division.
• Second, we factor the denominator. We can immediately pull out a factor
of x2 ; then we’re left with the quadratic polynomial x2 + 5x + 10. Using the
quadratic equation, we check that this has no real roots, so it is irreducible.
• Once we know the factorization of the denominator, we can set up our de-
composition.
3x3 + 15x2 + 35x + 10 3x3 + 15x2 + 35x + 10
=
x4 + 5x3 + 10x2 x2 (x2 + 5x + 10)
A B Cx + D
= + 2+ 2
x x x + 5x + 10
We multiply both sides by the original denominator.
3x3 + 15x2 + 35x + 10
= Ax(x2 + 5x + 10) + B(x2 + 5x + 10) + (Cx + D)x2 (1)
Following the “Sneaky Method,” we plug in x = 0.
0 + 10 = A(0) + B(10) + (C(0) + D)(0)
B= 1

• Knowing B allows us to simplify our Equation (1).

3x3 + 15x2 + 35x + 10 = Ax(x2 + 5x + 10) + 1(x2 + 5x + 10)
+ (Cx + D)x2
3x3 + 14x2 + 30x = Ax(x2 + 5x + 10) + (Cx + D)x2
We can factor x out of both sides of the equation.
3x2 + 14x + 30 = A(x2 + 5x + 10) + (Cx + D)x (2)

852
S OLUTIONS TO E XERCISES

• Again, we set x = 0.

0 + 30 = A(10) + (C(0 + D)(0)

A= 3

• We simplify Equation (2), using A = 3.

3x2 + 14x + 30 = 3(x2 + 5x + 10) + (Cx + D)x

−x = Cx2 + Dx
C= 0, D = −1

• Now that we have our coefficients, we can re-write our integral in a

friendlier form.
Z 2 3
3x + 15x2 + 35x + 10
dx
1 x4 + 5x + 10x2
Z 2 
3 1 1
= + − dx
1 x x2 x2 + 5x + 10
 2 Z 2
1 1
= 3 log |x| − − 2
dx
x 1 1 x + 5x + 10
Z 2
1 1
= 3 log 2 + − 2
dx
2 1 x + 5x + 10

The remaining integral is the reciprocal of a quadratic polynomial, much

1
like , whose antiderivative is arctangent. We complete the square
1 + x2  
2x+5
and use the substitution u = √
15
, du = √215 dx.

2 2
1 1
Z Z
dx = dx
2
x + 5x + 10 5 2 15


1 1 x+ 2
+ 4
2
4 1
Z
= 2 dx
15

1 2x+5
√
15
+1
√
9/ 15
2 1
Z
=√ √ 2
du
15 7/ 15 u + 1
2 h i9/√15
=√ arctan u √
15 7/ 15
    
2 9 7
=√ arctan √ −arctan √
15 15 15
So, all together,
2
3x3 + 15x2 + 35x + 10
Z
dx
1 x4 + 5x3 + 10x2

853
S OLUTIONS TO E XERCISES
  
1 2 9
= 3 log 2 + − √ arctan √
2 15 15
 
7
− arctan √
15
1.10.4.21. Solution. Our integrand is already in the nice form that would come
out of a partial fractions decomposition. Let’s consider its different pieces.

• First piece: x23+2 dx. The fraction looks somewhat like the derivative of
R

arctangent, so we can massage it to find an appropriate substitution.

3 3 1
Z Z
2
dx = 2 dx
x +2 2

√x +1
2

Use the substitution u = √x , du = √1 dx.

2 2

3 1
Z
=√ 2
du
2 u +1
3
= √ arctan u + C
2
 
3 x
= √ arctan √ +C
2 2

• The next piece is (xx−3

R
2 +2)2 dx. If the numerator were only x (and no con-

stant), we could use the substitution u = x2 + 2, du = 2xdx. So, to that

x 3
end, we can break up that fraction into (x2 +2) 2 − (x2 +2)2 . For now, we only

evaluate the first half.

x 1 1 1
Z Z
dx = du = − +C
(x2 + 2)2 2 u2 2u
1
=− 2 +C
2x + 4

3
• That leaves us with the final piece, (x2 +2) 2 , which is the hardest. We saw

something √similar in Question

√ 1.9.2.20 in Section 1.9: we can use the substi-
2
tution x = 2 tan θ, dx = 2 sec θdθ.
Z
3
Z
3 √
2 2
dx = 2 2
2 sec2 θdθ
(x + 2) (2 tan θ + 2)
3 √
Z
= 4
2 sec2 θdθ
4 sec θ
3
Z
= √ cos2 θdθ
2 2

854
S OLUTIONS TO E XERCISES

3
Z


= √ 1 + cos(2θ) dθ
4 2
 
3 1
= √ θ + sin(2θ) + C
4 2 2
3
= √ (θ + sin θ cos θ) + C
4 2
  √ !
3 x x 2
= √ arctan √ + 2 +C
4 2 2 x +2

2
√ x2 +
x
θ
√
2

From our substitution, tan θ = √x2 . So, we can draw a right triangle with
√
angle θ, opposite side x, and adjacent√side 2. Then by the Pythagorean
Theorem, the hypotenuse has length x2 + 2, and this gives us sin θ and
cos θ.

Now we have our integral.

Z  
3 x−3
+ dx
x2 + 2 (x2 + 2)2
3 x 3
Z Z Z
= 2
dx + 2 2
dx − dx
x +2 (x + 2) (x + 2)2
2
 
3 x 1
= √ arctan √ − 2
2 2 2x + 4
  √ !
3 x x 2
− √ arctan √ + 2 +C
4 2 2 x +2
 
9 x 1 3x
= √ arctan √ − − +C
4 2 2 2(x + 2) 4(x2 + 2)
2
 
9 x 2 + 3x
= √ arctan √ − +C
4 2 2 4(x2 + 2)
1.10.4.22. Solution. This is already as simplified as we can make it using partial
fraction. Indeed, this is the kind of term that could likely come out of the partial
fraction decomposition of a scarier rational function. So, we need to know how
to integrate it. Similar to the last piece we integrated in Question 21, we can use

855
S OLUTIONS TO E XERCISES

the substitution x = tan θ, dx = sec2 θdθ.

1 sec2 θ sec2 θ
Z Z Z
dx = dθ = dθ
(1 + x2 )3 (1 + tan2 θ)3 (sec2 θ)3
Z  2
1 + cos(2θ)
Z
4
= cos θdθ = dθ
2
1
Z
= (1 + cos(2θ))2 dθ
4
1
Z
1 + 2 cos(2θ) + cos2 (2θ) dθ


=
4
Z  
1 1
= 1 + 2 cos(2θ) + (1 + cos(4θ)) dθ
4 2
Z  
1 3 1
= + 2 cos(2θ) + cos(4θ)) dθ
4 2 2
 
1 3 1
= θ + sin(2θ) + sin(4θ) + C
4 2 8
3 1 1
= θ + sin(2θ) + sin(4θ) + C
8 4 32
3 1 1
= θ + sin θ cos θ + sin(2θ) cos(2θ) + C
8 2 16
3 1 1
= θ + sin θ cos θ + sin θ cos θ(cos2 θ − sin2 θ) + C
8 2 8
1 − x2
  
3 x 1 x
= arctan x + + +C
8 2(1 + x2 ) 8 1 + x2 1 + x2
3 3x3 + 5x
= arctan x + +C
8 8(1 + x2 )2

2
x
√ 1+
x
θ
1

To change our variables from θ to x, recall we used the substitution x = tan θ. So,
we draw a right triangle with angle θ, opposite side length x, and √ adjacent side
length 1. By the Pythagorean Theorem, the hypotenuse has length 1 + x2 . This
allows us to find sin θ and cos θ.
1.10.4.23. Solution. Our integrand is already as simplified as the method of
partial fractions can make it. The first term is easy to antidifferentiate. The second
term would be easier if it were broken into two pieces: one where the numerator
is a constant, and one where the numerator is a multiple of x.
Z  
3x + 1 3x
3x + 2 + dx
x + 5 (x2 + 5)2

856
S OLUTIONS TO E XERCISES
Z  
3 2 1 3x 3x
= x + + + dx
2 x2 + 5 x2 + 5 (x2 + 5)2
Z  
3 2 1 3x 3x
Z
= x + dx + + dx
2 x2 + 5 x2 + 5 (x2 + 5)2

The first integral looks similar to the derivative of arctangent. For the second
integral, we use the substitution u = x2 + 5, du = 2xdx.
Z  
3 2 1 1 3/2 3/2
Z
= x +  2 dx + + 2 du
2 5 √x
u u
5
+1

For the first integral, use the substitution w = √x , dw = √1 dx.

5 5

3 1 1 3 3
Z
= x2 + √ 2
dw + log |u| −
2 5 w +1 2 2u
3 1 3 3
= x2 + √ arctan w + log |x2 + 5| − 2 +C
2 5 2 2x + 10
 
3 1 x 3 3
= x2 + √ arctan √ + log |x2 + 5| − 2 +C
2 5 5 2 2x + 10
1.10.4.24. Solution. If our denominator were all sines, we could use the substi-
tution x = sin θ. To that end, we apply the identity cos2 θ = 1 − sin2 θ.

cos θ cos θ
Z Z
2
dθ = 2 dθ
3 sin θ + cos θ − 3 3 sin θ + 1 − sin θ − 3
cos θ
Z
= dθ
3 sin θ − sin2 θ − 2
We use the substitution x = sin θ, dx = cos θdθ.
1 −1 −1
Z Z Z
= 2
dx = 2
dx = dx
3x − x − 2 x − 3x + 2 (x − 1)(x − 2)

Now we can find a partial fraction decomposition.

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

Setting x = 1 and x = 2, we see

A= 1, B = −1

Now, we can evaluate our integral.

cos θ −1
Z Z
dθ = dx
3 sin θ + cos2 θ − 3 (x − 1)(x − 2)

857
S OLUTIONS TO E XERCISES
Z  
1 1
= − dx
x−1 x−2
x−1
= log |x − 1| − log |x − 2| + C = log +C
x−2
sin θ − 1
= log +C
sin θ − 2
1.10.4.25. Solution. This looks a lot like a rational function, but with the func-
tion et in place of the variable. So, we would like to make the substitution x = et ,
dx = et dt. Then dt = e1t dx = x1 dx.

1 1
Z Z
2t t
dt = 2
dx
e +e +1 x (x + x + 1)

The factor x2 + x + 1 is an irreducible quadratic, so the denominator is completely

factored. Now we can use partial fraction decomposition.

1 A Bx + C
= + 2
x (x2 + x + 1) x x +x+1
1 = A(x2 + x + 1) + (Bx + C)x
1 = (A + B)x2 + (A + C)x + A

The constant terms tell us A = 1; then the coefficient of x tells us C = −A = −1.

Finally, the coefficient of x2 tells us B = −A = −1. Now we can evaluate our
integral.

1 1
Z Z
dt = dx
e2t + et + 1 x (x2 + x + 1)
Z  
1 x+1
= − dx
x x2 + x + 1
 
1 x + 1/2 + 1/2
Z
= − dx (∗)
x x2 + x + 1
1 x + 1/2 1/2
Z Z Z
= dx − 2
dx − 2
dx
x x +x+1 x +x+1
1 1/2
Z
2
= log |x| − log |x + x + 1| − 2
dx
2 x +x+1
In step (∗), we set ourselves up so that we could evaluate the second integral
with the substitution u = x2 + x + 1. For the remaining integral, we complete the
square, so that the integrand looks something like the derivative of arctangent.

1 1/2
Z
2
= log |x| − log |x + x + 1| −
2 dx
2 x + 12 + 43
1 2 1
Z
2
= log |x| − log |x + x + 1| − 2 dx
2 3

2x+1
√
3
+ 1

858
S OLUTIONS TO E XERCISES

2x+1 √2 .
We use the substitution u = √ ,
3
du = 3

1 1 1
Z
2
= log |x| − log |x + x + 1| − √ du
2 3 u2 + 1
1 1
= log |x| − log |x2 + x + 1| − √ arctan u + C
2 3
 
1 2 1 2x + 1
= log |x| − log |x + x + 1| − √ arctan √ +C
2 3 3
 t 
t 1 2t t 1 2e + 1
= log |e | − log |e + e + 1| − √ arctan √ +C
2 3 3
 t 
1 2t t 1 2e + 1
= t − log |e + e + 1| − √ arctan √ +C
2 3 3
1.10.4.26. Solution.
√
• Solution 1: We use the substitution u = 1 + ex .
ex 2u
Then du = √ dx, so dx = du.
2 1 + ex u2 − 1
√ 2u 2u2
Z Z Z
x
1 + e dx = u · 2 du = du
u −1 u2 − 1
2(u2 − 1) + 2
Z  
2
Z
= du = 2+ 2 du
u2 − 1 u −1

We use a partial fraction decomposition on the fractional part of the inte-

grand.

2 2 A B
= = +
u2 −1 (u − 1)(u + 1) u−1 u+1
(A + B)u + (A − B)
=
(u − 1)(u + 1)

For the right hand side to match the left hand side, we need

A + B = 0, A − B = 2 =⇒ A = 1, B = −1

So the integral
√
Z  
2
Z
1 + ex dx = 2+ 2 du
u −1
Z  
1 1
= 2+ − du
u−1 u+1
u−1
= 2u + log |u − 1| − log |u + 1| + C = 2u + log +C
u+1
√
√ 1 + ex − 1
= 2 1 + ex + log √ +C
1 + ex + 1

859
S OLUTIONS TO E XERCISES

• Solution 2: It might not occur to us right away to use the fruitful substitu-
tion in Solution 1. More realistically, we might start with the “inside func-
1
tion,” u = 1 + ex . Then du = ex dx, so dx = u−1 du.
Z √
√ u
Z
x
1 + e dx = du
u−1
This isn’t quite a rational function, because we have a square root on top. If
we could turn it into
√ a rational function, we could use partial fraction. To
1
that end, let w = u, dw = 2√u du, so du = 2wdw.

w 2w2
Z Z
= 2wdw = dw
w2 − 1 w2 − 1
2(w2 − 1) + 2 2
Z Z
= 2
dw = 2 + 2 dw
w −1 w −1
Now we can use partial fraction decomposition.
2 2 A B
= = +
w2 −1 (w − 1)(w + 1) w−1 w+1
(A + B)w + (A − B)
=
(w − 1)(w + 1)
For the left and right hand sides to match, we need

A + B = 0, A − B = 2 =⇒ A = 1, B = −1

This allows us to antidifferentiate.

√
Z  
2
Z
x
1 + e dx = 2+ 2 dw
w −1
Z  
1 1
= 2+ − dw
w−1 w+1
= 2w + log |w − 1| − log |w + 1| + C
w−1
= 2w + log +C
w+1
√
√ u−1
= 2 u + log √ +C
u+1
√
√ 1 + ex − 1
= 2 1 + ex + log √ +C
1 + ex + 1

Remark: we also evaluated this integral using trigonometric substitution in Sec-

tion
√ 1.9, Question 1.9.2.26.
√ In that question, we found the antiderivative to be
x x
2 1 + e + 2 log 1 − 1 + e − x + C. These expressions are equivalent:
√
1 + ex − 1 √ 1
log √ x
= log 1 + ex − 1 + log √
1+e +1 1 + ex + 1

860
S OLUTIONS TO E XERCISES
√
√
  
1 1 − 1 + ex
= log 1 + ex − 1 + log √ √
1 + ex + 1 1 − 1 + ex
√
√ 1 − 1 + ex
x
= log 1 + e − 1 + log
1 − (1 + ex )
√
√ 1 − 1 + ex
= log 1 + ex − 1 + log
−ex
√ √
= log 1 + ex − 1 + log 1 − 1 + ex − log | − ex |
√
= 2 log 1 + ex − 1 − x

10
1.10.4.27. *. Solution. (a) Let’s graph y = √ . We start with the end-
25 − x2
points: (3, 52 ) and (4, 10
3
). Then we consider the first derivative:
 
d 10 10x
√ =√ 3
dx 25 − x2 25 − x2
Over the interval [3, 4], this is always positive, so our function is increasing over
the entire interval. The second derivative,
( )
d2 10(2x2 + 25)
 
10 d 10x
√ = √ 3 = √ 5 ,
dx2 25 − x2 dx 25 − x2 25 − x2
is always positive, so our function is concave up over the entire interval. So, the
region R is:

(b) Let V1 be the solid obtained by revolving R about the x-axis. The portion of V1
with x-coordinate between x and x + dx is obtained by rotating the red vertical
strip in the figure on the left below about the x-axis. That portion is a disk of
10

2
√ 10
radius √25−x 2 and thickness dx. The volume of this disk is π 25−x 2 dx. So the
total volume of V1 is
Z 4  2 Z 4
10 1
π √ dx = 100π 2
dx
3 25 − x2 3 25 − x
Z 4
1
= 100π dx
3 (5 − x)(5 + x)
Z 4
1 1 
= 10π + dx
3 5−x 5+x

861
S OLUTIONS TO E XERCISES
h i4
= 10π − log(5 − x) + log(5 + x)
3
h i
= 10π − log 1 + log 9 + log 2 − log 8
9 3
= 10π log = 20π log
4 2

(c) We’ll use horizontal washers as in Example 1.6.5.

• We cut R into thin horizontal strips of width dy as in the figure on the right
above.

• When we rotate R about the y-axis, each strip sweeps out a thin washer

◦ whose outer radius is rout = 4, and

q
◦ whose inner radius is rin = 25 − 100 y2
10
when y ≥ √25−3 10 5
2 = 4 = 2 (see

the red strip in the figure on the right above), and whose inner radius
is rin = 3 when y ≤ 25 (see the blue strip in the figure on the right
above) and
◦ whose thickness is dy and hence
2 2
)dy = π 100 5


◦ whose volume is π(rout − rin y2
− 9 dy when y ≥ 2
and whose
2
volume is π(rout 2
− rin )dy = 7π dy when y ≤ 25 and
10
• As our bottommost strip is at y = 0 and our topmost strip is at y = 3
(since
10 √ 10
at the top x = 4 and y = √25−x2 = 25−42
= 10
3
), the volume is
Z 10/3  100  Z 5/2
π − 9 dy + 7π dy
5/2 y2 0
h 100 i10/3 35
=π − − 9y + π
y 5/2 2
h 45 i 35
= π − 30 + 40 − 30 + + π
2 2
= 20π
1.10.4.28. Solution. In order to find the area between the curves, we need to
know which one is on top, and which on the bottom. Let’s start by finding where

862
S OLUTIONS TO E XERCISES

they meet.
4 2
2
=
3+x x(x + 1)
2x + 2x = 3 + x2
2

x2 + 2x − 3 = 0
(x − 1)(x + 3) = 0

In the interval [ 14 , 3], the curves only meet at x = 1. So, to find which is on top
and on bottom in the intervals [ 41 , 1) and (1, 3], it suffices to check some point in
each interval.
4 2
x 3+x2 x(x+1)
Top:
2
1/2 16/13 8/3 x(x+1)
4
2 4/7 1/3 3+x2

2
So, x(x+1) is the top function when 14 ≤ x < 1, and 3+x
4
2 is the top function when

1 < x ≤ 3. Then the area we want to find is:

Z 1  Z 3 
2 4 4 2
Area = − dx + − dx
1
4
x(x + 1) 3 + x2 1 3 + x2 x(x + 1)

We’ll need to antidifferentiate both these functions. We can antidifferentiate

2
using partial fraction decomposition.
x(x + 1)

2 A B (A + B)x + A
= + =
x(x + 1) x x+1 x(x + 1)
Matching coefficients

A + B = 0, A = 2 =⇒ A = 2, B = −2

and the integral

Z  
2 2 2
Z
dx = − dx = 2 log |x| − 2 log |x + 1| + C
x(x + 1) x x+1
x
= 2 log +C
x+1
4
We can antidifferentiate 2
using the substitution u = √x3 , du = √13 dx.
3+x
√
4 4 4 3
Z Z Z
dx =   2  dx = du
3 + x2 x 3 (1 + u2 )
3 1 + √3
 
4 4 x
= √ arctan u + C = √ arctan √ +C
3 3 3

863
S OLUTIONS TO E XERCISES

Now, we can find our area.

Z 1  Z 3 
2 4 4 2
Area = − dx + − dx
1
4
x(x + 1) 3 + x2 1 3 + x2 x(x + 1)
  1
x 4 x
= 2 log − √ arctan √
x+1 3 3 1/4
   3
4 x x
+ √ arctan √ − 2 log
3 3 x+1 1
 
1 4 π 1 4 1
= 2 log − √ · − 2 log + √ arctan √ +
2 3 6 5 3 4 3
 
4 π 3 4 π 1
√ · − 2 log − √ · + 2 log
3 3 4 3 6 2
5 4 1
= 2 log + √ arctan √
3 3 4 3
1
1.10.4.29. Solution. (a) To antidifferentiate , we use a partial fraction de-
t2 −9
composition.

1 1 A B (A + B)t + 3(A − B)
= = + =
t2−9 (t − 3)(t + 3) t−3 t+3 (t − 3)(t + 3)
1
A + B = 0, A − B =
3
1 1
A= , B = −
6Z 6 Z 
x x 
1 1/6 1/6
F (x) = 2
dx = − dx
1 t −9 1 t−3 t+3
 x
1 1
= log |t − 3| − log |t + 3|
6 6 1
 
1 1 1 1
= log |x − 3| − log |x + 3| − log 2 + log 4
6 6 6 6
 
1 x−3
= log 2 ·
6 x+3

(b) Rather than differentiate our answer from (a), we use the Fundamental Theo-
rem of Calculus Part 1 to conclude
Z x 
0 d 1 1
F (x) = 2
dt = 2
dx 1 t −9 x −9

1.11 · Numerical Integration

1.11.6 · Exercises

Exercises — Stage 1

864
S OLUTIONS TO E XERCISES

1.11.6.1. Solution. The absolute error is the difference between the two values:

|1.387 − 1.5| = 0.113

The relative error is the absolute error divided by the exact value:

0.113
≈ 0.08147
1.387
The percent error is 100 times the relative error:

≈ 8.147%
1.11.6.2. Solution. Midpoint rule:

x
2 10

Trapezoidal rule:

x
2 10

1.11.6.3. Solution.

a Differentiating, we find f 00 (x) = −x2 + 7x − 6. Since f 00 (x) is quadratic, we

have a pretty good idea of what it looks like.

• It factors as f (x) = −(x − 6)(x − 1), so its two roots are at x = 6 and
x = 1.
• The “flat part” of the parabola is at x = 3.5 (since this is exactly half
way between x = 1 and x = 6; alternately, we can check that f 000 (3.5) =
0).

865
S OLUTIONS TO E XERCISES

• Since the coefficient of x2 is negative, f (x) is increasing from −∞ to

3.5, then decreasing from 3.5 to ∞.

Therefore, over the interval [1, 6], the largest positive value of f 00 (x) occurs
when x = 3.5, and this is f 00 (3.5) = −(3.5 − 6)(3.5 − 1) = 6.25.

y
6.25

x
1 6

y = f 00 (x)

So, we take M = 6.25.

b We differentiate further to find f (4) (x) = −2. This is constant everywhere,

so we take L = | − 2| = 2.
1.11.6.4. Solution. Let’s start by differentiating.

f (x) = x sin x + 2 cos x

f 0 (x) = x cos x + sin x − 2 sin x = x cos x − sin x
f 00 (x) = −x sin x + cos x − cos x = −x sin x

For any value of x, | sin x| ≤ 1. When −3 ≤ x ≤ 2, then |x| ≤ 3. So, it is true (and
not unreasonably sloppy) that
f 00 (x) ≤ 3
whenever x is in the interval [−3, 2]. So, we can take M = 3.
Note that |f 00 (x)| is actually smaller than 3 whenever x is in the interval [−3, 2],
because when x = −3, sin x 6= 1. In fact, since 3 is pretty close to π, sin 3 is
pretty small. (The actual maximum value of |f 00 (x)| when −3 ≤ x ≤ 2 is about
1.8.) However, we find parameters like M for the purpose of computing error
bounds. There is often not much to be gained from taking the time to find the
actual maximum of a function, so we content ourselves with reasonable upper
bounds. Question 31 has a further investigation of “sloppy” bounds like this.
1.11.6.5. Solution.

a Let f (x) = cos x. Then f (4) (x) = cos x, so |f (4) (x)| ≤ 1 when −π ≤ x ≤ π.
So, using L = 1, we find the upper bound of the error using Simpson’s rule
with n = 4 is:
L(b − a)5 (2π)5 π5
= = ≈ 0.2
180n4 180 · 44 180 · 8

866
S OLUTIONS TO E XERCISES

The error bound comes from Theorem 1.11.13 in the text. We used a calcu-
lator to find the approximate decimal value.

b We use the general form of Simpson’s rule (Equation 1.11.9 in the text) with
∆x = b−a
n
= 2π4
= π2 .

∆x
A≈ (f (x0 ) + 4f (x1 ) + 2f (x2 ) + 4f (x3 ) + f (x4 ))
3
π/2
f (−π) + 4f ( −π ) + 2f (0) + 4f ( π2 ) + f (π)


= 2
3
π
= (−1 + 4(0) + 2(1) + 4(0) − 1) = 0
6

c To find the actual error in our approximation, we compare the approxima-

tion from (b) to the exact value of A. In fact, A = 0: this is a fact you’ve
probably seen before by considering the symmetry of cosine, but it’s easy
enough to calculate:
Z π
A= cos xdx = sin π − sin(−π) = 0
−π

So, our approximation was exactly the same as our exact value. The abso-
lute error is 0.

Remark: the purpose of this question was to remind you that the error bounds
we calculate are not (usually) the same as the actual error. Often our approx-
imations are better than we give them credit for. In normal circumstances, we
would be approximating an integral precisely to avoid evaluating it exactly, so
we wouldn’t find our exact error. The bound is a quick way of ensuring that our
approximation is not too far off.
1.11.6.6. Solution. Using Theorem 1.11.13 in the text, the error using the trape-
zoidal rule as described is at most
M (b − a)3 M 3 1
= ≤ = .
12 · n2 48 48 16
So, we’re really being asked to find a function with the maximum possible error
using the trapezoidal rule, given its second derivative.
With that in mind, our function should have the largest second derivative possi-
ble: let’s set f 00 (x) = 3 for every x. Then:

f 00 (x) = 3
f 0 (x) = 3x + C
3
f (x) = x2 + Cx + D
2

867
S OLUTIONS TO E XERCISES

for some constants C and D. Now we can find the exact and approximate values
Z 1
of f (x)dx.
0

1 1  
3 2
Z Z
Exact: f (x)dx = x + Cx + D dx
0 0 2
 1
1 3 C 2
= x + x + Dx
2 2 0
1 C
= + +D
Z 1 2 2 
1 1 1
Approximate: f (x)dx ≈ ∆x f (0) + f ( 2 ) + f (1)
0 2 2
  
1 1 3 C
= (D) + + +D
2 2 8 2
 
1 3
+ +C +D
2 2
 
1 9
= + C + 2D
2 8
9 C
= + +D
16 2
So, the absolute error associated with the trapezoidal approximation is:
   
1 C 9 C 1
+ +D − + +D =
2 2 16 2 16

So, for any constants C and D, f (x) = 32 x2 + Cx + D has the desired error.
Remark: contrast this question with Question 5. In this problem, our absolute
error was exactly as bad as the bound predicted, but sometimes it is much better.
The thing to remember is that, in general, we don’t know our absolute error. We
only guarantee that it’s not any worse than some worst-case-scenario bound.
a
1.11.6.7. Solution. Under any reasonable assumptions , my mother is older
than I am.

a Anyone caught trying to come up with a scenario in which I am older than my mother will
be sent to maximum security grad school.
1 1
1.11.6.8. Solution. (a) Since both expressions are positive, and 24 ≤ 12 , the
inequality is true.
(b) False. The reasoning is the same as in Question 7. The error bound given by
Theorem 1.11.13 is always better for the trapezoid rule, but this doesn’t necessar-
ily mean the error is better.
To see how the trapezoid approximation could be better than the correspond-
ing midpoint approximation in some cases, consider the function f (x) sketched

868
S OLUTIONS TO E XERCISES

below.
y

x
a b

Z b
The trapezoidal approximation of f (x)dx with n = 1 misses the thin spike,
a
and gives a mild underapproximation. By contrast, the midpoint approxima-
tion with n = 1 takes the spike as the height of the entire region, giving a vast
overapproximation.

y y

x x
a b a b
trapezoidal midpoint

1.11.6.9. *. Solution. True. Because f (x) is positive and concave up, the graph
of f (x) is always below the top edges of the trapezoids used in the trapezoidal
rule.
y
y = f (x)

1.11.6.10. Solution. According to Theorem 1.11.13 in the text, the error associ-
L (b − a)5
ated with the Simpson’s rule approximation is no more than , where
180 n4
L is a constant such that |f (4) (x)| ≤ L for all x in [a, b]. If L = 0, then the error is

869
S OLUTIONS TO E XERCISES

no more than 0 regardless of a, b, or n–that is, the approximation is exact.

Any polynomial f (x) of degree at most 3 has f (4) (x) = 0 for all x. So, any poly-
nomial of degree at most 3 is an acceptable answer. For example, f (x) = 5x3 − 27,
or f (x) = x2 .

Exercises — Stage 2
1.11.6.11. Solution.
b−a 30 − 0
• For all three approximations, ∆x = = = 5.
n 6
• For the trapezoidal rule and Simpson’s rule, the x-values where we eval-
1
uate 3 start at x = a = 0 and move up by ∆x = 5: x0 = 0, x1 = 5,
x +1
x2 = 10, x3 = 15, x4 = 20, x5 = 25, and x6 = 30.

0 5 10 15 20 25 30
x0 x1 x2 x3 x4 x5 x6

1
• For the midpoint rule, the x-values where we evaluate start at x =
+1 x3
2.5 = x0 +x
2
1
and move up by ∆x = 5: x̄1 = 2.5, x̄2 = 7.5, x̄3 = 12.5, x̄4 = 17.5,
x̄5 = 22.5, and x̄6 = 27.5.

0 2.5 5 7.5 10 12.5 15 17.5 20 22.5 25 27.5 30

x0 x̄1 x1 x̄2 x2 x̄3 x3 x̄4 x4 x̄5 x5 x̄6 x6

• Following Equation 1.11.2, the midpoint rule approximation is:

Z 30
1 h i
dx ≈ f (x̄ 1 ) + f (x̄ 2 ) + · · · + f (x̄ n ∆x
)
0 x3 + 1

1 1 1
= (2.5)13 +1 + + +
(7.5) + 1 (12.5) + 1 (17.5)3 + 1
3 3

1 1
+ + 5
(22.5)3 + 1 (27.5)3 + 1

• Following Equation 1.11.6, the trapezoidal rule approximation is:

Z 30
1
3
dx
0 x +1
h1 1 i
≈ f (x0 ) + f (x1 ) + f (x2 ) + · · · + f (xn−1 ) + f (xn ) ∆x
2 2
1/2 1 1 1 1
= 3 + 3 + 3 + 3 + 3
0 + 1 5 + 1 10 + 1 15 + 1 20 + 1

1 1/2
+ 3 + 5
25 + 1 303 + 1

870
S OLUTIONS TO E XERCISES

• Following Equation 1.11.9, the Simpson’s rule approximation is:

30
1
Z
dx
0 x3 +1
h i ∆x
≈ f (x0 )+4f (x1 )+2f (x2 )+4f (x3 )+2f (x4 )+4f (x5 )+f (x6 )
 3
1 4 2 4 2
= 3 + + + +
0 + 1 53 + 1 103 + 1 153 + 1 203 + 1

4 1 5
+ 3 + 3
25 + 1 30 + 1 3
1.11.6.12. *. Solution. By Equation 1.11.2, the midpoint rule approximation to
Rb
a
f (x)dx with n = 3 is
Z b  
f (x) dx ≈ f (x̄1 ) + f (x̄2 ) + f (x̄3 ) ∆x
a

b−a
where ∆x = 3
and

x0 = a x1 = a + ∆x x2 = a + 2∆x x3 = b
x̄1 = x0 +x
2
1
x̄2 = x1 +x
2
2
x̄3 = x2 +x
2
3

π
For this problem, a = 0, b = π and f (x) = sin x, so that ∆x = 3
and
π 2π
x0 = 0 x1 = 3
x2 = 3
x3 = π
x̄1 = π
6
x̄2 = π
2
x̄3 = 5π
6

0 π/6 π/3 π/2 2π/3 5π/6 π

x0 x̄1 x1 x̄2 x2 x̄3 x3

Therefore,
π    
π π 5π π 1 1 π 2π
Z
sin xdx ≈ sin + sin + sin = +1+ =
0 6 2 6 3 2 2 3 3
1.11.6.13. *. Solution. Let f (x) denote the diameter at height x. As in Example
1.6.6, we slice V into thin horizontal “pancakes”, which in this case are circular.

871
S OLUTIONS TO E XERCISES

dx

f (x)

• We are told that the pancake at height x is a circular disk of diameter f (x)
and so

2
• has cross-sectional area π f (x)
2
and thickness dx and hence

2
• has volume π f (x)
2
dx.

Hence the volume of V is

f (x) i2
Z 40 h
π dx
0 2
π h1 1 i
≈ 10 f (0)2 + f (10)2 + f (20)2 + f (30)2 + f (40)2
4 2 2
π h1 2 1 i
= 10 24 + 162 + 102 + 62 + 42
4 2 2
= 688 × 2.5π = 1720π ≈ 5403.5

where we have approximated the integral using the trapezoidal rule with ∆x =
10, and used a calculator to get a decimal approximation.
1.11.6.14. *. Solution. Let f (x) be the diameter a distance x from the left end of
the log. If we slice our log into thin disks, the disks x metres from the left end of
the log has
f (x)
• radius 2
,

• width dx, and so

 2
• volume π f (x)
2
dx = π4 f (x)2 dx.

872
S OLUTIONS TO E XERCISES

dx

f (x)

Using Simpson’s Rule with ∆x = 1, the volume of the log is:

6
π
Z
V = f (x)2 dx
0 4
π 1h i
≈ f (0)2 + 4f (1)2 + 2f (2)2 + 4f (3)2 + 2f (4)2 + 4f (5)2 + f (6)2
4 3h
π i
= 1.22 + 4(1)2 + 2(0.8)2 + 4(0.8)2 + 2(1)2 + 4(1)2 + 1.22
12
π
= (16.72)
12
≈ 4.377 m3

where we used a calculator to approximate the decimal value.

1.11.6.15. *. Solution. At height x metres, let the circumference of the tree be
c(x)
c(x). The corresponding radius is , so the corresponding cross–sectional area
2 2π
c(x)2

c(x)
is π = .
2π 4π

873
S OLUTIONS TO E XERCISES

dx

c(x)

The height of a very thin cross–sectional disk is dx, so the volume of a cross-
c(x)2
sectional disk is dx. Therefore, total volume of the tree is:
4π
Z 8
c(x)2 1 2h 2 2 2 2 2
i
dx ≈ c(0) + 4c(2) + 2c(4) + 4c(6) + c(8)
0 4π 4π 3
1h 2 i
= 1.2 + 4(1.1)2 + 2(1.3)2 + 4(0.9)2 + 0.22
6π
12.94
= ≈ 0.6865
6π
where we used Simpson’s rule with ∆x = 2 and n = 4 to approximate the value
of the integral based on the values of c(x) given in the table.
1.11.6.16. *. Solution. For both approximations, ∆x = 10 and n = 6.
(a) The Trapezoidal Rule gives
Z 60
V = A(h) dh
0
h1 1 i
≈ 10 A(0) + A(10) + A(20) + A(30) + A(40) + A(50) + A(60)
2 2
= 363,500

874
S OLUTIONS TO E XERCISES

(b) Simpson’s Rule gives

Z 60
V = A(h) dh
0
10 h i
≈ A(0)+4A(10)+2A(20)+4A(30)+2A(40)+4A(50)+A(60)
3
= 367,000
1.11.6.17. *. Solution. Call the curve in the graph y = f (x). It looks like

f (2) = 3 f (3) = 8 f (4) = 7 f (5) = 6 f (6) = 4

R6
We’re estimating 2 f (x)dx with n = 4, so ∆x = 6−2
4
= 1.
(a) The trapezoidal rule gives
 
3 4 49
T4 = +8+7+6+ ×1=
2 2 2

(b) Simpson’s rule gives

1  77
S4 = 3+4×8+2×7+4×6+4 ×1=
3 3
1.11.6.18. *. Solution. Let f (x) = sin(x2 ). Then f 0 (x) = 2x cos(x2 ) and

f 00 (x) = 2 cos(x2 ) − 4x2 sin(x2 ).

Since |x2 | ≤ 1 when |x| ≤ 1, and |sin θ| ≤ 1 and |cos θ| ≤ 1 for all θ, we have

2 cos(x2 ) − 4x2 sin(x2 ) ≤ 2| cos(x2 )| + 4x2 | sin(x2 )|

≤2×1+4×1×1=2+4=6

We can therefore choose M = 6, and it follows that the error is at most

M [b − a]3 6 · [1 − (−1)]3 2
2
≤ 2
= 6 = 2 · 10−6
24n 24 · 1000 10
1.11.6.19. *. Solution. Setting f (x) = 2x4 and b − a = 1 − (−2) = 3, we compute
f 00 (x) = 24x2 . The largest value of 24x2 on the interval [−2, 1] occurs at x = −2,
so we can take M = 24 · (−2)2 = 96. Thus the total error for the midpoint rule
with n = 60 points is bounded by
M (b − a)3 96 × 33 3
= =
24n2 24 × 60 × 60 100
3
That is: we are guaranteed our absolute error is certainly no more a than 100 ,
and using the bound stated in the problem we cannot give a better guarantee.
(The second part of the previous sentence comes from the fact that we used the
smallest possible M : if we had used a larger value of M , we would still have
5
some true statement about the error, for example “the error is no more than 100 ,”

875
S OLUTIONS TO E XERCISES

but it would not be the best true statement we could make.)

a This is what the error bound always tells us.

1.11.6.20. *. Solution. (a) Since a = 0, b = 2 and n = 6, we have ∆x = b−a n

=
2−0 1 1 2 4 5
6
= 3 , and so x0 = 0, x1 = 3 , x2 = 3 , x3 = 1, x4 = 3 , x5 = 3 , and x6 = 2. Since
Simpson’s Rule with n = 6 in general is

∆x  
f (x0 ) + 4f (x1 ) + 2f (x2 ) + 4f (x3 ) + 2f (x4 ) + 4f (x5 ) + f (x6 ) ,
3
the desired approximation is

1/3 5
1 5 2 5
5
4 5
(−3) + 4 −3 +2 − 3 + 4(−2) + 2 −3
3 3 3 3
5 5 
5
+4 − 3 + (−1)
3

(b) Here f (x) = (x − 3)5 , which has derivatives

f 0 (x) = 5(x − 3)4 f 00 (x) = 20(x − 3)3

f (3) (x) = 60(x − 3)2 f (4) (x) = 120(x − 3).

For 0 ≤ x ≤ 2, (x − 3) runs from −3 to −1, so the maximum absolute values

are found at x = 0, giving M = 20 · |0 − 3|3 = 540 and L = 120 · |0 − 3| = 360.
Consequently, for the Midpoint Rule with n = 100,

M (b − a)3 540 × 23 180

|EM | ≤ 2
= 4
= 4;
24n 24 × 10 10
whereas for Simpson’s Rule with n = 10,

360 × 25 64
|ES | ≤ 4
= 4.
180 × 10 10
Since 64 < 180, Simpson’s Rule results in a smaller error bound.
Rb
1.11.6.21. *. Solution. In general the error in approximating a f (x)dx using
Simpson’s rule with n steps is bounded by L(b−a) 180
(∆x)4 where ∆x = b−a n
and
(4) 1
L ≥ |f (x)| for all a ≤ x ≤ b. In this case, a = 1, b = 5, n = 4 and f (x) = x . We
need to find L, so we differentiate.

1 2 6 24
f 0 (x) = − f 00 (x) = f (3) (x) = − f (4) (x) =
x2 x3 x4 x5
and
f (4) (x) ≤ 24 for all x ≥ 1

876
S OLUTIONS TO E XERCISES

5−1
So we may take L = 24 and ∆x = 4
= 1, which leads to

24(5 − 1) 4 24 8
|Error | ≤ (1) = =
180 45 15
Rb
1.11.6.22. *. Solution. In general, the error in approximating a f (x)dx using
L(b − a) b−a
Simpson’s rule with n steps is bounded by (∆x)4 where ∆x =
180 n
and L ≥ |f (4) (x)| for all a ≤ x ≤ b. In this case, a = 0, b = 1, n = 6 and
f (x) = e−2x + 3x3 . We need to find L, so we differentiate.

f 0 (x) = −2e−2x + 9x2 f 00 (x) = 4e−2x + 18x

f (3) (x) = −8e−2x + 18 f (4) (x) = 16e−2x

1
Since e−2x = , we see f (4) (x) is a positive, decreasing function. So, its max-
e2x
imum occurs when x is as small as possible. In the interval [0, 1], that means
x = 0.

f (4) (x) ≤ f (0) = 16 for all x ≥ 0

1−0
So, we take L = 16 and ∆x = 6
= 16 .

L(b − a) 16(1 − 0) 16
|Error| ≤ (∆x)4 = (1/6)4 =
180 180 180 × 64
1 1
= =
180 × 34 14580
1
1.11.6.23. *. Solution. For both approximations, a = 1, b = 2, n = 4, f (x) = x
and ∆x = b−a n
= 14 .
Then x0 = 1, x1 = 54 , x2 = 23 , x3 = 74 , and x4 = 2.

1 5/4 3/2 7/4 2

x0 x1 x2 x3 x4

(a)
 
1 1
T4 = ∆x f (x0 ) + f (x1 ) + f (x2 ) + f (x3 ) + f (x4 )
2 2
 
1 1
= ∆x f (1) + f (5/4) + f (3/2) + f (7/4) + f (2)
2 2
   
1 1 4 2 4 1 1
= ×1 + + + + ×
4 2 5 3 7 2 2

(b)
∆x  
S4 = f (x0 ) + 4f (x1 ) + 2f (x2 ) + 4f (x3 ) + f (x4 )
3

877
S OLUTIONS TO E XERCISES

∆x  
= f (1) + 4f (5/4) + 2f (3/2) + 4f (7/4) + f (2)
3       
1 4 2 4 1
= 1+ 4× + 2× + 4× +
12 5 3 7 2
1
(c) In this case, a = 1, b = 2, n = 4 and f (x) = x
. We need to find L, so we
differentiate.
1 2 6 24
f 0 (x) = − f 00 (x) = f (3) (x) = − f (4) (x) =
x2 x3 x4 x5
So,

f (4) (x) ≤ 24 for all x in the interval [1, 2]

We take L = 24.
L(b − a)5 24(2 − 1)5 24 3
|Error | ≤ 4
≤ 4
= 4
=
180 × n 180 × 4 180 × 4 5760
1.11.6.24. *. Solution. Set a = 0 and b = 8. Since we have information about
s(x) when x is 0, 2, 4, 6, and 8, we set ∆x = b−a
n
= 2, so n = 4. (Recall with the
trapezoid rule and Simpson’s rule, n = 4 intervals actually uses the value of the
function at 5 points.)
We could perform the trapezoidal approximations with fewer intervals, for ex-
ample n = 2, but this would involve ignoring some of the points we’re given.
Since the question asks for the best estimation we can give, we use n = 4 inter-
vals and no fewer.

a
 
1 1
T4 = ∆x s(0) + s(2) + s(4) + s(6) + s(8)
2 2
 
1.00664 1.00233
=2 + 1.00543 + 1.00435 + 1.00331 +
2 2
= 8.03515
∆x  
S4 = s(0) + 4s(2) + 2s(4) + 4s(6) + s(8)
3
2
= 1.00664 + 4 × 1.00543 + 2 × 1.00435 + 4 × 1.00331
3 
+ 1.00233
≈ 8.03509

k 2
b The information s(k) (x) ≤ , with k = 2, tells us |s00 (x)| ≤ 1000 for all x
1000
2
in the interval [0, 8]. So, we take K2 (also called M in your text) to be 1000 .
Then the absolute error associated with our trapezoid rule approximation

878
S OLUTIONS TO E XERCISES

is at most
b
K2 (b − a)3 2 83
Z
f (x) dx − Tn ≤ ≤ · ≤ 0.00533
a 12n2 1000 12(4)2

4
For k = 4, we see |s(4) (x)| ≤ 1000 for all x in the interval [0, 8]. So, we take
4
K4 (also called L in your text) to be 1000 .
Then the absolute error associated with our Simpson’s rule approximation
is at most
Z b
K4 (b − a)5 4 85
f (x) dx − Sn ≤ ≤ · ≤ 0.00284
a 180n4 1000 180(4)4
1.11.6.25. *. Solution. In this case, a = 1, b = 4. Since −2 ≤ f 00 (x) ≤ 0 over the
relevant interval, we take M = 2. (Remember M is an upper bound on |f 00 (x)|,
not f 00 (x).) So we need n to obey

2(4 − 1)3 2 2(3)3 27000 9000

2
≤ 0.001 ⇐⇒ n ≥ 1000 = = = 4500
12n 12 6 2
√
One obvious allowed n is 100. Since 4500 ≈ 67.01, and n has to be a whole
number, any n ≥ 68 works.

Exercises — Stage 3
1.11.6.26. *. Solution. Denote by f (x) the width of the pool x feet from the left-
hand end. From the sketch, f (0) = 0, f (2) = 10, f (4) = 12, f (6) = 10, f (8) = 8,
f (10) = 6, f (12) = 8, f (14) = 10 and f (16) = 0.
A cross-section of the pool x feet from the left end is half of a circular disk with
diameter f (x) (so, radius f (x)
2
) and thickness dx. So, the volume of the part of the
 2
1 f (x)
pool with x–coordinate running from x to x + dx is 2 π 2 dx = π8 [f (x)]2 dx.
The total volume is given by the following integral.

π 16
Z
V = f (x)2 dx
8 0
π ∆x h
≈ · f (0)2 + 4f (2)2 + 2f (4)2 + 4f (6)2 + 2f (8)2 + 4f (10)2
8 3 i
+ 2f (12)2 + 4f (14)2 + f (16)2
π 2h
= · 0 + 4(10)2 + 2(12)2 + 4(10)2 + 2(8)2 + 4(6)2 + 2(8)2
8 3 i
+ 4(10)2 + 0
472
= π ≈ 494 ft3
3

879
S OLUTIONS TO E XERCISES

1.11.6.27. *. Solution. (a) The Trapezoidal Rule with n = 4, a = 0, b = 1, and

∆x = 41 gives:
Z 1
−6
W = 2π10 rg(r) dr
0
 
−6 1 1
≈ 2π10 ∆x x0 g(x0 ) + x1 g(x1 ) + x2 g(x2 ) + x3 g(x3 ) + x4 g(x4 )
2 2
−6 1 1 1 1 1 1 3 3 1
h       i
= 2π10 0g(0) + g + g + g + g(1)
4 2 4 4 2 2 4 4 2
−6 1 8100 8144 3 · 8170 8190
h i
= π10 + + +
2 4 2 4 2
32639π
= ≈ 0.025635
4 · 106
(b) Using the product rule, the integrand f (r) = 2π10−6 rg(r) obeys

d
f 00 (r) = 2π10−6 g(r) + rg 0 (r) = 2π10−6 2g 0 (r) + rg 00 (r)
  
dr
and hence, for 0 ≤ r ≤ 1,

f 00 (r) ≤ 2π10−6 2 × 200 + 1 × 150 = 1.1π10−3

 

So,

1.1π10−3 (1 − 0)3
|Error| ≤ ≤ 1.8 × 10−5
12(4)2
1.11.6.28. *. Solution. (a) Let f (x) = x1 , a = 1, b = 2 and ∆x = b−a 6
= 16 . Using
Simpson’s rule:
Z 2
1 ∆x h 7 8 9  10 
dx ≈ f (1) + 4f + 2f + 4f + 2f
1 x 3 6 6 6 6
 11  i
+ 4f + f (2)
6
1 h 24 12 24 12 24 1 i
= 1+ + + + + + ≈ 0.6931698
18 7 8 9 10 11 2
(b) The integrand is f (x) = x1 . The first four derivatives of f (x) are:
1 2 6 24
f 0 (x) = −2
, f 00 (x) = 3 , f (3) (x) = − 4 , f (4) (x) = 5
x x x x
On the interval 1 ≤ x ≤ 2, the fourth derivative is never bigger in magnitude
than L = 24.
L(b − a)5 24(2 − 1)5 4
|En | ≤ 4
= 4
=
180n 180n 30n4
So, we want an even number n such that
4 1
≤ 0.00001 =
30n4 105

880
S OLUTIONS TO E XERCISES

40000
n4 ≥
r3
4 40000
n≥ ≈ 10.7
3
So, any even number greater than or equal to 12 will do.
1.11.6.29. *. Solution. (a) From the figure, we see that the magnitude of |f 0000 (x)|
never exceeds 310 for 0 ≤ x ≤ 2. So, the absolute error is bounded by

310(2 − 0)5
≤ 0.01345
180 × 84
(b) We want to choose n such that:

310(2 − 0)5
4
≤ 10−4
180 × n
310 × 25 4
n4 ≥ 10
180
r
4 310 × 32
n ≥ 10 ≈ 27.2
180
For Simpson’s rule, n must be even, so any even integer obeying n ≥ 28 will
guarantee us the requisite accuracy.
Z x √
1.11.6.30. *. Solution. Let g(x) = sin( t) dt. By the Fundamental Theorem
√ 0
of Calculus Part 1, g 0 (x) = sin( x). By its definition, f (x) = g(x2 ), so we use the
chain rule to differentiate f (x).

f 0 (x) = 2xg 0 (x2 ) = 2x sin x f 00 (x) = 2 sin x + 2x cos x

Since | sin x|, | cos x| ≤ 1, we have |f 00 (x)| ≤ 2 + 2|x| and, for 0 ≤ t ≤ 1, |f 00 (t)| ≤ 4.
When the trapezoidal rule with n subintervals is applied, the resulting error En
obeys

4(1 − 0)3 1
En ≤ 2
= 2
12n 3n
We want an integer n such that

1
≤ 0.000005
3n2
4
n2 ≥
12 × 0.000005
r
1
n≥ ≈ 258.2
3 × 0.000005
Any integer n ≥ 259 will do.

881
S OLUTIONS TO E XERCISES

1.11.6.31. Solution.
x2 1
a When 0 ≤ x ≤ 1, then x2 ≤ 1 and x + 1 ≥ 1, so |f 00 (x)| = ≤ = 1.
|x + 1| 1
b To find the maximum value of a function over a closed interval, we test the
function’s values at the endpoints of the interval and at its critical points
inside the interval. The critical points are where the function’s derivative is
zero or does not exist.
2 2
The function we’re trying to maximize is |f 00 (x)| = |x+1|
x
= x+1 x
= f 00 (x)
(since our interval only contains nonnegative numbers). So, the critical
points occur when f 000 (x) = 0 or does not exist. We find f 000 (x) Using the
quotient rule.

(x + 1)(2x) − x2 x2 + 2x
f 000 (x) = =
(x + 1)2 (x + 1)2
x(x + 2)
0=
x+1
0 = x or x = −1 or x = −2

The only critical point in [0, 1] is x = 0. So, the extrema of f 00 (x) over [0, 1]
will occur at its endpoints. Indeed, since f 000 (x) ≥ 0 for all x in [0, 1], f 00 (x)
is increasing over this interval, so its maximum occurs at x = 1. That is,

1
|f 00 (x)| ≤ f 00 (1) =
2

M (b − a)3
c The absolute error using the midpoint rule is at most . Using
24n2
M = 1, if we want this to be no more than 10−5 , we find an acceptable
value of n with the following calculation:

M (b − a)3
≤ 10−5
24n2
1
≤ 10−5 (b − a = 1, M = 1)
24n2
105
≤ n2
24
n ≥ 65

M (b − a)3
d The absolute error using the midpoint rule is at most . Using
24n2
1
M = 2 , if we want this to be no more than 10−5 , we find an acceptable
value of n with the following calculation:

M (b − a)3
2
≤ 10−5
24n

882
S OLUTIONS TO E XERCISES

1 1
≤ 10−5 (b − a = 1, M = )
48n2 2
105
≤ n2
48
n ≥ 46

Remark: how accurate you want to be in these calculations depends a lot on

your circumstances. Imagine, for instance, that you were finding M by hand,
using this to find n by hand, then programming a computer to evaluate the ap-
proximation. For a simple integral like this, the difference between computing
time for 65 intervals versus 46 is likely to be miniscule. So, there’s not much to
be gained by the extra work in (b). However, if your original sloppy M gave you
something like n = 1000000, you might want to put some time into improving it,
to shorten computation time. Moreover, if you were finding the approximation
by hand, the difference between adding 46 terms and adding 65 terms would be
considerable, and you would probably want to put in the effort up front to find
the most accurate M possible.
1.11.6.32.
Z x Solution. Before we can take our Simpson’s rule approximation of
1
dt, we need to know how many intervals to use. That means we need to
1 t
d4
1
bound our error, which means we need to bound dt 4 t
.
d2 1
   
d 1 1 2
=− 2 =
dt t t dt2 t t3
d3 1 d4 1
   
6 24
3
= − 4 4
= 5
dt t t dt t t
d4 1
 
So, over the interval [1, 3], ≤ 24.
dt4 t
Now, we can find an appropriate n to ensure our error will be be less than 0.1 for
any x in [1, 3]:
L(b − a)5
< 0.1
180n4
24(x − 1)5 1
<
180n4 10
24 · (x − 1)5
n4 >
18
24 · 25 24 · (x − 1)5
Because x − 1 ≤ 2 for every x in [1, 3], if n4 > , then n4 > for
18 18
every allowed x.
24 · 25 128
n4 > =
r18 3
4 128
n> ≈ 2.6
3

883
S OLUTIONS TO E XERCISES

Since n must be even, n = 4 is enough intervals to guarantee our error is not

too high for any x in [1, 3]. Now we find our Simpson’s rule approximation with
x−1
n = 4, a = 1, b = x, and ∆x = . The points where we evaluate 1t are:
4
x−1 x−1
x0 =1 x1 = 1 + x2 = 1 + 2 ·
4 4
x+3 x+1
= =
4 2
x−1 x−1
x3 = 1 + 3 · x4 = 1 + 4 ·
4 4
3x + 1
= =x
4

1 x+3 x+1 3x+1 x

4 2 4 x3
x0 x1 x2 x3

x  
1 ∆x 1 4 2 4 1
Z
log x = dt ≈ + + + +
1 t 3 x0 x1 x 2 x3 x4
 
x−1 16 4 16 1
= 1+ + + +
12 x + 3 x + 1 3x + 1 x
= f (x)
Below is a graph of our approximation f (x) and natural logarithm on the same
axes. The natural logarithm function is shown red and dashed, while our approx-
imating function is solid blue. Our approximation appears to be quite accurate
for small, positive values of x.

3
y = f (x)

y = log x
2

x
2 4 6 8 10

884
S OLUTIONS TO E XERCISES

1.11.6.33. Solution. First, we want a strategy for approximating arctan 2. Our

1
hints are that involves integrating , which is the antiderivative of arctan-
1 + x2
π
gent, and the number , which is the same as arctan(1). With that in mind:
4
Z 2
1 π
2
dx = arctan(2) − arctan(1) = arctan(2) −
1 1+x 4
Z 2
π 1
So, arctan(2) = + 2
dx (∗)
4 1 1+x

We won’t know the value of the integral exactly, but we’ll have an approximation
A bounded by some positive error bound ε. Then,
Z 2 
1
−ε ≤ 2
dx − A ≤ ε
1 1+x
Z 2 
1
A−ε≤ 2
dx ≤ A + ε
1 1+x
π π
So, from (∗), + A − ε ≤ arctan(2) ≤ + A + ε
4 4
Which approximation should we use? We’re given the fourth derivative of
1
, which is the derivative we need for Simpson’s rule. Simpson’s rule is
1 + x2
also usually quite efficient, and we’re very interested in not adding up dozens of
terms, so we choose Simpson’s rule.
Now that we’ve chosen Simpson’s rule, we should decide how many intervals
to use. In order to bound our error, we need to find a bound for the fourth
derivative. To that end, define N (x) = 24(5x4 −10x2 +1). Then N 0 (x) = 24(20x3 −
20x) = 480x(x2 − 1), which is positive over the interval [1, 2]. So, N (x) ≤ N (2) =
24(5 · 24 − 10 · 22 + 1) = 984 when 1 ≤ x ≤ 2. Furthermore, let D(x) = (x2 + 1)5 .
If 1 ≤ x ≤ 2, then D(x) ≥ 25 . Now we can find a reasonable value of L:

25(5x4 − 10x2 + 1) N (x) 984 123

|f (4) (x)| = 2 5
= ≤ 5 = = 30.75
(x + 1) D(x) 2 4

h πL = 30.75. π
So, we take i hπ π i
We want + A − ε, + A + ε to look something like + 0.321, + 0.323 .
4 4 4 4
Note ε is half the length of the first interval. Half the length of the second interval
1
is 0.001 = 1000 . So, we want a value of ε that is no larger than this. Now we can
find our n:
L(b − a)5 1
4
≤
180 · n 1000
30.75 1
4
≤
180 · n 1000
30.75 × 1000
n4 ≥
180

885
S OLUTIONS TO E XERCISES
r
4 30750
n≥ ≈ 3.62
180
So, we choose n = 4), and are guaranteed that the absolute error in our approxi-
30.75
mation will be no more than < 0.00067.
180 · 44
b−a 1
Since n = 4, then ∆x = = , so:
n 4
5 3 7
x0 = 1 x1 = x2 = x3 = x4 = 2
4 2 4
Now we can find our Simpson’s rule approximation A:
1
1 ∆x 
Z

dx ≈ f (x 0 ) + 4f (x 1 ) + 2f (x 2 ) + 4f (x 3 ) + f (x 4 )
0 1 + x2 3
1/4  
= f (1) + 4f (5/4) + 2f (3/2) + 4f (7/4) + f (2)
3 
1 1 4 2 4 1
= + + + +
12 1 + 1 25/16 + 1 9/4 + 1 49/16 + 1 4 + 1
 
1 1 4 · 16 2·4 4 · 16 1
= + + + +
12 2 25 + 16 9 + 4 49 + 16 5
 
1 1 64 8 64 1
= + + + +
12 2 41 13 65 5
≈ 0.321748 = A

As we saw before, the error associated with this approximation is at most

30.75
< 0.00067 = ε. So,
180 · 44
Z 2
1
A−ε≤ 2
dx ≤ A + ε
1 1+x
Z 2
1
⇒ 0.321748 − 0.00067 ≤ 2
dx ≤ 0.321748 + 0.00067
1 1+x
Z 2
1
⇒ 0.321078 ≤ 2
dx ≤ 0.322418
1 1+x
Z 2
1
⇒ 0.321 ≤ 2
dx ≤ 0.323
1 1+x
Z 2
π 1 π π
⇒ + 0.321 ≤ 2
dx + ≤ + 0.323
4 1 1+x 4 4
π π
⇒ + 0.321 ≤ arctan(2) ≤ + 0.323;
4 4
This is precisely what we wanted to show.

1.12 · Improper Integrals

886
S OLUTIONS TO E XERCISES

1.12.4 · Exercises

Exercises — Stage 1
1.12.4.1. Solution. If b = ±∞, then our integral is improper because one limit
is not a real number.
Furthermore, our integral will be improper if its domain of integration contains
either of its infinite discontinuities, x = 1 and x = −1. Since one limit of integra-
tion is 0, the integral is improper if b ≥ 1 or if b ≤ −1.
Below, we’ve graphed x21−1 to make it clearer why values of b in (−1, 1) are the
only values that don’t result in an improper integral when the other limit of inte-
gration is a = 0.

1
y= x2 −1
x
−1 1

1.12.4.2. Solution. Since the integrand is continuous for all real x, the only kind
of impropriety available to us is to set b = ±∞.
1.12.4.3. Solution. For large values of x, |red function| ≤ (blue function) and
0 ≤ (blue function). If the blue function’s integral converged, then the red func-
tion’s integral would as well (by the comparison test, Theorem 1.12.17 in the text).
Since one integral converges and the other diverges, the blue function is g(x) and
the red function is f (x).
1.12.4.4. *. Solution. False. The inequality goes the “wrong way” for Theorem
1.12.17: the area under the curve f (x) is finite, but the area under g(x) could be
much larger, even infinitely larger. Z ∞
For example, if f (x) = e−x and g(x) = 1, then 0 ≤ f (x) ≤ g(x) and f (x) dx
Z ∞ 1

converges, but g(x) dx diverges.

1

1.12.4.5. Solution.

a Not enough information to decide. For example, consider h(x) = 0 versus

Z ∞
h(x) = −1. In both cases, h(x) ≤ f (x). However, 0dx converges to 0,
0

887
S OLUTIONS TO E XERCISES
Z ∞
while −1dx diverges.
0
Note:R if we had also specified 0 ≤ h(x), then we would be able to conclude
∞
that 0 h(x)dx converges by the comparison test.

b Not enough information to decide. For example, consider h(x) = f (x) ver-
sus h(x) = g(x). In both cases, f (x) ≤ h(x) ≤ g(x).
Z ∞
c h(x)dx converges.
0

• From the given information, |h(x)| ≤ 2f (x).

Z ∞
• We claim 2f (x)dx converges.
0
Z ∞ Z ∞
◦ We can see this by writing 2f (x)dx = 2 f (x)dx and noting
0 0
that the second integral converges.
◦ Alternately, we can use the limitingZ comparison test, Theo- ∞
rem 1.12.22. Since f (x) ≥ 0,
f (x)dx converges, and
Z ∞ 0
2f (x)
lim = 2 (the limit exists), we conclude 2f (x)dx con-
x→∞ f (x) 0
verges.
• So, comparing h(x) to 2f (x), by the comparison test (Theorem 1.12.17)
Z ∞
h(x)dx converges.
0

Exercises — Stage 2
1.12.4.6. *. Solution. The denominator is zero when x = 1, but the numerator
is not, so the integrand has a singularity (infinite discontinuity) at x = 1. Let’s
replace the limit x = 1 with a variable that creeps toward 1.
1 t
x4 x4
Z Z
dx = lim− dx
0 x5 − 1 t→1 0 x5 − 1

To evaluate this integral we use the substitution u = x5 , du = 5x4 dx. When x = 0

we have u = 0, and when x = t we have u = t5 , so
1 t Z u=t5
x4 x4 1
Z Z
dx = lim− dx = lim du
0 x5 − 1 t→1 5
0 x −1 t→1− u=0 5(u − 1)
 t5 !
1 1
= lim− log |u − 1| = lim− log |t5 − 1| = −∞
t→1 5 0 t→1 5

The limit diverges, so the integral diverges as well.

888
S OLUTIONS TO E XERCISES

1.12.4.7. *. Solution. The denominator of the integrand is zero when x = −1,

but the numerator is not. So, the integrand has a singularity (infinite discontinu-
ity) at x = −1. This is the only “source of impropriety” in this integral, so we
only need to make one break in the domain of integration.
2 t 2
1 1 1
Z Z Z
dx = lim− dx + lim+ dx
−2 (x + 1)4/3 t→−1 −2 (x + 1)4/3 t→−1 t (x + 1)4/3

Let’s start by considering the left limit.

!
t  t
1 3
Z
lim dx = lim− −
t→−1− −2 (x + 1)4/3 t→−1 (x + 1)1/3 −2
 
3 3
= lim− − + =∞
t→−1 (t + 1)1/3 (−1)1/3

Since this limit diverges, the integral diverges. (A similar argument shows that
the second integral diverges. Either one of them diverging is enough to conclude
that the original integral diverges.)
1.12.4.8. *. Solution. First, let’s identify all “sources of impropriety.” The inte-
grand has a singularity when 4x2 − x = 0, that is, when x(4x − 1) = 0, so at x = 0
and x = 41 . Neither of these are in our domain of integration, so the only “source
of impropriety” is the unbounded domain of integration.
We could antidifferentiate this function (it looks like a nice candidate for a trig
substitution), but is seems easier to use a comparison. For large values of x, the
term x2 willRbe much largerR ∞than x, so we might guess that our integral behaves
∞ 1 1
similarly to 1 √4x2 dx = 1 2x dx.
√ √ R∞ 1
For all x ≥ 1, 4x2 − x ≤ 4x2 = 2x. So, √4x12 −x ≥ 2x 1
. Note 1 2x dx diverges:

t  
1 1 1
Z
t
lim dx = lim log x 1 = lim log t = ∞
t→∞ 1 2x t→∞ 2 t→∞ 2

So:
1 √ 1
• 2x
and 4x2 −x
are defined and continuous for all x ≥ 1,
1
• 2x
≥ 0 for all x ≥ 1,

• √ 1 ≥ √1 = 1
for all x ≥ 1, and
4x2 −x 4x2 2x
R∞ 1
• 1 2x
dx diverges.

By the comparison test, Theorem 1.12.17 in the text, the integral does not con-
verge.
1.12.4.9. *. Solution. The integrand is positive everywhere. So, either the in-
tegral converges to some finite number, or it is infinite. We want to generate a

889
S OLUTIONS TO E XERCISES

guess as to which√ it is.

When x is small, x > x2 , so we might guess that our integral behaves√like the
integral of √1x when x is near to 0. On the other hand, when x is large, x < x2 ,
so we might guess that our integral behaves like the integral of x12 as x goes to
infinity. This is the hunch that drives the following work:

1 1
0≤ √ ≤ √
x2 + x x
1
dx
Z
and the integral √ converges by Example 1.12.9, and
0 x
1 1
0≤ 2 √ ≤ 2
x + x x
∞
dx
Z
and the integral converges by Example 1.12.8
1 x2
dx√
Note x2 + x
is defined and continuous everywhere, √1x is defined and continuous
for x > 0, and x12 is defined and continuous for x ≥ 1. So, the integral converges
by the comparison test, Theorem 1.12.17 in the text, together with Remark 1.12.16.
1.12.4.10. Solution. There are two “sources of impropriety:” the two (infinite)
limits of integration. So, we break our integral into two pieces.
Z ∞ Z 0 Z ∞
cos xdx = cos xdx + cos xdx
−∞ −∞ 0
Z 0  Z b 
= lim cos xdx + lim cos xdx
a→∞ −a b→∞ 0

These are easy enough to antdifferentiate.

= lim [sin 0 − sin(−a)] + lim [sin b − sin 0]

a→∞ b→∞
= DNE

Since the limits don’t exist, the integral diverges. (It happens that both limits
don’t exist; even if only one failed to exist, the integral would still diverge.)
1.12.4.11. Solution. There are two “sources of impropriety:” the two bounds.
So, we break our integral into two pieces.
Z ∞ Z 0 Z ∞
sin xdx = sin xdx + sin xdx
−∞ −∞ 0
Z 0  Z b 
= lim sin xdx + lim sin xdx
a→∞ −a b→∞ 0
= lim [− cos 0 + cos(−a)] + lim [− cos b + cos 0]
a→∞ b→∞
= DNE
Since the limits don’t exist, the integral diverges. (It happens that both limits

890
S OLUTIONS TO E XERCISES

don’t exist; even if only one failed to exist, the integral would diverge.)
Remark: it’s very tempting to think that this integral should converge, because
as an odd function the area to the right of the x-axis “cancels out” the area to
the left when the limits of integration are symmetric. One justification for not
using this intuition is given in RExample 1.12.11 in the text. Here’s another: In
∞
Question 1.12.4.10 we saw that −∞ cos xdx diverges. Since sin x = cos(x − π/2),
the area bounded by sine and the area boundedR by cosine over an infinite region
∞
seem to be the same–only shifted by π/2. So if −∞ sin xdx = 0, then we ought to
R∞
also have −∞ cos xdx = 0, but we saw in Question 1.12.4.10 this is not the case.

y = cos x
x

y = sin x

1.12.4.12. Solution. First, we check that the integrand has no singularities. The
denominator is always positive when x ≥ 10, so our only “source of impropriety”
is the infinite limit of integration.
x4 1
We further note that, for large values of x, the integrand resembles 5 = . So,
x x
we have a two-part hunch: Zthat the integral diverges, and that we can show it
∞
1
diverges by comparing it to dx.
10 x
x4 − 5x3 + 2x − 7
In order to use the comparison test, we’d need to show that ≥
x5 + 3x + 8
1
. If this is true, it will be difficult to prove–and it’s not at all clear that it’s
x
true. So, we will use the limiting comparison test instead, Theorem 1.12.22, with
1 x4 − 5x3 + 2x − 7
g(x) = , f (x) = , and a = 10.
x x5 + 3x + 8

• Both f (x) and g(x) are defined and continuous for all x > 0, so in particular
they are defined and continuous for x ≥ 10.
• g(x) ≥ 0 for all x ≥ 10
Z ∞
• g(x)dx diverges.
10

• Using l’Hôpital’s rule (5 times!), or simply dividing both the numerator and
denominator by x5 (the common leading term), tells us:
x4 −5x3 +2x−7
f (x) x5 +3x+8 x4 − 5x3 + 2x − 7
lim = lim 1 = lim x ·
x→∞ g(x) x→∞
x
x→∞ x5 + 3x + 8

891
S OLUTIONS TO E XERCISES

x5 − 5x4 + 2x2 − 7x
= lim =1
x→∞ x5 + 3x + 8
That is, the limit exists and is nonzero.
Z ∞
By the limiting comparison test, we conclude f (x)dx diverges.
10

1.12.4.13. Solution. Our domain of integration is finite, so the only potential

“sources of impropriety” are infinite discontinuities in the integrand. To find
these, we factor.
Z 10 Z 10
x−1 x−1
2
dx = dx
0 x − 11x + 10 0 (x − 1)(x − 10)

A removable discontinuity doesn’t affect the integral.

10
1
Z
= dx
0 x − 10

Use the substitution u = x−10, du = dx. When x = 0, u = −10, and when x = 10,
u = 0.
Z 0
1
= du
−10 u

This is a p-integral with p = 1. From Example 1.12.9 and Theorem 1.12.20, we

know it diverges.
1.12.4.14. *. Solution. You might think that, because the integrand is odd, the
integral converges to 0. This is a common mistake– see Example 1.12.11 in the
text, or Question 11 in this section. In the absence of such a shortcut, we use our
standard procedure: identifying problem spots over the domain of integration,
and replacing them with limits.
There are two “sources of impropriety,” namely x → +∞ and x → −∞. So, we
split the integral in two, and treat the two halves separately. The integrals below
can be evaluated with the substitution u = x2 + 1, 12 du = xdx.
+∞ Z 0 Z +∞
x x x
Z
2
dx = 2
dx + 2
dx
−∞ x +1 −∞ x + 1 0 x +1
0 Z 0
x x 1
Z 0
2
dx = lim 2
dx = lim log(x2 + 1)
−∞ x +1 R→∞ −R x + 1 R→∞ 2 −R
1 1
log 1 − log(R2 + 1) = lim − log(R2 +1)

= lim
R→∞ 2 R→∞ 2
= −∞
+∞ Z R
x x 1
Z R
2
dx = lim dx = lim log(x + 1)
0 x2 + 1 R→∞ 0 x2 + 1 R→∞ 2 0

892
S OLUTIONS TO E XERCISES

1 1
log(R2 + 1) − log 1 = lim log(R2 + 1)

= lim
R→∞ 2 R→∞ 2
= +∞

Both halves diverge, so the whole integral diverges.

Once again: after we found that one of the limits diverged, we could have
stopped and concluded that the original integrand diverges. Don’t make the
mistake of thinking that ∞ − ∞ = 0. That can get you into big trouble. ∞ is not a
normal number. For example 2∞ = ∞. So if ∞ were a normal number we would
have both ∞ − ∞ = 0 and ∞ − ∞ = 2∞ − ∞ = ∞.
1.12.4.15. *. Solution. We don’t want to antidifferentiate this integrand, so let’s
use a comparison. Note the integrand is positive when x > 0.
| sin x| 1
For any x, | sin x| ≤ 1, so 3/2 1/2
≤ 3/2 .
x +x x + x1/2
Since x = 0 and x → ∞ both cause the integral to be improper, we need to break
it into two pieces. Since both terms in the denominator give positive numbers
1 1 1 1
when x is positive, 3/2 1/2
≤ 3/2 and 3/2 1/2
≤ 1/2 . That gives us two
x +x x x +x x
options for comparison.
When x is positive and close to zero, x1/2 ≥ x3/2 , so we guess that we should
1
compare our integrand to x1/2 near the limit x = 0. In contrast, when x is very
1
large, x1/2 ≤ x3/2 , so we guess that we should compare our integrand to x3/2 as x
goes to infinity.

| sin x| 1
≤ 1/2
x3/2 +x 1/2 x
1
dx
Z
and the integral converges by the p-test, Example 1.12.9
0 x1/2
| sin x| 1
3/2 1/2
≤ 3/2
x +x x
∞
dx
Z
and the integral converges by the p-test, Example 1.12.8
1 x3/2
Now we have all the data we need to apply the comparison test, Theorem 1.12.17.
| sin x| 1 1
• , 1/2 , and 3/2 are defined and continuous for x > 0
x3/2 +x 1/2 x x
1 1
• and are nonnegative for x ≥ 0
x1/2 x3/2
Z 1
| sin x| 1 1
• 3/2 1/2
≤ 1/2 for all x > 0 and 1/2
dx converges, so (using Re-
x +x x 0 x
Z 1
| sin x|
mark 1.12.16) 3/2
dx converges.
0 x + x1/2
Z ∞
| sin x| 1 1
• 3/2 1/2
≤ 3/2
for all x ≥ 1 and 3/2
dx converges, so
x +x x 1 x

893
S OLUTIONS TO E XERCISES

∞
| sin x|
Z
dx converges.
1 x3/2
+ x1/2
Z ∞
| sin x|
Therefore, our integral dx converges.
0 x + x1/2
3/2

1.12.4.16. *. Solution. The integrand is positive everywhere, so either the in-

tegral converges to some finite number or it is infinite. There are two potential
“sources of impropriety” — a possible singularity at x = 0 and the fact that the
domain of integration extends to ∞. So we split up the integral.
∞ 1
x+1 x+1
Z Z
dx = dx
0 x (x2 + x + 1)
1/3
0 x1/3 (x2 + x + 1)
∞
x+1
Z
+ dx
1 x (x2 + x + 1)
1/3

Let’s develop a hunch about whether the integral converges or diverges. When
x ≈ 0, x2 and x are both a lot smaller than 1, so we guess we should compare the
1
integrand to x1/3 .
x+1 1 1
1/3 2
≈ 1/3 = 1/3
x (x + x + 1) x (1) x
R1 1
Note 0 x1/3 dx converges by Example 1.12.9 (it’s a p-type integral), so we guess
R1 x+1
0 x1/3 (x2 +x+1)
dx converges as well.
When x is very large, x2 is much bigger than x, which is much bigger than 1, so
1
we guess we should compare the integrand to x4/3 .

x+1 x 1
≈ 1/3 2 = 4/3
x1/3 (x2 + x + 1) x (x ) x
R∞ 1
Note
R∞ 1 x4/3
dx converges by Example 1.12.8 (it’s a p-type integral), so we guess
x+1
1 x1/3 (x2 +x+1)
dx converges as well.
Now it’s time to verify our guesses with the limiting comparison test, Theo-
rem 1.12.22. Be careful: our “≈” signs are not strong enough to use either the
limiting comparison test or the comparison test, they are only enough to suggest
a reasonable function to compare to.

x+1 1 1
• x1/3 (x2 +x+1)
, x1/3 , and x4/3
are defined and continuous for all x > 0
1 1
• x1/3
are positive for all x > 0
and x4/3
R1 1 R∞ 1
• 0 x1/3 dx and 1 x4/3 dx both converge
x+1
x1/3 (x2 +x+1) x+1 0+1
• lim 1 = lim = = 1; in particular, this limit
x→0 x→0 x2 + x + 1 0+0+1
x1/3
exists.

894
S OLUTIONS TO E XERCISES

• Using the limiting comparison test (Theorem 1.12.22, together with

Remark 1.12.16 because our impropriety is due to a singularity),
R1 x+1
0 x1/3 (x2 +x+1)
dx converges.
x+1
x1/3 (x2 +x+1) x(x + 1)
• lim 1 = lim = 1; in particular, this limit exists.
x→∞
x4/3
x2 + x + 1
x→0

R∞
• Using the limiting comparison test (Theorem 1.12.22), 1 x1/3 (xx+1 2 +x+1) dx

converges.
Z ∞
x+1
We conclude dx converges.
0 x1/3 (x2 + x + 1)

Exercises — Stage 3
1.12.4.17. Solution. To find the volume of the solid, we cut it into horizontal
slices, which are thin circular disks. At height y, the disk has radius x = y1 and
thickness dy, so its volume is yπ2 dy. The base of the solid is at height y = 1, and
its top is at height y = a1 . So, the volume of the entire solid is:

1/a  1/a
π π
Z
dy = − = π(1 − a)
1 y2 y 1

If we imagine sliding a closer and closer to 0, the volume increases, getting closer
and closer to π units, but never quite reaching it.
So, the statement is false. For example, if we set M = 4, no matter which a we
choose our solid has volume strictly less than M .
R1
Remark: we’ve seen before that 0 x1 dx diverges. If we imagine the solid that
would result from choosing a = 0, it would have a scant volume of π cubic units,
but a silhouette (side view) of infinite area.
1.12.4.18. *. Solution. Our goal is to decide when this integral diverges, and
where it converges. We will leave q as a variable, and antidifferentiate. In order
to antidifferentiate without knowing q, we’ll need different cases. The integrand
n+1
is x−5q , so when −5q 6= −1, we use the power rule (that is, xn dx = xn+1 ) to
R
1
antidifferentiate. Note x(−5q)+1 = x1−5q = 5q−1 .
x
h it
x1−5q
Z t

 1−5q
with 1 − 5q > 0 if q < 15
1

1  t
if q = 15

5q
dx = log x 1
1 x 
 h it
1
with 5q − 1 > 0 if q > 15


(1−5q)x 5q−1
 1
1
 1−5q (t

 1−5q
− 1) with 1 − 5q > 0 if q < 15
= 1
log t if q = 5

1 1 1
(1 − ) with 5q − 1 > 0 if q > .


5q−1 t5q−1 5

895
S OLUTIONS TO E XERCISES

Therefore,
∞ Z t 
1 1
Z
dx = lim dx
1 x5q t→∞ x5q
 1 
1 1−5q 1


 1−5q t→∞
lim t −1 =∞ if q < 5

1

lim log t = ∞
= t→∞ if q = 5
 
1

 1 1
 5q−1 1 − lim 5q−1 = 5q−1 if q > 15 .


t→∞ t

The first two cases are divergent, and so the largest such value is q = 15 . (Alterna-
tively, we might recognize this as a “p-integral” with p = 5q, and recall that the
p-integral diverges precisely when p ≤ 1.)
1.12.4.19. Solution. This integrand is a nice candidate for the substitution u =
x2 + 1, 21 du = xdx. Remember when we use substitution on a definite integral,
we also need to adjust the limits of integration.
∞ t
x x
Z Z
2
dx = lim dx
0 (x + 1)p t→∞ 0 (x2 + 1)p
t2 +1
1 1
Z
= lim du
t→∞ 2 1 up
2
1 t +1 −p
Z
= lim u du
t→∞ 2 1
  1−p t2 +1
1
 u
 lim
2 t→∞
if p 6= 1
= 1−p 1
h it2 +1
 1 lim log |u|

if p = 1

2
 t→∞ 1

 1 lim 1
 2 1−p


2 t→∞
(t + 1) − 1 if p 6= 1
= h1 − p i
 12 lim log(t2 + 1) = ∞
 if p = 1
t→∞

At this point, we can see that the integral diverges when p = 1. When p 6= 1, we
have the limit
1/2  2 1−p
 1/2 h 2 1−p
i 1/2
lim (t + 1) −1 = lim (t + 1) −
t→∞ 1 − p 1−p t→∞ 1−p

Since t2 +1 → ∞, this limit converges exactly when the exponent 1−p is negative;
that is, it converges when p > 1, and diverges when p < 1.
So, the integral in the question converges when p > 1.
1.12.4.20. Solution.
• First, we notice there is only one “source of impropriety:” the domain of
integration is infinite. (The integrand has a singularity at t = 1, but this is
not in the domain of integration, so it’s not a problem for us.)

896
S OLUTIONS TO E XERCISES

• We should try to get some intuition about whether the integral converges
or diverges.
R ∞When t → ∞, notice the integrand “looks like” the function t14 .
1
We know 1 t4 dt converges, because it’s a p-integral with p = 4 > 1 (see
Example 1.12.8). So, our integral probably converges as well. If we were
only asked show it converges, we could use a comparison test, but we’re
asked more than that.

• Since we guess the integral converges, we’ll need to evaluate it. The inte-
grand is a rational function, and there’s no obvious substitution, so we use
partial fractions.

1 1 1
= 2 = 2
t4 −1 2
(t + 1)(t − 1) (t + 1)(t + 1)(t − 1)
At + B C D
= 2 + +
t +1 t+1 t−1
Multiply by the original denominator.

1 = (At+B)(t+1)(t−1) + C(t2 +1)(t−1) + D(t2 +1)(t+1) (∗)

Set t = 1.
1
1 = 0 + 0 + D(2)(2) ⇒ D=
4
Set t = −1.
1
1 = 0 + C(2)(−2) + 0 ⇒ C=−
4
1
Simplify (∗) using D = 4
and C = − 14 .

1 1
1 = (At + B)(t + 1)(t − 1)− (t2 + 1)(t − 1) + (t2 + 1)(t + 1)
4 4
1 2
= (At + B)(t + 1)(t − 1) + (t + 1)
  2 
3 1 2 1
= At + B + t − At + −B
2 2

By matching up coefficients of corresponding powers of t, we find A = 0 and

B = − 12 .
Z ∞ Z ∞ 
1 −1/2 1/4 1/4
dt = − + dt
2 t4 − 1 2 t2 + 1 t + 1 t − 1
Z R 
−1/2 1/4 1/4
= lim − + dt
R→∞ 2 t2 + 1 t + 1 t − 1
 R
1 1 1
= lim − arctan t − log |t + 1| + log |t − 1|
R→∞ 2 4 4 2

897
S OLUTIONS TO E XERCISES
 R
1 1 t−1
= lim − arctan t + log
R→∞ 2 4 t+1 2
 1 1 1 R−1
= lim − arctan R + arctan 2 + log
R→∞ 2 2 4 R+1
1 2−1 
− log
4 2+1

R−1
We can use l’Hôpital’s rule to see lim = 1. Also note − log(1/3) = log 3.
R→∞ R + 1

1 π  1 1 1
=− + arctan 2 + log 1 + log 3
2 2 2 4 4
log 3 − π 1
= + arctan 2
4 2
1.12.4.21. Solution. There are three singularities in the integrand: x = 0, x = 1,
and x = 2. We’ll need to break up the integral at each of these places.
Z 5 !
1 1 1
p +p +p dx
−5 |x| |x − 1| |x − 2|
Z 0 !
1 1 1
= p +p +p dx
−5 |x| |x − 1| |x − 2|
Z 1 !
1 1 1
+ p +p +p dx
0 |x| |x − 1| |x − 2|
Z 2 !
1 1 1
+ p +p +p dx
1 |x| |x − 1| |x − 2|
Z 5 !
1 1 1
+ p +p +p dx
2 |x| |x − 1| |x − 2|

This looks rather unfortunate. Let’s think again. If all of the integrals below
converge, then we can write:
Z 5 !
1 1 1
p +p +p dx
−5 |x| |x − 1| |x − 2|
Z 5 Z 5 Z 5
1 1 1
= p dx + p dx + p dx
−5 |x| −5 |x − 1| −5 |x − 2|

That looks a lot better. Also, we have a good reason to guess these integrals
converge–they look like p-integrals with p = 21 . Let’s take a closer look at each
one.
Z 5 Z 0 Z 5
1 1 1
p dx = p dx + p dx
−5 |x| −5 |x| 0 |x|

898
S OLUTIONS TO E XERCISES

5
1
Z
=2 p dx (even function)
0 |x|
5
1
Z
=2 √ dx
0 x

This is a p-integral, with p = 12 . By Example 1.12.9 (and Theorem 1.12.20, since

the upper limit of integration is not 1), it converges. The other two pieces behave
similarly.
5 1 5
1 1 1
Z Z Z
p dx = p dx + p dx
−5 |x − 1| −5 |x − 1| 1 |x − 1|

Use u = x − 1, du = dx
0 Z 4
1 1
Z
= p du + p dx
−6 |u| 0 |u|
Z 6 Z 4
1 1
= √ du + √ dx
0 u 0 u

Since our function is even, we use the reasoning of Example 1.2.10 in the text to
consider the area under the curve when x ≥ 0, rather than when x ≤ 0. Again,
these are p-integrals with p = 21 , so they both converge. Finally:
5 2 5
1 1 1
Z Z Z
p dx = p dx + p dx
−5 |x − 2| −5 |x − 2| 2 |x − 2|

Use u = x − 2, du = dx.
0 Z 3
1 1
Z
= p du + p du
−7 |u| 0 |u|
Z 7 Z 3
1 1
= √ du + √ du
0 u 0 u

Since p = 12 , so they both converge. We conclude our original integral, as the sum
of convergent integrals, converges.
1.12.4.22. Solution. We can use integration by parts twice to find the antideriva-
tive of e−x sin x, as in Example 1.7.10. To keep our work a little simpler, we’ll find
the antiderivative first, then take the limit.
Let u = e−x , dv = sin xdx, so du = −e−x dx and v = − cos x.
Z Z
e sin xdx = −e cos x − e−x cos xdx
−x −x

Now let u = e−x , dv = cos xdx, so du = −e−x dx and v = sin x.

 Z 
−x −x −x
= −e cos x − e sin x + e sin xdx

899
S OLUTIONS TO E XERCISES
Z
−x −x
= −e cos x − e sin x − e−x sin xdx

All together, we found

Z Z
e sin xdx = −e cos x − e sin x − e−x sin xdx + C
−x −x −x

Z
2 e−x sin xdx = −e−x cos x − e−x sin x + C
1
Z
e−x sin xdx = − x (cos x + sin x) + C
2e
C
(Remember, since C is an arbitrary constant, we can rename 2
to simply C.) Now
we can evaluate our improper integral.
Z ∞ Z b
−x
e sin xdx = lim e−x sin xdx
0 b→∞ 0
 b
1
= lim − x (cos x + sin x)
b→∞ 2e 0
 
1 1
= lim − (cos b + sin b)
b→∞ 2 2eb

To find the limit, we use the Squeeze Theorem (see the CLP-1 text). Since
| sin b|, | cos b| ≤ 1 for any b, we can use the fact that −2 ≤ cos b + sin x ≤ 2 for
any b.

−2 1 2
b
≤ b (cos b + sin b) ≤ b
2e 2e 2e
−2 2
lim =0= b
b→∞ 2eb
 2e 
1
So, lim (cos b + sin b) = 0
b→∞ 2eb
 
1 1 1
Therefore, = lim − (cos b + sin b)
2 b→∞ 2 2eb
∞
1
Z
That is, e−x sin xdx = .
0 2
1.12.4.23. *. Solution. The integrand is positive everywhere. So either the in-
tegral converges to some finite number or it is infinite. There are two potential
“sources of impropriety” — a possible singularity at x = 0 and the fact that the
domain of integration extends to ∞. So, we split up the integral.
∞ 1 ∞
sin4 x sin4 x sin4 x
Z Z Z
dx = dx + dx
0 x2 0 x2 1 x2

900
S OLUTIONS TO E XERCISES

Let’s consider the first integral. By l’Hôpital’s rule (see the CLP-1 text),

sin x cos x
lim = lim = cos 0 = 1
x→0 x x→0 1
Consequently,

sin4 x  2
 sin x  sin x 
lim = lim sin x lim lim =0×1×1=0
x→0 x2 x→0 x→0 x x→0 x

and the first integral is not even improper. R∞ 1

Now for the second integral. Since | sin x| ≤ 1, we’ll compare it to 1 x2
.
sin4 x 1
• x2
and x2
are defined and continuous for every x ≥ 1
4 4
• 0 ≤ sinx2 x ≤ x12 = x12 for every x ≥ 1
R∞
• 1 x12 dx converges by Example 1.12.8 (it’s a p-type integral with p > 1)
Z ∞
sin4 x
By the comparison test, Theorem 1.12.17, 2
dx converges.
Z 1 Z ∞ 1 x Z ∞
sin4 x sin4 x sin4 x
Since dx and dx both converge, we conclude dx
0 x2 1 x2 0 x2
converges as well.
1.12.4.24. Solution. Since the denominator is positive for all x ≥ 0, the in-
tegrand is continuous over [0, ∞). So, the only “source of impropriety” is the
infinite domain of integration.

x
• Solution 1: Let’s try to use a direct comparison. Note √ ≥ 0 when-
+ x ex √
ever x ≥ 0. Also note that, for large values of x, ex is much larger than x.
That leads us to consider the following inequalty:
x x
0≤ x √ ≤ x
e + x e
R∞
If 0 exx dx converges, we’re in business. Let’s figure it out. The integrand
looks like a candidate for integration by parts: take u = x, dv = e−x dx, so
du = dx and v = −e−x .
Z ∞ Z b h Z b 
x x x ib −x
dx = lim dx = lim − x + e dx
0 ex b→∞ 0 ex b→∞ e 0 0
   
b  −x b b 1
= lim − b + −e 0 = lim − b − b + 1
b→∞ e b→∞ e e
   
b+1 1
= lim 1 − b
= lim 1 − b = 1
b→∞ e }
| {z b→∞ e
num→∞
den→∞
R∞ x
Using l’Hôpital’s rule, we see 0 ex
dx converges. All together:

901
S OLUTIONS TO E XERCISES
x x√
◦ ex
and ex + x
are defined and continuous for all x ≥ 0,
x√ x
◦ ex + x
≤ ex
, and
R∞ x
◦ 0 ex
dx converges.
∞
x
Z
So, by Theorem 1.12.17, our integral √ dx converges.
0 ex + x
• Solution 2: Let’s try to use a different direct comparison from Solution 1,
1
and avoid integration by parts. We’d like to compare to something like x ,
e
but the inequality goes the wrong way. So, we make a slight modification:
we consider 2e−x/2 . To that end, we claim x < 2ex/2 for all x ≥ 0. We can
prove this by noting the following two facts:

◦ 0 < 2 = 2e0/2 , and

d d
◦ dx
{x} = 1 ≤ ex/2 = dx
{2ex/2 }.

So, when x = 0, x < 2ex/2 , and then as x increases, 2ex/2 grows faster than
x.
Now we can make the following comparison:

x x 2ex/2 2
√
0≤ ≤ x
< x
= x/2
x
e + x e e e
R∞ 2 R∞
We have a hunch that 0 ex/2 dx converges, just like 0 e1x dx. This is easy
enough to prove. We can guess an antiderivative, or use the substitution
u = x/2.
∞ R  R
2 2 4
Z Z
dx = lim dx = lim −
0 ex/2 R→∞ 0 ex/2 R→∞ ex/2 0
 R
4 4
= lim 0
− R/2 =4
R→∞ e e 0

Now we know:
2
◦ 0 ≤ ex +x√x ≤ ex/2 , and
R∞ 2
◦ 0 ex/2 dx converges.
x√ 2
◦ Furthermore, ex + x
and ex/2
are defined and continuous for all x ≥ 0.

By the comparison test (Theorem 1.12.17), we conclude the integral con-

verges.

• Solution 3: Let’s use the limiting comparison test (Theorem

R∞ 1.12.22). We
have a hunch that our integral behaves similarly to 0 e1x dx, which con-
verges (see Example 1.12.18). Unfortunately, if we choose g(x) = e1x (and,

902
S OLUTIONS TO E XERCISES
x√
of course, f (x) = ex + x
), then
f (x) x x
lim = lim x √ · ex = lim √ =∞
x→∞ g(x) x→∞ e + x x→∞ 1 + x
e x
|{z}
→0

That is, the limit does

√
not exist, so the limiting comparison test does not
x
apply. (To find lim ex , you can use l’Hôpital’s rule.)
x→∞
This setback encourages us to try a slightly different angle. If g(x) gave
(x)
larger values, then we could decrease fg(x) 1
. So, let’s try g(x) = ex/2 = e−x/2 .
Now,
f (x) x 1 x
lim = lim x √ ÷ x/2 = lim √
x→∞ g(x) x→∞ e + x e x→∞ ex/2 + x
ex/2

Hmm... this looks hard. Instead of dealing with it directly, let’s use the
squeeze theorem (see CLP-1 notes).
x x
0≤ √ ≤ x/2
x
ex/2 + ex/2 e

Using l’Hôpital’s rule,

x 1
lim = lim = 0 = lim 0
x→∞ ex/2 x→∞ 1 ex/2 x→∞
|{z} 2
num→∞
den→∞

x√
ex + x 1
So, by the squeeze theorem lim 1 = 0. Since this limit exists, ex/2
is
x→0 ex/2
a reasonable function to use in the limiting comparison
R∞ 1 test (provided its
integral converges). So, we need to show that 0 ex/2 dx converges. This
can be done by simply evaluating it:

∞ b
1 1
Z Z
b
dx = lim e−x/2 dx = lim − e−x/2 0
0 ex/2 b→∞ 0 b→∞ 2
 
1 1 1
= lim − b/2
−1 =
b→∞ 2 e 2

So, all together:

1
◦ The functions ex +x√x and ex/2 are defined and continuous for all x ≥ 0,
1
and ex/2 ≥ 0 for all x ≥ 0.
R∞ 1
◦ 0 ex/2 dx converges.
x √
+ x
◦ The limit lim ex 1 exists (it’s equal to 0).
x→∞ ex/2
1.12.4.25. *. Solution. There are two sources of error: the upper bound is t,
◦ So,
rather than the limiting
R infinity, and comparison
we’re test (Theorem
using an approximation 1.12.17)
with tells number
some finite us that
∞ x√
0 ex + x
dx converges as well.

903
S OLUTIONS TO E XERCISES

of intervals, n. Our plan is to first find a value of t that introduces an error of

R ∞ e−x
no more than 21 10−4 . That is, we’ll find a value of t such that t 1+x dx ≤ 12 10−4 .
R t e−x
After that, we’ll find a value of n that approximates 0 x+1 dx to within 21 10−4 .
Then, all together, our error will be at most 12 10−4 + 12 10−4 = 10−4 , as desired.
(Note we could have broken up the error in another way—it didn’t have to be
1
2
10−4 and 21 10−4 . This will give us one of many possible answers.)
R ∞ e−x e−x
Let’s find a t such that t 1+x dx ≤ 21 10−4 . For all x ≥ 0, 0 < 1+x ≤ e−x , so
∞ ∞
e−x (∗)1 −4
Z Z
dx ≤ e−x dx = e−t ≤ 10
t 1+x t 2
1 
where (∗) is true if t ≥ − log 10−4 ≈ 9.90
2
Choose, for example, t = 10.
Now it’s time to decide how many intervals we’re going to use to approximate
Z t −x
e
dx. Again, we want our error to be less than 12 10−4 . To bound our error,
0 x + 1
e−x
we need to know the second derivative of x+1 .

e−x e−x e−x

f (x) = =⇒ f 0 (x) = − −
1+x 1 + x (1 + x)2
e−x e−x e−x
=⇒ f 00 (x) = +2 + 2
1+x (1 + x)2 (1 + x)3

Since f 00 (x) is positive, and decreases as x increases,

00 00 5(10 − 0)3 5000 625

|f (x)| ≤ f (0) = 5 =⇒ |En | ≤ 2
= 2
= 2
24n 24n 3n
and |En | ≤ 21 10−4 if

625 1
2
≤ 10−4
3n 2
1250 × 104
⇐⇒ n2 ≥
r 3
1.25 × 107
⇐⇒ n≥ ≈ 2041.2
3
So t = 10 and n = 2042 will do the job. There are many other correct answers.
1.12.4.26. Solution.
a Since f (x) is odd, using the reasoning of Example 1.2.11,
Z −1 Z −1 Z t
f (x)dx = lim f (x)dx = lim − f (x)dx
−∞ t→∞ −t t→∞ 1
Z t
= − lim f (x)dx
t→∞ 1

904
S OLUTIONS TO E XERCISES
Z ∞
Since f (x)dx converges, the last limit above converges. Therefore,
Z −1 1

f (x)dx converges.
−∞

b Since f (x) is even, using the reasoning of Example 1.2.10,

Z −1 Z −1 Z t
f (x)dx = lim f (x)dx = lim f (x)dx
−∞ t→∞ −t t→∞ 1
Z t
= lim f (x)dx
t→∞ 1
Z ∞
Since f (x)dx converges, the last limit above converges. Since f (x) is
1 Z ∞
continuous everywhere, by Theorem 1.12.20, f (x)dx converges (note
−1
the adjusted lower limit). Then, since
Z ∞ Z −1 Z ∞
f (x)dx = f (x)dx + f (x)dx
−∞ −∞ −1

and both terms converge, our original integral converges as well.

Rx
1.12.4.27. Solution. Define F (x) = 0 e1t dt.
Z x  x
1 1 1 1 1
F (x) = dt = − = − < =1
0 e
t et 0 e0 ex e0

So, the statement is false: there is no x such that F (x) = 1. For every real x,
F (x) < e10 = 1. Z x
1
We note here that lim dt = 1. So, as x grows larger, the gap between F (x)
x→∞ 0 et
and 1 grows infintesimally small. But there is no real value of x where F (x) is
exactly equal to 1.

1.13 · More Integration Examples

· Exercises

Exercises — Stage 1
1.13.1. Solution. (A) Note f 0 (x)f (x) dx =
R R
u du if we substitute u =
f (x). This is the kind of integrand described in (I). It’s quite possible that a u-
substitution would work on the others, as well, but (I) is the most reliable kind
of integrand for a u-substitution.
(B) A trigonometric substitution usually allows us to cancel out a square root
containing a quadratic function, as in (IV).
(C) We can often antidifferentiate the product of a polynomial with an exponen-

905
S OLUTIONS TO E XERCISES

tial function using integration by parts: see Examples 1.7.1, 1.7.6. If we let u be
the polynomial function and dv be the exponential, as long as we can antidif-
ferentiate dv, we can repeatedly apply integration by parts until the polynomial
function goes away. So, we go with (II)
(D) We apply partial fractions to rational functions, (III).
Note: without knowing more about the functions, there’s no guarantee that the
methods we chose will be the best methods, or even that they will work (with
the exception of (I)). With practice, you gain intuition about likely methods for
different integrals. Luckily for you, there’s lots of practice below.

Exercises — Stage 2
1.13.2. Solution. The integrand is a product of powers of sine and cosine. Since
cosine has an odd power, we want to substitute u = sin x, du = cos xdx. There-
fore, we should:

• reserve one cosine for the derivative of sine in our substitution, and

• change the rest of the cosines to sines using the identity sin2 x + cos2 x = 1.

Z π/2 Z π/2
4 5
sin x cos xdx = sin4 x(cos2 x)2 cos xdx
0 0
Z π/2
= sin4 x(1 − sin2 x)2 cos
| {zxdx}
0
du
Z sin(π/2)
= u4 (1 − u2 )2 du
sin(0)
Z 1
= 4
u (1 − 2u2 + u4 )du
Z0 1
= (u4 − 2u6 + u8 )du
0
 u=1
1 5 2 7 1 9
= u − u + u
5 7 9 u=0
 
1 2 1
= − + −0
5 7 9
8
=
315
1.13.3. Solution. We notice that there is a quadratic equation under the square
root. If that equation were a perfect square, we could get rid of the square root:
so we’ll mould it into a perfect square using a trig substitution.
Our candidates will use one of the following identities:

1 − sin2 θ = cos2 θ tan2 θ + 1 = sec2 θ sec2 θ − 1 = tan2 θ

We’ll be substituting x =(something), so we notice that 3 − 5x2 has the general

906
S OLUTIONS TO E XERCISES

form of (constant)−(function), as does 1−sin2 θ. In order to get the constant right,

we multiply through by three:

3 − 3 sin2 θ = 3 cos2 θ

Our goal is to get 3 − 5x2 = 3 − 3 sin2 θ; so we solve this equation for x and decide
on the substitution
r r
3 3
x= sin θ, dx = cos θdθ
5 5
Now we evaluate our integral.
v !2 r
Z √ Z u r
u 3 3
3 − 5x2 dx = t3 − 5 sin θ cos θdθ
5 5
Z p p
= 3 − 3 sin2 θ 3/5 cos θdθ
Z √ p
= 3 cos2 θ 3/5 cos θdθ
Z √ p
= 3 cos θ 3/5 cos θdθ
3
Z
=√ cos2 θdθ
5Z
3 1 + cos 2θ
=√ dθ
5 Z 2
3
= √ (1 + cos 2θ)dθ
2 5
 
3 1
= √ θ + sin(2θ) + C
2 5 2
3
= √ [θ + sin θ cos θ] + C
2 5
p p
From our substitution x = 3/5 sin θ, we glean sin θ = x 5/3, and θ =
p 
arcsin x 5/3 . To figure out cos θ, we draw a right triangle. Let θ be one angle,
√
x 5 √
and since sin θ = √ , we let the hypotenuse be 3 and the side opposite θ be
√ 3 √
x 5. By Pythagorus, the missing side (adjacent to θ) has length 3 − 5x2 .

√ 3 √
x 5
θ
√
3 − 5x2

907
S OLUTIONS TO E XERCISES
√
adj 3 − 5x2
Therefore, cos θ = = √ . So our integral evaluates to:
hyp 3

3
√ [θ + sin θ cos θ] + C
2 5
√
3 − 5x2
 
3 p p
= √ arcsin(x 5/3) + x 5/3 · √ +C
2 5 3
3 p x √
= √ arcsin(x 5/3) + · 3 − 5x2 + C
2 5 2
1.13.4. Solution. First, we note the integral is improper. So, we’ll need to re-
place the top bound with a variable, and take a limit. Second, we’re going to
have to antidifferentiate. The integrand is the product of an exponential func-
tion, e−x , with a polynomial function, x − 1, so we use integration by parts with
u = x − 1, dv = e−x du, du = dx, and v = −e−x .
x−1
Z Z
x
dx = −(x − 1)e + e−x dx
−x
e
= −(x − 1)e−x − e−x + C = −xe−x + C
Z ∞ Z b
x−1 x−1
So, dx = lim dx
0 ex b→∞ 0 ex
h x ib  
b
= lim − x = lim −
b→∞ e 0 b→∞ eb
|{z}
num→∞
den→∞

(∗) 1
= lim =0
b→∞ eb

(In the
Z ∞equality marked (∗), we used l’Hôpital’s rule.)
x−1
So, dx = 0.
0 ex Z ∞ Z ∞
x 1
Remark: this shows that, interestingly, dx = dx.
0 ex 0 ex
1.13.5. Solution.
• Solution 1: Notice the denominator factors as (x + 1)(3x + 1). Since the
integrand is a rational function (the quotient of two polynomials), we can
use partial fraction decomposition.
−2 −2
2
=
3x + 4x + 1 (x + 1)(3x + 1)
A B
= +
x + 1 3x + 1
A(3x + 1) + B(x + 1)
=
(x + 1)(3x + 1)
(3A + B)x + (A + B)
=
(x + 1)(3x + 1)

908
S OLUTIONS TO E XERCISES

So:
−2 = (3A + B)x + (A + B)
0 = 3A + B and − 2 = A + B
B = −3A and hence − 2 = A + (−3A)
A= 1 so then B = −3
So now:
−2 1 3
= −
3x2
+ 4x + 1 Zx +1 3x + 1 
−2 1 3
Z
dx = − dx
3x2 + 4x + 1 x + 1 3x + 1
= log |x + 1| − log |3x + 1| + C
x+1
= log +C
3x + 1

• Solution 2: The previous solution is probably the nicest. However, for the
foolhardy or the brave, this integral can also be evaluated using trigono-
metric substitution.
We start by completing the square on the denominator.
 
2 2 4 1
3x + 4x + 1 = 3 x + x +
3 3
 
2 2 4 4 1
=3 x +2· x+ − +
3 9 9 3
 2 !
2 4 3
=3 x+ − +
3 9 9
 2 !
2 1
=3 x+ −
3 9
 2
2 1
=3 x+ −
3 3
This has the form of a function minus a constant, which matches the
trigonometric identity sec2 θ − 1 = tan2 θ. Multiplying through by 13 , we
see we can use the identity 13 sec2 θ − 31 = 13 tan2 θ. So, to get the substitution
right, we want to choose a substitution that makes the following true:
 2
2 1 1 1
3 x+ − = sec2 θ −
3 3 3 3
 2
2 1
3 x+ = sec2 θ
3 3
 2
2
9 x+ = sec2 θ
3

909
S OLUTIONS TO E XERCISES

3x + 2 = sec θ

And, accordingly:

3dx = sec θ tan θdθ

Now, let’s simplify a little and use this substitution on our integral:
−2 −2
Z Z
2
dx =
2 dx
3x + 4x + 1 3 x + 32 − 13
−2
Z
=
2 3dx
9 x + 23 − 1
−2
Z
= 3dx
(3x + 2)2 − 1
−2
Z
= sec θ tan θdθ
(sec θ)2 − 1
−2
Z
= sec θ tan θdθ
tan2 θ
sec θ
Z
= −2 dθ
tan θ
1 cos θ
Z
= −2 · dθ
cos θ sin θ
1
Z
= −2 dθ
sin θ
Z
= −2 csc θdθ

Using the result of Example 1.8.21, or a table of integrals:

= 2 log |csc θ + cot θ| + C

Our final task is to translate this back from θ to x. Recall we used the sub-
hypotenuse
stitution 3x + 2 = sec θ. Using this information, and sec θ = ,
adjacent
we can fill in two sides of a right triangle with angle θ. p The Pythagorean
theorem
√ tells us the third side (opposite to θ) has measure (3x + 2)2 − 1 =
9x2 − 12x + 3.

+ 2
√
3x 9x2 + 12x + 3
θ
1

2 log |csc θ + cot θ| + C

910
S OLUTIONS TO E XERCISES

3x + 2 1
= 2 log √ +√ +C
9x2 + 12x + 3 9x2 + 12x + 3
3x + 3
= 2 log √ +C
2
9x + 12x + 3
(3x + 3)2
= log √ 2 +C
9x2 + 12x + 3
(3x + 3)2
= log +C
9x2 + 12x + 3
9(x + 1)2
= log +C
3(3x + 1)(x + 1)
3(x + 1)2
= log +C
(3x + 1)(x + 1)
3(x + 1)
= log +C
3x + 1
x+1
= log + log 3 + C
3x + 1

Since C is an arbitrary constant, we can write our final answer as

x+1
log +C
3x + 1
1.13.6. Solution. We see that we have two functions multiplied, but they don’t
simplify nicely with each other. However, if we differentiate logarithm, and in-
tegrate x2 , we’ll get a polynomial. So, let’s use integration by parts.

u = log x dv = x2 dx

du = (1/x)dx v = x3 /3
First, let’s antidifferentiate. We’ll deal with the limits of integration later.
Z Z
x log xdx = (log x) (x /3)) − (x3 /3) (1/x)dx
2 3
| {z } | {z } | {z } | {z }
u v v du
1 1
Z
= x3 log x − x2 dx
3 3
1 3 1 1
= x log x − · x3 + C
3 3 3
1 3 1
= x log x − x3 + C
3 9
We use the Fundamental Theorem of Calculus Part 2 to evaluate the definite
integral.
Z 2  2
3 1 3 1 3
x log xdx = x log x − x
1 3 9 1

911
S OLUTIONS TO E XERCISES
   
1 3 1 3 1 3 1 3
= 2 log 2 − 2 − 1 log 1 − 1
3 9 3 9
8 log 2 8 1
= − −0+
3 9 9
8 7
= log 2 −
3 9
1.13.7. *. Solution. The derivative of the denominator shows up in the nu-
merator, only differing by a constant, so we perform a substitution. Specifically,
substitute u = x2 − 3, du = 2x dx. This gives

x du/2 1 1
Z Z
2
dx = = log |u| + C = log x2 − 3 + C
x −3 u 2 2
1.13.8. *. Solution. (a) Although a quadratic under a square root often suggests
trigonometric substitution, in this case we have an easier substitution. Specifi-
cally, let y = 9 + x2 . Then dy = 2xdx, xdx = dy
2
, y(0) = 9, and y(4) = 25.
4 25 √
x 1 dy 1 y
Z Z 25
√ dx = √ = · =5−3=2
0 9 + x2 9 y 2 2 1/2 9

(b) The power of cosine is odd, so we can reserve one cosine for the differential
and change the rest to sines. Substituting y = sin x, dy = cos x, dx, y(0) = 0,
y(π/2) = 1, cos2 x = 1 − y 2 :
Z π/2 Z π/2
3 2
cos x sin x dx = cos2 x sin2 x cos x dx
0
Z0 1 Z 1
2 2
= (1 − y )y dy = (y 2 − y 4 ) dy
0 0
5 1
 3 
y y 1 1
= − = −
3 5 0 3 5
2
=
15
(c) The integrand is the product of two different kinds of functions, with no ob-
vious substitution or simplification. If we differentiate log x, it will match better
with the polynomial nature of the rest of the integrand. So, integrate by parts
with u(x) = log x and dv = x3 dx, then du = x1 dx and v = x4 /4.
e e Z e 4 Z e 3
x4 x 1 e4 x
Z
3
x log x dx = log x − · dx = − dx
1 4 1 1 4 x 4 1 4
e
e 4 x4 3e4 1
= − = +
4 16 1 16 16
1.13.9. *. Solution. (a) Integrate by parts with u = x and dv = sin x dx so that

912
S OLUTIONS TO E XERCISES

du = dx and v = − cos x.
Z Z
x sin x dx = −x cos x − (− cos x) dx = −x cos x + sin x + C

So
Z π/2 h iπ/2
x sin x dx = − x cos x + sin x =1
0 0

(b) The power of cosine is odd, so we can reserve one cosine for du and change
the rest into sines. Make the substitution u = sin x, du = cos x dx.
Z π/2 Z π/2
5

2
cos x dx = 1 − sin2 x cos x dx
0
Z0 1 Z 1
2 2
1 − 2u2 + u4 du



= 1 − u du =
0 0
 1
2 1 2 1 8
= u − u3 + u5 = 1 − + =
3 5 0 3 5 15
1.13.10. *. Solution. (a) This is a classic integration-by-parts example. If we
integrate ex , it doesn’t change, and if we differentiate x it becomes a constant. So,
let u = x and dv = ex dx, so that du = dx and v = ex .
Z 2 h i2 Z 2 h i2
xe dx = xex −
x
ex dx = 2e2 − ex = e2 + 1
0 0 0 0

(b) We have a quadratic function underneath a square root. In the absence of

an easier substitution, we can get rid of the square root with a trigonometric
substitution. Substitute x = tan y, dx =√sec2 y dy. When
p x = 0, tanpy = 0 so y = 0.
π
When x = 1, tan y = 1 so y = 4 . Also 1 + x = 1 + tan2 y = sec2 y = sec y,
2

since sec y ≥ 0 for all 0 ≤ y ≤ π4 .

1 π/4 π/4
1 sec2 y dy
Z Z Z
√ dx = = sec y dy
0 1 + x2 0 sec y 0
h iπ/4
= log | sec y + tan y|
0
π π
= log sec + tan − log |sec 0 + tan 0|
4 4
√ √
= log 2 + 1 − log |1 + 0| = log( 2 + 1)
Z 1
1 √
So, √ dx = log( 2 + 1)
0 1 + x2
(c) The integral is a rational function. In the absence of an obvious substitution,
we use partial fractions.
4x 4x
=
(x2 2
− 1)(x + 1) (x − 1)(x + 1)(x2 + 1)

913
S OLUTIONS TO E XERCISES

a b cx + d
= + + 2
x−1 x+1 x +1
Multiplying by the denominator,

4x = a(x + 1)(x2 + 1) + b(x − 1)(x2 + 1)

+ (cx + d)(x − 1)(x + 1) (∗)

Setting x = 1 gives 4a = 4, so a = 1. Setting x = −1 gives −4b = −4, so b = 1.

Substituting in a = b = 1 in (∗) gives:

4x = (x + 1)(x2 + 1) + (x − 1)(x2 + 1)
+ (cx + d)(x − 1)(x + 1)
2
4x = 2x(x + 1) + (cx + d)(x − 1)(x + 1)
4x − 2x(x2 + 1) = (cx + d)(x − 1)(x + 1)
−2x(x2 − 1) = (cx + d)(x2 − 1)
−2x = cx + d
c= −2, d = 0

So,
5 Z 5
4x 1 1 2x 
Z
dx = + − dx
3 (x2 − 1)(x2 + 1) 3 x − 1 x + 1 x2 + 1
h i5
= log |x − 1| + log |x + 1| − log(x2 + 1)
3
= log 4 + log 6 − log 26 − log 2 − log 4 + log 10
6 × 10 15
= log = log ≈ 0.1431
26 × 2 13
R3√
1.13.11. *. Solution. (a) 0 9 − x2 dx is the area of the portion of the disk
x2 + y 2 ≤ 9 that lies in the first quadrant. It is 14 π33 = 94 π . Alternatively, you
could also evaluate this integral using the substitution x = 3 sin y, dx = 3 cos y dy.
Z 3 √ Z π/2 q Z π/2
9 − x2 dx = 2
9 − 9 sin y (3 cos y) dy = 9 cos2 y dy
0 0 0
9 π/2
9h sin(2y) iπ/2
Z
= [1 + cos(2y)] dy = y +
2 0 2 2 0
9
= π
4

914
S OLUTIONS TO E XERCISES

√
y= 9 − x2

x
3

(b) It’s Rnot immediately obvious what to do with this one, but remember we
found log xdx using integration by parts with u = log x and dv = dx. Let’s
hope a similar trick works here. Integrate by parts, using u = log(1 + x2 ) and
2x
dv = dx, so that du = 1+x 2 dx, v = x.

1 1
2x
Z h i1 Z
2 2
log(1 + x ) dx = x log(1 + x ) − x dx
0 0 0 1 + x2
Z 1
x2
= log 2 − 2 2
dx
0 1+x
Z 1
1 
= log 2 − 2 1− dx
0 1 + x2
 1
= log 2 − 2 x − arctan x 0
π
= log 2 − 2 + ≈ 0.264
2
(c) The integrand is a rational function with no obvious substitution, so we use
partial fractions.
x a b c
2
= 2
+ +
(x − 1) (x − 2) (x − 1) x−1 x−2
a(x − 2) + b(x − 1)(x − 2) + c(x − 1)2
=
(x − 1)2 (x − 2)
Multiply by the denominator.

x = a(x − 2) + b(x − 1)(x − 2) + c(x − 1)2

Setting x = 1 gives a = −1. Setting x = 2 gives c = 2. Substituting in a = −1 and

c = 2 gives

b(x − 1)(x − 2) = x + (x − 2) − 2(x − 1)2

= −2x2 + 6x − 4 = −2(x − 1)(x − 2)
=⇒ b= −2

915
S OLUTIONS TO E XERCISES

Hence
∞
x
Z
dx
3 (x − 1)2 (x − 2)
Z M 
1 2 2
= lim − − + dx
M →∞ 3 (x − 1)2 x − 1 x − 2
 M
1
= lim − 2 log |x − 1| + 2 log |x − 2|
M →∞ x − 1
3
 M
1 x−2
= lim + 2 log
M →∞ x − 1 x−1 3
   
1 M −2 1 3−2
= lim + 2 log − + 2 log
M →∞ M − 1 M −1 3−1 3−1
1
= 2 log 2 − ≈ 0.886
2
since
M −2 1 − 2/M
lim log = lim log = log 1 = 0
M →∞ M − 1 M →∞ 1 − 1/M
1
and log = − log 2
2
1.13.12. Solution. This looks quite a lot like a rational function, but with vari-
able sin θ instead of x. So, we use the substitution x = sin θ, dx = cos θdθ.

sin4 θ − 5 sin3 θ + 4 sin2 θ + 10 sin θ

Z
cos θdθ
sin2 θ − 5 sin θ + 6
Z 4
x − 5x3 + 4x2 + 10x
= dx
x2 − 5x + 6
Since the numerator does not have smaller degree than the denominator, we need
to do some long division before we can set up our partial fractions decomposi-
tion.
x2 −2
2 4 3 2


x − 5x + 6 x − 5x + 4x + 10x
− x4 + 5x3 − 6x2
− 2x2 + 10x
2x2 − 10x + 12
12

That is,

x4 − 5x3 + 4x2 + 10x 12

2
= x2 − 2 + 2
x − 5x + 6 x − 5x + 6
12
= x2 − 2 +
(x − 2)(x − 3)

916
S OLUTIONS TO E XERCISES

We use partial fractions decomposition on the rightmost term.

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

Setting x = 3 and x = 2 gives us

B= 12, A = −12

Now we can evaluate our integral.

sin4 θ − 5 sin3 θ + 4 sin2 θ + 10 sin θ

Z
cos θdθ
sin2 θ − 5 sin θ + 6
Z 4
x − 5x3 + 4x2 + 10x
= dx
x2 − 5x + 6
Z  
2 12
= x −2+ dx
(x − 2)(x − 3)
Z  
2 12 12
= x − 2− + dx
x−2 x−3
1
= x3 − 2x − 12 log |x − 2| + 12 log |x − 3| + C
3
1 x−3
= x3 − 2x + 12 log +C
3 x−2
1 sin θ − 3
= sin3 θ − 2 sin θ + 12 log +C
3 sin θ − 2
1.13.13. *. Solution. (a) It doesn’t matter to us right now that the arguments of
sine and cosine are 2x rather than x. This is still the integral of powers of products
of sines and cosines. Since cosine has an odd power, we make the substitution
u = sin(2x), du = 2 cos(2x) dx.
Z π Z π
4 4
2 3
sin2 (2x) 1 − sin2 (2x) cos(2x) dx
 
sin (2x) cos (2x) dx =
0 0
1 1 2
Z
u 1 − u2 du

=
2 0
1 1 2 1 h 1 3 1 5 i1
Z
u − u4 du =


= u − u
2 0 2 3 5 0
1
=
15
(b) Make the substitution x = 3 tan t, dx = 3 sec2 t dt and use the trig identity
9 + 9 tan2 t = 9 sec2 t.

Z Z

3
2 −2

− 3
9+x dx = 9 + 9 tan2 t 2 3 sec2 t dt

917
S OLUTIONS TO E XERCISES
Z

−3
= 3 sec t 3 sec2 t dt
1 1
Z
= cos t dt = sin t + C
9 9
1 x
= √ +C
9 x2 + 9
To convert back to x, in the last step, we used the triangle below, which is rigged
to have tan t = x3 .

9
√ x2 +
x
t
3

(c) Seeing a rational function with no obvious substitutions, we use partial frac-
tions.
1 a bx + c a(x2 + 1) + (bx + c)(x − 1)
= + =
(x − 1)(x2 + 1) x − 1 x2 + 1 (x − 1)(x2 + 1)
Multiply by the original denominator.

1 = a(x2 + 1) + (bx + c)(x − 1) (∗)

Setting x = 1 gives 2a = 1 or a = 21 . Substituting in a = 1

2
in (∗) gives
1 2
(x + 1) + (bx + c)(x − 1) = 1
2
1 1
⇐⇒ (bx + c)(x − 1) = (1 − x2 ) = − (x − 1)(x + 1)
2 2
1
⇐⇒ (bx + c) = − (x + 1)
2
1
⇐⇒ b=c=−
2

So,
1
dx 1/2 (x + 1) i
Z Z h
2
= − 2 dx
(x − 1)(x2 + 1) x−1 x +1
To antidifferentiate the second piece, we split it into two integrals: one that can
be handled with the substitution u = x2 + 1, and another that looks like the
derivative of arctangent.

1/2 x/2 1/2 

Z 
= − 2 − 2 dx
x−1 x +1 x +1

918
S OLUTIONS TO E XERCISES

1/2 1 2x 1/2 
Z 
= − · 2 − 2 dx
x−1 4 x +1 x +1
1 1 1
= log |x − 1| − log(x2 + 1) − arctan x + C
2 4 2
(d) We know the derivative of arctangent, and it would integrate nicely if multi-
plied to the antiderivative of x. So, we integrate by parts with u = arctan x and
1 1 2
dv = x dx so that du = 1+x 2 dx and v = 2 x . Then

1 2 1 x2
Z Z
x arctan x dx = x arctan x − dx
2 2 1 + x2
1 2 1 1 + x2 − 1
Z
= x arctan x − dx
2 2 1 + x2
Z  
1 2 1 1
= x arctan x − 1− dx
2 2 1 + x2
1 2 
= x arctan x − x + arctan x + C
2
1.13.14. *. Solution. (a) We substitute y = sin(2x), dy = 2 cos(2x) dx. Note
sin(2 · 0) = 0 and sin(2 · π4 ) = 1.
π/4 1  1
dy 1 6 1
Z Z
5 5
sin (2x) cos(2x) dx = y = y =
0 0 2 12 0 12

(b) We can get rid of the square root with a trig substitution. Substituting x =
2 sin y, dx = 2 cos y dy,
Z √ Z q Z
2
4 − x dx = 4 − 4 sin y 2 cos y dy = 4 cos2 y dy
2

Z
 
=2 1 + cos(2y) dy

= 2y + sin(2y) + C = 2y + 2 sin y cos y + C

r
−1 x x2
= 2 sin +x 1− +C
2 4
p q
2 2
since sin y = 2 and cos y = 1 − sin y = 1 − x4 . Alternately, we can draw
x

a triangle with sin y = x2 , and use the Pythagorean theorem to find the adjacent
side.

2
x
y
√
4 − x2

(c) Seeing a rational function with no obvious substitution, we use the method of

919
S OLUTIONS TO E XERCISES

partial fractions. The denominator is already completely factored.

x+1 A B C
= + +
x2 (x − 1) x x2 x − 1
x + 1 = Ax(x − 1) + B(x − 1) + Cx2

Setting x = 1 gives us C = 2. Setting x = 0 gives us B = −1. Furthermore, the

coefficient of x2 on the left hand side (after collecting like terms), namely A + C,
must be the same as the coefficient of x2 on the right hand side, namely 0. So
A + C = 0 and A = −2. Checking,

−2x(x − 1) − (x − 1) + 2x2 = −2x2 + 2x − x + 1 + 2x2 = x + 1

as desired. Thus,
x+1 2 1 2 i
Z Z h
2
dx = − − 2+ dx
x (x − 1) x x x−1
1
= −2 log |x| + + 2 log |x − 1| + C
x
1.13.15. *. Solution. (a) Define
Z ∞ Z ∞
−x
I1 = e sin(2x) dx I2 = e−x cos(2x) dx
0 0

We integrate by parts, with u = sin(2x) or cos(2x) and dv = e−x dx. That is,
v = −e−x .
Z ∞ Z R
−x
I1 = e sin(2x) dx = lim e−x sin(2x) dx
0 R→∞ 0
h iR Z R 
−x −x
= lim − e sin(2x) + 2 e cos(2x) dx
R→∞ 0 0
= 2I2
Z ∞ Z R
−x
I2 = e cos(2x) dx = lim e−x cos(2x) dx
0 R→∞ 0
h iR Z R 
−x −x
= lim − e cos(2x) − 2 e sin(2x) dx
R→∞ 0 0
= 1 − 2I1
Z ∞
1 5 2
Substituting I2 = I1 into I2 = 1 − 2I1 gives I1 = 1, or e−x sin(2x) dx = .
2 2 0 5
(b) We
√ can cancel out
√ the square root if we use a trig substitution. Substitute
x = 2 tan y, dx = 2 sec2 y dy.
Z √2 √ Z π/4
1 sec2 y
dx = 2 dy
0 (2 + x2 )3/2 0 (2 + 2 tan2 y)3/2
π/4
1 π/4

1
Z
= cos y dy = sin y
2 0 2 0

920
S OLUTIONS TO E XERCISES

1
= √
2 2
(c)
• Solution 1: Integrate by parts, using u = log(1 + x2 ) and dv = x dx, so that
2x x2
du = 1+x 2, v = 2 .

Z 1 h1 i 1 Z 1 x3
2 2 2
x log(1 + x ) dx = x log(1 + x ) − 2
dy
0 2 0 0 1+x
Z 1h
1 x i
= log 2 − x− dx
2 0 1 + x2
1 h x2 1 i1
2
= log 2 − − log(1 + x )
2 2 2 0
1
= log 2 − ≈ 0.193
2

• Solution 2: First substitute y = 1 + x2 , dy = 2x dx.

Z 1
1 2
Z
2
x log(1 + x ) dx = log y dy
0 2 1
Then integrate by parts, using u = log y and dv = dy, so that du = y1 , v = y.
Z 1
1 2
Z
2
x log(1 + x ) dx = log y dy
0 2 1
h1 i2 1 Z 2 1 1
= y log y − y dy = log 2 −
2 1 2 1 y 2
≈ 0.193

(d) Seeing a rational function with no obvious substitution, we use partial frac-
tions.
1 a b c
2
= 2
+ +
(x − 1) (x − 2) (x − 1) x−1 x−2
1 = a(x − 2) + b(x − 1)(x − 2) + c(x − 1)2 (∗)
Setting x = 1 gives a = −1. Setting x = 2 gives c = 1. Substituting in a = −1 and
c = 1 to (∗) gives
b(x − 1)(x − 2) = 1 + (x − 2) − (x − 1)2
= −x2 + 3x − 2
= −(x − 1)(x − 2)
=⇒ b= −1
Hence:
∞
x
Z
dx
3 (x − 1)2 (x − 2)

921
S OLUTIONS TO E XERCISES

M
1 1 1 
Z 
= lim − − + dx
M →∞ 3 (x − 1)2 x − 1 x − 2
h 1 iM
= lim − log(x − 1) + log(x − 2)
M →∞ x − 1 3
h 1 M − 2i h 1 3 − 2i
= lim + log − + log
M →∞ M − 1 M −1 3−1 3−1
1
= log 2 − ≈ 0.193
2
since
M −2 1 − 2/M
lim log = lim log = log 1 = 0
M →∞ M − 1 M →∞ 1 − 1/M
1.13.16. *. Solution. (a) Integrate by parts with u = log x and dv = x dx, so that
du = dx
x
and v = 12 x2 .
1 2 1 1 1 1
Z Z
x log x dx = x log x − x2 · dx = x2 log x − x2 + C
2 2 x 2 4
(b) The denominator is an irreducible quadratic, so partial fractions can’t get us
any further. To integrate a function whose denominator is quadratic, we split the
numerator up so that one piece can be evaluated with a u-substitution, and the
other piece looks like arctangent.
(x − 1)dx x+2−3
Z Z
= dx
x2 + 4x + 5 x2 + 4x + 5
1 2x + 4 3
Z Z
= 2
dx − 2
dx
2 x + 4x + 5 x + 4x + 5
1 2x + 4 1
Z Z
= 2
dx − 3 dx
2 x + 4x + 5 (x + 2)2 + 1
1
= log[x2 + 4x + 5] − 3 arctan(x + 2) + C
2
For the last step, you can guess the antiderivative, or use the substitutions u1 =
x2 + 4x + 5 and u2 = x + 2, respectively, for the two integrals.
(c) We use partial fractions.
1 1 a b
2
= = +
x − 4x + 3 (x − 3)(x − 1) x−3 x−1
1 = a(x − 1) + b(x − 3)
Setting x = 3 gives a = 21 . Setting x = 1 gives b = − 12 . So,
dx 1/2 1/2 
Z Z 
= − dx
x2 − 4x + 3 x−3 x−1
1 1
= log |x − 3| − log |x − 1| + C
2 2
3 2
(d) Substitute y = x , dy = 3x dx.
Z 2
x dx 1 dy 1 1
Z
6
= 2
= arctan y + C = arctan x3 + C
1+x 3 1+y 3 3

922
S OLUTIONS TO E XERCISES

dx
1.13.17. *. Solution. (a) Integrate by parts with u = arctan x, dv = dx, du = 1+x2
and v = x. This gives
1 Z 1
x
Z
 1
arctan x dx = x arctan x 0 − 2
dx
0 0 1+x
h1 i1
2
= arctan 1 − log(1 + x )
2 0
π 1
= − log 2
4 2
(b) Note that the derivative of the denominator is 2x − 2, which differs from the
numerator only by 1.

2x − 1 2x − 2 1
Z Z Z
2
dx = 2
dx + 2
dx
x − 2x + 5 x − 2x + 5 x − 2x + 5
2x − 2 1
Z Z
= dx + dx
x2 − 2x + 5 (x − 1)2 + 4
1 x−1
= log |x2 − 2x + 5| + arctan +C
2 2
In the last step, you can guess the antiderivative, or use the substitutions u1 =
x2 − 2x + 5 and u2 = (x − 1)/2, respectively.
1.13.18. *. Solution. (a) Substituting u = x3 + 1, du = 3x2 dx

x2 1 du u−100 1
Z Z
dx = · = · +C
(x3 + 1)101 u101 3 −100 3
1
=− +C
300(x + 1)100
3

(b) Substituting u = sin x, du = cos x dx, cos2 x = 1 − sin2 x = 1 − u2 ,

Z Z Z
cos x sin x dx = cos x sin x cos x dx = (1 − u2 )u4 du
3 4 2 4

u5 u7
Z
= (u4 − u6 ) du = − +C
5 7
sin5 x sin7 x
= − +C
5 7
1.13.19. Solution. First, we note that the integral is improper, because sin π = 0.
So, we’ll have to use a limit.
Second, we need to antidifferentiate. The substitution u = sin x, du = cos xdx fits
just right.
Z π Z b Z sin b
cos x cos x 1
√ dx = lim− √ dx = lim− √ du
π/2 sin x b→π π/2 sin x b→π 1 u
h √ isin b √ √
= lim− 2 u = 2 0 − 2 1 = −2
b→π 1

923
S OLUTIONS TO E XERCISES

1.13.20. *. Solution. (a) If the integrand had x’s instead of ex ’s it would be a

rational function, ripe for the application of partial fractions. So let’s start by
making the substitution u = ex , du = ex dx:

ex du
Z Z
x x
dx =
(e + 1)(e − 3) (u + 1)(u − 3)

Now, we follow the partial fractions protocol, starting with expressing

1 A B
= +
(u + 1)(u − 3) u+1 u−3

To find A and B, the sneaky way, we cross multiply by the denominator

1 = A(u − 3) + B(u + 1)

and find A and B by evaluating at u = −1 and u = 3, respectively.

1
1 = A(−1 − 3) + B(−1 + 1) ⇐⇒ A = −
4
1
1 = A(3 − 3) + B(3 + 1) ⇐⇒ B =
4
Finally, we can do the integral:

ex du −1/4 1/4 
Z Z Z 
dx = = + du
(ex + 1)(ex − 3) (u + 1)(u − 3) u+1 u−3
1 1
= − log |u + 1| + log |u − 3| + C
4 4
1 1
= − log |ex + 1| + log |ex − 3| + C
4 4
(b) The argument of the square root is

12 + 4x − x2 = 12 − (x − 2)2 + 4 = 16 − (x − 2)2

Hmmm. The numerator is x2 − 4x + 4 = (x − 2)2 . So let’s make the integral look

somewhat simpler by substituting u = x − 2, du = dx. When x = 2 we have
u = 0, and when x = 4 we have u = 2, so:
Z x=4 Z u=2
x2 − 4x + 4 u2
√ dx = √ du
x=2 12 + 4x − x2 u=0 16 − u2
This is perfect for the trig substitution u = 4 sin θ, du = 4 cos(θ) dθ. When u = 0
we have 4 sin θ = 0 and hence θ = 0. When u = 2 we have 4 sin θ = 2 and hence
θ = π6 . So
u=2
u2 θ=π/6
16 sin2 θ
Z Z
√ du = p 4 cos θdθ
u=0 16 − u2 θ=0 16 − 16 sin2 θ

924
S OLUTIONS TO E XERCISES
Z π/6
= 16 sin2 θ dθ
0
Z π/6

=8 1 − cos(2θ) dθ
0
 π/6  √ 
1 π 1 3
= 8 θ − sin(2θ) =8 − ·
2 0 6 2 2
4π √
= −2 3
3
1.13.21. *. Solution. (a) Substituting y = cos x, dy = − sin x dx, sin2 x = 1 −
cos2 x = 1 − y 2

sin3 x sin2 x 1 − y2
Z Z Z
dx = sin x dx = (−dy)
cos3 x cos3 x y3
y −2
Z
y −3 − y −1 dy = −


=− + log |y| + C
−2
1
= sec2 x + log | cos x| + C
2
(b) The integrand is an even function, and the limits of integration are symmetric.
So, we can slightly simplify the integral by replacing the lower limit with 0, and
doubling the integral.
We’d rather not use partial fractions here, because it would be pretty compli-
cated. Instead, notice that the numerator is only off by a constant from the deriva-
tive of x5 . Substituting x5 = 4y, 5x4 dx = 4 dy, and using that x = 2 =⇒ 25 =
4y =⇒ y = 8,
Z 2 Z 2
x4 x4 4 8 1
Z
10
dx = 2 10
dx = 2 · 2
dy
−2 x + 16 0 x + 16 5 0 16y + 16
Z 8
1 1
= dy
10 0 y 2 + 1
1
= arctan 8 ≈ 0.1446
10
1.13.22. Solution.
• Solution 1: Let’s use the substitution u = x − 1, du = dx.
√ √
Z Z
x x − 1dx = (u + 1) udu
Z
u3/2 + u1/2 du


=
2 2
= u5/2 + u3/2 + C
5 3
2 2
= (x − 1)5/2 + (x − 1)3/2 + C
5 3
• Solution 2: We have an integrand with x multiplied by something inte-

925
S OLUTIONS TO E XERCISES
√
grable. So, if we use integration by parts with u = x and dv = x − 1dx,
then du = dx (that is, the x goes away) and v = 23 (x − 1)3/2 .
√ 2 √ 2
Z Z
3
x x − 1dx = x x − 1 − (x − 1)3/2 dx
3 3
√
 
2 3 2 2 5/2
= x x−1 − (x − 1) +C
3 3 5
2√
 
2 2
= x − 1 x(x − 1) − (x − 1) + C
3 5
2 √
= x − 1 · (3x2 − x − 2) + C
15
2√
= x − 1 · (3(x2 − 2x + 1) + 5x − 5) + C
15
2√
= x − 1 · (3(x − 1)2 + 5(x − 1)) + C
15
2 √ 5 2 √ 3
= ·3 x−1 + ·5 x−1 +C
15 15
2√ 5 2√ 3
= x−1 + x−1 +C
5 3
1.13.23. Solution. Z √
x2 − 2
dx
x2
We notice that there is a quadratic function under the square root. If that equation
were a perfect square, we could get rid of the square root: so we’ll mould it into
a perfect square using a trig substitution.
Our candidates will use one of the following identities:

1 − sin2 θ = cos2 θ tan2 θ + 1 = sec2 θ sec2 θ − 1 = tan2 θ

We’ll be substituting x =(something), so we notice that x2 − 2 has the general

form of (function)−(constant), as does sec2 θ − 1. In order to get the constant
right, we multiply through by two:

2 sec2 θ − 2 = 2 tan2 θ

or: √
( 2 sec θ)2 − 2 = 2 tan2 θ
so we decide to use the substitution
√ √
x = 2 sec θ dx = 2 sec θ tan θdθ

Now that we’ve chosen the substitution, we evaluate the integral.

Z √ 2 Z √
x −2 2 sec2 θ − 2 √
dx = 2 sec θ tan θdθ
x2 2 sec2 θ

926
S OLUTIONS TO E XERCISES
Z √
2 tan2 θ √
= 2 sec θ tan θdθ
2 sec2 θ
√
2 tan θ √
Z
= 2 sec θ tan θdθ
2 sec2 θ
tan2 θ
Z
= dθ
sec θ
sec2 θ − 1
Z
= dθ
sec θ
Z


= sec θ − cos θ dθ

= log | sec θ + tan θ| − sin θ + C

Now
√ we need everything back in terms of x. We need a triangle. Since x =
2 sec θ, that means if we label an angle θ, its secant (hypotenuse over adjacent
x √
side) is √ . By Pythagoras, the opposite side is x2 − 2.
2

x √
x2 − 2
θ
√
2
√ √
opp x2 − 2 opp x2 − 2
So tan θ = = √ , and sin θ = = . Then the value of the
adj 2 hyp x
integral is:

√ √
x x2 − 2 x2 − 2
log | sec θ + tan θ| − sin θ + C = log √ + √ − +C
2 2 x
√
√ √ x2 − 2
= log x + x2 − 2 − log 2 − +C
√ x
√ x2 − 2
= log x + x2 − 2 − +C
x

√ last step is due to our convention that C is an arbi-

Note the simplification in the
trary constant. So, C − log 2 can be re-written as simply C.
1.13.24. Solution. This is the product of secants and tangents, as in Section 1.8.2.
If u = tan x, then du = sec2 xdx. We can get the remaining two secants to turn
into tangents with the identity sec2 x = 1 + tan2 x, so we’ll use this substitution.

Z π/4 Z π/4
4 5
sec x tan xdx = sec2 x tan5 x sec2 xdx
0 0

927
S OLUTIONS TO E XERCISES
Z π/4
= (1 + tan2 x) tan5 x sec 2
| {zxdx}
0
du
Z tan(π/4)
= (1 + u2 )u5 du
tan(0)
Z 1
= (u + u7 )du
5
0
 1
1 6 1 8
= u + u
6 8 0
1 1 7
= + −0=
6 8 24
1.13.25. Solution. We can use partial fraction decomposition to break this into
chunks that we can deal with. The denominator has a repeated linear factor, so it
can be decomposed as the sum of constants divided by powers of that factor.

3x2 + 4x + 6 A B C
3
= + 2
+
(x + 1) x + 1 (x + 1) (x + 1)3
A(x + 1)2 + B(x + 1) + C
=
(x + 1)3
⇒ 3x2 + 4x + 6 = A(x + 1)2 + B(x + 1) + C
= Ax2 + (2A + B)x + (A + B + C)

So, by matching coefficients:

A = 3, 2A + B = 4, and A + B + C = 6
A = 3, B = −2, C = 5

Therefore:
3x2 + 4x + 6 3 −2 5
3
= + 2
+
(x + 1) x + 1 (x + 1) (x + 1)3

Now, the integration is easy, with a substitution of u = x + 1 and du = dx:

3x2 + 4x + 6
Z  
3 −2 5
Z
dx = + + dx
(x + 1)3 x + 1 (x + 1)2 (x + 1)3
Z
3u−1 − 2u−2 + 5u−3 du


=
5
= 3 log |u| + 2u−1 − u−2 + C
2
2 5
= 3 log |x + 1| + − +C
x + 1 2(x + 1)2
1.13.26. Solution. If the denominator were x2 + 1, the antiderivative would
be arctangent. So, by completing the square, let’s aim for the fraction to look like

928
S OLUTIONS TO E XERCISES

1
, for some u. This is a good strategy for integrating an irreducible quadratic
u2+1
under a constant.

First: complete the square

1 1 1
Z Z Z
dx = dx = dx
x2 + x + 1 x2 + x + 14 + 34 1 2 3


x+ 2
+ 4

Second: get the denominator in the form u2 + 1. To do this, we need to fix the
constant
! 
4
1
Z
3
= dx
1 2 3 4


x+ 2 + 4 3
4 1
Z
=
2 dx
3 4
· x+ 1 +1
3 2

Now a quick wiggle to make that first part of the denominator into something
squared again:

4 1
Z
= 2 dx
3

√2 x + √1 +1
3 3

2 1 2
Now we see that u = √ x + √ , du = √ dx will do the job
3 3 3
√
4 1 3 2 1
Z Z
= · du = √ du
3 u2 + 1 2 3 u2 + 1
2
= √ arctan u + C
3
 
2 2 1
= √ arctan √ x + √ +C
3 3 3
sin x
1.13.27. Solution. Since tan x = cos x
,

1
Z Z Z
2


sin x cos x tan xdx = sin xdx = 1 − cos(2x) dx
2
 
1 1
= x − sin(2x) + C
2 2
1
= (x − sin x cos x) + C
2

929
S OLUTIONS TO E XERCISES

1.13.28. Solution. We have the integral of a rational function with no obvious

substitution, so we use partial fractions. That means we need to factor the de-
nominator. We see that x = −1 is a root of the denominator, so x + 1 is a factor.
You might be able to figure out the rest of the factorization by inspection, or from
having seen this common expression before; alternately, we can use long divi-
sion.
x2 − x + 1
3


x+1 x +1
− x3 − x2
− x2
x2 + x
x+1
−x−1
0

Note x2 − x + 1 is an irreducible quadratic.

1 1 A Bx + C
= = + 2
x3 +1 2
(x + 1)(x − x + 1) x+1 x −x+1
2
1 = A(x − x + 1) + (Bx + C)(x + 1) (∗)
1
When x = −1, we see 1 = 3A, so 3
= A. We plug this into (∗).

1
1 = (x2 − x + 1) + (Bx + C)(x + 1)
3
1 2 1 2
− x + x + = Bx2 + (B + C)x + C
3 3 3
Matching up coefficients of corresponding power of x, we see B = − 13 and C = 32 .
1
x − 23
Z  
1 1/3
Z
dx = − 23 dx
x3 + 1 x+1 x −x+1

To integrate the second fraction, we break it up into two pieces: one we can inte-
grate using the substitution u = x2 − x + 1, the other will look like the derivative
of arctangent.
1
1 x − 16
− 12
Z
3
= log |x + 1| − dx
3 x2 − x
+1
1 1 2x − 1 1 1
Z Z
= log |x + 1| − dx + dx
3 6 x2 − x + 1 2 (x − 2 )2 + 34
1

1 1 1 1
Z
2
= log |x + 1| − log |x − x + 1| +   dx
3 6 2 3 2x−1
2
4
√
3
+1
1 1 2 1
Z
= log |x + 1| − log |x2 − x + 1| + 2 dx
3 6 3

2x−1
√
3
+1

930
S OLUTIONS TO E XERCISES

2x−1 √2 dx.
Let u = √ ,
3
du = 3

1 1 1 1
Z
= log |x + 1| − log |x2 − x + 1| + √ dx
3 6 3 u2 + 1
 
1 1 1 2x − 1
= log |x + 1| − log |x2 − x + 1| + √ arctan √ +C
3 6 3 3
1.13.29. Solution. By process of elimination, we decide to use integration by
parts. We won’t get anything better by antidifferentiating arcsine, so let’s plan
on differentiating it:

u = arcsin x dv = (3x)2 dx
1
du = √ dx v = 3x3
1 − x2

1
Z Z
2 3
(3x) arcsin xdx = |arcsin
{z x} · |{z}
3x − 3x3 · √ dx
|{z} 1 − x 2
u v v | {z }
du
3x3
Z
= 3x3 arcsin x − √ dx
1 − x2
So: we’ve gotten rid of the ugly pairing of arcsine with a polynomial, but now
we’re in another pickle. From here, two options present themselves. We could
use the substitution u = 1 − x2 , or we could use a trig substitution.

• Option 1: Let u = 1 − x2 . Then − 12 du = dx, and x2 = 1 − u.

3x3
Z Z
2 3
(3x) arcsin xdx = 3x arcsin x − √
dx
1 − x2
x2
Z
3
= 3x arcsin x − 3 √ · xdx
1 − x 2

3 1−u
Z
3
= 3x arcsin x + √ du
2 u
3
Z
= 3x3 arcsin x + u−1/2 − u1/2 du


2
 
3 3 1/2 2 3/2
= 3x arcsin x + 2u − u +C
2 3
√ √ 3
= 3x3 arcsin x + 3 1 − x2 − 1 − x2 + C
√ √
• Option 2: If we let x = sin θ, then 1 − x2 = cos2 θ = cos θ. So let’s use the
substitution x = sin θ, dx = cos θdθ.

3x3
Z Z
2 3
(3x) arcsin xdx = 3x arcsin x − √ dx
1 − x2

931
S OLUTIONS TO E XERCISES

3 sin3 θ
Z
3
= 3x arcsin x − p cos θdθ
1 − sin2 θ
Z
3
= 3x arcsin x − 3 sin3 θdθ

And now: a substitution from Section 1.8.1, u = cos x and du = − sin xdx
Z Z
3x arcsin x − 3 sin θdθ = 3x arcsin x − 3 sin2 θ sin θdθ
3 3 3

Z
= 3x arcsin x − 3 (1 − cos2 θ) sin θdθ
3

Z
= 3x arcsin x + 3 (1 − u2 )du
3

 
3 1 3
= 3x arcsin x + 3 u − u + C
3
3 3
= 3x arcsin x + 3u − u + C
= 3x3 arcsin x + 3 cos θ − cos3 θ + C

Recall x = sin θ; so we draw a triangle with angle √ θ, opposite side x,

hypotenuse
√ 1. Then by Pythagoras, adjacent side is 1 − x2 , so cos θ =
1 − x2 .
Z
(3x)2 arcsin xdx
√
= 3x3 arcsin x + 3 1 − x2 − (1 − x2 )3/2 + C

1
x
θ
√
1 − x2

Exercises — Stage 3
1.13.30. Solution. We would like to not have that square root there. Luck-
ily, there’s a way of turning cosine into cosine squared: the identity cos(2x) =
2 cos2 x − 1. If we take 2x = t, then cos t = 2 cos2 (t/2) − 1.
Z π/2 √
Z π/2 p √ Z π/2
cos t + 1 dt = 2 cos2 (t/2)dt = 2 | cos(t/2)|dt
0 0 0

Over the interval [0, π2 ], cos(t/2) > 0, so we can drop the absolute values.
 π/2
√ Z π/2 √

t
= 2 cos(t/2)dt = 2 2 sin
0 2 0

932
S OLUTIONS TO E XERCISES
√ π 
= 2 2 sin =2
4
1.13.31. Solution.
√ 1
• Solution 1: Using logarithm rules, log x = log x1/2 =


2
log x, so we can
simplify: Z e √ Z e
log x log x
dx = dx
1 x 1 2x
We use the substitution u = log x, du = x1 dx:
Z e
log x 1 e 1
Z
dx = log(x) · dx
1 2x 2 1 | {z } |x{z }
u
du
log(e)
1
Z
= u du
2 log(1)
Z 1
1
= u du
2 0
 1
1 1 2
= u
2 2 0
 
1 1 1
= −0 =
2 2 4

√ du 1 1 1
• Solution 2: We use the substitution u = log x. Then =√ · √ = ,
dx x 2 x 2x
1
hence 2du = dx. This fits our integral nicely!
x
Z e √ Z log √e
log x
dx = √
u · 2du
1 x log 1
h i1/2
= u2
0
 2
1 1
= − 02 =
2 4
1.13.32. Solution.
0.2
tan x
Z
dx
0.1 log(cos x)
It might not be immediately obvious how to proceed on this one, so this is an-
other example of an integral where you should not be discouraged by finding
methods that don’t work. One thing that’s worked for us in the past is to use a
u-substitution with the denominator. With that in mind, let’s find the derivative
of the denominator.
d 1 − sin x
{log(cos x)} = · (− sin x) = = − tan x
dx cos x cos x

933
S OLUTIONS TO E XERCISES

So, if we let u = log(cos x), we see −du = tan xdx, which will work for a substitu-
tion.

0.2 log(cos(0.2))
tan x −du
Z Z
dx =
0.1 log(cos x) log(cos(0.1)) u
h ilog(cos(0.2))
= − log |u|
log(cos(0.1))

= − log | log(cos 0.2)| + log | log(cos 0.1)|

log(cos(0.1))
= log
log(cos(0.2))
 
log(cos(0.1))
= log
log(cos(0.2))

Things to notice: the integrand is only defined when log(cos x) exists AND is
nonzero. So, for instance, it is not defined when x = 0, because then log cos x =
log 1 = 0, and we can’t divide by zero.
In the final simplification, since 0.1 and 0.2 are between 0 and π/2, the cosine term
is positive but less than one, so log(cos 0.1) and log(cos 0.2) are both negative; then
their quotient is positive, so we can drop the absolute value signs.
Using the base change
formula, we can also write the final answer as
log logcos(0.2) cos(0.1) .

1.13.33. *. Solution. (a) Without any other ideas, we see we have a compound
function—a function of a function. We often find it useful to substitute for the
“inside” function. So, we substitute u = log x, du = x1 dx. Then dx = x du = eu du.
Z Z
sin(log x) dx = sin(u) eu du

We have already seen, in Example 1.7.11, that

1
Z
sin(u) eu du = eu sin u − cos u + C


2
So,
1 
Z

sin(log x) dx = x sin(log x) − cos(log x) + C
2
(b) The integrand is of the form N (x)/D(x) with N (x) of lower degree than D(x).
So we factor D(x) = (x − 2)(x − 3) and look for a partial fractions decomposition:

1 A B
= + .
(x − 2)(x − 3) x−2 x−3

Multiplying through by the denominator yields

1 = A(x − 3) + B(x − 2)

934
S OLUTIONS TO E XERCISES

Setting x = 2 we find:

1 = A(2 − 3) + 0 =⇒ A = −1

Setting x = 3 we find:

1 = 0 + B(3 − 2) =⇒ B = 1

So we have found that A = −1 and B = 1. Therefore

Z  
1 1 1
Z
dx = − dx
(x − 2)(x − 3) x−3 x−2
= log |x − 3| − log |x − 2| + C

and the definite integral

1
1
Z h i1
dx = log |x − 3| − log |x − 2|
0 (x − 2)(x − 3) 0
   
= log 2 − log 1 − log 3 − log 2
4
= 2 log 2 − log 3 = log
3
1.13.34. *. Solution. (a) If we expand the integrand, one part of it is quite
familiar—a portion of a circle. So, we split the specified integral in two.
Z 3 √ Z 3 √ Z 3 √
(x + 1) 9 − x2 dx = 9− x2 dx + x 9 − x2 dx
0 0 0

The
√ first piece2 represents the area above the x-axis and below the curve y =
9 − x2 , i.e. x + y 2 = 9, with 0 ≤ x ≤ 3. That’s the area of one quadrant of a disk
of radius 3. So Z 3√
1 9
9 − x2 dx = (π · 32 ) = π
0 4 4
For the second part, we substitute u = 9 − x2 , du = −2x dx. Note u(0) = 9 and
u(3) = 0. So,
0
3 √ 0 √ du 1 u3/2
  
1 27
Z Z
x 9 − x2 dx = u =− =− − =9
0 9 −2 2 3/2 9 2 3/2

All together,
Z 3 √ 9
(x + 1) 9 − x2 dx = π + 9
0 4
(b) The integrand is of the form N (x)/D(x) with D(x) already factored and N (x)
of lower degree. We immediately look for a partial fractions decomposition:

4x + 8 A Bx + C
2
= + 2 .
(x − 2)(x + 4) x−2 x +4

935
S OLUTIONS TO E XERCISES

Multiplying through by the denominator yields

4x + 8 = A(x2 + 4) + (Bx + C)(x − 2) (∗)

Setting x = 2 we find:

8 + 8 = A(4 + 4) + 0 =⇒ 16 = 8A =⇒ A = 2

Substituting A = 2 in (∗) gives

4x + 8 = A(x2 + 4) + (Bx + C)(x − 2)

=⇒ −2x2 + 4x = (x − 2)(Bx + C)
=⇒ (−2x)(x − 2) = (Bx + C)(x − 2)
=⇒ B= −2, C = 0

So we have found that A = 2, B = −2, and C = 0. Therefore

Z  
4x + 8 2 2x
Z
dx = − dx
(x − 2)(x2 + 4) x − 2 x2 + 4
= 2 log |x − 2| − log(x2 + 4) + C

Here the second integral was found just by guessing an antiderivative. Alterna-
tively, one could use the substitution u = x2 + 4, du = 2x dx.
(c) The given integral is improper, but only because of its infinite limits of inte-
gration. (The integrand is continuous for all real numbers.) So, we’ll have to take
two limits. Before we do that, though, let’s find the antiderivative. We would
like to use the substitution u = ex , du = ex dx. That is, u1 du = dx.

1 1 1
Z Z Z
dx = 1 du = = arctan u + C
ex + e−x u(u + u ) u2 + 1
= arctan(ex ) + C

Now we can deal with the limits of integration.

∞ Z 0 Z ∞
1 1 1
Z
−x
dx = −x
dx + dx
x
−∞ e + e
x
−∞ e + e 0 ex + e−x
Z 0  Z b 
1 1
= lim x −x
dx + lim x −x
dx
a→−∞ a e +e b→∞ 0 e +e
0 b
= lim arctan(ex ) a + lim arctan(ex ) 0
 
a→−∞ b→∞

= lim arctan(e ) − arctan(ea ) + lim arctan(eb ) − arctan(e0 )

0
   
a→−∞ b→∞

= lim − arctan(e ) + lim arctan(eb )

a
   
a→−∞ b→∞
π π
= − arctan(0) + =
2 2

936
S OLUTIONS TO E XERCISES

1.13.35. Solution. It’s not immediately clear where to start, but a common
method we’ve seen is to use the denominator in a u-substitution, especially when
square roots
√ are involved.1
Let u = 1 − x, du = − 2√1−x dx. Then u2 = 1 − x, so x = 1 − u2 .

Z r √ Z √
x x
Z
dx = 2 √ dx = −2 1 − u2 du
1−x 2 1−x
Now we’re back in familiar territory. Let u = sin θ, du = cos θdθ.
Z p
= −2 1 − sin2 θ cos θdθ
Z
= −2 cos2 θdθ
Z


=− 1 + cos(2θ) dθ
1
= −θ − sin(2θ) + C
2
= −θ − sin θ cos θ + C
√
= − arcsin u − u 1 − u2 + C (∗)
√ √ √
= − arcsin( 1 − x) − 1 − x x + C

1
u
θ
√
1 − u2

In (∗), to convert from θ to u, our substitution u = sin θ tells us θ = arcsin u. To find

cos
√ θ, we can either trace our work backwards to see that we already simplified
1 − u2 into cos θ, or we can draw a right triangle with angle θ and sin θ = u,
then use the Pythagorean theorem to find the length of the adjacent side of the
triangle and cos θ.
1.13.36. Solution. Let’s use the substitution u = ex . There are a few reasons to
think this is a good choice. It’s an “inside function,” in that if we let f (x) = ex ,
x
then f (ex ) = ee , which is a piece of our integrand. Also its derivative, ex , is
multiplied by the rest of the integrand, since e2x = ex · ex .
Let u = ex , du = ex dx. When x = 0, u = 1, and when x = 1, u = e.
Z 1 Z 1 Z e
2x ex x ex x
e e dx = e e e dx = ueu du
0 0 1

937
S OLUTIONS TO E XERCISES

This is more familiar. We use integration by parts with dv = eu du, v = eu . Con-

veniently, the “u” we brought in with the substitution is what we want to use
for the “u” in integration by parts, so we don’t have to change the names of our
variables.
Z e
 u e
= ue 1 − eu du
1
= e · e − e − ee + e = ee (e − 1)
e

1.13.37. Solution. The substitution u = x + 1 looks promising at first, but

doesn’t result in something easily integrable. We can’t use partial fractions be-
cause our integration isn’t rational. This doesn’t look like something from the
trig-substitution family. So, let’s think about integration by parts. There’s a lot of
different ways we couldbreak up the integrand into two parts.
 For  example, we
x x ex
could view it as (x+1)2
ex , or we could view it as x+1 x+1
. After some
−2
trial and error, we settle on u = xe and dv = (x + 1) dx. Then du = ex (x + 1)
x
−1
and v = x+1 .

xex xex
Z x
e (x + 1)
Z
2
dx = − + dx
(x + 1) x+1 x+1
xex
Z
=− + ex dx
x+1
xex
=− + ex + C
x+1
ex
= +C
x+1
1.13.38. Solution. It would be niceR to use integration by parts with u = x,
because then we would integrate vdu, and du = dx. That is, the x would go
sin x
away, and we’d be left with a pure trig integral. If we use u = x, then dv = cos 2 x.

We need to find v:
sin x
Z Z
v= dx = tan x sec xdx = sec x
cos2 x
Now we use integration by parts.

x sin x
Z Z
dx = x sec x − sec xdx = x sec x − log | sec x + tan x| + C
cos2 x
1.13.39. Solution. If the unknown exponent gives you the jitters, think about
what this looks like in easier cases. If n is a whole number, the integrand is a
polynomial. Not so scary, right? However, it’s a little complicated to expand.
(You can do it using the very handy binomial theorem.) Let’s think of an easier
way.
If we had simply the variable x raised to the power n, rather than the binomial
x + a, that might be nicer. So, let’s use the substitution u = x + a, du = dx. Note

938
S OLUTIONS TO E XERCISES

x = u − a.
Z Z Z
n n
un+1 − aun du


x(x + a) dx = (u − a)u dx =

Now, if n 6= −1 and n 6= −2, we can just use the power rule:

u(n+2) un+1
= −a +C
n+2 n+1
(x + a)(n+2) (x + a)n+1
= −a +C
n+2 n+1
If n = −1, then
a
Z Z Z 
n n+1 n


x(x + a) dx = u − au du = du 1−
u
= u − a log |u| + C = (x + a) − a log |x + a| + C

If n = −2, then
Z 

1
Z Z
n n+1 n −2


x(x + a) dx = u − au du = − au du
u
a a
= log |u| + + C = log |x + a| + +C
u x+a
All together,

(x+a)(n+2) (x+a)n+1
 n+2 − a n+1 + C if n 6= −1, −2
Z 

n
x(x + a) dx = (x + a) − a log |a + x| + C if n = −1

log |x + a| + a + C

if n = −2
x+a

1.13.40. Solution. We’ve seen how to antidifferentiate arctan x: integration by

parts. Let’s hope the same thing will work here.
Step 1: integration by parts.
Let u = arctan(x2 ) and dv = dx. Then du = x42x+1 du and v = x.

2x2
Z Z
2 2
arctan(x )dx = x arctan(x ) − dx
x4 + 1
Now we have a rational function. There’s no obvious substitution, but we can
use partial fractions. The degree of the numerator is strictly less than the degree
of the denominator, so we don’t need to long divide first. We do, however, need
to factor the denominator. It’s a common function, so you might already know
the factorization, or you might be able to guess it. Below, we show another way
to find the factorization, similar to the method of partial fractions.
Step 2: factor x4 + 1. For any real x, note x4 +1 > 0. Since it has no roots, it has no
linear factors. That means it factors as the product of two irreducible quadratics.

939
S OLUTIONS TO E XERCISES

That is,
x4 + 1 = (ax2 + bx + c)(dx2 + ex + f )
Since the coefficient of x4 on the left-hand is 1, we may assume a = d = 1.
x4 + 1 = (x2 + bx + c)(x2 + ex + f )
Since the constant term is 1, cf = 1. That is, f = 1c .
x4 + 1 = (x2 + bx + c)(x2 + ex + 1/c)
   
4 3 1 2 b
= x + (b + e) x + + be + c x + + ec x + 1
c c
| {z } | {z } | {z }
(1) (3) (2)

1 The coefficient of x3 tells us e = −b.

2 Then the coefficient of x tells us 0 = cb + ec = cb − bc. So, c = 1c , hence c = ±1.
3 Finally, the coefficient of x2 tells us 0 = 1c + be + c = 1c − b2 + c. Since −b2
is negative (or zero), 1c + c is positive, so c = 1. That is, 0 = 1 − b2 + 1. So,
√
b = 2.
All together, √ √
x4 + 1 = (x2 + 2x + 1)(x2 − 2x + 1)
Step 3: partial fraction decomposition.
Now that we have the denominator factored into irreducible quadratics, we can
find the partial fraction decomposition of the integrand.
2x2 Ax + B Cx + D
4
= √ + √
x +1 2 2
x + 2x + 1 x − 2x + 1
√ √
2x2 = (Ax + B)(x2 − 2x + 1) + (Cx + D)(x2 + 2x + 1)
√ √
= (A + C)x3 + (B + D − 2A + 2C)x2
√ √
+ (A + C − 2B + 2D)x + (B + D)
From the coefficient of x3 , we see C = −A.
√ √ √
2x2 = (B + D − 2 2A)x2 + (− 2B + 2D)x + (B + D)
From the constant term, we see D = −B.
√ √
2x2 = (−2 2A)x2 + (−2 2B)x
2
√ √
From the coefficient
√ of x , we see −2 2A = 2, so A = −1/ 2. Since C = −A,
then C = 1/ 2.
From the coefficient of x, we see B = 0. Since D = −B, also D = 0.
Step 4: integration.
√ √ !
2x2 (−1/ 2)x (1/ 2)x
Z Z
dx = √ + √ dx
x4 + 1 x2 + 2x + 1 x2 − 2x + 1

940
S OLUTIONS TO E XERCISES
Z  
1 −x x
=√ √ + √ dx
2 x2 + 2x + 1 x2 − 2x + 1
To integrate, we want to break the fractions √ into two pieces each:
√
one we can
2
integrate with a substitution u = x ± 2x + 1 , du = 2x ± 2 dx (shown in
blue), and one that looks like the derivative of arctangent (shown in red).
√ √ √ √ !
1 −x − 22 + 22 x − 22 + 22
Z
=√ √ + √ dx
2 x2 + 2x + 1 x2 − 2x + 1
Z  1 √ √
2
1 − 2 (2x + 2) 2
=√ √ + √
2 x2 + 2x + 1 x2 + 2x + 1
1
√ √
2
(2x − 2)

+ 2 √ + √2 dx
x2 − 2x + 1 x2 − 2x + 1
√
√ 2
1 1
Z
2 2
= √ − log x + 2x + 1 + √ dx
2 2 x2 + 2x + 1
√ !
√ 2
1
Z
2
+ log |x − 2x + 1| + √2 dx
2 x2 − 2x + 1

We use logarithm rules to compress our work. In order to evaluate the remaining
integrals, we complete the squares of the denominators.
√
√ 2
1 1
Z
2 −√2x+1 2
=√ log xx2 + 2x+1
+ 2 dx
2 2

1 1
x+ 2 + 2
√
√ !
Z 2
2
+  2 dx
1 1
x− 2 + 2
√

√
1 1 √ 2
Z
2 −√2x+1
=√ log xx2 + 2x+1
+ √
2 dx
2 2 2x + 1 + 1
√ !
2
Z
+ √
2 dx
2x − 1 + 1
Now, we can either guess√ the antiderivatives of the remaining integrals, or use
the substitutions u = ( 2x ± 1).
!
1 1 √
2 − 2x+1
 √   √ 
=√ log xx2 + √
2x+1
+ arctan 2x+1 + arctan 2x−1 +C
2 2
Step 5: finishing touches.
Finally, we can put our work together. (Remember way back in Step 1, we used
integration by parts.)
2x2
Z Z
2 2
arctan(x )dx = x arctan(x ) − dx
x4 − 1

941
S OLUTIONS TO E XERCISES

1 1 √
x2 −√2x+1
= x arctan(x2 ) − √ log 2
x + 2x+1
2 2
!
√  √ 
+ arctan 2x + 1 + arctan 2x − 1 +C

Remark: although this integral calculation was longer than average, it didn’t use
any new ideas (except for the factoring of x4 + 1 mentioned in the hint). It’s good
exercise to apply familiar techniques in challenging situations, to deepen your
mastery.

2 · Applications of Integration
2.1 · Work
2.1.2 · Exercises

Exercises — Stage 1
2.1.2.1. Solution. Force is mass × acceleration (with acceleration equal to g in
this problem), and both in this scenario are constant, so we don’t need an integral
— only a product — to calculate the force acting on the block.
To find the force in newtons, recall one newton is one kg·m
sec2
, so we need the mass
3
of our block in kg. Specifically, our block has mass 1000 kg. So, the force involved
is   
3 m  kg · m
F = kg × 9.8 2 = 0.0294 = 0.0294N
1000 sec sec2
To find the work in joules, recall one joule is one newton-metre: that is, one
newton of force acting over one metre. So, we need our distance in metres.
 

1
W = 0.0294N × m = 0.00294N · m = 0.00294J
10
2.1.2.2. Solution. The force of the rock is one newton, or one kilogram-metre
per second squared, so

kg · m  m 
1 = (xkg) 9.8 2
sec2 sec
1
Therefore, the mass of the rock is 9.8 kg, or about 102 grams.
Now, since one joule is one newton-metre, the amount of work required to coun-
teract 1 N of gravitational force for one metre is precisely one joule.
Remark: having an idea of how much work a joule is, and how much force a
newton is, is a good tool for checking the reasonableness of your work. For
example, after this question, if you calculate that a marble weighs 100 N, you
can be pretty sure there’s an error in your calculation.
2.1.2.3. Solution.
b−a
a We defined ∆x = n
: that is, the length of one interval, when we chop [a, b]

942
S OLUTIONS TO E XERCISES

into n of them. If b and a are measured in metres, then ∆x is measured in

metres as well. So, the units of ∆x are metres.
Put another way, since a and b both describe a quantity in metres, b − a
describes a quantity in metres as well. (When we add or subtract quantities
of the same units, their sum or difference is given in the same units.) Since
n is a unitless quantity (simply a number: not “n kg” or “n m”), b−a n
still
describes a quantity in metres. (If I have 6 metres of cloth, and I cut it into
3 pieces, each piece has 36 = 2 metres — not 2 kilograms, or 2 metres per
second.)

b Since F (x) is measured in kilogram-metres per second squared (newtons),

the units of F (xi ) are kilogram-metres per second squared (newtons).

c W is calculated by adding up summands of the form F (xi )∆x. The units of

F (xi )∆x are the productsof the
 units of F (xi ) with the units of ∆x. That is,
2
the units of F (xi )∆x are kg·msec2
(m) = kg·m
sec2
= J. The sum of terms given in
joules is itself given in joules, so the units of W are joules.
Rb
2.1.2.4. Solution. As we saw in Question 3, the units of a f (x) dx are simply the
units of the integrand, f (x), multiplied by the units of the variable of integration,
smoot·barn
x. In this case, that yields megaFonzie (that is, smoot-barns per megaFonzie).

2.1.2.5. Solution. Hooke’s law says that the force required to stretch a spring x
units past its natural length is proportional to x; that is, there is some constant k
associated with the individual spring such that the force required to stretch it x
m past its natural length is kx.

• Solution 1: Since the force required to stretch the spring is proportional to

the amount stretched, and the force acting on the spring is proportional to
the mass hanging from it, we conclude the amount the spring stretches is
proportional to the mass hung from it. So, if 1 kg stretches it 1 cm, then 10
kg will stretch it 10 cm. We should mark the wall 10 cm below the bottom
of the spring as it hangs unloaded.

• Solution 2: We can find k from the test with the bag of water. The force
exerted by the bag of water was (1 kg)(9.8 m/sec2 ) = 9.8 N= k(1 cm). So,

9.8 kg·m
sec2 kg
k= = 980 2
0.01 m sec

If we hang 10 kg from the spring, gravity exerts a force of

(10 kg)(9.8 m/sec2 ) = 98 kg·m
sec2
. This will be matched by the spring with a
force of kx newtons, where k is the spring constant and x is the amount
stretched.
kg · m
kx = 98
sec2

943
S OLUTIONS TO E XERCISES
 
kg kg · m
980 2 (x m) = 98
sec sec2
1
x= m = 10 cm
10
So, we should put the mark at 10 cm below the natural length of the spring.
2.1.2.6. Solution. Definition 2.1.1 tells us the work done by the force is W (b) =
Rb
1
F (x)dx, where F (x) is the force on the object at position x. So, by the Funda-
mental Theorem of Calculus Part 1,
Z b 
d d
{W (b)} = F (x)dx = F (b)
db db 1
d  3
−b + 6b2 − 9b + 4 = F (b)
db
−3b2 + 12b − 9 = F (b)
−3(b − 1)(b − 3) = F (b)

So, F (x) is the quadratic polynomial −3(x − 1)(x − 3).

y = F (x)

x
1 2 3

The largest absolute value of F (x) over [1, 3] occurs at x = 2. At this point, we
have our strongest force.

Exercises — Stage 2
2.1.2.7. *. Solution. By Definition 2.1.1, the work done in moving the object
from x = 1 meters to x = 16 meters by the force F (x) is
16 16 h √ ix=16
a
Z Z
W = F (x)dx = √ dx = 2a x = 6a
1 1 x x=1

To have W = 18, we need a = 3.

As a side remark, F (x) = √ax should have units Newtons. Since x, a distance, is
√
measured in meters, a has to have the bizarre units newton- meters.
2.1.2.8. Solution.
c
a Since `−x
is measured in newtons, and ` and x (and therefore ` − x) are

944
S OLUTIONS TO E XERCISES

measured in metres, the units of c are newton-metres, i.e. joules.

b Following Definition 2.1.1, the work done compressing the air is

Z 1.5
W = F (x) dx
1

where F (x) is the amount of force applied when the plunger is x metres past
its natural position. The amount of force applied is equal in magnitude to
c
the amount of force supplied by the tube: `−x N. Note ` and c are constants.
We can guess the antiderivative, or use the substitution u = ` − x, du =
−x dx.
Z 1.5
c  1.5
W = dx = − c log |` − x| 1
1 `−x
 
= −c log |` − 1.5| − log |` − 1|
 
` − 1.5
= −c log
`−1
 
`−1
= c log J
` − 1.5

Note that, because ` > 1.5, the argument of logarithm is positive, so we

`−1
don’t need the absolute value signs. Furthermore, `−1 > `−1.5, so `−1.5 > 1,
`−1


hence log `−1.5 > 0.

2.1.2.9. *. Solution. By Hooke’s Law, the force exerted by the spring at dis-
placement x m from its natural length is F = kx, where k is the spring con-
stant. Measuring distance in meters and force in newtons (since one joule is one
newton-metre), the total work is
Z 0.1 m  0.1 m
1 2 1 1
kx dx = kx = · 50 · (0.1)2 = J.
0 2 0 2 |{z} | {z } 4
N/m m2

Note the units of the integrand (kx) are newtons, and the units of the variable of
integration, x, are metres. So, the evaluated integral has units newton-metres, or
joules.
2.1.2.10. *. Solution. First note that newtons and joules are SI units with one
joule equal to one newton-metre, so we should measure distances in meters
rather than centimeters. Next recall that a spring with spring constant k exerts a
force F (x) = kx when the spring is stretched x m beyond its natural length. So
in this case (0.05 m)(k) = 10 N, or k = 200 N/m. The work done is:
Z 0.5 m Z 0.5 h i0.5
F (x)dx = 200xdx = 100x2 = 25 J
0 0 0

Note the units of the integrand (F (x) = kx = 200x) are newtons (k is given in

945
S OLUTIONS TO E XERCISES

N/m, and x is given in m). The units of the variable of integration, x are metres.
So, the evaluated integral has units newton-metres, or joules.

2.1.2.11. *. Solution. Note that the cable has mass density 85 kg/m. When
the bucket is at height y, the cable

that remains to be lifted has length (5 − y)
m and mass 85 (5 − y) = 8 1 − y5 kg. So,
at height y, the cable is subject to a
y
downward gravitational force of 8 1 − 5 · 9.8
N; to raise the cable we need to
apply a compensating upward force of 8 1 − y5 · 9.8 N. So, the work required is
5 5
y2 

y
Z 
8 1− · 9.8 dy = 8 y − · 9.8
0 5 10 0
= 8 · 2.5 · 9.8 N · m = 196 J.

Alternatively, the cable has linear density 8 kg/5 m = 1.6 kg/m, and so the work
required to lift a small piece of the cable (of length ∆y) from height y m to height
5 m is
1.6∆y · 9.8 · (5 − y) .
| {z } |{z} | {z }
mass gravity distance
| {z }
force

The total work required is therefore

5  5
1 2
Z
1.6 · 9.8(5 − y) dy = 1.6 · 9.8 5y − y
0 2 0
 
25
= 1.6 · 9.8 · 25 − = 196 J
2

as before.
2.1.2.12. Solution. Imagine pumping out a thin, horizontal layer of water that
is at height y — that is, y metres above the bottom of the tank. Let the width of
the layer be dy.

dy

• The volume of water in the layer is 3dy m3 (since the cross-section has area
3 m3 ).

• One cubic metre is equal to 1003 cubic centimetres. So, the mass of water in

946
S OLUTIONS TO E XERCISES

1003
one cubic metre is = 1 000 kg.
1000
• Therefore, the mass of water in our layer is (3 000 dy) kg.

• The force of gravity acting on it is (−9.8 × 3 000 dy) N, so we need to pump

with a compensating force of (9.8 × 3 000 dy) N.

• The water needs to be pumped a distance of 1 − y metres.

• So, the work required to pump out the thin layer of water at height y is
(9.8 × 3 000 × (1 − y) dy) J.

So, all together, the work to pump out the entire tank is
1  1
1
Z
9.8 × 3 000 × (1 − y) dy = 9.8 × 3000 × y − y 2 = 14 700 J
0 2 0

2.1.2.13. *. Solution. We can model the sculpture as a collection of thin hori-

zontal plates of width dz. Remember work is force times distance; a horizontal
plate at height z moved z + 2 metres from the basement to its final position. So,
we need to know the force acting on the plate, which is the product of the mass
of the plate with the acceleration due to gravity. Since we are given the density
of iron, if we find the volume of the plate, then we can calculate its mass.

dz

The plate at height z

• has side length 3 − z m and hence
• has area (3 − z)2 m2 and hence
• has volume (3 − z)2 dz m3 and hence
• has mass 8000(3 − z)2 dz kg and hence
• is subject to a gravitational force of 9.8 × 8000(3 − z)2 dz N and hence
• requires work 9.8×8000(2+z)(3−z)2 dz J to raise it from 2 m below ground
level to z m above ground level.
So the total work is
Z 3
9.8 × 8000(2 + z)(3 − z)2 dz joules
0

947
S OLUTIONS TO E XERCISES

2.1.2.14. Solution. From the information given about the hanging kilogram,
we can find the spring constant k. One kilogram generates a force of 9.8 N under
gravity. (We find this by the calculation (1 kg) × (9.8m/sec2 ) = 9.8 N.) This force
1
is matched by the force of the spring, which by Hooke’s law is equal to k( 20 m).
So,
9.8 N N
k= 1 = 196
20
m m
Again by Hooke’s law, the force required to stretch the spring x metres past its
natural length is 196x N (when x is measured in metres).
So, the work required to stretch the spring from 5 cm past its natural length to 7
cm past its natural length is
Z 0.07 0.07
196x dx = 98x2 0.05 = 0.2352 J

W =
0.05

2.1.2.15. Solution. Let M be the mass of the rope. Then its density is M4 kg/m.
Following the method of Example 2.1.6, we let y be the height of the firewood
above the ground, so the wood is raised from y = 0 to y = 4. When the wood is
at height y,
M
• the rope that remains to be lifted has length 4−y, and so it has mass 4
(4−y)
kg,
• and the firewood still has mass 10 kg.
• The remaining ropeand the wood are subject to a downward gravitational
M
force of magnitude (4 − y) + 10 × 9.8 N.
4
| {z }
mass

 M y to height(y + dy), we need to apply

• So, to raise the firewood from height
a compensating upward
M force  4 (4 − y) + 10 × 9.8 through distance dy.
of
This takes work 4 (4 − y) + 10 × 9.8 dy J.
All together, the work involved in hauling up the wood is
Z 4    Z 4 
M M
(4 − y) + 10 × 9.8 dy = 9.8 (M + 10) − y dy
0 4 0 4
= 9.8(2M + 40) J
Since the work was 400 joules, solving 400 = 9.8(2M + 40) for M tells us the mass
of the rope is 200
9.8
− 20 = 20
49
kg, or about 408 g.
Alternately, the work involved in lifting up the wood is 10 × 9.8 × 4 = 392 J, so
the work in lifting up the rope is 8 J. A small section of rope of length dy, that
starts at height y above the ground, has mass M4 dy kg and is lifted (4 − y) metres,
so the work involved in lifting this section of rope is 9.8 × (4 − y) × M4 dy. Then
the amount of work to lift the whole rope (but not the wood) is
Z 4
9.8 × M 4

M
Z
8J= 9.8 × (4 − y) × dy = (4 − y)dy
0 4 4 0

948
S OLUTIONS TO E XERCISES

9.8 × M
= ×8
4
4 20
which again results in M = 9.8
= 49
kg.
2.1.2.16. Solution. For Questions 11 and 15 in this section, we gave two meth-
ods for finding the work involved in pulling up a cable: one where we consider
pulling up the entire remaining cable a tiny distance of dy, and one where we
consider pulling a tiny slice of cable of length dy the entire distance up.
There is another variation we can consider with the weight: we can either calcu-
late the work done on the weight and the work done on the rope separately, or
we can calculate them together. If we calculate them together, then there are two
cases to consider: the work done pulling up the first 5 metres of rope involves
the weight, while the last 5 metres does not. These two choices (how to model
the rope, and how to deal with the weight) actually lead to four solutions, but to
avoid unnecessary repetition only two are presented below.
• Solution 1: In this solution, we consider the work on the rope separately
from work on the weight, and we imagine lifting a tiny piece of rope the
entire distance to the window.
The weight has a mass of 5 kg, and is lifted a distance of 5 m to the window.
The force of gravity acting on the weight is (5 kg)(9.8 m/sec2 ) = 49 N, so
the work to lift it 5 metres is (49 N)(5 m) = 245J.

dy

1
The density of the rope is 10 kg/m. A tiny piece of rope of length dy,
1
hanging y metres from the window, has mass ( 10 dy) kg, and needs to
be lifted y metres. So, the force of gravity acting on the piece of rope is
1 2


( 10 dy kg) 9.8 m/sec = 0.98 dy N, and the work to pull it up to the win-
dow is (0.98y dy) J. So, the total work to pull up the rope is
Z 10  2 10
y
0.98y dy = 0.98 = 49 J
0 2 0

All together, the work to pull up the rope with the weight is 245 + 49 = 294
J.

949
S OLUTIONS TO E XERCISES

• Solution 2: In this solution, we consider the work on the rope together with
the weight, and we imagine lifting the remaining rope a tiny distance to the
window.
Suppose y metres of the rope have been pulled in, and 0 ≤ y ≤ 5 (shown on
the left, below). Then the remaining rope has length 10−y, and contains the
1 y
weight, so the mass remaining to be pulled up is (10 − y) + 5 = 6 − 10
10
| {z } |{z}
rope weight
kg. Then the force of gravity acting on the dangling rope and weight is
(9.8 m/sec2 )((6 − 10
y
) kg) = (58.8 − 0.98y) N. The work needed to lift this
rope dy metres is (58.8 − 0.98y) dyJ.

10 − y
10 − y

Now, suppose y metres of the rope have been pulled in, and 5 < y ≤ 10
(shown above, right). Then the remaining rope has length 10 − y, but does
1
not contain the weight, so the mass remaining to be pulled up is 10 (10 −
y
y) = 1 − 10 kg. Then the force of gravity acting on the dangling rope is
(9.8 m/sec2 )((1 − 10
y
) kg) = (9.8 − 0.98y) N. The work needed to lift this
rope dy metres is (9.8 − 0.98y) dyJ.
All together, the work needed to lift the rope is
Z 10
W = F (y)dy
0
Z 5 Z 10
= (58.8 − 0.98y) dy + (9.8 − 0.98y) dy
0 5
5  10
= 58.8y − 0.49y 2 0 + 9.8y − 0.49y 2 5


= 294 J
2.1.2.17. Solution.
m
= 39.2 kg·m


a The frictional force is µ × m × g = 0.4 (10 kg) 9.8 sec2 sec2
= 39.2 N.
Since this constant force acts over a distance of 3 metres, the work is 3 ×

950
S OLUTIONS TO E XERCISES

39.2 = 117.6 J.
In the case of a constant force, we don’t need to use an integral, but we
could if we wanted:
Z 3
 3
W = 39.2dx = 39.2x 0 = 39.2 × 3 = 117.6 J.
0

b Since the box is moving at a speed of 1 m/sec, at time t we can say the √ box
is at position t, 0 ≤ t ≤ 3. At position t, the mass√of the

box mis
(10 − t) kg,
so the frictional force is 0.4 × m × g = 0.4 10 − tkg 9.8 sec = 3.92(10 −
√ 2

t)N. Now that we know the force, to find the work we simply integrate,
following Definition 2.1.1:
3
3 √

2 3/2
Z
W = 3.92(10 − t)dt = 3.92 10t − t
0 3 0
√ √
 
2 3 h i
= 3.92 30 − 3 = 3.92 30 − 2 3 ≈ 104 J
3
2.1.2.18. Solution. Definition 2.1.1 in the text is justified by showing that the
work done by a force acting on a particle is equal to the change in the kinetic en-
ergy of that particle. We can use Hooke’s law to calculate the work done stretch-
ing the spring. That work will be equal to the change in kinetic energy of the
ball.
v2 2 v02
The ball initially has kinetic energy 12 (1 kg)(v0 m/sec)2 = 20 kg·m sec 2 = 2
J. At the
time the spring is stretched its farthest, the ball’s velocity is 0 m/sec, so its kinetic
v2
energy is 21 (1 kg)(0 m/sec)2 = 0J. So, the change in kinetic energy of the ball is 20
J.
Now let’s find the work done by the spring. Its spring constant is k = 5 N/m, so,
the force on the spring when it is stretched x metres past its natural length is 5x
N. The spring is stretched from its natural length to 10 cm, which is 0.1 m. Then
the work done by the spring is
0.1 0.1  0.1
5 1
Z Z
W = kxdx = 5x dx = x2 = J
0 0 2 0 40

Now we can find v0 .

v02 1 1
= =⇒ v0 = √ m/sec ≈ 22.36 cm/sec
2 40 20
2.1.2.19. Solution. The setup to answer this question is similar to Question 18
in this section: the work done by a spring on the occupied vehicle will be equal
to the change in kinetic energy of that occupied vehicle. So, we need to find the
work done by the spring, and the kinetic energy lost by the falling car. In order
to find the work done by the spring, we need to find the spring constant.

951
S OLUTIONS TO E XERCISES

• Spring constant: The car’s mass of 2000 kg compresses the struts 2 cm

past their natural length. The force of the car under gravity is (2000 kg) ×
(9.8 m/sec2 ) = 19 600 N. This force is exactly the same as that exerted by
the spring, k(0.02 m). So, k = 980 000 N/m.

• Work done by spring: The spring can safely compress 20 cm. So, the
amount of work done by the spring compressing that far gives us the max-
imum amount of work the spring can safely do. While the car is falling, the
spring is at its natural length, so the work done to compress it to 20 cm (0.2
m) shorter is:
Z 0.2 Z 0.2
0.2
980 000x dx = 490 000x2 0 = 19 600 J

kx dx =
0 0

• Change in kinetic energy: When the car first hits the pavement, it’s falling
at 4 m/sec, so it has kinetic energy 21 (2100 kg)(4 m/sec)2 = 16 800 J. When
the car compresses the springs as far as they go and it starts to rebound, it
has kinetic energy 0, since its instantaneous velocity is zero. So, the change
in kinetic energy is 16 800 J.

Since the change in kinetic energy is 16 800 J, and the struts can safely do a work
of (up to) 19 600 J, the jump is within the (meagre) safety limits set by the ques-
tion.

Exercises — Stage 3
2.1.2.20. Solution. Let’s consider sucking up a flat, horizontal layer of water. If
the water is y metres above bottom of the cone, then it needs to be raised 0.15 − y
metres. So, if its mass is m kg, then the force of gravity acting on it is 9.8m N and
the work involved in slurping it to the top of the cone is 9.8m(0.15 − y) J. So, what
we need to find is the mass of a layer of water y metres from the bottom of the
cone.
0.05

r
0.15
r
dy y

A horizontal cross-section of the cone is a circle. To find its radius, we use similar
triangles: yr = 0.05
0.15
, so r = 31 y. Therefore, the area of the cross-section of the cone

952
S OLUTIONS TO E XERCISES

2
y metres above its bottom is π 31 y = π9 y 2 m2 . If this layer has height dy, then its
volume is π9 y 2 dy m3 , and its mass is 1000 π9 y 2 dy kg.
Now, we know that the work to suck up the
π 2

layer of water y metres from the
bottom of the cup is 9.8(0.15−y) 1000 9 y dy J. So, the work involved in drinking
all the water is:
Z 0.15  π 
W = 9.8(0.15 − y) 1000 y 2 dy
0 9
Z 0.15
9800π
0.15y 2 − y 3 dy


=
9 0
 0.15
9800π 0.15 3 1 4
= y − y
9 3 4 0
 
9800π 3 1 4
= 0.05(0.15) − (0.15)
9 4
≈ 0.144 J

Even drinking water takes work. Life is hard.

2.1.2.21. *. Solution. Imagine slicing the water into horizontal pancakes of
thickness dx as in the sketch below.

Denote by x the distance of a pancake below the surface of the water. (So, x runs
from 0 to 3.) Each pancake:
√
• has radius 32 − x2 m (by Pythagoras) and hence

• has cross-sectional area π(9 − x2 ) m2 and hence

• has volume π(9 − x2 ) dx m3 and hence

• has mass 1000π(9 − x2 ) dx kg and hence

• is subject to a gravitational force of 9.8 × 1000π(9 − x2 ) dx N and hence

• requires work 9800π(9 − x2 )(x + 4) dx J to raise it to the spout. (It has to

be raised x m to bring it to the height of the centre of the sphere, then 3 m

953
S OLUTIONS TO E XERCISES

more to bring it to the top of the sphere, and finally 1 m more to bring it to
the spout.)

The total work is:

Z 3 Z 3
2
9800π 36 + 9x − 4x2 − x3 dx


9800π(9 − x )(x + 4) dx =
0 0
h 9 4 1 i3
= 9800π 36x + x2 − x3 − x4
2 3 4 0
369
= 9800 · π = 904,050π joules
4
2.1.2.22. Solution.

• Solution 1: Let’s consider the work involved in lifting up a small section of

cable, with length dy, distance y from the bottom end of the cable.

5−y

y dy

The distance this section must travel is (5 − y) metres, so if its mass is M (y),
then the work involved is
Z 5
W = 9.8 × (5 − y) × M (y)
0

So, we need to find M (y). The length of the section of cable is dy, and
its distance from the end of the cable is y, so the mass of the section is
(10 − y) dy. Therefore,
Z 5
W = 9.8 × (5 − y) × M (y)
0
Z 5
= 9.8 × (5 − y) × (10 − y) dy
0

954
S OLUTIONS TO E XERCISES
Z 5
50 − 15y + y 2 dy


= 9.8
0
 5
15 2 1 3
= 9.8 50y − y + y
2 3 0
6125
= = 1020 65 J
6

• Solution 2: Alternately, we can continue to use the basic method of Ex-

ample 2.1.6 in the text, noticing that the density of the cable is no longer
constant.
Let’s consider pulling the cable up a tiny distance of dy metres, after we
have already lifted it y metres (so (5 − y) metres of the cable is still in the
hole).

5−y

If R(y) is the mass of the remaining cable (in kg), then the force of gravity
is −9.8 × R(y), so the work done is 9.8 × R(y) × dy. Once we find R(y), we
can calculate the total work done:
Z 5
W = 9.8 × R(y) × dy (∗)
0

As given in the question statement, the density of the cable is (10−x) kg/m,
where x is the distance from the bottom end of the cable. Consider a tiny
section of cable x metres from the bottom end, of length dx.

955
S OLUTIONS TO E XERCISES

x dx

 
kg
The mass of this tiny section is 10 − x m × (dx m) = (10 − x)dx kg. The
section of cable dangling is the last (5 − y) metres of cable. So, the com-
bined mass of the section of cable dangling, after we’ve already pulled up
y metres of it, is
Z 5−y  5−y
1 2 75 1
R(y) = (10 − x)dx = 10x − x = − 5y − y 2
0 2 0 2 2

Now we can calculate the total work involved in pulling up the entire cable,
using equation (∗).
Z 5
W = 9.8 × R(y) × dy
0
Z 5  
75 1 2
= 9.8 × − 5y − y × dy
0 2 2
 5
75 5 1
= 9.8 y − y2 − y3
2 2 6 0
6125 5
= = 1020 J
6 6
2.1.2.23. Solution.
a The force of the depends on depth, which varies. So, consider a thin rect-
angle of the plunger at depth y, with height dy and width 1 m (the width of
the entire plunger). Let the area of this rectangle be dA.

depth

y dy

956
S OLUTIONS TO E XERCISES

The area of this rectangle is 1dy m2 ,
so the force of the water acting on it is
F = P · dA = 9800 mN3 (y m) dy m2 = 9800y dy N.


| {z } | {z } | {z }
c d dA
The depth at the top of the plunger is y = 0. To find the depth at the
bottom of the plunger, note that the water has a volume of 3 m3 , and is in a
rectangular container with base 1 m by 3 m. So, its height is 1 m.
The force over the entire plunger, from depth y = 0 to y = 1, is
Z 1 1
9800y dy = 4900y 2 0 = 4900 N

0

b Let’s follow our work from part (a), but with the width of the length of the
base as x m.
Still, a thin rectangle of plunger has width 1 m and height dy m, so it has
area dy m2 . At depth y, it has a force from the water of 9800y dy N. This
hasn’t changed from (a).
Now, let’s consider the depth of the water. The volume of water is 3 m3 ,
and it is in a rectangular container with base 1 m by x m. So, its depth is
3/x m. Therefore, the force on the entire plunger must be calculated from
y = 0 to y = 3/x.
3/x
9 44100
Z
3/x
9800y dy = 4900y 2 0 = 2 4900 =

F (x) = N
0 x x2

Let’s check that this answer makes sense: F (3) = 4900 N, which matches
our answer from (a).

c If the force of water acting on the plunger, when the length of the base is x
metres, is given by F (x), then we push the plunger with a force of −F (x).
Then the work we’re looking for is
Z 1 Z 3
W = −F (x) dx = F (x) dx
3 1

44100
F (x) is exactly what we found in (b): F (x) = x2
N.
3 3  3
44100 1
Z Z
W = F (x) dx = dx = 44100 −
1 1 x2 x 1
2
= 44100 · = 29 400 J
3
2.1.2.24. Solution. Let’s start by converting from time spent pulling to amount
pulled. When y metres of rope have been pulled up, 2y seconds have passed, so
1
5
y litres of water have leaked out of the bucket, leaving 5 − 15 y litres. (This only
makes sense when 15 y ≤ 5, but we only consider values of y from 0 to 5, so it’s

957
S OLUTIONS TO E XERCISES

not a problem. That is, we’re never hauling up an empty bucket that can’t leak
any more.)
When we’ve pulled up y metres of rope, the mass in the bucket is (5 − 15 y) kg,
so the force of gravity acting on it is 9.8(5 − 15 y) N. Since we pull up 5 metres of
rope, the work done is:
5    5
1 1 2
Z
W = 9.8 5 − y dy = 9.8 5y − y = 9.8 [25 − 2.5]
0 5 10 0
= 220.5 J
2.1.2.25. Solution. According to the formula for gravity between two objects,
the earth and moon will gravitationally attract one another no matter how far
apart they are, so what we’re looking for is the work to separate them infinitely
Z ∞
far. That is, we want to calculate F (r) dr, where a = 400 000 000 m.
a
If we take m1 and m2 to be the mass of the earth and moon as given in the ques-
tion statement, then:

m3
 
−11 24 22



Gm1 m2 = 6.7 × 10 6 × 10 kg 7 × 10 kg
kg · sec2
kg · m3
≈ 2.8 × 1037
sec2
With that out of the way, let’s calculate our work.
∞ b
m1 m2
Z Z
W = F (r) dr = lim G 2 dr
a b→∞ a r
 b
1
= (Gm1 m2 ) lim −
b→∞ r a
 
1 1
= (Gm1 m2 ) lim −
b→∞ a b
Gm1 m2
=
a
3
2.8 × 1037 kg·m
sec2
≈
4 × 108 m
= 7 × 1028 J

Remark: since the force of gravity between the earth and the moon gets weaker
as they are farther apart, it takes less and less work to move them each kilometre.
If we move them a finite distance apart, the work involved will always be less
than 7 × 1028 joules, no matter how huge that finite distance is. If we move them
a very, very long (but finite) distance apart, the work we did will be quite close
to (but still less than) 7 × 1028 joules.
2.1.2.26. Solution. A ball of mass m experiences a gravitational force of mg, so

958
S OLUTIONS TO E XERCISES

lifting it a height of `/2 involves a work of 21 mg`.

The cable has density m/`. A tiny section of cable with length dy has mass m` dy,
and so gravity acts on it with a force of mg
`
dy. If the tiny section of cable is y units
from the top of the cable, it needs to be pulled up y units, so the work on that
section is mg`
y dy. Therefore, the work to pull up the entire cable is
`
mg h mg i` mg 2 1
Z
y dy = y2 = ` = mg`
0 ` 2` 0 2` 2

So, the work to pull up a cable with uniform density is the same as the work to
pull up a ball with the same mass from the middle height of the cable.
Remark: this is a nice fact to use when you’re checking your computations for
“pulling up cable” problems, but keep in mind it depends on the cable being of
uniform density.
2.1.2.27. Solution. Like our other tank-pumping problems (e.g. Questions 12,
20, and 21 in this section, and Example 2.1.4 in the text), we can find the work
done by considering thin layers of liquid. If the layer of liquid h metres above
the bottom of the tank with thickness dh has mass M (h), then the force of gravity
acting on it is −9.8M (h) N and the work required to pump it to the top of the
tank (1 − h metres away) is 9.8(1 − h)M (h) J. So, the work to empty the entire
tank is Z 1
W = 9.8(1 − h)M (h) (∗)
0

Our remaining task is to find M (h). There are two things that vary with height:
the density of the liquid, and the area of the cross-section of the tank.
At height h metres, the cross-section of the tank is shaped like the finite region
bounded by the curves y = x2 and y = 2 − h − 3x2 . To find this area, we need an
integral (see Section 1.5 for a refresher), and to find the limits of integration, we
need to know where the √ two curves meet. By solving x2 = 2 − h√− 3x2 , we find
1
that they meet at x = ± 2 2 − h. (Recall h is between 0 and 1, so 2 − h is a real
√
number, i.e. the curves do indeed meet.) Furthermore, when − 12 2 − h ≤ x ≤
1
√
2
2 − h, then x2 ≤ 2 − h − 3x2 , so y = x2 is the bottom function and 2 − h − 3x2
is the top function.

959
S OLUTIONS TO E XERCISES

y = x2

x
√
− 12 2−h

y = 2 − h − 3x2

So, (taking advantage of the fact that our region has even symmetry) the area of
the cross-section of the tank at height h is
1
√
Z
2
2−h
2 − h − 3x2 − x2 dx
 

A(h) = √
− 12 2−h
1
√
Z
2
2−h
2 − h − 4x2 dx


=2
0
  1 √2−h
4 3 2
= 2 (2 − h)x − x
3 0
√ 4 1√
 
1 3
= 2 (2 − h) · 2−h− · 2−h
2 3 8
 
3/2 1 2
= (2 − h) 1− = (2 − h)3/2
3 3

Now, we can calculate the volume of a slice at height h of thickness dh.

2
V (h) = (2 − h)3/2 dh
3
√
The density of the liquid at height h is 1000 2 − h kg/m3 , so
√ 2
M (h) = 1000 2 − h × (2 − h)3/2 dh
3
2000
= (2 − h)2 dh
3
Now we use (∗) to find the work done pumping out the tank.
1 1
2000
Z Z
W = 9.8(1 − h)M (h) = 9.8(1 − h) · (2 − h)2 dh
0 0 3
19600 1
Z
4 − 8h + 5h2 − h3 dh


=
3 0

960
S OLUTIONS TO E XERCISES
 1
19600 2 5 3 1 4
= 4h − 4h + h − h
3 3 4 0
 
19600 5 1
= 4−4+ −
3 3 4
19600 17
= ×
3 12
83300
= = 9255 59 J
9
2.1.2.28. Solution. Since the only work done is against the force of gravity, we
only need to know how high the sand was lifted, not how it got there. So, we
don’t really need to worry about its semicircular path: we can imagine that every
grain of sand was lifted from its old position to its new position.
Consider a thin, horizontal layer of sand in the hourglass, y metres below the
vertical centre of the hourglass.
• Its final position is y metres above the centre of the hourglass. That is, it
was lifted 2y metres against the force of gravity.
• The layer is shaped like a circle with radius y 2 + 0.01 and height dy, so its
2
volume is π (y 2 + 0.01) dy cubic metres.
• To find the mass of the layer, we need to know the density of the sand. Let
the volume of sand in the hourglass be V . We are given its mass M . Then
2 2
π (y 2 + 0.01) dy cubic metres has a mass of M
V
π (y 2 + 0.01) dy kilograms.
2
• So, the force of gravity acting on the layer is 9.8 M
V
π (y 2 + 0.01) dy N, acting
vertically downwards.
• To lift the layer to its final position, we apply a compensating force over a
2
distance of 2y metres, for a total work of 9.8 M
V
π (y 2 + 0.01) 2y dy J.
• Since the hourglass has height 0.2 m, and exactly half of it is filled with
sand, the top layer of sand is exactly at the vertical centre of the hourglass,
and the bottom layer of sand is 0.1 metres below.
Using V is the volume of sand in the hourglass, and M is its mass (we’re given
M = 71 kg) then the total work flipping all the sand is:

0.1
M
Z

2
W = π y 2 + 0.01 2y dy
9.8
0 V
19.6M π 0.1 5
Z
y + 0.02y 3 + 0.0001y dy


=
V 0
Z 1/10  
19.6M π 5 2 3 1
= y + 2 y + 4 y dy
V 0 10 10
 1/10
19.6M π 1 6 1 4 1 2
= y + y + y
V 6 2 · 102 2 · 104 0

961
S OLUTIONS TO E XERCISES
  
19.6M π 1 1 1
= + + −0
V 6 · 106 2 · 106 2 · 106
 
19.6M π 7
= (∗)
V 6 · 106

It remains to find V : the volume of sand in the hourglass. We know the sand is
in the shape of a solid of rotation. Recall from Section 1.6 that we can find the
volume of such shapes by slicing them into thin disks.

x dx
0.1

x2 + 0.01

In the picture above, we’ve used an axis that matches the way we’ve been de-
scribing our solid: 0 is the vertical centre of the hourglass, which is where the top
of the sand is, and the bottom of the sand is 0.1 metres from 0.
To find the volume of this solid, we slice it into horizontal disks. The disk that is
x metres from the centre of the hourglass has radius x2 + 0.01 and thickness dx,
so it has volume π(x2 + 0.01)2 dx. The volume of the entire solid, i.e. the volume
of the sand, is:
Z 0.1 Z 0.1
2 2
V = π(x + 0.01) dx = π (x4 + 0.02x2 + 0.0001) dx
0 0
 0.1
1 5 2 3 1
=π x + x + 4x
5 3 × 102 10 0
  
1 2 1
=π + + −0
5 × 105 3 × 105 105
28π 28π
= =
15 × 105 1.5 × 106
28π
So, the volume of sand in the hourglass is V = 1.5×10 6 cubic metres.

Using (∗), we can find the total work done quickly flipping the hourglass.
 
19.6M π 7
W =
V 6 · 106
19.6 × 17 π
 
7
= 28π
1.5×106
6 · 106
19.6 × 1.5 29.4 7
= = = = 0.175 J
28 × 6 168 40

962
S OLUTIONS TO E XERCISES

2.1.2.29. Solution. Using Definition 2.1.1, the work involved is

Z 1/2 √
W = 1 − x4 dx J.
0
√
However, the function F (x) = 1 − x4 happens to not have an antiderivative
that can be expressed as an elementary function. That means we can’t use the
Fundamental Theorem of Calculus Part 2 to evaluate this integral (at least, not
without knowing a bit more about functions than is prerequisite for this course).
Instead, we can use numerical methods, like the midpoint rule or Simpson’s rule,
to approximate its value.
It’s not immediately clear which rule (Simpson’s, midpoint, or trapezoidal) will
lead us down the easiest path. For Simpson’s rule, we need to know the fourth
derivative of F (x), which is not a simple task. But, we often need fewer intervals
for Simpson’s rule than for the midpoint or trapezoid rules. In this case, we’ll
show below that n = 2 intervals suffice to guarantee a low enough error using
the midpoint rule, so Simpson’s rule won’t let us get away with fewer intervals.
Below, we find the approximation using the midpoint rule — but there are other
ways as well.
In order to decide how many intervals we should use with the midpoint rule, we
need to know the second derivative of F (x).

F (x) = (1 − x4 )1/2
1
−1/2
−1/2
F 0 (x) = 1 − x4 · (−4x3 ) = −2x3 1 − x4
2  
00 3 1
−3/2
−1/2
F (x) = (−2x ) − 1 − x4 (−4x3 ) + (−6x2 ) 1 − x4
2

−3/2
−1/2
= −4x6 1 − x4 + (−6x2 ) 1 − x4

−3/2
= 2x2 1 − x4 −2x4 − 3 1 − x4




−3/2 4
= 2x2 1 − x4


x −3
2x2 (x4 − 3)
=
(1 − x4 )3/2

For values of x between 0 and 21 ,

3/2

3/2
1 4
• the denominator 1 − x4 is always at least as big as 1 − 2
,

2
• the factor x2 in the numerator is never bigger than 12 , and

• the factor x4 − 3 in the numerator has magnitude at most3,

so that

00 2x2 |x4 − 3| 2( 12 )2 (3) 3

2 32
|F (x)| = ≤
3/2 = = √ <2
(1 − x4 )3/2 1 − ( 12 )4
153/2
64
5 15

963
S OLUTIONS TO E XERCISES

Using Theorem 1.11.13 with M = 2, a = 0, and b = 1/2, the error in a midpoint

approximation with n intervals is at most

M (b − a)3 2 1/8 1
· = · =
24 n2 24 n2 96n2
1
If n ≥ 2, then our error is certainly less than 100 .
1
If we use the midpoint rule with n = 2, then x̄1 = 8
and x̄2 = 38 .

x
1 1 3 1
0 8 4 8 2

R 1/2 √ 1
The midpoint rule approximation of 0
1 − x4 dx with n = 2 and ∆x = 4
is:
1/2 √
Z q q 
4 4
1− x4 dx ≈ ∆x 1 − x̄1 + 1 − x̄2
0
s 
 4 s  4
1 1 3 
=  1− + 1−
4 8 8

Since we used n = 2, by our previous work the error in this approximation is less
1 1
than 96×22 = 384 , which is certainly less than 0.01, as required.

2.2 · Averages
2.2.2 · Exercises

Exercises — Stage 1
2.2.2.1. Solution. Since the average of f (x) on the interval [0, 5] is A, using
Definition 2.2.2,

1 5
Z
A= f (x) dx
5 0
Z 5
5A = f (x) dx
0
R5
So, a rectangle with width 5 and height A has area 0 f (x) dx.
That is: if we replace f (x) with the constant function g(x) = A, then on the
interval [0, 5], the area under the curve is unchanged.

964
S OLUTIONS TO E XERCISES

x
5

(There are many rectangles with area 5A; we drew the one we consider to be the
most straightforward in this context.)
2.2.2.2. Solution. Average velocity, as discussed in Example 2.2.5, is change
in position divided by change in time. So, the change in position (i.e. distance
travelled) is (100 km/h)(5 h) = 500 km.
2.2.2.3. Solution. The work done is
Z b
W = F (x) dx
a

so the average value of F (x) is

b
1 1
Z
F (x) dx = (W ).
b−a a b−a

We can quickly check our units: since W is in joules (that is, newton-metres), and
W
b − a is in metres, so b−a is in newtons.
2.2.2.4. Solution.
a The entire interval has length b − a, and we’re cutting it into n pieces, so
the length of one piece (and hence the distance between two consecutive
samples) is b−a
n
.
b The first sample, as given in the question statement, is taken at x = a. The
second sample, then, is at x = a + b−an
, this third is at x = 1 + 2 b−an
, and the
fourth is at a + 3 b−a
n
.
c The y-value of the fourth sample is simply f a + 3 b−a


n
. Note this is the
number we use in our average, not the x-value.
d Our samples are f (a), f a + b−a , f a + 2 b−a , f a + 3 b−a




n n n
, etc. Since
there are n of them, we divide their sum by n. So, the average is:
f (a) + f a+ b−a + f a+2 b−a + · · · + f a+(n − 1) b−a




n n n
n

965
S OLUTIONS TO E XERCISES
    
1 b−a b−a
= f (a) + f a + f a+2 + ···
n n n
 
b−a
+ f a + (n − 1)
n
n  
1X b−a
= f a + (i − 1)
n i=1 n

Remark: if we multiply and divide by b − a, we see this expression is equiv-

alent to a left Riemann sum, divided by the length of our interval.
n  
1 X b−a b−a
= f a + (i − 1)
b − a i=1 n n
n
1 X
= f (a + (i − 1)∆x) ∆x
b − a i=1

As n gets larger and larger, using the definition of a definite integral, this
1
Rb
expression gets closer and closer to b−a a
f (x) dx. This is one way of justi-
fying our definition of an average of a function on an interval.
2.2.2.5. Solution.

a Yes, the average of f (x) is less than or equal to the average of g(x) on [0, 10].
The reason is that, if f (x) ≤ g(x) for all x in [0, 10], then:
Z 10 Z 10
1 1
f (x) dx ≤ g(x) dx.
10 0 10 0

b There is not enough information to tell. It’s certainly possible: for instance,
take f (x) = 0 and g(x) = 1 for all x in [0, 10]. Then f (x) ≤ g(x) and the
average of f (x) is 0, which is less than 1, the average of g(x).
(
100 if 0 ≤ x ≤ 0.01
However, consider f (x) = and g(x) = 0. Then
0 else
f (x) ≤ g(x) for all x in [0.01, 10], but the average of f (x) is 0.1, while the
average of g(x) is 0.
2.2.2.6. Solution. Recall the definition of an odd function: f (−x) = −f (x).
Since the domain of integration is symmetric, the signed area on one side of the
y-axis “cancels out” the signed area on the other — this is Theorem 1.2.12 in the
text.
Z 10
1 1
f (x) dx = (0) = 0
20 −10 20

Exercises — Stage 2

966
S OLUTIONS TO E XERCISES

2.2.2.7. *. Solution. By definition, the average value is

π/2
1
Z


sin(5x) + 1 dx
π −π/2

We now observe that sin(5x) is an odd function, and hence its integral over the
symmetric interval [− π2 , π2 ] equals zero. So the average value of f (x) on this in-
terval is 1:

1 π/2 1 π/2 1 π/2

Z Z Z


sin(5x) + 1 dx = sin(5x) dx + 1 dx
π −π/2 π −π/2 π −π/2
1 π/2
Z
= 1 dx = 1
π −π/2

Alternatively, using the fundamental theorem of calculus, the average equals:

 π/2
1 − cos(5x)
+x
π 5 −π/2
   
1 − cos(5π/2) π − cos(−5π/2) −π
= + − +
π 5 2 5 2
π
= =1
π
2.2.2.8. *. Solution. By definition, the average is
Z e
1
x2 log xdx
e−1 1

To antidifferentiate, we use integration by parts with u = log x and dv = x2 dx,

hence du = x1 dx and v = 31 x3 .
Z e  e Z e 
1 2 1 1 3 1 2
x log xdx = x log x − x dx
e−1 1 e−1 3 1 1 3
 3 x=e
1 x x3
= log x −
e−1 3 9 x=1
 3 3

1 e e 1
= − +
e−1 3 9 9
 
1 2 3 1
= e +
e−1 9 9
2.2.2.9. *. Solution. By definition, the average value in question equals
π/2
1
Z
(3 cos3 x + 2 cos2 x) dx
π/2 − 0 0
 Z π/2 Z π/2 
2 3 2
= 3 cos x dx + 2 cos x dx
π 0 0

967
S OLUTIONS TO E XERCISES

For the first integral we use the substitution u = sin x, du = cos x dx, cos2 x =
1 − sin2 x = 1 − u2 . Note that the endpoints x = 0 and x = π2 become u = 0 and
u = 1, respectively.
Z π/2 Z π/2
3
3 cos x dx = 3 cos2 x cos x dx
0
Z0 1
= 3(1 − u2 ) du
0
1
= (3u − u3 ) = 2.
0

1+cos(2x)
For the second integral we use the trigonometric identity cos2 x dx = 2
.
Z π/2 Z π/2
2


2 cos x dx = 1 + cos(2x) dx
0 0
 π/2
1 π
= x + sin(2x) =
2 0 2

Therefore, the average value in question is

Z π/2 π/2   
2 2 π 4
Z
3 2
3 cos x dx + 2 cos x dx = 2+ = + 1.
π 0 0 π 2 π
2.2.2.10. *. Solution. By definition, the average value in question equals
Z π/k
1
Ave = sin(kx) dx
π/k − 0 0
To evaluate the integral, we use the substitution u = kx, du = k dx. Note that the
endpoints x = 0 and x = π/k become u = 0 and u = π, respectively. So
k π du 1h 2
Z iπ
Ave = sin(u) = − cos(u) =
π 0 k π 0 π
Remark: the average does not depend on k. To see why this is, note that sin(kx)
runs between −1 and 1 as x changes. When x = 0, kx = 0, and when x = π/k,
kx = π. So, our function sin(kx) runs exactly from sin 0 = 0 to sin(π/2) = 1, then
back down to sin π = 0.
y

y = sin(kx)
x
π π
2k k

968
S OLUTIONS TO E XERCISES

2.2.2.11. *. Solution. By definition, the average temperature is

3 3
1 1 80
Z Z
T (x) dx = dx
3 0 3 0 16 − x2

We don’t see an obvious substitution, but integrand is a rational function. The

degree of the numerator is strictly less than the degree of the denominator, so we
factor the denominator and use a partial fraction decomposition.

80 80 A B
2
= = +
16 − x (4 − x)(4 + x) 4−x 4+x
80 = A(4 + x) + B(4 − x)

Setting x = 4, we see 80 = 8A, so A = 10. Setting x = −4, we see 80 = 8B, so

B = 10.
1 3 80 1 3 80
Z Z
dx = dx
3 0 16 − x2 3 0 (4 − x)(4 + x)
1 3 h 10 10 i
Z
= + dx
3 0 4−x 4+x
1 3h 10 10 i
Z
= − + dx
3 0 x−4 4+x
10 h i3
= − log |x − 4| + log |x + 4|
3 0
3
10 x+4 10
= log = [log 7 − log 1]
3 x−4 0 3
10
= log 7 degrees Celsius
3
2.2.2.12. *. Solution. By definition, the average value is
Z e
1 log x
dx
e−1 1 x
1
To integrate, we use the substitution u = log x, du = x
dx. Then the limits of
integration become 0 and 1, respectively.
e 1  2 1
1 log x 1 1 u 1
Z Z
dx = udu = =
e−1 1 x e−1 0 e − 1 2 0 2(e − 1)
2.2.2.13. *. Solution. By definition, the average value is:
Z 2π
1 1 1 2π
Z
2


cos x dx = · cos(2x) + 1 dx
2π 0 2π 2 0
1 h sin(2x) i2π
= +x
4π 2 0

969
S OLUTIONS TO E XERCISES

1 1
= · 2π =
4π 2
2.2.2.14. Solution. Before we start answering
questions, let’s look at our func-
tion a little more carefully. The term 50 cos 12t π has a period of 24 hours, while
t


the term 200 cos 4380 π has a period of one year. So, the former term describes a
standard daily variation, while the latter gives a seasonal variation over the year.
(a) Using the definition of an average, the average concentration over one year
(t = 0 to 8760) is:
Z 8760     
1 t t
400 + 50 cos π + 200 cos π dt
8760 0 12 4380
Z 8760 Z 8760  
1 50 t
= 400 dt + cos π dt
8760 0 8760 0 12
Z 8760  
200 t
+ cos π dt
8760 0 4380
  8760   8760
5 12 t 5 4380 t
= 400 + sin π + sin π
876 π 12 0 219 π 4380 0

Since 8760 8760 8760

12
= 730, which is even, sin 12
π = sin(0) = 0. Also, sin 4380
π =
sin(2π) = 0.
5 5
= 400 + (0) + (0)
876 219
= 400 ppm

Remark: for the portions of the integral in red and blue, we also could have
noticed that the integrand goes through a whole (integer) number of periods.
For every period, the net signed area between the curve and the x-axis is zero, so
we could have seen from the very beginning these terms would contribute 0 to
the final average.
(b) Using the definition of an average, the average concentration over the first
day (t = 0 to t = 24) is:
Z 24     
1 t t
400 + 50 cos π + 200 cos π dt
24 0 12 4380
Z 24
50 24 200 24
   
1 t t
Z Z
= 400 dt + cos π dt + cos π dt
24 0 24 0 12 24 0 4380

Note t = 0 to t = 24 is one complete period for the integrand in red, so the red
integral will evaluate to zero. However, t = 0 to t = 24 is less than one cycle for
the integrand in blue, so we expect this will contribute some nonzero quantity to
the average.
  24
200 4380 t
= 400 + 0 + sin π
24 π 4380 0

970
S OLUTIONS TO E XERCISES
 
25 4380 24
= 400 + · sin π
3 π 4380
 
25 4380 2
= 400 + · sin π
3 π 365
≈ 400 + 199.99
= 599.99 ppm

Remark: C(0) = 400 + 50 + 200. The red term comes from the daily variation,
and over the first day this will have an average of 0. The blue term comes from
the seasonal variation, which changes dramatically over the course of an entire
year but won’t change very much over the course of a single day. So, it is reason-
able that the average concentration over the first day should be close to (but not
exactly) 400 + 200 ppm.
(c) The average of N (t) over [0, 8760] is:
Z 8760   
1 t
350 + 200 cos π dt
8760 0 4380
   8760
200 4380 t
= 350 + sin π
8760 π 4380
  0
200 4380 8760
= 350 + sin π
8760 π 4380
100
= 350 + sin (2π)
π
= 350

Since the average of C(t) was 400, this gives us an absolute error of |400 − 350| =
50 ppm, for a relative error of

50
= 0.125,
400
or 12.5%.
That is: sampling at the same time every day, rather than throughout the day,
lead to an error of 12.5% in the yearly average concentration of carbon dioxide.
2.2.2.15. Solution.
a The cross-section of S at x is a circle with radius x2 , so area πx4 . The average
of these values, 0 ≤ x ≤ 2, is
Z 2
1 1 h π 5 i2 16π
A= πx4 dx = x =
2−0 0 2 5 0 5

b To find the volume of S, imagine cutting it into thin circular disks of radius
x2 and thickness dx. The volume of one such disk is πx4 dx, so the volume
of S is Z 2 h π i2 32π
πx4 dx = x5 =
0 5 0 5

971
S OLUTIONS TO E XERCISES

c The volume of a cylinder is the product of its base area with its length. A
cylinder with circular cross-sections of area 16π
5
and length 2 has volume
32π
5
.
Remark: this is the same as the volume of S, so the average cross-sectional
area of S tells us the cross-sectional area of a cylinder with the same length
and volume as S. Compare this to Question 1, where we saw the average
value of a function gave the height of a rectangle with the same area as the
function over the given interval.
2.2.2.16. Solution.

a We can see without calculation that the average will be zero, since f (x) = x
is an odd function and [−3, 3] is a symmetric interval. Alternately, we can
use the definition of an average to calculate
3  3
1 1 2 1
Z
x dx = x = (9 − 9) = 0
6 −3 12 −3 12

b Using the definition provided for root mean square:

s Z s 3
27 −27 √
r
1 3 2 1 3
RMS = x dx = x = − = 3
6 −3 18 −3 18 18
Using the definition provided,
2.2.2.17. Solution.
s Z s Z
π/4
2 2 π/4
RMS = tan2 x dx = (sec2 x − 1) dx
π −π/4 π −π/4
s r h
2h iπ/4 2 π  π i
= tan x − x = 1− − −1 +
π −π/4 π 4 4
r  r
2 π 4
= 2− = − 1 ≈ 0.52
π 2 π
2.2.2.18. Solution.
a Using Hooke’s law, when the spring is stretched (or compressed) f (t) me-
tres past its natural length, the force exerted is kf (t), where k is the spring
constant. In this case, the force is

F (x) = (3 N/cm)(f (t) cm) = 3 sin (tπ) N

b Our interval encompasses three full periods of sine, so the average will be
zero.
Alternately, we can compute, using the definition of an average:
6
1 6

1 3
Z
Avg = 3 sin(tπ) dt = − cos(tπ)
6 0 6 π 0

972
S OLUTIONS TO E XERCISES

1
= [cos 0 − cos(6π)] = 0
2π

This it doesn’t tell us very much about the “normal” amount of force from
the spring during our time period. It only tells us that force in one direction
at is “cancelled out” by force in the opposite direction at another time.

c Using the definition given for root mean square,

s s
Z 6
1 3 6 2
Z
2
RMS = (3 sin(tπ)) dt = sin (tπ) dt
6 0 2 0
s
3 6
Z
= (1 − cos(2tπ)) dt
4 0
s  6
3 1
= t− sin(2tπ)
4 2π 0
s  
3 1
= 6− sin(12π) − 0
4 2π
r
3 3
= (6) = √ ≈ 2.12
4 2

Exercises — Stage 3
2.2.2.19. *. Solution. (a) Let v(t) be the speed of the car at time t. Then, by the
trapezoidal rule with a = 0, b = 2, ∆t = 1/3, the distance traveled is
Z 2 h1
v(t) dt ≈ ∆t v(0) + v(1/3) + v(2/3) + v(3/3) + v(4/3)
0 2
1 i
+ v(5/3) + v(2)
2
1 1
h 1 i
= 50 + 70 + 80 + 55 + 60 + 80 + 40 = 130 km
3 2 2
dist 130 km
(b) The average speed is time
≈ 2 hr
= 65 km/hr.
2.2.2.20. Solution.

a Using the definition of an average,

1
1
Z
A= et dt = e − 1
1−0 0

b Since s(t) − A = et − e + 1, its average on [0, 1] is

1
1
Z
1
et − e + 1 dt = et − et + t 0


1−0 0

973
S OLUTIONS TO E XERCISES

= (e − e + 1) − (1) = 0

Remark: what’s happening here is that the average difference between s(t)
and A is zero, because the values of s(t) that are larger than A (and give
a positive value of s(t) − A) exactly cancel out the values of s(t) that are
smaller than A (and give a negative value of s(t) − A). However, knowing
how far the average value is from our calculated average is a reasonable
thing to measure. That’s where (c) comes in.
c Using the definition of an average, the quantity we want is:
Z 1
1
et − e + 1 dt
1−0 0

To deal with the absolute value, we consider the integral over two intervals:
one where et −e+1 is positive, and one where it’s negative. To decide where
to break the limits of integration, notice et −e+1 > 0 exactly when et > e−1,
so t > log(e − 1).
Z 1
1
et − e + 1 dt
1−0 0
Z log(e−1) Z 1
t
= | e − e + 1 | dt + | et − e + 1 | dt
0 log(e−1)
| {z } | {z }
negative positive
Z log(e−1) Z 1
− et + e − 1 dt + et − e + 1 dt



=
0 log(e−1)
h ilog(e−1) h i1
= − et + (e − 1)t + et − (e − 1)t
0 log(e−1)

= [−(e − 1) + (e − 1) log(e − 1) + 1]
+ [e − (e − 1) − (e − 1) + (e − 1) log(e − 1)]
= 4 − 2e + 2(e − 1) log(e − 1)
≈ 0.42
Remark: what we just measured is how far s(t) is, on average, from A. We
had to neglect whether s(t) was above or below A, because (as we saw in
(b)) the values above A “cancel out” the values below A. That’s where the
absolute value came in.
Knowing how well most of your function’s values match the average is
an important measure, but dealing with absolute values can be a little
clumsy. Therefore, the variance of a function squares the differences, rather
than taking their absolute value. (In our example, that means looking at
(s(t) − A)2 , rather than |s(t) − A|.) To compensate for the change in mag-
nitude involve in squaring, the standard deviation is the square root of the
variance. These are two very commonly used measures of how similar a
function is to its average. Compare standard deviation to root-square-mean
voltage from Example 2.2.6 and Questions 16 to 18.

974
S OLUTIONS TO E XERCISES

2.2.2.21. Solution.

a Neither: the average of both these functions is zero. We saw this with a
particular function in Question 20 (b), but it’s actually true in general. It’s a
quick calculation to prove.
The average of f (x) − A is:
4 4
1 1
Z Z


f (x) − A dx = f (x) dx −A = A − A = 0
4−0 0 4 0
| {z }
A

Similarly, the average of g(x) − A is:

4 4
1 1
Z Z


g(x) − A dx = g(x) dx −A = A − A = 0
4−0 0 4 0
| {z }
A

b The function |f (x) − A| tells us how far f (x) is from A, without worrying
whether f (x) is larger or smaller. Looking at our graph, for most values of
x in [0, 4], f (x) is quite far away from A, so |f (x) − A| is usually a large,
positive quantity.
By contrast, |g(x) − A| is a small positive quantity for most values of x. The
function g(x) is quite close to A for all values of x in [0, 4].
So, since |g(x) − A| generally has much smaller values than |f (x) − A|, the
average of |f (x) − A| on [0, 4] will be larger than the average of |g(x) − A|
on [0, 4].
As discussed in Question 20(c), the average of |f (x) − A| is a measure of
how closely f (x) resembles its average. We see from the graph that f (x)
doesn’t resemble the constant function y = A much at all, while g(x) seems
much more similar to the constant function y = A.
This kind of measure — how similar a function is to its average — is also
the idea behind the root square mean.
2.2.2.22. Solution. When we rotate f (x) about the x-axis, we form a solid whose
radius at x is |f (x)|. So, its circular cross-sections have area π|f (x)|2 = πf 2 (x). If
we slice this solid into circular disks of thickness dx, then
Z the disks have volume b
πf 2 (x) dx. Therefore, the volume of the entire solid is πf 2 (x) dx. All we need
a
to do now is get this into a form where we can replace the integral with the root
mean square, R.
b b
b−a
Z Z
2
V = πf (x) dx = π f 2 (x) dx
a b−a a

975
S OLUTIONS TO E XERCISES
s 2
b
1
Z
= π(b − a)  f 2 (x) dx
b−a a

= π(b − a)R2

Remark: the volume of a cylinder with length b − a and radius r is π(b − a)r2 .
So, the root mean square of f (x) gave us the radius of a cylinder with the same
volume as the solid formed by rotating f (x). Recall the average of f (x) gave
us the height of a rectangle with the same area as f (x). Compare this to the
geometric interpretations of averages in Questions 1 and 15.
R1
2.2.2.23. Solution. The question tells you 1−0 1
0
f (x) dx = f (0)+f
2
(1)
. Note f (0) =
c, and f (1) = a + b + c.
Z 1
1 f (0) + f (1) c + (a + b + c)
f (x) dx = =
1−0 0 2 2
Z 1
a + b + 2c
ax2 + bx + c dx =


0 2
 1
a 3 b 2 a b
x + x + cx = + + c
3 2 0 2 2
a b a b
+ +c= + +c
3 2 2 2
a
=a
3
a=0

That is, f (x) is linear.

2.2.2.24. Solution. The information given in the question is:

as2 + bs + c + at2 + bt + c


Z t
1
ax2 + bx + c dx


=
2 t−s s
 t
1 a 3 b 2
= x + x + cx
t−s 3 2 s

1 a 3 b
= (t − s3 ) + (t2 − s2 )
t−s 3 2

+ c(t − s)
a b
= (t2 + st + s2 ) + (t + s) + c
3 2
a 2 b a b
(s + t2 ) + (s + t) + c = (t2 + st + s2 ) + (t + s) + c
2 2 3 2
a 2 a
(s + t2 ) = (t2 + st + s2 )
2 3

976
S OLUTIONS TO E XERCISES

s2 + t2 t2 + st + s2
 
a − =0
2 3
 2
s − 2st + t2

a =0
6
a(s − t)2 = 0
a = 0 OR s = t

So, unless s = t (and we’re taking the very boring average of a single point!) then
a = 0. That is: f (x) is linear whenever s 6= t.
2.2.2.25. Solution. The function g(x) = f (a + b − x), on the interval [a, b], is
Rb
a mirror of the function f (x), with g(a) = f (b) and g(b) = f (a). So, a f (a +
Rb 1
Rb 1
Rb
b − x) dx = a f (x) dx, and hence b−a a
f (a + b − x) dx = b−a a
f (x) dx, so the
average value of f (a + b − x) on [a, b] is A.
1
Rb
Alternately, we can evaluate b−a a
f (a + b − x) dx directly, using the substitution
u = a + b − x, dx = −dx:
Z b Z u(b)
1 −1
f (a + b − x) dx = f (u) du
b−a a b − a u(a)
Z a
−1
= f (u) du
b−a b
Z b
1
= f (u) du
b−a a
=A
2.2.2.26. Solution.

a The function A(x) only gives us information about an integral when one
limit of integration is zero. We can get around this by using properties of
definite integrals from Section 1.2 to break our integral into two integrals,
each of which has 0 as one limit of integration. So, we find the average of
f (t) on [a, b] as follows:
Z b Z 0 Z b 
1 1
f (t) dt = f (t) dt + f (t) dt
b−a a b−a a 0
 Z a Z b 
1
= − f (t) dt + f (t) dt
b−a 0 0
 

1  1 a 1 b
Z Z 
= −a · f (t) dt +b · f (t) dt
 
b−a a b
| 0 {z | 0 {z

} }
A(a) A(b)

1 bA(b) − aA(a)
= (−aA(a) + bA(b)) =
b−a b−a

977
S OLUTIONS TO E XERCISES

b From the definition of A(x), we know

1 x
Z
A(x) = f (t) dt
x 0

That is, Z x
xA(x) = f (t) dt
0

To find f (x), we differentiate both sides. For the left side, we use the prod-
uct rule; for the right side, we use the Fundamental Theorem of Calculus
part 1.
A(x) + xA0 (x) = f (x)

So, f (t) = A(t) + tA0 (t).

2.2.2.27. Solution.
(
−1 if x ≤ 0
a One of many possible answers: f (x) = .
1 if x > 0

b No such function exists.

R1
• Note 1: Suppose f (x) > 0 for all x in [−1, 1]. Then 12 −1 f (x) dx >
1 1
R
2 −1
0 dx = 0. That is, the average value of f (x) on the interval [−1, 1]
is not zero — it’s something greater than zero.
R1
• Note 2: Suppose f (x) < 0 for all x in [−1, 1]. Then 21 −1 f (x) dx <
1 1
R
2 −1
0 dx = 0. That is, the average value of f (x) on the interval [−1, 1]
is not zero — it’s something less than zero.

So, if the average value of f (x) is zero, then f (x) ≥ 0 for some x in [−1, 1],
and f (y) ≤ 0 for some y ∈ [−1, 1]. Since f is a continuous function, and 0
is between f (x) and f (y), by the intermediate value theorem (see the CLP-
1 text) there is some value c between x and y such that f (c) = 0. Since x
and y are both in [−1, 1], then c is as well. Therefore, no function exists as
described in the question.
2.2.2.28. Solution. This seems like it might be true: if f is getting closer and
closer to zero, as x grows towards infinity, then over time the later values will
become a larger and larger portion of the total interval we’re looking at, and so
the average should look more and more like f (x) when x is large — that is, like
0. That’s some intuition to start us out, but it isn’t a rigorous argument. To be
sure we haven’t overlooked something, let’s use the definition of an average to
express A(x).

1 x
Z
A(x) = f (t) dt
x 0

978
S OLUTIONS TO E XERCISES
x
1
Z
lim A(x) = lim f (t) dt
x→∞ x→∞ x 0
R∞
If 0 f (t) dt converges, then this limit is 0, and the statementRis true. So, suppose
x
it does not converge. Since f (x) is positive, that means lim 0 f (t) d(t) = ∞, so
x→∞
we can use l’Hôpital’s rule. To differentiate the numerator, we use the Funda-
mental Theorem of Calculus part 1.
Rx
f (t) dt f (t)
lim A(x) = lim 0 = lim =0
x→∞ x→∞ x x→∞ 1
Rx
So, the statement is true whether 0 f (t) dt converges or not.
2.2.2.29. Solution. Note f (t) is a continuous function that takes only positive
values, and lim f (t) = 0. By the result of Question 28, lim A(x) = 0.
t→∞ x→∞

2.3 · Centre of Mass and Torque

2.3.3 · Exercises

Exercises — Stage 1
2.3.3.1. Solution. Note −x2 + 2x + 1 = 2 − (x − 1)2 . So, both parabolas are
symmetric about the line x = 1, and the x-coordinate of the centroid is x = 1.

y = (x − 1)2

x
y = −x2 + 2x + 1
x=1

The parabolas meet when:

(x − 1)2 = 2 − (x − 1)2
2(x − 1)2 = 2
|x − 1| = 1
x = 0, x = 2

At both these points, y = 1, so we see the figure is symmetric about the line y = 1.
Then the y-coordinate of the centroid is 1.

979
S OLUTIONS TO E XERCISES

y = (x − 1)2

y=1

x
y = −x2 + 2x + 1

Therefore, the centroid is at (1, 1).

2.3.3.2. Solution. The circle and the cut-out rectangle are symmetric about the
x-axis, and about the y-axis, so the centroid is the origin.
Remark: the centroid of a region doesn’t have to be a point in the region!
2.3.3.3. Solution. In general, this is false: weights farther out from the centre
“count more” when we calculate the centre of mass. For instance, a rod with a
1-kg weight at x = −10 and a 10-kg weight at x = 1 will balance at x = 0. There’s
far more mass to one side of x = 0 than the other.
2.3.3.4. Solution. Following Equation 2.3.1, the centre of mass of the rod is at:
P
(mass) × (position) 1×1+2×3+2×4+1×6 21 7
x̄ = P = = =
(mass) 1+2+2+1 6 2

That is, the centre of mass is 3.5 metres from the left end.
2.3.3.5. Solution. (a) If we were to set this figure on a pencil lined up along the
vertical line x = a, it seems pretty clear that it would fall to the left. So, the centre
of mass is to the left of the line x = a. The same is true in (b): the added density
on the left makes it only more lopsided. However, in (c), the right point is denser
than the left, which could counterbalance the left. Without knowing more about
the dimensions and the density, we can’t say where the centre of gravity is in
relation to the line x = a.
(d) Consider a section of the figure along the interval [b, c], and its “mirror” sec-
tion on the other side of the line x = a. These two sections will have the same
area, at the same distance from x = a. Since we only care about the x-coordinate
of the centre of gravity, it doesn’t matter that the two halves are at different y-
coordinates. The centroid falls along the line x = a.

980
S OLUTIONS TO E XERCISES

y B y B

A A
x x
a a
A A

B B

(e) There is the same amount of area to the left and right of the line x = a, as in
part (d). However, the area to the right is “stretched out” more, so that it occupies
space farther away from the line x = a. So, the centre of mass will be to the right
of the line x = a.
2.3.3.6. Solution.

• The volume of water in Tank A is 43 π(1)3 = 43 π cubic metres.

4000
• The mass of water is 3
π kg.

• By symmetry, the centre of mass of the water when it fills Tank A is exactly
in the centre of the sphere, at height ȳ1 = 4 metres above the ground (one
metre above the bottom of Tank A, which is three metres above the ground).

• When the water is entirely in Tank B, its height is 43 π metres. (The base of
Tank B has area 1 m2 , and the volume of water is 34 π m3 .) By symmetry, the
centre of mass is exactly halfway up, at height ȳ2 = 23 π metres.

• So, the point mass in our model is moved from ȳ1 = 4 to ȳ2 = 32 π, a distance
of 4 − 23 π metres, by gravity.

• The work involved is:

    
4000 2 m  78400π
πkg × 4 − πm × 9.8 2 = (6 − π)
3 3 sec 9
≈ 78, 225J
2.3.3.7. Solution.

a A thin slice of S at position x has height x1 , so if its width is dx, its area is
1
x
dx.

b A small piece of R at position x has density x1 , so if its length is dx, its mass
is x1 dx.

c Adding up all our tiny slices from (a) gives us the total area of S:
3
1
Z
dx = log 3
1 x

981
S OLUTIONS TO E XERCISES

d Adding up all our tiny pieces from (b) gives us the total mass of R:
3
1
Z
dx = log 3
1 x

e Using Equation 2.3.5, the x-coordinate of the centroid of S is

R3 R3
1
x · x1 dx 1
1 dx 2
R3 = =
1
dx log 3 log 3
1 x

f Using Equation 2.3.4, the centre of mass of R is

R3 R3
1
x · x1 dx 1
1 dx 2
R3 = =
1
dx log 3 log 3
1 x

Remark: following the derivation of Equation 2.3.5, if we wanted to find the

x-coordinate of the centroid of S, we would set up a rod that had exactly the
characteristics of R. That’s why all the answers were repeated.
2.3.3.8. Solution.
a If we chop R into n pieces, each piece has length b−a n
. Then our ith cut is at
position a + i b−a b−a b−a




n
, so our ith piece runs from a + (i − 1) n
to a + i n
.
The approximation of the mass of this piece comes from the density at its
midpoint,
a + (i − 1) b−a b−a

 

n
+ a + i n
mi =
 2 
1 b−a
= a + (i − 2 )
n

b−a
n

a m1 m2 mn b

So, the ith piece has length b−a n

, with approximate density ρ (mi ) =
1
ρ a + (i − 2 ) n . We approximate that the ith piece has mass b−a
b−a




n
·
ρ (mi ) and position mi . Using Equation 2.3.1, the centre of mass of R is
approximately at position:
n
P
(mass of ith piece) × (position of ith piece)
i=1
x̄n = n
P
(mass of ith piece)
i=1

982
S OLUTIONS TO E XERCISES

n 
b−a
P 
n
ρ(mi ) × mi
i=1
= n
b−a
P
n
ρ(mi )
i=1
n 
1
b−a
( b−a × a + (i − 21 ) b−a
P




n
ρ a+ i− 2 n
) n
i=1
= n
b−a
a + (i − 12 ) b−a
P


n
ρ n
i=1

b Remember the definition of a midpoint Riemann sum:

b n   
b−a b−a
Z X
1
f (x) dx ≈ · f a + (i − 2 )
a i=1
n n

The numerator of our approximation in part (a) is, therefore, a midpoint

Rb
Riemann sum of a ρ(x)×x dx, and the denominator is a midpoint Riemann
Rb
sum of a ρ(x) dx.
Using the definition of a definite integral (Definition 1.1.9), we see the limit
of the approximation in (a) as x goes to infinity is
Rb
xρ(x) dx
x̄ = Ra b
a
ρ(x) dx

This gives us the exact centre of mass of our rod.

Remark: this is Equation 2.3.4 in the text.
2.3.3.9. Solution.

a On the left-most corner of S, T (x) = B(x), so the height of S is zero; that

is, the area of a very small vertical strip is very close to zero, so the density
of R is close to 0. As we move closer to the position labeled a0 , the height
of the strips increases, so the areas of the strips increases, so the density
of R increases. Then, between the points labeled a0 and b0 , the height of S
remains constant, since T (x) and B(x) are parallel here, so the areas of the
strips of S remain constant, and the density of R remains constant. Then,
between b0 and b, the height of S decreases, so the area of the strips decrease,
so the density of R decreases.

983
S OLUTIONS TO E XERCISES

y T (x)

B(x)
x
a a 0
b 0 b

b At position x, the height of S is T (x) − B(x), so a rectangle with width dx

and this height would have area (T (x) − B(x)) dx.

c According to our model, the tiny section of R at position x with width dx

has mass (T (x)−B(x)) dx (that is, the area of S over this same tiny interval),
mass
so its density is ρ(x) = length = (T (x)−B(x))
dx
dx
= T (x) − B(x).

d Imagine S were a solid, of constant density. The the mass of a portion of S

is proportional to the area of that portion. To find the x-coordinate where
the solid would balance, we imagine compressing together the vertical di-
mension of S until it’s a rod. That is, we would take a very thin vertical
strip of S, and turn it into a small segment of a rod, with the same mass.
Then the centre of mass of that rod would be exactly the x-coordinate of the
centre of mass of the solid — that is, the x-coordinate of the centroid of S.
The compressed rod we form in this way is exactly R (perhaps multiplied
by a constant, to account for the density of S, but this doesn’t affect where
R balances). So, the x-coordinate of the centroid has the same position as
the centre of mass of R.
Our result from Question 8(b) tells us the centre of mass of R is
Rb
xρ(x) dx
Ra b
a
ρ(x) dx

In (c), we found ρ(x) = T (x) − B(x). So, for the solid S bounded by T (x)
and B(x) on the interval [a, b],
Rb
x(T (x) − B(x)) dx
x̄ = Ra b
a
(T (x) − B(x)) dx

Remark: the denominator is the area of S. This formula is the same as the
formula found in Equation 2.3.5.

984
S OLUTIONS TO E XERCISES

2.3.3.10. Solution.

a To begin with, we’ll sketch some strips, and put a dot at the centre of mass
of each one (its vertical centre).

y T (x)

B(x)
x
a a 0
b 0 b

In our model, each of these strips corresponds to a weight on R, positioned

at its centre of mass (the height of the dot), and with a mass equal to the
strip’s area. For the portion of S with a0 ≤ x ≤ b0 , each centre of mass is
at a slightly different height, but the areas of the slices are the same. So,
the corresponding weights along R are at different heights, but all have
the same mass, as shown below. (Note the rod R below only contains the
weights from the middle of S — we’ll add the rest later.)
For clarity, the diagrams below are zoomed in.

y T (x) R

B(x)

x
a a0 b0 b

By contrast to the slices in the interval [a0 , b0 ], the slices of S along [a, a0 ] all
have the same centre of mass, but different areas. So, there is one position
along R that has a number of weights all stacked on top of one another, of
varying masses.
The same situation applies to the slices of S along [b, b0 ]. So, all together, our
rod looks something like this:

985
S OLUTIONS TO E XERCISES

y T (x) R

B(x)

x
a a 0
b 0 b

Remark: if we had sketched the density of R, it would have looked some-

thing like this:

y T (x) R

B(x)
x
a a 0
b 0 b

because from our sketch, we see that the density of R:

• is 0 at either end,
• is suddenly very high where the blue weights are, and
• is constant and lower between the blue weights.

b At position x, the height of S is T (x) − B(x), and the width of the strip is
dx, so the area of the strip is (T (x) − B(x)) dx.
Since the density of S is uniform, the centre of mass of the strip is halfway
T (x) + B(x)
up: at .
2
c If we cut S into n strips, then the strip at position xi has area (T (xi ) −
T (xi ) + B(xi )
B(xi ))∆x, where ∆x = b−a
n
, and its centre of mass is at height .
2

986
S OLUTIONS TO E XERCISES

So, our approximation of the centre of mass of the rod is:

n
P
(Mi × yi )
i=1
ȳn = n
P
Mi
i=1
 
n
P T (xi ) + B(xi )
((T (xi ) − B(xi ))∆x) ×
i=1 2
= n
P
(T (xi ) − B(xi ))∆x
i=1
n
(T (xi )2 − B(xi )2 )∆x
P
i=1
= Pn
2 (T (xi ) − B(xi ))∆x
i=1

We use the definition of a definite integral (Definition 1.1.9) to re-write the

limit of the above function.
n
(T (xi )2 − B(xi )2 )∆x
P

ȳ = lim i=1P
n
n→∞
2
(T (xi ) − B(xi ))∆x
i=1
Rb
T (x)2 − B(x)2 dx


a
= Rb

2 a T (x) − B(x) dx

Remark: the denominator is twice the area of S. This equation for the y-
coordinate of the centroid is the same as the one given in Equation 2.3.5.
2.3.3.11. *. Solution. We use vertical strips, as in the sketch below. (To use
horizontal strips we would have to split the domain of integration in two: −3 ≤
y ≤ 0 and 0 ≤ y ≤ 3.)

The equations of the top and bottom of the triangle are

y = T (x) = −3x and y = B(x) = 3x.
The area of the triangle is A = 12 (6)(1) = 3. Now, we can apply the vertical-slice

987
S OLUTIONS TO E XERCISES

versions of Equation 2.3.5.

0 0
1 1
Z Z
   
x̄ = x T (x) − B(x) dx = x (−3x) − (3x) dx
A −1 3 −1
Z 0
1
=− 6x2 dx
3 −1

Exercises — Stage 2
2.3.3.12. Solution. Applying Equation 2.3.4,
R7 1 7
0
x · x dx 3
x3 1 3
(7 ) 14
x̄ = R7 = 07 = 3
1 2 =
x dx 1
x2 2
(7 ) 3
0 2 0

2.3.3.13. Solution. Applying Equation 2.3.4,

R 10 1
−3
x · 1+x 2 dx
x̄ = R 10 1
−3 1+x2
dx

For the numerator, we use the substitution u = 1 + x2 , du = 2x dx.

h i101
1
1 101 1
R
du 2
log u
= h 2 10 u i10 = 10
arctan 10 − arctan(−3)
arctan x
−3
h i
log 101 − log 10 log 10.1
= = ≈ 0.43
2(arctan 10 + arctan(3)) 2(arctan 10 + arctan(3))

Since arctangent is an odd function, arctan(−3) = − arctan(3); using logarithm

rules, log 101 − log 10 = log 101
10
= log 10.1.
2.3.3.14. *. Solution. If we use horizontal strips, then we need to break the
region into two pieces: y ≥ −1 = −e0 , and y ≤ −1. However, if we use vertical
strips, the equation of the top of the region is y = T (x) = 1, and the equation of
the bottom of the region is y = B(x) = −ex , for all x from a = 0 to b = 1. So, we
use vertical strips.

988
S OLUTIONS TO E XERCISES

y = −ex

Using Equation 2.3.5, the y-coordinate of the centre of mass is

Z 1
1 1 1
Z 
2 2 2x
 
ȳ = T (x) − B(x) dx = 1−e dx
2A 0 2e 0
1
e2

1 1 2x 1h 1i
= x− e = 1− −0+
2e 2 0 2e 2 2
3 e
= −
4e 4
2.3.3.15. *. Solution. (a) The lines y = 0, x = 0, and x = 2 are easy enough to
1
sketch. Let’s get some basic information about y = T (x) = √16−x 2 on the interval

[0, 2].

• For all x in its domain, T (x) ≥ 0. In particular, it’s always the top of our
region (so T (x) is a reasonable name for it), while the bottom is B(x) = 0.

• T (0) = 14 , and T (2) = 1

√
2 3

• T 0 (x) = x
(16−x2 )3/2
, which is positive on [0, 2], so T (x) is increasing.
Remark: to see that T (x) is increasing, we can also just break it into pieces:

◦ When x ≥ 0, x2 is increasing, so
◦ 16 − x2 is decreasing, so
√
◦ 16 − x2 is decreasing, so
◦ √ 1 = T (x) is increasing.
16−x2

2x2 +16
• T 00 (x) = (16−x2 )5/2
, which is positive, so T (x) is concave up.

989
S OLUTIONS TO E XERCISES

Remark: If we only wanted to solve (b), it would still be nice to have a sketch of
the region, but it wouldn’t need to be so detailed. Knowing that T (x) is always
greater than 0 would be enough to tell us we could use vertical slices with T (x)
as the top and y = 0 as the bottom.
If we wanted to use horizontal slices (we don’t... but we could!) we would addi-
tionally want to know that T (x) is increasing over [0, 2], T (0) = 41 , and T (2) = 2√1 3 .
This would tell us that:
• the right endpoint of a horizontal strip is always x = 2,
1 1
• the left endpoint is determined by T (x) from y = 4
to y = √
2 3
, and

• the left endpoint is x = 0 for 0 ≤ y ≤ 41 .

(b)

The part of the region with x coordinate between x and x + dx is a strip of width
1
dx running from y = 0 to y = √16−x 2 . It is illustrated in red in the figure above.

So, the area of the region is

Z 2 Z arcsin(1/2)
1 1 1 π
A= √ dx = 4 cos t dt = arcsin =
0 16 − x 2
0 4 cos t 2 6
√
where we made the substitution x = 4 sin t, dx = 4 cos tdt, 16 − x2 = 4 cos t.
Using Equation 2.3.5,
Z 2 " 2 #
1
R2 √ − 02 dx
2 2 16 − x 2

T (x) − B(x) dx 0
ȳ = 0 =
2A 2A

990
S OLUTIONS TO E XERCISES

2 2
1 1 1 1
Z Z
= 2
dx = dx
2A 0 16 − x 2A 0 (4 − x)(4 + x)

1 1/8 1/8
Using the method of partial fractions, we see 2
= + .
16 − x 4+x 4−x
Z 2h Z 2h
1 1/8 1/8 i 1 1 1 i
= + dx = − dx
2A 0 4 + x 4 − x 16A 0 x + 4 x − 4
1 h i2 6  
= log |x + 4| − log |x − 4| = log 6 − log 2 − log 4 + log 4
16A 0 16π
3 log 3
=
8π
2.3.3.16. *. Solution.
y

y = cos x

y = sin x

x
π
4

The top of the region is y = T (x) = cos(x) and the bottom of the region is y =
B(x) = sin(x). So, the area of the region is
Z π/4 Z π/4



A= T (x) − B(x) dx = cos(x) − sin(x) dx
0 0
h iπ/4
= sin(x) + cos(x)
0
√
 
1 1
= √ + √ − [0 + 1] = 2 − 1
2 2
If we use horizontal slices, we’ll need to break up the object into two regions, so
let’s use vertical slices. Using Equation 2.3.5, the region has centroid (x̄, ȳ) with:

1 π/4 1 π/4
Z Z



x̄ = x T (x) − B(x) dx = x cos(x) − sin(x) dx
A 0 A 0
We use integration by parts with u = x, dv = (cos x − sin x)dx; du = dx, v =
sin x + cos x.
!
1 h iπ/4 Z π/4
= x(sin x + cos x) − (sin x + cos x) dx
A 0 0

991
S OLUTIONS TO E XERCISES

1h iπ/4
= x sin(x) + x cos(x) + cos x − sin x
A  0 
1 π 1 π 1 1 1
= ·√ + ·√ +√ −√ −1
A 4 2 4 2 2 2
π
√ π
√
4
2−1 4
2−1
= = √
A 2−1
Again using Equation 2.3.5,
Z π/4 Z π/4
1 2 2 1
cos2 (x) − sin2 (x) dx



ȳ = T (x) − B(x) dx =
2A 0 2A 0
Z π/4
1 1 h1 iπ/4 1
= cos(2x) dx = sin(2x) = √
2A 0 2A 2 0 4( 2 − 1)
k
2.3.3.17. *. Solution. (a) Since k is positive, √1+x 2 > 0 for every x. Then the top
k
of our region is defined by T (x) = √1+x2 , and the bottom is defined by B(x) = 0.
If we make vertical slices, we don’t have to turn our region into two parts, so let’s
use vertical slices. The question asks for our final answer in terms of the area A
of the region, so we don’t need to find A explicitly.
Using Equation 2.3.5, the x-coordinate of the centroid is
1 2 1 1 k
Z Z
x̄ = x(T (x) − B(x)) dx = x√ dx
A 0 A 0 1 + x2
Although we have a quadratic function underneath a square root, we find an
easier method than a trig substitution: the substitution u = 1 + x2 , du = 2x dx.
This changes the limits of integration to 1 + 02 = 1 and 1 + 12 = 2, respectively.
 √ 2
1 2 k du k u k √
Z

= √ = = 2−1
A 1 u 2 2A 1/2 1 A
Again using Equation 2.3.5, the y-coordinate of the centroid is
Z 1 Z 1
1 2 2 1 k2
ȳ = (T (x) − B(x) ) dx = dx
2A 0 2A 0 1 + x2
k2 1 1 k2 h k2 π k2π
Z i
= dx = arctan 1 − arctan 0 = · =
2A 0 1 + x2 2A 2A 4 8A
(b) We have x̄ = ȳ if and only if
k √  k2π
2−1 =
A 8A
Since k and A are a positive constants (hence neither is equal to 0), we can divide
both sides by k and multiply both sides by A:
√ kπ
2−1=
8
8 √ 
k= 2−1
π

992
S OLUTIONS TO E XERCISES

2.3.3.18. *. Solution. (a)

The curve y = x2 − 3x is a parabola, pointing up, with x-intercepts at x = 0 and
x = 3.
The curve y = x − x2 is a parabola, pointing down, with x-intercepts at x = 0 and
x = 1.
To find where the two curves meet, we set them equal to each other:

x2 − 3x = x − x2
2x2 − 4x = 0
2x(x − 2) = 0
x = 0 and x = 2

This is enough information to sketch the figure, on the left below.

(b) As we found in (a), the curves cross when x = 0, x = 2. The corresponding

values of y are y = 0 and y = 2 − 22 = −2. Note the top curve is T (x) = x − x2 ,
and the bottom curve is B(x) = x2 − 3x. Using vertical strips, as in the figure on
the right above, the area of R is
2 2  2
2 3
Z Z
(x − x2 ) − (x2 − 3x) dx = 2 2
   
4x − 2x dx = 2x − x
0 0 3 0
16 8
=8− =
3 3
(c) Using Equation 2.3.5, the x-coordinate of the centroid of R (i.e. the weighted
average of x over R) is

3 2  3 2 2
Z Z
2 2
4x − 2x3 dx
 
x̄ = x (x−x ) − (x −3x) dx =
8 0 8 0
3 4 3 1 4
h i 2 3 32
h i
= x − x = −8
8 3 2 0 8 3
=1
2.3.3.19. *. Solution. Using Equation 2.3.5, the x-coordinate of the centroid is
R1 1
0
x 1+x 2 dx
x̄ = R 1 1
0 1+x2
dx

993
S OLUTIONS TO E XERCISES

We can guess the antiderivative in the numerator, or use the substitution u =

1 + x2 , du = 2x dx.
1 1
2
log(1 + x2 ) 0
1
2
log 2 2
= 1 = = log 2 ≈ 0.44127
arctan x π/4 π
0

2.3.3.20. *. Solution. By symmetry, the centroid lies on the y-axis, so x̄ = 0.

The area of the figure is the area of a half-circle of radius 3, and a rectangle of
width 6 and height 2. So, A = 21 π(9) + 6 × 2 = 29 π + 12.
We’ll use vertical strips as in the sketch below.

√
The top function of our figure is T (x) = 9 − x2 , and the bottom function of our
figure is B(x) = −2. Using Equation 2.3.5, the y-coordinate of the centroid is:
Z b
1
T (x)2 − B(x)2 dx


ȳ =
2A a
Z 3 √
1 2

= 9 − x2 − (−2)2 dx
2A −3
Z 3
1
5 − x2 dx


=
2A −3
 3
1 1 3
= 5x − x
2A 3 −3
1
= [15 − 9 + 15 − 9]
2A
6 6 12
= = 9 =
A 2
π + 12 9π + 24

2.3.3.21. *. Solution. (a) Notice that when x = 0, y = 3 and as x2 increases,

y decreases until y hits zero at x2 = 94 , i.e. at x = ± 23 . For x2 > 49 , y is not
even defined. So, on D, x runs from − 32 to + 23 and, for each x, y runs from 0 to
√
9 − 4x2 . Here is a sketch of D.

994
S OLUTIONS TO E XERCISES

√
As an aside, we can rewrite y = 9 − 4x2 as 4x2 + y 2 = 9, y ≥ 0, which is the top
half of the ellipse which passes through (±a, 0) and (0, ±b) with a = 32 and b = 3.
The area of the full ellipse is πab = 92 π. The area of D is half of that, which is 94 π.
But we are told to use an integral, so we will do so.
The area is
Z 3/2 √
Area = 9 − 4x2 dx
−3/2

We can evaluate this integral by substituting x = 32 sin θ, dx = 32 cos θ dθ and using

3
x=± ⇐⇒ sin θ = ±1
2
So − π2 ≤ θ ≤ π
2
and
Z π/2 q
2 3
3
Area = 9−4 2
sin θ · cos θ dθ
−π/2 2
π/2
3
Z p
= 9 − 9 sin2 θ · cos θ dθ
−π/2 2
Z π/2
9 9 π/2 cos(2θ) + 1
Z
2
= cos θ dθ = dθ
2 −π/2 2 −π/2 2
9 h sin(2θ) iπ/2 9
= +θ = π
4 2 −π/2 4
(b) The region D is symmetric about the y axis. So the centre of mass lies√on the y
axis. That is, x̄ = 0. Since D has area A = 94 π, top equation y = T (x) = 9 − 4x2
and bottom equation y = B(x) = 0, with x running from a = − 23 to b = 32 ,
Equation 2.3.5 gives us ȳ:
Z b Z 3/2
1 2 2 2
9 − 4x2 dx
   
ȳ = T (x) − B(x) dx =
2A a 9π −3/2
Z 3/2
4 4h 4 i3/2
9 − 4x2 dx = 9x − x3
 
=
9π 0 9π 3 0
3i
4 h 3 4 3 4 h 3 1 i
= 9· − · 3 = 9· −9·
9π 2 3 2 9π 2 2
2.3.3.22. Solution.4 Let’s start by sketching the region at hand. We know the
general shape of=arcsine (it’s like half a period of sine, if you swapped the x and
π

995
S OLUTIONS TO E XERCISES

y axes); we can sketch the curve y = arcsin(2 − x) by mirroring y = arcsin x about

the line x = 1.
y
π
2
y = arcsin(2 − x)

x
−1 1 2 3
y = arcsin x
− π2

If we use vertical strips, then we need two separate regions, because T (x) =
arcsin x when x ≤ 1, and T (x) = arcsin(2 − x) when x > 1. Also, we’d have to
antidifferentiate functions that have arcsine in them. Let’s think about horizontal
strips. If y = arcsin x, then x = sin y, and if y = arcsin(2 − x) then x = 2 − sin y.
For all y from − π2 to π2 , the left endpoint of a strip is given by L(y) = sin y, and
the right endpoint is given by R(y) = 2 − sin y.
First, let’s use our horizontal slices a to find the area of our region, A.
Z π/2 Z π/2



A= (2 − sin y) − (sin y) dy = 2 − 2 sin y dy
−π/2 −π/2
h iπ/2
= 2y + 2 cos y = (π + 0) − (−π + 0) = 2π
−π/2

From symmetry, it is clear that x̄ = 1. We find ȳ using Equation 2.3.5.

R π/2
−π/2
y [R(y) − L(y)] dy
ȳ =
A
R π/2
−π/2
y [(2 − sin y) − (sin y)] dy
=
2π
Z π/2
1
= y(2 − 2 sin y) dy
2π −π/2
1 π/2 1 π/2
Z Z
= y dy − (y sin y) dy
π −π/2 π −π/2
Since y is an odd function, and the domain of integration is symmetric, the first
integral evaluates to 0. Since y sin y is an even function (recall the product of two
odd functions is an even function), we can simplify our limits of integration.
2 π/2
Z
=− y sin y dy
π 0
We use integration by parts with u = y, dv = sin y dy; du = dy, v = − cos y.
Z π/2 !
2  π/2
=− − y cos y 0 + cos y dy
π 0

996
S OLUTIONS TO E XERCISES

2 π/2
=− − y cos y + sin y 0
π
2 2
= − [(0 + 1) − 0] = −
π π

a There’s also a sneaky way to find the area of A: look for a way to snip and rearrange bits of
the figure to turn it into a rectangle!

2.3.3.23. Solution. We’ll start by sketching the region.

y
y = ex

3 y = 3(x − 1)

x
1 2

If we use horizontal slices, we need to divide our figure into three regions, as in
the figure below, because the left and right functions change at the dashed lines.

997
S OLUTIONS TO E XERCISES

y
y=ex
x=log y

R(y) = 2

L(y) = log y

y=3(x−1)
3 x=1+ y3

y
R(y) = 1 + 3
1
L(x) = 0
x
1 2

If we use vertical slices, we only need two regions (shown below) to account for
the different top and bottom functions. This seems easier than three regions, so
we use vertical slices.

998
S OLUTIONS TO E XERCISES

T (x) = ex
y
y = ex

3 y = 3(x − 1)

x
1 2
B(x) = 0
B(x) = 3(x − 1)

When 0 ≤ x ≤ 2, T (x) = ex . When 0 ≤ x ≤ 1, B(x) = 0, and when 1 ≤ x ≤ 2,

B(x) = 3(x − 1).
The area of the figure is:
Z 2 Z 1 Z 2
x
A= (T (x) − B(x)) dx = (e − 0) dx + (ex − 3(x − 1)) dx
Z0 2 Z 2 0 1

= ex dx − 3(x − 1) dx
0 1
h i2  3 2
x 2
= e − (x − 1)
0 2 1
3 5
= e2 − 1 − = e2 −
2 2
Using Equation 2.3.5:
R2
x(T (x) − B(x)) dx
x̄ = 0
AZ 1 Z 2 
1 x x
= 2 x (e − 0) dx + x (e − 3(x − 1)) dx
e − 5/2 0 1
Z 2 Z 2 
1 x
= 2 xe dx − 3x(x − 1) dx
e − 5/2 0 1

999
S OLUTIONS TO E XERCISES

For the left integral, we use integration by parts with u = x, dv = ex dx; du = dx,
v = ex .
 Z 2 Z 2 
1 x 2 x 2
= 2 [xe ]0 − e dx − 3 (x − x) dx
e − 5/2 0 1
 2 !
1 2 1 1
= 2 [xex − ex ]0 − 3 x3 − x2
e − 5/2 3 2 1
  
1 2 2 8 1 1
= 2 (2e − e ) − (−1) − 3 −2− +
e − 5/2 3 3 2
2
e − 3/2
= 2 ≈ 1.2
e − 5/2

Using Equation 2.3.5 again:

R2
0
(T (x)2 − B(x)2 ) dx
ȳ =
2AZ
1 Z 2 
1 2x

2x 2


= e − 0 dx + e − 9(x − 1) dx
2(e2 − 5/2) 0 1
Z 2 Z 2 
1 2x 2
= e dx − 9(x − 1) dx
2(e2 − 5/2) 0 1
 2 h !
1 1 2x i2
= 2
e − 3(x − 1)3
2(e − 5/2) 2 0 1
 
1 1 4 1
= e − −3
2e2 − 5 2 2
4
e −7
= ≈ 2.4
4e2 − 10

Exercises — Stage 3
2.3.3.24. *. Solution. The area of the region is
∞ Z t   t
8 8 4
Z
A= dx = lim dx = lim − 2
1 x3 t→∞ 1 x
3 t→∞ x 1
 
4 4
= lim − 2 + 2 = 0 + 4
t→∞ t 1

We’ll now compute ȳ twice, once with vertical strips, as in the figure in the left
below, and once with horizontal strips as in the figure on the right below.

1000
S OLUTIONS TO E XERCISES

8
Vertical strips: The equation of the top of the region is y = T (x) = and the
x3
equation of the bottom of the region is y = B(x) = 0. Using vertical strips, as in
the figure on the left above, the y-coordinate of the centre of mass is
Z ∞
1
T (x)2 − B(x)2 dx
 
ȳ =
2A 1
Z  2
1 ∞ 8
= dx
8 1 x3
Z t 
8
= lim 6
dx
t→∞ 1 x
 t
8
= lim − 5
t→∞ 5x 1
 
8 8 8
= lim − 5 + =
t→∞ 5t 5 × 15 5
r
8 8
Vertical strips: Since y = 3 is equivalent to x = 3 , the equation of the right-
x y
2
hand side of the region is x = R(y) = 1/3 and the equation of the left hand side
y
of the region is x = L(y) = 1. The point at the top of the region is (1, 8). Thus y
runs from 0 to 8. So, using horizontal strips, as in the figure on the right above,
the y-coordinate of the centre of mass is
1 8 
Z

ȳ = y R(y) − L(y) dy
A 0
1 8  −1/3
Z

= y 2y − 1 dy
4 0
1 8  2/3
Z

= 2y − y dy
4 0
8
1 6 5/3 y 2

= y −
4 5 2 0
   
1 6 × 32 8 × 8 6 8
= − =8 −1 =
4 5 2 5 5

1001
S OLUTIONS TO E XERCISES

2.3.3.25. *. Solution. (a) The two curves cross at points (x, y) that satisfy both
y = x2 and y = 6 − x, and hence

x2 = 6 − x ⇐⇒ x2 + x − 6 = 0 ⇐⇒ (x + 3)(x − 2) = 0

So we see that the two curves intersect at x = 2 (as well as x = −3, which is to
the left of the y-axis and therefore irrelevant). Here is a sketch of A.

The top of A has equation y = T (x) = 6 − x, the bottom has equation y = B(x) =
x2 and x runs from 0 to 2. So, using vertical strips,
2
1
Z
 
x̄ = x T (x) − B(x) dx
A 0
Z 2 Z 2
1 2 3
(6x − x2 − x3 ) dx
 
= x (6 − x) − x dx =
22/3 0 22 0
2
x3 x4

3 2
= 3x − −
22 3 4 0
3 h 8 i 3 16 8
= 12 − − 4 = =
22 3 22 3 11
and
Z 2
1
T (x)2 − B(x)2 dx
 
ȳ =
2A 0
Z 2 2
(6 − x)3 x5

1 1 2 4

3
= · (6 − x) − x dx = − −
2 22/3 0 44 3 5 0
 
3 64 − 216 32 3 664 166
= − − = · =
44 3 5 44 15 55

The integral was evaluated by guessing an antiderivative for the integrand. It

could also be evaluated as
Z 2 2
x3 x5

3 2 4

3 2
36 − 12x + x − x dx = 36x − 6x + −
44 0 44 3 5
  0
3 8 32
= 72 − 24 + −
44 3 5

1002
S OLUTIONS TO E XERCISES

3 664 166
= =
44 15 55
(b) The question specifies the use of horizontal slices (as in Example 1.6.5). The
radius of the slice at height y is the x-value of the right-hand boundary of the
region at that point. So, we start by converting both equations y = 6 − x and
y = x2 into equations of the form x = f (y). To do so we solve for x in both
√
equations, yielding x = y and x = 6 − y.

• We use thin horizontal strips of width dy as in the figure above.

• When we rotate about the y-axis, each strip sweeps out a thin disk

◦ whose radius is r = 6 − y when 4 ≤ y ≤ 6 (see the blue strip in the

√
figure above), and whose radius is r = y when 0 ≤ y ≤ 4 (see the red
strip in the figure above) and
◦ whose thickness is dy and hence
◦ whose volume is πr2 dy = π(6 − y)2 dy when 4 ≤ y ≤ 6 and whose
volume is πr2 dy = πy dy when 0 ≤ y ≤ 4.

• As our bottommost strip is at y = 0 and our topmost strip is at y = 6, the

total volume is
Z 4 Z 6
π y dy + π (6 − y)2 dy
0 4

2.3.3.26. *. Solution. (a) Here is a sketch of the specified region, which we shall
call R.

1003
S OLUTIONS TO E XERCISES

The top of R has equation y = T (x) = ex , the bottom has equation y = B(x) = −1
and x runs from 0 to 1. So, using vertical strips, we see that R has area
Z 1 Z 1 Z 1
   x   x 
A= T (x) − B(x) dx = e − (−1) dx = e + 1 dx
0 0 0
 x 1
= e +x 0 =e

and
Z 1
1
T (x)2 − B(x)2 dx
 
ȳ =
2A 0
1
1 1  2x 1 e2x
Z 

= e − 1 dx = −x
2e 0 2e 2 0
1 e2
 
1 e 3
= −1− = −
2e 2 2 4 4e
(b) To compute the volume when R is rotated about the line y = −1
• we use thin vertical strips of width dx as in the figure above.

• When we rotate about the line y = −1, each strip sweeps out a thin disk

◦ whose radius is r = T (x) − B(x) = ex + 1 and

◦ whose thickness is dx and hence
◦ whose volume is πr2 dx = π(ex + 1)2 dx.

• As our leftmost strip is at x = 0 and our rightmost strip is at x = 1, the total

volume is
Z 1 Z 1
2
π x
(e + 1) dx = π (e2x + 2ex + 1) dx
0 0
 2x 1
e x
=π + 2e + x
2 0
 2 
e  1
=π + 2e + 1 − +2+0
2 2
 e2 3
=π + 2e −
2 2

1004
S OLUTIONS TO E XERCISES

2.3.3.27. Solution. By symmetry, ȳ = 1.5. We can’t immediately use Equa-

tion 2.3.5 to find x̄, because the density is not constant. Instead, we’ll go through
the derivation of Equation 2.3.5, to figure out what to do with a non-constant
density. (This is a good time to review Questions 9 and 10 in this section.)
Our model is that we’re making a rod R that reaches from x = 0 to x = 4, and
the mass of the section of the rod along [a, b] is equal to the mass of the strip of
our rectangle along [a, b]. If we have a formula ρ(x) for the density of R, we can
find the centre of mass of R, which is also the x-coordinate of the centre of mass
of the rectangle.
A thin vertical strip of the rectangle with length dx at position x has area 3dx m2
and density x2 kg/m2 , so it has mass 3x2 dx kg. Therefore, a short section of R at
position x with length dx ought to have mass 3x2 dx kg as well. Then its density
2
at x is ρ(x) = 3xdxdxmkg = 3x2 kg/m.
Now, we can use Equation 2.3.4 to find the centre of mass of the rod, which is
also the x-coordinate of the centre of mass of our rectangle:

4 4
R4 R4 3 
xρ(x) dx 3x3 dx x 3 · 43
x̄ = R0 4 = R04 = 4 40 = 3 = 3
ρ(x) dx 3x2 dx x3 0 4
0 0

The centre of mass of our rectangle is (3, 1.5).

2.3.3.28. Solution. By symmetry, the x-coordinate of the centre of mass will be
x̄ = 0; that is, exactly in the middle, horizontally. To find the y-coordinate of the
centre of mass, we need to consider the origin of Equation 2.3.5.
We can make vertical strips or horizontal strips. A vertical strip of the circle has
a density that varies from the bottom of the strip to the top, but a horizontal
strip has a constant density (assuming the strip is very thin). So it seems that
horizontal strips in this case will be the easier route.
Following the derivation of Equation 2.3.5, we model our circle as a vertical rod
R, filling the y-interval [0, 6]. A portion of the rod with a ≤ y ≤ b should have the
same mass as the portion of the circle with a ≤ y ≤ b. To achieve this, we slice
the circle into thin horizontal strips of thickness dy, calculate their mass, then use
that to find ρ(y), the density of R.
{First, let’s find a formula for the mass of a thin horizontal strip of the circle at
position y with height dy.}

y dy

x
x

1005
S OLUTIONS TO E XERCISES

The circle with radius 3 centred at (0, 3) hasp equation x2 + (y − 3)2 = 9. So,
the right half of the circle haspequation x = 9 − (y − 3)2 , and the left half of
circle has equation x = − 9 − (y − 3)2 . So, the width
the p p of a strip at height y
2 2 2
is 2 9 − (y − 3) m. Its height is dy m, so its area is 2 9 − (y − 3) dy m . Its
density is 2 + y mkg2 , so its mass is 2(2 + y) 9 − (y − 3)2 dy kg.
p

Now we can find ρ(y), the density of R at position p y. The mass of the section
of R at position y with length dy is 2(2 + y) 9 − (y − 3) 2
√ dy kg (the mass of
2(2+y) 9−(y−3)2 dykg
the strip in the paragraph above), so its density is dym
= 2(2 +
kg
p
y) 9 − (y − 3)2 m = ρ(y).
Now, Equation 2.3.4 will tell us the centre of mass of R, which is also the y-
coordinate of the centre of mass of the circle.

Rb R6 p
yρ(y) dy y × 2(2 + y) 9 − (y − 3)2 dy
ȳ = Ra b = 0R 6 p
ρ(y) dy 0
2(2 + y) 9 − (y − 3)2 dy
R 6a p
y(2 + y) 9 − (y − 3)2 dy
= R0 6 p
0
(2 + y) 9 − (y − 3)2 dy

To make things look a little cleaner, we use the substitution u = y − 3, du = dy.

Then the limits of integration become −3 and 3, respectively, and y = u + 3.
(Geometrically, we’re re-centring the circle at the origin, instead of at the point
(0,3).)
R3 √
−3
(u + 3)(2 + u + 3) 9 − u2 du
= R3
√
−3
2 + u + 3 9 − u2 du
R3 2

√
u + 8u + 15 9 − u2 du N
= −3 R 3 = (∗)
D

√
u + 5 9 − u2 du
−3

Let’s start by finding D, the integral of the denominator. If we break it into two
pieces, we can use symmetry and geometry to evaluate it.
Z 3 √ Z 3 √
D= u 9 − u2 du + 5 9 − u2 du
−3 −3

The left integrand is odd, so its integral over a symmetric interval is 0. (You can
also evaluate this using the substitution w = 9 − u2 , dw = −2u du.) The right
integral represents the area underneath half a circle of radius 3, centred at the
origin.

1 45
D = 0 + 5 · π · 32 = π
2 2
R3
Now, let’s evaluate our numerator integral from (∗), N = −3 u2 + 8u +

√
15 9 − u2 du. If we break it into three pieces, we can simplify the integration

1006
S OLUTIONS TO E XERCISES

somewhat.
Z 3 √
2
Z 3 √ Z 3 √
N= 2
u 9 − u du + 8 2
u 9 − u du + 15 9 − u2 du
−3 −3 −3

The first integrand is even, with a symmetric interval of integration, so we can

simplify its limits of integration a little bit. The middle integrand is odd, so its
integral over the symmetric interval [−3, 3] is zero. The last integral is the area of
half a circle of radius 3.
Z 3 √
N =2 u2 9 − u2 du + 0 + 15 · π · 32
0
Z 3 √
135
= π+2 u2 9 − u2 du
2 0

The remaining integral has a quadratic function underneath a square root with
no obvious substitution, so we use a trigonometric substitution. Let u = 3 sin θ,
du = 3 cos θ dθ. Note 3 sin(0) = 0 and 3 sin(π/2) = 3, so the limits of integration
become 0 and π2 .
Z π/2
135
2 q
2
N= π+2 3 sin θ 9 − 3 sin θ · 3 cos θ dθ
2 0
Z π/2
135 p
= π+2 9 sin2 θ · 9 − 9 sin2 θ · 3 cos θ dθ
2 0
135
Z π/2 √
= π + 54 sin2 θ · 9 cos2 θ · cos θ dθ
2 0
Z π/2
135
= π + 54 sin2 θ · 3 cos θ · cos θ dθ
2 0
Z π/2
135
= π + 162 sin2 θ · cos2 θ dθ
2 0

2
Using the identity sin(2θ) = 2 sin θ cos θ, we see sin2 θ cos2 θ = sin θ cos θ =
1
4
sin2 (2θ)
π/2
135 1 2
Z
N= π + 162 sin (2θ) dθ
2 0 4

Now, we use the identity sin2 x = 12 (1 − cos(2x)), with x = 2θ.

Z π/2
135 1

N= π + 162 1 − cos(4θ) dθ
2 0 8
Z π/2
135 81
= π+ 1 − cos(4θ) dθ
2 4 0
 π/2
135 81 1
= π+ θ − sin(4θ)
2 4 4 0

1007
S OLUTIONS TO E XERCISES

135 81  π 
= π+
2 4 2
621
= π
8
Now, using equation (∗), we find ȳ:
621
N 8
π 69
ȳ = = 45 = = 3.45
D 2
π 20

Let’s quickly check that this makes sense: if the circle has uniform density, its
centre of mass would lie at (0, 3). Since it’s denser at the top, the centre of mass
should be higher, and indeed 3.45 is higher than 3 (without being so high it’s
above the entire circle).
2.3.3.29. Solution.

a To find the centre of mass of the rod R, we need to know its density at
height y, ρ(y). Since the mass of a section of R is the same as the volume
of a section of the cone, let’s find the volume of a thin horizontal slice of
the cone at height y, with thickness dy. To find its radius s, we use similar
triangles. The diagram below represents a vertical cross-section of the cone.

y
s

x
r

r s
Since h
= the radius of our slice at height y is s = hr (h − y). Then the
h−y
,

2
volume of the slice is πs2 dy = π hr (h − y) dy. Correspondingly, the mass

2
of the piece of the rod at position y with length dy is π hr (h − y) dy, so its
density is

2
π hr (h − y) dy r 2
ρ(y) = =π (h − y) .
dy h

1008
S OLUTIONS TO E XERCISES

Now, we can find the centre of mass of R:

Rh Rh
2
0
yρ(y) dy 0
yπ hr (h − y) dy
ȳ = R h = Rh
2
0
ρ(y) dy 0
π hr (h − y) dy
Rh
r2
h2
π 0 y (h − y)2 dy
= 2 Rh
r
h2
π 0 (h − y)2 dy
Rh 2
(h y − 2hy 2 + y 3 ) dy
= R0 h
0
(h2 − 2hy + y 2 ) dy
h 2 ih
h 2 2h 3 1 4
2
y − 3
y + 4
y
0
= 
1 3 h

2 2
h y − hy + 3 y 0
h4 4 4
2
− 2h3 + h4
=
h3 − h3 + 13 h3
1
h4 2
− 32 + 1
4 h
= · 1 =
h3 3
4
h
So, the centre of mass of the cone occurs 4
metres above its base.
Remark: it is quite interesting that the centre of mass does not depend on
the radius of the cone!

b To find the centre of mass of a truncated cone, we simply consider a trun-

cated rod. If the top h − k metres are missing, then the height of the cone
(and also the rod) is k. Then the centre of mass has height:
Rk Rk r

2
yρ(y) dy yπ (h − y) dy
ȳ = R0 h = R0 h h
2
r


0
ρ(y) dy 0
π h
(h − y) dy
r2
Rk 2
2π y (h − y) dy
= h 2 R0 k
r
h2
π 0 (h − y)2 dy
Rk 2
(h y − 2hy 2 + y 3 ) dy
= R0 k
0
(h2 − 2hy + y 2 ) dy
h 2 ik
h 2 2h 3 1 4
2
y − 3 y + 4y
0
= 
1 3 k

2 2
h y − hy + 3 y 0
1 2 2
2
h k − 32 hk 3 + 14 k 4
=
h2 k − hk 2 + 13 k 3
1 2
2
h k − 23 hk 2 + 41 k 3
=
h2 − hk + 13 k 2
2.3.3.30. Solution. To use the result of Question 29, we need to now the dimen-
sions of the cone that was truncated to make the hourglass. The bottom (or top)

1009
S OLUTIONS TO E XERCISES

half of our hourglass has base radius 5 cm, height 9 cm, and top radius 0.5 cm.
Imagine extending it to a full cone. Let t be the distance from the top of the half
hourglass to the tip of the full cone.

0.5

Using similar triangles,

t t+9
=
0.5 5
1
so 5t = (t + 9)
2
4.5t = 4.5
t=1

Then the height of the full cone (that we imagined truncating to make half of the
hourglass) is h = 10 cm.
Before the hourglass is turned over, the sand forms a truncated cone of height 6
cm. So, it’s the bottom k = 6 cm of a cone of height h = 10 cm. Using the result
of Question 29, its centre of mass is at height:
1 2
2
h k − 23 hk 2 + 14 k 3 1
2
102· 6 − 32 10 · 62 + 14 63 57
= = ≈ 2.2
h2 − hk + 31 k 2 2
10 − 10 · 6 + 3 6 1 2
26

Next, let’s find the centre of mass of the sand after it’s been rotated. We have to be
a little careful with our vocabulary here: usually we imagine a cone sitting on its
base, with its tip pointing up. The upturned sand is in the opposite configuration.
When we say the “base” of the cone, we mean the larger horizontal face — the
top of the sand as it sits in the hourglass.
The formula we have from Question 29 gives us our centre of mass as a distance
from the base of the truncated cone (that is, the distance from the top of the
upturned sand). If k is the height the sand actually occupies, then we were told
we may assume k = 8.8 cm. It’s missing its “tip” of height 1 cm, so h, the height
of the “untruncated” cone, is 9.8 cm. Using our model from Question 29, we
don’t care about the empty, uppermost piece of the hourglass. The shape of the
sand is of a cone of height 9.8 cm (not 10 cm), with a tip of height 1 cm chopped
off.

1010
S OLUTIONS TO E XERCISES

0.2 cm of empty hourglass:

ignored in our calculation of sand’s centre of mass

sand: height k = 8.8 cm

h = 9.8

cut-off tip: height 1 cm

1 2
2
h k − 32 hk 2 + 14 k 3 1
2
9.82· 8.8 − 23 9.8 · 8.82 + 41 8.83
= ≈ 2.443
h2 − hk + 31 k 2 9.82 − 9.8 · 8.8 + 13 8.82

That is, the centre of mass of the upturned sand is about 2.443 centimetres be-
low its top, which is at height 8.8 + 10 = 18.8 cm above the very bottom
of the hourglass. So, the centre of mass of the upturned sand is at height
y = 18.8 − 2.443 = 16.357 cm.
57
Now, we have our model: the sand, viewed as a point mass, is moved from y = 26
to y = 16.357 cm. That is, it moved about 14.165 cm, or about 0.14165 m. It has a
mass of 0.6 kg, so the force required to lift it against gravity is
 m 
(0.6kg) × 9.8 2 × (0.14165m) ≈ 0.833newtons
sec
2.3.3.31. Solution. The techniques of Section 2.1 get pretty complicated here, so
we will use the techniques we developed in Questions 6, 29 and 30 in this section.
That is, (1) find the centre of mass of the water in its starting and ending positions,
and then (2) compute the work done as the work moving a point mass with the
weight of the water from the first centre of mass to the second. For the centre
of mass, all we need to know is the height — for one thing, we could find the
other coordinates by symmetry, but we don’t need them. The height moved by
the water is all that matters if we’re calculating the work done opposing gravity.
Let’s start by calculating the volume of the water. The volume of a sphere of
radius 1 is 43 π · 13 , so the volume of water is 32 π m3 .
Then the mass of the water is 2000 3
π kg.
Next, we calculate the centre of mass of Tank A, and the work done to pump the
water out of Tank A to a height of 3 metres. Symmetry alone won’t tell us the
height of the centre of mass. We’ll show you two ways to go about this.

• Option 1: As in Question 29, we’ll model the tank of water as a vertical rod,
along the y-axis spanning the interval [0, 1], such that the mass of a piece of
the rod along [a, b] is the same as the mass of the water from height y = a
to height y = b. Then, the centre of mass of the rod will be the same as the
centre of mass of the water.
Consider a horizontal slice of water at height y, with thickness dy. If the

1011
S OLUTIONS TO E XERCISES

radius of this slice is r(y), then the volume of the slice is πr(y)2 dy m3 , so its
mass is 1000πr(y)2 dy kg. Then the mass of the slice of the rod at position y
with length dy is 1000πr(y)2 dy kg, so its density ρ(y) is

1000πr(y)2 dy kg kg
ρ(y) = = 1000πr(y)2 .
dy m m

So, let’s find r(y), the radius of the slice of water at height y.

r
y

1 x

p
Using the Pythagorean Theorem, r = 1 − y 2 . Therefore,

ρ(y) = 1000π(1 − y 2 )

We use Equation 2.3.4 to calculate the centre of mass of the rod, which is
the height of the centre of mass of Tank A:
R1 R1
yρ(y) dy 1000πy(1 − y 2 ) dy
ȳA = R0 1 = R0 1
0
ρ(y) dy 0
1000π(1 − y 2 ) dy
1 4 1
R1  1 2
 1
0
(y − y 3
) dy 2
y − 4
y 0 2
− 41 3
= R1 =  1 = 1 = m
1 1− 3 8

0
(1 − y 2 ) dy y − y3
3 0

From here, we can find the work done moving pumping the water to a
height of 3 metres. We’ve moved the centre of mass from ȳA = 83 metres to
3 metres.
    
2000 3 m 
W = πkg × 3 − m × 9.8 2
3 8 sec
= 17, 150πJ

• Option 2: We can use the techniques of Section 2.1 to calculate the amount
of work it takes to pump the water from tank A to a height of 3 metres. That

1012
S OLUTIONS TO E XERCISES

solves part (a), and we can use the amount of work to figure out the centre
of gravity of the water in Tank A to help us solve part (b).
At height y, a horizontal
p layer of water in Tank A forms a disk with thick-
ness dy and radius 1 − y 2 . (The radius comes from the Pythagorean The-
orem — see the diagram below.)

y
1 x

p 2
The volume of the layer at height y is π 1 − y 2 dy = π(1 − y 2 ) dy m3 ,
so its mass is 1000π(1 − y 2 ) dy kg.
The layer at height y needs to be pumped a distance of 3 − y metres. So, the
work involved pumping the layer at height y is:
dW = 1000π(1 − y 2 ) dykg × (3 − ym) × 9.8m/sec2

= 9800π(y 3 − 3y 2 − y + 3) dy J
Then the work involved pumping out the entire tank to a height of 3 metres
is:
Z 1
W = 9800π(y 3 − 3y 2 − y + 3) dy
0
 1
1 4 3 1 2
= 9800π y − y − y + 3y
4 2 0
= 17, 150πJ

This gives us an answer to part (a). To find the centre of mass of the water
in Tank A, note that the work done is equivalent to moving a point mass
from the centre of mass of the tank to a height of 3 metres. We know the
water in Tank A has mass 2000 3
π kg. So, if ȳA is the centre of mass of the
water in Tank A:
 
2000
πkg × (3 − ȳA m) × 9.8m/sec2


W =
3
 
2000
17, 150π = π (3 − ȳA ) (9.8)
3
21
= 3 − ȳA
8
3
ȳA = m
8
1013
S OLUTIONS TO E XERCISES

Next let’s calculate the centre of mass of the water in Tank B. Since the volume
of the water in Tank B is 23 π m3 , and the base of Tank B has area 1 m2 , the height
of the water in Tank B is 32 π m. Since the water is of uniform density, and Tank
B has uniform horizontal cross-sections, by symmetry the centre of mass of the
water in Tank B is at
1
ȳB = π m.
3
Now, we can calculate the work done by moving the water directly from Tank A
to its final position in Tank B. The work done moving a point mass of 2000 3
π kg
a distance of ȳB − ȳA = 31 π − 83 m against the gravity, g = 9.8 m/sec2 , is:
   
2000 1 3
π − m × 9.8m/sec2


W = π kg ×
3 3 8
2450
= π (8π − 9) ≈ 13, 797 J
9
Finally, the “wasted” work is:

2450
∆W = 17, 150π − π (8π − 9)
 9 
8π − 9
= 2450π 7 −
9
 
8π
= 2450π 8 −
9
 π
= 19, 600π 1 −
9
As a percentage of 17,150π, this is:

!
19, 600π 1 − π9
waste = × 100
17, 150π
8 π
= 1− × 100 ≈ 74%
7 9
2.3.3.32. Solution. Using Equation 2.3.5 with T (x) = 2x sin(x2 ) and B(x) = 0,
Z √ π/2
2x2 sin(x2 ) dx
x̄ = Z0 √
π/2
2x sin(x2 ) dx
0
2
p with the substitution u = x , du =
We can evaluate the bottom integral exactly
2xdx. When x = 0, u = 0, and when x = π/2, u = π/2.

Z √π/2 Z π/2 h iπ/2

2
2x sin(x ) dx = sin u du = − cos u =1
0 0 0

1014
S OLUTIONS TO E XERCISES

So,
Z √π/2
x̄ = 2x2 sin(x2 ) dx
0
2 2
R
Evaluating the integral x sin(x ) dx is not so simple a , so we use a numerical
approximation. Since we’re given an upper bound on the fourth derivative, we
decide to use Simpson’s rule. The error involved using Simpson’s rule with n in-
5
tervals is at most L(b−a)
p
180n4 . For our approximation, a = 0 and b = π/2. According
d 4
2 2
to the information given in the problem statement, dx 4 {2x sin(x )} ≤ 415 over

the interval 0, π2 , so we set L = 415.

 p 
1
We want our final error to be no more than 100 , so we want to find an even n such
that:
pπ
5
415 2
−0 1
≤
180n4 100
π 5/2


415 · 100 2075π 5/2
n4 ≥ 2
= √
180 36 2
s
2075π 5/2
n≥ 4 √ ≈ 5.17
36 2
1
So, n = 6 intervals suffices. Then ∆x = b−a
pπ
6
= and our grid points are
1 1 1 2
p π6 2 5 p π
x0 = 0, x1 = 6 2 , x2 = 3 2 , x3 = 2 2 , x4 = 3 2 , x5 = 6 2 , and , x6 = π2 .
pπ pπ pπ p

0 1 1 1 2 5
pπ pπ pπ pπ pπ pπ
6 2 3 2 2 2 3 2 6 2 2

Following Equation 1.11.9, the Simpson’s rule approximation of

Z √π/2
2x2 sin(x2 ) dx is:
0

∆x h i
f (x0 ) + 4f (x1 ) + 2f (x2 ) + 4f (x3 ) + 2f (x4 ) + 4f (x5 ) + f (x6 )
3 r 
1 π 1 2π π 2π π
= · 0+4× sin +2× sin
6 2 3 72 72 18 18
 
2π  π  8π 4π
+4× sin +2× sin
8 8 18 18
  
50π 25π 2π  π
+4× sin + sin
72 72 2 2
r   
1 π π  π  2π  π   π  8π 2π
= sin + sin + π sin + sin
18 2 9 72 9 18 8 9 9
  
25π 25π
+ sin +π
9 72

1015
S OLUTIONS TO E XERCISES
r   
π π 1 π 2 π π  8 2π
= sin + sin + sin + sin
18 2 9 72 9 18 8 9 9
  
25 25π
+ sin +1
9 72
r   
π π π π π  2π
= sin + 2 sin + 9 sin + 8 sin
162 2 72 18 8 9
  
25π
+ 25 sin +9
72
≈ 0.976

The absolute error in our answer is at most:

p 5 √ 5
L(b − a)5 415 × π2 82 π
= = √ ≈ 0.005
180n4 180 × 64 186624 2
Remark: combining the error with our approximation, we see the actual value of
x̄ is in the interval

[0.976 − 0.005, 0.976 + 0.005] = [0.971, .981]

A computer algebra system approximates x̄ as 0.977451.

a Indeed, the antiderivative of 2x2 sin(x2 ) is not expressible as an elementary function.

2.4 · Separable Differential Equations

2.4.7 · Exercises

Exercises — Stage 1
2.4.7.1. Solution.
dy
a If y = 5(ex − 3x2 − 6x − 6), then dx
= 5(ex − 6x − 6). Let’s see whether this
is equal to y + 15x2 :

y + 15x2 = 5(ex − 3x2 − 6x − 6) + 15x2

= 5(ex − 3x2 − 6x − 6 + 3x2 )
= 5(ex − 6x − 6)
dy
=
dx
So, y = 5(ex − 3x2 − 6x − 6) is indeed a solution to the differential equation
dy
dx
= y + 15x2 .

1016
S OLUTIONS TO E XERCISES

−2 dy 4x
b If y = , then dx
= . Let’s see whether this is equal to xy 2 :
x2 +1 (x + 1)2
 2
2 −2
xy = x 2
x +1
4x
= 2
(x + 1)2
dy
=
dx
−2 dy
So, y = is indeed a solution to the differential equation dx
= yx2 .
x2
+1
dy √
c If y = x3/2 + x, then dx = 32 x + 1.
2 2
3√ 3√
 
dy dy
+ = x+1 + x+1
dx dx 2 2
9 9√
= x+ x+2
4 2
dy
6=
dx
dy 2 dy
So, y = x3/2 + x is not a solution to the differential equation


dx
+ dx
= y.
2.4.7.2. Solution.
 
dy dy sin y
a 3y dx = x sin y can be written as dx
= x 3y
, which fits the form of a
sin y
separable equation with f (x) = x, g(y) = 3y
.
dy
b dx
= ex+y = ex ey which fits the form of a separable equation using f (x) =
e , g(y) = ey .
x

dy dy
c + 1 = x can be written as dx
dx
= (x − 1), which fits the form of a sepa-
rable equation using f (x) = x − 1, g(y) = 1. (We can solve it by simply
antidifferentiating.)
dy 2 dy
+ x2 = 0 is a perfect square.


d Notice the left side of the equation dx − 2x dx
dy

2 dy
So, this equation is equivalent to dx − x = 0, that is, dx = x. This has the
form of a separable equation with f (x) = x, g(y) = 1.
2.4.7.3. Solution. The mnemonic allows us to skip from the separable differen-
tial equation we want to solve (very first line) to the equation
1
Z Z
dy = f (x) dx
g(y)
So, the mnemonic is just a shortcut for the substitution we performed to get this
point.
We also generally skip the explanation about C1 and C2 being replaced with C.

1017
S OLUTIONS TO E XERCISES

2.4.7.4. Solution. To say y = f (x) + C is a solution to the differential equation

means:
d
{f (x) + C} = x(f (x) + C)
dx
d d
Since y = f (x) is a solution, we know dx
{f (x)} = xf (x). Also, dx
{f (x) + C} =
d d
dx
{f (x)}. So, dx {f (x) + C} = xf (x).

xf (x) = x(f (x) + C)

0 = xC

Our equation should hold for all x in our domain, and for the derivative to y with
respect to x to make sense, our domain should not be a single point. So, there is
some x in our domain such that x 6= 0. Therefore, the C must be zero. So, f (x) + C
is not a solution to the differential equation for any constant C.
When we’re finding a general antiderivative, we add “+C” at the end. When
we’re finding a general solution to a differential equation, the “+C” gets added
when we antidifferentiate–we don’t add another one at the end of our work.
2.4.7.5. Solution.

a Since |y| ≥ 0 no matter what y is, we see Cx ≥ 0 for all x in the domain
of f (x). Since C is positive, that means the domain of f (x) only includes
nonnegative numbers. So, the largest possible domain of f (x) is [0, ∞).

b None exists.
The graph of Cx is given below for some positive constant C, also with the
graph of −Cx. If y = f (x) were sometimes the top function, and other times
the bottom function, then there would be a jump discontinuity where it
switched. Then the derivative of f (x) would not exist, violating the second
property.

y
y = Cx

y = −Cx

A tiny technical note is that it’s possible that f (x) = Cx when x = 0 and

1018
S OLUTIONS TO E XERCISES

f (x) = −Cx when x > 0 (or vice-versa). This would not introduce a jump
discontinuity, but it also does not satisfy that f (x) > 0 for some values of x.

Remark: in several instances below, solving a differential equation will lead us

to conclude something like |y| = g(x). In these cases, we choose either y = g(x),
or y = −g(x), but not y = ±g(x) (which is not a function) or that y is sometimes
g(x), and other times −g(x). The reasoning above somewhat explains this choice:
dy
if y were sometimes positive and sometimes negative, then dx would not exist at
the values of x where the sign of y switches, unless that switch occurrs at a root
of g(x). Since that’s a pretty specific occurrence, we usually feel safe ignoring it
to avoid getting bogged down in technical details.
2.4.7.6. Solution. Let Q(t) be the quantity of morphine in a patient’s blood-
stream at time t, where t is measured in minutes.
Using the definition of a derivative,

dQ Q(t + h) − Q(t) Q(t + 1) − Q(t)

= lim ≈
dt h→0 h 1

So, dQ
dt
is roughly the change in the amount of morphine in one minute, from t to
t + 1.
The sentence tells us that the change in the amount of morphine in one minute is
about −0.003Q, where Q is the quantity in the bloodstream. That is:

dQ
= −0.003Q(t)
dt
2.4.7.7. Solution. If p(t) is the proportion of times speakers use the new form,
measured between 0 and 1, then 1 − p(t) is the proportion of times speakers use
the old form.
The law, then, states that dp


dt
is proportional to p(t) × 1 − p(t) . When we say two
quantities are proportional, we mean that one is a constant multiple of the other.
So, the law says
dp

= αp(t) 1 − p(t)
dt
for some constant α.
Remark: it follows from this model that, when a new form is either very rare or
entirely ubiquitous, the rate of change of its adoption is small. This makes sense:
if the new form is used all the time (p(t) ≈ 1), there’s nobody left to convert; if
the new form is almost never used (p(t) ≈ 0) then people don’t know about it, so
they won’t pick it up.
2.4.7.8. Solution.
0
a When y = 0, y 0 = 2
− 1 = −1.
2
b When y = 2, y 0 = 2
− 1 = 0.
3
c When y = 3, y 0 = 2
− 1 = 0.5.

1019
S OLUTIONS TO E XERCISES

d The small red lines have varying slopes. The red lines on points with y-
coordinate 2 have slopes of 0; this matches y 0 when y = 0, as we saw above.
The red lines on points with y-coordinate 0 have slopes of approximately
−1; again, this matches what we found for y 0 when y = 0.
The red lines correspond to a tiny section of y(x), if y(x) passes through
that point. So, we can sketch a possible curve y(x) satisfying the equation
by starting somewhere, then following the slopes.
For example, suppose we start at the origin.

x
1

Then our function is decreasing at that point, which leads us to a coordinate

where (as we see from the red marks) the function is decreasing slightly
faster.

x
1

Following the red marks leads us down even further, so our function y(x)
might look something like this:

1020
S OLUTIONS TO E XERCISES

x
1

However, we didn’t have to start at the origin. Suppose y(0) = 3. Then at

x = 0, y is increasing, with slope 12 .

x
1

Our red marks run out that high up, but we now y 0 = 12 y − 1, so y 0 increases
as y increases. That means our function keeps getting steeper and steeper,
possibly something like this:

1021
S OLUTIONS TO E XERCISES

x
1

If y(0) = 2, we see another possible curve is the constant function y(x) = 2.

Remark: from Theorem 2.4.4, we see the solutions to the equation y 0 = 21 y − 1 =

1
2
(y − 2) are of the form y(x) = Cex/2 + 2 for some constant C. Check that the
curves you’re sketching look exponential.
2.4.7.9. Solution.
1
a If y(1) = 0, then y 0 (1) = 0 − 2
= − 12 .
1
b If y(1) = 2, then y 0 (1) = 2 − 2
= 32 .
1
c If y(1) = −2, then y 0 (1) = −2 − 2
= − 52 .

d There are 7 × 7 = 49 points on the grid; we don’t want to make 49 separate

calculations. Let’s find some shortcuts.

• If y 0 (x) = 0, then y = x2 , which applies to the points (0, 0), (2, 1), (4, 2)
and (6, 3). These are the orange dots in the sketch below.
• If y 0 (x) = 1, then y = 1 + x2 , which applies to the points (0, 1), (2, 2),
and (4, 3). (Note these are exactly 1 unit above the points with y 0 = 0.)
These are the red dots in the sketch below.
• If y 0 (x) = −1, then y = −1 + x2 , which applies to the points (0, −1),
(2, −2), and (4, −3). (Note these are exactly 1 unit below the points
with y 0 = 0.) These are the yellow dots in the sketch below.
• If x increases and y stays the same, y decreases.
• If y increases and x stays the same, y increases.
1
• If we draw a straight line of slope on our sketch, for every point on
2
that line, our mark has the same slope: for instance, the points where
we draw a mark with slope 0 are (0, 0), (2, 1), and (4, 2), and these all
lie on the line f (x) = x2 .

1022
S OLUTIONS TO E XERCISES

This is enough to give us a pretty good sketch. The points whose slopes we
found explicitly have dots; the rest can be sketched as either steeper or less
steep than what’s near them.

x
1

e To sketch a possible graph of y(x), we choose a point (x, y(x)), then follow
the red lines.
For example, if we suppose that y(4) = 2, then near (4, 2), the lines tell us
y(x) is fairly flat; and it is increasing to the left of x = 4, and decreasing to
the right.

x
1

Following the red lines a little farther in each direction brings us some-
where like this:

1023
S OLUTIONS TO E XERCISES

x
1

Extending yet further, we might sketch something like the following:

x
1

By choosing another point (x, y(x)) to be on the curve, we might find other
potential curves. Some examples are shown below.

y Passing through (0, 0):

x
1

1024
S OLUTIONS TO E XERCISES

y Passing through (0, 1):

x
1

y Passing through (6, 3):

x
1

y Passing through (6, −3):

x
1

Remark: the differential equation y 0 = y − x2 is not separable, so we haven’t talked

about how to solve it. The solutions have the form y(x) = Cex + x+1 2
. You can
verify that these functions satisfy y 0 = y − x2 .

1025
S OLUTIONS TO E XERCISES

Exercises — Stage 2
2.4.7.10. *. Solution. Rearranging, we have:

ey dy = 2x dx.

Integrating both sides:

Z Z
y
e dy = 2x dx

e y = x2 + C

Since y = log 2 when x = 0, we have

elog 2 = 02 + C
2 = C,

and therefore

e y = x2 + 2
y = log(x2 + 2)
2.4.7.11. *. Solution. Using separation of variables:

dy xy
= 2
dx x +1
dy x
= 2 dx
y x +1
dy x
Z Z
= 2
dx
y x +1
1
log |y| = log(1 + x2 ) + C
2
To satisfy y(0) = 3, we need log 3 = 21 log(1 + 0) + C, so C = log 3. Thus:

1
log |y| = log(1 + x2 ) + log 3
2 √
= log 1 + x2 + log 3
√
= log 3 1 + x2

So,
√
|y| = 3 1 + x2

1026
S OLUTIONS TO E XERCISES

We are told to find a functon√ y(x). So far, we have two possible

√ functions from the
2 2
work above: maybe √ y = 3 1 + x , and maybe y = −3 1 + x . It’s important to
note that y = ±3 1 + x2 is not a function: for an equation to represent a function,
for every input in the domain, there must only be one output. That is, functions
pass the vertical line test. (See the CLP-1 text for a definition of the vertical line
test and a formal√definition of a function.)
√ So, we need to decide whether our
2 2
function is y = 3 1 + x or y = −3 1 + x . Since y(0) = 3, we conclude
√
y(x) = 3 1 + x2
2.4.7.12. *. Solution. The given differential equation is separable and we solve
it accordingly.
y
y 0 = e 3 cos t
e−y/3 dy = cos t dt
Z Z
−y/3
e dy = cos t dt

−3e−y/3 = sin t + C
1 sin t + C
=
ey/3 −3
y/3 −3
e =
C +sin t 
y −3
= log
3 C + sin t
 
−3
y(t) = 3 log
C + sin t

for any constant C.

Since the domain of logarithm is (0, ∞), the solution only exists when C + sin t <
0.
2.4.7.13. *. Solution. The given differential equation is separable and we solve
it accordingly.
2
dy 2 2 xex
= xex −log(y ) = 2
dx y
2
y 2 dy = xex dx
Z Z
2 2
y dy = xex dx

2
We can guess the antiderivative of xex , or use the substitution u = x2 , du = 2xdx.

y3 1 2
= ex + C 0
3 2
3 2
y 3 = ex + 3C 0
2

1027
S OLUTIONS TO E XERCISES

Since C 0 can be any constant in (−∞, ∞), then also 3C 0 can be any constant in
(−∞, ∞), so we replace 3C 0 with the arbitrary constant C.

3 2
y 3 = ex + C
2
r
3 3 x2
y= e +C
2
for any constant C.
2.4.7.14. *. Solution. The given differential equation is separable and we solve
it accordingly.

dy
= xey
dx
dy
y
= x dx
Z e
dy
Z
= x dx
ey
1
−e−y = x2 + C
2
1
e−y = − x2 − C
2
Since C can be any constant in (−∞, ∞), then also −C can be any constant in
(−∞, ∞), so we write C instead of −C.
1
e−y = C − x2
2
x2

−y = log C −
2
x2
 
y = − log C −
2

for any constant C.

2
The solution only exists for C − x2 > 0. For this to happen, we need C > 0, and
√
then the domain of the function is those values x for which |x| < 2C.
2.4.7.15. *. Solution. The given differential equation is separable and we solve
it accordingly. Cross–multiplying, we rewrite the equation as

dy
y2 = ex − 2x
dx
y 2 dy = (ex − 2x) dx.

Integrating both sides, we find

Z Z
2
y dy = (ex − 2x) dx

1028
S OLUTIONS TO E XERCISES

1 3
y = ex − x2 + C
3
Setting x = 0 and y = 3, we find 13 33 = e0 − 02 + C and hence C = 8.

1 3
y = ex − x2 + 8
3
y = (3ex − 3x2 + 24)1/3
2.4.7.16. *. Solution. This is a separable differential equation that we solve in
the usual way.

dy
= −xy 3
dx
dy
− 3 = xdx
y
dy
Z Z
− 3 = xdx
y
y −2 x2
− = +C
−2 2
y −2 = x2 + 2C. (∗)

To have y = − 14 when x = 0, we must choose C to obey

 1 −2
− = 0 + 2C
4
16 = 2C

So, from (∗),

y −2 = x2 + 2C = x2 + 16
1
y2 = 2
x + 16
Now, we have two potential candidates for y(x):

1 1
y=√ OR y = −√
x2 + 16 x2 + 16

We know y = − 41 when x = 0. The only function above that fits this is

1
y = −√
x2 + 16
1
So, f (x) = − √ .
x2 + 16

1029
S OLUTIONS TO E XERCISES

2.4.7.17. *. Solution. This is a separable differential equation that we solve in

the usual way. Cross-multiplying and integrating,

y dy = (15x2 + 4x + 3) dx
Z Z
y dy = (15x2 + 4x + 3) dx
y2
= 5x3 + 2x2 + 3x + C.
2
42
Plugging in x = 1 and y = 4 gives 2
= 5 + 2 + 3 + C, and so C = −2. Therefore

y2
= 5x3 + 2x2 + 3x − 2
2
y 2 = 10x3 + 4x2 + 6x − 4

This leaves us with two possible functions for y:

√ √
y = 10x3 + 4x2 + 6x − 4 or y = − 10x3 + 4x2 + 6x − 4

When x = 1, y = 4. This only fits the first equation, so

√
y = 10x3 + 4x2 + 6x − 4
2.4.7.18. *. Solution. The given differential equation is separable and we solve
it accordingly.

dy
= x3 y
dx
dy
= x3 dx
y
dy
Z Z
= x3 dx
y
x4
log |y| = +C
4
4 4
|y| = ex /4+C = ex /4 eC

We are told that y = 1 when x = 0. That is, 1 = e0 eC , so eC = 1. That is, C = 0.

4 /4
|y| = ex

This leaves us with two potential functions:

4 /4 4 /4
y = ex or y = −ex

The first is always positive, and the second is always negative. Since y = 1 (a
positive number) when x = 0, we see
4 /4
y = ex

1030
S OLUTIONS TO E XERCISES

2.4.7.19. *. Solution. This is a separable differential equation, even if it doesn’t

quite look like it. First move the y from the left hand side to the right hand side.

dy
x + y = y2
dx
dy
x = y 2 − y = y(y − 1)
dx
dy dx
=
y(y − 1) x
1 1
Using the method of partial fractions, we see y(y−1)
= y−1
− y1 .
 
1 1 dx
− dy =
y−1 y x
Z  
1 1 dx
Z
− dy =
y−1 y x
log |y − 1| − log |y| = log |x| + C
|y − 1|
log = log |x| + C (∗)
|y|

To determine C we set x = 1 and y = −1.

| − 2|
log = log |1| + C
| − 1|
log 2 = C

Returning to (∗),

|y − 1|
log = log |x| + log 2
|y|
y−1
log = log |2x|
y
y−1
= |2x|
y

As y(1) = −1 is an initial condition, we have that x ≥ 1 and |2x| = 2x. For x = 1,

we have y = −1. So at least for x near 1, we have y near −1, so that y−1 y
is positive
and we may drop the absolute value signs. There remains the possibility that
y(x)−1
y(x)
changes sign for some larger x > 1. For now, we will simply ignore that
possibility. At the end, we will explicitly check that the y(x) we come up with
dy
really does satisfy the differential equation x dx + y = y 2 and the initial condition
y(1) = −1.
y−1
= 2x
y
y − 1 = 2xy

1031
S OLUTIONS TO E XERCISES

y − 2xy = 1
y(1 − 2x) = 1
1
y=
1 − 2x
As a check, we compute:
 
dy d 1
x +y =x +y
dx dx 1 − 2x
2 1
=x 2
+
(1 − 2x) 1 − 2x
2x + (1 − 2x)
=
(1 − 2x)2
1
=
(1 − 2x)2
= y2

So, our differential equation is satisfied. Furthermore:

1
y(1) = = −1
1−2×1
as desired. This confirms that our solution is correct.
2.4.7.20. *. Solution. The unknown function f (x) satisfies an equation that
involves the derivative of f . That means we’re in differential equation territory.
Specifically, we are told that y = f (x) obeys the separable differential equation
dy
dx
= xy.

dy
= xy
dx
dy
= x dx
y
dy
Z Z
= x dx
y
x2
log |y| = +C
2
To determine C we set x = 0 and y = e.

02
log e = +C
2
1=C

So, the solution is

x2
log |y| = +1
2

1032
S OLUTIONS TO E XERCISES

We are told that y = f (x) > 0, so may drop the absolute value signs.

x2
log y = +1
2
1 2 2
y = e1+ 2 x = e · ex /2
2.4.7.21. *. Solution. This is a separable differential equation.
dy 1
= 2
dx (x + x)y
dx
y dy =
x(x + 1)
1
Using partial fractions decomposition, we find x(x+1) = x1 − x+1
1
.
 
1 1
y dy = − dx
x x+1
Z  
1 1
Z
y dy = − dx
x x+1
y2 x
= log |x| − log |x + 1| + C = log +C
2 x+1
To satisfy the initial condition y(1) = 2 we must choose C to obey
22 1
= log +C
2 1+1
1
2 = log + C
2
1
C = 2 − log
2
So,
y2 x 1
= log + 2 − log
2 x+1 2
x 1
y 2 = 2 log + 4 − 2 log
x+1 2
Note that the question specifies that y(1) = 2 is an initial condition. So we always
x
have x ≥ 1. Then x+1 is positive, and we can drop the absolute values.
x 1
y 2 = 2 log + 4 − 2 log
x+1 2
This leaves two options for y(x): the positive or negative square root of the right
hand side above. Since y(1) = 1, which is positive, we must choose the positive
square root.
r 
x 1 
y(x) = 2 log − log + 2
x+1 2

1033
S OLUTIONS TO E XERCISES
r
2x
= 4 + 2 log
x+1
You might worry that y(x) could pass through zero, changing sign, at some x > 1.
dy 1
But the differential equation says that dx = (x2 +x)y is positive whenever y > 0 and
x ≥ 1. So y(x) is an increasing function whenever y > 0 and x ≥ 1. As y(1) = 2,
we have y(x) ≥ 2 for all x ≥ 1.
2.4.7.22. *. Solution. This is a separable differential equation.
p
1 + y 2 − 4 dy sec x
=
tan x dx y
 p 
2
y 1 + y − 4 dy = sec x tan xdx
Z p Z
 
2
y 1 + y − 4 dy = sec x tan xdx

For the integral on the left, we use the substitution u = y 2 − 4, 12 du = y dy.

1 √

Z
1 + u du = sec x + C
2
 
1 2 3/2
u+ u = sec x + C
2 3
 
1 2 2 2 3/2
y − 4 + (y − 4) = sec x + C
2 3
2
y 2 + (y 2 − 4)3/2 = 2 sec x + 2C + 4
3
To find C we set x = 0 and y = 2.
2√ 3
4+ 4 − 4 = 2 sec(0) + 2C + 4
3
4 = 2 + 2C + 4
2 = 2C + 4

So,
2
y 2 + (y 2 − 4)3/2 = 2 sec x + 2
3
2.4.7.23. *. Solution. The given differential equation is separable and we solve
it accordingly.
dP √
= −k P
dt
dP
√ = −k dt
P
dP
Z Z
√ = −k dt
P

1034
S OLUTIONS TO E XERCISES
√
2 P = −kt + C

At t = 0, P = 90, 000 so
p
2 90, 000 = −k × 0 + C
C = 2 × 300 = 600

Therefore,
√
2 P = −kt + 600 (∗)

Now, we find k. Let t be measured in weeks. Then when t = 6, P = 40, 000.

p
2 40, 000 = −6k + 600
2 · 200 = −6k + 600
200 100
k= =
6 3
Substituting our value of k into (∗):
√ 100
2 P =− t + 600
3
To find when the population will be 10,000, we set P = 10, 000 and solve for t.
p 100
2 10, 000 = − t + 600
3
100
2 · 100 = − t + 600
3
100
t = 400
3
t = 12

Since we measured t in weeks when we found k, we see that in 12 weeks the

population will decrease to 10,000 individuals.
2.4.7.24. *. Solution. The given differential equation is separable and we solve
it accordingly.

dv
m = −(mg + kv 2 )
dt
m
dv = −dt
mg + kv 2
m
Z Z
dv = −dt
mg + kv 2

1035
S OLUTIONS TO E XERCISES

The left integral looks something like the antiderivative of arctangent. Let’s fac-
tor out that mg from the denominator.
1 m
Z
k 2
dv = −t + C
mg 1 + mg v
1 1
Z
2 dv = −t + C
g
q
k
1+ mg
v

Now it looks even more like the derivative of arctangent.

q We can
q guess the an-
k k
tiderivative from here, or use the substitution u = mg v, du = mg dv.

r s !
1 mg k
arctan v = −t + C
g k mg
s !
m k
r
arctan v = −t + C (∗)
gk mg

At t = 0, v = v0 , so:
s !
m k
r
arctan v0 =C
gk mg

Plug C into (∗).

s ! s !
m k m k
r r
arctan v = arctan v0 −t
gk mg gk mg

At its highest point, the object has velocity v = 0. This happens when t obeys:
s ! r s !
m k m k
r
arctan 0 = arctan v0 − t
gk mg gk mg
s !
m k
r
0= arctan v0 − t
gk mg
s !
m k
r
t= arctan v0
gk mg

2.4.7.25. *. Solution. (a) The given differential equation is separable and we

solve it accordingly.

dv
= −k v 2
dt
dv
− 2 = k dt
v

1036
S OLUTIONS TO E XERCISES

dv
Z Z
− 2 = k dt
v
1
= kt + C
v
At t = 0, v = 40 so
1
=k×0+C
40
1
C=
40
Therefore,
1 1 40
v(t) = = = (∗)
kt + C kt + 1/40 40kt + 1

The constant of proportionality k is determined by

v(10) = 20
40
20 =
40k × 10 + 1
1 1
=
2 400k + 1
400k + 1 = 2
1
k=
400
(b) Subbing in the value of k to (∗),

40 40
v(t) = =
40kt + 1 t/10 + 1

We want to know the value of t that gives v(t) = 5.

40
5=
t/10 + 1
t
+1=8
10
t = 70 sec
2.4.7.26. *. Solution. (a) The given differential equation is separable and we
solve it accordingly.

dx
= k(3 − x)(2 − x)
dt
dx
= kdt
(x − 2)(x − 3)

1037
S OLUTIONS TO E XERCISES

1 1 1
Using the method of partial fractions, we find (x−2)(x−3)
= x−3
− x−2
.

1 1 i
Z h Z
− dx = kdt
x−3 x−2
log |x − 3| − log |x − 2| = kt + C
x−3
log = kt + C
x−2
x−3
= ekt+C = ekt eC
x−2
x−3
= Dekt
x−2
where D = ±eC . When t = 0, x = 1, forcing

1−3
= De0
1−2
D=2

Hence
x−3
= 2ekt
x−2
x − 3 = 2ekt (x − 2)
x − 2ekt x = 3 − 4ekt
3 − 4ekt
x(t) =
1 − 2ekt
(b) To evaluate the limit, we could use l’Hôpital’s rule, but we could also just
multiply the numerator and denominator by e−kt . Note lim e−tk = 0.
t→∞

3 − 4ekt 3 − 4ekt e−kt

lim x(t) = lim = lim ·
t→∞ t→∞ 1 − 2ekt t→∞ 1 − 2ekt e−kt
| {z }
num→−∞
den→−∞

3e−kt − 4 0−4
= lim −kt
= =2
t→∞ e −2 0−2
2.4.7.27. *. Solution. (a) The given differential equation is separable and we
solve it accordingly.

dP
= 4P − P 2
dt
dP
= dt
4P − P 2
dP
= dt
P (4 − P )

1038
S OLUTIONS TO E XERCISES

1 1/4 1/4
Using the method of partial fractions, we see P (4−P )
= P
+ 4−P
.

1h 1 1 i
+ dP = dt
Z 4hP 4−P
1 1 1 i
Z
+ dP = dt
4 P 4−P
1 i
log |P | − log |4 − P | = t + C
4
When t = 0, P = 2, so 14 log |2| − log |2| = C =⇒ C = 0. So,
 

1 P
log =t
4 4−P
P
At time t = 0, 4−P = 1 > 0. The ratio may not change sign at any finite time,
because this could only happen if at some finite time P took either the value 0 or
the value 4. But at this time t = 14 log 4−P
P P
would have to be infinite. So 4−P >0
for all time and:
1 P
log =t
4 4−P
P
log = 4t
4−P
P
= e4t
4−P
P = (4 − P )e4t
P + P e4t= 4e4t
4e4t 4
P = =
1 + e4t 1 + e−4t
(b) At t = 12 , P = 4
1+e−2
≈ 3.523.
4 4
lim P (t) = lim −4t
= =4
t→∞ t→∞ 1 + e 1+0
2.4.7.28. *. Solution.

a The rate of change of speed at time t is −kv(t)2 for some constant of propor-
tionality k (to be determined–but we assume it is positive, since the speed
is decreasing). So v(t) obeys the differential equation dv
dt
= −kv 2 .

b The equation dvdt

= −kv 2 is a separable differential equation, which we can
solve in the usual way.

dv
= −kv 2
dt
dv
= kdt
−v 2

1039
S OLUTIONS TO E XERCISES

dv
Z Z
− 2 = kdt
v
1
= kt + C
v
1
At time t = 0, v = 400, so C = 400
. Then:

1 1
= kt + (∗)
v 400
At time t = 1, v = 200, so
1 1
=k+
200 400
1
k=
400
Therefore, from (∗),
1 t 1 t+1
= + =
v 400 400 400
400
v=
t+1

c To find when the speed is 50, we set v = 50 in the equation from (b) and
solve for t.
400
50 =
t+1
50(t + 1) = 400
t+1=8
t=7

Exercises — Stage 3
2.4.7.29. *. Solution. (a) The given differential equation is separable and we
solve it accordingly.
dB
= (0.06 + 0.02 sin t)B
dt
dB
= (0.06 + 0.02 sin t) dt
Z B
dB
Z
= (0.06 + 0.02 sin t) dt
B
log |B(t)| = 0.06t − 0.02 cos t + C 0
Since B(t) is our bank account balance and we’re not withdrawing money, B(t)
is positive, so we can drop the absolute value signs.
log B(t) = 0.06t − 0.02 cos t + C 0

1040
S OLUTIONS TO E XERCISES
0
B(t) = e0.06t−0.02 cos t eC
B(t) = Ce0.06t−0.02 cos t
0
for arbitrary constants C 0 and C = eC ≥ 0.
Remark: the function B(t) = 0 obeys the differential equation so that C = 0 is
0
allowed, even though it is not of the form C = eC . This seeming discrepancy
arose because, in our very first step of part (a), we divided both sides of the
differential equation by B, which is only allowable if B 6= 0. So, in this step, we
implicitly assumed B was nonzero.
(b) We are told that B(0) = 1000. This allows us to find C.

1000 = B(0) = Ce0−0.02 cos 0 = Ce−0.02

C = 1000e0.02

So, when t = 2,

B(2) = 1000e0.02 e0.06×2−0.02 cos 2 = $1159.89

| {z }
C

rounded to the nearest cent.

Note that cos 2 is the cosine of 2 radians, cos 2 ≈ −0.416.
2.4.7.30. *. Solution. (a) The given differential equation is separable and we
could solve it accordingly. In fact we have already done so. If we rewrite the
equation in the form

dB  m
=a B−
dt a
it is of the form covered by Theorem 2.4.4. So that theorem tells us that the solu-
tion is
 m  at m
B(t) = B(0) − e +
a a
1
In this problem we are told that a = 0.02 = 50
, so

B(t) = {B(0) − 50m} et/50 + 50m = {30000 − 50m} et/50 + 50m

(b) The solution of part (a) is independent of time if and only if 30000 − 50m = 0.
So we need
30000
m= = $600
50
2.4.7.31. *. Solution. What we’re given is an equation relating y to the integral
of a function of y. What we know how to solve is an equation relating the deriva-
tive of y to a function of y. We can create this by differentiating the given integral

1041
S OLUTIONS TO E XERCISES

equation. By the Fundamental Theorem of Calculus, part 1:

Z x 
0 d 2


y (x) = y(t) − 3y(t) + 2 sin tdt
dx 0
= y(x)2 − 3y(x) + 2 sin x

So y(x) satisfies the differential equation y 0 = y 2 −3y+2 sin x = (y−2)(y−1) sin x

and the initial equation y(0) = 3 (just substitute x = 0 into (∗)). For y 6= 1, 2:

dy
= (y − 2)(y − 1) sin x
dx
dy
= sin xdx
(y − 2)(y − 1)
dy
Z Z
= sin xdx
(y − 2)(y − 1)
1 1 1
Using the method of partial fractions, we see (y−2)(y−1)
= y−2
− y−1
.

1 1 i
Z h Z
− dy = sin xdx
y−2 y−1
log |y − 2| − log |y − 1| = − cos x + c
y−2
log = − cos x + c
y−1
y−2
= ec−cos x
y−1
3−2
The condition y(0) = 3 forces 3−1
= ec−1 or ec = 12 e, hence

y−2 1
= e1−cos x
y−1 2

Observe that, when x = 0, y−2

y−1
= 21 > 0. Furthermore 21 e1−cos x , and hence y−2
y−1
,
can never take the value zero. As y(x) varies continuously with x, y(x) must
remain larger than 2. Consquently, y−2
y−1
remains positive and we may drop the
absolute value signs. Hence

y−2 1
= e1−cos x
y−1 2

Solving for y,

y−2 1
= e1−cos x
y−1 2
2(y − 2) = e1−cos x (y − 1)
2y − 4 = ye1−cos x − e1−cos x
y 2 − e1−cos x = 4 − e1−cos x

1042
S OLUTIONS TO E XERCISES

4 − e1−cos x
y=
2 − e1−cos x
To avoid division by zero in the last step, we need

e1−cos x 6= 2
1 − cos x 6= log 2
cos x 6= 1 − log 2

Let L = 1 − log 2, for brevity, and note that L > 0. (This can be seen by observing
2 < e, so, log 2 < log e = 1, hence 1 − log 2 > 0.)

L
x
− arccos(L) arccos(L)

Y = cos x

We know x = 0 is in the domain of our function, but the points x = ± arccos(L) =

± arccos(1 − log 2) are not.

x
− arccos(1 − log 2) 0 arccos(1 − log 2)

not in domain in domain not in domain

Therefore, the largest interval for which our answer makes sense is

− arccos(1 − log 2)) > x > arccos(1 − log 2)

or approximately −1.259 < x < 1.259.

2.4.7.32. *. Solution. Suppose that in a very short time interval dt, the height
of water in the tank changes by dh (which is negative). Then in this time interval
the amount of the water in the tank decreases by dV = −π(3)2 dh. This must be
the same as the amount of water that flows through the hole in this time interval.
The water flowing through the hole makes a cylinder of radius 1 cm (that is,
0.01 m) with length v(t)dt, the distance the water moves out of the hole in dt
seconds. So, the amountpof water leaving the hole over the time interval dt is
π(0.01)2 v(t) dt = π(0.01)2 2gh(t) dt.

1043
S OLUTIONS TO E XERCISES

dh

dV

This gives us a separable differential equation. Recall g is a constant.

p
−π(3)2 dh = π(0.01)2 2gh(t) dt
dh  0.01 2 p
√ =− 2g dt
h 3
dh 0.01 2 p
Z Z 
√ = − 2g dt
h 3
√  0.01 2 p
2 h=− 2g t + C
3
√
At time 0, the height is 6, so C = 2 6 and
√  0.01 2 p √
2 h=− 2g t + 2 6
3
We want to know when the height of the water in the tank is 0.
 0.01 2 p √
0=− 2g t + 2 6
3
 0.01 2 p √
2g t = 2 6
3 √
2 6
t =  2 √
0.01
3
2g
 3 2 r 3
=2
0.01 g
r
3
= 180, 000 ≈ 99, 591 sec ≈ 27.66 hr
g
2.4.7.33. *. Solution. Suppose that at time t, the mercury in the tank has height
h, which is between 0 and 12 feet.

1044
S OLUTIONS TO E XERCISES

At
p that time, the top surface of the mercury forms a circular disk of radius
62 − (h − 6)2 . (We found this by applying the Pythagorean Theorem to the
triangle in the diagram above. In the diagram, h is shown as being larger than
6, but the same equation holds for all h in [0, 12].) Now suppose that in a very
short time interval dt, the height of mercury in the tank changes by dh (which is
negative). Then in this time interval the amount of the mercury in the tank de-
p
2
creases by −π 62 − (h − 6)2 dh. (That’s the volume of the red disk in the fig-
ure above.) This must be the same as the amount of mercury that flows through
the hole in this time interval. The mercury comes out of the hole as a cylin-
1
der. Its radius is the radius of the hole, 12 foot, and its length is the distance the
mercury travels in dt seconds, v(t)dt feet. So, the volume of escaped mercury is
1 2


√
1 2
π 12 v dt = π 12 2gh dt. This gives us a separable differential equation.
p
2  1 2 p
2 2
−π 6 − (h − 6) dh = π 2gh dt
12
 1 2p

− 36 − (h2 − 12h + 36) dh =


2gh dt
12
1 p √
h2 − 12h dh =


2g h dt
144
1 p
h3/2 − 12h1/2 dh =


2g dt
144
1 p
Z Z
3/2 1/2


h − 12h dh = 2g dt
144
h5/2 h3/2 1 p
− 12 = 2g t + C
5/2 3/2 144
5/2 3/2
At time 0, the height is 12, so C = 125/2 − 12 123/2 = 125/2 25 − 23 = − 15 4
125/2 , which

yields

h5/2 h3/2 1 p 4
− 12 = 2g t − 125/2
5/2 3/2 144 15
We want to find the time t when the height is h = 0.
1 p 4
0= 2g t − 125/2
144 15
1 p 4 5/2
2g t = 12
144 15

1045
S OLUTIONS TO E XERCISES
s
4 × 144 125
t=
15 2g
r
124416
= 38.4 ≈ 2, 394 sec ≈ 0.665 hr
g
2.4.7.34. *. Solution. (a) Setting x = 0 gives
Z 0



f (0) = 3 + f (t) − 1 f (t) − 2 dt = 3
0

(b) By the Fundamental Theorem of Calculus part 1,

Z x
0 d




f (x) = f (t) − 1 f (t) − 2 dt = f (x) − 1 f (x) − 2
dx 0
Thus y = f (x) obeys the differential equation y 0 = (y − 1)(y − 2).
(c) If y 6= 1, 2,
dy
= (y − 1)(y − 2)
dx
dy
= dx
(y − 1)(y − 2)
dy
Z Z
= dx
(y − 1)(y − 2)
Using the method of partial fractions,
Z  
1 1
Z
− dy = dx
y−2 y−1
log |y − 2| − log |y − 1| = x + C
y−2
log =x+C
y−1
dy
Observe that dx = (y − 1)(y − 2) > 0 for all y ≥ 2. That is, f (x) is increasing at all
x for which f (x) > 2. As f (0) = 3, f (x) increases for all x ≥ 0, and f (x) ≥ 3 for
all x ≥ 0. So we may drop the absolute value signs.
f (x) − 2
log =x+C
f (x) − 1
f (x) − 2
= eC ex
f (x) − 1
f (x)−2 1
At x = 0, f (x)−1
= 2
so eC = 12 .
f (x) − 2 1
= ex
f (x) − 1 2
2f (x) − 4 = [f (x) − 1]ex
[2 − ex ]f (x) = 4 − ex
4 − ex
f (x) =
2 − ex

1046
S OLUTIONS TO E XERCISES

2.4.7.35. *. Solution. Suppose that at time t (measured in hours starting at, say,
noon), the water in the tank has height y, which is between 0 and 2 metres. At
that time, the top surface of the water forms a circular disk of radius r = y p and
area A(y) = πy 2p . Thus, by Torricelli’s law,

dy √
πy 2p = −c y
dt
π 2p− 1
− · y 2 dy = dt
c
π 2p− 1
Z Z
− ·y 2 dy = dt
c
1
π y 2p+ 2
− · +d=t
c 2p + 12
1
π 22p+ 2
for some constant d. At time t = 0, the height is y = 2, so d = · .
c 2p + 12
1 1 
π 22p+ 2 y 2p+ 2

t= −
c 2p + 21 2p + 12
π 
2p+ 12 2p+ 12

= 2 − y
c(2p + 12 )

The time at which the height is 1 is obtained by subbing y = 1 into this formula.
The time at which the height is 0 is obtained by subbing y = 0 into this formula.
Thus the condition that the top half (y = 2 to y = 1) takes exactly the same
amount of time to drain as the bottom half (y = 1 to y = 0) is:

t(2) − t(1) = t(1) − t(0)

0 − t(1) = t(1) − t(0)
t(0) = 2t(1)
π 
2p+ 21 2p+ 12
 π 
2p+ 12 2p+ 12

2 −0 =2 2 −1
c(2p + 12 ) c(2p + 12 )
1
 1

22p+ 2 = 2 22p+ 2 − 1
1 1
22p+ 2 = 2 · 22p+ 2 − 2
1
2 = 22p+ 2
1
1 = 2p +
2
1
p=
4
2.4.7.36. Solution.

a If we let f (t) = 0 for all t, then its average over any interval is 0, as is its
root mean square.

1047
S OLUTIONS TO E XERCISES

b Let’s start by simplifying the given equation.

Z x s Z x
1 1
f (t) dt = f 2 (t) dt
x−a a x−a a
Z x sZ
x
1
√ f (t) dt = f 2 (t) dt (E1)
x−a a a
Z x (sZ )
  x
d 1 d
√ f (t) dt = f 2 (t) dt (E2)
dx x−a a dx a

For the derivative on the left, we use the product rule and the Fundamental
Theorem of Calculus, part 1.
 Z x 
d 1
√ f (t) dt
dx x−a a
 Z x Z x 
d 1 1 d
= √ f (t) dt + √ · f (t) dt
dx x−a a x − a dx a
Z x
1 f (x)
=− √ 3 f (t) dt + √
2 x−a a x−a
 Z x 
1 1
=√ f (x) − f (t) dt
x−a 2(x − a) a

For the derivative on the right in Equation (E2) we use the chain rule and
the Fundamental Theorem of Calculus part 1
(sZ ) − 12
x Z x Z x 
d 2
1 2 d 2
f (t) dt = f (t) dt · f (t) dt
dx a 2 a dx a

f 2 (x)
= qR
x
2 a f 2 (t) dt

So, Equation (E2) yields the following:

Z x
f 2 (x)
 
1 1
√ f (x) − f (t) dt = qR (E3)
x−a 2(x!a) a x
2 a f 2 (t) dt

qR
x Rx
c From Equation (E1), f 2 (t) dt = √1 f (t) dt.
a x−a a

Z x
f 2 (x)
 
1 1
√ f (x) − f (t) dt = 1
Rx
x−a 2(x − a) a 2 √x−a a
f (t) dt

and x  x 
2 1
Z Z
f (t) dt f (x) − f (t) dt = f 2 (x)
x−a a 2(x − a) a

1048
S OLUTIONS TO E XERCISES

d Now what we have R x is a differential equation, although it might not look

dY
like it. Let Y (x) = a f (t) dt. Then dx (x) = f (x).
   2
2 dY 1 dY
Y − Y = (E4)
x−a dx 2(x − a) dx

We’re used to solving differential equations of the form dYdx

=(something).
So, let’s manipulate our equation until it has this form.
 2     2
dY 2Y dY Y
− + =0
dx x−a dx x−a

This is a quadratic equation, with variable dY

dx
. Its solutions are:

2Y

q 2Y
2 Y

2
dY x−a
± x−a
− 4 · x−a
=
dx 2
2Y
± 0
= x−a
2
Y
=
x−a
This gives us the separable differential equation
dY Y
=
dx x−a
dY dx
= (E5)
Z Y Zx − a
dY dx
=
Y x−a
log |Y | = log |x − a| + C
|Y | = elog |x−a|+C = |x − a|eC
Y = D(x − a)

where D is some constant, eC or −eC . Note this covers all real constants
except D = 0. If D = 0, then Y (x) = 0 for all x. This function also satisfies
Equation (E4), so indeed,

Y (x) = D(x − a) (E6)

for any constant D is the family of equations satisfying our differential

equation.
Remark: the reason we “lost” the solution Y (x) = 0 is that in Equation (E5),
we divided by Y , thus tacitly assuming it was not identically 0.
Rx
e Remember Y = a f (t) dt. So, Equation (E6) tells us:
Z x
f (t) dt = D(x − a)
a

1049
S OLUTIONS TO E XERCISES
Z x 
d d
f (t) dt = {D(x − a)}
dx a dx
f (x) = D

We should check that this function works.

Z x
1 1 h it=x Dx − Da
favg = D dt = Dt = =D
x−a a x−a t=a x−a
s Z x r
1 2
1 h 2 it=x
fRMS = D dt = D t
x−a a x−a t=a

D2 x − D2 a √ 2
r
= = D = |D|
x−a
So, f (x) = D works only if D is nonnegative.
That is: the only functions whose average matches their root mean square
over every interval are constant, nonnegative functions.
Remark: it was step (c) where we introduced the erroneous answer f (x) =
D, D < 0 to our solution. In Equation (E3), f (x) = D is not a solution if
D < 0:
Z x
f 2 (x)
 
1 1
√ f (x) − f (t) dt = qR
x−a 2(x − a) a x
2 a f 2 (t) dt
Z x
D2
 
1 1
√ D− D dt = qR
x−a 2(x − a) a x
2 a D2 dt
D2
 
1 1
√ D− D(x − a) = p
x−a 2(x − a) 2 D2 (x − a)
D2
 
1 1
√ D = √
x−a 2 2|D| x − a
D2
D= = |D|
|D|
qR
x
In (c), we replace a
f 2 (t) dt, which cannot be negative, with
x
√1
R
x−a a
f (t) dt, which could be negative if f (t) = D < 0. Indeed, if
qR
x 2 √ Rx √
f (t) = D, then f (t) dt = |D| x − a, while √1 f (t) dt = D x − a.
a x−a a
It is at this point that negative functions creep into our solution.
2.4.7.37. Solution. We start by antidifferentiating both sides with respect to x.
Z  2  Z  
dy 2 dy
dx = · dx
dx2 y 3 dx

1050
S OLUTIONS TO E XERCISES

The right integral is in exactly the form we would use for a change of variables
(substitution) to y.
Z  
dy 2 1
= 3
dy = − 2 + C
dx y y
dy
When y = 1, dx
= 3.

1
3=− +C
1
C=4

So,
dy 1
=− 2 +4
dx y
This is a separable differential equation.

dy 4y 2 − 1
=
dx y2
y2
dy = dx
4y 2 − 1
y2
Z Z
dy = dx (∗)
4y 2 − 1
We can evaluate the left integral with partial fractions, but because the numerator
has the same degree as the denominator, we have to simplify first. We do this by
inspection, but you can also use long division.
1
y2 4
(4y 2 − 1) + 14
=
4y 2 − 1 4y 2 − 1
 
1 1
= 1+ 2
4 4y − 1
 
1 1
= 1+
4 (2y − 1)(2y + 1)
 
1 1/2 1/2
= 1+ −
4 2y − 1 2y + 1

Now, we return to (∗).

y2
Z Z
dx = dy
4y 2 − 1
Z  
1 1/2 1/2
= 1+ − dy
4 2y − 1 2y + 1
 
1 1 1
= y + log |2y − 1| − log |2y + 1|
4 4 4

1051
S OLUTIONS TO E XERCISES
 
1 1 2y − 1
= y+ log
4 4 2y + 1
 
1 1 2y − 1
x+C = y+ log
4 4 2y + 1
1
When x = − 16 log 3, y = 1.
 
1 1 1 2−1 1 1 1
− log 3 + C = 1 + log = + log
16 4 4 2+1 4 16 3
1
C=
4
So,
 
1 1 1 2y − 1
x+ = y + log
4 4 4 2y + 1
 
1 1 2y − 1
x= y − 1 + log
4 4 2y + 1

We can check our answer by differentiating with respect to x.

 
1 1 2y − 1
x= y − 1 + log
4 4 2y + 1
1 1
4x = y − 1 + log |2y − 1| − log |2y + 1|
 4 4 
d d 1 1
{4x} = y − 1 + log |2y − 1| − log |2y + 1|
dx dx 4 4
dy dy
dy 1 2 dx 1 2 dx
4= + · − ·
dx 4 2y − 1 4 2y + 1
4y 2
   
dy 1/2 1/2 dy
4= 1+ − =
dx 2y − 1 2y + 1 dx 4y 2 − 1
dy 4y 2 − 1 1
= 2
=4− 2 (∗∗)
dx y y

Differentiating with respect to x again, using the chain rule,

d2 y 2 dy
= ·
dx2 y 3 dx

This is exactly the differential equation we were meant to solve.

3 · Sequence and series

3.1 · Sequences
3.1.2 · Exercises

1052
S OLUTIONS TO E XERCISES

Exercises — Stage 1
3.1.2.1. Solution. (a) The values of the sequence seem to be getting closer and
closer to -2, so we guess the limit of this sequence is -2.
(b) Overall, the values of the sequence seem to be getting extremely close to 0,
so we approximate the limit of this sequence as 0. It doesn’t matter that the se-
quence changes signs, or that the numbers are sometimes farther from 0, some-
times closer.
(c) This limit does not exist. The sequence is sometimes 0, sometimes -2, and not
consistently staying extremely near to either one.
3.1.2.2. Solution. True. We consider the end behaviour of the sequences, which
does not depend on any finite number of terms at their beginning.
A−B
3.1.2.3. Solution. (a) We follow the arithmetic of limits, Theorem 3.1.8:
C
cn
(b) Since lim cn is some real number, and n grows without bound, lim = 0.
n→∞ n→∞ n
a2n+5 A
(c) We note lim a2n+5 = lim an , so = .
n→∞ n→∞ bn B
3.1.2.4. Solution. There are many possible answers. One is:
(
3000 − n if n ≤ 1000
an =
−2 + n1 if n > 1000

where we have a series that looks different before and after its thousandth term.
Note every term is smaller than the term preceding it.
Another sequence with the desired properties is:

1, 002, 001
an = −2
n
When n ≤ 1000, an ≥ 1,002,001
1000
− 2 > 1,002,000
1,000
− 2 = 1000. That is, an > 1000
when n ≤ 1000. As n gets larger, an gets smaller, so an+1 < an for all n. Finally,
lim an = 0 − 2 = −2.
n→∞

3.1.2.5. Solution. One possible answer is an = (−1)n =

{−1, 1, −1, 1, −1, 1, −1, . . .}.
Another is an = n(−1)n = {−1, 2, −3, 4, −5, 6, −7, . . .}.
3.1.2.6. Solution. If the terms of a sequence are alternating sign, but the limit
of the sequence exists, the limit must be zero. (If it were a positive number, the
negative terms would not get very close to it; if it were a negative number, the
positive terms would not get very close to it.)
This gives us the idea to modify an answer from Question 5. One possible se-
quence:
(−1)n
 
1 1 1 1 1
an = = −1, , − , , − , , . . .
n 2 3 4 5 6

1053
S OLUTIONS TO E XERCISES

3.1.2.7. Solution.

a Since −1 ≤ sin n ≤ 1 for all n, one potential set of upper and lower bound
is
−1 sin n 1
≤ ≤
n n n
−1 1
Note lim = lim , so these are valid comparison sequences for the
n→∞ n n→∞ n
squeeze theorem.

b Since −1 ≤ sin n ≤ 1 and −5 ≤ −5 cos n ≤ 5 for all n, we see

7−1−5 ≤ 7 + sin n − 5 cos n ≤ 7 + 1 + 5

1 ≤ 7 + sin n − 5 cos n ≤ 13

This gives us the idea to try the bounds

n2 n2 n2
≤ ≤
13en en (7 + sin n − 5 cos n) en

n2 n2
We check that lim = lim (they’re both 0 — you can verify using
n→∞ 13en n→∞ en
l’Hôpital’s rule), so these are indeed reasonable bounds to choose to use
with the squeeze theorem.
1 (−1)n
c Since (−n)−n = = , we see
(−n)n nn
−1 1
n
≤ (−n)−n ≤ n
n n
−1 1
Since both lim and lim are 0, these are reasonable bounds to use
n→∞ nn n→∞ nn
with the squeeze theorem.
3.1.2.8. Solution.
a • Note an = bn , since (in the absence of evidence to the contrary) we
1 n+1
assume n begins at one, hence n = |n|. Then an = bn = 1 + = .
n n
So, whenever n is a whole number, an and bn are the same as h(n) and
i(n). (Be careful here: h(x) 6= i(x) when x is not a whole number.)
1
• cn = e−n = n = j(n)
e
• For any integer n, cos(πn) = (−1)n . So, dn = f (n).
• Similarly, en = g(n).

b According to Theorem 3.1.6, if any of the functions on the right have limits
that exist as x → ∞, then these limits match the limits of their correspond-
ing sequences. So, we only have to be suspicious of f (x) and i(x), since

1054
S OLUTIONS TO E XERCISES

these do not converge.

The limit lim f (x) does not exist, and f (n) = dn ; the limit lim dn also does
x→∞ n→∞
not exist. (We generally don’t write equality for two things that don’t exist:
equality refers to numerical value, and these have none. a )
The limit lim i(x) does not exist, because i(x) = 0 when x is not a whole
x→∞
number, while i(x) approaches 1 when x is a whole number. However,
lim lim an = lim bn = 1.
n→∞ n→∞
So, using our answers from part (a), we match the following:

• lim an = lim bn = lim h(x) = 1

n→∞ n→∞ x→∞

• lim cn = lim en = lim g(x) = lim j(x) = 0

n→∞ n→∞ x→∞ x→∞

• lim dn , lim f (x) and lim i(x) do not exist.

n→∞ x→∞ x→∞

a The idea “two things that both don’t exist are equal”
√ is also rejected
√ because it can lead to
contradictions.
√ √ For example, in the real numbers −1 and −2 don’t exist; if we write
−1 = −2, then squaring both sides yields the inanity −1 = −2.

3.1.2.9. Solution. (a) We want to find odd multiples of π that are close to inte-
gers.

• Solution 1: One way to do that is to remember that π is somewhat close to

22
. Then when we multiply π by a multiple of 7, we should get something
7
close to an integer.
 In particular,
  7π, 21π, and
 35π
 should be reasonably
22 22 22
close to 7 = 22, 21 = 66, and 35 = 110, respectively. We
7 7 7
check whether they are close enough:

7π ≈ 21.99 21π ≈ 65.97 35π ≈ 109.96

So indeed, 22, 66, and 110 are all within 0.1 of some odd multiple of π.
Since the cosine of an odd multiple of π is −1, we expect all of the sequence
values to be close to −1. Using a calculator:

a22 = cos(22) ≈ −0.99996,

a66 = cos(66) ≈ −0.99965,
a110 = cos(110) ≈ −0.99902

• Solution 2: Alternately, we could have just listed odd multiple of π until we

1055
S OLUTIONS TO E XERCISES

found three that are close to integers.

2k + 1 (2k + 1)π
1 3.14
3 9.42
5 15.71
7 21.99
9 28.27
11 34.56
13 40.84
15 47.12
17 53.41
19 59.69
21 65.97
23 72.26
25 78.54
27 84.82
29 91.11
31 97.39
33 103.67
35 109.96

Some earlier odd multiples of π (like 15π and 29π) get fairly close to inte-
gers, but not within 0.1.
2k + 1
(b) If x = π for some integer k (that is, x is an odd multiple of π/2), then
2
cos x = 0. So, we can either list out the first few terms of an until we find three
22
that are very close to 0, or we can use our approximation π ≈ to choose values
7
2k + 1
of n that are close to π.
2

• Solution 1:
2k + 1 (2k + 1) × 22 2k + 1
π≈ = 11
2 2×7 7
So, we expect our values to be close to integers when 2k + 1 is a multiple of
7. For example, 2k + 1 = 7, 2k + 1 = 21, and 2k + 1 = 35.
We check:
x n an
π
7 × ≈ 10.99557 11 a11 ≈ 0.0044
2π
21 × ≈ 32.98672 33 a33 ≈ −0.0133
2
π
35 × ≈ 54.97787 55 a55 ≈ 0.0221
2
These seem like values of an that are all pretty close to 0.

1056
S OLUTIONS TO E XERCISES

• Solution 2: We could have listed the first several values of an , and looked
for some that are close to 0.

n an
1 0.54
2 −0.42
3 −0.99
4 −0.65
5 0.28
6 0.96
7 0.75
8 −0.15
9 −0.91
10 −0.84
2k+1
Oof. Nothing very close yet. Maybe a better way is to list values of 2
π,
and see which ones are close to integers.

2k + 1 2k+1
2
π
1 1.57
3 4.71
5 7.85
7 10.996
9 14.14
11 17.28
13 20.42
15 23.56
17 26.70
19 29.85
21 32.99
23 36.13
25 39.27
27 42.41
29 45.55
31 48.69
33 51.84
35 54.98

We find roughly the same candidates we did in Solution 1, depending on

what we’re ready to accept as “close”.

Remark: it is possible to turn the ideas of this question into a rigorous proof that
lim cos n is undefined.
n→∞

1057
S OLUTIONS TO E XERCISES

• Let, for each integer k ≥ 1, nk be the integer that is closest to 2kπ. Then
2kπ − 21 ≤ nk ≤ 2kπ + 12 so that cos(nk ) ≥ cos 21 ≥ 0.8. Consequently, if
lim cos n = c exists, we must have c ≥ 0.8.
n→∞

• Let, for each integer k ≥ 1, n0k be the integer that is closest to (2k + 1)π.
Then (2k + 1)π − 21 ≤ n0k ≤ (2k + 1)π + 12 so that cos(n0k ) ≤ − cos 12 ≤ −0.8.
Consequently, if lim cos n = c exists, we must have c ≤ −0.8.
n→∞

• It is impossible to have both c ≥ 0.8 and c ≤ −0.8, so lim cos n does not
n→∞
exist.

Exercises — Stage 2
3.1.2.10. Solution. When determining the end behaviour of rational functions,
recall from last semester that we can either cancel out the highest power of n
from the numerator and denominator, or skip this step and compare the highest
powers of the numerator and denominator.

a Since the numerator has a higher degree than the denominator, this se-
quence will diverge to positive or negative infinity; since its terms are pos-
itive for large n, its limit is (positive) infinity. (You can imagine that the
numerator is growing much, much faster than the denominator, leading
the terms to have a very, very large absolute value.)
Calculating the longer way:
1 5
3n2 − 2n + 5 n
3n − 2 + n
an = 1 =
4n + 3 n
4 + n3
5
3n − 2 + n 3n − 2 + 0
lim an = lim = lim =∞
n→∞ n→∞ 4 + n3 n→∞ 4+0

b Since the numerator has the same degree as the denominator, as n goes to
infinity, this sequence will converge to the ratio of their leading coefficients:
3
. (You can imagine that the numerator is growing at roughly the same rate
4
as the denominator, so the terms settle into an almost-constant ratio.)
Calculating the longer way:
1
3n2 − 2n + 5 3 − n2 + n52
 
n2
bn = 1 =
4n2 + 3 n2
4 + n32
3 − n2 + n52 3−0+0 3
lim bn = lim 3 = =
n→∞ n→∞ 4 + n2 4+0 4

c Since the numerator has a lower degree than the denominator, this se-
quence will converge to 0 as n goes to infinity. (You can imagine that the

1058
S OLUTIONS TO E XERCISES

denominator is growing much, much faster than the numerator, leading the
terms to be very, very small.)
Calculating the longer way:
1 3
3n2 − 2n + 5 − n22 + 5
 
n3 n n3
cn = 1 =
4n3 + 3 n3
4 + n33
3
n
− n22 + 5
n3 0−0+0
lim cn = lim = =0
n→∞ n→∞ 4 + n33 4+0
3.1.2.11. Solution. At first glance, we see both the numerator and denominator
grow huge as n increases, so we’ll need to think a little further to find the limit.
We don’t have a rational function, but we can still divide the top and bottom by
ne to get a clearer picture.
1
4n3 − 21 4n3−e − n21e
 
ne
an = e 1 1 = 1
n +n ne
1 + ne+1

Since e < 3, we see 3 − e is positive, so lim n3−e = ∞.

n→∞

4n3−e − n21e 4n3−e − 0

lim an = lim 1 = lim =∞
n→∞ n→∞ 1 + e+1
n
n→∞ 1 + 0
√
3.1.2.12. Solution. This isn’t a rational sequence, but factoring out n from the
top and bottom will still clear things up.
√ 1
! 1 √1
4
n + 1 √n 4n +
√
n
bn = √ 1 = q
9n + 3 √
n 9 + n3
1 √1
√
4n + n 0+0
lim bn = lim q =√ =0
n→∞ n→∞
9+ 3 9+0
n

3.1.2.13. Solution. First, let’s start with a tempting fallacy.

sin n
The denominator grows without bound, so lim = 0.
n→∞ n
It’s certainly true that if the limit of the numerator is a real number, and the denom-
inator grows without bound, then the limit of the sequence is zero. However, in
our case, the limit of the numerator does not exist. To apply the limit arithmetic
rules from Theorem 3.1.8, our limits must actually exist.
A better reasoning looks something like this:
The denominator grows without bound, and the numerator never
sin n
gets very large, so lim = 0.
n→∞ n
To quantify this reasoning more precisely, we use the squeeze theorem, Theo-
rem 3.1.10. There are two parts to the squeeze theorem: finding two bounding
functions, and making sure these functions have the same limit.

1059
S OLUTIONS TO E XERCISES

−1 1
• Since −1 ≤ sin n ≤ 1 for all n, we choose functions an = n
and bn = n
.
Then an ≤ cn ≤ bn for all n.

• Both lim an = 0 and lim bn = 0.

n→∞ n→∞

sin n
So, by the squeeze theorem, lim = 0.
n→∞ n

3.1.2.14. Solution. The denominator of this sequence grows without bound.

The numerator is unpredictable: imagine that n is large. When sin n is close to
−1, nsin n puts a power of n “in the denominator,” so we can have nsin n very close
to 0. When sin n is close to 1, nsin n is close to n, which is large.
To control for these variations, we’ll use the squeeze theorem.
n−1 1 n 1
• Since −1 ≤ sin n ≤ n for all n, let bn = n2
= n3
and cn = n2
= n
. Then
b n ≤ an ≤ c n .

• Both lim bn = 0 and lim cn = 0.

n→∞ n→∞

nsin n
So, by the squeeze theorem, lim = 0 as well.
n→∞ n2
Remark: we also could have used bn = 0 for our lower bound, since an ≥ 0 for
all n.
3.1.2.15. Solution.
1
dn = e−1/n =
e1/n
1 1 1
lim dn = lim = = =1
n→∞ n→∞ e1/n e 0 1
3.1.2.16. Solution.
• Solution 1: Let’s use the squeeze theorem. Since sin(n2 ) and sin n are both
between −1 and 1 for all n, we note:
1 + 3(−1) − 2(1) ≤ 1 + 3 sin(n2 ) − 2 sin n ≤ 1 + 3(1) − 2(−1)
−4 ≤ 1 + 3 sin(n2 ) − 2 sin n ≤ 6
This allows us to choose suitable bounding functions for the squeeze theo-
rem.
4 6
◦ Let bn = − and cn = . From the work above, we see bn ≤ an ≤ cn
n n
for all n.
◦ Both lim bn = 0 and lim cn = 0.
n→∞ n→∞

1 + 3 sin(n2 ) − 2 sin n
So, by the squeeze theorem, lim = 0.
n→∞ n
• Solution 2: We simplify slightly to begin.
1 + 3 sin(n2 ) − 2 sin n 1 sin(n2 ) sin n
an = = +3· −2·
n n n n

1060
S OLUTIONS TO E XERCISES

sin(n2 ) sin n
We apply the squeeze theorem to the pieces and .
n n
−1 1 sin(n2 ) sin n
◦ Let bn = and cn = . Then bn ≤ ≤ cn , and bn ≤ ≤ cn .
n n n n
◦ Both lim bn = 0 and lim cn = 0.
n→∞ n→∞

sin(n2 ) sin n
So, by the squeeze theorem, lim = 0 and lim = 0.
n→∞ n n→∞ n
Now, using the arithmetic of limits from Theorem 3.1.8,

sin(n2 )
 
1 sin n
lim an = lim +3· −2·
n→∞ n→∞ n n n
=0+3·0−2·0=0
3.1.2.17. Solution. First, we note that both numerator and denominator grow
without bound. So, we have to decide whether one outstrips the other, or
whether they reach a stable ratio.

• Solution 1: Let’s try dividing the numerator and denominator by 2n (the

dominant term in the denominator; this is the same idea behind factoring
out the leading term in rational expressions).
 1  e n


en n
bn = n 2
= 2 n2
2 + n2 21n 1 + 2n
e  e n
Since e > 2, we see > 1, and so lim = ∞. Since exponential
2 n→∞ 2
functions grow much, much faster than polynomial functions, we also see
2
lim 2nn = 0. So,
n→∞
e n e n



2 2
lim bn = lim 2 = lim =∞
n→∞ n→∞ 1 + 2nn n→∞ 1+0

• Solution 2: Since the numerator and denominator both increase without

d
bound, we apply l’Hôpital’s rule. Recall dx {2x } = 2x log 2.
en
lim bn = lim
n→∞ n→∞ 2n + n2
| {z }
num→∞
den→∞

en
= lim
n→∞ 2n log 2 + 2n
| {z }
num→∞
den→∞

en
= lim n
n→∞ 2 (log 2)2 + 2
| {z }
num→∞
den→∞

1061
S OLUTIONS TO E XERCISES

en
= lim
n→∞ 2n (log 2)3
1  e n
= lim
(log 2)3 n→∞ 2
=∞
e  e n
Since e > 2, we see > 1, and so lim = ∞.
2 n→∞ 2

3.1.2.18. *. Solution. First, we simplify. Remember n! = n(n−1)(n−2) · · · (2)(1)

for any whole number n, so (k + 1)! = (k + 1)k! .

k! sin3 k k! sin3 k sin3 k

ak = = =
(k + 1)! (k + 1)k! k+1

Now, we can use the squeeze theorem.

• −1 ≤ sin k ≤ 1 for all k, so −1 ≤ sin3 k ≤ 1. Let bk = −1

k+1
and ck = 1
k+1
. Then
b k ≤ ak ≤ c k .

• Both lim bk = 0 and lim ck = 0.

k→∞ k→∞

So, by the squeeze theorem, also lim ak = 0.

k→∞

3.1.2.19. *. Solution. Note lim (−1)n doesn’t exist, but −1 ≤ (−1)n ≤ 1 for all
n→∞
n. Let’s use the squeeze theorem.

• Let an = − sin n1 and bn = sin n1 . Then an ≤ (−1)n sin n1 ≤ bn .

• Both lim − sin n1 = 0 and lim sin n1 = 0, since lim n1 = 0 and sin 0 = 0.



n→∞ n→∞ n→∞

By the squeeze theorem, the sequence (−1)n sin n1 converges to 0.



6n2 + 5n
3.1.2.20. *. Solution. First, we note that lim = 6. We see this either by
n→∞ n2 + 1
comparing the leading terms in the numerator and denominator, or by factoring
out n2 from the top and the bottom.  
1 1
Second, since lim 2 = 0, we see lim cos = cos 0 = 1.
n→∞ n n→∞ n2
Using arithmetic of limits, Theorem 3.1.8, we conclude
 2 
6n + 5n 2
lim + 3 cos(1/n ) = 6 + 3(1) = 9.
n→∞ n2 + 1

Exercises — Stage 3
3.1.2.21. *. Solution. Let’s take stock: sin(1/n) → sin(0) = 0 as n → ∞, so
log (sin(1/n)) → −∞. However, log(2n) → ∞. So, we have some tension here:
the two pieces behave in ways that pull the terms of the sequence in different

1062
S OLUTIONS TO E XERCISES

directions. (Recall we cannot conclude anything like “−∞ + ∞ = 0.”)

We try using logarithm rules to get a clearer picture.
    
1 1
log sin + log(2n) = log 2n sin
n n

Still, we have indeterminate behaviour: 2n sin(1/n) is the product of 2n, which

grows without bound, and sin(1/n), which approaches zero. In the past, we
learned that we can handle the indeterminate form 0 · ∞ with l’Hôpital’s rule
(after a little algebra), but there’s a slicker way. Note 1/n → 0 as n → ∞. If we
write n1 = x, then this piece of our limit resembles something familiar.

   
1 sin x
2n sin =2
n x
1
If n → ∞, then x = n
→ 0.
 
1 sin x
lim 2n sin = 2 lim
n→∞ n x→0 x

That limit is familiar:

= 2(1) = 2

Then:
  
1
lim log 2n sin = log 2
n→∞ n

sin x
Note: if you have forgotten that lim = 1, you can also evaluate this limit
x→0 x
using l’Hôpital’s rule:

sin x cos x
lim = lim = cos 0 = 1
x→0 x x→0 1
| {z }
num→0
den→0

3.1.2.22. Solution. First, although this sequence is not defined for some small
values of n, it is defined as long as n ≥ 5, so it’s not a problem to take the limit
as n → ∞. Second, we notice that our limit has the indeterminate form ∞ − ∞.
Since this form is indeterminate, more work is needed to find our limit, if it exists.
A standard trick we saw last semester with functions
√ of this form
√ was to multiply
and divide by the conjugate of the expression, n + 5n + n2 − 5n. Then the
2

denominator will be the sum of two similar things, rather than their difference.
See the work below to find out why that is helpful.
√ √
n2 + 5n − n2 − 5n

1063
S OLUTIONS TO E XERCISES
√ √ !
√ √
n2 + 5n + n2 − 5n
= n2 + 5n − n2 − 5n √ √
n2 + 5n + n2 − 5n
(n2 + 5n) − (n2 − 5n)
=√ √
n2 + 5n + n2 − 5n
10n
=√ √
n2 + 5n + n2 − 5n
√
Now, we’ll cancel out n from the top and the bottom. Note n = n2 .
1
10n n
=√ √ 1
n + 5n + n2 − 5n
2
n
!
1
10n n
=√ √
n2 + 5n + n2 − 5n √1
n2
10
=q q
1 + n5 + 1 − n5

Now, the limit is clear.

10 10 10
lim q q =√ √ = =5
n→∞
1+ 5
+ 1− 5 1+0+ 1+0 1+1
n n

3.1.2.23. Solution. First, although this sequence

√ is not defined for some small
values of n, it is defined as long as n ≥ 2.5, so it’s not a problem to take the
limit as n → ∞. Second, we notice that our limit has the indeterminate form
∞ − ∞. Since this form is indeterminate, more work is needed to find our limit,
if it exists.
In Question 22, we saw a similar limit, and made use of the conjugate. However,
in this case, there’s an easier path: let’s factor out n from each term.
s   s 
√ √

2 2 2
5 2
5
n + 5n − 2n − 5 = n 1 + − n 2− 2
n n
r r
5 5
=n 1+ −n 2− 2
n n !
r r
5 5
=n 1+ − 2− 2
n n

Now, the limit is clear.

" !#
h√ √ i r
5
r
5
lim n2 + 5n − 2n2 − 5 = lim n 1+ − 2− 2
n→∞ n→∞ n n
h √ √ i
= lim n 1+0− 2−0
n→∞
= lim [n (−1)] = −∞
n→∞

1064
S OLUTIONS TO E XERCISES

Remark: check Question 22 to see whether a similar trick would work there. Why
or why not?
3.1.2.24. Solution. First, h we
note that iwe have in indeterminate form: as n
100
grows, 2 + n → 2, so n 2 + n1
1
− 2100 has the form ∞ · 0. To overcome this
difficulty, we could use some algebra and l’Hôpital’s rule, but there’s a slicker
way. If we let h = n1 , then h → 0 as n → ∞, and our limit looks like:
" 100 #
1 (2 + h)100 − 2100
lim n 2 + − 2100 = lim
n→∞ n h→0 h

This reminds us of the definition of a derivative.

d  100 (x + h)100 − x100
x = lim
dx h→0 h
So, if we set f (x) = x100 , our limit is simply f 0 (2). That is, [100x99 ]x=2 = 100 · 299 .
3.1.2.25. Solution. Using the definition of a derivative,

f (a + h) − f (a)
f 0 (a) = lim
h→0 h
We want n → ∞, so we set h = n1 .
1


f a+ n
− f (a)
= 1lim 1
n
→0 n
   
1
= lim n f a + − f (a)
n→∞ n

We also could have chosen h = − n1 , which leads to the following:

f a − n1 − f (a)


f (a + h) − f (a)
lim = lim
h→0 h −n1
→0 −1/n
   
1
= lim −n f a − − f (a)
n→∞ n
  
1
= lim n f (a) − f a −
n→∞ n
3.1.2.26. Solution. (a) To find the area An , note that the figure with n sides can
be divided up into n isosceles triangles, each with two sides of length 1 and angle
between them of 2πn
:

1065
S OLUTIONS TO E XERCISES

1
2π
n

Each of these triangles has area 12 sin 2π

n
:

sin 2π
n
2π
n

1
 
n 2π
All together, the area of the n-sided figure is An = sin .
2 n
(b) We will discuss two ways to find lim An , which has the indeterminate form
n→∞
∞ × 0.
First, note that as n → ∞, our figures look more and more like a circle of radius 1.
So, we see An is approaching the area of a circle of radius 1. That is, lim An = π.
n→∞
sin x 2π
Alternately, we can make use of the limit lim = 1. Let x = . Note if n → ∞,
x→0 x n
then x → 0.
   
n 2π π 2π
lim An = lim sin = lim 2π sin
n→∞ n→∞ 2 n n→∞ n
n
sin x
= lim π =π×1=π
x→0 x
3.1.2.27. Solution.
(
1 2≤x<3
a f2 (x) =
0 else

1066
S OLUTIONS TO E XERCISES

x
2 3
(
1 3≤x<4
b f3 (x) =
0 else

x
3 4

c For any n, fn (x) = 1 for an interval of length 1, and fn (x) = 0 for all other
x. So, the area under the curve is a square of side length one.

R∞
Then An = 0 fn (x) dx = 1 for all n. That is, the sequence {An } is simply
{1, 1, . . . , 1}, a sequence of all 1s.

d Given the description above, lim An = 1.

n→∞

e For any fixed x, recall {fn (x)} = {0, . . . , 0, 1, 0, . . . 0, 0, 0, 0, 0, . . .}. In partic-

ular, there are infinitely many zeroes at its end. So, lim fn (x) = 0. Then
n→∞
g(x) = 0 for every x.

1067
S OLUTIONS TO E XERCISES
Z ∞ Z ∞
f Given the description above, g(x) dx = 0 dx = 0.
0 0

Remark: what we’ve shown here is that, for this particular fn (x),
Z ∞ Z ∞
lim fn (x) dx 6= lim fn (x) dx
n→∞ 0 0 n→∞

That is, we can’t necessarily swap a limit with an integral (which is, in this case,
another limit, since the integral is improper). The interested reader can look up
“uniform convergence” to learn about the conditions under which these can be
swapped.
3.1.2.28. Solution. If we naively try to find the limit, we run up against the
indeterminate form 1∞ . We’d like to use l’Hôpital’s rule, but we don’t have the
∞
form ∞ or 00 — we’ll need to use a logarithm. Additionally, l’Hôpital’s rule ap-
plies to differentiable functions defined for real numbers — so we’ll consider a
function, rather than the sequence.
Note the terms of the sequence are all positive.

1/x
• Solution 1: Define x = n1 , and f (x) = (1 + 3x + 5x2 ) 1


. Then bn = f n
=
f (x), and  
1
lim f = lim+ f (x).
n→∞ n x→0

If this limit exists, it is equal to lim bn .

n→∞

1/x
lim+ f (x) = lim+ 1 + 3x + 5x2
x→0 x→0
h
i
2 1/x
lim+ log[f (x)] = lim+ log 1 + 3x + 5x
x→0 x→0
log [1 + 3x + 5x2 ]
= lim+
x→0 x
| {z }
num→0
den→0
3+10x
1+3x+5x2
= lim+ =3
x→0 1
3
lim f (x) = e
x→0+

Since the limit exists, lim bn = e3 .

n→∞

1
3

of letting x =
• Solution 2: If we didn’t see the nice simplifying trick
5 x
n
, we
can still solve the problem using g(x) = 1 + x + x2 :
 x
3 5
g(x) = 1 + + 2
x x

1068
S OLUTIONS TO E XERCISES

log 1 + x3 + 5
   
3 5 x2
log[g(x)] = x log 1 + + 2 =
x x 1/x
| {z }
num→0
den→0
3
− − 103
x2 x
3
1+ x + 52
3
x2
+ x103
lim log[g(x)] = lim −1
x
= lim x2
x→∞ x→∞
x2
x→∞ 1 + x3 + x52
3 + 10
x 3+0
= lim 3 5 =
x→∞ 1 +
x
+ x2 1+0+0
=3
lim f (x) = e3
x→∞

Since the limit exists, lim bn = e3

n→∞

3.1.2.29. Solution.
4+8
a When a1 = 4, we see a2 = = 4, and so on. That is, an = 4 for every n.
3
So, lim an = 4.
n→∞

b Cross-multiplying, we see 3x = x + 8, hence x = 4.

c In order for our sequence to converge to 4, the terms should be getting in-
finitely close to 4. So, we find the relationship between an+1 − 4 and an − 4.

an + 8
an+1 =
3
an + 8 an − 4
an+1 − 4 = −4=
3 8
So, the distance between our sequence terms and the number 4 is decreas-
ing by a factor of 8 each term. This implies that the terms get infinitely close
to 4 as n grows. That is, lim an = 4.
n→∞

3.1.2.30. Solution.
a Since w1 has the highest frequency, w2 has the next-highest frequency, and
so on, we know f1 is larger than the other members of its sequence, f2 is the
next largest, etc. So, {fn } is a decreasing sequence.

b The most-used word in a language is w1 , while the n-th most used word in
a language is wn . So, we re-state the law as:

f1 = nfn

Then we can rewrite this fomula a little more naturally as fn = n1 f1 .

c Then f3 = 31 f1 . In this case, we expect the third-most used word to account

for 31 (6%) = 2% of all words.

1069
S OLUTIONS TO E XERCISES

1
d From (b), we know f10 = f.
10 1
Note f1 = 6f6 = 6(0.3%). Then:

1 1 1 1.8
f10 = f1 = 6f6 = (6)(0.3%) = % = 0.18%
10 10 10 10

So, f10 should be 0.18% of all words.

e The use of the word “frequency” in the statement of Zipf’s law implies
uses of wn
fn = total number of words
. The question asks for the total uses of wn . If we call
this quantity tn , and the total number of all words is T , then Zipf’s law tells
us tTn = n1 tT1 , hence tn = n1 t1 .
With this notation, the problem states t1 = 22, 038, 615, w1 = the, w2 = be,
and w3 = and.
Following Zipf’s law, tn = n1 t1 . So, we expect t2 = t21 = 11, 019, 307.5;
since this isn’t an integer, let’s say we expect t2 ≈ 11, 019, 308. Similarly, we
expect t3 = t31 = 7, 346, 205.
Remark: The 450-million-word source material that used “the” 22,038,615
times also contained 12,545,825 instances of “be,” and 10,741,073 instances
of “and.” While Zipf’s Law might be a nice model for our data overall, in
these few instances it does not appear to be extremely accurate.

3.2 · Series
3.2.2 · Exercises

Exercises — Stage 1
3.2.2.1. Solution. The N th term of the sequence of partial sums, SN , is the sum
∞
X 1
of the first N terms of the series .
n=1
n

N SN
1 1
2 1 + 21
3 1 + 21 + 13
4 1 + 21 + 13 + 41
5 1 + 21 + 13 + 41 + 1
5

3.2.2.2. Solution. If there were a total of 17 cookies before Student 11 came, and
20 cookies after, then Student 11 brought 3 cookies.

1070
S OLUTIONS TO E XERCISES

1 2 3 4 5 6 7 8 9 10 11

C10

C11

3.2.2.3. Solution.
a We find {an } from {SN } using the same logic as Question 2. SN is the sum
of the first N terms of {an }, and SN −1 is the sum of all the same terms except
aN . So, aN = SN − SN −1 when N ≥ 2. Written another way:
SN = a1 + a2 + a3 + · · · + aN −2 + aN −1 + aN
SN −1 = a1 + a2 + a3 + · · · + aN −2 + aN −1

So,
h i
SN − SN −1 = a1 + a2 + a3 + · · · + aN −2 + aN −1 + aN
h i
− a1 + a2 + a3 + · · · + aN −2 + aN −1
= aN
So, we calculate
   
N N −1
aN = SN − SN −1 = −
N +1 N −1+1
2 2
N N −1
= −
N (N + 1) N (N + 1)
1
=
N (N + 1)
Therefore,
1
an =
n(n + 1)
Remark: the formula given for SN has S0 = 0, which makes sense: the sum
of no terms at all should be 0. However, it is common for a sequence of
partial sums to start at N = 1. (This fits our definition of a partial sum–
we don’t really define the “sum of no terms.”) In this case, a1 must be
calculated separately from the other terms of {an }. To find a1 , we simply
set a1 = S1 , which (to reiterate) might not be the same as S1 − S0 .

1071
S OLUTIONS TO E XERCISES

b
1
lim an = lim = 0.
n→∞ n→∞ n(n + 1)

That is, the terms we’re adding up are getting very, very small as we go
along.

c By Definition 3.2.3,
∞
X N
an = lim SN = lim =1
N →∞ N →∞ N + 1
n=1

That is, as we add more and more terms of our series, our cumulative sum
gets very, very close to 1.
3.2.2.4. Solution. As in Question 3,
   
N 1 N −1 1
aN = SN − SN −1 = (−1) + − (−1) +
N N −1
1 1
= (−1)N − (−1)N −1 + −
N N −1
N − 1 N
= (−1)N + (−1)N + −
N (N − 1) N (N − 1)
1
= 2(−1)N −
N (N − 1)

Note, however, that aN is only the same as SN − SN −1 when N ≥ 2: otherwise,

we’re trying to calculate S1 − S0 , but S0 is not defined. So, we find a1 separately:

1
a1 = S1 = (−1)1 + =0
1
All together:
(
0 if n = 1
an = 1
2(−1)n − n(n−1)
else

3.2.2.5. Solution. If f 0 (N ) < 0, that means f (N ) is decreasing. So, adding more

terms makes for a smaller sum. That means the terms we’re adding are negative.
That is, an < 0 for all n ≥ 2.

3.2.2.6. Solution. (a) To generate the pattern, we repeat the following steps:

• divide the top triangle into four triangles of equal area,

• colour the bottom two of them black, and

• leave the middle one white.

1072
S OLUTIONS TO E XERCISES

Every time we repeat this sequence, we divide up a triangle with an area one-
quarter the size of our previous triangle, and take two of the four resulting pieces.
1
So, our area should end up as a geometric sum with common ratio r = , and
4
coefficient a = 2. This is shown more explicitly below.
Since the entire triangle (outlined in red) has area 1, the four smaller triangles
1
below each have area . The two black triangles will be added to our total black
4
area; the blue triangle will be subdivided.

1 1
4 4

1
The blue triangle had area , so each of the small black triangles below has area
   4
1 1 1
= 2.
4 4 4

1 1
42 42

1 1
4 4

1073
S OLUTIONS TO E XERCISES

Each time we make another subdivision, we add two black triangles, each with
1
the area of the previous black triangles. So, our total black area is:
4
        ∞
1 1 1 1 X 2
2 + 2 2 + 2 3 + 2 4 + ··· =
4 4 4 4 n=1
4n

(b) To evalutate the series, we imagine gathering up all our little triangles and
sorting them into three identical piles: the bottom three triangles go in three
different piles, the three triangles directly above them go in three different piles,
etc. (In the picture below, different colours correspond to different piles.)

1
Since the piles all have equal area, each pile has a total area of . The black area
3
shaded in the problem corresponds to two piles (red and blue above), so
∞
X 2 2
n
=
n=1
4 3

3.2.2.7. Solution. (a) The pattern can be described as follows: divide the inner-
most square into 9 equal parts (a 3 × 3 grid), choose one square to be black, and
another square to subdivide.
The area of the red (outermost) square is 1, so the area of the largest black square
1 1
is . The area of the central, blue square below is also .
9 9

1074
S OLUTIONS TO E XERCISES

1
9

1
9

When we subdivide the blue square, the subdivisions each have one-ninth its
1
area, or 2 .
9

1
92
1
92

1
9

We continue taking squares that are one-ninth the area of the previous square.
So, our total black area is
∞
1 1 1 X 1
+ 2 + 3 + ··· =
9 9 9 n=1
9n

(b) If we cut up this square along the marks, we can easily share it equally among
1
8 friends: there are eight squares of area along the outer ring, eight squares of
9
1
area 2 along the next ring in, and so on.
9

1075
S OLUTIONS TO E XERCISES

1
Since the eight friends all get the same total area, the area each friend gets is .
8
The area shaded in black in the question corresponds to the pile given to one
friend. So,
∞
X 1 1
=
n=1
9n 8

3.2.2.8. Solution. If we start with a shape of area 1, and iteratively divide it into
thirds, taking one of the three newly created pieces each time, then the area we
∞
X 1
take will be equal to the desired series, .
n=1
3n
One way to do this is to start with a rectangle, make three vertical strips, then
keep the left strip and subdivide the middle strip.

We see that the total area we take approaches one-half the total area of the figure,
∞
X 1 1
so n
= .
n=1
3 2
Alternately, instead of always taking vertical strips, we could alternate vertical
and horizontal slices.

1076
S OLUTIONS TO E XERCISES

In this setup, we notice that our strips come in pairs: two large vertical strips,
two smaller horizontal strips, two smaller vertical strips, etc. We shaed exactly
∞
X 1 1
one of each, so the shaded area is one-half the total area: n
= .
n=1
3 2
Other solutions are possible, as well.
N
X 1 − rN +1
3.2.2.9. Solution. Lemma 3.2.5 tells us arn = a , for r 6= 1. Our
n=0
1−r
geometric sum has a = 1, r = 15 , and N = 100. So:
100 1
X 1 1 − 5101 5101 − 1
= =
n=0
5n 1 − 15 4 · 5100

3.2.2.10. Solution. After twenty students have brought their cookies, the pile
numbers 53 cookies. 17 of these cookies were brought by students one through
ten. So, the remainder (53−17 = 36) is the number of cookies brought by students
11, 12, 13, 14, 15, 16, 17, 18, 19, and 20, together.

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

C10

C20

1077
S OLUTIONS TO E XERCISES

3.2.2.11. Solution.

• Solution 1: Using the ideas of Question 10, we see:

100 100 49
X 1 X 1 X 1
n
= n
−
n=50
5 n=0
5 n=0
5n

1
That is, we want start with the sum of all the terms up to , and then
5100
1
subtract off the ones we actually don’t want, which is everything up to 49 .
5
Now, both series are in a form appropriate for Lemma 3.2.5.
100 49 1
X 1 X 1 1 − 5101 1 − 5150
− = −
n=0
5n n=0 5n 1 − 15 1 − 51
5101 − 1 550 − 1 551
 
= −
4 · 5100 4 · 549 551
5101 − 1 5101 − 551
= −
4 · 5100 4 · 5100
551 − 1
=
4 · 5100

• Solution 2: If we write out the first few terms of our series, we see we can
factor out a constant to change the starting index.
100
X 1 1 1 1 1 1
= + + + + · · · +
n=50
5n 550 551 552 553 5100
 
1 1 1 1 1 1
= 50 + + + + · · · + 50
5 50 51 52 53 5
50
X 1 1
= 50
· n
n=0
5 5

1 1
Now, our sum is in the form of Lemma 3.2.5 with a = 50 , r = , and
5 5
N = 50.
50
1 1 − 5151 1 − 5151 551
 
X 1 1
·
50 5n
= 50 · 1 =
n=0
5 5 1 − 5
4 · 549 551
551 − 1
=
4 · 5100
3.2.2.12. Solution. (a) The table below is a record of our account, with black en-
tries representing the money your friend gives you, and red entries representing

1078
S OLUTIONS TO E XERCISES

the money you give them (which is why the red entries are negative).
1 1
d − d+1 d
1 − 12 1 1
2
2 − 13 1
2
2
3
3 − 14 1
3
3
4
4 − 15 1
4
4
5
5 − 16 1
5
5
6
6 − 17 1
6
6
7

1


After the exchange of day n, the amount you’re left with is 1 − n+1 . We see
this by the cancellation in the table: the $ 21 you gave your friend on day 1 was
returned on day 2; the $ 13 you gave your friend on day 2 was returned on day 3,
etc.
So, after a long time, you’ll have gained close to (but always slightly less than)
one dollar. ∞  
X 1 1
(b) The series − describes the scenario in (a), so by our reasoning
d=1
d (d + 1)
there,
∞    
X 1 1 1
− = lim 1 − =1
d=1
d (d + 1) n→∞ n + 1
(c) Again, let’s set up an account book.

d d + 1 −(d + 2)
1 2 −3 −1
2 3 −4 −2
3 4 −5 −3
4 5 −6 −4
5 6 −7 −5
6 7 −8 −6
By day d, you’ve lost $ d to your so-called friend. As time goes on, you lose more
and more. ∞
X
(d) The series ((d + 1) − (d + 2)) exactly describes the scenario in part (c), so it
d=1
∞
X ∞
X
diverges to −∞. You can also see this by writing ((d+1)−(d+2)) = (−1) =
d=1 d=1
−1 − 1 − 1 − 1 − 1 − · · ·.
Be careful to avoid a common mistake with telescoping series: if we look back
at our account book, we see that every negative term will cancel with a positive
term, with the initial +2 as the only term that never cancels. Your friend takes
$3, which they return the next day; then they take $4, which they return the next
day; then they take $5, which they return the next day, and so on. It’s extremely
tempting to say that the series adds up to $2, since every other term cancels out
eventually. This is where we lean on Definition 3.2.3: we evaluate the partial

1079
S OLUTIONS TO E XERCISES

sums, which always leave your friend’s last withdrawal unreturned. This defini-
tion makes sense: saying “I gained two bucks from this exchange” doesn’t really
capture the reality of your increasing debt.
3.2.2.13. Solution. Using arithmetic of series, Theorem 3.2.9, we see
∞
X ∞
X
(an + bn + cn+1 ) = A + B + cn+1
n=1 n=1

The question remaining is what do to with the last series. If we write out the
∞
X X∞
terms, we see the difference between cn and cn+1 is simply that the latter
n=1 n=1
is missing c1 :
∞
X
cn+1 = c2 + c3 + c4 + c5 + · · ·
n=1
= −c1 + c1 + c2 + c3 + c4 + c5 + · · ·
X∞
= −c1 + cn
n=1

So,
∞
X
(an + bn + cn+1 ) = A + B + C − c1
n=1

3.2.2.14. Solution. Theorem 3.2.9, arithmetic of series, doesn’t mention divi-

sion, because in general it doesn’t work the way the question suggests. For ex-
ample, let {an } = {bn } = 21n . Then:
P∞ P∞ 1
• n=0 an = n=0 bn = 1− 1 = 2, while
2

P∞ an P∞
• n=0 = n=0 1 = ∞.
bn
For the statement in the question, we can take {an } = {bn } = 21n , A = B = 2,
{cn } = {0, 0, 0, . . .}, and C = 0. We see the statement is false in this case.
So, in general, the statement given is false.

Exercises — Stage 2
3.2.2.15. *. Solution. We recognize that this is a geometric series:

1 1 1 1 1
1+ + + + + + ···
3 9 27 81 243
1 1 1 1 1 1
= 0 + 1 + 2 + 3 + 4 + 5 +·
3 3 3 3 3 3

1080
S OLUTIONS TO E XERCISES

∞
X 1
=
n=0
3n

1
Using Lemma 3.2.5 with r = and a = 1,
3
1 3
= 1 = .
1− 3
2
1
3.2.2.16. *. Solution. This is a geometric series, with ratio r = . However, it
8
doesn’t start at k = 0, which is what we’re used to.

• Solution 1: We write out the first few terms of the series to figure out a
convenient constant to factor out.
∞
X 1 1 1 1
k
= 7
+ 8
+ 9
+ ···
k=7
8 8 8 8
 
1 1 1 1
= 7 + + + ···
8 80 81 82
∞
X 1 1
= ·
7 8n
k=0
8

1 1
We now evaluate the series using Lemma 3.2.5 with r = , a = 7 .
8 8
1 1 1
= · 1 =
87 1 − 8
7 × 86

• Solution 2: Using the idea of Question 10, we express the series we’re inter-
ested in as the difference of two series that we can easily evaluate.
∞ ∞ 6
X 1 X 1 X 1
= −
k=7
8k k=0
8k k=0 8k

Using Lemma 3.2.5,

1 1 − 817
= −
1 − 18 1 − 18
1
=
7 × 86

1081
S OLUTIONS TO E XERCISES

3.2.2.17. *. Solution. We recognize this as a telescoping series.

6 6
k k2
− (k+1) 2 sk
6
1 6 −4 6 − 64
6
2 4
− 69 6 − 69
6 6 6
3 9
− 16 6 − 16
6 6 6
4 16
− 25 6 − 25
6 6 6
5 25
− 36 6 − 36
6 6 6
6 36
− 47 6 − 47
..
.
When we compute the nth partial sum, i.e. the sum of of the
 first n terms,
 succes-
6
sive terms cancel and only the first half of the first term, k62 − (k+1)2 , and
  k=1
the second half of the nth term, k62 − (k+1)
6
2 , survive. That is:
k=n

n  
X 6 6 6 6
sn = 2
− 2
= 2−
k=1
k (k + 1) 1 (n + 1)2

Therefore, we can see directly that the sequence of partial sums {sn } is conver-
gent:
 
6
lim sn = lim 6 − =6
n→∞ n→∞ (n + 1)2

By Definition 3.2.3 the series is also convergent, with limit 6.

3.2.2.18. *. Solution. We recognize that this is a telescoping series, and set up a
table to find the sequence of partial sums.

n cos nπ − cos n+1 π

sn
π π 1 π




3 cos 3
− cos 4
2
− cos 4

1
4 cos π4
− cos π5
2
− cos π5

1
5 cos π5
− cos π6
2
− cos π6

1
6 cos π6
− cos π7
2
− cos π7

π π 1
7 cos 7
− cos 8
2
− cos π8

1
8 cos π8 − cos π9 2
− cos π9
..
.
The N th partial sum sees every term cancel except the first part of the first term
( 21 ) and the second part of the last term (− cos( n+1
π
)).

N 
X π   π 
sN = cos − cos
n=3
n n+1
π   π 
= cos − cos
3 N +1

1082
S OLUTIONS TO E XERCISES

1  π 
= − cos .
2 N +1
As N → ∞, the argument Nπ+1 converges to 0, and cos x is continuous at x = 0.
By Definition 3.2.3, the value of the series is
  π 
1
lim sN = lim − cos
N →∞ N →∞ 2 N +1
1 1
= − cos(0) = −
2 2
3.2.2.19. *. Solution. (a) As in Question 2, since

sn−1 = a1 + a2 + · · · + an−1
sn = a1 + a2 + · · · + an−1 + an

we can find an by subtracting:

an = sn − sn−1
1 + 3n 1 + 3(n − 1) 3n + 1 3n − 2
= − = −
5 + 4n 5 + 4(n − 1) 4n + 5 4n + 1
(3n + 1)(4n + 1) − (3n − 2)(4n + 5)
=
(4n + 1)(4n + 5)
11
= 2
16n + 24n + 5
(b) Using Definition 3.2.3,
∞
X 1 + 3n 1/n + 3 0+3 3
an = lim sn = lim = lim = =
n→∞ n→∞ 5 + 4n n→∞ 5/n + 4 0+4 4
n=1

3
The series converges to .
4
3.2.2.20. *. Solution. What we have is a geometric series, but we need to get it
into the proper form before we can evaluate it.
∞ ∞ ∞
X 3 · 4n+1 X 3 · 4 · 4n 3 X  4 n
= =
n=2
8 · 5n n=2
8 · 5n 2 n=2 5

2
• Solution 1: If we factor our 45 , we can change our index to something
more convenient.
∞  n ∞  2  n−2
3X 4 3X 4 4
=
2 n=2 5 2 n=2 5 5
∞   2  n
3X 4 4
=
2 n=0 5 5

1083
S OLUTIONS TO E XERCISES

4
We use Lemma 3.2.5 with r = .
5
 2
3 4 1 24
= · 4 =
2 5 1− 5
5

• Solution 2: Using the idea of Question 10, we view our series as a more
convenient series, minus a few initial terms.
∞  n
"∞   #    0 !
n 1
3 X 4 3 X 4 4 4
= − −
2 n=2 5 2 n=0
5 5 5
∞
!
  n
3 X 4 9
= −
2 n=0 5 5

4
We use Lemma 3.2.5 with r = .
5
 
3 1 9 24
= 4 − =
2 1− 5
5 5
3.2.2.21. *. Solution. The number is:
3 3 3 1 3 3 3
0.2 + + + + ··· = + 2 + 3 + 4 + ···
100 1000 10000 5 10 10  10
1 3 1 1 1
= + 2 0
+ 1 + 2 + ···
5 10 10 10 10
∞
1 3 X 1
= + 2
5 10 n=0 10n

1
We use Lemma 3.2.5 with r = .
10
1 3 1
= + 2· 1
5 10 1 − 10
1 1 7
= + =
5 30 30
3.2.2.22. *. Solution. The number is:
65 65 65 65 65 65
2+ + + + ··· = 2 + + 2
+ + ···
100 10000 1000000 100 100 1003
∞
65 X 1
=2+
100 n=0 100n

1084
S OLUTIONS TO E XERCISES

1
We use Lemma 3.2.5 with r = .
100
65 1
=2+ · 1
100 1 − 100
65 263
=2+ =
99 99
3.2.2.23. *. Solution. The number is:

0.321 = 0.321321321 . . .
321 321 321
= + 6 + 9 + ···
1000 10 10 !
 0  1  2
321 1 1 1
= + + + ···
1000 103 103 103
∞  n
321 X 1
=
1000 n=0 103

1
We use Lemma 3.2.5 with r = .
103
321 1 321 107
= · 1 = =
1000 1 − 103 999 333
3.2.2.24. *. Solution. We split the sum into two parts.

∞  n+1 
X 2 1 1
+ −
n=2
3n 2n − 1 2n + 1
∞ ∞ 
2n+1

X X 1 1
= + −
n=2
3n n=2
2n − 1 2n + 1

The first part is a geometric series.

∞ ∞ ∞  n
X 2n+1 X 2n+3 X 23 2
= = 2
·
n=2
3n n=0
3n+2 n=0
3 3

2 8
We use Lemma 3.2.5 with r = and a = .
3 9
8 1 8
= · 2 =
9 1− 3
3

1085
S OLUTIONS TO E XERCISES

The second part is a telescoping series. Let’s make a table to see how it cancels.
1 1
n 2n−1
− 2n+1 sn
1
2 3
− 51 1
− 15
3
1
3 5
− 71 − 17
1
3
1
4 7
− 91 1
− 19
3
1 1 1 1
5 9
− 11 3
− 11
1 1 1 1
6 11
− 13 3
− 13
1 1 1 1
7 13
− 15 3
− 15
..
.
After adding terms n = 2 through n = N , the partial sum is

1 1
sN = −
3 2N + 1
because all the terms except the first part of the n = 2 term, and the last part of
the n = N term, cancel. Then:
∞  
X 1 1 1 1
− = lim sN = lim −
n=2
2n − 1 2n + 1 N →∞ N →∞ 3 2N + 1
1
=
3
All together,
∞  n+1 
X 2 1 1
+ −
n=2
3n 2n − 1 2n + 1
∞ ∞ 
2n+1

X X 1 1
= + −
n=2
3n n=2
2n − 1 2n + 1
8 1
= + =3
3 3
3.2.2.25. *. Solution. We split the sum into two parts.
∞  n  2 n−1  X ∞  n ∞ 
X 1 1 X 2 n−1
+ − = + −
n=1
3 5 n=1
3 n=1
5

Both are geometric series.

∞  n+1 ∞ 
X 1 2 n X
= + −
n=0
3 n=0
5
∞ ∞
1 X  1 n X  2 n
= + −
3 n=0 3 n=0
5

1086
S OLUTIONS TO E XERCISES

1 1 2
We use Lemma 3.2.5 with a1 = and r1 = , then with a2 = 1 and r2 = − .
3 3 5
1 1 1
= · 1 + 2
3 1− 3 1+ 5
1 5 17
= + =
2 7 14
3.2.2.26. *. Solution. We split the sum into two parts.
∞ ∞ ∞
X 1 + 3n+1 X 1 X 3n+1
= +
n=0
4n n=0
4n n=0 4n
∞ ∞  n
X 1 X 3
= n
+3
n=0
4 n=0
4

Using Lemma 3.2.5,

1 3
= 1 + 3
1− 4 1− 4
4 40
= + 12 =
3 3
3.2.2.27. Solution. Using logarithm rules, we see
∞   ∞
X n−3 X  
log = log(n − 3) − log n
n=5
n n=5

which looks like a telescoping series. Let’s make a table to figure out the partial
sums.

n log(n−3) − log n sn
5 log 2 − log 5 log 2 − log 5
6 log 3 − log 6 log 2 + log 3 − log 5− log 6
7 log 4 − log 7 log 2 + log 3 + log 4 − log 5− log 6− log 7
8 log 5 − log 8 log 2 + log 3 + log 4− log 6− log 7− log 8
9 log 6 − log 9 log 2 + log 3 + log 4− log 7− log 8− log 9
10 log 7 − log 10 log 2 + log 3 + log 4− log 8− log 9− log 10
11 log 8 − log 11 log 2 + log 3 + log 4− log 9− log 10− log 11
..
.
There is a “lag” before the terms cancel, which is why they “build up” more than
we saw in past examples. Still, we can clearly see the N th partial sum:
N 
X 
log(n − 3) − log(n)
n=5
= log 2 + log 3 + log 4 − log(N − 2) − log(N − 1) − log(N )

1087
S OLUTIONS TO E XERCISES
 
24
= log
N (N − 1)(N − 2)

when N ≥ 7. So,
∞ 
X 
log(n − 3) − log(n) = lim sN
N →∞
n=5
 
24
= lim log
N →∞ N (N − 1)(N − 2)
= −∞
3.2.2.28. Solution. This is a telescoping series. Let’s investigate it in the usual
way. We leave the fractions in the middle of the table unsimplified, to make the
pattern of cancellation clearer, since terms with the same denominator cancel.
2 1 1
n − − sn
n n+1 n−1
2 1 1 1
2 − − −
2 3 1 3
2 1 1 1 1 1
3 − − − −
3 4 2 3 4 2
2 1 1 1 1 1
4 − − − −
4 5 3 4 5 2
2 1 1 1 1 1
5 − − − −
5 6 4 5 6 2
2 1 1 1 1 1
6 − − − −
6 7 5 6 7 2
2 1 1 1 1 1
7 − − − −
7 8 6 7 8 2
2 1 1 1 1 1
8 − − − −
8 9 7 8 9 2
..
.
In general, there are three terms with the same denominator, and these cancel out
to zero, but it takes a while to gather all three. So, some are left over in the partial
sum.

N  
X 2 1 1 1 1 1
sN = − − = − −
n=2
n n+1 n−1 N N +1 2

Therefore,
∞  
X 2 1 1
− − = lim sN
n=2
n n+1 n−1 N →∞
 
1 1 1
= lim − −
N →∞ N N +1 2
1
=−
2

1088
S OLUTIONS TO E XERCISES

Exercises — Stage 3
1
3.2.2.29. Solution. The stone at position x has mass x kg, and we have to pull
4
it a distance of 2x metres, so the work involved in moving that one stone is
 
1 x
 m  9.8
kg (2 m) 9.8 2 = x J
4x sec 2

Therefore, the work to move all the stones is:

∞ ∞
X 9.8 X 9.8
x
=
x=1
2 x=0
2x+1
∞
X 9.8 1 9.8 1
= · x
= · 1 = 9.8 J
x=0
2 2 2 1− 2

1
3.2.2.30. Solution. The volume of a sphere of radius is
πn
 3  n
4 1 4π 1
vn = π =
3 πn 3 π3

So, the volume of all the spheres together is:

∞ ∞  n
X X 4π 1
vn =
n=1 n=1
3 π3
∞  n+1
X 4π 1
=
n=0
3 π3
∞  n
X 4 1
= 2
n=0
3π π3

4 1
We use Lemma 3.2.5 with a = 2
and r = 3 .
3π π
4 1 4π
= · 1 =
3π 2 1 − π3 3 (π 3 − 1)

1089
S OLUTIONS TO E XERCISES

3.2.2.31. Solution. Let’s make a table. Keep in mind cos2 θ + sin2 θ = 1.

sin2 n cos2 (n + 1)
n sn
2n2 2n+1
sin 3 cos2 4 sin2 3 cos2 4
3 +
23 24 23 24 2
sin2 4 cos2 5 2
sin 3 1 cos 5
4 + +
24 25 23 24 25 2
sin2 5 cos2 6 sin2 3 1 1 cos 6
5 + 4+ +
25 26 23 2 25 26 2
sin2 6 cos2 7 2
sin 3 1 1 1 cos 7
6 + + + 6+
26 27 23 24 25 2 27 2
sin2 7 cos2 8 sin2 3 1 1 1 1 cos 8
7 + 4+ 5+ + 7+
27 28 23 2 2 26 2 28
..
.

This gives us an equation for the partial sum sN , when N ≥ 4:

N 
sin2 n cos2 (n + 1)
X 
sN = +
n=3
2n 2n+1
N
!
sin2 3 X 1 cos2 (N + 1)
= + +
23 n=4
2n 2N +1

Using Definition 3.2.3, our series evaluates to:

" N
! #
sin2 3 X 1 cos2 (N + 1)
lim sN = lim + +
N →∞ N →∞ 23 n=4
2n 2N +1
∞
sin2 3 cos2 (N + 1)
  X
1
= + lim N +1
+
8 N →∞ 2 n=4
2n

We evaluate the limit using the squeeze theorem; the series is geometric.
∞
sin2 3 X 1
= +0+
8 n=0
2n+4
∞
sin2 3 1 X 1
= + 4
8 2 n=0 2n

Using Lemma 3.2.5,

sin2 3 1 1
= + 4 1
8 2 1− 2
sin2 3 1
= + ≈ 0.1275
8 8

1090
S OLUTIONS TO E XERCISES

∞
X
3.2.2.32. Solution. Since {SM } is the sequence of partial sums of SN , we
N =1
can find {SN } from {SM } as in Question 3:

N +1 N 1
SN = SN − SN −1 = − =− if N ≥ 2,
N N −1 N (N − 1)
S1 = S1 = 2

Similarly, we find {an } from {SN }. Do be careful: SN only follows the formula
we found above when N ≥ 2. In the next line, we use an expression containing
Sn−1 ; in order for the subscript to be at least two (so the formula fits), we need
n ≥ 3.
1 −1
an = Sn − Sn−1 = − −
n(n − 1) (n − 1)(n − 2)
2
= if n ≥ 3,
n(n − 1)(n − 2)
1 5
a2 = S2 − S1 = − −2=− ,
2(2 − 1) 2
a1 = S1 = 2

All together,

2

 n(n−1)(n−2)
 if n ≥ 3,
an = − 52 if n = 2,


2 if n = 1

3.2.2.33. Solution. We consider a circle of radius R, with an “inner ring” from

R
3
to 2R
3
and an “outer ring” from 2R
3
to R.
The area of the outer ring is:
 2
2 2R 5
πR − π = πR2
3 9
The area of the inner ring is:
 2  2
2R R 3
π −π = πR2
3 3 9
So, the ratio of the inner ring’s area to the outer ring’s area is 35 .
In our bullseye diagram, if we pair up any red ring with the blue ring just inside
it, the blue ring has 35 the area of the red ring. So, the blue portion of the bullseye
has 53 the area of the red portion.
Since the circle has area 1, if we let the red portion have area A, then
3 8
1=A+ A= A
5 5
So, the red portion has area 85 .

1091
S OLUTIONS TO E XERCISES

3.3 · Convergence Tests

3.3.11 · Exercises

Exercises — Stage 1
3.3.11.1. Solution.
1
A lim = 0, so the divergence test is inconclusive. It’s true that this series
n→∞ n
diverges, but we can’t show it using the divergence test.
n2
B lim = ∞, which is not zero, so the divergence test tells us this series
n→∞ n + 1
diverges.

C lim sin n does not exist, so in particular it is not zero. Therefore, the diver-
n→∞
gence test tells us this series diverges.

D For all whole numbers n, sin(πn) = 0, so lim sin(πn) = 0 and the diver-
n→∞
gence test is inconclusive.
3.3.11.2. Solution. Let f (x) be a function with f (n) = an for all whole numbers
n. In order to apply the integral test (Theorem 3.3.5) we need f (x) to be positive
and decreasing for all sufficiently large values of n.

A f (x) = x1 , which is positive and decreasing for all x ≥ 1, so the integral test
does apply here.
x 2
B f (x) = x+1 , which is not decreasing — in fact, it goes to infinity. So, the
integral test does not apply here. (The divergence test tells us the series
diverges, though.)

C f (x) = sin x, which is neither consistently positive nor consistently decreas-

ing, so the integral test does not apply. (The divergence test tells us the
series diverges, though.)

D f (x) = sinxx+1
2 is positive for all whole numbers n. To determine whether it
is decreasing, we consider its derivative.

x2 (cos x) − (sin x + 1)(2x) x cos x − 2 sin x − 2

f 0 (x) = 4
=
x x3
This is sometimes positive, and sometimes negative. (For example, if x =
100π, f 0 (x) = 100π−0−2
(100π)3
> 0, but if x = 101π then f 0 (x) = 101π(−1)−0−2
(101π)3
< 0.)
Then f (x) is not a decreasing function, so the integral test does not apply.
3.3.11.3. Solution. (a) If Olaf is old, and I am even older, then I am old as well.
(b) If Olaf is old, and I am {not as old}, then perhaps I am old as well (just slightly
less so), or perhaps I am young. There is not enough information to tell.
(c) If Yuan is young, and I am older, then perhaps I am much older and I am

1092
S OLUTIONS TO E XERCISES

old, or perhaps I am only a little older, and I am young. There is not enough
information to tell.
(d) If Yuan is young, and I am even younger, then I must also be young.
Another way to think about this is with a timeline of birthdates. People born
before the threshold are old, and people born after it are young.

Olaf threshold Yuan

If I’m born before (older than) Olaf, I’m born before the threshold, so I’m old. If
I’m born after (younger than) Yuan I’m born after the threshold, so I’m young.

Olaf threshold Yuan

definitely old definitely young

If I’m born after Olaf or before Yuan, I don’t know which side of the threshold
I’m on. I could be old or I could be young.

also older than Yuan older than Yuan

threshold Yuan

3.3.11.4. Solution. The comparison test is Theorem 3.3.8. However, rather than
trying to memorize which way the inequalities go in all cases, we use the same
reasoning as Question 3.
If a sequence has positive terms, it either converges, or it diverges to infinity,
with the partial sums increasing and increasing without bound. If one sequence
diverges, and the other sequence is larger, then the other sequence diverges —
just
Plike being older than an old person makes you old. P
If an converges, and {an } is the red (larger) series, then bn converges: it’s
smaller than a sequence that doesn’t add up to infinity, so it too does not add up
to P
infinity. P
If an diverges, and {an } is the blue (smaller) series, then bn diverges: it’s
larger than a sequence that adds up to infinity, so it too adds up to infinity.
P the other cases, we can’t say anything. If {an } is the red (larger)
In P series, and
an diverges, then perhaps {bn } behaves similarly toP{an } and bn diverges,
or perhaps {bn } is much, much smaller than {an } and P bn converges.
Similarly, if {an } is the blue (smaller)Pseries, and an converges, then perhaps
{bn } behaves similarly to {a Pn } and bn converges, or perhaps {bn } is much,
much bigger than {an } and bn diverges.

1093
S OLUTIONS TO E XERCISES

P P
if an converges if an diverges
P
and if {an } is the red series then bn CONVERGES inconclusive
P
and if {an } is the blue series inconclusive then bn DIVERGES

P1
3.3.11.5. Solution. (a) Since n
is divergent, we can only use it to prove series
1
with larger terms are divergent. This is the case here, since n−1 > n1 . So, the direct
comparison test is valid.
For the limit comparison test, we calculate:
1
an n−1 1
lim = lim 1 = lim 1 =1
n→∞ bn n→∞ n→∞ 1 −
n n

Since the limit exists

Pand is not zero, the limit comparison test is also valid.
1
(b) Since the series n2
converges, we can only use the direct comparison test to
show the convergence of a series if its terms have smaller absolute values. Indeed,
sin n | sin n| 1
2
= 2 < 2
n +1 n +1 n
so the series are set for a direct comparison.
To check whether a limit comparison will work, we compute:
sin n
an n2 +1 n2
lim = lim 1 = lim sin n = lim (1) sin n
n→∞ bn n→∞ n→∞ n2 + 1 n→∞
n2

The limit does not exist, so the limit comparison test is not a valid test to compare
these two series. P
1
(c) Since the series n3
converges, we can only use the direct comparison test to
conclude something about a series with smaller terms. However,
n3 + 5n + 1 n3 1
> = .
n6 − 2 n6 n3
Therefore the direct comparison test does not apply to this pair of series.
For the limit comparison test, we calculate:
n3 +5n+1
n3 + 5n + 1 n13
 
an n6 −2
lim = lim 1 = lim
n→∞ bn n→∞
n3
n→∞ n3 − n23 1
n3
1 + n52 + 1
n3
= lim =1
n→∞ 1 − n26
Since the limit is a nonzero real number, we can use the limit comparison test to
compare this pair ofPseries.
1
(d) Since the series 4 n diverges, we can only use the direct comparison test to
√

show that a series with larger terms diverges. However,

1 1
√ < √
n 4
n

1094
S OLUTIONS TO E XERCISES

so the direct comparison test isn’t valid with this pair of series.
For the limit comparison test, we calculate:

√1
an n 1
lim = lim 1 = lim √ =0
n→∞ bn n→∞ √
4n n→∞ 4
n

Since the limit is zero, the limit comparison test doesn’t apply.
3.3.11.6. Solution. It diverges by the divergence test, because lim an 6= 0.
n→∞

3.3.11.7. Solution. The divergence test (Theorem 3.3.1) is inconclusive when

lim an = 0. We cannot use the divergence test to show that a series converges.
n→∞

3.3.11.8. Solution. The integral test does not apply because f (x) is not decreas-
ing.
3.3.11.9. Solution. The inequality goes the wrong way, so the direct comparison
test (with this comparison series) is inconclusive.
3.3.11.10. Solution. Although the terms of (A) are sometimes negative and
sometimes positive, they are not strictly alternating in a positive-negative-
positive-negative pattern. For instance, sin 1 and sin 2 are both postive. So, (A) is
not an alternating series.
When n is a whole number, cos(πn) = (−1)n , so (B) is alternating.
Since the exponent of (−n) in (C) is even, the terms are always positive. Therefore
(C) is not alternating.
(D) is an alternating series.
∞
X 1
3.3.11.11. Solution. One possible answer: . This series converges (it’s a
n=1
n2
p-series with p = 2 > 1), but if we take the ratio of consecutive terms:

an+1 n2
lim = lim =1
n→∞ an n→∞ (n + 1)2

The limit of the ratio is 1, so the ratio test is inconclusive.

P
3.3.11.12. Solution. By the divergence test, for a series an to converge, we
need lim an = 0. That is, the magnitude (absolute value) of the terms needs to be
n→∞
an an+1
getting smaller. If lim < 1 or (equivalently) lim > 1, then |an+1 | >
n→∞ an+1 n→∞ an
|an | for sufficiently large n, so the terms are actually growing in magnitude. That
means the series diverges, by the divergence test.
3.3.11.13. Solution. The terms of the series only see a small portion of the do-
main of the integral. We can try to think of a function f (x) that behaves “nicely”
when x is a whole number (that is, it produces a sequence whose sum converges),
but is more unruly when x is not a whole number.
For example, suppose f (x) = sin(πx). Then f (x) = 0 for every integer x, but this

1095
S OLUTIONS TO E XERCISES

is not representative of the function as a whole. Indeed, our corresponding series

has terms {an } = {0, 0, 0, . . .}.
Z ∞  R
1 h i 1
• sin(πx) dx = lim − cos(πx) = lim − cos(πR) − Since the
1 R→∞ π 1
R→∞ π
limit does not exist, the integral diverges.
∞
X ∞
X
• sin(πn) = 0 = 0. The series converges.
n=1 n=1

3.3.11.14. *. Solution. When n is very large, the term 2n dominates the numera-
n
tor, and the term 3n dominates the denominator. So when n is very large an ≈ 23n .
2n
Therefore we should take bn = n . Note that, with this choice of bn ,
3
an 2n + n 3n 1 + n/2n
lim = lim n = lim =1
n→∞ bn n→∞ 3 + 1 2n n→∞ 1 + 1/3n

as desired.
∞
1
P
3.3.11.15. *. Solution. (a) In general false. The harmonic series n
diverges
n=1
by the p-test with p = 1.
(b) Be careful. You were not told that the an ’s are positive. So this is false in
∞ ∞
general. If an = (−1)n n1 , then (−1)n an is again the harmonic series 1
P P
n
, which
n=1 n=1
diverges.
(c) In general false. Take, for example, an = 0 and bn = 1.

Exercises — Stage 2
3.3.11.16. *. Solution. First, we’ll check the divergence test. It doesn’t always
work, but if it does, it’s likely the easiest path.
1
n2
 
n2 1 1
lim √ 1 = lim 1 = 6= 0
n→∞ 3n2 + n n→∞ 3+ n n
√ 3
n2

Since the limit of the terms being added is not zero, the series diverges by the
divergence test.
3.3.11.17. *. Solution. This precise question was asked on a 2014 final exam.
k
Note that the nth term in the series is an = 4k5+3k and does not depend on n! There
are two possibilities. Either this was intentional (and the instructor was being
n
particularly nasty) or it was a typo and the intention was to have an = 4n5+3n . In
both cases, the limit

5k 5k
lim an = lim = 6= 0
n→∞ n→∞ 4k + 3k 4k + 3k

1096
S OLUTIONS TO E XERCISES

5n (5/4)n
lim an = lim = lim = +∞ =
6 0
n→∞ n→∞ 4n + 3n n→∞ 1 + (3/4)n

is nonzero, so the series diverges by the divergence test.

3.3.11.18. *. Solution. We usually check the divergence test first, to look for
low-hanging fruit. The limit of the terms being added is zero:

1
lim 1 =0
n→∞ n+ 2

so the divergence test is inconclusive. That is, we need to look harder.

Next, we might consider a comparison test — these can also provide us (if we’re
lucky) with an easy path. The terms we’re adding look somewhat like n1 , but
our terms are smaller than these terms, which form the terms of the divergent
harmonic series. So, a direct comparison seems unlikely. Now we search for
more exotic tests.
1
Let f (x) = . Note f (x) is positive and decreases as x increases. So, by the
x + 12
integral test,
R ∞ which is Theorem 3.3.5, the given series converges if and only if the
1
integral 0 x+ 1 dx converges. Since
2

∞ R  x=R
1 1 1
Z Z 
1 dx = lim 1 dx = lim log x +
0 x+ 2
R→∞ 0 x+ 2
R→∞ 2 x=0
 
 1 1
= lim log R + − log
R→∞ 2 2

diverges, the series diverges.

3.3.11.19. Solution. The terms of the series tend to 0, so we can’t use the diver-
gence test.
To generate a guess about its convergence, we do the following:
X 1 X 1 X 1 X1
√ √ = √ ≈ √ =
k k+1 k2 + k k2 k
We guess that our series behaves like the harmonic series, and the harmonic se-
ries diverges (which can be demonstrated by p-test or integral test). So, we guess
that our series diverges. However, in order to directly compare our series to the
harmonic series and show our series diverges, our terms would have to be bigger
than the terms in the harmonic series, and this is not the case. So, we use limit
comparison.

1
√ √ r r
k k2 + k k2 + k k2 + k 1
= = √ = = 1 + , so
√ 1 k k2 k2 k
k2 +k
1
r
k 1
lim = lim 1+ =1
k→∞ √ 12 k→∞ k
k +k

1097
S OLUTIONS TO E XERCISES

√ √1
P
Since 1 is a real number greater than 0, by the Limit Comparison Test, k k+1
P1
diverges, like k
.
3.3.11.20. Solution. This is a geometric series with r = 1.001. Since |r| > 1, it is
divergent.

3.3.11.21. Solution. This is a geometric series with r = −1 5

. Since |r| < 1, it is
convergent.
We want to use the formula ∞ n 1
P
n=0 r = 1−r , but our series does not start at 0, so
we re-write it:
∞  n X ∞  n X 2  n
X −1 −1 −1
= −
n=3
5 n=0
5 n=0
5
 
1 1 1
= − 1− +
1 − (−1/5) 5 25
1 1 1 5 −25 + 5 − 1
= −1+ − = +
6/5 5 25 6 25
1
=−
150
P P
3.3.11.22. Solution. For any integer n, sin(πn) = 0, so sin(πn) = 0 = 0. So,
this series converges.
3.3.11.23. Solution. For any integer n, cos(πn) = ±1, so lim cos(πn) 6= 0. By the
n→∞
divergence test, this series diverges.
3.3.11.24. Solution. Factorials grow super fast. Like, wow, really fast. Even
faster than exponentials. So the terms are going to zero, and the divergence test
won’t help us. Let’s use ratio — it’s a good go-to test with factorials.
ek+1
ak+1 (k+1)! ek+1 k!
= ek
= k
·
ak k!
e (k + 1)!
k(k − 1) · · · (1) 1 e
=e· =e· =
(k + 1)(k)(k − 1) · · · (1) k+1 k+1

Since e is a constant,
ak+1 e
lim = lim =0
k→∞ ak k→∞ k + 1

Since 0 < 1, by the ratio test, the series converges.

3.3.11.25. Solution. This is close to being in the form of a geometric series. First,
we should have our powers be k, not k + 2, but we notice 3k+2 = 3k 32 = 9 · 32 , so:
∞ ∞ ∞ ∞  k
X 2k X 2k 1 X 2k 1X 2
k+2
= k
= k
=
k=0
3 k=0
9 · 3 9 k=0
3 9 k=0 3

1098
S OLUTIONS TO E XERCISES

Now it looks like a geometric series with r = 23

 
1 1 1
= =
9 1 − (2/3) 3
1
In conclusion: this (geometric) series is convergent, and its sum is .
3
3.3.11.26. Solution. Usually with factorials, we want to use the divergence test
or the ratio test. Since the terms are indeed tending towards zero, we are left with
the ratio test.
(n+1)!(n+1)!
an+1 (2n+2)! (n + 1)!(n + 1)! (2n)!
= n!n!
= ·
an (2n)!
n!n! (2n + 2)!
(n+1)(n)(n−1)···(1) (n+1)(n)(n−1)···(1) (2n)(2n−1)(2n−2)···(1)
= n(n−1)···(1)
· n(n−1)···(1)
· (2n+2)(2n+1)(2n)(2n−1)(2n−2)···(1)
1
= (n + 1)(n + 1) ·
(2n + 2)(2n + 1)

So
an+1 (n + 1)(n + 1) (n + 1)(n + 1)
lim = lim = lim
n→∞ an n→∞ (2n + 2)(2n + 1) n→∞ 2(n + 1)(2n + 1)
n+1 1
= lim =
n→∞ 4n + 2 4
Since the limit is a number less than 1, the series converges by the ratio test.
3.3.11.27. Solution. We want to make an estimation, when n gets big:

n2 + 1 n2 1
4
≈ 4
= 2
2n + n 2n 2n
P 1
Since 2n2
is a convergent series (by p-test, or integral test), we guess that our
series is convergent as well. If we wanted to use comparison test, we should have
n2 +1 1
to show 2n 4 +n < 2n2 , which seems unpleasant, so let’s use limit comparison.

n2 +1
(n2 + 1)2n2 2n4 + 2n2 1/n4
 
2n4 +n
lim 1 = lim = lim
n→∞
2n2
n→∞ 2n4 + n n→∞ 2n4 + n 1/n4
2
2+ n2
= lim 1 =1
n→∞ 2 +
n3

n1 +1
P
Since the limit is a positive finite number, by the Limit Comparison Test, 2n4 +n
P 1
does the same thing 2n2
does: it converges.
3.3.11.28. *. Solution. First, we rule out some of the easier tests. The limit of
the terms being added is zero, so the divergence test is inconclusive. The terms
P1
being added are smaller than the terms of the (divergent) harmonic series, n
,

1099
S OLUTIONS TO E XERCISES

so we can’t directly compare these two series, and there isn’t another obvious
series to compare ours to. However, the terms being added seem like a function
we could integrate.
5
Let f (x) = . Then f (x) is positive and decreases as x increases. So
x(log x)3/2
X∞ Z ∞
the sum f (n) and the integral f (x) dx either both converge or both di-
3 3
verge, by the integral test, which is Theorem 3.3.5. For the integral, we use the
substitution u = log x, du = dx
x
to get
Z ∞ Z ∞
5 dx 5 du
3/2
= 3/2
3 x(log x) log 3 u

3
which converges by the p-test (which is Example 1.12.8) with p = 2
> 1.
3.3.11.29. *. Solution. Let f (x) = x(log1 x)p . Then f (x) is positive for n ≥ 3, and
f (x) decreases as x increases. So, we can use the integral test, Theorem 3.3.5.

∞ R log R
1 1 dx 1
Z Z Z
dx = lim = lim du with ,
2 x(log x)p R→∞ 2
p
(log x) x R→∞ log 2 up

Using the results about p-series, Example 3.3.6, we know this integral converges
if and only if p > 1, so the same is true for the series by the integral test.
3.3.11.30. *. Solution. As usual, let’s see whether the “easy” tests work. The
terms we’re adding converge to zero:
√
e− n
1
lim √ = lim √ √n = 0
n→∞ n n→∞ ne
so the divergence test is inconclusive. Our series isn’t geometric, and it doesn’t
seem obvious how to compare it to a geometric series. However, the terms we’re
adding seem like√
they would make an integrable function.
e− x
Set f (x) = x . For x ≥ 1, this function is positive and decreasing (since it is
√
√
the product of the two positive decreasing functions e− x and √1x ). We use the
√
integral test with this function. Using the substitution u = x, so that du =
1
√
2 x
dx, we see that
Z ∞ Z R  − √x 
e
f (x) dx = lim √ dx
1 R→∞ 1 x
 Z √R 
−u
= lim e · 2 du
R→∞ 1
 √ 
R
−u
= lim −2e
R→∞ 1
√ √
 
− R − 1
= lim −2e + 2e = 0 + 2e−1 ,
R→∞

1100
S OLUTIONS TO E XERCISES

and so this improper integral converges. By the integral test, the given series also
converges.
3.3.11.31. *. Solution. We first develop some intuition. For very large n, 3n2
dominates 7 so that
√ √ √
3n2 − 7 3n2 3
≈ =
n3 n3 n2
∞
X 1
The series converges by the p-test with p = 2, so we expect the given series
n=2
n2
to converge too.
To verify that our intuition is correct, it suffices to observe that
√ √ √
3n2 − 7 3n2 3
0 < an = 3
< 3
= 2 = cn
n n n
∞
P ∞
P
for all n ≥ 2. As the series cn converges, the comparison test says that an
n=2 n=2
converges too.
3.3.11.32. *. Solution. We first develop
√ some
√ intuition. For very large k, k 4
3 4 3 4 4/3 5
dominates 1 so that the √ k + 1 ≈ k = k , and k dominates 9 so
√ numerator
5 5 5/2
that the denominator k + 9 ≈ k = k and the summand
√3
k4 + 1 k 4/3 1
√ ≈ 5/2 = 7/6
5
k +9 k k
∞
X 1
The series 7/6
converges by the p-test with p = 67 > 1, so we expect the given
n=1
k
series to converge too.
To verify that our intuition is correct, we apply the limit comparison test with
√
3
k4 + 1 1 k 4/3
ak = √ and bk = 7/6 = 5/2
k5 + 9 k k
which is valid since
√ p
ak 3
k 4 + 1/k 4/3 3
1 + 1/k 4
lim = lim √ = lim p =1
k→∞ bk k→∞ k 5 + 9/k 5/2 k→∞ 1 + 9/k 5
∞
7
P
exists. Since the series bk is a convergent p-series (with ratio p = 6
> 1), the
k=1
given series converges.
Note: to apply the direct comparison test with our chosen comparison series, we
would need to show that √
3
k4 + 1 1
√ ≤ 7/6
5
k +9 k
for all k sufficiently large. However, this is not true: the opposite inequality holds
when k is large.

1101
S OLUTIONS TO E XERCISES

3.3.11.33. *. Solution.
• Solution 1: Let’s see whether the divergence test works here.
 1 
n4 2n/3 n4 2n/3 2n/3
lim 1 = lim = lim =∞
n→∞ (2n + 7)4 4
n→∞ (2 + 7/n)4 n→∞ (2 + 0)4
n

The summands of our series do not converge to zero. By the divergence

test, the series diverges.

• Solution 2: Let’s develop some intuition for a comparison. For very large
n, 2n dominates 7 so that
n4 2n/3 n4 2n/3 1
4
≈ 4
= 2n/3
(2n + 7) (2n) 16
∞
X
The series 2n/3 is a geometric series with ratio r = 21/3 > 1 and so
n=1
diverges. (It also fails the divergence test.) We expect the given series to
diverge too.
To verify that our intuition is correct, we apply the limit comparison test
with
n4 2n/3
an = and bn = 2n/3
(2n + 7)4
which is valid since
an n4 1 1
lim = lim 4
= lim 4 = 4
n→∞ bn n→∞ (2n + 7) n→∞ (2 + 7/n) 2
∞
P
exists and is nonzero. Since the series bn is a divergent geometric series
n=1
(with ratio r = 21/3 > 1), the given series diverges.
(It is possible to use the plain comparison test as well. One needs to show
n4 2n/3 n4 2n/3 1
something like an = (2n+7) 4 ≥ (2n+7n)4 = 94 bn .)

• Solution 3: Alternately, one can apply the ratio test:

an+1 (n + 1)4 2(n+1)/3 /(2(n + 1) + 7)4

lim = lim
n→∞ an n→∞ n4 2n/3 /(2n + 7)4
(n + 1)4 (2n + 7)4 2(n+1)/3
= lim
n→∞ n4 (2n + 9)4 2n/3
(1 + 1/n)4 (2 + 7/n)4 1/3
= lim · 2 = 1 · 21/3 > 1.
n→∞ (2 + 9/n)4
Since the ratio of consecutive terms is greater than one, by the ratio test, the
series diverges.

1102
S OLUTIONS TO E XERCISES
√ √
3.3.11.34. *. Solution. (a) For large n, n2  1 and so n2 + 1 ≈ n2 = n. This
suggests that we apply the limit comparison test with an = √n12 +1 and bn = n1 .
Since √
an 1/ n2 + 1 1
lim = lim = lim p =1
n→∞ bn n→∞ 1/n n→∞ 1 + 1/n2
∞
1
P
and since n
diverges, the given series diverges.
n=1
(b) Since cos(nπ) = (−1)n , the given series converges by the alternating series
test. To check that an = 2nn decreases to 0 as n tends to infinity, note that

an+1 (n + 1)2−(n+1)  1 1
= = 1 +
an n2−n n 2
3
is smaller than 1 (so that an+1 ≤ an ) for all n ≥ 1, and is smaller than 4
(so
an+1 ≤ 43 an ) for all n ≥ 2.
3.3.11.35. *. Solution. For large k, k 4  2k 3 − 2 and k 5  k 2 + k so

k 4 − 2k 3 + 2 k4 1
≈ = .
k5 + k2 + k k5 k
k4 −2k3 +2
This suggests that we apply the limit comparison test with ak = k5 +k2 +k
and
bk = k1 . Since

ak k 4 − 2k 3 + 2 k k 5 − 2k 4 + k 2
lim = lim 5 · = lim
k→∞ bk k→∞ k + k 2 + k 1 k→∞ k 5 + k 2 + k
3
1 − 2/k + 1/k
= lim
k→∞ 1 + 1/k 3 + 1/k 4

=1
∞
1
P
and since k
diverges (by the p-test with p = 1), the given series diverges.
k=1

3.3.11.36. *. Solution. (a) For large n, n2  n + 1 and so the numerator n2 + n +

1 ≈ n2 . For large n, n5  n and so the denominator n5 − n ≈ n5 . So, for large n,
n2 + n + 1 n2 1
5
≈ 5
= 3.
n −n n n
n2 +n+1
This suggests that we apply the limit comparison test with an = n5 −n
and bn =
1
n3
. Since

an (n2 + n + 1)/(n5 − n) n5 + n4 + n3
lim = lim = lim
n→∞ bn n→∞ 1/n3 n→∞ n5 − n
1 + 1/n + 1/n2
= lim
n→∞ 1 − 1/n4
=1

1103
S OLUTIONS TO E XERCISES

∞
1
P
exists and is nonzero, and since n3
converges (by the p-test with p = 3 > 1),
n=1
the given series converges.√
(b) For large m, 3m  | sin m| and so
√
3m + sin m 3m 3
≈ = .
m2 m2 m
√
3m+sin m
This suggests that we apply the limit comparison test with am = m2
and
bm = m1 . (We could also use bm = m3 .) Since
√ √
am (3m + sin m)/m2 3m + sin m
lim = lim = lim
m→∞ bm m→∞ 1/m m→∞ m
√
sin m
= lim 3 +
m→∞ m
=3
∞
1
P
exists and is nonzero, and since m
diverges (by the p-test with p = 1), the
m=1
given series diverges.
3.3.11.37. Solution.
∞ ∞  n
X 1 X 1
=
n=5
en n=5
e
∞  n 4  n
X 1 X 1
= −
n=0
e n=0
e

5
1 1 − 1e
= −
1 − 1e 1 − 1e
1 5


1
= e 1 = 5
e 1 − 1e


1− e
1
= 5
e − e4
3.3.11.38. *. Solution. This is a geometric series.
∞ ∞ ∞
X 6 X 6 X 6 1
n
= n+2
= ·
2 7n
n=2
7 n=0
7 n=0
7

6
We use Lemma 3.2.5 with a = 72
and r = 17 .

6 1 6 1
= 2
· 1 = =
7 1− 7
42 7

1104
S OLUTIONS TO E XERCISES

3.3.11.39. *. Solution. (a)

• Solution 1: The given series is

∞
1 1 1 1 X
1+ + + + + ··· = an with
3 5 7 9 n=1

First we’ll develop some intuition by observing that, for very large n, an ≈
∞
1 1
P
2n
. We know that the series n
diverges by the p-test with p = 1. So let’s
n=1
apply the limit comparison test with bn = n1 . Since

an n 1 1
lim = lim = lim 1 =
n→∞ bn n→∞ 2n − 1 n→∞ 2 − 2
n

∞
P ∞
P
the series an converges if and only if the series bn converges. So the
n=1 n=1
given series diverges.

• Solution 2: The series

1 1 1 1 1 1 1 1 1
1+ + + + + ··· ≥ + + + + + ···
3 5 7 9 2 4 6 8 10
1 1 1 1 1 
= 1 + + + + + ···
2 2 3 4 5
The series in the brackets is the harmonic series which we know diverges,
by the p-test with p = 1. So the series on the right hand side diverges. By
the direct comparison test, the series on the left hand side diverges too.
(2n + 1)
(b) We’ll use the ratio test with an = . Since
22n+1
an+1 (2n + 3) 22n+1 1 (2n + 3) 1 (2 + 3/n)
= 2n+3
= =
an 2 (2n + 1) 4 (2n + 1) 4 (2 + 1/n)
1
→ < 1 as n → ∞
4
the series converges.
3.3.11.40. *. Solution. (a) For very large k, k  k 2 so that
√
3
√3
k k 1
an = 2 ≈ 2 = 5/3 .
k −k k k
1
We apply the limit comparison test with bk = k5/3
. Since
√
ak 3
k/(k 2 − k) k2 1
lim = lim 5/3
= lim 2
= lim =1
k→∞ bk k→∞ 1/k k→∞ k − k k→∞ 1 − 1/k

1105
S OLUTIONS TO E XERCISES

∞
1 5
P
exists and is nonzero, and k5/3
converges (by the p-test with p = 3
> 1), the
k=1
given series converges by the limit comparison test.
10 k (k!)2
(b) The k th term in this series is ak = k 10
(2k)!
. Factorials often work well with
the ratio test, because they simplify so nicely in quotients.

ak+1 (k + 1)10 10k+1 ((k + 1)!)2 (2k)!

= · 10 k
ak (2k + 2)! k 10 (k!)2
 k + 1 10 (k + 1)2
= 10
k (2k + 2)(2k + 1)
 1 10
 (1 + 1/k)2
= 10 1 +
k (2 + 2/k)(2 + 1/k)
1
As k tends to ∞, this converges to 10 × 1 × 2×2 > 1. So the series diverges by the
ratio test.
1
(c) We’ll use the integal test. The k th term in the series is ak = k(log k)(log log k)
= f (k)
1
with f (x) = x(log x)(log log x) , which is continuous, positive and decreasing for x ≥ 3.

∞ ∞ Z R
dx dx
Z Z
f (x) dx = = lim
3 3 x(log x)(log log x) R→∞ 3 x(log x)(log log x)
Z log R
dy dx
= lim with y = log x, dy =
R→∞ log 3 y log y x
Z log log R
dt dy
= lim with t = log y, dt =
R→∞ log log 3 t y
h ilog log R
= lim log t =∞
R→∞ log log 3

Since the integral is divergent, the series is divergent as well by the integral test.
3.3.11.41. *. Solution. For large n, the numerator n3 − 4 ≈ n3 and the denom-
n3 1
inator 2n5 − 6n ≈ 2n5 , so the nth term is approximately 2n 5 = 2n2 . So we apply
3 −4
the limit comparison test with an = 2nn5 −6n and bn = n12 . Since
4
an (n3 − 4)/(2n5 − 6n) 1− n3 1
lim = lim 2
= lim 6 =
n→∞ bn n→∞ 1/n n→∞ 2 − 2
n4

∞
P
exists and is nonzero, the given series an converges if and only if the series
n=1
∞ ∞ ∞
1
P P P
bn converges. Since the series bn = n2
is a convergent p-series (with
n=1 n=1 n=1
p = 2), both series converge.
3.3.11.42. *. Solution. By the alternating series test, the error introduced when

1106
S OLUTIONS TO E XERCISES

∞ N
X (−1)n X (−1)n
we approximate the series by is at most the magnitude
n=1
n · 10n n=1
n · 10n
1
of the first omitted term, . By trial and error, we find that this
(N + 1)10(N +1)
expression becomes smaller than 10−6 when N + 1 ≥ 6. So the smallest allowable
value is N = 5.
3.3.11.43. *. Solution. The sequence { n12 } decreases to zero as n increases to
infinity. So, by the alternating series error bound, which is given in Theorem
2 (−1)N
3.3.14, π12 − SN lies between zero and the first omitted term, (N +1)2
. We therefore
1 −6 3
need (N +1)2 ≤ 10 , which is equivalent to N + 1 ≥ 10 and N ≥ 999.

3.3.11.44. *. Solution. The error introduced when we approximate S by the N th

(−1)n+1
partial sum SN = N
P
n=1 (2n+1)2 lies between 0 and the first term dropped, which
(−1)n+1 (−1)N +2
is (2n+1)2
= (2N +3)2
. So we need the smallest positive integer N obeying
n=N +1

1 1
2
≤
(2N + 3) 100
2
(2N + 3) ≥ 100
2N + 3 ≥ 10
7
N≥
2
So we need N = 4 and then
1 1 1 1
S4 = 2
− 2+ 2− 2
3 5 7 9

Exercises — Stage 3
3.3.11.45. *. Solution. (a) There are plenty of powers/factorials. So let’s try the
n
ratio test with an = 9nn n! .

an+1 (n + 1)n+1 9n n! (n + 1)n+1

lim = lim n+1 = lim
n→∞ an n→∞ 9 (n + 1)! nn n→∞ nn 9 (n + 1)

(1 + 1/n)n e
= lim =
n→∞ 9 9
Here we have used that lim (1 + 1/n)n = e. See Example 3.7.20 in the CLP-1 text,
n→∞
with x = n1 and a = 1. As e < P9, our1 series converges.
(b) We know that the series ∞ n=1 n2 converges, by the p-test with p = 2, and also
2 1
that log n ≥ 2 for all n ≥ e . So let’s use the limit comparison test with an = nlog n
1
and bn = n2 .
an 1 n2 1
lim = lim log n · = lim log n−2 = 0
n→∞ bn n→∞ n 1 n→∞ n

So our series converges, by the limit comparison test.

1107
S OLUTIONS TO E XERCISES

3.3.11.46. *. Solution. (a)

• Solution 1:
◦ Our first task is to identify the potential sources of impropriety for this
integral.
◦ The domain of integration extends to +∞. On the domain of inte-
gration the denominator is never zero so the integrand is continuous.
Thus the only problem is at +∞.
◦ Our second task is to develop some intuition about the behavior of the
integrand for very large x. When x is very large:
 | sin x| ≤ 1  x, so that the numerator x + sin x ≈ x, and
 1  x2 , so that denominator 1 + x2 ≈ x2 , and
x + sin x x 1
 the integrand ≈ =
1 + x2 x2 x
Z ∞ Z ∞
dx x + sin x
◦ Now, since diverges, we would expect dx to di-
2 x 2 1 + x2
verge too.
◦ Our final task is to verify that our intuition is correct. To do so, we set
x + sin x 1
f (x) = g(x) =
1 + x2 x
and compute
f (x) x + sin x 1
lim = lim ÷
x→∞ g(x) x→∞ 1 + x2 x
(1 + sin x/x)x
= lim ×x
x→∞ (1/x2 + 1)x2

1 + sin x/x
= lim
x→∞ 1/x2 + 1

=1
Z ∞ Z ∞
dx
◦ Since g(x) dx = diverges, by Example 1.12.8 a , with p = 1,
2 2 x Z ∞ Z ∞
x + sin x
Theorem 1.12.22(b) now tells us that f (x) dx = dx
2 2 1 + x2
diverges too.
x + sin x x sin x
• Solution 2: Let’s break up the integrand as 2
= 2
+ .
Z ∞ 1+x 1+x 1 + x2
sin x
First, we consider the integral dx.
2 1 + x2
| sin x| 1 1
Z
◦ 2
≤ 2
, so if we can show dx converges, we can
1+x 1 +Zx 1 + x2
| sin x|
conclude that dx converges as well by the comparison test.
1 + x2

1108
S OLUTIONS TO E XERCISES
∞ Z ∞
1 1
Z
◦ 2
dx ≤ dx
2 1+x 2 x2
Z ∞
1
◦ dx converges (by the p-test with p = 2)
2 x2
Z ∞
sin x
◦ So the integral dx converges by the comparison test, and
2 1 + x2
hence
Z ∞
sin x
◦ dx converges as well.
2 1 + x2
Z ∞ Z ∞
x + sin x x
Therefore, 2
dx converges if and only if dx con-
2 1+x 2 1 + x2
verges. But
Z ∞ Z r
x x h1
2
ir
dx = lim dx = lim log(1+x ) =∞
2 1+x2 r→∞ 2 1+x2 r→∞ 2 2
Z ∞
x + sin x
diverges, so dx diverges.
2 1 + x2
x + sin x
(b) The problem is that f (x) = is not a decreasing function. To see this,
1 + x2
compute the derivative:

(1 + cos x)(1 + x2 ) − (x + sin x)(2x)

f 0 (x) =
(1 + x2 )2
(cos x − 1)x2 − 2x sin x + 1 + cos x
=
(1 + x2 )2

If x = 2mπ, the numerator is 0 − 0 + 1 + 1 > 0.

Therefore, the integral test does not apply.
(c)
• Solution 1: Set an = n+sin
1+n2
n
. We first try to develop some intuition about
the behaviour of an for large n and then we confirm that our intuition was
correct.

◦ Step 1: Develop intuition. When n  1, the numerator n + sin n ≈ n,

and the denominator 1 + n2 ≈ n2 so that an ≈ nn2 = n1 and it looks like
our series should diverge by the p-test (Example 3.3.6) with p = 1.
1
◦ Step 2: Verify intuition. To confirm our intuition we set bn = n
and
compute the limit
n+sin n
an 1+n2 n[n + sin n]
lim = lim 1 = lim
n→∞ bn n→∞
n
n→∞ 1 + n2
1 + sin n
= lim 1 n = 1
n→∞
n2
+1

1109
S OLUTIONS TO E XERCISES

∞ ∞
1
P P
We already know that the series bn = n
diverges by the p-test
n=1 n=1
with p = 1. So our series diverges by the limit comparison test, Theo-
rem 3.3.11.
∞
sin n 1 1
P
• Solution 2: Since 1+n2
≤ n2
and the series n2
converges by the p-test
n=1
∞ ∞
sin n n+sin n
P P
with p = 2, the series 1+n2
converges. Hence 1+n2
converges if and
n=1 n=1
∞
n x
P
only if the series 1+n2
converges. Now f (x) = 1+x2
is a continuous,
n=1
positive, decreasing function on [1, ∞) since

(1 + x2 ) − x(2x) 1 − x2
f 0 (x) = =
(1 + x2 )2 (1 + x2 )2
R∞ x
is negative for all x > 1. We saw in part (a) that the integral 2 1+x2 dx
R∞ x ∞
n
P
diverges. So the integral 1 1+x2 dx diverges too and the sum 1+n2
di-
n=1
∞
n+sin n
P
verges by the integral test. So 1+n2
diverges.
n=1

a To change the lower limit of integration from 1 to 2, just apply Theorem 1.12.20.
√
e− x
1
3.3.11.47. *. Solution. Note that √ = √ √x decreases as x increases.
x xe
Hence, for every n ≥ 1,
√ √
e− x
e n
√ ≥ √ for x in the interval [n − 1, n]
x n
√ √
n n
e− x e− n
Z Z
So, √ dx ≥ √ dx
n−1 x n−1 n
 −√n x=n
e
= √ x
n x=n−1
√
e− n
= √
n
Then, for every N ≥ 1,
∞ √ ∞ Z n −√x
X e− n X e
EN = √ ≤ √ dx
n=N +1
n n=N +1 n−1 x
Z N +1 − x√ √
Z N +2 − x
e e
= √ dx + √ dx + · · ·
N x N +1 x
Z ∞ −√x
e
= √ dx
N x

1110
S OLUTIONS TO E XERCISES
√ 1√
dx
Substituting y = x, dy = 2 x
,
√
∞ ∞
e− x √
Z Z ∞
√ dx = 2 √
e−y dy = −2e−y √ = 2e− N
N x N N

√ √
e− n
P∞
This shows that n=N +1
√
n
converges and is between 0 and 2e− N
. Since E14 =
√
2e− 14
= 0.047, we may truncate the series at n = 14.
∞ √ 14 √
X e− n X e− n
√ = √ + E14
n=1
n n=1
n
= 0.3679 + 0.1719 + 0.1021 + 0.0677 + 0.0478
+ 0.0352 + 0.0268 + 0.0209 + 0.0166 + 0.0134
+ 0.0109 + 0.0090 + 0.0075 + 0.0063 + E14
= 0.9042 + E14

The sum is between 0.9035 and 0.9535. (This even allows for a roundoff error of
0.00005 in each term as we were calculating the partial sum.)
3.3.11.48. *. Solution. Let’s get some intuition to guide us through a proof.
∞
P
Since an , converges an must converge to zero as n → ∞. So, when n is quite
n=1
an an
= a1n , and we know
P
large, 1−a n
≈ 1−0 an converges. So, we want to separate
the “large” indices from a finite number of smaller ones.
Since lim an = 0, there must be a some integer N such that 21 > an ≥ 0 for all
n→∞
n > N . Then, for n > N ,
an an
≤ = 2an
1 − an 1 − 1/2
From the information in the problem statement, we know
∞
X ∞
X
2an = 2 an converges.
n=N +1 n=N +1

So, by the direct comparison test,

∞
X an
converges as well.
n=N +1
1 − an

Since the convergence of a series is not affected by its first N terms, as long as N
is finite, we conclude
∞
X an
converges.
n=1
1 − an

a We could have chosen any positive number strictly less than 1, not only 12 .

1111
S OLUTIONS TO E XERCISES

∞
P
3.3.11.49. *. Solution. By the divergence test, the fact that (1 − an ) converges
n=0
guarantees that lim (1 − an ) = 0, or equivalently, that lim an = 1. So, by the
n→∞ n→∞
divergence test, a second time, the fact that

lim 2n an = +∞
n→∞

∞
2n an diverges too.
P
guarantees that
n=0

∞
X nan − 2n + 1
3.3.11.50. *. Solution. By the divergence test, the fact that
n=1
n+1
nan − 2n + 1
converges guarantees that lim = 0, or equivalently, that
n→∞ n+1
n 2n − 1
0 = lim an − lim = lim an − 2 ⇐⇒ lim an = 2
n→∞ n + 1 n→∞ n + 1 n→∞ n→∞

∞ 
P 
The series of interest can be written − log a1 + log(an ) − log(an+1 ) which
n=1
looks like a telescoping series. So we’ll compute the partial sum
N
X  
SN = − log a1 + log(an ) − log(an+1 )
n=1   
= − log a1 + log(a1 ) − log(a2 ) + log(a2 ) − log(a3 ) + · · ·
 
+ log(aN ) − log(aN +1 )
= − log(aN +1 )

and then take the limit N → ∞

∞
X  
− log a1 + log(an ) − log(an+1 ) = lim SN = − lim log(aN +1 )
N →∞ N →∞
n=1
1
= − log 2 = log
2
3.3.11.51. *. Solution. We are told that ∞
P
n=1 an converges. Thus we must have
that lim an = 0. In particular, there is an index N such that 0 ≤ an ≤ 1 for all
n→∞
n ≥ N . Then:

0 ≤ a2n ≤ an for

By the direct comparison test,

∞
X
a2n converges.
n=N +1

1112
S OLUTIONS TO E XERCISES

Since convergence doesn’t depend on the first N terms of a series for any finite
N,
∞
X
a2n converges as well.
n=1

3.3.11.52. Solution. The most-commonly used word makes up α percent of all

the words. So, we want to find α.
If we add together the frequencies of all the words, they should amount to 100%.
That is,
20,000
X α
= 100
n=1
n
We can approximate the sum (with α left as a parameter) using the ideas behind
the integral test. (See Example 3.3.4.)

α
f (x) = x

x
1 2 3 4 5
N
X α
As we see in the diagram above, (which is the sum of the areas of the
n=1
n
N +1
α
Z
rectangles) is greater than dx (the area under the curve). That is,
1 x

N +1 N
α Xα
Z
dx < .
1 x n=1
n

Using the fact that our language’s 20,000 words make up 100% of the words used,
we can find a lower bound for α.
20,000
X α Z 20,001 α h i20,001
100 = > dx = α log(x) = α log(20, 001)
n=1
n 1 x 1

100
α<
log(20, 001)

We can find an upper bound for α in a similar manner.

1113
S OLUTIONS TO E XERCISES

α
f (x) = x

x
1 2 3 4 5
N
X α
From the diagram, we see (which is the sum of the areas of the rectan-
n=2
n
N
α
Z
gles, excluding the first) is less than dx. (The reason for excluding the first
1 x
rectangle is to avoid comparing our series to an integral that diverges.) That is,
N N
α α
X Z
< dx .
n=2
n 1 x

Therefore,
20,000 20,000
X α X α
100 = =α+
n=1
n n=2
n
Z 20,00
α  
<α+ dx = α + α log(20, 000) = α 1 + log(20, 000)
1 x
100
α>
1 + log(20, 000)

Using a calculator, we see

9.17 < α < 10.01
So, the most-commonly used word makes up about 9-10 percent of the total
words.
3.3.11.53. Solution. Generalizing our work in Question 52, we find the approx-
imations:
b+1 b b
1 X1 1
Z Z
dx < < dx
a x n=a
n a−1 x
b+1 b
1 1
R P
when a ≥ 2. The inequality x
dx < n
can be read off of the sketch
a n=a

1114
S OLUTIONS TO E XERCISES

1
f (x) = x
x
a−1 a b b+1

b
P 1
Rb 1
and the inequality n
< x
dx can be read off of the sketch
n=a a−1

1
f (x) = x
x
a−1 a b b+1

We will evaluate the total population by writing

2×106 a−1 2×10 6
X 2 × 106 X 2 × 106 X 2 × 106
= +
n=1
n n=1
n n=a
n

and applying the above integral approximations to the second sum. We want our
error to be less than one million, so we need to choose a value of a such that:

2×106 2×106 +1
1 1
Z Z
6
2 × 10 dx − 2 × 106 dx < 106
a−1 x a x
| {z } | {z }
upper bound lower bound
2×106 2×106 +1
1 1
1
Z Z
dx < dx −
a−1 a x x
2
1
log 2 × 106 − log(a − 1) − log 2 × 106 + 1 − log(a) <

 

2
 6

6

 1
log 2 × 10 − log 2 × 10 + 1 + [log(a) − log(a − 1)] <
 2
2 × 106
  
a 1
log 6
+ log <
2 × 10 + 1 a−1 2
The first term is extremely close to 0, so we ignore it.
 
a 1
log <
a−1 2
a √
< e1/2 = e
a−1

1115
S OLUTIONS TO E XERCISES
√ √
a<a e− e
√ √
e < a( e − 1)
√
e
√ <a
e−1
√
e
Since √ ≈ 2.5, we use a = 3. That is, we will approximate the value of
e−1
2×106
X 1
using an integral. Then, we will use that approximation to estimate our
n=3
n
total population.

2×10 6
2×106 +1 X 1 Z 2×106 1
1
Z
dx < < dx
3 x n=3
n 2 x
2×106

6
X 1
< log 2 × 106 − log(2)



log 2 × 10 + 1 − log(3) <
n=3
n
2×106
1 X 1 1
+ log 2 × 106 + 1 − log(3) < < 1 + + log 2 × 106



1+
2 n=1
n 2
− log(2)
2×106
2 × 106 + 1
 
3 X 1 3
+ log < < + 6 log(10)
2 3 n=1
n 2
P2×106 2×106
So the population, namely n=1 n
, is between
3

2 × 106 + log 2
3
× 106 + 1
3
= 29, 820, 091
2
and 3 
6
2 × 10 + 6 log (10) = 30, 631, 021
2

3.4 · Absolute and Conditional Convergence

3.4.3 · Exercises

Exercises — Stage 1
∞
3.4.3.1. *. Solution. False. For example if bn = n1 , then (−1)n+1 bn =
P
n=1
∞ ∞
(−1)n+1 n1 converges by the alternating series test, but 1
P P
n
diverges by the
n=1 n=1
p–test.
Remark: if we had added that {bn } is a sequence of alternating terms, then by

1116
S OLUTIONS TO E XERCISES

∞
X
Theorem 3.4.2, the statement would have been true. This is because (−1)n+1 bn
n=1
∞
X ∞
X
would either be equal to |bn | or − |bn |.
n=1 n=1
P
3.4.3.2. Solution. Absolute convergence describes the situation where P|an |
converges (see Definition 3.4.1). By Theorem 3.4.2, this guarantees that also an
converges. P P
Conditional convergence describes the situation where |an | diverges but an
converges
P (see again Definition 3.4.1). P
If an diverges, we justP say it diverges. The reason is that if an diverges,
we automatically know |an | diverges as well, so there’s no need for a special
name.
P P
an converges an diverges
P
|a | converges converges absolutely not possible
P n
|an | diverges converges conditionally diverges

Exercises — Stage 2
∞
P (−1)n
3.4.3.3. *. Solution. The series 9n+5
converges by the alternating series test.
n=1
∞
P (−1)n P∞ 1
On the other hand the series 9n+5
= n=1 9n+5 diverges by the limit com-
n=1
parison test with bn = n1 . So the given series is conditionally convergent.
3.4.3.4. *. Solution. Note that (−1)2n+1 = (−1)·(−1)2n = −1. So we can simplify
∞ ∞
X (−1)2n+1 X 1
=−
n=1
1+n n=1
1+n

∞
1 1 1 X 1
Since ≥ = , diverges by the comparison test with the
1+n n+n 2n n=1 1 + n
∞
1
P
divergent harmonic series n
. The extra overall factor of −1 in the original
n=1
series does not change the conclusion of divergence.
3.4.3.5. *. Solution. Since
1 + 4n 1 + 4n
lim = lim =1
n→∞ 3 + 22n n→∞ 3 + 4n

1 + 4n
the alternating series test cannot be used. Indeed, lim (−1)n−1 does not
n
n→∞ 3 + 22n
1+4
exist (for very large n, (−1)n−1 3+22n alternates between a number close to +1 and

a number close to −1) so the divergence test says that the series diverges. (Note
that “none of the above” cannot possibly be the correct answer — every series

1117
S OLUTIONS TO E XERCISES

either converges absolutely, converges conditionally, or diverges.)

3.4.3.6. *. Solution. First, we’ll develop some intuition. For very large n
√ √
n cos(n) n cos(n) cos(n) 1
2
≈ 2
= 3/2
≤ 3/2
n −1 n n n
∞
X 1
since |cos(n)| ≤ 1 for all n. By the p–test, the series converges for all p > 1.
n=5
np
So we would expect the given series to converge absolutely.
√
n cos(n)
Now, to confirm that our intuition is correct, we’ll first try setting an = n2 −1
1
and bn = n3/2
.
√
n cos(n) √ √ 3
an n2 −1 n · n | cos n|
lim = lim 1 = lim
n→∞ bn n→∞
n3/2
n→∞ n2 − 1
n2 | cos n| n2
 
= lim = lim | cos n|
n→∞ n2 − 1 n→∞ n2 − 1
= lim 1 · | cos n|
n→∞

Unfortunately, this limit doesn’t exist, so we can’t use the limit comparison theo-
rem. We’ll need a slightly different tactic.

• Solution 1: Let’s try the limit comparison test with a different p-series. The
question is, which one. Let’s start by leaving p as a variable, then see what
happens.
√
We apply the limit comparison test with an = nn2cos(n) −1
and bn = 1
np
. We’ll
choose a specific p shortly.
√
an n cos n/(n2 − 1)
lim = lim
n→∞ bn n→∞ 1/np
np+0.5 | cos n| | cos n|
= lim 2 2
= lim (3/2−p)
n→∞ n (1 − 1/n ) n→∞ n (1 − 1/n2 )
= 0 if
∞
3
P
the limit comparison test says that if p < 2
and the series bn converges
n=5
∞ √
P n cos(n)
(which is the case if p > 1) then the series n2 −1
also converges. So
n=5
choosing any 1 < p < 23 , for example p = 54 , we conclude that the given
series converges absolutely.
• Solution 2: Let’s try to use the direct comparison test. When we were trying
to develop intuition, we noticed the following:
√ √
n cos(n) n cos(n) cos(n) 1
2
≈ 2
= 3/2
≤ 3/2
n −1 n n n

1118
S OLUTIONS TO E XERCISES

1
It’s not the case that our terms are less than n3/2 , but perhaps they would be
2
less than, say n3/2 . Let’s trace our reasoning above backwards.
√ √
2 2 cos n n cos n n cos n
≥ = ≥
n3/2 n 3/2 (1/2)n 2 n2 − 1

where the final inequality holds for all n ≥ 2.

∞
2
P
Since n3/2
converges by the p-test, our series converges as well by the
n=1
direct comparison test.
∞
X n2 − sin n
3.4.3.7. *. Solution. We first develop some intuition about 6 + n2
,
n=1
n
where we take the absolute value of the summands to consider whether the se-
ries converges absolutely. For very large n, n2 dominates sin n and n6 dominates
n2 so that
n2 − sin n n2 1
≈ =
n6 + n2 n6 n4
∞
X 1
The series 4
converges by the p-test with p = 4 > 1. We expect the given
n=1
n
series to converge too.
To verify that our intuition is correct, we apply the limit comparison test with

n2 − sin n 1
an = and bn =
n6 + n2 n4
which is valid since
an (n2 − sin n) n4 |n6 − n4 sin n|
lim = lim · = lim
n→∞ bn n→∞ n6 + n2 1 n→∞ n6 + n2
1 − n−2 sin n
= lim =1
n→∞ 1 + n−4
∞
P P∞ |n2 − sin n|
exists and is nonzero. Since the series bn converges, the series
n=1 n=1 n6 + n2
P∞ n2 − sin n
converges too. Therefore, the series 6 2
converges absolutely.
n=1 n + n

3.4.3.8. *. Solution. You might think that this series converges by the alter-
nating series test. But you would be wrong. The problem is that {an } does not
converge to zero as n → ∞, so that the series actually diverges by the divergence
test. To verify that the nth term does not converge to zero as n → ∞ let’s write
(2n)! th
an = (n2 +1)(n!) 2 (i.e. an is the n term without the sign) and check to see whether

1119
S OLUTIONS TO E XERCISES

an+1 is bigger than or smaller than an .

an+1 (2n + 2)! (n2 + 1)(n!)2

=
an ((n + 1)2 + 1)((n + 1)!)2 (2n)!
2
(2n + 2)(2n + 1) n + 1
=
(n + 1)2 (n + 1)2 + 1
2(2n + 1) 1 + 1/n2
=
(n + 1) (1 + 1/n)2 + 1/n2
1 + 1/2n 1 + 1/n2
=4
1 + 1/n (1 + 1/n)2 + 1/n2

So
an+1
=4lim
n→∞ an

and, in particular, for large n, an+1 > an . Thus, for large n, an increases with n
and so cannot converge to 0. So the series diverges by the divergence test.
3.4.3.9. *. Solution. This series converges by the alternating series test.
We want to know whether it converges absolutely, so we consider the
∞ ∞
X (−1)n X 1
seris 101
= 101
.
n=2
n(log n) n=2
n(log n)
We’ve seen similar function before (e.g. Example 3.3.7, with p = 101 > 1) and
it yields nicely to the integral test. Let f (x) = x(log1x)101 . Note f (x) is positive
and decreasing for n ≥ 3. Then by the integral test, the series ∞ 1
P
R∞ n=2 n(log n)101
converges if and only if the integral 2 x(log1x)101 dx does. We evaluate the integral
using the substitution u = log x, du = x1 dx.
∞ b
1 1
Z Z
dx = lim dx
2 x(log x)101 b→∞ 2 x(log x)
101
Z log b
1
= lim du
b→∞ log 2 u101
 log b
−1
= lim
b→∞ 100u100
log 2

1
=
100(log 2)100
∞
1
P
Since the integral converges, the series n(log n)101
converges, and therefore the
n=2
∞
P (−1)n
series n(log n)101
converges absolutely.
n=2

3.4.3.10. Solution. The sequence has some positive terms and some negative
terms, which limits the tests we can use. However, if we consider the series

1120
S OLUTIONS TO E XERCISES

∞
X sin n
, we can use the direct comparison test.
n=1
n2
∞
sin n 1 X 1
For every n, | sin n| < 1, so 0 ≤ < . Since converges, then by the
n2 n2 n=1
n2
∞ ∞
X sin n X sin n
direct comparison test, 2
converges as well. Then converges
n=1
n n=1
n2
absolutely — in particular, it converges.
3.4.3.11. Solution. The terms of this series are sometimes negative (for odd
values of n where sin n < 12 ) and sometimes positive. But, they are not strictly
alternating, so we can’t use the alternating series test. Instead, we use a direct
comparison test to show the series converges absolutely.

1 sin n 1
− ≤ ≤
 4 4
 4   
1 1 sin x 1 1 1
⇒ − − ≤ − < −
4 8 4 8 4 8
 
3 sin x 1 1
⇒ − ≤ − <
8 4 8 8
sin x 1 3
⇒ 0≤ − <
4 8 8
 n  n
sin x 1 3
⇒ 0≤ − <
4 8 8
∞  n
X 3
Since converges (it’s a geometric sum with |r| < 1), by the direct com-
n=1
8
∞  n
X sin x 1
parison test, − converges as well.
n=1
4 8
∞  n
X sin x 1
Then − converges absolutely — and so it converges.
n=1
4 8

3.4.3.12. Solution. The terms of this series are sometimes negative and some-
times positive. But, they are not strictly alternating, so we can’t use the alter-
nating series test. Instead, we use a direct comparison test to show the series
converges absolutely.

0 ≤ sin2 n ≤ 1
0 ≤ cos2 n ≤ 1
So, − 1 ≤ sin2 n − cos2 n ≤ 1
     
1 1 2 2 1 1 3
− = −1 + ≤ sin n − cos n + ≤ 1+ =
2 2 2 2 2

1121
S OLUTIONS TO E XERCISES

1 1 sin2 n − cos2 n + 21 3 1
− · n ≤ n
≤ · n
2 2 2 2 2
sin2 n − cos2 n + 21 3
0≤ n
≤ n+1
2 2
∞
X 3
The series n+1
converges, because it’s a geometric series with r = 12 . By
n=1
2
∞
X sin2 n − cos2 n + 21
the direct comparison test, n
converges as well. Then
n=1
2
∞
X sin2 n − cos2 n + 12
n
converges absolutely, so it converges.
n=1
2

Exercises — Stage 3
3.4.3.13. *. Solution. (a)
∞
3
24n2 e−n converges. If we replace n
P
• Solution 1: We need to show that
n=1
3
by x in the summand, we get f (x) = 24x2 e−x , which we can integate. (Just
substitute u = x3 .) So let’s try the integral test. First, we have to check that
f (x) is positive and decreasing. It is certainly positive. To determine if it is
decreasing, we compute
df 3 3 3
= 48xe−x − 24 × 3x4 e−x = 24x(2 − 3x3 )e−x
dx
which is negative for x ≥ 1. Therefore f (x) is decreasing for x ≥ 1, and the
integral test applies. The substitution u = x3 , du = 3x2 dx, yields
Z Z Z
2 −x3
f (x) dx = 24x e dx = 8e−u du = −8e−u + C
3
= −8e−x + C.
Therefore
Z ∞ Z R  R
−x3
f (x) dx = lim f (x) dx = lim −8e
1 R→∞ 1 R→∞
1
−R3 −1 −1
= lim (−8e + 8e ) = 8e
R→∞
∞
3
24n2 e−n converges and the
P
Since the integral is convergent, the series
n=1
∞
3
X
series (−1)n−1 24n2 e−n converges absolutely.
n=1
3
• Solution 2: Alternatively, we can use the ratio test with an = 24n2 e−n . We
calculate
3
an+1 24(n + 1)2 e−(n+1)
lim = lim
n→∞ an n→∞ 24n2 e−n3

1122
S OLUTIONS TO E XERCISES

3
!
(n + 1)2 en
= lim
n→∞ n2 e(n+1)3
 2
1 2
= lim 1 + e−(3n +3n+1) = 1 · 0 = 0 < 1,
n→∞ n
and therefore the series converges absolutely.
• Solution 3: Alternatively, alternatively, we can use the limiting compari-
son test. First a little intuition building. Recall that we need to show that
∞
3
24n2 e−n converges. The nth term in this series is
P
n=1

3 24n2
an = 24n2 e−n =
en3
It is a ratio with both the numerator and denominator growing with n. A
good rule of thumb is that exponentials grow a lot faster than powers. For
example, if n = 10 the numerator is 2400 = 2.4 × 103 and the denominator
is about 2 × 10434 . So we would guess that an tends to zero as n → ∞. The
question is “does an tend to zero fastPenough with n that our series con-
verges?”. For example, we know that ∞ 1
n=1 n2 converges (by the p-test with
p = 2). So if an tends to zero faster than n12 does, our series will converge.
3 2
So let’s try the limiting convergence test with an = 24n2 e−n = 24n en3
and
1
bn = n2 .
3
an 24n2 e−n 24n4
lim = lim = lim 3
n→∞ bn n→∞ 1/n2 n→∞ en

By l’Hôpital’s rule, twice,

24x4 4 × 24x3
lim 3 = lim 3 by l’Hôpital
x→∞ ex x→∞ 3x2 ex
32x
= lim x3 just cleaning up
x→∞ e
32
= lim 3 by l’Hôpital, again
x→∞ 3x2 ex
=0
P∞
That’s it. The limit comparison test now tells us that n=1 an converges.
3
(b) In part (a) we saw that 24n2 e−n is positive and decreasing. The limit of
this sequence equals 0 (as can be shown with l’Hôpital’s Rule, just as we did
at the end of the third solution of part (a)). Therefore, we can use the alter-
nating series test, so that the error made in approximating the infinite sum
∞ ∞ N
3
(−1)n−1 24n2 e−n by the sum of its first N terms, SN =
P P P
S = an = an ,
n=1 n=1 n=1
lies between 0 and the first omitted term, aN +1 . If we use 5 terms, the error satis-
fies
3.4.3.14. Solution. The error in our approximation using through term N is at
most (2(N1+1))! . We want
|S − S(2(N
1
| ≤+1))!
1
|a |<= 1000
24 × . By checking
36e−6
3
10−91values of N , we see
small
≈ 1.3 ×
5 6

1123
S OLUTIONS TO E XERCISES

that 8! = 40320 > 1000, so if N = 3, then 2(N1+1)! = 40320 1 1

< 1000 . So, for our
approximation, it suffices to consider the first four terms of our series.

3
X (−1)n 1 1 1 1
cos(2) ≈ = − + −
N =0
(2n)! 0! 2! 4! 6!
1 1 1
=1− + −
2 24 720
720 − 360 + 30 − 1 379
= =
720 720
When we use a calculator, we see
389
= 0.540277
720
cos(1) ≈ 0.540302
389 1
cos(1) − ≈ 0.000024528 ≈
720 40770
1 1
So, our error is reasonably close to our bound of 40320
, and far smaller than 1000
.
3.4.3.15. Solution. The terms of this series are sometimes negative and some-
times positive. But, they are not strictly alternating, so we can’t use the alter-
nating series test. Instead, we use a direct comparison test to show the series
converges absolutely.
If n is prime, then  n
an en/2 1 1
n
= − n = n/2 = √
e e e e
If n is not prime, then
an n2 n2
= − =
en en en
For n sufficiently large, n2 < en/2 , so for n sufficiently large,
 n
n2 1
n
≤ √ .
e e

√ X  1 n
Since e > 1, then e > 1, so the geometric series √ has |r| = r = √1e < 1,
e
∞
X an
so it converges. By the direct comparison test, converges as well. Then
n=1
en
∞
X an
n
converges absolutely, so it converges.
n=1
e

3.5 · Power Series

3.5.3 · Exercises
1124
S OLUTIONS TO E XERCISES

Exercises — Stage 1
3.5.3.1. Solution.
∞  n
X 3−1
f (1) =
n=0
4
∞  n
X 1
=
n=0
2

This is a geometric series with r = 12 .

1
= 1 =2
1− 2

3.5.3.2. Solution. Following Theorem 3.5.13, we differentiate our function

term-by-term.
∞
X (x − 5)n
f (x) =
n=1
n! + 2
∞
d (x − 5)n
X  
f 0 (x) =
n=1
dx n! + 2
∞
X n(x − 5)n−1
=
n=1
n! + 2

Keep in mind that x is our variable, and for each term, n is constant.
3.5.3.3. Solution. If x = c, then

f (x) = Aa (c − c)a + Aa+1 (c − c)a+1 + Aa+2 (c − c)a+2 + · · ·

= Aa · 0 + Aa+1 · 0 + Aa+2 · 0 + · · ·
=0

So, f (x) converges (to the constant 0) when x = c. (Had we allowed a = 0, it

would be possible for f (x) to converge to a nonzero number A0 , because we use
the convention 00 = 1.)
Depending on the sequence {An }, it’s possible that f (x) diverges for all x 6= c.
X∞
For example, suppose An = n!, so f (x) = n!(x − c)n . If x 6= c, then the limit
n=0
(n + 1)!(x − c)n+1
lim = lim (n + 1)|x − c| is infinity, since x − c 6= 0. So, the
n→∞ n!(x − c)n n→∞
series diverges.
We’ve now shown that the series definitely converges at x = c, but at any other
point, it may fail to converge.
3.5.3.4. Solution. According to Theorem 3.5.9, because f (x) diverges some-
where, and because it converges at a point other than its centre, f (x) has a posi-

1125
S OLUTIONS TO E XERCISES

tive radius of convergence R. That is, f (x) converges whenever |x − 5| < R, and
it diverges whenever |x − 5| > R.
If R > 6, then |11 − 5| < R, so f (x) converges at x = 11; since we are told f (x)
diverges at x = 11, we see R ≤ 6.
If R < 6, then | − 1 − 5| > R, so f (x) diverges at x = −1; since we are told f (x)
converges at x = −1, we see R ≥ 6.
Therefore, R = 6.

Exercises — Stage 2
3.5.3.5. *. Solution. (a) We apply the ratio test for the series whose k th term is
ak = (−1)k 2k+1 xk . Then

ak+1 (−1)k+1 2k+2 xk+1

lim = lim
k→∞ ak k→∞ (−1)k 2k+1 xk
= lim |2x| = |2x|
k→∞

Therefore, by the ratio test, the series converges for all x obeying |2x| < 1, i.e.
|x| < 12 , and diverges for all x obeying |2x| > 1, i.e. |x| > 12 . So the radius of
convergence is R = 12 .
Alternatively, one can set Ak = (−1)k 2k+1 and compute

Ak+1 (−1)k+1 2k+2

A = lim = lim = lim 2 = 2
k→∞ Ak k→∞ (−1)k 2k+1 k→∞

so that R = A1 = 21 , again.
(b) The series is
∞ ∞ ∞
X
k k+1 k
X
k
X 1
(−1) 2 x =2 (−2x) = 2 rk =2×
k=0 k=0 k=0
r=−2x 1−r
2
=
1 + 2x
for all |r| = |2x| < 1, i.e. all |x| < 21 .

3.5.3.6. *. Solution. We apply the ratio test for the series whose k th term is
k
ak = 10k+1x(k+1)! . Then

ak+1 xk+1 10k+1 (k + 1)!

lim = lim ·
k→∞ ak k→∞ 10k+2 (k + 2)! xk
10k+1 (k + 1)! xk+1
= lim · ·
k→∞ 10k+2 (k + 2)! xk
1
= lim |x| = 0 < 1
k→∞ 10(k + 2)

for all x. Therefore, by the ratio test, the series converges for all x and the radius

1126
S OLUTIONS TO E XERCISES

of convergence is R = ∞.
1 Ak+1
Alternatively, one can set Ak = and compute A = lim = 0,
10k+1 (k + 1)! k→∞ Ak
so that R is again +∞.
(x−2)n
3.5.3.7. *. Solution. We apply the ratio test with an = n2 +1
.

an+1 (x − 2)n+1 n2 + 1
lim = lim ·
n→∞ an n→∞ (n + 1)2 + 1 (x − 2)n

n2 + 1
= lim |x − 2|
n→∞ (n + 1)2 + 1

1 + 1/n2
= lim |x − 2|
n→∞ (1 + 1/n)2 + 1/n2

= |x − 2|

So, the series converges if |x − 2| < 1 and diverges if |x − 2| > 1. That is, the
radius of convergence is 1.

3.5.3.8. *. Solution. We apply the ratio test for the series whose nth term is
(−1)n (x + 2)n
an = √ . Then
n
√
an+1 (x + 2)n+1 n
lim = lim √
n→∞ an n→∞ n + 1 (x + 2)n
√
n
= lim |x + 2| √
n→∞ n+1
1
= lim |x + 2| p
n→∞ 1 + 1/n
= |x + 2|

So the series must converge when |x + 2| < 1 and must diverge when |x + 2| > 1.
When x + 2 = 1, the series reduces to
∞
X (−1)n
√
n=1
n

which converges by the alternating series test. When x + 2 = −1, the series
reduces to
∞
X 1
√
n=1
n

which diverges by the p–series test with p = 21 . So the interval of convergence is

−1 < x + 2 ≤ 1 or (−3, −1].

1127
S OLUTIONS TO E XERCISES

3.5.3.9. *. Solution. We apply the ratio test for the series whose nth term is
n
x+1 n
an = (−1)


n+1 3
.

(−1)n+1 x+1 n+1

an+1 n+2 3
lim = lim (−1)n x+1
n
n→∞ an n→∞
n+1 3
n+1
(−1) n+1 (x + 1)n+1 3n
= lim · · ·
n→∞ (−1)n n+2 (x + 1)n 3n+1
 
n+1 x+1 |x + 1|
= lim · =
n→∞ n+2 3 3
|x+1|
Therefore, by the ratio test, the series converges when 3
< 1 and diverges
when |x+1|
3
> 1. In particular, it converges when

|x + 1| < 3 ⇐⇒ −3 < x + 1 < 3 ⇐⇒ −4 < x < 2

(−1)n
and the radius of convergence is R = 3. (Alternatively, one can set An = (n+1)3n
and compute A = limn→∞ AAn+1 n
= 13 , so that R = A1 = 3.)
Next, we consider the endpoints 2 and −4. At x = 2, i.e. x + 1 = 3, the series
(−1)n
is simply ∞
P
n=0 n+1 , which is an alternating series: the signs alternate, and the
unsigned terms decrease to zero. Therefore the series converges at x = 2 by the
alternating series test.
At x = −4 the series is
∞ n X ∞ ∞
(−1)n −4 + 1 (−1)n

X X 1
= (−1)n = ,
n=0
n + 1 3 n=0
n + 1 n=0
n + 1
n
since (−1)n · (−1)n = (−1)2n = ((−1)2 ) = 1. This series diverges, either by
comparison or limit comparison with the harmonic series (the p-series with
P∞ p =
1). (For that matter, it is exactly equal to the standard harmonic series n=1 n1 ,
re-indexed to start at n = 0.)
In summary, the interval of convergence is −4 < x ≤ 2, or simply (−4, 2].
(x−2)n
3.5.3.10. *. Solution. We first apply the ratio test with an = n4/5 (5n −4)
.

an+1 (x − 2)n+1 n4/5 (5n − 4)

lim = lim ·
n→∞ an n→∞ (n + 1)4/5 (5n+1 − 4) (x − 2)n
n4/5 (5n − 4)
= lim |x − 2|
n→∞ (n + 1)4/5 (5n+1 − 4)

(1 − 4/5n )
= lim |x − 2|
n→∞ (1 + 1/n)4/5 (5 − 4/5n )

|x − 2|
=
5
Therefore the series converges if |x − 2| < 5 and diverges if |x − 2| > 5. When

1128
S OLUTIONS TO E XERCISES

∞ ∞
5n 1
P P
x − 2 = +5, i.e. x = 7, the series reduces to n4/5 (5n −4)
= n4/5 (1−4/5n )
which
n=1 n=1
1
diverges by the limit comparison test with bn = n4/5 . When x − 2 = −5, i.e.
∞
P (−5)n ∞
P (−1)n
x = −3, the series reduces to n4/5 (5n −4)
= n4/5 (1−4/5n )
which converges by
n=1 n=1
the alternating series test. So the interval of convergence is −3 ≤ x < 7 or [−3, 7).
(x+2)n
3.5.3.11. *. Solution. We apply the ratio test with an = n2
. Since

(x+2)n+1
an+1 (n+1)2 n2
lim = lim = lim |x + 2|
n→∞ an n→∞ (x+2)n n→∞ (n + 1)2
n2
1
= lim |x + 2| = |x + 2|
n→∞ (1 + 1/n)2

we have convergence for

|x + 2| < 1 ⇐⇒ −1 < x + 2 < 1 ⇐⇒ −3 < x < −1

and divergence for |x + 2| > 1. For |x + 2| = 1, i.e. for x + 2 = ±1, i.e. for
∞
P (±1)n
x = −3, −1, the series reduces to n2
, which converges absolutely, because
n=1
∞
1
P
np
converges for p = 2 > 1. So the given series converges if and only if
n=1
−3 ≤ x ≤ −1.
4n
3.5.3.12. *. Solution. We apply the ratio test with an = n
(x − 1)n . Since
an+1 4n+1 (x − 1)n+1 /(n + 1)
lim = lim
n→∞ an n→∞ 4n (x − 1)n /n
n
= lim 4|x − 1|
n→∞ n+1
n
= 4|x − 1| lim = 4|x − 1| · 1.
n→∞ n + 1

the series converges if

1 1
4|x − 1| < 1 ⇐⇒ −1 < 4(x − 1) < 1 ⇐⇒ − < x − 1 <
4 4
3 5
⇐⇒ <x<
4 4
and diverges if 4|x − 1| > 1. Checking the right endpoint x = 54 , we see that
∞ n X ∞
4n 5

X 1
−1 =
n=1
n 4 n=1
n

is the divergent harmonic series. At the left endpoint x = 34 ,

∞ n X ∞
4n 3 (−1)n
X 
−1 =
n=1
n 4 n=1
n

1129
S OLUTIONS TO E XERCISES

converges by the alternating series test.
Therefore the interval of convergence of

the original series is 43 ≤ x < 45 , or 43 , 54 .
n
3.5.3.13. *. Solution. We apply the ratio test with an = (−1)n 2(x−1)
n (n+2) . Since

an+1 (x − 1)n+1 2n (n + 2)
lim = lim n+1
n→∞ an n→∞ 2 (n + 3) (x − 1)n
|x − 1| n + 2
= lim
n→∞ 2 n+3
|x − 1| 1 + 2/n |x − 1|
= lim =
2 n→∞ 1 + 3/n 2

the series converges if

|x − 1|
< 1 ⇐⇒ |x − 1| < 2 ⇐⇒ −2 < (x − 1) < 2
2
⇐⇒ −1 < x < 3

and diverges if |x − 1| > 2. So the series has radius of convergence 2. Checking

the left endpoint x = −1, so that x−1
2
= −1, we see that
∞ ∞
X (−1 − 1)n X 1
(−1)n n =
n=0
2 (n + 2) n=0
n+2

x−1
is the divergent harmonic series. At the right endpoint x = 3, so that 2
= +1
and
∞ ∞
X (3 − 1)n
n
X (−1)n
(−1) n =
n=0
2 (n + 2) n=0 n + 2

converges by the alternating series test. Therefore

 the interval of convergence of
the original series is −1 < x ≤ 3, or − 1, 3 .

3.5.3.14. *. Solution. We apply the ratio test with an = (−1)n n2 (x − a)2n . Since

an+1 (−1)n+1 (n + 1)2 (x − a)2(n+1)

lim = lim
n→∞ an n→∞ (−1)n n2 (x − a)2n
(n + 1)2
= lim |x − a|2
n→∞ n2
1
2
= |x − a|2 lim 1 + = |x − a|2 · 1.
n→∞ n
the series converges if

|x − a|2 < 1 ⇐⇒ |x − a| < 1 ⇐⇒ −1 < x − a < 1

⇐⇒ a − 1 < x < a + 1

1130
S OLUTIONS TO E XERCISES

and diverges if |x − a| > 1. Checking both endpoints x − a = ±1, we see that

∞
X ∞
X
(−1)n n2 (x − a)2n = (−1)n n2
n=1 x−a=±1 n=1

fails the divergence test — the nth term does not converge to zero as n → ∞.
Therefore the interval

of convergence of the original series is a − 1 < x < a + 1,
or a − 1, a + 1 .
3.5.3.15. *. Solution. (a) We apply the ratio test for the series whose k th term is
k
Ak = (x+1)
k2 9k
. Then
Ak+1 (x + 1)k+1 k 2 9k
lim = lim
k→∞ Ak k→∞ (k + 1)2 9k+1 (x + 1)k

1 k2
= lim |x + 1|
k→∞ 9 (k + 1)2
1 1
= lim |x + 1|
k→∞ 9 (1 + 1/k)2
|x + 1|
=
9
So the series must converge when |x + 1| < 9 and must diverge when |x + 1| > 9.
When x + 1 = ±9, the series reduces to
∞ ∞
X (±9)k X (±1)k
2 9k
=
k=1
k k=1
k2
which converges (since, by the p–test, ∞ 1
P
k=1 kp converges for any p > 1). So the
interval of covnergence is |x + 1| ≤ 9 or −10 ≤ x ≤ 8 or [−10, 8].
(b) The partial sum
N 
X ak ak+1 
−
k=1
ak+1 ak+2
a a2   a2 a3   a aN +1 
1 N
= − + − + ··· + −
a2 a3 a3 a4 aN +1 aN +2
a1 aN +1
= −
a2 aN +2
∞ 
X ak ak+1  a1
We are told that − = . This means that the above partial sum
k=1
ak+1 ak+2 a2
a1
converges to a2 as N → ∞, or equivalently, that
aN +1
lim =0
N →∞ aN +2

and hence that

|ak+1 (x − 1)k+1 | |ak+1 |
lim k
= |x − 1| lim
k→∞ |ak (x − 1) | k→∞ |ak |

is infinite for any x 6= 1. So, by the ratio test, this series converges only for x = 1.

1131
S OLUTIONS TO E XERCISES

∞
1
xn =
P
3.5.3.16. *. Solution. Using the geometric series 1−x
,
n=0

∞ ∞ ∞
x3 3
X
n
X
n+3
X
=x x = x = xn
1−x n=0 n=0 n=3

3.5.3.17. Solution. We can find f (x) by differentiating its integral, or antidif-

ferentiating its derivative. In the latter case, we’ll have to solve for the arbitrary
constant of integration; in the former case, we do not. (Remember that many
different functions have the same derivative, but a single function has only one
derivative.) To avoid the necessity of finding the arbitrary constant, we can ig-
nore the given equation for f 0 (x), which makes the problem much simpler. This
is the method used in Solution 1.
• Solution 1: Using the Fundamental Theorem of Calculus Part 1:
Z x 
d
f (t)dt = f (x)
dx 5
( ∞
)
d X (x − 1)n+1
So, f (x) = 3x +
dx n=0
n(n + 1)2
∞
X (n + 1)(x − 1)n
=3+
n=1
n(n + 1)2
∞
X (x − 1)n
=3+
n=1
n(n + 1)

• Solution 2: Suppose we had used f 0 (x) instead. We would antidifferentiate

to find:
∞
!
(x − 1)n
Z X
f (x) = dx
n=0
n + 2
∞
!
X (x − 1)n+1
= +C
n=0
(n + 1)(n + 2)
∞
!
X (x − 1)n
= +C
n=1
n(n + 1)

Notice f (1) = 0 + C. So, to find C, we must find f (1). We can’t get that
information
Rx from f 0 (x), so our only option is to consider the given formula
for 5 f (t)dt. Using the Fundamental Theorem of Calculus Part 1:
Z x 
d
f (1) = f (t)dt
dx 5 x=1
( ∞
)
n+1
d X (x − 1)
= 3x +
dx n=1
n(n + 1)2
x=1

1132
S OLUTIONS TO E XERCISES
" ∞
#
X (n + 1)(x − 1)n
= 3+
n=1
n(n + 1)2
x=1
" ∞
#
X (x − 1)n
= 3+
n=1
n(n + 1)
x=1
∞ n
X 0
=3+
n=1
n(n + 1)
=3
∞
X (x − 1)n
So, f (x) = 3 + .
n=1
n(n + 1)
Note that in Solution 2, we did the same calculation as Solution 1, and more.

Exercises — Stage 3
3.5.3.18. *. Solution. We apply the ratio test for the series whose nth term is
n n
either an = 32nxlog n or an = 32nxlog n . For both series
an+1 xn+1 32n log n
lim = lim 2(n+1)
n→∞ an n→∞ 3 log(n + 1) xn
x log n x log n
= lim 2 = lim 2
n→∞ 3 log(n + 1) n→∞ 3 [log(n) + log(1 + 1/n)]

x
= lim 2
n→∞ 3 [1 + log(1 + 1/n)/ log(n)]

|x|
=
9
Therefore, by the ratio test, our series converges absolutely when |x| < 9 and
diverges when |x| > 9.
∞ ∞
X xn X (−1)n
For x = −9, 2n log n
= which converges by the alternating series
n=2
3 n=2
log n
test. ∞ ∞ ∞
X xn X 1 X (−1)n
For x = +9, 2n log n
= which is the same series as .
n=2
3 n=2
log n n=2
log n
We shall shortly show that n ≥ log n, and hence log1 n ≥ n1 for all n ≥ 1. This
∞
X 1
implies that the series diverges by comparison with the divergent series
n=2
log n
∞
X 1
. This yelds both divergence for x = 9 and also the failure of absolute
n=2
np p=1
convergence for x = −9.
Finally, we show that n − log n > 0, for all n ≥ 1. Set f (x) = x − log x. Then
f (1) = 1 > 0 and
1
f 0 (x) = 1 − ≥ 0 for all x ≥ 1
x

1133
S OLUTIONS TO E XERCISES

So f (x) is (strictly) positive when x = 1 and is increasing for all x ≥ 1. So f (x) is

(strictly) positive for all x ≥ 1.
∞
1 X
3.5.3.19. *. Solution. (a) Applying = (−1)n rn with r = x3 gives
1 + r n=0

∞ ∞
1 x3n+1
Z X Z X
3n
dx = (−1)n x dx = (−1)n +C
1 + x3 n=0 n=0
3n + 1

(b) By part (a),

1/4 ∞ 3n+1 1/4 ∞
1 n x 1
Z X X
dx = (−1) = (−1)n
0 1 + x3 n=0
3n + 1 0 n=0
(3n + 1)43n+1

This is an alternating series with successively smaller terms that converge to zero
as n → ∞. So truncating it introduces an error no larger than the magnitude of
the first dropped term. We want that first dropped term to obey

1 1
3n+1
< 10−5 = 5
(3n + 1)4 10

So let’s check the first few terms.

1 1 1
= > 5
(3n + 1)43n+1 n=0 4 10
1 1 1
= 5 > 5
(3n + 1)43n+1 n=1 4 10
1 1 1 1
= = =
(3n + 1)43n+1 n=2 7×4 7 7×2 14 7 × 16 × 1024
1 1
= < 5
112 × 1024 10
So we need to keep two terms (the n = 0 and n = 1 terms).
3.5.3.20. *. Solution. (a) Differentiating both sides of
∞
X 1
xn =
n=0
1−x

gives
∞
X 1
nxn−1 =
n=0
(1 − x)2
Now multiplying both sides by x gives
∞
X x
nxn =
n=0
(1 − x)2

1134
S OLUTIONS TO E XERCISES

as desired.
(b) Differentiating both sides of the conclusion of part (a) gives
∞
X (1 − x)2 − 2x(x − 1) (1 − x)(1 − x + 2x)
n2 xn−1 = 4
=
n=0
(1 − x) (1 − x)4
1+x
=
(1 − x)3

Now multiplying both sides by x gives

∞
X x(1 + x)
n 2 xn =
n=0
(1 − x)3

We know that differentiation preserves the radius of convergence of power se-

ries. So this series has radius of convergence 1 (the radius of convergence of the
original geometric series). At x = ±1 the series diverges by the divergence test.
So the series converges for −1 < x < 1.
∞
P
3.5.3.21. *. Solution. By the divergence test, the fact that (1 − bn ) converges
n=0
guarantees that lim (1 − bn ) = 0, or equivalently, that lim bn = 1. So, by equation
n→∞ n→∞
(3.5.2), the radius of convergence is
 −1  −1
bn+1 1
R = lim = =1
n→∞ bn 1
3.5.3.22. *. Solution. (a) We know that the radius of convergence R obeys

1 an+1 n (n + 1)an+1 C
= lim = lim =1 =1
R n→∞ an n→∞ n + 1 nan C
because we are told that lim nan = C. So R = 1.
n→∞
(b) Just knowing that the radius of convergence is 1, we know that the series
converges for |x| < 1 and diverges for |x| > 1. That leaves x ± 1.
∞
P
When x = +1, the series reduces to an . We are told that nan decreases to C > 0.
n=1
∞
C 1
P
So an ≥ n
. By the comparison test with the harmonic series n
, which diverges
n=1
by the p–test with p = 1, our series diverges when x = 1.
∞
(−1)n an . We are told that nan decreases to
P
When x = −1, the series reduces to
n=1
∞
(−1)n an
P
C > 0. So an > 0 and an converges to 0 as n → ∞. Consequently
n=1
converges by the alternating series test.
∞
an xn converges when −1 ≤ x < 1.
P
In conclusion
n=1

1135
S OLUTIONS TO E XERCISES

3.5.3.23. Solution. Equation 2.3.1 tells

P us the centre of mass of a rod with
mn xn
weights {mn } at positions {xn } is x̄ = P .
mn
We find the combined mass of our weights using Lemma 3.2.5 with r = 12 and
r = 13 , respectively.
∞ ∞ ∞ ∞
X 1 X 1 X 1 1 X 1 1
n
+ n
= · n
+ · n
n=1
2 n=1
3 n=0
2 2 n=0
3 3
1 1 1 1
·
= 1 + ·
2 1− 2 3 1 − 13
1 3
=1+ =
2 2
Now, we want to calculate the sum of the products of the masses and their posi-
tions. ∞ ∞
X 1 X 1
n
·n+ · (−n)
n=1
2 n=1
3n
We don’t have such a nice formula for this, but we can make one by differentiat-
ing.
The following formula is true for any x with |x| < 1:
∞
X 1
xn =
n=0
1−x
Differentiating both sides with respect to x:
∞
X 1
nxn−1 =
n=0
(1 − x)2
∞
X 1
nxn−1 =
n=1
(1 − x)2
Multiplying both sides by x:
∞
X x
nxn =
n=1
(1 − x)2
This allows us to evaluate our series.
∞ ∞ 1 1
X n X n 2 3
− =
−
1 2
n n
2
n=1
2 n=1
3 1− 2 1 − 13
4 2
=2− =
3 3
Therefore,
2/3 4
x̄ = = = 0.44
3/2 9
Remark: we can check that this makes some sense. Since the weights to the right
of x = 0 are heavier than those to the left, but spaced the same, we would expect
our rod to balance to the right of x = 0.

1136
S OLUTIONS TO E XERCISES

3.5.3.24. Solution. First, we differentiate.

∞
X
f (x) = An (x − c)n
n=0
X∞
f 0 (x) = nAn (x − c)n−1
n=0
∞
X
= nAn (x − c)n−1
n=1
∞
X
f 0 (c) = nAn · 0n−1
n=1
= A1 · 1 + 2A2 · 0 + 3A3 · 0 + · · ·
= A1

So, if A1 = 0, then f 0 (c) = 0. That is, f (x) has a critical point at x = c.

To determine the behaviour of this critical point, we use the second derivative
test.
∞
X
f 0 (x) = nAn (x − c)n−1
n=1
∞
X
f 00 (x) = n(n − 1)An (x − c)n−2
n=1
X∞
= n(n − 1)An (x − c)n−2
n=2
∞
X
f 00 (c) = n(n − 1)An · 0n−2
n=2
= 2(1)A2 · 00 + 3(2)A3 · 01 + 4(3)A4 · 02 + · · ·
= 2A2

Following the second derivative test, x = c is the location of a local maximum if

A2 < 0, and it is the location of a local minimum if A2 > 0. (If A2 = 0, the critical
point may or may not be a local extremum.)
∞ ∞
X n X
3.5.3.25. Solution. We recognize n−1
as f (x) = n · xn−1 , evaluated at
n=3
5 n=3
1
x = . We should figure out what f (x) is in equation form (as opposed to power
5
series X
form). Notice that this looks similar to the derivative of the geometric
series xn .

∞
1 X
= xn when
1 − x n=0

1137
S OLUTIONS TO E XERCISES
  (∞ )
d 1 d X
= xn
dx 1−x dx n=0
∞
1 X
= nxn−1
(1 − x)2 n=1
∞
X
0 1
= 1x + 2x + nxn−1
n=3
∞
X
= 1 + 2x + nxn−1
n=3
∞
1 X
So, − 1 − 2x = nxn−1
(1 − x)2 n=3
∞  n−1
1 2 X 1
Setting : −1− = n
(1 − 1/5)2 5 n=3
5
 2 ∞
5 2 X n
−1− =
4 5 n=3 5n−1

25 2 13
So, our series evaluates to −1− = .
16 5 80
3.5.3.26. Solution. As we saw in in Example 3.5.20,
∞
X xn+1
log(1 + x) = (−1)n
n=0
n+1

which is an alternating series when x is positive. If we use its partial sum SN to

approximate log(1 + x), the absolute error involved is no more than

x(N +1)+1 xN +2
=
(N + 1) + 1 N +2
1
We want this error to be at most 10−5 whenever 0 < x < 10 . For this range of x
N +2
x 1
values, < , so we want N that satisfies the inequality:
N +2 (N + 2)10N +2
1 1
N +2
≤ 5
(N + 2)10 10
N +2
⇒ (N + 2)10 ≥ 105

We see N = 3 suffices.
So, the partial sum
3
X
n x
n+1
x2 x3 x4
(−1) =x− + −
n=0
n+1 2 3 4

1138
S OLUTIONS TO E XERCISES

x5
approximates log(1 + x) to within an error of .
5
1 1
When x is between 0 and 10 , that error is at most < 10−5 , as desired.
5 · 105
Now we can approximate log(1.05).
 
1
log(1.05) = log 1 +
20
1

2 1 3

1 4


1
− 20 + 20 − 20


≈ 20
2 3 4
12 × 203 − 6 × 202 + 4 × 20 − 3 93677
= 4
=
12 × 20 1920000
93677
We note that a computer approximates 1920000 ≈ 0.04879010 and log(1.05) ≈
−8
0.04879016. So, our actual error is around 6 × 10 .
3.5.3.27. Solution. As we saw in in Example 3.5.21,
∞
X x2n+1
arctan x = (−1)n
n=0
2n + 1

which is an alternating series when x is nonzero. If we use its partial sum SN to

approximate arctan x, the absolute error involved is no more than

|x|2(N +1)+1 |x|2N +3

=
N (n + 1) + 1 2N + 3

We want this error to be at most 10−6 whenever − 14 < x < 14 . For this range of x
|x|2N +3 1
values, < , so we want N that satisfies the inequality:
2N + 3 (2N + 3)42N +3
1 1
2N +3
≤ 5
(2N + 3)4 10

A quick check with a calculator shows that N = 2 suffices.

So, the partial sum
2
X x2n+1 x3 x5
(−1)n =x− +
n=0
2n + 1 3 5
x7
approximates arctan x to within an error of .
7
1 1 1
When x is between − 41 and 14 , that error is at most 7
= < =
7·4 114688 100000
10−5 , as desired. (When x = 0, our approximation is 0, the exact value of arctan 0.)

3.6 · Taylor Series

3.6.7 · Exercises

1139
S OLUTIONS TO E XERCISES

Exercises — Stage 1
3.6.7.1. Solution. All functions A, B, and C intersect the function y = f (x)
when x = 2. B is a constant function, so this is the constant approximation. A is
the tangent line, so A is the linear approximation. C is a tangent parabola, so C
is the quadratic approximation.
3.6.7.2. Solution. Following how a Taylor series is constructed, the Taylor
series and the function agree at the point chosen as the centre. So, T (5) =
arctan3 (e5 + 7).
If we were evaluating a Taylor series at a point other than its centre, we would
generally need to check that (a) the series converges, and (b) it converges to the
same value as the function we used to create it.
3.6.7.3. Solution. These are listed in Theorem 3.6.7. However, it’s possi-
ble to figure out many of them without a lot of memorization. For example,
1
e0 = cos(0) = 1−0 = 1, while sin(0) = log(1 + 0) = arctan(0) = 0. So by plugging
in x = 0 to the series listed, we can divide them into these two categories.
The derivative of sine is cosine, so we can also look for one series that is the
derivative of another. The derivative of ex is ex , so we can look for a series that is
its own derivative.
Furthermore, sine and arctangent are odd functions and only II and IV are odd.
Cosine is an even function and only III is even.
Alternately, we can find the first few terms of each series using the definition of
a Taylor series, and match them up.
All together, the functions correspond to the following series:
A-V
B-I
C - IV
D - VI
E - II
F - III
3.6.7.4. Solution.

a Using the definition of a Taylor series, we know

∞ ∞
X n2 X f (n) (3)
(x − 3)n = (x − 3)n
n=0
(n! + 1) n=0
n!

(20)
So, the coefficient of (x − 3)20 is f 20!(3) (using the definition). Using the
202
given series, the coefficient of (x − 3)20 is 20!+1 . So,

f (20) (3) 202

=
20! 20! 
+1 
(20) 2 20!
⇒ f (3) = 20
20! + 1

(which is extremely close to 202 ).

1140
S OLUTIONS TO E XERCISES

b Using the definition of a Taylor series, we know

∞ ∞
X n2 X g (k) (3)
(x − 3)2n = (x − 3)k
n=0
(n! + 1) k=0
k!

(20)
So, the coefficient of (x − 3)20 is g 20!(3) (using the definition). Looking at the
102
given series, the coefficient of (x − 3)20 occurs when n = 10, so it is 10!+1 . So,

g (20) (3) 102

=
20! 10! 
+1 
(20) 2 20!
⇒ g (3) = 10
10! + 1

c With the previous two examples in mind, we find the Maclaurin series for
h(x). (Using the series representation will be much easier than differenti-
ating h(x) directly twenty times.) Recall from the text that we know the
Maclaurin series for arctan x.
∞
X x2n+1
arctan(x) = (−1)n
n=0
2n + 1
∞ 2 2n+1 ∞ 2n+1
n (5x ) n 5
X X
2
arctan(5x ) = (−1) = (−1) x4n+2
n=0
2n + 1 n=0
2n + 1
∞
arctan(5x2 ) X n 5
2n+1
= (−1) x4n−2
x4 n=0
2n + 1
∞ ∞
X h(k) (0) X
k 52n+1 4n−2
x = (−1)n x
k=0
k! n=0
2n + 1

h(22) (0)
Using the definition of a Maclaurin series, the coefficient of x22 is .
22!
This occurs in the given series when n = 6, so

h(22) (0) 6 5
2×6+1
513
= (−1) =
22! 2×6+1 13
13
22! · 5
⇒ h(22) (0) =
13
h(20)(0)
Similarly, the coefficient of x20 in the Maclaurin series is . Since no
20!
term x20 occurs in our series, that coefficient is 0, so h(20) (0) = 0.

Exercises — Stage 2
3.6.7.5. Solution. The definition of a Taylor series tells us we will be computing

1141
S OLUTIONS TO E XERCISES

the coefficients in the series

∞
X f (n) (1)
(x − 1)n
n=0
n!

That is, we need a general description of f (n) (1). To find this, we take a few
derivatives, and look for a pattern.

f (x) = log(x) f (1) = 0

f 0 (x) = x−1 f 0 (1) = 1
f 00 (x) = (−1)x−2 f 00 (1) = −1
f (3) (x) = (−2)(−1)x−3 f (3) (1) = 2!
f (4) (x) = (−3)(−2)(−1)x−4 f (4) (1) = −3!
f (5) (x) = (−4)(−3)(−2)(−1)x−5 f (5) (1) = 4!
f (6) (x) = (−5)(−4)(−3)(−2)(−1)x−6 f (6) (1) = −5!
.. ..
. .
f (n) (x) = (−1)n−1 (n − 1)! x−n f (n) (1) = (−1)n−1 (n − 1)!

Using the convention 0! = 1, our pattern for f (n) (1) begins when n = 1.
∞ ∞
X f (n) (1) n
X (−1)n−1 (n − 1)!
(x − 1) = 0 + (x − 1)n
n=0
n! n=1
n!
∞
X (−1)n−1
= (x − 1)n
n=1
n

3.6.7.6. Solution. To find the Taylor series for sine, centred at a = π, we’ll need
to know the various derivatives of sine at π.

f (x) = sin x f (π) = 0

f 0 (x) = cos x f 0 (π) = −1
f 00 (x) = − sin x f 00 (π) = 0
f 000 (x) = − cos x f 000 (π) = 1
f (4) (x) = sin x = f (x) f (4) (π) = 0

Even derivatives are 0; odd derivatives alternate between −1 and +1. (If you’re
following along with the derivation of the Maclaurin series for sine in the text,
note f (n) (π) = −f (n) (0).)
In our Taylor series, every even-indexed term will be zero, and we will be left
with only odd-indexed terms. If we let n be our index, then the term 2n + 1 will
capture all the odd numbers. Since the signs alternate, f (2n+1) (π) = (−1)n+1 . So,

1142
S OLUTIONS TO E XERCISES

our Taylor series is:

∞ ∞
X f (k) (π) X f (2n+1) (π)
k
(x − π) = (x − π)2n+1
k=0
k! n=0
(2n + 1)!
(since the even terms are all zero)
∞
X (−1)n+1
= (x − π)2n+1
n=0
(2n + 1)!

3.6.7.7. Solution. The definition of a Taylor series tells us we will be computing

the coefficients in the series
∞
X g (n) (10)
(x − 10)n
n=0
n!

That is, we need a general description of g (n) (10). To find this, we take a few
derivatives, and look for a pattern.

1
g(x) = x−1 g(10) =
10
−1
g 0 (x) = (−1)x−2 g 0 (10) =
102
(−1)2 2!
g 00 (x) = (−2)(−1)x−3 g 00 (10) =
103
(−1)3 3!
g (3) (x) = (−3)(−2)(−1)x−4 g (3) (10) =
104
(−1)4 4!
g (4) (x) = (−4)(−3)(−2)(−1)x−5 g (4) (10) =
105
(−1)5 5!
g (5) (x) = (−5)(−4)(−3)(−2)(−1)x−6 g (5) (10) =
106
.. ..
. .
(−1)n n!
g (n) (x) = (−1)n n!x−(n+1) g (n) (10) =
10n+1
Using the convention 0! = 1, our pattern for g (n) (10) begins when n = 0.

∞ ∞
X g (n) (1) X (−1)n n!
(x − 10)n = (x − 10)n
n=0
n! n=0
n!10n+1
∞
X (x − 10)n
=−
n=0
(−10)n+1
∞  n
1 X 10 − x
=
10 n=0 10

1143
S OLUTIONS TO E XERCISES

10−x
For fixed x, we recognize this as a geometric series with r = 10
. So it converges
precisely when |r| < 1, i.e.

10 − x
<1
10
|10 − x| < 10
−10 < x − 10 < 10
0 < x < 20

So, its interval of convergence is (0, 20).

3.6.7.8. Solution. The definition of a Taylor series tells us we will be computing
the coefficients in the series
∞
X h(n) (a)
(x − a)n
n=0
n!

That is, we need a general description of h(n) (a). To find this, we take a few
derivatives, and look for a pattern.

h(x) = e3x h(a) = e3a

h0 (x) = 3e3x h0 (a) = 3e3a
h00 (x) = 32 e3x h00 (a) = 32 e3a
h000 (x) = 33 e3x h000 (a) = 33 e3a
.. ..
. .
h(n) (x) = 3n e3x h(n) (a) = 3n e3a

The pattern for h(n) (a) holds for all (whole numbers) n ≥ 0. So, our Taylor series
for h(x) is
∞
X 3n e3a
(x − a)n
n=0
n!
To find its radius of convergence, we use the ratio test.

an+1 3n+1 e3a (x − a)n+1 n!

= · n 3a
an (n + 1)! 3 e (x − a)n
3n+1 e3a n! (x − a)n+1
= · · ·
3n e3a (n + 1)! (x − a)n
1
=3· · |x − a|
n + 1 
an+1 3
lim = lim · |x − a| = 0
n→∞ an n→∞ n + 1

Our series converges for every value of x, so its radius of convergence is ∞.

1144
S OLUTIONS TO E XERCISES

∞
1 X
3.6.7.9. *. Solution. Substituting y = 2x into = y n (which is valid for
1−y n=0
all −1 < y < 1) gives
∞ ∞
1 1 X X
f (x) = =− =− (2x)n = − 2n xn
2x − 1 1 − 2x n=0 n=0

for all − 12 < x < 21 .

1
3.6.7.10. *. Solution. Substituting first y = −x and then y = 2x into =
1−y
X∞
y n (which is valid for all −1 < y < 1) gives
n=0

∞ ∞
1 X X
= (−x)n = (−1)n xn
1 − (−x) n=0 n=0
∞ ∞
1 X X
= (2x)n = 2n xn
1 − (2x) n=0 n=0

for all − 12 < x < 21 . Hence, for all − 21 < x < 12 ,

3 1 3 1
f (x) = − = +
x + 1 2x − 1 1 − (−x) 1 − 2x
∞
X ∞
X ∞
X
n n n n
3(−1)n + 2n xn


=3 (−1) x + 2 x =
n=0 n=0 n=0

So bn = 3(−1)n + 2n .
3.6.7.11. *. Solution. We found the Taylor series for e3x from scratch in Ques-
tion 8. If we hadn’t just done that, we could easily find it by modifying the series
for ex .
Substituting y = 3x into the exponential series
∞
y
X yn
e =
n=0
n!

gives
∞ ∞
3x
X (3x)n X 3n xn
e = =
n=0
n! n=0
n!
35
so that c5 , the coefficient of x5 , which appears only in the n = 5 term, is c5 =
5!

1145
S OLUTIONS TO E XERCISES

3.6.7.12. *. Solution. Since

∞
d 2 X 1
f 0 (t) = log(1 + 2t) = =2 (−2t)n if |2t| < 1 i.e. |t| <
dt 1 + 2t n=0
2

and f (0) = 0, we have

x ∞ Z x ∞
xn+1
Z X X
0 n n n n n+1
f (x) = f (t) dt = 2 (−1) 2 t dt = (−1) 2
0 n=0 0 n=0
n+1
1
for all |x| < .
2
3.6.7.13. *. Solution. We just need to substitute y = x3 into the known Maclau-
rin series for sin y, to get the Maclaurin series for sin(x3 ), and then multiply the
result by x2 .

y3
sin y = y − + ···
3!
3 3 x9
sin(x ) = x − + ···
3!
x11
x2 sin(x3 ) = x5 − + ···
3!

so a = 1 and b = − 3!1 = − 16 .
3.6.7.14. *. Solution. Recall that
∞
y
X yn y2 y3
e = =1+y+ + + ···
n=0
n! 2 3!

Setting y = −x2 , we have

2 x4 x6
e−x = 1 − x2 + − + ···
2 3!
2 x4 x6
e−x − 1 = −x2 + − + ···
2 6
2
e−x − 1 x3 x5
= −x + − + ···
x 2 6
Z −x2
e −1 x2 x4 x6
dx = C − + − + ···
x 2 8 36
3.6.7.15. *. Solution. Recall that
∞
X y 2n+1
arctan(y) = (−1)n
n=0
2n + 1

1146
S OLUTIONS TO E XERCISES

Setting y = 2x, we have

∞
!
2n+1
n (2x)
Z Z X
4 4
x arctan(2x) dx = x (−1) dx
n=0
2n + 1
∞
!
2n+1 2n+5
2 x
Z X
= (−1)n dx
n=0
2n + 1
∞
X 22n+1 x2n+6
= (−1)n +C
n=0
(2n + 1)(2n + 6)
∞
X 22n x2n+6
= (−1)n +C
n=0
(2n + 1)(n + 3)

1 ∞
Substituting y = −3x3 into y n gives
P
3.6.7.16. *. Solution. =
1 − y n=0
∞ ∞
df 1 X
3 n
X
(−1)n 3n x3n+1


=x· =x − 3x =
dx 1 + 3x3 n=0 n=0

Now integrating,
∞
X x3n+2
f (x) = (−1)n 3n +C
n=0
3n + 2

To have f (0) = 1, we need C = 1. So, finally

∞
X 3n
f (x) = 1 + (−1)n x3n+2
n=0
3n + 2

3.6.7.17. *. Solution. We’re given a big hint: that our series resembles the Taylor
series for arctangent.
x2n+1
The terms of arctangent are (−1)n . Our terms resemble those terms, with
2n + 1
1
x2n+1 replaced by n .
√
2n 3 1 √
2n+1
Since 3n = 3 = √3 3 :
∞ ∞
X (−1)n √ X (−1)n
= 3 √
2n+1
n=0
(2n + 1)3n n=0 (2n + 1) 3
∞
√ X x2n+1 √ 1
= 3 (−1)n = 3 arctan √
n=0
2n + 1 x= √1 3
3
√ π π
= 3 = √
6 2 3

1147
S OLUTIONS TO E XERCISES

∞
x
X xn
3.6.7.18. *. Solution. Recall that e = . So
n=0
n!

∞ ∞
X (−1)n hX xn i h i
= = ex = e−1
n=0
n! n=0
n! x=−1 x=−1

∞
x
X xk
3.6.7.19. *. Solution. Recall that e = . So
k=0
k!

∞ ∞
X 1 hX xk i h i
k k!
= = ex = e1/e
k=0
e k=0
k! x=1/e x=1/e

∞
x
X xk
3.6.7.20. *. Solution. Recall that e = . So
k=0
k!

∞ ∞
X 1 hX xk i h i
k
= = ex = e1/π
k=0
π k! k=0
k! x=1/π x=1/π

This series differs from the given one only in that it starts with k = 0 while the
given series starts with k = 1. So
∞ ∞
X 1 X 1
k k!
= k k!
1 = e1/π − 1
− |{z}
k=1
π k=0
π
k=0

3.6.7.21. *. Solution. Recall, from Theorem 3.6.7, that, for all −1 < x ≤ 1,
∞ k+1 ∞
X
k x
X xn
log(1 + x) = (−1) = (−1)n−1
k=0
k + 1 n=1 n

(To get from the first sum to the second sum we substituted n = k + 1. If you
don’t see why the two sums are equal, write out the first few terms of each.) So
∞ ∞
X (−1)n−1 hX
n−1 x
ni h i
= (−1) = log(1 + x) = log(3/2)
n=1
n 2n n=1
n x=1/2 x=1/2

3.6.7.22. *. Solution. Write

∞ ∞ ∞
X n+2 n
X n n X 2 n
e = e + e
n=1
n! n=1
n! n=1
n!
∞ ∞
X en X en
= +2
n=1
(n − 1)! n=1
n!
∞ ∞
X en−1 X en
=e +2
n=1
(n − 1)! n=1
n!

1148
S OLUTIONS TO E XERCISES

∞ ∞
X en X en
=e +2
n=0
n! n=1
n!

∞
x
X xn
Recall that e = . So
n=0
n!

∞ ∞ ∞
X n+2 hX
n xn i hX xn i
e =e +2
n=1
n! n=0
n! x=e n=1
n! x=e
h i h i
= e ex + 2 ex − 1 = ee+1 + 2(ee − 1)
x=e x=e
e
= (e + 2)e − 2
3.6.7.23. Solution. Let’s use the ratio test:
2n+1
an+1 n+1
lim = lim 2n
n→∞ an n→∞
n
n
= lim 2 =2>1
n→∞ n+1
So, the series diverges.
∞
∞
y n+1 XX (−y)n n
Remark: it’s tempting to note that log(1 + y) = (−1) = − ,
n=0
n+1 n=1
n
and try to substitute in y = −2. But, the Maclaurin series for log(1 + y) has
radius of convergence R = 1, so it doesn’t converge at y = −2. Furthermore,
log(1 + (−2)) = log(−1), but this is undefined.
3.6.7.24. Solution. Our series looks something like the Taylor series for sine,
∞
X (−1)n 2n+1
sin x = x .
n=0
(2n + 1)!

∞
X (−1)n  π 2n+1
1 + 22n+1


n=0
(2n + 1)! 4
∞
(−1)n  π 2n+1  π 2n+1
X  
= +
n=0
(2n + 1)! 4 2
∞ ∞
X (−1)n  π 2n+1 X (−1)n  π 2n+1
= +
n=0
(2n + 1)! 4 n=0
(2n + 1)! 2
π  π 
= sin + sin
4 √2
1 1+ 2
= √ +1= √
2 2
3.6.7.25. *. Solution. (a)

1149
S OLUTIONS TO E XERCISES

x2n
• Solution 1: The naive strategy is to set an = and apply the ratio test.
(2n)!
x2n+2
an+1 (2n+2)! x2n+2 (2n)!
lim = lim x2n
= 2n
·
n→∞ an n→∞
(2n)!
x (2n + 2)(2n + 1)(2n)!
x2
= lim
n→∞ (2n + 2)(2n + 1)

=0

This is smaller than 1 no matter what x is. So the series converges for all x.
∞
x
X xn
• Solution 2: Alternatively, the sneaky way is to observe that both e =
n=0
n!
∞
X (−x)n
and e−x = are known to converge for all x. So
n=0
n!

∞
1 x −x

X xn X x2n
e +e = =
2 n even
n! n=0
(2n)!

also converges for all x.

∞
x
X xn
(b) Recall that e = . Then:
n=0
n!

∞
X 1
e=
n=0
n!
∞
−1
X (−1)n
e =
n=0
n!
∞ ∞ ∞
X 1 + (−1)n X 1 X 1
e + e−1 = =2 =2
n=0
n! n even
n! n=0
(2n)!

∞  
X 1 1 1
Hence = e+ .
n=0
(2n)! 2 e

3.6.7.26. Solution. All three series we’re adding up are alternating, so we can
bound the absolute error in the approximation SN (the N -th partial sum) by
|aN +1 |.
The Taylor series for arctangent is
∞
X x2n+1
arctan(x) = (−1)n
n=0
2n + 1
for every real x.

1150
S OLUTIONS TO E XERCISES

a Using the Taylor series for arctangent when x = 1, we see

∞
π X 1
= arctan(1) = (−1)n
4 n=0
2n + 1
∞
X 4
π= (−1)n
n=0
2n + 1

The error involved in approximating π with the partial sum SN is at most

|aN +1 | = 2N4+3 . In order for this to be at most 4 × 10−5 , we need:

4
≤ 4 × 10−5
2N + 3
2N + 3 ≥ 105
105 − 3 3
N≥ = 5 × 104 − = 50, 000 − 1.5
2 2
Since n must be an integer, we need to add up the terms from n = 0 to
n = 49, 999. That is, we add up the first 50,000 terms.

b Using the Taylor series for arctangent:

1 1
π = 16 arctan − 4 arctan
5 239
∞ ∞
X 1 X 1
= 16 (−1)n − 4 (−1)n

n=0
(2n + 1)52n+1 n=0
(2n + 1) · 2392n+1
∞
(−1)n
 
X 16 4
= 2n+1
−
n=0
2n + 1 5 2392n+1

This is an alternating sum, so the absolute error in using the partial sum SN
is at most:  
1 16 4
|aN +1 | = −
2N + 3 52N +3 2392N +3
So, we want to find a value of N that makes this at most 4 × 10−5 . Several
values of N are given below.

N |aN+1 | 
1 16 4
1 − ≈ 0.001
5  55 2395 
1 16 4
2 7
− ≈ 0.000029 < 4 × 10−5
7 5 2397

So, it suffices to add up the first three terms (n = 0, n = 1, and n = 2) of the

series.

1151
S OLUTIONS TO E XERCISES

c Again, we use the Taylor series for arctangent.

 
1 1 3+2 π
arctan + arctan = arctan = arctan(1) =
2 3 2·3−1 4

so that
 
1 1
π = 4 arctan + arctan
2 3
∞ ∞
X 1 X 1
=4 (−1)n + 4 (−1) n

n=0
(2n + 1)22n+1 n=0
(2n + 1)32n+1
∞  
X
n 4 1 1
= (−1) +
n=0
2n + 1 22n+1 32n+1

If we use the partial sum SN , our absolute error is at most

 
4 1 1
|aN +1 | = + .
2N + 3 22N +3 32N +3

Several of these values are given below.

N |aN+1 | 
4 1 1
1 + ≈ 0.028
5  25 35 
4 1 1
2 7
+ 7 ≈ 0.0047
7 2 3 
4 1 1
3 9
+ 9 ≈ 0.00089
9 2 3 
4 1 1
4 + ≈ 0.00018
11  211 311 
4 1 1
5 13
+ 13 ≈ 0.000038 < 4 × 10−5
13 2 3

So, it suffices to add the first six terms (n = 0 to n = 5) of the series.

Remark: if we actually wanted to approximate π this way, the series from part
(a) is probably not ideal — adding 50,000 terms sounds rough. The series from
(b) and (c) seem much more practical.
3.6.7.27. Solution. Using the Taylor series for log(1 + x):
∞
X xn
log(1 + x) = (−1)n+1
n=1
n
  X ∞
1 1
log(1.5) = log 1 + = (−1)n+1 n
2 n=1
n2

1152
S OLUTIONS TO E XERCISES

Since this is an alternating series, the error involved in using the partial sum SN
is at most
1
|aN +1 | = .
(N + 1)2N +1
We want this to be at most 5 × 10−11 .

N |aN +1 |
1
10 ≈ 4 × 10−5
11 · 211
1
15 ≈ 9.5 × 10−7
16 · 216
1
20 ≈ 2 × 10−8
21 · 221
1
25 ≈ 6 × 10−10
26 · 226
1
26 ≈ 3 × 10−10
27 · 227
1
27 ≈ 1 × 10−10
28 · 228
1
28 ≈ 6 × 10−11
29 · 229
1
29 ≈ 3 × 10−11
30 · 230

So, it suffices to add up the first 29 terms.

3.6.7.28. Solution. The Taylor Series for ex is not alternating, so we’ll use The-
orem 3.6.3 to bound the error in a partial-sum approximation. The error in the
partial-sum approximation SN is

f (N +1) (c)
EN = (x − a)N +1
(N + 1)!

for some c strictly between a and x. In our case, a = 0 and x = 1. So, we want to
find a value of N such that
f (N +1) (c) N +1 ec
(1 − 0) = < 5 × 10−11
(N + 1)! (N + 1)!

for all c in (0, 1).

If c is between 0 and 1, then ec is between 1 and e. However, since the purpose
of this problem is to approximate e precisely, it doesn’t make much sense to use
e in our bound. Since e is less than 3, then ec < 3 for all c in (0, 1). Now we can
search for an appropriate value of N .

1153
S OLUTIONS TO E XERCISES

3
N
(N + 1)!
3 1
10 = 10 ≈ 8 × 10−8
11! 9
3
11 ≈ 6 × 10−9
12!
3
12 ≈ 5 × 10−10
13!
3
13 ≈ 3 × 10−11
14!

So, it suffices to use the partial sum S13 .

3.6.7.29. Solution. The Taylor Series for log(1 − x) is not alternating, so we’ll
use Theorem 3.6.3 to bound the error in a partial-sum approximation. The error
in the partial-sum approximation SN is
f (N +1) (c)
EN = (x − a)N +1
(N + 1)!
1
for some c strictly between a and x. In our case, a = 0 and x = . So, we want
10
to find a value of N such that
 N +1
f (N +1) (c) 1
< 5 × 10−11
(N + 1)! 10
1
for all c in (0, 10 ).
To find this N , we to know f (N +1) (x). Just like when we create a Taylor polyno-
mial from scratch, we’ll differentiate f (x) several times, and look for a pattern.

−2(3)(4)(5)
f (x) = log(1 − x) f (6) (x) =
(1 − x)6
−1 −2(3)(4)(5)(6)
f 0 (x) = f (7) (x) =
1−x (1 − x)7
−1 ..
f 00 (x) = .
(1 − x)2
−2 −N !
f 000 (x) = f (N +1) (x) =
(1 − x)3 (1 − x)N +1
−2(3)
f (4) (x) =
(1 − x)4
−2(3)(4)
f (5) (x) =
(1 − x)5
1
Now we want a reasonable bound on f (N +1) (c), when c is in (0, 10 ). Note that in
this range, 1 − c > 0.
1
0<c<
10

1154
S OLUTIONS TO E XERCISES

9
⇒ <1−c<1
10
 N +1
9
⇒ < (1 − c)N +1 < 1
10
 N +1
1 10
⇒ 1< <
(1 − c)N +1 9
 N +1
N! 10
⇒ N! < < N!
(1 − c)N +1 9

This bound provides us with a “worst-case scenario” error. We don’t know ex-
actly what c is, but we don’t need to — the bound above holds for all c between
1
0 and 10 .
Now we’re ready to choose an N that results in a sufficiently small error bound.

N +1
N +1  N +1
f (N +1) (c) N ! 10

1 9 1
<
(N + 1)! 10 (N + 1)! 10
1
=
9N +1 · (N + 1)
1
So, we want: < 5 × 10−11
9N +1 · (N + 1)

To find an appropriate N , we test several values.

1
N
9N +1 · (N + 1)
1 1
8 = ≈ 3 × 10−10
9 · 99 910
1
9 10
≈ 3 × 10−11
10 · 9

So, it suffices to use the partial sum S9 .

3.6.7.30. Solution. We’ll use Theorem 3.6.3 to bound the error of a partial-sum
approximation. The error in the partial-sum approximation SN is

f (N +1) (c)
EN = (x − a)N +1
(N + 1)!

for some c strictly between a and x. In our case, a = 0 and x is in (−2, 1). So, we
want to find a value of N such that
f (N +1) (c)
(x)N +1 < 5 × 10−11
(N + 1)!

for all x in (−2, 1), and all c in (−2, 1).

To find this N , we to know f (N +1) (x). Just like when we create a Taylor polyno-

1155
S OLUTIONS TO E XERCISES

mial from scratch, we’ll differentiate f (x) several times, and look for a pattern.

ex − e−x
f (x) = sinh(x) =
2
ex
+ e−x
f 0 (x) =
2
ex
− e−x
f 00 (x) =
2
000 e + e−x
x
f (x) =
2
x −x
That is, even derivatives of f (x) are f (x), and odd derivatives of f (x) are e +e2
(which, incidentally, is the function called cosh x).
Now we want a reasonable bound on f (N +1) (c), when c is in (−2, 1). Since powers
x −x x −x
of e are always positive, we begin by noting that 0 < e −e 2
< e +e2
. So, all
ex +e−x
derivatives of f (x) are bounded above by 2 .
−2 < c < 1
⇒ e−2 < ec < e and e−1 < e−c < e2
ec + e−c e2 + e2
⇒ f (N +1) (c) < < = e2 < 9
2 2
This bound provides us with a “worst-case scenario” error. We don’t know ex-
actly what c is, but we don’t need to — the bound above holds for all c between
−2 and 1.
We also don’t know exactly what x will be, only that it’s between −2 and 1. So,
we note |x|N +1 < 2N +1 .
Now we’re ready to choose an N that results in a sufficiently small error bound.

f (N +1) (c) N +1 9 · 2N +1
(x) <
(N + 1)! (N + 1)!
N +1
9·2
So, we want: < 5 × 10−11
(N + 1)!
To find an appropriate N , we test several values.

9 · 2N +1
N
(N + 1)!
9 · 211
10 ≈ 5 × 10−4
(11)!
9 · 216
15 ≈ 3 × 10−8
(16)!
9 · 218
17 ≈ 4 × 10−10
(18)!
9 · 219
18 ≈ 4 × 10−11 < 5 × 10−11
(19)!

1156
S OLUTIONS TO E XERCISES

So, it suffices to use the partial sum S18 .

3.6.7.31. Solution. We’ll use Theorem 3.6.3 to bound the error in a partial-sum
approximation. The error in the partial-sum approximation SN is

f (N +1) (c)
EN = (x − a)N +1
(N + 1)!
1 1
for some c strictly between a and x. In our case, a = , x = − , and we are given
2 3
the nth derivative of f (x):
7
f (7) (c)

1 1
E6 = − −
7! 3 2
 7
1 6!  −7 6 −7  5
= · (1 − c) + (−1) (1 + c) −
7! 2 6
7
−5  −7 −7 
= · (1 − c) + (1 + c)
14 · 67
for some c in (− 31 , 12 ).
We want to provide actual numeric bounds for this expression. That is, we want
to find the absolute max and min of
−57  −7 −7 
E(c) = · (1 − c) + (1 + c)
14 · 67
over the interval − 31 , 12 . Absolute extrema occur at endpoints and critical

points. So, we’ll start by
differentiating E(c), and finding its critical points (if
any) in the interval − 13 , 12 .

−57  −7 −7 
E(c) = · (1 − c) + (1 + c)
14 · 67
−57  −8 −8 
E 0 (c) = · 7 (1 − c) − 7 (1 + c) =0
14 · 67
(1 − c)−8 = (1 + c)−8
1−c=1+c
c=0

Since E 0 (c) is defined over our entire interval, its only critical point is c = 0.

−57
• E(0) = [2]
14 · 67
−57 h 4
−7
i
2 −7 −57 h 3
7
i
3 7
• E − 13 =


+ = +
14 · 67 3 3
14 · 67 4 2

−57 h 1
−7
i
3 −7 −57 h 7
i
2 7
• E 21 =


+ = 2 +
14 · 67 2 2
14 · 67 3

1157
S OLUTIONS TO E XERCISES

We want to decide which of these numbers is biggest, and which smallest.

Note that 27 is much, much bigger than (3/2)7 , and both (3/4)7 and (2/3)7 are
less than one. Furthermore, (3/2)7 is much larger than 2. So: [27 + (2/3)7 ] >
[(3/2)7 + (3/4)7 ] > 2. Therefore,
"  7 # "   7 #
7
−57 7 2 −5 7
3 3 −57
2 + < + < [2]
14 · 67 3 14 · 67 4 2 14 · 67

We conclude that the error E6 is in the interval

"  7 # !
−57 7 2 −57
2 + , [2]
14 · 67 3 14 · 67

or, equivalently,

−57 −57
   
1
1+ 7 ,
14 · 37 3 7 · 67
which is approximately (−0.199, −0.040).

Exercises — Stage 3
3.6.7.32. *. Solution. Using the Maclaurin series expansions of cos x and ex ,

x2 x4
cos x = 1 − + + ···
2! 4!
x2 x4
1 − cos x = − + ···
2! 4!
x 2 x3
ex = 1 + x + + + ···
2! 3!
x2 x3
1 + x − ex = − − + ···
2! 3!
x2 4 1 2
1 − cos x 2!
− x4! + · · · 2!
− x4! + · · ·
= x2 x3 = 1
1 + x − ex − 2! − 3! + · · · − 2! − 3!x + · · ·

we have 2
1 1
1 − cos x 2!
− x4! + · · · 2!
lim = lim = = −1
x→0 1 + x − ex x→0 − 1 − x + · · · − 2!1
2! 3!

3.6.7.33. *. Solution. Using the Maclaurin series expansion of sin x,

x3 x5 x7
sin x = x − + − + ···
3! 5! 7!
x3 x5 x7
sin x − x + = − + ···
6 5! 7!
3
sin x − x + x6 1 x2
= − + ···
x5 5! 7!

1158
S OLUTIONS TO E XERCISES

we have 3
sin x − x + x6 1 x2  1 1
lim 5
= lim − + ··· = =
x→0 x x→0 5! 7! 5! 120
Remark: to solve this using l’Hôpital’s rule we would differentiate five times,
making series a practical alternative.
3.6.7.34. Solution. Our limit has the indeterminate form 1∞ ; as with l’Hôpital’s
rule, we can change it to a friendlier form using the natural logarithm.

2/x
f (x) = 1 + x + x2
h
2/x i 2
log(f (x)) = log 1 + x + x2 = log 1 + x + x2


x
∞
X (−1)n+1 y n
Recall log(1 + y) = , and set y = x + x2 . The series converges when
n=1
n
|y| < 1, and since we only consider values of x that are very close to 0, we can
assume |x + x2 | < 1.
∞
2 2

2X (−1)n+1 (x + x2 )n
log(f (x)) = log 1 + (x + x ) =
x x n=1 n
(x + x2 )2 (x + x2 )3
 
2 2
= (x + x ) − + − ···
x 2 3
(x + x2 )2 (x + x2 )3
= 2 + 2x − + − ···
2x 3x
(x2 + x)(1 + x) (x2 + x)2 (1 + x)
= 2 + 2x − + − ···
2 3
so that

lim log(f (x))

x→0
(x2 + x)(1 + x) (x2 + x)2 (1 + x)
 
= lim 2 + 2x − + − ···
x→0 2 3
= 2 + 0 + 0··· = 2

and

lim f (x) = lim elog f (x) = e2

x→0 x→0

3.6.7.35. Solution. We have an indeterminate form 1∞ . We can use a natural

logarithm to change this to a friendlier form. Furthermore, to avoid negative
1
powers, we substitute y = . As x grows larger and larger, y gets closer and
2x
closer to zero, while staying positive.
 x   
1 1 1
log 1 + = x log 1 + = log(1 + y)
2x 2x 2y

1159
S OLUTIONS TO E XERCISES

∞
1 X (−1)n+1 n
= y
2y n=1 n
y2 y3 y4
 
1
= y− + − + ···
2y 2 3 4
1 y y2 y3
 
= − + − + ···
2 4 6 8
x 
1 y y2 y3
  
1 1
lim log 1 + = lim+ − + − + ··· =
x→∞ 2x y→0 2 4 6 8 2
x 
√

1
lim 1+ = e1/2 = e
x→∞ 2x
3.6.7.36. Solution. The factor (n + 1)(n + 2) reminds us of a derivative. Start
with the geometric series.
∞
1 X
= xn
1 − x n=0
  (∞ )
d 1 d X n
= x
dx 1 − x dx n=0
∞ ∞
1 X
n−1
X
2
= nx = nxn−1
(1 − x) n=0 n=1
  (∞ )
d 1 d X
2
= nxn−1
dx (1 − x) dx n=1
∞ ∞
2 X
n−2
X
= n(n − 1)x = n(n − 1)xn−2
(1 − x)3 n=1 n=2
∞
X
= (n + 2)(n + 1)xn
n=0

Let x = 71 . Then |x| < 1, so our series converges.

∞  n
2 X 1
= (n + 2)(n + 1)
(1 − 1/7)3 n=0
7
∞
2 X (n + 2)(n + 1)
=
(6/7)3 n=0
7n

3.6.7.37. Solution. Recall the Taylor series for arctangent is:

∞
X x2n+1
arctan x = (−1)n
n=0
2n + 1

There are similarities between this and our given series: skipping powers of x,
and a denominator that’s not factorial. We’ll try to manipulate it to look like our

1160
S OLUTIONS TO E XERCISES

series. First, we antidifferentiate, to get a factor of (2n + 2) on the bottom.

∞
x2n+2
Z X
arctan x dx = (−1)n +C
n=0
(2n + 1)(2n + 2)

We can find the antiderivative of arctangent using integration by parts. Let u =

1
arctan x and dv = dx; then du = 1+x 2 dx and v = x.

x
Z Z
arctan x dx = x arctan x − dx + C
1 + x2

Now, we use the substitution w = 1 + x2 , dw = 2xdx.

1
log(1 + x2 ) + C
= x arctan x −
2
∞ 2n+2
X x 1
So, (−1)n = x arctan x − log(1 + x2 ) + C
n=0
(2n + 1)(2n + 2) 2

To find C, we evaluate both sides of the equation at x = 0.

1
log(1) + C = C
0 = 0 arctan 0 −
2
∞
X x2n+2 1
Therefore, (−1) n
= x arctan x − log(1 + x2 )
n=0
(2n + 1)(2n + 2) 2

Multiplying both sides by x2 ,

∞
X x2n+4 x2
(−1)n = x3 arctan x − log(1 + x2 )
n=0
(2n + 1)(2n + 2) 2

3.6.7.38. Solution.

(a) We’ll start, as we usually do, by finding a pattern for f (n) (0).

f (x) = (1 − x)−1/2
1
f 0 (x) = (1 − x)−3/2
2
1·3
f 00 (x) = 2 (1 − x)−5/2
2
1 ·3·5
f 000 (x) = (1 − x)−7/2
23
1·3·5·7
f (4) (x) = (1 − x)−9/2
24
..
.
1 · 3 · 5 · . . . · (2n − 1)
f (n) (x) = (1 − x)−(2n+1)/2
2n

1161
S OLUTIONS TO E XERCISES

1 · 3 · 5 · . . . · (2n − 1)
f (n) (0) =
2n
We could leave it like this, but we simplify, to make our work cleaner later
on.
1 (2n)!
= n
·
2 2 · 4 · 6 · . . . · (2n)
1 (2n)!
= n· n
2 2 · n!
(2n)!
= 2n
2 n!
This pattern holds for n ≥ 0. Now, we can write our Maclaurin series for
f (x).
∞ ∞
X f (n) (0) n X (2n)! n
(1 − x)−1/2 = x = 2n (n!)2
x
n=0
n! n=0
2
To find the radius of convergence, we use the ratio test.
an+1 (2n + 2)! 22n (n!)2
= 2n+2 · · |x|
an 2 ((n + 1)!)2 (2n)!
2
22n

(2n + 2)! n!
= · 2n+2 |x|
(2n)! (n + 1)! 2
 2
1 1
= (2n + 2)(2n + 1) · |x|
n+1 4
2
4n + 4n + 2
= 2 |x|
4n + 8n
 2 + 4 
an+1 4n + 4n + 2
lim = lim |x| = |x|
n→∞ an n→∞ 4n2 + 8n + 4

So, the radius of convergence is R = 1.

1
(b) We note the derivative of the arcsine function is √ = f (x2 ). With this
1 − x2
insight, we can manipulate our Taylor series for f (x) into a Taylor series for
arcsine.
∞
1 X (2n)!
√ = 2n 2
xn
1 − x n=0 2 (n!)
∞
1 X (2n)!
√ = 2n (n!)2
x2n
1 − x2 n=0
2
∞
!
1 (2n)! 2n
Z Z X
√ dx = x dx
1 − x2 n=0
22n (n!)2
∞
X (2n)!
arcsin x = x2n+1 + C
n=0
22n 2
(n!) (2n + 1)

1162
S OLUTIONS TO E XERCISES

∞
X (2n)!
arcsin x = x2n+1
n=0
22n (n!)2 (2n + 1)

where we found the value of C by setting x = 0. Its radius of convergence

is also 1, by Theorem 3.5.13.
3.6.7.39. *. Solution. We use that
∞
X yn
log(1 + y) = (−1)n−1 for all − 1 < y ≤ 1
n=1
n

x−2
with y = to give
2
  
x−2
log(x) = log(2 + x − 2) = log 2 1 +
2
∞
 x − 2 X (−1)n−1
= log 2 + log 1 + = log 2 + n
(x − 2)n
2 n=1
n 2

It converges when −1 < y ≤ 1, or equivalently, 0 < x ≤ 4.

3.6.7.40. *. Solution. (a) Using the geometric series expansion with r = −t4 ,
∞ ∞ ∞
1 X 1 X
4 n
X
= rn =⇒ = (−t ) = (−1)n t4n
1 − r n=0 1 + t4 n=0 n=0

Substituting this into our integral,

x
1
Z
I(x) = dt
0 1 + t4
∞
!
Z x X
= (−1)n t4n dt
0 n=0
" ∞
#t=x
X t4n+1 n
= (−1)
n=0
4n + 1
t=0
∞
X x4n+1
= (−1)n
n=0
4n + 1

(b) Substituting in x = 12 ,
∞
X 1
I(1/2) = (−1)n
n=0
(4n + 1)24n+1
1 1 1 1
= − 5
+ 9
− + ···
2 5×2 9×2 13 × 213
= 0.5 − 0.00625 + 0.000217 − 0.0000094 + · · ·

1163
S OLUTIONS TO E XERCISES

= 0.493967 − 0.0000094 + · · ·

See part (c) for the error analysis.

(c) The series for I(x) is an alternating series (that is, the sign alternates) with
successively smaller terms that converge to zero. So the error introduced by trun-
cating the series is between zero and the first omitted term. In this case, the first
omitted term was negative (−0.0000094). So the exact value of I(1/2) is the ap-
proximate value found in part (b) plus a negative number whose magnitude is
smaller than 0.00001 = 10−5 . So the approximate value of part (b) is larger than
the true value of I(1/2).
3.6.7.41. *. Solution. Expanding the exponential using its Maclaurin series,
Z 1 ∞ Z 1 2 n ∞
4 (−x ) (−1)n 1 2n+4
Z
4 −x2
X X
I= xe dx = x dx = x dx
0 n=0 0 n! n=0
n! 0
∞
X (−1)n 1 1 1 1
= = − + − +···
n=0
n!(2n + 5) |{z}
5 7
|{z} 18
|{z} 3!(11)
| {z }
n=0 n=1 n=2 n=3

The signs of successive terms in this series alternate. Futhermore the magnitude
of the nth term decreases with n. Hence, by the alternating series test, I lies
between 51 − 17 + 18
1
and 15 − 71 + 18
1 1
− 3!(11) . So

1 1
|I − a| ≤ =
3!(11) 66
3.6.7.42. *. Solution. Expanding the exponential using its Taylor series,
Z 1 ∞ Z 1 2 n ∞ Z 1
2
2 −x2
X 2
2 (−x )
X (−1)n 2 2n+2
I= xe dx = x dx = x dx
0 n=0 0
n! n=0
n! 0
∞
X (−1)n 1
= 2n+3
n=0
n!(2n + 3) 2

The signs of successive terms in this series alternate. Futhermore the magnitude
of the nth term decreases with n. Hence, by the alternating series test, I lies
N +1
NP
P (−1)n 1 (−1)n 1
between n!(2n+3) 22n+3 and n!(2n+3) 22n+3
, for every N . The first few terms
n=0 n=0
are, to five decimal places,

n 0 1 2 3
n
(−1) 1
22n+3
0.04167 -0.00625 0.00056 -0.00004
n!(2n + 3)

Allowing for a roundoff error of 0.000005 in each of these, I must be between

0.04167 − 0.00625 + 0.00056 + 0.000005 × 3 = 0.035995

1164
S OLUTIONS TO E XERCISES

and

0.04167 − 0.00625 + 0.00056 − 0.00004 − 0.000005 × 4 = 0.035920

where the multiples of 0.000005 are the maximum possible accumulated roundoff
errors in the added terms.
3.6.7.43. *. Solution. (a) Using the Taylor series expansion of ex with x = −t,
∞ ∞
−t
X (−t)n −t
X tn
e = =⇒ e −1= (−1)n
n=0
n! n=1
n!
−t ∞
e −1 X tn−1
=⇒ = (−1)n
t n=1
n!

Substituting this into our integral,

x ∞ x n−1 ∞
e−t − 1 t n x
n
Z X Z X
I(x) = dt = (−1)n dt = (−1)
0 t n=1 0 n! n=1
n · n!

(b) Substituting in x = 1,
∞
X 1
I(1) = (−1)n
n=1
n · n!
1 1 1 1
= −1 + − + − + ···
2 · 2! 3 · 3! 4 · 4! 5 · 5!
= −1 + 0.25 − 0.0556 + 0.0104 − 0.0017 + · · · = −0.80

See part (c) for the error analysis.

(c) The series for I(x) is an alternating series (that is, the sign alternates) with suc-
cessively smaller terms that converge to zero. So the error introduced by trun-
cating the series is no larger than the first omitted term. So the magnitude of
− 515! + · · · is no larger than 0.0017. Allowing for a roundoff error of at most 0.0001
in each of the two terms −0.0556 + 0.0104

I(1) = −1 + 0.25 − 0.0556 + 0.0104 ± 0.0019 = −0.7952 ± 0.0019

3.6.7.44. *. Solution. (a) Using the Taylor series expansion of sin x with x = t,
∞ ∞
X t2n+1 n sin t X t2n
sin t = (−1) =⇒ = (−1)n
n=0
(2n + 1)! t n=0
(2n + 1)!

So
x ∞ x
sin t t2n
Z X Z
Σ(x) = dt = (−1)n dt
0 t n=0 0 (2n + 1)!
∞
X x2n+1
= (−1)n
n=0
(2n + 1)(2n + 1)!

1165
S OLUTIONS TO E XERCISES

(b) The critical points of Σ(x) are the solutions of Σ0 (x) = 0. By the fun-
damental theorem of calculus Σ0 (x) = sinx x , so the critical points of Σ(x) are
x = ±π, ±2π, · · ·. The absolute maximum occurs at x = π.
(c) Substituting in x = π,
∞
X π 2n+1
Σ(π) = (−1)n
n=0
(2n + 1)(2n + 1)!
π3 π5 π7
=π− + − + ···
3 · 3! 5 · 5! 7 · 7!
= 3.1416 − 1.7226 + 0.5100 − 0.0856 + 0.0091 − 0.0007 + · · ·

The series for Σ(π) is an alternating series (that is, the sign alternates) with succes-
sively smaller terms that converge to zero. So the error introduced by truncating
the series is no larger than the first omitted term. So

Σ(π) = 3.1416 − 1.7226 + 0.5100 − 0.0856 + 0.0091 = 1.8525

with an error of magnitude at most 0.0007 + 0.0005 (the 0.0005 is the maximum
possible accumulated roundoff error in all five retained terms).
3.6.7.45. *. Solution. (a) Using the Taylor series expansion of cos t,

t2 t4 t6
cos t = 1 − + − + ···
2! 4! 6!
∞
X t2n
= (−1)n
n=0
(2n)!
cos t − 1 1 t2 t4
= − + − + ···
t2 2! 4! 6!
∞
X t2n−2
= (−1)n
n=1
(2n)!
x
cos t − 1 x x3 x5
Z
I(x) = dt = − + − + ···
0 t2 2! 4!3 6!5
∞
X x2n−1
= (−1)n
n=1
(2n)!(2n − 1)

(b), (c) Substituting in x = 1,

1 1 1
I(1) = − + − + ···
2 4!3 6!5
= −0.5 + 0.0139 − 0.0003 − · · ·
= −0.486 ± 0.001

The series for I(1) is an alternating series with decreasing successive terms that
converge to zero. So approximating I(1) by − 12 + 4!3
1
introduces an error between
1 1 1
0 and − 6!5 . Hence I(1) < − 2 + 4!3 .

1166
S OLUTIONS TO E XERCISES

3.6.7.46. *. Solution. (a) Using the Taylor series expansions of sin x and cos x
with x = t,
∞
X t2n+1
n
sin t = (−1)
n=0
(2n + 1)!
t3 t5 t7
=t− + − + ···
3! 5! 7!
∞
X
n t2n+2
t sin t = (−1)
n=0
(2n + 1)!
t4 t6 t8
= t2 −
+ − + ···
3! 5! 7!
∞
X t2n
=− (−1)n
n=1
(2n − 1)!
∞
X t2n
cos t = (−1)n
n=0
(2n)!
t2 t4 t6 t8
=1− + − + + ···
2! 4! 6! 8!
∞
X t2n
cos t − 1 = (−1)n
n=1
(2n)!
t2 t4 t6 t8
=− + − + + ···
2! 4! 6! 8!
∞ ∞
X t2n X t2n
cos t + t sin t−1 = (−1)n − (−1)n
n=1
(2n)! n=1 (2n − 1)!
∞  
X
n 2n 1 1
= (−1) t −
n=1
(2n)! (2n − 1)!
 1 2 1 1 4
= 1− t − − t + ···
2! 3! 4!
∞  
X
n 2n 1 2n
= (−1) t −
n=1
(2n)! (2n)!
∞  
X
n 2n 1 − 2n
= (−1) t
n=1
(2n)!
2 1 2 4 1 4
= − t − − t + ···
2! 2! 4! 4!
∞  
X
n+1 2n 2n − 1
= (−1) t
n=1
(2n)!
1 2 3 4 5 6 7 8
= t − t + t − t + ···
2! 4! 6! 8!
∞  
cos t + t sin t − 1 X n+1 2n−2 2n −1
= (−1) t
t2 n=1
(2n)!

1167
S OLUTIONS TO E XERCISES

1 3 5 7
= t − t2 + t4 − t6 + · · ·
2! 4! 6! 8!
Now, we’re ready to integrate.
Z x 
cos t + t sin t − 1
I(x) =
0 t2
Z x X ∞  !
2n − 1
= (−1)n+1 t2n−2 dt
0 n=1
(2n)!
"∞ #x
2n−1
X t
= (−1)n+1
n=1
(2n)!
0
∞ 2n−1
X x
= (−1)n+1
n=1
(2n)!

(b)
1 1 1 1
I(1) = − + − + ···
2! 4! 6! 8!
= 0.5 − 0.0416̇ + 0.00139 − 0.000024 + · · ·
= 0.460

The error analysis is in part (c).

(c) The series for I(1) is an alternating series with decreasing successive terms
that convege to zero. So approximating I(1) by 2!1 − 4!1 + 6!1 introduces an error
between 0 and − 8!1 . So I(1) < 2!1 − 4!1 + 6!1 < 0.460.
3.6.7.47. *. Solution. (a) Substituting x = −t into the known power series
2 3 4
ex = 1 + x + x2! + x3! + x4! + · · ·, we see that:

t2 t3 t4
e−t = 1 − t + − + − ···
2! 3! 4!
t2 t3 t4
1 − e−t = t − + − + · · ·
2! 3! 4!
1 − e−t t t2 t3
= 1 − + − + ···
t 2! 3! 4!
1 − e−t x2 x3 x4
Z
dt = C + x − + − + ···
t 2 · 2! 3 · 3! 4 · 4!

Finally, f (0) = 0 (since f (0) is an integral from 0 to 0) and so C = 0. Therefore

Z x
1 − e−t x2 x3 x4
f (x) = dt = x − + − + ··· .
0 t 2 · 2! 3 · 3! 4 · 4!
∞
x
X xn
We can also do this calculation entirely in summation notation: e = , and
n=0
n!

1168
S OLUTIONS TO E XERCISES

so
∞ ∞
−t
X (−t)n X (−1)n tn
e = =1+
n=0
n! n=1
n!
∞ ∞
X (−1)n tn X (−1)n−1 tn
1 − e−t = − =
n=1
n! n=1
n!
−t ∞
1−e X (−1)n−1 tn−1
=
t n=1
n!
x ∞
1 − e−t (−1)n−1 xn
Z X
f (x) = dt =
0 t n=1
n · n!

(−1)n−1 n
(b) We set an = An xn = x and apply the ratio test.
n · n!
an+1 (−1)n xn+1 /((n + 1) · (n + 1)!)
lim = lim
n→∞ an n→∞ (−1)n−1 xn /(n · n!)
 n+1 
|x| n · n!
= lim
n→∞ |x|n (n + 1) · (n + 1)!
 
n
= lim |x| since (n + 1)! = (n + 1) n!
n→∞ (n + 1)2
=0

This is smaller than 1 no matter what x is. So the series converges for all x.
3.6.7.48. *. Solution.
x2 x3
ex = 1 + x + + + ··· ≥ 1 + x for all x ≥ 0
2! 3!
=⇒ ex − 1 ≥ x
x3 x3
=⇒ x ≤ = x2
e −1 x
Z 1 Z 1
x3 2 1
=⇒ x
dx ≤ x dx =
0 e −1 0 3
∞ n
x
3.6.7.49. *. Solution. (a) We know that ex =
P
n!
for all x. Replacing x by −x,
n=0
∞
(−x)n
we also have e−x =
P
n!
for all x and hence
n=0

∞ ∞ ∞
1 x −x
 1h X xn X (−x)n i X xn
cosh(x) = e + e = + =
2 2 n=0 n! n=0 n! n=0
n!
n even
∞
X x2n
=
n=0
(2n)!

1169
S OLUTIONS TO E XERCISES

for all x. In particular, the interval of convergence is all real numbers.

(b) Using the power series expansion of part (a),
∞ ∞
22 24 X 22n 2 X 22n
cosh(2) = 1 + + + =3 +
2! 4! n=3 (2n)! 3 n=3 (2n)!
P∞ 22n 22n
So it suffices to show that n=3 (2n)! ≤ 0.1. Let’s write bn = (2n)!
. The first term in
P∞ 22n
n=3 (2n)! is
26 26 4
b3 =
= =
6! 6×5×4×3×2 45
P∞ 22n
The ratio between successive terms in n=3 (2n)! is

bn+1 22n+2 /22n 4

= =
bn (2n + 2)!/(2n)! (2n + 2)(2n + 1)
4 1
≤ = for all n ≥ 3
8×7 14
Hence
3 b 4 5 b ≤
6 b ≤ b ≤
∞
z}|{ z }| { z }| { z }| {
2n
X 2 4 4 1 4 1 4 1
≤ + × + × + × +···
n=3
(2n)! 45 45 14 45 142 45 143
4 1 4 14 56 1
= 1 = = <
45 1 − 14 45 13 585 10

(c) Comparing
∞ ∞ n
X t2n X (t2 )
cosh(t) = =
n=0
(2n)! n=0 (2n)!
∞ n ∞ n
1 2
t
X ( 1 t2 )
2
X (t2 )
and e 2 = =
n=0
n! n=0
2n n!

we see that it suffices to show that (2n)! ≥ 2n n!. Now. for all n ≥ 1,

n factors n factors
z }| {z }| {
(2n)! = 1 × 2 × · · · × n (n + 1) × (n + 2) × · · · × 2n
n factors n factors
z }| {z }| {
≥ 1 × 2 × ··· × n2 × 2 × ··· × 2
= 2n n!
3.6.7.50. Solution.

(a) For Newton’s method, recall we approximate a root of the function g(x) in
iterations: given an approximation xn , our next approximation is xn+1 =

1170
S OLUTIONS TO E XERCISES

g(xn )
xn − . In our case,
g 0 (xn )

x3 − 2
 
2 1
xn+1 = xn − n 2 = xn + 2 .
3xn 3 xn

We want to start somewhere reasonably close to the actual root we want, so

let’s set x0 = 1. (Your starting point may vary.)
 
2 1 4
x0 = 1 =⇒ x1 = 1+ =
3 1 3
≈ 1.3333
 
4 2 4 9 91
x1 = =⇒ x2 = + =
3 3 3 16 72
≈ 1.2639
2 91 722
 
91 1126819
x2 = =⇒ x3 = + 2 =
72 3 72 91 894348
≈ 1.2599
8943482
 
1126819 2 1126819
x3 = =⇒ x4 = +
894348 3 894348 11268192
≈ 1.2599
√
3
So, 2 ≈ 1.26.

(b) We’ll evaluate the given series at x = 2. This yields the series
∞
√ 1 X (2)(5)(8) · · · (3n − 4)
(−1)n−1
3
2=1+ + .
6 n=2 3n n!

This series is alternating, so if we use the partial sum SN , our absolute error
is at most
(2)(5)(8) · · · (3N − 1)
|aN +1 | =
3N +1 (N + 1)!
(if N ≥ 2). We want to know which value of N makes this at most 0.01. We
test several values.

1171
S OLUTIONS TO E XERCISES

N |aN +1 |
(2)(5)(8)
3 ≈ 0.04
34 · 4!
(2)(5)(8)(11)
4 ≈ 0.03
35 5!
(2)(5)(8)(11)(14)
5 ≈ 0.023
36 6!
(2)(5)(8)(11)(14)(17)
6 ≈ 0.019
37 7!
(2)(5)(8)(11)(14)(17)(20)
7 ≈ 0.016
38 8!
(2)(5)(8)(11)(14)(17)(20)(23)
8 ≈ 0.013
39 9!
(2)(5)(8)(11)(14)(17)(20)(23)(26)
9 ≈ 0.012
310 10!
(2)(5)(8)(11)(14)(17)(20)(23)(26)(29)
10 ≈ 0.0103
311 11!
(2)(5)(8)(11)(14)(17)(20)(23)(26)(29)(32)
11 ≈ 0.009
312 12!

So, the approximation S11 has a sufficiently small error. That is, we would
add up the first twelve terms.
3.6.7.51. Solution. Our plan is as follows:

• Make a Taylor series for f (x)

• Calculate the tenth derivative of the Taylor series of f (x).

• Decide how many terms we need to add to achieve the desired accuracy.

• Approximate f (10) 15 with a partial sum.

∞ 2n+1
n x
X
We know that the Taylor series for arctan x is (−1) , which converges
n=0
2n + 1
for −1 ≤ x ≤ 1. So, the Taylor series for arctan(x3 ) is
∞ 3 2n+1 ∞
X
3 n (x )
X x6n+3
f (x) = arctan(x ) = (−1) = (−1)n
n=0
2n + 1 n=0
2n + 1

It is much easier to differentiate this series many times than it is to differentiate

arctan(x3 ) directly many times.
∞
0
X
n (6n + 3)x6n+2
f (x) = (−1)
n=0
2n + 1
∞
X (6n + 3)(6n + 2)x6n+1
f 00 (x) = (−1)n
n=0
2n + 1

1172
S OLUTIONS TO E XERCISES

∞
000
X (6n + 3)(6n + 2)(6n + 1)x6n
f (x) = (−1)n
n=0
2n + 1
..
.
∞
(10)
X
n (6n + 3)(6n + 2)(6n + 1) · · · (6n − 6)x6n−7
f (x) = (−1)
n=0
2n + 1
∞
X (6n + 3)(6n + 2)(6n + 1) · · · (6n − 6)x6n−7
= (−1)n
n=2
2n + 1
∞
X (6n + 3)!
= (−1)n x6n−7
n=2
(2n + 1)(6n − 7)!
  X ∞
(10) 1 (6n + 3)!
f = (−1)n
5 n=2
(2n + 1)(6n − 7)! · 56n−7

(Notice, after ten differentiations, the terms a0 and a1 are both zero.)
Since this is an alternating series, the absolute error involved in using the ap-
proximation SN is at most

(6N + 9)!
|aN +1 | =
(2N + 3)(6N − 1)! · 56N −1

By testing a few values of N , we find

39!
|a6 | = |a5+1 | = 29
≈ 0.00000095 < 10−6
(13)(29!) · 5

So, S5 is a sufficient approximation. That is,

  X 5
(10) 1 (6n + 3)!
f ≈ (−1)n
5 n=2
(2n + 1)(6n − 7)! · 56n−7
15! 21! 27!
= (−1)2 5
+ (−1)3 11
+ (−1)4
5 · 5! · 5 7! · 11! · 5 9! · 17! · 517
33!
+ (−1)5
11! · 23! · 523
15! 21! 27! 33!
= 6
− 11
+ 17
−
5! · 5 7! · 11! · 5 9! · 17! · 5 11! · 23! · 523
Remark: if we had calculated f (10) (1/5) directly, using derivative rules instead of
series, we would have found an exact value; however, our value here is easier to
find, and is highly accurate (if not exact).
3.6.7.52. Solution.

(a) To sketch y = f (x), we note the following:

• f (x) is never negative.

1173
S OLUTIONS TO E XERCISES

• lim f (x) = e0 = 1, so the curve has horizontal asymptotes in both

x→±∞
directions at y = 1.
1
• lim f (x) = lim e1/x 2 = lim 1u = 0 = f (0), so the curve is continu-
x→±0 x→±0 u→+∞ e
ous at x = 0.
2
• For x 6= 0, f 0 (x) = x23 e−1/x , so our curve is decreasing on (−∞, 0) and
increasing on (0, ∞)
2
• Forpx 6= 0,pf 00 (x) = 2x−6 (2 − 3x2 )e−1/x , so our curve is concave up on
(− 2/3, 2/3), and concave down elsewhere.

1 y = f (x)

x
p p
− 2/3 2/3

(b) Since f (n) (0) = 0 for all whole n (that is, the graph is really quite flat at the
∞
X 0 n
origin), and since f (0) = 0, the Maclaurin series for f (x) is x = 0.
n=0
n!

(c) The Maclaurin series converges for all real values of x (to the constant 0).

(d) Since ey > 0 for any real y, we see f (x) = 0 only when x = 0. So, f (x) is
only equal to its Maclaurin series at the single point x = 0.

Remark: the function f (x) is an example of a function whose Maclaurin series

converges, but not to f (x)! To describe this behaviour, we say f (x) is non-analytic.
3.6.7.53. Solution.
• Solution 1: Since f (x) is odd, f (−x) = −f (x) for all x in its domain. We
plug this into our power series, then consider the even-indexed terms and
the odd-indexed terms separately.
f (−x) = −f (x)
∞ ∞
X f (n) (0) X f (n) (0) n
(−x)n = − x
n=0
n! n=0
n!
Now separating the even powered and odd powered terms
∞ ∞
X f (2n+1) (0) 2n+1
X f (2n) (0)
(−x) + (−x)2n
n=0
(2n + 1)! n=0
(2n)!
∞ ∞
X f (2n+1) (0) 2n+1 X f (2n) (0) 2n
=− x − x
n=0
n! n=0
n!

1174
S OLUTIONS TO E XERCISES

For any integer n, we have that (−1)2n = 1 and (−1)2n+1 = −1 so that

∞ ∞
X f (2n+1) (0) X f (2n) (0)
− x2n+1 + x2n
n=0
(2n + 1)! n=0
(2n)!
∞ ∞
X f (2n+1) (0) 2n+1 X f (2n) (0) 2n
=− x − x
n=0
n! n=0
n!

and
∞ ∞
X f (2n) (0) X f (2n) (0)
x2n = − x2n
n=0
(2n)! n=0
n!

and
∞
X f (2n) (0) 2n
2 x =0
n=0
(2n)!
∞
X f (2n) (0)
x2n = 0
n=0
(2n)!

• Solution 2: Alternately, we could note the following:

◦ Since all derivative of f (x) exist, all its derivatives are continuous.
◦ The derivative of an odd function is even, and the derivative of an
even function is odd.
◦ So, the even-indexed derivatives of f (x) are continuous, odd func-
tions.
◦ Every continuous, odd function passes through the origin. That is,
f (2n) (0) = 0.
◦ So, every term in the series is 0.

1175