0% found this document useful (0 votes)
128 views

Paper Code MATH M6202

This document provides the solutions to an examination on functional analysis. It addresses several problems related to concepts like the Cauchy-Schwarz inequality, completeness of subspaces, bounded linear operators, and convergence in different topological spaces. The problems range from basic definitions and theorems to more complex unseen questions. Overall, the solutions demonstrate a thorough understanding of key topics in functional analysis.
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
128 views

Paper Code MATH M6202

This document provides the solutions to an examination on functional analysis. It addresses several problems related to concepts like the Cauchy-Schwarz inequality, completeness of subspaces, bounded linear operators, and convergence in different topological spaces. The problems range from basic definitions and theorems to more complex unseen questions. Overall, the solutions demonstrate a thorough understanding of key topics in functional analysis.
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 6

EXAMINATION SOLUTIONS

FUNCTIONAL ANALYSIS
MATH 36202
(Paper Code MATH M6202)

May-June 2015, 2 hours and 30 minutes

1. (a) (5 marks, bookwork)


Using the linearity of the inner product,

kx + tyk2 = hx + ty, x + tyi = kxk2 + 2t hx, yi + t2 kyk2 ≥ 0.

Since this quadratic polynomial is always non-negative, its discriminant satisfies 4 hx, yi2 −
4kxk2 kyk2 ≤ 0. Hence, | hx, yi | ≤ kxk kyk, as required.
(b) (5 marks, similar to homework)
By the Cauchy-Schwarz inequality,

| hxn , yn i − hx, yi | ≤ | hxn , yn − yi | + | hxn − x, yi | ≤ kxn kkyn − yk + kxn − xkkyk.

We observe that kxn k ≤ kxk + kxn − xk → kxk, so that kxn k is bounded. Since
kxn − xk → 0 and kyn − yk → 0, this implies that the above expression converges to
zero.
(c) (5 marks, unseen)
Observe that
ktx + (1 − t)yk ≤ ktxk + k(1 − t)yk = 1,
for any 0 < t < 1.
Suppose in contrary that the equality holds for some t in (0, 1). Then

1 = htx + (1 − t)y, tx + (1 − t)yi = t2 + 2t(1 − t) hx, yi + (1 − t)2 ,

and we obtain hx, yi = 1 = kxkkyk — the equality case in the Cauchy–Schwarz


inequality. This is only possible when the vectors x and y are linearly dependent,
which gives a contradiction.
(d) (i) (2 marks, similar to homework)
If u, v ∈ H0 and a, b ∈ F, then hau + bv, x0 i = a hu, x0 i + b hv, x0 i = 0. Hence, H0
is a subspace.
Suppose that xn ∈ H0 and xn → x ∈ H. Then

| hxn , x0 i − hxn , x0 i | = | hxn − x, x0 i | ≤ kxn − xkkx0 k → 0.

This implies that hx, x0 i = 0 as required.


(ii) (3 marks, similar to bookwork)
Given y ∈ H, we write y = y0 + cx0 where c = hy,x 0i
kx0 k2
and y0 = y − cx0 . Then for
z ∈ H0 , we have y − z = (y0 − z) + cx0 where (y0 − z) ⊥ cx0 . Hence
ky − zk2 = ky0 − zk2 + |c|2 kx0 k2 .
|hy,x0 i|
This shows that ky − zk ≥ |c|kx0 k = kx0 k
, where equality holds when y = z.
|hy,x0 i|
Thus, the distance from y to H0 is kx0 k
.
(e) (i) (2 marks, bookwork)
Let (en ) be an orthonormal system in H. Then for every x ∈ H,
X
| hx, en i |2 ≤ kxk2 .
n

(ii) (3 marks, unseen)


Suppose that the Bessel inequality holds. Taking x = e1 , we deduce that
X
| he1 , e1 i |2 + | he1 , en i |2 ≤ ke1 k2
n≥2

and this implies that he1 , en i = 0 for all n ≥ 2. The same argument shows that
hek , el i = 0 for all k 6= l.
2. (a) (4 marks, bookwork)
(n)
Let x(n) = (xk )k≥1 be a Cauchy sequence in `∞ . This means that for every  > 0,
and n, m ≥ n0 (),
(n) (m)
kx(n) − x(m) k∞ = sup |xk − xk | < .
k

This, in particular, implies that for every k and n, m ≥ n0 (),


(n) (m)
|xk − xk | < .
(n)
In particular, each sequence (xk )n≥1 is a Cauchy sequence. Since R is complete,
(n)
xk → xk as n → ∞ for some xk ∈ R. Passing to the limit as m → ∞ in the above
inequality, we deduce that for every k and n ≥ n0 (),
(n)
|xk − xk | ≤ .
This implies that for every k, |xk | <  + kx(n) k∞ , and x = (xk )k≥1 belongs to `∞ . The
above inequality also implies that kx(n) − xk∞ → 0 as n → ∞.
(b) (5 marks, similar to bookwork)
Let S be complete subspace of a Banach space X. Let xn ∈ S such that xn → x ∈ X.
Then kxn − xm k ≤ kxn − xk + kxm − xk → 0 as n, m → ∞, so that the sequence xn
is Cauchy. Hence, xn → y for some y ∈ S. Since the limit is unique x = y ∈ S. This
shows that S is closed.
Now suppose that S is a closed subspace of X, and xn is a Cauchy sequence in S.
Since X in Banach space, by completeness xn → x for some x ∈ X, but since S is
closed, x ∈ S. This proves that xn converges in S, and S is complete.
(c) (i) (4 marks, similar to homework)
(n)
We consider x = (1/k)k≥1 ∈ `∞ and x(n) ∈ X such that xk = 1/k for k ≤ n and
(n)
xk = 0 for k > n. Then kx(n) − xk ≤ 1/(n + 1) → 0. This shows that x belongs
to the closure of X, and X is not closed (hence, not complete) in `∞ .
(ii) (5 marks, similar to homework)
We claim that X̄ consists of sequences x = (xk )k≥1 such that xk → 0. Let x be
such a sequence. Then for every  > 0 and n ≥ n0 (), we have |xk | < . We take
(n) (n)
xk = xk for k ≤ n0 () and xk = 0 for k > n. Then kx(n) − xk∞ < . This
proves that x ∈ X̄. Conversely, suppose that x(n) → x for some x(n) ∈ X. Then
for every  > 0 and sufficiently large n, kx(n) − xk∞ < . Since x(n) ∈ X, for all
(n)
sufficiently large k, |xk − xk | = |xk |. This implies that for all sufficiently large
k, |xk | < . Hence, xk → 0, as claimed.
(iii) (4 marks, similar to homework)
Consider the operator S : X → X defined by (xn ) → (2n xn ). Then T S = ST = I.
(k) (k)
Take x(k) ∈ X such that xn = 1 for n ≤ k and xn = 0 for n > k. Then
kx(k) k∞ = 1 and kSx(k) k∞ = 2k . Since kSk = sup{kSxk∞ : kxk∞ = 1}, this
shows that kSk = ∞.
(d) (3 marks, bookwork)
The Bounded Inverse Theorem says that if T : X → X is a bounded bijective linear
map where X is a Banach space, then the inverse map T −1 is also bounded. This
theorem does not apply to (c)(iii) because X in (c) is not not complete as shown in
(c)(i) so it is not a Banach space.
3. (a) (2 marks, bookwork)
The dual space X ∗ is the space of bounded linear maps f : X → C.
(b) (4 marks, bookwork)
1 ∗ ∞ ∞ 1
We claimPthat (` ) ' ` . Let a = (ak )k≥1 ∈ ` . We defined fa : ` → C by
fa (x) = k≥1 ak xk . It is easy to check that fa is linear. Also
X
|fa (x)| ≤ |ak xk | ≤ kak∞ kxk1 .
k≥1

Hence, fa ∈ X ∗ . Now take any f ∈ X ∗ . We denote by ek ∈ `1 the vector whose


k’s coordinate is 1 and the other coordinates are 0. Let ak = f (ek ). Then |ak | ≤
kf kkek k1 = kf k, so that a = (ak )k≥1 ∈ `∞ . By linearity f (x) = fa (x) for all x in the
subspace X of `1 consisting of x that have only finitely many non-zero coordinates.
This subspace is dense in `1 . Hence, by continuity f (x) = fa (x) for all x ∈ `∞ .
(c) (3 marks,
P similar to homework) P
Since n≥1 kxnP k < ∞, for every  > 0 and m ≥ m0 (), n≥m kx
Pnmk2 < . Consider the
m
sequence sm = n=1 xn . For m1 < m2 we have ksm1 − sm2 k ≤ n=m1 +1 kxn k. Hence,
if m1 ≥ m0 (), then ksm1 − sm2 k < . This shows that the sequence sm is Cauchy.
Since X is a Banach space, this sequence must converge.
(d) (3 marks, similar to homework)
We apply the Hahn-Banach theorem. Let V be the subspace of X spanned by x. We
define f ∈ X ∗ by f (ax) = a. Note that |f (ax)| = |a| = kaxk. By the Hahn-Banach
theorem f can be expended to a linear map X → C such that |f (x)| ≤ kxk for all
x ∈ X. In particular, |f | ≤ 1 on B.
(e) (5 marks, unseen)
Each element xn defines a linear map Ln : X ∗ → C by Ln (f ) = f (xn ). Since the
sequence f (xn ) is Cauchy, it is bounded. Hence, there exists c = c(f ) > 0 such that
|Ln (f )| ≤ c for all n. We note that X ∗ is also a Banach space. By the Uniform
Boundedness Theorem, supn kLn k < ∞. Finally,
kLn k = sup{|Ln (f )| : kf k = 1} = sup{|f (xn )| : kf k = 1} = kxn k.
This proves that the sequence xn is bounded.
(f) (i) (3 marks, bookwork)
An → A in norm topology if kAn − Ak → 0. An → A in strong topology if
kAn x − Axk → 0 for every x ∈ X. An → A in weak topology if f (An x − Ax) → 0
for every x ∈ X and f ∈ X ∗ .
(ii) (5 marks, unseen)
For k > n, S n ek = en−k , so that for k > n2 > n1 ,

k(S n1 − S n2 )ek k1 = kek−n1 − ek−n2 k1 = 2.

Hence, kS n1 − S n2 k ≥ 2. This shows that S n is not a Cauchy sequence, and it


cannot converge in the norm topology.
Take x ∈ `1 . For every  > 0, there exists x0 that has only finitely many nonzero
coordinates such that kx − x0 k < . Then for all sufficiently large n, S n x0 = 0,
and kS n xk1 ≤ kS n x0 k1 +kS n kkx−x0 k1 ≤ kx−x0 k1 < . This proves that S n → 0
in strong topology.
Alternatively, to prove that S n does not converge in norm topology, one notes
that kS n k = 1. If S n → T for some operator T , then kT k = 1, but T = 0 because
S n → 0 in weak topology. This gives a contradiction.

4. (a) (2 marks, bookwork)


Every bounded linear map f : H → C is of the form f (x) = hx, yi for a uniquely
defined y ∈ H. Moreover, kf k = kyk.
(b) (2 marks, bookwork)
xn → x weakly if hxn − x, zi → 0 for every z ∈ H.
(c) (3 marks, unseen)
hxn − y, wi → 0 and hxn − z, wi → 0 for every w ∈ H. Then by subtracting we obtain
hy − z, wi = 0. Since this holds for all w ∈ H, this implies that y = z.
(d) (i) (4 marks, unseen)
We need to show that for every f ∈ H ∗ , f (Axn −Ax) → 0. The map g(x) = f (Ax)
is a bounded linear map. Hence, g(xn − x) → 0 as required.
(ii) (5 marks, unseen)
Suppose that the operator A is unbounded. Then there a sequence of unit vectors
xn such kAxn k ≥ n2 . Let yn = xn /n. Then kyn k → 0 and kAyn k → ∞. By the
assumption Ayn → 0 weakly. In particular, the sequence hAyn , yi is bounded for
every y ∈ H. Let Ln (y) = hAyn , yi. Note that Ln ∈ H ∗ and kLn k = kAyn k. By
the Uniform Boundedness Principle, supn kAyn k < ∞. This gives a contradiction.
(e) (4 marks, homework)
Recall that (ST )∗ = T ∗ S ∗ . We have AA−1 = A−1 A = I and hence (A−1 )∗ A∗ =
A∗ (A−1 )∗ = I. This proves that (A−1 )∗ is the inverse of A∗ . Since k(A−1 )∗ k ≤ kA−1 k,
the inverse is bounded.
(f) (5 marks, homework)
For f ∈ H, kAφ f k2 ≤ kφk∞ kf k2 . Hence, kAφ k2 ≤ kφk∞ . We claim that the equality
holds. Let m = kφk∞ . Since φ is continuous, there exists x0 ∈ [0, 1] such |φ(x0 )| = m.
For every  > 0, there exists δ > 0 such that |φ(x0 ) − φ(x)| <  for all x ∈ Bδ (x0 ). We
take f ∈ H such that supp(f ) ⊂ Bδ (x0 ). Then

kAφ f k2 ≥ kAm f k2 − k(Aφ − Am )f k2 ≥ mkf k2 − kf k2 .


This shows that kAφ k2 ≥ m −  for every  > 0. Hence, kAφ k2 = m.
For f1 , f2 ∈ H, hAφ f1 , f2 i = hφf1 , f2 i = f1 , φ̄f2 = f1 , Aφ̄ f2 . Hence, A∗φ = Aφ̄ .


5. (a) (3 marks, bookwork)


A subset is meager if it is a countable union of sets whose closure has empty interior.
The Baire Category Theorem states that a complete metric space is not meager.
(b) (3 marks, similar to homework)R
1
Consider the linear maps Ln (f ) = 0 φn (x)f (x)dx. We claim that kLn k = kφn k2 .
Indeed, |Ln (f )| ≤ kφn k2 kf k2 , so that kLn k ≤ kφn k2 . Taking f = φ¯n , |Ln (f )| =
kφk2 kf k2 , so that we deduce that kLn k ≥ kφn k2 which proves the claim. The space
L2 ([0, 1]) is complete, and we can apply the Uniform Boundedness Theorem to the
linear maps. Suppose that for every f ∈ L2 ([0, 1]), the sequence Ln (f ) is bounded.
Then by the Uniform Boundedness Theorem, the norms kLn k are also uniformly
bounded. This contradicts our assumption.
(c) (i) (3 marks, bookwork and homework) P
Let x = (xk )k≥1 ∈ `1 . This means that k≥1 |xk | < ∞. Then |xk | → 0; in
particular, |xk |2 ≤ |xk | for all sufficiently large k. Hence, it follows that that
2
`1 ⊂ `2 .
P
k≥1 |xk | < ∞. This shows that
Take x = (xk )k≥1 ∈ `2 . Since k≥1 |xk |2 < ∞, for all  > 0 and n ≥ n0 ()
P
2
P
k≥n+1 |xk | < . Take y = (yk )k≥1 such that yk = xk for k ≤ n0 () and
yk = 0 otherwise. Clearly, yk ∈ `1 and kx − yk22 < . This shows that x can be
approximated by elements from `1 .
(ii) (4 marks, unseen)
(n)
Suppose that x(n) = (xk )k≥1 ∈ BR and x(n) → x in `2 . This implies that for
(n)
every k, xk → xk as n → ∞. We have
N
X N
X N
X
(n) (n)
kxk1 = lim |xk | = lim lim |xk | = lim lim |xk | ≤ R.
N →∞ N →∞ n→∞ N →∞ n→∞
k=1 k=1 k=1

(iii) (4 marks, unseen)


Suppose that `1 has non-empty interior in `2 , i.e., there exists x ∈ `1 and  > 0

such that x + y ∈ `1 for every y with kyk2 < . Then for every z ∈ `2 , x + 2kzk2
z∈
1 1 1 1 2
` . Since x ∈ ` , this implies that z ∈ ` . We have shown that ` = ` , but
this is not true. For instance, x = (1/k)k≥1 belongs to `2 , but not to `1 . This
contradiction implies that `1 has empty interior in `2 .
Now `1 = ∪∞ R=1 BR where each BR is closed and has empty interior, so that ` is
1

meager in `2 .
(d) (i) (4 marks, similar to homework)
We write fn (x) = (e2πinx + e−2πinx )/2. By orthogonality for n 6= m,
1
kfn − fm k22 = ke2πinx + e−2πinx − e2πimx − e−2πimx k22 = 1.
4
This implies that fn is not a Cauchy sequence in H, so that it doesn’t converge.
(ii) (4 marks, similar to homework)
We use that the space of trigonometric polynomials is dense in H. For every
 > 0, there exists a trigonometric polynomial such that kg − pk2 < . It follows
from orthogonality that for sufficiently large n, hfn , pi = 0. We obtain

| hfn , gi | ≤ | hfn , pi | + | hfn , g − pi | ≤ kfn k2 kg − pk2 < / 2.
This proves that hfn , gi → 0.

End of solutions.

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy