CALCULUS II (Math 102)

Text: Essential Calculus 2/e by James Stewart

Calculus: The study of real functions f , and their derivatives f 0 and integrals
R
f.

Let f : [a, b] → R be continuous.

Definition The definite integral of f from a to b is

Z b Z b n
X
f = f (x) dx := lim f (xi ) ∆x
a a n→∞
i=1
where ∆x = (b − a)/n and xi = a + i ∆x. Note: x is a “dummy” variable, replaced in many
Rb Rb
physical applications by time t, and soi a f = a f (t) dt.

f (xi )

a ∆x b
xi

The base of the shaded rectangle is ∆x, and f (xi ) is its height.

Physical Meaning distance traveled (during time a ≤ t ≤ b, where f (t) = velocity at time t)
Geometric Meaning signed area (between the graph of f and [a, b] ⊂ x-axis)

a b

Fundamental Formula of Calculus (FFC)

Z b  b 
f (x) dx = F (b) − F (a) also written F (x)
a a
†
for any antiderivative F of f on [a, b], meaning F0
= f on [a, b]. Since any two such antideriva-
tives differ by a constant (a consequence of the Mean Value Theorem (MVT)) it follows that
any antiderivative can be used; the constants will cancel!
† Antiderivatives are also called primitives or indefinite integrals .
1
 Bryn Mawr College

This wonderful formula, which reduces the problem of computing definite integrals to the prob-
lem of computing antiderivatives, follows from the following remarkable result:

Fundamental Theorem of Calculus (FTC) The function F : [a, b] → R defined by

Z x (
signed area between the graph of f
F (x) = f (x) dx =
a and the interval [a, x] on the x-axis
Z x
d
is an antiderivative of f . In other words, f (x) dx = f (x).
dx a
This theorem can be understood geometrically using the area interpretation of the definite
integral, and proved rigorously using – once again – the MVT. The FFC follows:
Z b Z a Z b
F (b) − F (a) = f (x) dx − f (x) dx = f (x) dx.
a a a
R
Because of this theorem, we often simplyR write f (x) dx to denote the antiderivative(s) of f .
Thus for example x dx = 12 x2 + C and x2 dx = 13 x3 + C.
R

Techniques of Integration

xp+1
Z
• Power Rule xp dx =
p+1

In words: to integrate a power, raise the power by 1, and then divide by the new power. This
follows from the power rule for derivatives (differentiate the RHS to verify it) so we also refer
to it as the backward power rule.
Z Z Z
• Sum and Difference Rule (f (x) ± g(x)) dx = f (x) dx ± g(x) dx

• Integration by Parts (backward product rule)

Z Z
f (x)g(x) dx = f (x)G(x) − f 0 (x)G(x) dx.

where G is any antiderivative of g (differentiate both sides to verify) More on this later.
• Integration by Substitution (backward chain rule)

If the substitution u = u(x) and du = u0 (x)dx converts

Z Z
f (x) dx into g(u) du

(that is, if f (x) = g(u(x)) u0 (x)) then the first integral can be computed by evaluating the
second, and then plugging u(x) back in for u. The key here is to find a u for which the new
integral is easier to evaluate than the old one. In particular, if u(x) and u0 (x) both appear in
f (x), then substituting u = u(x) is potentially productive.
For definite integrals, one can avoid plugging back in by substituting for the bounds:
Z b Z u(b)
f (x) dx = g(u) du
a u(a)

2
 Calculus II

4.5 Integration by Substitution

Z Z
1k
p
Examples 2
x x + 1 dx = x(x2 + 1)1/2 dx.

Substitute u = x2 + 1, du = 2x dx (so 21 du = x dx) to get

Z
1 1 2 3/2 1
u1/2 du = u + C = u3/2 + C
2 2 3 3
Z
1
and, plugging x2 + 1 back in for u, x(x2 + 1)1/2 dx = (x2 + 1)3/2 + C .
3
Z 1 p
2i x x2 + 1 dx.
0

With the same substitution, and noting that u(0) = 1 and u(1) = 2, we get
1 2 1/2 1 3/2 2 1 √
Z
u du = u = ( 8 − 1) .
2 1 3 1 3
Z
3i
p
x3 x2 + 1 dx.

With the same substitution, there will be an extra x2 = u − 1, so get

Z Z
1 1/2 1 1 2 2
(u − 1)u du = (u3/2 − u1/2 ) du = ( u5/2 − u3/2 ) + C
2 2 2 5 3
and, plugging x2 + 1 back in for u,
Z p
1 1
x x2 + 1 dx = (x2 + 1)5/2 − (x2 + 1)3/2 + C .
5 3
Z
4i
p
x2 x2 + 1 dx.

With the same substitution, there will be now be an extra x = (u − 1)1/2 , so get
Z Z
1 1/2 1/2 1
(u − 1) u du = (u2 − u)1/2 du
2 2
and its not clear that the substitution has helped. We’re stuck, at least for now.
Z
5i sec3 x tan x dx.

Substitute u = sec x, du = sec x tan x to get

Z
1 1
u2 du = u3 + C = sec3 x + C .
3 3
What happens if you substitute u = tan x? End up with
Z
1
u(u2 + 1)1/2 du = (u2 + 1)3/2 + C
3
1
from example (1) above, which equals 3 sec3 x + C since tan2 x + 1 = sec2 x.

3
 Bryn Mawr College
Z 2
6i (like HW 4.5 (50))
p
x2 64 − x6 dx.
0
p
If you substitute u = 64 − x6 this becomes the integral of u/(64 − u), up to a constant. Ugh!
But if you substitute u = x3 it becomes

1 8p
Z
64 − u2 du .
3 0
√
Now observe that the graph of v = 64 − u2 in the uv-plane is the upper half of the circle
u2 + v 2 = 82 . Thus the last integral can be interpreted as the area 16π of a quarter of that
circle, and so the original integral has value 16π/3.

7.1 Areas Between Curves

Let R be a bounded region in the plane. Choose an axis, parametrized say by t. For each value
of t, let L(t) denote the corresponding cross-sectional length of R, and suppose that L(t) is
positive for all t in (a, b), and zero outside of [a, b]. Then the area of R is given by
Z b
A = L(t) dt
a

Special case : R is bounded by curves

Z b
I) the x-axis, the graph of f (x) > 0 and the vertical lines x = a and x = b : A = f (x) dx
Z b a

II) the graph of y = u(x) and of y = `(x) ≤ u(x) : A = (u(x) − `(x)) dx

a
• Find the limits a, b of integration by solving f (x) = g(x).
• If you don’t know which of u or ` is larger, take the absolute value at the end.

Examples Find the areas of the regions bounded by

1i y = x and y = x2 : The curves intersect where x = x2 , that is, where x = 0 and x = 1.

Clearly x is larger than x2 along [0, 1], and so
Z 1
1 1 3 1 1
A = (x − x2 ) dx = ( x2 − x ) =
0 2 3 0 6

2i y = x2 − 3, y = 5 − x2 : The curves intersect where x2 − 3 = 5 − x2 , that is, where 2x2 = 8,

so x = −2 and x = 2. The region is clearly symmetric about the y-axis, so we can (to simplify
the math) compute twice the area along the interval [0, 2]. But suppose we don’t realize that
the second curve is on top. Then
Z 2 Z 2
2 2 2 2 32
A = 2 ((x − 3) − (5 − x )) dx = (2x2 − 8) dx = ( x3 − 8x) = .
0 0 3 0 3

3i x + 1 = 2(y − 2)2 and x + 6y = 7 (integrate with respect to y)

4
 Calculus II

5.1 Inverse Functions

In one-variable calculus we study functions f : X → R whose domain X is a subset of the

real numbers R. The function assigns a real number f (x) to each number x in X.
The domain is an indispensable part of the function. If not specified, it is understood to be
the set of all numbers x for which f (x) makes sense. For example the domain of the function
f (x) = 1/(x − 1) is the set of all real numbers not equal to 1.
The range of f : X → R is the set of all the numbers f (x) that arise as x ranges over X. For
example the range of f (x) = 1/(x − 1) is the set of nonzero real numbers.

Examples Find the domain and range of

1k the sine function : The domain is R, while the range is the closed interval [−1, 1].
x2 + 1
2k the function f (x) = : The domain is the set of all real numbers 6= ±1.
x2 − 1
The range is the set of all real numbers ≤ −1 or > 1. This takes a little work to verify.

Definition A function f : X → R is one-to-one (also written as 1-1) provided it never takes on

the same value twice, that is f (x1 ) 6= f (x2 ) whenever x1 6= x2 .

Equivalently, this says that every horizontal line intersects the graph of f in at most one point.
This is the horizontal line test for checking if f is 1-1.†
Here is a useful sufficient condition for a differentiable function defined on an interval to be 1-1;
the interval can be closed, half open, or open (which includes R = (−∞, ∞)):

Theorem (Derivative Test) If X is an interval, and f : X → R is differentiable with f 0 (x)

always of the same sign (either always positive or always negative), then f is 1-1.

This follows from the Mean Value Theorem, which implies that f is monotonic (i.e. either
increasing or decreasing) on X, and so certainly 1-1.

Examples Which of the following functions are 1-1?

1k the sine function : this is not 1-1, for example sin(0) = sin(π)
2k f (x) = x3 + 2x + 1 : the derivative is 3x2 + 2, which is positive for all x, so f is 1-1
3k T (t) = temperature in this room as a function of time t : this is surely not 1-1

Definition If f : X → R is a 1-1 function, then it has an inverse function

f −1 : Y −→ R (where Y is the range of f )

sending each y in Y to the unique x in X for which f (x) = y. Thus the domain of f −1 is the
range of f , and the range of f −1 is the domain of f . We often simply write
f
X Y
f −1

† Compare this with the vertical line test for checking if a curve in the plane is the graph of a function.

5
 Bryn Mawr College

Clearly f does not have an inverse function if it is not 1-1. Thus

f is 1-1 ⇐⇒ f −1 exists ⇐⇒ f satisfies the horizontal line test

How to try to find a formula for f −1

• Set y = f (x)
• Try to solve for x in terms of y x = f −1 (y)
• Swap x and y to give the conventional form y = f −1 (x).

Examples Find a formula for the inverse functions of the following 1-1 functions:
√ √
1kf (x) = x3 : y = x3 =⇒ x = 3 y. Therefore f −1 (x) = 3 x
2kf (x) = x3 + 2x + 1 : stuck
2x + 1 2x + 1
3kf (x) = : y= =⇒ y(3x − 5) = 2x + 1 =⇒ (3y − 2)x = 1 + 5y
3x − 5 3x − 5
1 + 5y 1 + 5x
=⇒ x = . Therefore f −1 (x) =
3y − 2 3x − 2

Properties The inverse function f −1 : Y → R of a 1-1 function f : X → R satisfies:

• y = f (x) if and only if f −1 (y) = x

• f −1 (f (x)) and f (f −1 (y)) = y, for all x ∈ X and y ∈ Y
• the graph of f −1 is the reflection of the graph of f through the line y = x
The last property shows geometrically (though it’s hard to prove analytically) that:
• If f is continuous, then so is f −1
• If f is differentiable with nonzero derivative everywhere, then so is f −1
This final fact is the first half of the very useful:

Inverse Function Theorem (IFT) Let f : X → R be a 1-1 differentiable function with f 0 (x) 6=
0 for all x in X. Then the inverse function f −1 : Y → R is differentiable, with derivative
(f −1 )0 (f (x)) = 1/f 0 (x).

Proof Differentiate f −1 (f (x)) = x (using the chain rule) to get (f −1 )0 (f (x)) f 0 (x) = 1. Now
divide by f 0 (x) and substitute y for f (x). 

Remark Practically speaking, the IFT shows

ak If you want to compute (f −1 )0 (b) for some particular number b, try to find a for which
f (a) = b. Then (f −1 )0 (b) = 1/f 0 (a).
bk If you want a general formula for (f −1 )0 (x), write y = f −1 (x) (reversing the roles of x
and y). Then x = f (y). Suppose that we know how to write f 0 (y) = dx/dy in terms of x (e.g.
replacing y by f −1 (x)). Then we get an explicit formula for
dy 1
= .
dx dx/dy

6
 Calculus II

We will give many examples of bkwhen we discuss inverse trig functions, and later the expo-
nential function, but first we illustrate ak:

Examples 1kGiven f (x) = x3 + 2x + 1, find (f −1 )0 (4). Solution: noting that f (1) = 4, we

have (f −1 )0 (4) = 1/f 0 (1) = 1/(3x2 + 2)|x=1 = 1/5.
2k(like HW 5.1 (44)) Let g be the inverse function of a differentiable 1-1 function f , and let
h = g 2 . If f (3) = 4 and f 0 (3) = 5, find h0 (4). Solution: By the chain rule, h0 (4) = 2g(4)g 0 (4) =
2 · 3/f 0 (3) = 6/5.

5.6 Inverse Trigonometric Functions

First observe that none of the trig functions sin, tan, sec, . . . are 1-1 (look at their graphs) so
we must restrict their domains in order to make them 1-1. These restrictions are denoted sin |,
tan |, sec |, . . . (shown in green below) and their inverses are what we simply call sin−1 , tan−1 ,
sec−1 , . . . :

Definition
sin | tan |
[− π2 , π2 ] [−1, 1] (− π2 , π2 ) (−∞, ∞)
sin−1 tan−1
and
sec |
[0, π2 ) ∪ [π, 3π
2 ) (−∞, −1] ∪ [1, ∞)
sec−1

sin | tan | sec |

sin−1 tan−1 sec−1

Formulas
Z
d 1 1
1k sin−1 x = √ and sin−1 x = √
dx
dx 1 − x2 1 − x2
Z
d 1 1
2k tan−1 x = 2 and tan−1 x = dx
dx x +1 x2 + 1
Z
d 1 1
3k sec−1 x = √ and −1
sec x = √ dx
dx x x2 − 1 x x2 − 1

Proofs (illustrating bkabove)

7
 Bryn Mawr College

1k y = sin−1 x. Thus x = sin y and so

dy 1 1 1 1
= = = p = √ .
dx dx/dy cos y 2
1 − sin x 1 − x2

2k y = tan−1 x. Thus x = tan y and so

dy 1 1 1 1
= = 2
= 2 = 2 .
dx dx/dy sec y tan x + 1 x +1

3k Homework 5.6 (#14)

Examples Using the formulas above (and the chain rule / substitution)
d 1 −c c
1k If c is constant, then tan−1 (c/x) = · = − 2
dx (c/x)2 + 1 x2 c + x2
d 1 3x2
2k tan−1 (sin−1 (x3 )) = · √
dx (sin−1 (x3 ))2 + 1 1 − x6
Z Z Z
k 1 1 1
3 √ dx = p dx = p dx.
2
x 4x − 9 2
x 9(4x /9 − 1) 3x (2x/3)2 − 1
Substituting u = 2x/3, du = (2/3)dx gives
Z
1 1 1 2x
√ du = sec−1 u + C = sec−1 ( ) + C.
2
3u u − 1 3 3 3

4k (like HW 5.6 (38)) You’re 5 ft tall, and in a museum looking at a 8 ft tall painting hanging
on the wall with its top at 14 ft above the floor. At what distance x away from the wall should
you stand to have the maximum viewing angle?
Solution: The viewing angle θ is given as a function of x by
θ = β − σ = tan−1 (9/x) − tan−1 (1/x)
as seen by drawing a picture.

β
θ
σ 1

We compute
dθ 9 1
= − 2
+
dx 81 + x 1 + x2
which is zero when the two fractions are equal, that is, when 9 + 9x2 = 81 + x2 , or x = 3. This
is evidently a maximum, and so the optimal distance is 3 feet.

8
 Calculus II

5.2 The Natural Logarithm

The natural logarithm function, denoted ln, is defined for all positive real numbers x by
Z x
1
ln x = dt
1 t
(we use t for the integration variable since x is being used as the indep. variable of ln). Note:
ln is not defined for x ≤ 0; its domain is the set R+ of all positive reals.
Geometrically, ln x represents a signed area under the graph of y = 1/t, as shown below:

y
−Area = ln x
y = 1/t Rx R1
(since 1 = − x )
Area = ln x

t
1 x x 1

Thus ln x is negative for x < 1 and positive for x > 1, and clearly zero when x = 1 (using either
the integral definition or the area interpretation).
We will show below that the range of ln is all of R. This is not obvious. For example, why
should one be able to achieve an arbitrarily large blue Rarea by choosing x large enough? (Think
x
about the analogous question for the function f (x) = 1 (1/t2 ) dt)

The derivative of ln
By the Fundamental Theorem, we compute
Z x
d d 1 1
ln x = dt = .
dx dx 1 t x
which holds for x > 0, where ln x is defined. There is an analogous formula for x < 0, when
using the chain rule we compute ln(−x) 0 = −1/(−x) = 1/x. Thus, in fact
d 1
ln |x| =
dx x
for all nonzero x, positive or negative.
The corresponding integration formula is
Z
1
dx = ln |x| + C.
x
Below, we will give explicit examples that use these derivative and integral formulas.
But first . . .

Algebraic laws of ln
For any a, b ∈ R+ and any rational number r
ak ln(ab) = ln a + ln b bk ln(a/b) = ln a − ln b ck ln(br ) = r ln b
Qualitatively, this says that the logarithm converts products into sums, quotients into differences,
and powers into products, which explains its great historical significance.

9
 Bryn Mawr College

Proofs
akSet f (x) = ln(xb). Then f 0 (x) = b/xb = 1/x, by the chain rule, so f (x) = ln x + C for some
constant C. Substituting x = 1 shows that C = ln b. Therefore ln(ab) = f (a) = ln a + ln b as
claimed.
bkFrom akwe have ln a = ln((a/b)b) = ln(a/b) + ln b, and the result follows.
ckSet g(x) = ln(xr ). Then g 0 (x) = rsr−1 /xr = r/x, so g(x) = ln x + C. Substituting x = 1
shows that C = 0. Therefore ln(br ) = f (b) = r ln b as claimed.

The graph of ln

Monotonicity ln0 x = 1/x > 0 for all x > 0, so ln x is an increasing function.†

Concavity ln00 x = (1/x)0 = −1/x2 < 0 for all x, so the graph of ln is concave down.
Thus, recalling that ln 1 = 0, we can give a rough sketch of the graph:
y
y = ln x

x
1

Asymptotic behavior
The logarithm function increases without bound as x → ∞, and decreases without bound as
x → 0, meaning
lim ln x = +∞ and lim ln x = −∞.
x→∞ x→0
Indeed ln 2±n = ±n ln 2 can be made as large in absolute value as we want by choosing n
sufficiently large. Since ln is continuous – indeed differentiable – it follows from the Intermediate
Value Theorem that the range of ln is all of R, as claimed above.
Later we will show – using L’Hôpital’s Rule for evaluating limits of quotients – that ln x “grows
slower” than any power xn (for n = 1, 2, . . . ), meaning
lim ln x/xn = 0 = lim xn ln x.
x→∞ x→0

Differentiation of logarithmic functions

Examples
1k Find an equation of the tangent line to y = sin(2 ln x) at (1, 0): The slope is y 0 (1) =
cos(2 ln x)(2/x)|x=1 = 2 cos 0 = 2, so the equation of the line is y = 2x − 2.
2k Implicit Differentiation Find y 0 if y is defined implicitly by y = ln(x + y 2 ): Differentiating
both sides w.r.t. x gives y 0 = (1 + 2yy 0 )/(x + y 2 ). To solve for y 0 , first cross multiply, giving
(x + y 2 )y 0 = 1 + 2yy 0 , and then collect the y 0 terms, giving (x + y 2 − 2y)y 0 = 1. Therefore
y 0 = 1/(x + y 2 − 2y).
† and therefore 1-1; we will discuss its inverse function “exp” in the next section

10
 Calculus II

3k Logarithmic Differentiation A special technique used to compute the derivative of compli-

cated functions based on the philosophy that the logarithm simplifies the algebra. If y = f (x),
then by the chain rule, (ln y)0 = y 0 /y, and so y 0 = y(ln y)0 . Thus
f 0 (x) = f (x) · (ln f (x))0 .
for any function f (x).
Examples Find y 0 for the given y.
s
(x2 + 1)3 tan x
ak y = √ =⇒ (ln y)0 = y · 12 (3 ln(x2 + 1) + ln tan x − 12 ln x)0 , so
x
s
(x2 + 1)3 tan x sec2 x
 
0 0 3x 3
y = y(ln y) = √ + − .
x x2 + 1 2 tan x 4x

bk y = xx =⇒ y 0 = y(ln y)0 = xx (x ln x)0 = xx (1 ln x + x(1/x)) = xx (ln x + 1).†

ck y = f (x)g(x) (a good exercise to test your understanding) †

Integration with logarithms

R
Recall that (1/x) dx = ln |x| (suppressing the ”+C” for convenience). Therefore
Z 0
f (x)
dx = ln |f (x)|
f (x)
for any function f (x), as seen by substituting u = f (x). For example
Z Z
cos x
cot x dx = dx = ln | sin x|
sin x
Z Z
sin x
tan x dx = dx = − ln | cos x| = ln | sec x|.
cos x

More Examples
Z Z
k 1 1/x
1 dx = dx = ln(ln x).
x ln x ln x
Z
2k sec x dx. Trick: multiply the integrand by 1 = (sec x + tan x)/(sec x + tan x),

sec2 x + sec x tan x

Z Z
sec x dx = dx = ln | sec x + tan x| .
sec x + tan x

3(ln x)2
Z Z
3k dx. Substituting u = ln x, we get 3u2 du = u3 + C = ln3 x.
x
2 ln 2
3(ln x)2
Z Z
4k dx. The substitution in 3kleads to
ln 2
u2 du = u3 0
= (ln 2)3 .
1 x 0

† In bkand ckwe allow arbitrary real exponents, so that xx and more generally f (x)g(x) should make sense
when x or g(x) are real. This will be justified when we discuss exponential functions below.

11
 Bryn Mawr College

5.3 The Exponential Function

The natural exponential function, denoted exp, is the inverse function of the natural logarithm
function:
ln
R+ R
exp
so exp has domain R and range R+ . Note that exp(0) = 1 since ln 1 = 0. We sketch the graph
of y = exp(x) by reflecting the graph of y = ln x about the line y = x:

y = exp(x)
1

The asymptotic behavior of ln translates into the corresponding behavior of exp:

lim exp(x) = ∞ and lim exp(x) = 0 .
x→∞ x→−∞

and as for logarithms, L’Hôpital’s rule will show that exponential functions of x “grow faster”
than any power xn (n = 1, 2, . . . ), i.e. limx→∞ xn / exp(x) = 0 = limx→−∞ xn exp(x).

The number e

RDefine
e
e := exp(1). Thus e is defined by the equation ln e = 1, and so is the number for which
1 (1/t) dt = 1.

y = 1/t
Area = 1

1 e

From the definition of the integral as a limit of Riemann sums, one can (with some work)
approximate e to any desired accuracy:
e ≈ 2.71828182845904523536028747135266249775724709369995957496697
In fact e is an irrational number. This can be shown using calculus, but we won’t take the time
to do so here.
Recall that if r is rational, then ln(er ) = r ln e = r, and so er = exp(r). We now define
ex := exp(x) or more generally bx := exp(x ln b) = ex ln b
for any real number x and any b > 0. (You should check that this agrees with the usual definition
when x is rational.) Taking the logarithm of both sides shows that the formula
ln(bx ) = x ln b
now holds for all x.

12
 Calculus II

A limit formula for e

It is a remarkably useful fact (used in financial applications) that
e = lim (1 + 1/n)n .
n→∞
To see why this is true, we compute the derivative of ln x at x = 1 from the definition, which
we also know from the formula ln0 x = 1/x must equal 1:
 
0 ln(1 + h) − ln 1 
1/h

1/h
1 = ln (1) = lim = lim ln (1 + h) = ln lim (1 + h)
h→0 h h→0 h→0

since ln is a continuous function. Taking exp of both sides yields the desired formula:
e = lim (1 + h)1/h = lim (1 + 1/n)n . †
h→0 n→∞

The derivative of exp

Set y = exp(x) = ex . Then ln y = x, and so
d x dy 1 1
e = = = = y = ex .
dx dx dx/dy 1/y
Thus ex is its own derivative ! And so also its own integral: ex dx = ex .
R

Algebraic laws of exp

The laws of logarithms: ln(ab) = ln a + ln b, ln(a/b) = ln a − ln b and ln(bx ) = x ln b, translate
into the familiar laws of exponents:
akep+q = ep eq bkep−q = ep /eq ck(eq )x = eqx
for any real numbers p, q and x. Indeed the latter follow from the former by setting p = ln a and
q = ln b, and then applying exp. And a quick check shows that these laws hold with e replaced
by any positive real number b:
akbp+q = bp bq bkbp−q = bp /bq ck(bq )x = bqx
as well as the familiar law dk(bc)x = bx cx for any positive b and c.

Differentiation and integration with exponential functions

Examples
1k Find an equation of the tangent line to y = ex +1 sin x at (0, 0): The slope is y 0 (0) =
2

2 2
(2xex +1 sin x + ex +1 cos x)|x=0 = e, so the equation of the line is y = ex.
2kFind the absolute maximum value of the function f (x) = 2x−ex . We compute f 0 (x) = 2−ex .
This is 0 when ex = 2, i.e. when x = ln 2, > 0 to the left of ln 2 and < 0 to the right. Therefore f
has an absolute maximum at x = ln 2 ≈ .7, with value f (ln 2) = 2 ln 2 − eln 2 = 2 ln 2 − 2 ≈ −.6.
√
Z 1
k
3 Evaluate ex 1 + ex dx. Substituting u = 1 + ex we obtain
0
Z 1
x
√
Z 1+e √
2 3/2
1+e √
e 1+ ex dx = u du = 3u = 2
3 ((1 + e)3/2 − 8) ≈ 2.89.
0 2 2

† This last limit converges very slowly – for example (1 + 1/400)400 = 2.715 to 3 decimal places, while the
correct value of e to 3 decimal places is 2.718 – so it cannot be used as a practical way to compute e.

13
 Bryn Mawr College

5.4 General Logarithmic and Exponential Functions

For any fixed b > 0 define the new functions

logb x = ln x/ ln b and expb (x) = bx = ex ln b .
Thus the logb x is obtained from ln x by dividing by the constant ln b, and expb (x) is obtained
by multiplying x by ln b and then applying exp. These are inverse functions
logb (bx ) = x and blogb x = x
as is readily verified.†

Algebraic laws

As noted above, bx satisfies the familiar algebraic laws, and a similar check shows that logb does
as well:
aklogb xy = logb x + logb y bklogb x/y = logb x − logb y cklogb xp = p logb x

Derivatives and Integrals

The derivatives of logb and expb are easily computed:

d 1 d x
logb x = and b = bx · ln b.
dx x ln b dx
For example, the second formula follows from the chain rule: (bx )0 = (ex ln b )0 = ex ln b (x ln b)0
= bx · ln b, and immediately yields the integration formula:
Z
bx dx = bx / ln b + C.
Z Z
x2 2
Example 4x · 5 dx = 2 · 5u du (where u = x2 ) = 2 · 5u / ln 5 = 2 · 5x / ln 5.

Graphs

These are discussed in some detail in the text. The basic principal is that the larger the base b,
the slower logb grows, and the faster bx grows.

increasing
b
y

increasing
b
y= bx

y = logb x

† log (bx ) = ln(bx )/ ln b = x ln b/ ln b = x and blogb x = e(ln x/ ln b) ln b = eln x = x.

b

14
 Calculus II

5.5 Applications: Exponential Growth and Decay

Population Growth

Let P (t) be the population at time t (of some city/animal tribe/bacteria culture, etc.), and set
P0 = P (0). With no environmental restrictions, we can reasonably assume the growth rate P 0 (t)
is proportional to P (t), i.e. P (t) satisfies the IVP (“initial value problem”)

P 0 (t) = k P (t) , P (0) = P0

for some k > 0 (generally not given; it depends on the nature of the population). One obvious
solution to these equations is

P (t) = P0 ekt (check this by differentiating).

In fact any solution must be of this form.†

Examples 1kA population was 1 million last year and 3 million this year. What do you expect
it to be next year?
Solution In the notation above P0 = P (0) = 3, so P (t) = 3ekt for some k. But we also know
P (−1) = 1, which means 3e−k = 1, or k = ln 3. Thus P (t) = 3et ln 3 = 3t+1 , and so the
population next year will be P (1) = 31+1 = 9 million.
2kIf a population doubles in 10 years, how long will it take to triple?
Solution We are given that P (10) = 2P0 = P0 e10k , and so e10k = 2. Thus k = ln 2/10, so
P (t) = P0 et ln 2/10 = P0 2t/10 . We want to find t for which P (t) = 3P0 , that is, 2t/10 = 3. Solving
we find t = 10 log2 3 ≈ 15.8 years.

Radioactive Decay

The mass m(t) of a physical substance at time t, with initial mass m0 = m(0), typically decays
at a rate m0 (t) proportional to m(t), with some proportionality constant k > 0 depending on
the substance. So it satisfies the IVP
m0 (t) = −k m(t) , m(0) = m0 with solution m(t) = m0 e−kt .

Often one knows the half life h of the substance, defined by m(h) = m0 /2 = m0 e−kh , and so k
and h are related by the equation kh = ln 2. It follows that k = ln 2−1/h , and so

m(t) = m0 2−t/h .

Example (Carbon dating of fossils – Willard Libby 1949) Living tissue has 2 isotopes of carbon:
C12 (stable) and C14 (radioactive, with a half life h = 5500 years). When it dies, the C12
remains and C14 decays.
For example if the C14 has decayed to 20% of the original, then the age t of the fossil satisfies
m(t) = m0 2−t/h = m0 /5, so −t/h = log2 1/5 =⇒ t = 5500 log2 5 ≈ 12, 270 years. In general, if
mass m remains, then the fossil is t = h log2 (m0 /m) years old.

† If Q(t) is another solution, then (P/Q)0 = (P 0 Q − P Q0 )/Q2 = ((kP )Q − P (kQ))/Q2 = 0, so P/Q is some
constant C. But P (0)/Q(0) = P0 /P0 = 1, so C = 1, whence Q = P .

15
 Bryn Mawr College

Cooling
Newton’s Law of Cooling says that the temperature T (t) at time t of a hot liquid sitting in a
room cools according to the IVP
T 0 (t) = −k(T (t) − Tr ) , T (0) = T0
for some k > 0 (depending on the liquid), where Tr is the room temperature. The solution

T (t) = Tr + (T0 − Tr )e−kt

is obtained as follows: Normalize the scale so the room temperature is zero, i.e. set Te(t) =
T (t) − Tr . The equations become Te0 (t) = −k Te(t) , Te(0) = T0 − Tr . Solve this as above:
Te(t) = (T0 − Tr )e−kt , and then add back in Tr .

Example (Coffee problem) If you pour your cup now, but plan to drink it later, should you add
milk now or later for a hotter cup?
Solution Let C and M be the temperature now of the coffee and the milk (in the frig), and p be
the portion the total liquid that is milk. For simplicity suppose Tr = 0, so C > 0 and M < 0.
The temperatures in the two scenarios are
• (add now) T n = T0 e−kt = (pM + (1 − p)C)e−kt
• (add later) T ` = pM + (1 − p)Ce−kt .
Subtracting we see that T n − T ` = pM (e−kt − 1) > 0, since both pM and e−kt − 1 are negative.
So add the milk now!

Compound Interest
An investment of P0 dollars at interest rate r (i.e. 100r percent per year) compounded continu-
ously obeys the IVP
P 0 (t) = r P (t) , P (0) = P0 with solution P (t) = P0 ert .
Alternate derivation: With yearly compounding, P0 will grow to P0 (1 + r)t in t years. If com-
pounded n times/yr, it’ll grow to P0 (1 + r/n)nt , and if compounded continuously, to
P0 lim (1 + r/n)nt = P0 lim (1 + 1/(n/r))(n/r)rt = P0 ert .
n→∞ n→∞

Example How long will it take for an investment of $1000 invested at an interest rate of .05
(i.e. r = .05, compounded continuously) to triple? What is the equivalent annual interest rate
during that time, i.e. at what rate s, compounded annually, would the investment have tripled
in that same amount of time?
Solution For the investment to triple, we must find t for which 1000e.05t = 3000, so t = 20 ln 3 ≈
22 years. The equivalent annual rate s satisfies 1000(1 + s)t = 3000, so (1 + s)t = e.05t , or
s = e.05 − 1 ≈ 5.13%.

A slightly harder problem (food for thought)

The snow plow problem: It started snowing one morning, and continued to snow steadily all
day. A snowplow, working steadily (i.e. removing a fixed volume of snow each hour) plowed 2
miles of the road between noon and 1 PM, and 1 more mile by 2 PM. At what time did it start
snowing?

16
 Calculus II

6.1 Integration by Parts

Recall the product 0 0 0

R 0 rule
R (F G) = F G + F G . Integrating (i.e. antidifferentiating) both sides
yields F G = F G + F G, or rearranging terms:
Z Z
F G = F G − F 0 G.
0

This is integration by parts, or “the backwards product rule”. In applying this method, one
must decide how to split the integrand as a product F G0 . The goal is to choose F and G0 so
that F 0 G is simpler to integrate than F G0 .

Remark Here are two other ways to view this technique:

• Write the integrand as a product F g of two functions F and g (the “parts”) where F is easy
to differentiate and g is easy to integrate. Then setting f = F 0 and G = g,
R
Z Z
F g = F G − f G.

In words, the integral of F times g is (F times the integral of g) minus (the integral of (the
derivative of F times the integral of g)).
R
• Using the standard notation for integrals using the differential dx, write the integral as udv
for some u = u(x), v = v(x), where dv = v 0 (x)dx as usual. Then
Z Z
u dv = uv − v du.

R
Examples Compute h(x) dx where h(x) =

1k xex (F = x, G0 = ex ) 4k sin2 x (F = sin x, G0 = sin x)

2k ln x (F = ln x, G0 = 1) 5k sec3 x (F = sec x, G0 = sec2 x)
3k x2 sin x (F = x2 , G0 = sin x)† 6k ex sin 2x (F = sin 2x, G0 = ex )

1kbecomes xex − ex dx = ex (x − 1) , 2kbecomes x ln x − 1 dx = x(ln x − 1)

R R

3kbecomes −x2 cos x + 2R x cos x dx. Integrating by parts again with F = x, G0 = cos x leads
R

to −x2 cos x + 2x sin x − 2 sin x dx = (2 − x2 ) cos x + 2x sin x.

4k becomes − sinRx cos x + cos2 x dx = − sin xRcos x + (1 − sin2 x) dx and so rearranging
R R

terms we see that 2 sin2 x dx = x − sin x cos x, so sin2 x dx = 12 (x − sin x cos x).
5k becomes x tan x − sec x tanR2 x dx = sec x tan x − (sec3 x − sec x) dx. RRearranging
R R
R sec
terms gives 2 sec3 x dx = sec x tan x+ sec x dx = sec x tan x+ln | sec x+tan x|, so sec3 x dx =
1
2 (sec x tan x + ln | sec x + tan x|).

6k becomes ex sin 2x − 2R ex cos 2x dx = ex sin 2x − 2ex cos 2xe x − 4 ex sin 2x dx.

R R

Now, rearranging terms gives 5ex sin 2x dx = ex (sin 2x−2 cos 2x), so ex sin 2x dx = 51 ex (sin 2x−
R
2 cos 2x).

† Apply the method twice

17
 Bryn Mawr College

Reduction Formulas Examples:

1k In := xn ex dx = xn ex − n xn−1 ex dx (using F = xn , G0 = ex ). Thus

R R

In = xn ex − nIn−1 .
Apply this repeatedly, noting that I0 = ex , to compute In . For example
Z
I3 = x3 ex dx = x3 ex − 3I2 = x3 ex − 3(x2 ex − 2I1 )

= x3 ex − 3(x2 ex − 2(xex − I0 )) = ex (x3 − 3x2 + 6x − 6).

2k Here is another way to think about this iterative process: For any function F and any
integer n ≥ 0, let F (n) denote the nth derivative of F , and F (−n) denote the nth integral of
F . In particular F (0) = F . Now given two functions F and G, arrange their derivatives and
integrals in two adjacent rows, as shown:
· · · F (−2) F (−1) F (0) F (1) F (2) · · ·
··· G(2) G(1) G(0) G(−1) G(−2) · · ·

Set In = F (n) G(1−n) , so in particular I0 = F G0 and I1 = F 0 G. The integration by parts

R R R

formula implies that In = F (n) G(−n) − In+1 for every n. Starting with I0 , and repeating, then
yields the formula I0 = F (0) G(0) − F (1) G(−1) + F (2) G(−2) − + · · · , that is:
Z
F G0 = F G − F 0 ( G) + F 00 ( G) − F 000 ( G) + · · ·
R RR RRR

Note that the series ends if F is a polynomial (so some derivative of F is zero). This gives a
quick alternative derivation of the integral x3 ex dx in 1k.
R

3k In := sinn = − sinn−1 cos +(n−1) sinn−2 cos2 (using F = sinn−1 , G0 = sin). Substituting
R R

1 − sin2 for cos2 and rearranging gives nIn = − sinn−1 cos +(n − 1)In−2 , so
1
In = n ((n − 1)In−2 − sinn−1 x cos x) .

Since I0 = x and I1 = − cos x, this allows R one to compute In for all n. For example I2 =
sin2 x dx = 21 (x − sin x cos x), and I3 = sin3 x dx = 13 (−2 cos x − sin2 x cos x)
R

6.2 Trigonometric Substitution

Integrals involving square roots of quadratic expressions in x often succumb to a trig substitution,
in which x is taken to be a trig function of the new variable u (so effectively we are substituting
u = an inverse trig function of x).
Here’s the scheme:

Expression Before Substitution Expression After

a2 − x2 x = a sin u, dx = a cos u du a2 cos2 u
a2 + x2 x = a tan u, dx = a sec2 u du a2 sec2 u
x2 − a2 x = a sec u, dx = a sec u tan u du a2 tan2 u

18
 Calculus II
R
Examples Compute f (x) dx where f (x) =
√
1k 1/ 4 − x2 5k 1/(9x2 + 4)5/2 cos3
R R
1
√ √
2k x/ 4 − x2 6k 1 + x2 sec3
R R
sin
√ √
3k x2 / 4 − x2 sin2 7k x2 25 − x2 sin2 cos2
R R
√ √
4k 1/ 9x2 + 4 8k x3 25 − x2 sin3 cos2
R R
sec

Integrating trig expressions

• cosp x
a) p odd : substitute u = sin x, cos2 = 1 − u2 polynomial in u
1
b) p even : apply trig identity cos2 x = 2 (1 + cos 2x) repeatedly / or use parts

Similarly for sinq x, q odd (u = cos x); q even (use sin2 x = 21 (1 − cos 2x) or parts)
• secp x
a) p even : substitute u = tan x, sec2 = 1 + u2 polynomial in u
b) p odd : parts
• cosp x sinq x
a) p odd : substitute u = sin x as above (similarly for q odd, substitute u = cos x)
b) p and q even : apply trig identity cos2 x = 21 (1 + cos 2x) repeatedly / or parts
• secp x tanq x
a) p even : u = tan x
b) q odd : u = sec x tan x

Examples 1k sin2 x dx 2k sin3 x dx 3k sin2 x cos2 x dx 4k sin3 x cos2 x dx

R R R R

The last resort (Weierstrass, mid 19th century)

To integrate any rational function of sin x and cos x, substitute
2
u = tan(x/2) =⇒ x = 2 tan−1 u , dx = du.
1 + u2
Then
2 tan(x/2) 2u
sin x = 2 sin(x/2) cos(x/2) = 2
=
sec (x/2) 1 + u2
1 − tan2 (x/2) 1 − u2
cos x = cos2 (x/2) − sin2 (x/2) = =
sec2 (x/2) 1 + u2
which leads to a rational function of u (discussed in the next section).
Z Z Z
1 1 2
Example dx = 1−u2
du = du = u = tan(x/2).†
1 + cos x 1+ 1 + u2
1+u2

† Note that the derivative of tan(x/2) = 1

sec2 (x/2). To see why this is equal to 1/(1 + cos x), use the half
2
2
angle trig identity: cos x = 2 cos (x/2) − 1.

19
 Bryn Mawr College

6.3 Integration of Rational Functions

A rational function is a quotient of polynomials.† For example

x+1 x3 + 2x + 1
f (x) = and g(x) =
x3 − 2x2 + x x2 + 1
√
are rational functions, while h(x) = x2 + 1 and sin(x2 − 1) are not. A rational function is
called proper if its numerator has a smaller degree (meaning its highest exponent) than its
denominator. Thus f is proper while g is not.
Goal: Learn how to integrate all rational functions – assuming we can first carry out the
appropriate algebra of factoring and dividing polynomials, and of decomposing rational functions
into simpler ones (see below).

Facts 1k Every polynomial can be factored into quadratic and linear polynomials.
2k Every improper rational function f (x) = n(x)/d(x) is a sum of a polynomial and a proper
rational function: Divide n(x) by d(x) to get a quotient q(x), with remainder r(x) satisfying
deg r < deg d. Then f = q + r/d, and r/d is proper.
3kpartial fractions Every proper rational function r(x)/d(x) is a sum of rational functions of
the form p(x)/q(x)k with deg p < deg q ≤ 2. Here the q(x)’s are the factors of the denominator
d(x), and k can take on any value ≤ the multiplicity of q(x) in d(x).

We assume you know how to do 1k(which is hard in general) and 2k(which is easy), and we’ll
learn through examples how to do 3k(after factoring both numerator and denominator using
1k). Since we know how to integrate polynomials, this reduces the problem of integrating any
rational function to the following special cases:
1 1 2x
ak k
bk 2 k
bk 2
(x − a) (x + bx + c) (x + bx + c)k

But akis easy (use the power rule, except when k = 1 where the integral is ln |x − a|) and ck
reduces to bkby the following trick:
2x 2x + b − b 2x + b 1 0 1
= = − b = −b k
k k k k k 
(substitute u =  = x2 + bx + c to integrate the next to last term). So that leaves bk, i.e.
integrating 1/(x2 + bx + c)k . This is accomplished in two steps:
• Complete the square in the denominator: x2 + bx +pc = (x + b/2)2 + (c − b2 /4). (Advice:
learn this technique!) Substituting u = x + b/2 and a = c − b2 /4 yields the integral
Z
du
(u ± a2 )k
2

which, after trig substitution, succumbs (painfully). It is worth memorizing the final formulas
for the case k = 1:
u−a
Z Z
du 1 −1 u du 1
= tan ( ) and = ln
u2 + a2 a a u2 − a2 2a u+a
These can be verified by differentiating the right hand sides, or derived by integrating using the
substitution u = a tan θ in the first and u = a sec θ in the second.
† We assume all the our polynomials a + a x + a x2 + · · · are real, meaning the a ’s are all real numbers.
0 1 2 i

20
 Calculus II

Partial Fraction Decompositions (PFDs)

Examples
1 1 a b
1k = = +
x2 −4 (x − 2)(x + 2) x−2 x+2
1 a(x + 2) + b(x − 2)
To find a and b, add the fractions on the right to get =
−4 x2 x2 − 4
Equate the numerators, a(x + 2) + b(x − 2) = 1, and then the coefficients of x and 1:
(
a+b =0
a = −b = 1/4
2a − 2b = 1
   
x−2
Z
1 1 1 1 dx 1 †
Therefore 2
= − 2
= ln
x −4 4 x−2 x+2 x −4 4 x+2
x x a b 2 1
2k = = + = ··· = −
x2 − 2x + 2 (x − 2)(x − 1) x−2 x−1 x−2 x−1

Short cut : the Heaviside method If (x−a)n is the largest power of x−a in the denominator
of a rational function f , then the coefficient of 1/(x−a)n in the PFD can be obtained by plugging
a into f (x)/(x − a)n :
p(x) p(a)/q(a)
f (x) = n
= + ··· (where p(a) and q(a) are both nonzero)
(x − a) q(x) (x − a)n
Why? Multiply by (x − a)n and then take the limit as x → a. This gives a quick way of redoing
the two examples above. For example in 2k, a = x/(x − 1)|x=2 = 2/1 = 2

More Examples
x+1 x+1 a b c 1 1 2
3k = = + + = ··· = − +
x3 2
− 2x + x x(x − 1)2 x x − 1 (x − 1)2 x x − 1 (x − 1)2
Note that Heaviside readily gives a and c, but not b.

4k Here’s an example with an irreducible quadratic factor in the denominator:

3 3 a bx + c 1 6 − 3x
= = + 2 = ··· = + 2
x3 +1 2
(x + 1)(x − x + 1) x+1 x −x+1 x+1 x −x+1

Note that Heaviside only gives us a; we find b and c by solving a system of equations.
Once we have computed these PFDs, we can (for example using 4k) compute
Z
3 1 √ 2x − 1
dx = ln |x + 1| − ln |x 2
− x − 1| + 3 tan−1 √ .
x3 + 1 2 3
The details are left to the reader!

† Try using this technique to give an alternative derivation of the last formula on page 20.

21
 Bryn Mawr College

5.8 L’Hôpital’s Rule (for evaluating limits of quotients)

Recall that limx→a f (x)/g(x) can be computed as limx→a f (x)/ limx→a g(x) provided the limit
in the denominator is nonzero. If it is zero but the limit in the numerator is nonzero, then the
original limit is ±∞ (assuming that f is continuous and nonzero near a).
But what if both limx→a f (x) and limx→a g(x) are zero (the “0/0 case”), or both are infinite
(the “∞/∞” case). This is when L’Hôpital’s Rule applies, asserting that
f (x) f 0 (x)
lim = lim 0
x→a g(x) x→a g (x)

provided f and g are differentiable and the right hand limit exists.†

Remarks
1kL’Hôpital’s Rule applies when a = ±∞ (analyzing the “asymptotic behavior” of f /g)
2kOne must always be in either a 0/0 or ∞/∞ case to apply the rule. For example
x2 2x
lim6= lim = 2
x→1 x x→1 1
since we’re in a “1/1 case”. In fact the left hand limit is 1 (always first try just plugging in when
investigating a limit).
3kThe rule can also be used to analyze limits of products f (x)h(x) in the “0 · ∞” case: Write
f · h as f /(1/h) or h/(1/f ), and apply the 0/0 or ∞/∞ case of L’Hôpital’s Rule.

Examples
sin x cos x
1k lim = lim = 1 (This is actually a circular reasoning, since we needed this
x→0 x 0/0 x→0 1
limit to show sin0 = cos)
x3 − 1 3x2
2k lim = lim = 3 (This can also be computed by factoring the numerator
x→1 x − 1 0/0 x→1 1
x3 − 1 = (x − 1)(x2 + x + 1), cancelling the (x − 1)’s, and then plugging in x = 1)
x 1
3k lim = lim = 0 (the last limit is a “1/∞” case)
x→∞ ex ∞/∞ x→∞ ex
x2 2x 2
4k lim = lim = lim = 0 (the rule is applied twice)
x→∞ ex ∞/∞ x→∞ ex ∞/∞ x→∞ ex

1 − cos x sin x cos x 1

5k lim = lim = lim =
x→0 x sin x 0/0 x→0 sin x + x cos x 0/0 x→0 2 cos x − sin x 2
ln x 1/x
6k lim x ln x = lim = lim = lim −x = 0
x→0 0·∞ x→0 1/x 0/0 x→0 −1/x2 x→0

† The proof of L’Hôpital’s Rule (in the 0/0 case, assuming f 0 and g 0 are continuous) goes roughly as follows:
We can assume f (a) = g(a) = 0 without affecting the limits. Then
f (x) − f (a) f (x) − f (a)
lim
f (x) x − a x→a x−a f 0 (a) f 0 (x)
lim = lim = = 0 = lim 0
x→a g(x) x→a g(x) − g(a) g(x) − g(a) g (a) z→a g (x)
lim
x−a x→a x−a

22
 Calculus II

6.6 Improper Integrals

These are definite integrals

Z b
f (x) dx
a
for which one or both of the bounds a, b are infinite, and/or for which f is ‘bad’ at finitely
many points in [a, b] (meaning f (x) is unbounded as x approaches those points). Here, we only
consider the following three cases:
• One infinite bound Then define
Z ∞ Z t Z b Z b
f (x) dx = lim f (x) dx and f (x) dx = lim f (x) dx
a t→∞ a −∞ t→−∞ t

If the limit exists, we say that the improper integral converges, and otherwise it diverges.
• Both bounds are infinite Then define
Z ∞ Z 0 Z ∞
†
f (x) dx = f (x) dx + f (x) dx
−∞ −∞ 0
provided both improper integrals on the right converge, in which case we say that the original
integral converges; otherwise it diverges.
• Finite bounds, but one of them is ‘bad’ (meaning f (x) is unbounded as x approaches the
bad endpoint from within the interval [a, b]). For example, if b is bad, then define
Z b Z t
f (x) dx = lim f (x) dx
a t↑b a

Examples (sketch the integrands to interpret the results in terms of areas)

Z ∞ Z ∞
dx dx
1k
t ∞
= lim (ln x 1 ) = lim ln t = ∞ (abbreviated = ln x 1 = ∞)
1 x t→∞ t→∞ 1 x
Z ∞
dx ∞
so the integral diverges, whereas 2
converges to (−1/x) 1 = 0 − (−1) = 1
1 x
Z ∞ ∞
dx π π
2k 2
= tan−1 (x) = −0 = ; it follows that
0 x +1 0 2 2
Z ∞ Z 0 Z ∞ Z ∞
dx dx dx dx π π
2+1
= 2+1
+ 2+1
= 2 2+1
dx = + = π
−∞ x −∞ x 0 x 0 x 2 2
Z ∞
3k 2 2 0 2 ∞
2xe1−x dx = (−e1−x ) −∞ + (−e1−x ) 0 = (−e − 0) + (0 − (−e)) = 0
−∞
1 1
x−1
Z
dx
4k 2
= ln = ln 0 − ln 1 = −∞, so the integral diverges
0 x −1 x+1 0

1 Z
dx
More generally both endpoints may be ‘bad’, e.g. 2
, or we may have internal ‘bad’
−1 x − 1
Z 2
dx
points, e.g. 2
at x = 1. Note: both of these integrals diverge.
0 x −1

† Note: You can replace 0 by any number a in the definition, i.e. R ∞ f = R a f + R ∞ f

−∞ −∞ a

23
 Bryn Mawr College

7.2–7.3 Computing Volumes

Solids of Revolution
Consider a solid S formed by rotating a domain D in the plane about a line L that meets the
domain – if at all – along a portion of its boundary. It may be difficult to draw the entire solid
S, but not so hard to sketch its intersections with planes perpendicular to L, called slices of S,
or with cylinders centered around L, called shells. We would like to express the volume V of S
as an integral in a suitable variable t.

Method of Slices
Choose the t-axis parallel to L. If a and b are the bounds of D, then
Z b
V = A(t) dt (a slice integral)
a
where A(t) is the area of the slice at level t. These slices will either be disks where R meets
L, or annuli (or unions of annuli) where R is disjoint from L, as shown below. Thus A(t) will
typically be of the form πr2 or π(R2 − r2 ), where r and R must be expressed in terms of t (often
a sketch will help in accomplishing this).

D
R
r
r
L

A(t) = πr2
A(t) = π(R2 − r2 )

a t t b

Examples
1k Let D be bounded by the x-axis, the line x = 2 and the curve y = x2 . Find the volume of
the solid obtained by rotating D about
Z 2
2
a) the x-axis: V = π(x2 )2 dx = (πx5 /5) 0 = 32π/5 4

0 Z 2
b) the line y = −1: V = π((1 + x2 )2 − 12 )dx y = x2
0 y
R2 √
= 0 π(2x2 + x4 ) dx = π(16/3 + 32/5) = 176π/15 2− y

√
c) the y-axis (note that the curve can be written as x = y): x2
Z 4
√ 4 x 2
V = π(22 − y 2 ) dy = π(4y − y 2 /2) 0 = 8π
0
(sketches for parts a & d)
4 4
√ √
Z Z
2
d) the line x = 2: V = π(2 − y) dy = π(4 − 4 y + y) dy = 8π/3
0 0

24
 Calculus II

2k Set up slice integrals (but do not evaluate) to compute the volumes of solids obtained by
rotating the domains bounded by
a) the x-axis and the graph of y = cos x for x in [0, π/2], about both axes
Z π/2 Z 1
x-axis: (cos x)2 dx y-axis: π(cos−1 x)2 dy
0 0
b) y = 2x and y = x2 ,about both axes
Z 2 Z 1
x-axis: π(x2 − x4 ) dx y-axis: π(y − y 2 ) dy
0 0
c) the x-axis, x = h and y = rx/h, about the x-axis (this is an ice cream cone)
Z h  
rx 2 πr2 h
π dx (= )
0 h 3
d) the circle of radius 1 about the point (2, 0), about the y-axis (this is a donut)
Z 1  p p  Z 1p
π (2 + 1 − y 2 )2 − (2 − 1 − y 2 )2 dy = 8π 1 − y 2 dy ( = 8π 2 )
−1 −1

Method of Shells
Choose the t-axis perpendicular to L. If a and b are the bounds of R, then
Z b
V⊥ = A⊥ (t) dt (a shell integral)
a
where A⊥ (t) is the area of the cylinder at level t, as shown below, and so will typically be of the
form 2πrh where r and y must be computed in terms of t as before:

r A⊥ (t) = 2πrh

at b

Examples
1k Repeat the problems in example 1kabove, but now using shells. The integrals are:
Z 4 Z 4
√ √
a) V⊥ = 2πy(2 − y) dy = 32π/5 b) V⊥ = 2π(1 + y)(2 − y) dy = 176π/15
0 0
Z 2 Z 2
c) V⊥ = 2πx · x2 dx = 8π d) V⊥ = 2π(2 − x)x2 dx = 8π/3
0 0

2k Should slices or shells be used to compute the volume of the solid obtained by rotating
the domain bounded by y = 2x − x2 , and the lines x = 0, x = 2 and y = 3, about the x-axis?
(Answer: slices) What about the y-axis? (Answer: shells) Do you see why?

25
 Bryn Mawr College

7.4, 9.2 Computing Arc Lengths and Surface Areas

Motivating discussion under construction . . .

Key observation: the ‘infinitessimal’ Pythogorean theorem shows that the ‘element’ ds of arc
length is
p
ds = dx2 + dy 2 .
For the graph of a function y = f (x) we would like to express this in terms of x, and so we
multiply and divide by dx :
r
dx2 + dy 2 p
ds = dx = 1 + f 0 (x)2 dx.
dx2
For a parametrized curve x = x(t), y = y(t), we multiply and divide by dt :
r
dx2 + dy 2 p
ds = dt = x0 (t)2 + y 0 (t)2 dt.
dt2
Z L
This leads to a formula for the length L = ds of the graph of y = f (x) for a ≤ x ≤ b :
0
Z bp
L = 1 + f 0 (x)2 dx
a
and the length of the parametrized curve x = x(t), y = y(t) for t0 ≤ t ≤ t1 :
Z t1 p
L = x0 (t)2 + y 0 (t)2 dt.
t0

Examples ...

7.5 Computing Surface Areas

Building on the arc length formulas, one can derive formulas for the areas of surfaces of revo-
lution. For example, the area A of the surface obtained by rotating the graph of y = f (x) for
a ≤ x ≤ b about the x-axis is given by
Z b p
A = 2πf (x) 1 + f 0 (x)2 dx
a
since each x in [a, b] corresponds to a circle on the surface of length 2πr = 2πf (x) (because
we are rotating about the x-axis), and when it is thickened p infinitessimally to dx, this circle
becomes a infinitessimal ‘slanted cylinder’ of width ds = 1 + f 0 (x)2 dx. If we were rotating
around the line y = −1 instead, then r would become
p f (x) + 1 while ds would remain the same,
so the area would be the integral of 2π(f (x) + 1) 1 + f 0 (x)2 .
Similar formulas can be written down for parametric curves. The principle is always the same.
The area is given by an integral of the form
Z ∗
A = 2πrds
∗
where r and ds are expressed in terms of the appropriate variable.

Examples . . .

26
 Calculus II

7.7 Differential Equations

Definition An ordinary differential equation, abbreviated ODE, is an equation in an indepen-

dent variable x, a dependent variable y, and the derivatives y 0 , y 00 , . . . of y with respect to x.
Thus a general ODE can be written in the form
F (x, y, y 0 , y 00 , . . . ) = 0
for some function F . The highest order of a derivative that appears is called the order of the
ODE. We will limit our discussion to first order equations F (x, y, y 0 ) = 0 that that can be
rewritten in the special form
(1) y 0 = f (x, y).
(excluding equations such as y 0 + sin(y 0 ) = xy). A solution to (1) is a function y = y(x) that
satisfies (1), i.e. y 0 (x) = f (x, y(x)).

Examples . . .
Equation (1) is called separable if f (x, y) is of the form h(x)/g(x), so it becomes
(2) g(y) y 0 = h(x).
To solve this we go through the following formal procedure: Substitute dy/dx for y 0 , then
multiply
R by Rdx on both sides . . . giving g(y)dy = h(x)dx . . . and finally integrate both sides:
g(y) dy = h(x) dx. Thus the solutions to (2) are exactly the solutions to
(3) G(y) = H(x) + C
where G and H are any chosen antiderivatives of g and h.† Do not forget to include the + C;
it is essential in what follows. To complete the solution to (2), one then tries to rewrite (3) so
as to solve for y explicitly in terms of x.

Examples . . .

Remark Not all first order ODEs are separable, in fact ‘most’ aren’t. Furthermore there is
no general method known for solving first order ODEs – even the ones in the special form (1)
– although solutions always exist in theory – see below. There are, however, some standard
techniques for attacking non-separable equations, e.g. the method of integrating factors. These
are treated in any introductory course on differential equations.
In applications, one often imposes an initial condition y = y0 when x = 0 (or more generally
y = y0 when x = x0 ), i.e. one seeks a solution Y = f (x) for which f (0) = y0 . The resulting set
up is called a first order initial value problem (IVP for short) :
(
y 0 = f (x, y)
(4)
y = y0 when x = x0
The initial condition allows us to solve for the unknown constant C, and thus to pin down a
unique solution.

Theorem Existence and Uniqueness Theorem (no proof given here) Under mild conditions on
the function f (which we don’t specify here), the IVP in (4) has a unique solution.
† This is rigorously proved as follows: y satisfies (3) ⇐⇒ G(y) = H(x) + C ⇐⇒ G0 (y)y 0 = H 0 (x) (since
functions have the same derivative if and only if they differ by a constant, by the Mean Value Theorem) ⇐⇒
g(y)y 0 = h(x) ⇐⇒ y satisfies (2).

27
 Bryn Mawr College

Example 1kSolve the IVP

(
y 0 = 2xy 2
y = 1 when x = 0 .
Separating variables we arrive at
−1
Z Z
1
dy = 2x dx which implies = x2 + C.
y2 y
Substituting in the initial condition gives −1 = 0 + C, so C = −1. Thus
−1 1
y = 2 = .
x −1 1 − x2

2k(The logistic equation) Consider a population P (t) constrained (by the environment or some
other factor) to a maximum carrying capacity of M . This is modeled by the IVP
(
P = kP (M − P )
P = P0 when t = 0 .
Separating variables, . . .

8.1 Sequences

A sequence is an ordered, infinite list of numbers: s1 , s2 , s3 , . . . indexed by the natural numbers,

or more formally, a function s : N → R, where sn = s(n). We say that sn converges to a number
s, or that s is the limit of the sequence sn , written
lim sn = s , or simply sn −→ s
n→∞
if we can make sn as close to s as we want by choosing n sufficiently large. More precisely, this
means that for every ε > 0, there is an N such that |sn − s| < ε for all n > N .
If sn does not converge, we say that it diverges, which can happen in a variety of ways. It can
‘diverge to ∞’ or ‘to −∞’, or ‘by oscillation’, or in more unpredictable ways.

Examples 1k1, 12 , 31 , 14 , . . . , n1 , . . . : sn converges to 0 (also written sn −→ 0)

2k−1, 1, −1, 1, . . . , (−1)n , . . . : sn diverges (by oscillation)
3k1, 4, 9, 16, . . . , n2 , . . . : sn diverges to ∞ (also written sn −→ ∞, by abuse of notation)
4kThe prime sequence : 2, 3, 5, 7, 11, 13, . . . , pn , . . . −→ ∞ (a formula for pn is unknown)
5kThe Fibonacci sequence : 1, 1, 2, 3, 5, 8, . . . , Fn , . . . −→ ∞ ; Do you see the pattern? Each
term is the sum of the previous two. There is a famous formula (called Binet’s formula) for the
nth Fibonacci number: √
n − rn
r+ − 1± 5
Fn = √ where r± = .
5 2
The number r+ is called the golden ratio.
Just as for limits of functions, limits of sequences obey the:

Limit laws If an −→ a and bn −→ b, then

• an ± bn −→ a ± b • an bn −→ ab • an /bn −→ a/b provided b 6= 0

28
 Calculus II

Remark If f : R → R is a function with f (n) = sn and limx→∞ f (x) = s, then sn −→ s. Thus

we can sometimes use calculus to compute limits of sequences. For example
4n2 + 5n − 1 4n2 + 5n − 1 8n + 5 8
2
−→ 2 since lim 2
= lim = lim = 2
2n + 9 x→∞ 2n + 9 L’Hop x→∞ 4n L’Hop x→∞ 4
This limit can also be computed by the trick of dividing numerator and denominator by the
largest power of n that appears, in this case n2 . This yields the expression:
4 + 5/n − 1/n2
2 + 9/n2
Since all the terms involving n go to zero, it follows from the basic fact that the limit of a below
that the limit is 4/2 = 2

DID NOT TEX UP NOTES FROM THE REST OF THE SEMESTER ...

29