18.

01A SUPPLEMENTARY NOTES,

EXERCISES, AND SOLUTIONS

Table of Contents

Notes

G. Graphing Functions
C. Continuity and Discontinuity
X. Exponentials and Logarithms
A. Approximation
MVT. Mean-Value Theorem
PI. Properties of Definite Integrals
FT. The Second Fundamental Theorem of Calculus
AV. Average Value
F. Heaviside’s Cover-up Method
INT. Improper Integrals
P. Probability

Exercises, Problems, and Solutions

E. Exercises Sections 1-8

(Starred exercises are not solved in section S.)
S. Solutions to Exercises

RP. Review Problems and Solutions RP1-RP5

Exercises and Solutions by David Jerison

Notes and Exercise Sections 7,8 by Arthur Mattuck
c
!M.I.T. and David Jerison 1996, 2006
 G. GRAPHING FUNCTIONS

To get a quick insight int o how the graph of a function looks, it is very helpful to know
how certain simple operations on the graph are related to the way the function expression
looks. We consider these here.

1. Right-left translation.
Let c > 0. Start with the graph of some function f (x). Keep the x-axis and y-axis fixed,
but move the graph c units to the right, or c units to the left. (See the pictures below.) You
get the graphs of two new functions:
right f (x − c)
! !
(1) Moving the f (x) graph c units to the gives the graph of .
left f (x + c)
If f (x) is given by a formula in x, then f (x − c) is the function obtained by replacing x by
x − c wherever it occurs in the formula. For instance,
f (x) = x2 + x ⇒ f (x − 1) = (x − 1)2 + (x − 1) = x2 − x, by algebra.

Example 1. Sketch the graph of f (x) = x2 − 2x + 1. 2

2
x2 (x 1)
Solution. By algebra, f (x) = (x − 1) . Therefore by (1), its graph is
the one obtained by moving the graph of x2 one unit to the right, as shown.
The result is a parabola touching the x-axis at x = 1. 1

To see the reason for the rule (1), suppose the graph of f (x) is moved c units to the right:
it becomes then the graph of a new function g(x), whose relation to f (x) is described by
(see the picture):

value of g(x) at x0 = value of f (x) at x0 − c = f (x0 − c) .

h(x) f(x) g(x)
This shows that g(x) = f (x − c). The reasoning is similar if the
graph is translated c units to the left. Try giving the argument c c
xo c x
yourself while referring to the picture. xo

The effect of up-down translation of the graph is much simpler to see. If c > 0,
up f (x) + c
! !
(2) Moving the f (x) graph c units gives the graph of .
down f (x) − c
since for example moving the graph up by c units has the effect of adding c units to each
function value, and therefore gives us the graph of the function f (x) + c
√
Example 2. Sketch the graph of 1 + x − 1. 1+ x 1
√ x
Solution Combine rules (1) and (2). First sketch √ x, then 1
move its graph 1 unit to the right √
to get the graph of x − 1, then
1
1 unit up to get the graph of 1 + x − 1, as shown.
1
2

Example 3. Sketch the curve y = x2 + 4x + 1.

Solution We “complete the square”:

x2 + 4x + 1 = (x2 + 4x + 4) − 3 = (x + 2)2 − 3 , 2

so we move the graph of x2 to the left 2 units, then 3 units down,

getting the graph shown. 3

2. Changing scale: stretching and shrinking.

Let c > 1. To stretch the x-axis by the factor c means to move the point 1 to the position
formerly occupied by c, and in general, the point x0 to the position formerly occupied by cx0 .
Similarly, to shrink the x-axis by the factor c means to move x0 to the position previously
taken by x0 /c. What happens to the graph of f (x) when we stretch or shrink the x-axis?

Stretching f (x/c)
! !
(3) the x-axis by c changes the graph of f (x) into that of .
Shrinking f (cx)
f( x ) f( x c)
The picture explains this rule; it illustrates stretching by the factor
c > 1. The new function has the same value at x0 that f (x) has
at x0 /c, so that it is given by the rule x0 → f (x0 /c), which means
x0 c x0
that it is the function f (x/c).
If the y-axis is stretched by the factor c > 1, each y-value is multiplied by c, so the new
graph is that of the function cf (x):

Stretching c f (x)
! !
(4) the y-axis by c changes the graph of f (x) into that of .
Shrinking f (x)/c

1
Example 4. Sketch the graph of .
2x − 1
Solution. Start with the graph of 1/x, move it 1 unit to the 1/2

right to get the graph of 1/(x − 1), then shrink the x-axis by the
factor 2 to get the graph of the given function. See the picture.

3. Reflecting in the x- and y -axes: even and odd functions.

To reflect the graph of f (x) in the y-axis, just flip the plane over around f( x )
f (- x)
the y-axis. This carries the point (x, y) into the point (−x, y), and the graph
of f (x) into the graph of f (−x). Namely, the new function has the same
y-value at x0 as f (x) has at −x0 , so it is given by the rule x0 → f (−x0 )
- x0 x0
and is the function f (−x).
Similarly, reflecting the xy-plane in the x-axis carries (x, y) to the point (x, −y) and the
graph of f (x) gets carried into that of −f (x).
Finally, relecting first in the y-axis and then in the x-axis carries the f(x)
point (x, y) into the point (−x, −y). This is called a reflection through
the origin. The graph of f (x) gets carried into the graph of −f (−x), by
-f(-x)
combining the above two results. Summarizing:
 G. GRAPHING FUNCTIONS 3

 
 y-axis
  f (−x)

(5) Reflecting in the x-axis moves the graph of f (x) into that of −f (x)
 
origin −f (−x).
 

Of importance are those functions f (x) whose graphs are symmetric with respect to the
y-axis — that is, reflection in the y-axis doesn’t change the graph; such functions are called
even. Functions whose graphs are symmetric with respect to the origin are called odd. In
terms of their expression in x,

(6) f (−x) = f (x) definition of even function

(7) f (−x) = −f (x) definition of odd function

Example 5. Show that a polynomial with only even powers, like x4 − 2x2 + 7, is an
even function, and a polynomial with only odd powers, like 3x5 − x3 + 2x, is an odd function
— this, by the way, explains the terminology “even” and “odd” used for functions.
Solution. We have to show (6) and (7) hold for polynomials with respectively only even
or odd powers, but this follows immediately from the fact that for any non-negative integer
n, we have ! n
x , if n is even,
(−x)n = (−1)n xn = !
−xn , if n is odd.

The following easily proved rules predict the odd- or even-ness of the product or quotient
of two odd or even functions:

(8) even · even = even odd · odd = even odd · even = odd
(9) even/even = even odd/odd = even odd/even = odd

x3
Example 6. is of the form odd/even, therefore it is odd;
1 − x2
(3 + x4 )1/2 (x − x3 ) has the form even · odd, so it is odd.

4. The trigonometric functions.

The trigonometric functions offer further illustrations of the ideas about translation,
change of scale, and symmetry that we have been discussing. Your book reviews the standard
facts about them in section 9.1, which you should refer to as needed.
The graphs of sin x and cos x are crudely sketched below. (In calculus, the variable x is
always to be in radians; review radian measure in section 9.1 if you have forgotten it. Briefly,
there are 2π radians in a 360o angle, so that for example a right angle is π/2 radians.)
As the graphs suggest and the unit circle picture shows,

(10) cos(−x) = cos x (even function) sin(−x) = − sin x (odd function).

1 ! /2 - x
From the standard triangle at the right, one sees that x

cos(π/2 − x) = sin x ,
 sin x
1
cos x x
4
x
-x
and since cos x is an even function, this shows that cos(- x) -x
sin (- x)
(11) cos(x − π/2) = sin x .
cos x sin x
From (11), we see that moving the graph ! 2!
of cos x to the right by π/2 units turns it
into the graph of sin x. (See picture.)
The trigonometric function
sin x "!
(12) tan x = !
cos x
is also important; its graph is sketched at the right. It is an odd
function, by (9) and (10), since it has the form odd/even.

Periodicity
An important property of the trigonometric functions is that they repeat their values:
(13) sin(x + 2π) = sin x, cos(x + 2π) = cos x .
This is so because x + 2π and x represent in radians the same angle.
From the graphical point of view, equations (13) say that if we move the graph of sin x
or cos x to the left by 2π units, it will coincide with itself.
From the function viewpoint, equations (13) say that sin x and cos x are periodic functions,
with period 2π. In general, let c > 0; we say that f (x) is periodic, with period c, if
(14) f (x + c) = f (x) for all x, and
(14’) c is the smallest positive number for which (14) is true.
By rule (1), the graph of a periodic function having period c coincides with itself when it is
translated c units to the left. If we replace x by x − c in (14), we see that the graph will also
coincide with itself if it is moved to the right by c units. But beware: if a function is made
by combining other periodic functions, you cannot always predict the period. For example,
although it is true that
tan(x + 2π) = tan x and cos2 (x + 2π) = cos2 x ,
the period of both tan x and cos2 x is actually π, as the above figure suggests for tan x.

The general sinusoidal wave.

The graph of sin x is referred to as a “pure wave” or a “sinusoidal oscillation”. We now
consider to what extent we can change how it looks by applying the geometric operations
of translation and scale change discussed earlier.
a) Start with sin x, which has period 2π and oscillates between ±1.
b) Stretch the y axis by the factor A > 0; by (4) this gives A sin x, which has period 2π
and oscillates between ±A.
c) Shrink the x-axis by the factor k > 0; by (3), this gives A sin kx, which has period
2π/k, since
2π
A sin k(x + ) = A sin(kx + 2π) = A sin kx.
k
d) Move the graph φ units to the right; by (1), this gives
 G. GRAPHING FUNCTIONS 5

(15) A sin k(x − φ) , A, k > 0, φ ≥ 0, general sinusoidal wave

A
which has
period 2π/k (the wave repeats itself every 2π/k units); #
angular frequency k (has k complete cycles as x goes from 0 to 2π); ! /k
amplitude A (the wave oscillates between A and −A);
phase angle φ (the midpoint of the wave is at x = φ). -A

Notice that the function (15) depends on three constants: k, A, and φ. We call such
constants parameters; their value determines the shape and position of the wave.
By using trigonometric identities, it is possible to write (15) in another form, which also
has three parameters:

(16) a sin kx + b cos kx

The relation between the parameters in the two forms is:

& b
(17) a = A cos kφ, b = −A sin kφ; A= a2 + b 2 , tan kφ = − .
a

Proof of the equivalence of (15) and (16).

(15) ⇒ (16): from the identity sin(α + β) = sin α cos β + cos α sin β, we get
A sin(k(x − φ)) = A sin(kx − kφ) = A cos kφ sin kx − A sin kφ cos kx
which has the form of (16), with the values for a and b given in (17).
(16) ⇒ (15): square the two equations on the left of (17) and add them; this gives
√
a2 + b2 = A2 (cos2 kφ + sin2 kφ) = A2 , showing that A = a2 + b2 .
If instead we take the ratio of the two equations on the left of (17), we get −b/a = tan kφ,
as promised. !

Example 7. Find the period, frequency, amplitude, and phase angle of the wave
represented by the functions
a) 2 sin(3x − π/6) b) −2 cos(2x − π/2)
Solution.
a) Writing the function in the form (15), we get 2 sin 3(x − π/18), which shows it has
period 2π/3, frequency 3, amplitude 2, and phase angle π/18 (or 10o ).

b) We get rid of the − sign by using − cos x = cos(x − π) — translating the cosine
curve π units to the right is the same as reflecting it in the x-axis (this is the best way to
remember such relations). We get then

−2 cos(2x − π/2) = 2 cos(2x − π/2 − π)

= 2 sin(2x − π), by (11);
= 2 sin 2(x − π/2).
6

Thus the period is π, the frequency 2, the amplitude 2, and the phase angle π/2. (Note
that the first three could have been read off immediately without making the above trans-
formation.)

Example 8. Sketch the curve sin 2x + cos 2x .

Solution Transforming it into the form (15), we can get A and φ by using (19):
√
A= 2; tan 2φ = −1 ⇒ 2φ = 135o = 3π/4, ⇒ φ = 3π/8 .
√ √
So the function is also representable as 2 sin 2(x − 3π/8); it is a wave of amplitude 2,
period π, frequncy 2, and phase angle 3π/8, and can be sketched using this data.

5. Reflection in the diagonal line; inverse functions.

As our final geometric operation on graphs, we consider the effect of reflecting a graph
in the diagonal line y = x. (b,a)
This reflection can be carried out by flipping the plane over about the y=x

diagonal line. Each point of the diagonal stays fixed; the x-and y-axes are a,b)
interchanged. The points (a, b) and (b, a) are interchanged, as the picture
shows, because the two rectangles are interchanged.
To see the effect of this on the function, let’s consider first a simple example.

Example 9. If the graph of f (x) = x2 , x ≥ 0 is reflected in the diagonal, what

function corresponds to the reflected graph?
Solution. The original curve is the graph of the equation: y = x2 , x≥0.
Reflection corresponds to interchanging the two axes; thus the reflected curve is the graph
of the equation: x = y 2 , y ≥ 0 .
To find the corresponding function, we have√to express y explicitly in terms of x, which
we do by solving the equation for y: y = x, x ≥ 0 ; the restriction on x follows
because if x = y 2 and y ≥ 0, then x ≥ 0 also.
y y y

reflect no change

x x x
y=x,2 x>0 x=y 2, y>0 y= x, x>0

Remarks.
1. When we flip the curve about the diagonal line, we do not interchange the labels
on the x- and y-axes. The coordinate axes remain the same — it is only the curve that
is moved (imagine it drawn on an overhead-projector transparency, and the transparency
flipped over). This is analogous to our discussion in section 1 of translation, where the curve
was moved to the right, but the coordinate axes themselves remained unchanged.

2. It was necessary in the previous example to restrict the domain of x in the original
function x2 , so that after being flipped, its graph was still the graph of a function. If we
 G. GRAPHING FUNCTIONS 7

hadn’t, the flipped curve would have been a parabola lying on its side; this is not the graph
of a function, since it has two y-values over each x-value.
√
The function having the reflected graph, y = x, x ≥ 0 is called the inverse func-
tion to the original function y = x2 , x ≥ 0. The general procedure may be represented
schematically by:
y = f (x) −→ x = f (y) −→ y = g(x)
original graph switch x and y reflected graph solve for y reflected graph

In this scheme, the equations x = f (y) and y = g(x) have the same graph; all that has been
done is to transform the equation algebraically, so that y appears as an explicit function
of x. This function g(x) is called the inverse function to f (x) over the given interval; in
general it will be necessary to restrict the domain of f (x) to an interval, so that the reflected
graph will be the graph of a function.
To summarize: f (x) and g(x) are inverse functions if
(i) geometrically, the graphs of f (x) and g(x) are reflections of each other in the diagonal
line y = x;
(ii) analytically, x = f (y) and y = g(x) are equivalent equations, either arising from the
other by solving explicitly for the relevant variable.

1
Example 10. Find the inverse function to , x>1.
x−1
Solution. We introduce a dependent variable y, then interchange x and y, getting

1
x = , y>1.
y−1

We solve this algebraically for y, getting

1 1/(x-1)
(20) y = 1+ , x>0.
x 1+1/x
1
(The domain is restricted because if y > 1, then equation (20) implies that
x > 0.) The right side of (20) is the desired inverse function. The graphs 1
are sketched.

It often happens that in determining the inverse to f (x), the equation

(21) x = f (y)

cannot be solved explicitly in terms of previously known functions. In that case, the corre-
sponding equation

(22) y = g(x)

is viewed as defining the inverse function to f (x), when taken with (21). Once again, care
must be taken to restrict the domain of f (x) as necessary to ensure that the relected will
indeed define a function g(x), i.e., will not be multiple-valued. A typical example is the
following.
8

Example 11. Find the inverse function to sin x.

Solution. Considering its graph, we see that for the reflected graph to define a function,
we have to restrict the domain. The most natural choice is to consider the restricted function

(23) y = sin x, −π/2 ≤ x ≤ π/2.

The inverse function is then denoted sin−1 x, or sometimes Arcsin x; it is defined by the
pair of equivalent equations

(24) x = sin y, −π/2 ≤ x ≤ π/2 ⇐⇒ y = sin−1 x, −1 ≤ x ≤ 1.

The domain [−1, 1] of sin−1 x is evident from the picture — it is the same as the range of
sin x over [−π/2, π/2].
As examples of its values, sin−1 = π/2, since sin π/2 = 1; similarly, sin−1 1/2 = π/6.
Care is needed in handling this function. For example, substituting the left equation in
(24) into the right equation says that

(25) sin−1 (sin y) = y, −π/2 ≤ y ≤ π/2 .

It is common to see the restriction on y carelessly omitted, since the equation by itself seems
“obvious”. But without the restriction, it is not even true; for example if y = π,
sin−1 (sin π) = 0.

! /2

-1
1
y=sin-1 x
! /2

Exercises: Section 1A
 C. CONTINUITY AND
DISCONTINUITY

1. One-sided limits
We begin by expanding the notion of limit to include what are called one-sided limits,
where x approaches a only from one side — the right or the left. The terminology and
notation is:.

right-hand limit lim f (x) (x comes from the right, x > a)

x→a+

left-hand limit lim f (x) (x comes from the left, x < a)

x→a−

Since we use limits informally, a few examples will be enough to indicate the usefulness of
this idea.

f(x)
1/ x 2
x
1/ x
-1 1 f(x)

Ex. 1 Ex.2 Ex. 3 Ex.4

! !
Example 1. lim 1 − x2 = 0 lim 1 − x2 = 0
x→1− x→−1+

(As the picture shows, at the two endpoints of the domain, we only have a one-sided limit.)

−1, x < 0
"
Example 2. Set f (x) = Then lim f (x) = −1, lim f (x) = 1.
1, x > 0. x→0− x→0+

1 1
Example 3. lim = ∞, lim = −∞
x→0+ x x→0− x
1 1
Example 4. lim = ∞, lim = ∞
x→0+ x2 x→0− x2
The relationship between the one-sided limits and the usual (two-sided) limit is given by

(1) lim f (x) = L ⇐⇒ lim f (x) = L and lim f (x) = L

x→a x→a− x→a+

In words, the (two-sided) limit exists if and only if both one-sided limits exist and are equal.
This shows for example that in Examples 2 and 3 above, lim f (x) does not exist.
x→0

Students often say carelessly that lim 1/x = ∞, but this is not sloppy, it is simply
x→0
wrong, as the picture for Example 3 shows. By contrast, lim 1/x2 = ∞ is correct and
x→0
acceptable terminology.
1
2

2. Continuity
To understand continuity, it helps to see how a function can fail to be continuous. All
of the important functions used in calculus and analysis are continuous except at isolated
points. Such points are called points of discontinuity. There are several types. Let’s begin
by first recalling the definition of continuity (cf. book, p. 75).

(2) f (x) is continuous at a if lim f (x) = f (a).

x→a

Thus, if a is a point of discontinuity, something about the limit statement in (2) must fail
to be true.
Types of Discontinuity

x2 - 1
x- 1

sin (1/ x)

removable removable jump infinite essential

In a removable discontinuity, lim f (x) exists, but lim f (x) %= f (a). This may be because
x→a x→a
f (a) is undefined, or because f (a) has the “wrong” value. The discontinuity can be removed
by changing the definition of f (x) at a so that its new value there is lim f (x). In the left-most
x→a
x2 − 1
picture, is undefined when x = 1, but if the definition of the function is completed
x−1
by setting f (1) = 2, it becomes continuous — the hole in its graph is “filled in”.

In a jump discontinuity (Example 2), the right- and left-hand limits both exist, but
are not equal. Thus, lim f (x) does not exist, according to (1). The size of the jump is
x→a
the difference between the right- and left-hand limits (it is 2 in Example 2, for instance).
Though jump discontinuities are not common in functions given by simple formulas, they
occur frequently in engineering — for example, the square waves in electrical engineering,
or the sudden discharge of a capacitor.

In an infinite discontinuity (Examples 3 and 4), the one-sided limits exist (perhaps as ∞
or −∞), and at least one of them is ±∞.

An essential discontinuity is one which isn’t of the three previous types — at least one of
the one-sided limits doesn’t exist (not even as ±∞). Though sin(1/x) is a standard simple
example of a function with an essential discontinuity at 0, in applications they arise rarely,
presumably because Mother Nature has no use for them.

We say a function is continuous on an interval [a, b] if it is defined on that interval and

continuous at every point of that interval. (At the endpoints, we only use the approrpiate
one-sided limit in applying the definition (2).)
 C. CONTINUITY AND DISCONTINUITY 3

We say a function is continuous if its domain is an interval, and it is continuous at every

point of that interval.
A point of discontinuity is always understood to be isolated, i.e., it is the only bad point
for the function on some interval.
We illustrate the point of these definitions. (They are slightly different from the ones in
your book, but are more consistent with standard terminology in calculus.)

f(x)
"!
!
x
1/ x
|x | f(x)

Example 5 Example 6 Example 7 Example 8

Example 5. The function 1/x is continuous on (0, ∞) and on (−∞, 0), i.e., for x > 0
and for x < 0, in other words, at every point in its domain. However, it is not a continuous
function since its domain is not an interval. It has a single point of discontinuity, namely
x = 0, and it has an infinite discontinuity there.

Example 6. The function tan x is not continuous, but is continuous on for example the
interval −π/2 < x < π/2. It has infinitely many points of discontinuity, at ±π/2, ±3π/2,
etc.; all are infinite discontinuities.

Example 7. f (x) = |x| is continuous, but f # (x) has a jump discontinuity at 0.

1, x > 0,
"
Example 8. The function in Example 2, f (x) = does not have a
−1, x < 0
continuous derivative f # (x) — students often think it does since it seems that f # (x) = 0
everywhere. However this is not so: f # (0) does not exist, since by definition,
f (0 + ∆x) − f (0)
f # (0) = lim ,
∆x→0 ∆x
but f (0) does not even exist. Even if f (0) is defined to be say 1, or 0, the derivative f # (0)
does not exist.

Remember the important little theorem (Simmons p. 75)

(3) f (x) differentiable at a ⇒ f (x) continuous at a

or to put it contrapositively,

f (x) discontinuous at a ⇒ f (x) not differentiable at a

The function in Example 8 is discontinuous at 0, so it has no derivative at 0; the discontinuity
of f # (x) at 0 is a removable discontinuity.

Exercises: Section 1D
 X. EXPONENTIALS
AND LOGARITHMS

1. The Exponential and Logarithm Functions.

We have so far worked with the algebraic functions — those involving polynomials and
root extractions – and with the trigonometric functions. We now have to add to our list the
exponential and logarithm functions, since these are used in your science and engineering
courses from the beginning. Your book will handle the calculus of these functions; here we
want to review briefly their algebraic properties, and look at applications; one or two of
them might be new to you.
Where does one encounter exponentials and logarithms? In general, exponentials are
used to express all sorts of simple growth and decay processes.
1. The growth of a bacteria colony which doubles in size every day: y = y0 2t ,
where t = time in days, y = population size, y0 = the initial size, i.e., size at t = 0.
2. Dollars in a bank account, at 5% interest compounded annually: A = A0 (1.05)n ,
where n = number of years, A = amount, A0 = initial amount (the “principal”).
3. Amount of radioactive substance, with a 1 year half-life: x = x0 (1/2)n = x0 2−n ,
where n = number of years, x = amount, x0 = initial amount.
There are many other examples: the decay of electric charge on a capacitor, the way
a hot and cold body come to the same temperature when they are brought together, the
falling of a body through a resisting medium (a steel ball dropped into oil, for instance) —
all involve exponentials when you express them in mathematical terms.

We use logarithms when the base of the exponential is unimportant and we want to focus
our attention on the exponents instead. Suppose the base is 10; writing simply “log” for
“log10 ” in what follows, we have

y = 10x ⇔ log y = x.

1. Star magnitude. The observed brightness B of a star is described by comparing it

to a standard brightness B0 , using the equation

B = B0 10−2m/5 .

The important thing is the number m called the magnitude; it is defined by the above
equation, or equivalently, taking logs, by
5 B
m = − log .
2 B0
Thus, the higher the magnitude, the fainter the star. A star of brightness B0 , if one existed,
would have magnitude zero. Thus the constant B0 gives the magnitude scale a zero point.
This point was chosen so that the brightest stars (other than the sun) — Sirius and Vega,
for example — would have magnitude near 1, i.e., be stars of the first magnitude.
1
2

2. Orders of magnitude. We speak of a kilometer as “three orders of magnitude”

greater than a meter. That is
! "
km km
= 103 , 3 = log .
meter meter

In general, when we compare two quantities A and B,

number of orders of magnitude by which A is greater than B = log(A/B) .

3. Chemical pH. In chemistry, the acidity or alkalinity of a solution is measured by

the concentration [ H+ ] of hydrogen ions in the solution, in units of moles/liter.
If for some solution the concentration [ H+ ] = 10n , then we define

pH of solution = −n = − log [ H+ ].

The negative sign is used to avoid always having negative pH’s, since the concentration [ H+ ]
is very small.

The algebraic laws.

We will begin with a brief review of the algebra of exponentials and logarithms. Please
study the examples given of calculations, since many students seem to have trouble with
this.
We fix a number a > 0, called the base. The pair of equations

(1) y = ax loga y = x

are equivalent; both express the same relation between x and y. If you substitute one
equation into the other, you get two other equations which are often more useful in practice
than those in (1) (the examples will illustrate):

(2) y = aloga y loga (ax ) = x .

In the first, we take the log then exponentiate; in the second, we exponentiate, then take
the log; in both cases, we end up where we started.
The two basic algebraic laws governing exponentials and logarithms are

(3) ax1 +x2 = ax1 ax2 loga y1 + loga y2 = loga y1 y2

(4) (ax1 )k = akx1 loga y1k = k loga y1

The laws on either side follow from those on the other side by substituting yi = axi or
xi = loga yi , and then using (1) or (2).
 X. EXPONENTIALS AND LOGARITHMS 3

Graphs of the exponential and log functions

The picture on the left below shows some graphs of the exponential function ax for
different choices of bases. (We will discuss the base e ≈ 2.718 presently.)
! "x
1
The graph rises if a > 1, falls if a < 1. Since = a−x , the graph of (1/a)x is
a
obtained by reflecting the graph of ax in the y-axis. (See Notes G, p. G.2.)
The pair of equivalent equations y = ax and x = loga y show that ax and loga x are
inverse functions, according to Notes G, pp. G.8,9. Thus the graph of loga x is obtained by
reflecting the graph of ax in the diagonal line y = x. The reasoning for this needs a lot of
repetition to sink in, so we repeat it briefly once more. Namely, consider in turn the three
equations
y = ax x = ay y = loga x
Since the second equation arises from the first by switching x and y, the graph of the second
equation is obtained by flipping the first graph around the diagonal line y = x; the third
equation has the same graph as the second, since it’s the same equation — we’ve only solved
it for y. To sum up: the first and third equations have graphs which are reflections of each
other about y = x.

y ex x y
2 log 2 x
x
10 logex
log10x
x x
(1/2)
x

The base e. Changing the base.

There are really only two or three bases in common use.
For areas in science and engineering where calculus does not enter, or where at least
historically it was not used, the base 10 is the usual choice, because it associates naturally
with the decimal system, making rough estimations of the size of exponentials a fairly easy
thing to do.
On the other hand, if you are going to have to differentiate or integrate, the base e ≈ 2.718
is the usual choice. (The symbol e was introduced by Euler — the first letter of both
“exponential” and “Euler”.) As you will see, it is the base which makes for the simplest-
looking differentiation and integration formulas for exponentials and logs.
(To the above two, one should perhaps add the base 2, which is useful in computer science,
with its bits and bytes.)
For base e and base 10, the corresponding notations for the logarithm are

loge x = ln x (the “natural logarithm”); log10 x = log x.

(Warning: Many pure mathematics books use log x for the natural logarithm; science and
engineering books generally use ln x. Some computer science books use log x to mean log2 x.)
4

Using science and engineering notation, the logarithm laws read, for the bases e and 10:

(5) ln(ab) = ln a + ln b log(ab) = log a + log b

b
ln(a ) = b ln a log(ab ) = b log a

and

(6) ln(ex ) = x, eln x = x log(10x ) = x, 10log x = x

ln e = 1, ln 1 = 0; e1 = e, e0 = 1 log 10 = 1, log 1 = 0; 101 = 10, 100 = 1.

Other bases. The reason why it is not necessary to use other bases is that all other
exponentials and logarithms can be written in terms of ex and ln x, or if one prefers, in
terms of 10x and log x. Since in calculus we use the base e, let’s use this base. Using (6),
we have for any base a > 0,
ax = (eln a )x = ex ln a .
Thus the functions ekx , for different constants k, represent all of the exponential functions
ax . no matter what the base a is.
As for logarithms, starting with the equations

ax = y, x = loga y ,

taking the natural log of the equation on the left gives

x(ln a) = ln y .

Solving for x and substituting into the second equation, we get

ln y
loga y = .
ln a

Finally, changing the variable from y to x, we have

ln x ln x
(7) loga x = ; in particular, log x = .
ln a ln 10

Study these transformations until you can reproduce them easily.

In calculations, students usually find it easy to use the laws of exponentials and logs.
The chief trouble is in passing from the log to the exponential and vice-versa. Study the
following examples, and don’t go further until you have tried some of the exercises. Notice
that you can almost always get through using (2), whereas the equations (1) sometimes
aren’t good enough, and even when they are, they require too much thinking. The general
principle which governs the following calculations is:

(8) To get rid of logs, exponentiate; to get rid of exponentials, take the logs.
 X. EXPONENTIALS AND LOGARITHMS 5

Example 1. If one liquid has a pH which is 4 greater than the pH of a second liquid,
what is the relation between the [ H+ ] concentrations of the two liquids? (cf. p.X.1)
First Solution. Using subscripts to distinguish the two liquids,

(pH)1 = (pH)2 + 4
− log [ H+ ]1 = − log [ H+ ] + 4

Multiplying through by −1, then exponentiating both sides (base 10) and using (6),

[ H+ ]1 = [ H+ ]2 · 10−4 .

Second Solution. Starting from the second line, put both logs on the left side, and
combine them using (5) (remember that − log a = ln(1/a)), then exponentiate both sides
(or use (1)).

Example 2 A radioactive substance decays according to the law y = y0 e−kt , k > 0.

Two measurements are made: at time ti , there is an amount yi present, where i = 1, 2.
Express k in terms of the yi and ti .
Solution. We have y2 = y0 e−kt2 and y1 = y0 e−kt1 . Dividing one equation by the other
to get rid of y0 ,

y2 /y1 = e−kt2 /e−kt1 = e−k(t2 −t1 ) ;

ln(y2 /y1 ) = −k(t2 − t1 ),
ln(y2 /y1 ) ln y2 − ln y1
k = − = − .
t2 − t1 t2 − t1

Example 3. Solve for y the equation ln(y + 1) = ln y + x − 2.

Solution. Exponentiating both sides, y + 1 = y · ex−2 ; this is of the form y + 1 = Ay,
whose solution for y is
1 1
y= = .
1−A 1 − ex−2

2
Example 4. Solve for y: (1 + x)3 = 4−y .
Solution. Take the log of both sides, to get the y downstairs:

3 ln(1 + x) = −y 2 ln 4
# $1/2
3
− ln(1 + x) = y.
ln 4

Exercises: Section 1H
 A. APPROXIMATIONS

In science and engineering, you often approximate complicated functions by simpler ones
which are easier to calculate with, and which show the relations between the variables more
clearly. Of course, the approximation must be close enough to give you reasonable accuracy.
For this reason, approximation is a skill, one your other teachers will expect you to have.
This is a good place to start acquiring it.
Throughout, we will use the symbol ≈ to mean “approximately equal to”; this is a bit
vague, but making approximations in engineering is more art than science.

1. The linear approximation; linearizations.

The simplest way to approximate a function f (x) for values of x near a f(a) f(x)
is to use a linear function. The linear function we shall use is the one whose
graph is the tangent line to f (x) at x = a. This makes sense because the
tangent line at (a, f (a)) gives a good approximation to the graph of f (x),
if x is close to a. That is, for x ≈ a, a

(1) height of the graph of f (x) ≈ height of the tangent at (a, f (a))

To turn (1) into calculus, we need the equation for the tangent line. Since the line goes
through (a, f (a)) and has slope f ! (a), its equation is

y = f (a) + f ! (a)(x − a),

and therefore (1) can be expressed as

(2) f (x) ≈ f (a) + f ! (a)(x − a), for x ≈ a .

This says that for x near a, the function f (x) can be approximated by the linear function
on the right of (2). This function — the one whose graph is the tangent line — is called the
linearization of f (x) at x = a.
The approximation (2) is often written in an equivalent form that you should become
familiar with; it makes use of a dependent variable. Writing

(3) y = f (x), ∆x = x − a, ∆y = f (x) − f (a),

y
the approximation (2) takes the form

(2’) ∆y ≈ f ! (a) ∆x, for ∆x ≈ 0 . !y f (a) ! x

f(x)
In this form, the quantity on the right represents the change in height of the !x
tangent line, while the left measures the change in height of the graph. x

Here are some examples of linear approximations. In all of them, we are taking a = 0,
this being the most important case. All can be found by using (2) above and calculating
the derivative. You should verify each of them, and memorize the approximation.
1
2 A. APPROXIMATIONS

Basic Linear Approximations

1
(4) ≈ 1 + x, for x ≈ 0 ;
1−x
(5) (1 + x)r ≈ 1 + rx, for x ≈ 0 ; r is any real number
(6) sin x ≈ x, for x ≈ 0 .

Note that (4) becomes a special case of (5) if we take r = −1 and replace x by −x;
nonetheless, learn (4) separately since it is very common. As an example of verification, let
us check (5):

f (x) = (1 + x)r ⇒ f ! (x) = r(1 + x)r−1 , for any real r ;

!
⇒ f (0) = r .

Therefore, (2) becomes

(1 + x)r ≈ f (0) + f ! (0) x

≈ 1+rx , which is (5).

2. The algebraic viewpoint; examples

Though the three basic approximations given above can be derived by using differen-
tiation, many people remember them better by relating them to high school algebra and
geometry. We show how.
The approximation (4) can be thought of as coming from the formula for the sum of a
geometric series (memorize this too, if you have forgotten it):

1
= 1 + x + x2 + . . . + xn + . . . , |x| < 1 .
1−x

If x is small, then the terms x2 , x3 , . . . on the right are all negligible compared with the
term x, so they can be ignored, and we get (4).
Similarly, the approximation (5) can be thought of as coming from the binomial theorem,
if r is a positive integer:

r(r − 1) 2
(1 + x)r = 1 + rx + x + . . . + xr .
2

As before, if x is small, we can neglect the terms in x2 , x3 , . . . , and we get the approximation
(5). Even if r is not an integer, you will learn when you study infinite series that the binomial
theorem is still formally true. Though it gives an infinite sum on the right, instead of a
finite sum, the coefficients are still calculated by the same formulas.
Finally, the linear approximation (6) for sin x should make sense if you 1 x
think of the trigonometric definition of sin x. Referring to the picture, it
sin x
says that a small arc 2x of the unit circle is approximately equal in length
to the chord 2 sin x it subtends.
 A. APPROXIMATIONS 3

Continuing this algebraic viewpoint, many other linear approximation formulas can be
derived from the basic ones above by using algebra, rather than by going back to (2) and
calculating derivatives. Here are some examples of this.

Example 1. In each of the following, we want a linear approximation valid for x ≈ 0.

Observe in the first two how the variable is divided by a number (or “scaled”, as one says,
since it amounts to a change of scale or change of units for the variable). The purpose is to
put the expression in a form where one of the basic approximations can be used.
! "
1 1/2 1 x
(a) = ≈ 1− , by scaling and using (4);
2+x 1 + x/2 2 2
√ √ #
(b) 9 + t = 9 1 + t/9 = 3(1 + t/9)1/2 , by scaling;
≈ 3(1 + t/18) = 3 + t/6, for t ≈ 0, by using (5) .

d√
$
$ 1
Example (b) above is just as easy to do by using (2), since 9 + t$$ = .
dt t=0 6
In example (c) below, however, using (2) would definitely require more work.
2+x 2+x
(c) √ ≈ ≈ (2 + x)(1 − x/2), using (5), then (4);
1+x 1 + x/2
≈ 2, multiplying out and neglecting terms in x2 .

Notice that in this example, the linearization 2 + 0 x turns out to be a constant function.

Approximations for x ≈ a, where a %= 0 .

Though it is most common to work near a = 0, sometimes one wants another value of a.
Either one can use the formula (2), or else one can make a change of variable: h, ∆x, ! are
all common choices, related to x by

(7) x = a + h, x = a + ∆x, x=a+! .

The new variable is then close to 0 when x is close to a. Here is an example.

Example 2. Approximate 3 + x4 for x ≈ 1.

Solution. Either use (2), or change variable; doing the latter, we put x = 1 + h. Then

3 + x4 = 3 + (1 + h)4 ,
≈ 3 + (1 + 4h), h ≈ 0, using (5);
≈ 4 + 4(x − 1), for x ≈ 1.

Applications. Here are a few typical uses of the linearization.

Example 3. In the theory of special relativity, the mass m of a body moving with speed
v is given by
m0 c
m = √ , m0 = mass at rest, c = velocity of light
c2 − v 2
What speed produces a 1% increase in mass?
4 A. APPROXIMATIONS

Solution. We could crank out the answer, using the formula for m, but in practise a
simplifying approximation would be used. To begin with, scale m and v, i.e., divide them
by suitable constants to make them dimensionless: m/m0 and v/c; this turns the above
formula into (dividing top and bottom by c):

m 1 1
= # = ,
m0 2
1 − v /c 2 (1 − u2 )1/2

where we have set v/c = u; when v is small compared with c, then u ≈ 0. We get, using (5)
with r = 1/2,
1 1 u2
≈ 2
≈ 1+ , u ≈ 0,
2
(1 − u ) 1/2 1 − (1/2)u 2
where the second approximation used (4), with x = u2 /2.
This approximation shows that to make m/m0 = 1.01 (this represents a 1% increase in
the mass), we want √
u2 /2 = .01, i.e., u = .02 ≈ 1/7 .
The corresponding velocity is (remember that u = v/c):

v ≈ c/7 ≈ (186, 000/7 mi/sec ≈ 27, 000 mi/sec.

Example 4. Give a useful approximate formula, valid for relatively small heights,
showing how the weight of a body decreases as it rises above the earth, and use your
formula to determine how high it must rise to experience a 1% loss in weight.
Solution. Let R be the radius of the earth. The force between two masses m1 and M2
with centers of mass separated by a distance d is

Gm1 m2
F = ,
d2

so if the earth weight is M and our body has weight m,

GM m
weight at surface =
R2
GM m
weight at height h above surface = , so that
(R + h)2
weight at height h R2 1
= 2
= ,
weight at surface (R + h) (1 + h/R)2

where in this last step we made the variable dimensionless by dividing numerator and
denominator by R2 ; this scaling also makes the expression simpler. We continue with
approximations:

≈ (1 − h/R)2 , using (4);

≈ 1 − 2h/R, using (5).

The approximation is valid if h/R ≈ 0, i.e., if h is very small compared to R.

 A. APPROXIMATIONS 5

Using our approximation, we see that to make the ratio of the weights ≈ .99, we want

2h .01R .01 · 4, 000

= .01, i.e., h = = = 20 miles.
R 2 2

4. Quadratic approximations.
To get greater accuracy, sometimes one wants to include higher-order terms in the ap-
proximating function. If we include second-order terms — that is, terms in (x − a)2 , we get
what is called a quadratic approximation for x ≈ a. It looks like

(8) f (x) ≈ A + B(x − a) + C(x − a)2 , x≈a.

There is a general formula for the coefficients A, B, C using calculus, but let’s work alge-
braically first, and consider the basic approximations.

Basic Quadratic Approximations

1
(9) ≈ 1 + x + x2 , for x ≈ 0 ;
1−x
r(r − 1) 2
(10) (1 + x)r ≈ 1 + rx + x , for x ≈ 0 ; r is any real number
2
(11) sin x ≈ x, for x ≈ 0 .
2
x
(12) cos x ≈ 1 − , for x ≈ 0 .
2
Discussion
Formula (9) comes as before from the sum of the geometric series.
Formula (10) is the beginning of the binomial theorem, if r is an integer.
Formula (11) looks like our earlier linear approximation, but the assertion here is that it
is also the best quadratic approximation — that is, the term in x2 has 0 for its coefficient.
This is so because sin x is an odd function, so the approximating polynomial should be odd
also, which means it cannot have any x2 term.
Formula (12) is derived from (11) and the identity sin2 x + cos2 x = 1; this is one of the
exercises.

Using these basic quadratic approximations, we can by algebra get quadratic approxima-
tions to more involved expressions. Examples are given below. In studying the examples,
notice that during the course of the calculation, all approximations must be quadratic. If
one of the approximations you use is only linear, then that contaminates the final result,
which probably will not have the correct x2 term. This is the same principle you meet in
adding numbers: if one of the numbers is only good to one decimal place, then no matter
how accurate all the other numbers are, the sum will only be good to one decimal place.
6 A. APPROXIMATIONS

Example 5. By using the basic approximations, give quadratic approximations valid

for x ≈ 0 for each of the following:
√ √ cos x
(a) sec x (b) 1 + 3x (c) 1 + x + x2 (d)
1−x

Solution.

1 1 x2
(a) sec x = ≈ 2
≈ 1+ , by (12) and (4).
cos x 1 − x /2 2

√ 1 1
(b) 1 + 3x = (1 + 3x)1/2 ≈ 1 + (3x) − (3x)2 , by (10);
2 8
3 9
≈ 1 + x − x2 .
2 8

# &1/2
1 + x + x2 = 1 + (x + x2 )
%
(c)
1 1 1 3
≈ 1 + (x + x2 ) − (x + x2 )2 ≈ 1 + x + x2 , by (10).
2 8 2 8

x2 x2
! "
cos x
(d) ≈ 1− (1 + x + x2 ) ≈ 1 + x + .
1−x 2 2

To illustrate what happens if you don’t keep enough terms during the calculation, observe
that if in (d) above we only used 1 + x in the right-hand factor, the answer would have been
1 + x − x2 /2, whose x2 term is incorrect.

6. The quadratic approximation formula.

f !! (a)
(13) f (x) ≈ f (a) + f ! (a)(x − a) + (x − a)2 , for x ≈ a.
2

Example 6. Check formulas (10) and (11) by using (13).

Solution. Since the first two terms of (13) are the linearization, we can build on our earlier
work, and have only to calculate the quadratic coefficient f !! (0)/2. We get

(a) sin x ≈ 0 + x + 0x2 , since sin!! (x) = − sin x ⇒ sin!! (0) = 0.

(b)
f !! (0) r(r − 1)
f (x) = (1 + x)r ⇒ f !! (x) = r(r − 1)(1 − x)r−2 ⇒ = , as in (10).
2 2

The usefulness of (13) is tempered by the fact that it requires calculation of second
derivatives. This can get rather tedious — function (d) in Example 5 is a good illustration
— so that using the algebraic techniques is often better.
 A. APPROXIMATIONS 7

The quadratic approximation formula (13) may be “derived” as follow. We are looking
for the right choice of coefficients in

(14) f (x) ≈ A + B(x − a) + C(x − a)2 , x ≈ a.

Let us denote by Q(x) the polynomial on the right of (14). Then it makes sense to choose
the coefficients A, B, C so that f (x) and Q(x) have the same value and the same first and
second derivatives at a, i.e., so that

(15) f (a) = Q(a), f ! (a) = Q! (a), f !! (a) = Q!! (a) .

Since Q! (x) = B + 2C(x − a) and Q!! (x) = 2C, equations (15) say that

(16) f (a) = A, f ! (a) = B, f !! (a) = 2C ;

these values for A, B, C turn (14) into (13), as promised. Note that the first two terms on
the right of (13) give the linearization at x = a; thus the quadratic approximation refines
the linear approximation by adding a quadratic term to it.

Exercises: Section 2A
 MVT. MEAN-VALUE THEOREM

There are two forms in which the Mean-value Theorem can appear;1 you should get
familiar with both of them. Assuming for simplicity that f (x) is differentiable on an interval
whose endpoints are a and b, or a and x, the theorem says

f (b) − f (a)
(1) = f ! (c), for some c between a and b;
b−a
(2) f (x) = f (a) + f ! (c)(x − a), for some c between a and x

The first form (1) has an intuitive geometric interpretation in terms of the slope of a
secant being equal to the slope of the graph at some point c. and in this form, it’s easy to
give an intuitive argument for the theorem.
The second form (2) looks less intuitive, but all that has been done is to multiply both
sides of (1) by b−a, transpose a term, and change the name of b to x. Now it’s not a theorem
about slopes; instead, it says that the value of f at some point x can be estimated, provided
you know the value of f at some fixed point a, and have information about the size of f ! on
the interval [a, x]. In other words, from information about f ! , we can get information about
f . (Such information can also be gotten by integration; one can think of the Mean-value
Theorem as a poor-person’s substitute for integration.)
The special case of (1) in which f (a) = f (b) = 0 is usually called Rolle’s theorem; it says
that if f is differentiable on [a, b],

(3) f (a) = f (b) = 0 ⇒ f ! (c) = 0 for some c, where a < c < b.

Exercises: Section 2G

1 see Simmons, p. 76
1
 PI. PROPERTIES OF INTEGRALS

For ease in using the definite integral, it is important to know its properties. Your book
lists the following1 (on the right, we give a name to the property):
! a ! b
(1) f (x) dx = − f (x) dx integrating backwards
!b a a

(2) f (x) dx = 0
a
! b ! c ! b
(3) f (x) dx = f (x) dx + f (x) dx interval addition
a a c
! b ! b ! b
(4) (f + g) dx = f (x) dx + g(x) dx linearity
a a a
! b ! b
C f (x) dx = C f (x) dx linearity
a a
! b ! b
(5) f (x) dx ≤ g(x) dx if f (x) ≤ g(x) on [a, b] estimation
a a

Property (5) is useful in estimating definite integrals that cannot be calculated exactly.
! 1 "
Example 1. Show that 1 + x3 dx < 1.3 .
0

Solution. We estimate the integrand, and then use (6). We have

x3 ≤ x on [0, 1];
#1
√ 2 √
! 1" 1
2
!
3
1 + x dx ≤ 1 + x dx = (1 + x)3/2 = (2 2 − 1) ≈ 1.22 < 1.3 .
0 0 3 0 3

We add two more properties to the above list.

$! b $ ! b
$ $
(6) $
$ f (x) dx$ ≤
$ |f (x)| dx . absolute value property
a a

Property (6) is used to estimate the size of an integral whose integrand is both positive and
negative (which often makes the direct use of (5) awkward). The idea behind (6) is that on
the left side, the intervals on which f (x) is negative give a negative value to the integral,
and these “negative” areas lower the overall value of the integral; on the right the integrand
has been changed so that it is always positive, which makes the integral larger.
! 100
Example 2. Estimate the size of e−x sin x dx .
0

1 see Simmons pp. 214-215

1
2 PI. PROPERTIES OF INTEGRALS

Solution. A crude estimate would be

$ ! 100 $ ! 100
$ $
$
$ e −x
sin x dx$ ≤
$ e−x | sin x| dx
0 0
! 100
≤ e−x dx, by (5), since | sin x| ≤ 1;
0
#100
= −e −x
= −e−100 + 1 < 1.
0
A final property tells one how to change the variable in a definite integral. The formula is
the most important reason for including dx in the notation for the definite integral, that is,
! b ! b
writing f (x) dx for the integral, rather than simply f (x), as some authors do.
a a

! d ! b
du  u = u(x),

(7) f (u) du = f (u(x)) dx, c = u(a), change of variables formula
c a dx 
d = u(b).


In words, we can change the variable from u to x, provided we

(i) express du in terms of dx; (ii) change the limits of integration.2
There are various possible hypotheses on u(x); the simplest is that it should be differen-
tiable, and either increasing or decreasing on the x-interval [a, b].
1
du
!
Example 3. Evaluate by substituting u = tan x.
0 (1 + u2 )3/2
Solution. u = 0, 1 corresponds to x = 0, π/4; du = sec2 x dx; 1 + tan2 x = sec2 x; thus
! 1 ! π/4
du sec2 x
= dx
0 (1 + u )
2 3/2
0 sec3 x
! π/4 #π/4 √
2
= cos x dx = sin x = .
0 0 2
Proof of (7). We use the First Fundamental Theorem.3 Let F (u) be an antiderivative of
f (u):
!
(8) F (u) = f (u) du;
d dF du du
F (u(x)) = · = f (u) , by the chain rule. So
dx du dx dx
du
!
(9) F (u(x)) = f (u(x)) dx . Therefore
dx
! d
f (u) du = F (d) − F (c), by the First Fundamental Theorem and (8);
c
b
du
!
= F (u(b)) − F (u(a)) = f (u(x)) dx,
a dx
by the First Fundamental Theorem and (9). !
. Exercises: Section 3E
2 see Simmons p.339 for a discussion and an example
3 see Section FT
 FT. SECOND FUNDAMENTAL THEOREM

1. The Two Fundamental Theorems of Calculus

The Fundamental Theorem of Calculus really consists of two closely related theorems,
usually called nowadays (not very imaginatively) the First and Second Fundamental Theo-
rems. Of the two, it is the First Fundamental Theorem that is the familiar one used all the
time. It is the theorem that tells you how to evaluate a definite integral without having to
go back to its definition as the limit of a sum of rectangles.

First Fundamental Theorem Let f (x) be continuous on [a, b]. Suppose there is a
function F (x) such that f (x) = F ! (x) . Then
! b "b
(1) f (x) dx = F (b) − F (a) = F (x) .
a a

(The last equality just gives another way of writing F (b) − F (a) that is in widespread use.)
Still another way of writing the theorem is to observe that F (x) is an antiderivative for f (x),
or as it is sometimes called, an indefinite integral for f (x); using the standard notation for
indefinite integral and the bracket notation given above, the theorem would be written
! b ! "b
(1! ) f (x) dx = f (x) dx .
a a

Writing the theorem this way makes it look sort of catchy, and more importantly, it avoids
having to introduce the new symbol F (x) for the antiderivative.

In contrast with the above theorem, which every calculus student knows, the Second
Fundamental Theorem is more obscure and seems less useful. The purpose of this chapter is
to explain it, show its use and importance, and to show how the two theorems are related.
To start things off, here it is.

Second Fundamental Theorem. Let f (x) be continuous, and fix a.

! x
(2) Set F (x) = f (t) dt ; then F ! (x) = f (x) .
a

We begin by interpreting (2) geometrically. Start with the graph of f (t) y f(t)
in the ty-plane. Then F (x) represents the area under f (t) between a and
x; it is a function1 of x. Its derivative — the rate of change of area with
F(x)
respect to x — is the length of the dark vertical line. This is what (2) says t
geometrically. a x

1 Simmons calls this function A(x) on p. 207 (2nd edition); this section of Simmons is another presentation

of much of the material given here.

1
2

Example 1. Verify (2) if f (x) = 2x sin x2 and a = 0.

Solution. Here we can integrate explicitly by finding an antiderivative (using the first
fundamental theorem):
! x "x
F (x) = 2t sin t2 dt = − cos t2 = − cos x2 + 1;
0 0

differentiating by the chain rule, we verify that indeed F ! (x) = 2x sin x2 , as predicted by
(2). !
x
sin t
!
Example 2. Let F (x) = dt . Find F ! (π/2).
1 1+t
Solution. Neither integration techniques nor integral tables will produce an explicit
antiderivative for the function in the integrand. So we cannot use (1). But we can use (2),
which says that
sin π/2 1
F ! (π/2) = = . !
1 + π/2 1 + π/2
Many students feel the Second Fundamental Theorem is “obvious”; these students are
confusing it with the similar-looking
!
(2! ) Let F (x) = f (x) dx ; then F ! (x) = f (x) .

Indeed, (2! ) is obvious. The “integral” in it is an indefinite integral, i.e., an antiderivative.

So what (2! ) says is: “Let F (x) be an antiderivative for f (x); then F (x) is an antiderivative
for f (x) — a true statement, but not a very exciting one (logicians call it a tautology.)
The Second Fundamental Theorem (2) looks almost the same as (2! ), but it is actually
entirely different, because F (x) is defined as a definite integral. The next section, which
continues the discussion, should help show the difference.

2. Do functions have antiderivatives?

! b
The First Fundamental Theorem tells us how to calculate f (x) dx by finding an anti-
a
derivative for f (x). But the theorem isn’t so useful if you can’t find an antiderivative. Most
2
calculus students think for example that e−x has no antiderivative — “the integral isn’t
in the tables”, “it can’t
√ be integrated” are some of the ways they express this. Even for a
simple function like 1 − x2 , it is not obvious what the antiderivative is. Perhaps it doesn’t
have any?
The Second Fundamental Theorem provides the answer; it says:
! x
Every continuous function f (x) has an antiderivative: f (t) dt .
a
The antiderivative may not be expressible in terms of elementary functions — this is the
2
difficulty with e−x — but it always exists.
 FT. THE SECOND FUNDAMENTAL THEOREM 3

1
Example 3. Find an antiderivative for # √ .
1+ x
Solution. This doesn’t look so easy to do explicitly. But the Second Fundamental Theorem
says the following function is an antiderivative:
x
1
!
(3) F (x) = # √ dt
0 1+ t

Discussion. You may feel that this doesn’t represent progress: the formula for the
antiderivative is useless. But that’s not so: the function F (x) can be calculated by numerical
integration. It can be programmed into a calculator so that when you press an x-value, the
screen will display the corresponding value of F (x) to 12 decimal digits. Pressing another
button will draw the graph of F (x) over any interval on the x-axis that you specify.

! Repeating what we said earlier, the integral in (3) should be carefully distinguished from
1
# √ dx — this “integral” is just another notation for the antiderivative, and is
1+ x
therefore not a solution to the problem. The integral in (3) by contrast is a perfectly
definite function, and it does solve the problem of finding an antiderivative.
In this case, it turns out that F (x) does have an expression in terms of elementary
functions. It is
4 √ √ 8
(4) F (x) = ( x + 1)3/2 − 4( x + 1)1/2 + .
3 3

(You can prove this is correct by differentiating it; the 8/3 is put in to ensure that F (0) = 0,
as definition (3) requires.)
The above way of writing F (x) is different from (3). Whether or not it is a better way
depends on what you want to know about F (x) and what use you want to make of it. For
instance,
Is F (x) > 0 when x > 0?
The answer is clearly “yes” if we look at the integral (3), since the integrand is positive; it
is not at all clear what the answer is if instead we look at (4), because of the − sign. As
another example, what is F ! (x) ? From (3) the answer is immediate, whereas from (4) you
would have to calculate for a while — as you will know if you took the trouble to check its
correctness!
4

3. Defining new functions: ln(x) and erf(x).

One important use of the Second Fundamental Theorem is to define new functions. Cal-
culus can then be used to study their properties.
To illustrate, we consider first an old function: ln x. Let’s pretend we know nothing of
logarithms. We do know that

xn+1
!
xn dx = , n= # −1 .
n+1

However, we know no explicit formula for an antiderivative of 1/x, i.e., when n = −1.
We therefore use the Second Fundamental Theorem to define an antiderivative of 1/x,
namely
! x
dt
(5) L(x) = .
1 t

(We use 1 as the lower limit of integration since the integrand is not defined at 0.) What
are the properties of this function? ! x
dt
Properties of L(x) =
1 t
L-1. L(x) is defined for x > 0 (since 1/t is continuous for t > 0);
L-2. L(1) = 0;
L-3. L! (x) = 1/x, by the Second Fundamental Theorem;
L-4. L!! (x) = −1/x2 , by differentiating 1/x;
L-5. L(x) is increasing for all x > 0, since L! (x) > 0; its graph is concave (i.e., concave
down) since L!! (x) < 0;
L-6. L(ab) = L(a) + L(b) .
Of course, it is this last which is the interesting property; the proof of it is elegant.
Proof of L-6. We break up the integral defining L(ab) into two parts, the first of which
is L(a): to do this, we use the interval addition rule (3) in Notes PI.1 .
ab a ab
dt dt dt
! ! !
(6) L(ab) = = + .
1 t 1 t a t

Comparing with Property L-6, we see we have to show the last integral on the right above
has the value L(b). To see this, make the change of variable t = au and apply the change
of variable rule (see (7), p. PI.2 in these notes). You get successively

dt a du du
t = au, dt = a du, = = .
t au u
We have to change the limits on the integral also: t = a and t = ab correspond respectively
to u = 1, u = b. Thus the rule for changing variable in a definite integral gives
ab b
dt du
! !
= = L(b),
a t 1 u
 FT. THE SECOND FUNDAMENTAL THEOREM 5

which proves L-6. !

Once we have this, the other properties of the logarithm follow in a standard way.
L-7. L(1/a) = −L(a), since L(1/a) + L(a) = L( a1 · a) = L(1) = 0 .
L-8. lim L(x) = ∞; namely, L(x) is increasing and L(2n ) increases without bound as
x→∞
n → ∞, since L(2n ) = nL(2), by Property 6; note that L(2) > 0 since L(x) is increasing. !
In our definition of L(x), the number e appears as the unique number such that
e
dt
!
L(e) = = 1.
1 t

Such a number exists by the Intermediate Value Theorem,2 since L(x) is increasing, contin-
uous (since it has a derivative), and gets bigger than 1.

We now turn to a second example of using the Second Fundamental Theorem to define
a function F (x) — this time, the function will be genuinely new. It is closely related to
an important function in probability and statistics, the error function erf x. (Statistics-
oriented calculators have a button for it.) The two functions differ only by a change of scale
on the x- and y-axes. There is no simpler or more elementary expression for F (x).
! x
2
Example 5. Define a function F (x) by F (x) = e−t dt .
0

Sketch the graph of F (x), indicating relative maxima and minima, points of inflection,
symmetries. Estimate F (1) roughly.
2
2 e -t
Solution. The graph of f (t) = e−t is shown. F(x)

F (x) is the indicated shaded area under the graph of f (t). x t

2
F ! (x) = e−x by the Second Fundamental Theorem; since the exponential is always F(x)
positive, this shows F (x) is increasing for all x, and therefore it has no relative maxima or
minima.
2 -1 1
F !! (x) = −2xe−x ; since F !! (x) < 0 for x > 0, the graph of F (x) is
concave (down) when x > 0. Similarly, it is convex (concave up) for x < 0,
and it has a point of inflection at x = 0.
2
F (x) is an odd function. To see this, we note that e−t is an even function. As the
picture shows, the two shaded areas are equal; the one on the left however must be counted
negatively, since the integration is backwards: if x > 0, then
! −x ! x
F (−x) = f (t)dt = − f (t) dt = −F (x) . -x x
0 0
1
This shows F (x) is an odd function. 1/e
F (1) can be estimated as the area under f (t) between 0 and 1; it is
roughly comparable to the area of the trapezoid shown, which about .7 . 1

2 Simmons, p. 78
6

4. Proof of the two Fundamental Theorems.

We will give an intuitive argument for the Second Fundamental Theorem, and then
deduce the First Fundamental Theorem from the Second. Though the argument for the
Second Theorem is only suggestive, it has the right ideas in it, and can be easily made
rigorous if you have available a precise definition of limit.3
Second Fundamental Theorem: Intuitive Argument
We wish to prove that if f (x) is continuous, then f(t)

! x
(8) F (x) = f (t) dt ⇒ F ! (x) = f (x) . !F
a
f(x)
We calculate F ! (x) using the definition of derivative. Let x change by ∆x, x t
and let ∆F be the corresponding change in F (x). From the picture, !x

! x+∆x
∆F = F (x + ∆x) − F (x) = f (t) dt
x
(9) ≈ f (x)∆x,

since the area of the vertical strip under the curve is approximately the same as the area of
the rectangle. Dividing, we have
∆F
≈ f (x),
∆x
where the error in the approximation is bounded by the height of the small curved triangle.
Since f (t) is continuous, the error is small compared with f (x), and disappears when we
pass to the limit as ∆x → 0; we get therefore

∆F
F ! (x) = lim = f (x).
∆x→0 ∆x

Note that if f (t) were discontinuous at the point x, the result would not be true; as the
picture shows, the approximation in (9) would not be true.

First Fundamental Theorem: Proof 4

We want to show that if f (x) is continuous and f (x) = F ! (x), then
! b
(10) f (x) dx = F (b) − F (a).
a

We begin by defining
! x
(11) G(x) = f (t) dt; then
a
G! (x) = f (x), by the Second Fundamental Theorem.
3A somewhat fuller argument is given in Simmons, Step 1, p. 206-7.
4 this same classical reasoning is given in Simmons: Steps 2 and 3, p. 207.
 FT. THE SECOND FUNDAMENTAL THEOREM 7
$ %!
Since G! (x) = f (x) = F ! (x), we have G(x) − F (x) = 0, which implies G(x) − F (x) = C,
i.e.,

(12) G(x) = F (x) + C, for some constant C.

To evaluate C, we put x = a in (12); since G(a) = 0, we get

C = −F (a) .

Finally, put x = b in equation (12), and use the above value for C:

G(b) = F (b) − F (a),

which is exactly (10), in view of the definition of G(x). !

Remark. Both fundamental theorems say that differentiation and definite integration
are inverse operations: each undoes what the other does. In the First Fundamental Theorem
we differentiate, then integrate:
! x
!
F (x) −→ F (x) −→ F ! (t) dt = F (x) − F (a);
a

In the Second Fundamental Theorem, we integrate, then differentiate:

x x
d
! !
f (x) −→ f (t) dt −→ f (t) dt = f (x) .
a dx a

In both cases, the theorem says that we end up essentially where we started – only “essen-
tially” because of the additive constant in F (x).
(Of course, differentiation and indefinite integration are also inverse operations, but this
is trivial – it’s just a restating of the definition of indefinite integration.)

Exercises: Section 3D
 AV. AVERAGE VALUE

What was the average temperature on July 4 in Boston? T T=f(x)

The temperature is a continuous function f (x), whose graph over the 80
24-hour period might look as shown. How should we define the average
value of such a function over the time interval [0, 24] — measuring time
x in hours, with x = 0 at 12:00AM ? 6 12 18 24 x

We could observe the temperature in the middle of every hour, that is, at the times
x1 = .5, x2 = 1.5, . . . , x24 = 23.5, then average these 24 observations, getting
24
1 !
f (xi ) .
24 i=1
To get a more accurate answer, we could average measurements made more frequently, say
every ten minutes.
For a general interval [a, b] and function f (x), the analogous procedure would be to divide
up the interval into n equal parts, each of length
b−a
(1) ∆x =
n
and average the values of the function f (x) at a succession of points xi , where xi lies in the
i-th interval. Then we ought to have

n
1!
(2) average of f (x) over [a, b] ≈ f (xi ) .
n 1

We can relate the sum on the right to a definite integral: using (1), (2) becomes
n
1 !
(3) average of f (x) over [a, b] ≈ f (xi )∆x .
b−a 1

As n → ∞, the sum on the right-hand side of (3) approaches the definite integral of f (x)
over [a, b], and we therefore define the average value of the function f (x) on [a, b] by
b
1 f(x)
"
(4) A = average of f (x) over [a, b] = f (x) dx . A
b−a a

Geometrically, the average value A can be thought of as the height of that

constant function A which has the same area over [a, b] as f (x) does. This
is so since (4) shows that a b
" b
A · (b − a) = f (x) dx .
a

1
2 AV. AVERAGE VALUE

Example 1. In alternating current, voltage is represented by a sine

wave with a frequency of 60 cycles/second, and a peak of 120 volts. What
is the average voltage? 1/60
2π 2π
Solution. The voltage function has frequency = = 120π,
period 1/60
and amplitude 120, so it is given by V (t) = 120 sin(120πt). Thus by (4),

1/120 #1/120
120 2
"
average V (t) = 120 V (t) dt = − cos(120πt) = · 120 .
0 π 0 π

(We integrate over [0, 1/120] rather than [0, 1/60] since we don’t want zero P
as the average.)
-1 ! x 1
Example 2. P
a) A point is chosen at random on the x-axis between −1 and 1; call
it P . What is the average length of the vertical chord to the unit circle
passing through P ?
b) Same question, but now the point P is chosen at random on the circumference.
√
Solution. a) If P is at x, the chord has length 2 1 − x2 , so we get

1 1 $ π
$ "
average of 2 1 − x2 over [−1, 1] = 2 1 − x2 dx = area of semicircle = ≈ 1.6 .
2 −1 2

b) By symmetry, we can suppose P is on the upper semicircle. If P is at the angle θ, the

chord has length 2 sin θ, so this time we get
π #π
1 2 4
"
average of 2 sin θ over [0, π] = 2 sin θ dθ = − cos θ = ≈ 1.3 .
π 0 π 0 π

(Intuitively, can you see why the average in part (b) should be less than the average in
part (a) — could you have predicted this would be so?)

Exercises: Section 4D
 F. HEAVISIDE’S COVER-UP METHOD

The eponymous method was introduced by Oliver Heaviside as a fast way to do a decom-
position into partial fractions. In 18.01 we need the partial fractions decomposition in order
to integrate rational functions (i.e., quotients of polynomials). In 18.03, it will be needed as
an essential step in using the Laplace transform to solve differential equations, and in fact
this was more or less Heaviside’s original motivation.
The cover-up method can be used to make a partial fractions decomposition of a rational
p(x)
function whenever the denominator can be factored into distinct linear factors.
q(x)
We illustrate with an example; though simple, it should convince you that the method is
worth learning.
x−7
Example 1. Decompose into partial fractions.
(x − 1)(x + 2)
Solution. We know the answer will have the form
x−7 A B
(1) = + .
(x − 1)(x + 2) x−1 x+2
To determine A by the cover-up method, on the left-hand side we mentally remove (or cover
up with a finger) the factor x − 1 associated with A, and substitute x = 1 into what’s left;
this gives A:
!
x−7 ! 1−7
(2) ! = = −2 = A .
(x + 2) !x=1 1+2
Similarly, B is found by covering up the factor x + 2 on the left, and substituting x = −2
into what’s left. This gives
!
x−7 ! −2 − 7
! = = 3 = B.
(x − 1) !
x=−2 −2 − 1
Thus, our answer is
x−7 −2 3
(3) = + .
(x − 1)(x + 2) x−1 x+2

Why does the method work? The reason is simple. The “right” way to determine A from
equation (1) would be to multiply both sides by (x − 1); this would give
x−7 B
(4) = A + (x − 1) .
(x + 2) x+2
Now if we substitute x = 1, what we get is exactly equation (2), since the term on the right
disappears. The cover-up method therefore is just an easy way of doing the calculation
without going to the fuss of writing (4) — it’s unnecessary to write the term containing B
since it will become 0 .
1
2 F. COVER-UP METHOD

In general, if the denominator of the rational function factors into the product of distinct
linear factors:
p(x) A1 Ar
= + ...+ , ai "= aj ,
(x − a1 )(x − a2 ) · · · (x − ar ) x − a1 x − ar

then Ai is found by covering up the factor x − ai on the left, and setting x = ai in the rest
of the expression.
1
Example 2. Decompose into partial fractions.
x3 −x
Solution. Factoring, x3 − x = x(x2 − 1) = x(x − 1)(x + 1). By the cover-up method,

1 −1 1/2 1/2
= + + .
x(x − 1)(x + 1) x x−1 x+1

To be honest, the real difficulty in all of the partial fractions methods (the cover-up
method being no exception) is in factoring the denominator. Even the programs which do
symbolic integration, like Macsyma, or Maple, can √ only factor polynomials whose factors
have integer coefficients, or “easy coefficients” like 2. and therefore they can only integrate
rational functions with “easily-factored” denominators. (Of course, these are the only kind
you’ll see in 18.01 or 18.03.)

Heaviside’s cover-up method also can be used even when the denominator doesn’t factor
into distinct linear factors. To be sure, it gives only partial results, but these can often be
a big help. We illustrate.
5x + 6
Example 3. Decompose .
(x2 + 4)(x − 2)
Solution. We write
5x + 6 Ax + B C
(5) = + .
(x2 + 4)(x − 2) 2
x +4 x−2

We first determine C by the cover-up method, getting C = 2 . Then A and B can be found
by the method of undetermined coefficients; the work is greatly reduced since we need to
solve only two simultaneous equations to find A and B, not three.
Following this plan, using C = 2, we combine terms on the right of (5) so that both sides
have the same denominator. The numerators must then also be equal, which gives us

(6) 5x + 6 = (Ax + B)(x − 2) + 2(x2 + 4) .

Comparing the coefficients say of x2 and of the constant terms on both sides of (6) then
gives respectively the two equations

0=A+2 and 6 = −2B + 8,

from which A = −2 and B = 1 .

In using (6), one could have instead compared the coefficients of x, getting 5 = −2A + B;
this provides a valuable check on the correctness of our values for A and B.
 F. COVER-UP METHOD 3

In Example 3, an alternative to undetermined coefficients would be to substitute two

numerical values for x into the original equation (5), say x = 0 and x = 1 (any values
other than x = 2 are usable). Again one gets two simultaneous equations for A and B.
This method requires addition of fractions, and is usually better when only one coefficient
remains to be determined (as in Example 4 below).
Still another method would be to factor the denominator completely into linear factors,
using complex coefficients, and then use the cover-up method, but with complex numbers. At
the end, conjugate complex terms have to be combined in pairs to produce real summands.
The calculations are sometimes longer, and require skill with complex numbers.

The cover-up method can also be used if a linear factor is repeated, but there too it gives
just partial results. It applies only to the highest power of the linear factor. Once again, we
illustrate.
1
Example 4. Decompose .
(x − 1)2 (x + 2)
Solution. We write
1 A B C
(7) 2
= 2
+ + .
(x − 1) (x + 2) (x − 1) x−1 x+2

To find A cover up (x − 1)2 and set x = 1; you get A = 1/3. To find C, cover up x + 2, and
set x = −2; you get C = 1/9.
This leaves B which cannot be found by the cover-up method. But since A and C are
already known in (7), B can be found by substituting any numerical value (other than 1 or
−2) for x in equation (7). For instance, if we put x = 0 and remember that A = 1/3 and
C = 1/9, we get
1 1/3 B 1/9
= + + ,
2 1 −1 2
from which we see that B = −1/9.
B could also be found by applying the method of undetermined coefficients to the equation
(7); note that since A and C are known, it is enough to get a single linear equation in order
to determine B — simultaneous equations are no longer needed.
The fact that the cover-up method works for just the highest power of the repeated linear
factor can be seen just as before. In the above example for instance, the cover-up method
for finding A is just a short way of multiplying equation (5) through by (x − 1)2 and then
substituting x = 1 into the resulting equation.

Exercises 5E
 INT. IMPROPER INTEGRALS

In deciding whether an improper integral converges or diverges, it is often awkward or

impossible to try to decide this by actually carrying out the integration, i.e., finding an
antiderivative explicitly. For example both of these two improper integrals converge:
! ∞ ! ∞
dx dx
6 + 3x3 + 2x2 + 1
and √ ,
1 x 1 x3+1

but there is no explicit antiderivative for the second integral, and finding one for the first
would be a hairy exercise in partial fractions, even if one were able to factor the denominator.
Instead of explicit integration, therefore, we show they converge by using estimation
instead, comparing them with simpler integrals which are known to converge. Thus, for the
first,
1 1
≤ 6, x > 0,
x6 + 3x3 + 2x2 + 1 x

so that
∞ ∞
1
! !
dx dx
≤ = .
1 x6 + 3x3 + 2x2 + 1 1 x6 5

Since the right hand integral converges, so does the left, which is smaller (but still positive).
In a similar way, for the second integral, we estimate
1 1
√ ≤ , x > 0,
x3 +1 x3/2

so that
! ∞ ! ∞
dx dx
√ ≤ = 2.
1 1 x3/2
x3 + 1
! ∞
x dx
In the same way we can show the divergence say of √ :
1 x3 + 1
x 1
√ ≥ √ , x > 1,
3
x +1 2 x
√
this last being equivalent to 2x3/2 ≥ x3 + 1, i.e., to 8x3 ≥ x3 + 1; thus we get
! ∞ ! ∞
x dx dx
√ ≥ √ = ∞.
1 x3 + 1 1 2 x

We call the general principle we are using here the

1
2 INT. IMPROPER INTEGRALS

Comparison Test for Improper Integrals. If 0 ≤ f (x) ≤ g(x) for a ≤ x < ∞,

! ∞ ! ∞
(1) g(x) dx converges ⇒ f (x) dx converges;
a a
! ∞ ! ∞
(1" ) f (x) dx diverges ⇒ g(x) dx diverges.
a a

In other words, if the area under g(x) is finite and f (x) lies below g(x) g(x)
(but still over the x-axis), then the area under f (x) will be finite also. Or f(x)
equivalently, if the area under f (x) is infinite, so is the area under g(x).
In using the test, the lower limit of integration is of no importance, since if a < b,
! ∞ ! b ! ∞
f (x) dx = f (x) dx + f (x) dx, so that
a a b

! ∞ ! ∞
(3) f (x) dx converges ⇔ f (x) dx converges.
a b

As another example of the use of the comparison test, we recall one used to illustrate the
Second Fundamental Theorem (cf. p. FT.3):
2
e−x ≤ e−xfor x ≥ 1, and therefore
! ∞ ! ∞
−x2
e dx converges, since e−x dx converges.
0 0

(Here we use (3) above to shift the lower limit from 1 to 0; also the comparison test (1).)
In using the comparison test, we have to have some standard integrals that we know
converge or diverge, to use for comparison purposes. The most useful are:
! ∞
dx
(4) converges if p > 1, diverges if p ≤ 1;
1 xp
! ∞
(5) e−ax converges if a > 0.
0

It is important to notice that in using the test, you must make the inequality go in the
right direction, which means you!must guess in advance whether the integral will converge
∞
x dx
or diverge. For example, to test 3−2
, we see that as x → ∞,
2 x
x x 1
∼ 3 = 2,
x3 − 2 x x
x 1
so from (4) we guess it will converge. Unfortunately, 3 > 2 , x ≥ 2, so we can’t
! ∞ x − 2 x
dx
use as the comparison integral; however, only a slight change is needed:
1 x2
x 2
< 2, x ≥ 2,
x3 −2 x
as we see by cross-multiplying: x3 < 2x3 − 4; thus our integral converges, using (1). !
 INT. IMPROPER INTEGRALS 3

Improper integrals of the second kind

The comparison test also works for improper integrals of the form
! b ! b−
f (x) dx, where lim f (x) = ∞, and f (x) dx, where lim f (x) = ∞.
a+ x→a+ a x→b−

Sometimes these are called improper integrals of the second kind

— the first kind being the previous type of improper integral, where
one of the limits of integration is ∞ or −∞. a b a b

For improper integrals of the second kind, useful standard comparison integrals are
! b− ! b
dx dx
(6) and ,
a (b − x)p a+ (x − a)p

which converge if p < 1, diverge if p ≥ 1 . Note that this is just the 1

opposite of (4). However, it’s easy to remember which is which if you 1 / /x
1/ x
think of the picture at the right, which compares the graphs for p < 1 1 / x2
1
and p > 1.
! 1
dx
Example. Test √ for convergence.
0 2x − x2
Solution. The integrand becomes infinite when x = 0 and when x = 2, but 2 is of no
importance, since it’s not in the interval over which we are integrating. In making our guess
as to convergence or divergence, we note that

1 1 1
√ ≈ √ ·√ , x ≈ 0+ .
2x − x2 2 x

Thus, according to (6) above, we expect convergence, and make the comparison

1 1
(7) √ ≤ √ .
2x − x2 x

Where is (7) valid (if at all)? Crossmultiplying, it claims that

√ "
x ≤ 2x − x2 , or squaring, x ≤ 2x − x2 , or x2 ≤ x .

This last inequality is true if 0 ≤ x ≤ 1, so we’re safe, since this is the region of integration.
So we see by the! comparison test above that our integral converges because according to
1
dx
(6), the integral √ converges.
0+ x
Remark. It isn’t customary to include the + and − symbols in a+ and b− when one
writes integrals of the second kind. We only did it here to point out where they are improper,
and on which side. It doesn’t appear in the solutions to the exercises on improper integrals.

Exercises: Section 6B
 P. Probability1
1. DISCRETE RANDOM VARIABLES
1.1 Probability laws.
Suppose a repeatable event can have n distinct possible numerical outcomes x1 , . . . , xn .
For example, the event could be the tossing of a die, with its face numbers 1, 2, 3, 4, 5, 6
as the possible outcomes; in symbols, xi = i, i = 1, . . . , 6.
We assign to each possible outcome xi a number P (xi ), the probability that xi is the
outcome of the event. P can be thought of as a function whose domain is the finite set
{x1 , . . . , xn } and whose values lie between 0 and 1.
For instance, in the die-tossing example, if the die is fair (i.e., unloaded), then P (i) = 1/6,
for i = 1, . . . , 6. If the die is loaded, however, the values P (i) will not all be equal. To say
for instance that P (2) = .4 means that in a large number N of trials (i.e., rollings of the
die), the face 2 will come up in about 4/10 (or 40%) of them. More precisely, we have

Definition 1.1A Given a repeatable event with a set {x1 , . . . , xn } of n possible numerical
outcomes, the associated probability function P is defined to be the function having domain
{x1 , . . . , xn } and values
# of times xi is the outcome in N trials
P (xi ) = lim .
N →∞ N
Several important properties of probability functions are given by the Range, Addition,
and Multiplication Laws.

Range Law. (i) 0 ≤ P (xi ) ≤ 1 for all xi .

(ii) xi is never the outcome ⇒ P (xi ) = 0; xi is always the outcome ⇒ P (xi ) = 1 .
(iii) P (x1 ) + . . . + P (xn ) = 1 .

Proof. The first two follow easily from Definition 1.1A; to prove (iii), in N trials, we have
(# of outcomes x1 ) + · · · + (# of outcomes xn ) = N,
so that dividing both sides by N and taking the limit as N → ∞ gives
# of outcomes x1 # of outcomes xn
lim + · · · + lim = 1,
N →∞ N N →∞ N
which is (iii), in view of Definition 1.1A.
To state the other two probability laws, we need to extend the meaning of P so it assigns
values also to combinations of outcomes. To illustrate once again with the die-tossing:
P (2 or 4) is the probability that the outcome of a trial will be either 2 or 4;
P (2, then 4) is the probability that two trials will have successive outcomes 2 and 4.
We shall continue to use the single word “outcome” to describe these combinations, and
denote them in symbols by capital letters, writing for instance “the outcomes A = (2 or 4)
and B = (2, then 4) .”

Addition law. If xi =
% xj , P (xi or xj ) = P (xi ) + P (xj );
if xi , xj , xk are distinct, P (xi or xj or xk ) = P (xi ) + P (xj ) + P (xk ), and so on.

These and their extensions to more outcomes are proved just like (iii) above.

We say two successive trials are independent if the outcome of the first has no influence
on the outcome of the second: successive rolls of a die, for example. By contrast, succes-
1 This section, with the accompanying exercises and solutions, is based on Notes by Frank Morgan Jr.
1
2 P. PROBABILITY

sive drawings without replacement of a card from a pack (outcomes: R, B) would not be
independent, since if you first draw a red card, P (R) < 12 < P (B) for the second drawing.

Multiplication law. If A1 and A2 are outcomes for two independent trials, then
P (A1 , then A2 ) = P (A1 ) · P (A2 ).
The formula extends to N successive trials, if they are independent.

To see this intuitively for two, say P (A1 ) = 1/3, and P (A2 ) = 1/4. Then in a large number
of trial pairs, about 1/3 will have A1 as their first outcome, and of these, about 1/4 will
have A2 as their second outcome, so about 13 · 14 = 12 1
of the trial pairs will have outcome
(A1 , then A2 ).

Example 1.1A A fair die is tossed three times. What is the probability of getting an odd
number each time?

Solution. Let A denote the outcome “odd number”. By the addition law,
P (A) = P (1 or 3 or 5) = P (1) + P (3) + P (5) = 16 + 16 + 61 = 1/2 ,
and by the (extended) multiplication law, P (A, then A, then A) = 12 · 21 · 21 = 1/8.

Most of our probability work will be expressed using the idea of random variable, which
summarizes and generalizes what we have done so far.

Definition 1.1B A finite random variable X for a repeatable trial is given by

(a) a finite set {x1 , . . . , xn } of numerical values that X can take on;
(b) A probability function P (X) whose domain is {x1 , . . . , xn } and whose values P (xi )
are given by Definition 1.1A.
A finite random variable X is thus something like an ordinary real variable x, except that
it can take on only a finite set of values (unlike x, which usually takes on all real values)
and in addition with each such value xi , there is an associated probability P (xi ) that X
takes on the value xi .
Sometimes P (xi ) is written P (X = xi ) for emphasis or clarity, or to remind you of
the symbol being used for the random variable..
The probability function P (X) associated with the finite random variable X always
satisfies the three laws: Range Law, Addition Law, and Multiplication Law, since these
follow directly from Definition 1.1A.

Example 1.1B If we tabulate the shoe sizes of 24 men on a high school football squad,
the results might be (the histogram summarizes the data): 6
5
shoe size : 7 8 9 10 11 12 13 4
no. of players : 1 2 4 6 6 3 2 3
2
Here the “shoesize” random variable S takes on values from
1
the set {7, 8, . . . , 13}, and the associated probability func-
tion P (S) is defined by 7 8 9 10 11 12 13

1 1 # players wearing size i

P (S = 7) = , P (9) = , and in general: P (i) = .
24 6 24
 P. PROBABILITY 3

1.2 Expectation. To get a quick feel for what a random variable X is telling you, it
is natural to ask what its “average value” is; the official designation for this is the mean or
expected value or expectation of X, in symbols, m(X) or E(X). Treating this intuitively for
the moment, for tosses of a fair die,
1+2+3+4+5+6
expectation = average value = = 3.5
6
(Note that, like the 2.1 children in the average American family, the average value of X
won’t usually be one of its possible values.) We will learn more by calculating the average
shoe size in Example 1.1B, a.k.a. the expectation of the shoesize random variable S.
7 + 8 + 8 + 9 + 9 + 9 + 9 + . . . + 13 + 13
E(S) =
24
7 + 8 · 2 + 9 · 4 + . . . + 13 · 2
=
24
1 2 4 2
= 7· +8· +9· + . . . + 13 · ,
24 24 24 24
which can be written more expressively as
= 7 · P (7) + 8 · P (8) + 9 · P (9) + · · · + 13 · P (13) ≈ 10.3 .
The reasoning leading up to this last line leads to the definition of expectation:

Definition 1.2 The expectation (or mean or expected value) of the finite random
variable X having values x1 , . . . , xn is defined to be
n
!
(2) E(X) = x1 P (x1 ) + . . . + xn P (xn ) = xi P (xi ) .
1

Example 1.2 A die is loaded so that 6 comes up twice as often as the other five numbers.
Describe the associated random variable Y , and calculate its expectation.

Solution. Y takes on the values 1, . . . , 6.

To determine its probability function P (Y ), we have
1
P (1) = . . . = P (5) = P (6), from the loading information;
2
P (1) + P (2) + P (3) + P (4) + P (5) + P (6) = 1, by the range law (iii);

Substituting from the first line and adding, we get

1 2 1
5 · P (6) + P (6) = 1, whence P (6) = , P (1) = . . . = P (5) = .
2 7 7
Using these values, and the definition (2), we get for the expectation
1+2+3+4+5+6·2 27
E(Y ) = = = 3 76 ,
7 7
a little more weighted toward 6 than the E(X) = 3.5 calculated for the fair die.

We can think of the expectation as a weighted average of the values x1 , . . . , xn . The

ordinary average would simply add up the values and divide by n (which gives each of them
the equal weight n1 ). Instead, we weight each xi , multiplying it by the likelihood of its
occurrence, and calculate this weighted average to get the expectation.
4 P. PROBABILITY

1.3 Discrete Random Variables. There are interesting random variables which take
on an infinity of values. Our work above can be easily extended to these, provided we
assume that the set of possible values can be arranged in a list x1 < x2 < . . . < xn < . . . .
(For example, if all the values are positive integers, they can be listed in this way; however,
if the values are all the real numbers in some interval of positive length, no such list exists.)
In the definition, we will also include the case of finitely many values.

Definition 1.3 A discrete random variable X for a repeatable trial consists of

(a) a finite or infinite list x1 < x2 < x3 < . . . < xn < . . . of numerical values that X can
take on;
(b) A probability function P (X), i.e., a function whose domain is the set {x1 , . . . , xn , . . . }
and whose values P (xi ) are given by Definition 1.1A.
As before, it follows from Definition 1.1A and the three probability laws (which remain
valid when the list of xi is infinite) that
(3) 0 ≤ P (xi ) ≤ 1 and P (x1 ) + P (x2 ) + . . . + P (xn ) + . . . = 1.
The new feature here is that if the list of possible numerical outcomes is infinite, then in
(3) the sum is to be interpreted as meaning that the infinite series of probabilities converges
to the sum 1. Similarly, by analogy to the finite case, we define the expectation E(X) to
be the sum of the following series (if it converges):
∞
!
(4) E(X) = xi P (xi ) .
1
Example 1.3 A fair coin is tossed until it comes up heads for the first time. Let X be the
random variable telling the number of tosses required. Find E(X), the average number of
tosses required if you make the trial many times.

Solution. The possible values of X are the positive integers. The corresponding probabil-
ities P (n) are given by:

n: 1 2 3 4 ...
toss pattern: H TH TTH TTTH ...
P (n) : 1/2 1/4 1/8 1/16 ...
For example, if n = 3, there are 8 possible patterns for the three tosses (two possibilities
for each toss), all equally likely, but only the pattern TTH produces the value 3 for X. Or
you can use the multiplication law for probabilities: the outcomes of the three tosses are
independent events, so the probability of getting T on the first toss, T on the second, and
H on the third is 12 · 12 · 21 = 18 . According to (4),
∞
1 2 3 4 ! n
E(X) = + + + + ... = .
2 4 8 16 1
2n
To sum the series, think of its successive terms as the successive sums of the horizontal rows
in the following pattern (only the first four rows are shown):
1/2
1/4 1/4
1/8 1/8 1/8
1/16 1/16 1/16 1/16
 P. PROBABILITY 5

If we sum up the terms in this pattern by columns instead, the first column sums to 1, the
second to 1/2 (since its terms are just half those of the first column), the third similarly to
1/4, so that adding up the sums of the successive columns gives
E(X) = 1 + 1/2 + 1/4 + . . . = 2 .

1.4 The Poisson Random Variable This is a discrete random variable associated
with events which occur sparsely, but whose frequency has to be estimated and allowed for:
defects in a manufacturing process, errors in transmission or recording of data, requests for
100-year eggs at a suburban Chinese restaurant, etc.

Definition 1.4 The Poisson random variable has as its domain the non-negative
integers k = 0, 1, 2, . . . , with associated probability function P (X) given by
mk
(5) P (X = k) = C , where m > 0 is fixed and C = e−m .
k!
The value of C is dictated by the fact that the probabilities must sum to 1:
mk mk
∞
! ∞
!
C =C = C em = 1 .
0
k! 0
k!
The interpretation of the parameter m in the definition is given by

Theorem 1.4 If X is a Poisson random variable with parameter m, then E(X) = m.

Proof. We use (4) and (5), dividing both sides by C to make calculating neater and
dropping the first term (since k = 0):
1 m m2 m3 m4
E(X) = 1 · +2· +3· +4· + ...
C "1! 2! 3! 4!
m m2 m3
#
=m 1+ + + + ... ;
1! 2! 3!
E(X) = C m em = m .

Example 1.4 On average, two people during the lunchtime hour show up at the Wilbur
box office to buy tickets for “Death of a Salesperson”. What’s the probability that in that
hour: (a) no one shows up? (b) more than two show up?

Solution. Since box office visits seem infrequent, we take the number of lunchtime hour
buyers to be a Poisson random variable X with mean (i.e., average value, or parameter
value) m = 2. Using (5), (note that 0! = 1),
20
(a) P (X = 0) = e−2 · = e−2 ≈ .14;
0!
(b) P (0) + P (1) + P (2) + P (X > 2) = 1, since these exhaust the possibilities. Therefore
" 0
21 22
#
−2 2
P (X > 2) = 1 − e + +
0! 1! 2!
= 1 − e−2 (1 + 2 + 2)
= 1 − 5e−2 ≈ .32

Exercises: Section 8A
6 P. PROBABILITY

2. CONTINUOUS RANDOM VARIABLES

2.1 Probability densities and distribution functions
The variables (like x, y, or t) used in calculus are usually not discrete, but rather con-
tinuous — that is, their values do not come from a finite or infinite list or numbers, but
rather can be any real number, or any real number in some interval. For instance, if x is
the height in inches of an M.I.T. undergraduate, its values could be any real number in say
the interval [30, 84], though the extreme values would be unlikely.
We want to turn x into a random variable X, that is, assign probabilities to the values
of x. We can do this in two ways.
a) Since as a practical matter, height can only be measured to the nearest half inch, we
can declare the possible values of x to be 30, 30.5, 31, . . . , 83.5, 84, thus turning it into a
discrete variable. Then by making a histogram of the actual student heights (as we did for
the football-players’ shoe sizes), we can assign a probability to each one of these possible
heights, and thus create a discrete random variable X having x as its associated ordinary
variable.
While this will be in some sense accurate, it has the disadvantage that we will not be
able to use the procedures of calculus to study it: all calculations will have to be done by
computer, and we might not see the general principles.
b) The other way is to find a continuous differentiable function h(x) that is a good
approximation to the histogram. Then
$ b
(6) no. of students such that a ≤ X ≤ b ≈ h(x) dx
a

h(x)
To convert to probabilities, divide by the number N of students; letting f (x) = ,
N
b
no. of students such that a ≤ X ≤ b
$
(7) P (a ≤ X ≤ b) = ≈ f (x) dx.
N a

Definition 2.1A We define a continuous random variable X to be a variable x whose

values lie in (−∞, ∞), together with a probability density function f (x), such that
$ ∞
(8a) f (x) ≥ 0 for all x, f (x) dx = 1 ,
−∞

and for any two numbers a < b,

$ b
(8b) P (a ≤ X ≤ b) = f (x) dx .
a

Sometimes the interval in which the random variable X takes its values (the domain of
f (x)) is not the whole x-axis, but some smaller finite interval [x1 , x2 ], or half-infinite interval
[x1 , ∞). Then [a, b] will lie inside this interval, and the second integral in (8a) will have
limits x1 and x2 (resp. x1 , ∞).
 P. PROBABILITY 7

In P (a ≤ X ≤ b), one can replace ≤ by < without altering the value, and one can write
x instead of X and it will be accepted (grudgingly).
The density function f (x) is usually continuous, but it can have a finite number of
discontinuities (or even an infinite number, as long as it is still integrable).

For calculating probabilities, (8b) shows that we have to know the area under the density
curve y = f (x). According to the first and second fundamental theorems of calculus,
$ b $ x
(9) P (a ≤ X ≤ b) = f (x) dx = F (b) − F (a), where F (x) = f (t) dt .
a −∞

Definition 2.1B The distribution function for the random variable X, or for its asso-
ciated density function f (x) is the function
$ x
(10) F (x) = f (t) dt .
−∞

The distribution function F (x) is a particular antiderivative of the density f (x), and as
(9) shows, it is the principal tool in calculating probabilities. If F (x) is not an elementary
function, (10) shows it can be calculated by numerical integration. Users obtain its values
from tables, calculators, or mathematics programs. From (8a) and (10), we see that

(11) F (x) is increasing; lim F (x) = 0, lim F (x) = 1 .

x→−∞ x→∞

If the domain of X is the interval [x1 , x2 ], then in (10) the lower limit of the integral will
be x1 , and in (11), the −∞ and ∞ will be replaced respectively by x1 and x2 .
Definition 2.1C If X is a continuous random variable with −∞ < x < ∞ and
probability density f (x), we define its expectation by
$ ∞
(12) E(X) = x f (x) dx
−∞

This is the continuous analog of the definition (4) of the expectation of a discrete random
variable — the summation is replaced by integration, and the range of x-values is now
(−∞, ∞), rather than the discrete list x1 , x2 , . . . . As with discrete random variables, the
expectation E(X) can be thought of as the “average value” of X.
The expectation integral (12) will have limits x1 , x2 if [x1 , x2 ] is the domain of X.

Example 2.1 Uniform random variables.

What is the simplest random variable?
Among the finite random variables X of section 1, it is the uniform random variable —
the one in which all of the values x1 , . . . , xn turn up with the same probability: P (xi ) = 1/n.
Tossing a fair die gives an example.
On the other hand, if X is an infinite discrete random variable, i.e., has an infinite list
x1 , x2 , . . . of possible values, it cannot %
be uniform. For if P (xi ) has the same positive
constant value k for all i, we would have ∞ 1 P (i) = ∞, whereas we know the probabilities
have to sum to 1.
8 P. PROBABILITY

A similar situation holds for the continuous case. If the values of the continuous random
variable X all lie in the finite interval [x1 , x2 ], then we call it uniform if its density function
is constant. Since (8a) tells us the total area under the probability density curve must be
1, we have for the continuous uniform random variable on [x1 , x2 ]:
1 b−a
f (x) = (density); P (a ≤ X ≤ b) = , [a, b] ⊆ [x1 , x2 ].
x2 − x1 x2 − x1
As the integral (9) shows, for the continuous uniform random variable X, it is still the case
that P (X = x0 ) = 0 for any single value x = x0 : it is not the probability that is a positive
constant, but rather the associated probability density.
Analogously to the discrete case, if the range of a continuous random variable X is an
interval of infinite length, X cannot be uniform.

2.2 Exponential Random Variable. This is the continuous random variable whose
domain is [0, ∞), and whose density function is
e−x/m
(13) f (x) = , x≥0. exponential density
m
It depends on a single positive parameter m which is chosen to fit the model, and turns
out to be the expectation of X.

Theorem 2.2 For the exponential random variable (13) with parameter m, the distri-
bution and expectation are respectively,
(14) F (x) = 1 − e−x/m ; E(X) = m .
Proof. From the definitions (10) and (12), we have
$ x −t/m &x
e
F (x) = dt = −e−t/m = −e−x/m + 1 ;
0 m 0

and using integration by parts,

$ ∞ &∞ $ ∞
x e−x/m
dx = −x e−x/m − e−x/m dx
0 m 0 0
&∞
= 0 − 0 −e −x/m
(−m) = m.
0

Example 2.2 A certain radioactive substance with a long half-life emits a β-particle on
the average every 10 seconds. What’s the probability of waiting more than a minute for the
next emission?

Solution. Radioactive waiting times are typically modeled by an exponential random

variable X, whose variable x represents the time to the next emission. Here the average
value of x is 10, so E(X) = m = 10, and we calculate from scratch:
e
$ ∞ −x/10 &∞
P (x > 60) = dx = −e−x/10 = e−6 ≈ .002,
60 10 60

or using (14),
= 1 − P (x < 60) = 1 − F (60) = e−60/10 .
Thus the probability of waiting more than a minute is .2%, to one significant figure.

Exercises: Section 8B
 P. PROBABILITY 9

3. STANDARD DEVIATION
3.1 Variance and Standard Deviation.
The mean or expectation of the random variable X can be thought of as a parameter asso-
ciated with X which summarizes in a single number m = E(X) some important information
about X — its “average value”.
In this section we define a second parameter, called the standard deviation of X, which
measures in a single number σ (“sigma”) how widely spread out the values of X are around
their average value, weighting them according to their probability of occurring. If X is a
continuous random variable, its probability density function will tend to look like a gentle
hill centered around its mean if σ is large, but like a sudden peak arising from a relatively
flat landscape if σ is small. If X is a discrete random variable, its histogram will have a
similar appearance according to whether σ is large or small.
Mathematically, it is a little easier to work with σ 2 rather than σ itself. This quantity is
called the variance of the random variable X, and denoted as above by σ 2 . Its definition
runs as follows. For convenience, we include the definition (4) of expectation given earlier.

Definition 3.1A For a discrete random variable X, having values x1 , x2 , . . . we define

!
(4) m = xi P (xi ) (mean, or expectation, of X)
i
! !
2
(15) σ = (xi − m)2 P (xi ) = x2i P (xi ) − m2 (variance of X)
i i
√
σ = σ2 (standard deviation of X)

Remarks. The sums will be from 1 to n or from 1 to ∞ according to whether X is

respectively a finite or infinite discrete random variable.
The two sums in (15) are equal, since (xi − m)2 = x2i − 2mxi + m2 , and therefore
∞
! ! ! !
(xi − m)2 P (xi ) = x2 P (xi ) − 2m xi P (xi ) + m2 P (xi )
1 i i i
!
= x2 P (xi ) − 2m · m + m2 · 1
i
!
= x2 P (xi ) − m2 ,
i

using the definition (4) of expectation, and the fact that the probabilities have to sum to 1.
Why do we give the variance in (15) using two forms for the sum?
The first form of the sum shows that variance is always non-negative (since (xi − m)2
and P (xi ) are), so that defining σ as its (positive) square root is a legal operation. The
first sum also shows the significance of the variance: if the xi that are far from the mean
m occur with high probability — in other words, if the spread is large — then the variance
will be large.
The second form of the sum in (15) leads to a simpler calculation, assuming that the
mean m will be calculated first.
10 P. PROBABILITY

Example 3.1 Find the standard deviation of the shoe-sizes in Example 1.1B.

Solution. Using the second sum in (15), with m = 10.3 (see p.3, line 13), we have

σ 2 = 72 · P (7) + 82 · P (8) + 92 · P (9) + . . . + 132 · P (13) − (10.3)2

1 2 4 2
= 72 · + 82 · + 92 · + . . . + 132 · − (10.3)2
24 24 24 24
= 2.118 ;
√
σ = 2.118 ≈ 1.45 = 1.5, to one decimal place.

The definition of variance and standard deviation for a continuous random variable is just
the natural analog of the above; again, we include the definition (12) of “mean”:

Definition 3.1B For a continuous random variable X with domain (−∞, ∞) and
probability density f (x), we define
$ ∞
(12) m = x f (x) dx (mean, or expectation, of X)
−∞
$ ∞ $ ∞
(16) σ2 = (x − m)2 f (x) dx = x2 f (x) dx − m2 (variance of X);
−∞ −∞
√
σ = σ2 (standard deviation of X)

The two integrals in (16) are equal by an argument which is the natural analog of the one
we gave for the sums in (15); and just as for the sums, the right- hand integral is usually
easier to calculate, but it is the left-hand integral that shows the variance is non-negative,
and therefore has a square root.

Theorem 3.1 For the Poisson random variable X with parameter m, (cf. (5)),
√
(17) E(X) = m, σ(X) = m.

For the exponential random variable X with parameter m, (cf. (13)),

(18) E(X) = m, σ(X) = m .

Proof. The values of the two expectations were calculated in Theorems 1.4 and 2.2,
respectively. The calculation of the standard deviations are left as exercises.

Exercises: Section 8C
 P. PROBABILITY 11

4. NORMAL RANDOM VARIABLES

4.1 The Standard Normal Random Variable. This is the most widely-used con-
tinuous random variable — it’s the one whose associated density function has the infamous
“Curve” as its graph. We will use the notations Z for this random variable, z for the as-
sociated numerical variable z, φ(z) for its density function, and Φ(z) for its distribution
function.
Though it has many applications, perhaps the most basic is to describe how a large
number of measurements of some physical quantity are distributed about the true value.
For example, suppose we use a large number of cheap thermometers to measure the tem-
perature (in o C) of a container of melting ice. The true value is 0, but some will read higher,
some lower. Most will read between −1 and 1; if the quality control at the thermometer
factory is just at the right level of incompetence, the histogram of the thermometer readings
will be approximated by the density function below, and the collection of readings will be a
typical set of outcomes for the associated random variable Z.

Definition 4.1 The standard normal random variable Z is the continuous random
variable having as its density, distribution, and associated probabilities (in z, w-coordinates)

1 2
(19) w = φ(z) = √ e−z /2 (standard normal density)
2π
$ z
1 2
(20) Φ(z) = √ e−t /2 dt; (standard normal distribution)
2π −∞
$ b
1 2
(21) P (a ≤ Z ≤ b) = √ e−z /2 dz = Φ(b) − Φ(a) .
2π a

√
The factor 1/ 2π is used in the density function (19) so that the total area under φ(z)
will equal 1. This follows from the statements in (22) below:
• the value of the integral on the left is “well-known” (it will be
√ calculated in 18.02);
• the value of the middle integral then follows by setting z = t 2;
• the value of the integral on the right follows since the integrand is an even function.

√
√
'
∞
π ∞
π ∞
$ $ $
2 2 2
/2 /2
(22) e−t dt = ⇒ e−z dz = ⇒ e−z dz = 2π .
0 2 0 2 −∞

The last equality shows that the total area under the graph of the density function (19) is
1, as required by (8a).

Since Φ(z) is not an elementary function, it must be calculated by numerical integration;

this means in practice that its values are obtained from tables, by pressing calculator buttons,
or from computer programs like Matlab, Maple, or Mathematica.

The table at the top of the next page gives values of the standard normal distribution
Φ(Z) to four decimal places, for z = 0 to z = 3, in increments of .1 . You will need these for
the problems to calculate numerical values of probability, by using (21). Values for z < 0
can be obtained by symmetry, as explained after the table.
12 P. PROBABILITY

Table of values for Φ(Z), Z≥0

z: 0 .1 .2 .3 .4 .5 .6 .7 .8 .9
Φ(z): .5000 .5398 .5793 .6179 .6554 .6915 .7257 .7580 .7881 .8159
z: 1.0 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9
Φ(z): .8413 .8643 .8849 .9032 .9192 .9332 .9452 .9554 .9641 .9713
z: 2.0 2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9 3.0
Φ(z): .9772 .9821 .9861 .9893 .9918 .9938 .9953 .9965 .9974 .9981 .9987

To get values of Φ(z) for negative z, we use the symmetry of φ(z) about z = 0. Still
assuming z ≥ 0, we have
$ −z $ ∞
(23) Φ(−z) = φ(t) dt = φ(t) dt = 1 − Φ(z)
−∞ z

The most frequent application is to intervals symmetric around z = 0; using (21) and (23),

(24) P (−a < Z < a) = Φ(a) − Φ(−a) = 2Φ(a) − 1 .

Using the table for z = 1.0 and z = 2.0, this gives the approximate values most widely used
in statistics (cf. Simmons, bottom p. 419 for a picture showing these values visually):

(25) P (−1 < Z < 1) = .68 ≈ 2/3; P (−2 < Z < 2) = .95 .

The intervals in (25) are chosen because of their relation to the mean and standard deviation
of Z:

Theorem 4.1 The mean and standard deviation of the standard normal random variable
Z are respectively

(26) E(Z) = 0; σ(Z) = 1 .

2
Proof. The integral (12) giving E(Z) has the integrand √12π z e−z /2 which is an odd
function; therefore its integral over the symmetric interval (−∞, ∞) is zero.
The √
integral for the variance can be calculated using integration by parts; the result is 1,
so σ = 1 = 1. We leave the details as an exercise.
Thus the values in (25) give the respective probabilities that Z is within one or two
standard deviations from its mean value.
 P. PROBABILITY 13

4.2 The Normal Random Variable.

To be adaptable to any situation, we need to modify the standard normal density function
w = φ(z) given by (19) in two ways.
First Modification. Suppose our factory starts producing less accurate thermome-
ters. The measurements of the melting ice should then still be clustered around 0, but
spread out more, though the density function (19) should still have the same general prop-
erties and shape as before. To achieve this, we stretch the z-axis by a positive scale factor
which we will call σ, since it will turn out to be the standard deviation of the new density
function. (In this example σ > 1, though in general it need not be). Using variables, the
way to do this stretching is to make a linear change of variable from z to x1 , say:
x1
z= (so that x1 = σ corresponds to z = 1) .
σ
At the same time, shrink the w-axis by the factor σ (the reason will be given in a moment)
by making the change of variable w = σ y (so that w = σ corresponds to y = 1). Under
these two operations, the standard normal density function φ(z) then becomes in x1 , y-
coordinates,

1 1 1 2
(27) y= φ(x1 /σ), i.e, √ e− 2 (x1 /σ) .
σ σ 2π

Remark. We shrink the w-axis by 1/σ so the total area under (19) will still be 1.
This can be proved formally by changing the variable in the integral for the area, but to
see it intuitively, imagine the area under φ(z) divided into equal tiny squares of side % say,
and area %2 . Stretching the z-axis and shrinking the w-axis by the same factor σ turns each
%-square into a rectangle of horizontal side σ% and vertical side %/σ, whose area is still %2 .
Since the stretching and shrinking changes each little square into a rectangle, but does
not change its area or the total number of little squares or rectangles, it does not change the
total area under φ(z): it started out 1 and remains 1 after the z-stretching and w-shrinking.
Second Modification. If we use these same thermometers to measure some other
temperature m, the graph of the density function should be moved so that its maximum
lies over the point m; to achieve this, we make the further change of variable x1 = x − m
(so that x1 = 0 corresponds to x = m), giving as our final result:
x−m
Definition-Theorem 4.2A Under the change of variables z = , w = σ y , the
σ
standard normal density function (19) becomes, in x, y-coordinates,

1 1 2 2
(28) y = φm,σ (x) = √ e− 2 (x−m) /σ , the normal density function
σ 2π

and its associated normal distribution function is

$ x
1 1 2 2
(29) Φm,σ (x) = √ e− 2 (t−m) /σ dt the normal distribution
σ 2π −∞

The normal density function (28) (as well as its associated random variable Xm,σ and
normal distribution function (29)) thus depends on two parameters m and σ, which turn
out to be the mean and standard deviation of Xm,σ .
14 P. PROBABILITY

1 1 2 2
Theorem 4.2B For the normal density function (19) φm,σ (x) = √ e− 2 (x−m) /σ ,
σ 2π

1
$ ∞
1 2 2
√ x e− 2 (x−m) /σ dx = m ;
σ 2π −∞
1
$ ∞
1 2 2
√ (x − m)2 e− 2 (x−m) /σ dx = σ 2 .
σ 2π −∞

Proofs. Use the change of variables in Theorem 4.2A to simplify both integrals; you’ll need
(20) for the first, and integration by parts for the second. It’s a good exercise in changing
variables and integration techniques; if you get stuck, it’s worked out in Simmons, p.418
(using t instead of z, and a slightly different change of variable, but the calculations are
basically the same).
The Simmons Calculus has in 12.5, p.419 several graphs illustrating how the normal
density curve looks for several different values of the standard deviation σ and the mean m.
Probability calculations for X = Xm,σ and the normal distribution Φm,σ are done using
the table on p. 11 for the standard Φ(Z) and the change of variable in Theorem 4.2A:

x−m
z= ; x = σz + m .
σ
This gives the basic formula relating the normal and standard distributions:
" #
a−m b−m
(30) P (a < X < b) = P <Z< ,
σ σ

or using the change of variable in the other direction, (25) turns into the rule-of-thumb
formulas most frequently used in statistical applications:

(31) P (m − σ < X < m + σ) = .68 ≈ 2/3

(32) P (m − 2σ < X < m + 2σ) = .95 .

Example 4.2A If the height in inches of a typical male Bostonian is a normal random
variable H having a mean of 68 and a standard deviation of 4, what percentage will be able
to walk through a 6’4” doorframe without having to stoop?

Solution. The doorframe is 76 inches. Since 76 = 68 + 2·4, the door height is two standard
deviations from the mean, so that (32) tells us

1
P (H > 76) ≈ (.05) = .025;
2

thus about 2 21 % will have to stoop, so that about 97 21 % will not have to.
 P. PROBABILITY 15

Example 4.2B In the end-of-term evaluation forms, a calculus lecturer had an average
rating of 4.5, with σ = 1.5. What fraction of the students gave the rating:
a) 4 or 5? b) between 3 and 6?

Solution. For (a), we use (30); letting X = X4.5,1.5 , we have

" #
.5 .5
P (4 < X < 5) = P − <Z<
1.5 1.5
= 2Φ(1/3) − 1 = .26 ≈ 1/4, by (24) and the table for Φ(Z).

For (b), we can use (31), since 3 and 6 are one standard deviation away from 4.5:

P (3 < X < 6) = .68 ≈ 2/3 .

Exercises: Section 8D
16 P. PROBABILITY

5. CENTRAL LIMIT THEOREM AND APPLICATIONS

5.1 Central limit theorem. This theorem is one of the main applications of the normal
distribution. It’s the basis for the polling and sampling techniques which try to predict how
a population thinks, looks, and behaves from questioning just a few people.
Suppose X is a random variable — discrete or continuous — giving the possible outcomes
of some experiment, with the associated probabilities for each outcome (or the probability
density, if it’s a continuous random variable). It has a mean m and standard deviation σ.
Common sense, or what is often called the “law of averages”, says that though the
outcome of any particular trial of the experiment may not come out close to m, if you
repeat the experiment many times and take the average of the results, that ought to be
close to the true mean. In fact, if you didn’t know in advance what the mean was, that
would be a good way of determining it: the more trials you average in, the closer your
answer should be to the mean.
The central limit theorem makes this idea more precise. In it, one can think of X as
the random variable describing the numerical outcome of an experiment, and X1 , . . . , Xn
as the identical random variables describing respectively the outcomes from the first trial of
the experiment, the second, and so on down to the n-th trial. Then X̄ is the new random
variable whose outcome is the average of the results of these n trials.

Theorem 5.1 The Central Limit Theorem. Let X be a discrete or continuous

random variable, with mean m and standard deviation σ, and let X1 , X2 , . . . , Xn be n
copies of X. Then if n is large,

X1 + . . . + Xn
(33) X̄ =
n
is approximately a normal random variable, whose mean and standard deviation are
σ
(34) E(X̄) = m, σ(X̄) = √ .
n

The definition of X̄ in the theorem is incomplete, since we haven’t said how to assign
probabilities or a probability density to the values of X̄, based on those of X. However, it
is clear from the experimental trial viewpoint that these probabilities or density function do
exist — we would need to know them if we were trying to prove the Central Limit Theorem,
but they aren’t necessary if we just want to use it.
The theorem has two points:
• even though X may not be a normal random variable, X̄ will be, if n is large enough,
at least approximately, so that we can apply the results of section 4 to it;
• when n is large, the standard deviation of X̄ is small, which says that the graph of the
density function for X̄ has most of its area in a sharp peak around the mean; so the average
of n trials will with high probability be close to the true mean.
 P. PROBABILITY 17

Example 5.1A “Press 0 to get a representative; this call may be monitored for quality”. . .
The average time a Sleep-Tight Motel representative spends on the telephone taking a
reservation is 5 minutes, with a standard deviation of 2 minutes.
One outsourced representative, Dora Chatemup, was monitored for 100 calls, and spent
an average of 5.5 minutes per call. What are the chances that she was following the guidelines
in the Motel’s How-to-Take-Reservations Manual, and just happened to get an unusually
large number of hard-to-satisfy callers?

Solution. Let X be the random variable giving the length of a call; according to the data
given, m(X) = 5, and σ(X) = 2. √
Let X̄ be the average of 100 calls; then m(X̄) = 5, and σ(X̄) = 2/ 100 = .2, by the
Central Limit Theorem. Using (30) and the table for Φ(Z),
" #
5.5 − 5
P (X̄ ≥ 5.5) = P Z ≥ = P (Z ≥ 2.5) = 1 − Φ(2.5) = .0062,
.2

so the probability is only around 12 % that her average time of 5.5 minutes was due just to
the luck of the draw (i.e., chance).

Example 5.1B We want to test whether a psychic has Extra-Sensory Perception by seeing
if he can guess whether the next number emitted by a random number generator will be
even or odd. How many numbers would he have to guess at, for a 51% correct guess rate
to have a 1% probability of being due to chance alone?

Solution. Let X be the random variable which has the value 1 if the guess is correct, 0
otherwise. Assuming ESP doesn’t exist, the associated probabilities are P (1) = P (0) = .5,
the mean and standard deviation are (using (15)):

m(X) = .5, σ 2 (X) = 12 (.5) − (.5)2 = .25, σ(X) = .5 .

If n guesses are made and averaged to get the random variable X̄, the Central Limit
Theorem says that if n is large,

.5
m(X̄) = .5 ; σ(X̄) = √
n

Using (30),
√
( .01 n ) √ √
P (X̄ ≥ .51) = P Z ≥ = P (Z ≥ .02 n) = 1 − Φ(.02 n)
.5

When is this probability ≤ 1%? From the table, calculating roughly,

√ √
1 − Φ(.02 n) = .01 ⇒ .02 n > 2.4 ⇒ n > (120)2 = 14, 400.

So a score of 51% correct out of say 15,000 guesses would have only a 1% probability of
being due to chance alone – or as one says, would confirm the existence of ESP at the 99%
confidence level (unless the psychic cheated, something professional magicians have been
better at detecting than professional scientists.)
18 P. PROBABILITY

5.2 Polling a random sample An application of statistics that everyone is familiar

with is polling — selecting a random sample of the population, asking who they favor in an
election, and from this concluding how the population as a whole will vote. How does one
determine how large the sample should be, and report the confidence level of the poll —
what the margin of error is, and perhaps what the chances of error are?
Assume there are two candidates — Hack and Shifty. Let p denote the unknown fraction
of the voting population that favors Hack; then 1 − p will be the fraction that favors Shifty.
Our aim is to find upper and lower limits on the value of p, with say 95% probability that
our limits are correct. How many voters will we have to poll to achieve this?
1, Hack favored, P (1) = p;
*
Consider the “polling” random variable defined by X =
0, Shifty favored, P (0) = 1 − p .
Using (15), we calculate its mean, variance, and standard deviation to be
+
(35) m(X) = p, σ 2 (X) = p − p2 , σ= p − p2

If we think of X1 , X2 , . . . , Xn , as the random variables representing the outcomes from

asking the first voter, the second, and so on up to the n-th in the sample, the average of
these X̄ represents the fraction of this sample of size n that favors Hack. By the Central
Limit Theorem
σ
m(X̄) = p, σ̄ = σ(X̄) = √
n
Therefore using (32), we get
, -
(36a) P m(X̄) − 2σ̄ < X̄ < m(X̄) + 2σ̄ = .95 ,

or rewriting this to make it more compact,

, -
(36b) P |X̄ − p| < 2σ̄ = .95 .

Or in the informal way the result is usually said (remember that p is what we want to know,
but the value of X̄ is the experimental outcome of polling the sample of size n)

(36c) p = X̄ ± 2σ̄ with 95% confidence .

√ +
We don’t know how big σ̄ = σ/ n is, since it depends on σ = p − p2 , and the value of p
is unknown. However, we can estimate it: by elementary calculus, p − p2 has its maximum
point at ( 12 , 41 ); using this, and then the Central Limit Theorem,

1 + 1 1
(37) p − p2 ≤ ⇒ σ= p − p2 ≤ ⇒ σ̄ ≤ √ .
4 2 2 n

Since the probability in (36) — the confidence level in (36c) — can only rise if we replace
2σ̄ by a larger number, (36c) and (37) imply our final answer:

1
(38) p = X̄ ± √ with 95% confidence .
n
 P. PROBABILITY 19

Example 5.1A 100 entering first-graders picked at random from the seven Cambridge
elementary schools were asked if they believed in the tooth fairy. 45 said “yes”, 55 said
“no”. With 95% confidence, what fraction p of Cambridge first-graders are believers?

Solution. Using (38) with n = 100, we get p = .45 ± .10; in other words,

.35 < p < .55 with 95% confidence,

or as the local news would report it, “at the 95% confidence level, 45% of the first-graders
are believers, with a 10% margin for error.” (They’d probably leave out the 95% confidence
clause.)

Example 5.1B How large a sample would be needed to use the results of a poll to predict
with 95% confidence the percentage of voters favoring Hack, with a 1% margin of error?
√
Solution. Again using (38), we want n = 100, so n = 10, 000.

Remark. The real difficulty in this work is in obtaining a sample that is truly random,
and not biased by neighborhood, social class, race, religion, etc. There is an equal problem
in getting reliable answers, which even if honestly given can depend on how the question is
phrased, the tone of voice, and even the expression on the pollster’s face, if it’s insufficiently
poker.

Exercises: Section 8E
 18.01 EXERCISES

Unit 1. Differentiation

1A. Graphing

1A-1 By completing the square, use translation and change of scale to sketch
a) y = x2 − 2x − 1 b) y = 3x2 + 6x + 2

1A-2 Sketch, using translation and change of scale

2
a) y = 1 + |x + 2| b) y =
(x − 1)2
1A-3 Identify each of the following as even, odd , or neither
x3 + 3x
a) b) sin2 x
1 − x4
tan x
c) d) (1 + x)4
1 + x2
e) J0 (x2 ), where J0 (x) is a function you never heard of

1A-4 a) Show that every polynomial is the sum of an even and an odd function.
b) Generalize part (a) to an arbitrary function f (x) by writing

f (x) + f (−x) f (x) − f (−x)

f (x) = +
2 2
Verify this equation, and then show that the two functions on the right are respectively even
and odd.
1
c) How would you write as the sum of an even and an odd function?
x+a
1A-5. Find the inverse to each of the following , and sketch both f (x) and the inverse
function g(x). Restrict the domain if necessary. (Write y = f (x) and solve for y; then
interchange x and y.)
x−1
a) ) b) x2 + 2x
2x + 3
1A-6 Express√in the form A sin (x + c)
a) sin x + 3 cos x b) sin x − cos x

1A-7 Find the period , amplitude , and phase angle, and use these to sketch
a) 3 sin (2x − π) b) −4 cos (x + π/2)

1A-8 Suppose f (x) is odd and periodic. Show that the graph of f (x) crosses the x-axis
infinitely often.

c
!Copyright David Jerison and MIT 1996, 2003
 E. 18.01 EXERCISES

1A-9 a) Graph the function f that consist of straight line segments joining the points
(−1, −1), (1, 2), (3, −1), and (5, 2). Such a function is called piecewise linear.
b) Extend the graph of f periodically. What is its period?
c) Graph the function g(x) = 3f ((x/2) − 1) − 3.

1B. Velocity and rates of change

1B-1 A test tube is knocked off a tower at the top of the Green building. (For the purposes
of this experiment the tower is 400 feet above the ground, and all the air in the vicinity of
the Green building was evacuated, so as to eliminate wind resistance.) The test tube drops
16t2 feet in t seconds. Calculate
a) the average speed in the first two seconds of the fall
b) the average speed in the last two seconds of the fall
c) the instantaneous speed at landing

1B-2 A tennis ball bounces so that its initial speed straight upwards is b feet per second.
Its height s in feet at time t seconds is given by s = bt − 16t2
a) Find the velocity v = ds/dt at time t.
b) Find the time at which the height of the ball is at its maximum height.
c) Find the maximum height.
d) Make a graph of v and directly below it a graph of s as a function of time. Be sure
to mark the maximum of s and the beginning and end of the bounce.
e) Suppose that when the ball bounces a second time it rises to half the height of the
first bounce. Make a graph of s and of v of both bounces, labelling the important points.
(You will have to decide how long the second bounce lasts and the initial velocity at the
start of the bounce.)
f) If the ball continues to bounce, how long does it take before it stops?

1C. Slope and derivative

1C-1 a) Use the difference quotient definition of derivative to calculate the rate of change
of the area of a disk with respect to its radius. (Your answer should be the circumference
of the disk.)
b) Use the difference quotient definition of derivative to calculate the rate of change
of the volume of a ball with respect to the radius. (Your answer should be the surface area
of the ball.)

1C-2 Let f (x) = (x − a)g(x). Use the definition of the derivative to calculate that f ! (a) =
g(a), assuming that g is continuous.

1C-3 Calculate the derivative of each of these functions directly from the definition.
a) f (x) = 1/(2x + 1) b) f (x) = 2x2 + 5x + 4
√
c) f (x) = 1/(x2 + 1) d) f (x) = 1/ x
e) For part (a) and (b) find points where the slope is +1, −1, 0.
 1. DIFFERENTIATION

1C-4 Write an equation for the tangent line for the following functions
a) f (x) = 1/(2x + 1) at x = 1 b) f (x) = 2x2 + 5x + 4 at x = a
√
c) f (x) = 1/(x2 + 1) at x = 0 d) f (x) = 1/ x at x = a

1C-5 Find all tangent lines through the origin to the graph of y = 1 + (x − 1)2 .

1C-6 Graph the derivative of the following functions directly below the graph of the
function. It is very helpful to know that the derivative of an odd function is even and the
derivative of an even function is odd (see 1F-6).

a) semicircle b) parabola c) odd function d) even function e) periodic; period = ?

2
-2 4
2
-2

1D. Limits and continuity

1D-1 Calculate the following limits if they exist. If they do not exist, then indicate whether
they are +∞, −∞ or undefined.
4 4x 4x2
a) lim b) lim c) lim
x→0 x − 1 x→2 x + 1 x→−2 x + 2
4x2 4x2 4x2
d) lim+ e) lim− f) lim
x→2 2 − x x→2 2 − x x→∞ x − 2
2 2
4x x + 2x + 3 x−2
g) lim − 4x i) lim j) lim 2
x→∞ x − 2 x→∞ 3x2 − 2x + 4 x→2 x − 4

1D-2 Repeat 1D-1 for each of the following, allowing +∞, −∞ as before. If the limit is
undefined, evaluate the corresponding one-sided limits that exist.

√ 1 1 |x|
a) lim x b) lim c) lim d) lim | sin x| e) lim
x→0 x→1 x−1 x→1 (x − 1)4 x→0 x→0 x

1D-3 Identify and give the type of the points of discontinuity of each of the following:

x4 x + a, x > 0
!
x−2 1
a) b) c) d) f (x) =
x2 − 4 sin x x3 a − x, x<0
d
e) f ! (x), for the f (x) in d) f) (f (x))2 , where f (x) = |x|
dx

4x2 1
1D-4 Graph the functions: a) (See 1D-1efg.) b)
x−2 x2 + 2x + 2
 E. 18.01 EXERCISES

ax + b, x ≥ 1;
!
1D-5 Define f (x) =
x2 , x < 1.
a) Find all values of a, b such that f (x) is continuous.
b) Find all values of a, b such that f ! (x) is continuous. (Be careful!)

1D-6 For each of the following functions, find all values of the constants a and b for which
the function is differentiable.
! 2 ! 2
x + 4x + 1, x ≥ 0; x + 4x + 1, x ≥ 1;
a) f (x) = b) f (x) =
ax + b, x < 0. ax + b, x < 1.
1D-7 Find the values of the constants a, b and c for which the following function is
differentiable. (Give a and b in terms of c.)

cx2 + 4x + 1, x ≥ 1;
!
f (x) =
ax + b, x < 1.

1D-8 For each of the following functions, find the values of the constants a and b for which
the function is continuous, but not differentiable.
ax + b, x > 0; ax + b, x > 0;
! !
a) f (x) = b) f (x) =
sin 2x, x ≤ 0. cos 2x, x ≤ 0.
1D-9 Find the values of the constants a and b for which the following function is differen-
tiable, but not continuous.
ax + b, x > 0;
!
f (x) =
cos 2x, x ≤ 0.
1D-10* Show that
f (a + h) − f (a)
g(h) = has a removable discontinuity at h = 0 ⇐⇒ f ! (a) exists.
h

1E. Differentiation formulas: polynomials, products, quotients

1E-1 Find the derivative of the following polynomials.

a) x10 + 3x5 + 2x3 + 4 b) e2 + 1 ( e = base of natural logs)
c) x/2 + π 3 d) (x3 + x)(x5 + x2 )

1E-2 Find the antiderivative of the following polynomials.

a) ax + b (a and b are constants.) b) x6 + 5x5 + 4x3
c) (x3 + 1)2

1E-3 Find the points (x, y) of the graph y = x3 + x2 − x + 2 at which the slope of the
tangent line is horizontal.

1E-4 For each of the following, find all values of a and b for which f (x) is differentiable.
! 2 ! 2
ax + bx + 4, x ≤ 0; ax + bx + 4, x ≤ 1;
a) f (x) = b) f (x) =
5x5 + 3x4 + 7x2 + 8x + 4, x > 0. 5x5 + 3x4 + 7x2 + 8x + 4, x > 1.
 1. DIFFERENTIATION

1E-5 Find the derivatives of the following rational functions.

x x+a
a) b) 2 (a is constant)
1+x x +1
4
x+2 x +1
c) 2 d)
x −1 x
1F. Chain rule, implicit differentiation

1F-1 Find the derivative of the following functions:

a) (x2 + 2)2 (two methods)
b) (x2 + 2)100 . Which of the two methods from part (a) do you prefer?

1F-2 Find the derivative of x10 (x2 + 1)10 .

1F-3 Find dy/dx for y = x1/n by implicit differentiation.

1F-4 Calculate dy/dx for x1/3 + y 1/3 = 1 by implicit differentiation. Then solve for y and
calculate y ! using the chain rule. Confirm that your two answers are the same.

1F-5 Find all points of the curve(s) sin x + sin y = 1/2 with horizontal tangent lines. (This
is a collection of curves with a periodic, repeated pattern because the equation is unchanged
under the transformations y → y + 2π and x → x + 2π.)

1F-6 Show that the derivative of an even function is odd and that the derivative of an odd
function is even.
(Write the equation that says f is even, and differentiate both sides, using the chain rule.)

1F-7 Evaluate the derivatives. Assume all letters represent constants, except for the
independent and dependent variables occurring in the derivative.
" dD m0 dm
a) D = (x − a)2 + y0 2 , b) m==? " , =?
dx 2
1 − v /c 2 dv
mg dF at dQ
c) F = , =d)
? Q= , =?
(1 + r2 )3/2 dr (1 + bt2 )3 dt

1F-8 Evaluate the derivative by implicit differentiation. (Same assumptions about the
letters as in the preceding exercise.)
1 dr dP
a) V = πr2 h, = ? b) P V c = nRT, =?
3 dh dV
da
c) c2 = a2 + b2 − 2ab cos θ, =?
db

1G. Higher derivatives

1G-1 Calculate y !! for the following functions.

√ x
a) 3x2 + 2x + 4 x b)
x+5
−5 x2 + 5x
c) d)
x+5 x+5
1G-2 Find all functions f (x) whose third derivative f !!! (x) is identically zero. (“Identically”
is math jargon for “always” or “for every value of x”.)
 E. 18.01 EXERCISES

1G-3 Calculate y !! using implicit differentiation and simplify as much as possible.

x2 a2 + y 2 b2 = 1

1G-4 Find the formula for the nth derivative y (n) of y = 1/(x + 1).

1G-5 Let y = u(x)v(x).

a) Find y ! , y !! , and y !!! .
b) The general formula for y (n) , the n-th derivative, is called Leibniz’ formula: it uses
the same coefficients as the binomial theorem , and looks like
# $ # $
(n) (n) n (n−1) (1) n (n−2) (2)
y =u v+ u v + u v + ... + uv (n)
1 2

Use this to check your answers in part (a), and use it to calculate y (p+q) , if y = xp (1 + x)q .

1H. Exponentials and Logarithms: Algebra

1H-1 The half-life λ of a radioactive substance decaying according to the law y = y0 e−kt
is defined to be the time it takes the amount to decrease to 1/2 of the initial amount y0 .
a) Express the half-life λ in terms of k. (Do this from scratch — don’t just plug into
formulas given here or elsewhere.)
b) Show using your expression for λ that if at time t1 the amount is y1 , then at time
t1 + λ it will be y1 /2, no matter what t1 is.

1H-2 If a solution containing a heavy concentration of hydrogen ions (i.e., a strong acid)
is diluted with an equal volume of water, by approximately how much is its pH changed?
(Express (pH)diluted in terms of (pH)original .)

1H-3 Solve the following for y:

a) ln(y + 1) + ln(y − 1) = 2x + ln x b) log(y + 1) = x2 + log(y − 1)
c) 2 ln y = ln(y + 1) + x
ln a
1H-4 Solve = c for a in terms of b and c; then repeat, replacing ln by log.
ln b
1H-5 Solve for x (hint: put u = ex , solve first for u):
ex + e−x
a) = y b) y = ex + e−x
ex − e−x
1H-6 Evaluate in a simpler form not involving logs the number A = log e · ln 10. Then
generalize the problem, replacing the numbers and log bases e and 10 by a and b respectively.

1H-7 The decibel scale of loudness is

L = 10 log10 (I/I0 )

where I, measured in watts per square meter, is the intensity of the sound and I0 = 10−12
watt/m2 is the softest audible sound at 1000 hertz. Classical music typically ranges from
30 to 100 decibels. The human ear’s pain threshold is about 120 decibels.
 1. DIFFERENTIATION

a) Suppose that a jet engine at 50 meters has a decibel level of 130, and a normal
conversation at 1 meter has a decibel level of 60. What is the ratio of the intensities of the
two sounds?
b) Suppose that the intensity of sound is proportional to the inverse square of the
distance from the sound. Based on this rule, calculate the decibel level of the sound from
the jet at a distance of 100 meters, at distance of 1 km.1
1H-8* The mean distance of each of the planets to the Sun and their mean period of
revolution is as follows.2 (Distance is measured in millions of kilometers and time in Earth
years.)

Mercury Venus Earth Mars Jupiter Saturn Uranus Neptune Pluto

57.9 108 150 228 778 1, 430 2, 870 4, 500 5, 900
0.241 0.615 1.00 1.88 11.9 29.5 84.0 165 248

a) Find the pattern in these data by making a graph of (ln x, ln y) where x is the
distance to the Sun and y is the period of revolution for the first four points (Mercury
through Mars). Observe that these points are nearly on a straight line. Plot a line with
ruler and estimate its slope. (You can check your estimated slope by calculating slopes of
lines connecting consecutive data points.)
b) Using an approximation to the slope m that you found in part (a) accurate to two
significant figures, give a formula for y in the form

ln y = m ln x + c

(Use the Earth to evaluate c.)

c) Solve for y and make a table for the predicted values of the periods of revolutions
of all the planets based on their distance to the Sun. (Your answers should be accurate to
one percent.)
d) The Earth has radius approximately 6, 000 km and the Moon is at a distance of
about 382, 000 km. The period of revolution of the Moon is a lunar month, say 28 days.
Assume that the slope m is the same for revolution around the Earth as the on you found
for revolution around the Sun in (a). Find the distance above the surface of the Earth of
geosynchronous orbit, that is, the altitude of the orbit of a satellite that stays above one
place on the equator. (For satellites this close to Earth it is important to know that y is
predicting the distance from the satellite to the center of the Earth. This is why you need
to know the radius of the Earth.)
e) Find the period of revolution of a satellite that circles at an altitude of 1, 000 km.

1 The inverse square law is justified by the fact that the intensity is measured in energy per unit time
per unit area. When the sound has travelled a distance r, the energy of a sound spread over a sphere of
radius r centered at the source. The area of that sphere is proportional to r 2 , so the average intensity is
proportional to 1/r2 . Fortunately for people who live near airports, sound doesn’t travel as well as this.
Part of the energy is dissipated into heating the air and another part into vibration of insulating materials
on the way to the listener’s ear.
2 from “Fundamentals of Physics, vol. 1,” by D. Halliday and R. Resnick
 E. 18.01 EXERCISES

1I. Exponential and Logarithms: Calculus

1I-1 Calculate the derivatives

2
a) xex b) (2x − 1)e2x c) e−x
d) x ln x − x e) ln(x2 ) f) (ln x)2
2
g) (ex )2 h) xx i) (ex + e−x )/2
j) (ex − e−x )/2 k) ln(1/x) l) 1/ ln x
m) (1 − ex )/(1 + ex )

1I-2 Graph the function y = (ex + e−x )/2 .

1
1I-3 a) Evaluate lim n ln(1 + ). Hint: Let h = 1/n, and use (d/dx) ln(1 + x)|x=0 = 1.
n→∞ n
1
b) Deduce that lim (1 + )n = e.
n→∞ n
1 n
1I-4 Using lim (1 + ) = e, calculate
n→∞ n
1 3n 2 5n 1 5n
a) lim (1 + ) b) lim (1 + ) c) lim (1 + )
n→∞ n n→∞ n n→∞ 2n
1I-5* If you invest P dollars at the annual interest rate r, then after one year the interest
is I = rP dollars, and the total amount is A = P + I = P (1 + r). This is simple interest.
For compound interest, the year is divided into k equal time periods and the interest is
calculated and added to the account at the end of each period. So at the end of the first
period, A = P (1 + r( k1 )); this is the new amount for the second period, at the end of which
A = P (1 + r( k1 ))(1 + r( k1 )), and continuing this way, at the end of the year the amount is
% r &k
A=P 1+ .
k
The compound interest rate r thus earns the same in a year as the simple interest rate of
% r &k
1+ −1 ;
k

this equivalent simple interest rate is in bank jargon the “annual percentage rate” or APR.3
a) Compute the APR of 5% compounded monthly, daily,4 and continuously. Contin-
uous compounding means the limit as k tends to infinity.
b) As in part (a), compute the APR of 10% compounded monthly, biweekly (k=26),
daily, and continuously. (We have thrown in the biweekly rate because loans can be paid
off biweekly.)

3 Banks are required to reveal this so-called APR when they offer loans. The APR also takes into account

certain bank fees known as points. Unfortunately, not all fees are included in it, and the true costs are higher
if the loan is paid off early.
4 For daily compounding assume that the year has 365 days, not 365.25. Banks are quite careful about

these subtle differences. If you look at official tables of rates from precalculator days you will find that they
are off by small amounts because U.S. regulations permitted banks to pretend that a year has 360 days.
 1. DIFFERENTIATION

1J. Trigonometric functions

1J-1 Calculate the derivatives of the following functions

a) sin(5x2 ) b) sin2 (3x) c) ln(cos(2x))
sin x
d) ln(2 cos x) e) f) cos(x + y); y = f (x)
x
2
g) cos(x + y); y constant h) esin x i) ln(x2 sin x)
2x 2
√
j) e sin(10x) k) tan (3x) l) sec 1 − x2

m) The following three functions have the same derivative: cos(2x), cos2 x − sin2 x,
and 2 cos2 x. Verify this. Are the three functions equal? Explain.
"
n) sec(5x) tan(5x) o) sec2 (3x) − tan2 (3x) p) sin( x2 + 1)
" x
q) cos2 ( 1 − x2 ) r) tan2 ( )
x+1
cos x
1J-2 Calculate lim by relating it to a value of (cos x)! .
x→π/2 x − π/2

1J-3 a) Let a > 0 be a given constant. Find in terms of a the value of k > 0 for which
y = sin(kx) and y = cos(kx) both satisfy the equation

y !! + ay = 0.

Use this value of k in each of the following parts.

b) Show that y = c1 sin(kx) + c2 cos(kx) is also a solution to the equation in (a), for
any constants c1 and c2 .
c) Show that the function y = sin(kx + φ) (whose graph is a sine wave with phase
shift φ) also satisfies the equation in (a), for any constant φ.
d) Show that the function in (c) is already included among the functions of part (b),
by using the trigonometric addition formula for the sine function. In other words, given k
and φ, find values of c1 and c2 for which

sin(kx + φ) = c1 sin(kx) + c2 cos(kx)

√
1J-4 a) Show that a chord of the unit circle with angle θ has length 2 − 2 cos θ. Deduce
from the half-angle formula '
1 − cos θ
sin(θ/2) =
2
that the length of the chord is
2 sin(θ/2)

b) Calculate the perimeter of an equilateral n-gon with vertices at a distance 1 from

the center. Show that as n tends to infinity, the perimeter tends to 2π, the circumference
of the unit circle.
 2. Applications of Differentiation
2A. Approximation
√
2A-1 Find the linearization of a + bx at 0, by using [(2), Notes A], and also by using the
basic approximation formulas. (Here a and b are constants; assume a > 0. Do not confuse
this a with the one in (2), which has the value 0.)
1
2A-2 Repeat Exercise 2A-1 for the function , a "= 0 .
a + bx

(1 + x)3/2
2A-3 Find the linearization at 0 of by using the basic approximation formulas,
1 + 2x
and also by using [(2), Notes A].
gW
2A-4 Find the linear approximation for h ≈ 0 for w = , the weight of a body
(1 + h/R)2
at altitude h above the earth’s surface, where W is the surface weight and R is the radius
of the earth. (Do this without referring to the notes.)

2A-5 Making reasonable assumptions, if a person 5 feet tall weighs on the average 120 lbs.,
approximately how much does a person 5! 1!! tall weigh?

2A-6 Find a quadratic approximation to tan θ, for θ ≈ 0.

sec x
2A-7 Find a quadratic approximation to √ , for x ≈ 0.
1 − x2
2A-8 Find the quadratic approximation to 1/(1 − x), for x ≈ 1/2 . (Either use [(13), Notes
A] or put x = 21 + h and use the basic approximations.)

2A-9* Derive [(12), Notes A] algebraically as suggested in Notes A.

2A-10 Derive [(9), (10), (12), Notes A] by using formula [(13), Notes A].

2A-11 For an ideal gas at constant temperature, the variables p (pressure) and v (volume)
are related by the equation pv k = C, where k and C are constants. If the volume is changed
slightly from v0 to v0 + ∆v, what quadratic approximation expressing p in terms of ∆v
would you use? (Find the approximation valid for ∆v ≈ 0.)

2A-12 Give the indicated type of approximation at the point indicated. (This is to be
done after studying the log and exponential functions.)
ex ln(1 + x)
a) (quadratic, x ≈ 0) b) (linear, x ≈ 0)
1−x xex
2
c) e−x (quadratic, x ≈ 0) d) ln cos x (quadratic, x ≈ 0)
e) x ln x (quadratic, x ≈ 1) (Hint: put x = 1 + h.)

2A-13 Find the linear and quadratic approximation to the following functions
a) sin(2x), near 0 b) cos(2x), near 0 c) sec(x), near 0
x2
d) e , near 0 e) 1/(a + bx), near 0; assuming a "= 0;
f) 1/(a + bx), near 1; what do you have to assume about the constants a and b?
 E. 18.01 EXERCISES

2A-14* Suppose that a piece of bubble gum has volume 4 cubic centimeters.
a) Use a linear approximation to calculate the thickness of a bubble of inner radius 10
centimeters.
(Start with the relation between the volume V of a sphere and the radius r, and derive
the approximate relation between ∆V and ∆r.)
b) Find the exact answer.
c) To how many significant figures is the linear approximation accurate? In other
words, find the order of magnitude of the difference between the approximation and the
exact answer. (Be sure you use enough digits of π to reflect correctly this accuracy!)
d) Use a quadratic approximation to the exact formula for the thickness that you
found in part (b) to get an even more accurate estimate.
e) Why is the quadratic term comparable to the error in the accuracy of the linear
approximation?

2A-15 Find the linear and quadratic approximations to cos(3x) near x = 0, π/6, and π/3.
!
2A-16 a) Use the law of cosines to find the formula 2 − cos(2π/n) for the length of the
side of the equilateral n-gon inscribed in the unit circle.
b) Compute the perimeter and then compute the limit as n tends to infinity using the
quadratic approximation to cos θ near θ = 0. (Compare with 1J-4.)

2B. Curve Sketching

2B-1 Sketch the graphs of the following. Find the intervals on which it is increasing and
decreasing and decide how many solutions there are to y = 0. (Graphs need not reflect
inflection points, which are discussed in 9-2).
a) y = x3 − 3x + 1 b) y = x4 − 4x + 1
c) y ! (x) = 1/(1 + x2 ) and y(0) = 0. d) y = x2 /(x − 1)
√
e) y = x/(x + 4) f) y = x + 1/(x − 3)
2
g) y = 3x4 − 16x3 + 18x2 + 1 h) y = e−x
2
i) y ! = e−x and y(0) = 0.

2B-2 Find the inflection points of the graphs in problem 1.

2B-3 Find the conditions on a, b and c for which the cubic

y = x3 + ax2 + bx + c

has a local maximum and a local minimum. Use the following two methods:
a) Find the condition under which y ! has two distinct real roots. Which of these roots
is at the local maximum and which is at the local minimum? (Draw a picture.)
b) Find the condition under which y ! < 0 at the inflection point. Why does this
property imply that there is a local maximum and a local minimum?

2B-4 Suppose that f is a continuous function on 0 ≤ x ≤ 10. Sketch the graph from the
following description: f is zero at 4, 7, and 9. f ! (x) > 0 on 0 < x < 5 and 8 < x < 10
 2. APPLICATIONS OF DIFFERENTIATION

and f ! (x) < 0 on 5 < x < 8. With the given information, can you say anything for certain
about the maximum value, the minimum value of f ? Can you say anything about the place
where the maximum is attained or the place where the minimum is attained?

2B-5 a) Trace a copy of the graph of the function below and draw the graph of the derivative
directly underneath. Connect the inflection point to the corresponding point on the graph
of the derivative with a vertical dotted line.
b) Find a rational function with a graph resembling the one below.

-7 -3 -1 5

2B-6 a) Find a cubic polynomial with a local maximum at x = −1 and a local minimum
at x = 1.
b) Draw the graph of the cubic on −3 ≤ x ≤ 3.
c) Draw a differentiable function on −3 ≤ x ≤ 3 that has an absolute maximum at
x = −1 and an absolute minimum at x = 1.

2B-7 a) Prove that if f (x) is increasing and it has a derivative at a, then f ! (a) ≥ 0. (You
may use the fact that a positive function has a limit ≥ 0.)
b) If the conclusion of part (a) is changed to : f ! (a) > 0, the statement becomes false.
Indicate why the proof of part (a) fails to show that f ! (a) > 0, and give a counterexample
to the conclusion f ! (a) > 0 (i.e., an example for which it is false).
c) Prove that if f (x) has relative maximum at a and it has a derivative at a, then
f ! (a) = 0. (Consider the right-and left-hand limits of ∆y/∆x; apply the ideas of part (a).)

2C. Max-min problems

2C-1 Cut four identical squares out of the corners of a 12 by 12 inch piece of cardboard
and fold the sides so as to make a box without a top. Find the size of the corner square
that maximizes the volume of the box.

2C-2 You are asked to design a rectangular barnyard enclosing 20,000 square feet with
fencing on three sides and the wall of a long barn on the fourth. Find the shortest length
of fence needed.

2C-3 What is the largest value of the product xy, if x and y are related by x + 2y = a,
where a is a fixed positive constant?
 E. 18.01 EXERCISES

2C-4 The U. S. Postal Service accepts boxes whose length plus girth equals at most 108
inches. What are the dimensions of the box of largest volume that is accepted? What is its
volume (in cubic feet)?
(“Length” is the longest of the three dimensions and “girth” is the sum of the lengths of
the four sides of the face perpendicular to the length, that is, the “waist-line” measurement
of the box. Note that the best shape for this rectangular face perpendicular to the length
must be a square.)

2C-5 Find the proportions of the open (i.e., topless) cylindrical can having the largest
volume inside, among all those having a fixed surface area A. (Use as the variable the
radius r.)

2C-6 What are the largest and the smallest possible values taken on by the product of
three distinct numbers spaced so the central number x has distance 1 from each of the other
two, if x lies between −2 and 2 inclusive?

2C-7 Find the dimensions of the rectangle of largest area that can be inscribed in a
semicircle of radius a.

2C-8 Find the dimensions of the rectangle of largest area in a right triangle, if
a) the sides of the rectangle are parallel to the legs;
b) one side of the rectangle is parallel to the hypotenuse.
(Of the two placements, which gives the rectangle of larger area?)

2C-9 A light ray reflected in a mirror travels the shortest distance between its starting
point and its endpoint.
Suppose the ray starts at (0, 1), and ends at the point (a, b) inside the first quadrant,
and being reflected when it hits the x-axis (at the point (x, 0), say).
Show that the two line segments forming its path make equal angles with the x-axis,
i.e., “the angle of incidence equals the angle of reflection.” B
2C-10 A swimmer is on the beach at a point A. The closest point on the straight shoreline 100
to A is called P . There is a platform in the water at B,
and the nearest pont on the shoreline to B is called Q. P x "
Suppose that the distance from A to P is 100 meters, the ! a x Q
distance from B to Q is 100 meters and the distance from
100
P to Q is a meters. Finally suppose that the swimmer can
run at 5 meters per second on the beach and swim at 2
meters per second in the water. A

Show that the path the swimmer should take to get to the platform in the least time has
the property that the ratio of the sines of the angles the path makes with the shoreline is
the reciprocal of the ratio of the speeds in the two regions:
sin α 5
=
sin β 2
(In optics, this is known as Snell’s law describing the path taken by a light ray through
two successive media. Snell discovered experimentally that the above ratio of sines was a
constant, not depending on the starting point and endpoint of the path. This problem shows
that Snell’s law follows from a minimum principle: the light ray takes the path minimizing
its total travel time. The ratio of sines is constant since it depends only on the speeds of
light in the two media.)
 2. APPLICATIONS OF DIFFERENTIATION

2C-11 A beam with a rectangular cross-section is cut from a log with a circular cross-
section. The strength S of the beam is proportional to the horizontal dimension x of the
rectangle and to the cube of the vertical dimension y of the rectangle, S = cxy 3 . Find the
ratio y/x which gives the strongest beam.

2C-12 You are going to mount a light on the wall behind your desk. The light at S
illuminates a point P on the horizontal surface of the desk with an S
intensity inversely proportional to the square of the distance from
P to S and proportional to sine of the angle between the ray from
P to S and the horizontal surface. Fix a point P on the desk 1
foot from the wall. Find the height of S above the desk for which #
the intensity at P is largest. P 1

2C-13 a) An airline will fill 100 seats of its aircraft at a fare of $200. For every $5 increase
in the fare, the plane loses two passengers. For every decrease of $5, the company gains two
passengers. What price maximizes revenue?
b) A utility company has a small power plant that can produce x kilowatt hours of
electricity daily at a cost of 10 − x/105 cents each for 0 ≤ x ≤ 8 × 105 . Consumers will
use 105 (10 − p/2) kilowatt hours of electricity daily at a price of p cents per kilowatt hour.
What price should the utility charge to maximize its profit?

2C-14 Find the maximum value and the location of the maximum for the following func-
tions. (When a > 0, xa ln x → 0 as x → 0+ . This follows from 2G-8 or from L’Hospital’s
rule in Section 6A.)
a) x2 ln(1/x), for x > 0 b) −x ln(2x), for x > 0

2C-15 Find the minimum of (x + 1)e−x , for 0 ≤ x < ∞.

2D. More Max-min Problems

2D-1* Consider a supersonic airplane wing with a cross-section in the shape of a thin
diamond (rhombus) in which the half-angle of the opening is τ and the attack angle α.
(The attack angle is the angle that the long diagonal of the
rhombus makes with the horizontal direction of motion of
the plane. See the picture.) The ratio of the lift to the !=$
drag is given by the formula (from the Course 16 Unified
Engineering notes on aerodynamics): direction of plane
lift α
= 2
drag α + τ2
a) For a given fixed τ , find the best attack angle α, that is, the one the maximizes
the ratio of lift to drag.
b) Find the minimum (largest negative) ratio. (This attack angle could be used in
the design of a winged car attempting to break the sound barrier, to prevent it from flying.)

2D-2* Consider a paper cup in the shape of a cone obtained by rotating the line segment
y = ax, 0 ≤ x ≤ r, around the y axis. For which slope a ≥ 0 will the paper cup hold the
most water, assuming its surface area A is held fixed. √
(Use the formulas: volume V = 13 (area of the base)(height); A = πr2 1 + a2 .)
 E. 18.01 EXERCISES

2D-3 Coffee in a cup at a temperature y(t0 ) at time t0 in a room at temperature a cools

according to the formula (derived in 3F-4); assume a = 20◦ C and c = 1/10:

y(t) = (y(t0 ) − a)e−c(t−t0 ) + a, t ≥ t0

You are going to add milk so that the cup has 10% milk and 90% coffee. If the coffee has
9 1
temperature T1 and the milk T2 , the temperature of the mixture will be 10 T1 + 10 T2 .
◦
The coffee temperature is 100 C at time t = 0, and you will drink the mixture at time
t = 10. The milk is refrigerated at 5◦ C. What is the best moment to add the milk so that
the coffee will be hottest when you drink it?

2D-4* a) Show that the shortest collection of roads joining four towns at the four corners of
a square of side 1 is given by roads that meet at 120◦ angles. Use the variable x representing
the length of the shorter dashed leg of the right triangle indicated in the left-hand picture.
(The right-hand one tells you what value you should find for x.)
b) Solve the same problem, but for towns at the corners of a rectangle. Use a for its
vertical height and 1 for its horizontal length, and x as in part (a). Write down the formula
for the length of the roads as a function of a. Hint: The answer is that for minimum total
length, the roads should meet at 120◦ angles, if a lies in a certain interval of positive real
numbers; if a is outside this interval, the pattern formed by the roads is simpler.
1/2
x 1
2 /3
1-2x 60
o

2D-5* Find the triangle of smallest area in the half-plane to the right of the y-axis whose
three sides are respectively segments of the x-axis, the diagonal y = x, and a line through
(2, 1).

2D-6* Find the point of the ellipse

73x2 − 72xy + 52y 2 = 2500

closest to (0, 0).

Hint: Use implicit differentiation to find a quadratic equation in x and y. This is the
type of equation (known as a homogeneous equation) that you faced in Problem 2C-11.
The homogeneous form makes it possible to solve for y/x. You can then find (x, y) by
substituting in the original equation.

2D-7* Find two positive numbers whose product is 10 and whose sum is as large as possible.

2E. Related Rates

2E-1 A robot going 20 ft/sec passes under a street light that is 30 feet above the ground.
If the robot is 5 feet tall, how fast is the tip of its shadow moving two seconds after passing
under the street light? How fast is the length of the shadow increasing at that moment?
 2. APPLICATIONS OF DIFFERENTIATION

2E-2 A beacon light 4 miles offshore (measured perpendicularly from a straight shoreline)
is rotating at 3 revolutions per minute. How fast is the spot of light on the shoreline moving
when the beam makes an angle of 60◦ with the shoreline?

2E-3 Two boats are travelling at 30 miles/hr, the first going north and the second going
east. The second crosses the path of the first 10 minutes after the first one was there. At
what rate is their distance increasing when the second has gone 10 miles beyond the crossing
point?

2E-4 Sand is pouring on a conical pile at a rate of 12 m3 per minute, in such a way that
the diameter of the base of the pile is always 3/2 the height. Find the rate at which the
height is increasing when the pile is 2 m tall.

2E-5 A person walks away from a pulley pulling a rope slung over it. The rope is being
held at a height 10 feet below the pulley. Suppose that the weight at the opposite end of
the rope is rising at 4 feet per second. At what rate is the person walking when s/he is 20
feet from being directly under the pulley?

2E-6 An airplane passes directly over a boat at a height of 2 miles. The plane is going
north at 400 mph and does not change its altitude. The boat is going west at 50 mph. How
rapidly is their distance from each other increasing after one hour?

2E-7 A trough is filled with water at a rate of 1 cubic meter per second. The trough has
a trapezoidal cross section with the lower base of length half a meter and one meter sides
opening outwards at an angle of 45◦ from the base. The length of the trough is 4 meters.
What is the rate at which the water level h is rising when h is one half meter?

2E-8 One ship is sailing east at 60km an hour and another is sailing south at 50km an
hour. The slower ship crosses the path of the faster ship at noon when the faster ship was
there one hour earlier. Find the time at which the two ships were closest to each other.

2E-9 A girl slides down a slide in the shape of the parabola y = (x − 1)2 for 0 ≤ x ≤ 1.
Her vertical speed is dy/dt = −y(1 − y). Find her horizontal speed dx/dt when y = 1/2.

2E-10 Oil spreads on a frying pan so that its radius is proportional to t1/2 , where t
represents the time from the moment when the oil is poured. Find the rate of change dT /dt
of the thickness T of the oil.

2F. Locating zeros; Newton’s method

2F-1 a) Graph the function y = cos x − x. Show using y ! that there is exactly one root to
the equation cos x = x, and give upper and lower bounds on the root.
b) Use Newton’s method to find the root to 3 decimal places.
c) Another way to find the root of cos x = x is to use what is called the fixed point
method. Starting with the value z1 = 1, press the cosine key on your calculator until the
answer stabilizes, i.e., until zn+1 = cos zn . How many iterations do you need until the first
nine digits stabilize? Which method takes fewer steps?

2F-2 Graph the function

1
y = 2x − 4 + −∞<x<∞
(x − 1)2
 E. 18.01 EXERCISES

In particular, count how many times y vanishes. (In math jargon, “vanishes” means y = 0,
not y → ∞. Another name for the values of x at which y = 0 is “the zeros” of y.) Give
reasons, based on the sign of y ! , that the zeros you have found must be there and that
there cannot be any more. (For example, “There must be exactly one zero in the interval
(−∞, −1) because . . . .”)

2F-3 Same problem as 2 for y = x2 + x−1

2F-4 The equation x5 − x = x(x2 − 1)(x2 + 1) = 0 has three roots, x = 0, 1, −1. How many
roots does the equation x5 − x − 1/2 = 0 have?

2F-5 a) Find an initial value x1 for the zero of x − x3 = 0 for which Newton’s method gives
an undefined quantity for x2 .
b) Find an initial value x1 for the zero of x − x3 = 0 such that Newton’s method
bounces back and forth between two values forever. Hint: use symmetry.
c) Find the largest interval around each of the roots of x − x3 = 0 such that Newton’s
method converges to that root for every initial value x1 in the interval. Hint: Parts (a) and
(b) should help.

2F-6 a) Suppose that a company manufacturing cylindrical beakers (of uniform thickness
on the bottom and the sides) is willing to use up to 10 percent more glass than the minimum
required to hold a particular volume. What proportions are permitted? (You will need to
use Newton’s method.)
b) What is the connection with Problem 2C-5?

2F-7 Find the point of the curve y = cos x closest to the origin. (Minimize the square of
the distance to the origin; prove you’ve found all the critical points.)

2G. Mean-value Theorem

2G-1 For each of these functions, on the indicated interval, find explicitly the point c whose
existence is predicted by the Mean-value Theorem; if there is more than one such c, find all
of them. Use the form (1).
(a) x2 on [0, 1] (b) ln x on [1, 2] (c) x3 − x on [−2, 2]

2G-2 Using the form (2), show that √

(a) sin x < x, if x > 0 (b) 1 + x < 1 + x/2 if x > 0.

2G-3 The Mass Turnpike is 121 miles long. An SUV enters at the Boston end at noon and
and emerges at the west end at 1:50. Prove that at some moment during the trip it was
speeding (i.e., over the 65 mph limit).

2G-4 A polynomial p(x) of degree n has at most n distinct real roots, but it may have
fewer — for instance, x2 + 1 has no real roots at all. However, show that if p(x) does have
n distinct real roots, then p! (x) has n − 1 distinct real roots.

2G-5 a) Suppose f !! (x) exists on an interval I and f (x) has a zero at three distinct points
a < b < c on I. Show there is a point p on [a, c] where f !! (p) = 0.
b) Illustrate part (a) on the cubic f (x) = (x − a)(x − b)(x − c).
 2. APPLICATIONS OF DIFFERENTIATION

2G-6 Using the form (2) of the Mean-value Theorem, prove that on an interval [a, b],
a) f ! (x) > 0 ⇒ f (x) increasing; (b) f ! (x) = 0 ⇒ f (x) constant.

2G-7* In what follows, use the following consequence of the Mean-value Theorem:
if f (a) ≥ g(a) and f ! (x) > g ! (x) for all x > a, then f (x) > g(x) for all x > a.
a) Starting with ex > 0, show that for all x > 0, we have ex > 1.
b) Use the same method as in part (a) to show that ex > 1 + x, and then from this
deduce that ex > 1 + x + x2 /2.
Remark This process can be continued. In the limit it leads to the infinite series which
represents ex :
x2 x3
ex = 1 + x + + + ···
2! 3!
This infinite series can be used to define ex , and it is a good way to compute it to high
accuracy.
c) Show that for each number n > 0

nx1/n < ln x for x > 1

2G-8* An analogous principle to the one in 2G-7 can be used if x < a:

if f (a) ≥ g(a) and f ! (x) < g ! (x) for all x < a, then f (x) > g(x) for all x < a.
(Why did the inequalities get reversed when x < a? Draw a graph to see.)
Use this principle with f (x) = ln x and a = 1 to show that

ln x > −nx−1/n for 0 < x < 1, n > 0.

 Unit 3. Integration

3A. Differentials, indefinite integration

3A-1 Compute the differentials df (x) of the following functions.

√
a) d(x7 + sin 1) b) d x c) d(x10 − 8x + 6)
√ √
d) d(e3x sin x) e) Express dy in terms of x and dx if x + y = 1

3A-2 Compute the following indefinite integrals

√
! " #
√ 1
! !
4 2
a) (2x + 3x + x + 8)dx b) x+ √ dx c) 8 + 9xdx
x
x
! ! !
d) x3 (1 − 12x4 )1/8 dx e) √ dx f) e7x dx
8 − 2x2
! √x
e dx
! !
4 x5
g) 7x e dx h) √ dx i)
x 3x + 2
x+5 x x
! !
j) dx. k) dx. (Write = 1 + . . . .)
x x+5 x+5
ln x dx
! !
l) dx m)
x x ln x

3A-3 Compute the following indefinite integrals.

! ! !
a) sin(5x)dx b) sin(x) cos(x)dx c) cos2 x sin xdx
cos x
! ! !
d) dx e) sec2 (x/5)dx f) tan6 x sec2 xdx
sin3 x
!
g) sec9 x tan xdx

3B. Definite Integrals

3B-1 Evaluate
$4 6
$
a) n2 b) 2j
n=1 j=1
5 4
$ $ 1
c) (−1)j j 2 d)
j=1 n=1
n

3B-2 Find a Σ notation expression for

a) 3 − 5 + 7 − 9 + 11 − 13 b) 1 + 1/4 + 1/9 + · · · + 1/n2
c) sin x/n + sin(2x/n) + · · · + sin((n − 1)x/n) + sin x

3B-3 Write the upper, lower, left and right Riemann sums for the following integrals, using
4 equal subintervals:
! 1 ! 3 ! 2π
3 2
a) x dx b) x dx c) sin xdx
0 −1 0
 E. 18.01 EXERCISES

3B-4 Calculate the difference between the upper and lower Riemann sums for the following
integrals with n intervals
! b ! b
a) x2 dx b) x3 dx
0 0
Does the difference tend to zero as n tends to infinity?

3B-5 Evaluate the limit, by relating it to a Riemann sum.

sin(b/n) + sin(2b/n) + · · · + sin((n − 1)b/n) + sin(nb/n)

lim
n→∞ n

! 1
3B-6* Calculate ex dx by using upper Riemann sums.
0

Hints: The sum is a geometric progression. You will need the limit lim n(e1/n − 1). This
n→∞
can be evaluated putting h = 1/n and relating the limit to the derivative of ex at x = 0.

3B-7* Evaluate the limit

2b/n + 22b/n + · · · + 2(n−1)b/n + 2nb/n

lim
n→∞ n

(See 3B-6.)

3C. Fundamental theorem of calculus

√
3C-1 Find the area under the graph of y = 1/ x − 2 for 3 ≤ x ≤ 6

3C-2 Calculate
! 2 2 π
√ sin xdx
! !
a) 3x + 5dx b) (3x + 5)n dx c)
0 0 3π/4 cos3 x

3C-3 Calculate
! 2 2b
xdx xdx
!
a) 2+1
b)
1 x b x2 + b2
! b
3C-4 Calculate lim x−10 dx. What area does this integral describe?
b→∞ 1

3C-5 Find the area

a) under one arch of sin x. b) under one arch of sin ax for a positive constant a.

3C-6 Find the area between the x-axis and

a) the curve y = x2 − 4 b) the curve y = x2 − a for a > 0.
 3. INTEGRATION

3D. Second fundamental theorem

x
dt
! %
3D-1 a) Prove that √ = ln(x + x2 + a2 ) − ln a, a > 0, x > 0.
0 t2 + a 2
! x
dt %
b) For what c is √ = ln(x + x2 + a2 ) ?
c t2 + a 2
! x%
3D-2* Show that the function y = 1 − t2 dt satisfies the differential equation with
0
side conditions:
y % y %% = −x; y(0) = 0, y % (0) = 1.
! x
1 − t2
3D-3 Discuss the function F (x) = 2
dt, including a sketch; describe
0 1+t

a) domain
b) relative maxima and minima, where increasing or decreasing, points of inflection
c) behavior as x → ∞ (Hint: Evaluate the integrand as t → ±∞.)
d) symmetry about y-axis or origin

3D-4 Find a function whose derivative is sin(x3 ) and whose value at 0 is

a) 0 b) 2 c) find one whose value at 1 is −1

1+∆x
1 t
!
3D-5 Evaluate lim √ dt two ways:
∆x→0 ∆x 1 1 + t4
a) by interpreting the integral as the area under a curve
! x
% t
b) by relating the limit to F (1), where F (x) = √ dt
0 1 + t4
! x
3D-6 For different values of a, the functions F (x) = dt differ from each other by
a
constants. Show this two ways:

a) directly

b) using the corollary to the mean-value theorem quoted ((8), p.FT.5)

3D-7 Evaluate F % (x) if F (x) =

! x2 sin x
√ dt
!
a) u sin udu b) √
0 0 1 − t2
! x2
c) tan udu
x

3D-8 Let f (x) be continuous. Find f (π/2) if :

! x ! x/2
a) f (t)dt = 2x(sin x + 1) b) f (t)dt = 2x(sin x + 1)
0 0
 E. 18.01 EXERCISES

3E. Change of variables; Estimating integrals

x
dt 1
!
3E-1 Prove directly from the definition L(x) = that L( ) = −L(a), by making a
1 t a
change of variables in the definite integral.

! x
1 2
3E-2 The function defined by E(x) = √ e−u /2 du is used in probability and statistics
2π 0
and has the same importance as sine and cosine functions have to trigonometry.
! x
2
a) Express E(x) in terms of the function of example 5 of Notes FT, F (x) = e−t dt,
0
√
by making a change of variable. It is known that lim F (x) = π/2. What is lim E(x)?
x→∞ x→∞

N
1
!
2
b) Evaluate lim √ e−u /2 du and lim E(x).
N →∞ 2π −N x→−∞

! b
1 2
c) Express √ e−u /2 du in terms of the function E and the constants a and
2π a
b, where a < b.

3E-3 Evaluate by making a substitution and changing both the variable and the limits of
integration.
! e√ π
ln x sin x
!
a) dx (u = ln x) b) dx
1 x 0 (2 + cos x)3
! 1 19
dx dx
!
c) √ (x = sin u) d) (x = z + 17)
0 1 − x2 18 x2 − 34x + 289

! 1 % ! a %
3E-4 From the definite integral 1 − x2 dx = π/2 deduce the value of a2 − x2 dx
−1 −a
by making a suitable change of variable of the form x = ct (c constant).

! x
3E-5 Let F (x) = f (t)dt.
0

a) Prove that if f (t) is even, then F (x) is odd.

b) Prove that if f (t) is odd, then F (x) is even.

Hint: Make the change of variable u = −t in the definite integral. (Compare with 4-6 .)

3E-6 By comparing the given integral with an integral that is easier to evaluate. establish
each of the following estimations:
! 1 ! π ! 20 %
dx 2
a) 3
> 0.65 b) sin x dx < 2 c) x2 + 1 dx > 150
0 1+x 0 10
 3. INTEGRATION
&! &
& N sin x &
3E-7 Show & dx& < 1
& &
& 1 x2 &

3F. Differential equations: separation of variables

3F-1 Solve the following differential equations

a) dy/dx = (2x + 5)4 b) dy/dx = (y + 1)−1
√
c) dy/dx = 3/ y d) dy/dx = xy 2

3F-2 Solve each differential equation with the given initial condition, and evaluate the
solution at the given value of x:
a) dy/dx = 4xy, y(1) = 3. Find y(3).
√
b) dy/dx = y + 1, y(0) = 1. Find y(3).
c) dy/dx = x2 y −1 , y(0) = 10. Find y(5).
d) dy/dx = 3y + 2, y(0) = 0. Find y(8).
y
e) dy/dx = e , y(3) = 0. Find y(0). For which values of x is the solution y
defined?

3F-3 a) Solve dy/dx = y 2 with y = 1 at x = 0. Evaluate y at x = 1/2, at x = −1, and at

x = 1.
b) Graph the solution and use the graph to discuss the range of validity of the formula
for y. In particular, explain why the apparent value at x = 3/2 is suspect.

3F-4 Newton’s law of cooling says that the rate of change of temperature is proportional
to the temperature difference. In symbols, if a body is at a temperature T at time t and the
surrounding region is at a constant temperature Te (e for external), then the rate of change
of T is given by
dT /dt = k(Te − T ).
The constant k > 0 is a constant of proportionality that depends properties of the body like
specific heat and surface area.
a) Why is k > 0 the only physically realistic choice?
b) Find the formula for T if the initial temperature at time t = 0 is T0 .
c) Show that T → Te as t → ∞.
d) Suppose that an ingot leaves the forge at a temperature of 680◦ Celsius in a room
at 40 Celsius. It cools to 200◦ in eight hours. How many hours does it take to cool from
◦

680◦ to 50◦ ? (It is simplest to keep track of the temperature difference T − Te , rather than
T . The temperature difference undergoes exponential decay.)
e) Suppose that an ingot at 1000◦ cools to 800◦ in one hour and to 700◦ in two hours.
Find the temperature of the surrounding air.
f) Show that y(t) = T (t − t0 ) also satisfies Newton’s law of cooling for any constant
t0 . Write out the formula for T (t− t0 ) and show that it is the same as the formula in E10/17
for y(t) by identifying the constants k, Te and T0 with their corresponding values in the
displayed formula in E10/17.
 E. 18.01 EXERCISES

3F-5* Air pressure satisfies the differential equation dp/dh = −(.13)p, where h is the
altitude from sea level measured in kilometers.
a) At sea level the pressure is1 1 kg/cm2 . Solve the equation and find the pressure at
the top of Mt. Everest (10 km).
b) Find the difference in pressure between the top and bottom of the Green Building.
(Pretend it’s 100 meters tall starting at sea level.) Compute the numerical value using a
calculator. Then use instead the linear approximation to ex near x = 0 to estimate the
percentage drop in pressure from the bottom to the top of the Green Building.
c) Use the linear approximation ∆p ≈ p% (0)∆h and compute p% (0) directly from the
differential equation to find the drop in pressure from the bottom to top of the Green Build-
ing. Notice that this gives an answer without even knowing the solution to the differential
equation. Compare with the approximation in part (b). What does the linear approximation
p% (0)∆h give for the pressure at the top of Mt. Everest?
d) What is the differential equation for p if altitude is measured in meters instead of
kilometers?

3F-6 Let y = cos3 u − 3 cos u, x = sin4 u. Find dy, dx, and dy/dx. Simplify.

3F-7 Solve:
a) y % = −xy, y(0) = 1 b) cos x sin y dy = sin x dx, y(0) = 0.

3F-8 a) Find all plane curves such that the tangent line at P intersects the x-axis 1 unit
to the left of the projection of P on the the x-axis.
b) Find all plane curves in the first quadrant such that for every point P on the curve,
P bisects the part of the tangent line at P that lies in the first quadrant.

3G. Numerical Integration

3G-1 Find approximations to the following integrals using four intervals using Riemann
sums with left endpoints, using the trapezoidal rule, and using Simpson’s rule. Also give
numerical approximations to the exact values of the integrals given to see how good these
approximation methods are.
! 1 ! π
√
a) xdx (= 2/3.) b) sin xdx (= 2.)
! 01 !0 2
dx dx
c) 2
(= π/4; cf. unit 5) d) (= ln 2)
0 1 + x 1 x

3G-2 Show that the value given by Simpson’s rule for two intervals for the integral
! b
f (x)dx
0

gives the exact answer when f (x) = x3 . (Since a cubic polynomial is a sum of a quadratic
polynomial and a polynomial ax3 , and Simpson’s rule is exact for any quadratic polynomial,
the result of this exercise implies by linearity (cf. Notes PI) that Simpson’s rule will also be
exact for any cubic polynomial.)
1 using the correspondence between weight and mass on Earth of F = ma with a = 10m/sec2
 3. INTEGRATION
√ √ √ √
3G-3 Use the trapezoidal rule to estimate 1 + 2 + 3 + ... + 10, 000. Is your estimate
too high or too low?

3G-4 Use the trapezoidal rule to estimate the sum of the reciprocals of the first n integers.
Is your estimate too high or too low?
! b
3G-5 If the trapezoidal rule is used to estimate the value of f (x)dx under what hy-
a
potheses on f (x) will the estimate be too low? too high?
 Unit 4. Applications of integration
4A. Areas between curves.

4A-1 Find the area between the following curves

a) y = 2x2 and y = 3x − 1 b) y = x3 and y = ax; assume a > 0
c) y = x + 1/x and y = 5/2. d) x = y 2 − y and the y axis.

4A-2 Find the area under the curve y = 1 − x2 in two ways.

4A-3 Find the area between the curves y = 4 − x2 and y = 3x in two ways.

4A-4 Find the area between y = sin x and y = cos x from one crossing to the next.

4B. Volumes by slicing; volumes of revolution

4B-1 Find the volume of the solid of revolution generated by rotating the regions bounded
by the curves given around the x-axis.
a) y = 1 − x2 , y = 0 b) y = a2 − x2 , y = 0 c) y = x, y = 0, x = 1
d) y = x, y = 0, x = a e) y = 2x − x2 , y = 0 f) y = 2ax − x2 , y = 0
√
g) y = ax, y = 0, x = a h) x2 /a2 + y 2 /b2 = 1, x = 0

4B-2 Find the volume of the solid of revolution generated by rotating the regions in 4B-1
around the y-axis.

4B-3 Show that the volume of a pyramid with a rectangular base is bh/3, where b is the
area of the base and h is the height. (Show in the process that the proportions of the
rectangle do not matter.)

4B-4 Consider (x, y, z) such that x2 + y 2 < 1, x > 0 and 0 ≤ z ≤ 5. This describes one half
of cylinder (a split log). Chop out a wedge out of the log along z = 2x. Find the volume of
the wedge.

4B-5 Find the volume of the solid obtained by revolving an equilateral triangle of sidelength
a around one of its sides.

4B-6 The base of a solid is the disk x2 + y 2 ≤ a2 . Planes perpendicular to the xy-plane
and perpendicular to the x-axis slice the solid in isoceles right triangles. The hypotenuse
of these trianglesis the segment where the plane meets the disk. What is the volume of the
solid?

4B-7 A tower is constructed with a square base and square horizontal cross-sections.
Viewed from any direction perpendicular to a side, the
tower has base y = 0 and profile lines y = (x − 1)2 and
y = (x + 1)2 . (See shaded region in picture.) Find the
volume of the solid.
 E. 18.01 EXERCISES

4C. Volumes by shells

4C-1 Assume that 0 < a < b. Revolve the disk (x − b)2 + y 2 ≤ a2 around the y axis. This
doughnut shape is known as a torus.
a) Set up the integral for volume using integration dx
b) Set up the integral for volume using integration dy
c) Evaluate (b).
d) (optional) Show that the (a) and (b) are the same using the substitution z = x − b.

4C-2 Find the volume of the region 0 ≤ y ≤ x2 , x ≤ 1 revolved around the y-axis.
√
4C-3 Find the volume of the region x ≤ y ≤ 1, x ≥ 0 revolved around the y-axis by both
the method of shells and the method of disks and washers.

4C-4 Set up the integrals for the volumes of the regions in 4B-1 by the method of shells.
(Do not evaluate.)

4C-5 Set up the integrals for the volumes of the regions in 4B-2 by the method of shells.
(Do not evaluate.)

4C-6 Let 0 < a < b. Consider a ball of radius b and a cylinder of radius a whose axis passes
through the center of the ball. Find the volume of the ball with the cylinder removed.

4D. Average value

4D-1 What is the average cross-sectional area of the solid obtained by revolving the region
bounded by x = 2, the x-axis, and the curve y = x2 about the x-axis? (Cross-sections are
taken perpendicular to the x-axis.)

4D-2 Show that the average value of 1/x over the interval [a, 2a] is of the form C/a, where
C is a constant independent of a. (Assume a > 0.)

4D-3 A point is moving along the x-axis; the functions x = s(t) and v(t) give respectively
its position and velocity at time t. Show that over a time interval a ≤ t ≤ b, the average
value of the function v(t) equals the average velocity of the point over this interval, as the
uncalculused would calculate it.

4D-4 What is the average value of the square of the distance of a point P from a fixed
point Q on the unit circle, where P is chosen at random on the circle? (Use coordinates;
place Q on the x-axis.) Check your answer for reasonableness.

4D-5 If the average value of f (t) between 0 and x is given by the function g(x), express
f (x) in terms of g(x).

4D-6 An amount of money A compounded continuously at interest rate r increases accord-

ing to the law
A(t) = A0 ert (t = time in years)
a) What is the average amount of money in the bank over the course of T years?
b) Suppose r and T are small. Give an approximate answer to part (a) by using the
quadratic approximation to your exact answer; check it for reasonableness.
 4. APPLICATIONS OF INTEGRATION

4D-7 Find the average value of x2 in 0 ≤ x ≤ b.

4D-8 Find the average distance from a point on the perimeter of a square of sidelength a
to the center. Find the average of the square of the distance.

4D-9 Find the average value of sin ax in its first hump.

4D’. Work

4D’-1 An extremely stiff spring is 12 inches long, and a force of 2,000 pounds extends it
1/2 inch. How many foot-pounds of work would be done in stretching it to 18 inches?

4D’-2 A heavy metal 2 pound pail initially is filled with 10 pounds of paint. Immediately
after it is filled, it is pulled up at a steady rate to the top of a building 30 feet high. While
being pulled, the paint leaks out through a hole in the pail at a steady rate so that by the
time it reaches the top, 1/5 of the paint has leaked out. How many foot-pounds of work
were done pulling the pail to the top of the building?

4D’-3 A heavy-duty rubber firehose hanging over the side of a building is 50 feet long and
weighs 2 lb./foot. How much work is done winding it up on a windlass on the top of the
building?

4D’-4 Two point-particles having respective masses m1 and m2 are at d units distance.
How much work is required to move them n times as far apart (i.e., to distance nd)? What
is the work to move them infinitely far apart?

4E. Parametric equations

4E-1 Find the rectangular equation for x = t + t2 , y = t + 2t2 .

4E-2 Find the rectangular equation for x = t + 1/t and y = t − 1/t (compute x2 and y 2 ).

4E-3 Find the rectangular equation for x = 1 + sin t, y = 4 + cos t.

4E-4 Find the rectangular equation for x = tan t, y = sec t.

4E-5 Find the rectangular equation for x = sin 2t, y = cos t.

4E-6 Consider the parabola y = x2 . Find the parametrization using the slope of the curve
at a point (x, y) as the parameter.

4E-7 Find the parametrization of the circle x2 + y 2 = a2 using the slope as the parameter.
Which portion of the circle do you obtain in this way?

4E-8 At noon, a snail starts at the center of an open clock face. It creeps at a steady rate
along the hour hand, reaching the end of the hand at 1:00 PM. The hour hand is 1 meter
long. Write parametric equations for the position of the snail at time t, in some reasonable
xy-coordinate system.

4E-9* a) What part of a train is moving backwards when the train moves forwards?
b) A circular disc has inner radius a and outer radius b. Its inner circle rolls along
the positive x-axis without slipping . Find parametric equations for the motion of a point P
 E. 18.01 EXERCISES

on its outer edge, assuming P starts at (0, b). Use θ as parameter. (Your equations should
reduce to those of the cycloid when a = b. Do they?)
c) Sketch the curve that P traces out.
d) Show from the parametric equations you found that P is moving backwards when-
ever it lies below the x-axis.

4F. Arclength

4F-1 Find the arclength of the following curves

a) y = 5x + 2, 0 ≤ x ≤ 1. b) y = x3/2 , 0 ≤ x ≤ 1.
2/3 3/2
c) y = (1 − x ) , 0 ≤ x ≤ 1. d) y = (1/3)(2 + x2 )3/2 , 1 ≤ x ≤ 2.

4F-2 Find the length of the curve y = (ex + e−x )/2 for 0 ≤ x ≤ b. Hint:
! x "2 ! x "2
e − e−x e + e−x
+1=
2 2

4F-3 Express the length of the parabola y = x2 for 0 ≤ x ≤ b as an integral. (Do not
evaluate.)

4F-4 Find the length of the curve x = t2 , y = t3 for 0 ≤ t ≤ 2.

4F-5 Find an integral for the length of the curve given parametrically in Exercise 4E-2 for
1 ≤ t ≤ 2. Simplify the integrand as much as possible but do not evaluate.

4F-6 a) The cycloid given parametrically by x = t − sin t, y = 1 − cos t describes the path
of a point on a rolling wheel. If t represents time, then the wheel is rotating at a constant
speed. How fast is the point moving at each time t? When is the forward motion (dx/dt)
largest and when is it smallest?
b) Find the length of the cycloid for one turn of the wheel. (Use a half angle formula.)

4F-7 Express the length of the ellipse x2 /a2 + y 2 /b2 = 1 using the parametrization x =
a cos t and y = b sin t. (Do not evaluate.)

4F-8 Find the length of the curve x = et cos t, y = et sin t for 0 ≤ t ≤ 10.

4G. Surface Area R

4G-1 Consider the sphere of radius R formed by revolving the circle

a b
x2 + y 2 = R2 around the x-axis. Show that for −R ≤ a < b ≤ R,
the portion of the sphere a ≤ x ≤ b has surface area 2πR(b − a). For
example, the hemisphere, a = 0, b = R has area 2πR2 .

4G-2 Find the area of the segment of y = 1 − 2x in the first quadrant revolved around the
x-axis.

4G-3 Find the area of the segment of y = 1 − 2x in the first quadrant revolved around the
y-axis.

4G-4 Find an integral formula for the area of y = x2 , 0 ≤ x ≤ 4 revolved around the
x-axis. (Do not evaluate.)
 4. APPLICATIONS OF INTEGRATION

4G-5 Find the area of y = x2 , 0 ≤ x ≤ 4 revolved around the y-axis.

4G-6 Find the area of the astroid x2/3 + y 2/3 = a2/3 revolved around the x-axis.

4G-7 Conside the torus of Problem 4C-1.

a) Set up the integral for surface area using integration dx
b) Set up the integral for surface area using integration dy
c) Evaluate (b) using the substitution y = a sin θ.

4H. Polar coordinate graphs

4H-1 For each of the following points given in rectangular coordinates, give its polar
coordinates. (For points below the x-axis, give two expressions for its polar coordinates,
using respectively positive and negative values for θ.)
√
a) (0, 3) b) (−2, 0) c) (1, 3) d) (−2, 2)
√
e) (1, −1) f) (0, −2) g) ( 3, −1) h) (−2, −2)

4H-2
a) Find using two different methods the equation in polar coordinates for the circle of
radius a with center at (a, 0) on the x-axis, as follows:
(i) write its equation in rectangular coordinates, and then change it to polar coordi-
nates (substitute x = r cos θ and y = r sin θ, and then simplify).
(ii) treat it as a locus problem: let OQ be the diameter lying along the x-axis, and
P : (r, θ) a point on the circle; use %OP Q and trigonometry to find the relation connecting
r and θ.
b) Carry out the analogue of 4H-2a for the circle of radius a with center at (0, a) on the
y-axis; OQ is now the diameter lying along the y-axis.
c) (i) Find the polar equation for the line intersecting the positive x- and
y-axes respectively at A and B, and having perpendicular distance a from
the origin.
(Let α = ∠DOA; use the right triangle DOP to get the equation B D
connecting r, θ, α and a. a P
(ii) Convert your polar equation to the usual rectangular equation r
involving A and B, by using trigonometry. O A

d) In the accompanying figure, the point Q moves around the circle of

radius a centered at the origin; QR is a perpendicular to the x-axis. P is Q
a point on ray OQ such that |QP | = |QR|: P is the point inside the circle P
a
in the first two quadrants, but outside the circle in the last two quadrants. O R
(i) Sketch the locus of P ; the locus is called a cardioid (cf. 4H-3c).
(ii) find the polar equation of this locus.
e) The point P moves in a locus so that the product of its distances from the two points
Q : (−a, 0) and R : (a, 0) is constant. Assuming the locus of P goes through the origin,
determine the value of the constant, and derive the polar equation of the locus of P .
(Work with the squares of the distances, rather than the distances themselves, and use
the law of cosines; the identities (A + B)(A − B) = A2 − B 2 and cos 2θ = 2 cos2 θ − 1 simplify
 E. 18.01 EXERCISES

the algebra and produce a simple answer at the end. The resulting curve is a lemniscate,
cf. 4H-3g.)

4H-3 For each of the following,

(i) give the corresponding equation in rectangular coordinates;
(ii) draw the graph; indicate the direction of increasing θ.

a) r = sec θ b) r = 2a cos θ

c) r = (a + b cos θ) (This figure is a cardioid for a = b, a limaçon with a loop for

0 < a < b, and a limaçon without a loop for a > b > 0.)
d) r = a/(b + c cos θ) (Assume the constants a and b are positive. This figure is an
ellipse for b > |c| > 0, a circle for c = 0, a parabola for b = |c|, and a hyperbola for b < |c|.)
e) r = a sin(2θ) (4-leaf rose) f) r = a cos(2θ) (4-leaf rose)
2 2
g) r = a sin(2θ) (lemniscate) h) r2 = a2 cos(2θ) (lemniscate)
i) r = eaθ (logarithmic spiral)
 4. APPLICATIONS OF INTEGRATION

4I. Area and arclength in polar coordinates

4I-1 Find the arclength element ds = w(θ)dθ for the curves of 4H-3.

4I-2 Find the area of one leaf of a three-leaf rose r = a cos(3θ).

4I-3 Find the area of the region 0 ≤ r ≤ e3θ for 0 ≤ θ ≤ π

4I-4 Find the area of one loop of the lemniscate r2 = a2 sin(2θ)

4I-5 What is the average distance of a point on a circle of radius a from a fixed point Q
on the circle? (Place the circle so Q is at the origin and use polar coordinates.)

4I-6 What is the average distance from the x-axis of a point chosen at random on the
cardioid r = a(1 − cos θ) , if the point is chosen
a) by letting a ray θ = c sweep around at uniform velocity, stopping at random and
taking the point where it intersects the cardioid;
b) by letting a point P travel around the cardioid at uniform velocity, stopping at
random; (the answers to (a) and (b) are different...)

4I-7 Calculate the area and arclength of a circle, parameterized by x = a cos θ, y = a sin θ.

4J. Other Applications

4J-1 Suppose it takes k units of energy to lift a cubic meter of water one meter. About
how much energy E will it take to pump dry a circular hole one meter in diameter and 100
meters deep that is filled with water? (Give reasoning.)

4J-2 The amount x (in grams) of a radioactive material declines exponentially over time
(in minutes), according to the law x = x0 e−kt , where x0 is the amount initially present
at time t = 0. If one gram of the material produces r units of radiation/minute, about how
much radiation R is produced over one hour by x0 grams of the material? (Give reasoning.)

4J-3 A very shallow circular reflecting pool has uniform depth D, and radius R (meters). A
disinfecting chemical is released at its center, and after a few hours of symmetrical diffusion
k
outwards, the concentration of chemical at a point r meters from the center is g/m3 .
1 + r2
What amount A of the chemical was released into the pool? (Give reasoning.)

4J-4 Assume a heated outdoor pool requires k units of heat/hour for each degree F it is
maintained above the external air temperature.
o o
If the external temperature T #varies between 50
$ and 70 over a 24 hour period starting
at midnight, according to T = 10 6 − cos(πt/12) , how many heat units will be required to
maintain the pool at a steady 75o temperature? (Give reasoning.)

4J-5 A manufacturers cost for storing one unit of inventory is c dollars/day for space and
insurance. Over the course of 30 days, production P rises from 10 to 40 units/day according
to P = 10 + t. Assuming no units are sold, what is the inventory cost for this period? (Give
reasoning.)

4J-6 A water tank for a town has the shape of a sphere of radius r feet, and its center is
at a height h above the ground. If the weight of a cubic foot of water is w lbs., how much
 E. 18.01 EXERCISES

work is required to fill the tank when empty by pumping water from the ground? (Give
reasoning using infinitesimals.)

4J-6 Divide the water in the tank into thin horizontal slices of width% dy.
If the slice is at height y above the center of the tank, its radius is r2 − y 2 .
volume of water in the slice = π(r2 − y 2 ) dy
weight of water in the slice = πw(r2 − y 2 ) dy
work to lift this slice from the ground = πw(r2 − y 2 ) dy (h + y).
r r (r
r2 y 2 y3 y4
& & '
Total work = πw(r2 −y 2 )(h+y) dy = πw (r2 h+r2 y−hy 2 −y 3 ) = πw r2 hy+ −h − .
−r −r 2 3 4 −r

The even powers of y have the same value at −r and r, so contribute 0 to the value; we get
(r
y3 r3
' ! "
4
= πwh r2 y − = 2πwh r3 − = πwhr3 .
3 −r 3 3
 Unit 5. Integration techniques
5A. Inverse trigonometric functions; Hyperbolic functions

5A-1 Evaluate
√ √
a) tan−1 3 b) sin−1 ( 3/2)
c) If θ = tan−1 5, then evaluate sin θ, cos θ, cot θ, csc θ, and sec θ.
d) sin−1 cos(π/6) e) tan−1 tan(π/3)
f) tan−1 tan(2π/3) g) lim tan−1 x.
x→−∞

5A-2 Calculate
! 2 2b 1
dx dx dx
! !
a) 2+1
b) c) √ .
1 x b x2 + b2 −1 1 − x2

5A-3 Calculate the derivative with respect to x of the following

" #
x−1
a) sin−1
b) tanh x
x+1
√
c) ln(x + x2 + 1) d) y such that cos y = x, 0 ≤ x ≤ 1 and 0 ≤ y ≤ π/2.
e) sin−1 (x/a) f) sin−1 (a/x)
√ √
g) tan−1 (x/ 1 − x2 ) h) sin−1 1 − x

5A-4 a) If the tangent line to y = cosh x at x = a goes through the origin, what equation
must a satisfy?
b) Solve for a using Newton’s method.

5A-5 a) Sketch the graph of y = sinh x, by finding its critical points, points of inflection,
symmetries, and limits as x → ∞ and −∞.
b) Give a suitable definition for sinh−1 x, and sketch its graph, indicating the domain
of definition. (The inverse hyperbolic sine.)
d
c) Find sinh−1 x.
dx
dx
!
d) Use your work to evaluate √
a + x2
2

5A-6 a) Find the average value of y with respect to arclength on the semicircle x2 + y 2 = 1,
y > 0, using polar coordinates.
b) A weighted average of a function is
$!
! b b
f (x)w(x)dx w(x)dx
a a

Do part (a) over again expressing arclength as ds = w(x)dx. The change of variables needed
to evaluate the numerator and denominator will bring back part (a).
√
c) Find the average height of 1 − x2 on −1 < x < 1 with respect to dx. Notice that
this differs from part (b) in both numerator and denominator.
 E. 18.01 EXERCISES

5B. Integration by direct substitution

Evaluate the following integrals

ln xdx
! % ! !
5B-1. x x2 − 1dx 5B-2. e8x dx 5B-3.
x
cos xdx
! ! !
2
5B-4. 5B-5. sin x cos xdx 5B-6. sin 7xdx
2 + 3 sin x
6xdx
! ! !
5B-7. √ 5B-8. tan 4xdx 5B-9. ex (1 + ex )−1/3 dx
x2 + 4
! ! !
2
2
5B-10. sec 9xdx 5B-11. sec 9xdx 5B-12. xe−x dx

x2 dx
!
5B-13. . Hint: Try u = x3 .
1 + x6
Evaluate the following integrals by substitution and changing the limits of integration.
! π/3 ! e ! 1
3 (ln x)3/2 dx tan−1 xdx
5B-14. sin x cos xdx 5B-15. 5B-16.
0 1 x −1 1 + x2

5C. Trigonometric integrals

Evaluate the following

! ! !
2 3
5C-1. sin xdx 5C-2. sin (x/2)dx 5C-3. sin4 xdx
! ! !
5C-4. cos3 (3x)dx 5C-5. sin3 x cos2 xdx 5C-6. sec4 xdx
! ! !
2
5C-7. 2
sin (4x) cos (4x)dx 5C-8. 2
tan (ax) cos(ax)dx 5C-9. sin3 x sec2 xdx
! !
5C-10. (tan x + cot x)2 dx 5C-11. sin x cos(2x)dx (Use double angle formula.)
! π
5C-12. sin x cos(2x)dx (See 27.)
0
5C-13. Find the length of the curve y = ln sin x for π/4 ≤ x ≤ π/2.

5C-14. Find the volume of one hump of y = sin ax revolved around the x-axis.

5D. Integration by inverse substitution

Evaluate the following integrals

dx x3 dx (x + 1)dx
! ! !
5D-1. 5D-2. √ 5D-3.
(a2 − x2 )3/2 2
a −x 2 4 + x2
! √ 2
a − x2 dx
! % ! %
5D-4. a2 + x2 dx 5D-5. 2
5D-6. x2 a2 + x2 dx
x
(For 5D-4,6 use x = a sinh y, and cosh2 y = (cosh(2y) + 1)/2, sinh 2y = 2 sinh y cosh y.)
! √ 2
x − a2 dx
! %
5D-7. 5D-8. x x2 − 9dx
x2
 5. INTEGRATION TECHNIQUES

5D-9. Find the arclength of y = ln x for 1 ≤ x ≤ b.

Completing the square

Calculate the following integrals

dx
! ! % ! %
5D-10. 5D-11. x −8 + 6x − x2 dx 5D-12. −8 + 6x − x2 dx
(x2 + 4x + 13)3/2
! √ 2
dx xdx 4x − 4x + 17dx
! !
5D-13. √ 5D-14. √ 5D-15.
2x − x 2 2
x + 4x + 13 2x − 1

5E. Integration by partial fractions

dx xdx xdx
! ! !
5E-1. dx 5E-2. dx 5E-3. dx
(x − 2)(x + 3) (x − 2)(x + 3) (x2 − 4)(x + 3)
3x2 + 4x − 11 3x + 2 2x − 9
! ! !
5E-4. dx 5E-5. dx 5E-6. dx
(x2 − 1)(x − 2) x(x + 1)2 2
(x + 9)(x + 2)
5E-7 The equality (1) of Notes F is valid for x &= 1, −2. Therefore, the equality (4) is
also valid only when x &= 1, −2, since it arises from (1) by multiplication. Why then is it
legitimate to substitute x = 1 into (4)?

5E-8 Express the following as a sum of a polynomial and a proper rational function
x2 x3 x2
a) 2 b) 2 c)
x −1 x −1 3x − 1
x+2 x8
d) e) (just give the form of the solution)
3x − 1 (x + 2)2 (x − 2)2

5E-9 Integrate the functions in Problem 5E-8.

5E-10 Evaluate the following integrals

dx (x + 1)dx (x2 + x + 1)dx
! ! !
a) b) c)
x3 − x (x − 2)(x − 3) x2 + 8x
(x2 + x + 1)dx dx 2
(x + 1)dx
! ! !
d) e) f)
x2 + 8x x3 + x2 x3 + 2x2 + x
x3 dx (x2 + 1)dx
! !
g) 2
h)
(x + 1) (x − 1) x2 + 2x + 2

5E-11 Solve the differential equation dy/dx = y(1 − y).

5E-12 This problem shows how to integrate any rational function of sin θ and cos θ using
the substitution z = tan(θ/2). The integrand is transformed into a rational function of z,
which can be integrated using the method of partial fractions.
a) Show that

1 − z2 2z 2dz
cos θ = , sin θ = , dθ = .
1 + z2 1 + z2 1 + z2
 E. 18.01 EXERCISES

Calculate the following integrals using the substitution z = tan(θ/2) of part (a).
! π ! π ! π
dθ dθ
b) c) 2
d) sin θdθ (Not the easiest way!)
0 1 + sin θ 0 (1 + sin θ) 0

5E-13 a) Use the polar coordinate formula for area to compute the area of the region
0 < r < 1/(1 + cos θ), 0 ≤ θ ≤ π/2. Hint: Problem 12 shows how the substitution
z = tan(θ/2) allows you to integrate any rational function of a trigonometric function.
b) Compute this same area using rectangular coordinates and compare your answers.

5F. Integration by parts. Reduction formulas

Evaluate the following integrals

!
5F-1 a) xa ln xdx (a &= −1) b) Evaluate the case a = −1 by substitution.
! ! !
x 2 x
5F-2 a) xe dx b) x e dx x3 ex dx
c)
! !
d) Derive the reduction formula expressing xn eax dx in terms of xn−1 eax dx.
!
5F-3 Evaluate sin−1 (4x)dx
!
5F-4 Evaluate ex cos xdx. (Integrate by parts twice.)
!
5F-5 Evaluate cos(ln x)dx. (Integrate by parts twice.)
! !
n x
x
5F-6 Show the substitution t = e transforms the integral x e dx, into (ln t)n dt. Use
a reduction procedure to evaluate this integral.
 Unit 6. Additional Topics

6A. Indeterminate forms; L’Hospital’s rule

6A-1 Find the following limits

sin 3x cos(x/2) − 1 ln x
a) lim b) lim c) lim
x→0 x x→0 x2 x→∞ x
x2 − 3x − 4 tan x
−1
x − sin x
d) lim e) lim f) lim
x→0 x+1 x→0 5x x→0 x3
xa − 1 tan(x) ln sin(x/2)
g) lim b h) lim i) lim
x→1 x − 1 x→1 sin(3x) x→π x−π
ln sin(x/2)
j) lim
x→π (x − π)2

6A-2 Evaluate the following limits.

a) lim+ xx b) lim+ x1/x c) lim+ (1/x)ln x
x→0 x→0 x→0

d) lim (cos x)1/x e) lim x1/x f) lim (1 + x2 )1/x

x→0+ x→∞ x→0+
x + cos x 1
g) lim+ (1 + 3x)10/x h) lim i) lim x sin
x→0 x→∞ x x→∞ x
! x "1/x2
j) lim+ k) lim xa (ln x)b . Consider all values of a and b.
x→0 sin x x→∞

6A-3 The power x−1 is the exceptional case among the integrals of the powers of x. It
would be nice if # #
lim xa dx = x−1 dx
a→−1

It seems hopeless for this to be true1 since

xa+1
#
xa dx = + c for a "= −1
a+1

involves only powers, yet the integral of x−1 is a logarithm. But it can be rescued using the
definite integral. Show using L’Hospital’s rule that
# x # x
lim ta dt = t−1 dt (= ln x)
a→−1 1 1

6A-4 Show that as a tends to −1 of a well-chosen solution to E30/1(a) tends to the answer
in part (b). Hint: Follow the method of the preceding problem.

6A-5 By repeated use of L’Hospital’s rule,

3x2 − 4x 6x − 4 6
lim 2
= lim = lim = −3,
x→0 2x − x x→0 2 − 2x x→0 −2

1 It seems hopeless because for almost all choices of c the indefinite integral has an infinite limit as

a → −1. The definite integral leads to the correct choice of c, namely, c = −1/(a + 1). The constant c is a
constant with respect to x, but there is no reason why it can’t vary with a. And the right choice of c makes
the limit as a → −1 finite.
 E. 19.01 EXERCISES

3x2 − 4x −4x
yet when x # 0 , # = −2. Resolve the contradiction.
2x − x2 2x
6A-6 Graph the following functions. (L’Hospital’s rule will help with some of the limiting
values at the ends.)
a) y = xe−x b) y = x ln x c) y = x/ ln x

6B. Improper integrals

Test the following improper integrals for convergence by using comparison with a simpler
integral.
1
∞
dx ∞
x2 dx dx
# # #
6B-1. √ 6B-2. 6B-3.
1 x3 + 5 0 x3 + 2 0 x3 + x2
1
dx ∞
e−x dx ∞
ln xdx
# # #
6B-4. √ 6B-5. 6B-6.
0 1 − x3 0 x 1 x2

6B-7 Decide whether the following integrals are convergent or divergent and evaluate if
convergent.
# ∞ # ∞ # ∞
a) e−8x dx b) x −n
dx, n > 1 c) x−n dx, 0 < n ≤ 1
0 1 1
2 # 2
xdx dx dx
# # ∞
d) √ e) √ f)
0 4 − x2 0 2−x e x(ln x)2
1 # 1 1
dx dx dx
# #
g) h) 3
i)
0 x1/3 0 x −1 x
1 # ∞ ∞
dx
# #
j) ln xdx k) e−2x cos xdx l) . (Use (f).)
0 0 e x(ln x)
10
∞
dx ∞
dx (ln x)2
# # #
m) n) o) dx
(x + 2)3 (x − 2)3 x
# 0π 0 0

p) sec xdx
0

6B-8 Find the following limits. (Use the fundamental theorem of calculus.)
# x # x # x
2 2 2 2 2 2
a) lim e−x et dt b) lim xe−x et dt c) lim ex e−t dt
x→∞ 0 x→∞ 0 x→∞ 0
√ 1 √ 1 b
dx dx dx
# # #
d) lim a √ e) lim a f) lim (b − π/2)
a→0+ a x a→0+ a x3/2 b→(π/2)+ 0 1 − sin x
 6. ADDITIONAL TOPICS

6C. Infinite Series

6C-1 Find the sum of the following geometric series:

a) 1 + 1/5 + 1/25 + · · · b) 8 + 2 + 1/2 + · · · c) 1/4 + 1/5 + · · ·

Write the two following infinite decimals as the quotient of two integers:
d) 0.4444 . . . e) 0.0602602602602 . . .

6C-2 Decide whether the following series are convergent or divergent; indicate reasoning.
(Do not evaluate the sum.)

a) 1 + 1/2 + 1/3 + 1/4 + 1/5 + · · · ; use comparison with an integral.

∞
$ 1
b) ; consider the cases p > 1 and p ≤ 1.
n=1
np

c) 1/2 + 1/4 + 1/6 + 1/8 + · · ·

d) 1 + 1/3 + 1/5 + 1/7 + · · ·

e) 1 − 1/2 + 1/3 − 1/4 + 1/5 − · · · Hint: Combine pairs of consecutive terms to take
advantage of the cancellation. Then use comparison.
∞ ∞
%√ &n ∞
%√ &n
$ n $ 5−1 $ 5+1
f) . g) . h) 5−n/2 .
n=1
n! n=1
2 n=1
2
∞ ∞ ∞
$ ln n $ ln n $ n+2
i) . j) . k) .
n=1
n n=1
n2 n=1
n 4−5

(n + 2)1/3
∞ ∞ ∞
$ $ 1 $
l) . m) ln(cos ) n) n2 e−n
n=1
(n4 + 5)1/3 n=1
n n=1
∞
$ √
2 − n
o) n e
n=1

6C-3 a) Use the upper and lower Riemann sums of

n
dx
#
ln n =
1 x

to show that
1 1 1
ln n < 1 + + + · · · + < 1 + ln n
2 3 n

' b) Suppose that it takes 10−10 seconds for a computer to add one term in the series
1/n. About how long would it take for the partial sum to reach 1000?
 7. Infinite Series
7A. Basic Definitions

7A-1 Do the following series converge or diverge? Give reason. If the series converges, find
its sum.
1 1 1 1
a) 1 + + + + ... + n + ... b) 1 − 1 + 1 − 1 + . . . + (−1)n + . . .
4 16 64 4
1 2 n √ √ √
c) 1 + + + . . . + + ... d) ln 2 + ln 2 + ln 3 2 + ln 4 2 + . . .
2 3 n+1
∞
2n−1
∞
! ! 1
e) f) (−1)n
1
3n 0
3n
7A-2 Find the rational number represented by the infinite decimal .21111 . . . .
∞ " #n
! x
7A-3 For which x does the series converge? For these values, find its sum f (x).
0
2
7A-4 Find the sum of these series by first finding the partial sum Sn .
∞ " #
! 1 1
a) √ −√
1
n n+1
∞
! 1 1 a b
b) . (Hint: = + for suitable a, b).
1
n(n + 2) n(n + 2) n n+2

7A-5 A ball is dropped from height h; each time it lands, it bounces back 2/3 of the height
from which it previously fell. What is the total distance (up and down) the ball travels?

7B: Convergence Tests

7B-1 Using the integral test, tell whether the following series converge or diverge; show
work or reasoning.
∞ ∞ ∞
! n ! 1 ! 1
a) 2
b) 2
c) √
0
n +4 0
n +1 0
n+1
∞ ∞ ∞
! ln n ! 1 ! 1
d) e) f)
1
n 2
(ln n)p · n 1
n p

(In the last two, the answer depends on the value of the parameter p.)
7B-2 Using the limit comparison test, tell whether each series converges or diverges; show
work or reasoning. (For some of them, simple comparison works.)
∞ ∞ ∞
! 1 ! 1 ! 1
a) 2 + 3n
b) √ c) √
1
n 1
n + n 1
n 2+n

∞ " # ∞ √ ∞
! 1 ! n ! ln n
d) sin 2
e) 2+1
f)
1
n 1
n 1
n
∞
n2
∞
! ! n3
g) h)
2
n4 − 1 1
4n4 + n2
 E. 1801 EXERCISES

$∞ $∞
7B-3 Prove that if an > 0 and 0 an converges, then 0 sin an also converges.
7B-4 Using the ratio test, or otherwise, determine whether or not each of these series is
absolutely convergent. (Note that 0! = 1.)
∞ ∞
2n
∞
! n ! ! 2n
a) n
b) c)
0
2 0
n! 1
1 · 3 · 5 · · · (2n − 1)
#n
(n!)2 (−1)n
∞ ∞ ∞ "
! ! ! n! 1
d) e) √ f) ; use lim 1 + =e
0
(2n)! 1
n 1
nn n→∞ n
∞
(−1)n
∞
(−1)n
∞
! ! ! n
g) h) √ i)
1
n2 0
n2 + 1 0
n+1
7B-5 For those series in 7B-4 which are not absolutely convergent, tell whether they are
conditionally convergent or divergent.
7B-6 By using the ratio test, determine the radius of convergence of each of the following
power series.
∞
xn
∞
2n xn
! ! !∞
a) b) 2
c) n!xn
1
n 1
n 0

(−1)n x2n (−1)n x2n+1 (2n)!x2n

∞
! ∞
! ∞
!
d) e) √ f)
0
3n 0
2n n 0
(n!)2
xn 22n xn
∞
! ∞
!
g) h)
2
ln n 0
n!

7C: Taylor Approximations and Power Series

7C-1 Using the general formula for the coefficients an , find the Taylor series at 0 for the
following functions; do the work systematically, calculating in order the f (n) , f (n) (0), and
then the an . √
a) cos x b) ln(1 + x) c) 1+x
7C-2 Calculate sin 1 using the Taylor series up to the term in x3 . Estimate the accuracy
using the remainder term. (The calculator value is .84147.) Use the remainder term R6 (x),
not R5 (x); why?
7C-3 Using the remainder term, tell for what value of n in the approximation
x2 xn
ex ≈ 1 + x + + ...+
2! n!
the resulting calculation will give e to 3 decimal places (by convention, this means: within
.0005).
x2
7C-4 By using the remainder term, tell whether cos x ≈ 1 − will be valid to within
2!
.001 over the interval |x| < .5 .
% .5
2 2
7C-5 Calculate e−x dx, using the approximation for e−x up to the term in x4 . Esti-
0
mate the error, using the correct remainder term (cf. 7B-3), and tell whether the answer
will be good to 3 decimal places.
 7. INFINITE SERIES

7D: General Power Series

7D-1 Find the power series around x = 0 for each of the following functions by using known
Taylor series: use substitution, addition, differentiation, integration, or anything else you
can think of:
√
a) e−2x b) cos x, x ≥ 0 c) sin2 x (use an identity)
1
d) e) tan−1 x (differentiate) f) ln(1 + x)
(1 + x)2
ex + e−x
g) cosh x =
2
7D-2 By using operations on power series (substitution, addition, integration, differenti-
ation, multiplication), find the power series for the following functions, and determine the
radius of convergence. (Where indicated, give just the first 2 or 3 non-zero terms.)
1 2
a) b) e−x c) ex cos x (3 terms)
x+9
% x % x
sin t 2 1
d) dt e) erf x = e−t /2 dt f) 3
0 t 0 x −1
g) cos2 x (differentiate; then use a trigonometric identity)
sin x
h) (3 terms); do it two ways: multiplication, and dividing sinx series by 1 − x
1−x
i) tan x (2 terms); do it two ways: Taylor series, and division of power series

7D-3 Find the following limits by using linear, quadratic, or cubic approximation (i.e. by
using the first few terms of the Taylor series), not by using L’Hospital’s rule.
1 − cos x x − sin x
a) lim 2
b) lim
x→0 x x→0 x3
√
1 + x − 1 − x/2 cos u − 1
c) lim d) lim
x→0 sin2 x u→0 ln(1 + u) − u
 8. Probability
Brief answers to the unstarred exercises are given at the end of this section.

8A. Discrete Random Variables

8A-1 Buck Fuller rolls a fair dodecahedral die: it has 12 faces, all regular pentagons. The
outcome is a random variable X, with integer values 1, 2, . . . , 12. Find
(a) P(X is divisible by 3) (b) P(X is divisible by 5)

8A-2 A fair pair of dice is rolled; the output is a random variable Y , with values 2, . . . , 12.
Answer the same two questions as in the preceding exercise.

8A-3 Referring to Example 1.1B: you pick one of the football team’s shoes at random.
Find the probability of getting an even size; deduce what the probability of getting an odd
size is.

8A-4 Assume that when asked to pick a random positive integer, a person picks the integer
n with probability 1/2n. Let X be the associated random variable giving this outcome. Find
(a) P(X is even); (b) P(X is odd).
(c) Show that 6/16 ≤ P(X is prime) ≤ 7/16.

8A-5 Say Mrs. Field’s chocolate chip cookies average 10 chips per cookie. What’s the
probability of getting 5 or less chips in a cookie? (Assume the number of chips is a Poisson
random variable.)

8A-6 Assume the number of calls per night to def-tuv-tuv-oper-oper is a Poisson random
variable with mean 5. What’s the likelihood that there will be at least three calls tonight?

8A-7 Tabitha is Latexing her thesis, proof-reading as she word-processes. When printed,
about 20% of the pages turn out to be error-free. What is the likelihood that
a) a single page has at most one error?
b) three pages have a total of at most three errors?

8A-8 Suppose a calculus textbook has a total of 600 misprints in its 950 pages. What is
the probability that
a) a chapter containing 10 pages has no misprints?
b) a chapter 5 pages long has at least one misprint?

8B. Continuous random variables.

8B-1 Let X be an exponential random variable with parameter m = 2.
a) Calculate the expectation of X directly from its definition.
b) Calculate P (1 ≤ X ≤ 3).

8B-2 The average time between sales at the Chinese pastry booth in the lobby of Building
10 is 4/5 minute (they wish). If we assume the time is an exponential random variable,
what is the probability that the time between successive sales is
a) greater than 2 minutes? b) less than 4 minutes?

8B-3 Say that on the average, a baby is born somewhere in the U.S. every 10 seconds (we’re
assuming it’s not 9 months after a massive nighttime power outage). What’s the probability
of a time gap between two successive births lasting between one and two minutes?
1
2 E. 1801 EXERCISES

8B-4 Assume the mean length of time between auto accidents on Southeast Expressway is
10 hours. Estimate the probability of no accidents for 24 hours.

8B-5 My city-tire bicycle seems to get a flat on the Charles River bike path on the average
every 100 days. For what length of time t0 will the probability be 90% that I won’t get a
flat during any time interval of that length?

8B-6 If X is an exponential random variable with parameter m, what is the probability

that X exceeds its mean?

8B-7* In Example 2.1,

a) verify the formulas given for the density function and P (a ≤ x ≤ b);
b) find the distribution function.

3. Standard deviation
8C-1* Find the standard deviation of X if:
a) X is the outcome of tossing a fair die
b) X is the uniform continuous random variable with range [x1 , x2 ].

8C-2* In Theorem 3.1, prove: (a) (17) (b) (18)

8C-3* Prove the equality of the two integral formulas in (16) for the variance of a continuous
random variable.

4. Normal random variables

8D-1 Let Z be the standard normal random variable. Using the table for the values of the
associated distribution function Φ(Z), extended by (23) and (24), calculate the value of:
a) P(1.5 < Z < 2.5) b) P(Z ≤ −1) c) P(−1 < Z < 1) d) P(−1 < Z < 2.5)

8D-2 Assume the lifetime in hours of a flashlight battery is a normal random variable X
with mean 120 and standard deviation 36. Find the probability that it lasts between 85 and
135 hours. How many batteries in a sample of 160 would you expect to last that long, on
the average?

8D-3 Suppose the grades on an .01A test have a normal distribution with mean 70 and
standard deviation 10. If 300 students take the test and passing is set at 55, how many
fail? A mean professor decides that “keeping up the standards” requires that 10% of the
students fail. What will she announce as the passing grade?

8D-4 For each of the following normal random variables, give an interval in which the
variable lies with probability 95%.
a) Lifetime in hours of a flashlight battery if m = 120, σ = 36;
b) grade on an exam for which m = 70, σ = 10;
c) annual snowfall in inches, if the mean is 46!! and the standard deviation 4!! .

8D-5* Prove in Theorem 4.1 that σ(Z) = 1 for the standard normal random variable Z.

8D-6* Prove the first implication in (22) by making the change of variable.

8D-7* Prove the total area under the normal density function (28) is 1 by making a change
of variable in the integral; you can use the results in (20).
 8. PROBABILITY 3

8E. Central Limit Theorem

8E-1 Suppose the average luggage weight for an airline passenger is 38 lbs. with a standard
deviation of 8 lbs. What is the probability that the luggage for 80 passengers will weigh
over 3200 pounds?

8E-2 In an R/O week contest, one hundred freshmen independently estimate the height in
meters of a picket fence. Assume that the standard deviation for the individual guesses is
less than 1 dm (.1 meter). Give a lower bound on the probability that the average of their
guesses is off by less than 1 cm.

8E-3 A national poll is to estimate the percentage of Americans who favor a draft over
a volunteer military . Copy and complete the table below so that if n people are polled
at random, we can say with approximately 95% confidence that our error is less than e
percentage points.
n: 50 100 625 10,000
e: 5 4 3 2

8E-4 The Today Show announces that in a poll of 900 randomly chosen Americans, 52%
favored college tuition tax credits. In what range can you say with approximately 95%
confidence that the actual percentage lies?

8E-5 To prove that a coin is unfair, a judge tosses it 2,000 times. How many heads would
he need to get to prove with 95% confidence that the coin is unfair?

8E-6 One hundred reservations have been confirmed for the 98-seat flight from Boston to
Bangor. If generally 3% of the confirmed passengers do not show up, what is the probability
that someone will be bumped from the flight?

8E-7 A poll of 10,000 Bostonians a week before a gubernatorial election gives the incumbent
52% of the vote. In what range can you put his support with approximately 95% confidence?
 SOLUTIONS TO 18.01 EXERCISES

Unit 1. Differentiation

1A. Graphing

1A-1,2 a) y = (x − 1)2 − 2
b) y = 3(x2 + 2x) + 2 = 3(x + 1)2 − 1

2 2
1 1

-2 -1 -2 1

1a 1b 2a 2b

(−x)3 − 3x −x3 − 3x
1A-3 a) f (−x) = 4
= = −f (x), so it is odd.
1 − (−x) 1 − x4
b) (sin(−x))2 = (sin x)2 , so it is even.
odd
c) , so it is odd
even
d) (1 − x)4 "= ±(1 + x)4 : neither.
e) J0 ((−x)2 ) = J0 (x2 ), so it is even.

1A-4 a) p(x) = pe (x) + po (x), where pe (x) is the sum of the even powers and po (x) is the
sum of the odd powers
f (x) + f (−x) f (x) − f (−x)
b) f (x) = +
2 2
f (x) + f (−x) f (x) − f (−x)
F (x) = is even and G(x) = is odd because
2 2
f (−x) + f (−(−x)) f (x) − f (−x)
F (−x) = = F (x); G(−x) = = −G(−x).
2 2
c) Use part b:
1 1 2a 2a
+ = = 2 even
x + a −x + a (x + a)(−x + a) a − x2
1 1 −2x −2x
− = = 2 odd
x + a −x + a (x + a)(−x + a) a − x2
1 a x
=⇒ = 2 − 2
x+a a − x2 a − x2
c
!David Jerison and MIT 1996, 2003
 S. 18.01 SOLUTIONS TO EXERCISES

x−1 3y + 1
1A-5 a) y = . Crossmultiply and solve for x, getting x = , so the inverse
2x + 3 1 − 2y
3x + 1
function is .
1 − 2x
b) y = x2 + 2x = (x + 1)2 − 1
(Restrict domain to x ≤ −1, so when it’s flipped about
√ the diagonal y = x, you’ll still
get the√graph of a function.) Solving for x, we get x = y + 1 − 1, so the inverse function
is y = x + 1 − 1 .

g(x) g(x)
f(x)

f(x)

5a 5b

√
√ 3 π
√
1A-6 a) A = 1 + 3 = 2, tan c = ,c= 3. So sin x + 3 cos x = 2 sin(x + π3 ) .
1
√ π
b) 2 sin(x − )
4
π
1A-7 a) 3 sin(2x − π) = 3 sin 2(x − ), amplitude 3, period π, phase angle π/2.
2
π
b) −4 cos(x + ) = 4 sin x amplitude 4, period 2π, phase angle 0.
2
3 4

! 2! ! 2!

-3 -4

7a 7b

1A-8
f (x) odd =⇒ f (0) = −f (0) =⇒ f (0) = 0.
So f (c) = f (2c) = · · · = 0, also (by periodicity, where c is the period).

1A-9
3
2
-8 -4 4 8 12
-7 -5 -3 -1 1 3 5

-1
9c -6
9ab period = 4

c) The graph is made up of segments joining (0, −6) to (4, 3) to (8, −6). It repeats in
a zigzag with period 8. * This can be derived using:
x/2 − 1 = −1 =⇒ x = 0 and g(0) = 3f (−1) − 3 = −6
x/2 − 1 = 1 =⇒ x = 4 and g(4) = 3f (1) − 3 = 3
x/2 − 1 = 3 =⇒ x = 8 and g(8) = 3f (3) − 3 = −6
 1. DIFFERENTIATION

1B. Velocity and rates of change

1B-1 a) h = height of tube = 400 − 16t2 .

h(2) − h(0) (400 − 16 · 22 ) − 400

average speed = = −32ft/sec
2 2

(The minus sign means the test tube is going down. You can also do this whole problem
using the function s(t) = 16t2 , representing the distance down measured from the top. Then
all the speeds are positive instead of negative.)
b) Solve h(t) = 0 (or s(t) = 400) to find landing time t = 5. Hence the average speed
for the last two seconds is

h(5) − h(3) 0 − (400 − 16 · 32 )

= = −128ft/sec
2 2

c)

h(t) − h(5) 400 − 16t2 − 0 16(5 − t)(5 + t)

= =
t−5 t−5 t−5
= −16(5 + t) → −160ft/sec as t → 5

1B-2 A tennis ball bounces so that its initial speed straight upwards is b feet per second.
Its height s in feet at time t seconds is

s = bt − 16t2

a)

s(t + h) − s(t) b(t + h) − 16(t + h)2 − (bt − 16t2 )

=
h h
bt + bh − 16t2 − 32th − 16h2 − bt + 16t2
=
h
bh − 32th − 16h2
=
h
= b − 32t − 16h → b − 32t as h → 0

Therefore, v = b − 32t.
b) The ball reaches its maximum height exactly when the ball has finished going up.
This is time at which v(t) = 0, namely, t = b/32. v
b
c) The maximum height is s(b/32) = b2 /64. s
d) The graph of v is a straight line with slope
−32. The graph of s is a parabola with maximum b/32
t b/32 b/16 t
at place where v = 0 at t = b/32 and landing time
at t = b/16. graph of velocity graph of position
 S. 18.01 SOLUTIONS TO EXERCISES

e) If the initial velocity on the first bounce was b1 = b, and

√ the velocity of the second
bounce is b2 , then b22 /64 = (1/2)b21 /64. Therefore, b2 = b1 / 2. The second bounce is at
b1 /16 + b2 /16. (continued →)
f) If the ball continues to bounce then the landing times form a geometric series
√ √
b1 /16 + b2 /16 + b3 /16 + · · · = b/16 + b/16 2 + b/16( 2)2 + · · ·
√ √
= (b/16)(1 + (1/ 2) + (1/ 2)2 + · · · )
b/16
= √
1 − (1/ 2)
√
Put another way, the ball stops bouncing after 1/(1 − (1/ 2)) ≈ 3.4 times the length of
time the first bounce.

1C. Slope and derivative.

1C-1 a)

π(r + h)2 − πr2 π(r2 + 2rh + h2 ) − πr2 π(2rh + h2 )

= =
h h h
= π(2r + h)
→ 2πr as h → 0

b)

(4π/3)(r + h)3 − (4π/3)r3 (4π/3)(r3 + 3r2 h + 3rh2 + h3 ) − (4π/3)r3

=
h h
(4π/3)(3r2 h + 3rh2 + h3 )
=
h
= (4π/3)(3r2 + 3rh + h2 )
→ 4πr2 as h → 0

f (x) − f (a) (x − a)g(x) − 0

1C-2 = = g(x) → g(a) as x → a.
x−a x−a
1C-3 a)
! " ! "
1 1 1 1 2x + 1 − (2(x + h) + 1)
− =
h 2(x + h) + 1 2x + 1 h (2(x + h) + 1)(2x + 1)
! "
1 −2h
=
h (2(x + h) + 1)(2x + 1)
−2
=
(2(x + h) + 1)(2x + 1)
−2
−→ as h → 0
(2x + 1)2
 1. DIFFERENTIATION

b)

2(x + h)2 + 5(x + h) + 4 − (2x2 + 5x + 4) 2x2 + 4xh + 2h2 + 5x + 5h − 2x2 − 5x

=
h h
4xh + 2h2 + 5h
= = 4x + 2h + 5
h
−→ 4x + 5 as h → 0

c)

1 (x2 + 1) − ((x + h)2 + 1)

! " ! "
1 1 1
− =
h (x + h)2 + 1 x2 + 1 h ((x + h)2 + 1)(x2 + 1)
1 x + 1 − x2 − 2xh − h2 − 1
! 2 "
=
h ((x + h)2 + 1)(x2 + 1)
−2xh − h2
! "
1
=
h ((x + h)2 + 1)(x2 + 1)
−2x − h
=
((x + h)2 + 1)(x2 + 1)
−2x
−→ 2 as h → 0
(x + 1)2

d) Common denominator:
! " !√ √ "
1 1 1 1 x− x+h
√ −√ = √ √
h x+h x h x+h x
√ √
Now simplify the numerator by multiplying numerator and denominator by x + x + h,
and using (a − b)(a + b) = a2 − b2 :
√ √
( x)2 − ( x + h)2
! " ! "
1 1 x − (x + h)
√ √ √ √ = √ √ √ √
h x + h x( x + x + h) h x + h x( x + x + h)
! "
1 −h
= √ √ √ √
h x + h x( x + x + h)
! "
−1
= √ √ √ √
x + h x( x + x + h)
−1 1
−→ √ 3 = − x−3/2 as h → 0
2( x) 2

e) For part (a), −2/(2x + 1)2 < 0, so there are no points where the slope is 1 or 0. For
slope −1,
√ √
−2/(2x + 1)2 = −1 =⇒ (2x + 1)2 = 2 =⇒ 2x + 1 = ± 2 =⇒ x = −1/2 ± 2/2

For part (b), the slope is 0 at x = −5/4, 1 at x = −1 and −1 at x = −3/2.

1C-4 Using Problem 3,

 S. 18.01 SOLUTIONS TO EXERCISES

a) f " (1) = −2/9 and f (1) = 1/3, so y = −(2/9)(x − 1) + 1/3 = (−2x + 5)/9
b) f (a) = 2a2 + 5a + 4 and f " (a) = 4a + 5, so

y = (4a + 5)(x − a) + 2a2 + 5a + 4 = (4a + 5)x − 2a2 + 4

c) f (0) = 1 and f " (0) = 0, so y = 0(x − 0) + 1, or y = 1.

√
d) f (a) = 1/ a and f " (a) = −(1/2)a−3/2 , so
√
y = −(1/2)a3/2 (x − a) + 1/ a = −a−3/2 x + (3/2)a−1/2

1C-5 Method 1. y " (x) = 2(x − 1), so the tangent line through (a, 1 + (a − 1)2 ) is

y = 2(a − 1)(x − a) + 1 + (a − 1)2

In order to see if the origin is on this line, plug in x = 0 and y = 0, to get the following
equation for a.

0 = 2(a − 1)(−a) + 1 + (a − 1)2 = −2a2 + 2a + 1 + a2 − 2a + 1 = −a2 + 2

√
Therefore a = ± 2 and the two tangent lines through the origin are
√ √
y = 2( 2 − 1)x and y = −2( 2 + 1)x

(Because these are lines throught the origin, the constant terms must cancel: this is a good
check of your algebra!)
Method 2. Seek tangent lines of the form y = mx. Suppose that y = mx meets
y = 1 + (x − 1)2 , at x = a, then ma = 1 + (a − 1)2 . In addition we want the slope
y " (a) = 2(a − 1) to be equal to m, so m = 2(a − 1). Substituting for m we find

2(a − 1)a = 1 + (a − 1)2

√ √
This is the same equation as in method 1: a2 − 2 = 0, so a = ± 2 and m = 2(± 2 − 1),
and the two tangent lines through the origin are as above,
√ √
y = 2( 2 − 1)x and y = −2( 2 + 1)x

1C-6

2
-2 4
2
-2

-2
(even) (odd) period = 6
5a 5b 5c 5d 5e
 1. DIFFERENTIATION

1D. Limits and continuity

1D-1 Calculate the following limits if they exist. If they do not exist, then indicate whether
they are +∞, −∞ or undefined.
a) −4
b) 8/3
c) undefined (both ±∞ are possible)
d) Note that 2 − x is negative when x > 2, so the limit is −∞
e) Note that 2 − x is positive when x < 2, so the limit is +∞ (can also be written ∞)
4x2 4x ∞
f) = → = ∞ as x → ∞
x−2 1 − (2/x) 1
4x2 4x2 − 4x(x − 2) 8x 8
g) − 4x = = = → 8 as x → ∞
x−2 x−2 x−2 1 − (2/x)
x2 + 2x + 3 1 + (2/x) + (3/x2 ) 1
i) = → as x → ∞
3x2 − 2x + 4 3 − (2/x) + 4/x2 ) 3
x−2 x−2 1 1
j) 2
= = → as x → 2
x −4 (x − 2)(x + 2) x+2 4
√ 1 1
1D-2 a) lim x=0 b) lim =∞ lim = −∞
x→0+ x→1+ x−1 x→1− x−1
c) lim (x − 1)−4 = ∞ (left and right hand limits are same)
x→1

d) lim | sin x| = 0 (left and right hand limits are same)

x→0

|x| |x|
e) lim =1 lim = −1
x→0+ x x→0− x
1D-3 a) x = 2 removable x = −2 infinite b) x = 0, ±π, ±2π, ... infinite
c) x = 0 removable d) x = 0 removable e) x = 0 jump f) x = 0 removable

1D-4

(!1,1)
2 (0,.5)

4a 4b

1D-5 a) for continuity, want ax + b = 1 when x = 1. Ans.: all a, b such that a + b = 1

2
dy d(x ) d(ax + b)
b) = = 2x = 2 when x = 1 . We have also = a. Therefore, to
"
dx dx dx
make f (x) continuous, we want a = 2.
Combining this with the condition a + b=1 from part (a), we get finally b = −1, a = 2.
 S. 18.01 SOLUTIONS TO EXERCISES

1D-6 a) f (0) = 02 + 4 · 0 + 1 = 1. Match the function values:

f (0− ) = lim ax + b = b, so b = 1 by continuity.

x→0

Next match the slopes:

f " (0+ ) = lim 2x + 4 = 4
x→0

and f " (0− ) = a. Therefore, a = 4, since f " (0) exists.

b)
f (1) = 12 + 4 · 1 + 1 = 6 and f (1− ) = lim ax + b = a + b
x→1

Therefore continuity implies a + b = 6. The slope from the right is

f " (1+ ) = lim 2x + 4 = 6

x→1

Therefore, this must equal the slope from the left, which is a. Thus, a = 6 and b = 0.

1D-7
f (1) = c12 + 4 · 1 + 1 = c + 5 and f (1− ) = lim ax + b = a + b
x→1

Therefore, by continuity, c + 5 = a + b. Next, match the slopes from left and right:

f " (1+ ) = lim 2cx + 4 = 2c + 4 and f " (1− ) = lim a = a

x→1 x→1

Therefore,
a = 2c + 4 and b = −c + 1.
1D-8
a)
f (0) = sin(2 · 0) = 0 and f (0+ ) = lim ax + b = b
x→0

Therefore, continuity implies b = 0. The slope from each side is

f " (0− ) = lim 2 cos(2x) = 2 and f " (0+ ) = lim a = a

x→0 x→0

Therefore, we need a "= 2 in order that f not be differentiable.

b)
f (0) = cos(2 · 0) = 1 and f (0+ ) = lim ax + b = b
x→0

Therefore, continuity implies b = 1. The slope from each side is

f " (0− ) = lim −2 sin(2x) = 0 and f " (0+ ) = lim a = a

x→0 x→0

Therefore, we need a "= 0 in order that f not be differentiable.

1D-9 There cannot be any such values because every differentiable function is continuous.
 1. DIFFERENTIATION

1E: Differentiation formulas: polynomials, products, quotients

1E-1 Find the derivative of the following polynomials

a) 10x9 + 15x4 + 6x2
b) 0 (e2 + 1 ≈ 8.4 is a constant and the derivative of a constant is zero.)
c) 1/2
d) By the product rule: (3x2 + 1)(x5 + x2 ) + (x3 + x)(5x4 + 2x) = 8x7 + 6x5 + 5x4 + 3x2 .
Alternatively, multiply out the polynomial first to get x8 +x6 +x5 +x3 and then differentiate.

1E-2 Find the antiderivative of the following polynomials

a) ax2 /2 + bx + c, where a and b are the given constants and c is a third constant.
b) x7 /7 + (5/6)x6 + x4 + c
c) The only way to get at this is to multiply it out: x6 + 2x3 + 1. Now you can take
the antiderivative of each separate term to get

x7 x4
+ +x+c
7 2

Warning: The answer is not (1/3)(x3 + 1)3 . (The derivative does not match if you apply
the chain rule, the rule to be treated below in E4.)

1E-3 y " = 3x2 + 2x − 1 = 0 =⇒ (3x − 1)(x + 1) = 0. Hence x = 1/3 or x = −1 and the

points are (1/3, 49/27) and (−1, 3)

1E-4 a) f (0) = 4, and f (0− ) = lim 5x5 + 3x4 + 7x2 + 8x + 4 = 4. Therefore the function
x→0
is continuous for all values of the parameters.

f " (0+ ) = lim 2ax + b = b and f " (0− ) = lim 25x4 + 12x3 + 14x + 8 = 8
x→0 x→0

Therefore, b = 8 and a can have any value.

b) f (1) = a + b + 4 and f (1+ ) = 5 + 3 + 7 + 8 + 4 = 27. So by continuity,

a + b = 23

f " (1− ) = lim 2ax + b = 2a + b; f " (1+ ) = lim 25x4 + 12x3 + 14x + 8 = 59.
x→1 x→1

Therefore, differentiability implies

2a + b = 59
Subtracting the first equation, a = 59 − 23 = 36 and hence b = −13.

1 1 − 2ax − x2 −x2 − 4x − 1
1E-5 a) b) c)
(1 + x)2 (x2 + 1)2 (x2 − 1)2
d) 3x2 − 1/x2
 S. 18.01 SOLUTIONS TO EXERCISES

1F. Chain rule, implicit differentiation

1F-1 a) Let u = (x2 + 2)

d 2 du d 2
u = u = (2x)(2u) = 4x(x2 + 2) = 4x3 + 8x
dx dx du
Alternatively,
d 2 d 4
(x + 2)2 = (x + 4x2 + 4) = 4x3 + 8x
dx dx
d 100 du d 100
b) Let u = (x2 + 2); then u = u = (2x)(100u99 ) = (200x)(x2 + 2)99 .
dx dx du
1F-2 Product rule and chain rule:

10x9 (x2 + 1)10 + x10 [10(x2 + 1)9 (2x)] = 10(3x2 + 1)x9 (x2 + 1)9

1F-3 y = x1/n =⇒ y n = x =⇒ ny n−1 y " = 1. Therefore,

1 1 1−n 1 1
y" = = y = x n −1
ny n−1 n n

1F-4 (1/3)x−2/3 + (1/3)y −2/3 y " = 0 implies

y " = −x−2/3 y 2/3

Put u = 1 − x1/3 . Then y = u3 , and the chain rule implies

dy du
= 3u2 = 3(1 − x1/3 )2 (−(1/3)x−2/3 ) = −x−2/3 (1 − x1/3 )2
dx dx
The chain rule answer is the same as the one using implicit differentiation because

y = (1 − x1/3 )3 =⇒ y 2/3 = (1 − x1/3 )2

1F-5 Implicit differentiation gives cos x + y " cos y = 0. Horizontal slope means y " = 0,
so that cos x = 0. These are the points x = π/2 + kπ for every integer k. Recall that
sin(π/2 + kπ) = (−1)k , i.e., 1 if k is even and −1 if k is odd. Thus at x = π/2 + kπ,
±1 + sin y = 1/2, or sin y = ∓1 + 1/2. But sin y = 3/2 has no solution, so the only
solutions are when k is even and in that case sin y = −1 + 1/2, so that y = −π/6 + 2nπ or
y = 7π/6 + 2nπ. In all there are two grids of points at the vertices of squares of side 2π,
namely the points

(π/2 + 2kπ, −π/6 + 2nπ) and (π/2 + 2kπ, 7π/6 + 2nπ); k, n any integers.

1F-6 Following the hint, let z = −x. If f is even, then f (x) = f (z) Differentiating and
using the chain rule:

f " (x) = f " (z)(dz/dx) = −f " (z) because dz/dx = −1

But this means that f " is odd. Similarly, if g is odd, then g(x = −g(z). Differentiating and
using the chain rule:

g " (x) = −g " (z)(dz/dx) = g " (z) because dz/dx = −1

 1. DIFFERENTIATION

dD 1 x−a
1F-7 a) = ((x − a)2 + y0 2 )−1/2 (2(x − a)) = #
dx 2 (x − a)2 + y0 2
dm −1 −2v m0 v
b) = m0 · (1 − v 2 /c2 )−3/2 · 2 = 2
dv 2 c c (1 − v 2 /c2 )3/2
dF 3 −3mgr
c) = mg · (− )(1 + r2 )−5/2 · 2r =
dr 2 (1 + r2 )5/2
dQ −6bt a a(1 − 5bt2 )
d) = at · 2 4
+ 2 3
=
dt (1 + bt ) (1 + bt ) (1 + bt2 )4

1 2 1 −r2 −r
1F-8 a) V = πr h =⇒ 0 = π(2rr" h + r2 ) =⇒ r" = =
3 3 2rh 2h
cP V c−1 cP
b) P V c = nRT =⇒ P " V c + P · cV c−1 = 0 =⇒ P " = − =−
Vc V
c) c2 = a2 + b2 − 2ab cos θ implies

−2b + 2 cos θ · a a cos θ − b

0 = 2aa" + 2b − 2(cos θ(a" b + a)) =⇒ a" = =
2a − 2 cos θ · b a − b cos θ

1G. Higher derivatives

−10 −10
1G-1 a) 6 − x−3/2 b) c) d) 0
(x + 5)3 (x + 5)3

1G-2 If y """ = 0, then y "" = c0 , a constant. Hence y " = c0 x + c1 , where c1 is some other
constant. Next, y = c0 x2 /2 + c1 x + c2 , where c2 is yet another constant. Thus, y must be
a quadratic polynomial, and any quadratic polynomial will have the property that its third
derivative is identically zero.

1G-3
x2 y2 2x 2yy "
+ = 1 =⇒ + 2 = 0 =⇒ y " = −(b2 /a2 )(x/y)
a2 b2 a2 b
Thus,

b2
$ 2%$
y + x(b2 /a2 )(x/y)
$
%$ % %
y − xy " b
y "" = − = −
a2 y2 a2 y2
$ 4 %
b b4
=− 3 2
(y 2 /b2 + x2 /a2 ) = − 2 3
y a a y

1G-4 y = (x + 1)−1 , so y (1) = −(x + 1)−2 , y (2) = (−1)(−2)(x + 1)−3 , and

y (3) = (−1)(−2)(−3)(x + 1)−4 .

The pattern is
y (n) = (−1)n (n!)(x + 1)−n−1
 S. 18.01 SOLUTIONS TO EXERCISES

1G-5 a) y " = u" v + uv " =⇒ y "" = u"" v + 2u" v " + uv ""

b) Formulas above do coincide with Leibniz’s formula for n = 1 and n = 2. To calculate
y (p+q) where y = xp (1 +$x)% q
, use u = xp and v = (1 + x)q . The only term in the Leibniz
n (p) (q)
formula that is not 0 is u v , since in all other terms either one factor or the other
k
is 0. If u = xp , u(p) = p!, so
$ %
(p+q) n n!
y = p!q! = · p!q! = n!
p p!q!

1H. Exponentials and Logarithms: Algebra

y0 1
1H-1 a) To see when y = y0 /2, we must solve the equation = y0 e−kt , or 2 = e−kt .
2
ln 2
Take ln of both sides: − ln 2 = −kt, from which t = .
k
− ln 2 1
b) y1 = y0 ekt1 by assumption, λ = y0 ek(t1 +λ) = y0 ekt1 · ekλ = y1 · e− ln 2 = y1 ·
k 2
1H-2 pH = − log10 [H + ]; by assumption, [H + ]dil = 12 [H + ]orig . Take − log10 of both sides
(note that log 2 ≈ .3):
− log [H + ]dil = log 2 − log [H + ]orig =⇒ pHdil = pHorig + log2 .

1H-3 a) ln(y + 1) + ln(y − 1) = 2x + ln x; exponentiating both sides and solving for y:

√
(y + 1) · (y − 1) = e2x · x =⇒ y 2 − 1 = xe2x =⇒ y = xe2x + 1, since y > 0.
y+1 2
b) log(y+1)−log(y−1) = −x2 ; exponentiating, = 10−x . Solve for y; to simplify
y−1
2
2 A+1 10−x + 1
the algebra, let A = 10−x . Crossmultiplying, y + 1 = Ay − A =⇒ y = = −x2
A−1 10 −1
c) 2 ln y − ln(y + 1) = x; exponentiating both sides and solving for y:
√
y2 x 2 x x ex e2x + 4ex
= e =⇒ y − e y − e = 0 =⇒ y = , since y − 1 > 0.
y+1 2
ln a log a
= c ⇒ ln a = c ln b ⇒ a = ec ln b = eln b = bc . Similarly, = c ⇒ a = bc .
c
1H-4
ln b log b

u2 + 1
1H-5 a) Put u = ex (multiply top and bottom by ex first): = y; this gives
u2 − 1
y+1 y+1 1 y+1
u2 = = e2x ; taking ln: 2x = ln( ), x= ln( )
y−1 y−1 2 y−1
1
b) ex +e−x = y; putting u = ex gives u+ = y ; solving for u gives u2 −yu+1 = 0
# u #
y ± y2 − 4 x y ± y2 − 4
so that u = = e ; taking ln: x = ln( )
2 2
1H-6 A = log e · ln 10 = ln(10log e ) = ln(e) = 1 ; similarly, logb a · loga b = 1
 1. DIFFERENTIATION

1H-7 a) If I1 is the intensity of the jet and I2 is the intensity of the conversation, then
$ %
I1 /I0
log10 (I1 /I2 ) = log10 = log10 (I1 /I0 ) − log10 (I2 /I0 ) = 13 − 6 = 7
I2 /I0

Therefore, I1 /I2 = 107 .

b) I = C/r2 and I = I1 when r = 50 implies

I1 = C/502 =⇒ C = I1 502 =⇒ I = I1 502 /r2

This shows that when r = 100, we have I = I1 502 /1002 = I1 /4 . It follows that

10 log10 (I/I0 ) = 10 log10 (I1 /4I0 ) = 10 log10 (I1 /I0 ) − 10 log10 4 ≈ 130 − 6.0 ≈ 124

The sound at 100 meters is 124 decibels.

The sound at 1 km has 1/100 the intensity of the sound at 100 meters, because 100m/1km =
1/10.
10 log10 (1/100) = 10(−2) = −20
so the decibel level is 124 − 20 = 104.

1I. Exponentials and Logarithms: Calculus

2 2
1I-1 a) (x + 1)ex b) 4xe2x c) (−2x)e−x d) ln x e) 2/x f) 2(ln x)/x g) 4xe2x
"
h) (xx )" = ex ln x = (x ln x)" ex ln x = (ln x + 1)ex ln x = (1 + ln x)xx
& '

i) (ex − e−x )/2 j) (ex + e−x )/2 k) −1/x l) −1/x(ln x)2 m) −2ex/(1 + ex )2

1I-2 1

(even)
1I-3 a) As n → ∞, h = 1/n → 0.
(
1 ln(1 + h) ln(1 + h) − ln(1) d (
n ln(1 + ) = = −→ ln(1 + x)(( =1
n h h h→0 dx x=0

Therefore,
1
lim n ln(1 + )=1
n→∞ n

b) Take the logarithm of both sides. We need to show

1 n
lim ln(1 + ) = ln e = 1
n→∞ n
But
1 n 1
ln(1 + ) = n ln(1 + )
n n
so the limit is the same as the one in part (a).
 S. 18.01 SOLUTIONS TO EXERCISES

1I-4 a)
$ %3n $$ %n %3
1 1
1+ = 1+ −→ e3 as n → ∞,
n n

b) Put m = n/2. Then

$ %5n $ %10m $$ %m %10
2 1 1
1+ = 1+ = 1+ −→ e10 as m → ∞
n m m

c) Put m = 2n. Then

$ %5n $ %5m/2 $$ %m %5/2
1 1 1
1+ = 1+ = 1+ −→ e5/2 as m → ∞
2n m m

1J. Trigonometric functions

1J-1 a) 10x cos(5x2 ) b) 6 sin(3x) cos(3x) c) −2 sin(2x)/ cos(2x) = −2 tan(2x)

d) −2 sin x/(2 cos x) = − tan x. (Why did the factor 2 disappear? Because ln(2 cos x) =
ln 2 + ln(cos x), and the derivative of the constant ln 2 is zero.)
x cos x − sin x 2
e) f) −(1 + y " ) sin(x + y) g) − sin(x + y) h) 2 sin x cos xesin x
x2
(x2 sin x)" 2x sin x + x2 cos x 2
i) 2
= = + cot x. Alternatively,
x sin x x2 sin x x

ln(x2 sin x) = ln(x2 ) + ln(sin x) = 2 ln x + ln sin x

2 cos x 2
Differentiating gives + = + cot x
x sin x x
j) 2e2x sin(10x) + 10e2x cos(10x) k) 6 tan(3x) sec2 (3x) = 6 sin x/ cos3 x
√ √
l) −x(1 − x2 )−1/2 sec( 1 − x2 ) tan( 1 − x2 )
m) Using the chain rule repeatedly and the trigonometric double angle formulas,

(cos2 x − sin2 x)" = −2 cos x sin x − 2 sin x cos x = −4 cos x sin x;

(2 cos2 x)" = −4 cos x sin x;
(cos(2x))" = −2 sin(2x) = −2(2 sin x cos x).

The three functions have the same derivative, so they differ by constants. And indeed,

cos(2x) = cos2 x − sin2 x = 2 cos2 x − 1, (using sin2 x = 1 − cos2 x).

n)

5(sec(5x) tan(5x)) tan(5x) + 5(sec(5x)(sec2 (5x)) = 5 sec(5x)(sec2 (5x) + tan2 (5x))

Other forms: 5 sec(5x)(2 sec2 (5x) − 1); 10 sec3 (5x) − 5 sec(5x)

 1. DIFFERENTIATION

o) 0 because sec2 (3x) − tan2 (3x) = 1, a constant — or carry it out for practice.
p) Successive use of the chain rule:
# # 1
(sin ( x2 + 1))" = cos ( x2 + 1) · (x2 + 1)−1/2 · 2x
2
x #
=√ cos ( x2 + 1)
x2 + 1

q) Chain rule several times in succession:

# # # −x
(cos2 1 − x2 )" = 2 cos 1 − x2 · (− sin 1 − x2 ) · √
1 − x2
x #
=√ sin(2 1 − x2 )
1 − x2

r) Chain rule again:

$ %
2 x x x x+1−x
tan ( ) = 2 tan( ) · sec2 ( )·
x+1 x+1 x+1 (x + 1)2
2 x x
= tan( ) sec2 ( )
(x + 1)2 x+1 x+1

1J-2 Because cos(π/2) = 0,

cos x cos x − cos(π/2) d

lim = lim = cos x|x=π/2 = − sin x|x=π/2 = −1
x→π/2 x − π/2 x→π/2 x − π/2 dx

1J-3 a) (sin(kx))" = k cos(kx). Hence

(sin(kx))"" = (k cos(kx))" = −k 2 sin(kx).

Similarly, differentiating cosine twice switches from sine and then back to cosine with only
one sign change, so
(cos(kx)"" = −k 2 cos(kx)
Therefore,
sin(kx)"" + k 2 sin(kx) = 0 and cos(kx)"" + k 2 cos(kx) = 0
√
Since we are assuming k > 0, k = a.
b) This follows from the linearity of the operation of differentiation. With k 2 = a,

(c1 sin(kx) + c2 cos(kx))"" + k 2 (c1 sin(kx) + c2 cos(kx))

= c1 (sin(kx))"" + c2 (cos(kx))"" + k 2 c1 sin(kx) + k 2 c2 cos(kx)
= c1 [(sin(kx))"" + k 2 sin(kx)] + c2 [(cos(kx))"" + k 2 cos(kx)]
= c1 · 0 + c2 · 0 = 0
 S. 18.01 SOLUTIONS TO EXERCISES

c) Since φ is a constant, d(kx + φ)/dx = k, and (sin(kx + φ)" = k cos(kx + φ),

(sin(kx + φ)"" = (k cos(kx + φ))" = −k 2 sin(kx + φ)

Therefore, if a = k 2 ,
(sin(kx + φ)"" + a sin(kx + φ) = 0

d) The sum formula for the sine function says

sin(kx + φ) = sin(kx) cos(φ) + cos(kx) sin(φ)

In other words
sin(kx + φ) = c1 sin(kx) + c2 cos(kx)
with c1 = cos(φ) and c2 = sin(φ).

1J-4 a) The Pythagorean theorem implies that

c2 = sin2 θ + (1 − cos θ)2 = sin2 θ + 1 − 2 cos θ + cos2 θ = 2 − 2 cos θ

Thus, )
√ 1 − cos θ
c = 2 − 2 cos θ = 2 = 2 sin(θ/2)
2

b) Each angle is θ = 2π/n, so the perimeter of the n-gon is

n sin(2π/n)

As n → ∞, h = 2π/n tends to 0, so

2π sin h − sin 0 d
n sin(2π/n) = sin h = 2π → 2π sin x|x=0 = 2π cos x|x=0 = 2π
h h dx
 2. Applications of Differentiation
2A. Approximation
d√ b √ b
2A-1 a + bx = √ ⇒ f (x) ≈ a + √ x by formula.
dx 2 a + bx 2 a
!
√ √ bx √ bx
By algebra: a + bx = a 1 + ≈ a(1 + ), same as above.
a 2a
1 −b 1 b 1 1/a 1 b
2A-2 D( )= 2
⇒ f (x) ≈ − 2 x; OR: = ≈ (1− x).
a + bx (a + bx) a a a + bx 1 + b/ax a a
3
(1 + x)3/2 (1 + 2x) · 2 · (1 + x)3/2 − (1 + x)3/2 · 2 1
2A-3 D( )= 2
⇒ f ! (0) = −
1 + 2x (1 + 2x) 2
1 (1 + x)3/2 3 1
⇒ f (x) ≈ 1 − x; OR, by algebra, ≈ (1 + x)(1 − 2x) ≈ 1 − x.
2 1 + 2x 2 2
h g 2h
2A-4 Put = !; then w = ≈ g(1 − !)2 ≈ g(1 − 2!) = g(1 − ).
R (1 + !)2 R
2A-5 A reasonable assumption is that w is propotional to volume v, which is in turn propor-
tional to the cube of a linear dimension, i.e., a given person remains similar to him/herself,
for small weight changes.) Thus w = Ch3 ; since 5 feet = 60 inches, we get
w(60 + !) C(60 + !)3 ! 3! 1
= 3
= (1 + )3 ⇒ w(60 + !) ≈ w(60) · (1 + ) ≈ 120 · (1 + ) ≈ 126.
w(60) C(60) 60 60 20
[Or you can calculate the linearization of w(h) arround h = 60 using derivatives, and
using the value w(60) to determine C. getting w(h) ≈ 120 + 6(h − 60)
sin θ θ
2A-6 tan θ = ≈ ≈ θ(1 + θ2 //2) ≈ θ
cos θ 1 − θ2 /2
sec x 1 1 1 2
2A-7 √ = √ ≈ 1 2 1 2 ≈ 1 − x2 ≈ 1 + x
1−x 2 cos x 1 − x2 (1 − 2 x )(1 − 2 x )
1 1 1 2
2A-8 = = =
1−x 1 − ( 12 + ∆x) 1
2 − ∆x 1 − 2∆x
≈ 2(1 + 2∆x + 4(∆x)2 ) ≈ 2 + 4(x − 12 ) + 8(x − 21 )2

2A-10 y = (1 + x)r , y ! = r(1 + x)r−1 , y !! = r(r − 1)(1 + x)r−2

r(r − 1) 2
Therefore y(0) = 1, y ! (0) = r, y !! (0) = r(r − 1), giving (1 + x)r ≈ 1 + rx + x .
2
∆v −k
2A-11 pv k = c ⇒ p = cv −k = c((v0 + ∆v)−k = cv0 −k (1 + )
v0
c ∆v k(k + 1) ∆v 2
≈ k
(1 − k + ( ) )
v0 v0 2 v0
x
e x2 5
2A-12 a) ≈ (1 + x + )(1 + x + x2 ) ≈ 1 + 2x + x2
1−x 2 2
 S. SOLUTIONS TO 18.01 EXERCISES

ln(1 + x) x
b) ≈ ≈1−x
xex x(1 + x)
2
c) e−x ≈ 1 − x2 [Substitute into ex ≈ 1 + x]
x2 x2
d) ln(cos x) ≈ ln(1 − )≈− [since ln(1 + h) ≈ h]
2 2
h2 h2 (x − 1)2
e) x ln x = (1 + h) ln(1 + h) ≈ (1 + h)(h − ) ≈ h+ ⇒ x ln x ≈ (x − 1) +
2 2 2
2A-13 Finding the linear and quadratic approximation
a) 2x (both linear and quadratic)
b) 1, 1 − 2x2
c) 1, 1 + x2 /2 (Use (1 + u)−1 ≈ 1 − u with u = x2 /2:
sec x = 1/ cos x ≈ 1/(1 − x2 /2) = (1 − x2 /2)−1 ≈ 1 + x2 /2

d) 1, 1 + x2
e) Use (1 + u)−1 ≈ 1 − u + u2 :
(a + bx)−1 = a−1 (1 + (bx/a))−1 ≈ a−1 (1 − bx/a + (bx/a)2 )
Linear approximation: (1/a) − (b/a2 )x
Quadratic approximation: (1/a) − (b/a2 )x + (b2 /a3 )x2
f) f (x) = 1/(a + bx) so that f ! (1) = −b(a + b)−2 and f !! (1) = 2b2 /(a + b)−3 . We need
to assume that these numbers are defined, in other words that a + b %= 0. Then the linear
approximation is
1/(a + b) − (b/(a + b)2 )(x − 1)
and the quadratic approximation is
1/(a + b) − (b/(a + b)2 )(x − 1) + (b/(a + b)3 )(x − 1)2

Method 2: Write
1/(a + bx) = 1/(a + b + b(x − 1))
Then use the expansion of problem (e) with a + b in place of a and b in place of b and (x − 1)
in place of x. The requirement a %= 0 in (e) corresponds to the restriction a + b %= 0 in (f).

2A-15 f (x) = cos(3x), f ! (x) = −3 sin(3x), f !! (x) = −9 cos(3x). Thus,

f (0) = 1, f (π/6) = cos(π/2) = 0, f (π/3) = cos π = −1

! !
f (0) = −3 sin 0 = 0, f (π/6) = −3 sin(π/2) = −3, f ! (π/3) = −3 sin π = 0
f !! (0) = −9, f !! (π/6) = 0, f !! (π/3) = 9
Using these values, the linear and quadratic approximations are respectively:
for x ≈ 0 : f (x) ≈ 1 and f (x) ≈ 1 − (9/2)x2
for x ≈ π/6 : both are f (x) ≈ −3(x − π/6)
for x ≈ π/3 : f (x) ≈ −1 and f (x) ≈ −1 + (9/2)(x − π/3)2
 2. APPLICATIONS OF DIFFERENTIATION

2A-16 a) The law of cosines says that for a triangle with sides a, b, and c, with θ opposite
the side of length c,
c2 = a2 + b2 − 2ab cos θ
Apply it to one of the n triangles with vertex at the origin: a = b = 1 and θ = 2π/n. So
the formula is "
c = 2 − 2 cos(2π/n)
"
b) The perimeter is n 2 − 2 cos(2π/n). The quadratic approximation to cos θ near 0
is
cos θ ≈ 1 − θ2 /2
Therefore, as n → ∞ and θ = 2π/n → 0,
" " "
n 2 − 2 cos(2π/n) ≈ n 2 − 2(1 − (1/2)(2π/n)2 ) = n (2π/n)2 = n(2π/n) = 2π
In other words, "
lim n 2 − 2 cos(2π/n) = 2π,
n→∞
the circumference of the circle of radius 1.

2B. Curve Sketching

2B-1 a) y = x3 − 3x + 1, y ! = 3x2 − 3 = 3(x − 1)(x + 1). y ! = 0 =⇒ x = ±1.

Endpoint values: y → −∞ as x → −∞, and y → ∞ as x → ∞.
Critical values: y(−1) = 3, y(1) = −1. (-1, 3)

Increasing on: −∞ < x < −1, 1 < x < ∞.

Decreasing on: −1 < x < 1.
Graph: (−∞, −∞) ( (−1, 3) ) (1, −1) ( (∞, ∞), 1a (1,-1)
crossing the x-axis three times.
b) y = x4 − 4x + 1, y ! = x3 − 4. y ! = 0 =⇒ x = 41/3 .
Increasing on: 41/3 < x < ∞; decreasing on: −∞ < x < 41/3 .
Endpoint values: y → ∞ as x → ±∞; critical value: y(41/3 ) = 1.
Graph: (−∞, ∞) ) (41//3 , 1) ( (∞, ∞), never crossing the x-axis. (See below.)

c) y ! (x) = 1/(1 + x2 ) and y(0) = 0. By inspection, y ! > 0 for all x, hence always
increasing.
Endpoint values: y → c as x → ∞ and by symmetry y → −c as x → −∞. (But it is
not clear at this point in the course whether c = ∞ or some finite value. It turns out (in
Lecture 26) that y → c = π/2.
Graph: (−∞, −c) ( (∞, c), crossing the x-axis once (at x = 0). (See below.)

(2,4)
(4/3, 1) (odd) 1
1
-4
1b 1c 1d 1e
 S. SOLUTIONS TO 18.01 EXERCISES

d) y = x2 /(x−1), y ! = (2x(x−1)−x2 )/(x−1)2 = (x2 −2x)/(x−1)2 = (x−2)x/(x−1)2 .

Endpoint values: y → ∞ as x → ∞ and y → −∞ as x → −∞.
Singular values: y(1+ ) = +∞ and y(1− ) = −∞.
Critical values: y(0) = 0 and y(2) = 4.
New feature: Pay attention to sign changes in the denominator of y ! .
Increasing on: −∞ < x < 0 and 2 < x < ∞
Decreasing on: 0 < x < 1 and 1 < x < 2
Graph: (−∞, −∞) ( (0, 0) ) (1, −∞) ↑ (1, ∞) ) (2, 4) ( (∞, ∞), crossing the x-axis
once (at x = 0).
Commentary on singularities: Look out for sign changes both where y ! is zero and also
where y ! is undefined: y ! = 0 indicates a possible sign change in the numerator and y !
undefined indicates a possible sign change in the denominator. In this case there was no
sign change in y ! at x = 1, but there would have been a sign change, if there had been an
odd power of (x − 1) in the denominator.

e) y = x/(x + 4), y ! = ((x + 4) − x)/(x + 4)2 = 4/(x + 4)2 . No critical points.

Endpoint values: y → 1 as x → ±∞.
Increasing on: −4 < x < ∞.
Decreasing on: −∞ < x < −4.
Singular values: y(−4+ ) = −∞, y(−4− ) = +∞.
Graph: (−∞, 1) ( (−4, ∞) ↓ (−4, −∞) ( (∞, 1), crossing the x-axis once (at x = 0).
√
f) y = x + 1/(x − 3), y ! = −(1/2)(x + 5)(x + 1)−1/2 (x − 3)−2 No critical points
because x = −5 is outside of the domain of definition, x ≥ −1.
Endpoint values: y(−1) = 0, and as x → ∞,

1 + x1 1
y= √ 3 → =0
x − √x ∞

Singular values: y(3+ ) = +∞, y(3− ) = −∞.

Increasing on: nowhere
Decreasing on: −1 < x < 3 and 3 < x < ∞.
Graph: (−1, 0) ) (3, −∞) ↑ (3, ∞) ) (∞, 0), crossing the x-axis once (at x = −1).

g) y = 3x4 − 16x3 + 18x2 + 1, y ! = 12x3 − 48x2 + 36x = 12x(x − 1)(x − 3). y ! = 0 =⇒

x = 0, 1, 3.
Endpoint values: y → ∞ as x → ±∞.
Critical values: y(0) = 1, y(1) = 6, and y(3) = −188.
Increasing on: 0 < x < 1 and 3 < x < ∞.
 2. APPLICATIONS OF DIFFERENTIATION
(1,6)

-1 3 1

1f 1g 1h
1i
(3,-26)

Decreasing on: −∞ < x < 0 and 1 < x < 3.

Graph: (−∞, ∞) ) (0, 1) ( (1, 6) ) (3, −188) ( (∞, ∞), crossing the x-axis once.
2 2
h) y = e−x , y ! = −2xe−x . y ! = 0 =⇒ x = 0.
Endpoint values: y → 0 as x → ±∞.
Critical value: y(0) = 1.
Increasing on: −∞ < x < 0
Decreasing on: 0 < x < ∞
Graph: (−∞, 0) ( (0, 1) ) (∞, 0), never crossing the x-axis. (The function is even.)

2
i) y ! = e−x and y(0) = 0. Because y ! is even and y(0) = 0, y is odd. No critical points.
Endpoint values: y → c as x → ∞ and by symmetry y → −c as x → −∞. It is not clear
at this point in the course whether c is finite or infinite. But we will be able to show that c
is finite when we discuss improper integrals√in Unit 6. (Using a trick with iterated integrals,
a subject in 18.02, one can show that c = π/2.)
Graph: (−∞, −c) ( (∞, c), crossing the x-axis once (at x = 0).

2B-2 a) One inflection point at x = 0. (y !! = 6x)

b) No inflection points. y !! = 3x2 , so the function is convex. x = 0 is not a point of
inflection because y !! > 0 on both sides of x = 0.
c) Inflection point at x = 0. (y !! = −2x/(1 + x2 )2 )
d) No inflection points. Reasoning: y !! = 2/(x − 1)3 . Thus y !! > 0 and the function
is concave up when x > 1, and y !! < 0 and the function is concave down when x < 1. But
x = 1 is not called an inflection point because the function is not continuous there. In fact,
x = 1 is a singular point.
e) No inflection points. y !! = −8/(x + 1)3 . As in part (d) there is a sign change in y !! ,
but at a singular point not an inflection point.
f) y !! = −(1/2)[(x + 1)(x − 3) − (1/2)(x + 5)(x − 3) − 2(x + 5)(x + 1)](x + 1)−3/2(x − 3)3
= −(1/2)[−(3/2)x2 − 15x − 11/2](x + 1)−3/2 (x − 3)3
√
Therefore there are two inflection points, x = (−30 ± 768)/6, ≈ 9.6, .38.
g) y !! = 12(3x2 − 8x + 36). Therefore there are no inflection points. The quadratic
equation has no real roots.
2 √
h) y !! = (−2 + 4x2 )e−x . Therefore there are two inflection points at x = ±1/ 2.
2
i) One inflection point at x = 0. (y !! = −2xe−x )
 S. SOLUTIONS TO 18.01 EXERCISES

2B-3 a) y ! = 3x2 +2ax+b. The roots of the quadratic polynomial are distinct real numbers
if the discriminant is positive. (The discriminant is defined as the number under the square
root in the quadratic formula.) Therefore there are distinct real roots if and only if

(2a)2 − 4(3)b > 0, or a2 − 3b > 0.

From the picture, since y → ∞ as x → ∞ and y → −∞ as

x → −∞, the larger root of 3x2 + 2ax + b = 0 (with the Since y <<-1 when x << -1
plus sign in the quadratic formula) must be the local min, and y >> 1 when x >> 1, the
and the smaller root must be the local max. local max. is to the left of the local min.

b) y !! = 6x + 2a, so the inflection point is at −a/3. Therefore the condition y ! < 0 at

the inflection point is

y ! (−a/3) = 3(−a/3)2 + 2a(−a/3) + b = −a2 /3 + b < 0,

which is the same as

a2 − 3b > 0.

If y ! < 0 at some point x0 , then the function is decreasing at that point. But y → ∞ as
x → ∞, so there must be a local minimum at a point x > x0 . Similarly, since y → −∞ as
x → −∞, there must be a local maximum at a point x < x0 .
Comment: We evaluate y ! at the inflection point of y (x = −a/3) since we are trying
to decide (cf. part (b)) whether y ! is ever negative. To do this, we find the minimum of y !
(which occurs where y !! = 0).

Max is at x = 5 or x = 10;
4 5 7 8 10
Min is at x = 0 or x = 8.
2B!4

Graph of function

!7 !3 !1 5

Graph of derivative; note that

local maximum point above corresponds
to zero below;
point of inflection above corresponds to
local minimum below.

2B!5
 2. APPLICATIONS OF DIFFERENTIATION

2B-6 a) Try y ! = (x + 1)(x − 1) = x2 − 1. Then y = x3 /3 − x + c. The constant c won’t

matter so set c = 0. It’s also more convenient to multiply by 3:

y = x3 − 3x

b) This is an odd function with local min and max: y(1) = −2 and y(−1) = 2. The
endpoints values are y(3) = 18 and y(−3) = −18. It is very steep: y ! (3) = 8 .
c)

-3 -1 1 3

∆y
2B-7 a) f ! (a) = lim
∆x→0 ∆x
∆y > 0 ⇒ ∆x > 0
#
∆y
If y increasing then . So in both cases > 0.
∆y < 0 ⇒ ∆x < 0 ∆x
∆y
Therefore, lim ≥ 0.
∆x→0 ∆x
∆y ∆x
b) Proof breaks down at the last step. Namely, > 0 doesn’t imply lim >0
∆x ∆x→0 ∆y

[Limits don’t preserve strict inequalities, only weak ones. For example, u2 > 0 for u %= 0,
but lim u2 = 0 ≥ 0, not > 0.]
u→0

Counterexample: f (x) = x3 is increasing for all x, but f ! (0) = 0.

c) Use f (a) ≥ f (x) to show that lim ∆y/∆x ≤ 0 and lim ∆y/∆x ≥ 0. Since
∆x→0+ ∆x→0−
the left and right limits are equal, the derivative must be zero.

2C. Max-min problems

x
2C-1 The base of the box has sidelength 12 − 2x and the
height is x, so the volume is V = x(12 − 2x)2 . 12-2x
At the endpoints x = 0 and x = 6, the volume is 0, so
the maximum must occur in between at a critical point. x

V ! = (12 − 2x)2 + x(2)(12 − 2x)(−2) = (12 − 2x)(12 − 2x − 4x) = (12 − 2x)(12 − 6x).
It follows that V ! = 0 when x = 6 or x = 2. At the endpoints x = 0 and x = 6 the volume
is 0, so the maximum occurs when x = 2.
y

2C-2 We want to minimize the fence length L = 2x + y, where x x

the variables x and y are related by xy = A = 20, 000.
Choosing x as the independent variable, we have y = A/x, so that L = 2x + A/x. At the
endpoints x = 0 and x = ∞ (it’s a long barn), we get L = ∞, so the minimumof L must
occur at a critical point.

A A
L! = 2 − ; L = 0 =⇒ x2 = = 10, 000 =⇒ x = 100 feet
x2 2
 S. SOLUTIONS TO 18.01 EXERCISES

2C-3 We have y = (a − x)/2, so xy = x(a − x)/2. At the endpoints x = 0 and x = a, the

product xy is zero (and beyond it is negative). Therefore, the maximum occurs at a critical
point. Taking the derivative,

d x(a − x) a − 2x
= ; this is 0 when x = a/2.
dx 2 2

2C-4 If the length is y and the cross-section is a square with sidelength x, then 4x+y = 108.
Therxefore the volume is V = x2 y = 108x2 − 4x3 . Find the critical points:

(108x2 − 4x3 )! = 216x − 12x2 = 0 =⇒ x = 18 or x = 0.

The critical point x = 18 (3/2 ft.) corresponds to the length y = 36 (3 ft.), giving therefore
a volume of (3/2)2 (3) = 27/4 = 6.75 cubic feet.
The endpoints are x = 0, which gives zero volume, and when x = y, i.e., x = 9/5 feet,
which gives a volume of (9/5)3 cubic feet, which is less than 6 cubic feet. So the critical
point gives the maximum volume.

2C-5 We let r = radius of bottom and h = height, then the volume is h

V = πr2 h, and the area is A = πr2 + 2πrh.
Using r as the independent variable, we have using the above formulas, r

A − πr2
$ %
A π dV A 3π 2
h= , V = πr2 h = r − r3 ; = − r .
2πr 2 2 dr 2 2
A − πr2
Therefore, dV /dr = 0 implies A = 3πr2 , from which h = = r.
2πr
Checking the endpoints, at one h = 0 and V = 0; at the other, lim V = 0 (using the
r→0+
expression above for V in terms of r); thus the critical point must occur at a maximum.
(Another way to do this problem is to use implicit differentiation with respect to r.
Briefly, since A is fixed, dA/dr = 0, and therefore

dA r+h
= 2πr + 2πh + 2πrh! = 0 =⇒ h! = − ;
dr r
dV
= 2πrh + πr2 h! = 2πrh − πr(r + h) = πr(h − r).
dr

It follows that V ! = 0 when r = h or r = 0, and the latter is a rejected endpoint.

2C-6 To get max and min of y = x(x + 1)(x − 1) = x3 − x, first find the critical points:
1
y ! = 3x2 − 1 = 0 if x = ± √ ;
3
1 1 2 −2 1 1 2 2
y( √ ) = √ (− ) = √ , rel. min. y(− √ ) = − √ (− ) = √ , rel. max.
3 3 3 3 3 3 3 3 3 3
Check endpoints: y(2) = 6 ⇒ 2 is absolute max.; y(−2) = −6 ⇒ −2 is absolute min.
(This is an endpoint problem. The endpoints should be tested unless the physical or
geometric picture already makes clear whether the max or min occurs at an endpoint.)
 2. APPLICATIONS OF DIFFERENTIATION

2C-7 Let r be the radius, which√ is fixed. Then the height a of the rectangle is in the
interval 0 ≤ a ≤ r. Since b = 2 r2 − a2 , the area A is given in terms of a by
"
A = 2a r2 − a2 .
r a
The value of A at the endpoints a = 0 and a = r is zero, so the
maximum occurs at a critical point in between. b

dA " 2a2 2(r2 − a2 ) − 2a2

= 2 r 2 − a2 − √ = √
da r 2 − a2 r 2 − a2
r √
Thus dA/da = 0 implies 2r2 = a2 , from which we get a = √ , b = r 2.
2
(We use the positive square root since a ≥ 0. Note that b = 2a and A = r2 .)

2C-8 a) Letting a and b be the two legs and x and y the sides of the rectangle, we have
y = −(b/a)(x − a) and the area A = xy = (b/a)x(a − x). The area is zero at the two ends
x = 0 and x = a, so the maximum occurs in between at a critical point:

A! = (b/a)((a − x) − x); = 0 if x = a/2.

Thus y = (b/a)(a − x) = b/2 and A = ab/4. b

y
b) This time let x be the point shown on the accompanying figure; x
using similar triangles, the sides of the rectangle are a

x" 2 b l
$1 = a + b2 and $2 = √ (a − x) 1
a b
a + b2
2
l2
Therefore the area is x a-x

A = $1 $2 = (b/a)x(a − x)
This is the same formula for area as in part (a), so the largest area is the same, occurring
when x = a/2, and the two maximal rectangles both have the same area; they have different
dimensions though, since in the present case, one side length is half the hypotenuse:
" "
$1 = a2 + b2 /2 and $2 = ab/2 a2 + b2 .

2C-9 The distance is

L=
" "
x2 + 1 + (a − x)2 + b2
b
1
! !2
The endpoint values are x → ±∞, for L → ∞, so the minimum 1
value is at a critical point. x a-x
" √
! x a−x x (a − x)2 + b2 − (a − x) x2 + 1
L =√ −" = √ "
x2 + 1 (a − x)2 + b2 x2 + 1 (a − x)2 + b2
Thus L! = 0 implies (after squaring both sides),

x2 ((a − x)2 + b2 ) = (a − x)2 (x2 + 1), or x2 b2 = (a − x)2 or bx = (a − x);

 S. SOLUTIONS TO 18.01 EXERCISES

we used the positive square roots since both sides must be positive. Rewriting the above,

b 1
= , or tan θ1 = tan θ2 .
a−x x

Thus θ1 = θ2 : the angle of incidence equals the angle of reflection.

2C-10 The total time is

√ "
1002 + x2 1002 + (a − x)2
T = +
5 2

As x → ±∞, T → ∞, so the minimum value will be at a critical point.

x (a − x) sin α sin β
T! = √ − " = − .
2
5 100 + x2 2
2 100 + (a − x)2 5 2

Therefore, if T ! = 0 , it follows that

sin α sin β sin α 5

= or = .
5 2 sin β 2

2C-11 Use implicit differentiation:

x2 + y 2 = d2 =⇒ 2x + 2yy ! = 0 =⇒ y ! = −x/y. d
y
We want to maximize xy 3 . At the endpoints x = 0 and y = 0, the strength
x
is zero, so there is a maximum at a critical point. Differentiating,

0 = (xy 3 )! = y 3 + 3xy 2 y ! = y 3 + 3xy 2 (−x/y) = y 3 − 3x2 y

Dividing by x3 ,
√
(y/x)3 − 3(y/x) = 0 =⇒ (y/x)2 = 3 =⇒ y/x = 3.

2C-12 The intensity is proportional to

√ S
sin θ x/ 1 + x2
y= = = x(1 + x2 )−3/2
1 + x2 1 + x2

Endpoints: y(0) = 0 and y → 0 as x → ∞, so the maximum will !

be at a critical point. Critical points satisfy P 1

√
y ! = (1 − 2x2 )(1 + x2 )−5/2 = 0 =⇒ 1 − 2x2 = 0 =⇒ x = 1/ 2
√
The best height is 1/ 2 feet above the desk. (It’s not worth it. Use a desk lamp.)

2C-13 a) Let p denote the price in dollars. Then there will be 100 + (2/5)(200 − p)
passengers. Therefore the total revenue is

R = p(100 + (2/5)(200 − p) = p(180 − (2/5)p)

 2. APPLICATIONS OF DIFFERENTIATION

At the “ends” zero price p = 0, and no passengers p = (5/2)180 = 450, the revenue is zero.
So the maximum occurs in between at a critical point.

R! = (180 − (2/5)p) − (2/5)p = 180 − (4/5)p = 0 =⇒ p = (5/4)180 = $225

b)
P = xp − x(10 − x/105 ) with x = 105 (10 − p/2)

Therefore, the profit is cents is

P = 105 (10 − p/2)(p − 10 + (10 − p/2)) = 105 (10 − p/2)(p/2) = (105 /4)p(20 − p)

dP
= (105 /2)(10 − p)
dp

The critical point at p = 10. This is x = 105 (10 − 5) = 5 × 105 kilowatt hours, which is
within the range available to the utility company. The function P has second derivative
−105 /2, so it is concave down and the critical point must be the maximum. (This is one of
those cases where checking the second derivative is easier than checking the endpoints.)
Alternatively, the endpoint values are:

x = 0 =⇒ 105 (10 − p/2) = 0 =⇒ p = 20 =⇒ P = 0.

x = 8 × 105 =⇒ 8 × 105 = 105 (10 − p/2)

=⇒ 10 − p/2 = 8 =⇒ p = 4
=⇒ P = (105 /4)4(20 − 4) = 16 × 105 cents = $160, 000

The profit at the crit. pt. was (105 /4)10(20 − 10) = 2.5 × 106 cents = $250, 000
(1//e, 1/2e)
2C-14 a) Endpoints: y = −x2 ln(x) → 0 as x → 0+ and y → −∞ as x → ∞.
√ 1
Critical points: y ! = −2x ln x − x = 0 =⇒ ln x = −1/2 =⇒ x = 1/ e.
√ 14a
Critical value: y(1/ e) = 1/2e.
√
Maximum value: 1/2e, attained when x = 1/ e. (min is not attained)
b) Endpoints: y = −x ln(2x) → 0 as x → 0+ and y → −∞ as x → ∞.
(1/2e, 1/2e)
Critical points: y ! = − ln(2x) − 1 = 0 =⇒ x = 1/2e.
1/2
Critical value: y(1/2e) = −(1/2e) ln(1/e) = 1/2e.
Maximum value: 1/2e, attained at x = 1/2e. (min is not attained) 14b

2C-15 No minimum. The derivative is −xe−x < 0, so the function decreases. (Not needed
here, but it will follows from 2G-8 or from L’Hospital’s rule in E31 that xe−x → 0 as
x → ∞.)
 S. SOLUTIONS TO 18.01 EXERCISES

2D. More Max-min Problems

2D-3 The milk will be added at some time t1 , such that 0 ≤ t1 ≤ 10. In the interval
0 ≤ t < t1 the temperature is

y(t) = (100 − 20)e−(t−0)/10 + 20 = 80e−t/10 + 20

Therefore,
−t1 /10
T1 = y(t−
1 ) = 80e + 20
We are adding milk at a temperature T2 = 5, so the temperature as we start the second
interval of cooling is
9 1 1
T1 + T2 = 72e−t1 /10 + 18 +
10 10 2
Let Y (t) be the coffee temperature in the interval t1 ≤ t ≤ 10. We have just calculated
Y (t1 ), so

5Y (t) = (Y (t1 ) − 20)e−(t−t1 )/10 + 20 = (72e−t1 /10 − 1.5)e−(t−t1 )/10 + 20

The final temperature is

& '
T = Y (10) = (72e−t1 /10 − 1.5)e−(10−t1 )/10 + 20 = e−1 72 − (1.5)et1 /10 + 20

We want to maximize this temperature, so we look for critical points:

dT
= −(1.5/10e)et1/10 < 0
dt1

Therefore the function T (t1 ) is decreasing and its maximum occurs at the left endpoint:
t1 = 0.
Conclusion: The coffee will be hottest if you put the milk in as soon as possible.

2E. Related Rates

2E-1 The distance from robot to the point on the ground directly below the street lamp is
x = 20t. Therefore, x! = 20.

x+y y
= (similar triangles) 30
30 5
5
Therefore, x y

(x! + y ! )/30 = y ! /5 =⇒ y ! = 4 and (x + y)! = 24

The tip of the shadow is moving at 24 feet per second and the length of the shadow is
increasing at 4 feet per second.
 2. APPLICATIONS OF DIFFERENTIATION

2E-2
x
tan θ = and dθ/dt = 3(2π) = 6π
4
with t is measured in minutes and θ measured in radians. The light makes an angle of 60◦
with the shore when θ is 30◦ or θ = π/6. Differentiate with respect to t to get
!
(sec2 θ)(dθ/dt) = (1/4)(dx/dt) y
x
Since sec2 (π/6) = 4/3, we get dx/dt = 32π miles per minute. shoreline

2E-3 The distance is x = 10, y = 15, x! = 30 and y ! = 30. Therefore,

& '!
(x2 + y 2 )1/2 = (1/2)(2xx! + 2yy ! )(x2 + y 2 )−1/2
"
= (10(30) + 15(30))/ 102 + 152 /x 2+ y 2
√ y
= 150/ 13miles per hour
x
2E-4 V = (π/3)r2 h and 2r = d = (3/2)h implies h = (4/3)r.
Therefore,
V = (π/3)r2 h = (4π/9)r3
Moreover, dV /dt = 12, hence
$ %
dV dV dr d dr
= = (4π/9)r3 ) = (4π/3)r2 (12) = 16πr2 . h = (4/3)r
dt dr dt dr dt

dV
When h = 2, r = 3/2, so that = 36πm3 /minute. r r
dt
2E-5 The information is

x2 + 102 = z 2 , z! = 4

We want to evaluate x! at x = 20. (Derivatives are with z

respect to time.) Thus 10

2xx! = 2zz ! and z 2 = 202 + 102 = 500 x

Therefore, weight

√ √
x! = (zz ! )/x = 4 500/20 = 2 5 z 2

2E-6 x! = 50 and y ! = 400 and y

z 2 = x2 + y 2 + 22 x 2 2
x+y
The problem is to evaluate z ! when x = 50 and y = 400. Thus boat

2zz ! = 2xx! + 2yy ! =⇒ z ! = (xx! + yy ! )/z

√ √ √
and z = 502 + 4002 + 4 = 162504. So z ! = 162500/ 162504 ≈ 403mph.
(The fact that the plane is 2 miles up rather than at sea level changes the answer by only
about 4/1000. Even the boat speed only affects the answer by about 3 miles per hour.)
 S. SOLUTIONS TO 18.01 EXERCISES

2E-7 V = 4(h2 + h/2), V ! = 1. To evaluate h! at h = 1/2,

h 1/2 h
! ! ! ! ! !
1 = V = 8hh + 2h = 8(1/2)h + 2h = 6h h h

Therefore, Cross-sectional area = h 2+ h/2

h! = 1/6 meters per second

2E-8 x! = 60, y ! = 50 and x = 60 + 60t, y = 50t. " Noon is t = 0 and t
is measured in hours. To find the time when z = x2 + y 2 is smallest, we z
may as well minimize z 2 = x2 + y 2 . We know that there will be a minimum x
at a critical point because when t → ±∞ the distance tends to infinity.
Taking the derivative with respect to t, the critical points satisfy y

2xx! + 2yy ! = 0
This equation says

2((60 + 60t)60 + (50t)50) = 0 =⇒ (602 + 502 )t = −602

Hence
t = −36/61 ≈ −35min
The ships were closest at around 11 : 25 am.

2E-9 dy/dt = 2(x − 1)dx/dt. Notice that in the range x < 1, x − 1 is negative and so
√
(x − 1) = − y. Therefore,
√ √ √
dx/dt = (1/2(x − 1))(dy/dt) = −(1/2 y)(dy/dt) = +( y)(1 − y)/2 = 1/4 2

Method 2: Doing this directly turns out to be faster:

√
x=1− y =⇒ dx/dt = 1 − (1/2)y −1/2 dy/dt

and the rest is as before.

2E-10 r = Ct1/2 . The implicit assumption is that the volume of oil is constant:

πr2 T = V or r2 T = (V /π) = const

Therefore, differentiating with respect to time t,

(r2 T )! = 2rr! T + r2 T ! = 0 =⇒ T ! = −2r! /r

But r! = (1/2)Ct−1/2 , so that r! /r = 1/2t. Therefore

T ! = −1/t

(Although we only know the rate of change of r up to a constant of proportionality, we can

compute the absolute rate of change of T .)
 2. APPLICATIONS OF DIFFERENTIATION

2F. Locating zeros; Newton’s method

2F-1 a) y ! = − sin x − 1 ≤ 0. Also, y ! < 0 except at a discrete list of points (where

sin x = −1). Therefore y is strictly decreasing, that is, x1 < x2 =⇒ y(x1 ) < y(x2 ). Thus
y crosses zero only once.
5#/2
Slopes at the points
Upper and lower bounds for z such that y(z) = 0: are 0 or -2.
3#/2
y(0) = 1 and y(π/2) = −π/2. Therefore, 0 < z < π/2.
b) xn+1 = xn − (cos xn − xn )/(sin xn + 1) #/2 3#/2 5#/2
"5#/2 "3#/2 "#/2

x1 = 1, x2 = .750363868,
"3#/2
x3 = .739112891, x4 = .739085133
"5#/2
Accurate to three decimals at x3 , the second step. Answer .739.
c) Fixed point method takes 53 steps to stabilize at .739085133. Newton’s method
takes only three steps to get to 9 digits of accuracy. (See x4 .)

2F-2
1
y = 2x − 4 + −∞<x<∞
(x − 1)2
2 2((x − 1)3 − 1)
y! = 2 − =
(x − 1)3 (x − 1)3
y ! = 0 implies (x − 1)3 = 1, which implies x − 1 = 1 and hence that x = 2. The sign changes
of y ! are at the critical point x = 2 and at the singularity x = 1. For x < 1, the numerator
and denominator are negative, so y ! > 0. For 1 < x < 2, the numerator is still negative,
but the denominator is positive, so y ! < 0. For 2 < x, both numerator and denominator are
positive, so y ! > 0.
6
y !! =
(x − 1)4
(2,1)
Therefore, y !! > 0 for x %= 1.
1
Critical value: y(2) = 1
Singular values: y(1− ) = y(1+ ) = ∞ -3

Endpoint values: y → ∞ as x → ∞ and y → −∞ as x → −∞.

Conclusion: The function increases from −∞ to ∞ on the interval (−∞, 1). Therefore,
the function vanishes exactly once in this interval. The function decreases from ∞ to 1 on
the interval (1, 2) and increases from 1 to ∞ on the interval (2, ∞). Therefore, the function
does not vanish at all in the interval (1, ∞). Finally, the function is concave up on the
intervals (−∞, 1) and (1, ∞)
 S. SOLUTIONS TO 18.01 EXERCISES

2F-3
2x3 − 1
y ! = 2x − x−2 =
x2
Therefore y ! = 0 implies x3 = 1/2 or x = 2−1/3 . Moreover, y ! > 0 when x > 2−1/3 , and
y ! < 0 when x < 2−1/3 and x %= 0. The sign does not change across the singular point x = 0
because the power in the denominator is even. (continued →)

2(x3 + 1)
y !! = 2 + 2x−3 =
x3
Therefore y !! = 0 implies x3 = −1, or x = −1. Keeping track of the sign change in the
denominator as well as the numerator we have that y !! > 0 when x > 0 and y !! < 0 when
−1 < x < 0. Finally, y !! > 0 when x < −1, and both numerator and denominator are
negative.

Critical value: y(2−1/3 ) = 2−2/3 + 21/3 ≈ 1.9 P

Singular value: y(0+ ) = +∞ and y(0− ) = −∞ P is the critical point

Endpoint values: y → ∞ as x → ±∞
Conclusions: The function decreases from ∞ to −∞ in the interval (−∞, 0). Therefore
it vanishes exactly once in this interval. It jumps to ∞ at 0 and decreases from ∞ to
2−2/3 + 21/3 in the interval (0, 2−1/3 ). Finally it increases from 2−2/3 + 21/3 to ∞ in the
interval (2−1/3 , ∞). Thus it does not vanish on the interval (0, ∞). The function is concave
up in the intervals (−∞, −1) and (0, ∞) and concave down in the interval (−1, 0), with an
inflection point at −1.

2F-4 From the graph, x5 − x − c = 0 has three roots for any small value of c. The value of
c gets too large if it exceeds the local maximum of x5 − x labelled. To calculate that local
maximum, consider y ! = 5x4 − 1 = 0, with solutions x = ±5−1/4 . The local maximum is at
x = −5−1/4 and the value is
P
(−5−1/4 )5 − (−5−1/4 ) = 5−1/4 − 5−5/4 ≈ .535 y=c

Since .535 > 1/2, there are three roots.

√
2F-5 a) Answer: x1 = ±1/ 3. f (x) = x − x3 , so f ! (x) = 1 − 3x2 and

xn+1 = xn − f (xn )/f ! (xn )

√
So x2 is undefined if f ! (x1 ) = 0, that is x1 = ±1/ 3.
√
b) Answer: x1 = ±1/ 5. (This value can be found by experimentation. It can be
also be found by iterating the inverse of the Newton method function.)
Here is an explanation: Using the fact that f is odd and that x3 = x1 suggests that
x2 = −x1 . This greatly simplifies the equation.

−2x3n
xn+1 = xn − (xn − x3n )/(1 − 3x2n ) =
1 − 3x2n
 2. APPLICATIONS OF DIFFERENTIATION

Therefore we want to find x satisfying

−2x3
−x =
1 − 3x2

This equation is√the same as x(1 − 3x2 ) = 2x3 , which√implies x = 0 or 5x2

√ = 1. In other
√
words, x = ±1/ 5. Now one can check that if x1 = 1/ 5, then x2 = −1/ 5, x3 = +1/ 5,
etc.
√ √
√ c) Answers: √If x1 < −1/ 3, then xn → −1. If x1 > 1/ 3, then xn → 1. If
−1/ 5 < x1 < 1/ 5, then xn → 0. This can be found experimentally, numerically. For a
complete analysis and proof one needs the methods of an upper level course like 18.100.

2F-6 a) To simplify this problem to its essence, let V = π. (We are looking for ratio r/h
and this will be the same no matter what value we pick for V .) Thus r2 h = 1 and

A = πr2 + 2π/r
Minimize B = A/π instead.
B = r2 + 2r−1 =⇒ B ! = 2r − 2r−2
and B ! = 0 implies r = 1. Endpoints: B → ∞ as r → 0 and as r → ∞, so we have found
the minimum at r = 1. (The constraint r2 h = 1 shows that this minimum is achieved when
r = h = 1. As a doublecheck, the fact that the minimum area is achieved for r/h = 1 follows
from 2C/5; see part (b).)
The minimum of B is 3 attained at r = 1. Ten percent more than the
minimum is 3.3, so we need to find all r such that
B B=3.3
B(r) ≤ 3.3 (1,3)

Use Newton’s method with F (r) = B(r) − 3.3. (It is unwise to start New-
ton’s method at r = 1. Why?) The roots of F are approximately r = 1.35 allowable
r
and r = .72. r-interval

Since r2 h = 1, h = 1/r2 and the ratio,

r/h = r3

Compute (1.35)3 ≈ 2.5 and (.72)3 ≈ .37. Therefore, the proportions with at most 10 percent
extra glass are approximately
.37 < r/h < 2.5

b) The connection with Problem 2C-5 is that the minimum area r = h is not entirely
obvious, and not just because we are dealing with glass beakers instead of tin cans. In 2C/5
the area is fixed whereas here the volume is held fixed. But because one needs a larger surface
area to hold a larger volume, maximizing volume with fixed area is the same problem as
minimizing surface area with fixed volume. This is an important so-called duality principle
often used in optimization problems. In Problem 2C-5 the answer was r = h, which is the
proportion with minimum surface area as confirmed in part (a).
 S. SOLUTIONS TO 18.01 EXERCISES

2F-7 Minimize the distance squared, x2 + y 2 . The critical points satisfy

2x + 2yy ! = 0

The constraint y = cos x implies y ! = − sin x. Therefore,

0 = x + yy ! = x − cos x sin x

There is one obvious solution x = 0. The reason why this problem is in this section is that
one needs the tools of inequalities to make sure that there are no other solutions. Indeed, if

f (x) = x − cos x sin x, then f ! (x) = 1 − cos2 x + sin2 x = 2 sin2 x ≥ 0

Furthermore, f ! (x)0 is strictly positive except at the points x = kπ, so f is increasing and
crosses zero exactly once.
There is only one critical point and the distance tends to infinity at the endpoints x →
±∞, so this point is the minimum. The point on the graph closest to the origin is (0, 1).
y = cos x

Alternative method: To show that (0, 1) is closest it suffices to show that for −1 ≤ x < 0
and 0 < x ≤ 1, "
1 − x2 < cos x
Squaring gives 1 − x2 < cos2 x. This can be proved using the principles of problems 6 and
7. The derivative of cos2 x − (1 − x2) is twice the function f above, so the methods are very
similar.
 2. APPLICATIONS OF DIFFERENTIATION

2G. Mean-value Theorem

1
2G-1 a) slope chord = 1; f ! (x) = 2x ⇒ f ! (c) = 1 if c = .
2
(2,6)
1 1
b) slope chord = ln 2; f ! (x) = ⇒ f ! (c) = ln 2 if c = .
x ln 2
f (2) − f (−2) 6 − (−6)
c) for x3 − x: slope chord = = = 3;
2 − (−2) 4 -c c
! 2 ! 2
f (x) = 3x − 1 ⇒ f (c) = 3c − 1 = 3 ⇒ c = ± √23
From the graph, it is clear you should get two values for c. (The
axes are not drawn to the same scale.) (-2,-6)

2G-2 a) f (x) = f (a) + f ! (c)(x − a) ; Take a = 0; f (ξ) = sin x, f ! (x) = cos x

⇒ f (x) = 0 + cos c · x ⇒ sin x < x (since cos c < 1 for 0 < c < 2π)
Thus the inequality is valid for 0 < x ≤ 2π; since the function is periodic, it is also valid
for all x > 0.
d √ 1 √ 1 1
b) 1+x= √ ⇒ 1+x = 1+ √ x < 1 + x, since c > 0.
dx 2 1+x 2 1+c 2
121
2G-3. Let s(t) = distance; then average velocity = slope of chord = = 66.
11/6
Therefore, by MVT, there is some time t = c such that s! (c) = 66 > 65.
(An application of the mean-value theorem to traffic enforcement...)

2G-4 According to Rolle’s Theorem (Thm.1 p.800 : an important special case of the M.V.T,
and a step in its proof), between two roots of p(x) lies at least one root of p! (x). Therefore,
between the n roots a1 , . . . , an of p(x), lie at least n − 1 roots of p! (x).
There are no more than n − 1 roots, since degree of p! (x) = n − 1; thus p! (x) has exactly
n − 1 roots.

2G-5 Assume f (x) = 0 at a, b, c.

By Rolle’s theorem (as in MVT-4), there are two points q1 , q2 where f ! (q1 ) = 0, f ! (q2 ) = 0.
By Rolle’s theorem again, applied to q1 and q2 and f ! (x), there is a point p where
!!
f (p) = 0. Since p is between q1 and q2 , it is also between a and c.

2G-6 a) Given two points xi such that a ≤ x1 < x2 ≤ b, we have

f (x2 ) = f (x1 ) + f ! (c)(x2 − x1 ), where x1 < c < x2 .

Since f ! (x) > 0 on [a, b], f ! (c) > 0; also x2 − x1 > 0. Therefore f (x2 ) > f (x1 ), which
shows f (x) is increasing .
b) We have f (x) = f (a) + f ! (c)(x − a) where a < c < x.
Since f ! (c) = 0, f (x) = f (a) for a ≤ x ≤ b, which shows f (x) is constant on [a, b].
 Unit 3. Integration
3A. Differentials, indefinite integration

3A-1 a) 7x6 dx. (d(sin 1) = 0 because sin 1 is a constant.)

b) (1/2)x−1/2 dx
c) (10x9 − 8)dx
d) (3e3x sin x + e3x cos x)dx
√ √
e) (1/2 x)dx + (1/2 y)dy = 0 implies
√ √ √
y
! "
1/2 xdx 1− x 1
dy = − √ = − √ dx = − √ dx = 1 − √ dx
1/2 y x x x

3A-2 a) (2/5)x5 + x3 + x2 /2 + 8x + c
b) (2/3)x3/2 + 2x1/2 + c
c) Method 1 (slow way) Substitute: u = 8 + 9x, du = 9dx. Therefore
√
# #
8 + 9xdx = u1/2 (1/9)du = (1/9)(2/3)u3/2 + c = (2/27)(8 + 9x)3/2 + c

Method 2 (guess and check): It’s often faster to guess the form of the antiderivative and
work out the constant factor afterwards:
d 27
Guess (8 + 9x)3/2 ; (8 + 9x)3/2 = (3/2)(9)(8 + 9x)1/2 = (8 + 9x)1/2 .
dx 2
2
So multiply the guess by to make the derivative come out right; the answer is then
27
2
(8 + 9x)3/2 + c
27

d) Method 1 (slow way) Use the substitution: u = 1 − 12x4 , du = −48x3 dx.

1 1
# #
x3 (1 − 12x4 )1/8 dx = u1/8 (−1/48)du = − (8/9)u9/8 + c = − (1 − 12x4 )9/8 + c
48 54

Method 2 (guess and check): guess (1 − 12x4 )9/8 ;

d 9
(1 − 12x4 )9/8 = (−48x3 )(1 − 12x4 )1/8 = −54(1 − 12x4 )1/8 .
dx 8
1
So multiply the guess by − to make the derivative come out right, getting the previous
54
answer.
e) Method 1 (slow way): Use substitution: u = 8 − 2x2 , du = −4xdx.

x 12 1
# #
√ dx = u1/2 (−1/4)du = − u3/2 + c = − (8 − 2x2 )3/2 + c
8 − 2x2 4 3 6
 S. SOLUTIONS TO 18.01 EXERCISES

Method 2 (guess and check): guess (8 − 2x2 )3/2 ; differentiating it:

d 3
(8 − 2x2 )3/2 = (−4x2 )(8 − 2x2 )1/2 = −6(8 − 2x2 )1/2 ;
dx 2
1
so multiply the guess by − to make the derivative come out right.
6
The next four questions you should try to do (by Method 2) in your head. Write down
the correct form of the solution and correct the factor in front.
f) (1/7)e7x + c
5
g) (7/5)ex + c
√
x
h) 2e +c
i) (1/3) ln(3x + 2) + c. For comparison, let’s see how much slower substitution is:

u = 3x + 2, du = 3dx, so

dx (1/3)du
# #
= = (1/3) ln u + c = (1/3) ln(3x + 2) + c
3x + 2 u

j) # ! "
x+5 5
#
dx = 1+ dx = x + 5 ln x + c
x x

k) # ! "
x 5
#
dx = 1− dx = x − 5 ln(x + 5) + c
x+5 x+5
In Unit 5 this sort of algebraic trick will be explained in detail as part of a general method.
What underlies the algebra in both (j) and (k) is the algorithm of long division for polyno-
mials.
l) u = ln x, du = dx/x, so

ln x
# #
dx = udu = (1/2)u2 + c = (1/2)(ln x)2 + c
x

m) u = ln x, du = dx/x.

dx du
# #
= = ln u + c = ln(ln x) + c
x ln x u

3A-3 a) −(1/5) cos(5x) + c

b) (1/2) sin2 x + c, coming from the substitution u = sin x or −(1/2) cos2 x + c, coming
from the substitution u = cos x. The two functions (1/2) sin2 x and −(1/2) cos2 x are not
the same. Nevertheless the two answers given are the same. Why? (See 1J-1(m).)
c) −(1/3) cos3 x + c
d) −(1/2)(sin x)−2 + c = −(1/2) csc2 x + c
 3. INTEGRATION

e) 5 tan(x/5) + c
f) (1/7) tan7 x + c.
g) u = sec x, du = sec x tan xdx,
# #
9
sec x tan xdx (sec x)8 sec x tan xdx = (1/9) sec9 x + c

3B. Definite Integrals

3B-1 a) 1 + 4 + 9 + 16 = 30 b) 2 + 4 + 8 + 16 + 32 + 64 = 126
c) −1 + 4 − 9 + 16 − 25 = −15 d) 1 + 1/2 + 1/3 + 1/4 = 25/12
6
$ n
$ n
$
3B-2 a) (−1)n+1 (2n + 1) b) 1/k 2 c) sin(kx/n)
n=1 k=1 k=1

3B-3 a) upper sum = right sum = (1/4)[(1/4)3 + (2/4)3 + (3/4)3 + (4/4)3 ] = 15/128
lower sum = left sum = (1/4)[03 + (1/4)3 + (2/4)3 + (3/4)3 ] = 7/128
b) left sum = (−1)2 + 02 + 12 + 22 = 6; right sum = 02 + 12 + 22 + 32 = 14;
upper sum = (−1)2 + 12 + 22 + 32 = 15; lower sum = 02 + 02 + 12 + 22 = 5.
c) left sum = (π/2)[sin 0 + sin(π/2) + sin(π) + sin(3π/2)] = (π/2)[0 + 1 + 0 − 1] = 0;
right sum = (π/2)[sin(π/2) + sin(π) + sin(3π/2) + sin(2π)] = (π/2)[1 + 0 − 1 + 0] = 0;
upper sum = (π/2)[sin(π/2) + sin(π/2) + sin(π) + sin(2π)] = (π/2)[1 + 1 + 0 + 0] = π;
lower sum = (π/2)[sin(0) + sin(π) + sin(3π/2) + sin(3π/2)] = (π/2)[0 + 0 − 1 − 1] = −π.

3B-4 Both x2 and x3 are increasing functions on 0 ≤ x ≤ b, so the upper sum is the right
sum and the lower sum is the left sum. The difference between the right and left Riemann
sums is

(b/n)[f (x1 + · · · + f (xn )] − (b/n)[f (x0 + · · · + f (xn−1 )] = (b/n)[f (xn ) − f (x0 )]

In both cases xn = b and x0 = 0, so the formula is

(b/n)(f (b) − f (0))

a) (b/n)(b2 − 0) = b3 /n. Yes, this tends to zero as n → ∞.

b) (b/n)(b3 − 0) = b4 /n. Yes, this tends to zero as n → ∞.

3B-5 The expression is the right Riemann sum for the integral
# 1
sin(bx)dx = −(1/b) cos(bx)|10 = (1 − cos b)/b
0

so this is the limit.

 S. SOLUTIONS TO 18.01 EXERCISES

3C. Fundamental theorem of calculus

3C-1 # 6 %6
(x − 2)−1/2 dx = 2(x − 2)1/2 % = 2[(4)1/2 − 11/2 ] = 2
%
3 3
%2
3C-2 a) (2/3)(1/3)(3x + 5)3/2 % = (2/9)(113/2 − 53/2 )
%
0

b) If n &= −1, then

%2
(1/(n + 1))(1/3)(3x + 5)n+1 %0 = (1/3(n + 1))((11n+1 − 5n+1 )

If n = −1, then the answer is (1/3) ln(11/5).

%π √
c) (1/2)(cos x)−2 %3π/4 = (1/2)[(−1)−2 − (−1/ 2)−2 ] = −1/2
%2
3C-3 a) (1/2) ln(x2 + 1)%1 = (1/2)[ln 5 − ln 2] = (1/2) ln(5/2)
%2b
b) (1/2) ln(x2 + b2 )%b = (1/2)[ln(5b2 ) − ln(2b2 )] = (1/2) ln(5/2)

3C-4 As b → ∞,
# b %b
x−10 dx = −(1/9)x−9 %1 = −(1/9)(b−9 − 1) → −(1/9)(0 − 1) = 1/9.
1

This integral is the area of the infinite region between the curve y = x−10 and the x-axis
for x > 0.
# π
3C-5 a) sin xdx = − cos x|π0 = −(cos π − cos 0) = 2
0
# π/a
π/a
b) sin(ax)dx = −(1/a) cos(ax)|0 = −(1/a)(cos π − cos 0) = 2/a
0

3C-6 a) x2 − 4 = 0 implies x = ±2. So the area is

2 2 %2
x3 8
# #
(x2 − 4)dx = 2 (x2 − 4)dx =
%
− 4x%% = − 4 · 2 = −16/3
−2 0 3 0 3

(We changed to the interval (0, 2) and doubled the integral because x2 − 4 is even.) Notice
that the integral gave the wrong answer! It’s negative. This is because the graph y = x2 − 4
is concave up and is below the x-axis in the interval −2 < x < 2. So the correct answer is
16/3.
√
b) Following part (a), x2 − a = 0 implies x = ± a. The area is
√
a
√
a %√a
x3 %% a3/2 ' 4 3/2
# #
2
(a − x )dx = 2ax − % = 2 a3/2 −
2
&
√
(a − x )dx = 2 = a
− a 0 3 0 3 3
 3. INTEGRATION

3D. Second fundamental theorem

3D-1 Differentiate both sides;

# x
d dt 1
left side L(x): L# (x) = √ = √ , by FT2;
dx 0 2
a +x 2 a + x2
2

d ( 1 + √a2x+x2 1
right side R(x): R# (x) = (ln(x + a2 + x2 ) − ln a) = √ = √
dx 2
x+ a +x 2 a + x2
2

Since L# (x) = R# (x), we have L(x) = R(x) + C for some constant C = L(x) − R(x). The
constant C may be evaluated by assigning a value to x; the most convenient choice is x = 0,
which gives
# 0 (
L(0) = = 0; R(0) = ln(0 + 0 + a2 ) − ln a = 0; therefore C = 0 and L(x) = R(x).
0
(
b) Put x = c; the equation becomes √ 0 = ln(c + c2 + a2 ); solve this for c by first
2 2
exponentiating both sides: 1 = c + c + a ; then subtract c and square both sides; after
2
some algebra one gets c = 21 (1 − a2 ). 1
y=1 -t 2
y 1 +t
-1 1 t

1 − t2 -1
3D-3 Sketch y = first, as shown at the right.
1 + t2
# x # x # x
3D-4 a) sin(t3 )dt, by the FT2. b) sin(t3 )dt + 2 c) sin(t3 )dt − 1
0 0 1

3D-5 This problem reviews the idea of the proof of the FT2. 2/2 f(t)

t
a) f (t) = √
1 + t4
# 1+∆x t
1 shaded area 1 1+!x
f (t)dt = ≈ height .
∆x 1 width
# 1+∆x
1 shaded area 1
lim f (t)dt = lim = height = f (1) = √ .
∆x→0 ∆x 1 ∆x→0 width 2
b) By definition of derivative,
# 1+∆x
# F (1 + ∆x) − F (1) 1
F (1) = lim = lim f (t)dt;
∆x→0 ∆x ∆x→0 ∆x 1

1
by FT2, F # (1) = f (1) = √ .
2
# x # x
3D-6 a) If F1 (x) = dt and F2 (x) = dt, then F1 (x) = x − a1 and F2 (x) = x − a2 .
a1 a2
Thus F1 (x) − F2 (x) = a2 − a1 , a constant.
b) By the FT2, F1# (x) = f (x) and F2# (x) = f (x); therefore F1 = F2 + C, for some
constant C.
 S. SOLUTIONS TO 18.01 EXERCISES

3D-7 a) Using the FT2 and the chain rule, as in the Notes,
# x2 √
d √ d(x2 )
u sin udu = x2 sin(x2 ) · = 2x2 sin(x2 )
dx 0 dx
# sin x
1 dt
b) = ( · cos x = 1. (So = x)
2
1 − sin x 0 1 − t2
x2
d
#
c) tan udu = tan(x2 ) · 2x − tan x
dx x

3D-8 a) Differentiate both sides using FT2, and substitute x = π/2: f (π/2) = 4.
b) Substitute x = 2u and follow the method of part (a); put u = π, get finally
f (π/2) = 4 − 4π.

3E. Change of Variables; Estimating Integrals

1/a
1 dt 1 1
#
3E-1 L( ) = . Put t = , dt = − 2 du. Then
a 1 t u u

1/a a
dt u 1 dt du
# #
= − 2 du =⇒ L( ) = =− = −L(a)
t u a 1 t 1 u
√ √
3E-2 a) We want −t2 = −u2 /2, so u = t 2, du = 2dt.
√ # x/√2
x x/ 2 √
1 1 1
# #
−u2 /2 −t2 2
√ e du = √ e 2dt = √ e−t dt
2π 0 2π 0 π 0
√
1 √ 1 π 1
=⇒ E(x) = √ F (x/ 2) and lim E(x) = · =
π x→∞ π 2 2

b) The integrand is even, so

N N
1 2
# #
−u2 /2 2
√ e du = √ e−u /2
du = 2E(N ) −→ 1 as N → ∞
2π −N 2π 0

lim E(x) = −1/2 because E(x) is odd.

x→−∞

b
1
#
2
√ e−u /2
du = E(b) − E(a) by FT1 or by “interval addition” Notes PI (3).
2π a

Commentary: The answer is consistent with the limit,

N
1
#
2
√ e−u /2
du = E(N ) − E(−N ) = 2E(N ) −→ 1 as N → ∞
2π −N
 3. INTEGRATION

e
√ # 1
dx ln x √ 2 3/2 %%1 2
#
3E-3 a) Using u = ln x, du = , dx = udu = u % = .
x 1 x 0 3 0 3
b) Using u = cos x, du = − sin x,
π # −1 %−1
sin x −du 1 1 1 1 4
#
dx = = = ( 2 − 2) = ..
%
(2 + cos x)3 (2 + u)3 2
2(2 + u) 1
%
2 1 3 9
0 1
# 1 # π2
dx cos u %π/2 π
c) Using x = sin u, dx = cos udu, √ = du = u% = .
%
0 1 − x2 0 cos u 0 2

3E-4 Substitute x = t/a; then x = ±1 ⇒ t = ±a. We then have a

# 1( # a) # a(
π t2 dt 1
= 1 − x2 dx = 1− 2 = 2 a2 − t2 dt.
2 −1 −a a a a −a
Multiplying by a2 gives the value πa2 /2 for the integral, which checks, -a a
since the integral represents the area of the semicircle.

3E-5 One can use informal reasoning based on areas (as in Ex. 5, Notes FT), but it is
better to use change of variable.
a) Goal: F (−x) = −F (x). Let t = −u, dt = −du, then
# −x # x
F (−x) = f (t)dt = f (−u)(−du)
0 0
# x
Since f is even (f (−u) = f (u)), F (−x) = − f (u)du = −F (x).
0

b) Goal: F (−x) = F (x). Let t = −u, dt = −du, then

# −x # x
F (−x) = f (t)dt = f (−u)(−du)
0 0
# x
Since f is odd ((f (−u) = −f (u)), F (−x) = f (u)du = F (x).
0

1 1
3E-6 a) x3 < x on (0,1) ⇒ > on (0,1); therefore
1 + x3 1+x
1 # 1
dx dx %1
#
> = ln(1 + x)% = ln 2 = .69
%
0 1 + x3 0 1+x 0

b) 0 < sin x < 1 on (0, π) ⇒ sin2 x < sin x on (0, π); therefore
# π # π %π
sin2 xdx < sin xdx = − cos x% = −(−1 − 1) = 2.
%
0 0 0
# 20 ( # 20 √ x % 1 2 %20
c) 2
x + 1dx > x2 dx = % = (400 − 100) = 150
10 10 2 10 2
%# % %N
% N sin x % *N *N
| sin x| 1 1%
= − N1 + 1 < 1.
%
3E-7 % dx% ≤ 1 x2 dx ≤ x2 dx = −x%
% %
% 1 x 2 1
% 1
 S. SOLUTIONS TO 18.01 EXERCISES

3F. Differential Equations: Separation of Variables. Applications

3F-1 a) y = (1/10)(2x + 5)5 + c

# #
b) (y + 1)dy = dx =⇒ (y + 1)dy = dx =⇒ (1/2)(y + 1)2 = x + c. You can leave
√
this in implicit form or solve for y: y = −1 ± 2x + a for any constant a (a = 2c)
c) y 1/2 dy = 3dx =⇒ (2/3)y 3/2 = 3x + c =⇒ y = (9x/2 + a)2/3 , with a = (3/2)c.
d) y −2 dy = xdx =⇒ −y −1 = x2 /2 + c =⇒ y = −1/(x2 /2 + c)

3F-2 a) Answer: 3e16 .

y −1 dy = 4xdx =⇒ ln y = 2x2 + c
y(1) = 3 =⇒ ln 3 = 2 + c =⇒ c = ln 3 − 2.
Therefore
ln y = 2x2 + (ln 3 − 2)
At x = 3, y = e18+ln 3−2 = 3e16
√
b) Answer: y = 11/2 + 3 2.

(y + 1)−1/2 dy = dx =⇒ 2(y + 1)1/2 = x + c

√
y(0) = 1 =⇒ 2(1 + 1)1/2 = c =⇒ c = 2 2
At x = 3,
√ √ √
2(y + 1)1/2 = 3 + 2 2 =⇒ y + 1 = (3/2 + 2)2 = 13/2 + 3 2
√
Thus, y = 11/2 + 3 2.
(
c) Answer: y = 550/3

ydy = x2 dx =⇒ y 2 /2 = (1/3)x3 + c

y(0) = 10 =⇒ c = 102 /2 = 50
Therefore, at x = 5, (
y 2 /2 = (1/3)53 + 50 =⇒ y = 550/3

d) Answer: y = (2/3)(e24 − 1)

(3y + 2)−1 dy = dx =⇒ (1/3) ln(3y + 2) = x + c

y(0) = 0 =⇒ (1/3) ln 2 = c
Therefore, at x = 8,

(1/3) ln(3y + 2) = 8 + (1/3) ln 2 =⇒ ln(3y + 2) = 24 + ln 2 =⇒ (3y + 2) = 2e24

Therefore, y = (2e24 − 2)/3

 3. INTEGRATION

e) Answer: y = − ln 4 at x = 0. Defined for −∞ < x < 4.

e−y dy = dx =⇒ −e−y = x + c

y(3) = 0 =⇒ −e0 = 3 + c =⇒ c = −4
Therefore,
y = − ln(4 − x), y(0) = − ln 4
The solution y is defined only if x < 4.

3F-3 a) Answers: y(1/2) = 2, y(−1) = 1/2, y(1) is undefined.

y −2 dy = dx =⇒ −y −1 = x + c

y(0) = 1 =⇒ −1 = 0 + c =⇒ c = −1
Therefore, −1/y = x − 1 and
1
y=
1−x
The values are y(1/2) = 2, y(−1) = −1/2 and y is undefined at x = 1.
b) Although the formula for y makes sense at x = 3/2, (y(3/2) = 1/(1 − 3/2) = −2),
it is not consistent with the rate of change interpretation of the differential equation. The
function is defined, continuous and differentiable for −∞ < x < 1. But at x = 1, y and
dy/dx are undefined. Since y = 1/(1 − x) is the only solution to the differential equation
in the interval (0, 1) that satisfies the initial condition y(0) = 1, it is impossible to define a
function that has the initial condition y(0) = 1 and also satisfies the differential equation in
any longer interval containing x = 1.
To ask what happens to y after x = 1, say at x = 3/2, is something like asking what
happened to a rocket ship after it fell into a black hole. There is no obvious reason why
one has to choose the formula y = 1/(1 − x) after the “explosion.” For example, one could
define y = 1/(2 − x) for 1 ≤ x < 2. In fact, any formula y = 1/(c − x) for c ≥ 1 satisfies the
differential equation at every point x > 1.

3F-4 a) If the surrounding air is cooler (Te −T < 0), then the object will cool, so dT /dt < 0.
Thus k > 0.
b) Separate variables and integrate.

(T − Te )−1 dT = −kdt =⇒ ln |T − Te | = −kt + c

Exponentiating,
T − Te = ±ec e−kt = Ae−kt
The initial condition T (0) = T0 implies A = T0 − Te . Thus

T = Te + (T0 − Te )e−kt

c) Since k > 0, e−kt → 0 as t → ∞. Therefore,

T = Te + (T0 − Te )e−kt −→ Te as t → ∞
 S. SOLUTIONS TO 18.01 EXERCISES

d)
T − Te = (T0 − Te )e−kt
The data are T0 = 680, Te = 40 and T (8) = 200. Therefore,

200 − 40 = (680 − 40)e−8k =⇒ e−8k = 160/640 = 1/4 =⇒ −8k = − ln 4.

The number of hours t that it takes to cool to 50◦ satisfies the equation

50 − 40 = (640)e−kt =⇒ e−kt = 1/64 =⇒ −kt = −3 ln 4.

To solve the two equations on the right above simultaneously for t, it is easiest just to divide
the bottom equation by the top equation, which gives

t
= 3, t = 24.
8

e)
T − Te = (T0 − Te )e−kt
The data at t = 1 and t = 2 are

800 − Te = (1000 − Te )e−k and 700 − Te = (1000 − Te )e−2k

Eliminating e−k from these two equations gives

! "2
700 − Te 800 − Te
=
1000 − Te 1000 − Te
(800 − Te )2 = (1000 − Te )(700 − Te )
8002 − 1600Te + Te2 = (1000)(700) − 1700Te + Te2
100Te = (1000)(700) − 8002
Te = 7000 − 6400 = 600

f) To confirm the differential equation:

y # (t) = T # (t − t0 ) = k(Te − T (t − t0 )) = k(Te − y(t))

The formula for y is

y(t) = T (t − t0 ) = Te + (T0 − Te )e−k(t−t0 ) = a + (y(t0 ) − a)e−c(t−t0 )

with k = c, Te = a and T0 = T (0) = y(t0 ).

3F-6 y = cos3 u − 3 cos u, x = sin4 u

dy = (3 cos2 u · (− sin u) + 3 sin u)du, dx = 4 sin3 u cos udu
dy 3 sin u(1 − cos2 u) 3
= 3 =
dx 4 sin u cos u 4 cos u
 3. INTEGRATION

3F-7 a) y # = −xy; y(0) = 1

dy 1
= −xdx =⇒ ln y = − x2 + c
y 2
To find c, put x = 0, y = 1: ln 1 = 0 + c =⇒ c = 0.
1 2
=⇒ ln y = − x2 =⇒ y = e−x /2
2
b) cos x sin ydy = sin xdx; y(0) = 0
sin x
sin ydy = dx =⇒ − cos y = − ln(cos x) + c
cos x
Find c: put x = 0, y = 0: − cos 0 = − ln(cos 0) + c =⇒ c = −1
=⇒ cos y = ln(cos x) + 1 P

y
3F-8 a) From the triangle, y # = slope tangent =
1
1
dy
=⇒ = dx =⇒ ln y = x + c1 =⇒ y = ex+c1 = Aex (A = ec1 )
y
b) If P bisects tangent, then P0 bisects OQ (by euclidean geometry)
So P0 Q = x ( since OP0 = x). P

−y dy dx
Slope tangent = y # = =⇒ =−
x y x
=⇒ ln y = − ln x + c1
1 c1 c
Exponentiate: y = ·e = ,c > 0
x x
c
Ans: The hyperbolas y = , c > 0
x
3G. Numerical Integration

3G-1 Left Riemann sum: (∆x)(y0 + y1 + y2 + y3 )

Trapezoidal rule: (∆x)((1/2)y0 + y1 + y2 + y3 + (1/2)y4 )
Simpson’s rule: (∆x/3)(y0 + 4y1 + 2y2 + 4y3 + y4 )

a) ∆x = 1/4 and
√ √
y0 = 0, y1 = 1/2, y2 = 1/ 2, y3 = 3/2, y4 = 1.
√ √
Left Riemann sum: (1/4)(0 + 1/2 + 1/ 2 + 3/2) ≈ .518
√ √
Trapezoidal rule: (1/4)((1/2) · 0 + 1/2 + 1/ 2 + 3/2 + (1/2)1) ≈ .643
√ √
Simpson’s rule: (1/12)(1 · 0 + 4(1/2) + 2(1/ 2) + 4( 3/2) + 1) ≈ .657
as compared to the exact answer .6666 . . .

b) ∆x = π/4
√ √
y0 = 0, y1 = 1/ 2, y2 = 1, y3 = 1/ 2, y4 = 0.
 S. SOLUTIONS TO 18.01 EXERCISES
√ √
Left Riemann sum: (π/4)(0 + 1/ 2 + 1 + 1/ 2) ≈ 1.896
√ √
Trapezoidal rule: (π/4)((1/2) · 0 + 1/ 2 + 1 + 1/ 2 + (1/2) · 0) ≈ 1.896 (same as Riemann
sum)
√ √
Simpson’s rule: (π/12)(1 · 0 + 4(1/ 2) + 2(1) + 4(1/ 2) + 1 · 0) ≈ 2.005
as compared to the exact answer 2
c) ∆x = 1/4

y0 = 1, y1 = 16/17, y2 = 4/5, y3 = 16/25, y4 = 1/2.

Left Riemann sum: (1/4)(1 + 16/17 + 4/5 + 16/25) ≈ .845

Trapezoidal rule: (1/4)((1/2) · 1 + 16/17 + 4/5 + 16/25 + (1/2)(1/2)) ≈ .8128
Simpson’s rule: (1/12)(1 · 1 + 4(16/17) + 2(4/5) + 4(16/25) + 1(1/2)) ≈ .785392
as compared to the exact answer π/4 ≈ .785398
(Multiplying the Simpson’s rule answer by 4 gives a passable approximation to π, of
3.14157, accurate to about 2 × 10−5 .)
d) ∆x = 1/4

y0 = 1, y1 = 4/5, y2 = 2/3, y3 = 4/7, y4 = 1/2.

Left Riemann sum: (1/4)(1 + 4/5 + 2/3 + 4/7) ≈ .76

Trapezoidal rule: (1/4)((1/2) · 1 + 4/5 + 2/3 + 4/7(1/2)(1/2)) ≈ .697
Simpson’s rule: (1/12)(1 · 1 + 4(4/5) + 2(2/3) + 4(4/7) + 1(1/2)) ≈ .69325
Compared with the exact answer ln 2 ≈ .69315, Simpson’s rule is accurate to about 10−4 .
# b
b4
3G-2 We have x3 dx = . Using Simpson’s rule with two subintervals, ∆x = b/2, so
0 4
that we get the same answer as above:

b4
! "
b b 3 3
S(x ) = (0 + 4(b/2)3 + b3 ) =
3
b = .
6 6 2 4

Remark. The fact that Simpson’s rule is exact on cubic polynomials is very significant to
its effectiveness as a numerical approximation. It implies that the approximation converges
at a rate proportional to the the fourth derivative of the function times (∆x)4 , which is fast
enough for many practical purposes.

3G-3 The sum √ √ ( y= x

S= 1 + 2 + ... + 10, 000
# 104 1
√
is related to the trapezoidal estimate of xdx :
0

1 2 3 10,000
 3. INTEGRATION

104 √ 1√ √ 1√ 4 1√ 4
#
(1) xdx ≈ 0 + 1 + ... + 10 = S − 10
0 2 2 2

But %104
104 √ 2 3/2 %% 2
#
xdx = x % = · 106
0 3 % 3
0

From (1),

2
(2) · 106 ≈ S − 50
3
Hence

(3) S ≈ 666, 717

In (1), we have >, as in the picture. Hence in (2), we have >, so in (3), we have <, Too
high.
1
y= x
1
3G-4 As in Problem 3 above, let

1 1 1
S= + + ... +
1 2 n
Then by trapezoidal rule, 1 2 3 n-1 n

n
dx 1 1 1 1 1 1 1 1
#
≈ · + + + ... + · = S − −
1 x 2 2 2 3 2 n 2 2n
n
dx 1 1
#
Since = ln n, we have S ≈ ln n + + . (Estimate is too low.)
1 x 2 2n

3G-5 Referring to the two pictures above, one can see that if f (x) is concave down on
[a, b], the trapezoidal rule gives too low an estimate; if f (x) is concave up, the trapezoidal
rule gives too high an estimate..
 Unit 4. Applications of integration
4A. Areas between curves.
! 1 "1
4A-1 a) (3x − 1 − 2x2 )dx = (3/2)x2 − x − (2/3)x3 "1/2 = 1/24
1/2
3
b) x = ax =⇒ x = ±a or x = 0. There are two enclosed pieces (−a < x < 0 and
0 < x < a) with the same area by symmetry. Thus the total area is:
√
! a "√a
2 (ax − x3 )dx = ax2 − (1/2)x4 "0 = a2 /2
0

5/2
(1,2) a3/2
1 1

(1/2,1/2) a1/2 a1/2 1 2

a 3/2

1a 1b 1c 1d

c) x + 1/x = 5/2 =⇒ x2 + 1 = 5x/2 =⇒ x = 2 or 1/2. Therefore, the area is

! 2 "2
[5/2 − (x + 1/x)]dx = 5x/2 − x2 /2 − ln x"1/2 = 15/8 − 2 ln 2
1/2

! 1 "1
d) (y − y 2 )dy = y 2 /2 − y 3 /3"0 = 1/6
0 y = 1-x 2
4A-2 First way (dx):
-1 1
! 1 ! 1 "1
2 2 3
(1 − x )dx = 2 (1 − x )dx = 2x − 2x /3 0 = 4/3
"
−1 0

√
Second way (dy): (x = ± 1 − y) 1
x = - 1-y
x = 1-y
! 1 # "1
2 1 − ydy = (4/3)(1 − y)3/2 " = 4/3
"
4
0 0
(1,3)
2 x=- 4-y x= 4-y
4A-3 4 − x = 3x =⇒ x = 1 or − 4. First way (dx):
! 1 "1
(4 − x2 − 3x)dx = 4x − x3 /3 − 3x2 /2"−4 = 125/6
−4

Second way (dy): Lower section has area x= y

3
! 3 # "3
(y/3 + 4 − ydy = y 2 /6 − (2/3)(4 − y)3/2 " = 117/6 (-4,-12)
"
−12 −12
 S. SOLUTIONS TO 18.01 EXERCISES

Upper section has area

! 4 # "4
2 4 − ydy = −(4/3)(4 − y)3/2 " = 4/3
"
3 3

(See picture for limits of integration.) Note that 117/6 + 4/3 = 125/6.

4A-4 sin x = cos x =⇒ x = π/4 + kπ. So the area is 5!/4

!/4
! 5π/4
5π/4
√
(sin x − cos x)dx = (− cos x − sin x)|π/4 = 2 2
π/4

4B. Volumes by slicing; volumes of revolution

! 1 ! 1 ! 1
2 2 2
4B-1 a) πy dx = π(1 − x ) dx = 2π (1 − 2x2 + x4 )dx
−1 −1 0
"1
= 2π(x − 2x3 /3 + x5 /5)"0 = 16π/15
$a $a $a
b) −a πy 2 dx = −a π(a2 − x2 )2 dx = 2π 0 (a4 − 2a2 x2 + x4 )dx
"a
= 2π(a4 x − 2a2 x3 /3 + x5 /5)"0 = 16πa5 /15
! 1
c) πx2 dx = π/3
0
! a
d) πx2 dx = πa3 /3
0
! 2 ! 2 "2
2 2
e) π(2x − x ) dx = π(4x2 − 4x3 + x4 )dx = π(4x3 /3 − x4 + x5 /5)"0 = 16π/15
0 0

(Why (e) the same as (a)? Complete the square and translate.)

1b 1d 1f 1g 1h
b

0 a
!a "x a 0 a 0 a
2
y = a2! x y=x y = 2ax ! x2 y= ax y 2= b 2(1 ! x 2/a 2)
(for 1a, set a = 1) (for 1c, set a = 1) (for 1e, set a = 1)

$ 2a $ 2a
f) 0
π(2ax − x2 )2 dx = π(4a2 x2 − 4ax3 + x4 )dx
0
"2
= π(4a2 x3 /3 − ax4 + x5 /5)"0 = 16πa5 /15
(Why is (f) the same as (b)? Complete the square and translate.)
! a
g) axdx = πa3 /2
0
 4. APPLICATIONS OF INTEGRATION
! a ! a "a
h) πy 2 dx = πb2 (1 − x2 /a2 )dx = πb2 (x − x3 /3a2 )"0 = 2πb2 a/3
0 0
! 1 ! a2
4B-2 a) π(1 − y)dy = π/2 b) π(a2 − y)dy = πa4 /2
! 1 0 !0 a
c) π(1 − y 2 )dy = 2π/3 d) π(a2 − y 2 )dy = 2πa3 /3
0 0
√
e) x2 − 2x + y = 0 =⇒ x = 1 ± 1 − y. Using the method of washers:
! 1 # # ! 1 #
π[(1 + 1 − y)2 − (1 − 1 − y)2 ]dy = 4π 1 − ydy
0 0
"1
= −(8/3)π(1 − y)3/2 " = 8π/3
"
0

(In contrast with 1(e) and 1(a), rotation around the y-axis makes the solid in 2(e) different
from 2(a).)

a 2 2b 2d a,2fa2) (a,a)
2g 2h
a b
"y

0 a
a 0

x = a2 ! y x=y x = a + a 2! y x = y2/a x2 = a2(1 ! y2 /b2 )

(for 2a, set a = 1) (for 2c, set a = 1) (for 2e, set a = 1)

#
f) x2 − 2ax + y = 0 =⇒ x = a ± a2 − y. Using the method of washers:
! a2 # # ! a2 #
π[(a + a2 − y)2 − (a − a2 − y)2 ]dy = 4πa a2 − ydy
0 0
"1
= −(8/3)πa(a2 − y)3/2 " = 8πa4 /3
"
0
! a "a
g) Using washers: π(a2 − (y 2 /a)2 )dy = π(a2 y − y 5 /5a2 )"0 = 4πa3 /5.
! b 0
! b "b
2
h) πx dy = 2π a2 (1 − y 2 /b2 )dy = 2π(a2 y − a2 y 3 /3b2)"0 = 4πa2 b/3 (The answer in
−b 0
2(h) is double the answer in 1(h), with a and b reversed. Can you see why?)
L M
4B-3 Put the pyramid upside-down. By similar triangles, the base of the
smaller bottom pyramid has sides of length (z/h)L and (z/h)M .
The base of the big pyramid has area b = LM ; the base of the smaller h
z
pyramid forms a cross-sectional slice, and has area

(z/h)L · (z/h)M = (z/h)2 LM = (z/h)2 b

Therefore, the volume is
! h "h
(z/h)2 bdz = bz 3 /3h2 "0 = bh/3
0
 x
z 2x
x
S. SOLUTIONS TO 18.01 EXERCISES 1
x
4B-4 The slice perpendicular to the xz-plane are right triangles 1 x
side view of top view of side view of
with base of length x and height z = 2x. Therefore the area of a wedge along wedge along slice along
slice is x2 . The volume is y-axis z-axis y-axis

! 1 ! 1
x2 dy = (1 − y 2 )dy = 4/3
−1 −1 y = 3x
√
4B-5 One side can be described by y = 3x for 0 ≤ x ≤ a/2.
Therefore, the volume is
! a/2 ! a/2 √
2 a/2
2 πy dx2 π( 3x)2 dx = πa3 /4
0 0
a2-x 2
4B-6 If the hypotenuse of an isoceles right triangle has length h,
2
then its area
√ is h /4. The endpoints
√ of the slice in the xy-plane a
2 2
are y = ± a − x2 , so h = 2 a2 − x2 . In all the volume is
2 x 2 a -x

! a ! a
(h2 /4)dx = (a2 − x2 )dx = 4a3 /3 top view slice
−a −a

4B-7 Solving for x in y = (x − 1)2 and y = (x + 1)2 gives the values x = - 1+ y x =1 - y

1
√ √
x=1± y and x = −1 ± y
-1 1
The hard part is deciding which sign of the square root representing
the endpoints of the square. 2 (x- y)
Method 1: The point (0, 1) has to be on the two curves. Plug in y = 1 and x = 0 to see
√ √
that the square root must have the opposite sign from 1: x = 1 − y and x = −1 + y.
√
Method 2: Look at the picture. x = 1 + y is the wrong choice because it is the right
√
half of the parabola with vertex (1, 0). We want the left half: x = 1 − y. Similarly, we
√
want x = −1 + y, the right half of the parabola with vertex (−1, 0). Hence, the side of
√ √ √
the square is the interval −1 + y ≤ x ≤ 1 − y, whose length is 2(1 − y), and the
1 1
√ 2 √
! !
Volume = (2(1 − y) dy = 4 (1 − 2 y + y)dy = 2/3 .
0 0

4C. Volumes by shells

4C-1 a)
! b+a ! b+a #
Shells: (2πx)(2y)dx = 4πx a2 − (x − b)2 dx
b−a b−a

#
b) (x − b)2 = a2 − y 2 =⇒ x = b ± a2 − y 2
! a ! a # #
2 2
Washers: π(x2 − x1 )dy = π((b + a2 − y 2 )2 − (b − a2 − y 2 )2 )dy
−a −a
! a #
=π 4b a2 − y 2 dy
−a
 4. APPLICATIONS OF INTEGRATION

y = a2 - (x - b) 2 x = b - a 2- y 2
a a
x = b + a2 - y 2

b-a b+a b-a b+a

-a -a
2
y = - a 2- (x - b}
Washers
Shells
! a #
c) a2 − y 2 dy = πa2 /2, because it’s the area of a semicircle of radius a.
−a

Thus (b) =⇒ Volume of torus = 2π 2 a2 b

d) z = x − b, dz = dx
! b+a # ! a # ! a #
2 2
4πx a − (x − b) dx = 2 2
4π(z + b) a − z dz = 4πb a2 − z 2 dz
b−a −a −a

because the part of the integrand with the factor z is odd, and so it integrates to 0.
! 1 ! 1
4C-2 2πxydx = 2πx3 dx = π/2
0 0

1 1

y = x2 y= x x = y2
4C!2 (shells) 4C!3a (shells) 4C!3b (discs)

! 1 ! 1 √
4C-3 Shells: 2πx(1 − y)dx = 2πx(1 − x)dx = π/5
0 0
! 1 ! 1
Disks: πx2 dy = πy 4 dy = π/5
0 0
! 1 ! 1 #
4C-4 a) 2πy(2x)dy = 4π y 1 − ydy
0 0
! a2 ! a2 #
b) 2πy(2x)dy = 4π y a2 − ydy
0 0
! 1
c) 2πy(1 − y)dy
0
! a
d) 2πy(a − y)dy
0
 S. SOLUTIONS TO 18.01 EXERCISES

√
e) x2 − 2x + y = 0 =⇒ x = 1 ± 1 − y.
# # #
The interval 1 − 1−y ≤x≤1+ 1−y has length 2 1 − y
! 1 # ! 1 #
=⇒ V = 2πy(2 1 − y)dy = 4π y 1 − ydy
0 0

#
f) x2 − 2ax + y = 0 =⇒ x = a ± a2 − y.
# # #
The interval a − a2 − y ≤ x ≤ a + a2 − y has length 2 a2 − y

! a2 # ! a2 #
=⇒ V = 2
2πy(2 a − ydy = 4π y a2 − ydy
0 0

4b 4d 4f 4g 4h
2
a2 (a, a )
(a,a)
a b

0 a
a 0

x = a + a 2- y 2 2 2
x = a 2 - y (right) x=y x = y /a x = a 1 - y /b
x = - a 2- y (left) x = a - a2 - y

! a
g) 2πy(a − y 2 /a)dy
0
! b ! b
h) 2πyxdy = 2πy(a2 (1 − y 2 /b2 )dy
0 0
(Why is the lower limit of integration 0 rather than −b?)
! 1 ! 1 ! 1
4C-5 a) 2πx(1 − x2 )dx c) 2πxydx = 2πx2 dx
0 0 0
! a ! a ! a
b) 2πx(a2 − x2 )dx d) 2πxydx = 2πx2 dx
0 0 0
! 2 ! 2
e) 2πxydx = 2πx(2x − x2 )dx
0 0

5b 5d 5f 5g b 5h

-a "x a a 0 a b
0
2
y=a-x
2 2 y=x y = 2ax - x2 y = ax y = b 1 - x 2/a
(for 5a, set a = 1 ) for 5c, set a = 1 ) (for 5e, set a =1 ) y = -b 1 - x 2/a 2
 4. APPLICATIONS OF INTEGRATION

! 2a ! 2a ! a ! a √
2
f) 2πxydx = 2πx(ax − x )dx g) 2πxydx = 2πx axdx
0 0 0 0
! a ! a
h) 2πx(2y)dx = 2πx(2b2 (1 − x2 /a2 ))dx
0 0 y= b2- x 2
(Why did y get doubled this time?)
Shells
4C-6 a b

! b ! b #
2πx(2y)dx = 2πx(2 b2 − x2 )dx 2 2
a a
y=- b - x
"b base of removed
= −(4/3)π(b2 − x2 )3/2 " = (4π/3)(b2 − a2 )3/2 cylinder
"
a

4D. Average value

4D-1 Cross-sectional area at x is = πy 2 = π · (x2 )2 = πx4 . Therefore,

"2
1 2 4 πx5 "" 16π
!
average cross-sectional area = πx dx = = .
2 0 10 "0 5
2a "2a % &
1 dx 1 1 1 2a ln 2
!
"
4D-2 Average = = ln x" = (ln 2a − ln a) =
" ln = .
a a x a a a a a a

4D-3 Let s(t) be the distance function; then the velocity is v(t) = s# (t)
! b
1 s(b) − s(a)
Average value of velocity = s# (t)dt = by FT1
b−a a b−a
= average velocity over time interval [a,b]

4D-4 By symmetry, we can restrict P to the upper semicircle.

P
By the law of cosines, we have |P Q|2 = 12 + 12 − 2 cos θ. Thus 1
2 1 π
!
1 π
#
average of |P Q| =
π 0
(2 − 2 cos θ)dθ =
π
[2θ − 2 sin θ]0 = 2 1 Q
(This is the value of |P Q|2 when θ = π/2, so the answer is reasonable.))

1 x
!
4D-5 By hypothesis, g(x) = f (t)dt To express f (x) in terms of g(x), multiply
x 0
thourgh by x and !apply the Sec. Fund. Thm:
x
f (t)dt = xg(x) ⇒ f (x) = g(x) + xg # (x) , by FT2..
0

T
1 1 A0 rt T A0 rT
!
4D-6 Average value of A(t) = A0 ert dt = e |0 = (e − 1)
T 0 T r rT
(rT )2
If rT is small, we can approximate: erT ≈ 1 + rT + , so we get
2
2
A0 (rT ) rT
A(t) ≈ (rT + ) = A0 (1 + ).
rT 2 2
 S. SOLUTIONS TO 18.01 EXERCISES

(If T ≈ 0, at the end of T years the interest added will be A0 rT ; thus the average is
approximately what the account grows to in T /2 years, which seems reasonable.)

1 b 2
!
4D-7 x dx = b2 /3
b 0
4D-8 The average on each side is the same as the average
2
over all four sides. Thus the average distance is x2+ (a/2)
a/2
a/2
1
! #
x2 + (a/2)2 dx x
a −a/2

Can’t be evaluated by a formula until Unit 5. The average of the square of the distance is
a/2 a/2
1 2
! !
2 2
(x + (a/2) )dx = (x2 + (a/2)2 )dx = a2 /3
a −a/2 a 0

π/a "π/a
1 1
!
"
4D-9 sin ax dx − cos(ax)"" = 2/π
π/a 0 π 0

4D’. Work

4D’-1 According to Hooke’s law, we have F = kx, where F is the force, x is the displace-
ment (i.e., the added length), and k is the Hooke’s law constant for the spring.
To find k, substitute into Hooke’s law: 2, 000 = k · (1/2) ⇒ k = 4000.
To find the work W , we have
! 6 ! 6
'6
W = F dx = 4000x dx = 2000x2 0 = 72, 000 inch-pounds = 6, 000 foot-pounds.
0 0

4D’-2 Let W (h) = weight of pail and paint at height h.

1
W (0) = 12, W (30) = 10 ⇒ W (h) = 12 − 15 h, since the pulling and leakage both occur
at a constant rate.
! 30 ! 30 (30
h h2
work = W (h) dh = (12 − ) dh = 12h − = 330 ft-lbs.
0 0 15 30 0

4D’-3 Think of the hose as divided into many equal little infinitesimal pieces, of length dh,
each of which must be hauled up to the top of the building.
The piece at distance h from the top end has weight 2 dh; to haul it up to the top requires
2h dh ft-lbs. Adding these up,
! 50
'50
total work = 2h dh = h2 0 = 2500ft-lbs.
0

g m1 m 2
4D’-4 If they are x units apart, the gravitational force between them is .
x2
nd (nd % & % &
g m1 m 2 g m1 m 2 1 1 g m1 m 2 n − 1
!
work = dx = − = −g m1 m2 − = .
d x2 x d nd d d n
 4. APPLICATIONS OF INTEGRATION

g m1 m 2
The limit as n → ∞ is .
d

4E. Parametric equations

4E-1 y − x = t2 , y − 2x = −t. Therefore,

y − x = (y − 2x)2 =⇒ y 2 − 4xy + 4x2 − y + x = 0 (parabola)

4E-2 x2 = t2 + 2 + 1/t2 and y 2 = t2 − 2 + 1/t2. Subtract, getting the hyperbola x2 − y 2 = 4

4E-3 (x − 1)2 + (y − 4)2 = sin2 θ + cos2 t = 1 (circle)

4E-4 1 + tan2 t = sec2 t =⇒ 1 + x2 = y 2 (hyperbola)

#
4E-5 x = sin 2t = 2 sin t cos t = ±2 1 − y 2 y. This gives x2 = 4y 2 − 4y 4 .

4E-6 y # = 2x, so t = 2x and

x = t/2, y = t2 /4
4E-7 Implicit differentiation gives 2x + 2yy # = 0, so that y # = −x/y. So the parameter is
t = −x/y. Substitute x = −ty in x2 + y 2 = a2 to get

t2 y 2 + y 2 = a2 =⇒ y 2 = a2 /(1 + t2 )

Thus
a −at
y= √ , x= √
1 + t2 1 + t2
For −∞ < t < ∞, this parametrization traverses the upper semicircle y > 0 (going clock-
wise). One can also get the lower semicircle (also clockwise) by taking the negative square
root when solving for y,
−a at
y= √ , x= √
1+t 2 1 + t2
4E-8 The tip Q of the hour hand is given in terms of the angle θ by Q = (cos θ, sin θ)
(units are meters). Q

Next we express θ in terms of the time parameter t (hours). We have

) * P
π/2, t = 0 #
θ= θ decreases linearly with t
π/3, t = 1
π
π − π2 · (t − 0) π
=⇒ θ − = 3 . Thus we get θ = 2 − π6 t.
2 1−0
Finally, for the snail’s position P , we have
P = (t cos θ, t sin θ) , where t increases from 0 to 1. So,
π π π
x = t cos( − t) = t sin t, y = t sin( π2 − π6 t) = t cos π6 t
2 6 6
 S. SOLUTIONS TO 18.01 EXERCISES

4F. Arclength
# √ ! 1 √ √
# 2
4F-1 a) ds = 1 + (y ) dx = 26dx. Arclength = 26dx = 26.
0
# #
b) ds = 1 + (y # )2 dx = 1 + (9/4)xdx.
! 1 # "1
Arclength = 1 + (9/4)xdx = (8/27)(1 + 9x/4)3/2 " = (8/27)((13/4)3/2 − 1)
"
0 0

#
c) y # = −x−1/3 (1 − x2/3 )1/2 = − x−2/3 − 1. Therefore, ds = x−1/3 dx, and
! 1 "1
Arclength = x−1/3 dx = (3/2)x2/3 " = 3/2
"
0 0

#
d) y # = x(2 + x2 )1/2 . Therefore, ds = 1 + 2x2 + x4 dx = (1 + x2 )dx and
! 2 "2
Arclength = (1 + x2 )dx = x + x3 /3"1 = 10/3
1
#
4F-2 y # = (ex − e−x )/2, so the hint says 1 + (y # )2 = y 2 and ds = 1 + (y # )2 dx = ydx.
Thus,
! b "b
Arclength = (1/2) (ex + e−x )dx = (1/2)(ex − e−x )"0 = (eb − e−b )/2
0

# √
! b #
4F-3 y = 2x, 1 + (y # )2 = 1 + 4x2 . Hence, arclength =
#
1 + 4x2 dx. 4F-4 ds =
# √ 0
(dx/dt)2 + (dy/dt)2 dt = 4t2 + 9t4 dt. Therefore,
! 2 # ! 2
Arclength = 4t2 + 9t4 dt = (4 + 9t2 )1/2 tdt
0 0
"2
= (1/27)(4 + 9t2 )3/2 " = (403/2 − 8)/27
"
0

4F-5 dx/dt = 1 − 1/t2 , dy/dt = 1 + 1/t2 . Thus

# #
ds = (dx/dt)2 + (dy/dt)2 dt = 2 + 2/t4 dt and
! 2 #
Arclength = 2 + 2/t4 dt
1

4F-6 a) dx/dt = 1 − cos t, dy/dt = sin t.

# √
ds/dt = (dx/dt)2 + (dy/dt)2 = 2 − 2 cos t (speed of the point)

Forward motion (dx/dt) is largest for t an odd multiple of π (cos t = −1). Forward motion
is smallest for t an even multiple of π (cos t = 1). (continued →)
 4. APPLICATIONS OF INTEGRATION

Remark: The largest forward motion is when the point is at the top of the wheel and the
smallest is when the point is at the bottom (since y = 1 − cos t.)
√
! 2π ! 2π
2π
b) 2 − 2 cos tdt = 2 sin(t/2)dt = −4 cos(t/2)|0 = 8
0 0

! 2π #
4F-7 a2 sin2 t + b2 cos2 tdt
0

4F-8 dx/dt = et (cos t − sin t), dy/dt = et (cos t + sin t).

# # √
ds = e2t (cos t − sin t)2 + e2t (cos t + sin t)2 dt = et 2 cos2 t + 2 sin2 tdt = 2et dt

Therefore, the arclength is

! 10 √ t √
2e dt = 2(e10 − 1)
0
y= R2- x 2

4G. Surface Area

a b
√
4G-1 The curve y = − R2 x2
for a ≤ x ≤ b is revolved around the x-axis.
√
Since we have y = −x/ R2 − x2 , we get
#

# # #
ds = 1 + (y # )2 dx = 1 + x2 /(R2 − x2 )dx = R2 /(R2 − x2 )dx = (R/y)dx

Therefore, the area element is

dA = 2πyds = 2πRdx

and the area is

! b
2πRdx = 2πR(b − a)
a

√
4G-2 Limits are 0 ≤ x ≤ 1/2. ds = 5dx, so

√ √ ! 1/2 √
dA = 2πyds = 2π(1−2x) 5dx =⇒ A = 2π 5 (1−2x)dx = 5π/2
0
 S. SOLUTIONS TO 18.01 EXERCISES 4G!3

4G-3 Limits are 0 ≤ y ≤ 1. x = (1 − y)/2, dx/dy = −1/2. Thus 1

y = 1 ! 2x
# # x = (1 ! y)/2
ds = 1 + (dx/dy)2 dy = 5/4dy;
√ √ $1 √ 1/2 4G!2
dA = 2πyds = π(1 − y)( 5/2)dx =⇒ A = ( 5π/2) 0 (1 − y)dy = 5π/4
! ! 4 #
4G-4 A = 2πyds = 2πx2 1 + 4x2 dx
0
√ √ #
4G-5 x = y, dx/dy = −1/2 y, and ds = 1 + 1/4ydy

2
√ #
! !
A= 2πxds = 2π y 1 + 1/4ydy
0
! 2 #
= 2π y + 1/4dy
0
"2
= (4π/3)(y + 1/4)3/2 " = (4π/3)((9/4)3/2 − (1/4)3/2 )
"
0
= 13π/3

4G-6 y = (a2/3 − x2/3 )3/2 =⇒ y # = −x−1/3 (a2/3 − x2/3 )1/2 . Hence

+
ds = 1 + x−2/3 (a2/3 − x2/3 )dx = a1/3 x−1/3 dx -a a
3/2
Therefore, (using symmetry on the interval −a ≤ x ≤ a) y = (a2/3- x 2/3)
! ! a
A= 2πyds = 2 2π(a2/3 − x2/3 )3/2 a1/3 x−1/3 dx
0
"a
= (4π)(2/5)(−3/2)a1/3 (a2/3 − x2/3 )5/2 "
"
0 2
y = a 2- (x-b)
= (12π/5)a2
# a a
4G-7 a) Top half: y = a2 − (x − b)2 , y # = (b − x)/y. Hence, b

2
y = - a 2- (x-b)
# #
ds = 1 + (b − x)2 /y 2 dx = (y 2 + (b − x)2 )/y 2 dx = (a/y)dx
upper and lower surfaces are
Since we are only covering the top half we double the integral for area: symmetrical and equal

b+a
xdx
! !
A= 2πxds = 4πa #
b−a a − (x − b)2
2
 4. APPLICATIONS OF INTEGRATION
#
b) We need to rotate two curves x2 = b + a2 − y 2
#
and x1 = b − a2 − y 2 around the y-axis. The value x = b + a 2- y 2
#
dx2 /dy = −(dx1 /dy) = −y/ a2 − y 2

So in both cases, # # x = b - a -2 y 2
ds = 1 + y 2 /(a2 − y 2 )dy = (a/ a2 − y 2 )dy
inner and outer surfaces are
The integral is not symmetrical and not equal

a
ady
! ! !
A= 2πx2 ds + 2πx1 ds = 2π(x1 + x2 ) #
−a a2 − y 2
But x1 + x2 = 2b, so
a
dy
!
A = 4πab #
−a a2 − y 2

c) Substitute y = a sin θ, dy = a cos θdθ to get

π/2 π/2
a cos θdθ
! !
A = 4πab = 4πab dθ = 4π 2 ab
−π/2 a cos θ −π/2

4H. Polar coordinate graphs

4H-1 We give the polar coordinates in the form (r, θ):

√
a) (3, π/2) b) (2, π) c) (2, π/3) d) (2 2, 3π/4)
√
e) ( 2, −π/4 or 7π/4) f) (2, −π/2 or 3π/2)
√
g) (2, −π/6 or 11π/6) h) (2 2, −3π/4 or 5π/4)

4H-2 a) (i) (x−a)2 +y 2 = a2 ⇒ x2 −2ax+y 2 = 0 ⇒ r2 −2ar cos θ = 0 ⇒ r = 2a cos θ.

(ii) ∠OP Q = 90o , since it is an angle inscribed in a semicircle.
In the right triangle OPQ, |OP | = |OQ| cos θ, i.e., r = 2a cos θ.
b) (i) Analogous to 4H-2a(i); ans: r = 2a sin θ.
(ii) analogous to 4H-2a(ii); note that ∠OQP = θ, since both angles are complements
of ∠P OQ.
c) (i) OQP is a right triangle, |OP | = r, and ∠P OQ = α − θ.
The polar equation is r cos(α − θ) = a, or in expanded form,
r(cos α cos θ + sin α sin θ) = a, or finally,
x y
+ = 1,
A B
a a
since from the right triangles OAQ and OBQ, we have cos α = , sin α = cos BOQ = .
A B
d) Since |OQ| = sin θ, we have:
if P is above the x-axis, sin θ > 0, OP | = |OQ| − |QR|, or r = a − a sin θ;
if P is below the x-axis, sin θ < 0, OP | = |OQ| + |QR|, or r = a + a| sin θ| = a − a sin θ.
Thus the equation is r = a(1 − sin θ).
 S. SOLUTIONS TO 18.01 EXERCISES

e) Briefly, when P = (0, 0), |P Q||P R| = a · a = a2 , the constant.

Using the law of cosines,
|P R|2 = r2 + a2 − 2ar cos θ;
|P Q|2 = r2 + a2 − 2ar cos(π − θ) = r2 + a2 + 2ar cos θ
Therefore
|P Q|2 |P R|2 = (r2 + a2 )2 − (2ar cos θ)2 = (a2 )2
which simplifies to
r2 = 2a2 cos 2θ.

4H-3 a) r = sec θ =⇒ r cos θ = 1 =⇒ x = 1 b) r = 2a cos θ =⇒ r2 = r · 2a cos θ =

2ax =⇒ x2 + y 2 = 2ax
c) r = (a + b cos θ) (This figure is a cardiod for a = b, a limaçon with a loop for
0 < a < b, and a limaçon without a loop for a > b > 0.)
#
r2 = ar + br · cos θ = ar + bx =⇒ x2 + y 2 = a x2 + y 2 + bx
b=|c| hyperbola b<|c|
y parabola

r r r b>|c| ellipse
r r
# a 2a # # #
# x
1

limacon a<b cardioid (a=b) limacon a>b

8a 8b 8c 8d

(d) r = a/(b + c cos θ) =⇒ r(b + c cos θ) = a =⇒ rb + cx = a

=⇒ rb = a − cx =⇒ r b = a2 − 2acx + c2 x2
2 2

=⇒ a2 − 2acx + (c2 − b2 )x2 − b2 y 2 = 0

(e) r = a sin(2θ) =⇒ r = 2a sin θ cos θ = 2axy/r2

=⇒ r3 = 2axy =⇒ (x2 + y 2 )3/2 = 2axy

r = a cos 2 # r = a sin 2# 2
r2 = a cos 2# r2 = a2sin 2#

2x2
f) r = a cos(2θ) = a(2 cos2 θ − 1) = a( − 1) =⇒ (x2 + y 2 )3/2 = a(x2 − y 2 )
x2 + y 2
xy
g) r2 = a2 sin(2θ) = 2a2 sin θ cos θ = 2a2 2 =⇒ r4 = 2a2 xy =⇒ (x2 + y 2 )2 = 2axy
r
2x2
h) r2 = a2 cos(2θ) = a2 ( − 1) =⇒ (x2 + y 2 )2 = a2 (x2 − y 2 )
x2 + y 2
# y
i) r = eaθ =⇒ ln r = aθ =⇒ ln x2 + y 2 = a tan−1
x
 4. APPLICATIONS OF INTEGRATION

4I. Area and arclength in polar coordinates

#
4I-1 (dr/dθ)2 + r2 dθ
a) sec2 θdθ
b) 2adθ
#
c) a2 + b2 + 2ab cos θdθ
√
a b2 + c2 + 2bc cos θ
d) dθ
(b + c cos θ)2
+
e) a 4 cos2 (2θ) + sin2 (2θ)dθ
+
f) a 4 sin2 (2θ) + cos2 (2θ)dθ
g) Use implicit differentiation:

2rr# = 2a2 cos(2θ) =⇒ r# = a2 cos(2θ)/r =⇒ (r# )2 = a2 cos2 (2θ)/ sin(2θ)

Hence, using a common denominator and cos2 + sin2 = 1,

# a
ds = a2 cos2 (2θ)/ sin(2θ) + a2 sin(2θ)dθ = # dθ
sin(2θ)

h) This is similar to (g):

a
ds = # dθ
cos(2θ)
#
i) 1 + a2 eaθ dθ

4I-2 dA = (r2 /2)dθ. The main difficulty is to decide on the endpoints of integration.
Endpoints are successive times when r = 0.

cos(3θ) = 0 =⇒ 3θ = π/2 + kπ =⇒ θ = π/6 + kπ/3, k an integer. # = !/6

! π/6 ! π/6
2 2 2
Thus, A = (a cos (3θ)/2)dθ = a cos2 (3θ)dθ. # =$!/6
−π/6 0
three-leaf rose
(Stop here in Unit 4. Evaluated in Unit 5.) three empty sectors

! ! π
2
"π
4I-3 A = (r /2)dθ = (e6θ /2)dθ = (1/12)e6θ "0 = (e6π −
0
1)/12 e3! 1

4I-4 Endpoints are successive time when r = 0.

sin(2θ) = 0 =⇒ 2θ = kπ, k an integer.

! ! π/2 "π/2
Thus, A = (r2 /2)dθ = (a2 /2) sin(2θ)dθ = −(a2 /4) cos(2θ)"0 =
0
a2 /2.
 S. SOLUTIONS TO 18.01 EXERCISES

4I-5 r = 2a cos θ, ds = 2adθ, −π/2 < θ < π/2. (The range was
chosen carefully so that r > 0.) Total length of the circle is 2πa. Since
the upper and lower semicircles are symmetric, it suffices to calculate 2a
the average over the upper semicircle: r = 2a cos #

π/2 "π/2
1 4a 4a
!
"
2a cos θ(2a)dθ = sin θ"" =
πa 0 π 0 π
P

4I-6 a) Since the upper and lower halves of the cardiod are symmetric, r
it suffices to calculate the average distance to the x-axis just for a point 0
Q
on the upper half. We have r = a(1 − cos θ), and the distance to the
x-axis is r sin θ, so
π π "π
1 1 a 2a
! !
r sin θdθ = a(1 − cos θ) sin θdθ = (1 − cos θ)2 " =
"
π 0 π 0 2π 0 π

# +
(b) ds = (dr/dθ)2 + r2 dθ = a (1 − cos θ)2 + sin2 θdθ
√
= a 2 − 2 cos θdθ = 2a sin(θ/2)dθ, using the half angle formula.
! 2π
2π
arclength = 2a sin(2θ)dθ = −4a cos(θ/2)|0 = 8a
0

For the average, don’t use the half-angle version of the formula for ds, and use the interval
−π < θ < π, where sin θ is odd:
! π √ ! π √
1 1
Average = |r sin θ|a 2 − 2 cos θdθ = | sin θ| 2a2 (1 − cos θ)3/2 dθ
8a −π 8a −π
√ ! π √ "π
2a 2a " 4
= (1 − cos θ)3/2 sin θdθ = (1 − cos θ)5/2 " = a
"
4 0 10 " 5
0

4I-7 dx = −a sin θdθ. So the semicircle y > 0 has area

! a ! 0 ! π
ydx = a sin θ(−a sin θ)dθ = a 2
sin2 θdθ
−a π 0

But π π
1
! !
2
sin θdθ = (1 − cos(2θ)dθ = π/2
0 2 0

So the area is πa2 /2 as it should be for a semicircle.

Arclength: ds2 = dx2 + dy 2 ds
d#
2 2 2 2 2 2 2
=⇒ (ds) = (−a sin θdθ) + (a cos θdθ) = a (sin dθ + cos dθ)(dθ)
#
=⇒ ds = adθ (obvious from picture). -a a
! ! 2π
ds = adθ = 2πa
0
 4. APPLICATIONS OF INTEGRATION

4J. Other applications

4J-1 Divide the water in the hole into n equal circular discs of thickness ∆y.
% &2
1
Volume of each disc: π ∆y
2
π
Energy to raise the disc of water at depth yi to surface: kyi ∆y.
4
Adding up the energies for the different discs, and passing to the limit,
n ! 100 (100
, π π πk y 2 πk104
E = lim kyi ∆y = ky dy = = .
n→∞
1
4 0 4 4 2 0 8

4J-2 Divide the hour into n equal small time intervals ∆t.
At time ti , i = 1, . . . , n, there are x0 e−kti grams of material, producing approximately
rx0 e−kti ∆t radiation units over the time interval [ti , ti + ∆t].
Adding and passing to the limit,
n ! 60 (60
, e−kt r x0 - .
R = lim r x0 e−kti ∆t = r x0 e−kt dt = r x0 = 1 − e−60k .
n→∞
1 0 −k 0 k

4J-3 Divide up the pool into n thin concentric cylindrical shells, of radius ri , i = 1, . . . , n,
and thickness ∆r.
The volume of the i-th shell is approximately 2π ri D ∆r.
k
The amount of chemical in the i-th shell is approximately 2π ri D ∆r.
1 + ri2
Adding, and passing to the limit,
n ! R
, k r
A = lim 2 2π ri D ∆r = 2πkD dr
n→∞
1
1 + ri 0 1 + r2
(R
= πkD ln(1 + r2 ) = πkD ln(1 + R2 ) gms.
0

4J-4 Divide the time interval into n equal small intervals of length ∆t by the points ti ,
i = 1, . . . , n.
The approximate number of heating units required to maintain the temperature at 75o
over the time interval [ti , ti + ∆t]: is
/ % &(
πti
75 − 10 6 − cos · k ∆t.
12
Adding over the time intervals and passing to the limit:
n / % &(
, πti
total heat = lim 75 − 10 6 − cos · k ∆t
n→∞
1
12
! 24 / % &(
πt
= k 75 − 10 6 − cos dt
0 12
! 24 % & / (24
πt 120 πt
= k 15 + 10 cos dt = k 15t + sin = 360k.
0 12 π 12 0
 S. SOLUTIONS TO 18.01 EXERCISES

4J-5 Divide the month into n equal intervals of length ∆t by the points ti , i = 1, . . . , n.
Over the time interval [ti .ti + ∆t], the number of units produced is about (10 + ti ) ∆t.
The cost of holding these in inventory until the end of the month is c(30 − ti )(10 + ti ) ∆t.
Adding and passing to the limit,
n
,
total cost = lim c(30 − ti )(10 + ti ) ∆t
n→∞
1
30 (30
t3
! /
= c(30 − t)(10 + t) dt = c 300t + 10t2 − = 9000c.
0 3 0

4J-6 Divide the water in the tank into thin horizontal slices of width# dy.
If the slice is at height y above the center of the tank, its radius is r2 − y 2 .
This formula for the radius of the slice is correct even if y < 0 – i.e., the slice is below the
center of the tank – as long as −r < y < r, so that there really is a slice at that height.
Volume of water in the slice = π(r2 − y 2 ) dy
Weight of water in the slice = πw(r2 − y 2 ) dy
Work to lift this slice from the ground to the height h + y = πw(r2 − y 2 ) dy (h + y).
! r
Total work = πw(r2 − y 2 )(h + y) dy
−r
! r
= πw (r2 h + r2 y − hy 2 − y 3 )
−r
(r
r2 y 2 hy 3 y4
/
2
= πw r hy + − − .
2 3 4 −r

In this last line, the even powers of y have the same value at −r and r, so contribute 0 when
it is evaluated; we get therefore
(r
y3 r3
/ % &
4
= πwh r2 y − = 2πwh r3 − = πwhr3 .
3 −r 3 3
 Unit 5. Integration techniques
5A. Inverse trigonometric functions; Hyperbolic functions
√
−1
√ π −1 3 π
5A-1 a) tan 3= b) sin ( )=
3 2 3
√ √ √
c)√tan θ = 5 implies sin θ = 5/ 26, cos θ = 1/ 26, cot θ = 1/5, csc θ = 26/5,
sec θ = 26 (from triangle)
√
π 3 π π π
d) sin−1 cos( ) = sin−1 ( )= e) tan−1 tan( ) =
6 2 3 3 3
2π −π −π −π
f) tan−1 tan( ) = tan−1 tan( )= g) lim tan−1 x = .
3 3 3 x→−∞ 2
! 2
dx "2 π
5A-2 a) 2
= tan−1 x"1 = tan−1 2 −
1 x + 1 4
! 2b ! 2b ! 2
dx d(by) dy 1 π
b) 2 + b2
= 2 + b2
(put x = by) = 2 + 1)
= (tan−1 2 − )
b x b (by) 1 b(y b 4
! 1
dx "1 π −π
c) √ = sin−1 x"−1 = − =π
−1 1 − x 2 2 2

x−1 1 (x + 1)
5A-3 a) y = , so 1 − y 2 = 4x/(x + 1)2 , and # = √ . Hence
x+1 1−y 2 2 x
dy 2
=
dx (x + 1)2
d dy/dx
sin−1 y = #
dx 1 − y2
2 (x + 1)
= · √
(x + 1)2 2 x
1
= √
(x + 1) x

b) sech2 x = 1/ cosh2 x = 4/(ex + e−x )2

√ √
c) y = x + x2 + 1, dy/dx = 1 + x/ x2 + 1.
√
d dy/dx 1 + x/ x2 + 1 1
ln y = = √ = √
dx y x + x2 + 1 x2 + 1

d) cos y = x =⇒ (− sin y)(dy/dx) = 1

dy −1 −1
= =√
dx sin y 1 − x2

e) Chain rule:
d 1 1 1
sin−1 (x/a) = # · = √
dx 1 − (x/a)2 a a − x2
2
 S. SOLUTIONS TO 18.01 EXERCISES

f) Chain rule:
d 1 −a −a
sin−1 (a/x) = # · = √
dx 1 − (a/x)2 x2 x x2 − a2
√
g) y = x/ 1 − x2 , dy/dx = (1 − x2 )−3/2 , 1 + y 2 = 1/(1 − x2 ). Thus
d dy/dx 1
tan−1 y = = (1 − x2 )−3/2 (1 − x2 ) = √
dx 1 + y2 1 − x2
Why is this the same as the derivative of sin−1 x?
√ √
h) y = 1 − x, dy/dx = −1/2 1 − x, 1 − y 2 = x. Thus,
d dy/dx −1
sin−1 y = # = #
dx 1−y 2 2 x(1 − x)
5A-4 a) y $ = sinh x. A tangent line through the origin has the equation y = mx. If it meets
the graph at x = a, then ma = cosh(a) and m = sinh(a). Therefore, a sinh(a) = cosh(a) .
b) Take the difference:
F (a) = a sinh(a) − cosh(a)
Newton’s method for finding F (a) = 0, is the iteration
an+1 = an − F (an )/F $ (an ) = an − tanh(an ) + 1/an
With a1 = 1, a2 = 1.2384, a3 = 1.2009, a4 = 1.19968. A serviceable approximation is
a ≈ 1.2
(The slope is m = sinh(a) ≈ 1.5.) The functions F and y are even. By symmetry, there is
another solution −a with slope − sinh a.

5A-5 a)
ex − e−x
y = sinh x =
2
x
e + e−x
y $ = cosh x =
2
y $$ = sinh x
y $ is never zero, so no critical points. Inflection point x = 0; slope
of y is 1 there. y is an odd function, like ex /2 for x >> 0. y = sinh x y = sinh 1x

b) y = sinh−1 x ⇐⇒ x = sinh y. Domain is the whole x-axis.

c) Differentiate x = sinh y implicitly with respect to x:
dy
1 = cosh y ·
dx
dy 1 1
= =#
dx cosh y sinh2 y + 1
d sinh−1 x 1
= √
dx 2
x +1
 5. INTEGRATION TECHNIQUES

d)

dx dx
! !
√ = #
x + a2
2 a x + a2 /a2
2

d(x/a)
!
= #
(x/a)2 + 1
= sinh−1 (x/a) + c

π
1
!
5A-6 a) sin θdθ = 2/π
π 0
√ √
b) y = 1 − x2 =⇒ y $ = −x/ 1 − x2 =⇒ 1 + (y $ )2 = 1/(1 − x2 ). Thus
#
ds = w(x)dx = dx/ 1 − x2 .

Therefore the average is

1 $! 1
dx dx
! #
1 − x2 √ √
1−x 2 1 − x2
−1 −1

! 1
The numerator is dx = 2. To see that these integrals are the same as the ones in part
−1
(a), take x = cos θ (as in polar coordinates). Then dx = − sin θdθ and the limits of integral
are from θ = π to θ = 0. Reversing the limits changes the minus back to plus:
1 ! π
dx
! #
1 − x2 √ = sin θdθ
−1 1 − x2 0
! 1 ! π
dx
√ = dθ = π
−1 1 − x2 0

(The substitution x = sin t works similarly, but the limits of integration are −π/2 and π/2.)
c) (x = sin t, dx = cos tdt)

1
1 π/2
! π/2
1
! # !
1 − x2 dx = cos2 tdt = cos2 tdt
2 −1 2 −π/2 0
! π/2
1 + cos 2t
= dt
0 2
= π/4

5B. Integration by direct substitution

Do these by guessing and correcting the factor out front. The substitution used implicitly
is given alongside the answer.
1
! #
3
5B-1 x x2 − 1dx = (x2 − 1) 2 + c (u = x2 − 1, du = 2xdx)
3
 S. SOLUTIONS TO 18.01 EXERCISES

1 8x
!
5B-2 e8x dx = e + c (u = 8x, du = 8dx)
8
ln xdx 1
!
5B-3 = (ln x)2 + c (u = ln x, du = dx/x)
x 2
cos xdx ln(2 + 3 sin x)
!
5B-4 = + c (u = 2 + 3 sin x, du = 3 cos xdx)
2 + 3 sin x 3

sin x3
!
5B-5 sin2 x cos xdx = + c (u = sin x, du = cos xdx)
3
− cos 7x
!
5B-6 sin 7xdx = + c (u = 7x, du = 7dx)
7
6xdx
! #
5B-7 √ = 6 x2 + 4 + c (u = x2 + 4, du = 2xdx)
x2 + 4
5B-8 Use u = cos(4x), du = −4 sin(4x)dx,

sin(4x)dx −du
! ! !
tan 4xdx = =
cos(4x) 4u
ln u ln(cos 4x)
=− +c=− +c
4 4
3
!
5B-9 ex (1 + ex )−1/3 dx = (1 + ex )2/3 + c (u = 1 + ex , du = ex dx)
2
1
!
5B-10 sec 9xdx = ln(sec(9x) + tan(9x)) + c (u = 9x, du = 9dx)
9
tan 9x
!
5B-11 sec2 9xdx = + c (u = 9x, du = 9dx)
9
2
−e−x
!
−x2
5B-12 xe dx = + c (u = x2 , du = 2xdx)
2
5B-13 u = x3 , du = 3x2 dx implies

x2 dx du tan−1 u
! !
= = +c
1 + x6 3(1 + u2 ) 3
tan−1 (x3 )
= +c
3
! π/3 ! sin π/3
3
5B-14 sin x cos xdx = u3 du (u = sin x, du = cos xdx)
0 sin 0
√ "√3/2
! 3/2 "
= u3 du = u4 /4" = 9
"
64
0 "
0

e ln e
(ln x)3/2 dx
! !
5B-15 = u3/2 du (u = ln x, du = dx/x)
1 x ln 1
 5. INTEGRATION TECHNIQUES

1 "1 2
!
= y 3/2 dy = (2/5)y 5/2 " =
"
0 0 5
1 ! tan−1 1
tan−1 xdx
!
5B-16 = udu (u = tan−1 x, du = dx/(1 + x2 )
−1 1 + x2 tan−1 (−1)
! π/4 "π/4
u2 ""
= udu = =0
−π/4 2 "−π/4

(tan x is odd and hence tan−1 x is also odd, so the integral had better be 0)

5C. Trigonometric integrals

1 − cos 2x x sin 2x
! !
5C-1 sin2 xdx = dx = − +c
2 2 4
! ! !
5C-2 sin3 (x/2)dx = (1 − cos2 (x/2)) sin(x/2)dx = −2(1 − u2 )du
(put u = cos(x/2), du = (−1/2) sin(x/2)dx)
3
2u3 2cos(x/2)
= −2u + + c = −2 cos(x/2) + +c
3 3
1 − cos 2x 2 1 − 2 cos 2x + cos2 2x
! ! !
5C-3 sin4 xdx = ( ) dx = dx
2 4
cos2 (2x) 1 + cos 4x x sin 4x
! !
dx = dx = + +c
4 8 8 32
Adding together all terms:
3x 1 1
!
sin4 xdx = − sin(2x) + sin(4x) + c
8 4 32
1 − u2
! ! !
5C-4 cos3 (3x)dx = (1 − sin2 (3x)) cos(3x)dx = du (u = sin(3x), du =
3
3 cos(3x)dx)
3
u u3 sin(3x) sin(3x)
= − +c= − +c
3 9 3 9
! ! !
3 2 2 2
5C-5 sin x cos xdx = (1 − cos x) cos x sin xdx = −(1 − u2 )u2 dy (u = cos x,
du = − sin xdx)
u3 u5 cos x3 cos x5
=− + +c=− + +c
3 5 3 5
! ! !
5C-6 sec4 xdx = (1 + tan2 x) sec2 xdx = (1 + u2 )du (u = tan x, du = sec2 xdx)

u3 tan3 x
=u+ + c = tan x + +c
3 3
sin2 8xdx (1 − cos 16x)dx 1 sin 16x
! ! !
2 2
5C-7 sin (4x) cos (4x)dx = = = − +c
4 8 8 128
 S. SOLUTIONS TO 18.01 EXERCISES

A slower way is to use

% &% &
2 2 1 − cos(8x) 1 + cos(8x)
sin (4x) cos (4x) =
2 2

multiply out and use a similar trick to handle cos2 (8x).

5C-8
sin2 (ax)
! !
tan2 (ax) cos(ax)dx = dx
cos(ax)
1 − cos2 (ax)
!
= dx
cos(ax)
!
= (sec(ax) − cos(ax))dx
1 1
= ln(sec(ax) + tan(ax)) − sin(ax) + c
a a
5C-9
1 − cos2 x
! !
sin3 x sec2 xdx = sin xdx
cos2 x
1 − u2
!
= − du (u = cos x, du = − sin xdx)
u2
1
= u + + c = cos x + sec x + c
u
5C-10
! ! !
(tan x + cot x)2 dx = tan2 x + 2 + cot2 xdx = sec2 x + csc2 xdx
= tan x − cot x + c
!
5C-11 sin x cos(2x)dx
! !
= sin x(2 cos2 x − 1)dx = (1 − 2u2 )du (u = cos x, du − sin xdx)

2 2
= u − u3 + c = cos x − cos3 x + c
3 3
! π "π
2 −2
sin x cos(2x)dx = cos x − cos3 x"" =
"
5C-12 (See 27.)
0 3 0 3
# #
5C-13 ds = 1 + (y $ )2 dx = 1 + cot2 xdx = csc xdx.
"π/2
! π/2 " √
arclength = csc xdx = − ln(csc x + cot x)" = ln(1 + 2)
"
π/4 "
π/4

! π/a ! π/a
5C-14 π sin2 (ax)dx = π (1/2)(1 − cos(2ax))dx = π 2 /2a
0 0
 5. INTEGRATION TECHNIQUES

5D. Integration by inverse substitution

5D-1 Put x = a sin θ, dx = a cos θdθ:

dx 1 1 x
! !
= 2 sec2 θdθ = 2 tan θ + c = √ +c
(a2 − x2 )3/2 a a a a2 − x2
2

5D-2 Put x = a sin θ, dx = a cos θdθ:

x3 dx
! ! !
3 3 3
√ =a sin θdθ = a (1 − cos2 θ) sin θdθ
a2 − x2
= a3 (− cos θ + (1/3) cos3 θ) + c
#
= −a2 a2 − x2 + (a2 − x2 )3/2 /3 + c
5D-3 By direct substitution (u = 4 + x2 ),
xdx
!
= (1/2) ln(4 + x2 ) + c
4 + x2
Put x = 2 tan θ, dx = 2 sec2 θdθ,
dx 1
! !
2
= dθ = θ/2 + c
4+x 2
In all,
(x + 1)dx
!
= (1/2) ln(4 + x2 ) + (1/2) tan−1 (x/2) + c
4 + x2
5D-4 Put x = a sinh y, dx = a cosh ydy. Since 1 + sinh2 y = cosh2 y,
a2
! # ! !
2 2 2 2
a + x dx = a cosh ydy = (cosh(2y) − 1)dy
2
= (a2 /4) sinh(2y) − a2 y/2 + c = (a2 /2) sinh y cosh y − a2 y/2 + c
#
= x a2 + x2 /2 − a2 sinh−1 (x/a) + c
5D-5 Put x = a sin θ, dx = a cos θdθ:
! √ 2
a − x2 dx
!
= cot2 θdθ
x2
!
= (csc2 θ − 1)dθ = − ln(csc θ + cot θ) − θ + c
#
= − ln(a/x + a2 − x2 /x) − sin−1 (x/a) + c
5D-6 Put x = a sinh y, dx = a cosh ydy.
! # !
x2 a2 + x2 dx = a4 sinh2 y cosh2 ydy
! !
4 2 4
= (a /2) sinh (2y)dy = a /4 (cosh(4y) − 1)dy

= (a4 /16) sinh(4y) − a4 y/4 + c

= (a4 /8) sinh(2y) cosh(2y) − a4 y/4 + c
= (a4 /4) sinh y cosh y(cosh2 y + sinh2 y) − a4 y/4 + c
#
= (1/4)x a2 + x2 (2x2 + a2 ) − (a4 /4) sinh−1 (x/a) + c
 S. SOLUTIONS TO 18.01 EXERCISES

5D-7 Put x = a sec θ, dx = a sec θ tan θdθ:

! √
x2 − a2 dx tan2 θdθ
!
=
x2 sec θ
(sec2 θ − 1)dθ
! !
= = (sec θ − cos θ)dθ
sec θ
= ln(sec θ + tan θ) − sin θ + c
# #
= ln(x/a + x2 − a2 /a) − x2 − a2 /x + c
# #
= ln(x + x2 − a2 ) − x2 − a2 /x + c1 (c1 = c − ln a)

5D-8 Short way: u = x2 − 9, du = 2xdx,

! #
x x2 − 9dx = (1/3)(x2 − 9)3/2 + c direct substitution

Long way (method of this section): Put x = 3 sec θ, dx = 3 sec θ tan θdθ.
! # !
x x2 − 9dx = 27 sec2 θ tan2 θdθ
!
= 27 tan2 θd(tan θ) = 9 tan3 θ + c

= (1/3)(x2 − 9)3/2 + c
√
(tan θ = x2 − 9/3). The trig substitution method does not lead to a dead end, but it’s
not always fastest.
#
5D-9 y $ = 1/x, ds = 1 + 1/x2 dx, so
! b #
arclength = 1 + 1/x2 dx
1

Put x = tan θ, dx = sec2 θdθ,

! √ 2
x + 1dx sec θ
!
= sec2 θdθ
x tan θ
sec θ(1 + tan2 θ)
!
= dθ
tan θ
!
= (csc θ + sec θ tan θ)dθ

= − ln(csc θ + cot θ) + sec θ + c

# #
= − ln( x2 + 1/x + 1/x) + x2 + 1 + c
# #
= − ln( x2 + 1 + 1) + ln x + x2 + 1 + c
# # √ √
arclength = − ln( b2 + 1 + 1) + ln b + b2 + 1 + ln( 2 + 1) − 2
 5. INTEGRATION TECHNIQUES

Completing the square

dx dx
! !
5D-10 = (x + 2 = 3 tan θ, dx = 3 sec2 θdθ)
(x2 + 4x + 13)3/2 ((x + 2)2 + 32 )3/2
1 1 (x + 2)
!
= cos θdθ = sin θ + c = √ +c
9 9 2
9 x + 4x + 13
5D-11
! # ! #
x −8 + 6x − x dx = x 1 − (x − 3)2 dx
2 (x − 3 = sin θ, dx = cos θdθ)
!
= (sin θ + 3) cos2 θdθ
!
= (−1/3) cos3 θ + (3/2) (cos 2θ + 1)dθ

= −(1/3) cos3 θ + (3/4) sin 2θ + (3/2)θ + c

= −(1/3) cos3 θ + (3/2) sin θ cos θ + (3/2)θ + c
= −(1/3)(−8 + 6x − x2 )3/2
#
+ (3/2)(x − 3) −8 + 6x − x2 + (3/2) sin−1 (x − 3) + c

5D-12
! # ! #
−8 + 6x − x2 dx = 1 − (x − 3)2 dx (x − 3 = sin θ, dx = cos θdθ)
!
= cos2 θdθ
1
!
= (cos 2θ + 1)dθ
2
1 θ
= sin 2θ + + c
4 2
1 θ
= sin θ cos θ + + c
2 √ 2
(x − 3) −8 + 6x − x2 sin−1 (x − 3)
= + +c
2 2
dx dx
! !
5D-13 √ = # . Put x − 1 = sin θ, dx = cos θdθ.
2x − x2 1 − (x − 1)2
!
= dθ = θ + c = sin−1 (x − 1) + c

xdx xdx
! !
5D-14 √ = # . Put x + 2 = 3 tan θ, dx = 3 sec2 θ.
2
x + 4x + 13 (x + 2)2 + 32
!
= (3 tan θ − 2) sec θdθ = 3 sec θ − 2 ln(sec θ + tan θ) + c
# #
= x2 + 4x + 13 − 2 ln( x2 + 4x + 13/3 + (x + 2)/3) + c
# #
= x2 + 4x + 13 − 2 ln( x2 + 4x + 13 + (x + 2)) + c1 (c1 = c − ln 3)
 S. SOLUTIONS TO 18.01 EXERCISES

! √ ! #
4x2 − 4x + 17dx (2x − 1)2 + 42 dx
5D-15 =
2x − 1 2x − 1
(put 2x − 1 = 4 tan θ, dx = 2 sec2 θdθ as in Problem 9)

sec θ
!
=2 sec2 θdθ
tan θ
sec θ(1 + tan2 θ)
!
=2 dθ
tan θ
!
= 2 (csc θ + sec θ tan θ)dθ

= −2 ln(csc θ + cot θ) + 2 sec θ + c

# #
= −2 ln( 4x2 − 4x + 17/(2x − 1) + 4/(2x − 1)) + 4x2 − 4x + 17/2 + c
# #
= −2 ln( 4x2 − 4x + 17 + 4) + 2 ln(2x − 1) + 4x2 − 4x + 17/2 + c

5E. Integration by partial fractions

1 1/5 −1/5
5E-1 = + (cover up)
(x − 2)(x + 3) x−2 x+3

dx
!
= (1/5) ln(x − 2) − (1/5) ln(x + 3) + c
(x − 2)(x + 3)

x 2/5 3/5
5E-2 = + (cover up)
(x − 2)(x + 3) x−2 x+3

xdx
!
= (2/5) ln(x − 2) + (3/5) ln(x + 3) + c
(x − 2)(x + 3)

x 1/10 1/2 −3/5

5E-3 = + + (cover up)
(x − 2)(x + 2)(x + 3) x−2 x+2 x+3

xdx
!
= (1/10) ln(x − 2) + (1/2) ln(x + 2) − (3/5) ln(x + 3)
(x2 − 4)(x + 3)

3x2 + 4x − 11 2 −2 3
5E-4 dx = + + (cover-up)
(x2 − 1)(x − 2) x−1 x+1 x−2

2dx −2dx 3dx

!
+ + = 2 ln(x − 1) − 2 ln(x + 1) + 3 ln(x − 2) + c
x−1 x+1 x−2
.
 5. INTEGRATION TECHNIQUES

3x + 2 2 B 1
5E-5 2
= + + (coverup); to get B, put say x = 1:
x(x + 1) x x + 1 (x + 1)2

5 B 1
=2+ + =⇒ B = −2
4 2 4
3x + 2 1
!
2
dx = 2 ln x − 2 ln(x + 1) − +c
x(x + 1) x+1
2x − 9 Ax + B C
5E-6 = 2 +
(x2 + 9)(x + 2) x +9 x+2
By cover-up, C = −1. To get B and A,

−9 B 1
x = 0 =⇒ = − =⇒ B = 0
9·2 9 2
−7 A 1
x = 1 =⇒ = − =⇒ A = 1
10 · 3 10 3
2x − 9 1
!
2
dx = ln(x2 + 9) − ln(x + 2) + c
(x + 9)(x + 2) 2
5E-7 Instead of thinking of (4) as arising from (1) by multiplication by x − 1, think of it
as arising from
x − 7 = A(x + 2) + B(x − 1)
by division by x + 2; since this new equation is valid for all x, the line (4) will be valid for
x &= −2, in particular it will be valid for x = 1 .

5E-8 Long division:

x2 1
a) =1+ 2
x2 − 1 x −1
x3 x
b) 2
=x+ 2
x −1 x −1
x2 1/9
c) = x/3 + 1/9 +
3x − 1 3x − 1
x+2 1 7/3
d) = +
3x − 1 3 3x − 1
x8 4 3 2 B3 x3 + B2 x2 + B1 x + B0
e) = A4 x + A3 x + A2 x + A1 x + A0 +
(x + 2)2 (x − 2)2 (x + 2)2 (x − 2)2
5E-9 a) Cover-up gives

1 1 1/2 −1/2
= = +
x2 −1 (x − 1)(x + 1) x−1 x+1

From 8a,
x2 1/2 −1/2
=1+ + and
x2−1 x−1 x+1
x2 dx
!
= x + (1/2) ln(x − 1) − (1/2) ln(x + 1) + c
x2 − 1
 S. SOLUTIONS TO 18.01 EXERCISES

b) Cover-up gives

x x 1/2 1/2
= = +
x2 − 1 (x − 1)(x + 1) x−1 x+1

From 8b,
x3 1/2 1/2
2
=x+ + and
x −1 x−1 x+1
x3 dx
!
= x2 /2 + (1/2) ln(x − 1) + (1/2) ln(x + 1) + c
x2 − 1
c) From 8c,
x2
!
dx = x2 /6 + x/9 + (1/27) ln(3x − 1) + c
3x − 1

d) From 8d,
x+2
!
dx = x/3 + (7/9) ln(3x − 1)
3x − 1
e) Cover-up says that the proper rational function will be written as

a1 a2 b1 b2
+ + +
x − 2 (x − 2)2 x + 2 (x + 2)2

where the coefficients a2 and b2 can be evaluted from the B’s using cover-up and the coef-
ficients a1 and b1 can then be evaluated using x = 0 and x = 1, say. Therefore, the integral
has the form

A4 x5 /5 + A3 x4 /4 + A2 x3 /3 + A1 x2 /2 + A0 x + c
a2 b2
+ a1 ln(x − 2) − + b1 ln(x + 2) −
x−2 x+2

5E-10 a) By cover-up,

1 1 −1 1/2 1/2
= = + +
x3 −x x(x − 1)(x + 1) x x−1 x+1

dx 1 1
!
= − ln x + ln(x − 1) + ln(x + 1) + c
x3−x 2 2

(x + 1) −3 4
b) By cover-up, = + . Therefore,
(x − 2)(x − 3) x−2 x−3

(x + 1)
!
dx = −3 ln(x − 2) + 4 ln(x − 3) + c
(x − 2)(x − 3)

(x2 + x + 1) −7x + 1
c) =1+ 2 . By cover-up,
x2 + 8x x + 8x

−7x + 1 −7x + 1 1/8 −57/8

= = + and
x2 + 8x x(x + 8) x x+8
 5. INTEGRATION TECHNIQUES

(x2 + x + 1)
!
= x + (1/8) ln x − (57/8) ln(x + 8) + c
x2 + 8x

d) Seeing double? It must be late.

1 1 A B C
e) = 2 = + 2+
x3 + x2 x (x + 1) x x x+1
Use the cover-up method to get B = 1 and C = 1. For A,
1 1
x = 1 =⇒ =A+1+ =⇒ A = −1
2 2
In all, ! % &
dx 1 1 1 1
!
= − + 2+ dx = − ln x + ln(x + 1) − + c
x + x2
3 x x x+1 x

x2 + 1 x2 + 1 A B C
f) = = + +
x3 2
+ 2x + x x(x + 1)2 x x + 1 (x + 1)2
By cover-up, A = 1 and C = −2. For B,
2 B 2
x = 1 =⇒ =1+ − =⇒ B = 0 and
4 2 4
x2 + 1
! % &
1 2 2
!
dx = − dx = ln x + +c
x3 + 2x2 + x x (x + 1)2 x+1

g) Multiply out denominator: (x+1)2 (x−1) = x3 +x2 −x−1. Divide into numerator:

x3 −x2 + x + 1
=1+ 3
x3 x2
+ −x−1 x + x2 − x − 1
Write the proper rational function as

−x2 + x + 1 A B C
= + +
(x + 1)2 (x − 1) x + 1 (x + 1)2 x−1

By cover-up, B = 1/2 and C = 1/4. For A,

1 1 5
x = 0 =⇒ −1 = A + − =⇒ A = − and
2 4 4

x3
! % &
1/2 1/4
!
−5/4
dx = 1+ + + dx
(x + 1)2 (x − 1) x + 1 (x + 1)2 x−1
1
= x − (5/4) ln(x + 1) − + (1/4) ln(x − 1) + c
2(x + 1)

(x2 + 1)dx 1 + 2x (2y − 1)dy

! ! !
h) = (1 − )dx = x − (put y = x + 1)
x2 + 2x + 2 2
x + 2x + 2 y2 + 1

= x − ln(y 2 + 1) + tan−1 y + c
= x − ln(x2 + 2x + 2) + tan−1 (x + 1) + c
 S. SOLUTIONS TO 18.01 EXERCISES

5E-11 Separate:
dy
= dx
y(1 − y)
Expand using partial fractions and integrate

1 1
! !
( − )dy = dx
y y−1

Hence,
ln y − ln(y − 1) = x + c

Exponentiate:
y
= ex+c = Aex (A = ec )
y−1

Aex
y=
Aex − 1
(If you integrated 1/(1 − y) to get − ln(1 − y) then you arrive at

Aex
y=
Aex + 1

This is the same family of answers with A and −A traded.)

5E-12 a) 1 + z 2 = 1 + tan2 (θ/2) = sec2 (θ/2). Therefore,

1 1 z2
cos2 (θ/2) = and sin2 (θ/2) = 1 − =
1 + z2 1+z 2 1 + z2

Next,
1 z2 1 − z2
cos θ = cos2 (θ/2) − sin2 (θ/2) = − = and
1 + z2 1 + z2 1 + z2
' '
1 z2 2z
sin θ = 2 sin(θ/2) cos(θ/2) = 2 =
1 + z2 1 + z2 1 + z2
Finally,
2dz
dz = (1/2) sec2 (θ/2)dθ = (1/2)(1 + z 2 )dθ =⇒ dθ =
1 + z2

b)

π tan π/2
dθ 2dz/(1 + z 2 )
! !
=
0 1 + sin θ 1 + 2z/(1 + z 2 )
!tan
∞
0
! ∞
2dz 2dz
= 2 + 1 + 2z
=
0 z 0 (z + 1)2
"∞
−2 ""
= =2
1 + z "0
 5. INTEGRATION TECHNIQUES

c)
π tan π/2
2dz/(1 + z 2 ) 2(1 + z 2 )dz
! ∞
dθ
! !
= =
0 (1 + sin θ)2 tan 0 (1 + 2z/(1 + z 2 ))2 0 (1 + z)4
! ∞ 2
2(1 + (y − 1) )dy
= (put y = z + 1)
1 y4
(2y 2 − 4y + 4)dy
! ∞ ! ∞
= = (2y −2 − 4y −3 + 4y −4 )dy
1 y4 1
"∞
= −2y −1 + 2y −2 − (4/3)y −3 " = 4/3 1

π ∞ ∞
2z 2dz 4zdz
! ! !
( d) sin θdθ = 2 1 + z2
=
0 0 1 + z 0 (1 + z 2 )2
"∞
−2 ""
= =2
1 + z2 " 0

5E-13 a) z = tan(θ/2) =⇒ 1 + cos θ = 2/(1 + z 2 ) and 0 ≤ θ ≤ π/2 corresponds to

0 ≤ z ≤ 1.
π/2 1
dθ2dz/(1 + z 2 )
! !
A= =
0 2(1 + cos θ)2
0 8/(1 + z 2 )2
! 1
"1
= (1/4)(1 + z 2 )dz = (1/4)(z + z 3 /3)"0 = 1/3
0

b) The curve r = 1/(1 + cos θ) is a parabola:

r + r cos θ = 1 =⇒ r + x = 1 =⇒ r2 = (1 − x)2 =⇒ y 2 = 1 − 2x
√
This is the region under y = 1 − 2x in the first quadrant:
! 1/2 √ "1/2
A= 1 − 2xdx = −(1/3)(1 − 2x)3/2 " = 1/3
"
0 0

5F. Integration by parts. Reduction formulas

xa+1 xa+1 xa+1 1

! ! !
a
5F-1 a) x ln xdx = ln xd( ) = ln x · − · dx
a+1 a+1 a+1 x
xa+1 ln x xa xa+1 ln x xa+1
!
= − dx = − + c (a &= −1)
a+1 a+1 a+1 (a + 1)2
!
b) x−1 ln xdx = (ln x)2 /2 + c (u = ln x, du = dx/x)
! ! !
x x x
5F-2 a) xe dx = xd(e ) = x · e − ex dx = x · ex − ex + c
! ! !
2 x 2 x 2 x
b) x e dx = x d(e ) = x · e − ex · 2xdx
 S. SOLUTIONS TO 18.01 EXERCISES
!
= x2 · ex − 2 xex dx = x2 · ex − 2x · ex + 2ex + c
! ! !
c) x3 ex dx = x3 d(ex ) = x3 · ex − ex · 3x2 dx

= x3 · ex − 3 x2 ex dx = x3 · ex − 3x2 · ex + 6x · ex − 6ex + c
(

eax eax n
! ax
e
! !
d) xn eax dx = xn d( )= ·x − · nxn−1 dx
a a a
eax n n
!
= ·x − xn−1 eax dx
a a
5F-3
4dx
! ! !
−1 −1 −1 −1
sin (4x)dx = x · sin (4x) − xd(sin (4x)) = x · sin (4x) − x· #
1 − (4x)2
du
!
= x · sin −1
(4x) +√ (put u = 1 − 16x2 , du = −32xdx)
8 u
1√
= x · sin−1 (4x) + u+c
4
1#
= x · sin−1 (4x) + 1 − 16x2 + c
4
5F-4
! ! !
ex cos xdx = ex d(sin x) = ex sin x − ex sin xdx
!
= e sin x − ex d(− cos x)
x

!
= ex sin x + ex cos x − ex cos xdx
!
Add ex cos xdx to both sides to get
!
2 ex cos xdx = ex sin x + ex cos x + c

Divide by 2 and replace the arbitrary constant c by c/2:

!
ex cos xdx = (ex sin x + ex cos x)/2 + c

5F-5
! !
cos(ln x)dx = x · cos(ln x) − xd(cos(ln x))
!
= x · cos(ln x) + sin(ln x)dx
!
= x · cos(ln x) + x · sin(ln x) − xd(sin(ln x))
!
= x · cos(ln x) + x · sin(ln x) − cos(ln x)dx
 5. INTEGRATION TECHNIQUES
!
Add cos(ln x)dx to both sides to get

!
2 cos(ln x)dx = x cos(ln x) + x sin(ln x) + c

Divide by 2 and replace the arbitrary constant c by c/2:

!
cos(ln x)dx = (x cos(ln x) + x sin(ln x))/2 + c

5F-6 Put t = ex =⇒ dt = ex dx and x = ln t. Therefore

! !
xn ex dx = (ln t)n dt

Integrate by parts:
! ! !
(ln t)n dt = t · (ln t)n − td(ln t)n = t(ln t)n − n (ln t)n−1 dt

because d(ln t)n = n(ln t)n−1 t−1 dt.

 Unit 6. Additional Topics
6A. Indeterminate forms; L’Hospital’s rule
sin 3x 3 cos 3x
6A-1 a) lim = lim =3
x→0 x x→0 1
cos(x/2) − 1 (−1/2) sin(x/2) (−1/4) cos(x/2)
b) lim = lim = lim = −1/8
x→0 x2 x→0 2x x→0 2
ln x 1/x
c) lim = lim =0
x→∞ x x→∞ 1

x2 − 3x − 4
d) lim = −4. Can’t use L’Hospital’s rule.
x→0 x+1
tan−1 x 1/(1 + x2 )
e) lim = lim = 1/5
x→0 5x x→0 5
x − sin x 1 − cos x sin x cos x
f) lim 3
= lim 2
= lim = lim = 1/6
x→0 x x→0 3x x→0 6x x→0 6
xa − 1 axa−1
g) lim = lim = a/b
x→1 xb − 1 x→1 bxb−1

tan(x) tan 1
h) lim = . Can’t use L’Hospital’s rule.
x→1 sin(3x) sin 3
ln sin(x/2) (1/2) cot(x/2)
i) lim = lim =0
x→π x−π x→π 1
ln sin(x/2) (1/2) cot(x/2) (−1/4) csc2 (x/2)
j) lim 2
= lim = lim = −1/8
x→π (x − π) x→π 2(x − π) x→π 2

6A-2 a) xx = ex ln x → e0 = 1 as x → 0+ because
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+
+

b) x1/x → 0 as x → 0+ because x → 0 and 1/x → ∞.

Slow way using logs:
ln x
x1/x = e x → e−∞ = 0 as x → 0+ because
ln x −∞
lim = + = −∞. (Can’t use L’Hospital’s rule.)
x→0 + x 0
c) Can’t use L’Hospital’s rule. Here are two ways:
2
(1/x)ln x → (∞)−∞ = 0 or (1/x)ln x = eln x ln(1/x) = e−(ln x) → e−∞ = 0
ln cos x
d) (cos x)1/x = e x → e0 = 1 as x → 0+ because
ln cos x − tan x
lim+ = lim+ =0
x→0 x x→0 1
ln x
e) x1/x = e x → e0 = 1 as x → ∞ because
ln x 1/x
lim = lim =0
x→∞ x x→∞ 1
 S. SOLUTIONS TO 18.01 EXERCISES

ln(1+x2 )
f) (1 + x2 )1/x = e x → e0 = 1 as x → 0+ because

ln(1 + x2 ) 2x/(1 + x2 )
lim = lim =0
x→0+ x x→0+ 1

10 ln(1+3x)
g) (1 + 3x)10/x = e x → e30 as x → 0+ because

10 ln(1 + 3x) 10 · 3/(1 + 3x)

lim+ = lim+ = 30
x→0 x x→0 1

x + cos x 1 − sin x
h) lim = (?) lim But the second limit does not exist, so L’Hospital’s
x→∞ x x→∞ 1
rule is inconclusive. But the first limit does exist after all:
x + cos x cos x
lim = lim 1 + =1
x→∞ x x→∞ x
because
| cos x| 1
≤ → 0 as x → ∞
x x
Commentary: L’Hospital’s rule does a poor job with oscillatory functions.
i) Fast way: Substitute u = 1/x.
1 sin u cos u
lim x sin = lim = lim =1
x→∞ x u→0 u u→0 1
Slower way:
1 sin(1/x) (−1/x2 ) cos(1/x)
lim x sin = lim = lim = cos 0 = 1
x→∞ x x→∞ 1/x x→∞ −1/x2
! x "1/x2 ln(x/ sin x) 1
j) = e x2 → e 6 because
sin x
ln(x/ sin x)
lim = 1/6
x→0+ x2

This is a difficult limit. Although it can be done by L’Hospital’s rule the easiest way to
work it out is with quadratic (and even cubic!) approximations:

x x 1
≈ = ≈ 1 + x2 /6
sin x x − x3 /6 1 − x2 /6

Hence,
ln(x/ sin x) ≈ ln(1 + x2 /6) ≈ x2 /6
Therefore,
1
ln(x/ sin x) → 1/6 as x → 0
x2

k) Obvious cases: If the exponents are positive (or one 0 and the other positive) then
the limit is infinite. If the exponents are both negative (or one 0 and the other negative)
then the limit is 0. Also if both exponents are 0 the limit is 1. (continued →)
 6. ADDITIONAL TOPICS

The remaining cases are the ones where a and b have opposite sign. In both cases a wins.
In other words, a < 0 implies the limit is 0 and a > 0 implies the limit is ∞. To show this
requires only one use of L’Hospital’s rule. For α > 0,

xα αxα−1
lim = lim = lim αxα = ∞
x→∞ ln x x→∞ 1/x x→∞

If a > 0 and b < 0, let c = −b > 0. Then

$c
xa/c
#
a b
x (ln x) = →∞ as x → ∞
ln x

using α = a/c > 0. The case a < 0 and b > 0 is the reciprocal so it tends to 0.
d a+1
6A-3 Using L’Hospital’s rule and x = xa+1 ln x,
da

xa+1 1 xa+1 − 1 xa+1 ln x

lim ( − ) = lim = lim = ln x
a→−1 a+1 a+1 a→−1 a + 1 a→−1 1

6A-4 x
xa+1 ln x xa+1 1
%
ta ln tdt = − 2
+
1 a+1 (a + 1) (a + 1)2
d a+1
Therefore, using L’Hospital’s rule and x = xa+1 ln x,
da
x
(a + 1)xa+1 ln x − xa+1 + 1
%
lim ta ln tdt = lim
a→−1 1 a→−1 (a + 1)2
(a + 1)x (ln x)2
a+1
= lim
a→−1 2(a + 1)
% x
2
= (ln x) /2 = t−1 ln tdt
1

6x − 4
6A-5 You can’t use L’Hospital’s rule for lim because the nominator and denom-
2 − 2x x→0
inator are not going to zero as x → 0. The first equality is true, but the second one is
false.

6A-6 a) y = xe−x is defined on −∞ < x < ∞.

y $ = (1 − x)e−x and y $$ = (−2 + x)e−x

Therefore, y $ > 0 for x < 1 and y $ < 0 for x > 1; y $$ > 0 for x > 2 and y $$ < 0 for x < 2.
Endpoint values: y → −∞ as x → −∞, because e−x → ∞ as x → −∞. By L’Hospital’s
rule,
x 1
lim y = lim x = lim x = 0
x→∞ x→∞ e x→∞ e

Critical value: y(1) = 1/e.

Graph: (−∞, −∞) & (1, 1/e) ' (∞, 0).
 1/ e
S. SOLUTIONS TO 18.01 EXERCISES
1 2

Concave up on: 2 < x < ∞, concave down on: −∞ < x < 2. y = xe-x
b) y = x ln x is defined on 0 < x < ∞.

y $ = ln x + 1, y $$ = 1/x

Therefore, y $ > 0 for x > 1/e and y $ < 0 for x < 1/e; y $$ > 0 for all x > 0.
Endpoint values: As x → ∞, both x and ln x tend to infinity, so y → ∞. By L’Hospital’s
rule,
ln x 1/x
lim x ln x = lim+ = lim+ =0 1/e
x→0+ x→0 x x→0 1 1
1/e
Critical value: y(1/e) = −1/e. y = x ln x

Graph: (0, 0) ' (1/e, −1/e) & (∞, ∞), crossing zero at x = e. Concave up for all x > 0.

c) y = x/ ln x is defined on 0 < x < ∞, except for x = 1.

ln x − 1
y$ =
(ln x)2

Thus, y $ < 0 for 0 < x < 1 and for 1 < x < e and y $ > 0 for x > e;
Endpoint values: y → 0 as x → 0+ because x → 0 and 1/ ln x → 0. L’Hôpital’s rule
implies
x 1
lim = lim =∞
x→∞ ln x x→∞ 1/x

Singular values: y(1+ ) = ∞ and y(1− ) = −∞.

Critical value: y(e) = e.
Graph: (0, 0) ' (1, −∞) ↑ (1, ∞) ' (e, e) & (∞, ∞).
To determine where it is convex and concave:
2 − ln x
y $$ =
x(ln x)3

We have y $$ = 0 when ln x = 2, i.e., when x = e2 . From this,

e
y $$ < 0 for 0 < x < 1 and for x > e2 and y $$ > 0 for 1 < x < e2 .
Concave (down) on: 0 < x < 1 and x > e2
Convex (concave up) on: 1 < x < e2 1 e

y= x
2 2
Inflection point: (e , e /2) (too far to the right to show on the
graph) ln x
 6. ADDITIONAL TOPICS

6B. Improper integrals

dx 1
6B-1 √ < √ for x > 0
3
x +5 x3
% ∞ % ∞
dx dx
√ < 3/2
which converges, by INT (4)
1
3
x +5 1 x
Answer: converges

x2 dx 1
6B-2 3
* if x >> 1, so we guess divergence.
x +2 x
x2 dx 1
3
> if 2x3 > x3 + 2 or x3 > 2 or x > 21/3
x +2 2x
% ∞ 2
x dx 1 ∞ dx
%
> , which diverges by INT (4).
2 x3 + 2 2 2 x
% ∞ 2 % ∞ 2
x dx x dx
3+2
diverges, by comp.test, and so does 3+2
by INT (3).
2 x 0 x
1
dx
%
6B-3 integrand blows up at x = 0
0 x3 + x2
1 1 1
3 2
= 2 ∼ 2 when x * 0
x +x x (x + 1) x
So we guess divergence.
1 1
> 2 if 2x2 > x3 + x2 or x2 > x3 ; true if 0 < x < 1.
x3 + x2 2x
% 1
dx 1 1 dx
%
=⇒ 3 2
> which diverges by INT (6)
0 x +x 2 0 x2
1
dx
%
6B-4 √ blows up at x = 1
0 1 − x3
1 1 1
√ = & ∼ √ √ for x * 1
1−x 3 2
(1 − x)(1 + x + x ) 3 1−x
So we guess convergence.
1 1
√ < √ if x3 < x OK if 0 < x < 1
1 − x3 1−x
1 1
√ converges by INT (6), so √ also converges by comp.test.
1−x 1 − x3
∞
e−x dx
%
6B-5 is improper at both ends.
0 x
At the ∞ end it converges, since
% ∞
e−x dx −x
< e if x > 1 and e−x converges.
x 0
 S. SOLUTIONS TO 18.01 EXERCISES

e−x dx 1
At the 0 end: trouble! ∼ . So we guess divergence.
x x
% ∞ −x
e−x dx 1 e dx 1 ∞ dx
%
> on 0 < x < 1 =⇒ > divergent.
x 4x 0 x 4 0 x
% ∞ −x
e dx
=⇒ diverges —one end is infinite (the 0 end!)
0 x

∞
ln xdx
%
6B-6
1 x2
Here ln x grows so slowly, that we suspect convergence.
ln x x
< 2 is not convergent.
x2 x
ln x 1 ln x
How about 2 < 3/2 ? if x >> 1. This says √ < 1 if x >> 1 and this is true, since
x x x
ln x 1/x 2
lim √ = lim √ = lim √ = 0
x→∞ x x→∞ 1/2 x x→∞ x
% ∞
ln xdx x
=⇒ 2
< 3/2 converges, by INT (4).
1 x x
% ∞
ln xdx
So converges by comp.test.
1 x2
These have been written out in detail, to review the reasoning. Your own solutions don’t
have to be so detailed.

% ∞ '∞
6B-7 a) e−8x dx = −(1/8)e−8x'0 = 1/8 convergent
0
∞ '∞
x−n+1 '' 1
%
b) x−n dx = = convergent (n > 1)
1 −n + 1 '1 n−1
c) divergent
% 2 '2
xdx
d) √ = −(4 − x2 )1/2 ' = 2 convergent
'
4−x 2 0
0
% 2 √
dx '2
e) √ = −2(2 − x)1/2 ' = 2 2 convergent
'
0 2−x 0
% ∞
dx '∞
f) 2
= −(ln x)−1 'e = 1 convergent
e x(ln x)
% 1 '1
dx 2/3 ' 3
g) 1/3
= (3/2)x ' = convergent
0 x 0 2
h) divergent (at x = 0)
i) divergent (at x = 0)
j) Convergent because ln x tends to −∞ more slowly than any power as x → 0+ .
 6. ADDITIONAL TOPICS

Integrate by parts % 1
1
ln xdx = x ln x − x|0 = −1
0

(Need L’Hospital’s rule to check that x ln x → 0 as x → 0+ .)

k) Convergent because |e−2x cos x| < e−2x . Evaluate by integrating by parts twice (as
in E30/4). % ∞ '∞
1 2 '
e−2x cos xdx = e−2x sin x − e−2x cos x'' = 2/5
0 5 5 0

∞
dx
%
∞
l) divergent ( = ln ln x|e = ∞)
e x ln x
∞
dx
%
'∞
m) = (−1/2)(x + 2)−2 '0 = 1/8 convergent
0 (x + 2)3
n) divergent (at x = 2)
o) divergent (at x = 0)
p) divergent (at x = π/2)
( x t2 2
0
e dt ex 1
6B-8 a) lim = lim 2 = lim = 0 (L’Hospital and FT2)
x→∞ ex2 x→∞ 2xex x→∞ 2x
( x t2 2
e dt ex 1 1
b) lim 0 x2 = lim 2 2 = lim =
x→∞ e /x x→∞ 2x2 ex − ex /x2 x→∞ 2 − (1/x2 ) 2
% x
2
c) lim e−t dt = A a finite number > 0 because the integral is convergent. But
x→∞ 0
x2
e → ∞, so the whole limit tends to infinity.
( 1 −1/2 √
a x dx −1/ a
d) = lim √ = lim = lim 2a = 0 (L’Hospital and FT2)
a→0+ 1/ a a→0+ (−1/2)a−3/2 a→0+
( 1 −3/2
x dx −a−3/2
e) = lim a √ = lim = 2 (L’Hospital and FT2)
a→0+ 1/ a a→0+ (−1/2)a−3/2

( b dx
b
dx
%
0 1−sin x
( f) lim (b − π/2) = lim
b→(π/2)+ 0 1 − sin x b→(π/2)+ 1/(b − π/2)
1/(1 − sin b)
= lim
b→(π/2)+ −1/(b − π/2)2

(b − π/2)2
= lim
b→(π/2)+ sin b − 1
2(b − π/2)
= lim
b→(π/2)+ cos b
2
= lim = −2
b→(π/2)+ − sin b
 S. SOLUTIONS TO 18.01 EXERCISES

6C. Infinite Series

1 1 1 1 1 5
6C-1 a) 1 + + + ··· = 1 + + 2 + ··· = 1 =
5 25 5 5 1− 5
4
1 1 1 1 6B
b) 8 + 2 + + · · · = 8(1 + + 2 + · · · ) = 8( )=
2 4 4 1 − 41 3
1 1 1 4 4 1 1 5
c) + + · · · = (1 + + ( )2 + ·) = ( )=
4 5 4 5 5 4 1 − 54 4
1 1 4
d) 0.4444 · · · = 0.4(1 + 0.1 + 0.12 + 0.13 + · · · ) = 0.4( ) = 0.4( )=
1 − 0.1 0.9 9
1
e)0.0602602602 · · · = 0.0602(1 + 0.001 + 0.000001 + · · · ) = 0.0602( )
1 − 0.001
0.0602 301
= =
0.999 4995

6C-2 a) 1 + 1/2 + 1/3 + 1/4 + · · ·

% 2 % 3
1 1 1
clearly, we have 1 > dx, > dx, · · ·
1 x 2 2 x
% 2 % 3 % 4 % 5
1 1 1 1 1 1 1
so we will have 1 + + + + · · · > dx + dx + dx + dx + · · · =
% ∞ 2 3 4 1 x 2 x 3 x 4 x
1
dx, which is divergent, so the infinite series is divergent.
1 x
∞
) 1
b) p
n=1
n
n+1
1 dx
%
Case 1: p ≤ 1. >
np n xp
∞ ∞
1 dx
) %
=⇒ p
> , which is divergent, so the infinite series is divergent.
n=1
n 1 xp
Case 2: p > 1
% n ∞ % ∞
1 dx ) 1 dx
p
< p
=⇒ p
< 1 + , which is convergent. So the infinite series is
n n−1 x n=1
n 1 xp
convergent.
c) 1/2 + 1/4 + 1/6 + 1/8 + · · · = (1/2)(1 + 1/2 + 1/3 + 1/4 + · · · ). So from a), the
series is divergent.
d) 1 + 1/3 + 1/5 + 1/7 + · · ·
1 > 1/2, 1/3 > 1/4, 1/5 > 1/6, 1/7 > 1/8, · · ·
So 1 + 1/3 + 1/5 + 1/7 + · · · > 1/2 + 1/4 + 1/6 + 1/8 + · · ·which is divergent from c) Thus
the series diverges.
 6. ADDITIONAL TOPICS

1 1 1 1 1 1 1 1 1
1− + − + − · · · = (1 − ) + ( − ) + ( − ) + · · ·
2 3 4 5 2 3 4 5 6
1 1 1
( e) = + + + ···
1·2 3·4 5·6
1 1 1
< 2 + 2 + 2 + ···
1 3 5
1 1 1 1 1
< 2 + 2 + 2 + 2 + 2 + ···
1 2 3 4 5

which is convergent by b). So the infinite series is convergent.

f) n/n! = 1/(n − 1)! < 1/(n − 1)(n − 2) * 1/n2 for n >> 1. So convergent by
comparison with b).
√
g) Geometric series with ratio ( 5 − 1)/2 < 1, so the series is convergent.
√ √
h) Geometric series with ratio ( 5 + 1)(2 5) < 1, so the series is convergent.
*
i) Larger than 1/n for n ≥ 3, so divergent by part b).
j) ln n grows more slowly than any power. For instance,

ln n
ln n < n1/2 =⇒ < n−3/2 for n >> 1
n2
*
The series n−3/2 converges by part b), so this series also converges.
n+2 1 * −3
k) Converges because * 3 , and n converges by part b).
n4 − 5 n
(n + 2)1/3 n1/3 1 *
l) 4 1/3
* 4/3 * . Therefore this series diverges by comparison with 1/n.
(n + 5) n n
m) Quadratic approximation implies cos(1/n) ≈ 1 − 1/2n2 and hence

1
ln(cos ) * −1/2n2 as n → ∞
n

1/n2 from part b).

*
Hence the series converges by comparison with
n) e−n beats n2 by a large margin. For example, L’Hospital’s rule implies

e−n/2 n2 → 0 as n → ∞
)
Therefore for large n, n2 e−n = n2 e−n/2 e−n/2 < e−n/2 and e−n/2 is a convergent geo-
metric series. Therefore the original series converges by comparison.
√
n
o) Just as in part (n), e− beats n2 by a large margin. L’Hospital’s rule implies

e−m/2 m4 → 0 as m → ∞
√
Put m = n to get √
n/2 2
e− n →0 as n → ∞
 S. SOLUTIONS TO 18.01 EXERCISES
√ √ √ √
Therefore for large n, n2 e− n
= n2 e − n/2 − n/2
e < e− n/2
. Moreover, we also have
√
n
e− < 1/n2 n large
) √ )
n/2
Thus the sum is dominated by e− < 1/n2 and is convergent by comparison with
part b).

6C-3 a)
n
dx 1 1 1 1
%
ln n = < Upper sum = 1 + + · · · < 1 + + ···
1 x 2 n−1 2 n

In other words,
1 1
ln n < 1 + + ···
2 n
On the other hand,
n
dx 1 1
%
ln n = > Lower sum = + · · ·
1 x 2 n
Adding 1 to both sides,
1 1
1 + ln n > 1 + + ···
2 n

b) Need at least ln n = 999

Time > 10−10 e999 ≈ 7 × 10423 seconds

This is far, far longer than the estimated time from the “big bang.”
 Unit 7. Infinite Series

7A: Basic Definitions

7A-1
∞ ∞ " #n
! 1 ! 1 1 4
a) Sum the geometric series: n
= = = .
0
4 0
4 1 − (1/4) 3

b) 1 − 1 + 1 − 1 + . . . + (−1)n + . . . diverges, since the partial sums sn are successively

1, 0, 1, 0, . . . , and therefore do not approach a limit.
n−1
c) Diverges, since the n-th term n does not tend to 0 (using the n-th term test for
divergence).
∞
!
1 1 1 1
d) The given series = ln 2 + 2 ln 2 + 3 ln 2 + . . . = ln 2(1 + 2 + 3 + . . . ); but 1/n
1
diverges; therefore the given series diverges.
∞ ∞
2n−1 1 ! 2n−1
" #
! 1 1 1
e) = , geometric series with sum = · 3 = 1.
1
3n 3 1 3n−1 3 1 − (2/3) 3
∞ " #n
! −1 1 3
f) series = = = (sum of a geometric series)
0
3 1 − (−1/3) 4
" #
1
7A-2 .21111 . . . = .2+.01+.001+. . . = .2+.01(1+ 10 + 1012 +. . . = .2+.01( 1−1/10
1
= 19
90 .

7A-3 Geometric series; converges if |x/2| < 1, i.e., if |x| < 2, or equivalently, −2 < x < 2.
7A-4 " # " # " #
1 1 1 1 1 1
a) Partial sum: sm = √ −√ + √ −√ + ...+ √ − √
1 2 2 3 m m+1
1
= 1− √ → 1 as m → ∞. Therefore the sum is 1.
m+1
∞ " ∞ " #
1 1/2 −1/2 ! 1 1 ! 1 1
b) = + ; therefore = − .
n(n + 2) n n+2 1
n(n + 2) 2 0
n n+2
The m-th partial sum of the series is
" # " #
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
sm = − + − + − + − +. . .+ − = 1+ − − ,
2 1 3 2 4 3 5 4 6 m m+2 2 2 m+1 m+2
since all other terms cancel.
3
Therefore sm → 4 as m → ∞, so the sum is 3/4.
" # " #
2 2 2 2 2 2
7A-5 The distance the ball travels is h+ h+ h+ h + h + ... ;
3 3 3 3 3 3
the successive terms give the first down, the first up, the second down, and so on. Add h
to the series to make the terms uniform; you get a geometric series to sum:
" " #
2 2 1
2 h + 2h/3 + (2/3) h + . . . ) = 2h(1 + 2/3 + (2/3) + . . . ) = 2h = 6h.
1 − 2/3
Subtracting the h that we added on gives: the total distance traveled = 5h.
 S. SOLUTIONS TO 18.01 EXERCISES

7B: Convergence Tests

7B-1
∞ %∞
x 1
$
2
a) = ln(x + 4) = ∞; divergent
0 x2 + 4 2 0
$ ∞ %∞
1 −1 π
b) 2+1
= tan x = ; convergent
0 x 0 2
$ ∞ %∞
1
c) √ = 2(x + 1)1/2 = ∞; divergent
0 x + 1 0
$ ∞ %∞
ln x 1
d) = (ln x)2 = ∞; divergent
1 x 2 1
%∞
(ln x)1−p
$ ∞
1
e) = , if p %= 1: divergent if p < 1, convergent if p > 1
2 (ln x)p · x 1−p 2
$ ∞ %∞
dx
If p = 1, = ln(ln x) = ∞. Thus series converges if p > 1, diverges if p ≤ 1.
2 ln x 2
%∞
x1−p
$ ∞
1
f) = , if p %= 1; diverges if p < 1, converges if p > 1.
1 xp 1−p 1
$ ∞ %∞
dx
If p = 1, = ln x = ∞; thus series converges if p > 1, diverges if p ≤ 1.
1 x 1
7B-2
∞
! 1 n2 1
a) Convergent; compare with : = → 1 as n → ∞
1
n2 2
n + 3n 1 + 3/n
!1 n 1
b) Divergent; compare with : √ = √ → 1, as n → ∞
n n+ n 1 + 1/ n
!1 n 1
c) Divergent; compare with : √ = & → 1, as n → ∞
n 2
n +n 1 + 1/n
∞ " #
! 1 2 1 sin h
d) Convergent; compare with 2
: lim n sin 2
= lim = 1
1
n n→∞ n h→0 h
∞ √
! 1 n3/2 n n2 1
e) Convergent; compare with : 2
= 2
= → 1 as
1
n 3/2 n + 1 n + 1 1 + 1/n2
n→∞
∞
ln n 1 ! 1
f) Divergent, by comparison test : > ; diverges
n n 1
n
! 1 n2 · n2 n4
g) Convergent; compare with : = 4 → 1 as n → ∞
n2 4
n −1 n −1
! 1 4n · n3 1
h) Divergent; compare with : = → 1
4n 4n4 + n2 1 + 1/4n2
 7. INFINITE SERIES

∞
! ∞
!
7B-3 By the mean-value theorem, sin x < x, if x > 0; therefore sin an < an ; so the
0 0
series converges by the comparison test.
7B-4
n + 1 2n
" #
n+1 1 1
a) By ratio test, n+1 · = · → as n → ∞; convergent
2 n n 2 2
2n+1 n! 2
b) By ratio test, · = → 0 as n → ∞; convergent
(n + 1)! 2n n+1
2n+1 1 · 3 · · · · · 2n − 1 2
c) By ratio test, · = → 0 as n → ∞;
1 · 3 · · · · · 2n + 1 2n 2n + 1
convergent
(n + 1)!2 (2n)! (n + 1)2 1
d) By ratio test, · 2
= → as n → ∞; convergent
(2n + 2)! n! (2n + 2)(2n + 1) 4
√
1 n ! 1
e) Ratio test fails: √ · → 1 as n → ∞; but √ diverges; therefore the
n+1 1 n
series is not absolutely convergent.
(n + 1)! nn nn 1 1
f) By ratio test, n+1
· = n
= n
→ < 1 as n → ∞;
(n + 1) n! (n + 1) (1 + 1/n) e
convergent
1 n2 ! 1
g) Ratio test fails: 2
· → 1 as n → ∞; but converges; therefore the
(n + 1) 1 n2
series is absolutely convergent.
! 1 !1
h) Ratio test fails: √ diverges, by limit comparison with ; therefore the
n2 + 1 n
series is not absolutely convergent.
! n
i) Ratio test fails: diverges by the n-th term test; therefore the series is not
n+1
absolutely convergent

7B-5
1
e) conditionally convergent: terms alternate in sign, √ → 0, decreasing;
n
1
h) conditionally convergent: terms alternate in sign, √ → 0, decreasing;
n2 + 1
(−1)n n
i) divergent, by the n-th term test: lim %= 0 .
n→∞ n+1
7B-6 In all of these, we are using the ratio test.
|x|n+1
" #
n n
a) · = |x| · → |x| as n → ∞; converges for |x| < 1; R = 1
n + 1 |x|n n+1
#2
2n+1 |x|n+1 n2
"
n
b) · = 2|x| · → 2|x| as n → ∞;
(n + 1)2 2n |x|n n+1
converges for 2|x| < 1 or |x| < 1/2; R = 1/2
 S. SOLUTIONS TO 18.01 EXERCISES

(n + 1)!|x|n+1
c) = (n + 1)|x| →∞ as n → ∞; converges only for |x| = 0; R = 0
n!|x|n
|x|2(n+1) 3n |x|2 |x|2 |x|2
d) n+1
· 2n = → as n → ∞; converges for < 1,
3 √|x| √ 3 3 3
that is, for |x| < 3; R = 3
√
|x|2n+3 2n n |x|2 |x|2
'
n
e) √ · 2n+1
= · → as n → ∞; converges for
2n+1 n + 1 |x| 2 n+1 2
|x|2 √ √
< 1 or |x| < 2; R = 2
2
(2n + 2)!|x|2n+2 n!2 (2n + 2)(2n + 1)
f) 2
· 2n
= |x|2 · → 4|x|2 as n → ∞;
(n + 1)! (2n)!|x| (n + 1)2
converges for 4|x|2 < 1, or |x| < 1/2; R = 1/2
|x|n+1 ln n ln n
g) · n = |x| · → |x| as n → ∞; converges for |x| < 1; R = 1
ln(n + 1) |x| ln(n + 1)
ln x 1/x
(By L’Hospital’s rule, lim = lim = 1.)
x→∞ ln(x + 1) x→∞ 1/(x + 1)

22n+2 |x|n+1 n! 22 |x|

h) · 2n n = → 0 as n → ∞; converges for all x; R = ∞
(n + 1)! 2 |x| n+1

7C: Taylor Approximations and Series

7C-1

(a) y = cos x y $ = − sin x y $$ = − cos x y (3) = sin x y (4) = cos x, . . .

y(0) = 1 y $ (0) = 0 y $$ (0) = −1 y (3) (0) = 0 y (4) (0) = 1, . . .
a0 = 1 a1 = 0 a2 = −1/2! a3 = 0 a4 = 1/4! . . .

The pattern then repeats with the higher coefficients, so we get finally

x2 x4 x6 (−1)n x2n
cos x = 1 − + − + ... + + ...
2! 4! 6! (2n)!
(b)

y = ln(1 + x) y $ = (1 + x)−1 y $$ = −(1 + x)−2 y (3) = 2!(1 + x)−3 y (4) = −3!(1 + x)−4 , . . .
y(0) = 0 y $ (0) = 1 y $$ (0) = −1 y (3) (0) = 2! y (4) (0) = −3!, . . .
a0 = 0 a1 = 1 a2 = −1/2 a3 = 1/3 a4 = −1/4 . . .

x2 x3 x4 (−1)n−1 xn
ln(1 + x) = x − + − + ...+ + ...
2 3 4 n
 7. INFINITE SERIES

(c) Typical terms in the calculation are given.

" #" #
1/2 1 −1 (−1)(−3)(−5)
y = (1 + x) $$
y = (1 + x)−3/2 y (4) = (1 + x)−7/2
2 2 24
−1 (−1)3 (1 · 3 · 5)
y(0) = 1 y $$ (0) = 2 y (4) (0) =
2 24
1·3·5
a0 = 1 a2 = −1/8 a4 = − 4
2 4!

√ x x2 1 · 3 · 5 · · · · (2n − 3) n
1+x=1+ − + . . . + (−1)n−1 x + ...
2 8 2n · n!
One gets the same answer by using the binomial formula; this is the way to remember the
series:
( 1 )(− 12 ) 2 ( 12 )(− 12 )(− 23 ) 3
" #
1
(1 + x)1/2 = 1 + x+ 2 x + x + ...
2 2! 3!

x3 x5
7C-2 sin x = x − + + R6 (x).
3! 5!
(We could use either R5 (x) or R6 (x), since the above polynomial is both T5 (x) and T6 (x),
but R6 (x) gives a smaller error estimation if |x| < 1, since it contains a higher power of x.)

sin(7) c 7 − cos c
R6 (1) = ·1 = , for some 0 < c < 1. Therefore
7! 7!
1 1
|R6 (1) ≤ = < .0002
7! 5040

1 1
Thus sin 1 ≈ 1 − + ≈ .84166; the true value is sin 1 = .84147, which is within the
3! 5!
error predicted by the Taylor remainder.

7C-3 Since f (x) = ex , the n-th remainder term is given by

f (n+1) (c) n+1 ec 3 5

Rn (1) = ·1 = < < 5 if n + 1 = 8.
(n + 1)! (n + 1)! (n + 1)! 10

Therefore we want n = 7, i.e., we should use the Taylor polynomial of degree 7; calculation
gives e ≈ 1 + 1 + 1/2 + 1/6 + 1/24 + 1/120 + 1/720 + 1/5040 = 2.71825 . . . , which is indeed
correct to 3 decimal places.

7C-4 Using as in 7C-2 the remainder R3 (x), rather than R2 (x), we have
( (4)
|x|4 (.5)4
( ( (
( cos (c) 4 ( ( cos c 4 (
|R3 (x)| = (( x (( = (( x (( ≤ ≤ = .0026.
4! 4! 4! 24

So the answer is no, if |x| < .5. (If the interval is shrunk to |x| < .3, the answer will be yes,
since (.3)4 /24 < .001.)
 S. SOLUTIONS TO 18.01 EXERCISES

7C-5 By Taylor’s formula for ex , substituting −x2 for x,

2 x4 ec (−x2 )3
e−x = 1 − x2 + + , 0 < c < .5
2! 3!

x6
Since 0 < ec < 2, the remainder term is < ; integrating,
3

.5 %.5
x3 x5
$ )
−x2
e dx = x− + + error = .461 + error;
0 3 10 0

.5 %.5
x6 x7
$
where |error| < = = .00028 < .0003; thus the answer .461 is good to 3
0 3 21 0
decimal places.

7D: Power Series

7D-1

22 2 2n
(a) e−2x = 1 − 2x + x + . . . + (−1)n xn + . . . ,
2! n!
by substituting −2x for x in the series for ex .

√ x x2 x3 (−1)n xn
(b) cos x = 1 − + − + ...+ + ...
2! 4! 6! (2n)!

(2x)2 (2x)4
" ) %#
1 1
(c) sin2 x = (1 − cos 2x) = 1− 1− + − ...
2 2 2! 4!
2 4 n−1 2n
" #
1 (2x) (2x) (−1) (2x)
= − + ...+ + ...
2 2! 4! (2n)!

(d) Write the series for 1/(1 + x), differentiate and multiply both sides by −1:

1
= 1 − x + x2 − x3 + . . . + (−1)n+1 xn+1 + . . .
1+x
1
= 1 − 2x + 3x2 + . . . + (−1)n (n + 1)xn + . . .
(1 + x)2

1
(e) D tan−1 x = = 1 − x2 + x4 − x6 + . . . + (−1)n x2n + . . . ,
1 + x2

by substituting x2 for x in the series for 1/(1 + x); (cf. (d) above). Now integrate both sides
of the above equation:

x3 x5 (−1)n x2n+1
tan−1 x = x − + − ...+ + . . . + C;
3 5 2n + 1
 7. INFINITE SERIES

Evaluate the constant of integration by putting x = 0, one gets 0 = 0 + C, so C = 0.

1
(f) D ln(1 + x) = = 1 − x + x2 − x3 + . . . + (−1)n+1 xn+1 + . . .
1+x
x2 x3 (−1)n xn+1
ln(1 + x) = x − + − ...+ + . . . + C,
2 3 n+1
by integrating both sides. Find C by putting x = 0, one gets C = 0.
x2 x3 x4
(g) ex = 1 + x + + + + ...
2! 3! 4!
x2 x3 x4
e−x = 1−x+ − + + ...
2! 3! 4!
x2 x4 x2n
Adding and dividing by 2 gives: cosh x = 1 + + + ...+ + ...
2! 4! (2n)!

7D-2
x x2 x3 x2
" #
1 1/9 1 1 x
a) = = 1 − + 2 − 3 + ... = − 2 + 3 − ...
x+9 1 + x/9 9 9 9 9 9 9 9
x2 x3 xn
b) ex = 1 + x + + + ...+ + . . . ; substituting −x2 for x gives
2! 3! n!
2 x4 (−1)n x2n
= 1 − x2 +
e−x − ...+ + ...
2! n!
x2 x3 x2
" 3
x3
" #" # #
x x
c) e cos x = 1 + x + + +... 1− + ... = 1+x+ − + ...
2 6 2 6 2
x3
= 1+x− + . . . ; the terms in x2 cancel.
3
sin t t2 t4 (−1)n t2n
d) = 1 − + + ... + + ...
t 3! 5! (2n + 1)!
x
sin t x3 x5 (−1)n x2n+1
$
dt = x − + − ...+ + ...
0 t 3 · 3! 5 · 5! (2n + 1) · ((2n + 1)!
2
/2 t2 t4 t6
e) e−t = 1− + 2 − 3 + ...
2 2 · 2! 2 · 3!
x
x3 x5 x7 (−1)n x2n+1
$
2
/2
e−t dt = x − + − + . . . + + ...
0 3 · 2 5 · 22 · 2! 7 · 23 · 3! (2n + 1) · 2n · n!
1 −1
f) = = −1 − x3 − x6 − . . . − x3n − . . .
x3 −1 1 − x3
g) y = cos2 x ⇒ y $ = −2 cos x sin x = − sin 2x; substituting 2x into the series for sin x,
23 x3 25 x5
−y $ = −2x +
+ . . . ; integrating,
3! 5!
23 x4 25 x6 (−1)n 22n−1 x2n
y = cos2 x = −x2 + − + ...+ + ...+ C ;
4! 6! (2n)!
x4
Since y(0) = 1, we see that C = 1, so cos2 x = 1 − x2 + − ...
3
 S. SOLUTIONS TO 18.01 EXERCISES

x3
" # " #
sin x 1
h) Method 1: = (sin x) = x− + . . . (1 + x + x2 + x3 + . . . )
1−x 1−x 6
x3
" #
5
= x + x2 + x3 − + ... = x + x2 + x3 + . . .
6 6
Method 2: divide 1 − x into x − x3 /6 + . . . , as done on the left below:

x + x2 + 5x3 /6 + . . . x + x3 /3 + . . .
1−x x − x3 /6 . . . 1 − x2 /2 x − x3 /6 + . . .
x − x2 x − x3 /2
x2 − x3 /6 + ... x3 /3 + . . .
x2 − x3
5x3 /6 + . . .

i) Method 1: Calculating successive derivatives gives:

y = tan x, y $ = sec2 x, y $$ = 2 sec2 x tan x, y (3) = 2(2 sec2 x tan x · tan x + sec2 x · sec2 x)
y(0) = 0, y $ (0) = 1, y $$ (0) = 0, y (3) (0) = 2,

so the Taylor series starts

2x3 x3
tan x = x + + ... = x + + ...
3! 3
sin x
Method 2: tan x = ; divide the cos x series into the sin x series (done on the
cos x
right above) — this turns out to be easier here than taking derivatives!

7D-3
1 − cos x 1 − (1 − x2 /2 + . . . ) x2 /2 + . . . 1
a) = = → as x → 0.
x2 x2 x2 2
x − sin x x − (x − x3 /6 + . . . ) x3 /6 + . . . 1
b) 3
= 3
= 3
→ as x → 0
x x x 6
c) (1 + x)1/2 = 1 + x/2 − x2 /8 + . . . ⇒ (1 + x)1/2 − 1 − x/2 = −x2 /8 + . . .
sin x = x − x3 /6 + . . . ⇒ sin2 x = x2 + . . .
(1 + x)1/2 − 1 − x/2 −x2 /8 + . . . −1
Therefore, 2 = → as x → 0.
sin x x2 + . . . 8
d) cos u − 1 = −u2 /2 + . . . ; ln(1 + u) − u = −u2 /2 + . . . ;
cos u − 1 −u2 /2 + . . .
Therefore, = → 1 as u → 0.
ln(1 + u) − u −u2 /2 + . . .
4 E. 1801 EXERCISES

Answers
8A-1 a) 1/3 b) 1/6
8A-2 a) 1/3 b) 7/36
8A-3 a) 11/24 b) 13/24
8A-4 a) 1/3 b) 2/3 c) P(X is 2 or 3)=6/16; P(X is neither 1 nor 4) = 7/16
8A-5 e−10 (1 + 10 + 102 /2! + . . . + 105 /5!) = .067
8A-6 e−5 (1 + 5 + 52 /2!) = .125 = P (X ≤ 2). So P (X ≥ 3) = .875
8A-7 The mean is .2, so 1.61 errors/page
a) e−1.61 (1 + 1.61) = .522 =P(at most 1 error/page)
b) average no. errors in 3 pages is 4.83; P(3 or less)= .29
8A-8 a) .0018 b) .957

8B-1 E(X) = 2; P (1 < X < 3) = e−1/2 − e−3/2 = 38%.

8B-2 e−5/2 ≈ 8%; 1 − e−5 ≈ 99%.
8B-3 e−6 − e−12 ≈ 0.2%.
8B-4 e−2.4 ≈ 9%.
8B-5 e−t0 /100 = .9 ⇒ t0 = −100 ln 9 ≈ 10.5 days.
8B-6 e−1 ≈ 37%.

8D-1 .9938 − .9332 = .0608 1 − .8413 = .1587;

.8413 − (1 − .8413) = .6826; .9938 − (1 − .8413) = .8351
8D-2 P (85 < X < 135) = P (−35/36 < Z < 15/36) = .6628 − .1660 = .4968.
Ans: about 50, about 80.
8D-3 P (X ≤ 55) = P (Z ≤ −15/10) = .0668. About 20 fail.
10% ≈ P (Z ≥ 1.3) = P (Z ≤ −1.3) = P (X < 57).
8D-4 48 to 192 hrs.; 50 to 90; 38 to 54 inches

√ √ √
8E-1 σ̄ = 8/ 80 = 2/ 5, P (X̄ ≥ 40) = P (Z ≥ 5 ≈ 1 − P (Z ≤ 2.24) = 1 − .987 = .013,
so about √
1.3%.
8E-2 σ̄ ≤ .1/ 100 = .01; P (|X̄ − σ̄| ≤ .01) = P (|Z| ≤ 1) = 2P (0 ≤ Z ≤ 1) = 2(P (Z ≤
1) − 1/2) = 2(.84
√ − .50) = .68; about 68%
8E-3 e = 100/ n ≈ 14, 10, 1; n = 10, 000/e2 ≈ 400, 1111.
8E-4 about 49%-55%
8E-5 Less than about 955 or more than about 1045 heads would show coin unfair.
8E-6 4e−3 ≈ 20%
8E-7 51%-53%
 18.01 REVIEW PROBLEMS AND SOLUTIONS

Unit I: Differentiation

R1-0 Evaluate the derivatives. Assume all letters represent constants, except for the
independent and dependent variables occurring in the derivative.
dp m0 dm
a) pV γ = nRT, =? b) m = ! , =?
dV 1 − v 2 /c2 dv
"
cω0 sin (2k + 1)α dR ""
c) R = , =?
α2 + β 2 dα " α=0

R1-1 Differentiate:
sin x √
a) b) sin2 ( x) c) x1/3 tan x
x+1
x2 + 2 ! !
d) √ e) cos( x2 + 1) f) cos3 ( x2 + 1)
x+1
g) tan(x3 ) h) x sec2 (3x + 1)

R1-2 Consider f (x) = 2x2 + 4x + 3. Where does the tangent line to the graph of f (x) at
x = 3 cross the y-axis?

R1-3 Find the equation of the tangent to the curve 2x2 + xy − y 2 + 2x − 3y = 20 at the
point (3,2).

R1-4 Define the derivative of f (x) . Directly from the definition, show that f ! (x) = cos x
sin h cos h − 1
if f (x) = sin x. (You may use without proof: lim = 1, lim = 0).
h→0 h h→0 h
R1-5 Find all real x0 such that f ! (x0 ) = 0:
x
a) f (x) = 2 b) f (x) = x2 + cos x
x +1
R1-6 At what points is the tangent to the curve y 2 + xy + x2 − 3 = 0 horizontal?

R1-7 State and prove the formula for (uv)! in terms of the derivatives of u and v. You
may assume any theorems about limits that you need.

R1-8 Derive a formula for (x1/5 )! .

R1-9 a) What is the rate of change of the area A of a square with respect to its side x ?

b) What is the rate of change of the area A of a circle with respect to its radius r?

c) Explain why one answer is the perimeter of the figure but the other answer is not.
 REVIEW PROBLEMS AND SOLUTIONS

x2 + 1, x ≥ 1
#
R1-10 Find all values of the constants c and d for which the function f (x) =
cx + d, x < 1
will be (a) continuous, (b) differentiable.

R1-11 Prove or give a counterexample :

a) If f (x) is differentiable then f (x) is continuous.

b) If f (x) is continuous then f (x) is differentiable.

#
sin x, x ≤ π
R1-12 Find all values of the constants a and b so that the function f (x) =
ax + b, x > π
will be (a) continuous; (b) differentiable.
sin(4x)
R1-13 Evaluate lim . (Hint: Let 4x = t.)
x→0 x

Unit 2: Applications of Differentiation

R2-1 Sketch the graphs of the following functions, indicating maxima, minima, points of
inflection, and concavity.
a) f (x) = (x − 1)2 (x + 2) b) f (x) = sin2 x, 0 ≤ x ≤ 2π
c) f (x) = x + 1/x2 d) f (x) = x + sin 2x

R2-2 A baseball diamond is a 90 ft. square. A ball is batted along the third base line at
a constant speed of 100 ft. per sec. How fast is its distance from first base changing when
a) it is halfway to third base, b) it reaches third base ?
√
R2-3 If x and y are the legs of a right triangle whose hypotenuse is 5, find the largest
value of 2x + y.

R2-4 Evaluate the following limits:

cos x sin x
a) limπ π b) lim
2 −x x
x→ 2 x→0

x17 − 4x3 + 2x2

c) lim
x→∞ 10x17 + 6x10 − x3 − 5x2

R2-5 Prove or give a counterexample:

a) If f ! (c) = 0 then f has a minimum or a maximum at c.

b) If f has a maximum at c and if f is differentiable at c, then f ! (c) = 0.

R2-6 Let f (x) = 1 − x2/3 . Then f (−1) = f (1) = 0 and yet f ! (x) %= 0 for 0 < x < 1. Find
the maximum value of f (x) on the real line, nevertheless.

Why did the standard method fail?

R2-7 A can is made in the shape of a right circular cylinder. What should its proportions
 REVIEW PROBLEMS AND SOLUTIONS

be, if its volume is to be 1 and one wants to use the least amount of metal?
 REVIEW PROBLEMS AND SOLUTIONS

R2-8 a) State the mean value theorem.

1
b) If f ! (x) = and f (1) = 1, use the mean value theorem to estimate f (2).
1 + x2
(Write your answer in the form α < f (2) < β.)

R2-9 One of these statements is false and one is true. Prove the true one, and give a
counterexample to the false one. (Both statements refer to all x in some interval a < x < b.)

a) If f ! (x) > 0, then f (x) is an increasing function.

b) If f (x) is an increasing function , then f ! (x) > 0.

R2-10 Give examples (either by giving a formula or by a carefully drawn graph ) of

a) A function with a relative minimum, but no absolute maximum on 0 < x < 1.

b) A function with a relative maximum but no absolute maximum on the interval

0 ≤ x ≤ 1.

c) A function f (x) defined on 0 ≤ x ≤ 1 , with f (0) < 0, f (1) > 0, yet with no root
on 0 ≤ x ≤ 1.

d) A function f (x) having a relative minimum at 0, but the following is false: f ! (0) = 0.

Unit 3: Integration
π 3 2
√ x2 + 1
$ $ $
R3-1 Evaluate: sin x dx, 1 + x dx, dx.
0 0 1 x2
R3-2 Egbert, an MIT nerd bicyclist, is going down a steep hill. At time t = 0, he starts
from rest at the top of the hill; his acceleration while going down is 3t2 ft./sec2 , and the hill
is 64 ft. long. If the fastest he can go without losing control is 64 ft./sec., will he survive
this harrowing experience? (A nerd bicycle has no brakes.)
%2
R3-3 Evaluate 0 x2 dx directly from the definition of the integral as the limit of a sum.
You may use the fact that
N
& N (N + 1)(2N + 1)
k2 =
6
k=1

R3-4 If f is a continuous function, find f (2) if:

$ x $ x2 $ f (x)
a) f (t)dt = x2 (1 + x) b) f (t)dt = x2 (1 + x) c) t2 dt = x2 (1 + x)
0 0 0

R3-5 The area under the graph of f (x) and over the interval 0 ≤ x ≤ a is

1 a2 a 1
− + + sin a + cos a
2 4 2 2
 REVIEW PROBLEMS AND SOLUTIONS

Find f (π/2).
√ √ √
3
R3-6 Use the trapezoidal rule to estimate the sum 3
1 + 3 2 + · · · + 106 . Is your estimate
high or low? Explain your reasoning.

R3- 7 Find the total area of the region above the graph of y = −2x and below the graph
of y = x − x2 .

R3-8 Use the trapezoidal rule with 6 subintervals to estimate the area under the curve
√ √ √ √
y = 1 + x2 , −3 ≤ x ≤ 3. (You may use: 2 ≈ 1.41, 5 ≈ 2.24, 10 ≈ 3.16.

Is your estimate too high or too low? Explain how you know.)
$ b
R3-9 Fill in this outline of a proof that F ! (x)dx = F (b) − F (a). Supply reasons.
a
$ x
a) Put G(x) = F ! (t)dt. Then G! (x) = F ! (x).
a

b) Therefore G(x) = F (x) + c, and one sees easily that c = −F (a). We’re done.

R3-10 The table below gives the known values of a function f (x):

x 0 1 2 3 4 5 6
f (x) 1 1.2 1.4 1.3 1.5 1.2 1.1

Use Simpson’s Rule to estimate the area under the curve y = f (x) between x = 0 and x = 6.

R3-11 Let f (t) be a function, continuous and positive for all t. Let A(x) be the area under
the graph of f, between t = 0 and t = x. Explain intuitively from the definition of derivative
dA
why = f (x).
dx
#
x + 1, 0 ≤ x ≤ 2 $ 4
R3-12 Let f (x) = Evaluate f (x)dx.
x − 2, 2 < x ≤ 4 0

sin x
R3-13 Suppose F (x) is a function such that F ! (x) = . In terms of F (x), evaluate the
x
sin 3x
$
indefinite integral dx ..
x
R3-14 Find a quadratic polynomial ax2 + bx + c = f (x) such that f (0) = 0, f (1) = 1, and
the area under the graph between x = 0 and x = 1 is 1.
 REVIEW PROBLEMS AND SOLUTIONS

Unit 4: Applications of integration.

R4-1 The area in the first quadrant bounded by the lines y = 1, x = 1, x = 3 and
f (x) = −x2 + 15 is rotated about the line y = 1. Find the volume of the solid thus obtained.

R4-2 Consider the circle x2 + y 2 = 4. A solid is formed with the given circle as base and
such that every cross-section cut by a plane perpendicular to the x-axis is a square. Find
the volume of this solid.
1 2
R4-3 Find the length of the arc of y = (x + 2)3/2 from x = 0 to x = 3
3
1
R4-4 For a freely falling body, s = gt2 , v = gt = 2gs. Show that:
!
2
a) the average value of v over the interval 0 ≤ t ≤ t1 is one-half the final velocity;

b) the average value of v over the interval 0 ≤ s ≤ s1 is two-thirds the final velocity.

R4-5 A bag of sand originally weighing 144 pounds is lifted at a constant rate of 3 ft./min.
the sand leaks out uniformly at such a rate that half the sand is lost when the bag has been
lifted 18 feet. find the work done in lifting the bag this distance.

R4-6 Find the area inside both loops of the lemniscate r2 = 2a2 cos 2θ .

R4-7 Calculate the volume obtained when the region (−2 ≤ x ≤ 2, 0 ≤ y ≤ x2 ) is rotated
about the y-axis.

R4-8 The table below gives the known values of a function f (x):

x 0 1 2 3 4 5 6
f (x) 1 1.2 1.4 1.3 1.5 1.2 1.1

Use Simpson’s Rule to estimate the volume obtained when the region below the graph of
y = f (x) and above the x-axis (0 ≤ x ≤ 6) is rotated about the x-axis.

R4-9 Winnie the Pooh eats honey at a rate proportional to the amount he has left. If it
takes him 1 hour to eat the first half of a pot of honey, how long will it take for him to eat
another quarter of a pot? When will he finish?

R4-10 a) Write down the definition of ln x as an integral.

b) Directly from the definition prove that:

i) ln(ax) = ln a + ln x; ii) ln x is an increasing function.

 REVIEW PROBLEMS AND SOLUTIONS

Unit 5: Integration Techniques

R5-1 Differentiate: ' (
2 1+x
a) x1/x , ex · ln(x2 ) b) tan−1 .
1−x
R5-2 Integrate:
$ $
a) sin 3 x cos 2 xdx b) ex sin xdx

R5-3 Integrate:
ex x+1 4x2
$ $ $
a) dx b) dx c) dx
1 + e2x x3 − 1 x−2
R5-4 Integrate:
x+1
$ $
a) dx b) x2 cos xdx
(1 + x2 )2
R5-5 a) Use the reduction formula

1 n−1
$ $
cosn xdx = cosn−1 x sin x + cosn−2 xdx
n n

$ π/2
to evaluate cos6 xdx.
0

b) Derive the formula for D tan−1 x from the formula for D tan x. What are the
domain and range of tan−1 x?
 SOLUTIONS TO 18.01 REVIEW PROBLEMS

Unit 1: Differentiation
R1-0.
−γnRT m0 v cω0 (2k + 1)
a) b) c)
V γ+1 c2 (1 − v 2 /c2 )3/2 β2
R1-1 √ √
(x + 1) cos x − sin x sin x cos x 1
a) b) √ c) x1/3 sec2 x + x−2/3 tan x
(x + 1)2 x 3
3x2
√ √ √
2+ 2x − 1 − sin x2 + 1 × x (3cos2 x2 + 1)(− sin x2 + 1)x
d) √ e) √ f) √
( x + 1)3 x2 + 1 x2 + 1
g) 3x2 sec2 (x3 ) h) sec2 (3x + 1) + 6x sec2 (3x + 1) tan(3x + 1)

R1-2 (0, −15)

R1-3 y = 4x − 10

R1-4 Hint: Use addition formula to expand sin(x + ∆x).

R1-5 a) x = ±1. b) x = 0

R1-6 (1, −2) and (−1, 2)

R1-7 See Simmons, sec. 3.2

R1-8 Hint: Differentiate implicitly the equation y 5 = x.

R1-10 a) c + d = 2 b) c = 2, d = 0

R1-11 a) cf. p. 75, Simmons. b) false; f (x) = |x|

R1-12 a) b = −aπ b) a = −1, b = π

sin(4x) 4 sin(t) sin(t)
R1-13 Let 4x = t. lim = tlim = 4 lim =4
x→0 x 4 →0 t t→0 t

Unit 2: Applications of Differentiation

√ √
R2-2 a) 20 5 ft/sec b) 50 2 ft/sec
1
R2-3 5 R2-4 a) 1 b) 1 c)
10
R2-5 a) false b) cf. p. 801—802 (1), (2), (3) Simmons. R2-6 1

R2-7 r = (2π)−1/3 , h = (4/π)1/3

 SOLUTIONS TO REVIEW PROBLEMS

6 3
R2-8 b) < f (2) <
5 2
R2-9 a) This is true, use mean value theorem. b) This is false ; try x3 .

R2-10 a) see graph b) f (x) must be discontinuous c) f (x) is discontinuous

d) |x|

Unit 3: Integration
14 3
R3-1 2, ,
3 2
R3-2 Yes. Hint: Find the time it takes him to reach the bottom of the hill, and find his
speed at that instant.
√
8 3 2
R3-3 R3-4 a) 16 b) 1 + c) (36)1/3
3 2
π 9
R3-5 R3-6 ≤ 75, 000, 050 R3-7 R3-8 11.46
4 2
R3-10 7.566... R3-12 6 R3-14 4x − 3x2

Unit 4: Applications of Integration

11 128
R4-1. π × 197 R4-2. R4-3. 12
15 3
R4-5. 1944 ft.lbs R4-6. 2a2 R4-7. 8π
R4-8. 8. π × 9.636̇... R4-9. Another hour; never.
! ab ! a ! ab
R4-10 b) hint: write f (t)dt = f (t)dt + f (t)dt
1 1 a

Unit 5: Integration Techniques

1 − ln x x2 2 1
R5-1 x1/x ( 2
), e ( + 2x ln(x2 )),
x x 1 + x2
cos5 x cos3 x ex
R5-2 a) − +c b) (sin x − cos x) + c
5 3 2
1 (x − 1)2
R5-3 a) tan−1 (ex ) + c b) ln 2 +c c) 2x2 + 8x + 16 ln(x − 2) + c
3 |x + x + 1|
1 x−1
R5-4 a) (tan−1 x + )+c b) x2 sin x + 2x cos x − 2 sin x + c
2 1 + x2
5π
R5-5
32