Calculus for Scientists and Engineers II:

Math 20B at UCSD

William Stein
1
March 15, 2006
1
I am attending John Eggers lectures, which have a very strong inuence on these notes.
2
Contents
1 Preface 5
1.1 Computers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2 Denite and Indenite Integrals 7
2.1 The Denite Integral . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.1.1 The denition of area under curve . . . . . . . . . . . . . . . . . 7
2.1.2 Relation between velocity and area . . . . . . . . . . . . . . . . . 7
2.1.3 Denition of Integral . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.1.4 The Fundamental Theorem of Calculus . . . . . . . . . . . . . . 8
2.2 Indenite Integrals and Change . . . . . . . . . . . . . . . . . . . . . . . 9
2.2.1 Indenite Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.2.2 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2.3 Physical Intuition . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.3 Substitution and Symmetry . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.3.1 The Substitution Rule . . . . . . . . . . . . . . . . . . . . . . . . 12
2.3.2 The Substitution Rule for Denite Integrals . . . . . . . . . . . . 14
2.3.3 Symmetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3 Applications to Areas, Volume, and Averages 17
3.1 Using Integration to Determine Areas Between Curves . . . . . . . . . . 17
3.1.1 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.2 Computing Volumes of Surfaces of Revolution . . . . . . . . . . . . . . . 20
3.3 Average Values . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
4 Polar Coordinates and Complex Numbers 25
4.1 Polar Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.2 Areas in Polar Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . 27
4.2.1 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
4.3 Complex Numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
4.3.1 Polar Form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
4.4 Complex Exponentials and Trig Identities . . . . . . . . . . . . . . . . . 32
4.4.1 Trigonometry and Complex Exponentials . . . . . . . . . . . . . 34
5 Integration Techniques 37
5.1 Integration By Parts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
5.2 Trigonometric Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
5.2.1 Some Remarks on Using Complex-Valued Functions . . . . . . . 43
5.3 Trigonometric Substitutions . . . . . . . . . . . . . . . . . . . . . . . . . 44
5.4 Factoring Polynomials . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
3
4 CONTENTS
5.5 Integration of Rational Functions Using Partial Fractions . . . . . . . . 49
5.6 Approximating Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
5.7 Improper Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
5.7.1 Convergence, Divergence, and Comparison . . . . . . . . . . . . . 59
6 Sequences and Series 63
6.1 Sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
6.2 Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
6.3 The Integral and Comparison Tests . . . . . . . . . . . . . . . . . . . . . 66
6.3.1 Estimating the Sum of a Series . . . . . . . . . . . . . . . . . . . 70
6.4 Tests for Convergence . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
6.4.1 The Comparison Test . . . . . . . . . . . . . . . . . . . . . . . . 71
6.4.2 Absolute and Conditional Convergence . . . . . . . . . . . . . . . 72
6.4.3 The Ratio Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
6.4.4 The Root Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
6.5 Power Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
6.5.1 Shift the Origin . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
6.5.2 Convergence of Power Series . . . . . . . . . . . . . . . . . . . . 77
6.6 Taylor Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
6.7 Applications of Taylor Series . . . . . . . . . . . . . . . . . . . . . . . . 82
6.7.1 Estimation of Taylor Series . . . . . . . . . . . . . . . . . . . . . 83
7 Some Dierential Equations 85
7.1 Separable Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
7.2 Logistic Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
Chapter 1
Preface
In order to learn Calculus its crucial for you to do all the assigned problems and then
some. When I was a student and started doing well in math (instead of poorly!), the
key dierence was that I started doing an insane number of problems (e.g., every single
problem in the book). Push yourself to the limit!
1.1 Computers
I think the best way to use a computer in learning Calculus is as a sort of solutions
manual, but better. Do a problem rst by hand. Then verify correctness of your
solution. This is way better than what you get by using a solutions manual!
You can try similar problems (not in the homework) and also verify your answers.
This is like playing solitaire, but is much more creative.
You can verify key steps of what you did by hand using the computer. E.g., if
youre confused about one of part of your approach to computing an integral, you
can compare what you get with the computer. Solution manuals either give you
only the solution or a particular sequence of steps to get there, which might have
little to do with the brilliantly original strategy you invented.
For this course its most useful to have a program that does symbolic integration.
I recommend maxima, which is a fairly simple completely free and open source
program written (initially) in the 1960s at MIT. Download it for free from
http://maxima.sourceforge.net
Its not insanely powerful, but itll instantly do (something with) pretty much any
integral in this class, and a lot more. Plus if you know lisp you can read the source
code. (You could also buy Maple or Mathematica, or use a TI-89 calculator.)
Here are some maxima examples:
(%i2) integrate(x^2 + 1 + 1/(x^2+1), x);
3
x
(%o2) atan(x) + -- + x
3
5
6 CHAPTER 1. PREFACE
(%i3) integrate(sqrt(5/x), x);
(%o3) 2 sqrt(5) sqrt(x)
(%i4) integrate(sin(2*x)/sin(x), x);
(%o4) 2 sin(x)
(%i5) integrate(sin(2*x)/sin(x), x, 0, %pi);
(%o5) 0
(%i6) integrate(sin(2*x)/sin(x), x, 0, %pi/2);
(%o6) 2
Chapter 2
Denite and Indenite
Integrals
2.1 The Denite Integral
2.1.1 The denition of area under curve
Let f be a continuous function on interval [a, b]. Divide [a, b] into n subintervals of
length x = (b a)/n. Choose (sample) points x

i
in ith interval, for each i. The
(signed) area between the graph of f and the x axis is approximately
A
n
f(x

1
)x + +f(x

n
)x
=
n

i=1
f(x

i
)x.
(The

is notation to make it easier to write down and think about the sum.)
Denition 2.1.1 (Signed Area). The (signed) area between the graph of f and the
x axis between a and b is
lim
n
_
n

i=1
f(x

i
)x
_
(Note that x = (b a)/n depends on n.)
It is a theorem that the area exists and doesnt depend on the choice of x

i
.
2.1.2 Relation between velocity and area
Suppose youre reading a car magazine and there is an article about a new sports car
that has this table in it:
Time (seconds) 0 1 2 3 4 5 6
Speed (mph) 0 5 15 25 40 50 60
They claim the car drove 1/8th of a mile after 6 seconds, but this just feels wrong...
Hmmm... Lets estimate the distance driven using the formula
distance = rate time.
7
8 CHAPTER 2. DEFINITE AND INDEFINITE INTEGRALS
We overestimate by assuming the velocity is a constant equal to the max on each interval:
estimate = 5 1 + 15 1 + 25 1 + 40 1 + 50 1 + 60 1 =
195
3600
miles = 0.054...
(Note: there are 3600 seconds in an hour.) But 1/8 0.125, so the article is inconsistent.
(Doesnt this sort of thing just bug you? By learning calculus youll be able to double-
check things like this much more easily.)
Insight! The formula for the estimate of distance traveled above looks exactly like
an approximation for the area under the graph of the speed of the car! In fact, if an
object has velocity v(t) at time t, then the net change in position from time a to b is
_
b
a
v(t)dt.
Well come back to this observation frequently.
2.1.3 Denition of Integral
Let f be a continuous function on the interval [a, b]. The denite integral is just the
signed area between the graph of f and the x axis:
Denition 2.1.2 (Denite Integral). The denite integral of f(x) from a to b is
_
b
a
f(x)dx = lim
n
_
n

i=1
f(x

i
)x
_
.
Properties of Integration:

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

_
b
a
c
1
f
1
(x) +c
2
f
2
(x)dx = c
1
_
b
a
f
1
(x) +c
2
_
b
a
f
2
(x)dx. (linearity)
If f(x) g(x) on for all x [a, b], then
_
b
a
f(x)dx
_
b
a
g(x)dx.
There are many other properties.
2.1.4 The Fundamental Theorem of Calculus
Let f be a continuous function on the interval [a, b]. The following theorem is incredibly
useful in mathematics, physics, biology, etc.
Theorem 2.1.3. If F(x) is any dierentiable function on [a, b] such that F

(x) = f(x),
then
_
b
a
f(x)dx = F(b) F(a).
One reason this is amazing, is because it says that the area under the entire curve is
completely determined by the values of a (magic) auxiliary function at only 2 points.
Its hard to believe. It reduces computing (2.1.2) to nding a single function F, which
one can often do algebraically, in practice.Whether or not one should use this theorem
to evaluate an integral depends a lot on the application at hand, of course. One can
also use a partial limit via a computer for certain applications (numerical integration).
2.2. INDEFINITE INTEGRALS AND CHANGE 9
Example 2.1.4. Ive always wondered exactly what the area is under a hump of the
graph of sin. Lets gure it out, using F(x) = cos(x).
_

0
sin(x)dx = cos() (cos(0)) = (1) (1) = 2.
But does such an F always exist? The surprising answer is yes.
Theorem 2.1.5. Let F(x) =
_
t
a
f(t)dt. Then F

(x) = f(x) for all x [a, b].

Note that a nice formula for F can be hard to nd or even provably non-existent.
The proof of Theorem 2.1.5 is somewhat complicated but is given in complete detail
in Stewarts book, and you should denitely read and understand it.
Sketch of Proof. We use the denition of derivative.
F

(x) = lim
h0
F(x +h) F(x)
h
= lim
h0
_
_
x+h
a
f(t)dt
_
x
a
f(t)dt
_
/h
= lim
h0
_
_
x+h
x
f(t)dt
_
/h
Intuitively, for h suciently small f is essentially constant, so
_
x+h
x
f(t)dt hf(x) (this
can be made precise using the extreme value theorem). Thus
lim
h0
_
_
x+h
x
f(t)dt
_
/h = f(x),
which proves the theorem.
2.2 Indenite Integrals and Change
(William Stein, Math 20b, Winter 2006)
Homework: Do the following by Tuesday, January 17.
* Section 5.3: 13, 37, 55, 67
* Section 5.4: 2, 9, 13, 27, 33, 39, 45, 47, 51, 53
* Section 5.5: 11, 23, 31, 37, 41, 55, 57, 63, 65, 75, 79
The rst quiz will be on Friday, Jan 20 and will consist of two problems from this homework.
Ace the rst quiz!
2.2.1 Indenite Integrals
The notation
_
f(x)dx = F(x) means that F

(x) = f(x) on some (usually specied)

domain of denition of f(x).
10 CHAPTER 2. DEFINITE AND INDEFINITE INTEGRALS
Denition 2.2.1 (Anti-derivative). We call F(x) an anti-derivative of f(x).
Proposition 2.2.2. Suppose f is a continuous function on an interval (a, b). Then
any two antiderivatives dier by a constant.
Proof. If F
1
(x) and F
2
(x) are both antiderivatives of a function f(x), then
(F
1
(x) F
2
(x))

= F

1
(x) F

2
(x) = f(x) f(x) = 0.
Thus F
1
(x)F
2
(x) = c from some constant c (since only constant functions have slope 0
everywhere). Thus F
1
(x) = F
2
(x) +c as claimed.
We thus often write _
f(x)dx = F(x) +c,
where c is an (unspecied xed) constant.
Note that the proposition need not be true if f is not dened on a whole interval.
For example, f(x) = 1/x is not dened at 0. For any pair of constants c
1
, c
2
, the
function
F(x) =
_
ln(|x|) +c
1
x < 0,
ln(x) +c
2
x > 0,
satises F

(x) = f(x) for all x = 0. We often still just write

_
1/x = ln(|x|)+c anyways,
meaning that this formula is supposed to hold only on one of the intervals on which 1/x
is dened (e.g., on (, 0) or (0, )).
We pause to emphasize the notation dierence between denite and indenite inte-
gration.
_
b
a
f(x)dx = a specic number
_
f(x)dx = a (family of) functions
One of the main goals of this course is to help you to get really good at computing
_
f(x)dx for various functions f(x). It is useful to memorize a table of examples (see,
e.g., page 406 of Stewart), since often the trick to integration is to relate a given integral
to a known one. Integration is like solving a puzzle or playing a game, and often you win
by moving into a position where you know how to defeat your opponent, e.g., relating
your integral to integrals that you already know how to do. If you know how to do
a basic collection of integrals, it will be easier for you to see how to get to a known
integral from an unknown one.
Whenever you successfully compute F(x) =
_
f(x)dx, then youve constructed a
mathematical gadget that allows you to very quickly compute
_
b
a
f(x)dx for any a, b (in
the interval of denition of f(x)). The gadget is F(b) F(a). This is really powerful.
2.2.2 Examples
Example 2.2.3.
_
x
2
+ 1 +
1
x
2
+ 1
dx =
_
x
2
dx +
_
1dx +
_
1
x
2
+ 1
dx
=
1
3
x
2
+x + tan
1
(x) +c.
2.2. INDEFINITE INTEGRALS AND CHANGE 11
Example 2.2.4.
_
_
5
x
dx =
_

5x
1/2
dx = 2

5x
1/2
+c.
Example 2.2.5.
_
sin(2x)
sin(x)
dx =
_
2 sin(x) cos(x)
sin(x)
=
_
2 cos(x) = 2 sin(x) +c
2.2.3 Physical Intuition
In the previous lecture we mentioned a relation between velocity, distance, and the
meaning of integration, which gave you a physical way of thinking about integration.
In this section we generalize our previous observation.
The following is a restatement of the fundamental theorem of calculus:
Theorem 2.2.6 (Net Change Theorem). The denite integral of the rate of change
F

(x) of some quantity F(x) is the net change in that quantity:

_
b
a
F

(x)dx = F(b) F(a).

For example, if p(t) is the population of students at UCSD at time t, then p

(t) is
the rate of change. Lately p

(t) has been positive since p(t) is growing (rapidly!). The

net change interpretation of integration is that
_
t2
t1
p

(t)dt = p(t
2
) p(t
1
) = change in number of students from time t
1
to t
2
.
Another very common example youll seen in problems involves water ow into or
out of something. If the volume of water in your bathtub is V (t) gallons at time t (in
seconds), then the rate at which your tub is draining is V

(t) gallons per second. If you

have the geekiest drain imaginable, it prints out the drainage rate V

(t). You can use

that printout to determine how much water drained out from time t
1
to t
2
:
_
t2
t1
V

(t)dt = water that drained out from time t

1
to t
2
Some problems will try to confuse you with dierent notions of change. A standard
example is that if a car has velocity v(t), and you drive forward, then slam it in reverse
and drive backward to where you start (say 10 seconds total elapse), then v(t) is positive
some of the time and negative some of the time. The integral
_
10
0
v(t)dt is not the
total distance registered on your odometer, since v(t) is partly positive and partly
negative. If you want to express how far you actually drove going back and forth,
compute
_
10
0
|v(t)|dt. The following example emphasizes this distinction:
Example 2.2.7. An ancient dragon is pacing on the clis in Del Mar, and has velocity
v(t) = t
2
2t 8. Find (1) the displacement of the dragon from time t = 1 until time
t = 6 (i.e., how far the dragon is at time 6 from where it was at time 1), and (2) the
total distance the dragon paced from time t = 1 to t = 6.
For (1), we compute
_
6
1
(t
2
2t 8)dt =
_
1
3
t
3
t
2
8t
_
6
1
=
10
3
.
12 CHAPTER 2. DEFINITE AND INDEFINITE INTEGRALS
For (2), we compute the integral of |v(t)|:
_
6
1
|t
2
2t 8|dt =
_

_
1
3
t
3
t
2
8t
__
4
1
+
_
1
3
t
3
t
2
8t
_
6
4
= 18 +
44
3
=
98
3
.
2.3 Substitution and Symmetry
Homework reminder.
Quiz reminder: Friday, Jan 20 (Ace the rst quiz!).
Oce Hours: Tue 11-1.
Monday is a holiday!
Wednesday areas between curves and volumes
First midterm: Wed Feb 1 at 7pm (review lecture during day!)
Quick 5 minute discussion of computers and Maxima.
Quiz format: one question on front; one on back.
Remarks:
1. The total distance traveled is
R
t
2
t
1
|v(t)|dt since |v(t)| is the rate of change of F(t) =
distance traveled (your speedometer displays the rate of change of your odometer).
2. How to compute
R
b
a
|f(x)|dx.
(a) Find the zeros of f(x) on [a, b], and use these to break the interval up into subin-
tervals on which f(x) is always 0 or always 0.
(b) On the intervals where f(x) 0, compute the integral of f, and on the intervals
where f(x) 0, compute the integral of f.
(c) The sum of the above integrals on intervals is
R
|f(x)|dx.
This section is primarly about a powerful technique for computing denite and
indenite integrals.
2.3.1 The Substitution Rule
In rst quarter calculus you learned numerous methods for computing derivatives of
functions. For example, the power rule asserts that
(x
a
)

= a x
a1
.
We can turn this into a way to compute certain integrals:
_
x
a
dx =
1
a + 1
x
a+1
if a = 1.
Just as with the power rule, many other rules and results that you already know
yield techniques for integration. In general integration is potentially much trickier than
dierentiation, because it is often not obvious which technique to use, or even how to
use it. Integration is a more exciting than dierentiation!
Recall the chain rule, which asserts that
d
dx
f(g(x)) = f

(g(x))g

(x).
We turn this into a technique for integration as follows:
2.3. SUBSTITUTION AND SYMMETRY 13
Proposition 2.3.1 (Substitution Rule). Let u = g(x), we have
_
f(g(x))g

(x)dx =
_
f(u)du,
assuming that g(x) is a function that is dierentiable and whose range is an interval on
which f is continuous.
Proof. Since f is continuous on the range of g, Theorem 2.1.5 (the fundamental theorem
of Calculus) implies that there is a function F such that F

= f. Then
_
f(g(x))g

(x)dx =
_
F

(g(x))g

(x)dx
=
_ _
d
dx
F(g(x))
_
dx
= F(g(x)) +C
= F(u) +C =
_
F

(u)du =
_
f(u)du.
If u = g(x) then du = g

(x)dx, and the substitution rule simply says if you let

u = g(x) formally in the integral everywhere, what you naturally would hope to be true
based on the notation actually is true. The substitution rule illustrates how the notation
Leibniz invented for Calculus is incredibly brilliant. It is said that Leibniz would often
spend days just trying to nd the right notation for a concept. He succeeded.
As with all of Calculus, the best way to start to get your head around a new concept
is to see severally clearly worked out examples. (And the best way to actually be able to
use the new idea is to do lots of problems yourself!) In this section we present examples
that illustrate how to apply the substituion rule to compute indenite integrals.
Example 2.3.2.
_
x
2
(x
3
+ 5)
9
dx
Let u = x
3
+ 5. Then du = 3x
2
dx, hence dx = du/(3x
2
). Now substitute it all in:
_
x
2
(x
3
+ 5)
9
dx =
_
1
3
u
9
=
1
30
u
10
=
1
30
(x
3
+ 5)
10
.
Theres no point in expanding this out: only simplify for a purpose!
Example 2.3.3.
_
e
x
1 +e
x
dx
Substitute u = 1 +e
x
. Then du = e
x
dx, and the integral above becomes
_
du
u
= ln |u| = ln |1 +e
x
| = ln(1 +e
x
).
Note that the absolute values are not needed, since 1 +e
x
> 0 for all x.
14 CHAPTER 2. DEFINITE AND INDEFINITE INTEGRALS
Example 2.3.4.
_
x
2

1 x
dx
Keeping in mind the power rule, we make the substitution u = 1 x. Then du = dx.
Noting that x = 1 u by solving for x in u = 1 x, we see that the above integral
becomes
_

(1 u)
2

u
du =
_
1 2u +u
2
u
1/2
du
=
_
u
1/2
2u
1/2
+u
3/2
du
=
_
2u
1/2

4
3
u
3/2
+
2
5
u
5/2
_
= 2(1 x)
1/2
+
4
3
(1 x)
3/2

2
5
(1 x)
5/2
.
2.3.2 The Substitution Rule for Denite Integrals
Proposition 2.3.5 (Substitution Rule for Denite Integrals). We have
_
b
a
f(g(x))g

(x)dx =
_
g(b)
g(a)
f(u)du,
assuming that u = g(x) is a function that is dierentiable and whose range is an interval
on which f is continuous.
Proof. If F

= f, then by the chain rule, F(g(x)) is an antiderivative of f(g(x))g

(x).
Thus
_
b
a
f(g(x))g

(x)dx =
_
F(g(x))
_
b
a
= F(g(b)) F(g(a)) =
_
g(b)
g(a)
f(u)du.
Example 2.3.6.
_

0
xcos(x
2
)dx
We let u = x
2
, so du = 2xdx and xdx =
1
2
du and the integral becomes
1
2

_
(

)
2
(0)
2
cos(u)du =
1
2
[sin(u)]

0
=
1
2
(0 0) = 0.
2.3.3 Symmetry
An odd function is a function f(x) such that f(x) = f(x), and an even function one
for which f(x) = f(x). If f is an odd function, then for any a,
_
a
a
f(x)dx = 0.
2.3. SUBSTITUTION AND SYMMETRY 15
If f is an even function, then for any a,
_
a
a
f(x)dx = 2
_
a
0
f(x)dx.
Both statements are clear if we view integrals as computing the signed area between
the graph of f(x) and the x-axis.
Example 2.3.7.
_
1
1
x
2
dx = 2
_
1
0
x
2
dx = 2
_
1
3
x
3
_
1
0
=
2
3
.
16 CHAPTER 2. DEFINITE AND INDEFINITE INTEGRALS
Chapter 3
Applications to Areas,
Volume, and Averages
3.1 Using Integration to Determine Areas Between
Curves
Today is 2006-01-18.
Quiz reminder: Friday, Jan 20 (describe format)
How was your weekend?
Mine was greatI wrote open source math software nonstop for days on end!
This section is about how to compute the area of fairly general regions in the plane.
Regions are often described as the area enclosed by the graphs of several curves. (My
land is the plot enclosed by that river, that fence, and the highway.)
Recall that the integral
_
b
a
f(x)dx has a geometric interpretation as the signed area
between the graph of f(x) and the x-axis. We dened area by subdividing, adding up
approximate areas (use points in the intervals) as Riemann sum, and taking the limit.
Thus we dened area as a limit of Riemann sums. The fundamental theorem of calculus
asserts that we can compute areas exactly when we can nding antiderivatives.
Instead of considering the area between the graph of f(x) and the x-axis, we consider
more generally two graphs, y = f(x), y = g(x), and assume for simplicity that f(x)
g(x) on an interval [a, b]. Again, we approximate the area between these two curves as
before using Riemann sums. Each approximating rectangle has width (b a)/n and
height f(x) g(x), so
Area bounded by graphs

[f(x
i
) g(x
i
)]x.
Note that f(x) g(x) 0, so the area is nonnegative. From the denition of integral
we see that the exact area is
Area bounded by graphs =
_
b
a
(f(x) g(x))dx. (3.1.1)
Why did we make a big deal about approximations instead of just writing down
(3.1.1)? Because having a sense of how this area comes directly from a Riemann sum is
17
18 CHAPTER 3. APPLICATIONS TO AREAS, VOLUME, AND AVERAGES
very important. But, what is the point of the Riemann sum if all were going to do is
write down the integral? The sum embodies the geometric manifestation of the integral.
If you have this picture in your mind, then the Riemann sum has done its job. If you
understand this, youre more likely to know what integral to write down; if you dont,
then you might not.
Remark 3.1.1. By the linearity property of integration, our sought for area is the
dierence
_
b
a
f(x)dx
_
b
a
g(x)dx,
of two signed areas.
3.1.1 Examples
Example 3.1.2. Find the area enclosed by y = x + 1, y = 9 x
2
, x = 1, x = 2.
0
1
2
3
4
5
6
7
8
9
-1 -0.5 0 0.5 1 1.5 2
x+1
9-x**2
Figure 3.1.1: What is the enclosed area?
Area =
_
2
1
_
(9 x
2
) (x + 1)
_
dx
We have reduced the problem to a computation:
_
2
1
[(9 x
2
) (x + 1)]dx =
_
2
1
(8 x x
2
)dx =
_
8x
1
2
x
2

1
3
x
3
_
2
1
=
39
2
.
The above example illustrates the simplest case. In practice more interesting situ-
ations often arise. The next example illustrates nding the boundary points a, b when
they are not explicitly given.
Example 3.1.3. Find area enclosed by the two parabolas y = 12 x
2
and y = x
2
6.
Problem: We didnt tell you what the boundary points a, b are. We have to gure
that out. How? We must nd exactly where the two curves intersect, by setting the two
curves equal and nding the solution. We have
x
2
6 = 12 x
2
,
3.1. USING INTEGRATION TO DETERMINE AREAS BETWEEN CURVES 19
-15
-10
-5
0
5
10
15
20
-4 -2 0 2 4
12-x**2
x**2-6
Figure 3.1.2: What is the enclosed area?
so 0 = 2x
2
18 = 2(x
2
9) = 2(x 3)(x +3), hence the intersect points are at a = 3
and b = 3. We thus nd the area by computing
_
3
3
_
12 x
2
(x
2
6)

dx =
_
3
3
(18 2x
2
)dx = 4
_
3
0
(9 x
2
)dx = 4 18 = 72.
Example 3.1.4. A common way in which you might be tested to see if you really
understand what is going on, is to be asked to nd the area between two graphs x = f(y)
and x = g(y). If the two graphs are vertical, subtract o the right-most curve. Or, just
switch x and y everywhere (i.e., reect about y = x). The area is unchanged.
Example 3.1.5. Find the area (not signed area!) enclosed by y = sin(x), y = x
2
x,
and x = 2.
-1
-0.5
0
0.5
1
1.5
2
2.5
3
3.5
4
-0.5 0 0.5 1 1.5 2 2.5
sin(pi*x)
x**2-x
0
Figure 3.1.3: Find the area
Write x
2
x = (x 1/2)
2
1/4, so that we can obtain the graph of the parabola
by shifting the standard graph. The area comes in two pieces, and the upper and lower
20 CHAPTER 3. APPLICATIONS TO AREAS, VOLUME, AND AVERAGES
curve switch in the middle. Technically, what were doing is integrating the absolute
value of the dierence. The area is
_
1
0
sin(x) (x
2
x)dx
_
2
1
(x
2
x) sin(x)dx =
4

+ 1
Something to take away from this is that in order to solve this sort of problem, you
need some facility with graphing functions. If you arent comfortable with this, review.
3.2 Computing Volumes of Surfaces of Revolution
Everybody knows that the voluem of a solid box is
volume = length width height.
More generally, the volume of cylinder is V = r
2
h (cross sectional area times height).
Even more generally, if the base of a prism has area A, the volume of the prism is
V = Ah.
But what if our solid object looks like a complicated blob? How would we compute
the volume? Well do something that by now should seem familiar, which is to chop the
object into small pieces and take the limit of approximations.
[[Picture of solid sliced vertically into a bunch of vertical thin solid discs.]]
Assume that we have a function
A(x) = cross sectional area at x.
The volume of our potentially complicated blob is approximately

A(x
i
)x. Thus
volume of blob = lim
n
n

i=1
A(x
i
)x
=
_
b
a
A(x)dx
Example 3.2.1. Find the volume of the pyramid with height H and square base with
sides of length L.
For convenience look at pyramid on its side, with the tip of the pyramid at the
origin. We need to gure out the cross sectional area as a function of x, for 0 x H.
The function that gives the distance s(x) from the x axis to the edge is a line, with
s(0) = 0 and s(H) = L/2. The equation of this line is thus s(x) =
L
2H
x. Thus the cross
sectional area is
A(x) = (2s(x))
2
=
x
2
L
2
H
2
.
The volume is then
_
H
0
A(x)dx =
_
H
0
x
2
L
2
H
2
dx =
_
x
3
L
2
3H
2
_
H
0
=
H
3
L
2
3H
2
=
1
3
HL
2
.
Today: Quiz!
Next: Polar coordinates, etc.
Questions:?
Recall: Find volume by integrating cross section of area. (draw picture)
3.2. COMPUTING VOLUMES OF SURFACES OF REVOLUTION 21
Figure 3.2.1: How Big is Pharaohs Place?
Example 3.2.2. Find the volume of the solid obtained by rotating the following region
about the x axis: the region enclosed by y = x
2
and y = x
3
between x = 0 and x = 1.
0
0.2
0.4
0.6
0.8
1
0 0.2 0.4 0.6 0.8 1
x**2
x**3
0
Figure 3.2.2: Find the volume of the ower pot
The cross section is a washer, and the area as a function of x is
A(x) = (r
o
(x)
2
r
i
(x)
2
) = (x
4
x
6
).
The volume is thus
_
1
0
A(x)dx =
_
1
0
_
1
5
x
5

1
7
x
7
_
dx =
_
1
5
x
5

1
7
x
7
_
1
0
=
2
35
.
Example 3.2.3. One of the most important examples of a volume is the volume V of
a sphere of radius r. Lets nd it! Well just compute the volume of a half and multiply
by 2. The cross sectional area is
22 CHAPTER 3. APPLICATIONS TO AREAS, VOLUME, AND AVERAGES
-1
-0.5
0
0.5
1
0 0.2 0.4 0.6 0.8 1
sqrt(1-x**2)
-sqrt(1-x**2)
0
Figure 3.2.3: Cross section of a half of sphere with radius 1
A(x) = r(x)
2
= (
_
r
2
x
2
)
2
= (r
2
x
2
).
Then
1
2
V =
_
r
0
(r
2
x
2
)dx =
_
r
2
x
1
3
x
3
_
r
0
= r
3

1
3
r
3
=
2
3
r
3
.
Thus V = (4/3)r
3
.
Example 3.2.4. Find volume of intersection of two spheres of radius r, where the
center of each sphere lies on the edge of the other sphere.
From the picture we see that the answer is
2
_
r
r/2
A(x),
where A(x) is exactly as in Example 3.2.3. We have
2
_
r
r/2
(r
2
x
2
)dx =
5
12
r
3
.
3.3 Average Values
Quiz Answers: (1) 29, (2)
1
2
ln

x
2
+ 1

+ tan
1
(x)
Exam 1: Wednesday, Feb 1, 7:00pm7:50pm, here.
Today: 6.5 Average Values
Today: 10.3 Polar coords
NEXT: 10.4 Areas in Polar coords
Why did we skip from 6.5 to 10.3? Later well go back and look at trig functions and complex
exponentials; these ideas will t together more than you might expect. Well go back to 7.1
on Feb 3.
3.3. AVERAGE VALUES 23
In this section we use Riemann sums to extend the familiar notion of an average,
which provides yet another physical interpretation of integration.
Recall: Suppose y
1
, . . . , y
n
are the amount of rain each day in La Jolla, since you
moved here. The average rainful per day is
y
avg
=
y
1
+ +y
n
n
=
1
n
n

i=1
y
i
.
Denition 3.3.1 (Average Value of Function). Suppose f is a continuous function
on an interval [a, b]. The average value of f on [a, b] is
f
avg
=
1
b a
_
b
a
f(x)dx.
Motivation: If we sample f at n points x
i
, then
f
avg

1
n
n

i=1
f(x
i
) =
(b a)
n(b a)
n

i=1
f(x
i
) =
1
(b a)
n

i=1
f(x
i
)x,
since x =
b a
n
. This is a Riemann sum!
1
(b a)
lim
n
n

i=1
f(x
i
)x =
1
(b a)
_
b
a
f(x)dx.
This explains why we dened f
avg
as above.
Example 3.3.2. What is the average value of sin(x) on the interval [0, ]?
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 0.5 1 1.5 2 2.5 3
sin(x)
2/pi
Figure 3.3.1: What is the average value of sin(x)?
1
0
_

0
sin(x)dx =
1
0
_
cos(x)
_

0
=
1

_
(1) (1)
_

0
=
2

24 CHAPTER 3. APPLICATIONS TO AREAS, VOLUME, AND AVERAGES

Observation: If you multiply both sides by (b a) in Denition 3.3.1, you see that
the average value times the length of the interval is the area, i.e., the average value
gives you a rectangle with the same area as the area under your function. In particular,
in Figure 3.3.1 the area between the x-axis and sin(x) is exactly the same as the area
between the horizontal line of height 2/ and the x-axis.
Example 3.3.3. What is the average value of sin(2x)e
1cos(2x)
on the interval [, ]?
-4
-3
-2
-1
0
1
2
3
4
-3 -2 -1 0 1 2 3
sin(2*x)*exp(1-cos(2*x))
0
Figure 3.3.2: What is the average value?
1
()
_

sin(2x)e
1cos(2x)
dx = 0 (since the function is odd!)
Theorem 3.3.4 (Mean Value Theorem). Suppose f is a continuous function on
[a, b]. Then there is a number c in [a, b] such that f(c) = f
avg
.
This says that f assumes its average value. It is a used very often in understanding
why certain statements are true. Notice that in Examples 3.3.2 and 3.3.3 it is just the
assertion that the graphs of the function and the horizontal line interesect.
Proof. Let F(x) =
_
x
a
f(t)dt. Then F

(x) = f(x). By the mean value theorem for

derivatives, there is c [a, b] such that f(c) = F

(c) = (F(b) F(a))/(b a). But by

the fundamental theorem of calculus,
f(c) =
F(b) F(a)
b a
=
1
b a
_
b
a
f(x)dx = f
avg
.
Chapter 4
Polar Coordinates and
Complex Numbers
4.1 Polar Coordinates
Rectangular coordinates allow us to describe a point (x, y) in the plane in a dierent
way, namely
(x, y) (r, ),
where r is any real number and is an angle.
Polar coordinates are extremely useful, especially when thinking about complex
numbers. Note, however, that the (r, ) representation of a point is very non-unique.
First, is not determined by the point. You could add 2 to it and get the same
point:
_
2,

4
_
=
_
2,
9
4
_
=
_
2,

4
+ 389 2
_
. =
_
2,
7
4
_
Also that r can be negative introduces further non-uniqueness:
_
1,

2
_
=
_
1,
3
2
_
.
Think about this as follows: facing in the direction 3/2 and backing up 1 meter gets
you to the same point as looking in the direction /2 and walking forward 1 meter.
We can convert back and forth between cartesian and polar coordinates using that
x = r cos() (4.1.1)
y = r sin(), (4.1.2)
and in the other direction
r
2
= x
2
+y
2
(4.1.3)
tan() =
y
x
(4.1.4)
(Thus r =
_
x
2
+y
2
and = tan
1
(y/x).)
Example 4.1.1. Sketch r = sin(), which is a circle sitting on top the x axis.
We plug in points for one period of the function we are graphingin this case [0, 2]:
25
26 CHAPTER 4. POLAR COORDINATES AND COMPLEX NUMBERS
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0.5 0.4 0.3 0.2 0.1 0 0.1 0.2 0.3 0.4 0.5
sin(t)
Figure 4.1.1: Graph of r = sin().
0 sin(0) = 0
/6 sin(/6) = 1/2
/4 sin(/4) =

2
2
/2 sin(/2) = 1
3/4 sin(3/4) =

2
2
sin() = 0
+/6 sin( +/6) = 1/2
Notice it is nice to allow r to be negative, so we dont have to restrict the input. BUT
it is really painful to draw this graph by hand.
To more accurately draw the graph, lets try converting the equation to one involving
polar coordinates. This is easier if we multiply both sides by r:
r
2
= r sin().
Note that the new equation has the extra solution (r = 0, = anything), so we have to
be careful not to include this point. Now convert to cartesian coordinates using (4.1.1)
to obtain (4.1.3):
x
2
+y
2
= y. (4.1.5)
The graph of (4.1.5) is the same as that of r = sin(). To conrm this we complete the
square:
x
2
+y
2
= y
x
2
+y
2
y = 0
x
2
+ (y 1/2)
2
= 1/4
Thus the graph of (4.1.5) is a circle of radius 1/2 centered at (0, 1/2).
Actually any polar graph of the form r = a sin() + b cos() is a circle, as you will
see in homework problem 67 by generalizing what we just did.
4.2. AREAS IN POLAR COORDINATES 27
4.2 Areas in Polar Coordinates
Exam 1 Wed Feb 1 7:00pm in Pepper Canyon 109 (not 106!! dierent class there!)
Oce hours: 2:45pm4:15pm
Next: Complex numbers (appendix G); complex exponentials (supplement, which is freely
available online).
We will not do arc length.
People were most confused last time by plotting curves in polar coordinates. (1) it is tedious,
but easier if you do a few and know what they look like (just plot some points and see); theres
not much to it, except plug in values and see what you get, and (2) can sometimes convert to
a curve in (x, y) coordinates, which might be easier.
GOAL for today: Integration in the context of polar coordinates. Get much better at working
with polar coordinates!
Example 4.2.1. (From Stewart.) Find the area enclosed by one leaf of the four-leaved
rose r = cos(2). To nd the area using the methods we know so far, we would need
Figure 4.2.1: Graph of y = cos(2x) and r = cos(2)
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
0 1 2 3 4 5 6
cos(2*x)
1
0.8
0.6
0.4
0.2
0
0.2
0.4
0.6
0.8
1
1 0.8 0.6 0.4 0.2 0 0.2 0.4 0.6 0.8 1
cos(2*t)
to nd a function y = f(x) that gives the height of the leaf.
Multiplying both sides of the equation r = cos(2) by r yields
r
2
= r cos(2) = r(cos
2
sin
2
) =
1
r
((r cos )
2
(r sin )
2
).
Because r
2
= x
2
+y
2
and x = r cos() and y = r sin(), we have
x
2
+y
2
=
1
_
x
2
+y
2
(x
2
y
2
).
Solving for y is a crazy mess, and then integrating? It seems impossible!
But it isnt... if we remember the basic idea of calculus: subdivide and take a limit.
[[Draw a section of a curve r = f() for in some interval [a, b], and shade in the
area of the arc.]]
Remark 4.2.2. We will almost never talk about angles in degreeswell almost always
use radians.
28 CHAPTER 4. POLAR COORDINATES AND COMPLEX NUMBERS
We know how to compute the area of a sector, i.e., piece of a circle with angle .
[[draw picture]]. This is the basic polar region. The area is
A = (fraction of the circle) (area of circle) =
_

2
_
r
2
=
1
2
r
2
.
We now imitate what we did before with Riemann sums. We chop up, approximate,
and take a limit. Break the interval of angles from a to b into n subintervals. Choose

i
in each interval. The area of each slice is approximately (1/2)f(

i
)
2

2
i
. Thus
A = Area of the shaded region
n

i=1
1
2
f(

i
)
2
().
Taking the limit, we see that
A = lim
n
n

i=1
1
2
f(

i
)
2
() =
1
2

_
b
a
f()
2
d.
Amazing! By understanding the denition of Riemann sum, weve derived a formula
for areas swept out by a polar graph. But does it work in practice? Lets revisit our
clover leaf.
4.2.1 Examples
Example 4.2.3. Find the area enclosed by one leaf of the four-leaved rose r = cos(2).
Solution: We need the boundaries of integration. Start at = /4 and go to = /4.
As a check, note that cos((/4) 2) = 0 = cos((/4) 2). We evaluate
1
2

_
/4
/4
cos(2)
2
d =
_
/4
0
cos(2)
2
d (even function)
=
1
2
_
/4
0
(1 + cos(4))d
=
1
2
_
+
1
4
sin(4)
_
/4
0
=

8
.
We used that
cos
2
(x) = (1 + cos(2x))/2 and sin
2
(x) = (1 sin(2x))/2, (4.2.1)
which follow from
cos(2x) = cos
2
(x) sin
2
(x) = 2 cos
2
(x) 1 = 1 2 sin
2
(x).
Example 4.2.4. Find area of region inside the curve r = 3 cos() and outside the
cardiod curve r = 1 + cos().
Solution: This is the same as before. Its the dierence of two areas. Figure out the
limits, which are where the curves intersect, i.e., the such that
3 cos() = 1 + cos().
4.3. COMPLEX NUMBERS 29
1.5
1
0.5
0
0.5
1
1.5
0.5 0 0.5 1 1.5 2 2.5 3
3*cos(t)
1+cos(t)
Figure 4.2.2: Graph of r = 3 cos() and r = 1 + cos()
Solving, 2 cos() = 1, so cos() = 1/2, hence = /3 and = /3. Thus the area is
A =
1
2
_
/3
/3
(3 cos())
2
(1 + cos())
2
d
=
_
/3
0
(3 cos())
2
(1 + cos())
2
d (even function)
=
_
/3
0
(8 cos
2
() 2 cos() 1)d
=
_
/3
0
_
8
1
2
(1 + cos(2)) 2 cos() 1
_
d
=
_
/3
0
3 + 4 cos(2) 2 cos()d
=
_
3 + 2 sin(2) 2 sin()
_
/3
0
= + 2
_
3
2
2
_
3
2
0 2 0 2 0
=
4.3 Complex Numbers
A complex number is an expression of the form a +bi, where a and b are real numbers,
and i
2
= 1. We add and multiply complex numbers as follows:
(a +bi) + (c +di) = (a +c) + (b +d)i
(a +bi) (c +di) = (ac bd) + (ad +bc)i
The complex conjugate of a complex number is
a +bi = a bi.
30 CHAPTER 4. POLAR COORDINATES AND COMPLEX NUMBERS
Note that
(a +bi)(a +bi) = a
2
+b
2
is a real number (has no complex part).
If c +di = 0, then
a +bi
c +di
=
(a +bi)(c di)
c
2
+d
2
=
1
c
2
+d
2
((ac +bd) + (bc ad)i).
Example 4.3.1. (1 2i)(8 3i) = 2 19i and 1/(1 +i) = (1 i)/2 = 1/2 (1/2)i.
Complex numbers are incredibly useful in providing better ways to understand ideas
in calculus, and more generally in many applications (e.g., electrical engineering, quan-
tum mechanics, fractals, etc.). For example,
Every polynomial f(x) factors as a product of linear factors (x ), if we allow
the s in the factorization to be complex numbers. For example,
f(x) = x
2
+ 1 = (x i)(x +i).
This will provide an easier to use variant of the partial fractions integration
technique, which we will see later.
Complex numbers are in correspondence with points in the plane via (x, y)
x+iy. Via this correspondence we obtain a way to add and multiply points in the
plane.
Similarly, points in polar coordinates correspond to complex numbers:
(r, ) r(cos() +i sin()).
Complex numbers provide a very nice way to remember and understand trig
identities.
4.3.1 Polar Form
The polar form of a complex number x + iy is r(cos() + i sin()) where (r, ) are any
choice of polar coordinates that represent the point (x, y) in rectangular coordinates.
Recall that you can nd the polar form of a point using that
r =
_
x
2
+y
2
and = tan
1
(y/x).
NOTE: The existence of complex numbers wasnt generally accepted until people
got used to a geometric interpretation of them.
Example 4.3.2. Find the polar form of 1 +i.
Solution. We have r =

2, so
1 +i =

2
_
1

2
+
i

2
_
=

2 (cos(/4) +i sin(/4)) .
Example 4.3.3. Find the polar form of

3 i.
Solution. We have r =

3 + 1 = 2, so

3 i = 2
_

3
2
+i
1
2
_
= 2 (cos(/6) +i sin(/6))
[[A picture is useful here.]]
4.3. COMPLEX NUMBERS 31
Finding the polar form of a complex number is exactly the same problem as nding
polar coordinates of a point in rectangular coordinates. The only hard part is guring
out what is.
If we write complex numbers in rectangular form, their sum is easy to compute:
(a +bi) + (c +di) = (a +c) + (b +d)i
The beauty of polar coordinates is that if we write two complex numbers in polar form,
then their product is very easy to compute:
r
1
(cos(
1
) +i sin(
1
)) r
2
(cos(
2
) +i sin(
2
)) = (r
1
r
2
)(cos(
1
+
2
) +i sin(
1
+
2
)).
The magnitudes multiply and the angles add. The above formula is true because of the
double angle identities for sin and cos (and it is how I remember those formulas!).
(cos(
1
) +i sin(
1
)) (cos(
2
) +i sin(
2
))
= (cos(
1
) cos(
2
) sin(
1
) sin(
2
)) +i(sin(
1
) cos(
2
) + cos(
1
) sin(
2
)).
For example, the power of a singular complex number in polar form is easy to
compute; just power the r and multiply the angle.
Theorem 4.3.4 (De Moivres). For any integer n we have
(r(cos() +i sin()))
n
= r
n
(cos(n) +i sin(n)).
Example 4.3.5. Compute (1 +i)
2006
.
Solution. We have
(1 +i)
2006
= (

2 (cos(/4) +i sin(/4)))
2006
=

2
2006
(cos(2006/4) +i sin(2006/4)))
= 2
1003
(cos(3/2) +i sin(3/2)))
= 2
1003
i
To get cos(2006/4) = cos(3/2) we use that 2006/4 = 501.5, so by periodicity of
cosine, we have
cos(2006/4) = cos((501.5) 250(2)) = cos(1.5) = cos(3/2).
EXAM 1: Wednesday 7:00-7:50pm in Pepper Canyon 109 (!)
Today: Supplement 1 (get online; also homework online)
Wednesday: Review
Bulletin board, online chat, directory, etc. see main course website.
Review day I will prepare no LECTURE; instead I will answer questions.
Your job is to have your most urgent questions ready to go!
Oce hours moved: NOT Tue 11-1 (since nobody ever comes then and Ill be at a conference);
instead Ill be in my oce to answer questions WED 1:30-4pm, and after class on WED too.
Oce: AP&M 5111
32 CHAPTER 4. POLAR COORDINATES AND COMPLEX NUMBERS
Quick review:
Given a point (x, y) in the plane, we can also view it as x + iy or in polar form as r(cos() +
i sin()). Polar form is great since its good for multiplication, powering, and for extracting
roots:
r1(cos(1) + i sin(1))r2(cos(2) + i sin(2)) = (r1r2)(cos(1 + 2) + i sin(1 + 2)).
(If you divide, you subtract the angle.) The point is that the polar form works better with
multiplication than the rectangular form.
Theorem 4.3.6 (De Moivres). For any integer n we have
(r(cos() + i sin()))
n
= r
n
(cos(n) + i sin(n)).
Since we know how to raise a complex number in polar form to the nth power, we
can nd all numbers with a given power, hence nd the nth roots of a complex number.
Proposition 4.3.7 (nth roots). A complex number z = r(cos() + i sin()) has n
distinct nth roots:
r
1/n
_
cos
_
+ 2k
n
_
+i sin
_
+ 2k
n
__
,
for k = 0, 1, . . . , n 1. Here r
1/n
is the real positive n-th root of r.
As a double-check, note that by De Moivre, each number listed in the proposition
has nth power equal to z.
An application of De Moivre is to computing sin(n) and cos(n) in terms of sin()
and cos(). For example,
cos(3) +i sin(3) = (cos() +i sin())
3
= (cos()
3
3 cos () sin()
2
) +i(3 cos()
2
sin() sin()
3
)
Equate real and imaginary parts to get formulas for cos(3) and sin(3). In the next
section we will discuss going in the other direction, i.e., writing powers of sin and cos
in terms of sin and cosine.
Example 4.3.8. Find the cube roots of 2.
Solution. Write 2 in polar form as
2 = 2(cos(0) +i sin(0)).
Then the three cube roots of 2 are
2
1/3
(cos(2k/3) +i sin(2k/3)),
for k = 0, 1, 2. I.e.,
2
1/3
, 2
1/3
(1/2 +i

3/2), 2
1/3
(1/2 i

3/2).
4.4 Complex Exponentials and Trig Identities
Recall that
r
1
(cos(
1
) +i sin(
1
))r
2
(cos(
2
) +i sin(
2
)) = (r
1
r
2
)(cos(
1
+
2
) +i sin(
1
+
2
)).
4.4. COMPLEX EXPONENTIALS AND TRIG IDENTITIES 33
The angles add. Youve seen something similar before:
e
a
e
b
= a
a+b
.
This connection between exponentiation and (4.4) gives us an idea!
If z = x +iy is a complex number, dene
e
z
= e
x
(cos(y) +i sin(y)).
We have just written polar coordinates in another form. Its a shorthand for the polar
form of a complex number:
r(cos() +i sin()) = re
i
.
Theorem 4.4.1. If z
1
, z
2
are two complex numbers, then
e
z1
e
z2
= e
z1+z2
Proof.
e
z1
e
z2
= e
a1
(cos(b
1
) +i sin(b
1
)) e
a2
(cos(b
2
) +i sin(b
2
))
= e
a1+a2
(cos(b
1
+b
2
) +i sin(b
1
+b
2
))
= e
z1+z2
.
Here we have just used (4.4).
The following theorem is amazing, since it involves calculus.
Theorem 4.4.2. If w is a complex number, then
d
dx
e
wx
= we
wx
,
for x real. In fact, this is even true for x a complex variable (but we havent dened
dierentiation for complex variables yet).
Proof. Write w = a +bi.
d
dx
e
wx
=
d
dx
e
ax+bix
=
d
dx
(e
ax
(cos(bx) +i sin(bx)))
=
d
dx
(e
ax
cos(bx) +ie
ax
sin(bx))
=
d
dx
(e
ax
cos(bx)) +i
d
dx
(e
ax
sin(bx))
Now we use the product rule to get
d
dx
(e
ax
cos(bx)) +i
d
dx
(e
ax
sin(bx))
= ae
ax
cos(bx) be
ax
sin(bx) +i(ae
ax
sin(bx) +be
ax
cos(bx))
= e
ax
(a cos(bx) b sin(bx) +i(a sin(bx) +b cos(bx))
34 CHAPTER 4. POLAR COORDINATES AND COMPLEX NUMBERS
On the other hand,
we
wx
= (a +bi)e
ax+bxi
= (a +bi)e
ax
(cos(bx) +i sin(bx))
= e
ax
(a +bi)(cos(bx) +i sin(bx))
= e
ax
((a cos(bx) b sin(bx)) +i(a sin(bx)) +b cos(bx))
Wow!! We did it!
That Theorem 4.4.2 is true is pretty amazing. Its what really gets complex analysis
going.
Example 4.4.3. Heres another fun fact: e
i
+ 1 = 0.
Solution. By denition, have e
i
= cos() +i sin() = 1 +i0 = 1.
4.4.1 Trigonometry and Complex Exponentials
Amazingly, trig functions can also be expressed back in terms of the complex expo-
nential. Then everything involving trig functions can be transformed into something
involving the exponential function. This is very surprising.
In order to easily obtain trig identities like cos(x)
2
+ sin(x)
2
= 1, lets write cos(x)
and sin(x) as complex exponentials. From the denitions we have
e
ix
= cos(x) +i sin(x),
so
e
ix
= cos(x) +i sin(x) = cos(x) i sin(x).
Adding these two equations and dividing by 2 yields a formula for cos(x), and subtract-
ing and dividing by 2i gives a formula for sin(x):
cos(x) =
e
ix
+e
ix
2
sin(x) =
e
ix
e
ix
2i
.
We can now derive trig identities. For example,
sin(2x) =
e
i2x
e
i2x
2i
=
(e
ix
e
ix
)(e
ix
+e
ix
)
2i
= 2
e
ix
e
ix
2i
e
ix
+e
ix
2
= 2 sin(x) cos(x).
Im unimpressed, given that you can get this much more directly using
(cos(2x) +i sin(2x)) = (cos(x) +i sin(x))
2
= cos
2
(x) sin
2
(x) +i2 cos(x) sin(x),
and equating imaginary parts. But there are more interesting examples.
Next we verify that (4.4.1) implies that cos(x)
2
+ sin(x)
2
= 1. We have
4(cos(x)
2
+ sin(x)
2
) =
_
e
ix
+e
ix
_
2
+
_
e
ix
e
ix
i
_
2
= e
2ix
+ 2 +e
2ix
(e
2ix
2 +e
2ix
) = 4.
The equality just appears as a follow-your-nose algebraic calculation.
4.4. COMPLEX EXPONENTIALS AND TRIG IDENTITIES 35
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
0 1 2 3 4 5 6
sin(x)**3
0
Figure 4.4.1: What is sin(x)
3
?
Example 4.4.4. Compute sin(x)
3
as a sum of sines and cosines with no powers.
Solution. We use (4.4.1):
sin(x)
3
=
_
e
ix
e
ix
2i
_
3
=
_
1
2i
_
3
(e
ix
e
ix
)
3
=
_
1
2i
_
3
(e
ix
e
ix
)(e
ix
e
ix
)(e
ix
e
ix
)
=
_
1
2i
_
3
(e
ix
e
ix
)(e
2ix
2 +e
2ix
)
=
_
1
2i
_
3
(e
3ix
2e
ix
+e
ix
e
ix
+ 2e
ix
e
3ix
)
=
_
1
2i
_
3
((e
3ix
e
3ix
) 3(e
ix
e
ix
))
=
_
1
4
__
e
3ix
e
3ix
2i
3
e
ix
e
ix
2i
_
=
3 sin(x) sin(3x)
4
.
36 CHAPTER 4. POLAR COORDINATES AND COMPLEX NUMBERS
Chapter 5
Integration Techniques
5.1 Integration By Parts
Quiz next Friday
Today: 7.1: integration by parts
Next: 7.2: trigonometric integrals and supplement 2functions with complex values
Exams: Average 19.68 (out of 34).
Tetrahedron problem:
Z
h
0
1
2

b
h
x + b

a
h
x + a

dx = =
abh
6
.
(The function that gives the base of the triangle cross section is a linear function that is b
at x = 0 and 0 at x = h, which allows you to easily determine it without thinking about
geometry.)
Dierentiation Integration
Chain Rule Substitution
Product Rule Integration by Parts
The product rule is that
d
dx
[f(x)g(x)] = f(x)g

(x) +f

(x)g(x).
Integrating both sides leads to a new fundamental technique for integration:
f(x)g(x) =
_
f(x)g

(x)dx +
_
g(x)f

(x)dx. (5.1.1)
Now rewrite (5.1.1) as
_
f(x)g

(x)dx = f(x)g(x)
_
g(x)f

(x)dx.
Shorthand notation:
u = f(x) du = f

(x)dx
v = g(x) dv = g

(x)dx
37
38 CHAPTER 5. INTEGRATION TECHNIQUES
Then have _
udv = uv
_
vdu.
So what! But whats the big deal? Integration by parts is a fundamental
technique of integration. It is also a key step in the proof of many theorems in calculus.
Example 5.1.1.
_
xcos(x)dx.
u = x v = sin(x)
du = dx dv = cos(x)dx
We get
_
xcos(x)dx = xsin(x)
_
sin(x)dx = xsin(x) + cos(x) +c.
Did this do anything for us? Indeed, it did.
Wait a minutehow did we know to pick u = x and v = sin(x)? We could have
picked them other way around and still written down true statements. Lets try that:
u = cos(x) v =
1
2
x
2
du = sin(x)dx dv = xdx
_
xcos(x)dx =
1
2
xcos(x) +
_
1
2
x
2
sin(x)dx.
Did this help!? NO. Integrating x
2
sin(x) is harder than integrating xcos(x). This
formula is completely correct, but is hampered by being useless in this case. So how do
you pick them?
Choose the u so that when you dierentiate it you get something simpler;
when you pick dv, try to choose something whose antiderivative is simpler.
Sometimes you have to try more than once. But with a good eraser nodoby will know
that it took you two tries.
Question 5.1.2. If integration by parts once is good, then sometimes twice is even
better? Yes, in some examples (see Example 5.1.5). But in the above example, you just
undo what you did and basically end up where you started, or you get something even
worse.
Example 5.1.3. Compute
_ 1
2
0
sin
1
(x)dx. Two points:
1. Its a denite integral.
2. There is only one function; would you think to do integration by parts? But it is
a product; it just doesnt look like it at rst glance.
Your choice is made for you, since wed be back where we started if we put dv =
sin
1
(x)dx.
u = sin
1
(x) v = x
du =
1

1 x
2
dv = dx
5.1. INTEGRATION BY PARTS 39
We get
_ 1
2
0
sin
1
(x)dx =
_
xsin
1
(x)
1
2
0

_ 1
2
0
x

1 x
2
dx.
Now we use substitution with w = 1 x
2
, dw = 2xdx, hence xdx =
1
2
dw.
_ 1
2
0
x

1 x
2
dx =
1
2
_
w

1
2
dw = w
1
2
+c =
_
1 x
2
+c.
Hence
_ 1
2
0
sin
1
(x)dx =
_
xsin
1
(x)
1
2
0
+
_
_
1 x
2
_1
2
0
=

12
+

3
2
1
But shouldnt we change the limits because we did a substitution? (No, since we com-
puted the indenite integral and put it back; this time we did the other option.)
Is there another way to do this? I dont know. But for any integral, there might be
several dierent techniques. If you can think of any other way to guess an antiderivative,
do it; you can always dierentiate as a check.
Note: Integration by parts is tailored toward doing indenite integrals.
Example 5.1.4. This example illustrates how to use integration by parts twice. We
compute
_
x
2
e
2x
dx
u = x
2
v =
1
2
e
2x
du = 2xdx dv = e
2x
dx
We have
_
x
2
e
2x
dx =
1
2
x
2
e
2x
+
_
xe
2x
dx.
Did this help? It helped, but it did not nish the integral o. However, we can deal
with the remaining integral, again using integration by parts. If you do it twice, you
what to keep going in the same direction. Do not switch your choice, or youll undo
what you just did.
u = x v =
1
2
e
2x
du = dx dv = e
2x
dx
_
xe
2x
dx =
1
2
xe
2x
+
1
2
_
e
2x
dx =
1
2
xe
2x

1
4
e
2x
+c.
Now putting this above, we have
_
x
2
e
2x
dx =
1
2
x
2
e
2x

1
2
xe
2x

1
4
e
2x
+c =
1
4
e
2x
(2x
2
+ 2x + 1) +c.
Do you think you might have to do integration by parts three times? What if it
were
_
x
3
e
2x
dx? Grrr youd have to do it three times.
40 CHAPTER 5. INTEGRATION TECHNIQUES
Example 5.1.5. Compute
_
e
x
cos(x)dx. Which should be u and which should be v?
Taking the derivatives of each type of function does not change the type. As a practical
matter, it doesnt matter. Which would you prefer to nd the antiderivative of? (Both
choices work, as long as you keep going in the same direction when you do the second
step.)
u = cos(x) v = e
x
du = sin(x)dx dv = e
x
dx
We get
_
e
x
cos(x)dx = e
x
cos(x) +
_
e
x
sin(x)dx.
We have to do it again. This time we choose (going in the same direction):
u = sin(x) v = e
x
du = cos(x)dx dv = e
x
dx
We get
_
e
x
cos(x)dx = e
x
cos(x) +e
x
sin(x)
_
e
x
cos(x)dx.
Did we get anywhere? Yes! No! First impression: all this work, and were back where
we started from! Yuck. Clearly we dont want to integrate by parts yet again. BUT.
Notice the minus sign in front of
_
e
x
cos(x)dx; You can add the integral to both sides
and get
2
_
e
x
cos(dx) = e
x
cos(x) +e
x
sin(x) +c.
Hence _
e
x
cos(dx) =
1
2
e
x
(cos(x) + sin(x)) +c.
5.2 Trigonometric Integrals
Friday: Quiz 2
Next: Trig subst.
cos
2
(x) =
1 + cos(2x)
2
and sin
2
(x) =
1 sin(2x)
2
.
(5.2.1)
Example 5.2.1. Compute
_
sin
3
(x)dx.
We use trig. identities and compute the integral directly as follows:
_
sin
3
(x)dx =
_
sin
2
(x) sin(x)dx
=
_
[1 cos
2
(x)] sin(x)dx
= cos(x) +
1
3
cos
3
(x) +c (substitution u = cos(x))
5.2. TRIGONOMETRIC INTEGRALS 41
This always works for odd powers of sin(x).
Example 5.2.2. What about even powers?! Compute
_
sin
4
(x)dx. We have
sin
4
(x) = [sin
2
(x)]
2
=
_
1 cos(2x)
2
_
2
=
1
4

_
1 2 cos(2x) + cos
2
(2x)

=
1
4
_
1 2 cos(2x) +
1
2
+
1
2
cos(4x)
_
Thus
_
sin
4
(x)dx =
_ _
3
8

1
2
cos(2x) +
1
8
cos(4x)
_
dx
=
3
8
x
1
4
sin(2x) +
1
32
sin(4x) +c.
Key Trick: Realize that we should write sin
4
(x) as (sin
2
(x))
2
. The rest is straightfor-
ward.
Example 5.2.3. This example illustrates a method for computing integrals of trig
functions that doesnt require knowing any trig identities at all or any tricks. It is very
tedious though. We compute
_
sin
3
(x)dx using complex exponentials. We have
cos(x) =
e
ix
+e
ix
2
sin(x) =
e
ix
e
ix
2i
.
hence
_
sin
3
(x)dx =
_ _
e
ix
e
ix
2i
_
3
dx
=
1
8i
_
(e
ix
e
ix
)
3
dx
=
1
8i
_
(e
ix
e
ix
)(e
ix
e
ix
)(e
ix
e
ix
)dx
=
1
8i
_
(e
2ix
2 +e
2ix
)(e
ix
e
ix
)dx
=
1
8i
_
e
3ix
e
ix
2e
ix
+ 2e
ix
+e
ix
e
3ix
dx
=
1
8i
_
e
3ix
e
3ix
+ 3e
ix
3e
ix
dx
=
1
8i
_
e
3ix
3i

e
3ix
3i
+
3e
ix
i

3e
ix
i
_
+c
=
1
4
_
1
3
cos(3x) 3 cos(x)
_
+c
=
1
12
cos(3x)
3
4
cos(x) +c
The answer looks totally dierent, but is in fact the same function.
42 CHAPTER 5. INTEGRATION TECHNIQUES
Here are some more identities that well use in illustrating some tricks below.
d
dx
tan(x) = sec
2
(x)
and
d
dx
sec(x) = sec(x) tan(x).
Also,
1 + tan
2
(x) = sec
2
(x).
Example 5.2.4. Compute
_
tan
3
(x)dx. We have
_
tan
3
(x)dx =
_
tan(x) tan
2
(x)dx
=
_
tan(x)
_
sec
2
(x) 1

dx
=
_
tan(x) sec
2
(x)dx
_
tan(x)dx
=
1
2
tan
2
(x) ln | sec(x)| +c
Here we used the substitution u = tan(x), so du = sec
2
(x)dx, so
_
tan(x) sec
2
(x)dx =
_
udu =
1
2
u
2
+c =
1
2
tan
2
(x) +c.
Also, with the substitution u = cos(x) and du = sin(x)dx we get
_
tan(x)dx =
_
sin(x)
cos(x)
dx =
_
1
u
du = ln |u| +c = ln | sec(x)| +c.
Key trick: Write tan
3
(x) as tan(x) tan
2
(x).
Example 5.2.5. Heres one that combines trig identities with the funnest variant of
integration by parts. Compute
_
sec
3
(x)dx.
We have _
sec
3
(x)dx =
_
sec(x) sec
2
(x)dx.
Lets use integration by parts.
u = sec(x) v = tan(x)
du = sec(x) tan(x)dx dv = sec
2
(x)dx
The above integral becomes
_
sec(x) sec
2
(x)dx = sec(x) tan(x)
_
sec(x) tan
2
(x)dx
= sec(x) tan(x)
_
sec(x)[sec
2
(x) 1]dx
= sec(x) tan(x)
_
sec
3
(x) +
_
sec(x)dx
= sec(x) tan(x)
_
sec
3
(x) + ln | sec(x) + tan(x)|
5.2. TRIGONOMETRIC INTEGRALS 43
This is familiar. Solve for
_
sec
3
(x). We get
_
sec
3
(x)dx =
1
2
_
sec(x) tan(x) + ln | sec(x) + tan(x)|
_
+c
5.2.1 Some Remarks on Using Complex-Valued Functions
Consider functions of the form
f(x) +ig(x), (5.2.2)
where x is a real variable and f, g are real-valued functions. For example,
e
ix
= cos(x) +i sin(x).
We observed before that
d
dx
e
wx
= we
wx
hence
_
e
wx
dx =
1
w
e
wx
+c.
For example, writing it e
ix
as in (5.2.2), we have
_
e
ix
dx =
_
cos(x)dx +i
_
sin(x)dx
= sin(x) i cos(x) +c
= i(cos(x) +i sin(x)) +c
=
1
i
e
ix
.
Example 5.2.6. Lets compute
_
1
x +i
dx. Wouldnt it be nice if we could just write
ln(x + i) + c? This is useless for us though, since we havent even dened ln(x + i)!
However, we can rationalize the denominator by writing
_
1
x +i
dx =
_
1
x +i

x i
x i
dx
=
_
x i
x
2
+ 1
dx
=
_
x
x
2
+ 1
dx i
_
1
x
2
+ 1
dx
=
1
2
ln |x
2
+ 1| i tan
1
(x) +c
This informs how we would dene ln(z) for z complex (which youll do if you take a
course in complex analysis). Key trick: Get the i in the numerator.
The next example illustrates an alternative to the method of Section 5.2.
44 CHAPTER 5. INTEGRATION TECHNIQUES
Example 5.2.7.
_
sin(5x) cos(3x)dx =
_ _
e
i5x
e
i5x
2i
__

e
i5x
+e
i5x
2
_
dx
=
1
4i
_
_
e
i8x
e
i8x
+e
i2x
e
i2x
_
dx +c
=
1
4i
_
e
i8x
8i
+
e
i8x
8i
+
e
i2x
2i
+
e
i2x
2i
_
+c
=
1
4
_
1
4
cos(8x) + cos(2x)
_
+c
This is more tedious than the method in 5.2. But it is completely straightforward. You
dont need any trig formulas or anything else. You just multiply it out, integrate, etc.,
and remember that i
2
= 1.
5.3 Trigonometric Substitutions
Return more midterms?
Rough meaning of grades:
2934 is A
2328 is B
1722 is C
1116 is D
Regarding the quizif you do every homework problem that was assigned, youll have a severe
case of deja vu on the quiz! On the exam, we do not restrict ourselves like this, but you get to
have a sheet of paper.
The rst homework problem is to compute
_
2

2
1
x
3

x
2
1
dx. (5.3.1)
Your rst idea might be to do some sort of substitution, e.g., u = x
2
1, but du = 2xdx
is nowhere to be seen and this simply doesnt work. Likewise, integration by parts gets
us nowhere. However, a technique called inverse trig substitutions and a trig identity
easily dispenses with the above integral and several similar ones! Heres the crucial
table:
Expression Inverse Substitution Relevant Trig Identity

a
2
x
2
x = a sin(),

2


2
1 sin
2
() = cos
2
()

a
2
+x
2
x = a tan(),

2
< <

2
1 + tan
2
() = sec
2
()

x
2
a
2
x = a sec(), 0 <

2
or <
3
2
sec
2
() 1 = tan
2
()
Inverse substitution works as follows. If we write x = g(t), then
_
f(x)dx =
_
f(g(t))g

(t)dt.
5.3. TRIGONOMETRIC SUBSTITUTIONS 45
This is not the same as substitution. You can just apply inverse substitution to any
integral directlyusually you get something even worse, but for the integrals in this
section using a substitution can vastly improve the situation.
If g is a 1 1 function, then you can even use inverse substitution for a denite
integral. The limits of integration are obtained as follows.
_
b
a
f(x)dx =
_
g
1
(b)
g
1
(a)
f(g(t))g

(t)dt. (5.3.2)
To help you understand this, note that as t varies from g
1
(a) to g
1
(b), the function
g(t) varies from a = g(g
1
(a) to b = g(g
1
(b)), so f is being integrated over exactly
the same values. Note also that (5.3.2) once again illustrates Leibnizs brilliance in
designing the notation for calculus.
Lets give it a shot with (5.3.1). From the table we use the inverse substition
x = sec().
We get
_
2

2
1
x
3

x
2
1
dx =
_
3

4
1
sec()
_
sec
2
() 1 sec() tan()d
=
_
3

4
1
sec()
tan() sec() tan()d
=
_
3

4
cos
(
)d
=
1
2
_
3

4
1 + cos(2)d
=
1
2
_
+
1
2
sin(2)
_
3

4
=

24
+

3
8

1
4
Wow! That was like magic. This is really an amazing technique. Lets use it again
to nd the area of an ellipse.
Example 5.3.1. Consider an ellipse with radii a and b, so it goes through (0, b) and
(a, 0). An equation for the part of an ellipse in the rst quadrant is
y = b
_
1
x
2
a
2
=
b
a
_
a
2
x
2
.
Thus the area of the entire ellipse is
A = 4
_
a
0
b
a
_
a
2
x
2
dx.
The 4 is because the integral computes 1/4th of the area of the whole ellipse. So we
need to compute
_
a
0
_
a
2
x
2
dx
46 CHAPTER 5. INTEGRATION TECHNIQUES
Obvious substitution with u = a
2
x
2
...? nope. Integration by parts...? nope.
Lets try inverse substitution. The table above suggests using x = a sin(), so
dx = a cos()d. We get
_
2
0
_
a
2
a
2
sin
2
()d = a
2
_
2
0
cos
2
()d (5.3.3)
=
a
2
2
_
2
0
1 + cos(2)d (5.3.4)
=
a
2
2
_
+
1
2
sin(2)
_
2
0
(5.3.5)
=
a
2
2


2
=
a
2
4
. (5.3.6)
Thus the area is
4
b
a
a
2
4
= ab.
Consistency Check: If the ellipse is a circle, i.e., a = b = r, this is r
2
, which is a
well-known formula for the area of a circle.
Remark 5.3.2. Trigonometric substitution is useful for functions that involve

a
2
x
2
,

x
2
+a
2
,

x
2
a, but not all at once!. See the above table for how to do each.
One other important technique is to use completing the square.
Example 5.3.3. Compute
_
5 + 4x x
2
dx. We complete the square:
5 + 4x x
2
= 5 (x 2)
2
+ 4 = 9 (x 2)
2
.
Thus _
_
5 + 4x x
2
dx =
_
_
9 (x 2)
2
dx.
We do a usual substitution to get rid of the x 2. Let u = x 2, so du = dx. Then
_
_
9 (x 2)
2
dx =
_
_
9 y
2
dy.
Now we have an integral that we can do; its almost identical to the previous example,
but with a = 9 (and this is an indenite integral). Let y = 3 sin(), so dy = 3 cos()d.
Then
_
_
9 (x 2)
2
dx =
_
_
9 y
2
dy
=
_ _
3
2
3
2
sin
2
()3 cos()d
= 9
_
cos
2
() d
=
9
2
_
1 + cos(2)d
=
9
2
_
+
1
2
sin(2)
_
+c
5.3. TRIGONOMETRIC SUBSTITUTIONS 47
Of course, we must transform back into a function in x, and thats a little tricky. Use
that
x 2 = y = 3 sin(),
so that
= sin
1
_
x 2
3
_
.
_
_
9 (x 2)
2
dx =
=
9
2
_
+
1
2
sin(2)
_
+c
=
9
2
_
sin
1
_
x 2
3
_
+ sin() cos()
_
+c
=
9
2
_
sin
1
_
x 2
3
_
+
_
x 2
3
_

_
_
9 (x 2)
2
3
__
+c.
Here we use that sin(2) = 2 sin() cos(). Also, to compute cos(sin
1
_
x2
3
_
), we draw
a right triangle with side lengths x 2 and
_
9 (x 2)
2
, and hypotenuse 3.
Example 5.3.4. Compute
_
1

t
2
6t + 13
dt
To compute this, we complete the square, etc.
_
1

t
2
6t + 13
dt =
_
1
_
(t 3)
2
+ 4
dt
[[Draw triangle with sides 2 and t 3 and hypotenuse
_
(t 3)
2
+ 4. Then
t 3 = 2 tan()
_
(t 3)
2
+ 4 = 2 sec() =
2
cos()
dt = 2 sec
2
()d
Back to the integral, we have
_
1
_
(t 3)
2
+ 4
dt =
_
2 sec
2
()
2 sec()
d
=
_
sec()d
= ln | sec() + tan()| +c
= ln

_
(t 3)
2
+ 42 +
t 3
2

+c.
48 CHAPTER 5. INTEGRATION TECHNIQUES
5.4 Factoring Polynomials
Quizes today!
How do you compute something like
_
x
2
+ 2
(x 1)(x + 2)(x + 3)
dx?
So far you have no method for doing this. The trick (which is called partial fraction
decomposition), is to write
_
x
2
+ 2
x
3
+ 4x
2
+x 6
dx =
_
1
4(x 1)

2
x + 2
+
11
4(x + 3)
dx (5.4.1)
The integral on the right is then easy to do (the answer involves lns).
But how on earth do you right the rational function on the left hand side as a
sum of the nice terms of the right hand side? Doing this is called partial fraction
decomposition, and it is a fundamental idea in mathematics. It relies on our ability to
factor polynomials and saolve linear equations. As a rst hint, notice that
x
3
+ 4x
2
+x 6 = (x 1) (x + 2) (x + 3),
so the denominators in the decomposition correspond to the factors of the denominator.
Before describing the secret behind (5.4.1), well discuss some background about
how polynomials and rational functions work.
Theorem 5.4.1 (Fundamental Theorem of Algebra). If f(x) = a
n
x
n
+ a
1
x+a
0
is a polynomial, then there are complex numbers c,
1
, . . .
n
such that
f(x) = c(x
1
)(x
2
) (x
n
).
Example 5.4.2. For example,
3x
2
+ 2x 1 = 3
_
x
1
3
_
(x + 1).
And
(x
2
+ 1) = (x +i)
2
(x i)
2
.
If f(x) is a polynomial, the roots of f correspond to the factors of f. Thus if
f(x) = c(x
1
)(x
2
) (x
n
),
then f(
i
) = 0 for each i (and nowhere else).
Denition 5.4.3 (Multiplicity of Zero). The multiplicity of a zero of f(x) is the
number of times that (x ) appears as a factor of f.
For example, if f(x) = 7(x2)
99
(x+17)
5
(x)
2
, then 2 is a zero with multiplicity
99, is a zero with multiplicity 2, and 1 is a zero multiplicity 0.
5.5. INTEGRATIONOF RATIONAL FUNCTIONS USINGPARTIAL FRACTIONS49
Denition 5.4.4 (Rational Function). A rational function is a quotient
f(x) =
g(x)
h(x)
,
where g(x) and h(x) are polynomials.
For example,
f(x) =
x
10
(x i)
2
(x +)(x 3)
3
(5.4.2)
is a rational function.
Denition 5.4.5 (Pole). A pole of a rational function f(x) is a complex number
such that |f(x)| is unbounded as x .
For example, for (5.4.2) the poles are at i, , and 3. They have multiplicity 2, 1,
and 3, respectively.
5.5 Integration of Rational Functions Using Partial
Fractions
Today: 7.4: Integration of rational functions and Supp. 4: Partial fraction expansion
Next: 7.7: Approximate integration
Our goal today is to compute integrals of the form
_
P(x)
Q(x)
dx
by decomposing f =
P(x)
Q(x)
. This is called partial fraction expansion.
Theorem 5.5.1 (Fundamental Theorem of Algebra over the Real Numbers).
A real polynomial of degree n 1 can be factored as a constant times a product of linear
factors x a and irreducible quadratic factors x
2
+bx +c.
Note that x
2
+bx +c = (x )(x ), where = z +iw, = z iw are complex
conjugates.
Types of rational functions f(x) =
P(x)
Q(x)
. To do a partial fraction expansion, rst
make sure deg(P(x)) < deg(Q(x)) using long division. Then there are four possible
situation, each of increasing generality (and diculty):
1. Q(x) is a product of distinct linear factors;
2. Q(x) is a product of linear factors, some of which are repeated;
3. Q(x) is a product of distinct irreducible quadratic factors, along with linear factors
some of which may be repeated; and,
4. Q(x) is has repeated irreducible quadratic factors, along with possibly some linear
factors which may be repeated.
50 CHAPTER 5. INTEGRATION TECHNIQUES
The general partial fraction expansion theorem is beyond the scope of this course.
However, you might nd the following special case and its proof interesting.
Theorem 5.5.2. Suppose p, q
1
and q
2
are polynomials that are relatively prime (have
no factor in common). Then there exists polynomials
1
and
2
such that
p
q
1
q
2
=

1
q
1
+

2
q
2
.
Proof. Since q
1
and q
2
are relatively prime, using the Euclidean algorithm (long divi-
sion), we can nd polynomials s
1
and s
2
such that
1 = s
1
q
1
+s
2
q
2
.
Dividing both sides by q
1
q
2
and multiplying by p yields
p
q
1
q
2
=

1
q
1
+

2
q
2
,
which completes the proof.
Example 5.5.3. Compute
_
x
3
4x 10
x
2
x 6
dx.
First do long division. Get quotient of x +1 and remainder of 3x 4. This means that
x
3
4x 10
x
2
x 6
= x + 1 +
3x 4
x
2
x 6
.
Since we have distinct linear factors, we know that we can write
f(x) =
3x 4
x
2
x 6
=
A
x 3
+
B
x + 2
,
for real numbers A, B. A clever way to nd A, B is to substitute appropriate values in,
as follows. We have
f(x)(x 3) =
3x 4
x + 2
= A+B
x 3
x + 2
.
Setting x = 3 on both sides we have (taking a limit):
A = f(3) =
3 3 4
3 + 2
=
5
5
= 1.
Likewise, we have
B = f(2) =
3 (2) 4
2 3
= 2.
Thus
_
x
3
4x 10
x
2
x 6
dx =
_
x + 1 +
1
x 3
+
2
x + 2
=
x
2
+ 2x
2
+ 2 log |x + 2| + log |x 3| +c.
5.5. INTEGRATIONOF RATIONAL FUNCTIONS USINGPARTIAL FRACTIONS51
Example 5.5.4. Compute the partial fraction expansion of
x
2
(x3)(x+2)
2
. By the partial
fraction theorem, there are constants A, B, C such that
x
2
(x 3)(x + 2)
2
=
A
x 3
+
B
x + 2
+
C
(x + 2)
2
.
Note that theres no possible way this could work without the (x + 2)
2
term, since
otherwise the common denominator would be (x 3)(x + 2). We have
A = [f(x)(x 3)]
x=3
=
x
2
(x + 2)
2
|
x=3
=
9
25
,
C =
_
f(x)(x + 2)
2

x=2
=
4
5
.
This method will not get us B! For example,
f(x)(x + 2) =
x
2
(x 3)(x + 2)
= A
x + 2
x 3
+B +
C
x + 2
.
While true this is useless.
Instead, we use that we know A and C, and evaluate at another value of x, say 0.
f(0) = 0 =
9
25
3
+
B
2
+

4
5
(2)
2
,
so B =
16
25
. Thus nally,
_
x
2
(x 3)(x + 2)
2
=
_
9
25
x 3
+
16
25
x + 2
+

4
5
(x + 2)
2
.
=
9
25
ln |x 3| +
16
25
ln |x + 2| +
4
5
x + 2
+ constant.
Example 5.5.5. Lets compute
_
1
x
3
+1
dx. Notice that x + 1 is a factor, since 1 is a
root. We have
x
3
+ 1 = (x + 1)
_
x
2
x + 1
_
.
There exist constants A, B, C such that
1
x
3
+ 1
=
A
x + 1
+
Bx +C
x
2
x + 1
.
Then
A = f(x)(x + 1)|
x=1
=
1
3
.
You could nd B, C by factoring the quadratic over the complex numbers and getting
complex number answers. Instead, we evaluate x at a couple of values. For example, at
x = 0 we get
f(0) = 1 =
1
3
+
C
1
,
so C =
2
3
. Next, use x = 1 to get B.
f(1) =
1
1
3
+ 1
=
1
3
(1) + 1
+
B(1) +
2
3
(1)
2
(1) + 1
1
2
=
1
6
+B +
2
3
,
52 CHAPTER 5. INTEGRATION TECHNIQUES
so
B =
3
6

1
6

4
6
=
1
3
.
Finally,
_
1
x
3
+ 1
dx =
_
1
3
x + 1

1
3
x
x
2
x 1
+
2
3
x
2
x 1
dx
=
1
3
ln |x + 1|
1
3
_
x 2
x
2
x + 1
dx
It remains to compute
_
x 2
x
2
x + 1
dx.
First, complete the square to get
x
2
x + 1 =
_
x
1
2
_
2
+
3
4
.
Let u = (x
1
2
), so du = dx and x = u +
1
2
. Then
_
u
3
2
u
2
+
3
4
du =
_
udu
u
2
+
3
4

3
2
_
1
u
2
+
_

3
2
_
2
du
=
1
2
ln

u
2
+
3
4

3
2

2

3
tan
1
_
2u

3
_
+c
=
1
2
ln

x
2
x + 1

3 tan
1
_
2x 1

3
_
+c
Finally, we put it all together and get
_
1
x
3
+ 1
dx =
1
3
ln |x + 1|
1
3
_
x 2
x
2
x + 1
dx
=
1
3
ln |x + 1|
1
6
ln

x
2
x + 1

3
3
tan
1
_
2x 1

3
_
+c
5.6. APPROXIMATING INTEGRALS 53
Discuss second quiz problem.
Problem: Compute
R
cos
2
(x)e
3x
dx using complex exponentials. The answer is

1
6
e
3x
+
1
13
e
3x
sin(2x)
3
26
e
3x
cos(2x) + c.
Heres how to get it.
Z
cos
2
(x)e
3x
dx =
Z
e
2ix
+ 2 + e
2ix
4
e
3x
dx
=
1
4

e
(2i3)x
2i 3

2
3
e
3x
+
e
(2i3)x
2i 3

+ c
=
1
6
e
3x
+
e
3x
4

e
2ix
2i 3

e
2ix
2i + 3

+ c
Simplify the inside part requires some imagination:
e
2ix
2i 3

e
2ix
2i + 3
=
1
13
(2ie
2ix
3e
2ix
+ 2ie
2ix
3e
2ix
)
=
1
13
(4 sin(2x) 6 cos(2x))
5.6 Approximating Integrals
Today: 7.7 approximating integrals
Friday: Third QUIZ and 7.8 improper integrals
Problem: Compute
_
1
0
e

x
dx.
Hmmm... Any ideas?
Today we will revisit Riemann sums in the context of nding numerical approxi-
mations to integrals, which we might not be able to compute exactly. Recall that if
y = f(x) then
_
b
a
f(x)dx = lim
n
n

i=1
f(x

i
)x.
The fundamental theorem of calculus says that if we can nd an antiderivative of f(x),
then we can compute
_
b
a
f(x)dx exactly. But antiderivatives can be either (1) hard to
nd, and sometimes worse (2) impossible to nd. However, we can always approximate
_
b
a
f(x)dx (possibly very badly).
For example, we could use Riemann sums to approximate
_
b
a
f(x)dx, say using left
endpoints. This gives the approximation:
L
n
=
n1

i=0
f(x
i
)x; x
0
, . . . , x
n1
left endpoints
54 CHAPTER 5. INTEGRATION TECHNIQUES
Using rightpoints gives
R
n
=
n

i=1
f(x
i
)x; x
1
, . . . , x
n
right endpoints
Using midpoints gives
M
n
=
n

i=1
f(x
i
)x; x
1
, . . . , x
n
midpoints,
where x
i
= (x
i1
+x
i
)/2. The midpoint is typically (but not always) much better than
the left or right endpoint approximations.
Yet another possibility is the trapezoid approximation, which is
T
n
=
1
2
(L
n
+R
n
);
this is just the average of the left and right approximations.
Question 5.6.1. But wouldnt the trapezoid and midpoint approximations be the
same?certainly not (see example below); interestingly, very often the midpoint approx-
imation is better.
Simpsons approximation
S
2n
=
1
3
T
n
+
2
3
M
n
gives the area under best-t parabolas that approximate our function on each interval.
The proof of this would be interesting but takes too much time for this course.
Many functions have no elementary antiderivatives:
_
1 +x
3
, e
x
2
,
1
log(x)
,
sin(x)
x
, . . . .
NOTE they do have antiderivatives; the problem is just that there is no simple formula
for them. Why are there no elementary antiderivatives?
Some of these functions are extremly important. For example, the integrals
_
x

e
u
2
/2
du
are extremely important in probability, even though there is no simple formula for the
antiderivative.
If you are doing scientic research you might spend months tediously computing
values of some function f(x), for which no formula is known.
Example 5.6.2. Compute
_
1
0
e

x
dx.
1. Trapezoid with n = 4
2. Midpoint with n = 4
3. Simpsons with with 2n = 8
The following is a table of the values of f(x) at k/8 for k = 0, . . . , 8.
5.6. APPROXIMATING INTEGRALS 55
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 0.2 0.4 0.6 0.8 1
exp(-sqrt(x))
Figure 5.6.1: Graph of e

x
k/8 f(k/8)
0 V
0
= 1.000000
1
8
V
1
= 0.702189
1
4
V
2
= 0.606531
3
8
V
3
= 0.542063
1
2
V
4
= 0.493069
5
8
V
5
= 0.453586
3
4
V
6
= 0.420620
7
8
V
7
= 0.392423
1 V
8
= 0.367879
L
4
= (V
0
+V
2
+V
4
+V
6
)
1
4
= 0.630055
R
4
= (V
2
+V
4
+V
6
+V
8
)
1
4
= 0.472025
M
4
= (V
1
+V
3
+V
5
+V
7
)
1
4
= 0.522565
T
4
=
1
2
(L
4
+R
4
) = 0.551040.
S
8
=
1
3
T
4
+
2
3
M
4
= 0.532057
Maxima gives 0.5284822353142306 and Mathematica gives 0.528482.
Note that Simpsonss is the best; it better be, since we worked the hardest to get it!
Method Error
|L
4
I| 0.101573
|R
4
I| 0.056458
|M
4
I| 0.005917
|T
4
I| 0.022558
|S
8
I| 0.003575
56 CHAPTER 5. INTEGRATION TECHNIQUES
5.7 Improper Integrals
Exam 2 Wed Mar 1: 7pm-7:50pm in ??
Today: 7.8 Improper Integrals
Monday presidents day holiday (and almost my bday)
Next 11.1 sequences
Example 5.7.1. Make sense of
_

0
e
x
dx. The integrals
_
t
0
e
x
dx
make sense for each real number t. So consider
lim
t
_
t
0
e
x
dx = lim
t
[e
x
]
t
0
= 1.
Geometrically the area under the whole curve is the limit of the areas for nite values
of t.
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 2 4 6 8 10
exp(-x)
Figure 5.7.1: Graph of e
x
Example 5.7.2. Consider
_
1
0
1

1x
2
dx (see Figure 5.7.2). Problem: The denominator
of the integrand tends to 0 as x approaches the upper endpoint. Dene
_
1
0
1

1 x
2
dx = lim
t1

_
t
0
1

1 x
2
dx
= lim
t1

_
sin
1
(t) sin
1
(0)
_
= sin
1
(1) =

2
Here t 1

means the limit as t tends to 1 from the left.

5.7. IMPROPER INTEGRALS 57
1
2
3
4
5
6
7
8
0 0.2 0.4 0.6 0.8 1
1/sqrt(1-x**2)
Figure 5.7.2: Graph of
1

1x
2
Example 5.7.3. There can be multiple points at which the integral is improper. For
example, consider
_

1
1 +x
2
dx.
A crucial point is that we take the limit for the left and right endpoints independently.
We use the point 0 (for convenience only!) to break the integral in half.
_

1
1 +x
2
dx =
_
0

1
1 +x
2
dx +
_

0
1
1 +x
2
dx
= lim
s
_
0
s
1
1 +x
2
dx + lim
t
_
t
0
1
1 +x
2
dx
= lim
s
(tan
1
(0) tan
1
(s)) + lim
t
(tan
1
(t) tan
1
(0))
= lim
s
(tan
1
(s)) + lim
t
(tan
1
(t))
=

2
+

2
= .
The graph of tan
1
(x) is in Figure 5.7.3.
Example 5.7.4. Brian Conrads paper on impossibility theorems for elementary in-
tegration begins: The Central Limit Theorem in probability theory assigns a special
signicance to the cumulative area function
(x) =
1

2
_
x

e
u
2
udu
under the Gaussian bell curve
y =
1

2
e
u
2
/2
.
It is known that () = 1.
58 CHAPTER 5. INTEGRATION TECHNIQUES
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
-20 -15 -10 -5 0 5 10 15 20
atan(x)
Figure 5.7.3: Graph of tan
1
(x)
What does this last statement mean? It means that
lim
t
1

2
_
0
t
e
u
2
udu + lim
x
1

2
_
x
0
e
u
2
udu = 1.
Example 5.7.5. Consider
_

xdx. Notice that

_

xdx = lim
s
_
0
s
xdx + lim
t
_
t
0
xdx.
This diverges since each factor diverges independtly. But notice that
lim
t
_
t
t
xdx = 0.
This is not what
_

xdx means (in this course in a later course it could be interpreted

this way)! This illustrates the importance of treating each bad point separately (since
Example 5.7.3) doesnt.
Example 5.7.6. Consider
_
1
1
1
3

x
dx. We have
_
1
1
1
3

x
dx = lim
s0

_
s
1
x

1
3
dx + lim
t0
+
_
1
t
x

1
3
dx
= lim
s0

_
3
2
s
2
3

3
2
_
+ lim
t0
+
_
3
2

3
2
t
2
3
_
= 0.
This illustrates how to be careful and break the function up into two pieces when there
is a discontinuity.
NOTES for 2006-02-22
Midterm 2: Wednesday, March 1, 2006, at 7pm in Pepper Canyon 109
Today: 7.8: Comparison of Improper integrals
11.1: Sequences
Next 11.2 Series
5.7. IMPROPER INTEGRALS 59
Example 5.7.7. Compute
_
3
1
1
x2
dx. A few weeks ago you might have done this:
_
3
1
1
x 2
dx = [ln |x 2|]
3
1
= ln(3) ln(1) (totally wrong!)
This is not valid because the function we are integrating has a pole at x = 2 (see
Figure 5.7.4). The integral is improper, and is only dened if both the following limits
exists:
lim
t2

_
t
1
1
x 2
dx and lim
t2
+
_
3
t
1
x 2
dx.
However, the limits diverge, e.g.,
lim
t2
+
_
3
t
1
x 2
dx = lim
t2
+
(ln |1| ln |t 2|) = lim
t2
+
ln |t 2| = .
Thus
_
3
1
1
x2
dx is divergent.
-100
-80
-60
-40
-20
0
20
40
-1 -0.5 0 0.5 1 1.5 2 2.5 3
1/(x-2)
Figure 5.7.4: Graph of
1
x2
5.7.1 Convergence, Divergence, and Comparison
In this section we discuss using comparison to determine if an improper integrals con-
verges or diverges. Recall that if f and g are continuous functions on an interval [a, b]
and g(x) f(x), then
_
b
a
g(x)dx
_
b
a
f(x)dx.
This observation can be incredibly useful in determining whether or not an improper
integral converges.
Not only does this technique help in determing whether integrals converge, but it
also gives you some information about their values, which is often much easier to obtain
than computing the exact integral.
60 CHAPTER 5. INTEGRATION TECHNIQUES
Theorem 5.7.8 (Comparison Theorem (special case)). Let f and g be continuous
functions with 0 g(x) f(x) for x a.
1. If
_

a
f(x)dx converges, then
_

a
g(x)dx converges.
2. If
_

a
g(x)dx diverges then
_

a
f(x)dx diverges.
Proof. Since g(x) 0 for all x, the function
G(t) =
_
t
a
g(x)dx
is a non-decreasing function. If
_

a
f(x)dx converges to some value B, then for any
t a we have
G(t) =
_
t
a
g(x)dx
_
t
a
f(x)dx B.
Thus in this case G(t) is a non-decreasing function bounded above, hence the limit
lim
t
G(t) exists. This proves the rst statement.
Likewise, the function
F(t) =
_
t
a
f(x)dx
is also a non-decreasing function. If
_

a
g(x)dx diverges then the function G(t) dened
above is still non-decreasing and lim
t
G(t) does not exist, so G(t) is not bounded.
Since g(x) f(x) we have G(t) F(t) for all a, hence F(t) is also unbounded, which
proves the second statement.
The theorem is very intuitive if you think about areas under a graph. If the bigger
integral converges then so does the smaller one, and if the smaller one diverges so does
the bigger ones.
Example 5.7.9. Does
_

0
cos
2
(x)
1+x
2
dx converge? Answer: YES.
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 2 4 6 8 10
cos(x)**2/(1+x**2)
1/(1+x**2)
Figure 5.7.5: Graph of
cos(x)
2
1+x
2
and
1
1+x
2
Since 0 cos
2
(x) 1, we really do have
0
cos
2
(x)
1 +x
2

1
1 +x
2
,
5.7. IMPROPER INTEGRALS 61
as illustrated in Figure 5.7.5. Thus
_

0
1
1 +x
2
dx = lim
t
tan
1
(t) =

2
,
so
_

0
cos
2
(x)
1+x
2
dx converges.
But why did we use
1
1+x
2
? Its a guess that turned out to work. You could have used
something else, e.g.,
c
x
2
for some constant c. This is an illustration of how in mathematics
sometimes you have to use your imagination or guess and see what happens. Dont get
anxiousinstead, relax, take a deep breath and explore.
For example, alternatively we could have done the following:
_

1
cos
2
(x)
1 +x
2
dx
_

1
1
x
2
dx = 1,
and this works just as well, since
_
1
0
cos
2
(x)
1+x
2
dx converges (as
cos
2
(x)
1+x
2
is continuous).
Example 5.7.10. Consider
_

0
1
x+e
2x
dx. Does it converge or diverge? For large values
of x, the term e
2x
very quickly goes to 0, so we expect this to diverge, since
_

1
1
x
dx
diverges. For x 0, we have e
2x
1, so for all x we have
1
x +e
2x

1
x + 1
(verify by cross multiplying).
But
_

1
1
x + 1
dx = lim
t
[ln(x + 1)]
t
1
=
Thus
_

0
1
x+e
2x
dx must also diverge.
Note that there is a natural analogue of Theorem 5.7.8 for integrals of functions that
blow up at a point, but we will not state it formally.
Example 5.7.11. Consider
_
1
0
e
x

x
dx = lim
t0
+
_
1
t
e
x

x
dx.
We have
e
x

x

1

x
.
(Coming up with this comparison might take some work, imagination, and trial and
error.) We have
_
1
0
e
x

x
dx
_
1
0
1

x
dx = lim
t0
+
2

1 2

t = 2.
thus
_
1
0
e
x

x
dx converges, even though we havent gured out its value. We just know
that it is 2. (In fact, it is 1.493648265 . . ..)
What if we found a function that is bigger than
e
x

x
and its integral diverges?? So
what! This does nothing for you. Bzzzt. Try again.
62 CHAPTER 5. INTEGRATION TECHNIQUES
Example 5.7.12. Consider the integral
_
1
0
e
x
x
dx.
This is an improper integral since f(x) =
e
x
x
has a pole at x = 0. Does it converge?
NO.
On the interal [0, 1] we have e
x
e
1
. Thus
lim
t0
+
_
1
t
e
x
x
dx lim
t0
+
_
1
t
e
1
x
dx
= e
1
lim
t0
+
_
1
t
1
x
dx
= e
1
lim
t0
+
ln(1) ln(t) = +
Thus
_
1
0
e
x
x
dx diverges.
Chapter 6
Sequences and Series
Exam 2: Wednesday at 7pm in PCYN 109
Today: Sequence and Series (11.1-11.2)
Next: 11.3 Integral Test, 11.4 Comparison Test
Our main goal in this chapter is to gain a working knowledge of power series and
Taylor series of function with just enough discussion of the details of convergence to get
by.
6.1 Sequences
What is
lim
n
1
2
n
?
You may have encountered sequences long ago in earlier courses and they seemed
very dicult. You know much more mathematics now, so they will probably seem
easier. On the other hand, were going to go very quickly.
We will completely skip several topics from Chapter 11. I will try to make what we skip clear.
Note that the homework has been modied to reect the omitted topics.
A sequence is an ordered list of numbers. These numbers may be real, complex, etc.,
etc., but in this book we will focus entirely on sequences of real numbers. For example,
1
2
,
1
4
,
1
8
,
1
16
,
1
32
,
1
64
,
1
128
, . . . ,
1
2
n
, . . .
Since the sequence is ordered, we can view it as a function with domain the natural
numbers = 1, 2, 3, . . ..
Denition 6.1.1 (Sequence). A sequence {a
n
} is a function a : N R that takes a
natural number n to a
n
= a(n). The number a
n
is the nth term.
63
64 CHAPTER 6. SEQUENCES AND SERIES
For example,
a(n) = a
n
=
1
2
n
,
which we write as {
1
2
n
}. Heres another example:
(b
n
)

n=1
=
_
n
n + 1
_

n=1
=
1
2
,
2
3
,
3
4
, . . .
Example 6.1.2. The Fibonacci sequence (F
n
)

n=1
is dened recursively as follows:
F
1
= 1, F
2
= 1, F
n
= F
n2
+F
n1
for n 3.
Lets return to the sequence
_
1
2
n
_

n=1
. We write lim
n
1
2
n
= 0, since the terms get
arbitrarily small.
Denition 6.1.3 (Limit of sequence). If (a
n
)

n=1
is a sequence then that sequence
converges to L, written lim
n
a
n
= L, if a
n
gets arbitrarily close to L as n get
suciently large. Secret rigorous definition: For every > 0 there exists B such
that for n B we have |a
n
L| < .
This is exactly like what we did in the previous course when we considered limits
of functions. If f(x) is a function, the meaning of lim
x
f(x) = L is essentially the
same. In fact, we have the following fact.
Proposition 6.1.4. If f is a function with lim
x
f(x) = L and (a
n
)

n=1
is the se-
quence given by a
n
= f(n), then lim
n
a
n
= L.
As a corollary, note that this implies that all the facts about limits that you know
from functions also apply to sequences!
Example 6.1.5.
lim
n
n
n + 1
= lim
x
x
x + 1
= 1
Example 6.1.6. The converse of Proposition 6.1.4 is false in general, i.e., knowing the
limit of the sequence converges doesnt imply that the limit of the function converges.
We have lim
n
cos(2n) = 1, but lim
x
cos(2x) diverges. The converse is OK if
the limit involving the function converges.
Example 6.1.7. Compute lim
n
n
3
+n + 5
17n
3
2006n + 15
. Answer:
1
17
.
6.2 Series
What is
1
2
+
1
4
+
1
8
+
1
16
+
1
32
+. . .?
What is
1
3
+
1
9
+
1
27
+
1
81
+
1
243
+. . .?
What is
1
1
+
1
4
+
1
9
+
1
16
+
1
25
+. . .?
6.2. SERIES 65
Consider the following sequence of partial sums:
a
N
=
N

n=1
1
2
n
=
1
2
+
1
4
+ +
1
2
N
.
Can we compute

n=1
1
2
n
?
These partial sums look as follows:
a
1
=
1
2
, a
2
=
3
4
, a
10
=
1023
1024
, a
20
=
1048575
1048576
It looks very likely that

n=1
1
2
n
= 1, if it makes any sense. But does it?
In a moment we will dene

n=1
1
2
n
= lim
N
N

n=1
1
2
n
= lim
N
a
N
.
A little later we will show that a
N
=
2
N
1
2
N
, hence indeed

n=1
1
2
n
= 1.
Denition 6.2.1 (Sum of series). If (a
n
)

n=1
is a sequence, then the sum of the series
is

n=1
a
n
= lim
N
N

n=1
a
n
= lim
N
s
N
provided the limit exists. Otherwise we say that

n=1
a
n
diverges.
Example 6.2.2 (Geometric series). Consider the geometric series

n=1
ar
n1
for
a = 0. Then
s
N
=
N

n=1
ar
n1
=
a(1 r
N
)
1 r
.
To see this, multiply both sides by 1 r and notice that all the terms in the middle
cancel out. For what values of r does lim
N
a(1r
N
)
1r
converge? If |r| < 1, then
lim
N
r
N
= 0 and
lim
N
a(1 r
N
)
1 r
=
a
1 r
.
If |r| > 1, then lim
N
r
N
diverges, so

n=1
ar
n1
diverges. If r = 1, its clear since
a = 0 that the series also diverges (since the partial sums are s
N
= Na).
For example, if a = 1 and r =
1
2
, we get

n=1
ar
n1
=
1
1
1
2
,
as claimed earlier.
66 CHAPTER 6. SEQUENCES AND SERIES
6.3 The Integral and Comparison Tests
Midterm Exam 2: Wednesday March 1 at 7pm in PCYNH 109 (up to last lecture)
Today: 7.37.4: Integral and comparison tests
Next: 7.6: Absolute convergence; ratio and root tests
Quiz 4 (last quiz): Friday March 10.
Final exam: Wednesday, March 22, 7-10pm in PCYNH 109.
What is

n=1
1
n
2
? What is

n=1
1
n
?
Recall that Section 6.2 began by asking for the sum of several series. We found the
rst two sums (which were geometric series) by nding an exact formula for the sum
s
N
of the rst N terms. The third series was
A =

n=1
1
n
2
=
1
1
+
1
4
+
1
9
+
1
16
+
1
25
+. . . . (6.3.1)
It is dicult to nd a nice formula for the sum of the rst n terms of this series (i.e., I
dont know how to do it).
Remark 6.3.1. Since Im a number theorist, I cant help but make some further re-
marks about sums of the form (6.3.1). In general, for any s > 1 one can consider the
sum
(s) =

n=1
1
n
s
.
The number A that we are interested in above is thus (2). The function (s) is called
the Riemann zeta function. There is a natural (but complicated) way of extending (s)
to a (dierentiable) function on all complex numbers with a pole at s = 1. The Riemann
Hypothesis asserts that if s is a complex number and (s) = 0 then either s is an even
negative integer or s =
1
2
+bi for some real number b. This is probably the most famous
unsolved problems in mathematics (e.g., its one of the Clay Math Institute million
dollar prize problems). Another famous open problem is to show that (3) is not a root
of any polynomial with integer coecients (it is a theorem of Apeery that zeta(3) is not
a fraction).
The function (s) is incredibly important in mathematics because it governs the
properties of prime numbers. The Euler product representation of (s) gives a hint as
to why this is the case:
(s) =

n=1
1
n
s
=

primes p
_
1
1 p
s
_
.
To see that this product equality holds when s is real with Re(s) > 1, use Example 6.2.2
with r = p
s
and a = 1 from the previous lecture. We have
1
1 p
s
= 1 +p
s
+p
2s
+ .
6.3. THE INTEGRAL AND COMPARISON TESTS 67
Thus

primes p
_
1
1 p
s
_
=

primes p
_
1 +
1
p
s
+
1
p
2s
+
_
=
_
1 +
1
2
s
+
1
2
2s
+
_

_
1 +
1
3
s
+
1
3
2s
+
_

=
_
1 +
1
2
s
+
1
3
s
+
1
4
s
+
_
=

n=1
1
n
s
,
where the last line uses the distributive law and that integers factor uniquely as a
product of primes.
Finally, Figure 6.3.1 is a graph (x) as a function of a real variable x, and Figure 6.3.2
is a graph of |(s)| for complex s.
Figure 6.3.1: Riemann Zeta Function: f(x) =

n=1
1
n
x
This section is how to leverage what youve learned so far in this book to say some-
thing about sums that are hard (or even impossibly dicult) to evaluate exactly. For
example, notice (by considering a graph of a step function) that if f(x) = 1/x
2
, then
for positive integer t we have
t

n=1
1
n
2

1
1
2
+
_
t
1
1
x
2
dx.
68 CHAPTER 6. SEQUENCES AND SERIES
Figure 6.3.2: Absolute Value of Riemann Zeta Function
Thus

n=1
1
n
2

1
1
2
+
_

1
1
x
2
dx
= 1 + lim
t
_
t
1
1
x
2
dx
= 1 + lim
t
_

1
x
_
t
1
= 1 + lim
t
_

1
t
+
1
1
_
= 2
We conclude that

n=1
converges, since the sequence of partial sums is getting bigger
and bigger and is always 2. And of course we also know something about

n=1
1
n
2
even though we do not know the exact value:

n=1
1
n
2
2. Using a computer we nd
that
t

t
n=1
1
n
2
1 1
2
5
4
= 1.25
5
5269
3600
= 1.46361
10
1968329
1270080
= 1.54976773117
100 1.63498390018
1000 1.64393456668
10000 1.64483407185
100000 1.6449240669
The table is consistent with the fact that

n=1
1
n
2
converges to a number 2. In fact
Euler was the rst to compute

n=1
exactly; he found that the exact value is

2
6
= 1.644934066848226436472415166646025189218949901206798437735557 . . .
There are many proofs of this fact, but they dont belong in this book; you can nd
them on the internet, and are likely to see one if you take more math classes.
6.3. THE INTEGRAL AND COMPARISON TESTS 69
We next consider the harmonic series

n=1
1
n
. (6.3.2)
Does it converge? Again by inspecting a graph and viewing an innite sum as the area
under a step function, we have

n=1
1
n

_

1
1
x
dx
= lim
t
[ln(x)]
t
1
= lim
t
ln(t) 0 = +.
Thus the innite sum (6.3.2) must also diverge.
We formalize the above two examples as a general test for convergence or divergence
of an innite sum.
Theorem 6.3.2 (Integral Test and Bound). Suppose f(x) is a continuous, positive,
decreasing function on [1, ) and let a
n
= f(n) for integers n 1. Then the series

n=1
a
n
converges if and only if the integral
_

1
f(x)dx converges. More generally, for
any positive integer k,
_

k
f(x)dx

n=k
a
n
a
k
+
_

k
f(x)dx. (6.3.3)
The proposition means that you can determine convergence of an innite series by
determining convergence of a corresponding integral. Thus you can apply the powerful
tools you know already for integrals to understanding innite sums. Also, you can use
integration along with computation of the rst few terms of a series to approximate a
series very precisely.
Remark 6.3.3. Sometimes the rst few terms of a series are funny or the series
doesnt even start at n = 1, e.g.,

n=4
1
(n 3)
3
.
In this case use (6.3.3) with any specic k > 1.
Proposition 6.3.4 (Comparison Test). Suppose

a
n
and

b
n
are two series with
positive terms. If

b
n
converges and a
n
b
n
for all n. then

a
n
converges. Likewise,
if

b
n
diverges and a
n
b
n
for all n. then

a
n
must also diverge.
Example 6.3.5. Does

n=1
1

n
converge? No. We have

n=1
1

n

_

1
1

x
dx = lim
t
(2

t 2

1) = +
Example 6.3.6. Does

n=1
1
n
2
+1
converge? Lets apply the comparison test: we have
1
n
2
+1
<
1
n
2
for every n, so

n=1
1
n
2
+ 1
<

n=1
1
n
2
.
70 CHAPTER 6. SEQUENCES AND SERIES
Alternatively, we can use the integral test, which also gives as a bonus an upper and
lower bound on the sum. Let f(x) = 1/(1 +x
2
). We have
_

1
1
1 +x
2
dx = lim
t
_
t
1
1
1 +x
2
dx
= lim
t
tan
1
(t)

4
=

2


4
=

4
Thus the sum converges. Moreover, taking k = 1 in Theorem 6.3.2 we have

n=1
1
n
2
+ 1

1
2
+

4
.
the actual sum is 1.07 . . ., which is much dierent than

1
n
2
= 1.64 . . ..
We could prove the following proposition using methods similar to those illustrated
in the examples above. Note that this is nicely illustrated in Figure 6.3.1.
Proposition 6.3.7. The series

n=1
1
n
p
is convergent if p > 1 and divergent if p 1.
6.3.1 Estimating the Sum of a Series
Suppose

a
n
is a convergent sequence of positive integers. Let
R
m
=

n=1
a
n

n=1
a
n
=

n=m+1
a
m
which is the error if you approximate

a
n
using the rst n terms. From Theorem 6.3.2
we get the following.
Proposition 6.3.8 (Remainder Bound). Suppose f is a continuous, positive, de-
creasing function on [m, ) and

a
n
is convergent. Then
_

m+1
f(x)dx R
m

_

m
f(x)dx.
Proof. In Theorem 6.3.2 set k = m+ 1. That gives
_

m+1
f(x)dx

n=m+1
a
n
a
m+1
+
_

m+1
f(x)dx.
But
a
m+1
+
_

m+1
f(x)dx
_

m
f(x)dx
since f is decreasing and f(m+ 1) = a
m+1
.
Example 6.3.9. Estimate (3) =

n=1
1
n
3
using the rst 10 terms of the series. We
have
10

n=1
=
19164113947
16003008000
= 1.197531985674193 . . .
6.4. TESTS FOR CONVERGENCE 71
The proposition above with m = 10 tells us that
0.00413223140495867 . . . =
_

11
1
x
3
dx (3)
10

n=1

_

10
1
x
3
dx =
1
2 10
2
=
1
200
= 0.005.
In fact,
(3) = 1.202056903159594285399738161511449990 . . .
and we hvae
(3)
10

n=1
= 0.0045249174854010 . . . ,
so the integral error bound was really good in this case.
Example 6.3.10. Determine if

n=1
2006
117n
2
+41n+3
convergers or diverges. Answer: It
converges, since
2006
117n
2
+ 41n + 3

2006
117n
2
=
2006
117

1
n
2
,
and

1
n
2
converges.
6.4 Tests for Convergence
Final exam: Wednesday, March 22, 7-10pm in PCYNH 109.
Quiz 4: Next Friday
Today: 11.6: Ratio and Root tests
Next: 11.8 Power Series
11.9 Functions dened by power series
6.4.1 The Comparison Test
Theorem 6.4.1 (The Comparison Test). Suppose

a
n
and

b
n
are series with
all a
n
and b
n
positive and a
n
b
n
for each n.
1. If

b
n
converges, then so does

a
n
.
2. If

a
n
diverges, then so does

b
n
.
Proof Sketch. The condition of the theorem implies that for any k,
k

n=1
a
n

k

n=1
b
n
,
from which each claim follows.
Example 6.4.2. Consider the series

n=1
7
3n
2
+2n
. For each n we have
7
3n
2
+ 2n

7
3

1
n
2
.
Since

n=1
1
n
2
converges, Theorem 6.4.1 implies that

n=1
7
3n
2
+2n
also converges.
72 CHAPTER 6. SEQUENCES AND SERIES
Example 6.4.3. Consider the series

n=1
ln(n)
n
. It diverges since for each n 3 we
have
ln(n)
n

1
n
,
and

n=3
1
n
diverges.
6.4.2 Absolute and Conditional Convergence
Denition 6.4.4 (Converges Absolutely). We say that

n=1
a
n
converges abso-
lutely if

n=1
|a
n
| converges.
For example,

n=1
(1)
n
1
n
converges, but does not converge absolutely (it converges conditionally, though we
will not explain why in this book).
6.4.3 The Ratio Test
Recall that

n=1
a
n
is a geometric series if and only if a
n
= ar
n1
for some xed a
and r. Here we call r the common ratio. Notice that the ratio of any two successive
terms is r:
a
n+1
a
n
=
ar
n
ar
n1
= r.
Moreover, we have

n=1
ar
n1
converges (to
a
1r
) if and only if |r| < 1 (and, of course
it diverges if |r| 1).
Example 6.4.5. For example,

n=1
3
_
2
3
_
n1
converges to
3
1
2
3
= 9. However,

n=1
3
_
3
2
_
n1
diverges.
Theorem 6.4.6 (Ratio Test). Consider a sum

n=1
a
n
. Then
1. If lim
n

an+1
an

= L < 1 then

n=1
a
n
is absolutely convergent.
2. If lim
n

an+1
an

= L > 1 then

n=1
a
n
diverges.
3. If lim
n

an+1
an

= L = 1 then we may conclude nothing from this!

Proof. We will only prove 1. Assume that we have lim
n

an+1
an

= L < 1. Let
r =
L+1
2
, and notice that L < r < 1 (since 0 L < 1, so 1 L+1 < 2, so 1/2 r < 1,
and also r L = (L + 1)/2 L = (1 L)/2 > 0).
Since lim
n

an+1
an

= L, there is an N such that for all n > N we have

a
n+1
a
n

< r, so |a
n+1
| < |a
n
| r.
Then we have

n=N+1
|a
n
| < |a
N+1
|

n=0
r
n
.
Here the common ratio for the second one is r < 1, hence thus the right-hand series
converges, so the left-hand series converges.
6.4. TESTS FOR CONVERGENCE 73
Example 6.4.7. Consider

n=1
(10)
n
n!
. The ratio of successive terms is

(10)
n+1
(n + 1)!
(10)
n
n!

=
10
n+1
(n + 1)n!

n!
10
n
=
10
n + 1
0 < 1.
Thus this series converges absolutely. Note, the minus sign is missing above since in the
ratio test we take the limit of the absolute values.
Example 6.4.8. Consider

n=1
n
n
3
1+3n
. We have

(n + 1)
n+1
3 (27)
n+1
n
n
3
1+3n

=
(n + 1)(n + 1)
n
27 27
n

27
n
n
n
=
n + 1
27

_
n + 1
n
_
n
+
Thus our series diverges. (Note here that we use that
_
n+1
n
_
n
e.)
Example 6.4.9. Lets apply the ratio test to

n=1
1
n
. We have
lim
n

1
n + 1
1
n

=
1
n + 1

n
1
=
n
n + 1
1.
This tells us nothing. If this happens... do something else! E.g., in this case, use the
integral test.
6.4.4 The Root Test
Since e and ln are inverses, we have x = e
ln(x)
. This implies the very useful fact that
x
a
= e
ln(x
a
)
= e
a ln(x)
.
As a sample application, notice that for any nonzero c,
lim
n
c
1
n
= lim
n
e
1
n
log(c)
= e
0
= 1.
Similarly,
lim
n
n
1
n
= lim
n
e
1
n
log(n)
= e
0
= 1,
where weve used that lim
n
log(n)
n
= 0, which we could prove using LHopitals rule.
Theorem 6.4.10 (Root Test). Consider the sum

n=1
a
n
.
1. If lim
n
|a
n
|
1
n
= L < 1, then

n=1
a
n
convergest absolutely.
2. If lim
n
|a
n
|
1
n
= L > 1, then

n=1
a
n
diverges.
3. If L = 1, then we may conclude nothing from this!
74 CHAPTER 6. SEQUENCES AND SERIES
Proof. We apply the comparison test (Theorem 6.4.1). First suppose lim
n
|a
n
|
1
n
=
L < 1. Then there is a N such that for n N we have |a
n
|
1
n
< k < 1. Thus for such
n we have |a
n
| < k
n
< 1. The geometric series

i=N
k
i
converges, so

i=N
|a
n
| also
does, by Theorem 6.4.1. If |a
n
|
1
n
> 1 for n N, then we see that

i=N
|a
n
| diverges
by comparing with

i=N
1.
Example 6.4.11. Lets apply the root test to

n=1
ar
n1
=
a
r

n=1
r
n
.
We have
lim
n
|r
n
|
1
n
= |r|.
Thus the root test tells us exactly what we already know about convergence of the
geometry series (except when |r| = 1).
Example 6.4.12. The sum

n=1
_
n
2
+1
2n
2
+1
_
n
is a candidate for the root test. We have
lim
n

_
n
2
+ 1
2n
2
+ 1
_
n

1
n
= lim
n
n
2
+ 1
2n
2
+ 1
= lim
n
1 +
1
n
2
2 +
1
n
2
=
1
2
.
Thus the series converges.
Example 6.4.13. The sum

n=1
_
2n
2
+1
n
2
+1
_
n
is a candidate for the root test. We have
lim
n

_
2n
2
+ 1
n
2
+ 1
_
n

1
n
= lim
n
2n
2
+ 1
n
2
+ 1
= lim
n
2 +
1
n
2
1 +
1
n
2
= 2,
hence the series diverges!
Example 6.4.14. Consider

n=1
1
n
. We have
lim
n

1
n

1
n
= 1,
so we conclude nothing!
Example 6.4.15. Consider

n=1
n
n
3(27
n
)
. To apply the root test, we compute
lim
n

n
n
3 (27
n
)

1
n
= lim
n
_
1
3
_1
n

n
27
= +.
Again, the limit diverges, as in Example 6.4.8.
6.5. POWER SERIES 75
6.5 Power Series
Final exam: Wednesday, March 22, 7-10pm in PCYNH 109. Bring ID!
Quiz 4: This Friday
Today: 11.8 Power Series, 11.9 Functions dened by power series
Next: 11.10 Taylor and Maclaurin series
Recall that a polynomial is a function of the form
f(x) = c
0
+c
1
x +c
2
x
2
+ +c
k
x
k
.
Polynomials are easy!!!
They are easy to integrate, dierentiate, etc.:
d
dx
_
k

n=0
c
n
x
n
_
=
k

n=1
nc
n
x
n1
_
k

n=0
c
n
x
n
dx = C +
k

n=0
c
n
x
n+1
n + 1
.
Denition 6.5.1 (Power Series). A power series is a series of the form
f(x) =

n=0
c
n
x
n
= c
0
+c
1
x +c
2
x
2
+ ,
where x is a variable and the c
n
are coecients.
A power series is a function of x for those x for which it converges.
Example 6.5.2. Consider
f(x) =

n=0
x
n
= 1 +x +x
2
+ .
When |x| < 1, i.e., 1 < x < 1, we have
f(x) =
1
1 x
.
But what good could this possibly be? Why is writing the simple function
1
1x
as
the complicated series

n=0
x
n
of any value?
1. Power series are relatively easy to work with. They are almost polynomials.
E.g.,
d
dx

n=0
x
n
=

n=1
nx
n1
= 1 + 2x + 3x
2
+ =

m=0
(m+ 1)x
m
,
where in the last step we re-indexed the series. Power series are only almost
polynomials, since they dont stop; they can go on forever. More precisely, a
76 CHAPTER 6. SEQUENCES AND SERIES
power series is a limit of polynomials. But in many cases we can treat them like
a polynomial. On the other hand, notice that
d
dx
_
1
1 x
_
=
1
(1 x)
2
=

m=0
(m+ 1)x
m
.
2. For many functions, a power series is the best explicit representation available.
Example 6.5.3. Consider J
0
(x), the Bessel function of order 0. It arises as a
solution to the dierential equation x
2
y

+ xy

+ x
2
y = 0, and has the following
power series expansion:
J
0
(x) =

n=1
(1)
n
x
2n
2
2n
(n!)
2
= 1
1
4
x
2
+
1
64
x
4

1
2304
x
6
+
1
147456
x
8

1
14745600
x
10
+ .
This series is nice since it converges for all x (one can prove this using the ratio
test). It is also one of the most explicit forms of J
0
(x).
6.5.1 Shift the Origin
It is often useful to shift the origin of a power series, i.e., consider a power series expanded
about a dierent point.
Denition 6.5.4. The series

n=0
c
n
(x a)
n
= c
0
+c
1
(x a) +c
2
(x a)
2
+
is called a power series centered at x = a, or a power series about x = a.
Example 6.5.5. Consider

n=0
(x 3)
n
= 1 + (x 3) + (x 3)
2
+
=
1
1 (x 3)
equality valid when |x 3| < 1
=
1
4 x
Here conceptually we are treating 3 like we treated 0 before.
Power series can be written in dierent ways, which have dierent advantages and
disadvantages. For example,
1
4 x
=
1
4

1
1 x/4
=
1
4

n=0
_
x
4
_
n
converges for all |x| < 4.
Notice that the second series converges for |x| < 4, whereas the rst converges only for
|x 3| < 1, which isnt nearly as good.
6.5. POWER SERIES 77
6.5.2 Convergence of Power Series
Theorem 6.5.6. Given a power series

n=0
c
n
(x a)
n
, there are exactly three possi-
bilities:
1. The series conveges only when x = a.
2. The series conveges for all x.
3. There is an R > 0 (called the radius of convergence) such that

n=0
c
n
(xa)
n
converges for |x a| < R and diverges for |x a| > R.
Example 6.5.7. For the power series

n=0
x
n
, the radius R of convergence is 1.
Denition 6.5.8 (Radius of Convergence). As mentioned in the theorem, R is
called the radius of convergence.
If the series converges only at x = a, we say R = 0, and if the series converges
everywhere we say that R = .
The interval of convergence is the set of x for which the series converges. It will be
one of the following:
(a R, a +R), [a R, a +R), (a R, a +R], [a R, a +R]
The point being that the statement of the theorem only asserts something about conver-
gence of the series on the open interval (a R, a +R). What happens at the endpoints
of the interval is not specied by the theorem; you can only gure it out by looking
explicitly at a given series.
Theorem 6.5.9. If

n=0
c
n
(x a)
n
has radius of convergence R > 0, then f(x) =

n=0
c
n
(x a)
n
is dierentiable on (a R, a +R), and
1. f

(x) =

n=1
n c
n
(x a)
n1
2.
_
f(x)dx = C +

n=0
c
n
n + 1
(x a)
n+1
,
and both the derivative and integral have the same radius of convergence as f.
Example 6.5.10. Find a power series representation for f(x) = tan
1
(x). Notice that
f

(x) =
1
1 +x
2
=
1
1 (x
2
)
=

n=0
(1)
n
x
2n
,
which has radius of convergence R = 1, since the above series is valid when | x
2
| < 1,
i.e., |x| < 1. Next integrating, we nd that
f(x) = c +

n=0
(1)
n
x
2n+1
2n + 1
,
for some constant c. To nd the constant, compute c = f(0) = tan
1
(0) = 0. We
conclude that
tan
1
(x) =

n=0
(1)
n
x
2n+1
2n + 1
.
78 CHAPTER 6. SEQUENCES AND SERIES
Example 6.5.11. We will see later that the function f(x) = e
x
2
has power series
e
x
2
= 1 x
2
+
1
2
x
4

1
6
x
6
+ .
Hence _
e
x
2
dx = c +x
1
3
x
3
+
1
10
x
5

1
42
x
7
+ .
This despite the fact that the antiderivative of e
x
2
is not an elementary function (see
Example 5.7.4).
6.6 Taylor Series
Final exam: Wednesday, March 22, 7-10pm in PCYNH 109. Bring ID!
Last Quiz 4: This Friday
Next: 11.10 Taylor and Maclaurin series
Next: 11.12 Applications of Taylor Polynomials
Midterm Letters:
A, 3238
B, 2631
C, 2025
D, 1419
Mean: 23.4, Standard Deviation: 7.8, High: 38, Low: 6.
Example 6.6.1. Suppose we have a degree-3 (cubic) polynomial p and we know that
p(0) = 4, p

(0) = 3, p

(0) = 4, and p

(0) = 6. Can we determine p? Answer: Yes! We

have
p(x) = a +bx +cx
2
+dx
3
p

(x) = b + 2cx + 3dx

2
p

(x) = 2c + 6dx
p

(x) = 6d
From what we mentioned above, we have:
a = p(0) = 4
b = p

(0) = 3
c =
p

(0)
2
= 2
d =
p

(0)
6
= 1
Thus
p(x) = 4 + 3x + 2x
2
+x
3
.
Amazingly, we can use the idea of Example 6.6.1 to compute power series expansions
of functions. E.g., we will show below that
e
x
=

n=0
x
n
n!
.
6.6. TAYLOR SERIES 79
Convergent series are determined by the values of their derivatives.
Consider a general power series
f(x) =

n=0
c
n
(x a)
n
= c
0
+c
1
(x a) +c
2
(x a)
2
+
We have
c
0
= f(a)
c
1
= f

(a)
c
2
=
f

(a)
2

c
n
=
f
(n)
(a)
n!
,
where for the last equality we use that
f
(n)
(x) = n!c
n
+ (x a)( + )
Remark 6.6.2. The denition of 0! is 1 (its the empty product). The empty sum is 0
and the empty product is 1.
Theorem 6.6.3 (Taylor Series). If f(x) is a function that equals a power series
centered about a, then that power series expansion is
f(x) =

n=0
f
(n)
(a)
n!
(x a)
n
= f(a) +f

(a)(x a) +
f

(a)
2
(x a)
2
+
Remark 6.6.4. WARNING: There are functions that have all derivatives dened, but
do not equal their Taylor expansion. E.g., f(x) = e
1/x
2
for x = 0 and f(0) = 0.
Its Taylor expansion is the 0 series (which converges everywhere), but it is not the 0
function.
Denition 6.6.5 (Maclaurin Series). A Maclaurin series is just a Taylor series with
a = 0. I will not use the term Maclaurin series ever again (its common in textbooks).
Example 6.6.6. Find the Taylor series for f(x) = e
x
about a = 0. We have f
(n)
(x) =
e
x
. Thus f
(n)
(0) = 1 for all n. Hence
e
x
=

n=0
1
n!
x
n
= 1 +x +
x
2
2
+
x
3
6
+
What is the radius of convergence? Use the ratio test:
lim
n

1
(n+1)!
x
n+1
1
n!
x
n

= lim
n
n!
(n + 1)!
|x|
= lim
n
|x|
n + 1
= 0, for any xed x.
Thus the radius of convergence is .
80 CHAPTER 6. SEQUENCES AND SERIES
Example 6.6.7. Find the Taylor series of f(x) = sin(x) about x =

2
.
1
We have
f(x) =

n=0
f
(n)
_

2
_
n!
_
x

2
_
n
.
To do this we have to puzzle out a pattern:
f(x) = sin(x)
f

(x) = cos(x)
f

(x) = sin(x)
f

(x) = cos(x)
f
(4)
(x) = sin(x)
First notice how the signs behave. For n = 2m even,
f
(n)
(x) = f
(2m)
(x) = (1)
n/2
sin(x)
and for n = 2m+ 1 odd,
f
(n)
(x) = f
(2m+1)
(x) = (1)
m
cos(x) = (1)
(n1)/2
cos(x)
For n = 2m even we have
f
(n)
(/2) = f
(2m)
_

2
_
= (1)
m
.
and for n = 2m+ 1 odd we have
f
(n)
(/2) = f
(2m+1)
_

2
_
= (1)
m
cos(/2) = 0.
Finally,
sin(x) =

n=0
f
(n)
(/2)
n!
(x /2)
n
=

m=0
(1)
m
(2m)!
_
x

2
_
2m
.
Next we use the ratio test to compute the radius of convergence. We have
lim
m

(1)
m+1
(2(m+ 1))!
_
x

2
_
2(m+1)

(1)
m
(2m)!
_
x

2
_
2m

= lim
m
(2m)!
(2m+ 2)!
_
x

2
_
2
= lim
m
_
x

2
_
2
(2m+ 2)(2m+ 1)
which converges for each x. Hence R = .
1
Evidently this expansion was rst found in India by Madhava of Sangamagrama (1350-1425).
6.6. TAYLOR SERIES 81
Example 6.6.8. Find the Taylor series for cos(x) about a = 0. We have cos(x) =
sin
_
x +

2
_
. Thus from Example 6.6.7 (with innite radius of convergence) and that
the Taylor expansion is unique, we have
cos(x) = sin
_
x +

2
_
=

n=0
(1)
n
(2n)!
_
x +

2


2
_
2n
=

n=0
(1)
n
(2n)!
x
2n
82 CHAPTER 6. SEQUENCES AND SERIES
6.7 Applications of Taylor Series
Final exam: Wednesday, March 22, 7-10pm in PCYNH 109. Bring ID!
Last Quiz 4: Today (last one)
Today: 11.12 Applications of Taylor Polynomials
Next; Dierential Equations
This section is about an example in the theory of relativity. Let mbe the (relativistic)
mass of an object and m
0
be the mass at rest (rest mass) of the object. Let v be the
velocity of the object relative to the observer, and let c be the speed of light. These
three quantities are related as follows:
m =
m
0
_
1
v
2
c
2
(relativistic) mass
The total energy of the object is mc
2
:
E = mc
2
.
In relativity we dene the kinetic energy to be
K = mc
2
m
0
c
2
. (6.7.1)
What? Isnt the kinetic energy
1
2
m
0
v
2
?
Notice that
mc
2
m
0
c
2
=
m
0
c
2
_
1
v
2
c
2
m
0
c
2
= m
0
c
2
_
_
1
v
2
c
2
_

1
2
1
_
.
Let
f (x) = (1 x)

1
2
1
Lets compute the Taylor series of f. We have
f(x) = (1 x)

1
2
1
f

(x) =
1
2
(1 x)

3
2
f

(x) =
1
2

3
2
(1 x)

5
2
f
(n)
(x) =
1 3 5 (2n 1)
2
n
(1 x)

2n+1
2
.
Thus
f
(n)
(0) =
1 3 5 (2n 1)
2
n
.
6.7. APPLICATIONS OF TAYLOR SERIES 83
Hence
f(x) =

n=1
f
(n)
(0)
n!
x
n
=

n=1
1 3 5 (2n 1)
2
n
n!
x
n
=
1
2
x +
3
8
x
2
+
5
16
x
3
+
35
128
x
4
+
We now use this to analyze the kinetic energy (6.7.1):
mc
2
m
0
c
2
= m
0
c
2
f
_
v
2
c
2
_
= m
0
c
2

_
1
2

v
2
c
2
+
3
8

v
2
c
2
+
_
=
1
2
m
0
v
2
+m
0
c
2

_
3
8
v
2
c
2
+
_
And we can ignore the higher order terms if
v
2
c
2
is small. But how small is small
enough, given that
v
2
c
2
appears in an innite sum?
6.7.1 Estimation of Taylor Series
Suppose
f(x) =

n=0
f
(n)
(a)
n!
(x a)
n
.
Write
R
N
(x) := f(x)
N

n=0
f
(n)
(a)
n!
(x a)
n
We call
T
N
(x) =
N

n=0
f
(n)
(a)
n!
(x a)
n
the Nth degree Taylor polynomial. Notice that
lim
N
T
N
(x) = f(x)
if and only if
lim
N
R
N
(x) = 0.
We would like to estimate f(x) with T
N
(x). We need an estimate for R
N
(x).
Theorem 6.7.1 (Taylors theorem). If |f
(N+1)
(x)| M for |x a| d, then
|R
N
(x)|
M
(N + 1)!
|x a|
N+1
for |x a| d.
84 CHAPTER 6. SEQUENCES AND SERIES
For example, if N = 0, this says that
|R(x)| = |f(x) f(a)| M|x a|,
i.e.,

f(x) f(a)
x a

M,
which should look familiar from a previous class (Mean Value Theorem).
Applications:
1. One can use Theorem 6.7.1 to prove that functions converge to their Taylor series.
2. Returning to the relativity example above, we apply Taylors theorem with N = 1
and a = 0. With x = v
2
/c
2
and M any number such that |f

(x)| M, we have
|R
1
(x)|
M
2
x
2
.
For example, if we assume that |v| 100m/s we use
|f

(x)|
3
2
(1 100
2
/c
2
)
5/2
= M.
Using c = 3 10
8
m/s, we get
|R
1
(x)| 4.17 10
10
m
0
.
Thus for v 100m/s 225mph, then the error in throwing away relativistic
factors is 10
10
m
0
. This is like 200 feet out of the distance to the sun (93 million
miles). So relativistic and Newtonian kinetic energies are almost the same for
reasonable speeds.
Chapter 7
Some Dierential Equations
Final exam: Wed March 22 7-10pm in Pepper canyon 109.
Today: Section 9.5
Friday: Review (with special guest John Eggers).
Extra Oce Hours: Monday 11-2pm
Introduction not written.
7.1 Separable Equations
A separable dierential equation is a rst order dierential equation that can be written
in the form
dy
dx
=
f(x)
h(y)
.
These can be solved by integration, by noting that
h(y)dy = f(x)dx,
hence _
h(y)dy =
_
f(x)dx.
This latter equation denes y implicitly as a function of x, and in some cases it is
possible to explicitly solve for y as a function of x.
7.2 Logistic Equation
The logistics equation is a dierential equation that models population growth. Often
in practice a dierential equation models some physical situtation, and you should read
it as doing so.
Exponential growth:
1
P
dP
dt
= k.
This says that the relative (percentage) growth rate is constant. As we saw before,
the solutions are
P
(t)
= P
0
e
kt
.
85
86 CHAPTER 7. SOME DIFFERENTIAL EQUATIONS
Note that this model only works for a little while. In everyday life the growth couldnt
actually continue at this rate indenitely. This exponential growth model ignores limi-
tations on resources, disease, etc. Perhaps there is a better model?
Over time we expect the growth rate should level o, i.e., decrease to 0. What about
1
P
dP
dt
= k
_
1
P
K
_
, (7.2.1)
where K is some large constant called the carrying capacity, which is much bigger than
P = P(t) at time 0. The carrying capacity is the maximum population that the environ-
ment can support. Note that if P > K, then dP/dt < 0 so the population declines. The
dierential equation (7.2.1) is called the logistic model (or logistic dierential equation).
There are, of course, other models one could use, e.g., the Gompertz equation.
First question: are there any equilibrium solutions to (7.2.1), i.e., solutions with
dP/dt = 0, i.e., constant solutions? In order that dP/dt = 0 then 0 = k
_
1
P
K
_
, so the
two equilibrium solutions are P(t) = 0 and P(t) = K.
The logistic dierential equation (7.2.1) is separable, so you can separate the vari-
ables with one variable on one side of the equality and one on the other. This means
we can easily solve the equation by integrating. We rewrite the equation as
dP
dt
=
k
K
P(P K).
Now separate:
KdP
P(P K)
= k dt,
and integrate both sides
_
KdP
P(P K)
=
_
k dt = kt +C.
On the left side we get
_
KdP
P(P K)
=
_ _
1
P K

1
P
_
dP = ln |P K| ln |P| +
Thus
ln |K P| ln |P| = kt +c,
so
ln |(K P)/P| = kt +c.
Now exponentiate both sides:
(K P)/P = e
kt+c
= Ae
kt
, where A = e
c
.
Thus
K = P(1 +Ae
kt
),
so
P(t) =
K
1 +Ae
kt
.
Note that A = 0 also makes sense and gives an equilibrium solution. In general we
have lim
t
P(t) = K. In any particular case we can determine A as a function of
P
0
= P(0) by using that
P(0) =
K
1 +A
so A =
K
P
0
1 =
K P
0
P
0
.