D-MATH Functional Analysis I HS 2012

Sample Solutions for Exercise Sheet 1

1. Let C([0, 1]) be the set of real-valued continuous functions on [0,1] endowed with the usual
topology which is induced by the metric

d(f, g) = sup |f (x) − g(x)|

x∈[0,1]

for f, g ∈ C([0, 1]). Prove that the set

R for all n and

X X
an xn  an ∈
 
A = x 7→ |an | < ∞ ⊂ C([0, 1])
n≥0 n≥0

is a set of the first Baire category.

Recall: A subset B of a topological space C is of the first Baire category if B is the union of
countably many nowhere dense subsets of C. Otherwise it is a set of the second category.

Let `1 be the sequence space of all real sequences a = (an )n≥0 satisfying
X
kak1 := |an | < ∞,
n≥0
P
endowed with the topology which is induced by the metric d(a, b) = n≥0 |an − bn | for
sequences a, b ∈ `1 . Consider the map
 
X
f : `1 → C([0, 1]), a 7→ x 7→ an xn  .
n≥0

One can easily check that f is both continuous and linear. Furthermore we have A = f (`1 ).

Consider the set

U = {a ∈ `1 | kak1 < 1}
and note that `1 =
S
k≥1 kU where kU := {(ka0 , ka1 , . . .) | (a0 , a1 , . . .) ∈ U } for all positive
integers k. Hence  
[ [
A = f (`1 ) = f  kU  = f (kU ).
k≥1 k≥1

Assume that A is a set of the second category. Thus, there exists a positive integer k0 such that
f (k0 U ) is not nowhere dense, i.e. the interior of the closure of f (k0 U ) is not empty. Since the

1
interior of a set is an open subset, it contains an open ball whenever it is not empty. Hence,
there exists a function g0 ∈ C([0, 1]) and a real number r > 0 such that

B(g0 , r) := {g ∈ C([0, 1]) : sup |g(x) − g0 (x)| < r} ⊂ f (k0 U ).

x∈[0,1]

Let g ∈ V := B(0, 1) (note that the first argument denotes the zero function). Since
g0 , g0 + rg ∈ B(g0 , r) and since C([0, 1]) is a linear space, we see that rg = (g0 + rg) − g0
is an element of
{g1 − g2 | g1 , g2 ∈ f (k0 U )}
which is a subset of 2f (k0 U ) = f (2k0 U ); here we have used the linearity of f . Thus,
rV ⊂ f (2k0 U ). Hence, again by the linearity of f , we have
r
· V ⊂ f (U ). (1)
2k0

r
Set δ = 2k0 and let h0 ∈ δV ⊂ f (U ). Then, by relation (1), there exists a sequence a(1) ∈ U
with
δ
sup |h0 (x) − (f (a(1) ))(x)| < ,
x∈[0,1] 2

i.e. h1 := h0 − f (a(1) ) ∈ 2δ V .
1
Inductively, we can choose a sequence of `1 -sequences (a(1) , a(2) , . . .) with a(n) ∈ 2n−1 ·U
and  
(n) δ 1
hn := hn−1 − f (a ) ∈ n · V ⊂ f ·U
2 2n
for all n ≥ 1. Thus hn = h0 − ni=1 f (a(i) ) → 0 as n → ∞.
P
By construction, we have
  
X∞  X 1 
ka(n) k1 ≤ sup kbk1 : b ∈  · U  ≤ sup {kbk1 : b ∈ 2U } = 2.
 2n 
n=1 n≥0
P 
k (n) k
Thus, the sequence n=1 ka 1 of real numbers is bounded and monotone, hence con-
k
vergent. Therefore, since `1 is complete, there exists a ∈ 2U with N (n) → a as N → ∞.
P
n=1 a
Note: Here we have used the fact that every absolutely convergent series in a complete normed
space is convergent. A proof of this statement will be given at the end of this solution.
Since f is continuous and linear, we also have
∞ ∞
!
X X
f (a) = f a(n) = f (a(n) ) = h0 .
n=1 n=1

Thus,
δV ⊂ f (2U ).
Therefore,  
[ [ 2k
C([0, 1]) = kV ⊂ f U = f (`1 ) = A. (2)
δ
k≥1 k≥1

2
 1
But this yields a contradiction: The function g : [0, 1] → , x 7→ 1+x R
is clearly in C([0, 1]).
By (2), we should have g ∈ A, i.e. there should exist a sequence (a0 , a1 , . . .) of real numbers
with X 1
an xn = (3)
1+x
n≥0

and X
|an | < ∞. (4)
n≥0

But (3) implies X

a0 + (an + an−1 )xn = 1
n≥1

for all x ∈ [0, 1], hence

(a0 , a1 , a2 , a3 , . . .) = (1, −1, 1, −1, . . .)

which contradicts (4).

It remains to prove the following proposition about complete normed spaces:

Let (X, k · k) bePa complete normed vector space and (xn ) a sequence in X. If ∞
P
n=1 kxn k
converges, then ∞ x converges.
N
n=1 P n
Proof: Let ε > 0. Since ∞ such that ∞
P
n=1 kxn k converges, there exists N ∈ n=m kxn k < ε
for all m > N . Thus, for all a, b > N with a < b, we have
b
X a
X b
X b
X ∞
X ∞
X
k xn − xn k = k xn k ≤ kxn k = kxn k − kxn k < ε;
n=1 n=1 n=a+1 n=a+1 n=b n=a

note that
Pthe first 
inequality follows from the triangle inequality
P which  holds for the norm k · k.
k k
Thus, n=1 xn is a Cauchy sequence in X. Hence, n=1 xn converges because X is
k k
complete.

2. Let g ∈ C([0, 1]) and

 
Ag = f g  f ∈ C([0, 1]) ⊂ C([0, 1]).

Give a necessary and sufficient condition on g for Ag to be a set of the second category.

We will show

Ag is of the second category in C([0, 1]) ⇔ g(x) 6= 0 for all x ∈ [0, 1].

If g(x) 6= 0 for all x ∈ [0, 1] then we can write every h ∈ C([0, 1]) as the product h = hg ·g and
we have hg ∈ C([0, 1]). Thus, Ag = C([0, 1]) which is a complete metric space and therefore
of the second category by the Baire category theorem.

3
 For the other direction, let g be a function in C([0, 1]) for which there exists y ∈ [0, 1] with
g(y) = 0. Let
Vy = {h ∈ C([0, 1]) | h(y) = 0},
the closed set of all functions which vanish in y. Note that Ag ⊂ Vy . Assume that Ag is a
subset of the second category in C([0, 1]). In particular, there exist f1 ∈ C([0, 1]) and r > 0
such that
B(f1 , r) ⊂ Ag ⊂ Vy = Vy .
Set f2 (x) := f1 (x) + 2r for all x ∈ [0, 1]. Then f2 ∈ B(f1 , r) but f2 ∈
/ Vy since f2 (y) 6= 0,
which yields a contradiction.

3. (a) Let
Ceven = {f | f is even}, Codd = {f | f is odd}
be two subsets of C([−1, 1]). Show that both sets are nowhere dense in C([−1, 1]).

Assume that Ceven is not nowhere dense in C([−1, 1]). Then there exist g0 ∈ C([−1, 1]) and
r > 0 such that the open ball B(g0 , r) is contained in the closure of Ceven . Since Ceven is a
closed set, we then have
B(g0 , r) ⊂ Ceven .
Let g1 ∈ B(0, r). Then g0 + g1 ∈ B(g0 , r) ⊂ Ceven , i.e. g0 + g1 is an even function. Since the
difference of any two even functions is also even, this implies that g1 = (g0 + g1 ) − g0 is even.
Therefore,
B(0, r) ⊂ Ceven .
R
However, if g ∈ Ceven , then cg is in Ceven for all c ∈ . This yields B(0, kr) ⊂ Ceven for all
integers k, hence [
C([−1, 1]) = B(0, kr) ⊂ Ceven .
k≥1

This is an obvious contradiction since there exist functions in C([−1, 1]) which are not even.
The proof for Codd is analogous.

(b) Find a topological vector space X and sets A, B ⊂ X of the first category such that the
sum set  
A + B = a + b  a ∈ A, b ∈ B
is a subset of the second category.

Let X = C([−1, 1]) and A = Ceven and B = Codd . Any f ∈ X can be written as
f = feven + fodd where

f (x) + f (−x) f (x) − f (−x)

feven (x) = , fodd =
2 2
for all x ∈ X. Note that feven ∈ A and fodd ∈ B.
X is a complete metric space and therefore of the second category by the Baire category theo-
rem. However, A and B are nowhere dense by (a) and therefore subsets of the first category.