Exercises to Section 10

Exercises in red are from the list of the typical exercises for the exam.
Exercises marked with a star ⇤ are for submission to your tutor.

Interchanging limiting operations

1. For each of the following relations, either justify the interchange of limiting operations
or show that the relation is false (or ill-defined).
Z 1 Z 1✓ ◆
n n
(a) lim dx = lim dx
n!1 0 n2 x2 + 1 0 n!1 n2 x2 + 1
Answer: The relation is false. Indeed, for the left hand side we have
Z n Z 1
1 1
lim dx = dx > 0.
n!1 0 x2 + 1 0 x 2+1

For the right hand side we have

n
lim = 0, 0 < x < 1,
n!1 n2 x 2 + 1
and so the integral in the right hand side equals zero. (A more careful inspection
also shows that the above limit does not exist when x = 0.)
Z 2 Z 2✓ ◆
n n
(b) lim dx = lim dx
n!1 1 n2 x2 + 1 1 n!1 n2 x2 + 1
Answer: The relation is true (we have zero on both sides). Justification:
n n
6 2 ! 0, n ! 1, 1 6 x 6 2,
n2 x 2 + 1 n +1
n
and so the sequence converges to the zero function uniformly on [1, 2].
n2 x 2 + 1
By a Theorem from the lectures, we can interchange integration with limit. (Alter-
natively, one can check directly that we have zero on both sides.)
Z 1 Z 1✓ ◆
1 1
(c) lim dx = lim dx
n!1 1 (x n2 ) 4 + 1 1 n!1 (x n2 ) 4 + 1
Answer: The relation is false. For the left hand side we have, changing the inte-
gration variable y = x n2
Z 1 Z 1
1 1
lim 2 4
dx = 4
dy > 0.
n!1 1 (x n ) +1 1 y +1

For the right hand side,

1
lim =0
n!1 (x n2 ) 4 + 1
for all x 2 R, and so the integral in the right hand side equals zero.
Z 1 Z 1✓ ◆
1 1
(d) lim dx = lim dx
n!1 1 (x n2 ) 4 + 1 1 n!1 (x n2 ) 4 + 1
Answer: The relation is true (we have zero on both sides). Justification:
1 1
6 ! 0, n ! 1, 1 6 x 6 1,
(x n2 ) 4 + 1 (1 n2 ) 4 + 1
1
and so the sequence converges to the zero function uniformly on
(x n2 )4 + 1
[ 1, 1]. By a theorem from the lectures, we can interchange integration with the
limit. (Alternatively, one can check directly that both sides equal zero.)
X1 Z 1 Z 1X1
cos nx cos nx
(e)⇤ 2+1
dx = 2+1
dx
n=1 0
n 0 n=1 n
Answer: The relation is true. Justification:
|cos nx| 1
6 2 , x 2 R,
n2 + 1 n +1
P1
and the series n=1 n2 +1 converges. By the Weierstrass M -test, the series
1
1
X cos nx
converges uniformly on [0, 1] and therefore, by a Corollary from the lec-
n=1
n2 + 1
tures, we can interchange integration with the series.
1 1
d X sin n3 x X d sin n3 x
(f) = ,x2R
dx n=1 n2 n=1
dx n2
Answer: The relation is false; more precisely, the right hand side is ill-defined.
Indeed, differentiation gives

X1 1
d sin n3 x X
2
= n cos n3 x;
n=1
dx n n=1

the series in the right hand side diverges — consider, for example x = 0.
1 1
d X sin nx X d sin nx
(g) = ,x2R
dx n=1 n3 n=1
dx n3
Answer: The relation is true. Justification: we have
|sin nx| 1 d sin nx cos nx 1
6 3, = 6 2,
n3 n dx n3 n2 n
P1 P1
and both series n=1 1/n2 , n=1 1/n3 converge. Thus, by the Weierstrass M -test,
both series
X1 X1
sin nx d sin nx
3
,
n=1
n n=1
dx n3

converge uniformly on R and so, by a Corollary from the lectures, differentiation can
be interchanged with summation here.
1 1
d X n X d n
(h) x = x , x 2 (0, 1). Hint: (0, 1) = [a<1 (0, a)
dx n=0 n=0
dx
Answer:
P1 n P Fix a < 1. We have xn < an and nxn 1 6 nan 1 , and both series
n=0 a , converge. By the Weierstrass M -test, it follows that both
1 n 1
n=0 na
series
1
X 1
X
xn , nxn 1
n=0 n=0

converge uniformly on (0, a), and so differentiation can be interchanged with the
series. As this is true on (0, a) for all a < 1, this is true on (0, 1).
2. Let {amn }1
n,m=1 be the following infinite matrix:

0 1 1 1 1 1 1
0 ···
B 2 4 8 16 32 C
B C
B C
B 1 1 1 1 1 C
B 0 · · ·C
B 2 2 4 8 16 C
B C
B C
B 1 1 1 1 1 C,
B 0 · · ·C
B 4 2 2 4 8 C
B C
B C
B C
B 1 1 1 1 1 C
B 0 · · ·C
@ 16 4 2 2 4 A
.. .. .. .. .. .. ..
. . . . . . .

where m labels the lines and n labels the columns. Compute and compare the quani-
tities
1 X
X 1 1 X
X 1
amn and amn .
m=1 n=1 n=1 m=1

Answer: We have
1
X X1
1 1
amn = n
= m 1
,
n=1 n=m
2 2

and so
1 X
X 1 1
X 1
amn = m 1
= 2.
m=1 n=1 m=1
2

Similarly,
1
X X1
1 1
amn = m
= ,
m=1 m=n
2 2n 1

and so
1 X
X 1 1
X 1
amn = n 1
= 2.
n=1 m=1 n=1
2

Obviously, 2 6= 2, so the order of summation cannot be interchanged here.

Power series
3. Determine the radius of convergence R of the given power series:
X1
xn
(a)
n=0
nn
Answer: Using the formula for the radius of convergence of power series, we have

(1/nn )1/n = 1/n ! 0,

and therefore R = 1.
X1 ✓ ◆n
n! n p n
(b) x Hint: use Stirling’s formula n! = 2⇡n (1 + o(1)), n ! 1.
n=0
(2n)! e
Answer: Using Stirling’s formula and c1/n = 1 + o(1), n ! 1, we find
✓ ◆1/n ✓ p ◆1/n
n! 2⇡n(n/e)n n/e
= p (1 + o(1)) = (1 + o(1)) ! 0
(2n)! 4⇡n(2n/e)2n (2n/e)2

as n ! 1, and therefore R = 1.
X1
n2 n
(c) x
n=0
2n
Answer: Using n1/n = 1 + o(1), we find

(n2 2 n 1/n
) = n2/n /2 ! 1/2, n ! 1,

and therefore R = 2.
X1
(d) (n!)xn
n=0
Answer: Using Stirling’s formula, we find

(n!)1/n = (2⇡n)1/2n (n/e)(1 + o(1)) ! 1, n ! 1,

and therefore R = 0.
X1
(e)⇤ (3 n + 4 n )xn
n=0
Answer: We have
n n 1/n 1
(3 +4 ) =3 (1 + (3/4)n )1/n ! 1/3, n ! 1,

and therefore R = 3.
X1 ✓ n ◆
a bn
(f) + 2 xn , a > 0, b > 0
n=1
n n
Answer: Suppose a > b. We have
✓ ◆1/n
an bn 1/n
+ 2 = an (1 + (b/a)n /n)1/n ! a, n ! 1,
n n
and so R = 1/a. Similarly, if b > a, we find
✓ ◆1/n
an bn 2/n
+ 2 = bn (1 + (a/b)n n)1/n ! b, n ! 1,
n n

and so R = 1/b. Conclusion: R = 1/ max{a, b}.

X1
(g) ↵ n xn , 0 < ↵ < 1
2

n=1
Answer: We have 2
(↵n )1/n = ↵n ! 0, n ! 1,
and so R = 1.
X1
xn
(h) p ,a>0
n=1
a n
Answer: We have
p p
n 1/n 1/ n
(1/a ) =a ! 1, n ! 1,

and so R = 1.
x2 x4 x6 x8
(i) 1 + x + 2 + x3 + 4 + x5 + 6 + x7 + 8 + · · · , where b > 0.
b b b b
Answer: We have (
b n if n is even
an =
1 if n is odd.
Thus, (
b 1
if n is even
a1/n =
if n is odd,
n
1

and so lim sup an , 1}. We obtain R = min{1, b}.

1/n 1
= max{b
n!1
1
X
(j) 2 n 2n
x
n=1
Answer: In order to relate this to more familiar notation, rewrite the series as
1
X
2 k x2 ; then an = 2 k if n = 2k and an = 0 otherwise. We have
k

k=1

k k
a1/n
n = (2 k 2
) =2 k2

if n = 2k and an = 0 otherwise. Finally, lim 2 = 1. It follows that

1/n k2 k

k!1
= 1, and so R = 1.
1/n
lim sup an
n!1
P1 P1
4. Let n=0 an xn and n=0 bn xn be two power series with radius of convergence R and
Prespectively.
S What can you say about the radius of convergence of the power series
P1
and n=0 an bn x ?
1 n n
(a
n=0 n + b n )x
P
Answer: Let us show that the radius of convergence of (an + bP n )x is min{R,
n
P S} nif
R 6= S. Suppose R > S (otherwise, swap the two series). Both an xn and bn x
converge for |x|P< S, and so the sum must also converge in this range. On the other
hand, suppose (an + bn )xn converges for some |x| > S; since the domain of con-
vergence
P of a power series is an Pinterval, we can choose x with S < |x| < R such that
(an + bn )xn converges. Since an xn also converges for this x, we conclude that
X X X
bn x n = (an + bn )xn an xn

also converges there, which contradicts our assumptions.

P
If R = S, then by the same reasoning the radius of convergence of (an + bn )xn is
> R. But in this case there is no upper bound on the radius of convergence of the sum;
indeed, if an = bn , then the sum converges everywhere.
P
Let us show that the radius of convergence of an bn xn is > RS. Indeed,
1/n 1/n 1/n 1/n 1/n
lim sup |an bn | = lim sup |an | |bn | 6 lim sup |an | lim sup |bn | = 1/(RS),
n!1 n!1 n!1

which gives the claim.

P
There is no upper bound on the radius of convergence
P of an bn xn . Indeed, if an vanish
for even n and bn vanish for odd n, then an bn xn converges everywhere, whatever
are R and S.
P1
5. Let n=0 an xn have the radiusP1 of convergence R > 0, and let k 2 N. Find the radius
of convergence of the series n=0 nk an xn .
Answer: Since n1/n ! 1 as n ! 1, we have
1/n 1/n 1/n
lim sup |nk an | = lim sup nk/n |an | = lim sup |an | .
n!1 n!1 n!1
P1
It follows that the radius of convergence of n=0 nk an xn is R.
Other series
6. Prove that the series
1
X
⇡n2 x
#(x) = e
n= 1

converges for x > 0 and its sum #(x) (called the theta function) is a continuous function
of x > 0. Theta function plays an important role in complex analysis and number theory.
Hint: consider convergence on intervals [a, 1) for a > 0.
Answer: Following the hint, consider the series on the interval [a, 1) for a > 0. We
have 2
e ⇡n x 6 e ⇡an , n > 1, x > a.
P1
Since n=1 e ⇡an converges, we conclude that our series converges absolutely and
uniformly on [a, 1) and defines a continuous function there. Since a > 0 is arbitrary,
we conclude that the series converges for all x > 0 and defines a continuous function
of x > 0. It is not difficult to see that the convergence on (0, 1) is not uniform.
P1
7. Let an ! 1 be a sequence of positive numbers such that 1
n=1 an < 1. Prove that
the series
1
X 1
n=1
x + an

converges uniformly for x > 0 and defines a continuous function on [0, 1). You will see
such series in the Complex Analysis module (it defines a function of complex variable
with singularities at the points an ).
Answer: We have
1 1
6 , x > 0,
x + an an
P1
and the series n=1 a1n converges by assumption. It follows that our series converges
absolutely and uniformly on [0, 1) and defines a continuous function there.
Challenging exercises
8. Prove that the function
1
X 1
f (x) =
n= 1
(x n)2

is well defined and continuous at all points x 2 R \ Z.

Answer: Fix N 2 N, consider x 2 ( N, N ), x 2
/ Z. Then

|x n| > ||n| |x|| > |n| N, |n| > N + 1,

and so
1 1
6 , |n| > N + 1.
(x n)2 (|n| N )2
Since the series X 1
(|n| N )2
|n|>N +1

converges, we see that the series for f (x) converges uniformly on ( N, N ) \ Z. Since
N 2 N can be chosen as large as we wish, we obtain the desired conclusion.
9. Let fn , f be functions on [0, 1) such that
• fn , f are Riemann integrable on each interval [0, b], b > 0;
• fn ! f as n ! 1 uniformly on each interval [0, b], b > 0;
• |fn (x)| 6 g(x), |fR(x)| 6 g(x) for all x > 0, where g is a function such that the
improper integral 0 g(x)dx converges.
1

Prove that Z 1 Z 1
lim fn (x)dx = f (x)dx.
n!1 0 0

Answer: Given " > 0, choose R > 0 such that

Z 1
g(x)dx < ".
R
By a theorem from the lectures, we have
Z R Z R
lim fn (x)dx = f (x)dx,
n!1 0 0

and so there exists N 2 N such that for all n > N we have

Z R Z R
fn (x)dx f (x)dx < ".
0 0

Then for all n > N we have

Z 1 Z 1 Z R Z R
fn (x)dx f (x)dx 6 fn (x)dx f (x)dx
0 0 0 0
Z 1 Z 1
+ fn (x)dx + f (x)dx ,
R R

where Z 1 Z 1 Z 1
fn (x)dx 6 |fn (x)|dx 6 |g(x)|dx < "
R R R

and similarly Z Z Z
1 1 1
f (x)dx 6 |f (x)|dx 6 |g(x)|dx < ".
R R R

Thus, for all n > N we have

Z 1 Z 1
fn (x)dx f (x)dx < 3",
0 0

which proves the claim.