Mathematisches Institut der Universität München 18.12.

2007
Prof. Laszlo Erdos, PhD

Solution to midterm exam

in functional analysis
Problem 1. Let A be a non-empty subset of a metric space (M, d). Prove that the function

f (x) = dist (x, A) := inf d(x, y)

y∈A

on M is continuous.
Solution. For any x, z ∈ M we have

dist (x, A) = inf d(x, y) ≤ inf (d(x, z) + d(z, y)) = d(x, z) + dist (z, A).
y∈A y∈A

It follows that |f (x) − f (z)| ≤ d(x, z), i.e. f is Lipschitz continuous.

Problem 2. Let K(x, y) be a continuous non-negative function on [0, 1] × [0, 1]. Define the
linear operator T by Z 1
T f (x) = K(x, y)f (y)dy.
0

a) Prove that if f ∈ L1 [0, 1],

then T f ∈ C[0, 1].
b) Consider T as a linear operator T : L1 [0, 1] → (C[0, 1], k · k∞ ). Prove that its norm

kT k = max K(x, y).

x,y∈[0,1]

Solution.
a) The functon K is uniformly continuous since it is continuous on a compact set. Then
Z 1 
x→x
|T f (x) − T f (x0 )| =  (K(x, y) − K(x0 , y))f (y)dy  ≤ kf k1 max |K(x, y) − K(x0 , y)| −→0 0.
 
0 y∈[0,1]

b) Let M = maxx,y∈[0,1] K(x, y). On the one hand,

Z 1 
kT f k∞ = sup  K(x, y)f (y)dy  ≤ M kf k1 ,
 
x∈[0,1] 0

which proves kT k ≤ M .
To prove that kT k ≥ M , let (x0 , y0 ) be the point such that K(x0 , y0 ) = M. For a given
 > 0, choose δ > 0 such that K(x0 , y) ≥ M −  for y ∈ (y0 − , y0 + ) (this is possible since K
is continuous). Then
Z 1
kT χ[y0 −,y0 +] k∞ ≥ |(T χ[y0 −,y0 +] )(x0 )| = K(x0 , y)χ[y0 −,y0 +] (y)dy ≥ (M −)kχ[y0 −,y0 +] k1 .
0

Since kχ[y0 −,y0 +] k =

6 0, it follows that kT k ≥ M − . Letting  → 0, we conclude that kT k = M .

Problem 3. Let f be a function on [0, 1]. Define the function g on [0, 1] by g(x) = f (x2 ).
Prove that if f ∈ L3 [0, 1], then g ∈ L1 [0, 1].
 1 1
Solution. By a change of variable and subsequent application of Hölder inequality with p = 3
and 1q = 23 (so that p = 3 and q = 32 ),
1 1
√ 1 1 |f (y)|
Z Z Z
2
kgk1 = |f (x )|dx = |f (y)|d y = √ dy
0 0 2 0 y
1  Z 1  1Z 1
3 dy 3 2
√
|f (y)|3 dy
3
≤ 3/4
= 2kf k3 .
2 0 0 y

Problem 4. Let A be the subset of the real Hilbert space l2 which consists of all sequences
(x1 , x2 , . . .) such that xk ≥ 0 for all k. Find the interior of A.
Solution. The interior Ao is empty. Suppose that there exists a sequence x = (x1 , x2 , . . .)
belonging to Ao . Then there exists  > 0 such that for any y ∈ l2 with kyk ≤  we have x+y ∈ A.
But any sequence in l2 is convergent; in particular for sufficiently large k we have xk < . It
follows that if y = −ek (where ek is the standard basis vector), then (x + y)k < 0 and hence
x+y ∈ / A. Contradiction.
R 2π
Problem 5. Let f : [0, 2π] → C be a periodic C 1 -function such that 0 f (x)dx = 0. Prove
that Z 2π Z 2π
2
|f (x)| dx ≤ |f 0 (x)|2 dx.
0 0
[Hint: Fourier series.]
P P
Solution. Let n∈Z cn en be the Fourier series for f . Then n∈Z incn en is the Fourier series
R 2π √ R 2π
for f 0 . Since 0 f (x)dx = 2πhe0 , f i, the condition 0 f (x)dx = 0 means that c0 = 0. Since
for any integer n 6= 0 we have n2 ≥ 1, it follows that
Z 2π X X Z 2π
|f (x)|2 dx = |cn |2 ≤ n2 |cn |2 = |f 0 (x)|2 dx.
0 06=n∈Z 06=n∈Z 0

Problem 6. Recall that lp = {(x1 , x2 , . . .) : p

P
n |xn | < ∞}.
a) Prove that l2 ⊂ l3 .
b) Prove that l3 \ l2 6= ∅.

Solution.
a) Let a sequence (x1 , x2 , . . .) belong to l2P
, i.e. k |xk |2 P
P
< ∞. Then in particular the sequence
is bounded: M = supk |xk |P < ∞. But then |x |3 ≤M 2 3
k k k |xk | < ∞, i.e. (x1 , x2 , . . .) ∈ l .
1 1
b) Let xk = √k . Then k |xk |2 = k k = ∞, i.e. (x1 , x2 , . . .) ∈ / l2 .
P
1
On the other hand, k |xk |3 = k k3/2 < ∞, i.e. (x1 , x2 , . . .) ∈ l2 .
P P