myweb.ttu.

edu/bban
[email protected]

Real Analysis
Byeong Ho Ban
Mathematics and Statistics
Texas Tech University

Chapter 5. Elements of Functional Analysis (Last update : March 30, 2018)

1. If X is a normed vector space over K (= R or C), then addition and scalar multiplication are continuous
from X × X and K × X to X . Moreover, the norm is continuous from X to [0, ∞); in fact, |kxk − kyk| ≤
kx − yk.
Proof. (Feb 3rd 2018)

N The addition is continuous.

Let  > 0 be given. And let δ = 2 . Then, for any (x, y) ∈ X × X , observe the below.


max(kx − vk, ky − wk) = k(x − v, y − w)k = k(x, y) − (v, w)k < δ =
2
 
=⇒ k(x + y) − (v + w)k = k(x − v) + (y − w)k ≤ kx − vk + ky − wk < + = 
2 2
Therefore, the addition operation is continuous.

N The Scalar multiplication is continuous.

 
Let  > 0 be given. Then, for any (α, x) ∈ K × X , let δ = min( 3kxk , 3|α| , 1, 3 ) and observe the below.

  
max(|α − β|, kx − yk) = k(α − β, x − y)k = k(α, x) − (β, y)k < δ = min( , , 1, )
3kxk 3|α| 3
=⇒ kαx − βyk ≤ kαx − βxk + kβx − βyk = |α − β|kxk + |β|kx − yk
 
≤ |α − β|kxk + |α|kx − yk + |α − β|kx − yk < + + kx − yk < 
3 3

N The norm operation is continuous.

Let  > 0 and x ∈ X be given. And let δ =  and, for any y ∈ X , by triangular inequality, note the
below.

kxk ≤ kx − yk + kyk =⇒ kxk − kyk ≤ kx − yk

kyk ≤ kx − yk + kxk =⇒ kyk − kxk ≤ kx − yk
Thus, |kxk − kyk| ≤ kx − yk.

Then observe the below.

1
2

kx − yk < δ =  =⇒ |kxk − kyk| ≤ kx − yk < 

Therefore, the mapping k · k : X → K is continuous.


2. L(X , Y) is a vector space and the function k · k defined by (5.3) is a norm on it. In particular, the three
expression on the right of (5.3) are always equal.
Proof. (Feb 4th 2018)

N L(X , Y) is a vector space.

Note that L(X , Y) be the space of all bounded linear maps from X to Y over a field K(= R or C). For
any T, W ∈ L(X , Y) and α, β ∈ K, noting that there exist CT ≥ 0 and CW ≥ 0 such that kT xk ≤ CT kxk
and kW xk ≤ CW kxk for all x ∈ X ,observe the below.

k(αT + βW )xk = kαT x + βW xk ≤ |α|kT xk + |β|kW xk ≤ (|α|CT + |β|CW )kxk ∀x ∈ X

Clearly, αT + βW is linear map, thus, by the relation above, L(X , Y) is a vector space.

N The three expressions on the right of (5.3) are always equal.

For convenience, let’s define the sets for the expressions as below.
A = {kT xk : kxk = 1}
kT xk
B := { : x 6= 0}
kxk
D := {C : kT xk ≤ Ckxk for all x}
Clearly, A ⊂ B, so sup A ≤ sup B.

And observe that if x = 0, kT xk ≤ Ckxk is always true for any C ∈ [0, ∞) which means the below.
{C : kT xk ≤ Ckxk for all x} = {C : kT xk ≤ Ckxk for all x 6= 0}
Let C ≥ 0 be any constant satisfying kT xk ≤ Ckxk for all x ∈ X with x 6= 0. Observe the below.
kT xk
kT xk ≤ Ckxk =⇒ ≤C
kxk
Since C is arbitrary in D,
kT xk kT xk
≤ inf D ∀x 6= 0 =⇒ sup B = sup ≤ inf D
kxk x6=0 kxk

Thus, sup B ≤ inf D.

For any nonzero vector x ∈ X , observe the below.

kT xk x ∈ A =⇒ kT xk ≤ sup A =⇒ kT xk ≤ sup Akxk =⇒ inf D ≤ sup A

= T
kxk kxk kxk
To sum up,
sup A ≤ sup B ≤ inf D ≤ sup A
=⇒ sup A = sup B = inf D
 3

Therefore, the three expressions are equivalent.

N The function k · k defined by (5.3) is a norm.

kT xk
Observe that kT k = 0 ⇐⇒ kxk
= 0 ∀x ∈ X \ {0} ⇐⇒ T x = 0 ∀x ∈ X \ {0} ⇐⇒ T = 0

For any T, W ∈ L(X , Y),

k(T + W )xk kT xk + kW xk kT xk kW xk
kT + W k = sup ≤ sup ≤ sup + sup = kT k + kW k
x6=0 kxk x6=0 kxk x6=0 kxk x6=0 kxk

Lastly, for any T ∈ L(X , Y) and α ∈ K,

kαT xk |α|kT xk kT xk
kαT k = sup = sup = |α| sup = |α|kT k
x6=0 kxk x6=0 kxk x6=0 kxk

Therefore, k · k defined by (5.3) is a norm. 

3. Complete the proof of Proposition 5.4.

Proposition 5.4
If Y is complete, so is L(X , Y).
So prove that T ∈ L(X , Y)(in fact, kT k = limn→∞ kTn k) and that kTn − T k → 0.
Proof. (Feb 4th 2018)

Note that we have defined T : X → Y as below.

T x = lim Tn x
n→∞

For any x, y ∈ X and any α ∈ K, observe the below.

T (x + y) = lim Tn (x + y) = lim [Tn x + Tn y] = lim Tn x + lim Tn y = T x + T y
n→∞ n→∞ n→∞ n→∞
T (αx) = lim Tn (αx) = lim αTn x = α lim Tn x = αT x
n→∞ n→∞ n→∞

Therefore, T is linear.

Also, since {Tn x}∞

n=1 is Cauchy, there is N ∈ N such that

kTn x − Tm xk < kxk ∀x ∈ X

Since k · k is continuous map, letting m → ∞,
kTn x − T xk < kxk ∀x ∈ X
=⇒ kT xk < kxk + kTn xk ≤ (1 + Cn )kxk
where kTn xk ≤ Cn kxk for any x ∈ X and ∀n ∈ N.
Therefore, T ∈ L(X , Y).
In fact, observe the below.
kT xk k limn→∞ Tn xk limn→∞ kTn xk kTn xk
kT k = sup = sup = sup = lim sup = lim kTn k
x6=0 kxk x6=0 kxk n6=0 kxk n→∞ x6=0 kxk n→∞

Lastly, observe the below.

4

k(Tn − T )xk kTn x − T xk 0

lim kTn − T k = lim sup = sup lim = sup =0
n→∞ n→∞ x6=0 kxk x6=0 n→∞ kxk x6=0 kxk

Therefore, kTn − T k → 0


4. If X , Y are normed vector spaces, the map (T, x) 7→ T x is continuous from L(X , Y) × X to Y. (That
is, if Tn → T and xn → x, then Tn xn → T x.)
Proof. (Feb 4th 2018)

Let {Tn } ⊂ L(X , Y) and {xn } ⊂ X be sequences such that

lim Tn = T lim xn = x
n→∞ n→∞

Thus, for given  > 0 there is N1 , N2 ∈ N such that

 
kTn − T k < ∀n > N1 kxn − xk < ∀n > N2
2 kxk 2 kT k
And, letting N = max(N1 , N2 ), for all n ≥ N , observe the below.

kTn xn T xk ≤ kTn xn + T xn k + kT xn − T xk ≤ kT − Tn k kxk + kT k kxn − xk < 

Therefore,
lim Tn xn = T x
n→∞

5. If X is a normed vector space, the closure of any subspace of X is a subspace.

Proof. (Feb 4th 2018)

Let X be a normed vector space over a field K and Z be any subspace of X . Let x ∈ X and y ∈ accZ.
Then there exists {yn } ⊂ Z such that yn → y. And for any α, β ∈ K,
αx + βy = lim (αx + βyn ) ∈ Z ∵ {αx + βyn }∞
n=1 ⊂ Z
n→∞

And if x, y ∈ accZ, there exist {xn }, {yn } ⊂ Z such that xn → x and yn → y.

αx + βy = lim (αxn + βyn ) ∈ Z {αxn + βyn }∞
n=1 ⊂ Z
n→∞

If x, y ∈ Z, clearly, αx + βy ∈ Z ⊂ Z.
Therefore, Z is a subspace of X . 

6. Suppose that X is a finite-dimensional vector space. Let e1 , . . . , en be a basis for X , and define
kΣn1 aj ej k1 = Σn1 |aj |.

a. k · k1 is a norm on X .
b. The map (a1 , . . . , an ) 7→ Σn1 aj ej is continuous from K n with the usual Euclidean topology to X with
the topology defined by k · k1 .
c. {x ∈ X : kxk1 = 1} is compact in the topology defined by k · k1 .
d. All norms on X are equivalent. (Compare any norm to k · k1 )
 5

Proof. (Feb 4th 2018)

Let X be a vector space over a field K.

a. For any Σnj=1 aj ej , Σnj=1 bj ej ∈ X and α ∈ K, observe the below.

n n
n n n n
X X X X X X
aj ej + bj ej = (aj + bj )ej = |aj + bj | ≤ |aj | + |bj |



j=1 j=1 1 j=1 j=1 j=1 j=1
n 1n
X X
= aj e j + bj e j


j=1 1 j=1 1

Xn Xn n
X n
X Xn
α a e = αa e = |αa | = |α| |a | = |α| a e

j j j j j j j j

j=1 1 j=1 1 j=1 j=1 j=1 1

Xn n
X n
X
a e = 0 ⇐⇒ |a | = 0 ⇐⇒ a = 0 ∀j ⇐⇒ aj e j = 0

j j j j

j=1 1 j=1 j=1

Therefore, k·k1 is a norm.

b. Let  > 0 be given and let δ = n . Then, for any (a1 , . . . , an ), (b1 , . . . , bn ) ∈ X observe the below.
k(a1 , . . . , an ) − (b1 , . . . , bn )k = k(a1 − b1 , . . . , an − bn )k < δ
n n

n
X X X
=⇒ aj e j − bj ej = |aj − bj | ≤ n max |aj − bj | ≤ nk(a1 − b1 , . . . , an − bn )k < 

1≤j≤n
j=1 j=1 1 j=1

Therefore, the mapping is continuous.

c. Let A = {x ∈ X : kxk1 = 1} and k · k1 = f . Then note that kxk1 ≤ 1 for all x ∈ A. Thus, A
is bounded. Also, note that f −1 ({1}) = A and {1} is closed since f is continuous. Thus, A is closed.
Therefore, A is closed and bounded, so it is compact.

d. Let k · k be a given norm on X . And observe the below.

n n
n
X X X
aj ej ≤ max kej k |aj | ≤ max kej k aj e j


1≤j≤n 1≤j≤n
j=1 j=1 j=1 1

Conversely, due to c, A is compact. Let x ∈ X be a vector such that kxk = minz∈A kzk since k · k is
continuous. Observe the below.

n n n
n
X
X X X ae

aj ej kxk ≤

aj ej
P j j = a e

j j
n
j=1 aj ej

j=1 1 j=1 1 j=1
j=1
1

Therefore, all norms on X are equivalent.



7. Let X be a Banach space.

6

a. If T ∈ L(X , X ) and kI − T k < 1 where I is the identity operator, then T is invertible; in fact, the
series Σ∞ n
0 (I − T ) converges in L(X , X ) to T
−1
.
b. If T ∈ L(X , X ) is invertible and kS − T k < kT −1 k−1 , then S is invertible. Thus, the set of invertible
operators is open in L(X , X ).
Proof. (Feb 5th 2018)

a. First, note that ∃α such that kI − T k ≤ α < 1, and observe the below.

X ∞ X ∞ X∞
n n
(I − T ) ≤ kI − T k < αn < ∞


n=0 n=0 n=0

Therefore, the series converges absolutely. Note that L(X , X ) is complete since X is complete. Thus,
the series below is convergent.
∞
X Xn
let n
L= (I − T ) = lim (I − T )k ∈ L(X , X )
n→∞
n=0 k=0

Now, observe the below.

n n n
" #
X X X
LT = lim (I − T )k T = lim k
(I − T ) (I + (T − I)) = lim (I − T )k − (I − T )k+1
n→∞ n→∞ n→∞
k=0 k=0 k=0
= lim I − (I − T )n+1
 
n→∞
n n
" n #
X X X
T L = lim T (I − T )k = lim (I + (T − I))(I − T )k = lim (I − T )k − (I − T )k+1
n→∞ n→∞ n→∞
k=0 k=0 k=0
= lim I − (I − T )n+1
 
n→∞

Since

lim I − I − (I − T )n+1 = lim (I − T )n+1 ≤ lim kI − T kn+1 < lim αn+1 = 0

 
n→∞ n→∞ n→∞ n→∞

, LT = T L = I which means T is bijection and L = T −1 . Lastly, observe there is C ≥ 0 such the below.
−1
T x ≤ C kxk ∀x ∈ X
Observe the below.

kxk = T −1 T x ≤ C kT xk ∀x ∈ X

Therefore, T is invertible.

−1
b. Suppose that T ∈ L(X , X ) is invertible and kS − T k < kT −1 k . And observe the below.
I − ST −1 ≤ kT − Sk T −1 = kS − T k < 1

kT −1 k−1
Since ST −1 ∈ L(X , X ), by a, ST −1 is invertiable, so T S −1 ∈ L(X , X ) which means S −1 = T −1 (T S −1 ) ∈
L(X , X ). Thus, S is bijection. Also, there is CST −1 ≥ 0 such that kST −1 xk ≥ CST −1 kxk for all x ∈ X .
Since T is invertible, there is C T ≤ 0 such that kT xk ≥ C kxk for all x ∈ X . It implies the below.
kSxk = ST −1 T x ≥ CST −1 kT xk ≥ CST −1 C kxk ∀x ∈ X

Therefore, S is invertible. 
 7

8. Let (X, M) be a measurable space, and let M (X) be the space of complex measures on (X, M). Then
kµk = |µ|(X) is a norm on M (X) that makes M (X) into a Banach space. (Use Theorem 5.1.)
Proof. (Feb 5th 2018)

N kµk = |µ|(X) is a norm.

(→)
For any µ, ν ∈ M (X) and α ∈ C, observe the below.

kµ + νk = |µ + ν|(X) ≤ |µ|(X) + |ν|(X) = kµk + kνk ∵ Prop. 3.14

kαµk = |αµ|(X) = |α||µ|(X) = |α| kµk
kµk = 0 ⇐⇒ |µ|(X) = 0 ⇐⇒ µ ≡ 0
Thus, the operation k·k is a norm on M (X), so M (X) is a normed vector space.

N M (X) is a complete space.

(→)
Let {µn }∞
n=1 ⊂ M (X) is a sequence such the below.
∞
X
kµn k < ∞
n=1

Let
∞
X
µ= µn
n=1

Then, for a sequence of disjoint sets {Xn }∞

n=1 ⊂ M, observe the below.
∞
X ∞
X
µ(∅) = µn (∅) = 0=0
n=1 n=1
∞
[ ∞
X [∞ ∞ X
X ∞ ∞ X
X ∞ ∞
X
µ( Xk ) = µn ( Xk ) = µn (Xk ) = µn (Xk ) = µ(Xk )
k=1 n=1 k=1 n=1 k=1 k=1 n=1 k=1

Therefore, µ ∈ M (X), so the series converges in M (X). Thus, by Proposition 5.1, M (X) is complete,
so is Banach space.


9. Let C k ([0, 1]) be the space of functions on [0, 1] possessing continuous derivatives up to order k on [0, 1],
including one-sided derivatives at the endpoints.
a. If f ∈ C([0, 1]), then f ∈ C k ([0, 1]) iff f is k times continuously differentiable on (0, 1) and limx&0 f (j) (x)
and limx%1 f (j) (x) exist for j ≤ k. (The mean value theorem is useful.)
Pk (j)
b. kf k = 0 f u is a norm on C k ([0, 1]) that makes C k ([0, 1]) into a Banach space. (Use induction

on k. The essential point is that if {fn } ⊂ C 1 ([0, 1]), fn → f uniformly, and fn0 R→ g uniformly, then
x
f ∈ C 1 ([0, 1]) and f 0 = g. The easy way to prove this is to show that f (x) − f (0) = 0 g(t)dt.)
Proof. (Feb 13rd 2018)

a. If f ∈ C k ([0, 1]), then clearly, by definition, f is k times continuously differentiable on [0, 1], so on
(0, 1). Also, it is clear that limx&0 f (j) (x) and limx%1 f (j) (x) exist for j ≤ k from the definition.
8

Conversely, suppose that f is k times continuously differentiable on (0, 1) and limx&0 f (j) (x) and limx%1 f (j) (x)
exist for j ≤ k. We need to verify if f is differentiable on the end points. Let’s define two values as below.
L := lim f (j) (x) R := lim f (j) (x) ∀j
x&0 x%1

Let  > 0 be given. Then there exist δL > 0 and δR > 0 such that
x − 0 < δL =⇒ |f (j) (x) − L| < 
1 − x < δR =⇒ |f (j) (x) − L| < 
By mean value theorem, ∀x ∈ (0, δL ) and ∀x ∈ (1 − δR , 1) there exist xL ∈ (0, δL ) and xR ∈ (1 − δR , 1)
such that
f (j−1) (x) − f (j−1) (0)
= f (j) (xL )
x−0
f (j−1) (x) − f (j−1) (1)
= f (j) (xR )
x−1
Therefore, ∀x ∈ (0, δL ) and ∀x ∈ (1 − δR , 1)
 (j−1) (j−1)

f (x) − f (0)   
 − L = f (j) (xL ) − L < 
 x−0
and
 (j−1)
(x) − f (j−1) (1)

f   
 − R = f (j) (xR ) − R < 
 x−1
Thus, each f (j) is differentiable on end points of [0, 1].

b. The given k·k is clearly a norm as can be observed as below.

For any f, g ∈ C k ([0, 1]) and λ ∈ C,

k
X k
X
(j) (j)
kλf k = λf =
u
|λ| f = |λ| kf k
u
j=0 j=0
k
X k
X k
X
(j) (j) (j)
kf + gk = f + g (j) ≤ f + g = kf k + kgk
u u u
j=0 j=0 j=0

kf k = 0 ⇐⇒ f (j) u = 0 ∀j ⇐⇒ f = 0
Thus, k·k is a norm.

Now we want to check if C k ([0, 1]) is a Banach Space with respect to the norm k·k. So, let {fn }∞
n=1 ⊂
k
C ([0, 1]) is a Cauchy sequence with the given norm. 

10. Let L1k ([0, 1]) be the space of all f ∈ C k−1 ([0, 1]) such that f (k−1) is absolutely continuous on [0, 1]
P R1
(and hence f (k) exists a.e. and is in L1 ([0, 1])). Then kf k = k0 0 |f (j) (x)|dx is a norm on L1k ([0, 1]) that
makes L1k ([0, 1]) into a Banach space.(See Exercise 9 and its hint.)
Proof. (March 8th 2018)
Pk R 1
(1) kf k = 0 0
|f (j) (x)|dx is a norm on L1k ([0, 1])
( =⇒ )
 9

k Z
X 1 k Z
X 1 k Z
X 1
(j) (j)
kλf k = |λf (x)|dx = |λ||f (x)|dx = |λ| |f (j) (x)|dx = |λ| kf k ∀λ ∈ C
j=0 0 j=0 0 j=0 0

k Z
X 1 k Z
X 1 Z 1 
(j) (j) (j) (j)
kf + gk = |f (x) + g (x)|dx ≤ |f (x)|dx + |g (x)|dx
0 0 j=0 0 0

k Z
X 1 k Z
X 1
(j)
= |f (x)|dx + |g (j) (x)|dx
j=0 0 j=0 0

= kf k + kgk ∀f, g ∈ L1k ([0, 1])

k Z
X 1
kf k = 0 ⇐⇒ |f (j) (x)|dx = 0
0 0
(j)
⇐⇒ f ≡ 0 ∀j ≤ k
⇐⇒ f ≡ 0

(2) L1k ([0, 1]) is a Banach Space with the given norm.
( =⇒ )
First of all, note that
Xk Z 1 k
X
(j)
(j)
kf k = |f (x)|dx = f 1
L
k=0 0 j=0

Suppose that {fn } ⊂ L11 ([0, 1]) is a Cauchy sequence. And observe the below.
kfn − fm k = kfn0 − fm
0
kL1 ([0,1]) + kfn − fm kL1 ([0,1])

Note that each fn is in L1 since fn is continuous on compact interval.

Thus, {fn } and {fn0 } are also Cauchy in L1 ([0, 1]). Since L1 ([0, 1]) is a Banach space, ∃f, g ∈ L1 ([0, 1])
such that fn → f and fn0 → g in L1 ([0, 1]).
Since each fn is absolutely continuous,
Z x Z x
0
fn (x) − fn (0) = fn dm = fn0 (y)dy
0 0

Let  > 0 be given and N = max(N1 , N2 ) where N1 , N2 ∈ N satisfies the following condition .

 
kfn − f kL1 < ∀n ≥ N1 kfn0 − gkL1 < ∀n ≥ N2
4 4
Then observe the below.

1 1 1
Z Z Z
fN0 (y)dy

f (x) − f (0) − g(y)dy ≤ kf (x) − f (0) − fN (x) + fN (0)kL1 + − g(y)dy

0 L1 0 0 L1
  
< + kkfN0 − gkL1 kL1 = + kfN0 − gkL1 <
4 4 2
10

Since  > 0 is arbitrary,

x x x
Z Z Z
0

f (x) − f (0) − g(y)dy = 0 =⇒ f (y)dy = f (x) − f (0) = g(y)dy a.e.

0 L1 0 0

Therefore, f 0 = g a.e., so

kf − fn k = kf − fn kL1 + kf 0 − fn0 kL1 ≤ + kf 0 − gkL1 + kg − fn0 kL1 <  ∀n ≥ N
4
Therefore, L11 ([0, 1]) is a Banach space.

Now suppose that L1k ([0, 1]) is a Banach space. And we need to show L1k+1 ([0, 1]) is a Banach space.
Suppose {fn } is a Cauchy sequence with respect to the norm
k+1
X (j)
kf k = f 1
L
j=0

(j)
Again, since L1 ([0, 1]) is a Banach Space and {fn } ⊂ L1k+1 ([0, 1]) ⊂ L1k ([0, 1]) is a Cauchy sequence in
L1 ([0, 1]),

∀j ∃gj such that fn(j) → gj

Therefore, by inductive assumption, gj = f (j) ∀j ≤ k.

Then with the same procedure with the above, since f (k+1) is absolutely continuous, gk+1 = f (k+1) a.e.,
so ∃M ∈ N such that kfn − f k <  ∀n ≥ M . In other words, fn → f in L1k+1 ([0, 1]), so it is a Banach Space.

In conclusion, ∀k ∈ N, L1k ([0, 1]) is a Banach space. 

11. If 0 < α ≤ 1, let Λα ([0, 1]) be the space of Holder continuous functions of exponent α on [0, 1]. That
is, f ∈ Λα ([0, 1]) iff kf kΛα < ∞, where

|f (x) − f (y)|
kf kΛα = |f (0)| + sup
x,y∈[0,1],x6=y |x − y|α

a. k·kΛα is a norm that makes Λα ([0, 1]) into a Banach space.

b. Let λα ([0, 1]) be the set of all f ∈ Λα ([0, 1]) such that

|f (x) − f (y)|
→ 0 as x → y ∀y ∈ [0, 1]
|x − y|α

If α < 1, λα ([0, 1]) is an infinite-dimensional closed subspace of Λα ([0, 1]). If α = 1, λα ([0, 1]) contains
only constant functions.

Proof. (Feb 24th 2018)

a.
N k·kΛα is a norm.
 11

( =⇒ )

|f (x) − f (y)|
kf kΛα = 0 ⇐⇒ |f (0)| + sup =0
x,y∈[0,1],x6=y |x − y|α
|f (x) − f (y)|
⇐⇒ f (0) = 0 ∧ = 0 ∀x, y ∈ [0, 1] with x 6= y
|x − y|α
⇐⇒ |f (0)| = 0 ∧ |f (x) − f (y)| = 0 ∀x, y ∈ [0, 1]
⇐⇒ f (x) = 0 ∀x ∈ [0, 1]
⇐⇒ f ≡ 0 on [0, 1]

For any f, g ∈ Λα ([0, 1]),

|f + g(x) − f + g(y)|
kf + gkΛα = |f (0) + g(0)| + sup
x,y∈[0,1],x6=y |x − y|α
|f (x) − f (y)| |g(x) − g(y)|
≤ |f (0)| + |g(0)| + sup α
+ sup
x,y∈[0,1],x6=y |x − y| x,y∈[0,1],x6=y |x − y|α
= kf kΛα + kgkΛα

For any f ∈ Λα ([0, 1]) and β ∈ C,

|βf (x) − βf (y)|
kβf kΛα = |βf (0)| + sup
x,y∈[0,1],x6=y |x − y|α
|f (x) − f (y)|
= |β||f (0)| + |β| sup
x,y∈[0,1],x6=y |x − y|α
!
|f (x) − f (y)|
= |β| |f (0)| + sup
x,y∈[0,1],x6=y |x − y|α
= |β| kf kΛα

N Λα ([0, 1]), kkΛα is a Banach space.

( =⇒ ) Let {fn }∞
n=1 ⊂ Λα be a Cauchy Sequence. Then, for given  > 0, ∃N ∈ N such that

|fn (x) + fm (x) − fn (y) − fm (y)|

n, m > N =⇒ kfn − fm kΛα = |fn (0) − fm (0)| + sup <
x,y∈[0,1],x6=y |x − y|α

Which means

n, m > N =⇒ |fn (x) − fm (0)| < 

Thus, {fn (0)}∞

n=1 is a Cauchy sequence in C, so there is f (0) such that

lim fn (0) = f (0)

n→∞
12

Also, noting that for any x ∈ [0, 1] there is M such that

|fn (x) + fm (x) − fn (y) − fm (y)|  
n, m > M =⇒ |fn (0) − fm (0)| + sup α
< and |fn (0) − fm (0)| <
x,y∈[0,1],x6=y |x − y| 2 2
 
=⇒ |fn (x) − fm (x)| − |fn (0) − fm (0)| ≤ |fn (x) + fm (x) − fn (0) − fm (0)| ≤ |x|α ≤
2 2
=⇒ |fn (x) − fm (x)| < 
Thus, there is f (x) such that
lim fn (x) = f (x) ∀x ∈ [0, 1]
n→∞

Clearly, f ∈ Λα ([0, 1]) since for any x, y ∈ [0, 1] there is N ∈ N such that
|f (x) + fN (x) − f (y) − fN (y)| |f (x) − fN (x)| + |fN (y) − f (y)| |x − y|α + |x − y|α
≤ < =1
|x − y|α |x − y|α |x − y|α
Thus,
|f (x) − f (y)| |fN (x) − fN (y) + f (x) − f (y)| + |fN (x) − fN (y)| 1 + |fN (x) − fN (y)|
α
≤ α
< <∞
|x − y| |x − y| |x − y|α

Similarly,
|f (0)| ≤ |fN (0) − f (0)| + |fN (0)| ≤ 1 + |fN (0)| < ∞

Therefore, f ∈ Λα ([0, 1]), so Λα ([0, 1]) is a Banach Space.

b.

N When α < 1

Let f ∈ acc(λα )([0, 1]), then there is a sequence {fn } ⊂ λα ([0, 1]) such that
lim kfn − f kΛalpha = 0
n→

And observe the below.

|f (x) − f (y)| limn→∞ |fn (x) − fn (y)|

lim α
= lim
x→y |x − y| x→y |x − y|α
|fn (x) − fn (y)|
= lim lim
n→∞ x→y |x − y|α
= lim 0
n→0
=0
Thus, f ∈ λα ([0, 1]) so λα ([0, 1]) is closed.

N When α = 1
 13

Note that for any f ∈ λ1 ([0, 1]) and any y ∈ [0, 1]

|f (x) − f (y)|
lim = f 0 (y) = 0
x→y |x − y|
And it means that f is constant function. Thus, λ1 ([0, 1]) consists only of constant functions. 

12. Let X be a normed vector space and M a proper closed subspace of X .

a. kx + Mk = inf y∈M kx + yk is a norm on X

b. For any  > 0 there exists x ∈ X such that kxk = 1 and kx + Mk ≥ 1 − 

c. The projection map π(x) = x + M from X to X /M has norm 1.

d. If X is complete, so is X /M. (Use Theorem 5.1)

e. The topology defined by the quotient norm is the quotient topology as defined in Exercise 28 in §4.2.
Proof. (Feb 25th 2018)

a.

kx + Mk = 0 ⇐⇒ inf kx + yk = 0
y∈M
⇐⇒ ∃y ∈ M such that y = −x
⇐⇒ x ∈ M
⇐⇒ x + M = 0 + M

For ∀x ∈ X /M and ∀a ∈ C
y
kax + Mk = inf kax + yk = |a| inf x + = |a| inf kx + yk = |a| kx + Mk

y∈M y∈M a y∈M

∀x, y ∈ X /M,
kx + y + Mk = inf kx + y + zk
z∈M
z z
≤ inf x + + inf y +

z∈M 2 z∈M 2
= inf kx + zk + inf ky + zk
z∈M z∈M
= kx + Mk + ky + Mk
Therefore, kx + Mk is a norm on X /M.

b.
14

If  ≥ 1, it is trivial. Thus, let’s assume that 1 >  > 0 be given. Then, for x ∈ X /M observe the
below.
kx + Mk kx + Mk k(x + y) + Mk
kx + Mk ≤ =⇒ ∃y ∈ M such that kx + yk ≤ =
1− 1− 1−
x+y
=⇒ 1 −  ≤ kx + yk + M

x+y
And kx+yk = 1.

c.

Observe that π is linear.

∀a, b ∈ C ∀x, y ∈ X
π(ax + by) = ax + by + M = ax + M + by + M = a(x + M) + b(y + M) = π(ax) + π(by)

And it is bounded, for any x ∈ X , due to below.

kx + Mk
kx + Mk = inf kx + yk ≤ kx + 0k = kxk =⇒ kπk = sup ≤1
y∈M x6=0 kxk
Also, from b.
kπk = sup kx + Mk ≥ 1
kxk=1

Therefore, kπk = 1.

d.

Suppose X is complete and that

∞
X
(xn + M)
n=1

is a absolutely convergent series in X /M. Then, there exists y ∈ M such that

∞ ∞  
X X 1
kxn + yk < kxn + Mk + n < ∞
n=1 n=1
2
Thus, {xn + y}∞n=1 ⊂ X is a sequence such that the series corresponding to the sequence is absolutely
convergent. Since X is complete, the series converges to (let)x ∈ X .

∞ ∞
! ∞ ∞
!
X X X X
(xn + M) = xn +M= (xn + y) − y +M=x+M
n=1 n=1 n=1 n=1

Thus, the absolutely convergent series converges, so X /M is complete.

 15

e.

I will come back after solving the corresponding problem. 

13. If k·k is a seminorm on the vector space X , let M = {x ∈ X : kxk = 0}.Then M is a subspace, and
the map x + M 7→ kxk is a norm on X /M.
Proof. (Feb 25th 2018)

N M is a subspace.

Clearly, 0 ∈ M.
For any a ∈ C and any x ∈ M,

kaxk = |a| kxk = 0

Thus, ax ∈ M

And for any x, y ∈ X ,

0 ≤ kx + yk ≤ kxk + kyk = 0 + 0 = 0
Thus, x + y ∈ M.

Therefore, M is a subspace of X .

N The map x + M 7→ kxk is a norm on X /M.

Since k·k is already seminorm, and

kxk = 0 ⇐⇒ x ∈ M ⇐⇒ x + M = 0 + M
, we can conclude that k·k is a norm on X /M 

14. If X is a normed vector space and M is a nonclosed subspace, then kx + Mk, as defined in Exercise
12, is a seminorm on X /M. If one divides by its nullspace as is Exercise 13, the resulting quotient space
is isometrically isomorphic to X /M. (Cf. Exercise 5)
Proof. (Feb 25th 2018)

N kx + Mk is a seminorm on X /M

∀a ∈ C and ∀x + M ∈ X /M
y
kax + Mk = inf kax + yk = |a| inf x + = |a| inf kx + yk = |a| kx + Mk

y∈M y∈M a y∈M

∀x, y ∈ X
16

z z
kx + y + Mk = inf kx + y + zk ≤ inf x + + inf y +

z∈M z∈M 2 z∈M 2
= inf kx + yk + inf ky + zk = kx + Mk + ky + Mk
z∈M z∈M

Let N = {x ∈ X : kx + Mk = 0}, then, by Exercise 13, k·k is a norm on X /N .

Let L(x + N ) = x + M, then it is well defined and injection.


15. Suppose that X and Y are normed vector spaces and T ∈ L(X , Y). Let N (T ) = {x ∈ X : T x = 0}
a. N (T ) is a closed subspace of X .
b. There is a unique S ∈ L(X /N (T ), Y) such that T = S ◦ π where π : X → X /N (T ) is the projection
π(x) = x + N (T ). Moreover, kSk = kT k.
Proof. (Feb 21st 2018)

a.
Observe the below.
T 0 = 0 =⇒ 0 ∈ N (T )
T (αx + βy) = αT x + βT y = 0 + 0 = 0 ∀x, y ∈ N (T )∀α, β ∈ C.
Thus, N (T ) is a subspace. And it is closed since T is bounded so continuous, {0} is closed in Y, and
N (T ) = T −1 ({0})

b.
Let’s define the S : X /N (T ) → Y as below.

S(x + N (T )) = T x ∀x ∈ X

It is well defined since the below.

x + N (T ) = y + N (T ) =⇒ (x − y) ∈ N (T )
=⇒ S(x + N (T )) = T x = T y + T (x − y) = T y = S(y + N (T ))

Then note the below.

S ◦ π(x) = S(x + N (T )) = T x ∀x ∈ X
Due to the following reasoning, S is linear.
∀α, β ∈ C∀x, y ∈ X
S((αx + βy) + N (T )) = T (αx + βy) = αT x + βT y = αS(x + N (T )) + βS(y + N (T ))
Also, it is bounded because T is bounded and of the below.
 17

∀x ∈ X
kS(x + N (T ))k = kT xk ≤ kxk
Therefore, S ∈ L(X /N (T ), Y).

Moreover,

kT xk kT xk kS(π(x))k kS(π(x))k kπ(x)k

kT k = sup = sup = sup = sup
x6=0 kxk x6∈N (T ) kxk x6∈N (T ) kxk x6∈N (T ) kπ(x)k kxk
kS(π(x))k kπ(x)k
= sup sup = kSk kπk = kSk
x6∈N (T ) kπ(x)k x6∈N (T ) kxk

Also, the S is unique. If there are S1 and S2 that satisfy the given conditions,

k(S1 − S2 )(π(x))k kS1 (π(x)) − S2 (π(x))k kT x − T xk

kS1 − S2 k = sup = sup = sup =0
x6∈N (T ) kπ(x)k x6∈N (T ) kπ(x)k x6∈N (T ) kπ(x)k
=⇒ S1 = S2


16.
Proof. 

17. A linear functional f on a normed vector space X is bounded iff f −1 ({0}) is closed. (Use Exercise
12.b)
Proof. (Feb 26th 2018)

( =⇒ )

If a linear functional f is bounded, f is continuous. Thus, the inverse image of closed set {0}, f −1 ({0})
is closed (C is Normal space).

(⇐=)

Note that f −1 ({0})(=let M) is a closed subspace of X . By Exercise 12 (b), there exists x ∈ X such
that kxk = 1 and
1 1
kx + Mk ≥ 1 − =
2 2
Note that x ∈ X \ M and let Mx = Cx + M
For any y ∈ X \ Mx , observe the below.
 
f (y) f (y)
y= x+ y− x ∈ Cx + M = Mx
f (x) f (x)
Therefore, X = Mx . Also, for any x ∈ X there exists y ∈ M, observe the below.
18

1 y
|f (z)| = |f (λx + y)| = |λ||f (x)| ≤ 2|λ| |f (x)| ≤ 2|λ| kx + Mk |f (x)| ≤ 2|λ| x + |f (x)| = 2|f (x)| kλx + yk

2 λ
Therefore, f is bounded.


18. Let X be a normed vector space.

a. If M is a closed subspace and x ∈ X \ M then M + Cx is closed. (Use Theorem 5.8 a.)

b. Every finite-dimensional subspace of X is closed.

Proof. (Feb 26th 2018)

a.

If X = M, then it is trivially true. Thus, let’s assume that M is a closed proper subspace of X .

Let z ∈ acc(M + Cx), then there exists a sequence {yn + λn x}∞ ∞

n=1 ⊂ M + Cx such that {yn }n=1 ⊂ M,
∞
{λn }n=1 ⊂ C, and
lim yn + λn x = z
n→∞

Note that, by Theorem 5.8a., there is a function f ∈ X ∗ such that

f (x) 6= 0 f |M = 0

And note the below.

f (z) f (z)
lim λn = lim = ∵ f (z) = f ( lim (yn + λn x)) = lim f (yn + λn x) = lim λn f (x)
n→∞ n→∞ f (x) f (x) n→∞ n→∞ n→∞

The exchanging the function and limit is valid here since f is bounded, thus continuous.
Now, observe the below.

f (z)
z = lim (yn + λn x) = lim (yn + λn x) − lim λn x + lim λn x = lim yn + lim λn x = lim yn + x
n→∞ n→∞ n→∞ n→∞ n→∞ n→∞ n→∞ f (x)
f (z)
=⇒ lim yn z − x =let y exists
n→∞ f (x)
Since M is closed, z − ff (x)
(z)
x = y ∈ M. And consequently, z = y + f (z)
f (x)
x ∈ M + Cx.
Therefore, M + Cx is closed.

b.

Note that {0} =let M0 is closed subspace of X . And from a. we know that {0} + Cx1 =let M1 is a
closed set for any x1 ∈ X . Assuming we have selected xn ∈ X \ Mn−1 where Mn−1 is a closed subspace,
we can create Mn = Mn−1 + Cxn . From a. we know Mn is a closed subspace. Therefore, any finite
dimensional subspace of X is closed. 
 19

19. Let X be an infinite-dimensional normed vector space.

a. There is a sequence {xj } in X such that kxj k = 1 for all j and kxj − xk k ≥ 1
2
for all k 6= j. (Con-
struct xj inductively, using Exercise 12b and 18.)

b. X is not locally compact.

Proof. (Feb 27th 2018)

a.

Let M0 = {0}, then M0 is a closed subspace of X .

Then, by Exercise 12, we can choose a vector x1 ∈ X \ M0 such that kx1 k = 1 and kx1 + M0 k ≥ 12 .

k=1 ⊂ X such that kxj k = 1 and kxj − xi k ≥

Suppose that we have chosen {xk }n−1 1
2
for all i, j ∈
{1, 2, . . . , n − 1}.

Let Mn = M0 + nk=1 Cxk , and it is a closed proper subspace of X from exercise 18. Thus, again by
P
Exercise 12, we can choose xn ∈ X \ Mn such that kxn k = 1 and kxn + Mn k ≥ 12 . Noting that −xi ∈ Mn
∀i ∈ {1, 2, . . . , n − 1}, observe the below.
1
kxn − xi k ≥ kxn + Mn k ≥
2

Thus, inductively, we can construct a sequence {xn }∞

n=1 as above.

b.

Assume that X is compact. Then there exists a compact neighborhood K0 of 0. Then we have r > 0
such that B(0, r) ⊂ K0 . Since B(0, r) is closed subset of a compact set, it is compact.

Firstly, note that {rxn }∞ ∞

n=1 where {xn }n=1 is from part a satisfy the following properties.

kxj k = r ∀j ∈ N
r
kxj − xi k ≥ ∀i, j ∈ N
2

And let’s construct a open cover as below.

r
C = {B(xk , ) : k ∈ N} ∪ U
2 !c
[  r
where U = B xk ,
k∈N
4

Note that C is an open cover of B(0, r). However, there exists no finite subcover since if you miss B(xk , 2r )
for some k ∈ N, we cannot cover xk . Therefore, B(0, r) is not a compact ans it is a contradiction. 
20

20. If M is a finite-dimensional subspace of a normed vector space X , there is a closed subspace N such
that M ∩ N = {0} and M + N = X .
Proof. (Mar 11st 2018)

If X = M, then N = {0} satisfies the conditions.

Since M is finite dimensional subspace of X , let

M
M= Cui
1≤i≤n

Note that M is closed (by Exercise 18). Thus, if

M
Bj = {ui }
1≤i≤n
i6=j

by a theorem,

∀j ∈ {1, . . . , n} ∃fj ∈ X ∗ such that fj (uj ) 6= 0 and fj |Bj = 0

Now let
n
\
N = fi−1 ({0})
i=1

Suppose that x ∈ M ∩ N . Since x ∈ M, x should be a linear combination of ui ’s. However, since x ∈ N

and ui 6∈ N for each i, we have x = 0. Therefore,
M ∩ N = {0}
Also, if x ∈ X , then
n
! n
!
X fi (x) X fi (x)
x= ui + x− ui ∈M∩N
f (u )
i=1 i i
f (u )
i=1 i i

Since
n
! n
X fi (x) X fi (x) fj (x)
fj x− ui = fj (x) − fj (ui ) = fj (x) − f (uj ) = 0
i=1
f i (ui ) i=1
f i (ui ) f j (uj )
Therefore,
X =M+N


∗ ∗
21. If X and Y are normed vector spaces, define α : X ×Y → (X × Y )∗ by
α(f, g)(x, y) = f (x) + g(y).
Then α is an isomorphism which is isometric if we use the norm k(x, y)k = max(kxk , kyk) on X × Y ,
the corresponding operator norm on (X × Y )∗ , and the norm k(f, g)k = kf k + kgk on X ∗ × Y ∗ .
Proof. (March 11st 2018)

Let β : (X × Y )∗ → X ∗ × Y ∗
be defined as below.
β(F (x, y)) = (F (x, 0), F (0, y))
 21

And observe the below.

∀(x, y) ∈ X × Y ∀F : (X × Y )∗ → X ∗ × Y ∗
∀f ∈ X ∗ , g ∈ Y ∗

(α ◦ β)(F (x, y)) = α(β(F (x, y))) = α(F (x, 0), F (0, y)) = F (x, 0) + F (0, y) = F (x, y)
(β ◦ α)(f (x), g(y)) = β(α(f (x), g(y))) = β(f (x) + g(y)) = (f (x) + g(0), f (0) + g(y)) = (f (x), g(y))
Therefore, α is a bijection. Also,
kα(f, g)(x, y)k = kf (x) + g(y)k ≤ kf (x)k + kg(y)k ≤ kf k kxk + kgk kyk
= (kf k + kgk) max(kxk , kyk) = (kf k + kgk) k(x, y)k
=⇒ kα(f, g)k ≤ k(f, g)k
so, α is bounded.
Conversely, let  > 0 be given. And note that there exists x ∈ X and y ∈ Y with kxk = kyk = 1 such
that
kf k < kf (x)k + 
kgk < kg(y)k + 
Now, observe the below.
k(f, g)k = kf k + kgk < kf (x)k + kg(y)k + 2 = k|f (x)| + |g(y)|k + 2
= kf (Sf x) + g(Sg y)k + 2 = kα(f, g)(Sf x, Sg y)k + 2
where
( (
f (x) g(y)
|f (x)|
f (x) 6= 0 |g(y)|
g(y) 6= 0
Sf = Sg =
0 f (x) = 0 0 g(y) = 0

Since  is arbitrary,
k(f, g)k < kα(f, g)(Sf x, Sg y)k ≤ kα(f, g)k max(kSf xk , kSg yk) = kα(f, g)k
Therefore,
kα(f, g)k = k(f, g)k
So, α is an isometry and so is an isomorphism. 

22. Suppose that X and Y are normed vector spaces and T ∈ L(X,Y).

a. Define T † : Y ∗ → X ∗ by T † f = f ◦ T . Then T † ∈ L(Y∗ , X∗ ) and T † = kT k.

T † is called the adjoint or transpose of T .

b. Applying the construction in (a) twice, one obtain T †† ∈ L(X∗∗ , Y∗∗ ). If X and Y are identified with
their natural image X
c and Y c in X ∗∗ and Y ∗∗ , then T †† |X = T .

c. T † is injective iff the range of T is dense in Y .

d. If the range of T † is dense in X ∗ , then T is injective; the converse is true if X is reflective.

Proof. (March 11st 2018)
22

a. For any f, g ∈ Y ∗ , λ ∈ C and x ∈ X , observe the linearity of T † as below.

T † (f + g)(x) = ((f + g) ◦ T )(x) = (f ◦ T )(x) + (g ◦ T )(x) = T † (f )(x) + T † (g)(x)
=⇒ T † (f + g) = T † (f ) + T † (g)

T † (λf )(x) = (λf ◦ T )(x) = λ(f ◦ T )(x) = λT † (f )(x)

=⇒ T † (λf ) = λT † (f )
Also, it is bounded as we can observe from below.
k(f ◦ T )(x)k = kf k kT xk ≤ kf k kT k kxk
†
=⇒ T (f ) = k(f ◦ T )k ≤ kf k kT k
Therefore, T † ∈ L(Y∗ , X∗ ).
Also, ∀x ∈ X with kxk = 1
†
T = sup T † (f ) ≤ sup kf ◦ T k ≤ sup kf k kT k = kT k

kf k=1 kf k=1 kf k=1
† †
T ≥ T (k·k) = kk·k ◦ T k = kT k
∗
since k·k ∈ Y , kk·kk = 1, and
k(k·k ◦ T )(x)k = kT xk
=⇒ kk·k ◦ T k = sup k(k·k ◦ T )(x)k = sup kT xk = kT k
kxk=1 kxk=1

Therefore, T † = kT k.

b.
Note that
∗
T †† (x̂)(f ) = (x̂ ◦ T † )(f ) = x̂ ◦ (f ◦ T ) = f (T x) = Tcx(f ) ∀f ∈ Y
Thus,
T †† |X = T †† |X
c = TX = T
d

c.
Suppose that ∃y ∈ Y \ T (X ). Note that T (X ) is a subspace, so M =let T (X ) is also a subspace (by
Exercise 5). Then observe the below.
∗
∃f ∈ Y such that f (y) 6= 0 and f |M = 0
∗
Let g ∈ Y be defined as g ≡ 0. Then observe the below.
T † (f )(x) = f (T x) = 0 = g(T x) = T † (g)(x) ∀x ∈ X
Thus, T † (f ) = T † (g) but f 6= g. Therefore, T † is not an injection.

Conversely, suppose that T † is not an injection. Then ∃f 6= g(∈ Y ∗ ) such that T † (f ) = T † (g). Since
f 6= g, ∃y ∈ Y such that f (y) 6= g(y). If T (X ) = Y , ∃{xn } such that
lim T xn = y
n→∞
 23

However,
f (y) = lim f (T (xn )) = lim T † (f )(xn ) = lim T † (g)(xn ) = lim g(T xn ) = g(y)
n→∞ n→∞ n→∞ n→∞
It shows a contradiction.

d. (Need to be fixed) If T † (Y ∗ ) = X ∗ , by c, T †† is injective. And since T †† = T †† |X

c = T . Thus,
T is an injection. If X is reflexive and T is injective, then T †† is also injective and, by c, T † has dense
image. 

23. Suppose X is a Banach space. If M is a closed subspace of X and N is a closed subspace of X ∗ , let
M0 = {f ∈ X ∗ : f |M = 0} and N ⊥ = {x ∈ X : f (x) = 0 ∀f ∈ N }.(Thus, if we identify X with its
image in X ∗∗ , N ⊥ = N 0 ∩ X )
∗
a. M0 and N ⊥ are closed subspaces of X and X , respectively.

b. (M0 )⊥ = M and (N ⊥ )0 ⊃ N . If X is reflective, (N ⊥ )0 = N .

c. Let π : X → X /M be the natural projection, and define α : (X /M) → X ∗ by α(f ) = f ◦ π. Then

α is an isometric isomorphism from (X /M)∗ onto M0 , where X /M has a quotient norm.

d. Define β : X ∗ → M∗ by β(f ) = f |M ; then β induces a map β : X ∗ /M0 → M∗ as in Exercise 15,

and β is an isometric isomorphism.
Proof. 

24. Suppose that X is a Banach space.

∗
a. Let X \
c , (X ) be the natural images of X , X ∗ in X ∗∗
,X ∗∗∗ d0 = {F ∈ X
, and let X ∗∗∗
: F |X
c = 0}.
\ ∗ d0 = {0} and (X \ ∗ d0 = X ∗∗∗ .
Then (X )∩X )+X
∗
b. X is reflexive iff X is reflexive.
Proof. 

25.
Proof. 

26.
Proof. 

27. There exists meager subsets of R whose complements have Lebesgue measure zero.
Proof. 

28. The Baire category theorem remains true if X is assumed to be LCH space rather than a complete
metric space. (The proof is similar, the substitute for completeness is Proposition 4.21.)
24

Proof. 

P∞
29. Let Y = L1 (µ) where µ is counting measure on N, and let X ={f ∈ Y : 1 n|f (n)| < ∞}, equipped
with the L1 norm.

a. X is a proper dense subspace of Y ; hence X is not complete.

b. Define T : X → Y by
T f (n) = nf (n)
Then T is closed but not bounded.

c. Let S = T −1 . Then S : Y → X is bounded and surjective but not open.

Proof. (March 12nd 2018)

a.
" ∞ ∞
#
X X
f ∈ X =⇒ |f (n)| ≤ n|f (n)| < ∞ =⇒ f ∈ Y =⇒ X ⊂ Y
n=1 n=1
" ∞ ∞ ∞ ∞ ∞
#
X X X X X
f, g ∈ X =⇒ |f (n)| < ∞ ∧ |g(n)| < ∞ =⇒ |f (n) + g(n)| ≤ |f (n)| + |g(n)| < ∞
n=1 n=1 n=1 n=1 n=1
=⇒ f + g ∈ X

" ∞ ∞
#
X X
λ ∈ C ∧ f ∈ X =⇒ |λf (n)| = |λ| |f (n)| < ∞
n=1 n=1
=⇒ λf ∈ X
1
Therefore, X is a proper subspace of Y since f (n) = n2
∈ Y but f 6∈ X .

Now, we will show that X is dense in Y .

Let  > 0 and f ∈ Y be given. Note that
∞
X
∃N ∈ N such that |f (n)| < 
n=N
If we define g as below,
(
f (n) n < N
g(n) = f (n)
n
n≥N
then

∞ ∞ ∞
X X 1 X
|f (n) − g(n)| = (1 − )|f (n)| ≤ |f (n)| < 
n=1 n=N
n n=1

Therefore, X is dense in Y .

b.
 25

Suppose that fm → f and T fm (n) → g(n) where g ∈ Y . Let  > 0 be given. Then there exist
N1 , N2 ∈ N such that
∞
X 
|fm (n) − f (n)| < ∀m ≥ N1
n=1
3
∞
X 
|T fm (n) − g(n)| < ∀m ≥ N2
n=1
3
Now, let N = max(N1 , N2 ). Note that it is possible that f ∈ Y because of a. However, since if f 6∈ X ,
we cannot define T f , we need to assume that f ∈ X . Then ∃N3 ∈ N such that
∞
X 
n|f (n) − fN (n)| <
n=N
3
3

Now, for all m ≥ N , observe the below.

∞
X ∞
X ∞
X
kT f − gk = |nf (n) − g(n)| ≤ n|f (n) − fN (n)| + |T fN (n) − g(n)|
n=1 n=1 n=1
N 3 −1 ∞
X X 
< n|f (n) − fN (n)| + n|f (n) − fN (n)| +
n=1 n=N3
3
 2
< N3 +
3 3
Since  is arbitrary, T f = g.
Therefore, T is closed. However, it is not bounded since, if we define f as below,
(
1 n=N
fN (n) =
0 Otherwise
for some N ∈ N, then fN ∈ X , kf k = 1, and

∞
X
kT fN k = n|f (n)| = N kfN k
n=1

Thus, for any constant C, there exists M ∈ N such that M > C, so that
kT fM k = M kfM k > C kfM k
Therefore, T is unbounded.

c.
f (n)
Note that Sf (n) = n
, and, for any f ∈ Y , observe the below.
∞ ∞ ∞
X X |f (n)| X
kSf k = |Sf (n)| = ≤ |f (n)| = kf k
n=1 n=1
n n=1

Thus, S is bounded.
Also, observe that, for any g(n) ∈ X , ng(n) ∈ Y , and
S(ng(n)) = g(n)
Therefore, S is a surjection. However, if S is open, then T is bounded. It is a contradiction to b.

26

30. Let Y = C([0, 1]) and X = C 1 ([0, 1]), both equipped with the uniform norm.

a. X is not complete.

b. The map
 
d
:X →Y
dx
is closed (see exercise 9) but not bounded.
Proof. (March 13rd 2018)

a. nq
1
o∞ √
Consider {fn } = x+ n ⊂ X . Then fn → x uniformly due to below.
n=1
r 
 1 √  1
sup  x + − x < √ → 0

x∈[0,1]  n  n
√
√ x is a continuous function. However, the derivative is not defined at x = 0.
Clearly,
Thus, x 6∈ Y , so X is not complete.

b.
If {fn } ⊂ X and fn → f pointwisely and fn0 → g uniformly, then f 0 is continuous and g = f 0 from basic
analysis course. Thus, the map is closed.

On the other hands, assume the map is bounded so there exists C > 0 such that

df
< C kf k ∀f ∈ X
dx
But we have fn (x) = xn which satisfies

df n−1
= nx = sup nxn−1 = n
dx
x∈[0,1]

Since ∃N ∈ N such that N > C, fN is an example that makes the bounded condition fail. 

31. Let X , Y be Banach spaces and let S : X → Y be an unbounded linear map (for the existence of
which, see §5.6). Let Γ(S) be the graph of S, a subspace of X × Y .

a. Γ(S) is not complete.

b. Define T : X → Γ(S) by T x = (x, Sx). Then T is closed but not bounded.

c. T −1 : Γ(S) → X is bounded and surjective but not open.

Proof. (March 13rd 2018)

a.
If Γ(S) is complete, S is closed. Then, by the Closed Graph Theorem, S should be bounded. It is a
contradiction.

b.
 27

Suppose xn → x in X and T xn → y in Γ(S). Since X is complete, x ∈ X , therefore, y = (x, Sx) ∈ Γ(S),

so T is closed.

Also, observe the below.

kT xk = max(kxk , kSxk) ≥ kSxk
Since S is unbounded, T is also unbounded.

c. For any y = (x, Sx) ∈ Γ(S),

−1
T y = kxk ≤ max(kxk , kSxk) = kyk

Thus, T −1 is bounded. Also, since, for any x ∈ X , we have (x, Sx) ∈ Γ(S), T −1 is surjective. However,
if T −1 is open, then T is continuous, so it is bounded, it is a contradiction to b. 

32. Let k·k1 and k·k2 be norms on the vector space X such that k·k1 ≤ k·k2 . If X is complete with respect
to both norms, then the norms are equivalent.
Proof. (March 13rd 2018)

Note that X 1 = (X , k·k1 ) and X 2 = (X , k·k2 ) are Banach spaces. Further note that T : X 2 → X 1
defined by T x = x is clearly a bijective linear map. Now, for any x ∈ X , with the given condition, observe
the below.
kT xk1 = kxk1 ≤ kxk2 = kxk2
Therefore, T is bounded. Then, by the open mapping theorem, T is open, so is an isomorphism.
Therefore, there exists C > 0 such that for any x ∈ X ,

kxk2 = T −1 x 2 ≤ C kxk1

Thus, the two norms are equivalent. 

33. There is no slowest rate of decay of the terms of an absolutely convergent series;
That is, there is no sequence {an } of positive numbers such that
∞
X
an |cn | < ∞ ⇐⇒ {cn } is bounded.
n=1

(The set of bounded sequences is the space B(N) of bounded functions on N, and the set of absolutely
summable sequences is L1 (µ) where µ is counting measure on N. If such an {an } exists, consider T :
B(N) → L1 (µ) defined by T f (n) = an f (n). The set of f such that f (n) = 0 for all but finitely many n is
dense in L1 (µ) but not in B(N).)
Proof. (March 13rd 2018)

Assume that such {an } exists. And let’s define a map T : B(N) → L1 ([0, 1]) as below.
T f (n) = an f (n)
Clearly, T is a linear map. 
28

34. With reference to Exercise 9 and 10, show that the inclusion map of L1k ([0, 1]) into C k−1 ([0, 1]) is
continuous (a) by using the closed graph theorem, and (b) by direct calculation. (This is to illustrate the
use of the closed graph theorem as a labor-saving device.)
Proof. 

35. Let X and Y be Banach spaces, T ∈ L(X , Y ), N (T ) = {x : T x = 0}, and M = range(T ). Then
X /N (T ) is isomorphic to M iff M is closed.(See Exercise 15.)
Proof. (March 13rd 2018)

From exercise 15, N (T ) is closed and there exists unique S ∈ L(X /N (T ), Y ) such that T = S ◦ π
defined as below.
S(x + N (T )) = T (x)
We now show that X /N (T ) is complete so is Banach space.
Let {xn + N (T )}∞
n=1 =
let
{yn } ⊂ X /N (T ) be a given sequence satisfying the following condition.
∞
X
kyn k < ∞
n=1
Note that for any n ∈ N we can choose x0n ∈ X satisfying the condition below.
kx0n k < kyn k + kyn k = 2 kyn k
Therefore,
∞
X ∞
X
kxn k ≤ 2 kyn k < ∞
n=1 n=1
Since X is a Banach space, ∃x ∈ X such that
X∞
xn = x
n=1
Now, observe that
∞ ! ∞
! ! ∞ !
X X X
yn − (x + N (T )) = xn −x + N (T ) ≤ xn − x = 0



n=1 n=1 n=1
Therefore,
∞
X
yn = x + N (T )
n=1

so X /N (T ) is a Banach space.

( =⇒ ) Now, suppose that X /N (T ) is isomorphic to M.

Then there is an isomorphism J : X /N (T ) → M. Let x ∈ M, then ∃{xn } ⊂ M such that xn → x.
Then, since J is an isomorphism, and
−1
J xn − J −1 (x) ≤ J −1 kxn − xk

J −1 xn → J −1 x. Now, since X /N (T ) is a Banach space, J −1 x ∈ X , so JJ −1 x = x ∈ M. Therefore, M

is closed.

(⇐=) Suppose that M is closed. Thus, M is a Banach space since Y is a Banach space.
 29

Consider the map S ∈ L(X /N (T ), Y ) as beginning of this proof. Clearly, S is an injection as we can
observe from below.
S(x + N (T )) = S(y + N (T )) ⇐⇒ T (x) = T (y)
⇐⇒ T (x − y) = 0
⇐⇒ x − y ∈ N (T )
⇐⇒ x + N (T ) = y + N (T )
Also, if y ∈ M, then ∃x ∈ X such that T x = y. Thus, from below, S is a surjection.
S(x + N (T )) = T x = y
Now, since X /N (T ) and M are Banach spaces and S ∈ L(X /N (T ), M), S is the isomorphism.
Therefore, X /N (T ) and M are isomorphic. 

36. Let X be a separable Banach

Proof. 

∗
37. Let X and Y be Banach spaces. If T : X → Y is a linear map such that f ◦ T ∈ X for every
f ∈ Y ∗ , then T is bounded.
Proof. (March 14th 2018)

In order to prove it, due to the Closed Graph Theorem, it is enough to prove that T is closed.
Let {xn } ⊂ X be a sequence such that xn → x. Since X is a Banach space, x ∈ X . Now, we need to
prove that T xn → T x. For any f ∈ Y ∗ , observe the below.
∗
lim f ◦ T (xn ) = f ◦ T ( lim xn ) = f ◦ T (x) ∵f ◦T ∈X
n→∞ n→∞
∗
lim f ◦ T (xn ) = f ( lim T xn ) ∵f ∈Y
n→∞ n→∞

Let I be identity mapping and note that I ∈ Y ∗ . Thus,

T x = I ◦ T (x) = lim I ◦ T (xn ) = I( lim T xn ) = lim T xn
n→∞ n→∞ n→∞

Therefore, T is closed, so is bounded. 

38. Let X and Y be Banach spaces, and let {Tn } be a sequence in L(X , Y ) such that lim Tn x exists for
every x ∈ X . Let T x = lim Tn x, then T ∈ L(X , Y ).
Proof. (March 14th 2018)

Clearly, T is linear as below.

T (x + y) = lim Tn (x + y) = lim Tn x + lim Tn y = T x + T y ∀x, y ∈ X
n→∞ n→∞ n→∞
T (αx) = lim Tn (αx) = α lim Tn (x) = αT (x) ∀x ∈ X , α ∈ C
n→∞ n→∞

Also, note that

lim Tn x exists ∀x ∈ X
n→∞

Thus, for any x ∈ X , {Tn x}∞

n=1 is bounded. In other words,

sup kTn xk < ∞ ∀x ∈ X

n∈N
30

Then, the Uniform Boundedness principle,

lim sup kTn k ≤ sup kTn k < ∞
n→∞ n∈N
Now, there exists C > 0 such that for any x ∈ X

kT xk = lim Tn x = lim kTn xk = lim sup kTn xk ≤ lim sup kTn k kxk ≤ C kxk

n→∞ n→∞ n→∞ n→∞

Therefore, T ∈ L(X , Y ). 

39. Let X ,Y ,Z be Banach spaces and let B : X × Y → Z be a separately continuous bilinear

map; that is, B(x, ·) ∈ L(Y , Z ) for each x ∈ X and B(·, y) ∈ L(Y , Z ) for each y ∈ Y . Then
B is jointly continuous, that is, continuous from X × Y to Z . (Reduce the problem to proving that
kB(x, y)k ≤ C kxk kyk for some C > 0.)
Proof. (March 14th 2018)

Let B1 (x) = B(x, ·) and B2 (y) = B(·, y), then B(x, y) = B1 ◦ 

40.
Proof. 

41.
Proof. 

42.
Proof. 

43.
Proof. 

44.
Proof. 

45.
Proof. 

46.
Proof. 

47.
 31

Proof. 

48.
Proof. 

49.
Proof. 

50.
Proof. 

51.
Proof. 

52.
Proof. 

53.
Proof. 

54.
Proof. 

55.
Proof. 

56.
Proof. 

57.
Proof. 

58.
Proof. 

59.
32

Proof. 

60.
Proof. 

61.
Proof. 

62.
Proof. 

63.
Proof. 

64.
Proof. 

65.
Proof. 

66.
Proof. 

67.
Proof. 

68.
Proof. 

69.
Proof. 

70.
Proof. 

71.
 33

Proof. 

72.
Proof. 

73.
Proof. 

74.
Proof. 

75.
Proof. 

76.
Proof. 

77.
Proof. 

78.
Proof. 

79.
Proof. 

80.
Proof. 

81.
Proof. 

82.
Proof. 
34

83.
Proof. 

84.
Proof. 

85.
Proof. 

86.
Proof. 

87.
Proof. 

88.
Proof. 

89.
Proof. 

90.
Proof. 

91.
Proof. 

92.
Proof. 

93.
Proof. 

94.
Proof. 
 35

95.
Proof. 

96.
Proof.