Calculus II Lecture Notes, Baylor

Jonathan Stanfill

August 20, 2021

Text: Single Variable Calculus: Early Transcendentals, 4th Edition, Jon Rogawski and Colin Adams

The course covers techniques of integration, applications of integration, and infinite series:

Techniques of Integration
• Review: Substitution [Section 5.7]
• Integration by parts [Section 7.1]
• Trigonometric integrals and trigonometric substitution [Sections 7.2 and 7.3]
• Partial fractions [Section 7.5]
• L’Hôpital’s rule [Section 4.5]
• Improper integrals [Section 7.7]
Applications of Integration
• Area between curves [Section 6.1]
• Integrals in geometry: volume, density, average value [Section 6.2]
• Integrals in geometry: volumes of revolution [Sections 6.3 and 6.4]
• Integrals in geometry: arc length and surface area [Section 8.2]
• Integrals in physics: work and energy [Section 6.5]
• Solving differential equations by separation of variables [Sections 9.1 and 9.4]
Infinite Series
• Sequences and series [Sections 10.1 and 10.2]
• Convergence tests for series [Section 10.3, 10.4, and 10.5]
• Power series [Section 10.6]
• Taylor polynomials [Section 10.7]
• Taylor series, complex numbers, and Euler’s formula [Section 10.8]

1
0 Lecture
Outline:
1. Welcome, syllabus
2. Calculus II in a Nutshell

0.1 Calculus II in a Nutshell

Students are often left with the impression that Calculus II is a hodgepodge of many unrelated topics and
ideas. However, Calculus II, or integral calculus of a single variable, is really only about two topics: integrals
and series, and the need for the latter can be motivated by the former. The purpose of this first lecture is
to explore this a little bit so that you should have a pretty good idea what Calculus II is really all about.

Let’s start in familiar territory: Integrals you have already encountered in Calculus I since Calculus II picks
up right were you left off in Calculus I. During the first weeks of Calculus II you are going to learn several
new techniques to evaluate integrals. For example, consider the integral
Z
e−x cos x dx.

This integral can’t be evaluated easily by using the methods you learned in Calculus I (try it!), but soon after
learning about a method called integration by parts (which is related to the product rule for derivatives) you
will be able to obtain that Z
1
e−x cos x dx = e−x (sin x − cos x) + C.
2
With all these new techniques, will we be able to evaluate (at least in principle) just about any integral, as
long as the integrand is a “nice” function? The answer may surprise you. Let’s look at the integral
Z
sin x2 dx.

Using Calculus I ideas, we could define a function S(x) as a definite integral as follows:
Z x
sin t2 dt.


S(x) =
0

By the Fundamental

Theorem of Calculus (FTC, Part II), the function S(x) is an antiderivative of the
function sin x2 and hence Z
sin x2 dx = S(x) + C.

Expressing an indefinite integral in terms of a definite integral feels like cheating! What we want is a formula
for S(x) that does not involve a definite integral. It turns out that there is no such formula in terms of any
of the functions that you have encountered in your career so far! But, as we will see later, there is a different
kind of formula for S(x) in terms of what is called a power series, the most important topic in Calculus II.

Before talking about power series, let’s return to familiar territory. Some of the simplest functions that
you are familiar with are polynomials. For example, f (x) = x − x3 /6 is a polynomial function. Amazingly,
most of the “higher” functions such as the trigonometric and exponential functions are essentially almost
polynomials, just with infinitely many terms. For example, we have

x3 x5 x7
sin x = x − + − + ··· ,
3! 5! 7!

2
where the three dots (· · · ) indicate that the terms go on forever in the obvious fashion. Here the denominators
(the numbers with the exclamation marks next to them) are simple products called factorials: 3! = 3·2·1 = 6,
5! = 5 · 4 · 3 · 2 · 1 = 120, etc.
The right-hand side of this mysterious formula for sin x is what we call a power series in x. Power series are
often presented in Σ-notation. (Σ is the capital Greek letter sigma, which should remind you of the initial
letter of sum.) In Σ-notation, the formula for sin x is as follows:
∞
x3 x5 x7 X x2n+1
sin x = x − + − + ··· = (−1)n .
3! 5! 7! n=0
(2n + 1)!

You may still wonder how to interpret this formula. Suppose that we only consider the first few terms of
the power series, say only up to the x7 -term. Then, as long as x is relatively close to 0, we get a good
approximation of sin x by a polynomial function of degree 7:

x3 x5 x7
sin x ≈ x − + − .
3! 5! 7!
Just how good is this polynomial approximation? Let’s compare the two graphs:
y
2

x3 x5 x7
f (x) = x − 3!
+ 5!
− 7!
1

−2π −π π 2π x

-1
f (x) = sin(x)

-2

Over the interval (−π/2, π/2), the two graphs are so close to each other that they are indistinguishable to
the naked eye. In fact, for −π/2 < x < π/2, it follows from Taylor’s theorem 1 that

x3 x5 x7 (π/2)9
 
sin x − x − + − < ≈ 0.000160441 .
3! 5! 7! 9!

As we include more and more terms of the power series, we get better and better approximations, even for
x that are not close to 0. Finally, we get the exact formula for sin x for every real number x if we use the
whole power series with infinitely many terms.

Let’s return to the integral of sin(x2 ). By replacing x by x2 in our power series formula for sin x we obtain

x6 x10 x14
sin(x2 ) = x2 − + − + ··· .
3! 5! 7!
1 You will encounter Taylor’s theorem in Calculus II as an application of integration by parts, the technique of integration

that was mentioned at the beginning.

3
Now integrating sin x2 is as easy as integrating a polynomial:

x6 x10 x14
Z Z  
2 2
sin x dx = x − + − + · · · dx
3! 5! 7!

x3 x7 x10 x15
 
= − + − + ··· +C
3 7 · 3! 11 · 5! 15 · 7!

Z x
This means that we found a formula for the function S(x) = sin(t2 ) dt after all:
0

x3 x7 x10 x15
S(x) = − + − + ··· ,
3 7 · 3! 11 · 5! 15 · 7!
or, in fancy pants Σ-notation,
∞
X x4n+3
S(x) = (−1)n .
n=0
(2n + 1)!(4n + 3)
By using the same power series trick, we can evaluate many other important integrals and special functions.
For example, we can obtain a power series formula for the Gauss’ error function
Z x
2 2
Erf(x) = √ e−t dt ,
π 0

which plays a fundamental role in statistics. The power series formula for Erf(x) is

x3 x5 x7
 
2
Erf(x) = √ x− + − + ··· .
π 3 5 · 2! 7 · 3!
√
Given this formula (and a calculator),
√ you may try to find the approximate value of Erf(1/ 2). If you do it
correctly, you will obtain Erf(1/ 2) ≈ 0.683, which has the following interpretation: In a data sample that
is normally distributed, about 68.3% of the data values are within one standard deviation of the mean.

The examples above may have already convinced you that power series will be useful in several applications
of mathematics. We will now look at the ultimate reason why you may want to learn about power series.
It is related to the fundamental reason why science students have to study calculus to begin with. In all
of the exact sciences, many (if not most) of the laws of nature can be expressed as differential equations.
Therefore, solving differential equations (and understanding their solutions!) is of paramount importance.

A differential equation is an equation involving an unknown function and its derivatives. An ordinary
differential equation (ODE) is a differential equation involving an unknown function, say y = F (x), of only
one independent variable. The simplest such equation is
dy
= f (x),
dx
where f (x) is some given function. A solution in this case is simply a function y = F (x) such that F 0 (x) =
f (x). In other words, solving this differential equation is equivalent to finding the indefinite integral of f (x).
In Calculus II, you will learn a few tricks to solve slightly more general equations by integration, but as you
might have guessed, not every differential equation can be solved by integration and we will often have to
look for power series solutions.

4
Of particular importance are second order2 equations that show up in the study of waves of various kinds.
For example, when studying the vibrations of a drum, one is led to the following second order equation:
d2 y dy
2
x2
+x + (x2 − α2 )y = 0,
dx dx
where α is a nonnegative number. This equation is called the Bessel equation and its canonical solutions are
the so-called Bessel functions. The easiest way to describe a Bessel function is as a power series:
∞
X (−1)n  x 2n+α
Jα (x) = .
n=0
n! Γ(n + α + 1) 2

Here the the expression Γ(n+α+1) is a value of the Gamma function, which you may encounter in Calculus II
(depending on the instructor) as part of the discussion of improper integrals. In the special case when α = 0,
the value Γ(n + α + 1) is equal to Γ(n + 1) = n! and it follows3 that
x2 x4 x6
J0 (x) = 1 − + − + ··· .
4 64 2304
Here is what the graphs of J0 (x), J1 (x), and J2 (x) look like:
y
1

f (x) = J0 (x)
J1 (x)

20 x

J2 (x)

-1/2

So, what should you take away from all this? After taking Calculus II, you will have gained a whole new
perspective on what we mean by a “function”. The functions from your youth (such as polynomials, trigono-
metric functions, exponential functions, and logarithmic functions) are but a small collection of functions
that arise in the sciences. If you want to understand the applications of calculus and differential equations
(even if it is just to understand a Wikipedia entry!), then you need to learn about series. Applications range
from Physical Chemistry (energies of reactions, entropy, heat, phase transitions, chemical potential), to phe-
nomena in Engineering and Physics (laws of nature), to Medicine (imagery, half-life), to general Mathematics
(including fractals!), and even arise in discussing some Greek paradoxes.
In other words, you need to take Calculus II.
2A second order equation involves the second derivative of the unknown function.
3 In case you want to verify this, you need to know that Γ(1) = 0! = 1.

5
1 Lecture
Outline:
1. Review of concepts: derivatives
2. Review basic rules: power, product, chain
3. Review of concepts: integrals
4. 5.7 The Substitution Method

1.1 Derivatives
Let y = f (x) be a function of real numbers with real values. Its derivative at x is

dy df d f (x + h) − f (x)
= = f (x) = f 0 (x) = lim = instantaneous rate of change at x
dx dx dx h→0 h or slope of the tangent line at x

provided this limit exists. For instance, if f (x) = 2x then f 0 (x) = 2 for every x because

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

1.1.1 Power rule

You have probably memorized that
d n
x = nxn−1 .
dx
For example, the derivative of x2 is 2x.

1.1.2 Product/quotient rule

More generally,
(f g)0 = f 0 g + f g 0 .
You can actually use this over and over again to get the power rule.
d 2 d d 3 d 2
x = x · x = 1x + x1 = 2x, x = x · x = 2x · x + x2 · 1 = 3x2 , etc.
dx dx dx dx
Another example: if f (x) = xex , then f 0 (x) = ex + xex = (1 + x)ex .
The quotient rule is similar:  0
f f 0 g − f g0
= .
g g2
For example,
d ln x (1/x)x1/2 − (1/2)x−1/2 ln x 2 − ln x
1/2
= = .
dx x x 2x3/2
(Note that the quotient rule is really just the product rule combined with the chain rule, since (d/dx)(1/x) =
−1/x2 .)

6
1.1.3 Chain rule
Recall that the composition of two functions f and g is given by

(f ◦ g)(x) = f (g(x)).

The chain rule is

dy dy dz
(f ◦ g)0 (x) = f 0 (g(x))g 0 (x), or
= ,
dx dz dx
whichever way helps you remember what’s going on. (Often taught as “inner” and “outer” functions.) For
example,
d x2 2
e = ex 2x, NOT ex 2x.
dx

1.2 Integrals
Let f be a continuous function on [a, b]. The definite integral of f over [a, b] is defined as the limit of
Riemann sums as the width of blocks goes to zero. By the Fundamental Theorem of Calculus,
Z b
f (x)dx = F (b) − F (a)
a

where F is any antiderivative of f . Remember that we can interpret the definite integral as the area
under the curve, f (x), on [a, b] or the net change of F (x) on [a, b]. We can think of the indefinite
integral of f as the set of all antiderivatives, meaning there is an unknown constant C. For example,
Z
1
x2 dx = x3 + C.
3
To compute a definite integral, it doesn’t matter which value of C we choose. For instance,
Z 1  1
2 1 3 1 1
x dx = x +C = +C −0−C = ,
0 3 0 3 3

from which we see that C is irrelevant for definite integrals.

1.3 Substitution
Thanks to the Fundamental Theorem of Calculus, integration is basically the opposite of differentiation, so
if you’re good with derivatives, you should be good at integrals. Still, it can be tricky to do, for example,
the chain rule backwards.
The chain rule backwards is called substitution.
Change of Variables Formula:
Z Z
0
f (u(x)) u (x) dx = f (u) du
| {z } | {z }
f (u) du

2 2
We saw that the derivative of ex is 2xex . Now suppose we want to compute an indefinite integral:
Z
2
xex dx.

2
The answer should have something to do with ex . How can we be systematic?

7
 2 2
Let’s set u = x2 . So ex = eu . What about du? Well, du/dx = 2x so du = 2x dx. So xex dx = 21 eu du. We
know how to integrate eu du. We get
Z Z
2 1 1 1 2
xex dx = eu du = eu + C = ex + C.
2 2 2

Do not forget that when using substitution for a definite integral we must consider the limits of integration!
Change of Variables Formula:
Z b Z u(b)
0
f (u(x)) u (x) dx = f (u) du
a | {z } | {z } u(a)
f (u) du

Now suppose we want to compute a definite integral:

Z 2
2
xex dx.
1

Then all our previous work holds, but we also need to consider the change to the limits of integration.
Since u = x2 when x = 1, we see u = 1, or more precisely, u(1) = 1 is the lower limit of integration after
substitution. Similarly, u(2) = 22 = 4 is the upper limit. We get
Z 2
1 4 u
Z
x2 1 4 1
e du = eu = e4 − e .


xe dx =
1 2 1 2 1 2

Example. Evaluate Z
x2 sin x3 dx.

We let u = x3 , so that du = 3x2 dx. So x2 sin x3 dx = 31 sin(u)du. We get

Z Z
1 1 1
x2 sin x3 dx = sin(u)du = − cos(u) + C = − cos x3 + C.



3 3 3

Example. Evaluate Z
tan θ dθ.

While this does not immediately look like it needs substitution, it indeed does! Let u = cos θ so that
du = − sin θ dθ. We get
Z Z Z
sin θ du
tan θ dθ = dθ = − = − ln |u| + C = − ln | cos θ| + C = ln | sec θ| + C.
cos θ u

The trick with all substitutions is to find the u which will make f (x)dx turn into g(u)du where g(u) is
nice and easy. Easier said than done! Indeed, substitution only works when you can find such a nice u.
Otherwise, you need to try other tricks.
Suggestions for Substitution
p
i) If a function is raised to a power or in a radical, i.e., [f (x)]2 , [f (x)]−1 , f (x), let u = f (x).
ii) If a function appears in the argument of another function, i.e., sin(f (x)), ln(f (x)), ef (x) , let u = f (x).
Note that i) is a particular case of ii) with the outer function being the power function, and these suggestions
do not always work!

8
2 Lecture
Outline:
1. 7.1 Integration by Parts

2.1 Integration by parts

By now you should be able to do the following techniques forwards (derivatives) and backwards (integrals):
Z
d n 1
power rule: x = nxn−1 → xn dx = xn+1 + C (n 6= −1)
dx Z n +
Z 1
df df du du
chain rule: = → g(u(x)) dx = g(u)du (substitution)
dx du dx dx
Now we extend this to the product rule:
Z Z Z
d du dv du dv dv du
(uv) = v+u → uv = vdx + u dx or u dx = uv − vdx.
dx dx dx dx dx dx dx
This is called integration by parts (or partial integration).
Another way to write the product rule for easier recall is to simply suppress dx to write
Z Z
d(uv) = u dv + v du → u dv = uv − v du.

Strategy: pick out a u and dv in your integral, such that it is advantageous to pass the derivative from v to
u in order to make integration easier.
R R
Example. Compute x cos xdx. Let u = x and dv = cos xdx. To find v, we integrate v = cos xdx = sin x.
(At this point we don’t need to worry about constants; we’ll make sure it’s there at the end.) So the product
rule tells us Z Z
x cos xdx = x sin x − sin xdx = x sin x + cos x + C.

It was advantageous to pass the derivative over to x because its derivative is just 1.
R
Example. Find ln xdx. This is a surprising example, because it doesn’t seem like there are any “parts”!
And yet, we don’t know how to antidifferentiate
R ln x off the top of our heads. So we take u = ln x, and as
for v we simply let dv = dx integrating v = 1dx to get v = x. Since du/dx = 1/x we get
Z Z Z
1
ln xdx = x ln x − x dx = x ln x − dx = x ln x − x + C.
x

Example. Find exRcos xdx. Here’s an example were something funny happens. Take u = ex and dv =
R

cos x dx, so that v = cos xdx = sin x. So

Z Z
e cos xdx = e sin x − ex sin xdx.
x x

ex sin xdx. Take u = ex and v = sin xdx = − cos x. So

R R
But now we need
Z Z
e sin xdx = −e cos x + ex cos xdx.
x x

It seems we’ve gone around in a circle! But in fact, putting these together we get
Z Z
ex cos xdx = ex sin x + ex cos x − ex cos xdx.

9
That means Z
1 x
ex cos xdx = (e sin x + ex cos x) + C.
2

Example. Compute xn exRdx for any positive integer n. Start with n = 1, i.e. xex dx. We take u = x
R R

and dv = ex dx, integrating ex dx to get v = ex . So

Z Z
xe dx = xe − ex dx = xex − ex + C = (x − 1)ex + C.
x x

x2 ex dx. We let u = x2 and v = ex . Then we get

R
Now let’s try
Z Z
x e dx = x e − 2xex dx.
2 x 2 x

But we just found xex dx. We can use that:

R
Z Z
x e dx = x e − 2 xex dx = x2 ex − 2(x − 1)ex + C = (x2 − 2x + 2)ex + C.
2 x 2 x

For any n, we use the following relation,

Z Z
xn ex dx = xn ex − n xn−1 ex dx,

which is called a recursive formula, or sometimes a reduction formula as it reduces the power under the
integral. It can be used repeatedly to find the answer for any n, because we can always get down to n = 1,
Z Z Z
n x n x n−1 x n n−1 x
x e dx = x e − n x e dx = (x − nx )e + n(n − 1) xn−2 ex dx

= · · · = (xn − nxn−1 + n(n − 1)xn−2 − · · · + (−1)n−1 n!x + (−1)n n!)ex .

Rπ
Example. Show Rthat 0 cos x cos 2x dx = 0. Limits of integration hold when doing integration by parts, so
let u = cos x, v = cos 2x dx = 21 sin 2x. Then
Z π π
1 π 1 π
 Z Z
1
cos x cos 2x dx = cos x sin 2x + sin x sin 2x dx = sin x sin 2x dx.
0 2 0 2 0 2 0
Let’s try integrating by parts again: u = sin x, v = sin 2x dx = − 21 cos 2x, so
R

Z π π
1 π 1 π
 Z Z
1
sin x sin 2x dx = − sin x cos 2x + cos x cos 2x dx = cos x cos 2x dx.
0 2 0 2 0 2 0
Plugging this into the first equation we get
Z π
1 π
Z
cos x cos 2x dx = cos x cos 2x dx.
0 4 0
But that means Z π Z π
3
cos x cos 2x dx = 0 ⇒ cos x cos 2x dx = 0.
4 0 0

Suggestions for Substitution If one of the functions in the integrand can be eliminated or made simpler
by taking a few derivatives (or integrals), choose this to be u (or dv). Example:
u : xn , ln x, (ln x)n , inverse trigonometric functions
If neither function will be made simpler by derivatives or integrals (multiplication of trigonometric and
exponential), you will have to use integration by parts until the original integral reappears and then solve
for it.

10
3 Lecture
Outline:
1. 7.2 Trigonometric Integrals, Part I

3.1 Trigonometric integrals

Many integrals involve trigonometric functions. To compute them, you will need to remember two or three
fundamental identities:
cos2 x + sin2 x = 1, sin 2x = 2 sin x cos x, cos 2x = cos2 x − sin2 x.
Note
cos 2x = cos2 x − (1 − cos2 x) = 2 cos2 x − 1, cos 2x = 1 − sin2 x − sin2 x = 1 − 2 sin2 x.
Together with the fact that (d/dx) sin x = cos x and (d/dx) cos x = − sin x, we can compute many integrals
of the form Z
sinm x cosn x dx.

Example. Compute cos100 x sin x dx. Use u = cos x, so du = − sin x dx. So

R

u101 cos101 x
Z Z
cos100 x sin x dx = − u100 du = − +C =− + C.
101 101

3.2 Odd powers of cosine or sine

If there’s an odd power of either cosine or sine, then we can always reduce down to a single cosine or sine
using cos2 x + sin2 x = 1, then use substitution.
Example. Compute sin3 x dx. Use sin2 x = 1 − cos2 x and then u = cos x to get
R

u3 cos3 x
Z Z Z
sin3 x dx = (1 − cos2 x) sin x dx = − (1 − u2 )du = −u+C = − cos x + C.
3 3

Example. Compute sin2 x cos3 x dx. Let u = sin x to get

R

u3 u5 sin3 x sin5 x
Z Z Z
2 2 2
sin x cos x dx = sin x(1 − sin x) cos x dx = u2 (1 − u2 )du =
3
− +C = − + C.
3 5 3 5

3.3 Even powers of cosine and sine

In general, even powers are more difficult to compute. How do we compute sin2 x dx? We could use
R

integration by parts, actually, but let’s use a different trick instead. We can use the double angle formula to
get sin2 x = 12 (1 − cos 2x). So
Z Z
2 1 1 1 x 1 x 1
sin x dx = (1 − cos 2x)dx = (x − sin 2x) + C = − sin 2x + C = − sin x cos x + C.
2 2 2 2 4 2 2
2 x 1
R
You can do the same to get cos x dx = 2 + 2 sin x cos x + C.
With higher powers it gets tougher. We use integration by parts to get a recursive formula. (Your book
calls it a reduction formula.) We want to find the integral of, say, cosn x. Let u = cosn−1 x, dv = cos x dx.
Then
Z Z
cosn x dx = cosn−1 x sin x + (n − 1) cosn−2 x sin2 x dx
Z Z
n−1 n−2
= cos x sin x + (n − 1) cos x dx − (n − 1) cosn x dx.

11
So Z Z
n n−1
n cos x dx = cos x sin x + (n − 1) cosn−2 x dx

n−1
Z Z
1
⇒ cosn x dx = cosn−1 x sin x + cosn−2 x dx.
n n
For example, Z Z
1 5
cos x dx = cos5 x sin x +
6
cos4 x dx.
6 6

We can use this to compute daunting integrals.

Example. Compute sin2 x cos4 xdx. First break it down into easier pieces
R

Z Z Z Z Z
1 5
sin2 x cos4 xdx = cos4 xdx − cos6 xdx = cos4 xdx − cos5 x sin x − cos4 x dx
6 6
Z
1 1
= cos4 xdx − cos5 x sin x
6 6
using the reduction formula. We can use the reduction formula again if we like:
Z Z
1 3 1 3 3x
cos4 xdx = cos3 x sin x + cos2 x dx = cos3 x sin x + cos x sin x + + C.
4 4 4 8 8
Thus Z
1 1 1 x
sin2 x cos4 xdx = − cos5 x sin x + cos3 x sin x + cos x sin x + + C.
6 24 16 16
Another way would be to use the double angle formula directly instead of relying on the reduction formula.
There are usually multiple options as you learn more techniques as seen here and in the next example.
Rπ
Interesting Example. Compute 0 sin2 x dx. From previous work we know
Z π π
x 1 π
sin2 x dx = − sin x cos x = .
0 2 2 0 2

But let us look at it a little differently which could help you in the future! Consider the following graphs
y
cos2 (x)
1

sin2 (x)

πx

Notice the area under each curve is equal, that is,

Z π Z π Z π Z π Z π  Z π 
1 1 π
sin2 x dx = cos2 x dx ⇒ sin2 x dx = sin2 x dx + cos2 x dx = 1 dx =
0 0 0 2 0 0 2 0 2

12
4 Lecture
Outline:
1. 7.2 Trigonometric Integrals, Part II

4.1 Trigonometric integrals: tangents and secants

Starting with cos2 x + sin2 x = 1, divide by cos2 x to get

1 + tan2 x = sec2 x.

We also have
d d sin x cos2 x + sin2 x d d 1 − sin x
tan x = = 2
= sec2 x = 1 + tan2 x, sec x = = = tan x sec x.
dx dx cos x cos x dx dx cos x − cos2 x
Remembering these two facts, as well as the basic facts we already know, we can compute lots of integrals
involving tangents and secants.
Example. Find tan2 x dx. This is very easy:
R

Z Z
tan2 x dx = (sec2 x − 1)dx = tan x − x + C.

R
Example. Find tan x dx. We use a substitution u = cos x, so du = − sin x. Thus
Z Z Z
sin x 1
tan x dx = dx = − du = − ln |u| + C = − ln | cos x| + C = ln | sec x| + C.
cos x u

Other examples involve reducing down in clever ways. Normally you substitute u = tan x or u = sec x, so
either du = sec2 x dx or du = tan x sec x dx. If what’s left over is reducible down to powers of u, then the
strategy works. We look for even powers of tan x and sec x, since tan2 x + 1 = sec2 x.
Example. Compute tan3 x sec5 x dx.
R

Let u = sec x, so du = tan x sec x dx. What’s left over is tan2 x sec4 x = (sec2 x − 1) sec4 x = (u2 − 1)u4 =
u6 − u4 . So
u7 u5 sec7 x sec5 x
Z Z
tan x sec x dx = (u6 − u4 )du =
3 5
− +C = − + C.
7 5 7 5

Example. Compute tan2 x sec4 xdx.

R

The previous substitution would leave odd powers, so let u = tan x so du = sec2 x dx. What’s left over is
tan2 x sec2 x = tan2 x(tan2 x + 1) = u2 (u2 + 1) = u4 + u2 . So

tan5 x tan3 x
Z Z
tan2 x sec4 xdx = (u4 + u2 )du = + + C.
5 3

Example. Find tan3 x dx. Here we see neither sec2 xdx nor sec x tan xdx, but if we recall that tan2 x =
R

sec2 x − 1, we get Z Z Z
tan3 x dx = tan x sec2 x dx − tan x dx.

The first one is done by setting u = tan x, and the second one was done at the beginning. We get
tan2 x
Z
tan3 x dx = + ln | cos x| + C.
2

13
 R
Example. Find sec x dx. One way to do this systematically requires a technique called partial fractions,
which we haven’t done yet. Still, I will only use it to get through one part; everything else is just trigonometric
identities. Let u = sin x.
Z Z Z Z Z  
cos x cos x 1 1 1 1
sec x dx = dx = dx = du = + du (partial frac.)
cos2 x 1 − sin2 x 1 − u2 2 1−u 1+u
(1 + sin x)2
     
1 1 1+u 1 1 + sin x 1
= (− ln(1 − u) + ln(1 + u)) + C = ln + C = ln + C = ln +C
2 2 1−u 2 1 − sin x 2 1 − sin2 x
(1 + sin x)2
 
1 1 + sin x
= ln + C = ln + C = ln |sec x + tan x| + C.
2 cos2 x cos x

A more clever solution from your book requires noticing that

d
(tan x + sec x) = sec2 x + tan x sec x = (tan x + sec x) sec x.
dx
So if we let u = tan x + sec x, we get du = u sec x dx, which means we get directly that
Z Z Z
u sec xdx du
sec x dx = = = ln |u| + C = ln | tan x + sec x| + C.
u u

Table of Trigonometric Integrals (will be given)

n−1
Z Z
n 1 n−1
sin x dx = − sin x cos x + sinn−2 x dx
n n

n−1
Z Z
1
cosn x dx = cosn−1 x sin x + cosn−2 x dx
n n
Z
sec x dx = ln | sec x + tan x| + C

n−2
Z Z
1
secn x dx = secn−2 x tan x + secn−2 x dx
n−1 n−1
Z
csc x dx = − ln | csc x + cot x| + C

n−2
Z Z
1
cscn x dx = − cscn−2 x cot x + cscn−2 x dx
n−1 n−1
Z
tan x dx = ln | sec x| + C = − ln | cos x| + C

Z Z
1
tann x dx = tann−1 x − tann−2 x dx
n−1
Z
cot x dx = − ln | csc x| + C = ln | sin x| + C

Z Z
1
cotn x dx = − cotn−1 x − cotn−2 x dx
n−1

14
5 Lecture
Outline:
1. 7.3 Trigonometric Substitution

5.1 Trigonometric substitution

Finding the area under part of a circle. Suppose I want the area under the top half √ of the unit circle,
but only on the interval 0 ≤ x ≤ 1/2. The equation for the top half of the circle is y = 1 − x2 . So I want
to evaluate Z 1/2 p
Area = 1 − x2 dx.
0
2 2
√
How? What if we could write 1 − x = w so that 1 − x2 = w?
With this idea in mind, we try a trigonometric substitution recalling that sin2 θ + cos2 θ = 1. Let
x = sin θ for −π/2 ≤ θ ≤ π/2, so that dx = cos θ dθ, noting that in this range, cos θ ≥ 0.
√
What about 1 − x2 ? This is made easy due to our substitution,
p p √
1 − x2 = 1 − sin2 θ = cos2 θ = cos θ.

So
sin−1 x x p
Z p Z
θ 1
1 − x2 dx = cos2 θdθ = + cos θ sin θ + C = + 1 − x2 + C.
2 2 2 2
To solve our problem, we get
1/2
sin−1 x x p 1√

π
Area = + 1 − x2 = + 3.
2 2 0 12 8

It is possible to verify this geometrically, by splitting the region into a triangle and a wedge.
√
If a circle has radius a instead of 1, then the top half is given by y = a2 − x2 . If we want to integrate this,
we use x = a sin θ instead of x = sin θ:
Z p Z p Z √ Z
a2 − x2 dx = a2 − a2 sin2 θa cos θdθ = a2 cos2 θa cos θdθ = a2 cos2 θdθ

a2 a2  x p  x 
= (sin θ cos θ + θ) + C = 1 − (x/a)2 + sin−1 + C.
2 2 a a

x2
Z
Example. Some tough integrals can be computed using the same substitution. Compute dx.
√ (4 − x2 )3/2
We set x = 2 sin θ, so 4 − x2 = 2 cos θ. Thus

x2 4 sin2 θ 8 sin2 θ sin2 θ

Z Z Z Z Z
dx = 2 cos θdθ = cos θdθ = dθ = tan2 θ dθ
(4 − x2 )3/2 (2 cos θ)3 8 cos3 θ cos2 θ
Z
sin θ x
= (sec2 θ − 1)dθ = tan θ − θ + C. = −θ+C = √ − sin−1 (x/2) + C.
cos θ 4 − x2

Tangents, secants Recall that

tan2 θ + 1 = sec2 θ ⇒ sec2 θ − 1 = tan2 θ.

15
 √ √ √
So if we get something like √ x2 + a2 we will try x = a tan θ where x2 + a2 = a sec θ. If we get x2 − a2
we will try x = a sec θ where x2 − a2 = a tan θ.
√
Z
1
Example. Evaluate √ dx. Take x = 3 tan θ, so dx = 3 sec2 θdθ and x2 + 9 = 3 sec θ. Thus
2
x +9
Z Z
1 1p 2 1 p
√ dx = sec θdθ = ln | sec θ+tan θ|+C = ln x + 9 + x +C = ln(1/3)+ln | x2 + 9+x|+C.
x2 + 9 3 3

Of course, ln(1/3) is just a constant, so we could just say

Z
1 p
√ dx = ln | x2 + 9 + x| + C.
2
x +9

√ √
Z
dx
Example. Evaluate √ . Now 4x2 − 36 = 2 x2 − 9. Let x = 3 sec θ so dx = 3 tan θ sec θdθ
2 2
x 4x − 36
√
and x2 − 9 = 3 tan θ. Then
√
1 x2 − 9
Z Z Z
dx 3 tan θ sec θ 1 1 1 tan θ
√ = dθ = cos θdθ = sin θ + C = +C = + C.
x2 4x2 − 36 9 sec2 θ6 tan θ 18 18 18 sec θ 18 x

Completing the square. Any quadratic polynomial can be converted into a square plus some number.
Then we can use trigonometric substitution (or partial fractions, as we will see).
dx 2
R
For example, suppose we want to compute (x2 −6x+11) 2 . Half of 6 is 3, and 3 = 9, so

x2 − 6x + 11 = (x − 3)2 + 2.

Let u = x − 3, so du = dx. Then the integral is just

Z
du
(u2 + 2)2

which can be solved with trig sub.

√ √ √ √
If there is time: Now let’s use trig subs. Set u = 2 tan θ so u2 + 2 = 2 sec θ and du = 2 sec2 θ.
Thus using our integral table,
Z √
2 sec2 θ
Z Z Z
du 1 2 1 1 1 θ
= dθ = √ cos θdθ = √ sin θ cos θ + √ dθ = √ sin θ cos θ + √ + C.
(u2 + 2)2 4 sec4 θ 2 2 4 2 4 2 4 2 4 2
Now √ √
tan θ 2u 2(x − 3)
sin θ cos θ = = = ,
sec2 θ u2 + 2 (x − 3)2 + 2
and    
−1 u −1 x−3
θ = tan √ = tan √ .
2 2
So we get  
x−3 x−3
Z
dx 1 1
2 2
= 2
+ √ tan−1 √ + C.
(x − 6x + 11) 4 (x − 3) + 2 4 2 2

16
6 Lecture
Outline:
1. 7.5 The Method of Partial Fractions; Part I

6.1 Partial fractions

Consider the integral Z
du
.
u2 − 1
We could use trig subs to do this, but that’s actually more trouble than it’s worth. Instead we notice that
 
1 1 1 1
= − .
u2 − 1 2 u−1 u+1

So then Z    
u−1
Z
du 1 1 1 1
2
= − du = ln + C.
u −1 2 u−1 u+1 2 u+1
We have used the method of partial fractions. There is a systematic technique we can use for any rational
function. (Recall that a rational function is just a fraction where top and bottom are polynomials.)
Any polynomial can be factored into linear (or first-order) and quadratic (or second-order) terms. For
example,
x3 − 2x2 + x = x(x − 1)2 .
If the polynomial appears in the denominator of a fraction, then the idea is to break up the fraction into a
sum where each term has one of the factors in the denominator. Think about this as the reverse (or inverse
operation) of finding common denominators.

6.2 Distinct linear factors

Let’s try x4x+1 2
R
2 −4 dx. The bottom needs to be factored. Notice that x − 4 = (x + 2)(x − 2). We try to find

the whole fraction as a sum

4x + 1 A B
2
= + .
x −4 x+2 x−2
By multiplying both sides by (x + 2)(x − 2) we get

4x + 1 A B
= + ⇒ 4x + 1 = A(x − 2) + B(x + 2).
(x + 2)(x − 2) x+2 x−2

This has to hold for every x. We can pick certain x cleverly to get A and B by themselves. Pick x = −2.
Then we get
7
4(−2) + 1 = A(−2 − 2) + B(−2 + 2) ⇒ −7 = −4A ⇒ A = .
4
Similarly take x = 2 to get 4B = 9 so B = 9/4. So we get
Z Z  
4x + 1 7 1 9 1 7 9
dx = + dx = ln |x − 2| + ln |x + 2| + C.
x2 − 4 4x−2 4x+2 4 4

17
6.3 Repeated linear factors
In the last example all three factors were distinct. When
R there is a repeated factor, it actually contributes
3x−9
more than one partial fraction. For example, let’s try (x−1)(x+2) 2 dx. We want to solve

3x − 9 A B C
= + +
(x − 1)(x + 2)2 x − 1 x + 2 (x + 2)2
for A, B, C. We can try proceeding as before:
3x − 9 = A(x + 2)2 + B(x − 1)(x + 2) + C(x − 1).
Plug in x = 1 to get A = −2/3. Plug in x = −2 to get C = 5. But what about B? There are several ways
to proceed, but let us start with a method that we will need if the problem is harder, comparing powers.
We start with the highest power of x in the equation we are consider,
3x − 9 = A(x + 2)2 + B(x − 1)(x + 2) + C(x − 1).
Note that the highest power of x is 2. So we compare the coefficient of the x2 term on the right and left-hand
sides to see  
2 2
0x2 = Ax2 + Bx2 ⇒ 0 = A + B ⇒ B = −A = − − = .
3 3
This method is great for comparing lead powers, but keep in mind for other powers we have to distribute
every term in order to collect coefficients.
Another way is just to try any other x not already tried, for example x = 0. Then we get
−9 = 4A − 2B − C ⇒ 2B = 9 − 8/3 − 5 = 4 − 8/3 = 4/3 ⇒ B = 2/3.
With either method we get
3x − 9
Z Z Z Z
2 dx 2 dx dx 2 2 5
dx = − + + 5 = − ln |x − 1| + ln |x + 2| − + C.
(x − 1)(x + 2)2 3 x−1 3 x+2 (x + 2)2 3 3 x+2

Remark. For repeated linear factors of degree higher than 2, write a fraction for each power, that is, you
should have the same number of fractions for the repeated factor as its power.
Example. In this example we will just focus on the partial fraction decomposition rather than the integral
as well so we can get a feel for how to solve something with more factors. Consider
1
.
x(x − 2)3
Since there is a power of 3 on x − 2, we will need to write 3 fractions for it. So we want to solve
1 A B C D
3
= + + 2
+
x(x − 2) x x − 2 (x − 2) (x − 2)3
for A, B, C, D. We proceed as before:
1 = A(x − 2)3 + Bx(x − 2)2 + Cx(x − 2) + Dx.
Plug in x = 0 to get A = −1/8. Plug in x = 2 to get D = 1/2. To find B we compare highest powers as
before to see B = −A = 1/8. Finally, choose x = 1 and plug in what we found for A, B, D to see
1 1 1 1 1 1 1
1 = − (1 − 2)3 + (1 − 2)2 + C(1 − 2) + ⇒ 1 = + − C + ⇒ C = − .
8 8 2 8 8 2 4
Hence,
1 1 1 1 1
3
=− + − 2
+
x(x − 2) 8x 8(x − 2) 4(x − 2) (x − 2)3

18
7 Lecture
Outline:
1. 7.5 The Method of Partial Fractions; Part II

7.1 Irreducible quadratic factors

Any factor of the form x2 + a2 is irreducible in the sense that it cannot be factored into linear terms.
We can deal with all irreducible quadratic factors by completing the square and either using substitution or
inverse tangents.
18
R
Example. Let’s try (x+3)(x 2 +9) dx. To find partial fractions, we solve the following:

18 A Bx + C
2
= + 2 ⇒ 18 = A(x2 + 9) + (Bx + C)(x + 3).
(x + 3)(x + 9) x+3 x +9
The Bx + C is generally necessary because it sits on top of a quadratic. Plug in x = −3 to get A = 1. Plug
in x = 0 to get 18 = 9 + 3C so C = 3. To find B there are a few strategies, but one way could just be to
multiply out and match coefficients:
18 = x2 + 9 + Bx2 + 3Bx + 3x + 9 = (B + 1)x2 + (3B + 3)x + 18.
It follows that B + 1 = 0 or B = −1. So
Z Z Z Z
18 dx x dx
2
dx = − 2
dx + 3 2
.
(x + 3)(x + 9) x+3 x +9 x +9
The second integral is solved by the substitution u = x2 + 9 so du = 2x dx:
Z
x 1 du 1
dx = = ln(x2 + 9) + C.
x2 + 9 2 u 2
The third integral is solved by the trig sub x = 3 tan θ. In fact,
3 sec2 θdθ
Z Z
dx 1 1 −1 x
 
= = θ + C = tan + C.
x2 + 9 9 sec2 θ 3 3 3
Putting these together we get
Z
18 1 x
2
dx = ln |x + 3| − ln(x2 + 9) + tan−1 + C.
(x + 3)(x + 9) 2 3

Remark. If there is a repeated irreducible factor, it again contributes more than one partial fraction.

7.2 Long division

R 3 +1
Let’s try xx2 −4 dx. Before getting partial fractions, we need to address the issue that the top has higher
degree (exponent) than the bottom. So first we’ll use long division to reduce the fraction down to a proper
fraction, where the top has lower degree than the bottom.
x3 + 1 x3 − 4x + 4x + 1 4x + 1
2
= =x+ 2 .
x −4 x2 − 4 x −4
Notice the first term is only x while the second term was solved in the previous lecture’s examples. So we
get Z 3
x2
Z  
x +1 7 1 9 1 7 9
dx = x + + dx = + ln |x − 2| + ln |x + 2| + C.
x2 − 4 4x−2 4x+2 2 4 4

19
8 Lecture
Outline:
1. 4.5 L’Hôpital’s Rule; Part I

8.1 L’Hôpital’s Rule

One interesting way to compare two functions f (x) and g(x) around a point x = a is to compute the limit

f (x)
lim .
x→a g(x)

This includes a = ±∞, that is the limit when x gets really large (or very negative).
Before getting to L’Hôpital’s Rule, let’s remember the basic techniques! It is important not to lose
these techniques, because L’Hôpital’s Rule doesn’t always apply! If limx→a f (x) exists, and if limx→a g(x)
exists and is not zero, then just plug in x = a:

x3 − 8 27 − 8
lim = = 19.
x→3 x − 2 3−2
If both limits exist but the bottom is zero, then the limit might not exist, or it may be infinity. The case
“0/0” is ambiguous, but sometimes cancellation is enough to find out what the limit is.

x2 − 4 (x − 2)(x + 2)
lim = lim = lim (x + 2) = 4.
x→2 x − 2 x→2 x−2 x→2

When you can’t easily find such cancellation, there’s L’Hôpital’s Rule.
Theorem 1. Suppose a is a number, that f and g are differentiable around a, and either f (a) = g(a) = 0 or
limx→a f (x) = ±∞ and limx→a g(x) = ±∞. Then if g 0 (x) is nonzero near a (but not necessarily at x = a)
we get
f (x) f 0 (x)
lim = lim 0 , (1)
x→a g(x) x→a g (x)

provided this limit exists or is ±∞.

Proof. We give a proof for the first case when f (a) = g(a) = 0 with g 0 (a) 6= 0. Then g(x) 6= g(a) near a and
f (x)−f (a)
f (x) f (x) − f (a) x−a f 0 (a) f 0 (x)
lim = lim = lim = = lim .
x→a g(x) x→a g(x) − g(a) x→a g(x)−g(a) g 0 (a) x→a g 0 (x)
x−a

You can use (1) for easy limits:

x2 − 4 2x
lim = lim = 4.
x→2 x − 2 x→2 1

But it’s more interesting for harder ones:

x2 − 4 2x 4
lim = lim = .
x→2 sin πx x→2 π cos πx π
In this case, we get a 0/0 limit, but cancellation looks pretty hopeless, so it’s a good thing we have L’Hôpital’s
Rule.

20
Remark. Don’t forget to check for 0/0 or ±∞/ ± ∞! Let us review the first example,
x3 − 8 27 − 8
lim = = 19.
x→3 x − 2 3−2
What if we had mistakenly used L’Hôpital’s Rule? We would have said
x3 − 8 3x2
lim “ = ” lim = 27 6= 19.
x→3 x − 2 x→3 1

WRONG! If the hypothesis of the theorem does not hold then we cannot apply the theorem!

8.2 Comparing two functions asymptotically

Quick Sort and Bubble Sort are two algorithms for sorting a list. If the list has size n, Quick Sort requires
a maximum of n ln n steps, whereas Bubble Sort requires a maximum of n2 steps. Which is more efficient?
In other words, we want to compare two functions asymptotically. Both f (x) = x ln x and g(x) = x2 go to
infinity as x goes to infinity. But one might go faster than the other. To make this precise, let’s compute
f (x) x ln x
lim = lim .
x→∞ g(x) x→∞ x2

We will need to use L’Hôpital’s Rule at infinity:

Theorem 2. Suppose f (x) and g(x) are both differentiable for x large and that g 0 (x) is nonzero for x large.
If limx→∞ f (x) and limx→∞ g(x) are either both zero or both infinite, then
f (x) f 0 (x)
lim = lim 0
x→∞ g(x) x→∞ g (x)

provided the limit on the right exists or is infinite.

Let’s apply it. Note that we can also perform some cancellation before applying the rule:
x ln x ln x 1/x
lim 2
= lim = lim = 0.
x→∞ x x→∞ x x→∞ 1
Now let’s interpret. As x gets very large, the ratio between x ln x and x2 gets very small. In other words,
x ln x is much smaller than x2 whenever x is big. We sometimes write this with the notation x ln x  x2 .
√
Example. Which is faster, an algorithm that takes (ln n)2 steps or one that takes n steps?
To answer this, we find
(ln x)2 2(ln x)(1/x) 4 ln x 4/x 8
lim√ = lim √ = lim √ = lim √ = lim √ = 0.
x→∞ x x→∞ 1/(2 x) x→∞ x x→∞ 1/(2 x) x→∞ x
Once again, the logarithmic function wins.
General fact. Exponentials grow faster than any polynomial. More formally,
xn  ex for any number n.
To see this, we can use L’Hôpital’s Rule.
xn nxn−1 n(n − 1)xn−2 n(n − 1) · · · (n − k + 1)xn−k
lim x
= lim x
= lim x
= · · · = lim
x→∞ e x→∞ e x→∞ e x→∞ ex
where k is the first whole number big enough so that n − k ≤ 0. But then the last limit is zero! So no matter
how big n is, this shows that ex is much bigger than xn , for x large.
It shouldn’t be surprising to know that ln x is much smaller than xn for x large, whenever n is a positive
number. You can also show this using L’Hôpital’s Rule.

21
9 Lecture
Outline:
1. 4.5 L’Hôpital’s Rule; Part II

9.1 More examples of L’Hôpital’s Rule

Motivated by the idea of comparing two functions, we can use L’Hôpital’s Rule to calculate lots of limits, not
just as x gets large. Remember, we always need to check that the fraction is of the form 0/0 or ±∞/ ± ∞
before we can use it.
Example. Find
cos2 x
lim .
x→π/2 1 − sin x

Both cos2 x and 1 − sin x go to zero as x → π/2, but how do they compare? We use L’Hôpital’s Rule to find

cos2 x −2 cos x sin x

lim = lim = lim 2 sin x = 2.
x→π/2 1 − sin x x→π/2 − cos x x→π/2

So around x = π/2, cos2 x is about twice as big as 1 − sin x.

The form 0 · ∞. When you get this form, you can rewrite as a fraction to get 0/0 or ±∞/ ± ∞ and apply
L’Hôpital’s Rule. For example,
ln x 1/x
lim x ln x = lim = lim = lim −x = 0.
x→0+ x→0+ 1/x x→0+ −1/x2 x→0+

The form ∞ − ∞. No, it’s not always zero! For example,

 
1 1 1−x
lim − = lim = ∞.
x→0 x2 x x→0 x2

On the other hand,

 
1 1 x − sin x 1 − cos x sin x
lim − = lim = lim = lim = 0.
x→0 sin x x x→0 x sin x x→0 sin x + x cos x x→0 2 cos x − x sin x

Using logarithms for the forms 00 and 1∞ . Let’s find the following limits:

lim xx , lim (1 + 4x)1/2x .

x→0+ x→0

For the first, note that ln(xx ) = x ln x. We know limx→0+ x ln x = 0 so

lim xx = lim ex ln x = e0 = 1.
x→0+ x→0+

1
For the second, we know ln(1 + 4x)1/2x = 2x ln(1 + 4x). By L’Hôpital’s Rule,

ln(1 + 4x) 4/(1 + 4x)

lim = lim = 2.
x→0 2x x→0 2
So limx→0 (1 + 4x)1/2x = e2 .
The form 0∞ in not indeterminate. If limx→a f (x)g(x) is of the form 0∞ , then limx→a eg(x) ln(f (x)) has
an exponent of the form −∞. Hence the limit is always 0 in this case!

22
10 Lecture
Outline:
1. 7.7 Improper Integrals; Part I

10.1 Improper integrals

We often think of integrals as areas of regions between curves. But some areas can be unbounded: they
stretch out to infinity. Still, the area could be finite or infinite. We call this an improper integral.
If the integral is finite, we say it converges. Otherwise, we say it diverges. To find out whether it converges,
we need to take the limit of proper integrals.
Z ∞ Z R Z a Z a
f (x)dx = lim f (x)dx, f (x)dx = lim f (x)dx,
a R→∞ a −∞ R→−∞ R
Z ∞ Z ∞ Z a
f (x)dx = f (x)dx + f (x)dx
−∞ a −∞

A very important function in many areas of science that is defined through improper integrals is the gamma
function which is define for s > 0 (actually for complex numbers whose real part is greater than zero) as
Z ∞
Γ(s) = ts−1 e−t dt.
0

See the extra credit problem on Homework 4 for more about it and to see how it generalizes the factorial.
R∞ −3
Example. Let’s find 2 dx x3 . In other words, we’re finding the area under the curve y = x over the
interval 2 ≤ x < ∞. So the region is unbounded and the graph is given below.

y But for any R > 2,

Z R R
dx 1 1 1
=− 2 =− + .
2 x3 2x 2 2R2 8
2
Therefore,
Z R
dx 1
lim 3
= .
1 R→∞ 2 x 8

x−3

1 2 x

23
 R∞ dx
Example. Does 1 x converge?

y For R > 1, we have

Z R R
dx
= ln(x) = ln(R).
1 x 1
2
But ln(R) → ∞ as R → ∞. So the integral diverges.
We write Z ∞
dx
1 = ∞.
1 x
x−1

1 2 x

R∞ dx
Example. What about −∞ x2 +1
? We need to split into two sides, positive and negative.

y For any R > 0,

Z R R
1 dx
= tan−1 (x) = tan−1 (R).
0 x2+1 0

Now as R → ∞, tan(R) → π/2. So in fact

Z ∞
dx π
2
= .
0 x +1 2
1
x2 +1
We can use symmetry to figure out that
-2 2 x
Z 0
dx π
= .
−∞ x2+1 2

So the integral converges and equals π.

p-integrals. We now generalize the first two examples. Suppose p is any number and a > 0. Consider
Z ∞
dx
.
a xp

If p is large then we expect this to converge. If p is small (or negative) we expect this to diverge. In fact,
for R > a, p 6= 1, we get
Z R R
dx 1 1 1
= = − .
a xp (1 − p)xp−1 a (1 − p)Rp−1 (1 − p)ap−1

If p > 1, then 1/Rp−1 → 0 as R → ∞. In this case the integral converges and we get (verify with first
example) Z ∞
dx 1
p
= p−1
.
a x (p − 1)a
R∞
If p < 1 then 1/Rp−1 = R1−p → ∞ as R → ∞. So the integral diverges, and we write a xdxp = ∞. If p = 1,
RR
we get a dx x = ln(R) − ln(a) → ∞ as R → ∞, so the integral again diverges (see second example).

24
11 Lecture
Outline:
1. 7.7 Improper Integrals; Part II

11.1 Unbounded functions

An area can be unbounded in the horizontal or vertical direction. So another kind of improper integral is
one for which the function itself blows up, i.e. goes to infinity. If f (x) goes to ±∞ as x → a, then we define
Z b Z b Z b Z R
f (x)dx = lim+ f (x)dx, f (x)dx = lim− f (x)dx,
a R→a R a R→b a

and if the function blows up at both ends, we add both limits together.
R 1 dx
Example. 0 √1−x 2
. The function blows up as x → 1.

y For 0 < R < 1 we have

Z R
dx
√ = sin−1 (R)
0 1 − x2
5 which goes to π/2 as R → 1. So the integral con-
verges and is just π/2.

√ 1
1−x2

1x

p-integrals. Consider Z a
dx
.
0 xp
1
The function blows up as x → 0. We expect the integral to converge for small or negative p, the opposite
xp
of what we expected before. If 0 < R < a then
Z a
dx 1 1
p
= p−1
− .
R x (1 − p)a (1 − p)Rp−1
1
If p > 1 then Rp−1 → ∞ as R → 0+ , so the integral diverges. If p < 1 then it goes to zero and we get
Z a
dx 1
p
= p−1
.
0 x (1 − p)a

If p = 1 then the integral also diverges because

Z a
dx
= ln(a) − ln(R) → ∞, R → 0+ .
R x

25
11.2 Comparison Test
Sometimes all we want to know is whether an integral converges or diverges; we don’t care about the value.
So we compare the integral to another one we know about already.
R∞ R∞ R∞
If f (x) ≥ g(x) ≥ 0, then if a f (x)dx converges, so does a g(x)dx. Logically, if a g(x)dx diverges, so
R∞
does a f (x)dx.
R∞
Example. Does 1 √xdx 3 +1
converge? A good comparison technique is to make the denominator smaller
to get a larger function. In this situation, try this:
1 1
√ ≤ .
x3 +1 x3/2
R∞ 1
We know that 1 x3/2
dx converges because 3/2 > 1. So the integral converges.
R ∞ dx
Example. Does 1
√
x+e3x
converge? Well, if we compare to √1x , then we might
think not, but that would
be a mistake. Exponentials go to zero super-fast, so their integrals converge. In fact, if R > 1 then
R
e−3R e−3
Z
dx
= − +
1 e3x 3 3
R∞ dx e −3
so = . Now since √ 1 3x ≤ 1
1 e3x 3 x+e e3x , we see that the integral converges.
R 1/2
Example. Does 0 x8dx +x2 converge? Probably not, because both 8 and 2 are larger than 1. We need to
compare to a smaller function whose integral diverges. So we make the denominator larger. Well, we’re
looking at 0 ≤ x ≤ 1/2. So x < 1 and x to any power is also less than 1. Thus x8 < x2 . We get
1 1
≥ 2.
x8 + x2 2x
R 1/2 dx
But 0 2x2 diverges because 2 > 1, so the original integral also diverges.
Remark. Comparisons can be difficult at first. You have to know which part of a function is bigger than
the others, and where that’s true! In general, we tend to care most about what happens for x small and
for x large. So you need to know the difference in behavior for functions as x → 0 and as x → ∞. And
when you’re integrating, it’s not enough just to know what the limits are; you have to know how much area
accumulates under the curve!

26
Exam 1 Review
The first exam will cover techniques of integration.

Substitution [Section 5.7]

This method is used when a function and its derivative appear in the integrand. Remember to return to the
original variable for the final answer!
Change of Variables Formula:
Z b Z u(b)
f (u(x)) u0 (x) dx = f (u) du
a | {z } | {z } u(a)
f (u) du

Suggestions for Substitution

p
i) If a function is raised to a power or in a radical, i.e., [f (x)]2 , [f (x)]−1 , f (x), let u = f (x).
ii) If a function appears in the argument of another function, i.e., sin(f (x)), ln(f (x)), ef (x) , let u = f (x).
Note that i) is a particular case of ii) with the outer function being the power function, and these suggestions
do not always work!
Z Z Z Z
2 x dx ln x
2xex dx, x2 sin x3 dx


Examples: √ , dx,
2
x +1 x

Integration by parts [Section 7.1]

Extending integrals to the product rule:
Z Z Z
d du dv du dv dv du
(uv) = v+u → uv = vdx + u dx or u dx = uv − vdx.
dx dx dx dx dx dx dx
This is called integration by parts.
Another way to write the product rule is simply
Z Z
d(uv) = u dv + v du → uv = u dv + v du.

Strategy: pick out a u and dv in your integral, such that it is advantageous to pass the derivative from v to
u in order to make integration easier.
Suggestions for Substitution If one of the functions in the integrand can be eliminated or made simpler
by taking a few derivatives (or integrals), choose this to be u (or dv). Example:

u : xn , ln x, (ln x)n , inverse trigonometric functions

If neither function will be made simpler by derivatives or integrals (multiplication of trigonometric and
exponential), you will have to use integration by parts until the original integral reappears and then solve
for it.
Z Z Z Z
2 x
Examples: x cos x dx, x e dx, x ln xdx, ex cos xdx

27
Trigonometric integrals [Section 7.2]
Many integrals involve trigonometric functions. To compute them, you will need to remember one funda-
mental identity (the Pythagorean identity):
cos2 x + sin2 x = 1.
Divide by cos2 x to get
1 + tan2 x = sec2 x.
We can then use the provided Table of Trigonometric Integrals to evaluate most of these integrals. Keep in
mind that if we have high powers of one or more trigonometric functions, it is most likely advantageous to
use the substitution method rather than applying the trigonometric integral identities three or more times!
Z Z
2 3
Examples: sin x cos x dx, tan20 x sec4 x dx

Trigonometric substitution [Section 7.3]

We can use the Pythagorean identity to evaluate many integrals that have terms that involve the square of
our variable and a constant by introducing trigonometric functions cleverly, then returning to our original
variable.
Z Z Z
dx dx x dx
Examples: √ , √ ,
x2 − 1 9 − x2 x2 + 1

Partial fractions [Section 7.5]

This method is used to turn a rational function into simpler rational functions that we can easily integrate.
In order to do this, we need to remember how to setup the different forms we might encounter.
4x + 1 3x + 9 18
Examples: , ,
(x + 2)(x − 2) (x − 1)(x + 2)2 (x + 3)(x2 + 9)

L’Hôpital’s rule [Section 4.5]

Loosely speaking, L’Hôpital’s rule allows us to compare two functions that approach zero or infinity at the
same point.
Remark. Don’t forget to check for 0/0 or ±∞/ ± ∞!
x ln x
Examples: lim+ x ln x, lim , lim xx
x→0 x→∞ x2 x→0+

Improper integrals [Section 7.7]

We often think of integrals as areas of regions between curves, but some areas can be unbounded with
the area being finite or infinite. We call this an improper integral. If the integral is finite, we say it
converges. Otherwise, we say it diverges. To find out whether it converges, we need to take the limit of
proper integrals.
Z ∞ Z R Z a Z a Z ∞ Z ∞ Z a
f (x)dx = lim f (x)dx, f (x)dx = lim f (x)dx, f (x)dx = f (x)dx+ f (x)dx
a R→∞ a −∞ R→−∞ R −∞ a −∞

If f (x) goes to ±∞ as x → a, then we define

Z b Z b Z b Z R
f (x)dx = lim+ f (x)dx, f (x)dx = lim− f (x)dx,
a R→a R a R→b a
and if the function blows up at both ends, we add both limits together.

28
12 Lecture
Outline:
1. 6.1 Area Between Two Curves

12.1 Area between two curves

If f > g then the curve y = f (x) is above the curve y = g(x). The area between the two curves can be
computed by integration.
Example. Find the area of the region between the graphs of f (x) = x2 − 4x + 10 and g(x) = 4x − x2 over
1 ≤ x ≤ 3.

y First, let’s find out which curve is above the other:

2
x − 4x + 10
f (x) − g(x) = 2x2 − 8x + 10 = 2(x2 − 4x + 5)

= 2((x − 2)2 + 1) > 0.

5
So f > g. So the area is
Z 3 Z 3
(f (x) − g(x))dx = (2x2 − 8x + 10)dx
2 1 1
4x − x
3
2x3

2
= − 4x + 10x
1 3 x 3 1
 
2 16
= [18 − 36 + 30] − − 4 + 10 = .
3 3

Changing sides. Let’s find the area between the graphs of f (x) = x2 − 5x − 7 and g(x) = x − 12 over
[−2, 5].

y First, let’s look at

f (x) − g(x) = x2 − 6x + 5 = (x − 5)(x − 1).

-2 1 5 x
So f > g when x > 5 or x < 1, and g > f when
1 < x < 5. (The graph change sides at x = 1 and
x = 5.) We only care about −2 ≤ x ≤ 5, but we still
-7
have to watch for x = 1. That means the area is

x2 − 5x − 7
x − 12 -12

Z 1 Z 5 Z 1 Z 5
2
[f (x) − g(x)]dx + [g(x) − f (x)]dx = [x − 6x + 5]dx + [−x2 + 6x − 5]dx
−2 1 −2 1
1 5
x3
  3  
x 124 113
= − 3x2 + 5x + − + 3x2 − 5x = [3 + 9 + 15] + − + 72 − 20 = .
3 −2 3 1 3 3

29
A region enclosed by several curves. Let’s find the area enclosed by the graphs y = 8/x2 , y = 8x, and
y = x.

y This is basically going to look like a wedge in the first

10
quadrant. Note that for x < 0, the curves don’t inter-
sect. The three vertices of this wedge are the points
of intersection of the curves: x = 0, where y = 8x
and y = x meet, and as for the other two:
5
8x 8/x2 8x = 8/x2 ⇒ x3 = 1 ⇒ x = 1,
x
x = 8/x2 ⇒ x3 = 8 ⇒ x = 2.
We need to divide up the region between 0 ≤ x ≤ 1
x and 1 < x ≤ 2. In the first region, the top side is
1 2
y = 8x and the bottom is y = x. In the second, the
top is y = 8/x2 and the bottom is y = x. So the total
area is

1 2 2
8 x2
Z Z     
8 7 7 1
7xdx + − x dx = + − − = + [−4 − 2] − −8 − = 6.
0 1 x2 2 x 2 1 2 2

Integrating along the y-axis. Sometimes it’s advantageous to switch roles of x and y and write x as a
function of y. For example, let’s find the area of the region enclosed by y = x2 , y = −x + 2, and y = 0.

y x2 If we draw the picture, we get a little region with

vertices at x = 0, x = 2 (the points where the curves
touch the x-axis) and at the intersection of the two
curves, which is x = 1. If we solve this by integrat-
1 ing along the x-axis, we have to divide the region be-
tween the part where x2 is the top and the part where
−x + 2 −x + 2 is the top. Or, we can integrate along the y-
√
axis. Then y = x2 becomes x = y and y = −x + 2
1 2 x becomes x = −y +2. The second is on top of the first,
over the interval 0 ≤ y ≤ 1. So the area is simply

1  1
√
Z
1 2 1 2 5
(−y + 2 − y)dy = − y 2 + 2y − y 3/2 =− +2− = .
0 2 3 0 2 3 6

Remark: We could also have solved this as before by evaluating

Z 1 Z 2  2
1 1 1 3 5
x2 dx + (−x + 2)dx = + − x2 + 2x = + 2 − = .
0 1 3 2 1 3 2 6

Summary: If you are given two or more functions without a specified interval, find their intersections to
determine the enclosed region and find its area.

30
13 Lecture
Outline:
1. 6.2 Setting Up Integrals: Volume, Density, Average Value; Part I
2. Volume

13.1 Volumes using cross sections

To compute the volume of a solid body, divide the body into N horizontal slices of thickness ∆y = (b − a)/N .
Then the ith slice extends from yi−1 to yi , and let Vi denote the volume of this slice.
Notice if N is very large, then ∆y is very small and the slices are very thin. Furthermore, the ith slice has
area A(yi−1 ) and height ∆y, and hence, Vi ≈ A(yi−1 )∆y. Then the entire volume is obtained by summing
all the volumes of the slices,
XN XN
V = Vi ≈ A(yi−1 )∆y.
i=1 i=1
Z b
This sum should look familiar as it is the left-endpoint approximation to the integral A(y) dy, where
a
A(y) denotes the area function. Assuming A is continuous, the approximation will converge to the integral
as N → ∞. This allows us to conclude that the volume of the solid is equal to the integral of its
cross-sectional area, that is,
Z b
Volume of the solid body = A(y) dy.
a

Because it might be visually easier, your book proposes using the y-axis on the graph with objects that stand
up straight. But this is just a matter of orientation.
Example. Find the volume of a pyramid of height 12 whose base a square of side 4.
We will line up the pyramid so that the y-axis goes through from top to bottom, hitting the base at a right
angle at the origin. So at y = 0 the cross sectional area is 42 = 16, while at y = 12 the cross sectional area
is 0 (the top is a point). At any y in between 0 and 12, the cross section is a square whose side should have
length in proportion to the distance from the top. The distance from the top is 12 − y. So if x is the side
length,
12 − y 12 1
= ⇒ x = (12 − y).
x 4 3
2 1 2
Thus the area of the cross section is A(y) = x = 9 (12 − y) . The total volume, then, is
Z 12  12
1 2 1
(y − 12) dy = (y − 12)3 = 64.
0 9 27 0

Example. Compute the volume of the solid whose base is the region between the parabola y = 4 − x2
and the x-axis, and whose vertical (in the “z direction”) cross sections are semicircles perpendicular to the
y-axis.
Since the cross sections are perpendicular to the y-axis, we have to integrate in that direction, with
√ bounds
0 ≤ y ≤ 4 (y = 0 is the x-axis, and y = 4 is the vertex of the parabola). The radius is x = 4 − y, so
A(y) = 12 πx2 = 12 π(4 − y). So the volume is
Z 4
π π 4
(4 − y)dy = 4y − y 2 /2 0 = 4π.
0 2 2

31
Example. Compute the volume of a sphere of radius R.
People knew how to do this long before calculus, but thankfully we came after them!
We can set up cross sections perpendicular to any axis we want, so why not the x-axis? So taking −R ≤
x ≤ R, the cross sections have area A(x) = πr2 where x2 + r2 = R2 , thus A(x) = π(R2 − x2 ) and so
Z R R 4π 3
π(R2 − x2 )dx = π R2 x − x3 /3 −R = πR3 − πR3 /3 − π[−R3 + R3 /3] =

V = R .
−R 3

Cavalieri’s principle: If two solids have equal cross-sectional areas, then they have equal volume. (For
example, just take two stacks with the same number of quarters. Even if the quarters aren’t stacked perfectly
one on top of the other, they have the same volume.)

32
14 Lecture (not covered on homework or exams)
Outline:
1. 6.2 Setting Up Integrals: Volume, Density, Average Value; Part II
2. Linear mass density
3. Flow rate
4. Average value of a function

14.1 Linear mass density and total mass

Density can vary as a continuously changing function. Some objects can be thought of as one-dimensional,
for instance a rod. If ρ(x) is the mass density at length x from one end of the rod, then the total mass of
the rod is Z L
ρ(x)dx
0
where L is the length of the rod. Note since dx came from the limit of ∆x, its unit of measure is length.
Example. If a 2 meter rod has linear mass density ρ(x) = 1 + x(2 − x) kg/m, then its total mass is
Z 2 Z 2  2
1 8 10
1 + 2x − x2 dx = x + x2 − x3 = 2 + 4 − =


(1 + x(2 − x))dx = kg.
0 0 3 0 3 3

There are other one-dimensional quantities. A population density could depend only on distance from
the city center, for example. To find the total population, we need to integrate circles:
Z R
Population with radius R = 2πrρ(r)dr.
0

Example. Suppose the population density depends on the distance from the city center like ρ(r) = 15(1 +
r2 )−1/2 thousands per square kilometer. This makes sense, since the farther out you go, the fewer people
there are. How many people are there between 10 and 30 km from the city center?
Z 30 Z 901 901
2 −1/2
(u)−1/2 du = 30πu1/2


2πr 15(1 + r ) dr = 15π ≈ 1881 thousand.
10 101 101

14.2 Flow rate for a laminar flow

The flow of a fluid through a pipe being laminar means the velocity of each particle depends only on its
distance from the center of the tube. So just as for population density, we get
Z R
Flow rate = 2πrv(r)dr
0

where R is the radius of the pipe and v(r) is the velocity of particles.
Example. According to Poiseuille’s Law, the velocity of blood flowing in a blood vessel of radius R is
v(r) = k(R2 − r2 ) (where k is a constant). What is the total flow rate?
Z R 2
R
r4 πkR4

2 2 2r
2πkr(R − r )dr = 2πk R − = .
0 2 4 0 2
This actually shows the danger of reduced blood flow, say from plaque buildup. The flow is proportional to
R4 , so if the radius, R, is reduced by say 1/2, the flow rate is reduced by a factor of 16.

33
14.3 Average value
The average value of N numbers is found by adding them and dividing by N :
a1 + · · · + aN
.
N
If we take N values of a function f from equally spaced intervals in [a, b], and then we let N → ∞, we get
the average value of f on [a, b]:
Z b
1
Average value = f (x)dx
b−a a

Example. Let’s find the average value of f (x) = sin(x) on the interval [0, π]. Recall that this is the interval
on which sin is non-negative. We get
π
1 π
Z
1 2
sin(x)dx = − cos(x) = .
π 0 π 0 π

Remark: Note with an average of a list of numbers, the average may never be obtained in the list but it
always is for a continuous function by the Mean Value Theorem.
Extra Example. A ball is thrown in the air vertically from ground level with initial velocity 18 m/s. Find
the average height and average speed over the time interval extending from the ball’s release to its return to
ground level. (The acceleration due to gravity is 9.8 m/s2 )
Step 1: find a function for the ball’s height and velocity at time t.
The acceleration is constant in the downward direction, and initial velocity is 18, so v(t) = 18 − 9.8t is the
velocity. The height is therefore h(t) = v(t)dt = 18t − 4.9t2 + C, where h(0) = 0 which is ground level. So
R

h(t) = 18t − 4.9t2 .

Step 2: find out when the ball returns to ground level.
We need to know when the height is zero again, i.e. we need to solve h(t) = 0 or 18t − 4.9t2 = 0. This factors
to (18 − 4.9t)t = 0. We want a time after t = 0 so the solution is tmax = 18/4.9 ≈ 3.67.
Step 3: find the averages over [0, tmax ].
Average height: Z 3.67
1 1  2 3.67
(18t − 4.9t2 )dt = 9t − 1.63t3 0 ≈ 11.08 m.
3.67 0 3.67
Average speed: We need to be mindful of the fact that speed is actually the absolute value of v(t). Since
v(t) ≥ 0 when t ≤ 18/(9.8), we compute
"Z #
3.67/2 Z 3.67
1 2  3.67/2
(18 − 9.8t)dt + (−18 + 9.8t)dt = 18t − 4.9t2 0 ≈ 9 m/s.
3.67 0 3.67/2 3.67

34
15 Lecture
Outline:
1. 6.3 Volumes of Revolution: Disks and Washers

15.1 Volumes of revolution using disks and washers

If f (x) ≥ 0 and we take the area under the graph above the x-axis, we can revolve it around the x-axis to
get interesting symmetrical solids. The volume can be found by integrating cross sections (which are disks)
with area A(x) = πr2 = πf 2 (x), since the radius of each cross section is f (x). Hence,
Z b
V =π f 2 (x)dx.
a

Example. Revolve the area under the curve y = x2 over 0 ≤ x ≤ 2 about the x-axis to get a shape similar
to the opening of a musical horn.

y What is the volume?

4 x2
Z 2 2
x5 32π
V = πx4 dx = π = .
0 5 0 5

1 2 x

We can do the same thing with areas between two curves y = f (x) and y = g(x) which are both above the
x-axis, but then the cross sections are washers–there’s a hole in the middle. We need to delete the hole
from the area: A(x) = πf 2 (x) − πg 2 (x).
Example. We make a lamp shade by revolving the area between y = x2 + 4 and y = 2 over 1 ≤ x ≤ 3
around the x-axis.

y The volume we get is

13 x2 + 4 Z 3
2126
V =π [(x2 + 4)2 − (22 )]dx = π.
1 15

y=2
2

1 2 3 x

35
Example. We make a wedding band by revolving the area enclosed by y = 4 − x2 and y = x2 + 2 around
the horizontal line y = −3. To find the limits of integration, and which curve goes on top, we compare:

4 − x2 − (x2 + 2) = −2x2 + 2 = −2(x2 − 1) = −2(x − 1)(x + 1).

The points of intersection are x = 1 and x = −1. If x is between 1 and −1, then −2(x − 1)(x + 1) > 0 (since
(−)(−)+ = +) so 4 − x2 > x2 + 2.

y Each cross section is a washer. The outer radius is

not 4 − x2 but rather 4 − x2 − (−3) = 7 − x2 , because
4 − x2 we revolve not around the x-axis but around y = −3.
4
Likewise the inner radius is not x2 + 2 but x2 + 5. So
3
Z 1
2
x2 + 2 V =π [(7 − x2 )2 − (x2 + 5)2 ]dx = 32π.
−1

-1 1 x

-3

Example. If we want to revolve around a vertical axis instead, we can still use washers or disks, but we
have to find x as a function of y. For instance, take the region between the graph y = 9 − x2 and the x-axis
over 0 ≤ x ≤ 3 and revolve it around the axis x = −2.

y √
Now x = 9 − y is the outside curve; the top of the
9 − x2
9
region is y = 9. Each
√ cross section is a washer whose
outside radius is 9 − y + 2 and whose inside radius
is just 2. So the volume is
Z 9 p
225
V =π [( 9 − y + 2)2 − 4]dy = π.
0 2

-2 3 x

36
16 Lecture
Outline:
1. 6.4 Volumes of Revolution: Cylindrical Shells

16.1 Volumes of revolution using cylindrical shells

The method of cylindrical shells is different from earlier methods, because we don’t look at cross sections
perpendicular to a central axis. Instead, we look at shells whose height is parallel to the central
axis. This is often useful if we take the region under a graph y = f (x) and revolve it around the y-axis.
The result is that we consider the surface areas of the different shells and add them up.
Notice that each of the surface areas can be found by “cutting” and flattening them to see they are simply
rectangles with width given by the circumference of the cylinder considered, 2πx, and height given by the
function, f (x). This gives
Z b Z b
V = 2π (radius)(height of shell)dx = 2π xf (x) dx.
a a

Example. Let’s take the region under the graph y = f (x) = x3 (and over the x-axis) between 0 ≤ x ≤ 1
and revolve it around the y-axis. We’ll get a sort of bowl with square corners.

y Each cylindrical shell has a height of f (x), and its

x3 radius is x. The circumference is 2πx, so the sur-
1 face area is 2πxf (x). But the shell actually has a
tiny width dx, so its volume is 2πxf (x)dx, and we
integrate to get the volume of the whole solid.
In this case we have
Z 1 Z 1
2π
V = 2πx · x3 dx = 2π x4 dx = .
0 0 5

1 x

√
Example. Take the region enclosed by y = x and y = x2 and revolve it around the y-axis. First, we need
to know what this region looks like.

y If we understand the shape of these functions, we’ll

√
see that x ≥ x2 when 0 ≤ x ≤ 1 but 2
√ after that x is
1 √ bigger and grows much faster than x. So the region
x
that gets revolved√is in 0 ≤ x ≤ 1. Then the height
of each “shell” is x − x2 . So the volume is
Z 1 √
Z 1
x2 V = 2π 2
x( x − x )dx = 2π (x3/2 − x3 )dx
0 0

 1
x 2 1 3π
1 = 2π x5/2 − x4 = .
5 4 0 10

37
Example. This time we’ll revolve around a different vertical axis, say x = −3. Take the region under the
graph y = x−1/2 over the interval 1 ≤ x ≤ 4.

y The height of each shell will be x−1/2 , but the radius

is no longer x but x − (−3) = x + 3. So
1
Z 4 Z 4
x−1/2
 
V = 2π (x+3)x−1/2 dx = 2π x1/2 + 3x−1/2 dx
1 1
 4
.5 2 3/2 64π
= 2π x + 6x1/2 = .
3 1 3

-3 1 4 x

Example. For our last example, we’ll go back to revolving around the x-axis. But we’ll see that using
the normal disk method will be very tricky, and the shell method will help. We take the region between the
graph y = sin−1 (x) and the x-axis over 0 ≤ x ≤ 1 and revolve it around the x-axis.

y If we used the disk method the volume would be cal-

culated
sin−1 (x)
π/2
Z 1
V = π(sin−1 (x))2 dx =???
0

(Two integrations by parts work, but seems hard!)

But by looking sideways, we can use the shell method.
We see that 0 ≤ y ≤ π/2, and the shells will have
height 1 − sin(y). So we get (using integration by
parts with u = y and dv = − sin(y) dy)
1 x Z π/2
V = 2π y(1 − sin(y))dy
0
Z π/2 Z π/2
= 2π ydy − 2π y sin(y)dy
0 0
π/2 Z π/2
y2

= 2π + y cos(y) − 2π cos(y)dy
2 0 0
 2 π/2
y
= 2π + y cos(y) − sin(y)
2 0

π3
= − 2π.
4

38
17 Lecture
Outline:
1. 8.2 Arc Length and Surface Area

17.1 A Short Warm Up

Let’s motivate the topics of arc length and surface area using geometric shapes we know first. Consider a
circle with radius r. Then we know that the area of the circle is Ac = πr2 , and the derivative of the area
with respect to the radius is A0c = 2πr = circumference = arc length of the circle. This is no coincidence
since as we add or subtract area, the circle changes by adding or subtracting portions around its boundary
or circumference (keeping the center fixed).
If the derivative of the area of a circle gave circumference, what about the derivative of the volume of a
sphere? Well, we know Vs = (4/3)πr3 , so Vs0 = 4πr2 = surface area of the sphere.
Remark: We need to be careful with other shapes as this is a result of changing the same in all directions,
that is, we need a notion similar to radius. For instance, to consider a square and a cube, we need to
label each side as 2x so that we change the same when fixing the center. Then we find As = (2x)2 = 4x2 ,
A0s = 8x = perimeter = arc length of the square, and Vc = (2x)3 = 8x3 , Vc0 = 24x2 = 6 · 4x2 = 6 · As =
surface area of the cube.

17.2 Arc length

Suppose we want to find the length of a curve given by y = f (x), with starting point x = a and end point
x = b. We break up the curve into pieces; each piece can be seen as the hypotenuse of a triangle whose
horizontal leg has length given by the change in x, ∆x, and whose p vertical leg has length
p approximately
f 0 (x)∆x. So the length of each hypotenuse is approximately ∆x2 + (f 0 (x))2 ∆x2 = 1 + (f 0 (x))2 ∆x.
Summing these up gives the total, therefore the whole length of the curve–its arc length–is given by the
integral
Z bp
L= 1 + (f 0 (x))2 dx.
a

Example. Let’s say I throw a ball into the air at an initial speed of 9.8 m/s vertically at 1 m/s horizontally.
Using some physics, I can determine that the path of the ball is given by y = 9.8x − 4.9x2 on the interval
0 ≤ x ≤ 2, where x is the horizontal distance traveled by the ball. What is the total distance the ball travels
through the air?
We want the arc length of the curve y = 9.8x − 4.9x2 on the interval [0, 2]. Now y 0 = 9.8 − 9.8x, so
Z 2 Z x=2 p
p
2
1
L= 1 + (9.8 − 9.8x) dx = 1 + u2 du, u = 9.8x − 9.8.
0 9.8 x=0

Now using the trig substitution u = tan θ eventually leads to

Z p
1 p 1 p
1 + u2 du = u 1 + u2 + ln |u + 1 + u2 | + C.
2 2
Therefore the length of the ball’s flight path is
 2
1 1 p 1 p
L= (9.8x − 9.8) 1 + (9.8x − 9.8)2 + ln |9.8x − 9.8 + 1 + (9.8x − 9.8)2 | ≈ 10.15 m
9.8 2 2 0

39
 √
Conceptual understanding. Consider the√two arcs y = x and y = x2 over the same interval 0 ≤ x ≤ 1.
Note that on this interval we can write y = x as x = y 2 with 0 ≤ y ≤ 1. In other words, the two arcs are
just mirror images of each other. They should have the same arc length.
√ 1
What does the formula say? On the one hand, since the derivative of x is 2√ x
, we get
Z 1
r
1
L= 1+ dx.
0 4x

On the other hand, since the derivative of x2 is 2x we get

Z 1p
L= 1 + 4x2 dx.
0
√ 1
To see that the two integrals are equal, let’s look at the first. Set u = x to get du = √
2 x
dx so dx = 2u du.
Thus Z 1r Z 1r Z 1p
1 1
1+ dx = 1 + 2 2u du = 4u2 + 1du,
0 4x 0 4u 0
which is exactly the same as the second integral.

17.3 Area of a surface of revolution

Let’s say we take a curve y = f (x) over an interval [a, b] like before, only this time we revolve it around
the x-axis to get a surface. What is its area?p Again, we break up the surface into pieces. Each piece is
a ring whose width can be approximated by 1 + (f 0p (x))2 ∆x and radius f (x). Therefore the area is the
circumference times the width, which is about 2πf (x) 1 + (f 0 (x))2 ∆x. We integrate this to get the total
surface area: Z b p
SA = 2π f (x) 1 + (f 0 (x))2 dx.
a
√
Example. What is the surface area of the paraboloid obtained by revolving the graph of y = x around
the x-axis over 0 ≤ x ≤ 1?
Well, y 0 = 1
√
2 x
so
1 "   3/2 #
3/2
1
r 1
r
√
Z Z 
1 1 2 3/2 4π 5 1
SA = 2π x 1+ dx = 2π x + dx = 2π (x + 1/4) = − .
0 4x 0 4 3 0 3 4 4

Example. What is the surface area√ of a sphere of radius R? Well, we can revolve a semicircle around the
2
x-axis to get the sphere. Let y = R2 − x2 . Then y 0 = − √R2x−x2 so (y 0 )2 = R2x−x2 . The surface area is

R
r R R
x2
Z p Z p Z
SA = 2π R 2 − x2 1+ dx = 2π R2 − x2 + x2 dx = 2π Rdx = 2πR(2R) = 4πR2 .
−R R − x2
2
−R −R

By simply changing the limits of integration, we get a simple formula for the area of any spherical
cap of height h. Indeed, if we integrate not from −R to R but instead from R − h to R, the area we get is
A = 2πRh. Which is kind of odd, when you think about it!
What it is really saying is the surface area is increasing with respect to the height by the circumference.

40
18 Lecture
Outline:
1. 6.5 Work and Energy
Great quote from the textbook:
For those who want some proof that physicists are human, the proof is in the idiocy of all the different units
which they use for measuring energy. –Richard Feynman, The Character of Physical Law

18.1 Work and Energy

A constant force F that pushes an object a straight distance d is said to perform a certain amount of work
given by
W = F · d.
Work is equivalent to energy; we can think of it as the energy expended by the trip.
In the International System (SI), the units for work are Newton meters (N · m), also known as Joules (J).
Recall that one Newton is the weight of one kilogram: 1N = 1kg × m/s2 . In the British system, the units
are simply foot pounds, because pounds are already a measure of weight, hence of force.
But if the force you apply changes at each point along the way, how do we compute the total work done?
Let’s say the path lies on the x-axis and the force is given by F (x). Loosely speaking, we divide up the
interval into little pieces of length dx and add up to get
Z b
W = F (x)dx
a

where a is the starting point and b is the end.

Hooke’s Law. If you stretch a spring out a distance x from its natural resting position, it will exert a
restoring force equal to −kx, where k > 0 is a constant depending on the spring. Let’s say, for example,
k = 200N/m. What is the amount of work required to stretch the spring a distance of 5 cm from rest?
We need to counteract the restoring force at each point using a force

F (x) = k (N/m) x (m) = 200x (N).

So the total work done will be

Z .05
.05
200x (N) dx (m) = 100x2 0
(N · m) = .25 N · m
0

Emptying out a tank. A spherical tank of radius 5 m is filled with water. Calculate the work performed
in pumping out the water through a spout of height 1 m at the top.
Note: the density of water is 1 g/cm3 = 1000 kg/m3 .
Let’s
p put the origin at the center of the tank. We divide up the tank into circular cross sections of radius
52 − y 2 and vanishing vertical width dy, for −5 ≤ y ≤ 5. Hence the volume is π(25 − y 2 )dy (m3 ). Very
loosely, let us think about the force required to move a volume of the water as how much water there is
(volume) times the density of the water times the force we are fighting (gravity). The force required to lift
this volume of water is then

F (y) =(gravity) × (density) × (volume)

=9.8 (m/s2 ) × 1000 (kg/m3 ) × π(25 − y 2 )dy (m3 ) = 9800π(25 − y 2 )dy (N).

41
The weight of this particular cross section of water will not change. The distance it has to travel is 6 − y
(the distance to the top of the spout). So the work done on the cross section is 9800π(25 − y 2 )(6 − y)dy J.
Therefore the total work done is
Z 5
9800π(25 − y 2 ) (6 − y) dy = 9.8 × 106 π J.
−5 | {z } | {z }
force per section distance

Building a pyramid. The Great Pyramid of Giza in Egypt is 146 m high and has a square base of side
230 m. The density of the stone is estimated at 2000 kg/m3 . Find the work required to build the pyramid.
Again, we find the work needed to lift each cross section into place. Put the origin at the center of the base.
At height y, the square cross section has length
230
x= (146 − y).
146
2
So the volume of the cross section with vertical width dy is 230 2
1462 (146 − y) dy. The force needed to lift it
2
230 2
is therefore 9.8 × 2000 × 1462 (146 − y) dy (acceleration times density times volume). It needs to travel a
2
distance of y. So the work performed on the cross section is 19600 × 230 2
1462 (146 − y) y dy. The total work is

146 146 146

2302 2302 2302 146 3 1 4
Z Z 
2 2
19600 (146 − y) y dy = 19600 u (146 − u)du = 19600 u − u
1462 0 1462 0 1462 3 4 0
1
= 19600 × 2302 × 1462 × ≈ 1.84 × 1012 J.
12

42
19 Lecture
Outline:
1. 9.1 Differential Equations, Part I

19.1 Differential Equations

A differential equation is any equation involving both a function and its derivatives. (We are only dealing with
ordinary differential equations. A partial differential equation involves a function and its partial derivatives
in several variables.) For example:
2
d2 y d3 y

dy dy
= y, + + y 3 = 0, = 0, etc.
dx dx2 dx dx3
Recall that this notation, the Leibniz notation, is used to emphasize the rates of change considered and will
be paramount in solving differential equations. A solution to a differential equation is any function that
dy
makes it true. For instance, y = ex is a solution of dx = y.
If all of the solutions to a differential equation are known, and we can write them down by a formula, we
call this the general solution. For example, if f (x) is a continuous function, then the general solution of
dy dy
= x is y = 12 x2 + C. That is, all of the
R
dx = f (x) is simply y = f (x)dx, e.g. the general solution of dx
solutions have this form, where C is any constant. We contrast general solutions with particular solutions.
dy
So y = 12 x2 would be just one particular solution of dx = x.
A differential equation is linear if it can be written like

an (x)y (n) + · · · + a1 (x)y 0 + a0 (x)y = b(x),

where a0 (x), a1 (x), . . . , an (x), b(x) are functions of the independent variable x and y, y 0 , y 00 , . . . , y (n) are the
unknown function y and its derivatives. This is analogous to a linear equation in n + 1 unknowns. Most
differential equations are nonlinear.
The order of a differential equation is the number of higher derivatives appearing in it. So the order of
y 00 + y 0 − 2y = 0 is 2, while the order of y 0 = y is 1. Most differential equations that are useful in applications
tend to be just first or second order, but not always.

Separable equations
We already know how to solve the simplest kind of differential equation, y 0 = f (x). A slightly less simple
kind looks like
y 0 = f (x)g(y).
This kind is called separable, because we can separate variables. To do this, write y 0 as dy/dx. Then
move everything with y in it to the left side, and everything with x in it to the right side. We get
dy
= f (x)dx.
g(y)
The next step is to integrate. If we’re lucky, we’ll be able to find a formula for y that looks something like

G(y) = F (x) + C ⇒ y = G−1 (F (x) + C).

In this case we’ve found the general solution to the separable equation.
dy
Example. Let’s find the general solution of y dx = x. In fact, it’s almost already separated for us; just
y2 x2
multiply by dx to get y dy = x dx. Now integrate to get 2 = 2 + C or just y 2 = x2 + C (remember C

43
 √
is any constant, and multiplying by 2 doesn’t change that fact). So y = ± x2 + C is the general solution
(note the plus and minus sign). Note that x2 + C must be non-negative in order for this to make sense. If
C is negative, this will put a restriction on the domain of y.
Remark: One needs to be very careful when solving a differential equation where either the logarithmic or
exponential equations appear at any step of the work as is seen in the next two examples.
Example. Let’s find the general solution of y 0 = ex+y . In fact this is separable because ex+y = ex ey . We
separate variables to get e−y dy = ex dx. Integrate to get −e−y = ex + C, so e−y = C − ex (again, C is
any constant). Thus −y = ln(C − ex ) which means y = − ln(C − ex ) is the general solution. Note that
this solution only exists on the interval where ex < C, since you can only take the logarithm of a positive
number; so in fact not all constants will give a solution, and even when a solution exists its domain isn’t the
whole number line!

Initial value problems

For a first-order differential equation, there is usually at most one solution if we also specify an initial value,
that is, if we specify that the solution must have some particular value at some particular value of x (or
time, t). In terms of physical applications, this usually means that we specify a starting point and then let
the dynamics tell us the trajectory from that initial value.
Example. Solve the initial value problem y 0 = −ty, y(0) = 3. Note that the independent variable is t
instead of x.
dy 2
First, we find the general solution of y 0 = −ty. Separate variables to get y = −t dt. Thus ln |y| = − t2 + C,
2
so |y| = eC−t /2
.
The question now becomes how do we remove the absolute value?
We do this by introducing both the positive and negative choices on the right-hand side, that is, we have
2 2 2
y = ±eC−t /2 = ±eC e−t /2 . Since y = 0 is also a solution, we can say y = Ae−t /2 is the general solution,
where A can be any constant. (Note: the reason y = 0 was not found by separation of variables is that we
had to assume y 6= 0 in order to integrate dy
y .)

Next, we take into account the initial value. We must have y(0) = 3. But then, y(0) = A. That means
2
A = 3. So we deduce y = 3e−t /2 .

44
20 Lecture
Outline:
1. 9.1 Differential Equations, Part II (A variety of applications)

20.1 Exponential growth and decay

Differential equations are often used to model real situations. This means a physical quantity is supposed
to solve a differential equation. A very typical model is
dy
= k(y − b).
dt
We can solve this by separating variables.
dy
= k dt ⇒ ln(|y − b|) = kt + C ⇒ y = Aekt + b.
y−b
Note that A can also be zero. In this case we get the equilibrium solution y = b.
This model describes exponential growth or decay.
• k > 0 implies exponential growth; y goes to ∞ or −∞ in the long run, depending on the sign of A.
• k < 0 implies exponential decay; y goes to b in the long run.

20.2 Short Term Population growth

We can use exponential growth to model, at least for an initial growth period, population growth as in the
following example.
Example. In the laboratory, the Escherichia coli bacteria grows such that the rate of change of the
population is proportional to the population present. Assume that 1000 bacteria are initially present, and
1500 are present after 1 hour. Determine the population, P (t), after t hours.
Since the rate of change, P 0 (t), is proportional to P (t), we have the differential equation

P 0 (t) = kP (t).

The general solution of this differential equation is P (t) = Aekt , and since P (0) = 1000, we easily conclude
that A = 1000. Furthermore, as P (1) = 1500, we have 1500 = 1000ek , so that we can see k = ln(1.5). Thus
we conclude that P (t) = 1000eln(1.5)t .

20.3 Half-Life and Doubling Time

Often it is beneficial to look at how long it will take for half of something to decay, or for it to double, which
is what we mean by half-life and doubling time respectively. One can find that from P (t) = P0 ekt , if k < 0,
half-life is given by ln(0.5)/k and if k > 0, the doubling time is given by ln(2)/k. This is straightforward
enough to check:
P (ln(0.5)/k) = P0 ek ln(0.5)/k = P0 eln(0.5) = 0.5P0
and
P (ln(2)/k) = P0 ek ln(2)/k = P0 eln(2) = 2P0 .

Example. A patient is administered 400 mg of penicillin, and after 50 minutes, 300 mg remain in her
bloodstream. Let A(t) represent the amount of penicillin (in mg) in her bloodstream t minutes after the

45
drug was administered. Assume that the drug is leaving her bloodstream at a rate proportional to the
amount in her bloodstream. Determine A(t) and the half-life of the resulting exponential decay.
Since the rate of change, A0 (t), is proportional to A(t), we have the differential equation

A0 (t) = kA(t).

The general solution of this differential equation is A(t) = A0 ekt , and since A(0) = 400, we easily conclude
that A0 = 400. Furthermore, as A(50) = 300, we have 300 = 400e50k , so that we can see k = ln(0.75)/50.
Thus we conclude that A(t) = 400et ln(0.75)/50 and the half-life is given by ln(0.5)/(ln(0.75)/50) ≈ 120
minutes, that is, it will take 120 minutes for half the medicine to leave her bloodstream.

20.4 Newton’s law of cooling

Let y be the temperature of an object in an ambient temperature T0 . Then there is a cooling constant k > 0
such that y 0 = −k(y − T0 ), so that the object’s temperature will tend toward T0 in the long run. The units
on k are inverse time, e.g. (min)−1 .
Example. A metal bar has cooling constant k = 2.1 min−1 . Suppose the bar is submerged in water held at
10 degrees Celsius. Then the temperature of the bar y(t) satisfies y 0 = −2.1(y − 10). So y(t) = 10 + Ae−2.1t .
Here the initial temperature is y(0) = 10 + A, so A = y(0) − 10.
If its initial temperature is 180 degrees, then y(t) = 10+170e−2.1t . After one minute it cools to 10+170e−2.1 ≈
30.8 degrees.
Suppose we ask the opposite problem. After 30 seconds, we know the bar has cooled to 80 degrees. How
warm was it to begin with? Well, y(.5) = 80 means 10+Ae−2.1·.5 = 80 or Ae−1.05 = 70, so A = 70e1.05 ≈ 200.
Thus the initial temperature is A + 10 ≈ 210 degrees.

20.5 Free fall with air resistance (terminal velocity)

Suppose you are skydiving. Your mass is m, so the size of the force of gravity downward is mg. But
thankfully air also resists proportionally to your speed, that is, if v is your velocity there’s a constant k > 0
(measured in units of mass/time, e.g. kg/s) such that the force of air resistance against you is −kv. If
positive v means you’re going up and negative v means you’re going down, the total force is

F = −mg − kv.

But F = ma = mv 0 , so mv 0 = −mg − kv or
k k mg
v0 = − v − g = − (v + ).
m m k
The general solution is v(t) = − mg k + Ae
−tk/m
. In the long run, your velocity will be − mg
k rather than
increase indefinitely; this is called terminal velocity. For more on this, see Homework 8, Exercise 66.
Example. Let’s say your mass is 80 kg and the air resistance constant is k = 8 kg/s. The terminal velocity
is then − 80×9.8
8 = −98 m/s, which means 98 m/s toward the ground. (This is very fast!!) To find your speed
30 seconds after you jump from the plane, note that your initial velocity is 0, so 0 = −98 + A i.e. A = 98.
So at t = 30 seconds we have v(30) = −98 + 98e−30·8/80 = −98 + 98e−3 ≈ −93.1 m/s. Thus your speed
increases quite quickly, but afterward it tapers off very quickly as well.

20.6 Annuities
An annuity is an investment which gains interest while money is also withdrawn. The initial investment P0
is called the principal. After time t the investment has grown (or shrunk) to P (t), satisfying the differential

46
equation  
N
P 0 = rP − N = r P − ,
r
where r is the interest rate (usually given as a percentage per unit time) and N is the withdrawal rate (given
as an amount of money per unit time). The general solution is P (t) = Nr + Aert . This suggests exponential
growth, but everything depends on the sign of A. If A > 0, we get endless growth, but if A < 0, we get total
disaster!
To see whether A > 0 or A < 0, we compare Nr to the principal. Indeed, P0 = P (0) = Nr + A, so A = P0 − Nr .
If P0 > Nr , then the investment grows forever. If not, we will quickly lose all our money! In other words,
do not withdraw too quickly. The rule we must follow is N < P0 r, meaning that the principal times the
interest rate determines how much we can withdraw.

47
21 Lecture
Outline:
1. 9.4 The Logistic Equation

21.1 The logistic equation

In the last section we saw differential equations that had solutions being exponential functions. These
differential equations were used to model population growth of bacteria, but we noted this was only valid
for short time periods as a population cannot grow indefinitely. That brings us to the idea of changing the
formulation of the problem considered previously in order to model both short and long term growth.
Let k > 0 be a growth constant, and A > 0 a carrying capacity (or saturation constant). If y(t) represents
the population at time t, we let A − y(t) represent the room available for growth. Now we assume that the
rate of change of y is proportional to the amount present, y(t), and the amount of room for growth, A − y(t).
Translating this into a differential equation, we obtain the logistic differential equation,
dy  y k
= ky 1 − = y(A − y).
dt A A
For small y, y grows approximately exponentially. The logistics equation has two constant solutions: the
equilibrium or steady state solutions y = 0 and y = A.

21.2 Solving the logistic equation

The logistic equation is separable (assuming y 6= 0, A) so that we can divide:

A dy
= k dt.
y(A − y)

We need to integrate both sides. Use partial fractions:

A a b
= + ⇒ A = a(A − y) + by = (b − a)y + aA.
y(A − y) y A−y

Thus b = a = 1. We get
Z    
|y|
Z
A dy 1 1
= + dy = ln(|y|) − ln(|A − y|) + C = ln + C.
y(A − y) y A−y |A − y|

It follows that  
|y| y
ln = kt + C ⇒ = Cekt ,
|A − y| A−y
where C always means some undetermined constant, but not necessarily the same one in each equation. We
solve for y:
ACekt AC
y = (A − y)Cekt ⇒ (1 + Cekt )y = ACekt ⇒ y = = .
1 + Cekt C + e−kt
Note: This is different from the solution given in the book:
A
y= .
1 − e−kt /B

However, the two are really not different if C 6= 0, since we can set B = −C and get the same formula.

48
21.3 Stable and unstable equilibrium
We can see that the solution has a limit as t → ∞. In fact,
AC AC
lim = = A.
t→∞ C + e−kt C
So as long as y0 6= 0, we know that y goes to A in the long run. This is why we call A a stable equilibrium.
On the other hand, 0 is an unstable equilibrium because any solution that begins at any other point, no
matter how close to 0, will quickly move away from 0.
This immediately implies that if the solution is ever equal to 0 or A, then the solution is the constant
solution y = 0 or y = A (i.e. it is always 0 or A).

21.4 Example: deer population

A deer population grows logistically with growth constant k = 0.4/year and carrying capacity A = 1000
deer. Starting at 100 deer, what is the population after t years? How long does it take to grow to 500 deer?
Let y(t) be the population after t years. We have
 
1000C 1000
y(t) = = .
C + e−0.4t 1 − e−0.4t /B
| {z }
using book form

Applying the initial condition shows

1000C 1
100 = y(0) = ⇒ 100C + 100 = 1000C ⇒ C = .
C +1 9
Hence,
1000
y(t) = .
1 + 9e−0.4t
To answer the second question, we solve y(t) = 500 for t. That is,

1000 1 ln(9)
= 500 ⇒ 2 = 1 + 9e−0.4t ⇒ = e−0.4t ⇒ − ln(9) = −0.4t ⇒ t = ≈ 5.5 years.
1 + 9e−0.4t 9 0.4

49
Exam 2 Review
Do not forget that integration techniques from the first test can still show up on this test!
The second exam will cover applications of integration:

Area Between Curves [Section 6.1]

If f > g then the curve y = f (x) is above the curve y = g(x). The area between the two curves can be
computed by integration. We can find the area enclosed by more curves by breaking the region into sections
where we only consider two curves at a time as seen in the following example.
A region enclosed by several curves. Let’s find the area enclosed by the graphs y = 8/x2 , y = 8x, and
y = x.

y This is basically going to look like a wedge in the first

10
quadrant. Note that for x < 0, the curves don’t inter-
sect. The three vertices of this wedge are the points
of intersection of the curves: x = 0, where y = 8x
and y = x meet, and as for the other two:
5
8x 8/x2 8x = 8/x2 ⇒ x3 = 1 ⇒ x = 1,
x
x = 8/x2 ⇒ x3 = 8 ⇒ x = 2.
We need to divide up the region between 0 ≤ x ≤ 1
x and 1 < x ≤ 2. In the first region, the top side is
1 2
y = 8x and the bottom is y = x. In the second, the
top is y = 8/x2 and the bottom is y = x. So the total
area is

1 2 2
8 x2
Z Z     
8 7 7 1
7xdx + − x dx = + − − = + [−4 − 2] − −8 − = 6.
0 1 x2 2 x 2 1 2 2

Volume [Section 6.2]

Z b
Volume of the solid body = A(y) dy.
a

This method is used to find the volume of a given shape (pyramid, sphere, cone, etc.) where cross sections
can be easily understood and have area A(y). Remember that often similar triangles are needed to find a
formula for the area.

Disk and Washer Method [Section 6.3]

If f (x) ≥ 0, we can revolve the area under the curve around the x-axis to get symmetrical solids. The
volume can be found by integrating cross sections (which are disks) with area A(x) = πr2 = πf 2 (x), since
the radius of each cross section is f (x). Hence,
Z b
V =π f 2 (x)dx.
a

50
Shell Method [Section 6.4]
The method of cylindrical shells is different from earlier methods, because we look at shells whose height
is parallel to the central axis. The result is that we consider the surface areas of the different shells and
add them up. Notice that each of the surface areas can be found by “cutting” and flattening them to see
they are simply rectangles with width given by the circumference of the cylinder considered, 2πx, and height
given by the function, f (x). This gives
Z b Z b
V = 2π (radius)(height of shell)dx = 2π xf (x) dx.
a a

Arc Length and Surface Area [Section 8.2]

The length of a curve–its arc length–is given by the integral
Z bp
L= 1 + (f 0 (x))2 dx.
a

The surface area of a surface produced by revolving a curve around the x-axis is given by
Z b p
SA = 2π f (x) 1 + (f 0 (x))2 dx.
a

Work and Energy [Section 6.5]

A constant force F that pushes an object a straight distance d is said to perform a certain amount of work
given by
W = F · d.
Units for work are Newton meters (N · m), also known as Joules (J). Recall that one Newton is the weight
of one kilogram: 1N = 1kg × m/s2 .
How do we compute the total work done? Let’s say the path lies on the x-axis and the force is given by
F (x). Loosely speaking, we divide up the interval into little pieces of length dx and add up to get
Z b
W = F (x)dx
a
where a is the starting point and b is the end.
You may be asked to set up, but not solve, problems involving springs and emptying tanks.
Emptying out a tank. A good way to remember how to approach this problem is to think of the integrand
as the force required to move a volume of the water, which is how much water there is (volume) times the
density of the water (given to you) times the force we are fighting (gravity), multiplied by how far we have
to move the water out of the tank and pipe above the tank.

Differential Equations [Section 9.1]

Find general solution of a differential equation using the separation of variables technique.
dy dy
Examples: y = x, = ex+y
dx dx
Solve initial value problems by applying initial condition to find the particular solution.
Example. Solve the initial value problem y 0 = −ty, y(0) = 3.
2
First, we find the general solution y = Ae−t /2
, where A can be any constant.
2
Next, we take into account the initial value to deduce y = 3e−t /2
is our particular solution.

51
22 Lecture
Outline:
Z
sin x2 dx, which we still cannot compute!


1. 10.1 Sequences, Recall: We wanted to find

22.1 Sequences
A sequence, often denoted {an }, is an infinite set of numbers placed in order:

a1 , a2 , a3 , . . .

For example, the Fibonacci sequence:

1, 1, 2, 3, 5, 8, 13, 21, . . .
The sequence goes on forever. To get the next number, you add the previous two together. If you take the
ratio of successive terms in this sequence, you get closer and closer to the golden ratio. This sequence also
appears in a surprisingly wide variety of situations, such as the number of spiral arms in a sunflower almost
always is a number from this sequence!
The nth term in the sequence is often written an . Sometimes it has a formula. For example, the sequence

2, 4, 8, 16, 32, . . .

can have its nth term written as an = 2n .

Or sometimes you can write an in terms of other members of the sequence. For the Fibonacci sequence,
an = an−1 + an−2 .
As sequences are infinite, the interesting question is whether they approach a certain limit.
We say limn→∞ an = L if for every  > 0 there exists N large enough such that |an − L| ≤  for every n ≥ N .
In other words, an gets closer and closer to L in a way that can be quantified. If no limit exists, we say that
{an } diverges.
Example. Consider the sequence
5 6 7 8 9 10
, , , , , ,...
2 3 4 5 6 7
This can be written an = n+4
n+1 . It looks like an → 1. Let’s prove it.

Let  > 0. We want to show that

n+4
−1 ≤
n+1
n+4 3 3 3 3
for n large enough. Now n+1 −1 = n+1 . So we want  ≥ n+1 which means n + 1 ≥  or n ≥  − 1. In
3
other words, |an − 1| <  for all n >  − 1, thus proving an → 1. Let’s try to understand what this means:
3
If you give me a value for , say .1, I will choose n > .1 − 1 = 30 − 1 = 29, say n = 30, so a30 will be within
.1 of 1, which you can easily check (a30 = 34/31 ≈ 1.097)!

The following is one of the most useful tools we have to quickly find the limit of a sequence.
Using what you know about functions. If we can find a formula for an , then finding the limit lim an
n→∞
n + ln n
is the same as finding lim f (x) for some function f . For example, to find lim , we can just use,
x→∞ n→∞ n2
x + ln x
say, L’Hôpital’s rule on lim to see the answer is zero. (Use L’Hôpital’s rule for sequences of
x→∞ x2
the form 0 · ∞ and always rewrite as a function of x)

52
Application. You may have learned that 1/9 = .1111111 . . .. What this means is that 1/9 is the limit of
a sequence that converges. Consider the sequence
1 11 111 1111
.1, .11, .111, .1111, . . . or , , , ,...
10 100 1000 10000
This can be written
1 + 10 + · + 10n−1
an = .
10n
We need a better formula. Notice that
10n − 1
(10−1)(1+10+·+10n−1 ) = 10+100+·+10n −(1+10+·+10n−1 ) = 10n −1 ⇒ 1+10+·+10n−1 = .
9
So
10n − 1 1 − 10−n
an = = .
9 · 10n 9
1 − 10−x 1
Since we know that lim = , it follows that an → 1/9.
x→∞ 9 9
Geometric sequences. If r ≥ 0, and c > 0, we call an = crn a geometric sequence. It is exactly like an
exponential function, except it is a sequence.

 0 0≤r<1
lim crn = c r=1
n→∞
∞ r>1


1  π n π
For example, n → 0, 2(1)n → 2, and diverges since > 1.
2 3 3
Squeeze Theorem If bn ≤ an ≤ cn and bn and cn both have the same limit, then an has the same limit.
 n
1
For example, intuitively, one would think that the sequence defined by an = − , d > 1 should converge
d
to 0, but how would you show it? We cannot use the function method as the negative would cause roots
of negative numbers (thus complex numbers) to appear! This is not a geometric sequence as r is negative.
Applying the squeeze theorem is our best choice. Notice that since the sequence alternates signs, the limit
(if it exists) will be somewhere between the sequences of all positive or negative signs, that is, for all n, an
agrees with either the positive or negative sequence’s nth term,
 n  x
1 1 1 1 1 1
− n ≤ − ≤ n ⇒ lim − x ≤ lim − ≤ lim x .
d d d x→∞ d x→∞ d x→∞ d
1
Now we can apply that lim = 0 on both sides (just an exponential function!) to conclude that an → 0.
x→∞ dx
Bounded and monotonic sequences. To say a sequence is bounded above or below just means it never
gets larger (or smaller) than a certain number. A monotonic sequence is one that either never decreases or
never increases.
1
For example, an = n is bounded and monotonic. It never gets below 0 nor above 1, and it never increases.
Every bounded monotonic sequence converges. To see this for a non-decreasing sequence, just take the
smallest possible upper bound; that is the limit. For a non-increasing sequence just take the largest possible
lower bound.
√ √
Example.
√ √an = n + 1 − n is0 decreasing and bounded below. To see this, consider the function f (x) =
x + 1 − x and notice that f (x) = (x + 1)−1/2 − x−1/2 < 0 for all positive x, hence the function, and
the sequence, is decreasing because the derivative
√ is
√ negative. In fact, its lower bound is zero, which we see
by using conjugates to rewrite an = 1/( n + 1 + n) and then considering the limit of the corresponding
function.

53
23 Lecture
Outline:
1. 10.2 Summing an Infinite Series; Part I

23.1 Series
∞
X n
X
A series is simply an infinite sum: ak = lim ak . You can think of them as having been obtained by
n→∞
k=1 k=1
adding up all of the terms in a sequence of numbers. For example:
1 1 1
1− + − + ···
2 3 4
is a series. We could write it as
∞
X 1
ak , ak = (−1)k−1 .
k
k=1

The question then becomes how do we make sense of adding infinitely many terms? The idea is to examine
finite sums of terms at the beginning of the series and see how they behave. A series is said to converge to
Xn
a number L if for any  > 0, there exists an N such that if n ≥ N then ak − L ≤ .
k=1

Note that whether or not the series converges might depend on how we label the terms! For instance,
1 − 1 + 1 − 1 + . . . alternates between one and zero if we start adding from left to right so it will not converge,
but if we group every two terms, it is identically 0 and hence converges to 0. Therefore, be careful to not
rearrange terms (add the first, then second, etc.) as grouping may change the series itself!
We can think of a series in a couple of different way: either as an improper integral or as a sequence of partial
sums.

23.2 Telescoping series

To illustrate how to use partial sums, consider the series
∞
1 1 1 1 X 1
+ + + + ··· = .
2 6 12 20 k(k + 1)
k=1

When we compute a few partial sums, Sn , we notice a pattern:

1 1 1 2 1 1 1 3 1 1 1 1 4
S1 = , S2 = + = , S3 = + + = , S4 = + + + = ,...
2 2 6 3 2 6 12 4 2 6 12 20 5
We can use partial fractions to investigate. We discover
1 1 1
= − .
k(k + 1) k k+1
This means the terms telescope, or, all the middle terms cancel leaving only the first and last! We have for
the nth partial sum,
n n          
X 1 X 1 1 1 1 1 1 1 1 1 1
Sn = = − = 1− + − + − +· · ·+ − = 1− .
k(k + 1) k k+1 2 2 3 3 4 n n+1 n+1
k=1 k=1
P∞ 1
Let n → ∞. Then we get k=1 k(k+1) = 1.

54
23.3 Linearity of Infinite Series
The following theorem shows that infinite series may be added or subtracted like ordinary sums, provided
that the series converge.
X X X X
Theorem (Linearity of Infinite Series) If ak and bk both converge, then (ak ± bk ) and cak
converge, where c is any constant. Furthermore
X X X
(ak ± bk ) = ak ± bk ,
X X
cak = c ak .

This is really helpful as it allows us to possibly rewrite a sum considered into sums that we already understand.
Remark. The fact that the individual series must converge is reminiscent of linearity of limits of functions.

23.4 Zeno of Elea’s paradox: Achilles and the tortoise

One of the first infinite series to be studied is called the geometric series (which we will discuss in detail next
lecture). It goes back to Zeno of Elea (c. 490–430 BC) and his paradox involving a race between Achilles
and a tortoise.
Achilles is in a footrace with the tortoise. Since Achilles is the fastest runner, the tortoise is allowed a
head start of some distance, say 1 kilometer. Suppose that each racer will run at some constant speed,
with Achilles being twice as fast. After some finite time, Achilles will have run 1 km, bringing him to the
tortoise’s starting point. During this time, the tortoise has run a much shorter distance of .5 km. It will
then take Achilles some further time to run that distance of .5 km, by which time the tortoise will have
advanced .25 km farther; and then more time still to reach this third point, while the tortoise moves ahead.
Thus, whenever Achilles arrives somewhere the tortoise has been, he still has some distance to go before he
can even reach the tortoise. After n iterations of this, Achilles has run
1 1 1
1+ + + · · · + n−1 km,
2 4 2
where as the tortoise has run
1 1 1 1
1+ + + · · · + n−1 + n km.
2 4 2 2
1
Taking the difference of these two finite sequences, one easily concludes that the tortoise will be km
2n
ahead of Achilles after n iterations. But then how can Achilles ever catch the tortoise (which we know
should happen...)? Mathematically speaking, this argument is considering the infinite geometric sum (which
is the same for both)
1 1
1 + + + ...
2 4
We will see that this sum adds to the finite number 2, so that both Achilles and the tortoise will be at the
2 km mark at the same time! Furthermore, we know this will happen in a finite amount of time, so there
is no conundrum after all. The idea that Zeno was grappling with was the idea of the limits of the partial
sums, and understanding that both series are the same, something that would take until the 19th century
to fully understand.

55
24 Lecture
Outline:
1. 10.2 Summing an Infinite Series; Part II

24.1 Geometric series

P∞
A geometric series has the form k=0 crk for some r, c 6= 0. (It is not a geometric sequence.) To find the
sum, we use a standard (and important) trick. We want to find
n
X
crk = c(1 + r + r2 + · · · + rn ).
k=0

If we multiply 1 − r we get a telescoping sum:

c(1 − r)(1 + r + r2 + · · · + rn ) = c(1 − r + r − r2 + r2 − r3 + · · · − rn + rn − rn+1 ) = c(1 − rn+1 ).

Therefore, as long as r 6= 1, we get

c(1 − rn+1 )
c(1 + r + r2 + · · · + rn ) = .
1−r
Of course if r = 1 then we get c(1 + r + r2 + · · · + rn ) = c(n + 1).
Now we let n → ∞. If |r| < 1 we get
∞
X c(1 − rn+1 ) c
crk = lim = .
n→∞ 1−r 1−r
k=0

If |r| > 1 we get the same limit, but it diverges (goes to infinity) since the numerator would grow without
bound. If r = ±1 the series also goes to infinity since it would grow like n.
If you start further out than at k = 0, then we can adjust by factoring:
∞ ∞
X X crm
crk = rm crk = .
1−r
k=m k=0

Example. Suppose I agree to pay you a dollar today, fifty cents tomorrow, and every day for the rest of
eternity I pay you half of what I paid you the day before. How much money will you earn?
Answer: two dollars. To see this, just recognize this as a geometric series (the same as Zeno’s paradox
actually!):
∞
X 1 1
= 1 = 2.
2k 1 − 2
k=0

Application: repeating decimals. Let us once again consider 1/9 = .111111... What does this repeating
decimal mean as a sum? It means
∞
X 1
.111111... = .
10k
k=1

When we compute the sum of this geometric series, we get 1/9 as expected. The same applies to any
repeating decimal.

56
Cantor’s disappearing table. Imagine a table 2 meters long. Let’s remove a quarter of it from the middle.
Each of the two remaining pieces is now less than 1 meter long (half the original). Now from each piece let’s
remove 1/16th, but remember there are two pieces so we really remove 1/8th of the remaining total length!
Each of the remaining four pieces is now less than 1/2 meter long (one fourth the original). Keep going in
like manner:
Step Number of Pieces left Size of each piece ≤ How much taken out
1 1 2 1/4
2 2 1 1/8
3 4 1/2 1/16
4 8 1/4 1/32
.. .. .. ..
. . . .
n 2n−1 2(1/2)n−1 1/2n .
By the end, the size of each piece goes to zero, since 2(1/2)n−1 → 0. But how much of the table was taken
out in total? For that we find the sum:
1 ∞
1 1 1 X 1
4 1
+ + + ··· = = 1 = .
4 8 16 2k 1 − 2
2
k=2

So only half of the table is gone! But there is not a single piece left with nonzero length! Where did it all
go.......?

24.2 The nth term divergence test

P∞
Suppose k=1 ak converges to L. Look at the nth term an . Then
n
X n−1
X
an = ak − ak → L − L = 0.
k=1 k=1

So the terms must go to zero. Therefore, if an doesn’t go to zero, it follows that the sum diverges.
P∞
In some cases this is really obvious. For example, the geometric series k=0 2k diverges because all the
terms are bigger than 1, so summing them all up clearly gives infinity. In other cases it’s somewhat less
obvious.
Example. The series
∞
1 2 3 4 X k
− + − + ··· = (−1)k−1
2 3 4 5 k+1
k=1

diverges, because lim an does not exist, hence the terms do not go to zero.
n→∞

On the other hand, if we group every two terms and then take the sum, the series does converge! The
problem here comes from the oscillation.
P∞
The nth term divergence test is not very powerful. Consider the series k=1 √1k . Then the nth term
goes to zero, but the series diverges, so the test was not very helpful. To see that the series diverges, look at
partial sums:
n
X 1 1 1 1 1 √
Sn = √ = √ + √ + · · · + √ ≥ n · √ = n → ∞.
k 1 2 n n
k=1

57
25 Lecture
Outline:
1. 10.3 Convergence of Series with Positive Terms; Part I

25.1 Series with positive terms

If a1 , a2 , a3 , . . . are all positive, then there are only two possibilities for their sum:
Pn P∞
1. k=1 ak is bounded above; in this case k=1 ak converges.
Pn
2. k=1 ak is unbounded; in this case the sum is infinity.

Note that if the series converges, then an → 0.

We are going to learn some different kinds of comparison tests to see whether a series converges.

25.2 Integral test

We can think of a series as an improper integral, so the comparison principle applies:
R∞
Suppose an = f (n) where f is positive, decreasing, and continuous function of x for x ≥ 1. If 1
f (x)dx
P∞ R∞ P∞
converges, then so does k=1 ak . Conversely, if 1 f (x)dx diverges, then so does k=1 ak .
The requirement that f be decreasing is only so that we know an ≥ f (x) ≥ an+1 for n ≤ x ≤ n + 1.
P∞ R∞
Harmonic series. The series 1 + 21 + 13 + · · · = k=1 k1 diverges, because 1 x1 dx diverges.
P∞
p-series. The series k=1 k1p converges when p > 1 and diverges when p ≤ 1. To see this when p > 0 just
use the integral test. When p ≤ 0 this is obvious, since the terms don’t even go to zero.
P∞
Example. Show that k=2 k(ln1k)2 converges. Use the integral test with f (x) = x(ln1x)2 . Note that f is
R∞ R
decreasing because x and ln x are both increasing. Since 2 f (x)dx = limR→∞ − ln1x 2 = limR→∞ ln12 −
1 1
ln R = ln 2 , we conclude that the sum converges.

25.3 Direct comparison test

Recalling how we compared improper integrals to see if they converged, we see the same applies for series:
P∞ P∞
If 0 ≤ an ≤ Pbn for large n and k=1 bk converges, then so does k=1 ak . Therefore the contrapositive is
∞ P∞
also true: if k=1 ak diverges, then so does k=1 bk .
P∞ 1
Example. Show that k=1 √k3 k
converges. The idea is to look at something larger that we know converges,
P∞
hence since √k3k ≤ 3k use the convergent geometric series k=1 31k as a comparison.
1 1

P∞ 1
Example. Show that k=1 (k2 +1) 1/3 diverges. To see this, we note that

1 1
(k 2 + 1)1/3 ≤ (k 2 + 2k + 1)1/3 = (k + 1)2/3 ⇒ ≥ .
(k 2 + 1)1/3 (k + 1)2/3
P∞ 1
P∞ 1
P∞ 1
But k=1 (k+1)2/3 = j=2 j 2/3 diverges by the p-series test. So by comparison, k=1 (k2 +1) 1/3 diverges.

Change of variables. The trick I just used is often quite useful: you can start Pthe series from P
any place
∞ 1 ∞ 1
you want, and it may help you in your calculations. For example, to see that k=1 (k+1) 2/3 = j=2 j 2/3
just let j = k + 1 and notice that we start counting j at j = 2. However, the only kind of change of variables
that will work is j = k + m, where m is some integer. (It’s not like integration...)

58
26 Lecture
Outline:
1. 10.3 Convergence of Series with Positive Terms; Part II

26.1 Limit comparison test

Sometimes we don’tP∞have 0 ≤ an ≤ bn , but Pbn is “really close” to an , and we want to be able to conclude
∞
something about k=1 ak because we know k=1 bk . For that we have the following:
an
Theorem 3. Suppose L = lim .
n→∞ bn
P∞ P∞
1. If 0 < L < ∞, then k=1 ak converges if and only if k=1 bk converges.
P∞ P∞
2. If L = 0 and k=1 bk converges, then so does k=1 ak .
P∞ P∞
3. If L = ∞ and k=1 ak converges, then so does k=1 bk .
k2
P∞ 1
Example. Show that k=2 k4 −k−1 converges. When we reduce the fraction we get k2 −1/k−1/k 2 which is
1
approximately k2 when k is large. In other words, let us compare our series terms with the simpler series
terms k12 . Then

n2 1 an 1
an = , bn = 2 ⇒ = → 1, n → ∞.
n4 − n − 1 n bn 1 − 1/n3 − 1/n4
k2
P∞ P∞
So the limit comparison test applies. Since k=1 k12 converges, so does k=2 k4 −k−1 .
P∞
Example. Show that k=3 √k12 +4 diverges. When k is large √k12 +4 is approximately k1 . Indeed,

1 1 an n 1
an = √ , bn = ⇒ =√ =p → 1.
n2 +4 n bn 2
n +4 1 + 4/n2
P∞ 1
P∞
Now since diverges, so does √ 1 .
k=1 k k=3 k2 +4

Making sense of limit comparisons. When you use this test, you need to look for what’s important in
a sequence of terms. For example, when we look at

n2
,
n4 − n − 1
we need to see that what’s important is
n2 1
= 2.
n4 n
This gives us the terms we want to compare.
The key is to know which terms dominate as n gets large, just like we did when comparing functions
asymptotically. Here’s a hierarchy starting with the fastest growing:
1. n!
2. epn (p > 0, the bigger p is, the higher up it is)
3. np (p > 0, the bigger p is, the higher up it is)
4. ln n
5. sin n, cos n, constants

59
The strategy: look at the dominant term on the top and bottom of your fraction (if the term is a fraction,
which it probably is). Then just forget the rest.
Example. If you want to compare the term

e2n + n + ln n + en
,
n2 − en
e2n
then you just need −en = −en , because the rest is lower on the hierarchy.

60
27 Lecture
Outline:
1. 10.4 Absolute and Conditional Convergence

27.1 Absolute versus conditional convergence

P∞ P∞
If k=1 |ak | converges, then the series k=1 ak also converges, and we say it converges absolutely.
For example, consider the series
∞
1 1 1 X 1
1− + − + ··· = (−1)k−1 2 .
4 9 16 k
k=1
P∞ P∞
This converges absolutely by the p-series test by considering k=1 (−1)k−1 k12 = k=1 k12 . What this is
really telling us here is that any cancellation occurring from the alternating plus/minus is not important for
the convergence of the series!
P∞ P∞
On the other hand, sometimes k=1 ak converges even though k=1 |ak | does not. This is called condi-
tional convergence. The reason is that if you change the order of the terms, you can change the limit of
the sum, whereas for an absolutely converging series you cannot.
For example, it turns out that
∞
1 1 1 X 1
+ − + ··· =
1− (−1)k−1
2 3 4 k
k=1
P∞
converges conditionally; it does not converge absolutely because k=1 k1 is harmonic, but it does converge
because it is alternating. Hence the cancellation is now the only reasoning the series converges!

27.2 Alternating series test

Suppose an is a sequence of decreasing positive numbers such that limn→∞ an = 0. Then the series
P ∞ k−1
k=1 (−1) ak = a1 − a2 + a3 − a4 + · · · converges.
P∞
Example. Let 0 < p ≤ 1. Then the series k=1 (−1)k−1 k1p converges by the alternating series test. It only
converges conditionally, however, by the p-series test.
Remark. See Exercise 35 to see the assumption that an is a decreasing sequence is necessary.

27.3 Examples
∞
X (−1)k k 4
1. . This diverges because the terms grow like k, thus do not go to zero but rather to ±∞.
k3 + 1
k=1
∞
X sin((k + 21 )π) P∞ 1
2. 2
. This converges absolutely because k=1 k2 converges.
k
k=1
∞
X (−1)k
3. 1 . This diverges because the terms do not go to zero but rather to ±1.
k=1
1+ k
∞
X (−1)k k n
4. . This converges by the alternating series test since n2 +1 is a sequence of decreasing positive
k2 + 1
k=1 P∞ k
numbers tending to 0. The convergence is only conditional, however; the series k=1 k2 +1 diverges by
the limit comparison test with the harmonic series which diverges.

61
28 Lecture
Outline:
1. 10.5 Ratio and Root Tests and Strategies for Choosing Tests; Part I
2. Ratio Test
3. Root Test

28.1 Ratio test

P∞ k
Does k=1 2k! converge? We can’t do an integral comparison, because of the factorial. What other kinds of
comparison can we make? Well, the series is not geometric, because the ratio between terms would have to
be constant. Recall that
X∞
rk
k=0
converges precisely when |r| < 1. It turns out this makes a good test even when the series isn’t precisely
geometric.

Theorem 4 (Ratio test). Let r = limn→∞ aan+1 n

.
P
1. If r < 1, then ak converges absolutely.
P
2. If r > 1, then ak diverges.
P
3. If r = 1, then ak may or may not converge.
P∞ 2k
Example. k=1 k! converges because

an+1 2n+1 n! 2
lim = lim = lim = 0.
n→∞ an n→∞ (n + 1)! 2n n→∞ n + 1

This is the smallest ratio you can get.

P∞ k3
Example. k=1 2k converges because

an+1 (n + 1)3 2n (n + 1)3 1

lim = lim n+1 3
= lim 3
= < 1.
n→∞ an n→∞ 2 n n→∞ 2n 2
P∞ P∞
Often inconclusive. Look at k=1 k12 and k=1 k 2 . The first converges by the p-series test, while the
second diverges (by common sense...or the p-series test with p = −2). But in either case, the ratio between
terms goes to 1:
1/(n + 1)2 n2 (n + 1)2
lim = lim = 1, lim = 1.
n→∞ 1/n2 n→∞ (n + 1)2 n→∞ n2
So when r = 1, you don’t learn anything.
P∞ 1
One more example. Can we learn anything about k=2 k ln k from the ratio test? No, because
an+1 n ln n
lim = lim = 1.
n→∞ an n→∞ (n + 1) ln(n + 1)

This can be seen by using L’Hôpital’s Rule twice, for example,

1
x ln x ln x + 1 x
lim = lim = lim 1 =1
x→∞ (x + 1) ln(x + 1) x→∞ ln(x + 1) + 1 x→∞
x+1
R∞ 1
So does it converge? As a matter of fact, it diverges by the integral comparison test: since 2 x ln x dx
P∞ 1
diverges, so does the series k=2 k ln k.

62
28.2 Root test
Another way to see if a series is close to a convergent geometric series is to take the nth root of an .
Theorem 5 (Root test). Let r = limn→∞ |an |1/n .
P
1. If r < 1, then ak converges absolutely.
P
2. If r > 1, then ak diverges.
P
3. If r = 1, then ak may or may not converge.
P∞  k k n 1
Example. k=1 2k+3 converges because |an |1/n = 2n+3 → 2 < 1.
P∞
2
1 −k
Example. Here’s a pretty cool example: k=1 1+ k converges because
 −n
1
|an |1/n = 1+ → e−1 < 1.
n

63
29 Lecture
Outline:
1. 10.5 Ratio and Root Tests and Strategies for Choosing Tests; Part II
2. So many tests, but how to choose?

29.1 Strategies for choosing tests

∞
X
Let ak be given. Keep in mind the series for which convergence or divergence is already known include
k=1
geometric series and p-series, so these will be very useful in comparison.
1. The nth Term Divergence Test.
Always check this test first as it could give an easy conclusion. If lim an 6= 0, then the series diverges and
n→∞
we are done! But if lim an = 0, we cannot conclude anything...move on to the next step.
n→∞

2. Positive Series. If all terms in the series are positive, we have the following tests:
(a) The Direct Comparison Test.
Consider whether dropping terms in the numerator or denominator gives a series that we already know
convergence or divergence for. If a large series converges, or a smaller series diverges, then the original series
does the same.
∞ ∞
X 1 X 1
For example, √ converges by comparison with the convergent p-series .
2
k + k k2
k=1 k=1
∞
X 1
However, if we try the same strategy on √ , we cannot conclude anything from this test. Maybe
k=2
k2
− k
the Limit Comparison Test will be better suited.
(b) The Limit Comparison Test.
Consider the dominant term in the numerator and denominator, and compare the original series to the ratio
of these terms.
∞
X 1 √ 1
For example, look again at √ . Since k 2 grows faster than k, we compare with 2 to conclude
2
k − k k
k=2
the series converges as
1√
k2 − k
lim 1 =1
k→∞
k2
∞
X 1
and is once again a convergent p-series.
k2
k=1

(c) The Ratio Test.

This is often useful when there is a factorial or a constant to the power n present, since the factorial or power
will disappear in the ratio.

64
 ∞
X 3k
For example, for , applying the Ratio Test yields
k!
k=1

3n+1
(n+1)! 3
lim 3n = lim = 0 < 1,
n→∞
n!
n→∞ n+1

hence the series converges.

(d) The Root Test.
This is often effective when there is a term of the form f (n)g(n) , that is, the base and power is some function
of n.
∞
X 2k
For example, consider . The Root Test yields
k 2k
k=1

1/n
2n

2
lim = lim = 0 < 1,
n→∞ n2n n→∞ n2

hence the series converges.

(e) The Integral Test.
When all else fails for a positive, decreasing series, consider the Integral Test by looking at what happens to
the integral of the corresponding function.
∞
X 1 1
For example, consider . While all the other tests do not apply easily, the function is decreasing
k ln k x ln x
Z ∞ k=2
dx
and = ∞. Thus the integral, and hence the series, diverges.
2 x ln x
3. Series That Are Not Positive Series.
(a) Alternating Series Test.
If we have an alternating series, show the positive sequence is decreasing and tends to 0. Then the Alternating
Series Test shows the series converges, albeit, this only allows one to conclude conditionally.
(b) Absolute Convergence.
X X
If |ak | (which is now a positive series so we have many test for) converges, then ak converges absolutely,
and therefore is convergent.

Helpful Table
The following is a table that describes the different series tests we have discussed, convergence and divergence
criteria, usefulness, and are given in a general good order to look at them if you are not sure what to apply.

65
Test Series Converges if Diverges if Useful for
Geometric Series Sum can start at k =
∞ m
X |r| ≥ 1 m. Converges to cr
crk |r| < 1 to
c 1−r

k=0 1−r for |r| < 1 by factor-

ing.
p-series Test Work the same as
∞
X 1 p>1 p≤1 improper p-integrals
kp did.
k=1

nth Term Diver- When terms do not

∞
gence Test X n/a lim an 6= 0 tend to zero.
ak n→∞
k=1

Positive Series All positive terms.

Direct Compari- Comparing to
∞
son Test X a
P∞ known series
ak 0 ≤ an ≤ bn and 0 ≤ bn ≤ an and
k=1 k=1 bk .
∞
X ∞
X
bk converges bk diverges
k=1 k=1

Limit Compari- Dominant term

∞
son Test X
an an gives a known series
ak lim ≥ 0 and lim > 0 and P∞
k=1 n→∞ bn n→∞ bn k=1 bk .

∞
X ∞
X
bk converges bk diverges
k=1 k=1

Ratio Test No info when equal

∞
X to one. Useful for
ak an+1 an+1
k=1
lim <1 lim >1 factorials or constant
n→∞ an n→∞ an raised to k.

Root Test No info when equal

∞
X to one. Useful for
ak lim |an |1/n < 1 lim |an |1/n > 1
k=1 n→∞ n→∞ terms of the form
f (n)g(n) .
Integral Test (let Useful for positive,
∞
f (n) = an be pos- X Z ∞ Z ∞ decreasing series
ak
itive and decreas- f (x) dx < ∞ f (x) dx = ∞ where integral can
k=1
ing) 1 1 be easily understood.

Not Positive Se-

ries
Absolute Conver- Can use all the above
∞
gence X
∞ n/a tests on the series
ak X
k=1 |ak | converges with positive terms.
k=1 No conclusion if pos-
itive series diverges.
Alternating Determining conver-
Series Test ∞ lim an 6= 0 gence of series that
n→∞
X lim an = 0, has cancellation from
(−1)k−1 ak n→∞
k=1 positive and negative
an is positive and de- terms occurring.
creasing

66
30 Lecture
Outline:
1. 10.6 Power Series; Part I

30.1 Power series

A power series is basically an infinitely long polynomial. So it looks like
∞
X
a0 + a1 x + a2 x2 + a3 x3 + a4 x4 + · · · = ak xk
k=0

where x is a variable.
Now, a regular polynomial is nice and continuous (and infinitely differentiable) everywhere. But a power series might
not even converge. How do we know it’s even defined?
Example. Consider the power series
∞
X
1 + x + x2 + x3 + x4 + · · · = xk .
k=0

For which x does this converge?

Well, it’s a geometric series. So
∞ n
X X 1 − xn+1
xk = lim xk = lim ,
n→∞ n→∞ 1−x
k=0 k=0

which converges when |x| < 1 but diverges when |x| ≥ 1. (The Ratio Test does not apply when |x| = 1, so to check
the case when x = 1, we just get limn→∞ n = ∞. Note that when x = −1 we get oscillating divergence rather than
divergence to infinity.)
In fact, when |x| < 1 we have
∞
X 1
xk = .
1−x
k=0
This is our first hint that a power series can represent some well-known function–which is not a polynomial!
Radius of convergence. In the last example, the series converged for |x| < 1 but diverged for |x| ≥ 1. The number
1 is called the radius of convergence. It turns out every power series has a radius of convergence, which could be
infinity or even zero (so the power series would only converge at one point).
Example. Find the radius of convergence of the power series
∞
x2 x3 x4 X xk
1+x+ + + + ··· = .
2 6 24 k!
k=0

Let’s use the ratio test.

xn+1 /(n + 1)! |x|
r = lim = lim =0
n→∞ xn /n! n→∞ n + 1

no matter what x is! Therefore the series always converges. The radius of convergence is ∞.
By the way, it turns out that
∞
X xk
= ex , for all x,
k!
k=0
another well-known function.
Example. Find the radius of convergence of the power series
∞
X
1 + 2x + 3x2 + 4x3 + 5x4 + · · · = (k + 1)xk .
k=0

67
Again, use the ratio test:
(n + 2)xn+1 (n + 2)|x|
r = lim = lim = |x|.
n→∞ (n + 1)xn n→∞ (n + 1)

If |x| < 1 the series converges, if |x| ≥ 1 the series diverges. The radius of convergence is 1.
Different center.
If a power series has the form
∞
X
a0 + a1 x + a2 x2 + a3 x3 + · · · = ak xk
k=0

then it is centered at zero. When we plug in x = 0, we just get one term. So the series always converges there
(but possibly nowhere else!).
If we want a series centered around another point, we just shift:
∞
X
a0 + a1 (x − c) + a2 (x − c)2 + a3 (x − c)3 + · · · = ak (x − c)k .
k=0

Now c is the center. This will not change the radius of convergence; it will only shift where the series converges.
Endpoints. If x is within the radius of convergence, then the series converges, and if it’s outside then the series
diverges, but what about the endpoints? You have to check those by plugging in each value for x and considering
both series separately. Don’t worry, it’s only two points!
Example. Find where the series ∞ k k
P
k=0 2 (x − 3) converges. Use the ratio test:

2n+1 |x − 3|n+1
r = lim = 2|x − 3|.
n→∞ 2n |x − 3|n
1
So we get convergence of |x − 3| < 2
and divergence if |x − 3| > 12 . The radius of convergence is still 1/2, but centered
at x = 3.
What
P∞ about the endpoints?
P∞ They are x = 3 ± 1/2, that is, 7/2P and 5/2. If we lookPat x = 7/2 we get the series
k k ∞ k k ∞ k
k=0 2 (1/2) P = k=0 1 = ∞. If we look at x = 5/2 we get k=0 2 (−1/2) = k=0 (−1) which diverges by
∞ k k
oscillating. So k=0 2 (x − 3) converges when 5/2 < x < 7/2 but diverges everywhere else. We also sometimes say
it’s interval of convergence is (5/2, 7/2) since this is more specific than the radius of convergence by specifying
the behavior at the end points.
Remark. Sometimes the endpoints both converge (closed interval of convergence [a, b]), both diverge (open interval
of convergence (a, b)), or have different behavior at each one (half-open interval of convergence (a, b] or [a, b)). It is
often difficult to intuitively see what will happen, so just check each one!

68
31 Lecture
Outline:
1. 10.6 Power Series; Part II

31.1 Power series that converge to functions

We can make substitutions (change of variables) to get power series formulas for other functions using
∞
1 X k
= x , |x| < 1.
1−x
k=0

Example. Change x to −x2 . Since | − x2 | < 1 is the same as |x| < 1, we get
∞ ∞
1 X X
= (−x2 )k = (−1)k x2k , |x| < 1.
1 + x2
k=0 k=0
1
Example. Expand in a power series centered at c = 0. We have
2 + 3x
1 1 1 1 1
= = ,
2 + 3x 2 1 + (3/2)x 2 1 − (−3/2)x
so we can simply use the substitution x 7→ (−3/2)x and multiply the entire series by 1/2 to find
∞
1 1X
= (−1)k (3/2)k xk , |x| < 2/3.
2 + 3x 2
k=0
Term by term integration and differentiation. If a power series centered at c converges in a certain radius R,
then you can take its derivative by differentiating each term (just like with polynomials) for x ∈ (c − R, c + R). The
same holds for integration. The radius of convergence remains the same.
1 1
Example. How would we find a power series for (1−x)2
? Differentiating 1−x
we get
  ∞
1 d 1 d X
= = (1 + x + x2 + x3 + x4 + · · · ) = 1 + 2x + 3x2 + 4x3 + · · · = (k + 1)xk , |x| < 1.
(1 − x)2 dx 1−x dx
k=0

What about for ln(1 − x)? Integrating, we get

∞
x2 x3 x4 X xk
− ln(1 − x) = x + + + + ··· = , |x| < 1.
2 3 4 k
k=1
Application. One way to try an solve a differential equation is to use a power series, and solve for the coefficients.
This application is only for enrichment; take a differential equations course if you want to know more.
Let’s try solving y 0 = y, y(0) = 1. Remember we have already seen the solution for this is y = ex . However, let’s
solve this differential equation by letting y be a power series:
∞
X
y = a0 + a1 x + a2 x2 + · · · = ak xk .
k=0

Now we differentiate to get

y 0 = a1 + 2a2 x + 3a3 x2 + · · · .
0
Since y = y we should have
a0 + a1 x + a2 x2 + a3 x3 + · · · = a1 + 2a2 x + 3a3 x2 + 4a4 x3 · · · .
1 1
We notice a pattern: an = a
n n−1
. But that means an = a .
n! 0
Note that y(0) = a0 , so by the initial condition we
get a0 = 1. Thus,
∞
X xk
y= .
k!
k=0
This is the formula for ex as a power series. We saw earlier that it converges everywhere.

69
32 Lecture
Outline:
1. 10.7 Taylor Polynomials
2. 10.8 Taylor Series; Part I

32.1 Taylor series

Every function with infinitely many derivatives has, in theory, a power series representation. It just might not
converge. Whether or not it converges, we call it the Taylor series associated with that function.
We pick a center x = c. Write
∞
X
f (x) = ak (x − c)k = a0 + a1 (x − c) + a2 (x − c)2 + a3 (x − c)3 + · · · .
k=0

Let’s assume the series converges. We want to solve for a0 , a1 , a2 , a3 , . . . in terms of f and its derivatives.
Plug in x = c, get f (c) = a0 .
Now take the derivative:
f 0 (x) = a1 + 2a2 (x − c) + 3a3 (x − c)2 + 4a4 (x − c)3 + · · · .
So f 0 (c) = a1 .
Take the second derivative:
f 00 (x) = 2a2 + 6a3 (x − c) + 12a4 (x − c)2 + 20a5 (x − c)3 + · · · .
So f 00 (c) = 2a2 .
Take the third derivative:
f 000 (x) = 6a3 + 24a4 (x − c) + 60a5 (x − c)2 + 6 · 5 · 4a6 (x − c)3 + · · · .
So f 000 (c) = 6a3 .
We continue the pattern to get
f (n) (c)
f (n) (c) = n!an ⇒ an = .
n!
In other words, the Taylor series of f centered at c is
∞
X f (k) (c)
(x − c)k .
k!
k=0

If the Taylor series converges with radius r > 0, we say f is analytic on the interval (c − r, c + r). Note that the
Taylor series must converge to f (x) if it converges at all!
If the center is c = 0, sometimes we call the Taylor series a Maclaurin series.

32.2 Taylor Polynomials

The partial sums of a Taylor series are known as Taylor Polynomials. They are a very useful tool for approximating
functions so that you consider only finitely many terms as was seen in the first lecture. We define the nth Taylor
polynomial, Tn , of f (x) centered at x = c as the nth partial sum
n
X f (k) (c)
Tn (x) = (x − c)k .
k!
k=0

There are lots of theorems on how good these approximations are including error bounds, remainders, and Taylor’s
theorem, however, we will not go into details here. Similar to series, a Maclaurin polynomial is simply a Taylor
polynomial with center c = 0.

70
32.3 Examples.
To find a Taylor series (or Taylor polynomial) you need to compute lots of derivatives. Sometimes that is not a
problem.
Take ex . All of its derivatives are also ex . If we center at c, the series is
∞
X ec
ex = (x − c)k .
k!
k=0

By the ratio test, this converges everywhere. So ex is analytic on the whole real line. The usual center is just c = 0
so we can write
∞
X xk
ex = , x ∈ (−∞, ∞),
k!
k=0

which we have seen before.

Take sin x. Its derivatives follow a repeating pattern: sin x, cos x, − sin x, − cos x, sin x, . . .. Let’s take the center to
be c = 0. Then the derivatives at this center are 0, 1, 0, −1, 0, . . .. So only the odd powers of x will remain, and the
terms will alternate:
∞
x x3 x5 x7 X x2k+1
sin x = − + − + ··· = (−1)k , x ∈ (−∞, ∞).
1 3! 5! 7! (2k + 1)!
k=0

For cos x we get

∞
x2 x4 x6 X x2k
cos x = 1 − + − + ··· = (−1)k , x ∈ (−∞, ∞).
2! 4! 6! (2k)!
k=0

By the ratio test, both of these converge everywhere, just like ex .

Take ln x. Now, there’s no way this is analytic everywhere, since it isn’t even defined everywhere. We can’t pick our
center at 0, because ln 0 doesn’t exist. Let’s pick our center to be c = 1 instead. The derivatives of ln x:
1 1 2 6 24
ln x, ,− 2, 3,− 4, 5,...
x x x x x
At the center c = 1:
0, 1, −1, 2, −6, 24, ...
n−1
which is (−1) (n − 1)! for n ≥ 1. We get
∞ ∞
X (x − 1)k X (1 − x)k
ln x = (−1)k−1 =− .
k k
k=1 k=1

Note that it starts at k = 1, not zero! What is the radius of convergence? Well, it can’t possibly be more than
only valid for x ∈ (0, 2). In
1, because ln(0) does not exist. By the ratio test, we find out it is exactly 1, so is P
1
fact, we have seen this series before, when we integrated the geometric series 1−x = ∞ k
k=0 x term by term to get
P∞ xk
− ln(1 − x) = k=1 k .
Taylor Polynomial Example. As an example for Taylor polynomials, let’s find T4 (x) centered at c = 2 for
f (x) = ln x. This is easy since we already have the derivatives we need above, but if not we would simply need to
find the first four derivatives. We find
f 00 (2) f 000 (2) f (4) (2)
T4 (x) = f (2) + f 0 (2)(x − 2) + (x − 2)2 + (x − 2)3 + (x − 2)4
2 3! 4!
1 1 1 1
= ln 2 + (x − 2) − (x − 2)2 + (x − 2)3 − (x − 2)4 .
2 8 24 64

71
33 Lecture
Outline:
1. 10.8 Taylor Series; Part II

33.1 Variations on these examples

3
Example. Find the Taylor series for x2 ex centered at zero. There is no reason to take all the derivatives; we just
use the series for ex and plug in x3 in place of x, then multiply by x2 :
∞ ∞
3 X (x3 )k X x3k+2
x2 ex = x2 = for all x.
k! k!
k=0 k=0

Example. Here’s a warning against forgetting things you know. Let’s try to find the Taylor series for ln(x2 ) centered
at x = 1. We get
∞
X (1 − x2 )k
ln(x2 ) = − , x ∈ (0, 2).
k
k=1

OK but wait a minute. That’s not actually a power series; we would have to multiply out (1 − x2 )k and regroup
terms. That looks hard!
But actually we know what we should get. The answer is that ln(x2 ) = 2 ln x. So
∞
X (1 − x)k
ln(x2 ) = 2 ln x = −2 , x ∈ (0, 2).
k
k=1

33.2 Application
Limits. I mentioned before that using series can sometimes simplify limits or asymptotics. Here’s an example limit
you might be tempted to calculate using L’Hôpital’s Rule:
x − sin x
lim .
x→0 x3 cos x
But you would need to differentiate twice, and that would get complicated. Instead, let’s use Taylor series to see that
3 5 3 5 2
x − sin x x − (x − x3! + x5! − · · · ) x
3!
− x5! + · · · 1
3!
− x5! + · · ·
3
= 2 4 = 5 7 = 2 4 .
x cos x x3 (1 − x2! + x4! + · · · ) x3 − x2! + x4! + · · · 1 − x2! + x4! + · · ·
Now we can take x → 0: 2
1
x − sin x 3!
− x5! + · · · 1 1
lim = lim 2 4 = = .
x→0 x3 cos x x→0 1 − x + x + · · · 3! 6
2! 4!

Integrals. Let’s return to the integral of sin x2 which was introduced in the first lecture of the course. By replacing
x by x2 in our power series formula for sin x we obtain
∞
X x4k+2
sin x2 = (−1)k for all x.
(2k + 1)!
k=0

Now integrating sin x2 is easy:

∞ ∞
(−1)k x4k+3
Z X Z X
sin x2 dx = x4k+2 dx = (−1)k + C for all x.
(2k + 1)! (2k + 1)!(4k + 3)
k=0 k=0

Certain Values of Derivatives. We can also use the power series of a function to quickly find the values of
derivatives at the center of the series. For example, let f (x) = sin x2 and find f (10) (0). By using the Taylor series
formula or taking 10 derivatives of the series for sin x2 we find f (10) (0) = 10!/5! = 30240.
Remark: The series here actually tells us that the only nonzero derivatives of sin x2 are 4k + 2, k = 0, 1, 2, . . .

72
 Z 1
Error Bound. We can also use alternating series to easily estimate definite integrals. For example, find sin x2 dx
0
within an error of 1/10000. From what we had before,
∞ 1 ∞
1
x4k+3 (−1)k
Z X X
sin x2 dx = (−1)k = .
0 (2k + 1)!(4k + 3) (2k + 1)!(4k + 3)
k=0 0 k=0

To easily estimate the integral, we simply need to find the first N such that the next term, aN +1 < 1/10000. N = 0, 1
do not work, but N = 2 does, so the integral is about 1/3 − 1/42 + 1/1320 = 2867/9240 (within 1/10000 of an error).
Special functions. Let us end our discussion on Taylor series by reminding ourselves about something else mentioned
in the first lecture, the so-called special functions. These are functions that appear in a wide range of physics and
engineering applications (such as Bessel, hypergeometric, and elliptic functions which are particularly interesting and
used to study pendulums), which will be discussed in most Differential Equations courses.
For example, when studying the vibrations of a drum, one is led to the following second order equation:

d2 y dy
x2 +x + (x2 − α2 )y = 0,
dx2 dx
where α is a nonnegative number. This equation is called the Bessel equation and its canonical solutions are the
so-called Bessel functions. The easiest way to describe a Bessel function is as a power series:
∞
X (−1)n  x 2n+α
Jα (x) = .
n=0
n! Γ(n + α + 1) 2

Here the the expression Γ(n + α + 1) is a value of the Gamma function defined as an improper integrals. In the
special case when α = 0, the value Γ(n + α + 1) is equal to Γ(n + 1) = n! and it follows that

x2 x4 x6
J0 (x) = 1 − + − + ··· .
4 64 2304
This looks very similar to the expansion for cos x, which is no coincidence. Bessel functions are oscillatory like sine
and cosine, however they decay as x grows large. To see this better, here is what the graph of J0 (x) (as well as J1 (x)
and J2 (x)) looks like:
y

f (x) = J0 (x)
J1 (x)

20 x

J2 (x)

-1

73
34 Lecture
Outline:
1. Complex Numbers

34.1 Complex Numbers

Let us begin by discussing the complex numbers, denoted C. The complex numbers are numbers of the form a + bi,
where a and b are any real number and i is defined to be the square root of −1, that is, i2 = −1. We can realize the
complex numbers in the plane by letting what is ordinarily labeled the x-axis correspond to the real part, a, and the
y-axis correspond to the imaginary part, b. We can add and multiply complex numbers:

(a + bi) + (c + di) = (a + c) + (b + d)i,

(a + bi)(c + di) = ac + adi + bci − bd = (ac − bd) + (ad + bc)i.

As a result, we can compute polynomial functions of a complex variable, usually denoted z,

f (z) = a0 + a1 z + a2 z 2 + · · · + an z n .

We can also measure the distance between two complex numbers as

p
d(a + bi, c + di) = (c − a)2 + (d − b)2 ,

which is exactly how we would measure the distance between two points in the regular plane (Cartesian coordinates)
using Pythagorean’s Theorem. This allows one to consider a series of complex numbers and its convergence. In this
way, the term radius of convergence now makes sense as it truly is looking at some circle in the complex plane in
which your series converges.
In particular, the series for the exponential function, sine function, and cosine function can immediately be extended
to the entire complex plane by simply replacing the real variable x with the complex variable z.

34.2 Euler’s Formula

This allows one to prove the following theorem, which is surprisingly easy to prove, but extremely useful in application.
Euler’s Formula. For all complex numbers z,

eiz = cos z + i sin z.

Proof. The key to the proof is noticing the pattern of powers of the complex unit i. Note

i0 = 1, i1 = i, i2 = −1, i3 = i2 i = −i, i4 = i2 i2 = (−1)(−1) = 1, . . . ,

and repeats from there. Now it remains to use the substitution z = iz in the series for ez , then group real and
imaginary parts. We find

(iz)2 (iz)3 (iz)4 (iz)5

eiz = 1 + iz + + + + + ...
2 3! 4! 5!
2 3 4 5
z z z z
= 1 + iz − −i + + i ...
2 3! 4! 5!
z2 z4 z3 z5
   
= 1− + + ... + i z − + + ...
2 4! 3! 5!
= cos z + i sin z.

74
Euler’s Formula is particularly useful in electrical engineering studying periodic signals.
If we substitute z = π into Euler’s Formula, we arrive at what is known as Euler’s Identity, which is probably
already familiar to many of you,
eiπ = cos π + i sin π = −1.
Rearranging yields
eiπ + 1 = 0.
While this is a useful identity for many reasons, if we take a step back it is quite an amazing equation as it relates
5 of the most important numbers in mathematics and its applications. Namely, and in no particular order, it relates
the mathematical constants 0, 1, e, π, and i. That is truly amazing if you really think about it!

34.3 Applications
If Euler’s Identity did not impress you, maybe seeing how easy trigonometric identities become using Euler’s Formula
might! It is fairly easy (and an extra credit problem) to show using Euler’s Formula that

eiz + e−iz eiz − e−iz

cos z = , sin z = ,
2 2i
which can now be used to find trig identities.
Let’s begin easy with Pythagorean’s Theorem, that is, sin2 x + cos2 x = 1. How hard is it to prove this using Euler’s
Formula? Well,
2 2
eiz − e−iz eiz + e−iz
 
1  2iz 
sin2 z + cos2 z = + = −e + 2 − e−2iz + e2iz + 2 + e−2iz = 1,
2i 2 4

so not bad at all!

Along these same lines, it is also easy to see
2 2
eiz + e−iz eiz − e−iz
 
1  2iz 
cos2 z − sin2 z = − = e + 2 + e−2iz + e2iz − 2 + e−2iz
2 2i 4

1  2iz  ei2z + e−i2z

2e + 2e−2iz =
= = cos(2z),
4 2
thus finding the double angle formula!
It is often much easier to find these formulas, or come up with new ones, using complex exponential functions as it
becomes a purely algebraic manipulation. Even the fact that when you take the derivative of cosine you introduce a
negative becomes clear by letting z = x, and differentiating since

d eix + e−ix ieix − ie−ix i ieix − ie−ix eix − e−ix

 
d
cos x = = = =− = − sin x.
dx dx 2 2 i 2 2i

75
Exam 3 Review
The third exam will cover sequences, series, and complex numbers:

Sequences [Section 10.1]

A sequence, often denoted {an }, is an infinite set of numbers placed in order:

a1 , a2 , a3 , . . .

Be able to find a formula for the nth term of a sequence as well as the limit of a sequence by considering the behavior
of an associated function (using L’Hôpital’s Rule if needed).

Series [Sections 10.2-10.5]

Be able to use the following techniques and tests in order to determine if a series converges or diverges. Do not
forget the table I provided that lists all of this on one page!
∞
X
Let ak be given. Keep in mind the series for which convergence or divergence is already known include geometric
k=1
series and p-series, so these will be very useful in comparison. If the series given is a geometric series, we even
have a formula for what the series converges to, if it converges.
1. The nth Term Divergence Test.
Always check this test first as it could give an easy conclusion. If lim an 6= 0, then the series diverges and we are
n→∞
done! But if lim an = 0, we cannot conclude anything...move on to the next step.
n→∞

2. Positive Series. If all terms in the series are positive, we have the following tests:
(a) The Direct Comparison Test.
Consider whether dropping terms in the numerator or denominator gives a series that we already know convergence
or divergence for. If a large series converges, or a smaller series diverges, then the original series does the same.
∞ ∞
X 1 X 1
For example, √ converges by comparison with the convergent p-series 2
.
2
k + k k
k=1 k=1
∞
X 1
However, if we try the same strategy on √ , we cannot conclude anything from this test. Maybe the Limit
k=2
k2 − k
Comparison Test will be better suited.
(b) The Limit Comparison Test.
Consider the dominant term in the numerator and denominator, and compare the original series to the ratio of these
terms.
∞
X 1 √ 1
For example, look again at √ . Since k2 grows faster than k, we compare with 2 to conclude the series
k 2 − k k
k=2
converges as
1√
k2 − k
lim 1 =1
k→∞
k2
∞
X 1
and is once again a convergent p-series.
k2
k=1

(c) The Ratio Test.

This is often useful when there is a factorial or a constant to the power n present, since the factorial or power will
disappear in the ratio.

76
 ∞
X 3k
For example, for , applying the Ratio Test yields
k!
k=1

3n+1
(n+1)! 3
lim 3n = lim = 0 < 1,
n→∞
n!
n→∞ n+1

hence the series converges.

(d) The Root Test.
Will not be covered on the exam.
(e) The Integral Test.
When all else fails for a positive, decreasing series, consider the Integral Test by looking at what happens to the
integral of the corresponding function.
∞
X 1 1
For example, consider . While all the other tests do not apply easily, the function is decreasing and
k ln k x ln x
Z ∞ k=2
dx
= ∞. Thus the integral, and hence the series, diverges.
2 x ln x
3. Series That Are Not Positive Series.
(a) Alternating Series Test.
If we have an alternating series, show the positive sequence is decreasing and tends to 0. Then the Alternating Series
Test shows the series converges, albeit, this only allows one to conclude conditionally.
(b) Absolute Convergence.
X X
If |ak | (which is now a positive series so we have many test for) converges, then ak converges absolutely, and
therefore is convergent.

Power Series [Section 10.6]

Remember that a power series is basically an infinitely long polynomial. So it look like
∞
X
a0 + a1 x + a2 x2 + a3 x3 + · · · = ak xk
k=0

where x is a variable. Be able to find the interval of convergence for a given power series.
Note that the most useful test for this is often the Ratio Test. Take the limit of the consecutive terms and then refer
to the Ratio Test’s criteria for convergence to understand where the series converges. If the limit of the ratio does
not depend on x and converges by the test, then it converges for all x. If there is x dependence left in the ratio, solve
for x in the inequality for convergence to find the radius of convergence. Keep in mind that the endpoints will always
have to be considered separately. Simply plug each in and see if the series (which no longer depends on x!) converges
or not using any test we know.

Taylor Polynomials [Section 10.7]

We define the nth Taylor polynomial, Tn , of f (x) centered at x = c as the nth partial sum
n
X f (k) (c)
Tn (x) = (x − c)k .
k!
k=0

77
Taylor Series and Complex Numbers [Section 10.8]
The Taylor series of f centered at c is
∞
X f (k) (c)
(x − c)k .
k!
k=0

If the Taylor series converges with radius r > 0, we say f is analytic on the interval (c − r, c + r). Note that the
Taylor series must converge to f (x) if it converges at all!
If the center is c = 0, sometimes we call the Taylor series Maclaurin series.
Be able to find a Taylor series for a given function. Note that this means you need to be able to find a “nice”
formula for the derivatives if possible. If you are given a function that looks complicated, that is, it is a composition
of functions or multiplication of a function by a power of x, remember to use what is known about the “simpler”
function instead of trying to find a formula for the derivatives directly.
3
For example, if you were asked to find the Taylor series for x2 ex centered at zero, there is no reason to take all the
derivatives (which would be really complicated!). We just use the series for ex and plug in x3 in place of x, then
multiply by x2 :
∞ ∞
3 X (x3 )k X x3k+2
x2 ex = x2 = for all x.
k! k!
k=0 k=0

Also be able to use term-by-term differentiation of integration to find a series representation for the derivative or
integral of a function. Recall that we used this technique to find
∞ ∞
(−1)k x4k+3
Z X Z X
sin x2 dx = x4k+2 dx = (−1)k for all x.
(2k + 1)! (2k + 1)!(4k + 3)
k=0 k=0

Euler’s Formula. For all complex numbers z,

eiz = cos z + i sin z.

Be able to use Euler’s Formula to write a complex exponential as the sum of its real and imaginary parts (and vice
versa). This is to say, if given the left-hand side of the above formula for a specific z, be able to rewrite in the form
of the right-hand side and simplify into a + bi, and vice versa.

78
35 Extra Lecture
35.1 Fluid pressure and force
The pressure p at depth h in a fluid of mass density ρ is

p = ρgh,

where g is the acceleration due to gravity. In other words, pressure is proportional to depth. This is one of the major
obstacle’s to deep sea diving!
Pascal’s principle says that the pressure acts perpendicular to any surface. To find the force acting on the surface,
we multiply pressure by area.
Units: pressure is usually measured in pascals (Pa), which are by definition Newtons per square meter (N/m2 ).
Example. Consider a 2 × 2 × 5 m box, whose top and bottom are square, submerged in a pool of water with its top
3 m below the surface. Recall the density of water is 1000 kg/m3 .

1. What is the force exerted on the top? Well, the depth is h = 3 m so the pressure is p = ρgh = 1000(9.8)3 =
29400 Pa. The area is 4 m2 so the force is 4 × 29400 = 117600 N.
2. What is the force exerted on the bottom? Here the depth is h = 8 m so the pressure is 1000(9.8)8 = 78400 Pa.
The are is the same; the force is 4 × 78400 = 313600 N.
3. What is the force exerted on one of the sides? Well, the depth changes from h = 3 to h = 8. At each depth,
a different pressure is exerted. We divide the side into strips of width dh; the length is always 2 m in this
case. So the pressure at each h is p(h) = 1000(9.8)h, the area is 2 dh, and the force is pressure times area or
1000(9.8)h(2)dh = 19600h dh. We integrate this to get the total force:
Z 8
8
19600h dh = 9800h2 3 = 9800(64 − 9) = 539000 N.

F =
3

Inclined surface. Consider a dam inclined at an angle of 45 degrees. We’ll say it has height 200 m and width 500
m. Assume the reservoir is filled to the top of the dam. Find the force exerted on the dam.
Our depth will vary from h = 0 all the way down to h = 200 m. At each depth h, we have a tiny strip 500
m across and the vertical distance from √top to bottom is dh. √ Since the strip is at an incline of 45 degrees, the
width of
√ that strip is not dh√but rather 2dh. The area is 500 2dh. The force exerted on that strip is therefore
ρgh500 2dh = 1000(9.8)500 2hdh. We integrate this to get total force exerted:
√ Z 200 √ 1
F = 490000 2 h dh = 490000 2 (200)2 = 1.39 × 1010 N.
0 2

General principle. At each depth h you know the pressure (p = ρgh). Imagine a strip of vertical width dh. To
get the force exerted on that strip, multiply by its area. To find the area, you need its length as a function of h; you
can call it l(h) or something. Finally, you integrate. So the total force exerted on a surface which extends downward
from depth a to b is Z b
F = ρghl(h)dh.
a

Not a rectangle. If the surface is not a rectangle, you have to do some geometry to see what the area of each strip
will be! For example, let’s say there’s an equilateral triangular plate of side 2 m submerged vertically in a tank of
water, with its top corner touching the surface and its bottom side perfectly horizontal (something must be holding
it in place!).
At each depth h, we need to know the distance l(h) across the equilateral triangle. By similar triangles we know
l(h)
√
h
= total 2height . The total height is 3 by the Pythagorean theorem. So l(h) = √23 h. Therefore
Z 2
2 9800
F = (1000)(9.8)h √ h dh = √ (2)2 = 22632 N.
0 3 3

79
36 Extra Lecture
36.1 Center of Mass and Centroids
Imagine a child of 80 lbs and an older child of 120 lbs on a see-saw. If the fulcrum is exactly in the middle, clearly
the bigger child will sink and the other will rise. Where can I put the fulcrum so that the two are evenly balanced?
Intuitively, we know it should be toward the bigger child. Where, exactly?
Let’s say the see-saw is 10 feet long. We’ll let x be the distance between the fulcrum and the bigger child. By the
principle of Archimedes, we discover that in order to find perfect balance we must have

120x = 80(10 − x),

where 10 − x is the distance from the fulcrum to the smaller child. Solving for x we get 200x = 800 or x = 4. So we
should put the fulcrum 4 feet from the bigger child.
In other words, this is the center of mass. In general, if we have two masses m1 and m2 lined up on the x-axis at
positions x1 and x2 , respectively, then the center of mass x̄ must satisfy
m1 x1 + m2 x2
m1 (x̄ − x1 ) = m2 (x2 − x̄) ⇒ (m1 + m2 )x̄ = m1 x1 + m2 x2 ⇒ x̄ = .
m1 + m2
The numbers m1 x1 , m2 x2 are called moments.
The center of mass is thus a kind of weighted average, much like the way your grade is calculated. It is defined as
the sum of the moments divided by the total mass. If all the masses are equal, then the center of mass is just the
average position, which is called the centroid.
The same formula holds for any number of masses m1 , . . . , mn located at points x1 , . . . , xn on the x-axis:
m1 x1 + · · · + mn xn
x̄ = .
m1 + · · · + mn

Now what if we want to find the center of mass of an object of variable density? We use our integration technique.
Example. Consider a 2 meter rod with linear mass density ρ(x) = 1 + x(2 − x) kg/m, where x is the distance from
one end of the rod in meters. Where is the center of mass?
Intuition: ρ(x) = 2 − (x − 1)2 , so the rod is heaviest in the middle at x = 1, and the density is perfectly symmetrical,
so one would think its center of mass is there in the middle. Let’s show that using calculus.
We break up the rod into an infinite number of pieces, each with length dx, so the mass is ρ(x)dx and the moment
is xρ(x)dx. To get the total mass and the total moment, we integrate each of these:
2 2 2
x2 2x3 x4
Z Z 
16 10
moment = xρ(x)dx = (x + 2x2 − x3 )dx = + − =2+ −4= ,
0 0 2 3 4 0 3 3
Z 2 Z 2  3
2
x 8 10
mass = ρ(x)dx = (1 + 2x − x2 )dx = x + x2 − =2+4− = .
0 0 3 0 3 3
So the center of mass is the moment divide by the total mass, which gives us x̄ = 1.
In general, to find center of mass of a linear object with linear mass density ρ(x) for x going from a to b, we have
Rb
xρ(x)dx
x̄ = Ra b .
a
ρ(x)dx

If the linear mass density is constant, the center of mass (or centroid, in this case) is just the midpoint between a
and b, i.e. a+b
2
.

80
Centroids in two dimensions
Given a finite number of masses m1 , . . . , mn at points (x1 , y1 ), . . . , (xn , yn ) on the plane, the center of mass can be
computed in the exact same way as in one dimension:
 
m1 (x1 , y1 ) + · · · + mn (xn , yn ) m1 x1 + · · · + mn xn m1 y1 + · · · + mn yn
(x̄, ȳ) = = , .
m1 + · · · + mn m1 + · · · + mn m1 + · · · + mn

For example, if three stones having masses 2, 4, and 8 kg are placed at points (0, 2), (3, 1), and (6, 4), respectively,
then the center of mass of the three stones is ( 30
7
, 20
7
).
Note that the x- and y-coordinates are calculated independently of one another! In fact, if we call the sum of all
moments m1 x1 + · · · + mn xn = Mhor the total horizontal moment, and if we call Mver = m1 y1 + · · · + mn yn the total
vertical moment, then we just have  
Mhor Mver
(x̄, ȳ) = , .
M M
(Note: I do not use the book’s notation My and Mx , because I find it too confusing.)
As the number of masses goes to infinity, finding the center of mass will in general require a double integral, which
is not part of this course. Still, we can use some tools we already know in order to find the centroid of some thin
plates in two dimensions.
Suppose we are looking at a region bounded by the graph of y = f (x), the x-axis, and the vertical axes x = a and
x = b. We want the centroid of this region. This is the same as the center of mass, assuming the density is uniformly
1; so mass is just area. Let M be the total mass, Mhor be the horizontal moment, and Mver be the vertical moment
as before. The total mass is just the area: Z b
M= f (x)dx.
a
The horizontal moment can be found by splitting the region into blocks of width dx and height f (x); the mass (area)
of each block is f (x)dx, so the moment is xf (x)dx. We add these moments together to get the total horizontal
moment: Z b
Mhor = xf (x)dx.
a
The vertical moment can be found in a similar way, if we can somehow find the horizontal distance across the region
at each height y. This is feasible in some cases, like when f (x) is strictly increasing or decreasing. Alternatively,
we can split once more into vertical blocks and find the vertical moment of each block. Each block is a rectangle of
height f (x) and width dx. By symmetry, its centroid is at height 12 f (x). On the other hand, its total mass is its
area, f (x)dx. So its vertical moment is the product of these, or 12 f 2 (x)dx. To get the total vertical moment we add
these up: Z b
1 2
Mver = f (x)dx.
a 2

Example. Let R be the region bounded by y = x2 , the x-axis, the y-axis, and the line x = 2. Let’s find the
R2 3 R2 R2
centroid. The total mass is M = 0 x2 dx = x3 |20 = 38 . The horizontal moment is Mhor = 0 x · x2 dx = 0 x3 dx =
x4 2
|
4 0
= 4. The vertical moment can be gotten in two different ways. First, we could use horizontal blocks with
√
width dy, going from y = 0 to y = 4. The distance across at each height y is 2 − y. So the vertical moment is
R4 4
Mver = 0 y(2 − y 1/2 )dy = 0 (2y − y 3/2 )dy = (y 2 − 25 y 5/2 )|40 = 16 − 52 (32) = 16
R
5
. The other way to see this is using
1 2
R2
the formula Mver = 2 0 (x ) dx = 12 0 x4 dx = 10
2 2 1 5 2 32
= 16 is (x̄, ȳ) = 38 (4, 16 ) = ( 32 , 65 ).
R
x |0 = 10 5
. So the centroid 5

Symmetry principle: If a region is symmetric about a line, then its centroid lies on that line.
Example. Let’s find the centroid of the upper half of the unit disk. Since it’s symmetric about the y-axis, the
centroid must lie on this axis, i.e. x̄ = 0.
As for ȳ, we need the vertical moment; we already know the total√
mass is π/2. The vertical moment can be computed
in two ways, but probably the simpler way is to use y = f (x) = 1 − x2 as a formula for the circle and then

1 1 2 1 1 x3 1
Z Z
1 1
Mver = f (x)dx = (1 − x2 )dx = [x − ]−1 = [2/3 − (−2/3)] = 2/3.
2 −1 2 −1 2 3 2

81
 4
So the centroid is found at (0, 3π
).
Additivity of moments and mass. If we put together two separate regions to make one region, clearly the mass
of the whole is obtained by adding. But the same is true of each of the moments. We can use that to be clever in
computing the centroid of composite regions.
Example. Let R be the region consisting of two disks, one of radius 2 centered at the origin and the other of radius
1 centered at (0, 3). (These touch each other at (0, 2)). Find the centroid.
Well, both disks have the y-axis going down the center, so the centroid is somewhere on the y-axis. Now the centroid
of the first circle is just (0, 0), so its vertical moment is just 0. The centroid of the second circle is just (0, 3), and its
mass is π, so its vertical moment is 3π. Thus the total vertical moment of R is 3π. Since its mass is 4π + π = 5π, its
centroid is at (0, 35 ).

82
37 Extra Lecture
37.1 Integrating factor
Warm-up: find the general solution of
dy
= p(x)y.
dx
This is separable. We get
dy R
= p(x)dx ⇒ y = Ce p(x)dx .
y
Now let’s consider a differential equation like
dy
+ p(x)y = q(x).
dx
R
p(x)dx
This is not separable. But we hope to use e somehow.
0
If P (x) is an antiderivative of p(x), i.e. P (x) = p(x), then notice, by the product rule,
 
d P (x) P (x) dy P (x) 0 P (x) dy
(e y) = e +e P (x)y = e + p(x) = eP (x) q(x).
dx dx dx

That’s a neat trick. Now we integrate both sides:

Z
P (x)
e y= eP (x) q(x)dx.

Then we can solve for y.

eP (x) is called an integrating factor.
NOTE: The integrating factor doesn’t need a “+C” in it, since it’s just a tool. In fact, you really don’t even need
the most general antiderivative; for example, if you integrate 1/x to get the integrating factor, you should just take
ln x rather than ln |x| (the integrating factor would then be eln x = x). The reason for this can sometimes be technical
(think about the domain of the solution y) but in any case, sometimes being a little lazy can pay off.
R
Example. Let’s find the general solution of y 0 = y+x. Rewrite as y 0 −y = x. The integrating factor is e− 1 dx
= e−x
(remember, not the most general antiderivative). Then we check that

d −x
(e y) = e−x y 0 − e−x y = e−x x.
dx
We need to integrate the right-hand side. Use integration by parts to get e−x x dx = −e−x (x + 1) + C. So
R

e−x y = −e−x (x + 1) + C ⇒ y = −(x + 1) + Cex .

It’s not an accident that the solution is y = Cex + g(x), where g(x) is some other function of x.
Check yourself: If you solve an equation like y 0 + p(x)y = q(x) using an integrating factor, check your solution by
solving y 0 + p(x)y = 0. Your solution to the first equation should look like the solution to the second plus an extra
term involving x.
Example. Solve xy 0 − 3y = x2 , y(1) = 2. This is an initial value problem, so there are two steps. First, the general
solution. Write as
3
y 0 − y = x.
x
R 3
Integrating factor: e− x
dx
= e−3 ln x = x−3 . Check that
d −3
(x y) = x−3 y 0 − 3x−4 y = x−3 (y 0 − 3x−1 y) = x−3 x = x−2 .
dx
So then Z
x−3 y = x−2 dx = −x−1 + C ⇒ y = −x2 + Cx3 .

83
Second, we use the initial condition y(1) = 2 to get 2 = −1 + C so C = 3. The solution is y = −x2 + 3x3 .
R
Example. Find the general solution of y 0 +(sec x)y = cos x. The integrating factor here is e sec x dx
= eln(sec x+tan x) =
sec x + tan x. We can check that
d dy
((sec x + tan x)y) = (sec x tan x + sec2 x)y + (sec x + tan x)
dx dx
= (sec x + tan x)(sec xy + y 0 ) = (sec x + tan x) cos x = 1 + sin x.
So Z
(sec x + tan x)y = (1 + sin x)dx = x − cos x + C.

Thus
x − cos x + C
y= .
sec x + tan x

Example. Solve y 0 + y = sin x, y(0) = 1. The integrating factor is ex , we get ex y = ex sin x dx. We do integration
R

by parts twice:
Z Z Z
ex sin x dx = ex sin x − ex cos x dx = ex sin x − ex cos x + ex (− sin x) dx
Z
1 x
⇒ ex sin x dx = e (sin x − cos x) + C.
2
Thus y = 21 (sin x − cos x) + Ce−x . Since y(0) = 1 we get 1 = − 12 + C or C = 32 . Therefore y = 21 (sin x − cos x) + 23 e−x .

84