Sample Calculus Problems

Part 1: Single Variable Functions 1

Part 2: Multi-Variable Functions 75
Part 3: Sequences and Series 126
Part 4: Vector Analysis 151

Last revision: August 25, 2021

[This page is intentionally left blank.]
 Part 1: Single Variable Functions

√
4 − 3x + 1
1. Evaluate the limit lim 2 . (Do not use L’Hôpital’s Rule.)
x→5 x − 7x + 10

Solution:
√ √ √
4 − 3x + 1 (4 − 3x + 1)(4 + 3x + 1)
lim 2 = lim √
x→5 x − 7x + 10 x→5 (x2 − 7x + 10)(4 + 3x + 1)
16 − (3x + 1)
= lim √
x→5 (x2 − 7x + 10)(4 + 3x + 1)
15 − 3x
= lim √
x→5 (x − 5)(x − 2)(4 + 3x + 1)
3(5 − x)
= lim √
x→5 (x − 5)(x − 2)(4 + 3x + 1)
−3
= lim √
x→5 (x − 2)(4 + 3x + 1)
−3
= √
(5 − 2)(4 + 3 ⋅ 5 + 1)
−3
=
3⋅8
1
=−
8

Remark: “=” is the most frequently used verb in mathematics. It was introduced in 1557 by
Robert Recorde “to avoid the tedious repetition of the words ‘is equal to’.” It is important to
use the equal sign correctly.

• To introduce “ ”, the phantom equal sign, to avoid the tedious repetition of the symbol
“=” is not a good idea. The solution above should not go like:
√ √ √
4 − 3x + 1 (4 − 3x + 1)(4 + 3x + 1)
lim 2 lim √
x→5 x − 7x + 10 x→5 (x2 − 7x + 10)(4 + 3x + 1)
15 − 3x
lim √
x→5 (x − 5)(x − 2)(4 + 3x + 1)
3(5 − x)
lim √
x→5 (x − 5)(x − 2)(4 + 3x + 1)

1
• One must also not use other symbols, which have completely different meanings, in place
of “=”. The solution above should not go like:
√ √ √
4 − 3x + 1 (4 − 3x + 1)(4 + 3x + 1)
lim 2 ⇒ lim √
x→5 x − 7x + 10 x→5 (x2 − 7x + 10)(4 + 3x + 1)
15 − 3x
→ lim √
x→5 (x − 5)(x − 2)(4 + 3x + 1)
3(5 − x)
 lim √
x→5 (x − 5)(x − 2)(4 + 3x + 1)

• The equal sign always stands between two things, although sometimes one of these things
are at the end of the previous line or at the beginning of the next line. The solution above
should not start like:
√ √
(4 − 3x + 1)(4 + 3x + 1)
= lim √
x→5 (x2 − 7x + 10)(4 + 3x + 1)
√ √
(4 − 3x + 1)(4 + 3x + 1)
This begs the question: What is equal to lim √ ?
x→5 (x2 − 7x + 10)(4 + 3x + 1)

• The equal sign can be used between two functions when we deal with identities, like
x2 − 1
= x + 1 for all x =/ 1
x−1
or when we deal with equations, like

Find all x such that x2 = 4.

Therefore we can not just drop some of the limit signs in the solution above to make it
look like:
√ √ √
4 − 3x + 1 (4 − 3x + 1)(4 + 3x + 1)
lim 2 = √ 7
x→5 x − 7x + 10 (x2 − 7x + 10)(4 + 3x + 1)
⋮
−3
= √
(x − 2)(4 + 3x + 1)
−3
= √ 7
(5 − 2)(4 + 3 ⋅ 5 + 1)
−3
=
3⋅8
1
=−
8
−3
The equalities on the lines marked with 7 are not correct. √ is
(x − 2)(4 + 3x + 1)
1 −3
not equal to − because, for instance, if we let x = 1 then √ =
8 (x − 2)(4 + 3x + 1)
−3 1 1
√ = =/ − .
(1 − 2)(4 + 3 ⋅ 1 + 1) 2 8

2
 x2
2. Let m be the slope of the tangent line to the graph of y = at the point (−3, −9).
x+2
Express m as a limit. (Do not compute m.)

Solution: The slope m of the tangent line to the graph of y = f (x) at the point
(x0 , f (x0 )) is given by the limit

f (x) − f (x0 )
m = lim
x→x0 x − x0
or equivalently by the limit

f (x0 + h) − f (x0 )
m = lim .
h→0 h
Therefore two possible answers are

x2 (−3 + h)2
− (−9) − (−9)
m = lim x + 2 = lim −3 + h + 2 .
x→−3 x − (−3) h→0 h

f (x)
3. Suppose that lim f (x) =/ 0 and lim g(x) = 0. Show that lim does not exist.
x→c x→c x→c g(x)

f (x) f (x)
Solution: Assume that lim exists, and let L = lim . Then by the product
x→c g(x) x→c g(x)
rule for limits we obtain
f (x) f (x)
lim f (x) = lim ( ⋅ g(x)) = lim ⋅ lim g(x) = L ⋅ 0 = 0 .
x→c x→c g(x) x→c g(x) x→c

This contradicts the fact that lim f (x) =/ 0. Therefore our assumption cannot be
x→c
f (x)
true: lim does not exist.
x→c g(x)

4. Suppose that lim f (x) = 0 and there exists a constant K such that ∣g(x)∣ ≤ K for all x =/ c in
x→c
some open interval containing c. Show that lim (f (x)g(x)) = 0.
x→c

Solution: We have ∣f (x)g(x)∣ = ∣f (x)∣ ⋅ ∣g(x)∣ ≤ ∣f (x)∣ ⋅ K in some open interval

around c. Therefore
−K∣f (x)∣ ≤ f (x)g(x) ≤ K∣f (x)∣ .
Now applying the Sandwich Theorem and using the fact that lim ∣f (x)∣ = 0 we obtain
x→c
the result.

3
5. Determine the following limits if lim+ f (x) = A and lim− f (x) = B.
x→0 x→0

a. lim f (x2 − x) b. lim (f (x2 ) − f (x)) c. lim f (x3 − x)

x→0− x→0− x→0+

d. lim (f (x3 ) − f (x)) e. lim f (x2 − x)

x→0− x→1−

Solution: a. If x < 0, then x2 > 0 and −x > 0. Therefore x2 − x > 0 for x < 0, and
x2 −x approaches 0 from the right as x approaches 0 from the left. lim− f (x2 − x) = A.
x→0

b. Since x2 > 0 for x < 0, x2 approaches 0 from the right as x approaches 0 from
the left. Hence lim− (f (x2 ) − f (x)) = lim− f (x2 ) − lim− f (x) = A − B.
x→0 x→0 x→0

c. For 0 < x < 1, we have x3 < x and x3 − x < 0. So x3 − x approaches 0 from the left
as x approaches 0 from the right. Therefore lim+ f (x3 − x) = B.
x→0

d. Since x3 < 0 for x < 0, x3 approaches 0 from the left as x approaches 0 from the
left. Hence lim− (f (x3 ) − f (x)) = lim− f (x3 ) − lim− f (x) = B − B = 0.
x→0 x→0 x→0

e. For 0 < x < 1 we have x2 < x and x2 − x < 0. x2 − x approaches 0 from the left as
x approaches 1 from the left. Hence lim− f (x2 − x) = B.
x→1

6. Let Q be the point of intersection in the first quadrant of the circle C1 with equation
(x − 1)2 + y 2 = 1 and the circle C2 with equation x2 + y 2 = r2 . Let R be the point where the line
passing through the points P (0, r) and Q intersects the x-axis. Determine what happens to R
as r → 0+ .

4
 Solution: Subtracting x2 + y 2 = r2 from (x − 1)2 + y 2√= 1 we obtain x = r2 /2, and
substituting this back in x2 + y 2 = r2 gives us Q(r2 /2, r2 − r4 /4).

Let R(a, 0) be the coordinates of R and let S be the foot of the perpendicular from
Q to the x-axis. Since the triangles RSQ and ROP are similar we have

a − r2 /2 a
√ =
r2 − r4 /4 r

and hence
r3 /2
a= √ .
r − r2 − r4 /4
Then
r3 /2
lim+ a = lim+ √
r→0 r→0 r − r2 − r4 /4
r3 /2 √
= lim+ ( ⋅ (r + r2 − r4 /4))
r→0 r2 − (r2 − r4 /4)
√
= 2 lim+ (1 + 1 − r2 /4)
r→0
√
= 2 ⋅ (1 + 1 − 02 /4) = 4 .

Therefore R approaches the point (4, 0) as r → 0+ .

1
7. Use the formal definition of the limit to show that lim = 2.
x→1/2 x

Solution: Given ε > 0 we want to find δ > 0 such that

1 1
0 < ∣x − ∣ < δ Ô⇒ ∣ − 2∣ < ε . (⋆)
2 x
We will do this in two different ways.

1
“Solve the Inequality” Method : First we solve ∣ − 2∣ < ε for x.
x
1 1
∣ − 2∣ < ε ⇐⇒ 2 − ε < < 2 + ε
x x
The next step depends on whether 2 − ε is positive, zero or negative.

If 2 − ε > 0, that is if ε < 2, then

1 1 1
2−ε< < 2 + ε ⇐⇒ >x> .
x 2−ε 2+ε

5
If 2 − ε = 0, that is if ε = 2, then
1 1
2−ε< < 2 + ε ⇐⇒ x > .
x 2+ε

If 2 − ε < 0, that is if ε > 2, then

1 1 1
2−ε< < 2 + ε ⇐⇒ x > or − >x.
x 2+ε ε−2

6
 1
Next we choose δ in such a way that every x satisfying the condition 0 < ∣x − ∣ < δ
2
1
lies in the solution set of ∣ − 2∣ < ε, and therefore the implication in (⋆) holds. In
x
all three cases choosing a delta such that 0 < δ ≤ 1/2 − 1/(2 + ε) = ε/(4 + 2ε) achieves
this.

The Estimation Method : Suppose 0 < ∣x − 1/2∣ < δ for some δ > 0. Then
1 2∣x − 1/2∣ 2δ
∣ − 2∣ = < .
x ∣x∣ ∣x∣
At this point let us also decide to choose δ to satisfy δ ≤ 1/4. (Why 1/4?) Then
0 < ∣x − 1/2∣ < δ Ô⇒ 1/2 − δ < x < 1/2 + δ Ô⇒ 1/4 < x < 3/4 Ô⇒ 4 > 1/x > 4/3 Ô⇒
1/∣x∣ < 4 and therefore
1 2δ
∣ − 2∣ < < 8δ .
x ∣x∣
Hence for a given ε > 0 if we choose δ to satisfy δ ≤ ε/8 (as well as δ ≤ 1/4) then
1 ε
we will have ∣ − 2∣ < 8δ ≤ 8 ⋅ = ε and (∗) will hold. In conclusion, any choice of δ
x 8
satisfying 0 < δ ≤ min{ε/8, 1/4} works.

8. Show that lim (x4 + 7x − 17) = 43 using the formal definition of the limit.
x→−3

Solution: For any given ε > 0 we have to find a δ > 0 so that for all x we have
0 < ∣x − (−3)∣ < δ Ô⇒ ∣x4 + 7x − 17 − 43∣ < ε .
We have (x4 + 7x − 17) − 43 = x4 + 7x − 60 = (x + 3)(x3 − 3x2 + 9x − 20). Suppose
that 0 < ∣x − (−3)∣ < δ and δ ≤ 1. Then −4 ≤ −3 − δ < x < −3 + δ ≤ −2. In particular,

7
 ∣x∣ < 4. Therefore, using the Triangle Inequality, we obtain ∣x3 − 3x2 + 9x − 20∣ ≤
∣x∣3 + 3∣x∣2 + 9∣x∣ + 20 < 43 + 3 ⋅ 42 + 9 ⋅ 4 + 20 = 168. Now if we choose δ to satisfy
0 < δ ≤ min{ε/168, 1}, then we have
ε
∣x4 + 7x − 17 − 43∣ = ∣x4 + 7x − 60∣ = ∣x + 3∣ ⋅ ∣x3 − 3x2 + 9x − 20∣ < δ ⋅ 168 ≤ ⋅ 168 = ε
168
whenever 0 < ∣x − (−3)∣ < δ. We are done.

9. Suppose that for all 0 < ε < 1,

ε2
∣x − 1∣ < Ô⇒ ∣f (x) − 3∣ < ε (⋆)
4
and
ε
∣x − 1∣ < Ô⇒ ∣g(x) − 4∣ < ε . (⋆⋆)
35
Find a real number δ > 0 such that
1
∣x − 1∣ < δ Ô⇒ ∣f (x) + g(x) − 7∣ < .
5
1
Solution: If we take ε = in (⋆) we get
10
1 1
∣x − 1∣ < Ô⇒ ∣f (x) − 3∣ < .
400 10
1
If we take ε = in (⋆⋆) we get
10
1 1
∣x − 1∣ < Ô⇒ ∣g(x) − 4∣ < .
350 10
1
Therefore if ∣x − 1∣ < , then we have
400
1 1 1
∣f (x) + g(x) − 7∣ = ∣f (x) − 3 + g(x) − 4∣ ≤ ∣f (x) − 3∣ + ∣g(x) − 4∣ < + =
10 10 5
1
by the Triangle Inequality. Hence we can take δ = .
400
√
9−4 2
Remark: In fact, any δ ≤ works.
1225

⎧
⎪ 1
⎪
⎪1 if x = where n is a positive integer,
⎪
⎪ n
10. Let f (x) = ⎨
⎪
⎪
⎪
⎪
⎪
⎩0 otherwise.

a. Show that if c =/ 0 then lim f (x) = 0.

x→c

b. Show that lim f (x) does not exist.

x→0

8
 Solution: a. Assume c > 0. Then there is a positive integer m such that 1/m is
the closest to c among all real numbers different from c and of the form 1/n where
1
n is a positive integer. (Why?) Let δ = ∣c − ∣ > 0. Then for any ε > 0, we have
m
1
0 < ∣x − c∣ < δ Ô⇒ x =/ for any positive integer n
n
Ô⇒ f (x) = 0 Ô⇒ ∣f (x) − 0∣ = ∣0 − 0∣ = 0 < ε .

Therefore lim f (x) = 0.

x→c

Assume c < 0. Take δ = ∣c∣. Then for any ε > 0, we have

1
0 < ∣x − c∣ < δ Ô⇒ x < 0 Ô⇒ x =/ for any positive integer n
n
Ô⇒ f (x) = 0 Ô⇒ ∣f (x) − 0∣ = ∣0 − 0∣ = 0 < ε .

Therefore lim f (x) = 0.

x→c

b. Let L be a real number and assume that lim f (x) = L. Then for every ε > 0,
x→0
there exists a δ > 0 such that for all x,

0 < ∣x − 0∣ < δ Ô⇒ ∣f (x) − L∣ < ε .

If L is not 0, let ε = ∣L∣/2 > 0. Then there is a δ > 0 such that

0 < ∣x∣ < δ Ô⇒ ∣f (x) − L∣ < ∣L∣/2 .

Take x = −δ/2. Then 0 < ∣x∣ < δ is true, but ∣L∣ = ∣0 − L∣ = ∣f (x) − L∣ < ∣L∣/2 is not
true. We have a contradiction.

On the other hand, if L = 0, let ε = 1/2. Then there is a δ > 0 such that

0 < ∣x∣ < δ Ô⇒ ∣f (x)∣ < 1/2 .

If n is a positive integer satisfying n > 1/δ, take x = 1/n. Then 0 < ∣x∣ < δ is true,
but 1 = ∣1∣ = ∣f (x)∣ < 1/2 is not true. Again we have a contradiction.

Hence lim f (x) cannot exist.

x→0

11. Show that the equation x2 − 10 = x sin x has a real solution.

Solution: Consider the function f (x) = x2 − 10 − x sin x. Then f (0) = −10 < 0 and
f (10) = 102 − 10 − 10 sin(10) = 90 − 10 sin(10) ≥ 90 − 10 = 80 > 0. Note that f is
continuous on [ 0, 10]. Therefore we can apply the Intermediate Value Theorem to
the function f on the interval [ 0, 10] for the value 0 to conclude that there is a point
c in ( 0, 10) such that f (c) = 0. This c is also a solution of the given equation.

9
 x2
12. Consider the equation 1 − = cos x .
4
a. Show that this equation has at least one real solution.
b. Show that this equation has at least two real solutions.
c. Show that this equation has at least three real solutions.

x2 x2
Solution: Let f (x) = 1 − − cos x. The solutions of the equation 1 − = cos x
4 4
correspond to the zeros of f .

As f (0) = 0, x = 0 is one zero.

Now observe that f (π/2) = 1 − π 2 /16 > 0 and f (π) = 2 − π 2 /4 < 0 as 4 > π > 3. As f
is continuous on the entire real line, applying the Intermediate Value Theorem to
the function f on the interval [π/2, π] we conclude that there is a point c in this
interval such that f (c) = 0. This is our second zero.

Finally, as the function f is even, we have f (−c) = f (c) = 0, and x = −c is our third
zero.

13. Show that at any moment there are two antipodal points on the equator of the Earth with
the same temperature.

Solution: First we will make a mathematical model of the problem. We will

consider the equator as a circle, and use the longitude as our coordinate θ. We
choose the positive direction for θ to correspond to the East, measure θ in radians,
and let it take any real value. So 45○ W corresponds to θ = −π/4, θ = 7π/4, and
θ = 15π/4, among other values. Note that θ + π corresponds to the antipode of
the point corresponding to θ. We let T (θ) denote the temperature at θ. We have
T (θ + 2π) = T (θ) for all θ. We will assume that T is a continuous function. We
want to show that there is a c such that T (c + π) = T (c).

Consider the function f (θ) = T (θ + π) − T (θ). Note that since T is continuous, f is

continuous. Our quest to find a c such that T (c + π) = T (c) is equivalent to find a
c such that f (c) = 0. If f (0) = 0, then we let c = 0 and we are done. Suppose that
f (0) =/ 0. Observe that f (0) = T (π) − T (0) = T (π) − T (2π) = −f (π). In other words,
f (0) and f (π) have opposite signs. Now we apply the Intermediate Value Theorem
to f on the interval [0, π] for the value 0, and conclude that there is a point c in
[0, π] such that f (c) = 0. We are done.

Remark: It is possible to show that at any moment there are two antipodal points on Earth
with the same temperature and the same pressure.

10
14. Find all tangent lines to the graph of y = x3 that pass through the point (2, 4).
Solution: As dy/dx = d(x3 )/dx = 3x2 , the equation of the tangent line through
a point (x0 , x30 ) on the graph is y − x30 = 3x20 (x − x0 ). This line passes through
(2, 4) exactly when 4 − x30 = 3x20 (2 − x0 ), or in other words, x30 − 3x20 + 2 = 0. We
observe that x0 = 1 is a root of this polynomial. Therefore we have the factorization √
x30 − 3x20 + 2 = (x0 − 1)(x20 − 2x0 − 2). The roots of the quadratic factor are x0 = 1 ± 3.
√ √
Therefore
√ the tangent
√ lines to y = x 3 at the points (1, 1), (1 + 3, 10 + 6 3), and
(1 − 3, 10√− 6 3) pass through √ (2, 4). The equations
√ of these
√ lines are y = 3x − 2,
y = (12 + 6 3)x − (20 + 12 3), and y = (12 − 6 3)x − (20 − 12 3), respectively.

√
1 + sin2 x2 − cos3 x2
15. Evaluate the limit lim .
x→0 x3 tan x
Solution:
√ √
1 + sin2 x2 − cos3 x2 ⎛⎛ 1 + sin2 x2 − 1 1 − cos3 x2 ⎞ x ⎞
lim = lim + ⋅
x→0 x3 tan x x→0 ⎝⎝ x4 x4 ⎠ tan x ⎠
sin x2 2 1 1 − cos3 x2 x
= lim ((( 2 ) ⋅ √ + 4
)⋅ )
x→0 x 1 + sin x2 + 1
2 x tan x
Now we observe that:
sin x2
lim =1
x→0 x2
1 1 1
lim √ =√ =
x→0
1 + sin x + 1
2 2
1 + sin 0 + 1 2
2 2
x
lim =1
x→0 tan x

11
 and
1 − cos3 x2 1 − cos x2
lim = lim ( ⋅ (1 + cos x2 + cos2 x2 ))
x→0 x4 x→0 x4
2 sin2 (x2 /2)
= lim ( ⋅ (1 + cos x2 + cos2 x2 ))
x→0 x4
2
1 sin(x2 /2)
= ⋅ (lim ) ⋅ lim(1 + cos x2 + cos2 x2 )
2 x→0 x2 /2 x→0

1
= ⋅ 12 ⋅ 3
2
3
= .
2
Therefore: √
1 + sin2 x2 − cos3 x2 1 3
lim 3
= (1 ⋅ + ) ⋅ 1 = 2 .
x→0 x tan x 2 2

16. Find the equation of the tangent line to the graph of y = sin2 (πx3 /6) at the point with
x = 1.
dy
Solution: y = sin2 (πx3 /6) ⇒ = 2 sin(πx3 /6) ⋅ cos(πx3 /6) ⋅ 3πx2 /6 . Therefore,
dx √
dy 3π
∣ = 2 sin(π/6) ⋅ cos(π/6) ⋅ π/2 = .
dx x=1 4

Since y∣x=1 = 1/4, using the point-slope formula we find the equation of the tangent
line as √
1 3π
y− = (x − 1)
4 4
or, after some reorganization,
√ √
3π 1 − 3π
y= x+ .
4 4

17. Let
⎧
⎪
⎪
⎪
⎪
1
2x + x2 sin( ) if x =/ 0,
⎪
⎪
⎪ x
f (x) = ⎨
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎩0 if x = 0.
a. Find f ′ (x) for all x.
b. Show that f ′ is not continuous at 0.

12
Solution: a. For x =/ 0 we compute the derivative using the rules of differentiation:
d
f ′ (x) = (2x + x2 sin(1/x)) = 2 + 2x sin(1/x) + x2 cos(1/x) ⋅ (−1/x2 )
dx
for x =/ 0.

For x = 0 we must use the definition of the derivative:

f (0 + h) − f (0) 2h + h2 sin(1/h)
f ′ (0) = lim = lim
h→0 h h→0 h
1
= lim 2 + lim h sin( ) = 2 + 0 = 2
h→0 h→0 h

Here we used the fact that lim h sin(1/h) = 0 whose proof uses the Sandwich (or
h→0
Squeeze) Theorem. Here is a recap of the proof: Since ∣ sin(1/h)∣ ≤ 1 for all h =/ 0
we have ∣h sin(1/h)∣ = ∣h∣ ⋅ ∣ sin(1/h)∣ ≤ ∣h∣ for all h =/ 0. Therefore

1
−∣h∣ ≤ h sin( ) ≤ ∣h∣ for all h =/ 0 .
h

As lim ∣h∣ = 0 = lim(−∣h∣), it follows by the Sandwich Theorem that lim h sin(1/h) = 0.
h→0 h→0 h→0

To summarize:
⎧
⎪
⎪
⎪
⎪
1 1
2 + 2x sin( ) − cos( ) if x =/ 0,
⎪
⎪
⎪ x x
f ′ (x) = ⎨
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎩2 if x = 0.

b. Consider lim f ′ (x). We have lim 2 = 2 and lim 2x sin(1/x) = 0 as in part (a).
x→0 x→0 x→0
However lim cos(1/x) does not exist. It follows that lim f ′ (x) does not exist and
x→0 x→0
hence f ′ is not continuous at 0.

13
 d2 y
18. Find ∣ if y is a differentiable function of x satisfying the equation x3 +2y 3 = 5xy.
dx2 (x,y)=(2,1)

Solution:

x3 + 2y 3 = 5xy
Ô⇒

d/dx

dy dy
3x2 + 6y 2 = 5y + 5x (⋆)
dx dx
Ô⇒

x = 2, y = 1

dy dy
12 + 6 = 5 + 10
dx dx
Ô⇒

dy
4 =7
dx
Ô⇒

dy 7
= at (x, y) = (2, 1)
dx 4

Now we differentiate the equation marked (⋆) with respect to x to find the second

14
 derivative.
dy dy
3x2 + 6y 2 = 5y + 5x
dx dx

Ô⇒
d/dx

2
dy d2 y dy dy d2 y
6x + 12y ( ) + 6y 2 2 = 5 +5 + 5x 2
dx dx dx dx dx
dy 7

Ô⇒
x = 2, y = 1, =
dx 4
2
7 d2 y 7 d2 y
12 + 12 ( ) + 6 2 = 10 ⋅ + 10 2
4 dx 4 dx

Ô⇒
d2 y 125
= at (x, y) = (2, 1)
dx2 16

5y − 3x2
Remark: An alternative approach is to solve y ′ from (⋆), viz. y ′ = , and then
6y 2 − 5x
differentiate this with respect to x to find y ′′ .

19. A piston is connected by a rod of length 14 cm to a crankshaft at a point 5 cm away from

the axis of rotation of the crankshaft. Determine how fast the crankshaft is rotating when the
piston is 11 cm away from the axis of rotation and is moving toward it at a speed of 1200
cm/sec.
Solution: Let P (x, y) and Q(a, 0) be the ends of the connecting rod as shown in
the picture. The axis of rotation of the crankshaft passes through the origin of the
xy-plane and is perpendicular to it. The point P where the rod is connected to
crankshaft moves on a circle with radius 5 cm and center at the origin. The point
Q where the rod is connected to the piston moves along the positive x-axis. θ is the
angle between the ray OP and the positive x-axis.
y

P (x, y)

14 c
m
θ Q(a, 0)
−5 5 x

x2 + y 2 = 25

The question is:

da dθ
a = 11 cm and = −1200 cm/sec Ô⇒ =?
dt dt
15
We have
x2 + y 2 = 5 2 (I)
and
(x − a)2 + y 2 = 142 . (II)

At the moment in question a = 11 cm. Substituting this in (I) and (II) we obtain
x2 + y 2 = 52 and (x − 11)2 + y 2 = 142 . Subtracting the second equation from the
√ first
gives 22x−11 = 5 −14 , and solving for x we get x = −25/11 cm. Then y = 20 6/11
2 2 2

cm.

Differentiating (I) and (II) with respect to time t we obtain

dx dy
x +y =0 (III)
dt dt
and
dx da dy
(x − a) ⋅ ( − )+y =0. (IV)
dt dt dt
da
At the moment in question a = 11 cm, = v = −1200 cm/sec, x = −25/11 cm and
√ dt
y = 20 6/11 cm. Substituting these in (III) and (IV) we get

dx √ dy
−5 +4 6 =0 (V)
dt dt
and
dx √ dy
−146 + 20 6 = −146v . (VI)
dt dt
dx dx 146
Subtracting 5 times (V) from (VI) we find −121 = −146 v, and hence = v.
dt dt 121
dy 365
Substituting this back in (V) gives = √ v.
dt 242 6

dθ
Now we are ready to compute . Since tan θ = y/x, differentiation gives
dt
dy dx
dθ x −y
sec2 θ = dt 2 dt
dt x
and using sec2 θ = 1 + tan2 θ = 1 + (y/x)2 we obtain

dy dx
dθ x dt − y dt
= .
dt x2 + y 2
√
25 20 6 dx 146 dy 365
Plugging x = − cm, y = cm, = v, and = √ v in this formula
11 11
√ dt 121 dt 242 6
dθ 73 1460 6
gives =− √ v= radian/sec.
dt 110 6 11

16
 √ √
1460 6 1460 6 60
Remark: radian/sec is ⋅ rpm or approximately 3105 rpm.
11 11 2π

Remark: This problem has a shorter solution if we use the law of cosines. Start with x2 + 52 −
2 ⋅ 5 x cos θ = 142 and differentiate with respect to t to obtain
dθ 5 cos θ − 11
= v.
dt 5 sin θ
√
Put x = 11 cm in the first equation to find cos θ = −5/11 and then sin θ = 4 6/11. Now
substituting these in the second equation gives the answer.

20. Determine how fast the length of an edge of a cube is changing at the moment when the
length of the edge is 5 cm and the volume of the cube is decreasing at a rate of 100 cm3/sec.

Solution: Let a denote the length of an edge of the cube, and V denote the volume
of the cube. Then we have V = a3 . Differentiating with respect to time t gives
dV da dV
= 3a2 . Substituting = −100 cm3/sec and a = 5 cm for the moment in
dt dt dt
da 4
question, we obtain = − cm/sec. Therefore the length of the edge is decreasing
dt 3
4
at a rate of cm/sec at that moment.
3

21. We measure the radius and the height of a cone with 1% and 2% errors, respectively. We
use these data to compute the volume of the cone. Estimate the percentage error in volume.

Solution: Let r, h, and V be the radius, the height and the volume of the cone,
respectively.
π 2 2π π dV dr dh
V = r h Ô⇒ dV = rhdr + r2 dh Ô⇒ =2 + .
3 3 3 V r h

dr dh
Since the error in r is 1% we have ∣ ∣ ≤ 1% . Similarly ∣ ∣ ≤ 2% . Now using the
r h
triangle inequality we obtain
dV dr dh dr dh
∣ ∣ = ∣2 + ∣ ≤ 2 ∣ ∣ + ∣ ∣ ≤ 2 ⋅ 1% + 2% = 4% .
V r h r h
The error in volume is 4% .

17
22. A cone of radius 2 cm and height 5 cm is lowered point first into a tall cylinder of radius
7 cm that is partially filled with water. Determine how fast the depth of the water is changing
at the moment when the cone is completely submerged if the cone is moving with a speed of 3
cm/s at that moment.

Solution: Let r and h be the radius and the height of the part of the cone that is
under the water level, respectively. Let L be the depth of the water in the cylinder
and let y be the vertical distance from the tip of the cone to the bottom of the
cylinder. Let V0 be the volume of the water.

Then
2
π π 2 4π
V0 = π ⋅ 7 ⋅ L − r2 h = π ⋅ 72 ⋅ L − ( h) h = 49πL − h3
2
3 3 5 75
where we used the fact that r/h = 2/5.

Now differentiating this with respect to time t gives

d dL 4π 2 dh
0= V0 = 49π − h .
dt dt 25 dt
In particular at the moment when the cone is completely submerged we have h = 5
cm and
dL dh
49 =4 .
dt dt
On the other hand, at the same moment
dh dL dy dh dL
h = L − y Ô⇒ = − Ô⇒ = +3
dt dt dt dt dt
because dy/dt = −3 cm/s as the cone is being lowered at a speed of 3 cm/s.

From these two equations we obtain dL/dt = 4/15 cm/s. In other words, the depth
of the water is increasing at a rate of 4/15 cm/s at that moment.

18
23. A water tank has the shape of an upside-down cone with radius 2 m and height 5 m. The
water is running out of the tank through a small hole at the bottom. Assume that the speed of
the water flowing through the hole is proportional to the square root of the depth of the water
in the tank.

a. In this part, suppose that the water is running out at a rate of 3 m3/min when the depth
of the water in the tank is 4 m. Find the rate at which the water level is changing at this
moment.
b. In this part, suppose that the water level is falling at a rate of 1/3 m/min when the tank
is full. Find the rate at which the water level is changing when the depth of the water in the
tank is 4 m.
c. In this part, suppose that it takes 3 minutes for the depth of the water to decrease from
5 m to 4 m. Find how long it takes for the full tank to completely drain.

Solution: Let r and h denote the radius and the height of the cone formed by the
water, and let V denote the volume of the water. Using the fact that r/h = 2/5, we
obtain
2
π π 2 4π 3
V = r2 h = ( h) h = h
3 3 5 75
and hence:
dV 4π 2 dh
= h
dt 25 dt

In part (a), we are given that dV /dt = −3 m3 /min when h = 4 m. Substituting these
in the equation above and solving for dh/dt we obtain dh/dt = −75/(64π) m/min.
Hence the water level is falling at a rate of 75/(64π) m/min in part (a).

Now we will use the condition that the speed of the water flowing through the hole
is proportional to the square root of the depth of the water in the tank. This means
dV √
= −k h
dt
for some positive constant k. Substituting this in the previous equation we obtain:
√ 4π 2 dh
−k h = h (⊛)
25 dt

19
In part (b), we are given that dh/dt = −1/3 m/min when h = 5 m. Substituting
these in (⊛) we obtain:
√ 4π 2 1
−k 5 = 5 (− )
25 3
On the other hand, (⊛) gives
√ 4π 2 dh
−k 4 = 4
25 dt

√ h = 4 m. Now solving for dh/dt from these two equations

when √ gives dh/dt =
−5 5/24 m/min. Hence the water level is falling at a rate of 5 5/24 m/min in part
(b).

For part (c), we rewrite (⊛) in the form

4π 3/2
h dh = −k dt
25
and integrate to obtain
8π 5/2
h = −kt + C
125
for some constant C. In other words, h5/2 = −at + b for some constants a and b. As
h = 5 m when t = 0 min, we have b = 55/2 . Using the condition in part (c) that h = 4
m when t = 3 min we obtain 45/2 = −a ⋅ 3 + 55/2 , so a = (55/2 − 45/2 )/3. Finally, the
tank is empty when h = 0 and this happens when t = b/a = 3 ⋅ 55/2 /(55/2 − 45/2 ) min.
Hence the tank drains in 3/(1 − (4/5)5/2 ) minutes in part (c).

20
24. Find the absolute maximum value and the absolute minimum value of f (x) = x4/3 − x − x1/3
on the interval [−1, 6].
Remark: How to find the absolute maximum and the absolute minimum values of a
continuous function f on a closed interval [ a, b] of finite length:
i. Compute f ′ .
ii. Find the critical points of f in (a, b).
iii. Add the endpoints a and b to this list.
iv. Compute the value of f at each point in the list.
v. The largest value is the the absolute maximum and the smallest value is the absolute
minimum of f on [a, b].

4 1
Solution: f ′ (x) = x1/3 −1− x−2/3 . The derivative is not defined at x = 0, therefore
3 3
x = 0 is a critical point. Next we solve f ′ (x) = 0.

4 1
In the equation x1/3 −1− x−2/3 = 0 we let z = x1/3 to obtain the equation 4z 3 −3z 2 −
3 3
1 = 0. Since z = 1 is a root, we have the factorization 4z 3 −3z 2 −1 = (z −1)(4z 2 +z +1).
As the quadratic factor has no real roots, z = 1 is the only solution. Therefore
x = z 3 = 13 = 1, which belongs to the interval [−1, 6], is the only other critical point.

We have f (0) = 0, f (1) = −1, f (−1) = 3, and f (6) = 5 ⋅ 61/3 − 6. Observe that
5 ⋅ 61/3 − 6 > 3 ⇐⇒ 5 ⋅ 61/3 > 9 ⇐⇒ 53 ⋅ 6 > 93 ⇐⇒ 725 > 721.

We conclude that the absolute maximum and minimum values of f (x) = x4/3 −x−x1/3
on the interval [−1, 6] are 5 ⋅ 61/3 − 6 and −1, respectively.

x+1
25. Find the absolute maximum and the absolute minimum values of f (x) = on the
x2 +x+9
interval [ 0, ∞).
Remark: When looking for the absolute maximum and the absolute minimum values of a
continuous function on an interval that is not necessarily closed or of finite length, a modified
version of the algorithm above can be used.
• In Step iv , if an endpoint does not belong to the interval, then we compute the appropriate
one-sided limit of the function at that point instead of the value of the function.
• In Step v , if the largest value (which can be ∞) occurs only as a limit, then we conclude
that there is no absolute maximum. Similarly for the smallest value.

x2 + 2x − 8
Solution: We compute f ′ (x) = − . The roots of f ′ (x) = 0 are x = −4
(x2 + x + 9)2
and x = 2. Only x = 2 is in the interval [ 0, ∞). So our list is 0, 2, and ∞.

1 1 1 1 1
f (0) = , f (2) = , and lim f (x) = 0. Since > > 0, the absolute maximum is
9 5 x→∞ 5 9 5
and there is no absolute minimum.

21
26. Suppose that a function f satisfies f ′ (x) = f (x/2) for all x and f (0) = 1. Show that if
f (x0 ) = 0 for some x0 > 0, then there is x1 such that 0 < x1 < x0 and f (x1 ) = 0.

Solution: Since f is differentiable, f is continuous. Suppose that f (x0 ) = 0 for some

x0 > 0. We apply the Mean Value Theorem to f on [ 0, x0 ] to conclude that there
f (x0 ) − f (0) 1
is c such that 0 < c < x0 and f ′ (c) = = − < 0. Then f ′ (x) = f (x/2)
x0 − 0 x0
gives f (c/2) = f ′ (c) < 0. We also have f (0) = 1 > 0. We apply the Intermediate
Value Theorem to f on [ 0, c/2] to conclude that there is x1 such that 0 < x1 < c/2
and f (x1 ) = 0. As 0 < x1 < c/2 < c < x0 we are done.

Remark: With a little bit more work it can be shown that f (x) > 1 for all x > 0.

27. Show that if f is a twice-differentiable function such that f (0) = 1, f ′ (0) = −1, f (1) = 2,
f ′ (1) = 5, and f ′′ (x) ≥ 0 for all x, then f (x) ≥ 1/3 for all 0 ≤ x ≤ 1.

Solution: Let x1 < x2 be in [0, 1]. The Mean Value Theorem applied to the
function f ′ on the interval [x1 , x2 ] says that there is a point c in (x1 , x2 ) such that
f ′ (x2 ) − f ′ (x1 )
= f ′′ (c). As f ′′ (c) ≥ 0 we have f ′ (x1 ) ≤ f ′ (x2 ). In particular, we
x2 − x1
have −1 = f ′ (0) ≤ f ′ (x) ≤ f ′ (1) = 5 for all x in (0, 1).

Let 0 < x < 1. By the Mean Value Theorem applied to the function f on the
f (x) − f (0)
interval [0, x], there exists a point c1 in (0, x) such that = f ′ (c1 ). Since
x−0
f ′ (c1 ) ≥ −1, it follows that
f (x) ≥ −x + 1 (Q)
for 0 < x < 1. Similarly, applying the Mean Value Theorem to the function
f on the interval [x, 1], we see that there exists a point c2 in (x, 1) such that
f (1) − f (x)
= f ′ (c2 ). Now using the fact that f ′ (c2 ) ≤ 5 we conclude that
1−x
f (x) ≥ 5x − 3 (QQ)

for 0 < x < 1.

Adding 5 times the inequality (Q) to the inequality (QQ) we get f (x) ≥ 1/3 for all x
in [0, 1].

Remark: In fact, it can be shown that f (x) > 1/3 for all x in [0, 1].

22
28. Sketch the graph of y = 5x2/3 − 2x5/3 .

10 −1/3 10 2/3 10 −1/3

Solution: y ′ = x − x = x (1 − x). Therefore y ′ > 0 on (0, 1), and
3 3 3
y ′ < 0 on (−∞, 0) and (1, ∞).

10 −4/3 20 −1/3 20
y ′′ = − x − x = − x−4/3 (x + 1/2). Therefore y ′′ > 0 on (−∞, −1/2), and
9 9 9
y ′′ < 0 on (−1/2, 0) and (0, ∞).

These give us the following table of signs and shapes.

x ∣ −1/2 0 1
y′ ∣ − ∣ − ^ + 0∣ −
^
y ′′ ∣ + 0∣ − ^ − ∣ −
^
↑ ↑ " ↑
"

inf.pt. loc.min. loc.max.

We compute the y-coordinates of the important points √ to get (0, 0) for the local
minimum, (1, 3)√for the local maximum,√ and (−1/2, 3 2) for the inflection point.
3

Note that since 3 2 > 1, we have 3 3 2 > 3. Also note that the function is continuous
at x = 0, but
10
lim y ′ = lim+ ( x−1/3 (1 − x)) = ∞
x→0+ x→0 3
and
10
lim− y ′ = lim− ( x−1/3 (1 − x)) = −∞ .
x→0 x→0 3
Therefore (0, 0) is a cusp.

Finally we find the x-intercepts: y = 0 ⇒ 5x2/3 − 2x5/3 = 0 ⇒ 2x2/3 (5 − 2x) = 0 ⇒ x =

0 or x = 5/2.

(The graph is on the next page.)

23
 Now we use these data to draw the graph:

Remark: Here is the same graph drawn by Maple:

24
29. Two corridors meet at a corner. One of the corridors is 2 m wide and the other one is 3
m wide. What is the length of the longest ladder that can be carried horizontally around this
corner?

Solution: Length of the longest ladder will be equal to the absolute minimum value
of L in the picture.

y
2m L

3m

Using the relation 3/x = y/2 we obtain

L = (x2 + 9)1/2 + (4 + y 2 )1/2 = (x2 + 9)1/2 + (4 + (6/x)2 )1/2 .

Therefore we have to minimize

2
L = (x2 + 9)1/2 ⋅ (1 + ) for 0 < x < ∞ .
x
We first look at the critical points:
dL 2 2
= (x2 + 9)−1/2 ⋅ x ⋅ (1 + ) + (x2 + 9)1/2 ⋅ (− 2 ) = 0
dx x x
18
Ô⇒ x + 2 − (2 + 2 ) = 0 Ô⇒ x = 181/3 m Ô⇒ L = (22/3 + 32/3 )3/2 m
x
Since lim+ L = ∞ and lim L = ∞, the value at the critical point is indeed the absolute
x→0 x→∞
minimum.

Hence the length of the longest ladder that can be carried around this corner is
(22/3 + 32/3 )3/2 m.

Remark: (22/3 + 32/3 )3/2 m is approximately 7.02 m.

25
30. Find the maximum possible total surface area of a cylinder inscribed in a hemisphere of
radius 1.

Solution: Let S be the total surface area of the cylinder, S = 2πr2 + 2πrh. We have
h = (1 − r2 )1/2 . Hence we want to maximize

S = 2πr2 + 2πr(1 − r2 )1/2 for 0 ≤ r ≤ 1 .

1 h

1 dS
First we find the critical points: = 2r + (1 − r2 )1/2 − r2 (1 − r2 )−1/2 = 0 Ô⇒
2π dr
2r(1 − r2√)1/2 = 2r2 − 1 Ô⇒ 4r2 (1 − r2 ) = (2r2 − 1)2 Ô⇒ 8r4 − 8r2 + 1 = 0 Ô⇒
8 ± 32 1
r2 = . Note that 2r(1 − r2 )1/2 = 2r2 − 1 implies r2 ≥ , and therefore
16
√ √ √ 2
2+ 2 2+ 2
r2 = and r = .
4 2
√ √
√ 2+ 2
We get S = (1 + 2)π at the critical point r = , S = 0 at the endpoint r = 0,
√ 2 √
and S = 2π at the endpoint r = 1. As 2 > 1 we have 1 + 2 > 2, and the absolute
maximum value occurs at the critical point.
√
The maximum possible surface area of the cylinder is (1 + 2)π.

26
31. A fold is formed on a 20 cm × 30 cm rectangular sheet of paper running from the short
side to the long side by placing a corner over the long side. Find the minimum possible length
of the fold.

Solution: Let ABCD be the sheet of paper and let P be the point on the edge
AB where the corner C is folded over. The fold runs from Q on the edge BC to R
on the edge CD. Let S be the projection of R on to the edge AB. Let L be the
length of the fold QR and let x = CQ.

D R C

L x

A S P B

Then P Q/P√ B = RP /RS by the similarity

√ of the triangles P BQ and RSP . Hence
RP = 20x/ x − (20 − x) = 20x/ 40x − 400 and L2 = RQ2 = RP 2 + P Q2 =
2 2

400x2 /(40x − 400) + x2 = x3 /(x − 10). The largest possible value of x is 20

cm. √The smallest possible value √of x occurs when RP = 30 cm; that is when
20x/ 40x − 400 = 30 or x = 45 − 15 5 cm.
√
Therefore we want to minimize L where L2 = x3 /(x − 10) for 45 − 15 5 ≤ x ≤ 20.

Differentiating 2LdL/dx = 3x2 /(x − 10) − x3 /(x − 10)2 and setting dL/dx = 0 we
obtain√x = 15 cm as the only critical point in the domain. For x = 15 cm we have
L = 15 3 cm.
√ √
At the endpoint x =√20 cm we have L = 20 2 cm and at the endpoint x = 45 − 15 5
√
cm we have L = 15 18 − 6 5 cm.
√
Since 15 3 cm is the smallest of these values, we conclude that this is the smallest
possible value for the length of the fold.

√
Remark: If the question is posed √ for a w × ℓ sheet of paper with w ≤ 2 2ℓ/3, then the length
of the shortest possible fold is 3 3w/4. Therefore, for a A4 size paper with w = 210 mm and
ℓ = 297 mm, the shortest fold is approximately 273 mm; and for a letter size paper with w = 8.5
in and ℓ = 11 in, the shortest fold is approximately 11.04 in.

27
32. The Rubber Duck is a sculpture designed by Florentijn Hofman and constructed from
PVC. For the purposes of this question, we consider the Rubber Duck to consist of a spherical
head of radius a and a spherical body of radius b. The research shows that the cuteness K of
the Rubber Duck is given by
⎧
⎪
⎪
⎪
a a
⎪ b (1 − b )(a + b) if 0 ≤ a < b,
⎪
K =⎨
⎪
⎪
⎪
⎪
⎪ if 0 < b ≤ a.
⎩0
Find the dimensions of the cutest Rubber Duck with a total surface area of 400π m2 .

Solution: The total surface area of the Rubber Duck is 4πa2 + 4πb2 . Hence 4πa2 +
4πb2 = 400π m2 , giving b2 = 100 − a2 . Therefore,
⎧
⎪ a(100 − 2a2 ) √
⎪
⎪
⎪ 100 − a2 if 0 ≤ a ≤ 5 2 ,
K =⎨
⎪
⎪
⎪
⎪
⎩0 otherwise,

and we have to find where the absolute maximum value of

100a − 2a3
K =
100 − a2
√
on the interval 0 ≤ a ≤ 5 2 occurs.

We first find the critical points. As

dK (100 − 6a2 )(100 − a2 ) − (100a − 2a3 )(−2a)

= ,
da (100 − a2 )2

dK /da = 0 gives a4 − 250a2 + 5000 = 0. Now using the quadratic formula we obtain
√
250 ± 2502 − 4 ⋅ 5000 √
a =
2
= 125 ± 25 17 ,
2

28
 which gives us four solutions
√ √ √ √ √ √ √ √
a = 5 5 + 17, 5 5 − 17, −5 5 + 17, −5 5 − 17 .
√
Only the first two√of these are positive and the first one is greater than 5 √
2. On the
√ √ √ √
other hand, 0 < 5 5 − 17 < 5 2 as 3 < 17 < 5. We conclude that a = 5 5 − 17
is the only critical point.

We have
√ √ √
17 − 3
K = 10 5 − 17 ⋅ √ >0
17 − 1
√ √ √
for a = 5 5 − 17. As K = 0 at the endpoints a = 0 and a = 5 2 of the interval,
this is the absolute maximum value.
√ √ √√
Therefore the cutest Rubber Duck has a = 5 5 − 17 m and b = 5 17 − 1 m.

33. A snowman is an anthropomorphic sculpture made from snow as well as some pieces of
coal, a carrot, a hat and a scarf. For the purposes of this question, we consider a snowman to
consist of a spherical head of radius a and a spherical body of radius b, and we also assume
that the snow does not melt and its density does not change while it is being sculpted.
The research shows that the cuteness K of a snowman is given by
⎧
⎪
2
⎪
⎪
⎪
a a
⎪( b ) (1 − b ) (a + ab + b ) if 0 ≤ a < b,
2 2
K =⎨
⎪
⎪
⎪
⎪
⎪ if 0 < b ≤ a.
⎩ 0
Find the dimensions of the cutest snowman that can be built with 4π/3 m3 of snow.

4π 3 4π 3 4π 3 4π 3
Solution: The total volume of the snowman is a + b . Hence a + b =
3 3 3 3

29
 4π 3
m , giving b3 = 1 − a3 . Therefore,
3
⎧
⎪ a2 (1 − 2a3 ) 1
⎪
⎪
⎪ if 0 ≤ a ≤ √ ,
K =⎨ 1−a
3 3
2
⎪
⎪
⎪
⎪
⎩0 otherwise,

and we have to find where the absolute maximum value of

a2 − 2a5
K =
1 − a3
1
on the interval 0 ≤ a ≤ √
3
occurs.
2

We first find the critical points. As

dK (2a − 10a4 )(1 − a3 ) − (a2 − 2a5 )(−3a2 )
= ,
da (1 − a3 )2
dK /da = 0 gives 4a7 − 9a4 + 2a =√0. As we have√4a7 − 9a4 + 2a = a(a3 − 2)(4a3 − 1),
the critical points are a = 0, a = 3 2 and √a = 1/ 3 4. Hence the only critical point in
the interval we are interested in is a = 1/ 3 4.
√ √
We have K = 1/(3 3 2) > 0 for a = 1/ 3 4. As K = 0 at the endpoints a = 0 and
1
a= √ 3
of the interval, this is the absolute maximum value.
2
√ √
Therefore the cutest snowman has a = 1/ 3 4 m and b = 3 3/4 m.

34. A dessert in the shape of a hemisphere with radius 1 dm is made by baking a cylindrical
cake of height h , and topping it with a spherical cap of ice cream and surrounding it with a
hemispherical ring of chocolate mousse as shown in the figure. If the cake costs 8/π /dm3 , the
ice cream costs 9/π /dm3 and the chocolate mousse costs 12/π /dm3 , determine the value of
h for (a) the least expensive and (b) the most expensive dessert that can be made.
You may use the fact that the volume of a hemispherical ring of height h is 2πh3 /3.

Ice cream
(9/π /dm3 )

Cake
(8/π /dm3 )

Chocolate
mousse
(12/π /dm3 )
1 h

30
 Solution: We have
12 8
Cost = ⋅ (Volume of the ring) + ⋅ (Volume of the cylinder)
π π
9
+ ⋅ (Volume of the cap)
π
12 2π 3 8 9 2π 2π
= ⋅ h + ⋅ π(1 − h2 )h + ⋅ ( ⋅ 13 − h3 − π(1 − h2 )h)
π 3 π π 3 3
= 6 − h + 3h3

and hence we want to find the absolute maximum and the absolute minimum values
of
Cost = 6 − h + 3h3 for 0 ≤ h ≤ 1 .

We first find the critical points. As d(Cost)/dh = −1 + 9h2 , setting this equal to zero
gives h = 1/3 dm and h = −1/3 dm. Only the first of these are in the interval [0, 1],
and at that point the cost is 52/9 .

Next we look at the endpoints of the interval. h = 0 dm gives Cost = 6 , and h = 1

dm gives Cost = 8 .

Therefore the least expensive dessert has h = 1/3 dm, and the most expensive dessert
has h = 1 dm.

35. We want to build a greenhouse that has a half cylinder roof of radius r and height r
mounted horizontally on top of four rectangular walls of height h as shown in the figure. We
have 200π m2 of plastic sheet to be used in the construction of this structure. Find the value
of r for the greenhouse with the largest possible volume we can build.

Solution: We have
1 1
200πm2 = Total Surface Area = 2 ⋅ (r + 2r)h + 2 ⋅ πr2 + ⋅ 2πr ⋅ r = 6rh + 2πr2
2
´¹¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹¸¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¶ ´¹¹ ¹ ¹ ¹ ¹ ¸ 2
walls ¹ ¹ ¹ ¹ ¹ ¹¶ ´¹¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¸ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¶
half-disks top

31
 π 100
and hence h = ( − r). In particular, 0 < r ≤ 10.
3 r

Let V denote the volume of the greenhouse. Then

1 π
V = r ⋅ 2r ⋅ h + ⋅ πr2 ⋅ r = 2r2 h + r3
´¹¹ ¹ ¹ ¹ ¹¸¹ ¹ ¹ ¹ ¹ ¶ 2 2
bottom ´¹¹ ¹ ¹ ¹ ¹ ¹ ¹ ¸¹¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹¶
top

and substituting h in terms of r, the problem becomes:

200π π
Maximize V = r − r3 for 0 < r ≤ 10 .
3 6

We first find the critical points in the interval (0, 10]:

dV 200π π 2 20
= − r = 0 Ô⇒ r = ± √ m
dr 3 2 3
√
As 3 < 2, neither of these satisfies 0 < r ≤ 10, and there are no critical points in
the interval.

Next we look at the endpoints of the interval (0, 10]: At r = 0 m we have

lim+ V = 0 m3 , and at r = 10 m we have V = 500π m3 .
r→0

Therefore the maximum possible volume is 500π m3 and occurs when r = 10 m (and
h = 0 m).

36. A pool, like the one in front of the Faculty of Science Building A, loses water from its sides
and its bottom due to seepage, and from its top due to evaporation. For a pool with radius R
and depth H in meters, the rate of this loss in m3/hour is given by an expression of the form

aR2 + bR2 h + cRh2

where h is the depth of the water in meters, and a, b, c are constants independent of R, H and
h. Due to this loss, water must be pumped into the pool to keep it at the same level even when
the drains are closed.
Suppose that a = 1/300 m/hour and b = c = 1/150 1/hour. Find the dimensions of the pool with
a volume of 45π m3 which will require the water to be pumped at the slowest rate to keep it
completely full.

H
h

32
 Solution: From πR2 H = 45π we have H = 45/R2 . When the pool is full, the water
is lost at the rate
2
1 2 1 2 45 1 45
L = aR2 + bR2 H + cRH 2 = R + R ⋅ 2+ R ⋅ ( 2) .
300 150 R 150 R

Hence we want to
1 2 3 27 1
Minimize L = R + + ⋅ for 0 < R < ∞ .
300 10 2 R3

We first find the critical points:

dL 1 81 1
= R− = 0 Ô⇒ R = 3 ⋅ 52/5 m
dR 150 2 R4
As lim+ L = ∞ and lim L = ∞ at the endpoints of the interval, the absolute
R→0 R→∞
minimum value of L occurs at R = 3 ⋅ 52/5 m, and the corresponding depth is
H = 45/(3 ⋅ 52/5 )2 = 51/5 m.

Hence the pool that requires the water to be pumped at the slowest rate has radius
R = 3 ⋅ 52/5 m and depth H = 45/(3 ⋅ 52/5 )2 = 51/5 m.

37. The towns A and B lie inland and the town C lies on the coast as shown in Figures 1 and
2 . A port P will be built on the coast and connected to the towns A, B, C with straight roads.
The cost of constructing the roads connecting P to A and P to B is 105 /km, and the cost of
constructing the road connecting P to C is k × 105 /km. The port will be built at the location
for which the total cost of constructing these roads is the smallest.
A B A B

10 km 10 km 10 km 10 km

C C P
10 km 30 km 10 km 30 km
P SEA SEA
Figure 1 Figure 2

For some values of k the smallest total cost will be achieved when P is built at C, as in Figure
1 ; whereas for others it will be achieved, if at all, when P is built somewhere else, as in Figure
2 . Determine the smallest value of k for which the lowest total cost is achieved by building P
at C.

Solution: Let x be the directed distance from C to the right along the coast. The
total cost K in units of 105 is given by

K(x) = (12 + (x + 1)2 )1/2 + (12 + (3 − x)2 )1/2 + k ∣x∣

for −∞ < x < ∞, where the distances are measured in units of 10 km. Then:
⎧
⎪
⎪(x + 1)(1 + (x + 1)2 )−1/2 + (x − 3)(1 + (x − 3)2 )−1/2 + k if x > 0
K ′ (x) = ⎨
⎪
⎩(x + 1)(1 + (x + 1) )
2 −1/2 + (x − 3)(1 + (x − 3)2 )−1/2 − k if x < 0
⎪

33
 √ √
In particular, if k < 3/ 10 − 1/ 2, then

3 1
lim+ K ′ (x) = −( √ − √ ) + k < 0
x→0 10 2

and therefore K ′ is negative on some interval (0, a). Hence K is decreasing on [0, a]
as it is continuous, and the absolute minimum value of K does not occur at x = 0.

Next we observe that

K ′′ (x) = (1 + (x + 1)2 )−1/2 − (x + 1)2 (1 + (x + 1)2 )−3/2

+ (1 + (x − 3)2 )−1/2 − (x − 3)2 (1 + (x − 3)2 )−3/2
= (1 + (x + 1)2 )−3/2 + (1 + (x − 3)2 )−3/2 > 0

for all x =/ 0, and therefore K ′ is increasing on (−∞, 0) and (0, ∞).

√ √
In particular, if k = 3/ 10−1/ 2, then lim+ K ′ (x) = 0, and hence K ′ (x) > 0 for x > 0;
√ √ x→0
and lim− K ′ (x) = −2(3/ 10 − 1/ 2) < 0, and hence K ′ (x) < 0 for x < 0. Therefore,
x→0
K has its absolute minimum value at x = 0.
√ √
We conclude that 3/ 10 − 1/ 2 is the smallest value of k for which the absolute
minimum of K occurs at C.

38. Let f be a continuous function.

x2
a. Find f (4) if ∫ f (t) dt = x sin πx for all x.
0
f (x)
b. Find f (4) if ∫ t2 dt = x sin πx for all x.
0

FTC1
x2 d x d
2
↓
Solution: a. ∫ f (t) dt = x sin πx Ô⇒ ∫ f (t) dt = (x sin πx) Ô⇒ f (x2 ) ⋅ 2x = sin πx + x ⋅
0 dx 0 dx
Now letting x = 2 we get f (4) = π/2.
f (x)
f (x) t3 f (x)3
b. ∫ t dt = x sin πx Ô⇒ ] = x sin πx Ô⇒
2
= x sin πx. Hence x = 4
0 3 0 3
gives f (4) = 0.

√
Remark: One might ask if such functions exist. In part (b), f (x) = 3 3x sin πx is the unique
function satisfying the given condition.
√
sin π x π √
In part (a), f (x) = √ + cos π x for x > 0, and f (0) = π by continuity; but it can be
2 x 2
anything for x < 0 so long as it is continuous.

34
 d2 y
39. Compute ∣ if y is a differentiable function of x satisfying the equation:
dx2 (x,y)=(0,0)
x+y
e−t dt = xy
2
∫0

Solution: Differentiating the equation with respect to x we obtain:

x+y
e−t dt = xy
2
∫0

Ô⇒
d/dx

d x+y d
e−t dt =
2
∫ (xy)
dx 0 dx

Ô⇒
FTC1

d d
e−(x+y) ⋅
2
(x + y) = (xy)
dx dx Ô⇒

dy dy
e−(x+y) (1 +
2
)=y+x (e)
dx dx
Ô⇒

x = 0, y = 0

dy
= −1 at (x, y) = (0, 0)
dx

Now differentiating (e) with respect to x again we get:

dy dy
e−(x+y) (1 +
2
)=y+x
dx dx
Ô⇒

d/dx

d dy d dy
(e−(x+y) (1 + )) =
2
(y + x )
dx dx dx dx
Ô⇒

2 2
−(x+y)2 dy 2 d y dy dy d2 y
e (−2(x + y))(1 + ) + e−(x+y) = + + x
dx dx2 dx dx dx2
Ô⇒

x = 0, y = 0, dy/dx = −1

d2 y
= −2 at (x, y) = (0, 0)
dx2

35
40. Suppose that f is a continuous function satisfying
x
f (x) = x ∫ f (t) dt + x3
0

for all x, and c is a real number such that f (c) = 1. Express f ′ (c) in terms of c only.

Solution: We have:
x x FTC1 x
d d ↓
f (x) = ∫ f (t) dt + x ∫ f (t) dt + 3x3 = ∫ f (t) dt + xf (x) + 3x2
dx 0 dx 0 0
c
Now substituting x = c and using the facts that f (c) = c ∫ f (t) dt + c3 and f (c) = 1
0
we obtain:
c f (c) − c3 1
f ′ (c) = ∫ f (t) dt + cf (c) + 3c2 = + c + 3c2 = + c + 2c2
0 c c

2 /2
Remark: It can be shown that f (x) = 2x(ex − 1) is the only function that satisfies the given
condition.

x
∫ sin(xt ) dt
3
41. Evaluate the limit lim 0 .
x→0 x5

Solution: We first make the change of variable t = x−1/3 u, dt = x−1/3 du, to obtain:
x x4/3
−1/3
∫0 sin(xt ) dt = x sin(u3 ) du
3
∫0

Now we have:
x x4/3
∫0 sin(xt ) dt x−1/3 ∫0 sin(u3 ) du
3
lim = lim
x→0 x5 x→0 x5
x4/3
∫ sin(u3 ) du
= lim 0
x→0 x16/3
d x4/3
∫0 sin(u ) du
L’H
3
↓ dx
= lim
x→0 16/3 x13/3
FTC1
↓ sin(x4 ) ⋅ 4/3 x1/3
= lim
x→0 16/3 x13/3
sin(x4 ) ⋅ 4/3 x1/3
= lim
x→0 16/3 x13/3
1 sin(x4 )
= lim
4 x→0 x4
1 1
= ⋅1=
4 4

36
42. Suppose that f is a twice-differentiable function satisfying f (0) = −2, f ′ (0) = 11, f ′′ (0) =
−8, f (2) = 5, f ′ (2) = −3, f ′′ (2) = 7 ; and also suppose that the function
1 x
g(x) = f (t) dt
x ∫0
has a critical point at x = 2 . Determine whether the critical point of g at x = 2 is a local
minimum, a local maximum or neither.

Solution: We first observe that

d d 1 x
g(x) = ( f (t) dt)
dx dx x ∫0
FTC1 x x
↓ 1 1d
= − ∫ f (t) dt + ∫ f (t) dt
x2 0 x dx 0
1 x 1
= − 2 ∫ f (t) dt + f (x)
x 0 x

2
and as g ′ (2) = 0 we must have −1/4 ∫ f (t) dt + 1/2 f (2) = 0, and hence
0
2
∫0 f (t) dt = 2f (2) = 10.

Now differentiating a second time we obtain that:

d2 d 1 x 1
g(x) = ( − ∫ f (t) dt + f (x))
dx2 dx x2 0 x
FTC1 x x
↓ 2 1 d 1 1 ′
= ∫ f (t) dt − ∫ f (t) dt − f (x) + f (x)
x3 0 x2 dx 0 x2 x
2 x 1 1 1
= 3 ∫ f (t) dt − 2 f (x) − 2 f (x) + f ′ (x)
x 0 x x x
Substituting x = 2 in this we get
1 2 1 1 3
g ′′ (2) = ∫ f (t) dt − f (2) + f ′ (2) = − < 0
4 0 2 2 2
and conclude that g has a local maximum at x = 2.

43. Suppose that f is a continuous and positive function on [0, 5], and the area between the
graph of y = f (x) and the x-axis for 0 ≤ x ≤ 5 is 8. Let A(c) denote the area between the graph
of y = f (x) and the x-axis for 0 ≤ x ≤ c, and let B(c) denote the area between the graph of
dR
y = f (x) and the x-axis for c ≤ x ≤ 5. Let R(c) = A(c)/B(c). If R(3) = 1 and ∣ = 7, find
dc c=3
f (3).

Solution: We have A(3) + B(3) = 8 and R(3) = 1 Ô⇒ A(3) = B(3), implying

A(3) = B(3) = 4.

37
 As
c 5
A(c) = ∫ f (t) dt and B(c) = ∫ f (t) dt ,
0 c
differentiating these with respect to c and using the Fundamental Theorem of
Calculus Part 1 we obtain A′ (c) = f (c) and B ′ (c) = −f (c). In particular,
A′ (3) = f (3) and B ′ (3) = −f (3).

On the other hand,

d d A(c) A′ (c)B(c) − A(c)B ′ (c)
R(c) = =
dc dc B(c) B(c)2
and letting c = 3 gives
d A′ (3)B(3) − A(3)B ′ (3) f (3)B(3) + A(3)f (3) f (3)
7= R(c)∣ = = =
dc c=3 B(3)2 B(3)2 2

and hence f (3) = 14.

44. Let f be a continuous function and let

1
g(x) = ∫ f (t) ∣x − t∣ dt .
−1

Express g ′′ (x) in terms of f (x) for −1 < x < 1.

Solution: For −1 < x < 1 we rewrite the definition of g(x) as follows:

x 1
g(x) = ∫ f (t) (x − t) dt + ∫ f (t) (t − x) dt
−1 x
x x 1 1
= x∫ f (t) dt − ∫ f (t)t dt + ∫ f (t)t dt − x ∫ f (t) dt
−1 −1 x x

Now we differentiate using the FTC1 to obtain:

x d x d x
g ′ (x) = ∫ f (t) dt + x ∫ f (t) dt − ∫ f (t)t dt
−1 dx −1 dx −1
d 1 1 d 1
+ ∫ f (t)t dt − ∫ f (t) dt − x ∫ f (t) dt
dx x x dx x
x 1
=∫ f (t) dt + xf (x) − f (x)x − f (x)x − ∫ f (t) dt + xf (x)
−1 x
x 1
=∫ f (t) dt − ∫ f (t) dt
−1 x

Using FTC1 again we compute g ′′ (x) to find

d x d 1
g ′′ (x) = ∫ f (t) dt − ∫ f (t) dt
dx −1 dx x
= f (x) + f (x)
= 2f (x)

for −1 < x < 1.

38
45. Suppose that a continuous function f satisfies the equation
x 1
f (x) = x2 − x + (1 − x) ∫ t2 f (t) dt + x ∫ (t − t2 )f (t) dt
0 x

for all x. Express f ′′ (1/2) in terms of A = f (1/2).

Solution: We have
x 1
f ′ (x) = 2x − 1 − ∫ t2 f (t) dt + (1 − x)x2 f (x) + ∫ (t − t2 )f (t) dt − x(x − x2 )f (x)
0 x
x 1
= 2x − 1 − ∫ t2 f (t) dt + ∫ (t − t2 )f (t) dt
0 x

for all x, where we used the Fundamental Theorem of Calculus Part 1 twice:
x FTC1
d ↓
∫ t 2
f (t) dt = x2 f (x)
dx 0
1 FTC1
d ↓
(t − 2
)f (t) = −(x − x2 )f (x)
dx ∫x
t dt

Differentiating again and using FTC1 similarly we obtain:

f ′′ (x) = 2 − x2 f (x) − (x − x2 )f (x) = 2 − xf (x)

for all x . This gives f ′′ (1/2) = 2 − 1/2 ⋅ f (1/2) = 2 − A/2 .

1 1 1
46. Evaluate the limit lim (n ( + +⋯+ )) .
n→∞ (2n + 1)2 (2n + 3)2 (4n − 1)2

1
Solution: Consider the function f (x) = on the interval [2, 4]. If we divide this
x2
2 2k
interval into n subintervals of equal length using the points xk = 2 + , 0 ≤ k ≤ n,
n n
2k − 1
and choose our sample points to be the midpoints ck = 2 + , 1 ≤ k ≤ n, of these
n
subintervals, the Riemann sum for these data becomes
n n
2k − 1 2 n
2n
∑ f (ck ) ∆xk = ∑ f (2 + )⋅ = ∑
k=1 k=1 n n k=1 (2n + 2k − 1)2

and the definition of the definite integral gives

n
2n 4 dx 1 4 1 1 1
lim ∑ = ∫ = − ] = − + = .
k=1 (2n + 2k − 1)
n→∞ 2 2 x2 x 2 4 2 4

Therefore
1 1 1 1
lim (n ( + +⋯+ )) = .
n→∞ (2n + 1)2 (2n + 3)2 (4n − 1)2 8

39
47. When pirates retire, they live on the Square Island which has the shape of a square with
10 hectometer (=hm) long sides. Having lived all their lives on it, the retired pirates want to
be as far away from the sea as possible. As a result, the pirate population density p(x) at a
point on the Square Island is proportional to the distance x of the point from the shore and
reaches its largest value of 15 pirate/hm2 at the center of the island. Find the total number N
of the pirates on the island.

Solution: Firstly, we have p(x) = 3x. Then we observe that the points on the
island whose distance to the shore are x hm lie on a square whose sides are parallel
to the sides of the island and 10 − 2x hm long.

Now consider another square which consists of points whose distances to the shore
are x+∆x hm for some small positive ∆x, and choose a x∗ which lies between x and
x + ∆x. Then the number of pirates living in the strip between these two squares is
approximately 4 ⋅ (10 − 2x) ⋅ p(x∗ ) ⋅ ∆x.

40
 Hence N can be approximated by the Riemann sum
n
N ≈ ∑ 4 ⋅ (10 − 2xi ) ⋅ p(x∗i ) ⋅ ∆xi
i=1

for a partition 0 = x0 < x1 < ⋯ < xn−1 < xn = 5 of the interval [0, 5] and sample points
x∗i in [xi−1 , xi ] for 1 ≤ i ≤ n, and in the limit we obtain:
5
N =∫ 4(10 − 2x)p(x) dx
0

This gives:
5 5 5 1 5
N = ∫ 4(10 − 2x)(3x) dx = 24 ∫ (5x − x2 ) dx = 24 [ x2 − x3 ] = 500
0 0 2 3 0

There are 500 pirates living on the island.

Remark: Imagine that the entire landscape of the Square Island consists of just a mountain
and the altitude of a point that is horizontally x hm away from the shore is p(x) hm. Then
the island has the shape of a square pyramid with a 10 hm by 10 hm base and 15 hm height,
and N is now the volume of the island. Hence we have:
1 1
N = (height)(area of the base) = ⋅ 15 ⋅ 102 = 500
3 3

Remark: Compare this example with Example 36 in Part 2.

41
48. An island has the shape of a 10 hm×10 hm square and its landscape consists of a mountain
whose height h at a horizontal distance x from the shore is given by h = x2 where both h and
x are measured in hectometers (=hm). Let V be the volume of the mountain.

a. Express V as an integral with respect to h by considering cutting the mountain into slices
as shown in Figure a.
b. Express V as an integral with respect to x by considering cutting the mountain into
shells as shown in Figure b.
Figure a: Figure b:

c. Compute V .

Solution: a. A horizontal slice at a height√ h is approximately a prism with a

square base of side length 10 − 2x = 10 − 2 h√and height ∆h , and therefore its
volume is approximately (10 − 2x)2 ∆h = (10 − 2 h)2 ∆h . Hence the volume V can
be approximated by the Riemann sum
n √
V ≈ ∑(10 − 2 h∗i )2 ∆hi
i=1

for a partition 0 = h0 < h1 < ⋯ < hn−1 < hn = 25 of the interval [0, 25] and sample
points h∗i in [hi−1 , hi ] for 1 ≤ i ≤ n , and in the limit we obtain:
25 √
V = ∫ (10 − 2 h)2 dh
0

42
 b. A shell at a horizontal distance x from the shore consists of four pieces each of
which is approximately a rectangular prism of dimensions x2 × (10 − 2x) × ∆x where
∆x is the thickness of the shell. Therefore its volume is approximately 4 (10 −
2x) x2 ∆x . Hence the volume V can be approximated by the Riemann sum
n
V ≈ ∑ 4 (10 − 2x∗i ) (x∗i )2 ∆xi
i=1

for a partition 0 = x0 < x1 < ⋯ < xn−1 < xn = 5 of the interval [0, 5] and sample points
x∗i in [xi−1 , xi ] for 1 ≤ i ≤ n , and in the limit we obtain:
5
V = ∫ 4 (10 − 2x) x2 dx
0

c. We use the integral in Part c to compute the volume:

5 5 5 1 5 1250
V = ∫ 4 (10 − 2x) x2 dx = 8 ∫ (5x2 − x3 ) dx = 8 [ x3 − x4 ] = hm3
0 0 3 4 0 3

49. Evaluate the following integrals.

a. ∫ x sin(x2 ) cos(x2 ) dx
1 √
b. ∫ x 1 − x dx
0

Solution: a. Let u = sin(x2 ). Then du = 2x cos(x2 ) dx, and

1 1 u2 1
∫ x sin(x 2
) cos(x 2
) dx = ∫ u du = ⋅ + C = sin2 (x2 ) + C .
2 2 2 4

b. Let u = 1 − x. Then du = −dx, and x = 0 ⇒ u = 1, x = 1 ⇒ u = 0. Therefore:

1 √ 0 1
∫ x 1 − x dx = ∫ (1 − u)u (−du) = ∫ (u − u ) du
1/2 1/2 3/2
0 1 0
1
u3/2 u5/2 2 2 4
=[ − ] = − =
3/2 5/2 0
3 5 15

43
Remark: In part (a), if we let u = cos(x2 ), then we obtain the answer
1
∫ x sin(x ) cos(x ) dx = − 4 cos (x ) + C ;
2 2 2 2

1
and if we first observe that sin(x2 ) cos(x2 ) = sin(2x2 ), and then let u = sin(2x2 ), then we
2
obtain the answer
1
∫ x sin(x ) cos(x ) dx = − 8 cos(2x ) + C .
2 2 2

These are all correct answers. If we write

1
∫ x sin(x ) cos(x ) dx = 4 sin (x ) + C1
2 2 2 2

1
∫ x sin(x ) cos(x ) dx = − 4 cos (x ) + C2
2 2 2 2

1
∫ x sin(x ) cos(x ) dx = − 8 cos(2x ) + C3
2 2 2

1 1
then C2 = C1 + and C3 = C1 + .
4 8

a f (x) a
50. Show that ∫ dx = for any positive continuous function on [0, a] .
0 f (x) + f (a − x) 2
a f (x)
Solution: Let I = ∫ dx .
0 f (x) + f (a − x)

Let u = a − x. Then du = −dx, x = 0 ⇒ u = a, x = a ⇒ u = 0, and

a f (x) 0 f (a − u) a f (a − x)
I =∫ dx = ∫ (−du) = ∫ dx .
0 f (x) + f (a − x) a f (a − u) + f (u) 0 f (x) + f (a − x)

Therefore
f (x)
a a f (a − x)
2I = ∫ dx + ∫ dx
0 f (x) + f (a − x) 0 f (x) + f (a − x)
a f (x) + f (a − x)
=∫ dx
0 f (x) + f (a − x)
a
=∫ dx = a
0

a
and I = .
2

(See the remark on the next page.)

44
Remark: Here is an explanation of what is going on with no integral signs: Consider the
rectangle with a vertex at the origin, and sides along the positive x- and the y-axes with
lengths a and 1, respectively. The graph of y = f (x)/(f (x) + f (a − x)) is symmetric with
respect to the center (a/2, 1/2) of this rectangle, and therefore divides it into two regions of
equal area. Since the area of the rectangle is a and I is the area of the lower half, we have
a
I= .
2

51. Let R be the region bounded by the parabola y = x − x2 and the x-axis, and let V be the
volume of the solid generated by revolving R about the x-axis.
a. Express V as an integral using the disk method. (Do not compute! )
b. Express V as an integral using the cylindrical shell method. (Do not compute! )

Solution: The parabola intersects the x-axis at x = 0 and x = 1. Therefore we have

1 1
V = π ∫ R(x)2 dx = ∫ (x − x2 )2 dx
0 0

for part (a).

For part (b), the cylindrical shell method gives a y-integral,

d
V = 2π ∫ (radius of the shell)(height of the shell) dy
c

45
as the region is revolved about the x-axis. We have c = 0 and d = 1/4 , the y-
coordinate of the highest point of the parabola. The radius of the shell is the vertical
distance from the red rectangle in the figure to the x-axis, which is y. The height
of the shell is the horizontal length of the rectangle; that is, the difference between
the x-coordinates of the right and the left sides of the √ rectangle. By solving
√ the
1 + 1 − 4y 1 − 1 − 4y
equation y = x−x2 for x, we find these values as x = and x = ,
2 2
respectively. Hence
√ √
1/4 1 + 1 − 4y 1 − 1 − 4y
V = 2π ∫ y( − ) dy .
0 2 2

46
52. Let R be the region bounded by the curve y 2 = x2 − x4 . Let V be the volume obtained by
rotating R about the x-axis. Let W be the volume obtained by rotating R about the y-axis.

a. Express V using both the disk method and the cylindrical shells method.
b. Express W using both the disk method and the cylindrical shells method.
c. Compute V and W .

Solution: By symmetry, V is 2 times the volume obtained by revolving the portion

of R lying in the first quadrant about the x-axis. In the disk method, the red
vertical rectangles, when revolved about the x-axis, form the disks that are
√ used in
the computation of V , and therefore the radii of the disks are given by x2 − x4 .
Hence:
1 1 √
V = 2 ⋅ π ∫ (radius of disk)2 dx = 2 ⋅ π ∫ ( x2 − x4 )2 dx
0 0

Again by symmetry, W is 2 times the volume generated by revolving the portion

of R lying in the first quadrant about the y-axis. This time in the washer method,
the green horizontal rectangles, when revolved about the y-axis, form the washers
that are used in the computation of W .

To find the outer and the inner radii of the washers we have to solve y 2 =√
x2 −x4 for x.
Applying the quadratic formula to (x ) −x +y = 0 we obtain x = (1± 1 − 4y 2 )/2,
2 2 2 2 2

47
and this gives us
√ √ √ √
x= (1 + 1 − 4y 2 )/2 and x = (1 − 1 − 4y 2 )/2

as the two nonnegative solutions. Hence:

1/2
W = 2 ⋅ π∫ ((outer radius of washer)2 − (inner radius of washer)2 ) dy
0
¿ ¿
Á 1 + √1 − 4y 2 2 Á 1 − √1 − 4y 2 2
1/2 Á
À Á
À
= 2 ⋅ π∫ (( ) −( ) ) dy
0 2 2

Now we consider the cylindrical shells method for both volumes. When revolved
about the y-axis the red vertical rectangles generate the cylindrical shells that are
used
√ in the computation of W , and therefore the heights of these shells are given
by x2 − x4 . Hence:
1 1 √
W = 2 ⋅ 2π ∫ (radius of shell)(height of shell) dx = 2 ⋅ 2π ∫ x x2 − x4 dx
0 0

When revolved about the x-axis the green horizontal rectangles generate the
cylindrical shells that are used
√ in√the computation
√ of√V . Therefore the heights
of these shells are given by (1 + 1 − 4y )/2 − (1 − 1 − 4y 2 )/2. Hence:
2

1/2
V = 2 ⋅ 2π ∫ (radius of shell)(height of shell) dy
0
¿ ¿
Á 1 + √1 − 4y 2 Á 1 − √1 − 4y 2
1/2 Á
À Á
À
= 2 ⋅ 2π ∫ y( − ) dy
0 2 2

Finally we compute the volumes. To compute V we use the integral we obtained

with the disk method,
1 x3 x5 1 4π
V = 2 ⋅ π ∫ (x − x ) dx = 2π [ − ] =
2 4
0 3 5 0 15

48
 and to compute W we use the integral we obtained with the washer method
√ √
1/2 1 + 1 − 4y 2 1 − 1 − 4y 2
W = 2 ⋅ π∫ ( − ) dy
0 2 2
1/2 √
= 2π ∫ 1 − 4y 2 dy
0
π/2 1
= 2π ∫ cos θ ⋅ cos θ dθ
0 2
π/2
= π∫ cos2 θ dθ
0
1 + cos 2θ
π/2
= π∫ dθ
0 2
θ sin 2θ π/2
= π[ + ]
2 4 0
π2
=
4
1 1
where we made the change of variable y = sin θ, dy = cos θ dθ.
2 2

Remark: Compare this example with Example 43 in Part 2.

53. A water tank has a bottom consisting of a disk of radius a with 0 ≤ a < 3, and a side surface
having the shape generated by revolving the graph of a continuous nonnegative function x = g(y)
for 0 ≤ y ≤ 5 with g(0) = a and g(5) = 3 about the y-axis where all units are in meters. Assume
that:
○1 As water runs out of a small hole at the bottom of tank, the speed of the water flowing
through the hole at any moment is proportional to the square root of the depth of the water
in the tank at that moment.
2 The function g and the constant a are chosen in such a way that the depth of the water
○
changes at a constant rate at all times.
Find the volume of the tank.

49
Solution: Let V (h) denote the volume of the water when the depth is h. Then:
h
V (h) = π ∫ g(y)2 dy
0

Differentiating this with respect to time t and using the Fundamental Theorem of
Calculus, Part 1 we obtain:
dV dh
= πg(h)2
dt dt
√
Using the condition ○ 1 which says that dV /dt = −k1 h for some positive constant
k1 and the condition ○ 2 which says that dh/dt = −k2 for some positive constant
k2 , we conclude that g(h) = k ⋅ h1/4 for some positive constant k. In particular,
a = g(0) = 0, and g(5) = 3 gives k = 3/51/4 . Therefore g(h) = 3h1/4 /51/4 .

Hence the volume of the tank is:

2 5
5 3 9π y 3/2
V (5) = π ∫ ( y 1/4
) dy = 1/2 ⋅ ] = 30π m3
0 51/4 5 3/2 0

50
54. Let V be the volume of the water-dropper shown in the figure on the next page which has
the shape obtained by revolving the curve x4 + y 4 = 1 about the line x = −5/2 where all units
are in centimeters.
a. Express V as an integral using the cylindrical shells method.
b. Express V as an integral using the washer method.
1
c. Show that the improper integral ∫ u−3/4 (1 − u)1/4 du converges.
0
1
d. Express V in terms of A = ∫ u−3/4 (1 − u)1/4 du .
0

Solution: a. The radius and height of the cylindrical shells are x − (−5/2) and
(1 − x4 )1/4 − (−(1 − x4 )1/4 ), respectively. Hence:
1 1 5
V = 2π ∫ (radius) ⋅ (height) dx = 2π ∫ (x + ) ⋅ ((1 − x4 )1/4 − (−(1 − x4 )1/4 ) dx
−1 −1 2

b. The outer and inner radii of the washers are 5/2 + (1 − y 4 )1/4 and 5/2 − (1 − y 4 )1/4 ,
respectively. Hence:
1
V = π ∫ ((outer radius)2 − (inner radius)2 ) dy
−1
2 2
1 5 5
= π∫ (( + (1 − y 4 )1/4 ) − ( − (1 − y 4 )1/4 ) ) dy
−1 2 2

c. We have 0 < u ≤ 1 Ô⇒ 0 ≤ 1 − u < 1 Ô⇒ 0 ≤ (1 − u)1/4 < 1 Ô⇒ 0 ≤ u−3/4 (1 − u)1/4 <

1
u−3/4 for all 0 < u ≤ 1. Since ∫ u−3/4 du is convergent (because p = 3/4 < 1), so is
0
1
−3/4
∫0 u (1 − u) du by the Comparison Test.
1/4

d. We use the integral in part (b) and symmetry

1
V = 10π ∫ (1 − y 4 )1/4 dy
−1
1
= 20π ∫ (1 − y 4 )1/4 dy
0
1 1 −3/4
= 20π ∫ (1 − u)1/4 ⋅ u du
0 4
1
= 5π ∫ u−3/4 (1 − u)1/4 du
0
= 5πA

as well as the change of variable y = u1/4 , dy = 1/4 ⋅ u−3/4 du.

51
52
55. Find the absolute maximum and the absolute minimum values of f (x) = (x2 − 3)ex on the
interval [−2, 2].

Solution: f ′ (x) = 2xex + (x2 − 3)ex = (x2 + 2x − 3)ex . We want to solve f ′ (x) = 0.

(x2 + 2x − 3)ex = 0 Ô⇒ x2 + 2x − 3 = 0 Ô⇒ x = 1, −3 .

Since −3 does not belong to the interval [−2, 2], the only critical point is x = 1.
Hence we are going to compute f at the points x = 1,−2, 2.

f (1) = −2e, f (−2) = e−2 , and f (2) = e2 . Since e > 1 we have e2 > e−2 > −2e. Therefore
the absolute maximum is e2 and the absolute minimum is −2e.

56. Find the absolute maximum and the absolute minimum values of x1/x .
ln x
Solution: Let y = x1/x . Then ln y = . Differentiating this with respect to x we
x
obtain
1
1 dy d d ln x x ⋅ x − ln x ⋅ 1 1 − ln x d 1/x 1 − ln x
= ln y = = 2
= 2
Ô⇒ x = x1/x .
y dx dx dx x x x dx x2
Since 1 − ln x > 0 for 0 < x < e and 1 − ln x < 0 for x > e, the absolute maximum value
of x1/x occurs at x = e and is e1/e . x1/x has no absolute minimum value.

Remark: Although the reasoning above does not require it, let us also look at what happens
at the endpoints of the domain. Since

ln x L’H
↓ 1/x
lim ln y = lim = lim =0,
x→∞ x→∞ x x→∞ 1

we have lim x1/x = lim y = lim eln y = e0 = 1. On the other hand, lim+ x1/x = 0 as 0∞ = 0.
x→∞ x→∞ x→∞ x→0

1 n k/n
57. Evaluate the limit lim ( ∑2 ).
n→∞ n k=1
n n
1
Solution: ∑ 2k/n is a Riemann sum ∑ f (ck ) ∆xk for f (x) = 2x on [0, 1] for the
k=1 n k=1
k k
partition xk = , 0 ≤ k ≤ n, and the sample points ck = , 1 ≤ k ≤ n. Therefore,
n n
1
1 n 1 2x 1
lim ( ∑ 2k/n ) = ∫ 2x dx = ] = .
n→∞ n 0 ln 2 0 ln 2
k=1

53
Remark: Alternatively, we can use the formula for the sum of a finite geometric series
n
2(n+1)/n − 21/n 21/n
∑ 2k/n = = ,
k=1 21/n − 1 21/n − 1

and then

1 n k/n 21/n t2t L’H

↓ 2t + t ⋅ ln 2 ⋅ 2t 1
lim ( ∑ 2 ) = n→∞
lim = lim = lim = .
n→∞ n k=1 n(2 − 1) t→0 2 − 1
1/n + t t→0 +
ln 2 ⋅ 2 t ln 2

2
sin x 1/x
58. Evaluate the limit lim ( ) .
x→0 x
sin x 1
Solution: Since lim = 1 and lim 2 = ∞, this is limit has the indeterminate
x→0 x x→0 x
form 1∞ .
sin x
1/x2 ln (
)
sin x x
Let y = ( ) . Then ln y = . As x → 0, this will have the indeterminate
x x2
0
form and we can use L’Hôpital’s Rule.
0
ln(sin x) − ln x
lim ln y = lim
x→0 x→0 x2
cos x 1
L’H − x cos x − sin x
↓
= lim sin x x = lim
x→0 2x x→0 2x2 sin x
L’H
↓ cos x − x sin x − cos x − sin x
= lim = lim
x→0 4x sin x + 2x cos x
2 x→0 4 sin x + 2x cos x
sin x
− −1 1
= lim x = =−
x→0 sin x 4+2 6
4 + 2 cos x
x
Here applications of L’Hôpital’s Rule are indicated with L’H . Then
2
sin x 1/x
lim ( ) = lim y = lim eln y = e−1/6
x→0 x x→0 x→0

using the continuity of the exponential function.

54
Remark: The first example of an application of L’Hôpital’s Rule in Guillaume François
Antoine Marquis de L’Hôpital’s book Analyse des Infiniment Petits pour l’Intelligence des
Lignes Courbes of 1696:

There is a typo. Can you find it?

cos(2x) − e−2x
2

59. Evaluate lim .

x→0 sin4 x
Solution:
cos(2x) − e−2x cos(2x) − e−2x
2 2
x4
lim = lim ( ⋅ )
x→0 sin4 x x→0 x4 sin4 x
cos(2x) − e−2x
2
x 4
= lim ⋅ (lim )
x→0 x4 x→0 sin x

cos(2x) − e−2x
2

= lim
x→0 x4
−2 sin(2x) + 4xe−2x
L’H 2
↓
= lim
x→0 4x3
−4 cos(2x) + 4e−2x − 16x2 e−2x
L’H 2 2
↓
= lim
x→0 12x2
8 sin(2x) − 48xe−2x + 64x3 e−2x
L’H 2 2
↓
= lim
x→0 24x
sin(2x) 8
− 2e−2x + x2 e−2x )
2 2
= lim (
x→0 3x 3
2 4
= −2+0=−
3 3

Remark: It is easier to solve this problem using the Taylor series, which will be seen in Calculus

55
II:
cos(2x) − e−2x
2

lim
x→0 sin4 x
(2x)2 (2x)4 (2x)6 (−2x2 )2 (−2x2 )3
(1 − + − + ⋯) − (1 + (−2x2 ) + + + ⋯)
2! 4! 6! 2! 3!
= lim 4
x→0 x3
(x − + ⋯)
3!
4 56
− x4 + x6 + ⋯
= lim 3 45
x→0 2
x4 − x6 + ⋯
3
4 56 2
− + x +⋯
= lim 3 45
x→0 2
1 − x2 + ⋯
3
4
=−
3

sin(x + ax3 ) − x
60. Find the value of the constant a for which the limit lim exists and compute
x→0 x5
the limit for this value of a.
Solution: We have
sin(x + ax3 ) − x
lim
x→0 x5
L’H
↓ cos(x + ax3 )(1 + 3ax2 ) − 1
= lim
x→0 5x4
L’H
↓ − sin(x + ax3 )(1 + 3ax2 )2 + cos(x + ax3 )(6ax)
= lim
x→0 20x3
L’H
↓ − cos(x + ax3 )(1 + 3ax2 )3 − sin(x + ax3 ) ⋅ 3(1 + 3ax2 )(6ax) + cos(x + ax3 )(6a)
= lim .
x→0 60x2
The numerator of the fraction inside this last limit goes to −1+6a as x → 0. Therefore
the limit does not exist unless a = 1/6. If a = 1/6 then
sin(x + x3 /6) − x
lim
x→0 x5
− cos(x + x3 /6)(1 + x2 /2)3 − sin(x + x3 /6) ⋅ 3(1 + x2 /2)x + cos(x + x3 /6)
= lim
x→0 60x2
is the sum of
cos(x + x3 /6)(1 − (1 + x2 /2)3 )
lim
x→0 60x2
cos(x + x3 /6)(3/2 + 3x2 /4 + x4 /8) 1
= − lim =−
x→0 60 40
and
− sin(x + x3 /6) ⋅ 3(1 + x2 /2)x
lim
x→0 60x2
sin(x + x3 /6) (1 + x2 /2)(1 + x2 /6) 1
= − lim ⋅ lim = − .
x→0 x + x3 /6 x→0 20 20

56
 Hence
sin(x + x3 /6) − x 3
lim 5
=− .
x→0 x 40

Remark: Once again there are shorter ways of doing this. If we use the Taylor series then
(x + ax3 )3 (x + ax3 )5
sin(x + ax3 ) = (x + ax3 ) − + −⋯
3! 5!
1 1 1
= x + (a − ) x3 + ( − a) x5 + ⋯
6 120 2
and it is immediate that the limit exists exactly when a = 1/6 and then its value is −3/40.

61. Let b > a > 0 be constants. Find the area of the surface generated by revolving the circle
(x − b)2 + y 2 = a2 about the y-axis.

Solution: We have
¿
Á 2
xÁ
À1 + ( dx ) dy
d
Surface Area = 2π ∫
c dy

√ by revolving a curve about the y-axis. We have x = b +

for a surface generated
√
a − y and x = b − a2 − y 2 for the right and left halves of the circle, respectively.
2 2

Then ¿
Á 2
dx
= ∓√
y
Ô⇒ Á
À dx
1+( ) = √
a
.
dy a −y
2 2 dy a − y2
2

Hence
a √ a
Surface Area = 2π ∫ (b + a2 − y 2 ) √ dy
−a a2 − y 2
a √ a
+ 2π ∫ (b − a2 − y 2 ) √ dy
−a a2 − y 2
a 1
= 4πab ∫ √ dy
−a a2 − y 2
a
y
= 4πab arcsin ( ) ]
a −a
π π
= 4πab (arcsin 1 − arcsin(−1)) = 4πab ( − ( − )) = 4π 2 ab
2 2

57
Remark: The surface generated by revolving a circle about a line (in the same plane) that
does not intersect it is called a torus.

dy
62. Find y(1) if = xy 2 and y(0) = 1.
dx
Solution: We have
dy dy dy 1 1
= xy 2 Ô⇒ 2 = x dx Ô⇒ ∫ 2 = ∫ x dx Ô⇒ − = x2 + C .
dx y y y 2
2
Then y(0) = 1 Ô⇒ −1 = C. Hence y = . This in turn gives y(1) = 2.
2 − x2

Remark: What would your answer be if the question asked y(2)?

63. Nitrogen dioxide is a reddish-brown gas that contributes to air pollution, and also gives the
smog its color. Under sunlight it decomposes producing other pollutants, one of which is ozone.
As nitrogen dioxide decomposes, its density decreases at a rate proportional to the square of
the density. Suppose that the density of nitrogen dioxide Q is 1/2 grams per liter at time t = 0
minutes and 6/25 grams per liter at time t = 3 minutes. Find Q when t = 15 minutes. (Assume
that no new nitrogen dioxide is added to the environment.)

Solution: We have dQ/dt = −kQ2 where k is a positive constant. Then dQ/Q2 =

−k dt and integrating we obtain −1/Q = −kt + C where C is a constant.

Letting t = 0 minutes we find −1/(1/2) = −1/Q(0) = C, hence C = −2 liters per gram.

Now letting t = 3 min gives −1/(6/25) = −1/Q(3) = −k ⋅ 3 − 2 and hence k = 13/18
liters per gram per minute. Finally we obtain Q = 18/(13t + 36).

Using this we compute the density of nitrogen dioxide as Q(15) = 6/77 grams per
liter after 15 minutes.

64. A pool, like the one in front of the Faculty of Science Building A, loses water from its sides
and its bottom due to seepage. For a pool with radius R and depth H in meters, the rate of
this loss in m3/hour is given by
dV
= −aR2 h − bRh2
dt
where V is the volume of the water in cubic meters, t is the time in hours, h is the depth of
the water in meters, and a and b are constants independent of R, H and h.
Consider a pool with H = 1 m, R = 6 m and a = b = π/500 1/hour.

H
h

58
 a. Find the depth of the water h as a function of time t if h = 1 m when t = 0 hour.
b. The term aR2 h on the right side of the equation represents the rate of loss due to seepage
through the bottom of the pool. Find the total volume of the water that seeps through the
bottom while the initially full pool completely empties.

Solution: a. We have:
π 2 π
V = πR2 h Ô⇒ πR2 h = − R h− Rh2
500 500
1 2 1
Ô⇒ 62 ⋅ h = − 6 ⋅h− 6 ⋅ h2
500 500
6 dh dt
Ô⇒ =−
h(h + 6) 500
1 1 dt
Ô⇒ ( − ) dh = −
h h+6 500
t
Ô⇒ ln h − ln(h + 6) = − +C
500
h
Ô⇒ = Ae−t/500
h+6
Substituting h = 1 m when t = 0 hours in this last equation we find A = 1/7 . Hence
h = 6/(7et/500 − 1) for t ≥ 0 .

b. The volume of the water lost through the bottom of the pool is
t=∞ π 2 ∞ 6
Volume = ∫ a R2 h dh = 6 ∫ dt
t=0 500 0 7et/500 − 1
∞ e−t/500 dt/500 7 du
= 216π ∫ = 216π ∫6 u = 216π ln(7/6) m
3
0 7 − e−t/500
where we used the substitution u = 7 − e−t/500 , du = e−t/500 dt/500 .

65. Some students believe that Bilkent Math 101 exams get more difficult as time passes. This
is in fact true. The difficulty level H(t) of these exams satisfies the differential equation

dH
= Ha
dt
with the initial condition H(0) = 1, where t is time measured from Fall 1986 in academic years
and a is a constant whose value is a secret.
There is a quatrain in Nostradamus’s Les Propheties which can be interpreted to be about
Bilkent Math 101 exams.
a. According to one interpretation, the exams will become infinitely difficult in Fall 2021.
Accepting this interpretation, find the difficulty level of the exams in Fall 2016.
b. According to another interpretation, the exams will be twice as difficult in Fall 2021 as
they were in Fall 1986. Show that a must satisfy −8 < a < −7 if this is the case.

59
 Solution: a. We have H −a dH = dt. If a = 1, then integration gives ln ∣H∣ = t + C
for some constant C, and the initial condition H(0) = 1 implies that C = 0. Hence
∣H∣ = et , and by continuity we have H = et . On the other hand, if a =/ 1, then
by integration we obtain H −a+1 /(−a + 1) = t + C for some constant C. Using the
condition that H(0) = 1 we get C = 1/(−a + 1) and hence H = ((1 − a)t + 1)1/(1−a) .

We want lim− H = ∞. Among these solutions this is possible only if (1−a)⋅35+1 = 0

t→35
and 1/(1 − a) < 0. So we must have a = 36/35. Therefore H = (1 − t/35)−35 . In
particular, H(30) = 735 , and the exams of Fall 2016 are 735 times as difficult as
those of Fall 1986.

b. This time we want H(35) = 2. Since e35 > e > 2, this is not possible if a = 1.
Therefore H = ((1−a)t+1)1/(1−a) for some constant a =/ 1 and ((1−a)⋅35+1)1/(1−a) = 2.
In other words, 36 − 35a = 21−a and a =/ 1.

Consider the function g(a) = 36 − 35a − 21−a . We have f ′ (a) = −35 + ln 2 ⋅ 21−a , which
is 0 only when a = log2 (2(ln 2)/35). Therefore, by Rolle’s Theorem, f can have
at most one more zero beside a = 1. On the other hand, f (−8) = −196 < 0 and
f (−7) = 25 > 0, and f must have a zero between −8 and −7 by the Intermediate
Value Theorem. We conclude that H(35) = 2 is possible for exactly one value of a
and this value lies in the interval (−8, −7).

66. Suppose that a continuous function f satisfies the equation

x
f ′ (x) = f (x) ∫ f (t) dt (⋆)
0

for all x.
a. In this part assume that f (0) = −1. Show that f has a critical point at x = 0, and
determine whether it is a local maximum, a local minimum, or neither.
b. In this part assume that f ′′ (2) = 0. Express f (2) in terms of A = f ′ (2) only.
c. Show that there are constants a and b such that
f ′ (x)2 = af (x)3 + bf (x)2
for all x.
k
d. Find all possible values of the constant k if f (x) = .
(ex + e−x )2
0
Solution: a. Setting x = 0 gives f ′ (0) = f (0) ∫ f (t) dt = f (0) ⋅ 0 = 0. Hence f has
0
a critical point at x = 0. Differentiating both sides of the equation (⋆) with respect
to x and then using the Fundamental Theorem of Calculus Part 1 we obtain
x x FTC1 x
′′ ′ d ↓
f (x) = f (x) ∫ f (t) dt + f (x) ( ∫ f (t) dt) = f ′ (x) ∫ f (t) dt + f (x) ⋅ f (x)
0 dx 0 0

for all x. Substituting x = 0 in this gives:

0
f ′′ (0) = f ′ (0) ∫ f (t) dt + f (0)2 = 0 ⋅ 0 + (−1)2 = 1 > 0
0

60
 Hence f has a local minimum at x = 0.

b. Multiplying the equation we obtained in part (a) by f (x) and then using (⋆)
gives us
f (x)f ′′ (x) = f ′ (x)2 + f (x)3
for all x. Substituting x = 2 in this we get f (2)f ′′ (2) = f ′ (2)2 + f (2)3 , which gives
f (2) = −f ′ (2)2/3 = −A2/3 if f ′′ (2) = 0.

c. Integrating (⋆) with respect to x, we obtain

2
1 x
f (x) = ( ∫ f (t) dt) + f (0)
2 0

for all x. Multiplying this by 2f (x)2 and using (⋆) again gives

f ′ (x)2 = 2f (x)3 − 2f (0) f (x)2

for all x.

d. Let f (x) = k (ex + e−x )−2 . If k = 0, then f (x) = 0 and f ′ (x) = 0 for all x, and (⋆)
is satisfied for all x. Now assume that k =/ 0. Then:
f ′ (x) ex − e−x
f ′ (x) = −2k (ex + e−x )−3 (ex − e−x ) and = −2 x −x
f (x) e +e
Hence:
d f ′ (x) (ex + e−x )(ex + e−x ) − (ex − e−x )(ex − e−x )
( ) = −2 = −8(ex + e−x )−2
dx f (x) (ex + e−x )2

f satisfies (⋆) for all x if and only if

f ′ (x) x
= ∫ f (t) dt for all x
f (x) 0

if and only if

d f ′ (x) f ′ (0) 0
( ) = f (x) for all x and = ∫ f (t) dt
dx f (x) f (0) 0

if and only if
0
−8(ex + e−x )−2 = k (ex + e−x )−2 for all x and =0
k/4
if and only if
k = −8 and 0 = 0 .
Hence f (x) = k (ex + e−x )−2 satisfies (⋆) for all x exactly when k = −8 or k = 0.

Remark: Note that in part (c), it is not sufficient to just check that (f ′ (x)/f (x))′ = f (x) for
all x. All of the functions f (x) = −8(ex+c + e−x−c )−2 , where c is a constant, satisfy this condition,
but only the one with c = 0 satisfies the original equation.

61
Remark: The only functions satisfying (⋆) for all x are f (x) = −8c2 (ecx + e−cx )−2 where c is a
nonnegative constant. Besides these, f (x) = 2c2 sec2 cx are the only functions satisfying (⋆) for
all x in (−π/(2c), π/(2c)) where c is a positive constant.

67. Suppose that f is a function that has a continuous second derivative and that satisfies
f (0) = 4 , f (1) = 3 , f ′ (0) = 5 , f ′ (1) = 7, f ′′ (0) = 8 and f ′′ (1) = 11 . Show that:
1
′′
∫0 f (x)f (x) dx ≤ 1

Solution: We first do an integration by parts with u = f (x), dv = f ′′ (x) dx, hence

du = f ′ (x) dx, v = f ′ (x), to obtain:
1
1 1
′′ ′ ′
∫0 f (x)f (x) dx = [f (x)f (x)] − ∫0 (f (x)) dx
2

The first term on the right is equal to f (1)f ′ (1) − f (0)f ′ (0) = 3 ⋅ 7 − 4 ⋅ 5 = 1, and
the second term is nonnegative as (f ′ (x))2 ≥ 0. The result follows.

Remark: It can be shown that:

1
′′
∫0 f (x)f (x) dx < 0

68. Evaluate the following integrals.

a. ∫ sin3 x sin 2x dx
e ln x
b. ∫ √ dx
1 x

Solution: a. We use the identity sin 2x = 2 sin cos x to obtain

∫ sin x sin 2x dx = ∫ sin x 2 sin x cos x dx

3 3

= 2 ∫ sin4 x cos x dx

= 2 ∫ u4 du
2
= u5 + C
5
2
= sin5 x + C
5
after the substitution u = sin x, du = cos x dx.

62
 √ √
b. The substitution u = x, du = dx/(2 x) gives:
√
e ln x e
∫1 √ dx = 4 ∫ ln u du
x 1
√
e
= 4 [u ln u − u]
√ √ 1√
= 4 ( e ln e − e + 1)
√
=4−2 e

Remark: There are other ways of doing these. Here are some:

Solution: a. Use the double-angle formula sin2 x = (1 − cos 2x)/2 and then the
substitution u = 1 − cos 2x, du = 2 sin 2x dx :
3
1 − cos 2x
∫ sin x sin 2x dx = ∫ ( ) sin 2x dx
3
2
1
= √ ∫ u3/2 du
4 2
1
= √ u5/2 + C
10 2
1
= √ (1 − cos 2x)5/2 + C
10 2

Or you can run this by simply saying u = sin2 x, du = 2 sin x cos x dx = sin 2x dx and:

∫ sin x sin 2x dx = ∫ u du
3 3/2

2
= u5/2 + C
5
2
= sin5 x + C
5
Or use the identity sin3 x = (3 sin x − sin 3x)/4 and the trigonometric product to sum
formulas:
1
∫ sin x sin 2x dx = ∫ 4 (3 sin x − sin 3x) sin 2x dx
3

1
= ∫ (3 sin x sin 2x − sin 3x sin 2x) dx
4
1
= ∫ (3(cos x − cos 3x) − (cos x − cos 5x)) dx
8
1
= ∫ (cos 5x − 3 cos 3x + 2 cos x) dx
8
1 1 1
= sin 5x − sin 3x + sin x + C
40 8 4

63
Or, if you are willing to go complex, use the identity sin θ = (eiθ − e−iθ )/(2i) :
3
eix − e−ix e2ix − e−2ix
∫ sin x sin 2x dx = ∫ ( )
3
dx
2i 2i
1
= ∫ (e5ix + e−5ix − 3e3ix − 3e−3ix + 2eix + 2e−ix ) dx
16
1 e5ix − e−5ix e3ix − e−3ix eix − e−ix
= ( − +2 )+C
16 5i i i
1 1 1
= sin 5x − sin 3x + sin x + C
40 8 4

b. Do integration by parts:
e ln x e √
∫1 √ dx = ∫1 ln x d(2 x)
x
√ e e √ dx
= [2 x ln x]1 − ∫ 2 x
1 x
√ e dx
= 2 e − 2∫ √
1 x
√ √ e
= 2 e − 4 [ x]1
√ √
=2 e−4 e+4
√
=4−2 e

Or do the other integration by parts first,

e ln x e 1
∫1 √ dx = ∫1 √ d(x ln x − x)
x x
e
x ln x − x 1 e x ln x − x
=[ √ ] + ∫ dx
x 1
2 1 x3/2
1 e ln x √ e
= 1 + ∫ √ dx − [ x]1
2 1 x
√ 1 e ln x
= 2 − e + ∫ √ dx
2 1 x

and then from

1 e ln x √
∫ √ dx = 2 − e
2 1 x
solve for
e ln x √
∫1 √ dx = 4 − 2 e .
x

64
 Or first do the substitution x = eu , dx = eu du, and then do an integration by parts:
e ln x 1 u
∫1 √ dx = ∫0 √ u e du
u
x e
1
=∫ u eu/2 du
0
1
=∫ u d(2eu/2 )
0
1 1
= [2ueu/2 ]0 − 2 ∫ eu/2 du
0
u/2 1
= 2e − 4[e ]0
1/2

= 2e− 4e
1/2 1/2
+4
√
=4−2 e

69. Evaluate the following integrals:

√
x
a. ∫ e dx
√
b. ∫ 1 − x2 dx

dx
c. ∫
(x + 1)2
2

Solution: a. We first do a change of variable x = t2 , dx = 2tdt,

√
∫ e
x
dx = 2 ∫ tet dt

and then do an integration by parts, u = t, dv = et dt Ô⇒ du = dt, v = et :

= 2tet − 2 ∫ et dt
= 2tet − 2et + C
√ √ √
= 2 xe x − 2e x + C

π π
b. We use the trigonometric substitution x = sin θ, − ≤ θ ≤ . Then dx = cos θ dθ
2 2

65
 √ √ √ π π
and 1 − x2 = 1 − sin2 θ = cos2 θ = ∣ cos θ∣ = cos θ as cos θ ≥ 0 for − ≤ θ ≤ .
2 2
√
∫ 1 − x2 dx = ∫ cos θ ⋅ cos θ dθ

= ∫ cos2 θ dθ
1
= (1 + cos 2θ) dθ
2∫
1 1
= θ + sin 2θ + C
2 4
1 1
= θ + sin θ cos θ + C
2 2
1 1 √
= arcsin x + x 1 − x2 + C
2 2

1
x

√
1 − x2
π π
c. We use the trigonometric substitution x = tan θ, − < θ < . Then dx = sec2 θ dθ
2 2
and x2 + 1 = tan2 θ + 1 = sec2 θ .
dx sec2 θ
=
∫ (x2 + 1)2 ∫ (sec2 θ)2 dθ

= ∫ cos2 θ dθ
1
= (1 + cos 2θ) dθ
2∫
1 1
= θ + sin 2θ + C
2 4
1 1
= θ + sin θ cos θ + C
2 2
1 1 x 1
= arctan x + ⋅ √ ⋅√ +C
2 2 x2 + 1 x2 + 1
1 1 x
= arctan x + ⋅ 2 +C
2 2 x +1

√
x2 + 1
x

66
Remark: Other methods can also be used. For instance, the integral in part (b) can be done
using integration by parts.

70. Evaluate the following integrals:

√
1/2 1−x
a. ∫ arcsin x dx
−1/2 1+x
1/2 dx
b. ∫ √
−1/2 x + 1 − x2
∞ dx
c. ∫
0 1 + ex

Solution: a. We first re-write the integral as follows:

√
1/2 1−x 1/2 1 1/2 −x
∫−1/2 arcsin x dx = ∫ √ arcsin x dx + ∫ √ arcsin x dx
1+x −1/2 1 − x2 −1/2 1 − x2
The first integral on the right vanishes as the integrand is odd and the integration
interval is symmetric about the origin. We do integration by parts for the second
−x 1
integral with u = arcsin x and dv = √ dx, and hence du = √ and v =
√ 1 − x2 1 − x2
1 − x2 :
1/2 −x √ 1/2 1/2 π
∫−1/2 arcsin x √ dx = [arcsin x 1 − x 2] − ∫ dx = √ − 1
1 − x2 −1/2 −1/2 2 3
Therefore: √
1/2 1−x π
∫−1/2 arcsin x dx = √ − 1
1+x 2 3

b. We start by changing variables t = sin θ, dt = cos θ dθ.

1/2 dx π/6 cos θ dθ
∫−1/2 √ =∫
x+ 1−x 2 −π/6 sin θ + cos θ
π/6 cos θ(cos θ − sin θ)
=∫ dθ
−π/6 cos2 θ − sin2 θ
1 π/6 1 + cos 2θ − sin 2θ
= ∫ dθ
2 −π/6 cos 2θ
1 π/6
= ∫ (sec 2θ + 1 − tan 2θ) dθ
2 −π/6
π/6
1 1 1
= [ ln ∣ tan 2θ + sec 2θ∣ + θ − ln ∣ sec 2θ∣]
2 2 2 −π/6
1 √ π
= ln( 3 + 2) +
2 6

c. Let u = e−x + 1, du = −e−x dx. Then

dx e−x dx du −x
∫ 1 + ex ∫ e−x + 1 = − ∫ u = − ln ∣u∣ + C = − ln(1 + e ) + C
=

67
 and from this
∞ dx c dx
∫0 = lim ∫ = − lim [ln(1 + e−x )]c0 = − lim (ln(1 + e−c ) − ln 2) = ln 2
1 + ex c→∞ 0 1 + ex c→∞ c→∞

follows.

Remark: Here is another way of doing part (c). Let u = ex , du = ex dx. Then

dx ex dx
=
∫ 1 + ex ∫ ex + (ex )2
du
=∫
u + u2
1 1
=∫ ( − ) du
u 1+u
= ln ∣u∣ − ln ∣1 + u∣ + C
= x − ln(1 + ex ) + C

and from this

∞ dx c dx
∫0 =
1 + ex c→∞ ∫0 1 + ex
lim
= lim [x − ln(1 + ex )]0c
c→∞
= lim (c − ln(1 + ec ) + ln 2)
c→∞
= lim (− ln(e−c ) − ln(1 + ec ) + ln 2)
c→∞
= lim (− ln(e−c + 1) + ln 2)
c→∞
= ln 2

follows.

71. Evaluate the improper integral

∞ dx
∫0 (ax + 1)(x2 + 1)

where a is a positive constant.

Solution: We have the partial fraction decomposition

1 1 a2 ax − 1
= ( − 2 ).
(ax + 1)(x + 1) a + 1 ax + 1 x + 1
2 2

Hence:
dx 1 a2 ax − 1
∫ (ax + 1)(x2 + 1) a2 + 1 ∫ ax + 1 dx − ∫ x2 + 1 dx)
= (

1 a
= (a ln ∣ax + 1∣ − ln(x2 + 1) + arctan(x)) + C .
a2 +1 2

68
 Therefore:
∞ dx c dx
∫0 = lim ∫
(ax + 1)(x + 1) c→∞ 0 (ax + 1)(x2 + 1)
2
c
1 a
= 2 lim [a ln ∣ax + 1∣ − ln(x2 + 1) + arctan(x)]
a + 1 c→∞ 2 0
1 a
= lim (a ln(ac + 1) − ln(c2 + 1) + arctan c
a2 + 1 c→∞ 2
a
− a ln 1 + ln 1 − arctan 0)
2
a ac + 1 1
= lim ln ( √ ) + lim arctan c
a2 + 1 c→∞ c2 + 1 a2 + 1 c→∞
1 π
= (a ln a + )
a2 +1 2
ac + 1 π
as lim √ = a and lim arctan c = .
c→∞ c2 + 1 c→∞ 2

72. The curve y = 1/x, x ≥ 1, is revolved about the x-axis to generate a surface S and the
region between the curve and the x-axis for x ≥ 1 is revolved about the x-axis to generate a
solid D.
a. Show that D has finite volume.
b. Show that S has infinite area.
Solution: a. Using the disk method we obtain
∞ ∞ dx
Volume = π ∫ R(x)2 dx = π ∫ .
1 1 x2
Since p = 2 > 1, this integral converges.

b. The surface area formula gives

√
∞ √ ∞ 1 1
Surface Area = 2π ∫ y 1 + (y ′ )2 dx = 2π ∫ 1+ dx .
1 1 x x4
√
1 1 1 ∞ dx
We have 1 + 4 ≥ ≥ 0 for x ≥ 1, and ∫ = ∞. The surface area is infinite
x x x 1 x
by the Direct Comparison Test.

Remark: We want to get the surface S painted for a reasonable, finite price. We offer the job
to Painter1 and Painter2.
Painter1 says: “It cannot be done. S has infinite area, it cannot be painted with
finite amount of paint.”

Painter2 says: “It can be done. D has finite volume. We can fill the inside of S
with volume(D) cubic units of paint and let the excess paint run out.”

69
Whom should we believe?

∞
73. Let n be a nonnegative integer. Show that ∫ tn e−t dt = n! .
0

Solution: We use induction on n.

∞
• Let n = 0. Then ∫ e−t dt = 1 = 0! .
0
∞
• Let n > 0 and assume that ∫ tn−1 e−t dt = (n − 1)! . Integration by parts gives
0

∞ c
∫0 tn e−t dt = lim ∫ tn e−t dt
c→∞ 0
c
= lim ([−tn e−t ]0 + n ∫ tn−1 e−t dt)
c
c→∞ 0
∞
lim [−tn e−t ]0 tn−1 e−t dt
c
= + n∫
c→∞ 0
∞
= n∫ tn−1 e−t dt
0
= n ⋅ (n − 1)!
= n!

where lim cn e−c = 0 can be seen after n applications of L’Hôpital’s Rule.

c→∞

Remark: The Gamma function is defined by

∞
Γ(x) = ∫ tx−1 e−t dt
0

for x > 0. It can be shown that the improper integral on the right converges if and only if x > 0.
Note that 0 is also a “bad point” besides ∞ for 0 < x < 1.
A calculation similar to the one in the solution above shows that Γ(x + 1) = xΓ(x) for x > 0.
This relation Γ(x) = Γ(x + 1)/x can be used repeatedly to define the Gamma function for all
real numbers which are not nonpositive integers.
Since Γ(n + 1) = n! for all nonnegative integers n, we can use the Gamma function to define the
factorials of all real numbers which are not negative integers by x! = Γ(x + 1). In particular,
√
1 ∞
−1/2 −t
∞
−u2 π √
(− )! = Γ(1/2) = ∫ t e dt = 2 ∫ e du = 2 ⋅ = π.
2 0 0 2

The volume of an n-dimensional ball with radius r is π n/2 rn /(n/2)! . Check this formula for
n = 1 (the interval [−r, r]), n = 2 (the disk with radius r), and n = 3 (the sphere with radius r).
The case n = 4 will be seen in Calculus II.

∞ ln x
74. Show that ∫ dx = 0 .
0 x2 + 1

70
 Solution: First of all, by definition,
∞ ln x 1 ln x ∞ ln x
∫0 dx = ∫ dx + ∫ dx
x2 + 1 0 x2 + 1 1 x2 + 1
and the improper integral on the left converges if and only if both basic improper
integrals on the right converge.
∞ ln x ln x ln x
Consider ∫ dx . Note that 0 ≤ 2 ≤ 2 for x ≥ 1. On the other hand,
1 x +1
2 x +1 x
∞ ln x ∞
−t
∫1 dx = ∫0 te dt
x2
c
= lim ∫ te−t dt
c→∞ 0
c
= lim ∫ t d(−e−t )
c→∞ 0
c c
= lim ([ − te−t ]0 + ∫ e−t dt)
c→∞ 0
−t c
−c
= lim (−ce − [e ]0 )
c→∞
= lim (−ce−c − e−c + 1)
c→∞
=1

where we used the substitution x = et , dx = et dt, followed by the definition of the

improper integral, then an integration by parts and finally the limit

−c c L’H
↓ 1
lim ce = lim c = lim c = 0 .
c→∞ c→∞ e c→∞ e

∞ ln x
Therefore by the Direct Comparison Test the improper integral ∫ dx
1 x2 + 1
converges.
1 ln x
Now the convergence of ∫ dx follows as
0 x2 + 1

1 ln x 1 ln(1/u) −du ∞ ln u
∫0 x2 + 1 dx = ∫∞ (1/u)2 + 1 ⋅ u2 = − ∫1 1 + u2 du

where we used the change of variable x = 1/u, dx = −du/u2 , and this also gives:
∞ ln x 1 ln x ∞ ln x
∫0 dx = ∫ dx + ∫ dx
x2 + 1 0 x2 + 1 1 x2 + 1
∞ ln x ∞ ln x
= −∫ dx + ∫1 x2 + 1 dx = 0
1 x2 + 1

∞ ln x
Remark: Instead of computing ∫ dx, one can show its convergence using the
1 x2
ln x 1 ∞ dx
comparison 0 ≤ 2 ≤ 3/2 for x ≥ 1 and the fact that ∫ is convergent as p = 3/2 > 1.
x x 1 x3/2

71
 ln x 1 √
Here the fact that ≤ for x ≥ 1 can be seen as follows: Consider f (x) = x − ln x on
x2 √
x3/2 √
[1, ∞). Then f (x) = 1/(2 x) − 1/x = ( x − 2)/(2x)
′
√ and x = 4 is the only critical point. Since
f (x) < 0 for x < 4 and f (x) > 0 for x > 4, f (4) = 4 − ln 4 = 2 − 2 ln 2 = 2(1 − ln 2) > 0 must be
′ ′

the absolute minimum value of f on [1, ∞), and we are done.

75. For a positive continuous function f on (−∞, ∞), let

● R(a) be the region between the graph of y = f (x) and the x-axis for x ≤ a,
● A(a) be the area of R(a), and
● V (a) be the volume of the solid generated by revolving R(a) about the x-axis.
a. Show that if f (x) = e2x/π , then V (a) = (A(a))2 for all a.
b. Show that f (x) = e2x/π is the only positive continuous function on (−∞, ∞) for which
A(a) and V (a) are finite and satisfy V (a) = (A(a))2 for all a, and f (0) = 1.
Solution: a. We observe that
a
a a ekx eka − ekc eka − 0 eka
∫−∞ e dx = c→−∞ e dx = lim [ ] = lim = =
kx kx
lim ∫
c c→−∞ k c c→−∞ k k k
for any positive constant k. Hence, for f (x) = e2x/π , from
a a e2a/π π 2a/π
R(a) = ∫ f (x) dx = ∫ e2x/π dx = = e
−∞ −∞ 2/π 2
and
a a a e4a/π π 2 4a/π
V (a) = π ∫ (f (x))2 dx = π ∫ (e2x/π )2 dx = π ∫ e4x/π dx = π = e
−∞ −∞ −∞ 4/π 4
we obtain
2
π 2 π2
(R(a)) = ( e2a/π ) = e4a/π = V (a)
2 4
for all a.

b. Now assume that V (a) = (A(a))2 for all a for some positive continuous function
on (−∞, ∞) for which A(a) and V (a) are finite and f (0) = 1. Then
2
a a
π∫ (f (x))2 dx = ( ∫ f (x) dx)
−∞ −∞

for all a. Differentiating both sides with respect to a and using the Fundamental
Theorem of Calculus, Part 1, we obtain
a
π(f (a))2 = 2 ( ∫ f (x) dx) f (a)
−∞

for all a. Cancelling a factor of f (a), differentiating with respect to a and using the
Fundamental Theorem of Calculus, Part 1 again, we obtain
πf ′ (a) = 2f (a)
for all a. Hence, f ′ (a)/f (a) = 2/π for all a. Integrating this with respect to a we
get ln ∣f (a)∣ = 2a/π + C for all a, for some constant C. Since f (0) = 1, C must be 0.
Therefore, f (a) = e2a/π for all a.

72
 ∞ dx
76. Determine whether the improper integral ∫ converges or diverges.
0 ex − e−x
Solution: By definition,
∞ dx 1 1 ∞ dx
∫0 = ∫ + ∫
e −e
x −x 0 e −e
x −x 1 e − e−x
x

and the given integral converges if and only if both of the integrals on the right
hand side converge.
1 dx
Let us consider ∫ first. Since we have the linearization
0 e − e−x
x

ex − e−x ≈ (1 + x) − (1 − x) = 2x ,

centered at x = 0, we expect 1/(ex − e−x ) to behave like 1/(2x) near the “bad point”
0, and therefore this integral to diverge.

In fact, we have

1/(ex − e−x ) x L’H

↓ 1 1
L = lim+ = lim+ x −x = lim+ x −x = .
x→0 1/x x→0 e − e x→0 e + e 2
1 dx
Since 0 < L < ∞ and ∫ is divergent (because p = 1 ≥ 1), we conclude
0 x
1 dx ∞ dx
that ∫ diverges by the Limit Comparison Test. Therefore ∫
0 e −e
x −x 0 e − e−x
x
diverges too.

Remark: The other improper integral on the right hand side converges. We have

1/(ex − e−x ) 1
L = lim = lim = 1.
x→∞ e−x x→∞ 1 − e−2x

Since 0 < L < ∞ and

∞ c
∫1 e−x dx = lim ∫ e−x dx = lim [−e−x ]c1 = lim (−e−c + e−1 ) = e−1 < ∞ ,
c→∞ 1 c→∞ c→∞

∞ dx
we conclude that ∫ converges by the Limit Comparison Test.
1 ex − e−x

∞ 1 − e−1/x
77. Determine whether the improper integral ∫ √ dx converges or diverges.
0 x

Solution: By definition,
∞ 1 − e−1/x 1 1 − e−1/x ∞ 1 − e−1/x
∫0 √ dx = ∫ √ dx + ∫ √ dx
x 0 x 1 x
and the given integral converges if and only if both integrals on the right hand side
converge.

73
On one hand, since

1 − e−1/x
√
x
L = lim+ = lim+ (1 − e−1/x ) = 1
x→0 1 x→0
√
x
1 dx
is a positive real number, and ∫ √ converges as p = 1/2 < 1; we conclude that
0 x
1 1 − e−1/x
∫0 √ dx converges by the Limit Comparison Test.
x

On the other hand, since

1 − e−1/x
√
x 1 − e−1/x 1 − e−t L’H
↓ e−t
L = lim = lim = lim+ = lim+ =1
x→∞ 1 x→∞ 1/x t→0 t t→0 1
x3/2
∞ dx
is a positive real number, and ∫ converges as p = 3/2 > 1; we conclude that
1 x3/2
∞ 1 − e−1/x
∫1 √ dx also converges by the Limit Comparison Test.
x
∞ 1 − e−1/x
Hence ∫ √ dx converges.
0 x

74
 Part 2: Multi-variable Functions

1. Consider the point P (3, −5, 1) and the line L ∶ x = 2t − 1 , y = −t + 2 , z = −2t ; −∞ < t < ∞ .
a. Find the equation of the plane passing through P perpendicular to L.
b. Find the equation of the plane passing through P and containing L.

Solution: a. The line L is parallel to the vector v = 2i − j − 2k . Therefore we can

take the normal vector of the plane to be n = 2i − j − 2k . Then the equation of the
plane is 2 ⋅ (x − 3) + (−1) ⋅ (y − (−5)) + (−2) ⋅ (z − 1) = 0 , or 2x − y − 2z = 9 .

b. A normal n to the plane containing the line L and the point P will be
# »
perpendicular to v, and it will also be perpendicular to P Q where Q is any point
on the line. We can take Q(−1, 2, 0), the point corresponding to t = 0 , and then
# »
P Q = −4i + 7j − k . Now we can take the normal vector v to be
RR i j kRRRR
# » RRRR R
n = v × P Q = RRR 2 −1 −2 RRRR = 15i + 10j + 10k ,
RRR R
RR−4 7 −1 RRRR
or in fact n = 3i + 2j + 2k . Then the equation of the plane is

3 ⋅ (x − 3) + 2 ⋅ (y − (−5)) + 2 ⋅ (z − 1) = 0 ,

or 3x + 2y + 2z = 1 .

2. Consider the plane P ∶ 3x − 4y + z = 10 , and the points P (2, 3, −1) and Q(1, 2, 2) .
a. Find the equation of the line passing through P perpendicular to P.
b. Find the equation of the plane passing through P and Q perpendicular to P.

Solution: a. Since n = 3i − 4j + k is normal to the plane, it will be parallel to any

line perpendicular to the plane. Hence we can take v = 3i − 4j + k , and the equation
of the line is x = 3t + 2, y = −4t + 3, z = t − 1; −∞ < t < ∞ .

b. A plane perpendicular to P will have a normal n′ perpendicular to the normal

n = 3i − 4j + k of P. Also a plane containing the points P and Q will have a normal
# »
n′ perpendicular to P Q = −i − j + 3k . Therefore we can take n′ to be
RR i kRRRR
# » RRRR
j
R
′ R
n = n × P Q = RR 3 −4 1 RRRR = −11i − 10j − 7k ,
RRR R
RR−1 −1 3 RRRR

75
 or rather n′ = 11i + 10j + 7k . This gives the equation of the plane as
11 ⋅ (x − 2) + 10 ⋅ (y − 3) + 7 ⋅ (z − (−1)) = 0 ,
or 11x + 10y + 7z = 45 .

3. Find a parametric equation of the line L that intersects both of the lines
L1 ∶ x = 2t − 1, y = −t + 2, z = 3t + 1
and
L2 ∶ x = s + 5, y = 2s + 3, z = −s
perpendicularly.
Solution: v1 = 2i − j + 3k and v2 = i + 2j − k are the velocity vectors of the lines L1
and L2 , respectively. Hence,
RRR i j kRRRR
RRR R
v = RRR2 −1 3 RRRR = −5i + 5j + 5k
RRR R
RR1 2 −1 RRRR
is a velocity vector for L. So we may take v = i − j − k.

Then
RRR i j kRRRR
RRR R
n = v×v1 = RRR1 −1 −1 RRRR = −4i − 5j + k
RRR RR
RR2 −1 3 RRR
is normal to the plane P containing the lines L and L1 . As P1 (−1, 2, 1) is in P,
an equation of P is
−4 ⋅ (x − (−1)) + (−5) ⋅ (y − 2) + 1 ⋅ (z − 1) = 0 ,
or 4x + 5y − z = 5. At the point P0 of intersection of P and L2 , s satisfies:
4 ⋅ (s + 5) + 5 ⋅ (2s + 3) − (−s) = 5
Hence s = −2. Substituting this back in the equations of L2 gives P0 (3, −1, 2).
Therefore an equation of L is:
x = t + 3, y = −t − 1 , z = −t + 2 ; (−∞ < t < ∞) .

76
4. Let c be a constant. Show the angle between the position and the velocity vectors along the
curve r = ect cos t i + ect sin t j , −∞ < t < ∞ , is constant.

Solution: We have

r = ect cos t i + ect sin t j

and

dr
v= = (c ect cos t − ect sin t)i + (c ect sin t + ect cos t)j .
dt

Then

∣r∣ = ((ect cos t)2 + (ect sin t)2 )1/2 = ect ,

√
∣v∣ = ((c ect cos t − ect sin t)2 + (c ect sin t + ect cos t)2 )1/2 = c2 + 1 ect ,
r ⋅ v = ect cos t ⋅ (c ect cos t − ect sin t) + ect sin t ⋅ (c ect sin t + ect cos t) = c e2t .

Therefore, if θ is the angle between r and v, we have

r⋅v c e2t c
cos θ = = √ =√ ,
∣r∣ ∣v∣ e ⋅ c + 1 e
ct 2 ct c +1
2

and we conclude that θ is constant.

77
5. In the xyz-space where a flyscreen lies along the plane with the equation

2x + y − 2z = 1 ,

the trajectory of a bee as a function of time t is given by

r = t i + t2 j + t3 k

for −∞ < t < ∞.

a. Find all times t when the bee is flying parallel to the screen.
b. Find all times t when the bee is flying perpendicular to the screen.
c. There are holes in the screen through which the bee passes. Find the coordinates of all
of these holes.

Solution: a. The normal vector of the plane along which the screen lies is n =
2i + j − 2k, and the velocity vector of the bee is v = i + 2tj + 3t2 k. The bee is flying
parallel to the screen whenever these two vectors are perpendicular to each other;
in other words,√whenever n ⋅ v = 0. As n ⋅ v = 2 + 2t − 6t2 , we want 3t2 − t − 1 = 0.
Hence t = (1 ± 13)/6 are the times when the bee is flying parallel to the screen.

b. The bee is flying perpendicular to the screen whenever n and v are parallel
to each other; in other words, whenever 1/2 = 2t/1 = 3t2 /(−2). This implies t2 =
−1/3 which is not possible. Therefore there is no moment when the bee is flying
perpendicular to the screen.

c. Substituting the coordinates x = t, y = t2 , z = t3 of the position of the bee

into the equation 2x + y − 2z = 1 of the plane we obtain 2t3 − t2 − 2t + 1 = 0. As
2t3 − t2 − 2t + 1 = (2t − 1)(t − 1)(t + 1), the fly is in the plane of the screen when
t = 1/2, t = 1, t = −1. These times correspond to the points (x, y, z) = (1/2, 1/4, 1/8),
(1, 1, 1) and (−1, 1, −1).

78
 xy 2
6. a. Show that lim =0.
(x,y)→(0,0) x6 + y 2

xy
b. Show that lim does not exist.
(x,y)→(0,0) x + y 2
6

Solution: a. We have 0 ≤ y 2 ≤ x6 + y 2 for all (x, y). Hence

xy 2 y2
0≤∣ ∣ = ∣x∣ ⋅ ≤ ∣x∣ ⋅ 1 = ∣x∣
x6 + y 2 x6 + y 2

for all (x, y) =/ (0, 0) . Since lim ∣x∣ = 0 , the Sandwich Theorem gives
(x,y)→(0,0)

xy 2
lim =0.
(x,y)→(0,0) x6 + y 2

b. The limit along the x-axis is

xy x⋅0
lim = lim 6 = lim 0 = 0 ,
(x,y)→(0,0) x6 +y 2 x→0 x + 02 x→0
along the x-axis

whereas the limit along the y = x line is

xy x⋅x 1
lim = lim 6 = lim 4 =1.
(x,y)→(0,0) x6+y 2 x→0 x + x 2 x→0 x + 1
along the line y = x

xy
Since these two limits are different, the two-variable limit lim does not
+ y2
(x,y)→(0,0) x6
exist by the Two-Path Test.

Remark: Sertöz Theorem tells exactly when such limits exist:

Let a and b be nonnegative real numbers, and c and d be positive real numbers.
a b ∣x∣a ∣y∣b
• If + > 1 , then lim =0.
c d (x,y)→(0,0) ∣x∣c + ∣y∣d

a b ∣x∣a ∣y∣b
• If + ≤ 1 , then lim does not exist.
c d (x,y)→(0,0) ∣x∣c + ∣y∣d

Let a and b be nonnegative integers, and c and d be positive even integers.

a b xa y b
• If + > 1 , then lim =0.
c d (x,y)→(0,0) xc + y d

a b xa y b
• If + ≤ 1 , then lim does not exist.
c d (x,y)→(0,0) xc + y d

79
A proof of the first version can be found in http://www.fen.bilkent.edu.tr/˜otekman/calc2
/sertoztheorem.pdf. A proof of the n-variable case of the second version can be found in
http://sertoz.bilkent.edu.tr/depo/limit.pdf.

7. Determine whether each of the following limits exists.

x sin y − y sin x
a. lim
(x,y)→(0,0) x2 + y 2
x sin y − y sin x
b. lim
(x,y)→(0,0) x4 + y 4
Solution: a. We have
x sin y − y sin x x(sin y − y) − y(sin x − x)
lim = lim
(x,y)→(0,0) x +y
2 2 (x,y)→(0,0) x2 + y 2
xy 3 sin y − y x3 y sin x − x
= lim ( 2 − )
(x,y)→(0,0) x + y 2 y3 x2 + y 2 x3
1 1
=0⋅(− )−0⋅(− )=0
6 6
where we used the facts that
sin t − t L’H
↓ cos t − 1 L’H
↓ − sin t 1
lim 3
= lim 2
= lim =−
t→0 t t→0 3t t→0 6t 6
and
xy 3 x3 y
lim = 0 = lim .
(x,y)→(0,0) x2 + y 2 (x,y)→(0,0) x2 + y 2
The first of these last two limits can be obtained by observing that
y2 xy 3
0 ≤ y 2 ≤ x2 + y 2 Ô⇒ 0 ≤ 2 ≤ 1 Ô⇒ 0 ≤ ∣ ∣ ≤ ∣xy∣ for (x, y) =/ (0, 0)
x + y2 x2 + y 2
and using the Squeeze Theorem. The other one follows by symmetry.

b. We observe that
x sin y − y sin x x sin 0 − 0 sin x
lim = lim = lim 0 = 0
(x,y)→(0,0) x +y
4 4 x→0 x4 + 04 x→0
along the x-axis

and
x sin y − y sin x x sin 2x − 2x sin x
lim = lim
(x,y)→(0,0) x +y
4 4 x→0 x4 + (2x)4
along the line y = 2x
1 sin 2x − 2 sin x
= lim
17 x→0 x3
L’H
↓ 1 2 cos 2x − 2 cos x
= lim
17 x→0 3x2
L’H
↓ 2 −2 sin 2x + sin x
= lim
3 ⋅ 17 x→0 2x
2 sin 2x 1 sin x
= lim ( − 2 + )
3 ⋅ 17 x→0 2x 2 x
2 1 1
= ( − 2 + ⋅ 1) = − .
3 ⋅ 17 2 17

80
 x sin y − y sin x
Since 0 =/ −1/17, the limit lim does not exist by the 2-Path
(x,y)→(0,0) x4 + y 4
Test.

8. Determine all values of the constant α > 0 for which the limit
x2 y 3
lim
(x,y)→(0,0) ∣x∣3 + ∣y∣α

exists.

Solution: We observe that

x2 y 3 (y 3 )2 y 3 y 1
lim = lim = lim ⋅ lim .
(x,y)→(0,0) ∣x∣ + ∣y∣
3 α y→0 ∣y ∣ + ∣y∣
3 3 α y→0 ∣y∣ y→0 1 + ∣y∣α−9
along the curve x = y 3

1 y
lim is 1 if α > 9 and 1/2 if α = 9. On the other hand lim does not
y→0 1 + ∣y∣α−9 y→0 ∣y∣
x2 y 3
exist. Therefore the limit of 3 along the curve x = y 3 does not exist, and
∣x∣ + ∣y∣α
x2 y 3
consequently the two-variable limit lim does not exist either for α ≥ 9.
(x,y)→(0,0) ∣x∣3 + ∣y∣α

Now we will show that the limit is 0 if α < 9. We have

x2 y 3 (∣x∣/∣y∣α/3 )2
0≤∣ ∣ = ⋅ ∣y∣3−α/3 ≤ ∣y∣3−α/3
∣x∣3 + ∣y∣α (∣x∣/∣y∣α/3 )3 + 1

for all (x, y) =/ (0, 0). Here we used the fact that if t ≥ 1, then t2 ≤ t3 and hence
t2 t2
≤ 1, and if 0 < t < 1, then t 2 < 1 gives ≤ 1. Since lim ∣y∣3−α/3 = 0 for
t3 + 1 t3 + 1 (x,y)→(0,0)
x2 y 3
α < 9, lim = 0 follows by the Sandwich Theorem.
(x,y)→(0,0) ∣x∣3 + ∣y∣α

9. Determine all values of the positive constant k for which the limit

∣x∣
lim
(x,y)→(0,0) (x2 + y 2 )k
exists.
1 1
Solution: The limit exists for k < , and does not exist for k ≥ .
2 2
1
Suppose k < . We have 0 ≤ x2 ≤ x2 + y 2 for all (x, y), and hence
2
k
∣x∣ x2
0≤ 2 ≤ ∣x∣1−2k
( ) ≤ ∣x∣1−2k
(x + y 2 )k x2 + y 2

81
 for all (x, y) =/ (0, 0) . Since 1 − 2k > 0, we have ∣x∣1−2k → 0 as (x, y) → (0, 0).
Therefore
∣x∣
lim =0
(x,y)→(0,0) (x + y 2 )k
2

by the Sandwich Theorem.

1
Suppose k = . Then
2
∣x∣ ∣x∣
lim = lim 2 = lim 1 = 1 ,
(x,y)→(0,0) (x2 +y )2 1/2 x→0 (x + 02 )1/2 x→0
along the x-axis

and
∣x∣ ∣0∣
lim = lim 2 = lim 0 = 0 .
(x,y)→(0,0) (x2 +y )2 1/2 y→0 (0 + y 2 )1/2 x→0
along the y-axis

∣x∣
Since these limits are different, the two-variable limit lim does not
(x,y)→(0,0) (x2 + y 2 )1/2
exist by the Two-Path Test.

Suppose k > 1/2 . Then

∣x∣ ∣x∣
lim = lim 2 = lim ∣x∣1−2k = ∞
(x,y)→(0,0) (x2 +y )2 k x→0 (x + 02 )k x→0
along the x-axis

∣x∣
does not exist. Therefore the two-variable limit lim does not exist
(x,y)→(0,0) (x2 + y 2 )k
either.

10. Let
⎧
⎪ xa y b
⎪
⎪
⎪ if (x, y) =/ (0, 0)
⎪
⎪ x4 + y 6
f (x, y) = ⎨
⎪
⎪
⎪
⎪
⎪
⎪
⎩0 if (x, y) = (0, 0)
where a and b are nonnegative integers. In each of (a-e), determine whether there exist values
of a and b for which f satisfies the given condition.
a. f (x, y) is continuous at (0, 0).
b. f (x, y) goes to 1 as (x, y) approaches (0, 0) along the line y = x, and f (x, y) goes to −1
as (x, y) approaches (0, 0) along the line y = −x.
c. f (x, y) goes to 0 as (x, y) approaches (0, 0) along any line through the origin, and the
limit lim f (x, y) does not exist.
(x,y)→(0,0)

d. f (x, y) goes to 0 as (x, y) approaches (0, 0) along any line through the origin except the
y-axis, and f (x, y) goes to 1 as (x, y) approaches (0, 0) along the y-axis.
e. fx (0, 0) and fy (0, 0) exist, and f (x, y) is not differentiable at (0, 0).

82
 Solution: a. If a = 4 and b = 1, then f is continuous at (0, 0). This follows from
the Sandwich Theorem as
x4 y x4
0 ≤ ∣f (x, y)∣ = ∣ ∣ ≤ ⋅ ∣y∣ ≤ 1 ⋅ ∣y∣ ≤ ∣y∣
x4 + y 6 x4 + y 6
for (x, y) =/ 0 implies that the limit of f (x, y) at (0, 0) is 0 = f (0, 0).

b. Let a = 3 and b = 1. Then

x4 1
lim f (x, y) = lim f (x, x) = lim = lim =1
(x,y)→(0,0) x→0 x→0 x + x
4 6 x→0 1 + x2
along the line y=x

and
−x4 −1
lim f (x, y) = lim f (x, −x) = lim = lim = −1 .
(x,y)→(0,0) x→0 x→0 x4 + x6 x→0 1 + x2
along the line y=−x

c. Let a = 2 and b = 3. Then we have

m3 x5 m3 x
lim f (x, y) = lim f (x, mx) = lim = lim =0
(x,y)→(0,0) x→0 x→0 x4 + m6 x6 x→0 1 + m6 x2
along the line y=mx

as well as
lim f (x, y) = lim f (0, y) = lim 0 = 0 .
(x,y)→(0,0) y→0 y→0
along the y-axis

However,
x4 1 1
lim f (x, y) = lim f (x, x2/3 ) = lim = lim = =/ 0
(x,y)→(0,0) x→0 x→0 x4 + x4 x→0 2 2
along the curve y=x2/3

and hence the limit of f (x, y) at (0, 0) does not exist by the 2-Path Test.

d. Let a = 0 and b = 6. Then we have

m6 x6 m6 x2
lim f (x, y) = lim f (x, mx) = lim = lim =0
(x,y)→(0,0) x→0 x→0 x4 + m6 x6 x→0 1 + m6 x2
along the line y=mx

and
y6
lim f (x, y) = lim f (0, y) = lim = lim 1 = 1 .
(x,y)→(0,0) y→0 y→0 y 6 y→0
along the y-axis

e. If a = 1 and b = 1, then f (x, y) is not differentiable at (0, 0) as it is not even

continuous there. This can be seen by considering its limit along the line y = x
which does not exist. On the other hand, fx (0, 0) and fy (0, 0) are both 0 as f is
identically zero on both axes.

Remark: The complete lists of ordered pairs (a, b) of nonnegative integers that satisfy the
given conditions are as follows:

83
 a. all (a, b) with 3a + 2b > 12

b. (3, 1), (1, 3)

c. (1, 4), (2, 3)

d. (0, 6)

e. (0, 7), (1, 1), (1, 2), (1, 3), (1, 4), (1, 5), (2, 1), (2, 2), (2, 3), (2, 4), (3, 1), (3, 2), (4, 1)

11. In Genetics, Fisher’s Equation,

∂p ∂ 2p
= p (1 − p) + 2
∂t ∂x
describes the spread of an advantageous allele in a population with uniform density along a
1-dimensional habitat, like a shoreline, as a result of both reproduction and dispersion of the
offspring. Here p(x, t) is the frequency of the allele as a function of the position x and the time
t.
Find all possible values of the pair of constants (a, b) for which the function
1
p(x, t) =
(1 + eax+bt )2

satisfies the Fisher’s Equation.

Solution: We have

pt = −2(1 + eax+bt )−3 ⋅ eax+bt ⋅ b

px = −2(1 + eax+bt )−3 ⋅ eax+bt ⋅ a
pxx = 6(1 + eax+bt )−4 ⋅ (eax+bt ⋅ a)2 − 2(1 + eax+bt )−3 ⋅ eax+bt ⋅ a2
p(1 − p) = (1 + eax+bt )−4 ⋅ (2eax+bt + (eax+bt )2 )

and substitution these in the Fisher’s Equation gives

−2b(1 + eax+bt ) = 2 + eax+bt + 6a2 eax+bt − 2a2 (1 + eax+bt )

or:
2a2 − 2b − 2 = (4a2 + 2b + 1)eax+bt

As the left hand side if this equality is constant, so must be the right hand side. This
is possible only if eax+bt is constant or 4a2 + 2b + 1 = 0. eax+bt is constant only if a = 0
and b = 0, and for these values the equation becomes −2 = 1 which is not possible.
In the second case, the equation becomes 2a2 − 2b − 2 = 0. Adding 4a2 + 2b +√ 1 = 0 and
2a − 2b −√
2 2 = 0 we find 6a = 1 and hence b = −5/6. Therefore (a, b) = (1/ 6, −5/6)
2

and (−1/ 6, −5/6) are the only values for which the given function satisfies the
Fisher’s Equation.

84
12. The delayed heat equation
ut (x, t + 1) = uxx (x, t)
where u(x, t) is the temperature as a function of the position x and the time t, arises in the
problems of heat conduction in media which react to spatial variations in temperature with
a time delay. For example, the temperature of meat as it is cooked can be modeled with the
delayed heat equation.
Find all possible values of the pair of positive constants (a, b) for which the function

u(x, t) = sin(ax − bt)

satisfies the delayed heat equation for all (x, t).

Solution: We have ux = a cos(ax − bt), uxx = −a2 sin(ax − bt), ut = −b cos(ax − bt).
Hence, ut (x, t + 1) = uxx (x, t) for all (x, t) means

−b cos(ax − bt − b) = −a2 sin(ax − bt)

for all (x, t). Letting (x, t) = (0, 0) we obtain b cos b = 0. Since b > 0, this implies
π 3π
b = + 2πn or b = + 2πn for some nonnegative integer n.
2 2
π
If b = + 2πn, then
2
π π
cos(ax − bt − b) = cos(ax − bt − − 2πn) = cos(ax − bt − ) = sin(ax − bt)
2 2
and ut (x, t + 1) = uxx (x, t) for all (x, t) holds exactly when b = a2 .

3π
On the other hand, if b = + 2πn, then
2
3π 3π
cos(ax − bt − b) = cos(ax − bt − − 2πn) = cos(ax − bt − ) = − sin(ax − bt)
2 2
and ut (x, t + 1) = uxx (x, t) for all (x, t) holds exactly when b = −a2 , which is
impossible as b > 0.
1/2
π π
Therefore, (a, b) = (( + 2πn) , + 2πn), where n is a nonnegative integer, are the
2 2
only pairs of positive constants for which u(x, t) satisfies the delayed heat equation.

85
13. Assume that z and w are differentiable functions of x and y satisfying the equations xw3 +
∂z
yz 2 + z 3 = −1 and zw3 − xz 3 + y 2 w = 1 . Find at (x, y, z, w) = (1, −1, −1, 1) .
∂x

Solution: Differentiating the equations with respect to x and keeping in mind that
z and w depend on x, but y does not; we obtain

w3 + x ⋅ 3w2 wx + y ⋅ 2zzx + 3z 2 zx = 0 ,

and
zx w3 + z ⋅ 3w2 wx − z 3 − x ⋅ 3z 2 zx + y 2 wx = 0 .
Substituting x = 1, y = −1, z = −1, w = 1, we get 5zx + 3wx = −1 and 2zx + 2wx = 1 .
Solving for zx we find
∂z 5
=− at (x, y, z, w) = (1, −1, −1, 1) .
∂x 4

14. Once upon a time there was an xy-plane. The temperature at each point of this xy-plane
changed as time passed. Bugs roamed this xy-plane. Each bug had a device that measured the
temperature in real time, and showed the rate of change of temperature with respect to time
on its screen.
One day four of these bugs met at a point P0 . At the moment they met:
• The first bug was moving with velocity v1 = 2i + j m/s and its device was showing 1○ C/s.

• The second bug was moving with velocity v2 = i − 5j m/s and its device was showing
−1○ C/s.

• The third bug was moving with velocity v3 = i + j m/s and its device was showing 2○ C/s.

• The fourth bug was sitting still.

What was the device of the fourth bug showing?

86
 Solution:

15. Let z = f (x, y) be a differentiable function such that

f (3, 3) = 1, fx (3, 3) = −2, fy (3, 3) = 11,

f (2, 5) = 1, fx (2, 5) = 7, fy (2, 5) = −3.

Suppose w is a differentiable function of u and v satisfying the equation
f (w, w) = f (uv, u2 + v 2 )
∂w
for all (u, v). Find at (u, v, w) = (1, 2, 3).
∂u

Solution: Differentiating the identity

f (w, w) = f (uv, u2 + v 2 )
with respect to u gives
∂w ∂w ∂(uv) ∂(u2 + v 2 )
fx (w, w) + fy (w, w) = fx (uv, u2 + v 2 ) + fy (uv, u2 + v 2 )
∂u ∂u ∂u ∂u
87
 by the Chain Rule. Hence
∂w
(fx (w, w) + fy (w, w)) = fx (uv, u2 + v 2 ) v + fy (uv, u2 + v 2 ) 2u
∂u
which leads to
∂w
(fx (3, 3) + fy (3, 3)) = 2fx (2, 5) + 2fy (2, 5)
∂u
after substituting (u, v, w) = (1, 2, 3). Now using fx (3, 3) = −2, fy (3, 3) = 11,
fx (2, 5) = 7, and fy (2, 5) = −3, we conclude that
∂w 8
= at (u, v, w) = (1, 2, 3) .
∂u 9

16. Let u = x + y + z, v = xy + yz + zx, w = xyz, and suppose that f (u, v, w) is a differentiable

function satisfying f (u, v, w) = x4 + y 4 + z 4 for all (x, y, z). Find fu (2, −1, −2).

Solution: We can take (x, y, z) = (1, −1, 2) as this gives (u, v, w) = (2, −1, −2).

Differentiating f (u, v, w) = x4 +y 4 +z 4 with respect to x, y, z, respectively, we obtain:

fu ⋅ ux + fv ⋅ vx + fw ⋅ wx = 4x3
fu ⋅ uy + fv ⋅ vy + fw ⋅ wy = 4y 3
fu ⋅ uz + fv ⋅ vz + fw ⋅ wz = 4z 3
Now using u = x + y + z, v = xy + yz + zx, w = xyz, these give:
fu ⋅ 1 + fv ⋅ (y + z) + fw ⋅ yz = 4x3
fu ⋅ 1 + fv ⋅ (x + z) + fw ⋅ xz = 4y 3
fu ⋅ 1 + fv ⋅ (x + y) + fw ⋅ xy = 4z 3
Substituting (x, y, z) = (1, −1, 2) we get:
fu + fv − 2fw = 4
fu + 3fv + 2fw = −4
fu − fw = 32
Subtracting 3 times the first equation from the second gives −2fu + 8fw = −16, and
adding 8 times the third equation to this gives 6fu = 240. So fu (2, −1, −2) = 40.

Remark: There are such f . In fact, f (u, v, w) = u4 − 4u2 + 2v 2 + 4uw is one. Also note that
for a given (u, v, w), the corresponding (x, y, z) must be the roots of T 3 − uT 2 + vT − w = 0 and
hence is determined up to a permutation of its entries, making the answer independent of the
choice.

17. Let z = f (x, y) be a twice-differentiable function and x = r cos θ, y = r sin θ . Show that
∂ 2f ∂ 2f ∂ 2z 1 ∂z 1 ∂ 2z
+ = + + .
∂x2 ∂y 2 ∂r2 r ∂r r2 ∂θ2

88
 Solution: If z = F (x, y) is a differentiable function of x and y, then by the chain
rule we have
∂z ∂F ∂x ∂F ∂y
= ⋅ + ⋅ = Fx ⋅ cos θ + Fy ⋅ sin θ , (i )
∂r ∂x ∂r ∂y ∂r
and similarly,
∂z ∂F ∂x ∂F ∂y
= ⋅ + ⋅ = Fx ⋅ (−r sin θ) + Fy ⋅ (r cos θ) . (ii )
∂θ ∂x ∂θ ∂y ∂θ
We use (i ) with F = f to obtain
∂z
= fx ⋅ cos θ + fy ⋅ sin θ . (A)
∂r
Then
∂ 2z ∂ ∂
2
= (fx ) ⋅ cos θ + (fy ) ⋅ sin θ .
∂r ∂r ∂r
∂ ∂
To compute (fx ) and (fy ) we use (i ) with F = fx and F = fy , respectively:
∂r ∂r
∂ 2z
= (fxx cos θ + fxy sin θ) cos θ + (fyx cos θ + fyy sin θ) sin θ
∂r2
= fxx cos2 θ + 2fxy cos θ sin θ + fyy sin2 θ (B)

Similarly, using (ii ) with F = f gives

∂z
= fx ⋅ (−r sin θ) + fy ⋅ (r cos θ) ,
∂θ
and differentiating this with respect to θ again gives
∂ 2z ∂ ∂
2
= (fx ) ⋅ (−r sin θ) + fx ⋅ (−r sin θ)
∂θ ∂θ ∂θ
∂ ∂
+ (fy ) ⋅ (r cos θ) + fy ⋅ (r cos θ)
∂θ ∂θ
= (fxx (−r sin θ) + fxy r cos θ)(−r sin θ) + fx (−r cos θ)
+ (fyx (−r sin θ) + fyy r cos θ)(r cos θ) + fy (−r sin θ)
= fxx r sin θ − 2fxy r2 cos θ sin θ + fyy r2 cos2 θ
2 2

− r(fx cos θ + fy sin θ) (C)

where we used (ii ) with F = fx and F = fy .

Now if we add (B), 1/r times (A), and 1/r2 times (C), we obtain fxx + fyy .

18. Suppose that f (x, y) is a twice-differentiable function with continuous derivatives satisfying

x y
f( , 2 ) = f (x, y)
x2 + y x + y2
2

for all (x, y) =/ (0, 0). Find fxx (3/10, 1/10) if fx (3, 1) = −8, fy (3, 1) = 7, fxx (3, 1) = 2, fxy (3, 1) =
5, fyy (3, 1) = −4.

89
Solution: We have:
∂ x y
fx (x, y) = f( 2 , )
∂x x + y 2 x2 + y 2
x y ∂ x
= fx ( , 2 )⋅ ( 2 )
x2 +y x +y
2 2 ∂x x + y 2
x y ∂ y
+ fy ( , 2 )⋅ ( 2 )
x2 +y x +y
2 2 ∂x x + y 2
x y 1 ⋅ (x2 + y 2 ) − x ⋅ 2x
= fx ( , ) ⋅
x2 + y 2 x2 + y 2 (x2 + y 2 )2
x y −2xy
+ fy ( , 2 )⋅ 2
x2 +y x +y
2 2 (x + y 2 )2

Differentiating this a second time we obtain:

∂ x y y 2 − x2
fxx (x, y) = (fx ( 2 , ) ⋅
∂x x + y 2 x2 + y 2 (x2 + y 2 )2
x y −2xy
+ fy ( , 2 )⋅ 2 )
x2 +y x +y
2 2 (x + y 2 )2
2
x y y 2 − x2
= fxx ( 2 , ) ⋅ ( )
x + y 2 x2 + y 2 (x2 + y 2 )2
x y −2xy y 2 − x2
+ fxy ( , ) ⋅ ⋅
x2 + y 2 x2 + y 2 (x2 + y 2 )2 (x2 + y 2 )2
x y ∂ y 2 − x2
+ fx ( , ) ⋅ ( )
x2 + y 2 x2 + y 2 ∂x (x2 + y 2 )2
x y y 2 − x2 −2xy
+ fyx ( , ) ⋅ ⋅ 2
x +y x +y
2 2 2 2 (x + y ) (x + y 2 )2
2 2 2

2
x y −2xy
+ fyy ( 2 , ) ⋅ ( )
x + y 2 x2 + y 2 (x2 + y 2 )2
x y ∂ −2xy
+ fy ( , 2 )⋅ ( 2 )
x2 +y x +y
2 2 ∂x (x + y 2 )2
2
x y y 2 − x2
= fxx ( 2 , ) ⋅ ( )
x + y 2 x2 + y 2 (x2 + y 2 )2
x y y 2 − x2 −2xy
+ 2fxy ( , ) ⋅ ⋅ 2
x +y x +y
2 2 2 2 (x + y ) (x + y 2 )2
2 2 2

2
x y −2xy
+ fyy ( 2 , 2 )⋅( 2 )
x +y x +y
2 2 (x + y 2 )2
x y 2x(x2 − 3y 2 )
+ fx ( , ) ⋅
x2 + y 2 x2 + y 2 (x2 + y 2 )3
x y 2y(3x2 − y 2 )
+ fy ( , ) ⋅
x2 + y 2 x2 + y 2 (x2 + y 2 )3

90
 Now letting (x, y) = (3/10, 1/10) gives:

fxx (3/10, 1/10) = fxx (3, 10) ⋅ (−8)2 + 2fxy (3, 10) ⋅ (−8) ⋅ (−6) + fyy (3, 10) ⋅ (−6)2
+ fx (3, 10) ⋅ 36 + fy (3, 10) ⋅ 52
= 2 ⋅ 64 + 2 ⋅ 5 ⋅ 48 + (−4) ⋅ 36 + (−8) ⋅ 36 + 7 ⋅ 52
= 540

19. Consider the following conditions for a differentiable function f (x, y) :

Ê f (2, 1) = 8

Ë An equation for the tangent line to the level curve f (x, y) = 8 in the xy-plane at the point
(2, 1) is 3x − 5y = 1

Let P be the tangent plane to the graph of z = f (x, y) at the point (2, 1, 8) .
In each of the parts (a-e) below a Ìrd condition is given. Determine whether

• there is no function satisfying the conditions Ê-Ì ,

• there are functions satisfying the conditions Ê-Ì , but they do not all have the same
tangent plane P, or

• there are functions satisfying the conditions Ê-Ì and all of these functions have the same
tangent plane P. (In this case find an equation of P too.)

a. Ì f (3, 2) = 11
b. Ì fx (2, 1) = −1
d
c. Ì f (t2 + 1, t3 )∣ = 6
dt t=1

d. Ì The line with parametric equations x = 4t + 2 , y = 2t + 1 , z = t + 8 , (−∞ < t < ∞) , lies

in P
e. Ì The line with parametric equations x = −t + 2 , y = 2t + 1 , z = t + 8 , (−∞ < t < ∞) , is
perpendicular to P

Solution: The condition Ë implies that fx (2, 1) = 3c and fy (2, 1) = −5c for some
constant c. c is uniquely determined by the condition Ì in parts (b-d) and there
is no such c in part (e):

b. If fx (2, 1) = −1 gives c = −1/3. Hence fy (2, 1) = 5/3. Therefore all functions

satisfying the conditions Ê-Ì have the same tangent plane P. An equation
for this common tangent plane P is 3x − 5y + 3z = 25, and an example of a
function satisfying the given conditions is f (x, y) = −x + 5/3 y + 25/3.

91
 c. As we have
d
f (t2 + 1, t3 ) = fx (t2 + 1, t3 ) ⋅ 2t + fy (t2 + 1, t3 ) ⋅ 3t2
dt
by the Chain Rule, substituting t = 1 we obtain 3c⋅2+(−5c)⋅3 = 6 and c = −2/3.
Hence fx (2, 1) = −2 and fy (2, 1) = 10/3. Therefore all functions satisfying the
conditions Ê-Ì have the same tangent plane P. An equation for this common
tangent plane P is 6x − 10y + 3z = 26, and an example of a function satisfying
the given conditions is f (x, y) = −2x + 10/3 y + 26/3.
d. As the tangent plane P will have an equation of the form z = 3c⋅(x−2)+(−5c)⋅
(y − 1) + 8, for the line with parametric equations x = 4t + 2 , y = 2t + 1 , z =
t + 8 , (−∞ < t < ∞) , to lie in P we must have 1 = (3c) ⋅ 4 + (−5c) ⋅ 2 and c = 1/2.
Hence fx (2, 1) = 3/2 and fy (2, 1) = −5/2. Therefore all functions satisfying
the conditions Ê-Ì have the same tangent plane P. An equation for this
common tangent plane P is 3x − 5y − 2z = −15, and an example of a function
satisfying the given conditions is f (x, y) = 3/2 x − 5/2 y + 15/2.
e. As the tangent plane P will have an equation of the form 3c ⋅ (x − 2) + (−5c) ⋅
(y − 1) − (z − 8) = 0, for the line with parametric equations x = −t + 2 , y =
2t + 1 , z = t + 8 , (−∞ < t < ∞) , to be perpendicular to P we must have the the
normal vector n = 3ci−5cj−k of the plane and the velocity vector v = −i+2j+k
of the line to be parallel. This requires 3c/(−1) = (−5c)/2 = −1/1, which is not
possible. Therefore there is no function f satisfying the conditions Ê-Ì .

Finally, in part (a), both of the functions f (x, y) = 3(x − 2) − 5(y − 1) + 8 + 5(x − 2)2
and f (x, y) = −3(x − 2) + 5(y − 1) + 8 + (x − 2)2 satisfy the conditions Ê-Ì , but
equations of their graphs’ tangent planes P at (2, 1, 8) are z = 3x − 5y + 7 and
z = −3x + 5y + 9, respectively. Therefore, in this case there are functions satisfying
the conditions Ê-Ì , but they do not all have the same tangent plane P.

20. Let f (x, y) = x3 y − xy 2 + cx2 where c is a constant. Find c if f increases fastest at the point
P0 (3, 2) in the direction of the vector A = 2i + 5j.

Solution: ∇f = (3x2 y −y 2 +2cx)i+(x3 −2xy)j Ô⇒ (∇f )P0 = (50+6c)i+15j . Since f

increases the fastest at P0 in the direction of (∇f )P0 , A = 2i + 5j must be a positive
multiple of (∇f )P0 . Hence (50 + 6c)/2 = 15/5 Ô⇒ c = −22/3 . Finally we check that
c = −22/3 gives (∇f )P0 = 6i + 15j = 3A which is indeed a positive multiple of A.

21. Find a vector that is tangent to the intersection curve of the surfaces x2 + y 2 + z 2 = 9 and
z = xy at the point P0 (1, 2, 2).

Solution: Let f (x, y, z) = x2 +y 2 +z 2 and g(x, y, z) = xy −z. Then the given surfaces
are level surfaces of f and g.

We have ∇f = 2xi + 2yj + 2zk Ô⇒ (∇f )P0 = 2i + 4j + 4k and ∇g = yi + xj − k Ô⇒

(∇g)P0 = 2i + j − k . (∇f )P0 is normal to the surface defined by x2 + y 2 + z 2 = 9 and

92
 (∇g)P0 is normal to the surface defined by z = xy. Therefore,
RRR i j k RRRR
RR R
(∇f )P0 × (∇g)P0 = RRRR 2 4 4 RRRR = −8i + 10j − 6k ,
RRR R
RR 2 1 −1 RRRR

or any multiple of it, is tangent to both of these surfaces, and hence, to their curve
of intersection at P0 .

22. Suppose that a differentiable function f (x, y, z) satisfies the following conditions:
∂f
Ê (3, 1, 3) = 1.
∂y
Ë The parametric curve

C1 ∶ r1 = (3 + 2t) i + (1 − t2 ) j + (3 − 5t + t2 ) k (−∞ < t < ∞)

is contained in the level surface S of f (x, y, z) passing through the point (3, 1, 3).
Ì The parametric curve

C2 ∶ r2 = (2 + t2 ) i + (2t3 − 1) j + (2t + 1) k (−∞ < t < ∞)

is also contained in the level surface S of f (x, y, z) passing through the point (3, 1, 3).
∂f
Find (3, 1, 3).
∂z
Solution: We compute the velocity vectors
dr1
v1 = = 2i − 2tj + (−5 + 2t)k Ô⇒ v1 ∣t=0 = 2i − 5k
dt
and
dr2
v2 = = 2ti + 6t2 j + 2k Ô⇒ v2 ∣t=1 = 2i + 6j + 2k
dt
for the curves C1 and C2 at the point (3, 1, 3). Since these curves are contained in
the level surface, their tangents at this point are tangent to the level surface at this
point. Therefore, their cross product
RRR i j k RRRR
RR R
n = v1 ∣t=0 × v2 ∣t=1 = RRRR 2 0 −5 RRRR = 30i − 14j + 12k
RRR R
RR 2 6 2 RRRR

is normal to the level surface at (3, 1, 3). Since (∇f )(3, 1, 3) is also perpendicular
to the level surface at (3, 1, 3), we must have (∇f )(3, 1, 3) = cn = 30ci − 14cj + 12ck
for some scalar c. As fy (3, 1, 3) = 1, we have −14c = 1 and c = −1/14. Therefore,
fz (3, 1, 3) = 12c = 12(−1/14) = −6/7.

93
23. Find all points P (a, b, c) on the paraboloid z = x2 + y 2 such that the tangent plane to the
paraboloid at the point P contains the line:

L ∶ x = 3t + 1 , y = t − 1 , z = 14t − 3 ; (−∞ < t < ∞)

Solution: Let f (x, y) = x2 + y 2 . Then an equation for the tangent plane is:

z = fx (a, b)(x − a) + fy (a, b)(y − b) + f (a, b)

Since fx = 2x and fy = 2y, this becomes:

z = 2a(x − a) + 2b(y − b) + a2 + b2

That is:
z = 2ax + 2by − a2 − b2
The line L is contained in this tangent plane if and only if

14t − 3 = 2a(3t + 1) + 2b(t − 1) − a2 − b2

for all t. This is equivalent to (a, b) satisfying both of the equations 6a + 2b = 14 and
2a − 2b − a2 − b2 = −3. From the first equation we have b = 7 − 3a„ and substituting
this to the second one gives 10a2 − 50a + 60 = 0. Therefore a must be either 2 or 3.
The corresponding values of b are 1 and −2.

Hence (a, b, c) = (2, 1, 5) and (3, −2, 13) are the only points satisfying the condition.

24. Three students are working on a problem involving a differentiable function f (x, y, z) at a
point P0 .
Student A says: An equation for the tangent plane to the level surface of f passing
through P0 is 2x + 3y − 6z = 11.

Student B says: The largest possible rate of change of f at P0 in any direction is 25.

Student C says: The directional derivative of f at P0 in the direction of the vector

A = 2i + 2j + k is 5.
Student D, who has been listening in, but does not know anything about the function f , decides
that at least one of the Students A, B, C must have made a mistake in their calculations.
Explain Student D’s reasoning.

Solution: If Student A’s statement is correct, then (∇f )(P0 ) = c(2i + 3j − 6k) for
some scalar c as the gradient of f at P0 is normal to the tangent plane to its level
surface through P0 .

If Student C ’s statement is correct, then Du (P0 ) = 5 where u = A/∣A∣ = (2i+2j+k)/3.

Since (Du f )(P0 ) = (∇f )(P0 ) ⋅ u, we obtain 5 = c(2i + 3j − 6k) ⋅ (2i + 2j + k)/3 Ô⇒ c =
15/4.

94
 √
Therefore, (∇f )(P0 ) = 15(2i + 3j − 6k)/4, and ∣(∇f )(P0 )∣ = 15 22 + 32 + (−6)2 /4 =
105/4. This contradicts Student B ’s statement as the largest possible rate of change
of f at P0 is ∣(∇f )(P0 )∣ and 25 =/ 105/4.

Hence, not all three statements can be correct.

25. You see the following question in an old Calculus book:

Find ∇f (P0 ) if f (x, y, z) is a differentiable function such that:

À f ( t2 + 1 , 5t , t3 − t ) = t − t2 for all t, and

Á At the point P0 (5, 10, 6), f increases the fastest in the direction of
the vector A = 83i + j − 2k.

The first component of A, which is only partly legible, seems to be either 3 or 8.

Show that one of these possibilities can be eliminated if the question is correct.

Solution: By differentiating the identity in À and using the Chain Rule we obtain

∇f ⋅ (2t i + 5 j + (3t2 − 1) k) = 1 − 2t

for all t. Substituting t = 2 in this gives:

∇f (P0 ) ⋅ (4i + 5j + 11k) = −3

Suppose that A = ai + j − 2k. Then, by Á, ∇f (P0 ) = bA for some b > 0. Substituting

this in the equation above we obtain b(ai+j−2k)⋅(4i+5j+11k) = −3 Ô⇒ b(4a−17) =
−3. Since b > 0, we must have 4a − 17 < 0. Therefore a < 17/4. Hence a cannot be
8.

Remark: There are many functions satisfying the conditions of this example.
6 2 2 3 7 6 3
f (x, y, z) = xy − y − x− y− z+
25 125 5 25 5 5
is one.

95
26. In the figure below some of the level curves and the corresponding values of a nice function
f (x, y) are shown.

a. Draw ∇f (0, 0) as best you can on the figure.

b. Determine the signs of the derivatives fx , fy , fxx , fyy , fxy at the origin.

Remark: This is an approximation problem and it should be solved under the assumption
that we are dealing with a function for which good approximations can be made using only the
given data.

(The solution starts on the next page.)

96
Solution: a. We know that ∇f (0, 0) is perpendicular to the level curve of f
passing through (0, 0) and points in the direction f increases, hence lies along the
pink vector in the figure.

We also know that the length of ∇f (0, 0) is the rate of change of f with respect to
distance in this direction. On the figure using the teal circle we measure that the
pink vector crosses the level curve f = 1 at a distance of approximately 0.44 units
from the origin. Hence along this direction f increases at an rate of 1/0.44 ≈ 2.27
in this direction. So now we draw a vector of this length in this direction as shown
in the next figure and this is our approximate ∇f (0, 0).

(The figure is on the next page.)

97
b. Considering the components of ∇f (0, 0) in part (a) it can be seen that fx (0, 0) >
0 and fy (0, 0) < 0. This can also be done in the following way: As we move along the
x-axis (the lime line in the figure on the next page) to the right, we meet level curves
belonging to higher values of f; hence fx (0, 0) > 0. Similarly, as we move along the
y-axis (the orange line) upwards, we meet level curves belonging to smaller values
of f values; hence fy (0, 0) < 0.

Moreover, note that as we move in the positive direction along both axes, the points
where we meet the level curves of f belonging to values with the same difference
become farther apart. This means that the absolute value of the rate of change of f
is getting smaller. This in turn means fxx (0, 0) < 0 as fx (0, 0) > 0, and fyy (0, 0) > 0
as fy (0, 0) < 0.

(Part (b) is continued on the next page.)

98
 Finally, we compare the rate of changes along the lime and turquoise lines. The
level curves intersect the turquoise line at points farther apart than the points they
intersect the lime line. Hence fx , which is positive, is getting smaller as we move
upwards along y-axis. In other words, fxy (0, 0) < 0.

Remark: It is also possible to find approximate values of these derivatives using the figure. In
fact, those of fx (0, 0) and fy (0, 0) can already be read off as the components of ∇f (0, 0).

99
27. Let a be a constant. Find and classify all critical points of f (x, y) = x3 − 3axy + y 3 .

Solution: At a critical point fx = 3x2 − 3ay = 0 and fy = −3ax + 3y 2 = 0. If a = 0,

then 3x2 = 0 ⇒ x = 0 and 3y 2 = 0 ⇒ y = 0, and (0, 0) is the only critical point.
If a =/ 0, then the first equation gives y = x2 /a, and substituting this in the second
equation we get x4 −a3 x = 0 whose solutions are x = 0 and x = a. Now using y = x2 /a,
we get (0, 0) and (a, a) as the critical points.

We compute the discriminant:

f f 6x −3a
∆ = ∣ xx xy ∣ = ∣ ∣
fyx fyy −3a 6y

As ∆(a, a) = 27a2 and fxx (a, a) = 6a, (a, a) is a local minimum for a > 0 and a
local maximum for a < 0. On the other hand, ∆(0, 0) = −9a2 implies that (0, 0) is a
saddle point for a =/ 0.

Now we look at the sole critical point (0, 0) in the case a = 0. As ∆(0, 0) = 0,
the second derivative test fails in this case. If we restrict f (x, y) = x3 + y 3 to the
x-axis we get f (x, 0) = x3 . Since this single variable function does not have a local
maximum or minimum at x = 0, f (x, y) cannot have a local maximum or minimum
at (0, 0) either. We conclude that (0, 0) is a saddle point when a = 0.

28. Find the absolute maximum and minimum values of the function

f (x, y) = 2x3 + 2xy 2 − x − y 2

on the unit disk D = {(x, y) ∶ x2 + y 2 ≤ 1}.

Solution: We first find the critical points of f (x, y) in the interior of D. At a

critical point we have fx = 6x2 + 2y 2 − 1 = 0 and fy = 4xy − 2y = 0. The second
equation implies√ that y = 0 or x = 1/2. Substituting these into the first equation we
obtain x = ±1/ 6 in the first case,√and no solution √ in the second case. Therefore
the critical points are (x, y) = (1/ 6, 0) and (−1/ 6, 0). Note that both of these
points lie in the interior of D.

√ boundary of D, that is, the unit circle x + y = 1. We can

Now we look at the 2 2

solve y as y = ± 1 − x2 , −1 ≤ x ≤ 1. √ These solutions correspond to the upper

and lower semicircles. Then f (x, ± 1 − x2 ) = x2 + x − 1 for −1 ≤ x ≤ 1, and
d √ 1
f (x, ± 1 − x2 ) = 2x + 1 = 0 ⇒ x = − . This gives the critical points (x, y) =
dx √ √ 2
(−1/2, 3/2) and (−1/2, − 3/2) of the restriction of f to the boundary of D. We
must also include the endpoints x = −1 and x = 1, in other words, the points
(x, y) = (1, 0) and (−1, 0) in our list.

Hence the absolute maximum and the absolute minimum of f on D occur at some
of the points
√ √ √ √
(1/ 6, 0), (−1/ 6, 0), (−1/2, 3/2), (−1/2, − 3/2), (1, 0), (−1, 0) .

100
 The values of f at these points are
√ √
1 2 1 2 5 5
− , , − , − , 1 , −1 ,
3 3 3 3 4 4
respectively. Therefore the absolute maximum value is 1 and the absolute minimum
value is −5/4.

Remark: Here are two more ways of dealing with the critical points of the restriction of f to
the boundary of D.
In the first one we parametrize the boundary, which is the unit circle, by x = cos t, y = sin t,
−∞ < t < ∞. Then
d d 1
f (cos t, sin t) = (cos2 t + cos t − 1) = −2 cos t sin t − sin t = 0 ⇒ cos t = − or sin t = 0 ,
dt dt 2
√ √
and these give us (x, y) = (−1/2, 3/2), (−1/2, − 3/2), (1, 0), (−1, 0).
g(x,y)
³¹¹ ¹ ¹ ·¹¹ ¹ ¹ µ
In the second we use the Lagrange Multipliers Method for the boundary x2 + y 2 = 1.
⎧
⎪ f = λgx ⎫⎪ ⎧
⎪ 6x2 + 2y 2 − 1 = λ 2x ○
1
∇f = λ∇g ⎪
⎪ x ⎪
⎪ ⎪
⎪
} Ô⇒ ⎨ fy = λgy ⎬ Ô⇒ ⎨ 4xy − 2y = λ 2y ○ 2
g = 1 ⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎩ g = 1 ⎪
⎭ ⎪
⎩ x 2 + y2 = 1 3
○

○2 gives y = 0 or x = (1 +λ)/2. If y = 0, then ○3 gives x = ±1 (and λ = ±5/2). On√the other hand,

if x = (1 + λ)/2, then from√○1 and ○, 3 4x√ 2 − 2λx + 1 = 0 ⇒ x = −1/2 and y = ± 3/2. Therefore

the points (x, y) = (−1/2, 3/2), (−1/2, − 3/2), (1, 0), (−1, 0) are added to the list.

101
29. Find the absolute maximum and minimum values of the function

f (x, y) = 2(x2 + y 2 − 1)2 + x2 − y 2

on the unit disk D = {(x, y) ∶ x2 + y 2 ≤ 1} .

Solution: We first find the critical points of f in the interior of D. We want:

fx = 4(x2 + y 2 − 1) ⋅ 2x + 2x = 0 and fy = 4(x2 + y 2 − 1) ⋅ 2y − 2y = 0 .

The first equation gives x = 0 or 4x2 + 4y 2 − 3 = 0, and the second one gives y = 0
or 4x2 + 4y 2 − 5 = 0. Therefore x = 0 and y = 0, or 4x2 + 4y 2 − 3 = 0 and y = 0, or
x = 0 and 4x2 + 4y 2 − 5 = 0, or 4x2 + 4y 2 − 3 = 0 and 4x2 + 4y 2 − 5 = 0. In the last
√ case
there is no
√ solution, and the first three cases leads to the points (0, 0), (± 3/2, 0)
and (0, ± 5/2). Only the first three of these critical points lie in the interior of D
as for the last two x2 + y 2 = 5/4 > 1.

Next we consider the restriction

√ of f to the boundary of D which √ is the unit circle.
As on the boundary y = ± 1 − x , we consider
2
√ the functions f (x, ± 1 − x ) = 2x −1
2 2

on the interval −1 ≤ x ≤ 1. As (d/dx)f (x, ± 1 − x2 ) = 4x, the only critical point of

both of these restrictions occur when x = 0. This leads the points (x, y) = (0, 1) and
(0, −1). On the other hand, for both functions the endpoints x = 1 and x = −1 of
the interval −1 ≤ x ≤ 1 give the points (x, y) = (1, 0) and (−1, 0).

Now we find the values of f at these seven points:

f (0, 0) = 2
√ √
f ( 3/2, 0) = f (− 3/2) = 7/8
f (0, 1) = f (0, −1) = −1
f (1, 0) = f (−1, 0) = 1

Therefore the absolute maximum and minimum values of f on D are 2 and −1,
respectively.

102
Remark: The boundary can also be dealt with using the parametrization x = cos t, y = sin t,
(−∞ < t < ∞), of the unit circle. Then the restriction of f to the unit circle gives the function
f (cos t, sin t) = cos 2t on the interval −∞ < t < ∞. (d/dt)f (cos t, sin t) = −2 sin 2t = 0 means that
t is an integer multiple of π/2 and this gives the critical points (x, y) = (±1, 0) and (0, ±1) for
the restriction of f to the boundary.
Remark: Consider the function g(x, y) = 2(x2 + y 2 − 1)2 + 2xy on the unit disk D. This is the
same function as f rotated 45○ counterclockwise. Hence the absolute maximum value of g on
D will be 2 assumed√ at √the point (x, y) = (0, 0), the absolute minimum value −1 at the
√ and √
points (x, y) = (1/ 2, 1/ 2) and (x, y) = (−1/ 2, −1/ 2). It is a good exercise to solve this
problem directly.

30. Three hemispheres with radiuses 1, x and y, where 1 ≥ x ≥ y ≥ 0, are stacked on top of each
other as shown in the figure. Find the largest possible value of the total height h.

103
Solution: We want to maximize
√ √
h(x, y) = 1 − x2 + x2 − y 2 + y

on the closed and bounded region D = {(x, y) ∶ 0 ≤ y ≤ x ≤ 1}.

We first find the critical points of h in the interior of D. We want

x x y
hx = − √ +√ = 0 and hy = − √ + 1 = 0.
1 − x2 x2 − y 2 x2 − y 2
√
From the second equation we obtain y 2 = x2 −y 2 Ô⇒ x2 = 2y 2 Ô⇒ x = 2y where we
used the fact that x > 0 and y > 0. Now√substituting this in the first
√ equation gives
1−2y = 2y −y Ô⇒ 3y = 1 Ô⇒
2 2 2 2
√ y = 1/√ 3 as y > 0 and hence x = 2/3. So
√the only
√
critical point of h is (x, y) = ( 2/3, 1/ 3) and it lies in D as 0 ≤ 1/ 3 ≤ 2/3 ≤ 1.

Now we consider the restriction of h to the boundary of D.

Side 1 : On the bottom edge of the triangle, we √ have y = 0 and 0 ≤ x ≤ 1, and
therefore we are considering the function h(x, 0) = 1 − x2 + x for 0 ≤ x ≤ 1.

d x 1
h(x, 0) = − √ + 1 = 0 Ô⇒ x2 = 1 − x2 Ô⇒ 2x2 = 1 Ô⇒ x = √
dx 1 − x2 2
as x > 0. Taking √ the endpoints x = 0 and x = 1 into account, Side 1 gives us the
points (x, y) = (1/ 2, 0), (0, 0), (1, 0).
we have x = 1 and 0 ≤ y ≤ 1, and therefore
Side 2 : On the right edge of the triangle, √
we are considering the function h(1, y) = 1√ − y 2 + y for 0 ≤ y ≤ 1. As in the case of
Side 1 this leads to the points (x, y) = (1, 1/ 2), (1, 0), (1, 1).
√ have y = x and 0 ≤ x ≤ 1, and therefore
Side 3 : On the top edge of the triangle, we
we are considering the function h(x, x) = 1 − x2 + x√for 0√≤ x ≤ 1. Once again as in
the case of Side 1 we obtain the points (x, y) = (1/ 2, 1/ 2), (0, 0), (1, 1).

104
 Now we find the values of h at these seven points:

h(0, 0) = h(1, 0) = h(1, 1) = 1

√ √ √ √ √
h(1/ 2, 0) = h(1, 1/ 2) = f (1/ 2, 1/ 2) = 2
√ √ √
h( 2/3, 1/ 3) = 3
√
Therefore the maximum possible total height of the three hemispheres is 3.

Remark: Here is a single variable argument which solves the problem √ for any number
hemispheres: Suppose the maximum possible height for k hemispheres is k for some k ≥ 1.
Then for k + 1 hemispheres, where
√ the√ second one from the bottom has radius r, the maximum
√
possible height will be H(r) = kr + 1 − r2 for 0 ≤ r ≤ 1. The√ value of H(r) is 1 and k at
endpoints r = 0 and r = 1, respectively; and its value is k + 1 at its only critical point
the √
r = k/(k + 1).√ Therefore we conclude inductively that the maximum possible height for n
hemispheres is n for n ≥ 1.

31. Find the absolute maximum and the absolute minimum of f (x, y, z) = x3 + yz on the unit
sphere x2 + y 2 + z 2 = 1.

Solution: We will use the Lagrange Multipliers Method. Let g(x, y, z) = x2 + y 2 +

z 2 − 1. Then
⎧
⎪ fx = λgx ⎫
⎪ ⎧
⎪ 3x2 = λ 2x 1
○
⎪
⎪
⎪ ⎪
⎪
⎪ ⎪
⎪
⎪
∇f = λ∇g ⎪ fy = λgy ⎪ ⎪ z = λ 2y 2
○
} Ô⇒ ⎨ ⎬ Ô⇒ ⎨
g = 0 ⎪
⎪ fz = λgz ⎪
⎪ ⎪
⎪ y = λ 2z 3
○
⎪
⎪
⎪ ⎪
⎪
⎪ ⎪
⎪
⎪
⎩ g = 0 ⎭ ⎩ x 2 + y2 + z2 = 1 4
○

3 we obtain z = 4λ2 z, and hence z = 0 or λ = 1/2 or λ = −1/2.

2 and ○
From ○

• If z = 0, then ○
3 gives y = 0, and then ○
4 gives x = ±1. In this case we have
the points (±1, 0, 0).
• If λ = 1/2, then from ○ 1 and ○,3 3x√2 = x and y = z. Therefore we either
have x = 0 in which case y = z = ±1/ 2 by ○, 4 or we have x = 1/3 √ and then
√
y = z = ±2/3 again by ○.
4 In this case the critical points are (0, ±1/ 2, ±1/ 2)
and (1/3, ±2/3, ±2/3).
• If λ = −1/2, then from ○ 1 and ○, 3 3x2 = −x √ and y√= −z. A reasoning similar to
the previous case gives the points (0, ±1/ 2, ∓1/ 2) and (−1/3, ±2/3, ∓2/3).
√ √ √ √
Hence the critical points are (±1, 0, 0), (0, ±1/ 2, ±1/ 2), (1/3, ±2/3, ±2/3), (0, ±1/ 2, ∓1/ 2), (−1/3,
and the values of f at these points are ±1, 1/2, −1/2, 13/27, −13/27 , respectively.
Therefore the absolute maximum is 1 and the absolute minimum is −1.

105
32. Evaluate the following integrals:
1 1
a. ∫ ∫√ sin(πx3 ) dx dy
0 y

b. ∬ y 2 e−x dA where R = {(x, y) ∶ 0 ≤ y ≤ x}

2

R
√
2 2y−y 2 xy
c. ∫ ∫ dx dy
0 0 x2+ y2
∞ ∞ dy dx
d. ∫ ∫0
0 (x2 + y 2 )2 + 1

Solution: a. We will first express the iterated integral as a double integral and
√
then reverse the order of integration. The x-integral goes from x = y to x = 1 as
shown by the red line segments in the figure in the xy-plane. Then the y-integral
goes from y = 0 to y = 1.

Therefore the intervals of the x-integral trace out the region R bounded by the
parabola y = x2 , the line x = 1, and the x-axis. Note that the x-integrals are always
from left to right in the interval 0 ≤ y ≤ 1. Hence we have
1 1
∫0 ∫√y sin(πx ) dx dy = ∬R sin(πx ) dA .
3 3

Now we express this double integral as an iterated integral with the y-integral first.
The green line segment in the figure shows the interval of integration for the y-
integral which goes from y = 0 to y = x2 . Then:

106
 1 1
∫0 ∫√y sin(πx ) dx dy = ∬R sin(πx ) dA
3 3

1 x2
=∫ ∫0 sin(πx3 ) dy dx
0
1
y=x2
=∫ sin(πx3 ) y]y=0 dx
0
1
=∫ sin(πx3 )x2 dx
0
1 1
=− cos(πx3 )]0
3π
2
=
3π

b. We integrate with respect to y first to obtain

∞ x
2 −x 2 2 −x 2
∬R y e dA = ∫0 ∫0 y e dy dx
1 ∞
= ∫ x3 e−x dx
2

3 0
1 ∞ −t
= ∫ te dt
6 0
1
=
6
where we used the integration by parts
∞ ∞
∫0 te−t dt = ∫ t d(−e−t )
0
c
= lim ([−te−t ]c0 + ∫ e−t dt)
c→∞ 0
c
= − lim c − lim [e−t ]c0
c→∞ e c→∞
L’H
↓ 1
= − lim c − lim (e−c − 1)
c→∞ e c→∞
=1

in the last step.

c. This time we will use the polar coordinates. To do so we first determine the
region of integration R.
√
In the iterated integral the x-integral goes from x√= 0 to x = 2y − y 2 as shown
√ in the figure. Since x = 2y − y Ô⇒ x = 2y − y Ô⇒
by the red line segment 2 2 2

x2 + (y − 1)2 = 12 , x = 2y − y 2 gives the right semicircle of the circle x2 + (y − 1)2 = 1.

On the other hand, the y-integral goes from y = 0 to y = 2. Therefore R is right half
of the disk x2 +(y −1)2 ≤ 1. Note that the polar equation of the circle x2 +(y −1)2 = 1
is r = 2 sin θ, and to obtain the right semicircle we vary θ from θ = 0 to θ = π/2.

107
Hence:
√
2 2y−y 2 xy xy
∫0 ∫0 dx dy = ∬ 2 dA
x2 +y 2 R x + y2
π/2 2 sin θ r cos θ ⋅ r sin θ
=∫ ∫ r dr dθ
0 0 r2
π/2 2 sin θ
=∫ ∫0 sin θ cos θ r dr dθ
0
π/2 r2 r=2 sin θ
=∫ sin θ cos θ [ ] dr dθ
0 2 r=0
π/2
=∫ 2 sin3 θ cos θ dθ
0
π/2
sin4 θ
= ]
2 0
1
=
2

d. Again we use the polar coordinates. This time the integration region is R =
{(x, y) ∶ x ≥ 0 and y ≥ 0}, that is, the first quadrant.
∞ ∞ dy dx 1
∫0 ∫0 = ∬ dA
(x2 + y 2 )2 + 1 R (x2 + y 2 )2 + 1
π/2 ∞ r dr dθ
=∫ ∫
0 0 r4 + 1

108
 π/2
1 r=c
=∫ lim [arctan(r2 )]r=0 dθ
0 2 c→∞
π/2 π
=∫ dθ
0 4
π2
=
8

1
33. Evaluate the double integral ∬ dA where R is the region shown in the figure.
R (x + y 2 )2
2

Solution: Let R′ be the portion of the region lying in the first quadrant between
the line y = x and the x-axis. By symmetry we have:
1 1
∬R (x2 + y 2 )2 dA = 8 ∬R′ (x2 + y 2 )2 dA
π/4 2 sec θ 1
= 8∫ ∫ r dr dθ
0 sec θ (r2 )2
π/4 1 r=2 sec θ
= 8∫ [− 2 ] dθ
0 2r r=sec θ
π/4
= 3∫ cos2 θ dθ
0
1 + cos 2θ
π/4
= 3∫ dθ
0 2
3 sin 2θ π/4
= [θ + ]
2 2 0
3π 3
= +
8 4

109
34. Evaluate the following integrals.

a. ∬ x2 y cos(πy 5 /2) dA where R1 is the region shown in the figure.

R 1

b. ∬ x2 y cos(πy 5 /2) dA where R2 is the region shown in the figure.

R 2

y y=x

y
y=1

R1 x
R2
x

y=x y = −x

x=1 y = −x

Solution: a. Note that the region R1 is symmetric with respect to the x-axis, but
the function f (x, y) = x2 y cos(πy 5 /2) changes sign under reflection with respect to
the x-axis: f (x, −y) = −f (x, y). Therefore ∬ x2 y cos(πy 5 /2) dA = 0.
R 1

b. We integrate with respect to x, then y:

1 y 1 x=y
1 3
∬R x2 y cos(πy 5 /2) dA = ∫∫−y x2 y cos(πy 5 /2) dx dy = ∫
[ x y cos(πy 5
/2)] dy
2 0 0 3 x=−y
1
2 1 2 2 4
= ∫ y 4 cos(πy 5 /2) dy = [ sin(πy 5 /2)] =
3 0 3 5π 0 15π

110
35. Evaluate the following integral ∬ (1 + x − y) dA where R = {(x, y) ∶ ∣x − y∣ ≤ 2/3 and 0 ≤
R
x ≤ 1 and 0 ≤ y ≤ 1}.

Solution: We first observe that

∬R (1 + x − y) dA = ∬R dA + ∬R (x − y) dA

and then note that the second integral on the right is zero because the region R is
symmetric with respect to the line y = x whereas the function f (x, y) = x−y changes
sign under the reflection with respect to this line: f (y, x) = −f (x, y).

On the other hand the first on the right is just the area of R, that is 12 −(1/3)2 = 8/9.
Hence:
8 8
∬R (1 + x − y) dA = ∬R dA + ∬R (x − y) dA = 9 + 0 = 9

111
36. When pirates retire, they live on the Square Island

S = {(x, y) ∶ ∣x∣ ≤ 5 and ∣y∣ ≤ 5}

in the Sea of xy-plane where all distances are measured in hectometers (=hm). Having lived all
their lives on it, the retired pirates want to be as far away from the sea as possible. As a result,
the pirate population density p(x, y) at a point (x, y) on the Square Island is proportional to
the distance of the point from the shore and reaches its largest value of 15 pirate/hm2 at the
center of the island. Find the total number N of pirates living on the Square Island.

Solution:

Remark: Compare this example with Example 47 in Part 1.

112
37. The region enclosed by the lemniscate r2 = 2 cos(2θ)
√ in the plane is the base of a solid right
cylinder whose top is bounded by the sphere z = 2 − r2 . Find the cylinder’s volume.

Solution: Let R be the the region enclosed by the lemniscate in the first quadrant.
By symmetry the volume is
√
√ π/4 2 cos 2θ √
4∬ 2 − r2 dA = 4 ∫ ∫0 2 − r2 r dr dθ
R 0
√
π/4 r= 2 cos 2θ
1 4 π/4 3/2
= 4∫ [− (2 − r2 )3/2 ] dθ = ∫ (2 − (2 − 2 cos 2θ)3/2 ) dθ
0 3 3 0
√ r=0
√
8 2 π/4 2 2π 32 π/4 3
= ∫ (1 − (2 sin2 θ)3/2 ) dθ = −
3 √ 3 ∫0
sin θ dθ
3
√ 0
2 2π 32 π/4 2 2π 32 1
= − ∫ (1 − cos2 θ) sin θ dθ = − ∫ √ (1 − u ) du
2
3 3 0 3 3 1/ 2
√ 1
√
2 2π 32 u 3 2 2π 32 2 5
= − [u − ] √ = − ( − √ )
3 3 3 1/ 2 3 3 3 6 2
√ √
2 2π 64 40 2
= − + .
3 9 9

113
38. Consider the iterated integral:
2 1− 12 z 3−2x− 23 z
I =∫ ∫ ∫0 f (x, y, z) dy dx dz
0 0

a. The iterated integral I corresponds to a triple integral

∭E f (x, y, z) dV

where E is a region in space. Draw a picture of the region E, and clearly label the surfaces
bounding it with their equations and clearly label the important points with their coordinates.
b. Express the iterated integral I in terms of iterated integrals with the order of integration
dz dy dx.

Solution: a.

b.
1 x 2−2x 1 3−2x 2−4x/3−2y/3
I =∫ ∫0 ∫0 f (x, y, z) dz dy dx + ∫ ∫x ∫0 f (x, y, z) dz dy dx
0 0

114
39. Let D be the region in space bounded by the plane y + z = 1 on the top, the parabolic
cylinder y = x2 on the sides, and the xy-plane at the bottom. Express the volume V of the
region D in terms of iterated integrals with orders of integration (a) dz dy dx and (b) dx dy dz.

Solution:

115
 1
40. Consider the triple integral ∭ dV where D is the region bounded by the
D (x2 + y 2 + z 2 )2
√
cylinder x2 + y 2 = 1 on the sides and by the hemisphere z = 4 − x2 − y 2 at the bottom. Express
this integral in terms of iterated integrals in
a. Cartesian coordinates,
b. cylindrical coordinates,
c. spherical coordinates, and
d. evaluate the integral in the coordinate system of your choice.

Solution: We observe that the projection of D to the xy-plane is the unit disk
x2 + y 2 ≤ 1. A vertical line passing through a point of the unit disk enters the region
at a point on the hemisphere
√ and remains in the region from there on. The
√ equations
of the hemisphere z = 4 − x − y and the cylinder x + y = 1 are z = 4 − r2 and
2 2 2 2

r = 1, respectively, in cylindrical coordinates. The answers to parts (a),

√
1 1 1−x2 ∞ 1
∭D (x2 + y 2 + z 2 )2 dV = ∫−1 ∫−√1−x2 ∫ 4−x2 −y2 (x2 + y 2 + z 2 )2 dz dy dx ,
√

and (b),

1 2π 1 ∞ 1
∭D (x2 + y 2 + z 2 )2 dV = ∫0 ∫0 ∫√4−r2 (r2 + z 2 )2 r dz dr dθ ,

immediately follow from these observations.

To do part (c) we further observe that

• the equations of the hemisphere and the cylinder are ρ = 2 and ρ sin ϕ = 1 ,
respectively,
• a ray starting at the origin enters the region at a point on the hemisphere and
leaves the region at a point on the cylinder, and
• such a ray intersects the region exactly when 0 ≤ ϕ ≤ π/6. (The ray passes
through a point on the intersection circle when ϕ = π/6.)

Therefore,

1 2π π/6 csc ϕ 1
∭D (x2 + y 2 + z 2 )2 dV = ∫0 ∫0 ∫2 ρ 4
ρ2 sin ϕ dρ dϕ dθ .

116
We use this last iterated integral for the computation (d):

1 2π π/6 csc ϕ 1
∭D (x2 + y 2 + z 2 )2 dV = ∫0 ∫0 ∫2 ρ2
sin ϕ dρ dϕ dθ
2π π/6 1 ρ=csc ϕ
=∫ ∫0 [− ] sin ϕ dϕ dθ
0 ρ ρ=2
2π π/6 1
=∫ ∫ (− sin ϕ + ) sin ϕ dϕ dθ
0 0 2
2π π/6 1 − cos 2ϕ sin ϕ
=∫ ∫0 (− + ) dϕ dθ
0 2 2
2π ϕ sin 2ϕ cos ϕ ϕ=π/6
=∫ [− + − ] dθ
0 2 4 2 ϕ=0
√ √
2π π 3 3 1
=∫ (− + − + ) dθ
0 12 8 4 2
√
π 3
= π (1 − − )
6 4

117
41. Let D be the region in space bounded on the top by the sphere x2 + y 2 + z 2 = 2 and on the
bottom by the paraboloid z = x2 + y 2 . Express the volume V of D in terms of iterated integrals
in the (a) Cartesian, (b) cylindrical, and (c) spherical coordinates.

Solution: Note that the projection of the curve of intersection of the sphere and
the paraboloid is the unit circle x2 + y 2 = 1 in the xy-plane. This curve also bounds
the projection of the solid D to the xy-plane. Hence we have
√ √
1 1−x2 2−x2 −y 2
V =∫ ∫−√1−x2 ∫x2 +y2 dz dy dx
−1

and √
2π 1 2−r 2
V =∫ ∫0 ∫r2 r dz dr dθ .
0

Now observe that a ray starting at the origin leaves D through the sphere if the
angle it makes with the positive z-axis is less than π/4, whereas it does so through
the paraboloid if this angle is between π/4 and π/2. Hence:
√
2π π/4 2
V =∫ ∫0 ∫0 ρ2 sin ϕ dρ dϕ dθ
0
2π π/2 cos ϕ/ sin2 ϕ
+∫ ∫π/4 ∫0 ρ2 sin ϕ dρ dϕ dθ
0

118
42. Consider the iterated integral
π/2 1 2−r2
∫0 ∫0 ∫r dz dr dθ

in cylindrical coordinates.
a. Change the order of integration into dr dz dθ.
b. Express the integral in spherical coordinates with order of integration dϕ dρ dθ .

Solution: a. z = r is the cone z 2 = x2 + y 2 , and z = 2 − r2 is the paraboloid

z = 2 − x2 − y 2 . These surfaces intersect along a circle that is also the curve of
intersection of the cylinder r = 1 and the horizontal plane z = 1. Since 0 ≤ r ≤ 1 and
0 ≤ θ ≤ π/2, the integration region D is the region in the first octant bounded by
the paraboloid on the top and the cone at the bottom.

When we integrate with respect to r first, we will be moving along a curve on

119
which z and θ are constant. These two conditions describe a horizontal plane and a
plane containing the z-axis, respectively. Therefore the r-integration is along a ray
perpendicular to the z-axis like the blue ones in the figure. Such a ray starts on the
z-axis in D, and leaves D through the cone if 0 ≤ z ≤ 1, and through the paraboloid
if 1 ≤ z ≤ 2. Therefore,
π/2 1 2−r 2 π/2 1 z
∫0 ∫0 ∫r dz dr dθ = ∫ ∫0 ∫0 dr dz dθ
0
√
π/2 2 2−z
+∫ ∫1 ∫0 dr dz dθ
0
√
where we used the fact that z = 2 − r2 and r ≥ 0 Ô⇒ r = 2−z.

b. We have
π/2 1 2−r2
1
∫0 ∫0 ∫r dV dz dr dθ = ∭
D r
as dV = r dz dr dθ in cylindrical coordinates where D is the region described in part
(a).

Since we want to integrate with respect to ϕ first, we will be moving along the curves
on which ρ and θ are constant. These two conditions describe a sphere with center
at the origin and a half-plane whose spine is the z-axis, respectively. Therefore the
ϕ-integration takes place along a vertical semicircle subtended by a diameter along
the z-axis and with center at the origin like the green ones in the figure. Such a
semicircle starts on the positive√z-axis in D, and leave the√region of integration
intersecting the cone for 0 ≤ ρ ≤ 2 and the paraboloid for 2 ≤ ρ ≤ 2. The upper
half of the cone z = r has the equation ϕ = π/4 in spherical coordinates.
√ On the
other hand, z = 2 − r Ô⇒ ρ cos ϕ = 2 − (ρ sin ϕ) Ô⇒ cos ϕ = (1 + 4ρ − 7)/(2ρ) for
2 2 2

a point on the paraboloid as ϕ ≤ π/4 . Therefore,

1 1
∭D r dV = ∭D ρ sin ϕ dV
√
π/2 2 π/4
=∫ ∫0 ∫0 ρ dϕ dρ dθ
0
√
π/2 2 arccos((1+ 4ρ2 −7)/(2ρ))
+∫ ∫√2 ∫0 ρ dϕ dρ dθ
0

where we substituted dV = ρ2 sin ϕ dϕ dρ dθ for the volume element in spherical

coordinates.

120
43. The region R bounded by the curve z 2 = y 2 − y 4 in the right half of the yz-plane is rotated
about the z-axis to obtain a solid D in the xyz-space. Express the volume V of D in terms of
iterated integrals in (a) the Cartesian, (b) the cylindrical and (c) the spherical coordinates.

Solution: Since the surface bounding the solid D is obtained by revolving a curve
about the z-axis, its equation in the cylindrical coordinates will not depend on θ.
Since r = y in the yz-plane and the curve has the equation z 2 = y 2 − y 4 , we conclude
that z 2 = r2 − r4 is the equation of the surface in the cylindrical coordinates. From
this √
2π 1 r 2 −r 4
V =∫ ∫0 ∫−√r2 −r4 r dz dr dθ
0

follows as the projection of D to the xy-plane is the unit disk.

The equation z 2 = r2 − r4 in the cylindrical coordinates transforms to z 2 = (x2 +

y 2 ) − (x2 + y 2 )2 in the Cartesian coordinates, and once again using the fact that the
projection of D to the xy-plane is the unit disk, one now obtains:
√ √
1 1−x2 (x2 +y 2 )−(x2 +y 2 )2
V =∫ ∫ √ ∫−√(x2 +y2 )−(x2 +y2 )2 dz dy dx
−1 − 1−x2

The spherical coordinates require a little bit more work. First note that the equation
z 2 = r2 −√
r4 in the cylindrical coordinates now gives (ρ cos ϕ)2 = (ρ sin ϕ)2 − (ρ sin ϕ)4 ,
or ρ = sin2 ϕ − cos2 ϕ/ sin2 ϕ, in the spherical coordinates. Next note that for
sin2 ϕ − cos2 ϕ = − cos 2ϕ to be nonnegative, ϕ must be between π/4 and 3π/4 in the
interval [0, π]. Therefore:
√
2π 3π/4 sin2 ϕ−cos2 ϕ/ sin2 ϕ
V =∫ ∫π/4 ∫0 ρ2 sin ϕ dρ dϕ dθ
0

Remark: Compare this example with Example 52 in Part 1.

121
 0 2y+3 x+y
44. Evaluate the iterated integral ∫ ∫ ex−2y dx dy .
−1 −y (x − 2y)2

Solution:

∂(x, y) ∂(u, v)
45. Let (u, v) and (x, y) be two coordinate systems. Show that ⋅ = 1.
∂(u, v) ∂(x, y)

Solution: By definition,

∂(x, y) xu xv
=∣ ∣ = xu yv − xv yu
∂(u, v) yu yv

and
∂(u, v) ux uy
=∣ ∣ = ux vy − uy v x .
∂(x, y) vx vy

122
 Therefore,

∂(x, y) ∂(u, v)
⋅ = (xu yv − xv yu )(ux vy − uy vx )
∂(u, v) ∂(x, y)
= x u ux y v v y + x v v x y u u y − x u u y y v v x − x v ux y u v y
= x u ux y v v y + x v v x y u u y − x u u y y v v x − x v ux y u v y
+ x v v x y v v y + x u ux y u uy − x v v x y v v y − x u ux y u uy
= (xu ux + xv vx ) (yu uy + yv vy )
− (xu uy + xv vy ) (yu ux + yv vx )
= xx yy − xy yx
=1⋅1−0⋅0
=1

2 /y √
46. Evaluate the double integral ∬ ex dA where R = {(x, y) ∶ x2 ≤ y ≤ x }.
R

Solution: We change the coordinates to u = y 2 /x, v = x2 /y. In this coordinate

system the region of integration becomes G = {(u, v) ∶ 0 ≤ u ≤ 1 and 0 ≤ v ≤ 1} .

We have
∂(u, v) ux uy −y 2 /x2 2y/x ∂(x, y) 1 1
=∣ ∣=∣ ∣ = −3 Ô⇒ = =− ,
∂(x, y) vx vy 2x/y −x /y2 2
∂(u, v) ∂(u, v) 3
∂(x, y)
and the change of variables formula gives

x2 /y ∂(x, y) 1 1 1 e−1
∬R e dx dy = ∬ ev ∣ ∣ du dv = ∫ ∫ ev ∣ − ∣ dv du = .
G ∂(u, v) 0 0 3 3

123
Remark: We can multiply differentials. The rules are du du = 0 = dv dv and dv du = −du dv in
any uv-coordinate system. For instance, when changing from Cartesian to polar coordinates,
we have

dx dy = d(r cos θ) d(r sin θ)

= (cos θ dr − r sin θ dθ)(sin θ dr + r cos θ dθ)
= −r sin2 θ dθ dr + r cos2 θ dr dθ
= r(sin2 θ + cos2 θ) dr dθ
= r dr dθ

and in Example 46 we have

du dv = d(y 2 /x) d(x2 /y)

= (2y/x dy − y 2 /x2 dx)(2x/y dx − x2 /y 2 dy)
= 4 dy dx + dx dy
= −4 dx dy + dx dy
= −3 dx dy ,

1
hence dx dy = − du dv . This can be used to keep track of how the area element changes under
3
a coordinate change, but note that the sign of the factor in front must be corrected by hand so
that it is positive on the region of integration.

u v
47. Consider the transformation T ∶ x = ,y= . Show that there is a constant
u+v+1 u+v+1
C such that the inequality
∂(x, y)
∬G ∣ ∂(u, v) ∣ du dv ≤ C

holds for all regions G contained in the first quadrant of the uv-plane.

u v u+v
Solution: Observe that x = ≥ 0, y = ≥ 0 and x + y = ≤1
u+v+1 u+v+1 u+v+1
x y
for u ≥ 0 and v ≥ 0. We also have u = and v = . Therefore T maps
1−x−y 1−x−y
the first quadrant of the uv-plane into the triangle R = {(x, y) ∶ x+y ≤ 1, x ≥ 0, y ≥ 0}
in a one-to-one manner.

124
 Hence:
∂(x, y) 1
∬G ∣ ∂(u, v) ∣ du dv = ∬T (G) dx dy = (Area of T (G)) ≤ (Area of R) = 2

Remark: One can also do a straightforward computation.

∂(x, y) xu xv (v + 1)/(u + v + 1)2 −u/(u + v + 1)2 1

=∣ ∣=∣ ∣=
∂(u, v) yu yv −v/(u + v + 1) 2 (u + 1)/(u + v + 1)2
(u + v + 1)3

and hence
∂(x, y) ∞ ∞ du dv 1 ∞ dv 1
∣
∬G ∂(u, v) ∣ du dv ≤ ∫0 ∫0 (u + v + 1)3 2 ∫0 (v + 1)2 = 2 .
=

∂(x, y, z)
48. Compute the Jacobian where (ρ, ϕ, θ) are the spherical coordinates and (x, y, z)
∂(ρ, ϕ, θ)
are the Cartesian coordinates.

Solution: We have x = ρ sin ϕ cos θ, y = ρ sin ϕ sin θ, z = ρ cos ϕ. Therefore

RRx x x RR
∂(x, y, z) RRRR ρ ϕ θ RRRR
= RR yρ yϕ yθ RRR
∂(ρ, ϕ, θ) RRRR R
RR zρ zϕ zθ RRRR
RRRsin ϕ cos θ ρ cos ϕ cos θ −ρ sin ϕ sin θRRR
RR RR
= RRRR sin ϕ sin θ ρ cos ϕ sin θ ρ sin ϕ cos θ RRRR
RRR RRR
RR cos ϕ −ρ sin ϕ 0 RR
sin ϕ sin θ ρ cos ϕ sin θ
= (−ρ sin ϕ sin θ) ∣ ∣
cos ϕ −ρ sin ϕ
sin ϕ cos θ ρ cos ϕ cos θ
− (ρ sin ϕ cos θ) ∣ ∣
cos ϕ −ρ sin ϕ
sin ϕ cos θ ρ cos ϕ cos θ
+0⋅∣ ∣
sin ϕ sin θ ρ cos ϕ sin θ
= (−ρ sin ϕ sin θ)(−ρ sin θ)(sin2 ϕ + cos2 ϕ)
− (ρ sin ϕ cos θ)(−ρ cos θ)(sin2 ϕ + cos2 ϕ)
+0
= ρ sin ϕ (sin θ + cos θ)
2 2 2

= ρ2 sin ϕ

where we used the cofactor expansion with respect to the third column.

125
 Part 3: Sequences and Series

1. Let a1 = 1, a2 = a3 = 2, a4 = a5 = a6 = 3, a7 = a8 = a9 = a10 = 4, and so on. That is,

an ∶ 1, 2, 2, 3, 3, 3, 4, 4, 4, 4, 5, 5, 5, 5, 5, 6, . . . . What is a2021 ?

Solution: an = k if 1 + 2 + ⋯ + (k − 1) < n ≤ 1 + 2 + ⋯ + k. In other words, an = k if

k(k − 1)/2 < n ≤ k(k + 1)/2. Since for k = 64, k(k − 1)/2 = 2016 and k(k + 1)/2 = 2080,
we have a2021 = 64.

√
8n + 1 − 1
Remark: A more explicit formula an = ⌈ ⌉ can be obtained.
2

2. Curve 1 is an equilateral triangle with unit sides. For n ≥ 2, Curve n is obtained from Curve n−1
by replacing the middle third of every edge with the other two sides of the outward pointing
equilateral triangle sitting on it. Let Ln be the length of Curve n and An be the area of the
region enclosed by Curve n . Find Ln , An , lim Ln and lim An .
n→∞ n→∞

Solution: Let en and dn denote the number of edges and the length of each edge
of Curve n , respectively. Since en = 4en−1 and dn = dn−1 /3 for n ≥ 2, and e1 = 3
and d1 = 1, we find that en = 3 ⋅ 4n−1 and dn = 1/3n−1 for n ≥ 1. It follows that
Ln = en dn = 3 ⋅ (4/3)n−1 for n ≥ 1, and lim Ln = lim 3 ⋅ (4/3)n−1 = ∞.
n→∞ n→∞

On the other hand, the nth region is obtained by adjoining en−1 equilateral triangles
of side length dn−1 /3 to the (n − 1)st region. Therefore,
√ √
3 (dn−1 /3)2 3 1 2
An = An−1 + en−1 ⋅ = An−1 + 3 ⋅ 4n−2 ⋅ ( )
4 4 3n−1
and then
1 4 n−2 4 n−3
An = √ (( ) + ( ) + ⋯ + 1) + A1
4 3 9 9
√
3
for n ≥ 2. Since A1 = , this gives
4
√ √
3 3 4 n−1 3
An = (1 − ( ) ) +
20 9 4

for n ≥ 1. We obtain √ √ √
3 3 3 2 3
lim An = + = .
n→∞ 20 4 5

126
 Curve 1 : Curve 2 :

Curve 3 : Curve 4 :

Remark: A magician puts 1 ball into an empty box at t = 0 sec, takes out 1 ball at t = 1/2 sec,
puts 2 balls at t = 3/4 sec, takes out 1 ball at t = 7/8 sec, puts 3 balls at t = 15/16 sec, takes out
1 ball at t = 31/32 sec, and so on. Then she challenges you to guess how many balls there are
in the box at t = 1 sec.
You ask for the advice of your friends.

Friend1 says: “The net result of 2k + 1st and 2k + 2nd steps for k ≥ 1 is to put at
least one more ball into the box. There are infinitely many balls in the box at t = 1
sec.”

Friend2 says: “Imagine that the balls are labeled as ball1, ball2, ball3, and so on
with invisible ink which only the magician can see. Then she puts ball1 at t = 0
sec, takes out ball1 at t = 1/2 sec, puts ball2 and ball3 at t = 3/4 sec, takes out
ball2 at t = 7/8 sec, puts ball4, ball5, and ball6 at t = 15/16 sec, takes out ball3
at t = 31/32 sec, and so on. For any given k, ballk is not in the box at time t = 1
sec, because it was taken out at time t = 1 − 1/22k−1 sec. There are no balls in the
box at time t = 1 sec.”

What do you think?

127
 1
3. Let the sequence {an } be defined by a1 = 1 and an = for n ≥ 1 . Show that the
1 + an−1
sequence converges and find its limit.

Solution: Observe that if 0 < an < an+1 , then 0 < 1 + an < 1 + an+1 and an+1 =
1/(1 + an ) > 1/(1 + an+1 ) = an+2 . Similarly, if an > an+1 > 0, then 1 + an > 1 + an+1 > 0
and an+1 = 1/(1 + an ) < 1/(1 + an+1 ) = an+2 . Since a1 = 1 > 1/2 = a2 , it follows that
1
1 = a1 > a3 > a5 > ⋯ > a6 > a4 > a2 = .
2
Therefore, the sequence {a2n }∞
n=1 is increasing and bounded from above by 1, and
∞
the sequence {a2n−1 }n=1 is decreasing and bounded from below by 1/2. By the
Monotonic Sequence Theorem, then both of these sequences are convergent.

Let lim a2n−1 = L and lim a2n = M . Taking the limit of a2n+1 = 1/(1 + a2n ) as n → ∞
n→∞ n→∞
we obtain L = 1/(1 + M ), and taking the limit of a2n = 1/(1 + a2n−1 ) as n → ∞ we
obtain M = 1/(1 + L). From M + M L = 1 and LM + L = 1 it follows that M = L.
Therefore the sequence {an }∞ n=1 converges to L = M .
√ √
Finally, L2 + L − 1 = 0 gives L = ( 5 − 1)/2
√ or L = −( 5 + 1)/2. Since 0 < an for all
n ≥ 1, we must have 0 ≤ L . Hence L = ( 5 − 1)/2 is the limit and
√
5−1
lim an = .
n→∞ 2

Remark: This result is sometimes expressed symbolically in the form:

√
5−1 1
=
2 1
1+
1
1+
1
1+
1
1+
1
1+
1+⋯

Remark: Note that xn = Fn /Fn+1 for n ≥ √1 where Fn ∶ 1, 1, 2, 3, 5, 8, . . . is the Fibonacci

sequence. So we showed that lim Fn /Fn+1 = ( 5 − 1)/2 .
n→∞

Remark: It can be showed with√a little bit more work√that any sequence satisfying the given
recursion relation converges to (√5 − 1)/2 if a1 =/ −(1 + 5)/2 and a1 =/ −Fn+1 /Fn for any n ≥ 1 .
On the other hand, if a1 = −(1 + 5)/2 , then the sequence is constant; and if a1 = −Fn+1 /Fn for
some n ≥ 1, then an = −1 and an+1 is undefined.

128
4. A sequence {an }∞
n=1 satisfies the recursion relation

10
an = max{an−1 , an−2 } − min{an−1 , an−2 }
9
for n ≥ 3.
a. Determine whether the sequence {an }∞
n=1 converges or diverges if a1 = 1 and a2 = 1/9.

b. Determine whether the sequence {an }∞

n=1 converges or diverges if a1 = 1 and a2 = 2/3.

Solution: a. We observe that:

10 10 1 1 10 1
a3 = max{a2 , a1 } − min{a2 , a1 } = max { , 1} − min { , 1} = ⋅1− =1
9 9 9 9 9 9
10 10 1 1 10 1
a4 = max{a3 , a2 } − min{a3 , a2 } = max {1, } − min {1, } = ⋅1− =1
9 9 9 9 9 9
10 10 10 1
a5 = max{a4 , a3 } − min{a4 , a3 } = max{1, 1} − min{1, 1} = ⋅1−1=
9 9 9 9
Since each terms depends only on the previous two terms, the pattern 1, 1/9, 1, 1,
1/9, 1, 1, 1/9, . . . repeats. Hence the sequence {an }∞
n=1 diverges.

b. We observe that

● a1 = 1 = (2/3)0 = (2/3)1−1 and a2 = 2/3 = (2/3)1 = (2/3)2−1 , and

● if an−2 = (2/3)n−3 and an−1 = (2/3)n−2 , then

10
an = max{an−1 , an−2 } − min{an−1 , an−2 }
9
10
= max{(2/3)n−2 , (2/3)n−3 } − min{(2/3)n−2 , (2/3)n−3 }
9
n−3 n−2
10 2 2
= ⋅( ) −( )
9 3 3
n−3
10 2 2
=( − )⋅( )
9 3 3
n−3 n−1
4 2 2
= ⋅( ) =( ) .
9 3 3

Hence an = (2/3)n−1 for all n ≥ 1. In particular, lim an = lim (2/3)n−1 = 0. Therefore

n→∞ n→∞
the sequence {an }∞
n=1 converges.

Remark: It can be shown that, depending on the values of a1 and a2 , the sequence {an }∞ n=1
either converges to 0 or is periodic after some index with the pattern c, c, c/9, where c is a
positive real number.

129
5. Let a0 be a rational number greater than 1, and consider the sequence {an }∞
n=0 satisfying

pn + 1
an+1 =
qn + 1

if an = pn /qn where pn and qn are positive integers with no common divisor greater than 1, for
n ≥ 0.
Find the limit lim (an )n .
n→∞

Solution: Let dn be the greatest common divisor of pn and qn . Then

pn + 1 qn + 1 p n − qn
pn+1 = , qn+1 = , and pn+1 − qn+1 =
dn dn dn
for n ≥ 0.

Hence {pn −qn }∞n=0 is a nonincreasing sequence of positive integers. Therefore, it must
be constant from some index on. That is, there are integers N ≥ 0 and c > 0 such
that pn − qn = c for all n ≥ N . This implies that dn = 1 for all n ≥ N . In particular,
qn+1 = qn + 1 for all n ≥ N , and hence qn = qN + n − N and pn = qN + n − N + c for all
n ≥ N.

Observe that c must divide one of the consecutive integers qn = qN + n − N for

N ≤ n ≤ N + c − 1. This will make pn = qn + c also divisible by c. Therefore, as dn = 1
for all n ≥ N , c must be 1.

Hence,
n
qN − N + 1
n (1 + )
qN + n − N + 1 n eqN −N +1
lim (an )n = lim ( ) = lim n = =e
n→∞ n→∞ qN + n − N n→∞ qN − N eqN −N
(1 + )
n
n
x
where we used the fact that lim (1 + ) = ex for all x.
n→∞ n

130
 xn−1 1
6. Let the sequence {xn } be defined by x0 = 2 and xn = + for n ≥ 1 .
2 xn−1
a. Find the limit of this sequence assuming it exists.
b. Show that the limit exists.

Solution: a. We assume that the limit L = lim xn exists. Then

xn−1 1 x2
xn = + for n ≥ 1 Ô⇒ xn xn−1 = n−1 + 1 for n ≥ 1
2 xn−1 2
2
x L2
Ô⇒ lim(xn xn−1 ) = lim n−1 + 1 Ô⇒ L2 = + 1 Ô⇒ L2 = 2
2 2
√ √
as lim xn−1 = lim xn = L . Therefore L = 2 or L = − 2.

x0 = 2 > 0 and if xn > 0 then xn+1 = xn /2 + 1/xn > 0 + 0 = 0. Hence by induction

√ xn > 0
for all n ≥ 0. It follows that L = lim xn ≥ 0. We conclude that the limit is 2 .

b. In part (a) we proved that xn > 0 for all n ≥ 0. Therefore {xn } is bounded from
below.
√
Now we will show by induction on n that 2 < xn+1 < xn for all n ≥ 0.
√
• Let n = 0. Since x0 = 2, x1 = 3/2, and x21 = 9/4 > 2, we have 2 < x1 < x0 .
√
• Let n > 0 and assume that 2 < xn+1 < xn . Then

xn+1 1 x2 − 2
xn+1 − xn+2 = − = n+1 >0
2 xn+1 2xn+1
and hence xn+2 < xn+1 . On the other hand,
2 2
xn+1 1 x2 − 2
x2n+2 −2=( + ) − 2 = ( n+1 ) >0
2 xn+1 2xn+1
√
and since xn+2 > 0, we have xn+2 > 2.

Therefore the sequence is decreasing.

Since the sequence is bounded from below and decreasing, we conclude that it
converges by the Monotonic Sequence Theorem.

131
7. We have a 1/n×1/n square for each positive integer n. For each of (a-c), determine whether
it is possible or not to place these squares in the xy-plane in such a way that they completely
cover the given set.

1
1/2
1/3
1/2 1/3
1/4
1/4
...
1

a. The entire xy-plane

b. The line defined by the equation y = x
c. The region between the graph of y = e−x and the x-axis for x ≥ 0

∞
1
Solution: a. The sum of the areas of the squares is ∑ 2
. This series is convergent
n=1 n
as it is the p-series with p = 2 > 1. (In fact, its sum is π 2 /6.) On the other hand,
the entire plane has infinite area. Hence it is not possible to cover the entire plane
using these squares.

b. We place the odd-numbered squares along the half of the line lying in the first
quadrant and the even-numbered √ ones along the half lying√in the third
√ quadrant
∞
2 1 ∞ 1 ∞
2 ∞
2
as shown in the figure. As ∑ = √ ∑ = ∞ and ∑ >∑ = ∞, the
k=1 2k 2 k=1 k k=1 2k − 1 k=1 2k
entire line is covered by the squares.

c. We place the squares along the positive x-axis as shown in the figure. Then the
∞
1
squares extend along the entire positive axis as ∑ = ∞. On the other hand, by
n=1 n

132
 the Integral Test Inequality we have
1 1 1 n+1 dx
1+ + +⋯+ >∫ = ln(n + 1)
2 3 n 1 x
for all n ≥ 1. (This inequality can be obtained by comparing the area covered by
rectangles erected on the unit intervals along the interval [1, n + 1] with upper left
corners on the graph of y = 1/x and the area under the graph on this interval.)
Hence
1
e−x ≤ e−(1+1/2+1/3+⋯+1/n) ≤
n+1
n n+1
1 1
for ∑ ≤ x ≤ ∑ for all n ≥ 0. In other words, the top side of the nth rectangle
k=1 k k=1 k
never lies lower than the graph of y = e−x . Therefore we can cover the region between
the graph of y = e−x and the positive x-axis with the squares.

8. For each of the series in (a-d), determine whether there exists a positive integer n such that
the nth partial sum sn of the series satisfies the condition 2014 ≤ sn ≤ 2015.
∞ ∞ ∞ n ∞
1 2999 1
a. ∑ n b. ∑ 5 n
c. ∑ ( ) d. ∑
n=1 2 n=1 n=1 3000 n=1 n

Solution: Note that all terms of these series are positive, and therefore, their partial
sums form increasing sequences. This will be used repeatedly in the following.

a. We have n ∞
1 1 1/2
sn = ∑ < ∑ = = 1 < 2014
k=1 2
k
n=1 2
n 1 − 1/2
for all n ≥ 1, where we used the geometric series sum formula, and hence there are
no partial sums lying between 2014 and 2015.

b. This time we have

n 4
sn = ∑ 5k ≤ ∑ 5k = 5 + 25 + 125 + 625 = 780 < 2014
k=1 k=1

133
 for n ≤ 4, and
n 5
sn = ∑ 5k ≥ ∑ 5k = 5 + 25 + 125 + 625 + 3125 = 3905 > 2015
k=1 k=1

for n ≥ 5. Therefore no sn lies between 2014 and 2015.

c. Every term of this series is between 0 and 1, hence its partial sums increase in
steps smaller than 1 starting with s1 = 2999/3000 < 2014. Moreover,
∞ n
2999 2999/3000
∑( ) = = 2999 > 2015
n=1 3000 1 − 2999/3000

by the geometric series sum formula. Hence there is at least one partial sum sn
satisfying the condition 2014 ≤ sn ≤ 2015.

d. The same reasoning as in part (c) works, this time with the observation that
∞
1
∑ = ∞ > 2015 ,
n=1 n

to give the existence of at least one partial sum sn satisfying the condition 2014 ≤
sn ≤ 2015.

9. Determine whether each of the following series converges or diverges.

∞ ∞ ∞
n ln n π
a. ∑ (21/n − 21/(n+1) ) b. ∑ n
c. ∑ (−1)n+1 cos ( )
n=1 n=2 3 n=1 n
∞ ∞ ∞ n
nn 1 5 −2 n
d. ∑ e. ∑ √ f. ∑ n
n=1 n + n=1 7 − 6
n
n=1 n! n sin n
∞ ∞
1 sin n
g. ∑ h. ∑ 2
n=3 n ln n ln(ln n) n=1 n

n n n
Solution: a. We have sn = ∑ (21/k − 21/(k+1) ) = ∑ 21/k − ∑ 21/(k+1) = 2 − 21/(n+1) .
k=1 k=1 k=1
∞
Hence lim sn = 1 . Therefore the series converges and ∑ (21/n − 21/(n+1) ) = 1 .
n→∞
n=1

b. We have an = n ln n/3n and

∣an+1 ∣ (n + 1) ln(n + 1)/3n+1 1 n+1 ln(n + 1)
ρ = lim = lim = ⋅ lim ⋅ lim
n→∞ ∣an ∣ n→∞ n ln n/3 n 3 n→∞ n n→∞ ln n
1 ln(x + 1) L’H
↓ 1 1/(x + 1) 1 1
= ⋅ 1 ⋅ lim = ⋅ lim = ⋅1= .
3 x→∞ ln x 3 x→∞ 1/x 3 3
∞
n ln n
Since ρ = 1/3 < 1 , the series ∑ n
converges by the Ratio Test.
n=2 3

134
c. Since an = (−1)n+1 cos(π/n), we have ∣an ∣ = cos(π/n) for n ≥ 2, and
lim ∣an ∣ = lim cos(π/n) = cos 0 = 1 . Therefore lim an =/ 0 , and the series diverges
n→∞ n→∞ n→∞
by the nth Term Test.

d. We have an = nn /n! and

n
∣an+1 ∣ (n + 1)n+1 /(n + 1)! 1
ρ = lim = lim = lim (1 + ) = e .
n→∞ ∣an ∣ n→∞ n /n!
n n→∞ n
∞
nn
Since ρ = e > 1 , the series ∑ diverges by the Ratio Test.
n=1 n!

e. We observe that
√
1/(n + n sin n) 1 1
L = lim = lim = =1.
n→∞ 1/n n→∞ sin n 1 + 0
1+ √
n
∞
1
Since 0 < L < ∞ and the harmonic series ∑ diverges, we conclude that the series
n=1 n
∞
1
∑ √ diverges by the Limit Comparison Test.
n=1 n + n sin n

∞
f. The geometric series ∑ (5/7)n converges as r = 5/7 Ô⇒ ∣r∣ = 5/7 < 1. We also
n=1
(5n − 2n )/(7n − 6n ) 1 − (2/5)n 1 − 0
have L = lim = lim = = 1 . Since 0 < L < ∞ , the
n→∞ 5n /7n n→∞ 1 − (6/7)n 1−0
∞ n
5 − 2n
series ∑ n converges by the Limit Comparison Test.
n=1 7 − 6
n

g. Since ln x is an increasing function on (0, ∞) , ln(ln x) is an increasing function

on (1, ∞). Both ln x and ln(ln x) are positive on [3, ∞) . Therefore x ln x ln(ln x) is
positive and increasing on [3, ∞) as it is the product of three positive and increasing
functions. This in turn implies that 1/(x ln x ln(ln x)) is a positive and decreasing
function on [3, ∞) . Since it is also continuous, we can apply the Integral Test. The
improper integral
∞ dx ∞ du
=
∫3 x ln x ln(ln x) ∫ln(ln 3) u

diverges, where we used the change of variable u = ln(lnx), du = dx/(x ln x) . Hence

∞
1
the series ∑ diverges.
n=3 n ln n ln(ln n)

∞
1 + sin n
h. Consider the series ∑ . Since 0 ≤ 1 + sin n ≤ 2 for n ≥ 1 , we have
n=1 n2
∞
1 + sin n 2 1
0≤ 2
≤ 2
for n ≥ 1 . The p-series ∑ 2
converges as p = 2 > 1 . Therefore
n n n=1 n

135
 ∞ ∞ ∞
2 1 1 + sin n
the series ∑ 2
= 2 ∑ 2
converges. Then the series ∑ converges by the
n=1 n n=1 n n=1 n2
∞ ∞
1 + sin n 1
Direct Comparison Test. Since both ∑ 2
and ∑ 2
converge, the series
n=1 n n=1 n
∞
sin n
∑ 2 , which is their difference, also converges.
n=1 n

Remark: Other tests can be used too. Here are some examples:
∞
1
In part (a), the Limit Comparison Test with the series ∑ 2
also works.
n=1 n
∞
1
In part (b), the Limit Comparison Test with the geometric series ∑ n
also works.
n=1 2

1/n
(n!) 1
In part (d), the Root Test also works where we use the fact that lim = .
n→∞ n e

136
Remark: The following was a bonus problem on Moodle in Spring 2010 Math 102 course.

Problem K: In class we showed that 0.12 = 0.121212. . . = 12/99 as an application

of the geometric series.
We learn the following trick in elementary school: If x = 0.121212 . . . , then 100x =
12.121212 . . . ; their difference gives 99x = 12 and therefore x = 12/99 . Of course
when we see this in elementary school, no one talks about convergence.
That is what we are going to do in this problem. We will forget about convergence
as we know it, and take a trip to Tersonia.
In Tersonia after they teach the students about integers and rational numbers,
they come to the decimal representations of numbers. As you know our decimal
expansions have the form

±dn dn−1 . . . d2 d1 d0 .d−1 d−2 . . .

where each di is in {0, 1, . . . , 9}. We can have infinitely many nonzero digits after
the decimal point, but we must have only finitely many nonzero digits before the
decimal point. In Tersonia they do just the opposite. Their decimal expansions
have the form
. . . t3 t2 t1 t0 .t−1 t−2 . . . t−n
where each ti is in {0, 1, . . . , 9}. Note that there is no minus sign. They can have
infinitely many nonzero digits before the decimal point, but they can only have
finitely many digits after the decimal point.
Take a few minutes to convince yourself that Tersonians can add and multiply
their decimal expansions just like we do.
Why no minus sign? Well, because Tersonians don’t need it. “Negative” numbers
are already there. For instance, consider the number y = 12.0 = . . . 121212.0 .
Then 100y = . . . 121200.0 and −99y = 12 . Therefore y = −12/99 . So in fact 12.0
is a “negative” number.
Here are some problems from Tersonian Elementary School Mathematics Book :
1 1 1
a. =? b. =? c. =? d. −1 = ?
2 3 7

e. Find two nonzero numbers A and B such that A B = 0.

Part (e) was later turned into a programming challenge. A Java applet that com-
putes the last n digits of A and B when their last digits are given can be found at
http://www.fen.bilkent.edu.tr/˜otekman/calc2/ters.html .

137
10. Determine the smallest of the real numbers A, B, C, D, E where :
∞ ∞ ∞ ∞ ∞
(−1)n 1 (−1)n+1 n (−1)n
A=∑ B=∑ C=∑ D=∑ E=∑
n=0 3 (2n + 1)
n n n n
n=0 2 n=1 n2 n=1 n! n=1 3

Solution: Firstly we have

∞ n
(−1)n ∞ −1 1 2
A=∑ = ∑( ) = =
n=0 2
n
n=0 2 1 − (−1/2) 3

by the geometric series sum formula. Next we observe that

∞ ∞
1 1 1 1 2 n 1 2 3 2
B=∑ > + + = and D = ∑ > + 2+ 3=
n=1 n2
n 2 2⋅2 2 3⋅2 3 3 n=1 3
n 3 3 3 3

as all the remaining terms of these series are positive. Finally,

∞ ∞
(−1)n 1 8 2 (−1)n+1 1 1 2
E=∑ > 1 − = > and C = ∑ < 1 − + =
n=0 3 (2n + 1) 3⋅3 9 3
n n! 2! 3! 3
n=1

by the Alternating Series Estimate, which can be applied to these series as {1/(3n ⋅
(2n + 1))}∞ ∞
n=0 and {1/n!}n=1 are decreasing sequences with limit 0 .

Hence C is the smallest.

√
Remark: One can also observe that B = ln 2, C = 1 − 1/e, D = 3/4, and E = π/(2 3) as
∞
xn
ln(1 + x) = ∑ (−1)n+1 for − 1 < x ≤ 1 ,
n=1 n

∞
xn
ex = ∑ for all x ,
n=0 n!

x d 1 d ∞ n ∞
= x = x ∑ x = ∑ n xn for ∣x∣ < 1 , and
(1 − x)2 dx 1 − x dx n=0 n=1

∞
x2n+1
arctan x = ∑ (−1)n for ∣x∣ ≤ 1 .
n=0 2n + 1

Therefore, C < A < B < D < E.

138
 ∞
xn
11. Consider f (x) = ∑ .
n=0 5 (n + 1)
n 2

4 3
a. Show that < f (3) < .
3 2
3 4
b. Show that < f (−3) < .
4 5

Solution: a. We have:
∞
3n 3 32 343 4
f (3) = ∑ > 1 + + = >
n=0 5 (n + 1) 5 ⋅ 2 5 ⋅ 5 250 3
n 2 2

and:
∞ ∞
3n 3 3n
f (3) = ∑ = 1 + + ∑
n=0 5 (n + 1) 5 ⋅ 2 n=2 5n (n2 + 1)
n 2

13 1 ∞ 3n 13 1 (3/5)2 37 3
< + ∑ = + ⋅ = <
10 5 n=2 5n 10 5 1 − 3/5 25 2

b. We have
∞
3n 3 32 193 4
f (−3) = ∑ (−1)n < 1 − + = <
n=0 5 (n + 1)
n 2 5 ⋅ 2 5 ⋅ 5 250 5
2

and ∞
3n 3 32 33 469 3
f (−3) = ∑ (−1)n > 1 − + − = >
n=0 5 (n + 1)
n 2 5 ⋅ 2 5 ⋅ 5 5 ⋅ 10 625 4
2 3

n ∞
3 1
by the Alternate Series Estimate as {( ) ⋅ 2 } is a decreasing sequence and
5 n + 1 n=0
n
3 1
lim ( ) ⋅ 2 = 0.
n→∞ 5 n +1

∞
(−4)n
12. Determine whether the sum of the series ∑ is positive or negative.
n=0 n!(n + 1)!

4n
Solution: Let bn = for n ≥ 0. Then:
n!(n + 1)!
• bn > 0 for all n ≥ 0.
4n 4n 1
• lim bn = lim = lim ⋅ lim = 0 as both limits are 0, the first
n→∞ n→∞ n!(n + 1)! n→∞ n! n→∞ (n + 1)!
one being one of the “Useful Limits”.
4n 4n+1
• bn ≥ bn+1 ⇐⇒ ≥ ⇐⇒ (n + 2)(n + 1) ≥ 4 ⇐⇒ n ≥ 1.
n!(n + 1)! (n + 1)!(n + 2)!

139
 Therefore the series satisfies the conditions of the Alternating Series Test. In
particular, it converges. Moreover,
∞
(−4)n 4 42 43 44 4 4 4
S=∑ =1− + − + −⋯=1−2+ − + −⋯
n=0 n!(n + 1)! 1!2! 2!3! 3!4! 4!5! 3 9 45
´¹¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¸¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¶ ¯
s3 =−1/9 b4

and, by the Alternating Series Estimate, ∣S − s3 ∣ < b4 . In other words, ∣S − (−1/9)∣ <
1/45, and S is negative.

∞ ∞
(−4)n 1 (−1)n (x/2)2n+1
Remark: ∑ is J1 (4), where J1 (x) = ∑ is the Bessel function of
n=0 n!(n + 1)! 2 n=0 n!(n + 1)!
the first kind of order 1, and its value is approximately −0.03302166401.

∞ n
1 1
13. Show that the series ∑ ( + ) converges and its sum is less than 4.
n=1 2 n

Solution: Since L = lim ∣an ∣1/n = lim ((1/2 + 1/n)n )1/n = lim (1/2 + 1/n) = 1/2 < 1,
n→∞ n→∞ n→∞
the series converges by the nth Root Test. Moreover:
∞ n ∞ 3 4 n
1 1 3 5 3 1 1
∑( + ) = +1+( ) +( ) + ∑( + )
n=1 2 n 2 6 4 n=5 2 n
∞ n
3 125 81 7 3 7 1 (7/10)5
< +1+ + + ∑( ) < +1+ + +
2 216 256 n=5 10 2 12 3 1 − (7/10)
5
41 7 10 41 7 ⋅ 49 ⋅ 49 41 7
= +( ) = + < + =4
12 10 3 12 3 ⋅ 100 ⋅ 100 12 12

Remark: An alternative approach is to observe that ln(1 + x) < x for x > 0 Ô⇒ (1 + 2/n)n < e2
for all positive integers n, and use the bound
∞ ∞ n ∞ n n n
1 1 2 1 1 1 9
∑ ( + ) = ∑ (1 + ) ( ) < e2 ∑ ( ) = e2 4 <
n=5 2 n n=5 n 2 n=5 2 2 16
for the tail.

140
 ∞
xn
14. Find the radius of convergence of the power series ∑ (−1)n .
n=0 (2n + 1)(n2 + 1)

Solution: We will give two different solutions.

Solution 1 : Here we will use the Ratio Test.

xn
We have an = (−1)n and
(2n + 1)(n2 + 1)

∣an+1 ∣
ρ = lim
n→∞ ∣an ∣
xn+1
∣(−1)n+1 ∣
(2n+1 + 1)((n + 1)2 + 1)
= lim
n→∞ xn
∣(−1)n ∣
(2n + 1)(n2 + 1)
∣x∣n+1
(2n+1 + 1)((n + 1)2 + 1)
= lim
n→∞ ∣x∣n
(2n + 1)(n2 + 1)
1 + 2−n n2 + 1
= lim ( ⋅ )∣x∣
n→∞ 2 + 2−n (n + 1)2 + 1

∣x∣
= .
2
If ∣x∣ < 2 , then ρ = ∣x∣/2 < 1 and the power series converges absolutely by the Ratio
Test. On the other hand, if ∣x∣ > 2 , then ρ = ∣x∣/2 > 1 and the power series diverges
by the Ratio Test.

It follows by the definition of the radius of convergence that R = 2 .

Solution 3 : Here we will use the radius of convergence formulas.

(−1)n
We have cn = , and the radius of convergence formula gives
(2n + 1)(n2 + 1)

1 ∣cn+1 ∣ 1/((2n+1 + 1)((n + 1)2 + 1))

= lim = lim
R n→∞ ∣cn ∣ n→∞ 1/((2n + 1)(n2 + 1))
1 + 2−n n2 + 1 1
= lim ( ⋅ ) = .
n→∞ 2 + 2−n (n + 1)2 + 1 2

Therefore R = 2.

141
 ∞
x2n+1
15. Consider the power series ∑ (−1)n .
n=0 (2n + 1)(n2 + 1)
a. Find the radius of convergence of the power series.
b. Determine whether the power series converges or diverges at the right endpoint of its
interval of convergence.
c. Determine whether the power series converges or diverges at the left endpoint of its
interval of convergence.

x2n+1
Solution: a. We have an = (−1)n and
(2n + 1)(n2 + 1)

ρ = lim ∣an ∣1/n

n→∞
1/n
∣x∣2n+1
= lim ( n )
n→∞ (2 + 1)(n2 + 1)

∣x∣2 ⋅ ∣x∣1/n
= lim
n→∞ 2 ⋅ (1 + 2−n )1/n ⋅ n1/n ⋅ (1 + n−2 )1/n
2
∣x∣
= .
2
√
If ∣x∣ < 2 , then ρ = ∣x∣
√ /2 < 1 and the power series converges absolutely by the
2

Root Test; and if ∣x∣ > 2 √, then ρ = ∣x∣2 /2 > 1 and the power series diverges by the
Root Test. Therefore R = 2 .

√
b. At x = 2 we have
∞ ∞
√
x 2n+1 ( 2)2n+1
∑ (−1)n n = ∑ (−1) n
n=0 (2 + 1)(n2 + 1) n=0 (2n + 1)(n2 + 1)
√ ∞ 2n
= 2 ∑ (−1)n n .
n=0 (2 + 1)(n2 + 1)
∞
2n
Consider the corresponding absolute value series ∑ . Since
n=0 (2 + 1)(n + 1)
n 2
∞
2n 1 1
0< < for all n ≥ 1 and the p-series ∑ 2 with p = 2 > 1
(2n + 1)(n2 + 1) n2 n=1 n
∞
2n
converges, ∑ n converges by the Direct Comparison Test; and then
n=0 (2 + 1)(n√ + 1)
2

the power series at x = 2 converges absolutely by the Absolute Convergence Test.

√
c. At x = − 2 we have
∞ ∞
√
x2n+1 (− 2)2n+1
∑ (−1) nn
= ∑ (−1) n
n
n=0 (2 + 1)(n2 + 1) n=0 (2 + 1)(n2 + 1)
√ ∞ 2n
= 2 ∑ (−1)n+1 n .
n=0 (2 + 1)(n2 + 1)

142
 This is just −1 times the series we considered in part (b), and therefore it converges
absolutely.

16. Determine the radius of convergence and the interval of convergence of the power series
∞
xn
∑ .
n=0 3n + (−1)
n+1

Also determine the type of convergence at each point of the interval of convergence.

1
Solution: As cn = , the formulas for the radius of convergence gives
3n + (−1)n+1

1 ∣cn+1 ∣ ∣1/(3(n + 1) + (−1)n+2 )∣ 3 + (−1)n+1 /n

= lim = lim = lim = 1.
R ∣cn ∣ ∣1/(3n + (−1)n+1 )∣ 3 + 3/n + (−1)n+2 /n

Therefore R = 1 . Hence we have absolute convergence for ∣x∣ < 1 and divergence for
∣x∣ > 1.

At x = 1 we have ∞ ∞
xn 1
∑ = ∑ .
n=0 3n + (−1) n=0 3n + (−1)
n+1 n+1

Since
1/(3n + (−1)n+1 ) 1 1
L = lim = lim =
n→∞ 1/n n→∞ 3 + (−1)n+1 /n 3
∞
1
is a positive real number and the harmonic series ∑ diverges, the power series
n=1 n
diverges by the Limit Comparison Test at x = 1 .

At x = −1 we have ∞ ∞
xn (−1)n
∑ = ∑ .
n=0 3n + (−1) n=0 3n + (−1)
n+1 n+1

1
Let un = .
3n + (−1)n+1
1
i. un = > 0 for all n ≥ 1 .
3n + (−1)n+1
ii. 0 < 3n + (−1)n+1 ≤ 3n + 1 < 3n + 2 ≤ 3(n + 1) + (−1)n+2 for all n ≥ 1. Hence
un > un+1 for n ≥ 1 .
1
iii. lim un = lim = 0.
n→∞ n→∞ 3n + (−1)n+1

∞
(−1)n
It follows that the series ∑ converges by the Alternating Series Test.
n=0 3n + (−1)
n+1
∞
1
Its absolute value series is the same as the series 1 + ∑ and we showed
n=1 3n + (−1)
n+1

that this series diverges. Hence the power series converges conditionally at x = −1 .

143
 To summarize, the radius of convergence is R = 1, the interval of convergence
is [−1, 1), the power series converges absolutely at every point of (−1, 1), and it
converges conditionally at x = −1 .

∞
1 ⋅ 3 ⋅ 5 ⋅ ⋯ ⋅ (2n − 1) n
17. Consider the power series f (x) = 1 + ∑ x .
n=1 2 ⋅ 4 ⋅ 6 ⋅ ⋯ ⋅ (2n)

a. Show that the radius of convergence of the power series is R = 1.

b. Determine the behavior of this power series at the endpoints of its interval of
convergence.
c. Show that 2(1 − x)f ′ (x) = f (x) for ∣x∣ < 1.
1
d. Solve this differential equation to show that f (x) = √ for ∣x∣ < 1.
1−x
e. Show that ∞
1 ⋅ 3 ⋅ 5 ⋅ ⋯ ⋅ (2n − 1) x2n+1
arcsin x = x + ∑ ⋅
n=1 2 ⋅ 4 ⋅ 6 ⋅ ⋯ ⋅ (2n) 2n + 1
for ∣x∣ < 1.

1 ⋅ 3 ⋅ 5 ⋅ ⋯ ⋅ (2n − 1)
Solution: a. We have cn = for n ≥ 1 . We use the radius of
2 ⋅ 4 ⋅ 6 ⋅ ⋯ ⋅ (2n)
convergence formula

1 ⋅ 3 ⋅ 5 ⋅ ⋯ ⋅ (2n − 1)(2n + 1)
1 ∣cn+1 ∣ 2 ⋅ 4 ⋅ 6 ⋅ ⋯ ⋅ (2n)(2n + 2) 2n + 1
= lim = lim = lim =1
R n→∞ ∣cn ∣ n→∞ 1 ⋅ 3 ⋅ 5 ⋅ ⋯ ⋅ (2n − 1) n→∞ 2n + 2

2 ⋅ 4 ⋅ 6 ⋅ ⋯ ⋅ (2n)

to obtain R = 1 .

∞
1 ⋅ 3 ⋅ 5 ⋅ ⋯ ⋅ (2n − 1)
b. At x = 1 we have the series 1 + ∑ . Since
n=1 2 ⋅ 4 ⋅ 6 ⋅ ⋯ ⋅ (2n)

1 ⋅ 3 ⋅ 5 ⋅ ⋯ ⋅ (2n − 1) 3 5 2n − 1 1 1
= ⋅ ⋅⋯⋅ ⋅ >
2 ⋅ 4 ⋅ 6 ⋅ ⋯ ⋅ (2n) 2 4 2n − 2 2n 2n
∞
1
for n > 1 and the harmonic series ∑ diverges, we conclude that the series at x = 1
n=1 n
diverges by the Direct Comparison Test.
∞
1 ⋅ 3 ⋅ 5 ⋅ ⋯ ⋅ (2n − 1)
At x = −1 we obtain the alternating series 1 + ∑ (−1)n with
n=1 2 ⋅ 4 ⋅ 6 ⋅ ⋯ ⋅ (2n)
1 ⋅ 3 ⋅ 5 ⋅ ⋯ ⋅ (2n − 1)
un = . We have (i ) un > 0 for all n ≥ 0, and we also have (ii )
2 ⋅ 4 ⋅ 6 ⋅ ⋯ ⋅ (2n)

144
un > un+1 for all n ≥ 0 as un /un+1 = (2n + 2)/(2n + 1) > 1 for n ≥ 0. On the other
hand,

12 ⋅ 32 ⋅ 52 ⋅ ⋯ ⋅ (2n − 1)2 1 ⋅ 3 3 ⋅ 5 (2n − 3)(2n − 1) 2n − 1 1 1

u2n = = 2 ⋅ 2 ⋅⋯⋅ ⋅ ⋅ <
2 ⋅ 4 ⋅ 6 ⋅ ⋯ ⋅ (2n)
2 2 2 2 2 4 (2n − 2)2 2n 2n 2n
√
for n > 1. Therefore 0 < un < 1/ 2n for n > 1, and the Sandwich Theorem gives
(iii ) lim un = 0 . We conclude that the series at x = −1 converges by the Alternating
n→∞
Series Test. The convergence is conditional as we have already seen that the absolute
value series diverges.

c. For ∣x∣ < 1 we have

∞
1 ⋅ 3 ⋅ 5 ⋅ ⋯ ⋅ (2n − 1) n
f (x) = 1 + ∑ x .
n=1 2 ⋅ 4 ⋅ 6 ⋅ ⋯ ⋅ (2n)

Differentiating this we get

1 ∞ 1 ⋅ 3 ⋅ 5 ⋅ ⋯ ⋅ (2n − 1) n−1
f ′ (x) = +∑ x
2 n=2 2 ⋅ 4 ⋅ 6 ⋅ ⋯ ⋅ (2n − 2) ⋅ 2

for ∣x∣ < 1 . Therefore

∞
1 ⋅ 3 ⋅ 5 ⋅ ⋯ ⋅ (2n − 1) n
2xf ′ (x) = x + ∑ x
n=2 2 ⋅ 4 ⋅ 6 ⋅ ⋯ ⋅ (2n − 2)

and ∞
3 1 ⋅ 3 ⋅ 5 ⋅ ⋯ ⋅ (2n + 1) n
2f ′ (x) = 1 + x+ ∑ x
2 n=2 2 ⋅ 4 ⋅ 6 ⋅ ⋯ ⋅ (2n)

for ∣x∣ < 1 . Taking the difference of these two, we obtain

∞
1 1 ⋅ 3 ⋅ 5 ⋅ ⋯ ⋅ (2n − 1) 2n + 1
2(1 − x)f ′ (x) = 1 + x+ ∑ ⋅( − 1) xn
2 n=2 2 ⋅ 4 ⋅ 6 ⋅ ⋯ ⋅ (2n − 2) 2n
∞
1 1 ⋅ 3 ⋅ 5 ⋅ ⋯ ⋅ (2n − 1) 1 n
=1+ x+ ∑ ⋅ x
2 n=2 2 ⋅ 4 ⋅ 6 ⋅ ⋯ ⋅ (2n − 2) 2n
∞
1 1 ⋅ 3 ⋅ 5 ⋅ ⋯ ⋅ (2n − 1) n
=1+ x+ ∑ x
2 n=2 2 ⋅ 4 ⋅ 6 ⋅ ⋯ ⋅ (2n)
∞
1 ⋅ 3 ⋅ 5 ⋅ ⋯ ⋅ (2n − 1) n
=1+ ∑ x
n=1 2 ⋅ 4 ⋅ 6 ⋅ ⋯ ⋅ (2n)
= f (x)

for ∣x∣ < 1.

145
 d. For ∣x∣ < 1,

f ′ (x) 1
2(1 − x)f ′ (x) = f (x) Ô⇒ =
f (x) 2(1 − x)
1
Ô⇒ ln ∣f (x)∣ = − ln ∣1 − x∣ + C for some constant C
2
A
Ô⇒ f (x) = √ for some constant A .
1−x
1
Now substituting x = 0 gives 1 = f (0) = A. Therefore, f (x) = √ for ∣x∣ < 1.
1−x

e. We have ∞
1 1 ⋅ 3 ⋅ 5 ⋅ ⋯ ⋅ (2n − 1) n
√ =1+ ∑ x
1−x n=1 2 ⋅ 4 ⋅ 6 ⋅ ⋯ ⋅ (2n)

for ∣x∣ < 1 . Substituting x2 for x gives

∞
1 1 ⋅ 3 ⋅ 5 ⋅ ⋯ ⋅ (2n − 1) 2n
√ =1+ ∑ x
1−x n=1 2 ⋅ 4 ⋅ 6 ⋅ ⋯ ⋅ (2n)
2

for ∣x2 ∣ < 1 . Integrating this we obtain

∞
1 ⋅ 3 ⋅ 5 ⋅ ⋯ ⋅ (2n − 1) x2n+1
arcsin x = x + ∑ ⋅ +C
n=1 2 ⋅ 4 ⋅ 6 ⋅ ⋯ ⋅ (2n) 2n + 1

for ∣x∣ < 1 . Substituting x = 0 gives 0 = arcsin 0 = C, and hence C = 0. Thus

∞
1 ⋅ 3 ⋅ 5 ⋅ ⋯ ⋅ (2n − 1) x2n+1
arcsin x = x + ∑ ⋅
n=1 2 ⋅ 4 ⋅ 6 ⋅ ⋯ ⋅ (2n) 2n + 1

for ∣x∣ < 1 .

Remark: It can be shown that

∞
1 1 ⋅ 3 ⋅ 5 ⋅ ⋯ ⋅ (2n − 1) n
√ =1+ ∑ x
1−x n=1 2 ⋅ 4 ⋅ 6 ⋅ ⋯ ⋅ (2n)

for −1 ≤ x < 1 , and

∞
1 ⋅ 3 ⋅ 5 ⋅ ⋯ ⋅ (2n − 1) x2n+1
arcsin x = x + ∑ ⋅
n=1 2 ⋅ 4 ⋅ 6 ⋅ ⋯ ⋅ (2n) 2n + 1
for ∣x∣ ≤ 1 .

146
 ∞
n(2x − 1)3n+1
18. Consider the power series ∑ .
n=1 5n
a. Find the radius of convergence of the power series.
b. Find the sum of the power series explicitly.

n(2x − 1)3n+1
Solution: a. an = gives
5n
∣an+1 ∣
ρ = lim
n→∞ ∣an ∣
∣(n + 1)(2x − 1)3(n+1)+1 /5n+1 ∣
= lim
n→∞ ∣n(2x − 1)3n+1 /5n ∣
n + 1 ∣2x − 1∣3
= lim ⋅
n→∞ n 5
∣2x − 1∣3
= .
5
If ∣2x − 1∣3 /5 < 1 , then ρ < 1 and the series converges by the Ratio Test; and
if ∣2x − 1∣3 /5 > 1 , then ρ > 1√ and the series diverges by the Ratio Test. Since
3
1 5
∣2x − 1∣3 /5 < 1 ⇐⇒ ∣x − ∣ < , it follows that the radius of convergence of the
√ 2 2
power series is R = 3 5/2 .

b. We know that ∞
1
∑ xn =
n=0 1−x
for ∣x∣ < 1 . Differentiating this we obtain
∞
1
∑ nxn−1 =
n=1 (1 − x)2

for ∣x∣ < 1 . Now we replace x with (2x − 1)3 /5 to get

∞
n(2x − 1)3(n−1) 1
∑ = 2
n=1 5n−1 (2x − 1)3
(1 − )
5

for ∣(2x − 1)3 /5∣ < 1 . Finally we multiply by (2x − 1)4 /5 to obtain
∞
n(2x − 1)3n+1 (2x − 1)4 /5 5 (2x − 1)4
∑ = = ⋅
n=1 5n (2x − 1)3
2
4 (4x3 − 6x2 + 3x − 3)2
(1 − )
5
√
for ∣x − 1/2∣ < 3
5/2 .

147
19. Consider the function defined by:
∞
xn
f (x) = ∑
n=0 n! 2n(n−1)/2

a. Find the domain of f .

f (x) − ex
b. Evaluate the limit lim .
x→0 1 − cos x

3
c. Show that f (2) < e + .
2
d. Show that f (−2) < 0 .

1
Solution: a. Since cn = , the radius of convergence formula gives
n! 2n(n−1)/2
1 ∣cn+1 ∣ 1/((n + 1)! 2(n+1)n/2 ) 1
= lim = lim = lim =0.
R n→∞ ∣cn ∣ n→∞ 1/(n! 2 n(n−1)/2 ) n→∞ (n + 1)2n

Therefore R = ∞ . This means that the domain of f is (−∞, ∞) .

b. Using
∞
xn x2 x3
f (x) = ∑ =1+x+ + 3 +⋯
n=0 n! 2n(n−1)/2 2 ⋅ 2! 2 ⋅ 3!
we obtain
x2 x3 x2 x3
(1 + x + + 3 + ⋯) − (1 + x + + + ⋯)
f (x) − ex 2 ⋅ 2! 2 ⋅ 3! 2! 3!
lim = lim
x→0 1 − cos x x→0 x2 x4
+ − ⋯)
1 − (1 −
2! 4!
1 7 3 1 7
− x2 − x −⋯ − − x−⋯
= lim 4 48 = lim 4 48
x→0 1 1 4 x→0 1 1 2
x2 − x +⋯ − x +⋯
2 24 2 24
1
− 1
= 4 =− .
1 2
2

c. We have
∞ ∞
2n 1 2 1 ∞ 1
f (2) = ∑ =∑ =1+2+ + +∑ .
n=0 n! 2n(n−1)/2 n=0 n! 2n(n−3)/2 2! 3! n=4 n! 2 n(n−3)/2

On the other hand,

∞
1 1 1 ∞ 1
e=∑ =1+1+ + + ∑ .
n=0 n! 2! 3! n=4 n!

148
 1 1
Since n(n − 3) > 0 for n ≥ 4, we have < for n ≥ 4 . Hence
∞ ∞
n! 2n(n−3)/2 n!
1 1 3
∑ <∑ . Therefore, f (2) − e < .
n=4 n! 2n(n−3)/2 n=4 n! 2

d. This time we have

∞
(−2)n
f (−2) = ∑ n(n−1)/2
n=0 n! 2
∞
(−1)n
=∑ n(n−3)/2
n=0 n! 2
2 ∞ (−1)n
=1−2+ + ∑
2! n=3 n! 2n(n−3)/2
∞
(−1)n
=∑ n(n−3)/2
n=3 n! 2
1 1 1 1
=− + 2 − 5 + 9 −⋯.
3! 2 ⋅ 4! 2 ⋅ 5! 2 ⋅ 6!
1 1
As (n + 1)2n−1 > 1 for n ≥ 3 , we have − + < 0 for n ≥ 3
n! 2n(n−3)/2 (n + 1)! 2(n+1)(n−2)/2
1 1 5
too. Therefore, f (−2) < − + 2 = − < 0.
3! 2 ⋅ 4! 32

2
20. Estimate ∫ e−x dx with error less than 0.01.
2

Solution: We have
x2 xn
ex = 1 + x + +⋯+ +⋯
2! n!
for all x. Therefore,
x4 x2n
e−x = 1 − x2 +
2
+ ⋯ + (−1)n +⋯
2! n!
for all x, and integration gives
2 23 25 22n+1
e−x dx = 2 −
2
∫0 + + ⋯ + (−1)n +⋯.
3 2! 5 n! (2n + 1)
22n+1
Let un = . Then:
n! (2n + 1)
22n+1
i. un = > 0 for all n ≥ 0 .
n! (2n + 1)
ii. 2n2 − 3n − 1 > 0 for n ≥ 2 Ô⇒ (n + 1)(2n + 1) > 22 (2n + 1) for n ≥ 2. Hence
un > un+1 for n ≥ 2 .
4n 2
iii. lim un = lim ( ⋅ ) = 0.
n→∞ n→∞ n! 2n + 1

149
 The series satisfies the conditions of Alternating Series Test for n ≥ 2. Since 223 /(23⋅
11!) = 32768/3586275 ≈ 0.009 < 0.01, it follows by the Alternating Series Estimate
that the sum
2 23 25 219 221 12223758182
e−x dx ≈ 2 −
2
∫0 + +⋯− + = ≈ 0.89
3 2! 5 9! 19 10! 21 13749310575
2
approximates ∫ e−x dx with error less than 0.01.
2

∞
1
21. Find the exact value of ∑ .
n=0 4 (2n + 1)
n

Solution: On one hand we have

∞ ∞
1 (1/4)n ∞ (1/2)2n ∞
(1/2)2n+1
∑ = ∑ = ∑ = 2 ∑
n=0 4 (2n + 1) n=0 2n + 1 n=0 2n + 1 n=0 2n + 1
n

(1/2) 3 (1/2)5
= 2 (1/2 + + + ⋯) .
3 5
On the other hand we have
∞
xn x2 x3 x4 x5
ln(1 + x) = ∑ (−1)n+1 =x− + − + −⋯
n=1 n 2 3 4 5

for −1 < x ≤ 1 . In particular, x = 1/2 gives

3 1 (1/2)2 (1/2)3 (1/2)4 (1/2)5

ln ( ) = ln (1 + ) = 1/2 − + − + −⋯,
2 2 2 3 4 5

and x = −1/2 gives

1 1 (1/2)2 (1/2)3 (1/2)4 (1/2)5

ln ( ) = ln (1 − ) = −1/2 − − − − −⋯.
2 2 2 3 4 5

Therefore
∞
1 3 1
∑ = ln ( ) − ln ( ) = ln 3 .
n=0 4n (2n + 1) 2 2

150
 Part 4: Vector Analysis

1. Find the value of the line integral

‰
(3x2 y 2 + y) dx + 2x3 y dy
C

where C is the cardioid r = 1 + cos θ parameterized counterclockwise.

Solution: We use the Green’s Theorem

‰
∂N ∂M
M dx + N dy = ∬ ( − ) dA
C R ∂x ∂y

where R is the region enclosed by the simple closed curve C. Therefore

‰ ³¹¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹·
M
¹¹¹¹¹¹¹¹¹¹¹¹ µ
N
¬
N
¬
M
³¹¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹·¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ µ
∂ ∂
(3x2 y 2 + y) dx + 2x3 y dy = ∬ ( ( 2x3 y ) − (3x2 y 2 + y)) dA
C R ∂x ∂y
= ∬ (6x2 y − (6x2 y + 1)) dA
R

= − ∬ dA
R
2π 1+cos θ
= −∫ ∫0 r dr dθ
0
2π r2 r=1+cos θ
= −∫ [ ] dθ
0 2 r=0
1 2π
= − ∫ (1 + 2 cos θ + cos2 θ) dθ
2 0
1 2π 1 + cos 2θ
=− ∫ (1 + 2 cos θ + ) dθ
2 0 2
1 3
= − ⋅ ⋅ 2π
2 2
3π
=− .
2

Remark We used the Circulation-Curl Form of the Green’s Theorem, but the computation is
exactly the same with the Flux-Divergence Form:
‰
∂M ∂N
M dy − N dx = ∬ ( + ) dA
C R ∂x ∂y

151
In fact, when expressed in terms of components and coordinates, both forms of the Green’s
Theorem can be summarized and most easily remembered as
‰
ω = ∬ dω
C R

where ω = A dx + B dy is a differential form and dω = dA dx + dB dy . The multiplication of

differentials here is as described in the Remark following Example 46 in Part 2. Indeed, if
ω = M dx + N dy, then
∂M ∂M ∂N ∂N
dω = dM dx + dN dy = ( dx + dy) dx + ( dx + dy) dy
∂x ∂y ∂x ∂y
∂M ∂N ∂M ∂N ∂N ∂M
= dy dx + dx dy = − dx dy + dx dy = ( − ) dx dy ,
∂y ∂x ∂y ∂x ∂x ∂y
and if ω = M dy − N dx, then
∂M ∂M ∂N ∂N
dω = dM dy − dN dx = ( dx + dy) dy − ( dx + dy) dx
∂x ∂y ∂x ∂y
∂M ∂N ∂M ∂N ∂M ∂N
= dx dy − dy dx = dx dy + dx dy = ( + ) dx dy
∂x ∂y ∂x ∂y ∂x ∂y
where we used dx dx = 0 = dy dy and dy dx = −dx dy. Unlike in the case of change of variables,
now we must always get a dx dy under the double integral and we must not get rid of the sign.
‰
2. Evaluate F ⋅ dr where
C
−y x
F= i+ 2 j
+ 9y
4x22 4x + 9y 2
and C is the unit circle parametrized in the counterclockwise direction.
Solution: Observe that
∂ x ∂ −y
curl F = ( 2 )− ( 2 )
∂x 4x + 9y 2 ∂y 4x + 9y 2
1 8x2 1 18y 2
= 2 − + − =0
4x + 9y 2 (4x2 + 9y 2 )2 4x2 + 9y 2 4x2 + 9y 2
at all points (x, y) =/ (0, 0).

Consider C0 ∶ r = (cos t)/2 i + (sin t)/3 j, 0 ≤ t ≤ 2π, the counterclockwise

parametrization of the ellipse 4x2 + 9y 2 = 1. Let R be the region lying inside the
unit circle and outside this ellipse.

Since curl F = 0 at all points of R, ∬ curl F dA = 0 . Therefore by the generalized

R
Green’s Theorem,
‰ ‰ ‰
−y dx + x dy
F ⋅ dr = F ⋅ dr =
C0 4x + 9y
2 2
C C0
t=2π −(sin t)/3 d((cos t)/2) + (cos t)/2 d((sin t)/3)
=∫
t=0 cos2 t + sin2 t
1 2π 1 π
= ∫ dt = ⋅ 2π = .
6 0 6 3

152
3. Find the surface area of the parameterized torus

r = (a + b cos u) cos v i + (a + b cos u) sin v j + b sin u k ,

0 ≤ u ≤ 2π, 0 ≤ v ≤ 2π, where 0 < b < a are constants.

Solution: We have:

ru = −b sin u cos v i − b sin u sin v j + b cos u k

rv = −(a + b cos u) sin v i + (a + b cos u) cos v j
ru × rv = −b(a + b cos u) cos v cos u i − b(a + b cos u) sin v cos u j − b(a + b cos u) sin u k
∣ru × rv ∣ = b(a + b cos u)(cos2 v cos2 u + sin2 v cos2 u + sin2 u)1/2 = b(a + b cos u)

Therefore

Surface Area = ∬ ∣ru × rv ∣ dA = ∬ b(a + b cos u) dA

R R
2π 2π
=∫ ∫0 b(a + b cos u) du dv = ab ⋅ 2π ⋅ 2π = 4π 2 ab
0

where R = {(u, v) ∶ 0 ≤ u ≤ 2π and 0 ≤ v ≤ 2π}.

153
4. Find the area of the portion S of the cylinder x2 + y 2 = 2y that lies inside the sphere
x2 + y 2 + z 2 = 4.

Solution: Let F (x, y, z) = x2 + y 2 − 2y. Then the cylinder is given by F (x, y, z) = 0.

We will use the projection to the yz-plane. This projection is 2-1 from the portion of
the cylinder inside the sphere to the region R given by the inequalities 0 ≤ 2y ≤ 4−z 2
in the yz-plane, and by symmetry

∣∇F ∣
Surface Area of S = 2 ∬ dA
R ∣∇F ⋅ p∣

where p = i. We have

∇F = 2xi + (2y − 2)j

∣∇F ∣ = (4x2 + (2y − 2)2 )1/2 = (4x2 + 4y 2 − 8y + 4)1/2 = 2 on S
∇F ⋅ p = ∇F ⋅ i = 2x

and therefore
∣∇F ∣ 2 1
Surface Area = 2 ∬ dA = 2 ∬ dA = 2 ∬ dA
R ∣∇F ⋅ p∣ R ∣2x∣ R ∣x∣
1 √
= 2∬ √ dA as ∣x∣ = 2y − y 2 on S
R 2y − y 2
√ √
2 4−2y 1 2 4 − 2y
= 2∫ ∫ √ √ dz dy = 2 ⋅ 2 ∫ √ dy
0 − 4−2y 2y − y 2 0 2y − y 2
√ 2 1 √ √
= 4 ⋅ 2 ∫ √ dy = 4 2 ⋅ 2 2 = 16 .
0 y

154
 √
5. Consider the parametrized surface S ∶ r = u2 i + 2 uvj + v 2 k, −∞ < u < ∞ , 0 ≤ v < ∞ . Find
the area of the portion of the surface S that lies inside the unit ball x2 + y 2 + z 2 ≤ 1.
√
Solution: We have x = u2 , √ y = 2uv and z = v 2 on the surface. Therefore,
x2 + y 2 + z 2 ≤ 1 means (u2 )2 + ( 2uv)2 + (v 2 )2 ≤ 1. In other words, (u2 + v 2 )2 ≤ 1, or
u2 + v 2 ≤ 1. Hence the part of the surface lying inside the sphere is the image of the
region R, the upper half of the unit disk, in the uv-plane.

To compute the surface area we first compute:

√
ru = 2ui + 2vj
√
rv = 2uj + 2vk
√ √
ru ×ru = 2 2v 2 i − 4uvj + 2 2u2 k
√ √ √
∣ru ×ru ∣ = ((2 2v 2 )2 + (4uv)2 + (2 2u2 )2 )1/2 = 2 2(u2 + v 2 )
Therefore,

Surface Area = ∬ ∣ru ×ru ∣ du dv

R
√
= 2 2 ∬ (u2 + v 2 ) du dv
R
√ π 1
= 2 2∫ ∫0 r ⋅ r dr dθ
2
0
1 π
= √ ∫ dθ
2 0
π
=√
2
where we used the polar coordinates in the uv-plane.

Remark: There is a shorter way of solving this problem which does not use Calculus. The
given parametrization maps the upper half-uv-plane onto the half-cone given by the equation
y 2 = 2xz, and the conditions x ≥ 0 and z ≥ 0 in a one-to-one manner (except on the u-axis).
This half-cone has its vertex at the origin, its axis lies along the bisector of the positive x- and
z-axes, and it has an opening angle of 45○ . The portion of this half-cone cut off by√the unit
√ surface of a right cone with slant height ℓ = 1 and radius r = 1/ 2, hence
sphere is the lateral
with area πrℓ = π/ 2.

155
6. Verify Stokes’s Theorem for the vector field F = y i + z k and the surface S, where S is the
portion of the paraboloid z = x2 +y 2 satisfying z ≤ 3, with the unit normal vector field n pointing
away from the z-axis.
√ √
Solution: S is bounded by the curve C ∶ r = 3 cos t i − 3 sin t j + 3 k, 0 ≤ t ≤ 2π.
Note that this parametrization is consistent with the direction of n. The circulation
of F around C is
‰ ‰
F ⋅ dr = (y i + z k) ⋅ (dx i + dy j + dz k)
C C
t=2π √ √ √
=∫ (− 3 sin t i + 3 k) ⋅ (d( 3 cos t) i + d(− 3 sin t) j + d(3) k)
t=0
2π √ √ 2π
=∫ (− 3 sin t)(− 3 sin t) dt = 3 ∫ sin2 t dt = 3π
0 0

On the other hand, the paraboloid is a level surface of f (x, y, z) = z − x2 − y 2 , and

∇f = −2x i − 2y j + k points in the opposite direction to n, so we will choose the
minus sign from ± in the flux integral. The projection of S into the xy-plane is the
disk R = {(x, y) ∶ x2 + y 2 ≤ 3}. Finally,
RRR i j k RRRR
RRR RRR
RRR RR
RRR ∂ ∂ ∂ RRRR
∇ × F = RRR R R = −k .
RRR ∂x ∂y ∂z RRRRR
RRR RRR
RRR R
RR y 0 z RRRR
The flux of ∇ × F across S is
±∇f
∬S (∇ × F) ⋅ n dσ = ∬R F ⋅ ∣k ⋅ ∇f ∣ dA

= ∬ (−k) ⋅ (2x i + 2y j − k) dA
R
√
= ∬ dA = Area of R = π( 3)2 = 3π .
R

‰
Hence ∬ (∇ × F) ⋅ n dσ = F ⋅ dr.
S C

7. Verify Divergence Theorem for the vector field F = xz i + yz j + z 3 k and the region D =
{(x, y, z) ∶ x2 + y 2 + z 2 ≤ 4} .
Solution: ∇ ⋅ F = ∂(xz)/∂x + ∂(yz)/∂y + ∂(z 3 )/∂z = z + z + 3z 2 = 2z + 3z 2 , and

∭D ∇ ⋅ F dV = ∭D (2z + 3z ) dV
2

2π π 2
=∫ ∫0 ∫0 (2ρ cos ϕ + 3ρ cos ϕ) ρ sin ϕ dρ dϕ dθ
2 2 2
0
2π π 96
=∫ ∫ (8 cos ϕ + cos2 ϕ) sin ϕ dϕ dθ
0 0 5
64 2π 128π
= ∫ dθ = .
5 0 5

156
To compute the outward flux through the sphere S = {(x, y, z) ∶ x2 + y 2 + z 2 = 4}, we
divide it as the upper hemisphere S1 = {(x, y, z) ∶ x2 + y 2 + z 2 = 4 and x ≥ 0} and the
lower hemisphere S2 = {(x, y, z) ∶ x2 + y 2 + z 2 = 4 and x ≤ 0}, and project both onto
the disk R = {(x, y) ∶ x2 + y 2 ≤ 4} in the xy-plane.

The sphere is a level surface of f (x, y, z) = x2 + y 2 + z 2 . ∇f = 2x i + 2y j + 2z k points

in the same direction as n, the outward pointing unit normal vector field on S, so
we will choose the plus sign from ± in the flux integral.

For the upper hemisphere we have

±∇f
∬S F ⋅ n dσ = ∬R F ⋅ ∣k ⋅ ∇f ∣ dA
1

2x i + 2y j + 2z k
= ∬ (xz i + yz j + z 3 k) ⋅ dA
R 2z
= ∬ (x2 + y 2 + z 3 ) dA
R

= ∬ (x2 + y 2 + (4 − x2 − y 2 )3/2 ) dA
R
2π 2
=∫ ∫0 (r + (4 − r ) ) r dr dθ
2 2 3/2
0
2π 32 64π
=∫ (4 + ) dθ = 8π + .
0 5 5
A similar computation for the lower hemisphere gives
64π
∬S F ⋅ n dσ = −8π + 5 .
2

Therefore

∬S F ⋅ n dσ = ∬S F ⋅ n dσ + ∬S F ⋅ n dσ
1 2
64π 64π 128π
= (8π + ) + (−8π + )= ,
5 5 5

and ∬ F ⋅ n dσ = ∭ ∇ ⋅ F dV .
S D

157