Scribd
Upload
70 views4 pages

Tenta 21 April

The document is an exam in functional analysis consisting of 3 problems. The solutions provided are: 1) A proof that the domain of the Hilbert adjoint operator T* is a vector subspace and that T* is a linear operator. 2) A proof that the image of a compact set C under a closed linear operator T is a closed subset. 3) The spectral family (Eλ) associated with the self-adjoint operator T defined on l2 is given by a formula and verified to satisfy the desired properties.

Uploaded by

Fabian Molina
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
70 views4 pages

Tenta 21 April

The document is an exam in functional analysis consisting of 3 problems. The solutions provided are: 1) A proof that the domain of the Hilbert adjoint operator T* is a vector subspace and that T* is a linear operator. 2) A proof that the image of a compact set C under a closed linear operator T is a closed subset. 3) The spectral family (Eλ) associated with the self-adjoint operator T defined on l2 is given by a formula and verified to satisfy the desired properties.

Uploaded by

Fabian Molina
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/ 4

Uppsala Universitet

Matematiska Institutionen
Andreas Strombergsson

Prov i matematik
Funktionalanalys
Kurs: NVP 6p-delen
2006-04-21

Skrivtid: 9.0011.30
Till
atna hj
alpmedel: Manuella skrivdon, Kreyszigs bok Introductory Functional Analysis with Applications och Strombergssons haften
Spectral theorem for compact, self-adjoint operators and Mathematical
statements and proofs.
1. Let T : D(T ) H be a (possibly unbounded) densely defined
linear operator in a complex Hilbert space H. Recall the definition of
the Hilbert-adjoint operator T : D(T ) H (Definition 10.1-2). Give
a careful proof that D(T ) is a vector subspace of H and that T is a
linear operator.
(6p)
2. Let T : D(T ) Y be a closed linear operator, where D(T ) X
and X and Y are normed spaces. Let C D(T ) be a compact set.
Prove that the image T (C) = {T (x) | x C} is a closed subset of Y .
(7p)
3. Let T : `2 `2 be the self-adjoint bounded linear operator


T (1 , 2 , 3 , . . .) = (1 , 12 2 , 3 , 14 4 , 5 , 16 6 , 7 , . . .).
What is the spectral family (E ) associated with T ?
(You get several points for giving the correct formula for E for each
R, even if found by an intuitive (non-rigorous) argument.RHowever,

for full score, please give a careful verification of the fact T = dE


for your family (E ).)
(7p)
GOOD LUCK!

Solutions
1. By Definition 10.1-2 we have
n
o
D(T ) = y H | y H : x D(T ) : hT x, yi = hx, y i ,
and for each y D(T ) we define T y := y where y H is the vector
as above, i.e. the vector which has the property x D(T ) : hT x, yi =
hx, y i (this vector y is unique since D(T ) is dense in H).
Now let y1 , y2 be arbitrary vectors in D(T ) and let , be arbitrary
complex numbers. Then by the definition of D(T ) there exist vectors
y1 , y2 such that
x D(T ) : hT x, y1 i = hx, y1 i and hT x, y2 i = hx, y2 i.
Now note that for all x D(T ) we have
hT x, y1 + y2 i = hT x, y1 i + hT x, y2 i = hx, y1 i + hx, y2 i = hx, y1 + y2 i.
This proves that y1 + y2 D(T ), and also that T (y1 + y2 ) =
y1 + y2 = T (y1 ) + T (y2 ). Since these two properties hold for
all y1 , y2 D(T ) and all , C it follows that D(T ) is a vector
subspace of H and that T is a linear operator.
2. Let y1 , y2 , y3 , . . . be a sequence of points in T (C) such that y =
limj yj exists in Y . We must prove that y T (C).
For each j = 1 there exists some xj C with yj = T (xj ), since
yj T (C). We now assume that such a vector xj has been chosen
for each j = 1. Since C is compact and x1 , x2 , x3 , . . . C there exists a subsequence xj1 , xj2 , xj3 , . . . (where 1 5 j1 < j2 < j3 < . . .)
which converges to an element x C, i.e. limn xjn = x C. Now
limn T (xjn ) = limn yjn = y (since limj yj = y). Hence by
Theorem 4.13-3, since T is a closed linear operator, we have T x = y.
But x C, hence y T (C).
Hence we have proved that for every sequence y1 , y2 , y3 , . . . in T (C)
which converges to some y Y , we actually have y T (C). Hence
T (C) is closed in Y , by Theorem 1.4-6.
3. Recall the intuitive formula E =[projection on all part of `2 which
have eigenvalues 5 ]. Note that the given operator T has the property that all of `2 is (Hilbert-)spanned by eigenvectors; for note that
all the vectors1 e1 , e2 , e3 , e4 , . . . are eigenvectors of T , and these vectors
span `2 in the Hilbert sense, i.e. `2 = Span{e1 , e2 , e3 , . . .}. Hence it
1We

use the standard notation ej = (0, 0, . . . , 0, 1, 0, 0, . . .), where the 1 is in


position j.

seems reasonable to expect that the above intuitive formula is in fact


rigorously true in our case.
This leads to the following guess: For 5 0: E = 0. For 0 < < 1:
1
Let k be the smallest positive integer such that = 2k
:


E (1 , 2 , 3 , . . .) = (0, 0, . . . , 0, 2k , 0, 2k+2 , 0, 2k+4 , 0, . . .)
(where the first non-zero entry is in position 2k). Finally for = 1 we
should have E := I.
We now prove that (E ) as specified above satisfies all the desired
properties.
One easily checks that (E ) is a spectral family on [0, 1]. Indeed,
properties (7) and (8*) on p. 495 are directly clear from our definition
of (E ). It thus remains to check (9) on p. 495. Note that by our
definition of (E ) we have that for each 6= 0 R there exists some
> 0 such that E = E for all [, + ], and this property
immediately implies lim+0 E x = E x for all x H. Hence it
only remains to verify that (9) on p. 495 holds when = 0, i.e. that
lim0+0 E x = E x holds for all x H. By our definition of (E ) this
is the same as proving:
lim (0, 0, . . . , 0, 2k , 0, 2k+2 , 0, . . .) = 0,

x = (1 , 2 , 3 , . . .) `2 .

This is clear from the fact that for all x = (1 , 2 , 3 , . . .) `2 we have

2 X



|2j |2 0,
as k .
(0, 0, . . . , 0, 2k , 0, 2k+2 , 0, . . .) =
j=k

Hence we have proved completely that (E ) is a spectral family on


[0, 1].
R1
We now prove T = 00 t dEt . Given n Z+ , let Pn : 0 = t0 < t1 <
1
and for
t2 < . . . < tm = 1 be any partition of [0, 1] which has t1 = 2n
1
1
1
1 1
which all the points 2n , 2n2 , 2n4 , . . . , 4 , 2 occur among the tj s, and
1
with more tj -points inserted in a way such that (Pn ) 5 2n
. Then the
R1
Riemann-Stieltjes sum for the integral 00 t dEt corresponding to Pn
is:
m
X

s(Pn ) =
tj Etj Etj1
j=1

(Since we have lower limit of integration 0 0 we should in fact add


a term t0 Et0 to this sum, but note that this term is anyway 0.) The
1
contribution from j = 1 in the above sum is t1 (Et1 Et0 ) = 2n
(E 1 0).
2n

Note that this operator acts as follows on `2 :




1
1
1
1
1 0) (1 , 2 , 3 , . . .)
(E
2n , 0, 2n
2n+2 , 0, 2n
2n+4 , 0, . . .).
= (0, 0, . . . , 0, 2n
2n
2n

1
Furthermore, if j in the above sum is such that tj = 2k
for some
1
1
,
k = 1, 2, . . . , n 1 then by construction we have 2k+2 5 tj1 < 2k
and thus Etj Etj1 is projection onto the 2k:th coordinate, and the
1
(Etj Etj1 ), which acts as follows:
contribution from j to our sum is 2k


1
1
(Etj Etj1 ) (1 , 2 , 3 , . . .) = (0, 0, . . . , 0, 2k
2k , 0, 0, 0, . . .).
2k

(where the non-zero entry is in position 2k). Next, for j = m we have


tm = 1 and tm1 = 12 , hence the contribution from this j is 1(E1 E 1 ).
2
Note that this operator acts as follows:


(E1 E 1 ) (1 , 2 , 3 , . . .) = (1 , 0, 3 , 0, 5 , 0, . . .).
2

Finally, for every other j in the sum we see from our construction that
Etj = Etj1 , hence the contribution for this j is 0. Hence:


s(Pn ) (1 , 2 , 3 , . . .)
1
1
1
2n , 2n+1 , 2n
2n+2 , 2n+3 , 2n
2n+4 , 2n+5 , . . .).
= (1 , 12 2 , 3 , 41 4 , 5 , . . . , 2n
1
1
Hence we easily see ||s(Pn ) T || 5 2n
2n+2
0 as n . Hence
R1
T = 00 t dEt .
We have now proved that (E ) is a spectral family on [0, 1] and that
R1
T = 00 t dEt . Hence by the uniqueness part of Theorem 9.9-1 (which
I told about in class) we have that (E ) is the spectral family associated
with T .
Answer: For 5 0: E = 0. For 0 < < 1: Let k be the smallest
1
:
positive integer such that = 2k


E (1 , 2 , 3 , . . .) = (0, 0, . . . , 0, 2k , 0, 2k+2 , 0, 2k+4 , 0, . . .)

(where the first non-zero entry is in position 2k). For = 1: E := I.

You might also like