Functional Analysis I

Assignment Solutions

James C. Robinson
 Contents

1 Examples I page 1

2 Examples II 5

3 Examples III 9

4 Examples IV 16

iii
 1

Examples I

1. Suppose that
n
X m
X
v= αj ej and v= βk fk .
j=1 k=1

with αj , βk ∈ K and ej , fk ∈ E. Relabel βk and fk so that fj = ej for

j = 1, . . . , n0 , and fj ∈
/ {e1 , . . . , en } for j > n0 , i.e.
n 0 m
X X
v= βk ek + βk fk ,
k=1 k=n0 +1

with the understanding that the first sum is zero if n0 = 0, and the second
zero if n0 = m.
It follows that
n n 0 m
X X X
αj ej − βk ek − βk fk = 0,
j=1 k=1 k=n0 +1

or
n0 n m
X X X
(αj − βj )ej − αj ej + βk fk = 0.
j=1 k=n0 +1 k=n0 +1

Since E is linearly independent, it follows that αj = βj = 0 for j > n0 ,

from which n0 = n, and that αj = βj for j = 1, . . . , n. So the expansion
is unique.
2. If x, y ∈ `lim (C) then x + y ∈ `lim (C) and αx ∈ `lim (C), since

lim (xn + yn ) = lim xn + lim yn and lim αxn = α lim xn .

n→∞ n→∞ n→∞ n→∞ n→∞

1
2 1 Examples I

3. The sum of two elements of `f (K) contains only a finite number of non-
zero terms, and multiplication by a scalar can only decrease the number
of non-zero terms (if that scalar is zero). The ej are clearly linearly
independent, and if x ∈ `f (K) has terms xj 6= 0 for j ∈ J , where J is
a finite subset of N, then clearly
X
x= xj ej ,
j∈J

so the collection of all the {ej }∞

j=1 forms a basis.
4.(i) If there exist a, b with 0 < a ≤ b such that

akxk1 ≤ kxk2 ≤ bkxk1

then
1 1
kxk2 ≤ kxk1 ≤ kxk2 ,
b a
while
(ii) if
akxk1 ≤ kxk2 ≤ bkxk1

and
αkxk2 ≤ kxk3 ≤ βkxk2

then
aαkxk1 ≤ kxk3 ≤ bβkxk1 .

5. See Corollary 3.12 in the printed lecture notes

6. If {un } is a Cauchy sequence in U then, since U ⊂ V and the norm on U
is simply the restriction of the map u 7→ kukV to U , it follows that {un }
is also a Cauchy sequence in V . Since V is complete, un → u for some
u ∈ V . But since U is closed, we know that if un → u then u ∈ U . So U
is complete.
7. Consider
P∞ ∞
j=1 |xj yj | |xj | |yj |
X
=
kxk`p kyk`q kxk`p kyk`q
j=1
∞ ∞
1 X |xj |p 1 X |yj |q
≤ p +
p kxk`p q kykq`q
j=1 j=1
1 1
= + = 1.
p q
 Examples I 3

8. That `p is a vector space is easy to see, since if x, y ∈ `p then

|xj + yj |p ≤ (|xj | + |yj |)p ≤ 2p (|xj |p + |yj |p ).
Now,
∞ ∞ ∞
kx + ykp`p =
X X X
|xj + yj |p ≤ |xj ||xj + yj |p−1 + |yj ||xj + yj |p−1
j=1 j=1 j=1

Using Hölder’s inequality on the terms on the right-hand side with p, q

where p−1 + q −1 = 1, we have
 1/q
∞
kx + ykp`p ≤ (kxk`p + kyk`p ) 
X
|xj + yj |q(p−1)  .
j=1

Note that the bracketed sum is indeed finite, since p−1 + q −1 = 1 implies
that q(p−1) = p, and we have already shown that x+y ∈ `p . So therefore
p/q
kx + ykp`p ≤ (kxk`p + kyk`p ) kx + yk`p ;
and since p−1 + q −1 = 1 implies that p − p/q = 1, this gives the triangle
inequality in `p as desired.
9. Suppose that xk = (xk1 , xk2 , · · · ) is a Cauchy sequence in `p . Then for every
 > 0 there exists an N such that
∞
kxn − xm kp`p =
X
|xnj − xm p
j | <
p
for all n, m ≥ N . (1.1)
j=1

In particular {xkj }∞
k=1 is a Cauchy sequence of real numbers for every fixed
j. It follows that for each k ∈ N
xkj → ak
for some ak ∈ R.
Set a = (a1 , a2 , · · · ). We want to show that a ∈ `p and that kxk −
ak`p → 0 as k → ∞. First, since {xk } is Cauchy we have from (1.1) that
kxn − xm k`p <  for all n, m ≥ N , and so in particular for any N ∈ N
N
X ∞
X
|xnj − xm p
j | ≤ |xnj − xm p p
j | ≤ .
j=1 j=1

Letting m → ∞ we obtain
N
X
|xnj − aj |p ≤ p ,
j=1
4 1 Examples I

and since this holds for all N it follows that

∞
X
|xnj − aj |p ≤ p ,
j=1

and so xk − a ∈ `p . But since `p is a vector space and xk ∈ `p , this implies

that a ∈ `p and kxk − ak`p ≤ .
 2

Examples II

1. If x = (x1 , x2 , . . .) ∈ `2 (R), then

∞
X
|xj |2 < ∞.
j=1

So
∞
X
kx − (x1 , x2 , . . . , xn , 0, 0, . . .)k2`2 = |xj |2 → 0
j=n+1

as n → ∞. So `f (R) is dense in `2 (R).

2. If x ∈ X ∩ Y then there exist X and Y such that

{y ∈ B : ky − xk < X } ⊂ X and {y ∈ B : ky − xk < Y } ⊂ Y.

Taking  = min(X , Y ) it follows that

{y ∈ B : ky − xk < } ⊂ X ∩ Y,

and so X ∩ Y is open.
Now, given z ∈ B and  > 0, since X is dense there exists an x ∈ X
such that kx − zk < /2. Since X is open, there is a δ < /2 such that

{x0 ∈ B : kx0 − xk < δ} ⊂ X.

Since Y is dense, there is a y ∈ Y such that ky − xk < δ/2. By the above,

it follows that we also have y ∈ X. So we have found y ∈ X ∩ Y such that
ky − zk ≤ ky − xk + kx − zk < δ/2 + /2 < , and so X ∩ Y is dense.
A refinement of this argument allows one to prove the powerful Baire
Category Theorem: a countable intersection of open and dense sets is
dense.

5
6 2 Examples II

3. We have to assume that (Y, k · kY ) is a Banach space. Then if xn ∈ X

and xn → x (with x ∈ V ) we know that {xn } is Cauchy in V . So, since

kF (xn ) − F (xm )kY ≤ Lkxn − xm k

it follows that {F (xn )} is a Cauchy sequence in Y . Since Y is complete,

we know that limn→∞ F (xn ) exists and is an element of Y .
If xn → v, and yn → v (where the two sequences in X, {xn } and {yn },
might be different) then given any  > 0 there exists a N such that

kxn − vk <  and kyn − vk < 

and

kF (xn ) − lim F (xn )k <  and kF (yn ) − lim F (yn )k < ,

n→∞ n→∞

for all n ≥ N . In particular, it follows that

k lim F (xn ) − lim F (yn )k ≤ k lim F (xn ) − F (xN )k + kF (xN ) − F (yN )k

n→∞ n→∞ n→∞
+ kF (yN ) − lim F (yn )k
n→∞
≤  + LkxN − yN k + 
≤ 2(1 + L),

since kxN − yN k ≤ kxN − vk + kv − yN k. Since this is true for all  > 0

it follows that
lim F (xn ) = lim F (yn ).
n→∞ n→∞

That F so defined is continuous is clear: for any v1 , v2 ∈ V , one can

find sequences in X that converge to v1 , v2 , and then from the definition

kF (v1 ) − F (v2 )k ≤ Lkv1 − v2 k.

4. Expanding the right-hand side gives

(x + y, x + y) − (x − y, x − y) + i(x + iy, x + iy) − i(x − iy, x − iy)

= kxk2 + (x, y) + (y, x) + kyk2 − kxk2 + (x, y) + (y, x) − kyk2
+ i(kxk2 + i(y, x) − i(x, y) + kyk2 − kxk2 + i(y, x) − i(x, y) − kyk2 )
= 4 real(x, y) + 4 imag(x, y) = 4(x, y)

5. Suppose that E is orthonormal and that

n
X
αj ej = 0
j=1
 Examples II 7

for some αj ∈ K and ej ∈ E. Then taking the inner product of both sides
with ek (with k = 1, . . . , n) gives
Xn
( αj ej , ek ) = 0 ⇒ αk = 0,
j=1

so αk = 0 for each k, i.e. E is linearly independent.

6. The sequence
1 1 1
x(n) = (1, , , . . . , , 0, 0, . . .)
2 3 n
of elements of `f (R) converge in `2 (R) to
1 1 1
x∗ = (1, , , . . . , , . . .)
2 3 n
(i.e. x∗ with x∗j = 1/j). Although x∗ ∈ `2 , clearly it has an infinite number
of non-zero terms.
7.(i) Suppose that d = 0. Then there exists a sequence yn ∈ Y such that

lim kx − yn k = 0,
n→∞

i.e. yn → x. Since Y is a closed linear subspace, it follows that x ∈ Y

– but this contradicts the fact that x ∈ / Y.
(ii) By definition there exists a sequence yn ∈ Y such that d = limn→∞ kx−
yn k. So for n sufficiently large, clearly kx − yn k < 2d. Let z be one
such yn .
(iii) We have

x−z 1 1 1
kx̂ − yk =
kx − zk − y = kx − zk x − (z + kx − zky) > 2d d = 2 ,

since z + kx − zky ∈ Y and d = inf{kx − yk : y ∈ Y }.

8. Take x1 ∈ B with kx1 k = 1. Let Y1 be the linear subspace {αx1 : α ∈ K}.
Using the result of question 7 there exists an x2 ∈ B with kx2 k = 1 such
that kx2 − x1 k > 12 . Now let Y2 = Span(x1 , x2 ), and use question 7 to find
x3 ∈ B with kx3 k = 1 with kx3 − yk > 21 for every y ∈ Span(x1 , x2 ) – in
particular
1 1
kx3 − x1 k > and kx3 − x2 k > .
2 2
Continuing in this way we obtain a sequence {xn } ∈ B such that kxn k = 1
and
1
kxn − xm k >
2
8 2 Examples II

for all n 6= m. It follows that the unit ball is not compact, since no
subsequence of the {xn } can be a Cauchy sequence.
9. The unit ball is closed and bounded. If B is finite-dimensional then this
implies that the unit ball is compact. We have just shown, conversely,
that if B is infinite-dimensional then the unit ball is not compact.
 3

Examples III

1. Take x, y ∈ A and suppose that λx + (1 − λ)y ∈ A for all λ ∈ [0, 1] such

that 2k λ ∈ N. Then given µ with 2k+1 µ ∈ N, 2k µ ∈
/ N we have

µ = l2−(k+1)

with l odd, so that

(l − 1) −(k+1) (l + 1) −(k+1)
µ= 2 + 2 .
2 2

With λ1 = (l − 1)2−(k+1) and λ2 = (l + 1)2−(k+1) we have 2k λ1 , 2k λ2 ∈ N,

and so

1
µx + (1 − µ)y = [(λ1 x + (1 − λ1 )y) + (λ2 x + (1 − λ2 )y)] ,
2

with λj x + (1 − λj )y ∈ A, i.e. µx + (1 − µ)y ∈ A.

It follows by induction that λx + (1 − λ)y ∈ A for any λ ∈ [0, 1] such
that 2k λ ∈ N for some k ∈ N. Since any λ ∈ [0, 1] can be approximated
by a sequence λj such that 2j λj ∈ N, it follows since A is closed that
λx + (1 − λy) ∈ A for all λ ∈ [0, 1], i.e. A is convex.
2. We have x ∈ (X ⊥ )⊥ if for every y ∈ X ⊥ , (x, y) = 0. So clearly X ⊆
(X ⊥ )⊥ . If we do not have equality here then there exists a z ∈ / X with
(z, y) = 0 for every y ∈ X ⊥ . But we know that z = x + ξ with x ∈ X and
ξ ∈ X ⊥ ; but then (x + ξ, ξ) = 0, which implies that kxik2 = 0, i.e. x ∈ X.
So (X ⊥ )⊥ = X as claimed.
3. See pages 52 and 53 in the printed notes

9
10 3 Examples III

4. We have
r!r r !r
5 5 3 3
e04 = x3 − x3 , (3x2 − 1) (3x2 − 1) − x3 , x x
8 8 2 2
 
1 1
− x3 , √ √
2 2
Z 1
3x 1 4 1 1 3
Z Z
5
= x3 − (3x2 − 1) (3t5 − t3 ) dt − t dt − t dt
8 −1 2 −1 2 −1
3x
= x3 − .
5
Then
1 2 1
t7 6t5 3t3
Z  
3t 8
ke04 k2 = 3
t − dt = − + =
−1 5 7 25 25 −1 7 × 25

and so
r
7
e4 = (5x3 − 3x).
8
5. Approximation of sin x by a third degree polynomial f3 (x): note that
(sin x, e1 ) and (sin x, e3 ) will be zero since sin x is odd and e1 and e3 are
even, so

3x 1 7(5x3 − 3x) 1 3
Z Z
f3 (x) = t sin t dt + (5t − 3t) sin t dt.
2 −1 8 −1

Now,
Z 1 Z 1
t sin t dt = [−t cos t]1−1 + cos t dt = [sin t]1−1 = 2 sin 1
−1 −1

and
Z 1 1
Z 1
t sin t dt = −t3 cos t −1 +
3
3t2 cos t dt

−1 −1
1
Z 1
= 3t2 sin t −1 − 6

t sin t dt
−1
= −6 sin 1,

giving
Z 1
(5t3 − 3t) sin t dt = −36 sin 1;
−1
 Examples III 11

it follows that
7(27x − 45x3 ) 195x − 315x3
   
f3 (x) = sin 1 3x + = sin 1 .
2 2

6. First way: given orthonormal polynomials ẽj (x) on [−1, 1] put

√
ej (x) = 2 ẽj (2x − 1),

since then
Z 1 Z 1 Z 1
ej (x)ek (x) dx = 2 ẽj (2x − 1)ẽk (2x − 1) dx = ẽj (y)ẽk (y) dy.
0 0 −1

Doing this gives

√
e1 (x) = 1, e2 (x) = 3 (2x − 1),
√
5 √
e3 (x) = [3(4x2 − 4x + 1) − 1] = 5 (6x2 − 6x + 1),
2
and
√
7
5(8x3 − 12x2 + 6x − 1) − 3(2x − 1)

e4 (x) =
√2
= 7 [20x3 − 30x2 + 12x − 1].

Or the painful way: e1 (x) = 1, then

Z 1
1
e02 (x) = x − t dt = x − ,
0 2
and then
" #1
1 (x − 12 )3
Z
1
ke02 k2 = (x − 12 )2 dx = =
0 3 12
0
√
which gives e2 (x) = 3 (2x − 1). Now set
Z 1  Z 1
0 2 2
e3 (x) = x − 3(2x − 1) t (2t − 1) dt − t2 dt
0 0
4
1
t3

1 t
= x2 − 3(2x − 1) − −
0 3 2 3
2x − 1 1 1
= x2 − − = x2 − x + ,
2 3 6
12 3 Examples III

and then
1 1
4t2
Z Z
1 t 1
ke03 k2 = 2
(t − t + )2 dt = t4 − 2t3 + − + dt
0 6 0 3 3 36
1
t5 t4 4t3 t2

t
= − + − +
5 2 9 6 36 0
1
= ,
180
√
and so e3 (x) = 5(6x2 − 6x + 1).
Finally, we have
Z 1 
e04 (x) = x3 − 5(6x2 − 6x + 1) t3 (6t2 − 6t + 1) dt
0
Z 1  Z 1
3
−3(2x − 1) t (2t − 1) dt − t3 dt
0 0
6x2 − 6x + 1 9 1
= x3 − − (2x − 1) −
4 20 4
20x3 − 30x2 + 12x − 1
=
20
with
Z 1
1
ke04 k2 = (20t3 − 30t2 + 12t − 1)2 dt
400 0
Z 1
1
= 400t6 + 900t4 + 144t2 + 1 − 1200t5 + 480t4 − 40t3 − 720t3
400 0
2
+60t − 24t dt 
1 400 900 144 1200 480 40 720 60 24
= + + +1− + − − + −
400 7 5 3 6 5 4 4 3 2
1
= ,
400 × 7
√
and so e4 (x) = 7 (20x3 − 30x2 + 12x − 1) as above.
1/2 1/2
7. Given x, y ∈ `2w , set x̂j = wj xj and ŷj = wj yj . Then
(x, y)`2w = (x̂, ŷ)`2 ,
so the fact that (·, ·)`2w is an inner product on `2w follows from the fact
that (·, ·)`2 is an inner product on `2 . Similarly the completeness of `2w
follows from the completeness of `2 – if {xn } is a Cauchy sequence in `2w
then {x̂n } is a Cauchy sequence in `2 , so there exists a y ∈ `2 such that
1/2
x̂n → y. Setting zj = yj /wj gives a z ∈ `2w such that xn → z in `2w .
 Examples III 13

8. See Proposition 9.6 in the printed notes.

9. Consider the functions fn = e−nx . Then kfn k∞ = 1, but fn0 (x) = −ne−nx ,
and so kfn0 k∞ = n. So the map f 7→ f 0 is not bounded.
10. If kxkX ≤ 1 then

x x
A kxkX ≤ A kxkX ,
kAxkY = kxkX

and so since k(x/kxkX )kX = 1 we have

sup kAxkY ≤ sup kAzkY ≤ sup kAzkY .

kxkX ≤1 kzkX =1 kzkX ≤1

Rearranging the above equality we have

kAxkY x
= A
,
kxkX kxkX Y

and so, again since k(x/kxkX )kX = 1,

kAxkY
sup = sup kAxkY .
x6=0 kxkX kxk=1

11.(i) If y ∈ `q then
 
∞
X 

|l(x)| = 
 xj yj  ≤ kxkp kykq
 j=1 

using Hölder’s inequality. So klkp∗ ≤ kykq as claimed.

(ii) Any x = (x1 , x2 , x3 , . . .) ∈ `p can be written as

∞
X
x= xj ej ,
j=1

for the elements ej defined in the question. Since l is bounded it is

14 3 Examples III

continuous, so
 
n
X
l(x) = l  lim xj ej 
n→∞
j=1
 
n
X
= lim l  xj ej 
n→∞
j=1
n
X
= lim xj l(ej )
n→∞
j=1
∞
X
= xj l(ej )
j=1
X∞
= xj yj ,
j=1

if we define yj = l(ej ). [Should you come across something similar in

an exam you should at least mention continuity of l as the justification
for taking it inside the infinite sum.]
(iii) We have
n  n n
 |yj |q p X
X  X
kx(n) kpp = 
 yj 
 = |y j |p
q − 1 = |yj |q ,
j=1 j=1 j=1

and so, as claimed,

 1/p
Xn
kx(n) kp =  |yj |q  .
j=1

Also
n n
(n)
X |yj |q X
l(x )= yj = |yj |q .
yj
j=1 j=1

Since |l(x(n) | ≤ klkp∗ kx(n) kp , we have

 1/p
n
X n
X
|yj |q ≤ klkp∗  |yj |q  ,
j=1 j=1
 Examples III 15

from which it follows, since 1 − 1/p = 1/q, that

 1/q
X n
 |yj |q  ≤ klkp∗ .
j=1

Since this holds for every n, we must have y ∈ `q with kykq ≤ klkp∗ .
 4

Examples IV

1. Clearly fφ is linear. Given u ∈ C 0 ([a, b]) we have

Z b 
 
|fφ (u)| = 
 φ(t)u(t) dt
a
 Z b
≤ max |u(t)| |φ(t)| dt
t∈[a,b] a
Z b
= kuk∞ |φ(t)| dt, (4.1)
a

Rb
and so kfφ k ≤ a |φ(t)| dt. Now since φ is continuous we can find a
sequence of continuous functions un that approximate the sign of φ (+1
if φ > 0, −1 if φ < 0) increasingly closely in the L2 norm: if φ(s) 6= 0
then there is an interval around s on which φ has the same sign. Since
this interval has non-zero length, and the union of these intervals lies
within [a, b], there are at most a countable number, {Ij }∞
j=1 . Let un be
the continuous function defined on Ij = (lj , rj ) as

(t − lj )/δj,n
 lj ≤ t ≤ lj + δj,n
un (t) = sign(φ(t)) × 1 lj + δj,n < t < rj − δj,n

(rj − t)/δj,n rj − δj,n ≤ t ≤ rj


where δj,n = min(2−(n+j) , (rj − lj )/2), and zero when φ(t) = 0. Then

Z b ∞
X
2
|un (t) − sign(φ(t))| dt ≤ 2−(n+j) = 2−(n−1) ,
a j=1

16
 Examples IV 17

i.e. un → sign(φ) in L2 (a, b). Note that kun k∞ = 1. It follows that

Z b 
 
|fφ (un )| = 
 φ(t)un (t) dt
a
Z b Z b 
 
= ≥ |φ(t)| dt − φ(t)(un (t) − sign(φ(t))) dt
a a
Z b Z b 
 
≥ |φ(t)| dt −  φ(t)(un (t) − sign(φ(t))) dt
a a
Z b
≥ |φ(t)| dt − kφkL2 kun − sign(φ)kL2 .
a

Since φ ∈ C 0 ([a, b]) we know that kφkL2 < ∞, and we have just shown
that kun − sign(φ)kL2 → 0 as n → ∞. It follows since kun k∞ = 1 that we
cannot improve the bound in (4.1) and therefore
Z b
kfφ k = |φ(t)| dt.
a

2. Take x, y ∈ L2 (0, 1), then

Z 1 Z t 
(T x, y) = K(t, s)x(s) ds y(t) dt
0 0
Z 1Z t
= K(t, s)x(s)y(t) ds dt
0 0
Z 1Z 1
= K(t, s)x(s)y(t) dt ds
0 t
Z 1 Z 1 
= K(t, s)y(t) dt x(s) ds
0 t
= (x, T ∗ y)

(draw a picture to see how the limits of integration change) where

Z 1
T ∗ (y)(s) = K(t, s)y(t) dt.
t

3. If x ∈ (range(T ))⊥ then

(x, y) = 0 for all y ∈ range(T ),

i.e.
(x, T z) = 0 for all z ∈ H.

Since (T ∗ )∗ = T , this is the same as (T ∗ x, z) = 0 for all z ∈ H. This

18 4 Examples IV

implies that T ∗ x = 0, i.e. that x ∈ Ker(T ∗ ). This argument can be

reversed, which gives the required equality.
Now – with apologies for the mistake in the question: we will show that
λ̄ is an eigenvalue of T ∗ – suppose that range(T − λI) 6= H. Then there
exists some non-zero u ∈ (range(T − λI))⊥ . By the equality we have just
proved, this gives a non-zero u in Ker((T − λI)∗ ), i.e. a non-zero u such
that
(T − λI)∗ u = 0.
Since (T − λI)∗ = T ∗ − λ̄I, this gives a non-zero u with T ∗ u = λ̄u. So λ̄
is an eigenvalue of T ∗ .
4.(i) If T ∈ B(H, H) then T ∗ ∈ B(H, H) and kT ∗ kop = kT kop . So if {xn } is
a bounded sequence in H, with kxn k ≤ M , say, it follows that
kT ∗ xn k ≤ kT ∗ kop kxn k ≤ M kT ∗ kop .
So {T ∗ xn } is also a bounded sequence in H. Since T is compact, it
follows that {T (T ∗ xn )} = {(T T ∗ )xn } has a convergent subsequence,
which shows that T T ∗ is compact.
(ii) Following the hint, we have
kT ∗ xk2 = (T ∗ x, T ∗ x) = (T T ∗ x, x) ≤ kT T ∗ xkkxk.
Now, if {T T ∗ xn } is Cauchy then given any  > 0, there exists a N such
that for all n, m ≥ N ,
kT T ∗ xn − T T ∗ xm k = kT T ∗ (xn − xm )k ≤ .
It follows that for all n, m ≥ N ,
kT ∗ xn − T ∗ xm k2 = kT ∗ (xn − xm )k2 ≤ kT T ∗ (xn − xm )kkxn − xm k.
Since {xn } is a bounded sequence, with kxn k ≤ M , say, we have
kT ∗ xn − T ∗ xm k2 ≤ M  for all n, m ≥ N.
It follows that {T ∗ xn } is Cauchy.
(iii) So suppose that {xn } is a bounded sequence in H. Part (i) shows
that T T ∗ is compact, so {T T ∗ xn } has a subsequence {T T ∗ xnj } that is
Cauchy. But part (ii) shows that this implies that {T ∗ xnj } is Cauchy
too. So T ∗ is compact.
5.(i) If for every z ∈ H we have
(xn , z) → (x, z) and (xn , z) → (y, z)
then clearly (x, z) = (y, z) for every z ∈ H. But then (x − y, z) = 0
 Examples IV 19

for every z ∈ H; in particular we can take z = x − y, which shows that

kx − yk2 = 0, i.e. that x = y.
(ii) Take y ∈ H, and suppose that (en , y) does not tend to zero as n →
∞. Then for some  > 0, thee exists a sequence nj → ∞ such that
|(enj , y)| > . But then
X X
kyk2 = |(y, ej )|2 ≥ |(y, enj )|2 = ∞,
n j

contradicting the fact that y ∈ H with {ej } an orthonormal basis.

(iii) For some fixed z ∈ H, consider the map f : H → K given by

u 7→ (T u, z).

This map is clearly linear,

u + λv 7→ (T (u + λv), z) = (T u, z) + λ(T v, z),

and it is bounded,
 
|(T u, z)| ≤ kT ukkzk ≤ kT kop kzk kuk.

So f ∈ H ∗ . It follows that there exists a y ∈ H such that

(T u, z) = f (u) = (u, y)

for every u ∈ H. So if un * u, in particularly for this choice of y we

have
(un , y) → (u, y) ⇒ (T un , z) → (T u, z).

Since this holds whatever our choice of z, it follows that T un * T u.

(iv) **This part requires you to know that any weakly convergence sequence
is bounded, which is not at all obvious (but true).
Suppose that T xn 6→ T x. Then there exists a δ > 0 and a subsequence
T xnj such that

kT xnj − T xk > δ for all j = 1, 2, 3, . . . . (4.2)

Since T is compact and {xnj } is bounded there exists a further subse-

quence such that T xnj → w for some w ∈ H. But since convergence
implies weak convergence, we must have T xnj * w. But we already
know that T xnj * T x, so it follows from the uniqueness of weak limits
that w = T x, and so T xnj → T x, contradicting (4.2).
20 4 Examples IV

6. Clearly
n
X
T nx = λnj (x, ej )ej .
j=1

If |λj | < 1 for all j = 1, . . . , n then Lemma ?? shows that kT k < 1, and
so (I − T )−1 = I + T + T 2 + · · · , i.e.
n
X n
X
(I − T )−1 = I + λj (x, ej )ej + λ2j (x, ej )ej + · · ·
j=1 j=1
Xn
= I+ (λj + λ2j + · · · )(x, ej )ej
j=1
n
X λj
= I+ (x, ej )ej .
1 − λj
j=1

7.(i) Since x = f + αT x, the solution is given by x = (I − αT )−1 f . So if

kαT kop < 1 we can write
∞
X
x = (I + αT + α2 T 2 + α3 T 3 + · · · )f = (αT )j f.
j=0

So this expansion is valid for |α| < 1/kT kop .

(ii) Suppose that
Z b
(T n−1 x)(t) = Kn−1 (t, s)x(s) ds.
a

Then
Z b
n
(T x)(t) = K(t, s)(T n−1 x)(s) ds
a
Z b Z b 
= K(t, s) Kn−1 (s, r)x(r) dr ds
a a
Z bZ b
= K(t, s)Kn−1 (s, r)x(r) dr ds
a a
Z b Z b 
= K(t, s)Kn−1 (s, r) ds x(r) dr,
a a

i.e.
Z b
n
(T x)(t) = Kn (t, r)x(r) dr.
a
 Examples IV 21

with
Z b
Kn (t, r) = K(t, s)Kn−1 (r, s) ds
a

as claimed.
8. If T is invertible then Ker(T ) = {0}, using Lemma 12.4. So all the
eigenvalues of T are non-zero. It follows from the Corollary to the Hilbert-
Schmidt Theorem that the eigenvectors of T (corresponding to non-zero
eigenvalues) form a basis for H.
9. We have
 
X∞ ∞
X
(T u, v) =  λj (u, ej )ej , v  = λj (u, ej )(ej , v)
j=1 j=1
   
∞
X ∞
X
= u, λj (ej , v)ej  = u, λj (v, ej )ej 
j=1 j=1

= (u, T v)
and so T is self-adjoint. Now consider Tn u defined by
n
X
Tn u = λj (u, ej )ej .
j=1

Since the range of T is contained in the span of {ej }nj=1 it is finite-

dimensional, and so Tn is compact. Since
2
∞
X
2

kTn u − T uk = λj (u, ej )ej

j=n+1
∞
X
= |λj |2 |(u, ej )|2 ,
j=n+1

we have from the fact that λn → 0 that given any  > 0 there exists an
N such that for all n ≥ N we have |λj | < , and hence
∞
X ∞
X
2 2 2 2
kTn u − T uk ≤  |(u, ej )| ≤  |(u, ej )|2 = 2 kuk2 ,
j=n+1 j=1

i.e.
kTn − T kop ≤ ,
and so Tn → T . It follows from Theorem 13.8 that T is compact.
 22 4 Examples IV

10.(i) If x 6= 0 then kxk > 0, from which it follows that kT xk > 0, and so
x ∈/ Ker(T ). So Ker(T ) = {0} and T −1 exists for all y ∈ range(T ).
Then we have
kT (T −1 x)k2 ≥ αkT −1 xk2 ,
i.e. kT −1 xk2 ≤ α−1 kxk2 , so T −1 is bounded.
(ii) Let λ = α + iβ. Then
k(T − λI)xk2 = (T x − αx − iβx, T x − αx − iβx)
= k(T x − αx)k2 − (T x − αx, iβx) − (iβx, T x − αx) + |β|2 kxk2
= kT x − αxk2 + i(T x − αx, βx) − i(βx, T x − αx) + |β|2 kxk2 .
Now, since T is self-adjoint and α is real, it follows that
(T x − αx, βx) = (βx, T x − αx),
and so
k(T − λI)xk2 = kT x − αxk2 + β 2 kxk2 ≥ β 2 kxk2 ,
i.e. T − λI is bounded below as claimed.
(iii) If λ ∈ C and T − λI has a densely defined inverse then (T − λI)−1 ∈
B(H, H), i.e. λ ∈ R(T ). So σc (T ) ⊆ R.
(iv) If the range of T − λI is not dense in H then by Question 3 we know
that λ̄ is an eigenvalue of T ∗ . If T is self-adjoint, we must have λ̄ an
eigenvalue of T . But all eigenvalues of T are real, so in fact λ is an
eigenvalue of T . So λ ∈ σp (T ) rather than σr (T ), which implies that
σr (T ) is empty.