Assignment 3

AMAT 617

Mark Girard
7 March 2014

Problem 1 (Problem 4, Chapter 4.1, p. 212). Find all maximal elements of M with respect to the partial
ordering m ≤ n whenever m divides n, where M is
(a) {2, 3, 4, 8}

(b) the set of all prime numbers.

Solution. .
(a) The maximal elements of M = {2, 3, 4, 8} are 3 and 8. Indeed, 3 does not divide any of the other
elements and 2 and 4 both divide 8, wheras 8 does not divide any of the other elements.

(b) Every element of M = {p | p prime} is maximal, since no prime number divides any other prime number.

1
Problem 2 (Problem 8, Chapter 4.2, p. 218). If a subadditive functional defined on a normed space X is
nonnegative outside of a sphere {x | kxk = r}, show that it is nonnegative for all x ∈ X.
Solution. Let p be a subadditive functional on X such that p(x) is nonnegative for all x ∈ X with kxk > r.
Note that p(y) = p(y + 0) ≤ p(y) + p(0) for all y ∈ X, and thus

0 = p(y) − p(y) ≤ p(0).

Hence 0 ≤ p(0) and thus p(0) is nonnegative. Let x ∈ X such that x 6= 0. If kxk > r then p(x) is nonnegative
r
by assumption, so suppose kxk ≤ r. Then there is an n ∈ N such that n > kxk and thus knxk > r. Then

p(x + · · · + x) ≤ p(x) + · · · + p(x) = np(x),

| {z } | {z }
n times n times

1
and thus n p(nx) ≤ p(x). But p(nx) ≥ 0, so p(x) ≥ 0 as well.

2
Problem 3 (Problem 8, Chapter 4.3, p. 224). Let X be a nomed space and X 0 its dual space. If X 6= {0},
show that X 0 cannot be {0}.
Solution. Let x ∈ X such that x 6= 0. By the Hahn-Banach Theorem, there exists a functional f ∈ X 0 such
that f (x) = kxk and kf k = 1. Hence X 0 6= {0}.

3
Problem 4 (Problem 8, Chapter 4.5, p. 238). Let X and Y be normed spaces and T ∈ B(X, Y ) such that
−1
×
T −1 ∈ B(Y, X) exists. Show that (T × ) = T −1 .
Solution. First note that (1X )× = 1X 0 for any normed space X, where 1X is the identity mapping on X.
Indeed, for any f ∈ X 0 and x ∈ X we have

(1X )× f (x) = f (1X x) = f (x)

and thus (1X )× f = f for all X 0 .

Recall that, for operators S and T , we have (ST )× = T × S × . Take S = T −1 , then

1X 0 = (1X )× = (T −1 T )× = T × (T −1 )× ,

so (T −1 )× is the right-inverse of T × . Similarly,

1Y 0 = (1Y )× = (T T −1 )× = (T −1 )× T × ,
−1
×
so (T −1 )× is also the left-inverse of T × . Hence, T × is invertible and (T × ) = T −1 .

4
Problem 5 (Problem 10, Chapter 4.5, p. 239 – Annihilator). Let B be a subset of the dual space X 0 of
a nomed space X. The annihilator a B of B is defined to be
a
B = {x ∈ X | f (x) = 0 for all f ∈ B} .

Let T : X −→ Y be a bounded linear operator. Show that R(T ) ⊂ a N (T × ). What does this mean with
respect to the task of solving an equation T x = y?
Solution. Let f ∈ N (T × ), then T × f = 0. That is,

(T × f )(x) = f (T x) = 0 for all x ∈ X.

Since this holds for all f ∈ N (T × ), we have T x ∈ a N (T × ) for all x ∈ X and thus R(T ) ⊂ a N (T × ).

This means that if there exists a functional f ∈ N (T × ) such that f (y) 6= 0, then y 6∈ a N (T × ) and thus y
is not in the range of T . Hence T x = y has no solution. This is equivalent to:

if f (y) 6= 0 for some functional f ∈ Y 0

such that f (T x) = 0 for all x ∈ X,
then T x = y has no solution.

5
Problem 6 (Problem 4, Chapter 4.6, p. 246). Show that a Banach space X is reflexive if and only if its
dual space X 0 is reflexive. (Hint: Show that a closed subspace of a reflexive Banach space is reflexive.)
Solution.
Lemma 1. Any closed subspace of a reflexive Banach space is reflexive.

Proof. Let X be a reflexive Banach space and Y ⊂ X be a closed subset. Let ϕ ∈ Y 00 and define ϕ̃ ∈ X 00 by

ϕ̃(f ) = ϕ(f |Y ) for all f ∈ X 0 .

Since X is reflexive, ϕ̃ is of the form ϕ̃ = ψx for some x ∈ X where

ψx (f ) = f (x) for all f ∈ X 0 .

So ϕ(f |Y ) = f (x). We claim that x ∈ Y . Indeed, otherwise x 6∈ Y and there exists a bounded linear
functional on X such that f |Y = 0 and f (x) 6= 0 (see Lemma 4.6-7 in Kreyszig). But this is a contradiction
to the fact that
f (x) = ϕ(f |Y ) = ϕ(0) = 0,
since f |Y = 0. By the Hahn-Banach Theorem, every linear functional g on Y ⊂ X can be written as g = f |Y
for some f ∈ X 0 , and thus

ϕ(g) = ϕ(f |Y ) = ϕ̃(f ) = hx (f ) = f (x) = f |Y (x) = g(x)

where x ∈ Y . Hence for each ϕ ∈ Y 00 there is an x ∈ Y such that ϕ(g) = g(x), and thus Y is reflexive.
Proposition 2. A Banach space X is reflexive if and only if its dual space X 0 is reflexive.
Proof. Assume that X is reflexive and let C1 : X −→ X 00 be the canonical isometry. Consider X 000 = (X 00 )0
and let C2 : X 0 −→ X 000 be the canonical embedding that maps f ∈ X 0 to functional ξf ∈ X 000 such that
ξf (ϕ) = ϕ(f ) for all ϕ ∈ X 00 . Let ξ ∈ X 000 and define a functional fξ ∈ X 0 by

fξ (x) = ξ(C1 (x)).

Since X is reflexive, for each ϕ ∈ X 00 there is an x ∈ X such that ϕ = C1 (x). Hence

(C2 (fξ ))(ϕ) = ϕ(fξ ) = fξ (x) = ξ(C1 (x)) = ξ(ϕ)

and thus ξ = C2 (fξ ). So the embedding C2 is surjective and is therefore an isometry, so X 0 is reflexive.
Now suppose that X 0 is reflexive. By the argument above, we have that X 00 is reflexive as well. Consider
R(C1 ) ⊂ X 00 , which is isometric to X. Since R(C1 ) is closed in X 00 , by the lemma above we have that R(C1 )
and thus X are also reflexive.

6
Problem 7 (Problem 6, Chapter 4.6, p. 246). Show that different closed subspaces Y1 and Y2 of a normed
space X have different annihilators.
Solution. Proof. Let Y1 , Y2 ⊂ X be two closed subspaces of X with Y1 6= Y2 . Without loss of generality, we
may assume that Y1 r Y2 6= ∅, then let y ∈ Y1 r Y2 . Since y 6∈ Y2 , by the Hahn-Banach Theorem (see Lemma
4.6-7) there exists a functional f ∈ X 0 such that f (y) 6= 0 and f |Y2 = 0. Thus f ∈ (Y2 )a but f 6∈ (Y1 )a ,
hence (Y1 )a 6= (Y2 )a .

7
Problem 8 (Problem 8, Chapter 4.6, p. 246). Let M be any subset of a normed space X. Show that an
x0 ∈ X is an element of A = span M if and only if f (x0 ) = 0 for every f ∈ X 0 such that f |M = 0.
Solution. Using the ‘annihilator’ notation, this is equivalent to saying that span M = a (Ma ).
Proof. Now suppose that x0 6∈ A = span M . By the Hahn-Banach Theorem (see Lemma 4.6-7), there is
a linear functional f ∈ X 0 such that f (x0 ) 6= 0 and f |A = 0. But M is a subset of A, so f |M = 0 with
f (x0 ) 6= 0 and thus x0 6∈ a (Ma ).

8
Problem 9 (Problem 4, Chapter 4.7, p. 254). Find a meager dense subset in R2 .
Solution. Note that Q is meager and dense in R. Futhermore, Q is countable and we may enumerate it as
Q = {q1 , q2 , q3 , . . . }. Consider the subset M ⊂ R2 defined by
∞
[
M= Mi , where Mi = {(qi , x) | x ∈ R} .
i=1

That is, M consists of the union of parallel lines intersecting the x-axis at each rational point. Then M is
clearly dense in R2 , and each Mi is rare is R2 . Since M is a countable union of rare sets, it is meager in R2 .

9
Problem 10 (Problem 6, Chapter 4.7, p. 255). Show that the complement M c of a meager subset M of a
complete metric space X is nonmeager.
Solution. If X = ∅ then this is true vacuously. Let X 6= ∅ be a complete metric space and let M ⊂ X be a
meager subset. Note that we may write X as a union

X = M ∪ M c.

If M c were meager, then it may be written as a countable union of rare sets. Hence X is a countable union
of rare sets (since it is a union of two sets that are countable unions of rare sets), and thus X is meager
in itself. But this is a contradiction to Baire’s Category Theorem, which states that a nonempty complete
normed space must be non-meager in itself.

10
Problem 11 (Problem 8, Chapter 4.7, p. 255). Show that completeness of X is essential in the statement
of the Uniform Boundedness Theorem. (Hint: Consider X ⊂ `∞ consisting of all finite sequences x = {xj },
that is, xj nonzero for only finitely many j, and consider the operator Tn x = fn x = nxn .)
Solution. Recall the statement of the theorem.

Theorem 3 (Uniform Boundedness Theorem). Let {Tn } be a sequence of bounded linear operators Tn :
X −→ Y from a Banach space X into a normed space Y such that {kTn xk} is bounded for every x ∈ X.
That is, for every x there exists a cx such that kTn xk ≤ cx for all n. Then the sequence of the norms kTn k
is bounded.
As in the problem statement, let X ⊂ `∞ be the subspace consisting of finite sequences. Then X is not
complete. For each n let Tn = fn be the linear functional defined by fn (x) = nxn . Let x ∈ X, then x = {xj }
and there exists an N ∈ N such that xj = 0 for all j > N . Then

kTn xk = |fn (x)| = |nxn | = n|xn | ≤ n kxk

where cx = kxk depends on the choice of x, and thus {kTn xk} is bounded for each x ∈ X. However, kTn k = n
for each n. Indeed, from the above analysis we see that kTn k ≤ n, and choosing x = {1, 1, . . . , 1, 0, 0, · · · }
| {z }
n
with kxk = 1 we have kTn xk = n so kTn k ≥ n for each n. Thus {kTn k} is not bounded.

11
Problem 12 (Problem 12, Chapter 4.7, p. 255). Let X and Y be Banach spaces and Tn ∈ B(X, Y ) such
that {Tn x} is Cauchy in Y for every x ∈ X. Show that Tn x −→ T x for some T ∈ B(X, Y ).
Solution. Since every Cauchy sequence is bounded, we have that {kTn xk} is bounded for all x ∈ X and
thus {kTn k} is bounded by the Uniform Boundedness Theorem. Thus sup kTn k is finite.
n∈N

Since Y is complete and the sequence {Tn x} is Cauchy in Y for each x ∈ X, each {Tn x} converges to an
element in Y . Define a mapping T : X −→ Y by

T x = lim Tn x.
n→∞

This is clearly linear, since by the linearity of Tn we have

T (ax + by) = lim Tn (ax + by) = lim (aTn x + bTn y) = aT x + bT y

n→∞ n→∞

for all x, y ∈ X and a, b in the field F. This operator is also bounded, since for all x ∈ X we have

kT xk = lim Tn x = lim kTn xk

n→∞ n→∞
≤ sup kTn k kxk ,
n∈N

kT xk
where we use the fact that the norm is continuous, so it commutes with limits. Hence kxk ≤ sup kTn k < ∞
n∈N
for all x 6= 0, but this is finite and thus kT k ≤ sup kTn k < ∞, so T is also bounded and thus T ∈ B(X, Y )
n∈N
as desired.

12
Problem 13 (Problem 14, Chapter 4.7, p. 255). If X and Y are Banach spaces and Tn ∈ B(X, Y ) for
each n, show that the following statements are equivalent:
(a) {kTn k} is bounded,
(b) {kTn xk} is bounded for all x ∈ X,

(c) {|g(Tn x)|} is bounded for all x ∈ X and g ∈ Y 0 .

Solution. We first prove the following lemma.
Lemma 4. If {xn } is a sequence in a Banach space X such that {f (xn )} is bounded for each functional
f ∈ Y 0 , then {kxn k} is bounded.

Proof. For each f ∈ X 0 , define the sequence {ϕn } in X 00 by ϕn (f ) = f (xn ). Then the sequence {ϕn (f )}
is bounded for each f ∈ X 0 by assumption, and hence {kϕn (f )k} is bounded for each f . By the Uniform
Boundedness Theorem, we have that {kϕn k} is also bounded, but kϕn k = kxn k for each n and thus {kxn k}
is bounded.

We now proceed to prove the problem statement.

• (a) ⇒ (b). Since {kTn k} is bounded, there is a constant c ≥ 0 such that kTn k ≤ c for all n. Define the
constant cx = c kxk for each x ∈ X. Then

kTn xk ≤ kTn k kxk ≤ cx ,

and thus {kTn xk} is bounded for each x ∈ X.

• (b) ⇒ (a). This follows from the Uniform Boundedness Theorem.
• (a) ⇒ (c). Let g ∈ Y 0 and define the functionals fn ∈ X 0 for each n by fn (x) = g(Tn x). Since {kTn k}
is bounded, there is a constant c such that kTn k ≤ c for all n. Then we have

|fn (x)| = |g(Tn x)| ≤ kgk kTn k kxk

≤ kgk c kxk

for each x ∈ X and thus kfn k ≤ c kgk, so the sequence {kfn k} is bounded. From the (a) ⇒ (b) part of
this problem, this implies that {|fn (x)|} = {|g(Tn x)|} is bounded for all x.
• (c) ⇒ (b). Let x ∈ X and define the sequence {yn } in Y by yn = Tn x. Then the sequence {g(yn )}
is bounded for each g ∈ Y 0 . Since Y is a Banach space, Lemma 4 from above says that {kyn k} is
bounded, and thus {kTn xk} is bounded.

13
Problem 14 (Problem 2, Chapter 4.8, p. 262). Let X and Y be normed spaces, T ∈ B(X, Y ) and {xn } a
w w
sequence in X. If xn −→ x, show that T xn −→ T x.
Solution. Let g ∈ Y 0 and define a functional f ∈ X 0 by f (x) = g(T x). Then g(T xn ) = f (xn ) and thus
w
g(T xn ) −→ g(T x) if and only if f (xn ) −→ f (x), which is true since xn −→ x. Hence g(T xn ) −→ g(T x) for
w
all g ∈ Y 0 and thus T xn −→ T x.

14
 w
Problem 15 (Problem 4, Chapter 4.8, p. 262). Show that xn −→ x implies lim inf kxn k ≥ kxk.
n→∞

Solution. By the Hahn-Banach theorem, there exists a functional f ∈ X 0 such that kf k = 1 and f (x) = kxk.
Since xn weakly converges to x, we have that f (xn ) −→ f (x), hence |f (xn )| −→ |f (x)| = kxk. Thus

kxk = |f (x)| = lim |f (xn )|

n→∞
= lim inf |f (xn )|
n→∞
≤ lim inf kf k kxn k
n→∞ |{z}
=1
= lim inf kxn k ,
n→∞

as desired.

15
Problem 16 (Problem 6, Chapter 4.8, p. 262). If {xn } is a weakly convergent sequence in a normed space
w
X, say xn −→ x, show that there is a sequence {ym } of linear combinations of elements of {xn } which
converges strongly to x.
Solution. We first prove the following lemma (see problem 4.8.1 in Kreeyszig).
w
Lemma 5. If xn −→ x in a normed space X, then x ∈ Y where Y = span{xn }.
Proof. Suppose otherwise that x 6∈ Y . Then the distance

δ = inf kx − yk
y∈Y

from x to Y is positive. By the theorem in class (Theorem 4.6-7 in Kreyszig), there exists a functional
f ∈ X 0 such that kf k = 1, f (x) = δ and f (y) = 0 for all y ∈ Y . Hence f (xn ) = 0 for each n, and thus f (xn )
w
does not converge to f (x) = δ, a contradiction to xn −→ x.

The proof of the problem statement follows trivially from this lemma. Indeed, since x ∈ Y , there is a
sequence {ym } in Y such that ym −→ x. But Y = span{xn }, and thus each ym is a linear combination of
elements in {xn }.

16
Problem 17 (Problem 8, Chapter 4.8, p. 262 – Weak Cauchy sequence). A weak Cauchy sequence in a
real or complex normed space X is a sequence {xn } in X such that for every f ∈ X 0 the sequence {f (xn )}
is Cauchy in R or C, respectively. Show that every weak Cauchy sequence is bounded.
Solution. For clarity, denote F = R or C.
Define the sequence {ϕn } in X 00 such that ϕn (f ) = f (xn ) for all f ∈ X 0 . Let f ∈ X 0 be arbitrary. Since
{xn } is weak Cauchy, we have that {f (xn )} is Cauchy in F and thus the sequence {|f (xn )|} is bounded.
Hence {|ϕn (f )|} is bounded for each f ∈ X 0 . But X 0 is Banach, so we have that {kϕn k} is bounded by the
Uniform Boundedness Theorem. Then {xn } is bounded since kxn k = kϕn k.

17
Problem 18 (Problem 10, Chapter 4.8, p. 263 – Weak completeness). A normed space X is said to be
weakly complete if each Cauchy sequence in X converges weakly in X. If X is reflexive, show that X is
weakly complete.
Solution. Let {xn } be a Cauchy sequence in X. For each xn ∈ X let ϕn ∈ X 00 be the functional defined
by ϕn (f ) = f (xn ) for all functionals f ∈ X 0 . Since X is reflexive, we have X ∼
= X 00 so {ϕn } is a Cauchy
00
sequence in X . Let ε > 0, then there exists an N ∈ N such that kϕn − ϕm k ≤ kfε k for all n, m ≥ N since
{ϕn } is Cauchy. Then for each n, m ≥ N we have

|f (xn ) − f (xm )| = |ϕn (f ) − ϕm (f )|

≤ kϕn − ϕm k kf k
< ε,

so the sequence {f (xn )} is Cauchy in R and thus {xn } is weakly convergent. Then X is weakly complete,
since every Cauchy sequence in X is weakly convergent.

18