Qualifying Examination (Analysis)

(Fall, 2024)

1. (10 points) Let {xn } be a sequence of real numbers with the property that
1
|xn+1 − xn | ≤ , for all n ≥ 1.
n2
Prove that the sequence {xn } is convergent.

Solution. We prove that the sequence {xn } is Cauchy. For n > m, we have, using the triangle inequality

|xn − xm | = |xn − xn−1 + xn−1 − xn−2 + · · · + xm+1 − xm | ≤ |xn − xn−1 | + |xn−1 − xn−2 | + · · · + |xm+1 − xm |
1 1 1
≤ + + ··· + 2.
(n − 1)2 (n − 2)2 m
P 1
Since the series n2 is convergent, its partial sums form a Cauchy sequence. Therefore, given ϵ > 0, there
exists N ∈ N such that, for n > m ≥ N ,
n−1
X 1
< ϵ.
k2
k=m

Therefore, for n > m ≥ N , we have

|xn − xm | < ϵ.

This means that the sequence {xn } is Cauchy and so it is convergent.

2. (10 points) Let (X, d) be a metric space. The distance bewtwen two nonempty subsets A and B of X is
defined to be
d(A, B) := inf{d(x, y) : x ∈ A, y ∈ B}.

Prove that if A and B are two nonempty compact subsets of X, then there exist two points x0 ∈ A and
y0 ∈ B such that
d(A, B) = d(x0 , y0 ).

Solution. By the definition of infimum, for any n ∈ N, there exist points xn ∈ A and yn ∈ B such that
1
d(xn , yn ) < d(A, B) + .
n
Since A is compact, there exists a convergent subsequence {xnk } with limit x0 ∈ A. Consider then the
subsequence {ynk } in the compact subset B. This has a convergent sub-subsequence, call it {yml } with limit
y0 in B. Then the sub-subsequence {xml } of {xnk } still has the limit x0 .
The function d on X × X is continuous. This can be seen from an easy application of the triangle
inequality:
|d(x1 , y1 ) − d(x2 , y2 )| ≤ d(x1 , x2 ) + d(y1 , y2 ), for all x1 , x2 , y1 , y2 ∈ X.

Therefore,
d(A, B) ≤ d(x0 , y0 ) = lim d(xml , yml ) ≤ d(A, B),
l→∞

so d(x0 , y0 ) = D(A, B).

Another simple solution is to use the continuity of the map d on the compact space A × B. By the
Extremum Value Theorem, the map has a minimum, which is the infimum d(A, B).
1
3. (10 points) Prove that the function
∞
X 1
f (x) = 2 + n2
, (x ∈ R)
n=1
x

is well defined and continuous on R. Is it differentiable?

Solution. We prove that the convergence of the series is uniform on the whole real line. We use the
Weierstrass M -test. We have
1 1
≤ 2 , (x ∈ R).
x2 + n2 n
Since n n12 is convergent, we get that the series is uniformly convergent, and therefore the function is well
P

defined and continuous.

To prove that the function is differentiable we show that the series of the derivatives is uniformly con-
vergent on any bounded interval. Let M > 0 and consider the series on the interval [−M, M ]. We evaluate
a bound for the derivative
−2x 2M
2 2 2
≤ 4 , (x ∈ [−M, M ]).
(x + n ) n
Since the series n n4 is convergent, by the Weierstrass M -test, the series of the derivatives n (x2−2x
P 2M P
+n2 )2 is
uniformly convergent on [−M, M ]. This implies that the function f is differentiable on [−M, M ], and since
M was arbitrary, f is differentiable everywhere.

4. (10 points) Determine for what values of p > 0 the series

∞   
X 1
n 1 − cos
n=1
np

is convergent.
1


Solution. First we have to see how fast 1 − cos np goes to zero. We use Taylor’s expansion for cos x:

x2 x4
cos x = 1 − + − ...
2 4!
x2
Therefore 1 − cos x goes to 0 like 2! . More precisely

1 − cos x
lim x2
= 1.
x→0
2

This can be checked either using L’Hopital’s rule twice, or using Lagrange’s remainder’s theorem for f (x) =
cos x:
f (2) (ξ) 2 − cos ξ 2
cos x − (1 + 0 · x) = x = x , for some ξ between 0 and x.
2! 2
Then
1 − cos x
x2
= cos ξ → 1.
2
1
Then, for our problem, let x = np , and

n 1 − cos n1p



lim = 1,
n→∞ n · 2n12p
1
P
By the Comparison Test, our series has the same nature as the series n n2p−1 , which is convergent iff
2p − 1 > 1, i.e., p > 1.
2
5. (10 points) Let f : [−1, 1] → R be continuous. Find
Z 1
lim n x2n f (x) dx.
n→∞ −1

Solution. Note first that nx2n → 0 for all x ∈ (−1, 1), but it blows up to infinity at x = ±1. Therefore,
we have to isolate the two singularities ±1 and split the integral into 3 pieces: around -1, around +1 and away
from ±1. To simplify, we can work first on the interval [0, 1] and then on [−1, 0] we can use an analogous
argument.
Let ϵ > 0, arbitrary and let 0 < δ < 1 which we will specify later, but think of it as being close to 1. We
have: between 0 and δ, nx2n ≤ nδ 2n , and the function f is bounded, |f (x)| ≤ M for all x ∈ [−1, 1] for some
M > 0, since it is continuous. Therefore, |nx2n f (x)| ≤ nδ 2n M → 0 as n → ∞, so the function goes to zero
uniformly on [0, δ]. Thus
Z δ
nx2n f (x) dx → 0.
0

For δ close to 1, since f is continuous, f (x) is close to f (1). From the previous paragraph, the values of
R1
f on [0, δ] do not really matter. Therefore, we compare our quantity with n 0 x2n f (1) dx:
Z 1 Z 1 Z δ Z 1
2n 2n 2n
An := nx f (x) dx − nx f (1) dx ≤ |f (x) − f (1)|nx dx + |f (x) − f (1)|nx2n dx.
0 0 0 δ

Now pick δ ∈ (0, 1) such that |f (x) − f (1)| < ϵ for x ∈ [δ, 1] (by continuity). Then pick N such that
Z δ
|f (x) − f (1)|nx2n dx < ϵ, for n ≥ N.
0

We have, for n ≥ N ,
Z 1 Z 1
2n n
An ≤ ϵ + ϵnx dx ≤ ϵ + nϵ x2n dx = ϵ + ϵ < 2ϵ.
δ 0 2n + 1

This shows that our limit on [0, 1] is the same as

Z 1
n f (1)
lim n x2n f (1) dx = lim f (1) = .
n→∞ 0 n→∞ 2n + 1 2

For the other interval [−1, 0] we obtain

Z 0
f (−1)
lim n x2n f (x) dx = .
n→∞ −1 2

Adding together we obtain

Z 1
f (−1) + f (1)
lim n x2n f (x) dx = .
n→∞ −1 2

6. (10 points) Let f : (0, ∞) → R be differentiable and let a be a real number. Prove that if limt→∞ f ′ (t) = a,
then

f (t)
lim = a.
t→∞ t
3
 Solution. Let ϵ > 0. Since limt→∞ f ′ (t) = a, there exists t0 > 0 such that |f ′ (t) − a| < ϵ for all t > t0 .
Let t > t0 . By the Mean Value Theorem on [t, t0 ], there exists a point ξ > t0 such that

f (t) − f (t0 )
= f ′ (ξ).
t − t0
Therefore
f (t) − f (t0 )
− a = |f ′ (ξ) − a| < ϵ.
t − t0
f (t0 )
We have limt→∞ t−t0 = 0, therefore, there exists t1 > t0 such that

f (t0 )
< ϵ, for t > t1 .
t − t0

Then, for t > t1 , we have

f (t) f (t) − f (t0 ) f (t0 )

−a ≤ −a + < 2ϵ.
t − t0 t − t0 t − t0

Then we evaluate
f (t) f (t) f (t) t − t0 f (t) f (t) t − t0
− = · − = −1
t t − t0 t − t0 t t − t0 t − t0 t

t0
≤ (|a| + 2ϵ) < ϵ,
t
for t large enough, t > t2 > t1 .
Therefore, for t > t2 ,
f (t) f (t) f (t) f (t)
−a ≤ − + − a < 3ϵ,
t t t − t0 t − t0
and this shows that
f (t)
lim = a.
t→∞ t

7. (10 points) Let {fn } be a uniformly bounded sequence of continuous functions on [a, b]. Let
Z x
Fn (x) := fn (t) dt, (a ≤ x ≤ b).
a

Prove that there exists a subsequence {Fnk } which converges uniformly on [a, b].

Solution. We use Arzela-Ascoli Theorem. We have to check that the sequence {Fn } is uniformly bounded
and equicontinuous.
Since the sequence {fn } is uniformly bounded, there exists a constant M > 0 such that |fn (x)| ≤ M for
all x ∈ [a, b] and n ∈ N.
We have Z x
|Fn (x)| ≤ M dt ≤ M (b − a), (x ∈ [a, b], n ∈ N).
a

Thus {Fn } is uniformly bounded.

By the Fundamental Theorem of Calculus,

|Fn′ (x)| = |fn (x)| ≤ M, (x ∈ [a, b], n ∈ N),

4
so the derivatives are uniformly bounded. This implies that the functions {Fn } are uniformly Lipschitz:
with the Mean Value Theorem, for some ξ ∈ [a, b],

|Fn (x) − Fn (y)| = |Fn′ (ξ)||x − y| ≤ M |x − y|, for all x, y ∈ [a, b] and n ∈ N.

ϵ
Then, the family {Fn } is equicontinuous: for ϵ > 0, take δ = M and, if |x − y| < δ, then

|Fn (x) − Fn (y)| ≤ M |x − y| < ϵ, for all n ∈ N.

Arzela-Ascoli’s Theorem implies that there is a convergent subsequence.