John Erdos 1

Introduction
These are notes for a King’s College course to fourth year undergraduates and MSc
students. They cover the theoretical development of operators on Hilbert space up
to the spectral theorem for bounded selfadjoint operators. I have tried to make the
treatment as elementary as possible and to include only what is essential to the
main development. The proofs are the simplest and most intuitive that I know. The
exercises are culled from various sources; some of them are more or less original. They
are designed to be worked as the course progresses and, in some cases, they anticipate
results and techniques that come in the later theorems of the course.
It should be emphasized that these notes are not complete. Although the theoretical
development is covered rather fully, examples, illustrations and applications which are
essential for the understanding of the subject, are left to be covered in the lectures.
There are good reasons for doing this. Experience has shown that audiences lose
concentration if they are provided with comprehensive notes which coincide with the
lectures. Also, in many cases examples and such are best treated in a less formal
way which is more suited to oral presentation. In this way it is possible to cater for
different sections of an audience with a mixed background. A formal proof may be
indicated to some while others may have to take certain statements on trust. This is
especially the case when integration spaces are involved.
I would like to thank the many students and colleagues who have pointed out errors
and obscurities in earlier versions of these notes. A few proofs contain some sentences
in square brackets. These indicate explanations that I consider rather obvious and
should be superfluous to a formal proof but were added in response to some query.
For the benefit of a wider audience, here is a brief indication of what might be covered
to supplement the notes and also a few comments.
Section 1. Examples of inner product spaces :
2
n
(= C
n
) and
2
. Continuous func-
tions on [a, b] with 'f, g` =

b
a
f(t)g(t) dt and problems with extending to larger classes
of functions (equivalence classes as elements of the space). Completeness of
2
n
(= C
n
)
and
2
(and continuous functions on [a, b] not complete).
L
2
[a, b]. Some brief discussion of the Lebesgue integral. The following statement to
be known or accepted: there is a definition of the integral such that the (equivalence
classes) of all functions f such that

b
a
[f(t)[
2
dt exists and is finite forms a Hilbert
space with the inner product 'f, g` =

b
a
f(t)g(t) dt (that is, it is complete). Some
more general L
2
spaces might be mentioned (e.g. L
2
(S) where S = [a, b] [a, b] or
some other subset of R
n
).
Examples of normed spaces which cannot be Hilbert spaces because they do not
satisfy the parallelogram law (C[a, b],
p
n
(n = 2) for p = 2).
Examples of normed spaces with closed convex sets where the distance from a point
is not attained uniquely (e.g. the unit ball in
1
2
with the point (1, 1)) or not attained
at all (e.g. the space X = ¦f ∈ C[0, 1] : f(0) = 0¦, the set ¦f ∈ X :

1
0
f(t) dt = 1¦
and the zero function as the point).
Some indications of the applications of the minimum distance theorem, e.g. to ap-
proximation theory and optimal control theory.
2 Operators on Hilbert space
Section 2. This section is supplemented by specific examples of operators on
2
and
L
2
[a, b]. These include diagonal operators, shifts (forward, backward and weighted)
on
2
, the bilateral shift on
2
(Z) and the following operators on L
2
[a, b].
Multiplication operator : (M
φ
f)(x) = φ(x).f(x) φ (essentially) bounded.
Fredholm integral operator : (Kf)(x) =

b
a
k(x, t)f(t) dt where

[k[
2
< ∞,
with the Volterra operator, (V f)(x) =

x
a
f(t) dt as a special case.
The boundedness of these operators should be established and the adjoints identified.
Other examples of finding adjoints (similar to those in the exercises) might be done.
Section 3. The main additional topic for this section is the connection to clas-
sical Fourier series. The fact that the normalized trigonometric functions for an
orthonormal basis of L
2
[−π, π] should be established or accepted. One route uses the
Stone-Weierstrass Theorem and the density of C[−π, π] in L
2
[−π, π]. Inevitably, this
requires background in metric spaces and Lebesgue theory. Note that this fact is also
established, albeit in a roundabout way, by the work in Section 7.
The projection onto a subspace can be written down in terms of an orthonormal basis
of the subspace : P
N
h =
¸
'h, y
i
`y
i
where ¦y
i
¦ is an orthonormal basis of N.
Applying the Gram-Schmidt process to the polynomials in the Hilbert space L
2
[−1, 1]
gives (apart from constant factors) the Legendre polynomials. Similarly the Hermite
and the Laguerre polynomials arise from orthonormalizing ¦x
n
e
−x
2
¦ and ¦x
n
e
−x
¦ in
the spaces L
2
[−∞, ∞] and L
2
[0, ∞] respectively.
Section 4. Some of the operators introduced in Section 2 should be examined for
compactness. In particular, the conditions for a diagonal operator on
2
to be compact
should be established.
Theorem 4.4 and Lemma 4.3 on which its proof depends are the only results in these
notes which are not strictly needed for what comes later. Note that this result is not
valid in general Banach spaces.
Section 5. The spectra of some specific operators should be identified. In particular,
the spectrum of M
φ
where φ(x) = x on L
2
[0, 1] should be identified as [0, 1] and the
fact that M
φ
has no eigenvalues should be noted.
Section 6. The fact that the Volterra operator has no eigenvalues should be es-
tablished, hence showing that some compact operators may have spectrum equal to
¦0¦.
It is useful to review the orthogonal diagonalization of real symmetric matrices and/or
unitary diagonalization of Hermitian (i.e. selfadjoint) matrices. It is instructive to
re-write both these results and Theorem 6.9 in terms of projections onto eigenspaces.
Section 7. This section is motivated by an informal discussion of of the Green’s
function as the response of a system to the input of a unit pulse. This is illustrated
by the elementary example of finding the shape under gravity of a heavy string (of
variable density) fixed at (0, 0) and (, 0). This is found by calculating the (triangular)
shape k(x, t) of a weightless string with a unit weight at x = t and then using an
integration process. The differential equation is also found and shown to give the
John Erdos 3
same answer. (Naturally, usual elementary applied mathematical assumptions - small
displacements, constant tension - apply.)
Additionally, a brief, very informal discussion of delta functions and the Green’s
function as the solution of the system with the function f being a delta function is
of interest.
It should be stressed, however, that the proof of Theorem 7.1 is purely elementary
and quite independent of the discussions above.
Section 8. The most important part of this final section is Theorem 8.3, the con-
tinuous functional calculus. This is sufficient for the vast majority of applications of
the spectral theorem for bounded self-adjoint operators. These include, for example,
the polar decomposition and the properties of e
At
and these are done in the course.
The approach here to the general spectral theorem is elementary and very pedestrian.
It should be noted that, given the appropriate background, there are more elegant
ways. These include using the identification of the dual of C[m, M] (actually of C(σ))
as a space of measures. There is also a Banach algebra treatment.
In an elementary course such as this, the technicalities of the spectral theorem need
not be strongly emphasized. However, a down to earth approach should clarify the
meaning of theorem and remove the mystery often attached by students to these
operator integrals.
4 Operators on Hilbert space
1 Elementary properties of Hilbert space
Definition A (complex) inner (scalar) product space is a vector space H together
with a map ' , ` : HH →C such that, for all x, y, z ∈ H and λ, µ ∈ C,
1. 'λx + µy, z` = λ'x, z` + µ'y, z`,
2. 'x, y` = 'y, x`,
3. 'x, x` ≥ 0, and 'x, x` = 0 ⇐⇒x = 0 .
Properties 1,2 and 3 imply
4. '(x, λy + µz` =
¯
λ'x, y` + ¯ µ'x, z`,
5. 'x, 0` = '0, x` = 0.
Theorem 1.1 (Cauchy-Schwartz inequality)
['x, y`[ ≤ 'x, x`
1/2
'y, y`
1/2
, ∀x, y ∈ H.
Proof. For all λ we have 'λx + y, λx + y` ≥ 0. That is, for real λ
λ
2
'x, x` + λ('x, y` +'y, x`) +'y, y` ≥ 0 .
In the case that 'x, y` is real, we have that the discriminant (“b
2
− 4ac”) of this
quadratic function of λ is negative which gives the result.
In general, put x
1
= e
−iθ
x where θ is the argument of the complex number 'x, y`.
Then 'x
1
, y` = e

'x, y` = ['x, y`[ is real and 'x
1
, x
1
` = 'x, x`. Applying the above to
x
1
, y gives the required result.
[Alternatively, put λ = −
y,x
x,x
in 'λx + y, λx + y` ≥ 0.]
Theorem 1.2
|x| = 'x, x`
1/2
is a norm on H.
Proof. The facts that |x| ≥ 0, |x| = 0 ⇐⇒x = 0 and |λx| = 'λx, λx`
1
2
= [λ[.|x|
are all clear from the equivalent properties of the inner product. For the triangle
inequality,
|x + y|
2
= |x|
2
+|y|
2
+'x, y` +'y, x`
≤ |x|
2
+|y|
2
+ 2['x, y`[
≤ |x|
2
+|y|
2
+ 2|x|.|y| using (1.1)
= (|x| +|y|)
2
.
Lemma 1.3 (Polarization identity)
'x, y` =
1
4

|x + y|
2
−|x −y|
2
+ i|x + iy|
2
−i|x −iy|
2

.
John Erdos 5
Proof.
|x + y|
2
= |x|
2
+|y|
2
+'x, y` +'y, x`
−|x −y|
2
= −|x|
2
+−|y|
2
+'x, y` +'y, x`
i|x + iy|
2
= i|x|
2
+ i|y|
2
+'x, y` −'y, x`
−i|x −iy|
2
= −i|x|
2
+−i|y|
2
+'x, y` −'y, x` .
Adding the above gives the result.
Lemma 1.4 (Paralellogram law)
|x + y|
2
+|x −y|
2
= 2|x|
2
+ 2|y|
2
.
Proof.
|x + y|
2
= |x|
2
+|y|
2
+'x, y` +'y, x`
|x −y|
2
= |x|
2
+|y|
2
−'x, y` −'y, x` .
Adding the above gives the result.
Definition x is said to be orthogonal to y if 'x, y` = 0; we write x ⊥ y.
Lemma 1.5 (Theorem of Pythagoras)
'x, y` = 0 =⇒|x|
2
+|y|
2
= |x + y|
2
.
Proof. Obvious.
Definition. If H is an inner product space and (H, | | ) is complete then H is called
a Hilbert space.
A set C (in a vector space) is convex if
x, y ∈ C =⇒αx + (1 −α)y ∈ C whenever 0 ≤ α ≤ 1 .
In a metric space, the distance from a point x to a set S is
d(x, S) = inf¦|x −s| : s ∈ S¦.
Theorem 1.6 If K is a closed convex set in a Hilbert space H and h ∈ H then there
exists a unique k ∈ K such that
d(h, K) = |h −k|.
Proof. Let C = K − h = ¦k − h : k ∈ K¦. Note that C is also closed and convex,
d(h, K) = d(0, C) and if c = k − h ∈ C is of minimal norm then k is the required
element of K. Therefore it is sufficient to prove the theorem for the case h = 0.
6 Operators on Hilbert space
Let d = d(0, C) = inf
c∈C
|c|. The |c| ≥ d for all c ∈ C. Choose a sequence (c
n
) such
that (|c
n
|) →d. Using the parallelogram law (Lemma 1.4),
|c
n
+ c
m
|
2
+|c
n
−c
m
|
2
= 2|c
n
|
2
+ 2|c
m
|
2
.
But, since C is convex,
c
n
+c
m
2
∈ C and so |
c
n
+c
m
2
| ≥ d; that is |c
n
+ c
m
|
2
≥ 4d
2
.
Therefore
0 ≤ |c
n
−c
m
|
2
= 2|c
n
|
2
+ 2|c
m
|
2
−|c
n
+ c
m
|
2
≤ 2(|c
n
|
2
+|c
m
|
2
) −4d
2
→0 (∗)
as n, m → ∞. It follows easily that (c
n
) is a Cauchy sequence. [ Since (|c
n
|) → d,
given > 0, there exists n
0
such that for n > n
0
, 2(|c
n
|
2
−d
2
) <

2
2
. Then (*) shows
that for n, m > n
0
, |c
n
−c
m
| < .] Since H is complete and C is closed, (c
n
) converges
to an element c ∈ C and |c| = lim
n→∞
|c
n
| = d.
To prove uniqueness, suppose also that c

∈ C with |c

| = d. The same calculation
as for (*) (with c
n
= c and c
m
= c

) shows that
0 ≤ |c −c

|
2
≤ 2|c|
2
+ 2|c

|
2
−|c + c

|
2
≤ 2d
2
+ 2d
2
−4d
2
= 0
and so c = c

.
Lemma 1.7 If N is a closed subspace of a Hilbert space H and h ∈ H then
d(h, N) = |h −n
0
| if and only if 'h −n
0
, n` = 0 for all n ∈ N .
Proof. Suppose d(h, N) = |h−n
0
|. Write z = h−n
0
. Then for all non-zero n ∈ N,
|z|
2

z −
'z, n`n
|n|
2

2
= |z|
2

2['z, n`[
2
|n|
2
+
['z, n`[
2
|n|
2
= |z|
2

['z, n`[
2
|n|
2
so 'z, n` = 0.
Conversely if h −n
0
⊥ N then, by Pythagoras (Lemma 1.5) for all n ∈ N,
|h −n|
2
= |h −n
0
+ n
0
−n|
2
= |h −n
0
|
2
+|n
0
−n|
2
≥ |h −n
0
|
2
.
Hence inf
n∈N
|h −n| is attained at n
0
.
Note that the above proof is putting a geometrical argument into symbolic form. The
quantity
z,nn
n
2
is the “resolution of the vector z in the direction of n”.
In these notes the term subspace (of a Hilbert space) will always mean a closed
subspace. The justification for this is that the prefix “sub” refers to a substructure;
so the subspace should be a Hilbert space in its own right, that is, it should be
John Erdos 7
complete. But it is an easy fact that a subset of a complete space is complete if and
only if it is closed.
Definition. Given a subset S of H the orthogonal complement S

is defined by
S

= ¦x : 'x, s` = 0 for all s ∈ S¦ .
Corollary 1.8 If N is a (closed) subspace of a Hilbert space H,
N

= (0) ⇐⇒N = H
Proof. Clearly, if N = H then N

= (0). For the converse, if N = H take h / ∈ N,
Then there is n
0
∈ N such that d(h, N) = |h − n
0
| and the Lemma shows that
0 = h −n
0
⊥ N, so N

= (0).
Lemma 1.9 For subsets of a Hilbert space H
(i) S

is a closed subspace,
(ii) S
1
⊇ S
2
=⇒S

1
⊆ S

2
,
(iii) S ⊆ S
⊥⊥
,
(iv) S

= S
⊥⊥⊥
,
(v) S ∩ S

= (0).
Proof. (i) Clearly S

is a vector subspace. To show it is closed, let t
n
∈ S

be a
sequence converging to t. Then, by the continuity of the inner product, for all s ∈ S,
't, s` = lim
n→∞
't
n
, s` = 0
so t ∈ S

. [In grim detail, ['t, s`[ = ['t − t
n
, s`[ ≤ |t − t
n
|.|s| → 0 , so, since 't, s`
does not depend on n, 't, s` = 0.]
(ii) and (iii) are clear. For (iv), apply (iii) to S

yields S

⊆ S
⊥⊥⊥
, and applying (ii)
to (iii) to gives the reverse inclusion. For (v), if x ∈ S ∩ S

then 'x, x` = 0 so x = 0.
Lemma 1.10 If M and N are orthogonal subspaces of a Hilbert space then M ⊕ N
is closed.
Proof. Note that since N ⊥ M, we have that N ∩ M = (0) and the sum M + N is
automatically direct. Let z
n
∈ M ⊕ N such that (z
n
) → z. We need to show that
z ∈ M⊕N. Now z
n
= x
n
+y
n
with x
n
∈ N and y
n
∈ M. Therefore, using Pythagoras
(Lemma 1.5) since M ⊥ N,
|z
n+p
−z
n
|
2
= |x
n+p
−x
n
|
2
+|y
n+p
−y
n
|
2
.
As (z
n
) is convergent, it is a Cauchy sequence. If follows easily from the above that
both (x
n
) and (y
n
) are Cauchy sequences so, since H is complete, (x
n
) and (y
n
) both
converge. Call the limits x and y. Then, since M and N are closed subspaces, x ∈ M
and y ∈ N. Thus z = lim(x
n
+ y
n
) = x + y ∈ M ⊕N.
8 Operators on Hilbert space
Theorem 1.11 If N is a subspace of a Hilbert space H then N ⊕N

= H.
Proof. From above, N ⊕ N

is a (closed) subspace. Also, if x ∈ (N ⊕ N

)

then
x ∈ N

∩ N
⊥⊥
so x = 0. Therefore, from Corollary 1.8, N ⊕N

= H.
Corollary 1.12
(i) If N is a subspace then N
⊥⊥
= N.
(ii) For any subset S of a Hilbert space H, S
⊥⊥
is the smallest subspace containing
S.
Proof. (i) From Lemma 1.9 (iii) N ⊆ N
⊥⊥
. Since H = N ⊕ N

, if x ∈ N
⊥⊥
then
x = s + t with s ∈ N and t ∈ N

. But then also t = x − s ∈ N
⊥⊥
, so t = 0 and
x = s ∈ N.
(ii) Clearly S
⊥⊥
is a subspace containing S. If M is any subspace containing S then
(Lemma 1.9 (ii)) S
⊥⊥
⊆ M
⊥⊥
= M.
John Erdos 9
Exercises 1
1. For a Hilbert space H, show that the inner product, considered as a map from HH
to C, is continuous.
2. Let H
1
and H
2
be Hilbert spaces. Let H be the set of ordered pairs H
1
H
2
with
addition and multiplication defined (in the usual way) as follows:
(h
1
, h
2
) + (g
1
, g
2
) = (h
1
+ g
1
, h
2
+ g
2
)
α(h
1
, h
2
) = (αh
1
, αh
2
).
Show that the inner product defined by
'(h
1
, h
2
), (g
1
, g
2
)` = 'h
1
, g
1
` +'h
2
, g
2
`
satisfies the axioms for an inner product and H with this inner product is a Hilbert
space. [H is called the (Hilbert space) direct sum of H
1
and H
2
. One writes H =
H
1
⊕H
2
.]
3. Prove that in the Cauchy–Schwarz inequality ['x, y`[ ≤ |x||y| the equality holds iff
the vectors x and y are linearly dependent.
4. For which real α does the function f(t) = t
α
belong to
(i) L
2
[0, 1] (ii) L
2
[1, ∞] (iii) L
2
[0, ∞] ?
5. Let x and y be vectors in an inner product space. Given that | x+y |=| x | + | y |,
show that x and y are linearly dependent.
6. Let W[0, 1] be the space of complex-valued functions which are continuously differ-
entiable on [0, 1]. Show that,
'f, g` =

1
0
¦f(t)g(t) + f

(t)g

(t)¦ dt
defines an inner product on W[0, 1].
7. Prove that in a complex inner product space the following equalities hold:
'x, y` =
1
N
N
¸
k=1
|x + e
2πik/N
y|
2
e
2πik/N
for N ≥ 3,
'x, y` =
1


0
|x + e

y|
2
e

dθ .
[This generalizes the polarization identity.]
8. Let M and N be closed subspaces of a Hilbert space. Show that
(i) (M + N)

= M

∩ N

(ii) (M ∩ N)

= M

+ N

.
9. Show that the vector subspace of
2
spanned by sequences of the form (1, α, α
2
, α
3
, . . .),
where 0 ≤ α < 1, is dense in
2
.
A challenging but not very important exercise :
10. Show that, for any four points of a Hilbert space,
| x −z | . | y −t |≤| x −y | . | z −t | + | y −z | . | x −t | .
10 Operators on Hilbert space
2 Linear Operators.
Some of the results in this section are stated for normed linear spaces but they will
be used in the sequel only for Hilbert spaces.
Lemma 2.1 Let X and Y be normed linear spaces and let L : X → Y be a linear
map. Then the following are equivalent :
1. L is continuous;
2. L is continuous at 0;
3. there exists a constant K such that |Lx| ≤ K|x| for all x ∈ X.
Proof. 1 implies 2 is obvious. If 2 holds, take any > 0. Continuity at 0 shows that
there is a corresponding δ > 0 such that |LX| < whenever |x| < δ. Take some c
with 0 < c < δ. Then for any x = 0,

cx
x

= c < δ and so

L

cx
|x|

= c
|Lx|
|x|
< .
This shows that |Lx| < K|x| where K =

c
.
If 3 holds, to show continuity at any point x
0
, note that
|Lx −Lx
0
| = |L(x −x
0
)| ≤ K|x −x
0
| .
Therefore, given any > 0, let δ =

K
. Then if |x−x
0
| < δ we have |Lx−Lx
0
| < .
The set of all continuous (bounded) linear maps X → Y is denoted by B(X, Y ).
When X = Y we write B(X).
For L ∈ B(X, Y ), define |L| = sup
x=0
Lx
x
.
Exercise. | | is a norm on B(X, Y ) and
|L| = sup
x=0
|Lx|
|x|
= sup
x≤1
|Lx| = sup
x=1
|Lx|.
If Y is complete then so is B(X, Y )
When Y = C then B(X, C) is called the dual of X and denoted by X

(sometimes by
X

). The elements of the dual are called (continuous) linear functionals.
We shall be concerned with Hilbert spaces; H will always denote a Hilbert space.
Theorem 2.2 (Riesz representation theorem) Every linear functional f on H is of
the form
f(x) = 'x, h`
for some h ∈ H, where |f| = |h|.
John Erdos 11
Proof. If f = 0, take h = 0. For f = 0 then N = f
−1
(0) = ¦x : f(x) = 0¦ = H.
Also, since f is continuous, N is closed. Thus N

= (0) so take y ⊥ N. Then
f(y) = 0. Write z =
y
f(y)
so that f(z) = 1 [using f(αx) = αf(x)]. For any x ∈ H
f (x −f(x)z) = f(x) −f(x).f(z) = 0 and so x −f(x)z ∈ N .
Since z ⊥ N,
'x −f(x)z, z` = 'x, z` −f(x)|z|
2
= 0.
Writing h =
z
z
2
we obtain
f(x) = 'x, h` .
For the norm, note that [f(x)[ = ['x, h`[ ≤ |x|.|h| so |f| ≤ |h|. Also
|f[ = sup
x=0
[f(x)[
|x|

[f(h)[
|h|
= |h| .
Note that the result |f| = |h| shows that the correspondence between H and its
dual is one to one.
Lemma 2.3 (Polarization identity for operators)
'Ax, y` =
1
4
['A(x + y), (x + y)` −'A(x −y), (x −y)`
+i'A(x + iy), (x + iy)` −i'A(x −iy), (x −iy)`] .
Proof.
'A(x + y), (x + y)` = 'Ax, x` +'Ay, y` +'Ax, y` +'Ay, x`
−'A(x −y), (x −y)` = −'Ax, x` −'Ay, y` +'Ax, y` +'Ay, x`
i'A(x + iy), (x + iy)` = i'Ax, x` + i'Ay, y` +'Ax, y` −'Ay, x`
−i'A(x −iy), (x −iy)` = −i'Ax, x` −i'Ay, y` +'Ax, y` −'Ay, x` .
Adding the above gives the result.
Corollary 2.4 If 'Ax, x` = 0 for all x ∈ H then A = 0.
Proof. If 'Ax, x` = 0 for all x ∈ H the above shows that 'Ax, y` = 0 for all x, y ∈ H
and so using y = Ax it follows that |Ax|
2
= 0 for all x ∈ H. Thus A = 0.
Definition Let H be a Hilbert space. A bilinear form (also called a sesquilinear
form) φ on H is a map φ : HH →C such that
φ(αx + βx

, y) = αφ(x, y) + βφ(x

, y)
φ(x, αy + βy

) = ¯ αφ(x, y) +
¯
βφ(x, y

) .
A bilinear form is said to be bounded if, for some constant K, [φ(x, y)[ ≤ K|x|.|y|
for all x, y ∈ H.
Theorem 2.5 (Riesz) Every bounded bilinear form φ on H is of the form
φ(x, y) = 'Ax, y`
for some A ∈ B(H).
12 Operators on Hilbert space
Proof. Consider x fixed for the moment. Then φ(x, y) is conjugate linear in y, so
that φ(x, y) is linear in y. Using Theorem 2.2 we have that there is a (unique) h ∈ H,
φ(x, y) = 'y, h` , that is φ(x, y) = 'h, y` .
One can find such an h corresponding to each x ∈ H. Define a function H → H by
Ax = h. Then A is linear since , for all x, x

, y,
'A(x + x

), y` = φ((x + x

), y) = φ(x, y) + φ(x

, y) = 'Ax, y` +'Ax

, y`
so A(x + x

) = Ax + Ax

[since A(x + x

) − Ax − Ax

∈ H

= (0)]. Similarly
A(αx) = αAx. Also,
|Ax| = sup
y=0
'Ax, y`
|y|
= sup
y=0
[φ(x, y)[
|y|
≤ K|x|
so A is continuous.
Definition The adjoint. Let A ∈ B(H). Then ψ(x, y) = 'x, Ay` is a bounded bilinear
form on H so, by Theorem 2.5 there is an operator A

∈ B(H) such that
'A

x, y` = ψ(x, y) = 'x, Ay` .
A

is called the adjoint of A.
Exercise.
(i) (A

)

= A, (ii) (λA)

=
¯
λA

,
(iii) (A + B)

= A

+ B

, (iv) (AB)

= B

A

,
(v) |A| = |A

|.
Note. Bilinear forms could have been defined as maps φ from H / to C where
H and / are different Hilbert spaces. All the above can be done with essentially no
change; (the adjoint of A ∈ B(H, /) is then an operator in B(/, H)).
Definition.
If A = A

then A is said to be selfadjoint.
If AA

= A

A then A is said to be normal.
If UU

= U

U = I then U is said to be unitary.
John Erdos 13
Projections.
Let N be a closed subspace of H. Then from Theorem 1.11,
H = N ⊕N

that is, any h ∈ H has a unique decomposition as h = x+y with x ∈ N and y ∈ N

.
The orthogonal projection P onto N is defined by Ph = x (where h = x + y is the
decomposition above). Note that then y = (I −P)h and I −P is the projection onto
N

.
In this course we shall not consider projections that are not orthogonal and usually
call these operators “projections”.
Lemma 2.6 Let N be a closed subspace of H and let P be the orthogonal projection
onto N. Then
(i) P is linear,
(ii) |P| = 1 (unless N = 0),
(iii) P
2
= P,
(iv) P

= P.
Also, if E ∈ B(H) satisfies E = E
2
= E

then E is the (orthogonal) projection onto
some (closed) subspace.
Proof. (i) Let h, h

∈ H and suppose h = x + y and h

= x

+ y

are the
unique decompositions of h and h

with x, x

∈ N and y, y

∈ N

. Then αh + βh

=
(αx + βx

) + (αy + βy

) is the decomposition of αh + βh

and
P(αh + βh

) = αx + βx

= αPh + βPh

.
(ii) If h = x + y with x ∈ N and y ∈ N

,
|Ph|
2
= |x|
2
≤ |x|
2
+|y|
2
= |h|
2
and so |P| ≤ 1. But if 0 = h ∈ N then Ph = h and so |P| = 1.
(iii) If h ∈ N [then h = h + 0 is the decomposition of h and] Ph = h. But for any
h ∈ H, Ph ∈ N so P(Ph) = Ph, that is, P
2
= P.
(iv) If h = x + y and h

= x

+ y

with x, x

∈ N and y, y

∈ N

.
'Ph, h

` = 'x, x

+ y

` = 'x, x

`
since x ∈ N and y

∈ N

. Similarly 'h, Ph

` = 'x, x

` and so P = P

.
Finally, if E ∈ B(H) satisfies E = E
2
= E

let N = ¦x : Ex = x¦. Then N =
ker(I −E), so N is closed. For any h ∈ H, write
h = Eh + (I −E)h.
Then Eh ∈ N since E(Eh) = E
2
h = Eh and (I −E)h ⊥ N since if x ∈ N, Ex = x
and
'(I −E)h, x` = '(I −E)h, Ex` = 'E

(I −E)h, x` = '(E
2
−E)h, x` = 0.
This shows that E is the projection onto N.
14 Operators on Hilbert space
Lemma 2.7 If P is the orthogonal projection onto a subspace N then for all h ∈ H,
d(h, N) = |(I −P)h|.
Proof. For any h ∈ H we have Ph ∈ N and '(I − P)h, n` = 0 for all n ∈ N.
Therefore from Lemma 1.7
d(h, N) = |h −Ph| = |(I −P)h| .
Lemma 2.8 Let A ∈ B(H) and P be the orthogonal projection onto a subspace N.
(i) N is invariant under A ⇐⇒AP = PAP.
(ii) N

is invariant under A ⇐⇒PA = PAP.
If A = A

then N is invariant under A ⇐⇒N

is invariant under A ⇐⇒PA = AP.
Proof. (i) =⇒ Suppose An ∈ N for all n ∈ N. Then since Ph ∈ N for all h ∈ H,
we have APh ∈ N. Therefore then PAPh = APh [since Pn = n for all n ∈ N].
⇐= If n ∈ N then Pn = n and so An = APn = PAPn ∈ N [since N is the range of
P].
(ii) The projection onto N

is I −P. Trivial algebra shows that
A(I −P) = (I −P)A(I −P) ⇐⇒PA = PAP
and so (ii) follows from (i).
Finally AP = PAP ⇐⇒ (AP)

= (PAP)

⇐⇒ PA = PAP since A = A

. If these
equalities hold then PA = AP.
John Erdos 15
Exercises 2
1. Let X ∈ B(H). Show that :
(i) X is selfadjoint ⇐⇒ 'Xx, x` is real for all x ∈ H,
(ii) X is normal ⇐⇒| Xx |=| X

x | for all x ∈ H,
(iii) X is unitary ⇐⇒| Xx |=| X

x |=| x | for all x ∈ H.
2. Let S be the one-dimensional subspace of
2
spanned by the element (1, −1, 0, 0, . . .).
Show explicitly that any element x = (ξ
k
) ∈
2
can be written as x = x
1
+ x
2
where
x
1
∈ S and x
2
⊥ S.
3. Let A be a selfadjoint operator such that for all x ∈ H, | Ax |≥ c | x |, where c is
a positive constant. Show that A has a continuous inverse.
[Hints : Show (i) A is injective, (ii) the range of A is closed (iii) (ran(A))

= (0).]
Note that the selfadjointness condition is needed – consider the operator S on
2
defined by S(ξ
1
, ξ
2
, ξ
3
, . . .) = (0, ξ
1
, ξ
2
, ξ
3
, . . .).
4. The operators D and W on
2
are defined by
D(ξ
1
, ξ
2
, ξ
3
, . . .) = (α
1
ξ
1
, α
2
ξ
2
, α
3
ξ
3
, . . .) ,
W(ξ
1
, ξ
2
, ξ
3
, . . .) = (0, α
1
ξ
1
, α
2
ξ
2
, α
3
ξ
3
, . . .) ,
where (α
n
) is a bounded sequence of complex numbers. Show that W and D are
bounded operators and find their adjoints.
5. Given that X ∈ B(H) is invertible (that is, there exists X
−1
∈ B(H) such that
XX
−1
= X
−1
X = I) prove that X

is invertible and (X

)
−1
= (X
−1
)

.
6. Find the adjoint of the operator V defined on L
2
[0, 1] by (V f)(x) =

x
0
f(t) dt .
7. Let T : L
2
[0, 1] →L
2
[0, 1] be defined by
(Tf)(x) =

2xf(x
2
) .
Find the adjoint of T and deduce that T is unitary.
8. Let E, F be the orthogonal projections onto subspaces M and N respectively. Prove
that,
(i) EF = F ⇐⇒ N ⊆ M ⇐⇒ E −F is an orthogonal projection,
(ii) EF = 0 ⇐⇒ N ⊆ M

⇐⇒ E + F is an orthogonal projection,
(iii) EF = FE ⇐⇒ E + F −FE is an orthogonal projection.
9. The operator A ∈ B(H) satisfies Ax = x for some x ∈ H. Prove that A

x = x + y
where y ⊥ x. If, further, |A| ≤ 1, show that A

x = x.
Suppose that E
2
= E and |E| = 1. Use the above to show that ran(E) = ran(E

)
and ker(E) = ker(E

) and deduce that E = E

(so that E is the orthogonal projection
onto some subspace of H).
16 Operators on Hilbert space
10. Let L
o
and L
e
be subspaces of L
2
[−1, 1] defined by
L
o
= ¦f : f(t) = −f(−t) (almost everywhere )¦
L
e
= ¦f : f(t) = f(−t) (almost everywhere )¦.
Show that L
o
⊕ L
e
= L
2
[−1, 1] and find the projections of L
2
[0, 1] onto L
o
and L
e
.
Find expressions for the distances of any element f to L
o
and L
e
. Calculate the values
in the specific case where f(t) = t
2
+ t.
11. Let M and N be vector subspaces of H such that M ⊥ N and M
˙
+N = H. Prove
that M and N are closed.
12. Show that the set of sequences x = (ξ
n
) such that
¸
n
n
2

n
[
2
converges, forms a dense
subset of l
2
.
Define the operator D on l
2
by
D(ξ
1
, ξ
2
, ξ
3
, . . .) = (ξ
1
,
1
2
ξ
2
,
1
3
ξ
3
, . . .) ,
and let M and N be linear subspaces of l
2
⊕l
2
defined by
M = ¦(x, 0) : x ∈ l
2
¦ N = ¦(x, Dx) : x ∈ l
2
¦ .
Observe that M is closed and use the continuity of D to show that N is also closed.
Show that M ∩ N = (0) and that the algebraic direct sum of M and N is dense in
l
2
⊕l
2
but is not equal to l
2
⊕l
2
(and so it is not closed).
John Erdos 17
3 Orthonormal Sets.
Definition. A set o of vectors of H is said to be orthonormal if
1. |x| = 1 for all x ∈ o,
2. 'x, y` = 0 if x = y and x, y ∈ o.
Lemma 3.1 (Bessel’s inequality) If ¦x
i
: 1 ≤ i ≤ n¦ is a finite orthonormal set then,
for any h ∈ H, writing α
i
= 'h, x
i
`,
n
¸
i=1

i
[
2
≤ |h|
2
.
(Note that the case n = 1 is the Cauchy-Schwartz inequality)
Proof. Let h ∈ H. Then
|h −
n
¸
i=1
α
i
x
i
|
2
= 'h −
n
¸
i=1
α
i
x
i
, h −
n
¸
i=1
α
i
x
i
`
= |h|
2
−'h,
n
¸
i=1
α
i
x
i
` −'
n
¸
i=1
α
i
x
i
, h` +
n
¸
i,j=1
α
i
α
j
'x
i
, x
j
`
= |h|
2
−'h,
n
¸
i=1
α
i
x
i
` −'
n
¸
i=1
α
i
x
i
, h` +
n
¸
i=1

i
[
2
= |h|
2

n
¸
i=1

i
[
2
≥ 0.
Lemma 3.2 If ¦x
i
: i = 1, 2, 3 ¦ is an orthonormal sequence then, for any h ∈ H,
writing α
i
= 'h, x
i
`,

¸
i=1
α
i
x
i
converges to a vector h

such that 'h −h

, x
i
` = 0 for all i.
Proof. Put h
r
=
¸
r
i=1
α
i
x
i
. Then
|h
r+p
−h
r
|
2
=

r+p
¸
i=r+1
α
i
x
i

2
=
r+p
¸
i=r+1

i
[
2
.
Now
¸
n
i=1

i
[
2
≤ |h|
2
for all n and so
¸

i=1

i
[
2
is convergent and so is a Cauchy
series. Hence, given any > 0 there exists n
0
such that for r > n
0
, p > 0 we have
¸
r+p
i=r+1

i
[
2
<
2
; that is, |h
r+p
−h
r
| < . Therefore (h
r
) is a Cauchy sequence and,
since H is complete, it is convergent. Call its limit h

.
For any fixed i, 'h − h
r
, x
i
` = 0 for all r > i. Now let r → ∞. Then it follows from
the continuity of the inner product that 'h −h

, x
i
` = 0 for all i.
18 Operators on Hilbert space
Theorem 3.3 Let ¦x
i
: i = 1, 2, 3 ¦ be an orthonormal sequence. The the following
statements are equivalent.
(i) ¦x
i
: i = 1, 2, 3 ¦ is maximal (that is, it is not a proper subset of any or-
thonormal set).
(ii) If α
i
= 'h, x
i
` = 0, for all i then h = 0.
(iii) (Fourier expansion) For all h ∈ H we have h =
¸

i=1
α
i
x
i
.
(iv) (Parseval’s relation) For all h, g ∈ H we have 'h, g` =
¸

i=1
α
i
β
i
.
(v) (Bessel’s equality) For all h ∈ H we have |h|
2
=
¸

i=1

i
[
2
.
(In the above, α
i
= 'h, x
i
` and β
i
= 'g, x
i
`.)
Proof. (i) =⇒ (ii). If (ii) is false then adding
h
h
to the set ¦x
i
: i = 1, 2, 3 ¦ gives
a larger orthonormal set, contradicting (i).
(ii) =⇒ (iii). Let h

=
¸

i=1
α
i
x
i
(this exists, by Lemma 3.2). Then 'h − h

, x
i
` = 0
for all i and so h = h

by (ii)¿
(iii) =⇒ (iv). Let h
r
=
¸
r
i=1
α
i
x
i
and g
s
=
¸
s
i=1
β
i
x
i
. Then
'h
r
, g
s
` =
min[r,s]
¸
i=1
α
i
β
i
.
Let r →∞ and s →∞. Using the continuity of the inner product, it follows that
'h, g` =

¸
i=1
α
i
β
i
.
(iv) =⇒ (v). Put g = h in (iv).
(v) =⇒ (i). If ¦x
i
: i = 1, 2, 3 ¦ is not maximal and can be enlarged by adding z
then 'z, x
i
` = 0 for all i but also
1 = |z|
2
=

¸
i=1
['z, x
i
`|
2
= 0
which give a contradiction.
Definition. A maximal orthonormal sequence is called an orthonormal basis.
Clearly the concept of an orthonormal basis refers to a set of vectors so that any
permutation of such a set is still an orthonormal basis. It follows that the series giving
the fourier expansion of a vector can be re-arranged arbitrarily without altering its
convergence or its sum. Such a series is said to be unconditionally convergent.
John Erdos 19
Theorem 3.4 (Gram-Schmidt process) Let ¦y
i
: i = 1, 2, 3 ¦ be a sequence of
vectors of H. Then there exists an orthonormal sequence ¦x
i
: i = 1, 2, 3 ¦ such
that, for each integer k
span¦x
1
, x
2
, x
3
x
k
¦ ⊇ span¦y
1
, y
2
, y
3
y
k
¦ .
If ¦y
i
: i = 1, 2, 3 ¦ is a linearly independent set, then the above inclusion is an
equality for each k.
Proof. First consider the case when ¦y
i
: i = 1, 2, 3 ¦ is a linearly independent set.
Define
u
1
= y
1
x
1
=
u
1
|u
1
|
,
u
2
= y
2
−'y
2
, x
1
`x
1
x
2
=
u
2
|u
2
|
,
.
.
.
.
.
.
u
n
= y
n

n−1
¸
i=1
'y
n
, x
i
`x
i
x
n
=
u
n
|u
n
|
.
Easy inductive arguments show that for each k, x
k−1
∈ span¦y
1
, y
2
, y
k−1
¦ and,
since ¦y
i
: i = 1, 2, 3 ¦ is linearly independent, that u
k
= 0. A further easy
induction shows that for r < n we have 'u
n
, x
r
` = 0 and so it follows easily that
¦x
i
: i = 1, 2, 3 ¦ is an orthonormal sequence. In the general case, the same
construction applies except that whenever y
r
is a linear combination of y
1
, y
2
, , y
r−1
this element is ignored.
It is clear in general that x
r
is a linear combination of on y
1
, y
2
, , y
r−1
and so the
inclusion
span¦x
1
, x
2
, x
3
x
k
¦ ⊇ span¦y
1
, y
2
, y
3
y
k
¦ .
is obvious. When ¦y
i
: i = 1, 2, 3 ¦ is a linearly independent set we have equality,
since both sides have dimension k.
For the rest of this course we shall often need to assume that the Hilbert space we
consider has a countable orthonormal basis. This is true for all spaces considered in
the applications. The restriction is a rather technical matter and could be avoided
but this would entail a discussion of uncountable sums. A few statements would also
need modification.
The appropriate way to state this restriction is to say that the Hilbert space we
consider is separable. For our purposes one could say that a Hilbert space is separable
if it has a countable orthonormal basis, and take this as the definition of separability.
However, this is a more general notion: recall that a metric space is defined to be
separable if it has a countable, dense subset. The proposition which follows connects
these ideas.
20 Operators on Hilbert space
Proposition
1. In a separable Hilbert space, every orthonormal set is countable.
2. A Hilbert space is separable if and only if it has a (countable) orthonormal basis.
We shall not prove this in detail, but here is a sketch. For each element x
α
of an
orthonormal set ¦x
α
¦
α∈A
, let B
α
be the open ball centre x
α
, radius

2
2
. Since these
balls are disjoint and since every open set must contain at least one element of a
dense subset, it is clear that if the space is separable the orthonormal set must be
countable. For 2, applying the Gram-Schmidt process to a countable dense subset
results in an orthonormal sequence that is easily proved to be a basis. Conversely, if
¦x
n
: n = 1, 2, 3, ¦ is a countable orthonormal basis, tedious but routine arguments
show that the set
S = ¦
N
¸
n=1
r
n
x
n
: N finite , r
n
rational ¦
is countable. It is clearly dense because the closure of S is a subspace containing an
orthonormal basis.
John Erdos 21
Exercises 3
1. Let ¦N
i
¦ be a sequence of mutually orthogonal subspaces of a Hilbert space H and
let ¦E
i
¦ be the sequence of projections onto ¦N
i
¦. Show that for each x ∈ H,
(i) for a finite subset ¦N
i
¦
n
i=1
of ¦N
i
¦,
¸
n
i=1
| E
i
x |
2
≤| x |
2
.
(ii)
¸

i=1
E
i
x converges to some h ∈ H which satisfies (x −h) ⊥ N
i
for each i.
Show further that the following are equivalent :
(a) ¦N
i
¦ is not a proper subset of any orthogonal set of subspaces of H,
(b) h ⊥ N
i
for all i ⇒h = 0,
(c) for each x ∈ H, x =
¸

i=1
E
i
x,
(d) for each x, y ∈ H, 'x, y` =
¸

i=1
'E
i
x, E
i
y`,
(e) for each x ∈ H, | x |
2
=
¸

i=1
| E
i
x |
2
.
[Hints : You will need Q. 8 (ii) of Sheet 2 or its equivalent. Note that under the given
conditions, since E
i
= E

i
= E
2
i
, 'E
i
x, E
j
y` = 'E
j
E
i
x, y` = 0 if i = j.]
2. Find the first three functions obtained by applying the Gram- Schmidt process to
the elements ¦t
n
: n = 0, 1, . . .¦ of L
2
[−1, 1]. [Note: apart from constant factors, this
process yields the Legendre polynomials.] Use your results and the theory developed
in lectures to find a, b and c which minimises the quantity

1
−1
[t
4
−a −bt −ct
2
[
2
dt .
3. Let N be a subspace of L
2
[0, 1] with the property that for some fixed constant C and
each f ∈ N,
[f(x)[ ≤ C|f| almost everywhere .
Prove that N is finite dimensional.
[Hint: for any orthonormal subset f
1
, f
2
, . . . f
n
, of N, evaluate, for any fixed y, the
norm of g where
g(x) =
n
¸
i=1
f
i
(y)f
i
(x) .
Deduce that
¸
n
i=1
[f
i
(y)[
2
≤ C
2
and integrate this relation with respect to y.]
22 Operators on Hilbert space
4 Compact Operators.
Definition. An operator K ∈ B(H) is said to be compact if for every bounded set o
of vectors of H the set ¦Ks : s ∈ o¦ is compact.
Equivalently :
Definition. An operator K ∈ B(H) is said to be compact if for every bounded
sequence (x
n
) of vectors of H the sequence (Kx
n
) has a convergent subsequence.
We shall denote the set of all compact operators on H by /(H).
Definition. The rank of an operator is the dimension of its range.
Note that every operator of finite rank is compact. This is an immediate consequence
of the Bolzano-Weierstrass theorem which states that every bounded sequence in C
n
has a convergent subsequence. Note also that the identity operator on a Hilbert space
H is compact if and only if H is finite-dimensional.
Theorem 4.1 /(H) is an ideal of B(H).
Proof. We need to show that, if A, B ∈ /(H) and T ∈ B(H) then αA, A + B, TA
and AT are all in /(H). That is, for any a bounded sequence (x
n
), we must show
that (αAx
n
), ([A + B]x
n
), (TAx
n
) and (ATx
n
) all have convergent subsequences.
Since A is compact, (Ax
n
) has a convergent subsequence (Ax
n
i
). Then clearly (αAx
n
i
)
is a convergent subsequence of (αAx
n
) showing that αA is compact. Also, (x
n
i
) is a
bounded sequence and so, since B is compact, (Bx
n
i
) has a convergent subsequence
(Bx
n
i
j
). Then ([A+B]x
n
i
j
) is a convergent subsequence of ([A+B]x
n
), showing that
A + B is compact.
Again, since T ∈ B(H), T is continuous and so (TAx
n
i
) is a convergent subsequence
of (TAx
n
) showing that TA is compact. The proof for AT is slightly different. Here,
since (x
n
) is bounded and |Tx
n
| ≤ |T|.|x
n
| we have that (Tx
n
) is bounded and
so, since A is compact, (ATx
n
) has a convergent subsequence, showing that TA is
compact.
A consequence of the above theorem is that, it H is infinite-dimensional then and
T ∈ B(H) has an inverse T
−1
∈ B(H) then T is not compact.
Theorem 4.2 /(H) is closed.
Proof. Let (K
n
) be a sequence of compact operators converging to K. To show that
K is compact, we need to show that if (x
i
) is a bounded sequence the (Kx
i
) has a
convergent subsequence.
Let (x
1
i
) be a subsequence of (x
i
) such that (K
1
x
1
i
) is convergent,
let (x
2
i
) be a subsequence of (x
1
i
) such that (K
2
x
2
i
) is convergent,
let (x
3
i
) be a subsequence of (x
2
i
) such that (K
3
x
3
i
) is convergent,
and continue in this way.
[The notation above is slightly unusual and is adopted to avoid having to use sub-
scripts on subscripts on .]
Let z
i
= x
i
i
. Then (z
i
) is a subsequence of (x
i
). Also, for each n, apart from the first
n terms, (z
i
) is a subsequence of (x
n
i
) and so (K
n
z
i
) is convergent.
John Erdos 23
We now show that (Kz
i
) is convergent by showing that it is a Cauchy sequence. For
all i, j, n we have
|Kz
i
−Kz
j
| = |(K −K
n
)z
i
+ K
n
z
i
−K
n
z
j
−(K −K
n
)z
j
|
≤ |K −K
n
|(|z
i
| +|z
j
|) +|K
n
(z
i
−z
j
)| .
Let > 0 be given. Since (K
n
) → K we can find n
0
such that |K − K
n
| <

4c
for
n > n
0
where c satisfies |x
i
| ≤ c for the bounded sequence (x
i
). Choose one fixed
such n. Now, since (K
n
z
i
) converges, it is a Cauchy sequence and so there is an
i
0
such that for i > i
0
, j > i
0
we have |K
n
z
i
− K
n
z
j
| <

2
. Combining these with
the displayed inequality shows that for i > i
0
, j > i
0
, |Kz
i
− Kz
j
| < so (Kz
i
) is
convergent as required.
Example. The operator K on L
2
[a, b] defined by
(Kf)(x) =

b
a
k(x, t)f(t) dt ,
where

b
a

b
a
[k(x, t)[
2
dxdt = M
2
< ∞, is compact.
We have already seen that operators of the above type are continuous with |K| ≤ M
(Recall that k(x, t) is called the kernel of the integral operator K). We shall show
that K is the norm limit of a sequence of finite rank operators. Note that if k(x, t) is
of the form u(x)v(t) then
(Rf)(x) =

b
a
u(x)v(t)f(t) dt = 'f, v`u = (v ⊗u)f
is a rank one operator.
Let S be the square [a, b] [a, b]. We shall apply Hilbert space theory to L
2
(S) which
is a Hilbert space of functions of 2 variables with the inner product
'φ, ψ` =

b
a

b
a
φ(x, t)ψ(x, t) dxdt .
Let (u
i
) be an orthonormal basis of L
2
[a, b]. Then (u
i
(x)u
j
(t))

i,j=1
is an orthonormal
basis of L
2
(S). Indeed,
'u
i
(x)u
j
(t), u
k
(x)u
l
(t)` =

b
a

b
a
u
i
(x)u
j
(t)u
k
(x)u
l
(t) dxdt
=

b
a
u
i
(x)u
k
(x) dx

b
a
u
j
(t)u
l
(t) dt = 0
unless i = k and j = l, in which case the integral is 1. Thus (u
i
(x)u
j
(t))

i,j=1
is an
orthonormal sequence. To show that it is a basis, suppose φ(x, t) ⊥ u
i
(x)u
j
(t) for all
i, j. Then
0 =

b
a

b
a
φ(x, t)u
i
(x)u
j
(t) dxdt =

b
a

b
a
φ(x, t)u
i
(x) dx

u
j
(t) dt .
This shows that, for each i, the function

b
a
φ(x, t)u
i
(x) dx of t is orthogonal to u
j
(t)
for each j. Therefore, since (u
j
) is a basis of L
2
[a, b], it is (equivalent to) the zero
function. Then, for fixed t the function φ(x, t) is orthogonal to u
i
(x) for each i and
so it is zero.
24 Operators on Hilbert space
Returning to the operator K, note that k ∈ L
2
(S). Therefore, by Theorem 3.3 (iii)
it has a fourier expansion using the basis (u
i
u
j
) of the type
k(x, t) =

¸
i,j=1
α
ij
u
i
(x)u
j
(t) .
Thus, writing k
n
(x, t) =
¸
n
i,j=1
α
ij
u
i
(x)u
j
(t) and
(K
n
f)(x) =

b
a
k
n
(x, t)f(t) dt ,
we have that K
n
is a finite rank operator (of rank at most n
2
). Note that K −K
n
is
an integral operator (of the same type as K) with kernel k(x, t) −k
n
(x, t). Thus
|K −K
n
|
2

b
a

b
a
[k(x, t) −k
n
(x, t)[
2
dxdt = |k −k
n
|
2
L
2
(S)
and the right hand side →0. Therefore Theorem 4.2 shows that K is compact.
Lemma 4.3 Let K be a compact operator on H and suppose (T
n
) is a bounded se-
quence in B(H) such that, for each x ∈ H the sequence (T
n
x) converges to Tx, where
T ∈ B(H). Then (T
n
K) converges to TK in norm.
Briefly, the above can be rephrased as :
If K ∈ /(H) and |T
n
x −Tx| →0 for all x ∈ H then |T
n
K −TK| →0.
In words : multiplying by a compact operator on the right converts a pointwise
convergent sequence of operators into a norm convergent one.
Proof. Since (T
n
) is a bounded sequence, |T
n
| ≤ C for some constant C. Then for
all x ∈ H, |Tx| = lim
n
|T
n
x| ≤ C|x| and so |T| ≤ C.
Let K be compact and suppose that |TK−T
n
K| →0. Then there exists some δ > 0
and a subsequence (T
n
i
K) such that |TK − T
n
i
K| > δ. Choose unit vectors (x
n
i
)
of H such that |(TK −T
n
i
K)x
n
i
| > δ. [That this can be done follows directly from
the definition of the norm of an operator.] Using the fact that K is compact, we can
find a subsequence (x
n
j
) of (x
n
i
) such that (Kx
n
j
) is convergent. Let the limit of this
sequence be y. Then for all j
δ < |(TK −T
n
j
K)x
n
j
| ≤ |(T −T
n
j
)(Kx
n
j
−y)| +|(T −T
n
j
)y| .
Now, using the convergence of (Kx
n
j
) to y, there exists n so that, for n
j
> n,
|Kx
n
j
−y| <
δ
8C
. Also, using the convergence of (T
n
j
) to T, there exists m so that,
for n
j
> m, |(T − T
n
j
)y| <
δ
4
. Then, for j > max[n, m] the right hand side of the
displayed inequality is less than
δ
2
, and this contradiction shows that the supposition
that |TK −T
n
K| →0 is false.
John Erdos 25
The theorem below is true for all Hilbert spaces, but we shall only prove it for the
case when the space is separable.
Theorem 4.4 Every compact operator on H is a norm limit of a sequence of finite
rank operators.
Proof. Let x
i
be an orthonormal basis of H. Define P
n
by
P
n
h =
n
¸
i=1
'h, x
i
`x
i
.
[Note that P
n
is the projection onto span x
1
, x
2
, , x
n
. Also, P
n
could be written
as P
n
h =
¸
n
i=1
x
i
⊗ x
i
.] From Theorem 3.3(iii), for all x ∈ H, P
n
x converges to x
(that is, P
n
converges pointwise to the identity operator I). Now, if K is any compact
operator, P
n
K is of finite rank and, from Theorem 4.3 (P
n
K) converges to K in norm.
26 Operators on Hilbert space
Exercises 4
1. Let T be the operator on l
2
⊕ l
2
defined by T(x, y) = (0, x). Show that T
2
= 0 and
that T is not compact.
2. Let (x
n
) be an orthonormal sequence in a Hilbert space H and let (α
n
) be a bounded
sequence of complex numbers. Prove that the operator A defined by
Ax =

¸
n=1
α
n
'x, x
n
`x
n
is bounded with
| A |≤ sup
n

n
[ .
Hence prove that if lim
n→∞

n
) = 0 then A is compact.
Show that, when m = n,
| Ax
m
−Ax
n
|
2
= [α
m
[
2
+[α
n
[
2
.
Hence prove that, conversely if lim
n→∞

n
) = 0 then A is not compact.
3. Given that K

K is compact, prove that K is compact.
[Hint: if (K

Kx
n
) is convergent, prove that (Kx
n
) is a Cauchy sequence.]
4. Let K be a compact operator. Using the hints below, prove that for any orthonormal
sequence ¦x
n
¦, (Kx
n
) →0 as n →∞
Hints: Observe that, for any vector z, 'x
n
, z` →0. [A result of the course states that
¸
['x
n
, z`[
2
is convergent.] Apply this, with z = K

y for any y, and show that no
subsequence of (Kx
n
) can converge to a non-zero vector.
5. Let A
n
be a bounded sequence in B(H) such that, for all x, y ∈ H, lim
n→∞
'A
n
x, y` =
0. Prove that, for any compact operator K,
lim
n→∞
|KA
n
K| = 0 .
[Use the ideas in the proof of Lemma 4.3.]
John Erdos 27
5 The Spectrum.
Definition. The spectrum of an operator T is the set of all complex numbers λ such
that λI −T has no inverse in B(H).
The spectrum of T is denoted by σ(T).
The complement (in C) of σ(T), that is, the set of all complex numbers λ such that
λI −T has an inverse in B(H), is called the resolvent set of T and is denoted by ρ(T).
For any element T of B(H), it is a fact that σ(T) is a non-empty compact subset of
C. We shall not need this general fact in this course. For the two classes of operators
that we shall be concerned with (compact operators and selfadjoint operators) the
required facts about the spectrum will be established by simple methods.
Note that every eigenvalue of an operator T is in the spectrum of T.
Also, if the K is a compact operator on an infinite-dimensional Hilbert space then
0 ∈ σ(K) (this merely repeats the fact that K does not have an inverse).
Lemma 5.1 Let T be an operator such that for all x ∈ H, | Tx |≥ c | x |, where c
is a positive constant. Then the range of T is closed.
Proof. Let (y
n
) be a convergent sequence of elements of ran(T) converging to y.
Then y
n
= Tx
n
for some sequence (x
n
) and we need to show that y = Tx for some x.
Since (y
n
) is convergent it is a Cauchy sequence. Now,
|x
n
−x
m
| ≤
1
c
|T(x
n
−x
m
)| =
1
c
|y
n
−y
m
)|
so it follows easily that (x
n
) is a Cauchy sequence and so convergent to some element
x. Then, since T is continuous, y = limy
n
= limTx
n
= limTx, as required.
Corollary 5.2 If T is an in the lemma, the range of T
n
is closed for each positive
integer n.
Proof. |T
n
x| ≥ c
n
|x| for all x ∈ H.
We now derive some simple properties of the spectrum of a selfadjoint operator. For
the rest of this section, A will denote a selfadjoint operator. Recall that 'Ax, x` is
real for all x since 'Ax, x` = 'x, Ax` = 'A

x, x` = 'Ax, x`.
Lemma 5.3
|A| = sup
x≤1
['Ax, x`[ .
Proof. Let k = sup
z≤1
['Az, z`[. Then ['Ax, x`[ ≤ k|x|
2
for all x and, from the
Cauchy-Schwatrz in equality, k ≤ |A|. Since
|A| = sup
x≤1
|Ax| = sup
x≤1
sup
y≤1
['Ax, y`[ ,
to show that |A| ≤ k, we need to show that ['Ax, y`[ ≤ k whenever |x| ≤ 1 and
|y| ≤ 1. It is sufficient to prove this when 'Ax, y` is real, since if ['Ax, y`[ = e

'Ax, y`
28 Operators on Hilbert space
then applying the result for the real case for 'Ax

, y` where x

= e

x, proves the
general result.
Now, using the polarization identity (Lemma 2.3) and the paralellogram law (Lemma 1.4),
4'Ax, y` = 'A(x + y), (x + y)` −'A(x −y), (x −y)`
+i['A(x + iy), (x + iy)` −'A(x −iy), (x −iy)`]
≤ k¦|x + y|
2
+|x −y|
2
¦
= k(2|x|
2
+ 2|y|
2
) ≤ 4k ,
(the expression in square brackets being zero since 'Ax, y` is real).
Note that sup
x≤1
['Ax, x`[ = sup
x=1
['Ax, x`[. We write
m = inf
x=1
'Ax, x` and M = sup
x=1
'Ax, x` .
Corollary 5.4 For all T ∈ B(H)
|T

T| = |T|
2
.
Proof. Since T

T is selfadjoint,
|T

T| = sup
x≤1
['T

Tx, x`[ = sup
x≤1
|Tx|
2
= |T|
2
.
The key to the next result is proving that | (λI−A)x |≥ c | x | whenever λ ∈ [m, M].
This is done by a single calculation in the body of the proof. However, it can also be
established by a sequence of simpler proofs as follows. Note that, if X is selfadjoint
then
|(iI −X)x|
2
= '(iI −X)x, (iI −X)x`
= |x|
2
+|Xx|
2
−i'x, Xx` + i'Xx, x`
= |Xx|
2
+|x|
2
≥ |x|
2
.
Thus, if λ = ξ + iη is not real (i.e. η = 0), then, using the above result for X =
1
η
(A −ξI), we have
| (λI −A)x |=| η(iI −X)x |≥ [η[|x| .
If λ is real with λ > M, we have that for |x| = 1,
|(λI −A)x| = sup
y≤1
['(λI −A)x, y`[ ≥ '(λI −A)x, x` ≥ λ −M
so that (dividing by |x|) we have |(λI − A)x| ≥ (λ − M)|x| for all x. A similar
proof holds when λ < m.
Theorem 5.5
(i) σ(A) ⊆ [m, M] ,
(ii) m ∈ σ(A) and M ∈ σ(A) .
John Erdos 29
Proof. (i) Suppose λ ∈ [m, M] and let d = dist(λ, [m, M]). Let x ∈ H be any
unit vector and write α = 'Ax, x`. Then '(αI −A)x, x` = 'x, (αI −A)x` = 0 and
|(λI −A)x|
2
= |[λI −αI + (αI −A)]x|
2
= '[λI −αI + (αI −A)]x, [λI −αI + (αI −A)]x`
= [λ −α[
2
|x|
2
+ (¯ α −
¯
λ)'(αI −A)x, x`
+(α −λ)'x, (αI −A)x` +|(αI −A)x|
2
≥ [λ −α[
2
≥ d
2
.
It follows that |(λI − A)x| ≥ d|x| [apply the above for
x
x
]. Hence λI − A is
injective and, by Lemma 5.1, it has closed range. Further, if 0 = z ⊥ ran(λI − A)
then 0 = '(λI − A)x, z` = 'x, (
¯
λI − A)z` for all x and so (
¯
λI − A)z = 0. But this
is impossible, since, from above, noting that d = dist(λ, [m, M]) = dist(
¯
λ, [m, M]),
we have |
¯
λI − A)z| ≥ d|z|. Therefore, ran(λI − A) = H, (being both dense and
closed).
Therefore, for any y ∈ H, there is a unique x ∈ H such that y = (λI − A)x. Define
(λI −A)
−1
y = x. Then |y| ≥ d|x| so
|(λI −A)
−1
y| = |x| ≤
1
d
|y|
showing that (λI −A)
−1
∈ B(H) (i.e. it is continuous). Thus λ ∈ σ(A), proving (i).
(ii) From Lemma 5.3, |A| is either M or −m. If |A| = M = sup
x=1
'Ax, x`;
(if |A| = −m the same proofs, applied to −A, hold) there exists a sequence (x
n
) of
unit vectors such that ('Ax
n
, x
n
`) →M. Then
|(A −MI)x
n
|
2
= |Ax
n
|
2
+ M
2
−2M'Ax
n
, x
n
` ≤ 2M
2
−2M'Ax
n
, x
n
` →0 .
Hence A−MI has no inverse in B(H) [since if X were such an operator, 1 = |x
n
| =
|X(A −MI)x
n
| ≤ |X|.|X(A −MI)x
n
| →0] and so M ∈ σ(A). For m, note that
sup
x=1
'(MI −A)x, x` = M −m = |MI −A|
since inf
x=1
'(MI − A)x, x` = 0. Applying the result just proved to the operator
MI −A shows that M −m ∈ σ(MI −A), that is, (M −m)I −(MI −A) = A−mI
has no inverse. Hence m ∈ σ(A).
The spectral radius, ν(T), of an operator T is defined to be
ν(T) = sup¦[λ[ : λ ∈ σ(T)¦ .
Thus we have shown that the spectrum of a selfadjoint operator is non-empty and
real and its norm is equal to its spectral radius.
30 Operators on Hilbert space
Exercises 5
1. Let X, T ∈ B(H) and suppose that X is invertible. Prove that σ(T) = σ(X
−1
TX).
2. Let A ∈ B(H) be a selfadjoint operator. Show that U = (A − iI)(A + iI)
−1
is a
unitary operator.
John Erdos 31
6 The Spectral analysis of compact operators.
In this section K will always denote a compact operator.
Theorem 6.1 If λ = 0 then either λ is an eigenvalue of K or λ ∈ ρ(K).
Proof. Suppose that λ = 0 is not an eigenvalue of K. We show that λ ∈ ρ(K). The
proof of this is in several stages.
(a) For some c > 0, we have that that |(λI −K)x| ≥ c|x| for all x ∈ H.
Suppose this is false. Then the inequality fails for c =
1
k
for k = 1, 2, . Therefore
there is a sequence of unit vectors such that
|(λI −K)x
k
| ≤
1
k
,
that is, ((λI −K)x
k
) → 0. Applying the condition that K is compact, there is a
subsequence (x
k
i
) such that (Kx
k
i
) is convergent. Call its limit y. Then
x
k
i
=
1
λ
((λI −K)x
k
i
+ Kx
k
i
)
and so (x
k
i
) →
y
λ
. Since (x
k
i
) is a sequence of unit vectors, y = 0. But then,
(λI −K)y = lim
i→∞
(λI −K)x
k
i
= λ
y
λ
−y = 0 .
This contradicts the fact that λ is not an eigenvalue, so (a) is established.
(b) ran(λI −K) = H.
Let H
n
= ran(λI −K)
n
and write H
0
= H. It follows from (a) using Lemma 5.1 that
(H
n
) is a sequence of closed subspaces. Also
(λI −K)H
n
= H
n+1
H
0
⊇ H
1
⊇ H
2
⊇ H
3
⊇ .
Note that, if y ∈ H
n
then Ky = ((K −λI)y + λy) ∈ H
n
so that K(H
n
) ⊆ H
n
.
We now use the compactness of K to show that the inclusion H
n
⊆ H
n+1
is not
always proper. Suppose, on the contrary that
H
0
⊃ H
1
⊃ H
2
⊃ H
3
⊃ .
Using Lemma 1.7, for each n we can find a unit vector x
n
such that x
n
∈ H
n
and
x
n
⊥ H
n+1
. We show that (Kx
n
) cannot have a Cauchy subsequence. Indeed, if
m > n
Kx
n
−Kx
m
= (K −λI)x
n
+ λx
n
−Kx
m
= λx
n
+ [(K −λI)x
n
−Kx
m
]
= λx
n
+ z
where z ∈ H
n+1
[Kx
m
∈ H
m
⊆ H
n+1
follows from m > n]. Thus
|Kx
n
−Kx
m
|
2
= [λ[
2
+|z|
2
≥ [λ[
2
32 Operators on Hilbert space
and so (Kx
n
) has no convergent Cauchy subsequence. Therefore, the inclusion is not
always proper. Let k be the smallest integer such that H
k
= H
k+1
. If k = 0 then
choose x ∈ H
k−1
` H
k
. Then (λI −K)x ∈ H
k
= H
k+1
and so, for some y,
(λI −K)x = (λI −K)
k+1
y = (λI −K)z
where z = (λI −K)
k
y ∈ H
k
. Now x ∈ H
k
so x −z = 0 and
(λI −K)(x −z) = 0
contradicting the fact that λ is not an eigenvalue. Therefore k = 0, that is ran(λI −
K) = H
1
= H
0
= H.
(c) Completing the proof. This is done exactly as in Theorem 5.5 (i). For any
y ∈ H, there is a unique x ∈ H such that y = (λI − K)x. Define (λI − K)
−1
y = x.
Then |y| ≥ c|x| so
|(λI −K)
−1
y| = |x| ≤
1
c
|y|
showing that (λI −K)
−1
∈ B(H) (i.e. it is continuous). Thus λ ∈ σ(K).
Lemma 6.2 If ¦x
n
¦ are eigenvectors of K corresponding to different eigenvalues
¦x
n
¦, then ¦x
n
¦ is a linearly independent set.
Proof. This is exactly as in an elementary linear algebra course. Suppose the state-
ment is false and k is the first integer such that x
1
, x
2
, , x
k
is linearly dependent.
Then
¸
k
i=1
α
i
x
i
= 0 and α
k
= 0. Also, by hypothesis Kx
i
= λ
i
x
i
with the λ
i
’s all
different. Now x
k
=
¸
k−1
i=1
β
i
x
i
(where β
i
= −α
i

k
) and so
0 = (λ
k
I −K)x
k
=
k−1
¸
i=1

k
−λ
i

i
x
i
showing that x
1
, x
2
, , x
k−1
is linearly dependent, contradicting the definition of k.
Theorem 6.3 σ(K)`¦0¦ consists of eigenvalues with finite-dimensional eigenspaces.
The only possible point of accumulation of σ(K) is 0.
Proof. Let λ be any non-zero eigenvalue and let N = ¦x : Kx = λx¦ be the
eigenspace of λ. If N is not finite-dimensional, we can find an orthonormal sequence
(x
n
) of elements of N [apply the Gram-Schmidt process (Theorem 3.4) to any linearly
independent sequence]. Then
|Kx
n
−Kx
m
|
2
= |λx
n
−λx
m
|
2
= 2[λ[
2
which is impossible, since K is compact.
To show that σ(K) has no points of accumulation other than (possibly) 0, we show
that ¦λ ∈ C : [λ[ > δ¦ ∩ σ(K) is finite for any δ > 0. Suppose this is false and there
is a sequence of distinct eigenvalues (λ
i
) with [λ
i
[ > δ for all i. Then we have vectors
x
i
with Kx
i
= λ
i
x
i
.
Let H
n
= span¦x
1
, x
2
, , x
n
¦. Then, since ¦x
n
¦ is a linearly independent set, we
have the proper inclusions
H
1
⊂ H
2
⊂ H
3
⊂ H
4
⊂ .
John Erdos 33
It is easy to see that K(H
n
) ⊆ H
n
and (λ
n
I − K)H
n
⊆ H
n−1
. Choose, as in The-
orem 6.1 a sequence of unit vectors (y
n
) with y
n
∈ H
n
and y
n
⊥ H
n−1
. Then, for
n > m,
Ky
n
−Ky
m
= λ
n
y
n
−[(λ
n
I −K)y
n
−Ky
m
] .
Since (λ
n
I − K)y
n
∈ H
n−1
and Ky
m
∈ H
m
⊆ H
n−1
, the vector in square brackets is
in H
n−1
. Therefore, since y
n
⊥ H
n−1
,
|Ky
n
−Ky
m
| > [λ
n
[ > δ
showing that (Ky
n
) has no convergent subsequence.
Corollary 6.4 The eigenvalues of K are countable and whenever they are put into
a sequence (λ
i
) we have that lim
→∞
λ
i
= 0.
Proof. [The set of all eigenvalues is (possibly) 0 together with the countable union
of the finite sets of eigenvalues >
1
n
, (n = 1, 2, ).
If > 0 is given then, since λ : λ an eigenvalue of K, [λ[ ≥ is finite, we have that

i
[ < for all but a finite number of values of i. Hence (λ
i
) →0. ]
Corollary 6.5 If A is a compact selfadjoint operator then |A| equals its eigenvalue
of largest modulus.
Proof. This is immediate from Theorem 5.5 (ii).
The Fredholm alternative. For any scalar µ, either
(I −µK)
−1
exists
or the equation
(I −µK)x = 0
has a finite number of linearly independent solutions.
(Fredholm formulated this result for the specific operator (Kf)(x) =

b
a
k(x, t)f(t) dt .
In fact, he said : EITHER the integral equation
f(x) −µ

b
a
k(x, t)f(t) dt = g(x)
has a unique solution, OR the associated homogeneous equation
f(x) −µ

b
a
k(x, t)f(t) dt = 0
has a finite number of linearly independent solutions.)
We now turn to compact selfadjoint operators. For the rest of this section A will
denote a compact selfadjoint operator.
Note that every eigenvalue of of A is real. This is immediate from Theorem 5.5, but
can be proved much more simply since if Ax = λx, where x is a unit eigenvector,
¯
λ = 'Ax, x` = 'x, Ax` = 'A

x, x` = 'Ax, x` = λ.
Lemma 6.6 Distinct eigenspaces of A are mutually orthogonal.
34 Operators on Hilbert space
Proof. Let x and y be eigenvectors corresponding to distinct eigenvalues λ and µ.
Then,
λ'x, y` = 'Ax, y` = 'x, A

y` = 'x, Ay` = ¯ µ'x, y` = µ'x, y`
(since µ is real) and so 'x, y` = 0.
Theorem 6.7 If A is a compact selfadjoint operator on a Hilbert space H then H
has an orthonormal basis consisting of eigenvectors of A.
Proof. Let (λ
i
)
i=1,2,···
be the sequence of all the non-zero eigenvalues of A and let N
i
be the eigenspace of λ
i
. Take an orthonormal basis of each N
i
and an orthonormal
basis of N
0
= ker A. Let (x
n
) be the union of all these, put into a sequence. It follows
from Lemma 6.6 that this sequence is orthonormal.
Let M = ¦z : z ⊥ x
n
for all n¦. Then, if y ∈ M we have that 'x
n
, Ay` = 'Ax
n
, y` =
λ
n
'x
n
, y` = 0 and so A(M) ⊆ M. Therefore A with its domain restricted to M
is a compact selfadjoint operator on the Hilbert space M. Clearly this operator is
selfadjoint ['Ax, y` = 'x, Ay` for all x, y ∈ H so certainly for all x, y ∈ M]. Also
it cannot be have a no-zero eigenvector [for then M ∩ N
k
= (0) for some k > 0].
Therefore, by Corollary 6.5, it is zero. But then M ⊆ N
0
. But also M ⊥ N
0
and so
M = (0). Therefore (x
n
) is a basis.
Corollary 6.8 Then there is an orthonormal basis ¦x
n
¦ of H such that, for all h,
Ah =

¸
n=1
λ
n
'h, x
n
`x
n
.
Proof. Let (x
n
) be the basis found in the Theorem and let λ
n
= 'Ax
n
, x
n
` (this is
merely re-labeling the eigenvalues. The from Theorem 3.3 (iii), for any h ∈ H,
h =

¸
n=1
'h, x
n
`x
n
.
Acting on this by A, since A is continuous and Ax
n
= λ
n
x
n
we have that
Ah =

¸
n=1
λ
n
'h, x
n
`x
n
.
Theorem 6.9 If A is a compact selfadjoint operator on a Hilbert space H then there
is an orthonormal basis ¦x
n
¦ of H such that
A =

¸
n=1
λ
n
(x
n
⊗x
n
)
where the series is convergent in norm.
Proof. Let ¦x
n
¦ be the basis found as above so that Ax
n
= λ
n
x
n
and
Ah =

¸
n=1
λ
n
'h, x
n
`x
n
.
John Erdos 35
Note that (λ
n
) →0. Let
A
k
=
k
¸
n=1
λ
n
(x
n
⊗x
n
) .
We need to show that |A −A
k
| →0 as k →∞.
Now
(A −A
k
)h =

¸
n=k+1
λ
n
'h, x
n
`x
n
.
and, using Theorem 3.3 (v)
|(A −A
k
)h|
2
=

¸
n=k+1

n
'h, x
n
`[
2
≤ sup
n≥k+1

n
[
2

¸
n=k+1
['h, x
n
`[
2
≤ sup
n≥k+1

n
[
2

¸
n=1
['h, x
n
`[
2
= sup
n≥k+1

n
[
2
|h|
2
.
Thus |(A−A
k
)| ≤ sup
n≥k+1

n
[, and so since (λ
n
) →0, we have that |A−A
k
| →0
as k →∞.
Alternatively, Theorem 4.4 may be used to prove the above result. Let ¦x
n
¦ and A
k
and A be as above and let
P
k
=
k
¸
n=1
(x
n
⊗x
n
) .
Then, since ¦x
n
¦ is a basis, Theorem 3.3 (iii) shows that (P
k
) converges pointwise to
the identity operator I. Since A
k
= AP
k
, Theorem 4.4 shows that (A
k
) converges to
A in norm.
36 Operators on Hilbert space
Exercises 6
1. Let K be a compact operator on a Hilbert space H and let λ = 0 be an eigenvalue
of K. Show that λI − K has closed range. [Hint : let N = ker(λI − K) and let
M = N

. If y ∈ ran(λI −K), show that y = lim
n→∞
(λI −K)z
n
with z
n
∈ M. Now
imitate the proof for the case when λ is not an eigenvalue.]
2. Find the norm of the compact operator V defined on L
2
[0, 1] by
(V f)(x) =

x
0
f(t) dt
.
Hints: Use Corollary 5.4 and the fact that the norm of the compact selfadjoint opera-
tor V

V is given by its largest eigenvalue. Now use the result of Exercises 2 Question
6 to show that if f satisfies V

V f = λf then it satisfies

λf

+ f = 0
f(1) = 0, f

(0) = 0.
[You may assume that any vector in the range of V

V (being in the range of two
integrations) is twice differentiable (almost everywhere).]
Note that a direct approach to evaluating | V | seems to be very difficult (try it !).
3. Let ¦x
n
¦ be an orthonormal basis of H and suppose that T ∈ B(H) is such that the
series
¸

n=1
|Tx
n
|
2
converges. Prove that
(i) T is compact,
(ii)
¸

n=1
|Ty
n
|
2
converges for every orthonormal basis ¦y
n
¦ of H and for the sum
is the same for every orthonormal basis.
Note : an operator satisfying the above is called a Hilbert-Schmidt operator.
Hints: (i) write h ∈ H as a Fourier series, h =
¸

i=1
α
i
x
i
where α
i
= 'h, x
i
`. Define
T
n
h =
¸
n
i=1
α
i
Tx
i
and show that
|(T −T
n
)h|
2

¸

¸
n+1

i
[.|Tx
i
|
¸

≤ |h|
2
.

¸

¸
n+1
|Tx
i
|
2
¸

.
(ii) Take an orthonormal basis φ
k
of H consisting of eigenvectors of the compact
operator T

T. Observe that if T


k
= µ
k
φ
k
then µ
k
= 'T


k
, φ
k
` = |Tφ
k
|
2
≥ 0.
Now use the spectral theorem for T

T to prove that if for any orthonormal basis
¦x
n
¦,
¸

n=1
|Tx
n
|
2
converges then

¸
n=1
|Tx
n
|
2
=

¸
n=1
'T

Tx
n
, x
n
` =

¸
k=1
µ
k
.
Note that for a double infinite series with all terms positive, the order of summation
may be interchanged.
John Erdos 37
7 The Sturm-Liouville problem.
In this section we shall discuss the differential operator
Ly =
d
dx

p(x)
dy
dx

+ q(x)y
acting on functions y defined on a closed bounded interval [a, b]. We shall assume
that p(x) > 0 and q(x) real for a ≤ x ≤ b.
We make further assumptions that may be summarized, broadly speaking, by saying
that “everything makes sense”. Specifically we need L to act on functions that are
twice differentiable and whose second derivatives are in L
2
[a, b]. We also need to have
that p is differentiable with p

continuous on [a, b].
We shall be concerned with solving the problem
Ly =
d
dx

p(x)
dy
dx

+ q(x)y = f(x) (∗)
subject to boundary conditions
α
1
y(a) + α
2
y

(a) = 0
β
1
y(b) + β
2
y

(b) = 0
¸
(†)
Where α
1
, α
2
, β
1
and β
2
are real and α
1
α
2
= 0, β
1
β
2
= 0.
Note. The following calculation is of interest because it shows that L satisfies a symmetry
condition that would, for a bounded operator, make it self-adjoint. However, it will not be
used in the sequel. If L is restricted to act on the set of functions that satisfy the boundary
conditions, then 'Ly, z` = 'y, Lz`. Indeed,
'−Ly, z` +'y, Lz` =

b
a
¸
d
dt

p(t)
d¯ z(t)
dt

y −
d
dt

p(t)
dy
dt

¯ z

dt +

b
a
(−qy¯ z +yq¯ z) dt
=
¸
p(t)
d¯ z
dt
y(t) −p(t)
dy
dt
¯ z(t)

b
a
+

b
a
p(t)

dy
dt
d¯ z
dt

d¯ z
dt
.
dy
dt

dt
and, as the integrals on the right of each line are 0, this will vanish if
p(b)

¯ z

(b)y(b) − ¯ z(b)y

(b)

= p(a)

¯ z

(a)y(a) − ¯ z(a)y

(a)

.
But ¯ z

(a)y(a) − ¯ z(a)y

(a) is the determinant of the 2 2 system
ξ¯ z(a) +η¯ z

(a) = 0 ,
ξy(a) +ηy

(a) = 0 ,
which has the non-trivial solution (ξ, η) = (α
1
, α
2
) when y and z satisfy the boundary
conditions (†). So ¯ z

(a)y(a)−¯ z(a)y

(a) = 0 and similarly ¯ z

(b)y(b)−¯ z(b)y

(b) = 0. Therefore
'Ly, z` = 'y, Lz`.
We shall be looking for eigenvalues and eigenfunctions of L that satisfy the conditions
(†); that is, for scalars λ and corresponding functions f that satisfy (†) and the equation
Lf = λf. We make the additional assumption that λ = 0 is not an eigenvalue of the system.
This is quite a reasonable assumption, since if it fails then the problem (*), subject to (†)
does not have a unique solution [an arbitrary multiple of the eigenfunction corresponding
to λ = 0 could be added to any solution to obtain another solution].
38 Operators on Hilbert space
Theorem 7.1 (Existence of the Green’s function.) Under the assumptions stated above,
the problem (*), subject to (†) has the solution
y(x) =

b
a
k(x, t) f(t) dt
where k(x, t) is real-valued and continuous on the square [a, b] [a, b].
Proof. From the elementary theory of the initial value problem for linear differential
equations, (also from Questions 1,2 and 3 of Exercises 6) we have that there is a unique
function u such that Lu = 0, u(a) = −α
2
, u

(a) = α
1
. It follows easily that every solution
of Ly = 0, α
1
y(a) + α
2
y

(a) = 0 is a scalar multiple of u. Similarly we have a unique
v such that Lv = 0, v(b) = −β
2
, v

(b) = β
1
. The assumption that 0 is not an eigenvalue
implies that u and v are linearly independent [if u were a multiple of v then it would be an
eigenfunction].
Let
k(x, t) =

l u(x) a ≤ x ≤ t
mv(x) t ≤ x ≤ b
where (for fixed t) l, m are constants to be chosen. [Our motivational work suggests that
we require k(x, t) to be continuous and p(x).

∂x
k(x, t) to have a unit discontinuity at x = t.]
Choose l, m such that
m.v(t) −l.u(t) = 0,
p(t)[m.v

(t) −l.u

(t)] = 1.
Solving for l, m gives
l =
v(t)

m =
u(t)

where ∆ = p(t)(v

u −u

v) = pJ(u, v), where J is the Jacobean and hence non-zero [since u
and v are independent]. Also,
d∆
dt
= u(pv

)

+u

(pv

) −v(pu

)

−v

(pu

) = −quv +vqu = 0 ,
so ∆ is a constant (i.e. also independent of t). [One can see, independently of the theory
of Jacobeans, that ∆ = 0 since otherwise, at some point t
0
ξ.u(t
0
) +η.v(t
0
) = 0,
ξ.u

(t
0
) +η.v

(t
0
) = 0
has a non-trivial solution (ξ, η). Then y = ξ.u+η.v is a solution of Ly = 0, y(t
0
) = y

(t
0
) = 0
and so ξ.u + η.v is identically 0, contradicting the linear independence of u and v.] Hence
we have that
k(x, t) =

u(x).v(t)

a ≤ x ≤ t ,
u(t).v(x)

t ≤ x ≤ b .
John Erdos 39
To complete the proof, we just verify directly that
y(x) =

b
a
k(x, t) f(t) dt
is the required solution. First note that when x = a we have x ≤ t throughout the range of
integration and so
y(a) =
1

b
a
u(a).v(t) f(t) dt
and, since y

(x) =

b
a
∂k(x, t)
∂x
f(t) dt,
y

(a) =
1

b
a
u

(a).v(t) f(t) dt .
Therefore α
1
y(a) +α
2
y

(a) = 0 since u satisfies the boundary condition at x = a. Similarly
β
1
y(b) +β
2
y

(b) = 0.
We now substitute into the equation. For notational convenience we substitute ∆y(x)
(remembering that ∆ is constant). Since u(x), v(x) can be taken outside the integration,
we obtain
∆y(x) = ∆

b
a
k(x, t) f(t) dt = v(x)

x
a
u(t) f(t) dt +u(x)

b
x
v(t) f(t) dt
(∆y(x))

= v(x)u(x)f(x) +v

(x)

x
a
u(t) f(t) dt
−u(x)v(x)f(x) +u

(x)

b
x
v(t) f(t) dt
(p(x)(∆y(x))

)

= (pv

)

x
a
u(t) f(t) dt +pv

uf + (pu

)

b
x
v(t) f(t) dt −pu

vf .
Therefore
(p(x)(∆y(x))

)

+q∆y = [(pv

)

+qv]

x
a
u(t) f(t) dt
+[(pu

)

+qu]

b
x
v(t) f(t) dt +pf(v

u −u

v)
= f.∆
since u and v are solutions of Ly = 0.
40 Operators on Hilbert space
We can now apply the results of Section 6 to draw conclusions about the eigenfunctions
and eigenvalues of the Sturm-Liouville system (*), (†).
Define the operator K by
(Kf)(x) =

b
a
k(x, t) f(t) dt .
Since k is continuous on [a, b] [a, b] it is clear that

[k[
2
< ∞and so, as shown in Section
4, K is compact.
Let T be the set of functions y such that Ly exists as a function in L
2
[a, b] and satisfies
the boundary conditions (†). (There is a little technical hand waving here. A more precise
statement is: y ∈ T (which is an equivalence class of functions) if there is a representative
y such that y is differentiable and p.(y)

has a derivative almost everywhere such that
(p.(y)

)

∈ L
2
[a, b]. Informally T is the all y ∈ L
2
[a, b] that qualify as solutions of (*),(†) for
some right hand side.) If y ∈ T and f = Ly then Theorem 7.1 shows that Kf = K(Ly) = y,
that is KL acts like the identity on T.
In the other order, it follows from the proof of Theorem 7.1, that for every f ∈ L
2
[a, b]
Kf ∈ T. (The verification that Kf is a solution of Ly = f explicitly shows this. Naturally,
for the most general f, the differentiation of expressions like

x
a
u(t) f(t) dt one requires the
relevant background from Lebesgue integration.) Also, from Theorem 7.1, LKf = f. Thus
LK = I.
Note that L fails to be an inverse of K since it is not defined on the whole of L
2
[a, b], the
Hilbert space in question. Indeed, since K is compact, it cannot be invertible. However, L
is defined on a dense subset.
We use these notations and observations in the statements a proofs below.
Theorem 7.2 (i) The operator K does not have λ = 0 as an eigenvalue.
(ii) λ is an eigenvalue of K if an only if µ =
1
λ
is an eigenvalue of the system (*),(†).
Consequently, the system (*),(†)
1. has a countable sequence (µ
i
) of real eigenvalues such that ([µ
i
[) →∞ ;
2. has eigenfunctions which form an orthonormal basis of L
2
[a, b];
3. has finite-dimensional eigenspaces.
Proof. (i) If f = 0 the solution of Ly = f is Kf and cannot be y = 0. Therefore 0 is not
an eigenvalue of K.
John Erdos 41
(ii) If λ is an eigenvalue of K the Kφ = λφ and since φ is in the range of K, from the
discussion above, φ ∈ T. Then λLφ = LKφ = φ so that
Lφ =
1
λ
φ = µφ,
and µ is an eigenvalue of (*),(†).
Conversely, if µ is an eigenvalue of (*),(†), by assumption µ = 0. We then have Lφ = µφ
and φ ∈ T. Then
KLφ = φ = µKφ
and so Kφ = λφ where λ =
1
µ
is an eigenvalue of K.
The consequences are immediate deductions from the results of Section 6 (principally 6.3,
6.4 and 6.7). Note that in this case the set of eigenvalues of K cannot be finite because this
would imply (by Corollary 6.8) that K vanishes on a non-zero (in fact, infinite-dimensional)
subspace.
The most important result arising from this is consequence 2, since, for example, this is
what justifies the expansions that are required in solving partial differential equations by
the method of separation of variables.
Note. The assumption that 0 is not an eigenvalue of the system is not an essential restric-
tion. For any constant c, the eigenfunctions of L and L+c are the same and the eigenvalues
of L + c are λ + c whenever λ is an eigenvalue L. It is a fact that, by adding a suitable
constant to q we can always ensure that λ = 0 is not an eigenvalue of the system. For
example, if the boundary conditions are y(a) = y(b) = 0, choose c so that c + q does not
change sign in [a, b]; for definiteness, assume c + q(t) < 0 for a ≤ t ≤ b. Let u be the
(unique) solution of
L
c
y =
d
dt

p(t)
dy
dt

+ [c +q(t)]y = 0, y(a) = 0, y

(a) = 1 .
Any solution of L
c
y = 0, y(a) = 0 is a multiple of u so to show that 0 is not an eigenvalue
we must show that u(b) = 0.
Since u

(a) > 0 and u(a) = 0, it follows that u is strictly positive in some interval (a, a+δ).
Suppose z is the smallest zero of u that is > a (if any). Then u

must vanish between a and
z. If z ≤ b then (pu

)

= −[c +q]u is positive in (a, z) and so pu

is increasing in (a, z). But
pu

is strictly positive at a so it is strictly positive in (a, z) contradicting that u

vanishes
between a and z. Hence 0 is not an eigenvalue of L
c
y = 0, y ∈ T.
Similar, but more complicated arguments can be used for other boundary conditions (see
Dieudonne, “Foundations of modern analysis”, Chapter XI, Section 7).
The trigonometric functions form an orthonormal basis of L
2
[−π, π]. This fact
can be deduced from the work of the present section. [The trigonometric functions are
actually the eigenfunctions of y

= 0 subject to periodic boundary conditions y(−π) =
y(π), y

(−π) = y

(π), and such systems are not covered by our work; however the device
below gives us the result.]
The eigenfunctions of the system y

= 0, y(0) = y(π) = 0 are the functions ¦sin nt :
n = 1, 2, 3, ¦. Therefore, from Theorem 7.2 these form an orthonormal basis of L
2
[0, π].
Similarly the functions ¦cos nt : n = 0, 1, 2, 3, ¦ are the eigenfunctions of the system
y
,
= 0, y

(0) = y

(π) = 0 and so also form an orthonormal basis of L
2
[0, π]. (It is true that
42 Operators on Hilbert space
0 is an eigenvalue of the latter system, but this is covered by the note above. Alternatively,
on can consider the system y
,
+ ky = 0, y

(0) = y

(π) = 0 for a suitable constant k – any
non-integral k will do).
Now suppose that f ∈ L
2
[−π, π] is orthogonal to all the trigonometric functions. Then a
simple change of variable shows that for each integer n,
0 =

π
−π
f(t) sin nt dt =

0
−π
f(t) sin nt dt +

π
0
f(t) sin nt dt =

π
0
[f(t) −f(−t)] sin nt dt .
Therefore f(t) = f(−t) (almost everywhere) for 0 ≤ t ≤ π, showing that f is an even
function on [−π, π]. A similar calculation with cosines shows that f is also an odd function
on [−π, π]. Thus f = 0 (almost everywhere) and the fact is proved.
John Erdos 43
Exercises 7
1. Let X be a Banach space (that is, a normed linear space that is complete). A series
¸
n
x
n
in X is said to be absolutely convergent if the series
¸
n
| x
n
| of real numbers is
converegent. Prove that an absolutely convergent series in a Banach space is convergent.
[Hint: prove that the sequence of partial sums is Cauchy.]
Existence theory for linear initial value problems using operator theory.
2. Let k(x, t) be bounded and square integrable over the square [a, b] [a, b], (in particular
this will hold if k is continuous on [a, b] [a, b]). Define K : L
2
[a, b] →L
2
[a, b] by
(Kf)(x) =

x
a
k(x, t)f(t) dt.
Prove that | K |≤ (b −a)M where M is a bound for k in [a, b] [a, b].
Let k
n
be defined inductively by k
1
= k and k
n
(x, t) =

x
t
k(x, s)k
n−1
(s, t) ds. Prove that
(K
n
f)(x) =

x
a
k
n
(x, t)f(t) dt.
Show by induction that
[k
n
(x, t)[ ≤ M
n
[x −t[
n−1
(n −1)!
.
Using this result and the formal binomial expansion of (I −K)
−1
, deduce from Question 1
that (I −K) has an inverse in B(H). [Hint : after proving absolute convergence, verify by
multiplication that the sum of the formal expansion is the inverse of (I − K).] For each
λ = 0, observe that
K
λ
is an operator of the same type as K and deduce that (λ −K) has
an inverse in B(H).
3. Consider the initial value problem
(∗)

y
(n)
(x) +p
1
(x)y
(n−1)
(x) +. . . +p
n
(x)y(x) = f(x)
y(0) = y

(0) = . . . = y
(n−1)
(0) = 0
where p
i
and f are continuous. By putting u(x) = y
(n)
(x), show that, for any b > 0, this
problem reduces to
(†) (I −K)u = f
where K is an operator on L
2
[0, b] of the type considered in Question 6.
[Hint : show that y
(n−r)
(x) =

x
0
(x−t)
r−1
(r−1)!
u(t) dt.]
Prove that (†) has a unique solution in L
2
[0, b] for each b > 0. By quoting appropriate
theorems show that this solution is continuous (strictly, that the equivalence class of this
solution contains a continuous function). Deduce that there is a unique continuous function
with n continuous derivatives that satisfies (∗) in [0, ∞).
Note that the general initial value problem
(∗∗)

y
(n)
(x) +p
1
(x)y
(n−1)
(x) +. . . +p
n
(x)y(x) = f(x)
y(0) = a
0
, y

(0) = a
1
, . . . = y
(n−1)
(0) = a
n−1
44 Operators on Hilbert space
can be transformed into the form (*) by changing the dependent variable from y to z where
z(x) = y(x) −
n−1
¸
k=0
a
k
x
k
and thus (∗∗) also has a unique solution.
4. Find a Green’s function for the system
y

= f, y(0) = y(1) = 0 .
Check your answer by verifying that it gives x(x −1) as the solution when f = 2.
Evaluate the eigenvalues and eigenfunctions of
y

= λy, y(0) = y(1) = 0 ,
and consequently find an orthonormal basis of L
2
[0, 1].
5. Repeat the above question with the system
y

= f, y(0) +y

(0) = y(1) −y

(1) = 0 .
John Erdos 45
8 The Spectral analysis of selfadjoint operators.
In this section A will always denote a selfadjoint operator with spectrum σ = σ(A) where
σ(A) ⊆ [m, M] as defined in Section 6.
Theorem 8.1 (Spectral Mapping Theorem) Let T ∈ B(H). For any polynomial p,
σ(p(T)) = ¦p(λ) : λ ∈ σ(T)¦ .
Proof. Let p be any polynomial. By the elementary scalar remainder theorem, x −λ is a
factor of p(x) −p(λ), that is, p(x) −p(λ) = (x −λ)q(x) for some polynomial q. Therefore,
p(T) −p(λ)I = (T −λI)q(T) .
Now if p(λ) ∈ σ(p(T)) then p(T) −p(λ)I has an inverse X and so
I = X.[p(T) −p(λ)I] = X.q(T).(T −λI) = [p(T) −p(λ)I].X = (T −λI).q(T).X .
Therefore (T − λI) has both a left inverse (namely X.q(T)) and a right inverse (q(T).X).
An easy algebraic argument shows that these are equal and (T − λI) has an inverse, that
is, λ ∈ σ(T). That is, if λ ∈ σ(T) then p(λ) ∈ σ(p(T)).
Conversely, if k ∈ σ(p(T)) then the polynomial p(x) −k factors over the complex field into
linear factors: p(x) −k = (x −λ
1
)(x −λ
2
)(x −λ
3
) (x −λ
n
) . Where λ
1
, λ
2
, λ
3
λ
n
are
the roots of p(x) = k. Then
p(T) −kI = (T −λ
1
I)(T −λ
2
I)(T −λ
3
I) (T −λ
n
I) ,
and these factors clearly commute. If each T − λ
i
I has an inverse then the product of all
these would be an inverse of p(T) − kI. Therefore T − λ
i
I fails to have an inverse for at
least one root λ
i
of p(x) = k. That is k = p(λ
i
) for some λ
i
∈ σ(T).
Corollary 8.2 If A is a selfadjoint operator then, for any polynomial p,
|p(A)| = sup¦[p(λ)[ : λ ∈ σ(A)¦
Proof. If p is real, p(A) is selfadjoint. We know from Theorem 5.5 that for selfadjoint
operators, the norm equals the spectral radius. Therefore, using the result of the theorem,
|p(A)| = sup¦[k[ : k ∈ σ(p(A))¦ = sup¦[p(λ)[ : λ ∈ σ(A)¦ .
For the general case, since (¯ p.p)(A) = ¯ p(A).p(A) = p(A)

.p(A) we have, using Corollary 5.4
|p(A)|
2
= |p(A)

.p(A)| = sup¦[ ¯ pp(λ)[ : λ ∈ σ(A)¦
= sup¦[p(λ)[
2
: λ ∈ σ(A)¦
= (sup¦[p(λ)[ : λ ∈ σ(A)¦)
2
.
Definition. The functional calculus. Let f ∈ C[m, M]. From Weierstrass’ approxima-
tion theorem there is a sequence (p
n
) of polynomials converging to f uniformly on [m, M].
[If f = g + ih is complex, (p
n
) is found by combining sequences approximating g and h.]
The operator f(A) is defined by
f(A) = lim
n→∞
p
n
(A) .
46 Operators on Hilbert space
Theorem 8.3 The operator f(A) ∈ B(H) and is well defined. The map f → f(A) is a
*-algebra homomorphism of C[m, M] into B(H) and
|f(A)| = sup¦[f(λ)[ : λ ∈ σ(A)¦ .
Proof. We first show that the above definition determines an operator f(A). Let p
n
be
a sequence of polynomials converging to f uniformly on [m, M]; that is, in the norm of
C[m, M]. Then Corollary 8.2 shows that
|p
n
(A) −p
m
(A)| = sup¦[p
n
(λ) −p
m
(λ)[ : λ ∈ σ(A)¦
≤ sup¦[p
n
(λ) −p
m
(λ)[ : λ ∈ [m, M]¦ = |p
n
−p
m
| .
Since (p
n
) is a Cauchy sequence in C[m, M], it is clear that (p
n
(A)) is a Cauchy sequence
in B(H) and so is convergent. To show that this defines a unique operator we must show
that it is independent of the choice of the sequence. Suppose that (p
n
) and (q
n
) are two
sequence of polynomials converging to f uniformly on [m, M]. Write limp
n
(A) = X and
limq
n
(A) = Y . Then
|X −Y | = |X −p
n
(A) +p
n
(A) −q
n
(A) +q
n
(A) −Y |
≤ |X −p
n
(A)| +|p
n
(A) −q
n
(A)| +|q
n
(A) −Y |
≤ |X −p
n
(A)| +|q
n
(A) −Y | + sup
m≤t≤M
[p
n
(t) −q
n
(t)[
and, since |X−Y | is independent of n and the right hand side →0 as n →∞, this implies
that X = Y = f(A) and that the limit depends only on the function f.
To demonstrate the *-algebra homomorphism statement, we need to show that,
(αf +βg)(A) = αf(A) +βg(A)
(f.g)(A) = f(A).g(A)
¯
f(A) = f(A)

.
These follow easily from the definitions since, if (p
n
) and (q
n
) are sequences of polynomials
converging uniformly on [m, M] to f and g respectively, then
(αf +βg)(A) = lim
n→∞
(αp
n
+βq
n
)(A) = α lim
n→∞
p
n
(A) +β lim
n→∞
q
n
(A) = αf(A) +βg(A)
and the other statements are easily proved in the same way.
Finally,
|f(A)| = | lim
n→∞
p
n
(A)| = lim
n→∞
|p
n
(A)| = lim
n→∞
sup
λ∈σ
[p
n
(λ)[ = sup
λ∈σ
[f(λ)[ .
The extension of the functional calculus. We wish to extend the functional calculus
to limits of pointwise monotonically convergent sequences of continuous functions. (The
immediate goal is to attach a meaning to χ
[−∞,t]
(A). Note that χ
[−∞,t]
is a real function
which is equal to its square. Our hope is that E
t
= χ
[−∞,t]
(A) will satisfy E
t
= E
2
t
= E

t
;
that is, that E
t
is a projection.)
We first need a definition some technical results.
Definition. An operator A is said to be positive (in symbols A ≥ 0) if 'Ax, x` ≥ 0 for all
x ∈ H. We write A ≥ B to mean A−B ≥ 0.
Note that, from the polarization identity, Theorem 2.3 and one of the exercises, any positive
operator is selfadjoint. Also, it follows from Theorem 5.5 that A ≥ 0 if and only if σ(A) ⊆
R
+
.
John Erdos 47
Lemma 8.4 Let A be a positive operator. Then
(i) ['Ax, y`[ ≤ 'Ax, x`.'Ay, y`
(ii) |Ax| ≤ |A|'Ax, x`.
Proof. (i) This is proved in exactly the same way as Theorem 1.1 from the fact that
'A(λx +y), (λx +y)` ≥ 0 for all λ.
(ii) The result is obvious if Ax = 0. If Ax = 0, using (i) with y = Ax we have
|Ax|
4
= ['Ax, Ax`[
2
≤ 'Ax, x`.'A
2
x, Ax`
≤ 'Ax, x`.|A
2
x|.|Ax|
≤ 'Ax, x`.|A|.|Ax|
2
and the result follows on dividing by |Ax|
2
.
Theorem 8.5 Let (A
n
) be a decreasing sequence of positive operators. Then (A
n
) converges
pointwise (strongly) to an operator A such that 0 ≤ A ≤ A
n
for all n.
Proof. Note that for each n, using Lemma normsa, |A
n
| = sup
x=1
'A
n
x, x` ≤ sup
x=1
'A
1
x, x` =
|A
1
|. The hypothesis shows that, for each x ∈ H, the sequence ('A
n
x, x`) is a decreasing
sequence of positive real numbers and hence is convergent. If m > n then A
m
− A
n
≥ 0
and from Lemma 8.4 (ii),
|(A
m
−A
n
)x|
2
≤ |A
m
−A
n
|.'(A
m
−A
n
)x, x`
≤ 2.|A
1
|.['A
m
x, x` −'A
n
x, x`]
and this shows that (A
n
) is a Cauchy sequence, and so convergent. Call the limit Ax. It is
routine to show that A is linear, bounded, selfadjoint and 0 ≤ A ≤ A
n
for all n. [E.g. to
show that A is selfadjoint, we write
'Ax, y` = lim
n→∞
'A
n
x, y` = lim
n→∞
'x, A
n
y` = 'x, Ay` .
Lemma 8.6 If the sequences (A
n
), (B
n
) converge pointwise (strongly) to an operators A, B
respectively (and if |A
n
| is bounded) then (A
n
.B
n
) →AB pointwise.
[Note that pointwise convergence is convergence in a topology on B(H) called the “strong
operator topology”; hence the alternative terminology. Note also that by a theorem of
Banach space theory (the Uniform Boundedness Theorem) the condition in brackets is
implied by the convergence of (A
n
).]
Proof. For any x ∈ H we have
|(A
n
B
n
−AB)x| = |(A
n
B
n
−A
n
B+A
n
B−AB)x| ≤ |A
n
|.|(B
n
−B)x| +|(A
n
−A)Bx|
and the right hand side →0 from the pointwise convergence of (A
n
) and (B
n
) [at the points
Bx and x respectively].
Definition. The extended functional calculus. Let φ be a positive function on [m, M]
that is the pointwise limit of a decreasing sequence (f
n
) of positive functions ∈ C[m, M].
The operator φ(A) is defined as the pointwise limit of the sequence (f
n
(A)).
It is a fact that φ(A) is well defined (that is, it depends only on φ and not on the approxi-
mating sequence).
48 Operators on Hilbert space
Lemma 8.7 Let φ, ψ be a positive functions on [m, M] that are the pointwise limits of a
decreasing sequences of positive functions ∈ C[m, M]. Then
(i) φ(A) +ψ(A) = (φ +ψ)(A) .
(ii) φ(A).ψ(A) = (φ.ψ)(A) .
(iii) If X commutes with A then X commutes with φ(A) .
Proof. (i) Let (f
n
) and (g
n
) be decreasing sequences of functions that are continuous on
[m, M] and converge pointwise [m, M] to φ and ψ respectively. Then (f
n
+g
n
) is a decreasing
sequence of continuous functions converging pointwise to φ +ψ and
(φ +ψ)(A) = lim
n→∞
(f
n
+g
n
)(A) = lim
n→∞
f
n
(A) + lim
n→∞
g
n
(A) = φ(A) +ψ(A)
where the limits indicate pointwise convergence in H.
(ii) As in (i) (f
n
.g
n
) is a decreasing sequence of continuous functions converging pointwise
to φ.ψ and
(φ.ψ)(A) = lim
n→∞
(f
n
.g
n
)(A) = lim
n→∞
f
n
(A). lim
n→∞
g
n
(A) = φ(A).ψ(A),
using Lemma 8.6, where the limits indicate pointwise convergence in H.
(iii) If X commutes with A then X commutes with f(A) for every f ∈ C[m, M] since if
(p
n
) is a sequence of polynomials converging uniformly to f on [m, M],
X.f(A) = lim
n→∞
X.p
n
(A) = lim
n→∞
p
n
(A).X = f(A).X .
Now if (f
n
) is a decreasing sequences of functions in C[m, M] that converge pointwise [m, M]
to φ then, using Lemma 8.6,
X.φ(A) = lim
n→∞
X.f
n
(A) = lim
n→∞
f
n
(A).X = φ(A).X .
For every real λ we define the operator E
λ
as follows: let
f
n,λ
=

1 t < λ
1 −n(t −λ) λ ≤ t ≤ λ +
1
n
0 t > λ +
1
n
Then (f
n,λ
) is is a decreasing sequence of continuous functions converging pointwise to
χ
[−∞,λ]
. Now let E
λ
= χ
[−∞,λ]
(A). If λ < m then for all t ∈ [m, M] we have f
n,λ
(t) = 0 for
all sufficiently large n and so E
λ
= 0. Similarly, E
λ
= I for λ ≥ M.
Note that
E
λ
.E
µ
= E
ν
where ν = min[λ, µ] .
It follows that E
λ
= E
2
λ
. Also E
λ
≥ 0 and so E
λ
= E

λ
.
A family of projections with these properties is called a bounded resolution of the
identity.
We call the family ¦E
λ
: −∞ < λ < ∞¦ as obtained above the bounded resolution of the
identity for the operator A. It is fact that it is (essentially) uniquely determined by A.
We say that the integral

M
m
f(λ) dE
λ
of a function with respect to a bounded resolution of the identity exists and is equal to T
if, given any > 0 there exists a partition
λ
0
< λ
1
< λ
2
< λ
n−1
< λ
n
John Erdos 49
with λ
0
< m and λ
n
> M such that, for any ξ ∈ (λ
i−1
, λ
i
],

T −
n
¸
i=1
f(ξ
i
)(E
λ
i
−E
λ
i−1
)

< .
For the proof of the Spectral Theorem, we need the following lemmas.
Lemma 8.8 Let ¦∆
i
: 1 ≤ i ≤ n¦ be orthogonal projections such that I =
¸
n
i=0

i
and

i
.∆
j
= 0 when i = j. If X ∈ B(H) commutes with each ∆
i
then
|X| = max
1≤i≤n
|∆
i
X∆
i
| .
Proof. Let h ∈ H. Then
|h|
2
= |
n
¸
i=0

i
h|
2
=

n
¸
i=0

i
h,
n
¸
j=0

j
h
¸
=
n
¸
i=0
|∆
i
h|
2
since the cross terms are zero. Therefore, since X∆
i
= X∆
2
i
= ∆
i
X∆
i
= ∆
i
X
|Xh|
2
=
n
¸
i=0
|∆
i
Xh|
2
=
n
¸
i=0
|∆
i
X∆
i
.∆
i
h|
2

n
¸
i=0
|∆
i
X∆
i
|
2
.|∆
i
h|
2
≤ max
1≤i≤n
|∆
i
X∆
i
|
2
.
n
¸
i=0
|∆
i
h|
2
= max
1≤i≤n
|∆
i
X∆
i
|
2
. |h|
2
.
Thus |X| ≤ max
1≤i≤n
|∆
i
X∆
i
| . But the opposite inequality is clear, since for each i we
have |∆
i
X∆
i
| ≤ |∆
i
|.|X|.|∆
i
| = |X|.
Corollary 8.9

M
m
f(λ) dE
λ

≤ sup
m≤t≤M
[f(t)[ .
Proof. From the lemma,

n
¸
i=1
f(ξ
i
)(E
λ
i
−E
λ
i−1
)

= max
1≤i≤n
|f(ξ
i
)(E
λ
i
−E
λ
i−1
)| = max
1≤i≤n
[f(ξ
i
)[ ≤ sup
m≤t≤M
[f(t)[ .
Since the integral is approximated arbitrarily closely in norm by these sums, the result
follows.
Lemma 8.10 For λ > µ,
µ(E
λ
−E
µ
) ≤ (E
λ
−E
µ
)A(E
λ
−E
µ
) ≤ λ(E
λ
−E
µ
) .
50 Operators on Hilbert space
Proof. First note the general fact that if S ≤ T then for any operator X we have X

SX ≤
X

TX. This is clear since for any h ∈ H,
'X

SXh, h` = 'SXh, Xh` ≤ 'TXh, Xh` = 'X

TXh, h` .
Next we claim that AE
λ
≤ λE
λ
. Let f
λ,n
be as in the definition of E
λ
and define g
n
(t) =
(λ − t)f
λ,n
+
1
n
. It is easy to see that g
n
≥ 0. [This is obvious when t ≤ λ, and also
when t ≥ λ +
1
n
, for then f
λ,n
(t) = 0; for t ∈ (λ, λ +
1
n
) we have λ − t > −
1
n
so, since
0 ≤ f
λ,n
(t) ≤ 1, it follows that g
n
(t) ≥ 0.] Also (g
n
) is decreasing (this is an elementary but
tedious exercise) and so the pointwise (strong) limit of (g
n
(A)) exists and is positive. But
(g
n
(t)) is pointwise convergent to λχ(−∞, λ](t) −tχ(−∞, λ](t). Therefore
λE
λ
−AE
λ
≥ 0
proving the claim.
Note that E
λ
commutes with A and AE
2
λ
= AE
λ
. Therefore we have that
E
λ
AE
λ
= AE
2
λ
≤ λE
λ
.
We now use that (I −E
µ
) = (I −E
µ
)

and the general fact from the start of the proof to
conclude that
(E
λ
−E
µ
)A(E
λ
−E
µ
) = (I −E
µ
)E
λ
AE
λ
(I −E
µ
) ≤ (I −E
µ
)λE
λ
(I −E
µ
) = λ(E
λ
−E
µ
) .
The fact that µ(E
λ
−E
µ
) ≤ (E
λ
−E
µ
)A(E
λ
−E
µ
) is proved in an exactly similar way.
Theorem 8.11 (The Spectral Theorem for bounded selfadjoint operators.) For any bounded
selfadjoint operator A there exists a bounded resolution of the identity E
λ
such that
A =

M
m
λdE
λ
.
Proof. Let E
λ
be the resolution of the identity as found above. Let > 0 be given.
Choose λ
0
< λ
1
< λ
2
< λ
n
with λ
0
< m, λ
n
> M such that 0 ≤ λ
i
− λ
i−1
< . Write

i
= E
λ
i
−E
λ
i−1
. From Lemma 8.10 we have that
λ
i−1

i
≤ ∆
i
A∆
i
≤ λ
i

i
.
Hence, for any ξ
i
∈ [λ
i−1
, λ
i
],

i−1
−ξ
i
)∆
i
≤ ∆
i
A∆
i
−ξ
i

i
≤ (λ
i
−ξ
i
)∆
i
.
Note that when |h| = 1, since ∆
i
is an orthogonal projection, 0 ≤ '∆
i
h, h` = |∆
i
h|
2
≤ 1 .
Therefore, using Theorem 5.3
|∆
i
A∆
i
−ξ
i

i
| ≤ max[[λ
i
−ξ
i
[, [λ
i−1
−ξ
i
[] < .
Observe that ¦∆
i
: 1 ≤ i ≤ n¦ satisfies the hypotheses of Lemma 8.8 and that X =
(A−
¸
n
i=1
ξ
i

i
) commutes with each ∆
i
. Therefore, applying Lemma 8.8 we have that
|(A−
n
¸
i=1
ξ
i

i
)| = max
1≤i≤n
|∆
i
A∆
i
−ξ
i

i
| <
and this is exactly what is required.

i
.∆
j
= 0 when i = j.
John Erdos 51
Corollary 8.12 If f is continuous on [m, M] then
f(A) =

M
m
f(λ) dE
λ
.
Proof. For any integer k, choose n, ξ
1
, ξ
2
ξ
n
, λ
0
, λ
1
, . . . , λ
n
as in the Theorem so that

(A−
n
¸
i=1
ξ
i

i
)

<
1
k
and write 1
k
=
¸
n
i=1
ξ
i

i
. Then lim
k→∞
1
k
= A. Therefore, for any integer r,
A
r
= lim
k→∞
1
r
k
.
But
1
r
k
=

n
¸
i=1
ξ
i

i

r
=
n
¸
i=1
ξ
r
i

i
,
and the right hand side is the approximating sum to

M
m
λ
r
dE
λ
. Therefore
A
r
=

M
m
λ
r
dE
λ
.
and, by taking linear combinations,
p(A) =

M
m
p(λ) dE
λ
for all polynomials p.
For f ∈ C[m, M], given any > 0, choose a polynomial p

such that
sup
m≤λ≤M
[f(λ) −p(λ)[ < .
Then, from Theorem 8.3 |f(A) −p(A)| < and

f(A) −

M
m
f(λ) dE
λ

f(A) −

M
m
p(λ) dE
λ

+

M
m
p(λ) dE
λ

M
m
f(λ) dE
λ

= |f(A) −p(A)| +

M
m
p(λ) −f(λ) dE
λ

< 2 ,
the last step using Corollary 8.9. Since is arbitrary, it follows that
f(A) =

M
m
f(λ) dE
λ
.
52 Operators on Hilbert space
Exercises 8
1. (More spectral mapping results.) If X ∈ B(H) show that
(i) σ(X

) = ¦λ :
¯
λ ∈ σ(X)¦,
(ii) if X is invertible then σ(X
−1
) = ¦λ
−1
: λ ∈ σ(X)¦.
Deduce that every member of the spectrum of a unitary operator has modulus 1.
2. For any selfadjoint operator A, prove that ker A = (ranA)

.
3. Let A and B be selfadjoint operators. Show that if there exists an invertible operator T
such that T
−1
AT = B then there exists a unitary operator U such that U

AU = B (that
is, if A and B are similar then they are unitarily equivalent).
[Hint: use the polar decomposition of T.]
4. Suppose X ∈ B(H) is selfadjoint and | X |≤ 1. Observe that X + i

I −X
2
can be
defined and prove that it is unitary. Deduce that any operator T can be written as a linear
combination of at most 4 unitary operators. [First write T = X+iY with X, Y selfadjoint.]
5. (i) Use results on the spectrum show that A ≥ kI (where k ∈ R) if and only if for all
λ ∈ σ(A), λ ≥ k. [Note that A is selfadjoint, since 'Ax, x` is real – make sure you know how
this follows!] Deduce that if A ≥ I then A
n
≥ I for every positive integer n. [Alternatively
factorise A
n
−I.]
(ii) If B and C commute and B ≥ C ≥ 0 then prove that B
n
≥ C
n
for every positive
integer n. [Factorise B
n
−C
n
.]
6. Let U be both selfadjoint and unitary. Prove that σ(U) = ¦−1, 1¦ (unless U = ±I).
[Use Question 1.] Use the spectral theorem to find an orthogonal projection E such that
U = 2E−I; (alternatively, if you are given the result it is trivial to verify that E =
1
2
(U +I)
is a suitable E). Note that a self adjoint isometry must be unitary [use Question 2]. Deduce
that the only positive isometry is I.
[Definition: V is an isometry if |V h| = |h| for all h ∈ H.]

2

Operators on Hilbert space

Section 2. This section is supplemented by specific examples of operators on 2 and L2 [a, b]. These include diagonal operators, shifts (forward, backward and weighted) on 2 , the bilateral shift on 2 (Z) and the following operators on L2 [a, b]. Multiplication operator : (Mφ f )(x) = φ(x).f (x) Fredholm integral operator : (Kf )(x) = with the Volterra operator, (V f )(x) =
x a b a

φ (essentially) bounded. |k|2 < ∞,

k(x, t)f (t) dt where

f (t) dt as a special case.

The boundedness of these operators should be established and the adjoints identified. Other examples of finding adjoints (similar to those in the exercises) might be done. Section 3. The main additional topic for this section is the connection to classical Fourier series. The fact that the normalized trigonometric functions for an orthonormal basis of L2 [−π, π] should be established or accepted. One route uses the Stone-Weierstrass Theorem and the density of C[−π, π] in L2 [−π, π]. Inevitably, this requires background in metric spaces and Lebesgue theory. Note that this fact is also established, albeit in a roundabout way, by the work in Section 7. The projection onto a subspace can be written down in terms of an orthonormal basis of the subspace : PN h = h, yi yi where {yi } is an orthonormal basis of N . Applying the Gram-Schmidt process to the polynomials in the Hilbert space L2 [−1, 1] gives (apart from constant factors) the Legendre polynomials. Similarly the Hermite 2 and the Laguerre polynomials arise from orthonormalizing {xn e−x } and {xn e−x } in the spaces L2 [−∞, ∞] and L2 [0, ∞] respectively. Section 4. Some of the operators introduced in Section 2 should be examined for compactness. In particular, the conditions for a diagonal operator on 2 to be compact should be established. Theorem 4.4 and Lemma 4.3 on which its proof depends are the only results in these notes which are not strictly needed for what comes later. Note that this result is not valid in general Banach spaces. Section 5. The spectra of some specific operators should be identified. In particular, the spectrum of Mφ where φ(x) = x on L2 [0, 1] should be identified as [0, 1] and the fact that Mφ has no eigenvalues should be noted. Section 6. The fact that the Volterra operator has no eigenvalues should be established, hence showing that some compact operators may have spectrum equal to {0}. It is useful to review the orthogonal diagonalization of real symmetric matrices and/or unitary diagonalization of Hermitian (i.e. selfadjoint) matrices. It is instructive to re-write both these results and Theorem 6.9 in terms of projections onto eigenspaces. Section 7. This section is motivated by an informal discussion of of the Green’s function as the response of a system to the input of a unit pulse. This is illustrated by the elementary example of finding the shape under gravity of a heavy string (of variable density) fixed at (0, 0) and ( , 0). This is found by calculating the (triangular) shape k(x, t) of a weightless string with a unit weight at x = t and then using an integration process. The differential equation is also found and shown to give the

John Erdos

3

same answer. (Naturally, usual elementary applied mathematical assumptions - small displacements, constant tension - apply.) Additionally, a brief, very informal discussion of delta functions and the Green’s function as the solution of the system with the function f being a delta function is of interest. It should be stressed, however, that the proof of Theorem 7.1 is purely elementary and quite independent of the discussions above. Section 8. The most important part of this final section is Theorem 8.3, the continuous functional calculus. This is sufficient for the vast majority of applications of the spectral theorem for bounded self-adjoint operators. These include, for example, the polar decomposition and the properties of eAt and these are done in the course. The approach here to the general spectral theorem is elementary and very pedestrian. It should be noted that, given the appropriate background, there are more elegant ways. These include using the identification of the dual of C[m, M ] (actually of C(σ)) as a space of measures. There is also a Banach algebra treatment. In an elementary course such as this, the technicalities of the spectral theorem need not be strongly emphasized. However, a down to earth approach should clarify the meaning of theorem and remove the mystery often attached by students to these operator integrals.

4

Operators on Hilbert space

1

Elementary properties of Hilbert space

Definition A (complex) inner (scalar) product space is a vector space H together with a map · , · : H × H → C such that, for all x, y, z ∈ H and λ, µ ∈ C, 1. λx + µy, z = λ x, z + µ y, z , 2. x, y = y, x , 3. x, x ≥ 0, and x, x = 0 ⇐⇒ x = 0 . Properties 1,2 and 3 imply ¯ 4. (x, λy + µz = λ x, y + µ x, z , ¯ 5. x, 0 = 0, x = 0. Theorem 1.1 (Cauchy-Schwartz inequality) | x, y | ≤ x, x
1/2

y, y

1/2

, ∀x, y ∈ H.

Proof. For all λ we have λx + y, λx + y ≥ 0. That is, for real λ λ2 x, x + λ( x, y + y, x ) + y, y ≥ 0 . In the case that x, y is real, we have that the discriminant (“b2 − 4ac”) of this quadratic function of λ is negative which gives the result. In general, put x1 = e−iθ x where θ is the argument of the complex number x, y . Then x1 , y = eiθ x, y = | x, y | is real and x1 , x1 = x, x . Applying the above to x1 , y gives the required result. [Alternatively, put λ = − Theorem 1.2 x = x, x
1/2 y,x x,x

in λx + y, λx + y ≥ 0.]

is a norm on H.
1

Proof. The facts that x ≥ 0, x = 0 ⇐⇒ x = 0 and λx = λx, λx 2 = |λ|. x are all clear from the equivalent properties of the inner product. For the triangle inequality, x+y
2

= x 2+ ≤ x 2+ ≤ x 2+ = ( x +

y 2 + x, y + y, x y 2 + 2| x, y | y 2+2 x . y using (1.1) 2 y ) .

Lemma 1.3 (Polarization identity) x, y = 1 4 x+y
2

− x−y

2

+ i x + iy

2

− i x − iy

2

.

John Erdos Proof. x+y − x−y i x + iy −i x − iy
2 2 2 2

5

= x 2 + y 2 + x, y + y, x = − x 2 + − y 2 + x, y + y, x = i x 2 + i y 2 + x, y − y, x = −i x 2 + −i y 2 + x, y − y, x .

Adding the above gives the result. Lemma 1.4 (Paralellogram law) x+y Proof. x+y x−y
2 2 2

+ x−y

2

=2 x

2

+2 y

2

.

= =

x x

+ y 2 + y

2

2 2

+ x, y + y, x − x, y − y, x .

Adding the above gives the result. Definition x is said to be orthogonal to y if x, y = 0; we write x ⊥ y. Lemma 1.5 (Theorem of Pythagoras) x, y = 0 =⇒ x Proof. Obvious. Definition. If H is an inner product space and (H, · ) is complete then H is called a Hilbert space. A set C (in a vector space) is convex if x, y ∈ C =⇒ αx + (1 − α)y ∈ C whenever 0 ≤ α ≤ 1 . In a metric space, the distance from a point x to a set S is d(x, S) = inf{ x − s : s ∈ S}. Theorem 1.6 If K is a closed convex set in a Hilbert space H and h ∈ H then there exists a unique k ∈ K such that d(h, K) = h − k . Proof. Let C = K − h = {k − h : k ∈ K}. Note that C is also closed and convex, d(h, K) = d(0, C) and if c = k − h ∈ C is of minimal norm then k is the required element of K. Therefore it is sufficient to prove the theorem for the case h = 0.
2

+ y

2

= x+y

2

.

] Since H is complete and C is closed.6 Operators on Hilbert space Let d = d(0. Using the parallelogram law (Lemma 1. m → ∞. Note that the above proof is putting a geometrical argument into symbolic form. = 2 cn 2 + 2 cm 2 − cn + cm ≤ 2( cn 2 + cm 2 ) − 4d2 → 0 (∗) as n. suppose also that c ∈ C with c as for (*) (with cn = c and cm = c ) shows that 0≤ c−c and so c = c . by Pythagoras (Lemma 1. h−n 2 = h − n0 + n0 − n 2 = h − n0 2 + n0 − n 2 ≥ h − n0 2 .5) for all n ∈ N . 2 cn +cm 2 ≥ d. The justification for this is that the prefix “sub” refers to a substructure. To prove uniqueness. (cn ) converges to an element c ∈ C and c = limn→∞ cn = d. Hence inf n∈N h − n is attained at n0 . it should be . The same calculation ≤2 c 2 +2 c 2 − c+c 2 ≤ 2d2 + 2d2 − 4d2 = 0 ≤ = = z. 2 given > 0. The c ≥ d for all c ∈ C. Choose a sequence (cn ) such that ( cn ) → d. n |2 z 2− n 2 2 so z. n = 0. that is. z 2 2 = d. Then for all non-zero n ∈ N . Lemma 1. cn + cm But. [ Since ( cn ) → d. Then (*) shows that for n. N ) = h − n0 if and only if h − n0 . Suppose d(h. C) = inf c∈C c . Conversely if h − n0 ⊥ N then. N ) = h − n0 .4). since C is convex. 2( cn 2 − d2 ) < 2 . n In these notes the term subspace (of a Hilbert space) will always mean a closed subspace. The quantity z. cn −cm < . m > n0 .n 2n is the “resolution of the vector z in the direction of n”.7 If N is a closed subspace of a Hilbert space H and h ∈ H then d(h. n n z− n 2 2| z. Therefore 0 ≤ cn − cm 2 2 + cn − cm ∈ C and so 2 = 2 cn cn +cm 2 2 + 2 cm 2 . there exists n0 such that for n > n0 . n = 0 for all n ∈ N . Write z = h − n0 . n |2 + z 2− n 2 n 2 | z. It follows easily that (cn ) is a Cauchy sequence. Proof. that is cn + cm 2 ≥ 4d2 . n |2 | z. so the subspace should be a Hilbert space in its own right.

If follows easily from the above that both (xn ) and (yn ) are Cauchy sequences so. Proof. s = 0. by the continuity of the inner product. Lemma 1. x ∈ M and y ∈ N . zn+p − zn 2 = xn+p − xn 2 + yn+p − yn 2 . (i) Clearly S ⊥ is a vector subspace. since M and N are closed subspaces. As (zn ) is convergent.John Erdos 7 complete. For the converse. But it is an easy fact that a subset of a complete space is complete if and only if it is closed. Lemma 1. s | ≤ t − tn . Let zn ∈ M ⊕ N such that (zn ) → z. Now zn = xn +yn with xn ∈ N and yn ∈ M . / Then there is n0 ∈ N such that d(h. | t. [In grim detail. For (v). if x ∈ S ∩ S ⊥ then x. s → 0 . let tn ∈ S ⊥ be a sequence converging to t.8 If N is a (closed) subspace of a Hilbert space H. N ) = h − n0 and the Lemma shows that 0 = h − n0 ⊥ N . using Pythagoras (Lemma 1. Proof.10 If M and N are orthogonal subspaces of a Hilbert space then M ⊕ N is closed. Thus z = lim(xn + yn ) = x + y ∈ M ⊕ N . s = 0 lim so t ∈ S ⊥ . Note that since N ⊥ M . For (iv). (xn ) and (yn ) both converge. To show it is closed. t. since t.5) since M ⊥ N . . Therefore. S ⊥ = S ⊥⊥⊥ . Then. apply (iii) to S ⊥ yields S ⊥ ⊆ S ⊥⊥⊥ . Given a subset S of H the orthogonal complement S ⊥ is defined by S ⊥ = {x : x.9 For subsets of a Hilbert space H (i) (ii) (iii) (iv) (v) S ⊥ is a closed subspace. Then. We need to show that z ∈ M ⊕N . N ⊥ = (0) ⇐⇒ N = H Proof. we have that N ∩ M = (0) and the sum M + N is automatically direct. Definition. x = 0 so x = 0. s = n→∞ tn . Clearly. s | = | t − tn . and applying (ii) to (iii) to gives the reverse inclusion. Call the limits x and y. for all s ∈ S. ⊥⊥ S⊆S . so N ⊥ = (0). s does not depend on n. so. if N = H take h ∈ N . s = 0 for all s ∈ S} . ⊥ ⊥ S1 ⊇ S2 =⇒ S1 ⊆ S2 .] (ii) and (iii) are clear. S ∩ S ⊥ = (0). since H is complete. it is a Cauchy sequence. t. if N = H then N ⊥ = (0). Corollary 1.

8. N ⊕ N ⊥ = H. so t = 0 and x = s ∈ N.12 (i) (ii) S. (ii) Clearly S ⊥⊥ is a subspace containing S. .8 Operators on Hilbert space Theorem 1. Therefore. Since H = N ⊕ N ⊥ . Proof.11 If N is a subspace of a Hilbert space H then N ⊕ N ⊥ = H. if x ∈ (N ⊕ N ⊥ )⊥ then x ∈ N ⊥ ∩ N ⊥⊥ so x = 0. from Corollary 1. (i) From Lemma 1.9 (ii)) S ⊥⊥ ⊆ M ⊥⊥ = M . if x ∈ N ⊥⊥ then x = s + t with s ∈ N and t ∈ N ⊥ . If M is any subspace containing S then (Lemma 1. Corollary 1. Also. N ⊕ N ⊥ is a (closed) subspace. From above. If N is a subspace then N ⊥⊥ = N . For any subset S of a Hilbert space H. But then also t = x − s ∈ N ⊥⊥ . S ⊥⊥ is the smallest subspace containing Proof.9 (iii) N ⊆ N ⊥⊥ .

). 1] be the space of complex-valued functions which are continuously differentiable on [0. g = 1 0 {f (t)g(t) + f (t)g (t)} dt defines an inner product on W [0. 9. h2 + g2 ) α(h1 . A challenging but not very important exercise : 10. Let x and y be vectors in an inner product space. One writes H = H1 ⊕ H2 . N k=1 1 2π x. is dense in 2 . α3 . Show that (i) (M + N )⊥ = M ⊥ ∩ N ⊥ 2 (ii) (M ∩ N )⊥ = M ⊥ + N ⊥ . x−t .] 8. y = x + eiθ y 2 eiθ dθ . α. Let W [0. spanned by sequences of the form (1. h2 ) = (αh1 . . z−t + y−z . 7. y−t ≤ x−y . g1 + h2 . Prove that in a complex inner product space the following equalities hold: 1 N x. 2. 2π 0 [This generalizes the polarization identity. For which real α does the function f (t) = tα belong to (i) L2 [0. αh2 ). [H is called the (Hilbert space) direct sum of H1 and H2 . ∞] ? x+y = x + y . ∞] (iii) L2 [0. . h2 ) + (g1 . y | ≤ x the vectors x and y are linearly dependent. Prove that in the Cauchy–Schwarz inequality | x.John Erdos 9 Exercises 1 1. 4. (g1 .] 3. considered as a map from H × H to C. 1]. . x−z . Show that. g2 ) = (h1 + g1 . 5. Let M and N be closed subspaces of a Hilbert space. for any four points of a Hilbert space. Let H be the set of ordered pairs H1 × H2 with addition and multiplication defined (in the usual way) as follows: (h1 . . Given that show that x and y are linearly dependent. show that the inner product. g2 ) = h1 . Show that the inner product defined by (h1 . y = x + e2πik/N y 2 e2πik/N for N ≥ 3. 1] (ii) L2 [1. 1]. α2 . Show that. f. Show that the vector subspace of where 0 ≤ α < 1. g2 satisfies the axioms for an inner product and H with this inner product is a Hilbert space. is continuous. y the equality holds iff 6. h2 ). For a Hilbert space H. Let H1 and H2 be Hilbert spaces.

For L ∈ B(X. define L = supx=0 Exercise. Therefore. x x ≤1 x =1 If Y is complete then so is B(X. The elements of the dual are called (continuous) linear functionals. H will always denote a Hilbert space. We shall be concerned with Hilbert spaces. 2. cx = c < δ and so x L cx x =c Lx < . to show continuity at any point x0 . If 3 holds. given any > 0. 1 implies 2 is obvious. Then the following are equivalent : 1. The set of all continuous (bounded) linear maps X → Y is denoted by B(X. Then if x − x0 < δ we have Lx − Lx0 < . x This shows that Lx < K x where K = c . Y ). Lemma 2.1 Let X and Y be normed linear spaces and let L : X → Y be a linear map. When X = Y we write B(X). let δ = K . . Some of the results in this section are stated for normed linear spaces but they will be used in the sequel only for Hilbert spaces. Then for any x = 0. Theorem 2. Continuity at 0 shows that there is a corresponding δ > 0 such that LX < whenever x < δ. L is continuous at 0. Proof. take any > 0. 3. · Lx x . Y ) When Y = C then B(X. h for some h ∈ H. C) is called the dual of X and denoted by X (sometimes by X ∗ ). is a norm on B(X. where f = h . there exists a constant K such that Lx ≤ K x for all x ∈ X.2 (Riesz representation theorem) Every linear functional f on H is of the form f (x) = x. If 2 holds.10 Operators on Hilbert space 2 Linear Operators. Y ). L is continuous. note that Lx − Lx0 = L(x − x0 ) ≤ K x − x0 . Y ) and L = sup x=0 Lx = sup Lx = sup Lx . Take some c with 0 < c < δ.

(x + y) − A(x − y). x + Ay. y) = αφ(x. Adding the above gives the result. z − f (x) z 2 = 0.John Erdos 11 Proof. y ∈ H. Since z ⊥ N . αy + βy ) = αφ(x. x = 0 for all x ∈ H then A = 0. y − Ay. x = 0 for all x ∈ H the above shows that Ax. A bilinear form (also called a sesquilinear form) φ on H is a map φ : H × H → C such that φ(αx + βx . y for some A ∈ B(H). y) + βφ(x. h .5 (Riesz) Every bounded bilinear form φ on H is of the form φ(x. y)| ≤ K x . y) + βφ(x . A(x + y). Thus A = 0. Proof. x + i Ay. y) = Ax. y for all x. Also f | = sup x=0 |f (x)| |f (h)| ≥ = h . y + Ay. y + Ax. h so f ≤ h . note that |f (x)| = | x. y) ¯ φ(x. y + Ax. (x − iy) = Ax. Writing h = zz 2 we obtain f (x) = x. If f = 0. y + Ax. h | ≤ x . If Ax. Then y f (y) = 0. Proof. x = −i Ax. (x + iy) −i A(x − iy). For f = 0 then N = f −1 (0) = {x : f (x) = 0} = H. x − f (x)z. y ) . y + Ax. |φ(x. (x − y) i A(x + iy). take h = 0. (x + y) − A(x − y).f (z) = 0 and so x − f (x)z ∈ N . (x − iy) ] . y = 0 for all x. y = 1 [ A(x + y). x − Ay. Write z = f (y) so that f (z) = 1 [using f (αx) = αf (x)]. x = − Ax. y + Ay. For the norm.3 (Polarization identity for operators) Ax. for some constant K. = h shows that the correspondence between H and its Lemma 2. y − Ay. x = i Ax.4 If Ax. z = x. N is closed. (x + iy) − i A(x − iy). Definition Let H be a Hilbert space. For any x ∈ H f (x − f (x)z) = f (x) − f (x). Also. . Thus N ⊥ = (0) so take y ⊥ N . Theorem 2. ¯ A bilinear form is said to be bounded if. x − i Ay. Corollary 2. y ∈ H and so using y = Ax it follows that Ax 2 = 0 for all x ∈ H. x h Note that the result f dual is one to one. since f is continuous. x . (x − y) 4 +i A(x + iy).

(the adjoint of A ∈ B(H. All the above can be done with essentially no change. Using Theorem 2. Then ψ(x. Definition The adjoint. so that φ(x. (ii) ¯ (λA)∗ = λA∗ . One can find such an h corresponding to each x ∈ H. Bilinear forms could have been defined as maps φ from H × K to C where H and K are different Hilbert spaces. Note. y |φ(x. h . then U is said to be unitary. y so A(x + x ) = Ax + Ax [since A(x + x ) − Ax − Ax ∈ H⊥ = (0)]. y) + φ(x . y) = φ(x. Then A is linear since . Let A ∈ B(H). Ay is a bounded bilinear form on H so. φ(x. Definition. for all x. H)). y) = Ax. Also. K) is then an operator in B(K. y . then A is said to be normal. Similarly A(αx) = αAx. y) = x. y) = y. y = φ((x + x ). Exercise. Ay .12 Operators on Hilbert space Proof. y + Ax . y) = x. y. (A + B)∗ = A∗ + B ∗ . y) is conjugate linear in y. y = ψ(x. x . A(x + x ).5 there is an operator A∗ ∈ B(H) such that A∗ x. y) = h. A∗ is called the adjoint of A. (i) (iii) (v) (A∗ )∗ = A. If If If A = A∗ AA∗ = A∗ A U U ∗ = U ∗U = I then A is said to be selfadjoint. Consider x fixed for the moment. Define a function H → H by Ax = h. y) is linear in y. y)| = sup ≤K x y y y=0 Ax = sup y=0 so A is continuous. that is φ(x. A = A∗ . Then φ(x.2 we have that there is a (unique) h ∈ H. . Ax. (iv) (AB)∗ = B ∗ A∗ . by Theorem 2.

But for any h ∈ H. (iii) If h ∈ N [then h = h + 0 is the decomposition of h and] P h = h. (iii) P 2 = P . Lemma 2. Ex = E ∗ (I − E)h. P h = x. Then N = ker(I − E). Then from Theorem 1. x = 0. In this course we shall not consider projections that are not orthogonal and usually call these operators “projections”. if E ∈ B(H) satisfies E = E 2 = E ∗ let N = {x : Ex = x}. P h. if E ∈ B(H) satisfies E = E 2 = E ∗ then E is the (orthogonal) projection onto some (closed) subspace. P 2 = P . so N is closed. (iv) If h = x + y and h = x + y with x. The orthogonal projection P onto N is defined by P h = x (where h = x + y is the decomposition above). For any h ∈ H. h ∈ H and suppose h = x + y and h = x + y are the unique decompositions of h and h with x. x and so P = P ∗ . Similarly h.John Erdos Projections. . (ii) P = 1 (unless N = 0). Also. This shows that E is the projection onto N . y ∈ N ⊥ . any h ∈ H has a unique decomposition as h = x + y with x ∈ N and y ∈ N ⊥ .11. But if 0 = h ∈ N then P h = h and so P = 1. x ∈ N and y. y ∈ N ⊥ . (i) Let h. Ph 2 = x 2 ≤ x 2 + y 2 = h 2 and so P ≤ 1. Let N be a closed subspace of H. Note that then y = (I − P )h and I − P is the projection onto N ⊥. x = (E 2 − E)h. Finally.6 Let N be a closed subspace of H and let P be the orthogonal projection onto N . x = (I − E)h. x since x ∈ N and y ∈ N ⊥ . Then αh + βh = (αx + βx ) + (αy + βy ) is the decomposition of αh + βh and P (αh + βh ) = αx + βx = αP h + βP h . Then Eh ∈ N since E(Eh) = E 2 h = Eh and (I − E)h ⊥ N since if x ∈ N . P h ∈ N so P (P h) = P h. Ex = x and (I − E)h. write h = Eh + (I − E)h . (iv) P ∗ = P . h = x. Then (i) P is linear. (ii) If h = x + y with x ∈ N and y ∈ N ⊥ . x + y = x. x ∈ N and y. H = N ⊕ N⊥ 13 that is. Proof. that is.

14 Operators on Hilbert space Lemma 2. N ) = (I − P )h . N ⊥ is invariant under A ⇐⇒ P A = P AP . If these equalities hold then P A = AP . Lemma 2. Proof. (i) (ii) N is invariant under A ⇐⇒ AP = P AP . Then since P h ∈ N for all h ∈ H. Finally AP = P AP ⇐⇒ (AP )∗ = (P AP )∗ ⇐⇒ P A = P AP since A = A∗ . n = 0 for all n ∈ N .7 d(h. Trivial algebra shows that A(I − P ) = (I − P )A(I − P ) ⇐⇒ P A = P AP and so (ii) follows from (i). we have AP h ∈ N . If A = A∗ then N is invariant under A ⇐⇒ N ⊥ is invariant under A ⇐⇒ P A = AP . . Proof. (ii) The projection onto N ⊥ is I − P . N ) = h − P h = (I − P )h .7 If P is the orthogonal projection onto a subspace N then for all h ∈ H. Therefore from Lemma 1. ⇐= If n ∈ N then P n = n and so An = AP n = P AP n ∈ N [since N is the range of P ]. d(h. Therefore then P AP h = AP h [since P n = n for all n ∈ N ]. (i) =⇒ Suppose An ∈ N for all n ∈ N . For any h ∈ H we have P h ∈ N and (I − P )h.8 Let A ∈ B(H) and P be the orthogonal projection onto a subspace N .

A ≤ 1. x is real for all x ∈ H. 5. . 9. . ξ2 . Prove that. 0. F be the orthogonal projections onto subspaces M and N respectively. . (ii) the range of A is closed (iii) (ran(A))⊥ = (0). ξ3 . 8. where (αn ) is a bounded sequence of complex numbers. 1] be defined by (T f )(x) = √ 2x f (x2 ) .) . Ax ≥ c a positive constant.) = (0. there exists X −1 ∈ B(H) such that XX −1 = X −1 X = I) prove that X ∗ is invertible and (X ∗ )−1 = (X −1 )∗ . ξ2 . . ξ1 .) = (0.). . ξ3 . Show that A has a continuous inverse. α3 ξ3 . . 6. . 0. .) = (α1 ξ1 . Given that X ∈ B(H) is invertible (that is. −1. Find the adjoint of T and deduce that T is unitary. . The operator A ∈ B(H) satisfies Ax = x for some x ∈ H. ξ3 . Let T : L2 [0. (iii) EF = F E ⇐⇒ E + F − F E is an orthogonal projection. . α3 ξ3 . (i) (ii) EF = F ⇐⇒ N ⊆ M ⇐⇒ E − F is an orthogonal projection. (iii) X is unitary ⇐⇒ Xx = X x = x 2. where c is [Hints : Show (i) A is injective. EF = 0 ⇐⇒ N ⊆ M ⊥ ⇐⇒ E + F is an orthogonal projection. α1 ξ1 . for all x ∈ H. . . 1] by (V f )(x) = 7. . 1] → L2 [0. x . . Prove that A∗ x = x + y where y ⊥ x. Let X ∈ B(H). X is normal ⇐⇒ Xx = X ∗ x ∗ for all x ∈ H. .] Note that the selfadjointness condition is needed – consider the operator S on 2 defined by S(ξ1 . x 0 f (t) dt .). Let A be a selfadjoint operator such that for all x ∈ H. α2 ξ2 . W (ξ1 .) .John Erdos 15 Exercises 2 1. ξ3 . ξ2 . . α2 ξ2 . . . Use the above to show that ran(E) = ran(E ∗ ) and ker(E) = ker(E ∗ ) and deduce that E = E ∗ (so that E is the orthogonal projection onto some subspace of H). ξ2 . Let S be the one-dimensional subspace of 2 spanned by the element (1. . 4. The operators D and W on 2 are defined by D(ξ1 . . Show that : (i) (ii) X is selfadjoint ⇐⇒ Xx. If. Show explicitly that any element x = (ξk ) ∈ 2 can be written as x = x1 + x2 where x1 ∈ S and x2 ⊥ S. Show that W and D are bounded operators and find their adjoints. . Find the adjoint of the operator V defined on L2 [0. further. . 3. show that A∗ x = x. Let E. Suppose that E 2 = E and E = 1.

2 and let M and N be linear subspaces of l2 ⊕ l2 defined by M = {(x. . . Calculate the values in the specific case where f (t) = t2 + t. .) . . Let Lo and Le be subspaces of L2 [−1. 1] onto Lo and Le . . . 3 ξ3 . 1 ξ2 . Show that the set of sequences x = (ξn ) such that subset of l2 . ξ2 . 1] and find the projections of L2 [0.) = (ξ1 . Show that Lo ⊕ Le = L2 [−1. 1] defined by Operators on Hilbert space Lo = {f : f (t) = −f (−t) (almost everywhere )} Le = {f : f (t) = f (−t) (almost everywhere )}. Dx) : x ∈ l2 } . 12. Observe that M is closed and use the continuity of D to show that N is also closed. Define the operator D on l2 by n n2 |ξn |2 converges. ˙ 11. 0) : x ∈ l2 } N = {(x.16 10. ξ3 . Find expressions for the distances of any element f to Lo and Le . Show that M ∩ N = (0) and that the algebraic direct sum of M and N is dense in l2 ⊕ l2 but is not equal to l2 ⊕ l2 (and so it is not closed). forms a dense 1 D(ξ1 . Prove that M and N are closed. Let M and N be vector subspaces of H such that M ⊥ N and M +N = H. .

n i=1 |αi |2 ≤ h 2 . 2. Put hr = r i=1 αi xi .j=1 n αi αj xi . Now let r → ∞. h + i. 2. Definition. for any h ∈ H. xi = 0 for all i. y = 0 if x = y and x. . For any fixed i. xi = 0 for all i. x. given any > 0 there exists n0 such that for r > n0 . i=1 n αi xi − i=1 n α i xi . p > 0 we have r+p 2 2 . (Note that the case n = 1 is the Cauchy-Schwartz inequality) Proof. y ∈ S. for any h ∈ H. i=r+1 |αi | < since H is complete. xi = 0 for all r > i. i=r+1 = Now n |αi |2 ≤ h 2 for all n and so ∞ |αi |2 is convergent and so is a Cauchy i=1 i=1 series. x = 1 for all x ∈ S. xi . Then r+p 2 hr+p − hr 2 = i=r+1 r+p αi xi |αi |2 . Lemma 3. h − hr . h + i=1 2 − i=1 |αi |2 ≥ 0. Let h ∈ H. h − i=1 n αi xi n n h h h 2 − h. 3 · · ·} is an orthonormal sequence then. writing αi = h. writing αi = h. Call its limit h . Hence. A set S of vectors of H is said to be orthonormal if 1. Then n n n h− i=1 α i xi 2 = = = = h− i=1 αi xi . xj |αi |2 2 − h. Then it follows from the continuity of the inner product that h − h .2 If {xi : i = 1. Proof. hr+p − hr < .John Erdos 17 3 Orthonormal Sets. it is convergent. i=1 n αi xi − i=1 α i xi .1 (Bessel’s inequality) If {xi : 1 ≤ i ≤ n} is a finite orthonormal set then. that is. ∞ αi xi i=1 converges to a vector h such that h − h . xi . Therefore (hr ) is a Cauchy sequence and. Lemma 3.

2. g = i=1 αi βi . 3 · · ·} be an orthonormal sequence. (i) =⇒ (ii). 2. Definition. 3 · · ·} is maximal (that is.3 Let {xi : i = 1. 3 · · ·} gives (ii) =⇒ (iii). 2. (In the above. If (ii) is false then adding a larger orthonormal set. (i) {xi : i = 1. xi .s] hr .) Proof. (Fourier expansion) For all h ∈ H we have h = (Bessel’s equality) For all h ∈ H we have h 2 ∞ i=1 αi xi . it is not a proper subset of any orthonormal set). 2. Using the continuity of the inner product. xi = 0. 3 · · ·} is not maximal and can be enlarged by adding z then z. xi = 0 i=1 for all i and so h = h by (ii)¿ (iii) =⇒ (iv). g ∈ H we have h. g = = ∞ i=1 αi βi . contradicting (i). (v) =⇒ (i). Such a series is said to be unconditionally convergent. A maximal orthonormal sequence is called an orthonormal basis. Then h − h . Let h = ∞ αi xi (this exists. (iv) =⇒ (v). h h to the set {xi : i = 1. Let r → ∞ and s → ∞. xi = 0 for all i but also ∞ 1= z 2 = i=1 | z. it follows that ∞ h. .18 Operators on Hilbert space Theorem 3. g s = i=1 αi βi . αi = h. xi 2 =0 which give a contradiction. If {xi : i = 1. xi and βi = g. ∞ i=1 (Parseval’s relation) For all h. for all i then h = 0. The the following statements are equivalent. by Lemma 3. Then min[r. (ii) (iii) (iv) (v) If αi = h. It follows that the series giving the fourier expansion of a vector can be re-arranged arbitrarily without altering its convergence or its sum. Clearly the concept of an orthonormal basis refers to a set of vectors so that any permutation of such a set is still an orthonormal basis. Put g = h in (iv).2). Let hr = r i=1 αi xi and gs = s i=1 βi xi . |αi |2 .

x1 x1 un = y n − i=1 y n . 2. In the general case. 2. y2 . A few statements would also need modification. y2 · · · . Then there exists an orthonormal sequence {xi : i = 1. since both sides have dimension k. y2 . and take this as the definition of separability. · · · . dense subset. yr−1 this element is ignored. x3 · · · xk } ⊇ span{y1 . For the rest of this course we shall often need to assume that the Hilbert space we consider has a countable orthonormal basis.4 (Gram-Schmidt process) Let {yi : i = 1. 3 · · ·} is linearly independent. . n−1 y1 y2 − y2 . . The appropriate way to state this restriction is to say that the Hilbert space we consider is separable. When {yi : i = 1. 3 · · ·} be a sequence of vectors of H. 2. xr = 0 and so it follows easily that {xi : i = 1. 3 · · ·} such that. . It is clear in general that xr is a linear combination of on y1 . 2. yk−1 } and. x3 · · · xk } ⊇ span{y1 . xk−1 ∈ span{y1 . for each integer k span{x1 . First consider the case when {yi : i = 1. x2 . x i xi u1 . y2 . y3 · · · yk } . If {yi : i = 1. · · · . 2. 3 · · ·} is a linearly independent set we have equality. This is true for all spaces considered in the applications. 3 · · ·} is a linearly independent set. y3 · · · yk } . The restriction is a rather technical matter and could be avoided but this would entail a discussion of uncountable sums. For our purposes one could say that a Hilbert space is separable if it has a countable orthonormal basis. u2 . 3 · · ·} is an orthonormal sequence. then the above inclusion is an equality for each k. this is a more general notion: recall that a metric space is defined to be separable if it has a countable. u1 u2 x2 = . A further easy induction shows that for r < n we have un . 2. y2 . Proof. The proposition which follows connects these ideas. the same construction applies except that whenever yr is a linear combination of y1 . is obvious. 2. However. xn = un x1 = Easy inductive arguments show that for each k. 3 · · ·} is a linearly independent set. Define u1 = u2 = . yr−1 and so the inclusion span{x1 . . that uk = 0.John Erdos 19 Theorem 3. un . . x2 . since {yi : i = 1.

We shall not prove this in detail. 3. Since these balls are disjoint and since every open set must contain at least one element of a dense subset. but here is a sketch. every orthonormal set is countable. Conversely. For each element xα of an √ orthonormal set {xα }α∈A . . rn rational } is countable. let Bα be the open ball centre xα . 2. In a separable Hilbert space. applying the Gram-Schmidt process to a countable dense subset results in an orthonormal sequence that is easily proved to be a basis. if {xn : n = 1. A Hilbert space is separable if and only if it has a (countable) orthonormal basis. 2. It is clearly dense because the closure of S is a subspace containing an orthonormal basis. · · ·} is a countable orthonormal basis. radius 22 . For 2. tedious but routine arguments show that the set N S={ n=1 rn xn : N finite . it is clear that if the space is separable the orthonormal set must be countable.20 Proposition Operators on Hilbert space 1.

Show further that the following are equivalent : {Ni } is not a proper subset of any orthogonal set of subspaces of H. . . x 2= ∞ Ei x 2 . ∞ i=1 x. x= ∞ i=1 Ei x. |f (x)| ≤ C f almost everywhere . Ei y . Find the first three functions obtained by applying the Gram. f2 . y = 0 if i = j. Note that under the given conditions. [Hint: for any orthonormal subset f1 . Deduce that n i=1 |fi (y)|2 ≤ C 2 and integrate this relation with respect to y. Prove that N is finite dimensional. . for each x. evaluate. for any fixed y. for each x ∈ H. 1] with the property that for some fixed constant C and each f ∈ N . Ei x converges to some h ∈ H which satisfies (x − h) ⊥ Ni for each i. (e) for each x ∈ H. 3. . y ∈ H. 1. 8 (ii) of Sheet 2 or its equivalent. this process yields the Legendre polynomials. of N . h ⊥ Ni for all i ⇒ h = 0. Ei x. fn . (i) (ii) (a) (b) (c) (d) for a finite subset {Ni }n of {Ni }. the norm of g where n g(x) = i=1 fi (y)fi (x) . Show that for each x ∈ H. Let {Ni } be a sequence of mutually orthogonal subspaces of a Hilbert space H and let {Ei } be the sequence of projections onto {Ni }.] Use your results and the theory developed in lectures to find a.Schmidt process to the elements {tn : n = 0. . b and c which minimises the quantity 1 −1 |t4 − a − bt − ct2 |2 dt . [Note: apart from constant factors. since Ei = Ei∗ = Ei2 .] 2. 1]. Let N be a subspace of L2 [0. i=1 ∞ i=1 n i=1 Ei x 2 ≤ x 2 . Ej y = Ej Ei x.] .} of L2 [−1. i=1 [Hints : You will need Q. y = Ei x.John Erdos 21 Exercises 3 1. .

Let (Kn ) be a sequence of compact operators converging to K. we must show that (αAxn ).1 K(H) is an ideal of B(H). Proof. since A is compact. (T Axn ) and (AT xn ) all have convergent subsequences. we need to show that if (xi ) is a bounded sequence the (Kxi ) has a convergent subsequence. B ∈ K(H) and T ∈ B(H) then αA. it H is infinite-dimensional then and T ∈ B(H) has an inverse T −1 ∈ B(H) then T is not compact. since T ∈ B(H). Also. for any a bounded sequence (xn ). for each n. Note that every operator of finite rank is compact. To show that K is compact. i i i let (x3 ) be a subsequence of (x2 ) such that (K3 x3 ) is convergent. Again. Proof. (Bxni ) has a convergent subsequence (Bxnij ). Then ([A + B]xnij ) is a convergent subsequence of ([A + B]xn ). An operator K ∈ B(H) is said to be compact if for every bounded set S of vectors of H the set {Ks : s ∈ S} is compact. We shall denote the set of all compact operators on H by K(H). i i let (x2 ) be a subsequence of (x1 ) such that (K2 x2 ) is convergent. xn we have that (T xn ) is bounded and so. A consequence of the above theorem is that. (xni ) is a bounded sequence and so. Theorem 4. (AT xn ) has a convergent subsequence. Note also that the identity operator on a Hilbert space H is compact if and only if H is finite-dimensional. Also. since (xn ) is bounded and T xn ≤ T . Since A is compact.22 Operators on Hilbert space 4 Compact Operators. Theorem 4. T A and AT are all in K(H). showing that A + B is compact. T is continuous and so (T Axni ) is a convergent subsequence of (T Axn ) showing that T A is compact. Here. Definition. Equivalently : Definition. i . That is. Definition. We need to show that. The proof for AT is slightly different. showing that T A is compact. if A. (zi ) is a subsequence of (xn ) and so (Kn zi ) is convergent. apart from the first i n terms. A + B. The rank of an operator is the dimension of its range. since B is compact. Then (zi ) is a subsequence of (xi ). i i i and continue in this way. (Axn ) has a convergent subsequence (Axni ).2 K(H) is closed. Let (x1 ) be a subsequence of (xi ) such that (K1 x1 ) is convergent. [The notation above is slightly unusual and is adopted to avoid having to use subscripts on subscripts on · · · .] Let zi = xi . Then clearly (αAxni ) is a convergent subsequence of (αAxn ) showing that αA is compact. An operator K ∈ B(H) is said to be compact if for every bounded sequence (xn ) of vectors of H the sequence (Kxn ) has a convergent subsequence. This is an immediate consequence of the Bolzano-Weierstrass theorem which states that every bounded sequence in Cn has a convergent subsequence. ([A + B]xn ).

Therefore. b]. We shall apply Hilbert space theory to L2 (S) which is a Hilbert space of functions of 2 variables with the inner product φ. Now. for fixed t the function φ(x. Combining these with the displayed inequality shows that for i > i0 . t) is of the form u(x)v(t) then (Rf )(x) = is a rank one operator. |k(x. For all i. n we have Kzi − Kzj = ≤ (K − Kn )zi + Kn zi − Kn zj − (K − Kn )zj K − Kn ( zi + zj ) + Kn (zi − zj ) . Then. t) is orthogonal to ui (x) for each i and so it is zero. Kzi − Kzj < so (Kzi ) is convergent as required.John Erdos 23 We now show that (Kzi ) is convergent by showing that it is a Cauchy sequence. ψ = b a b a b a u(x)v(t)f (t) dt = f. j > i0 . Let > 0 be given. since (uj ) is a basis of L2 [a. We shall show that K is the norm limit of a sequence of finite rank operators. t)ui (x) dx of t is orthogonal to uj (t) for each j.j=1 orthonormal sequence. Since (Kn ) → K we can find n0 such that K − Kn < 4c for n > n0 where c satisfies xi ≤ c for the bounded sequence (xi ). ui (x)uj (t). t) ⊥ ui (x)uj (t) for all i. b]. . t)ψ(x. Thus (ui (x)uj (t))∞ is an i. t)f (t) dt . it is (equivalent to) the zero function. t) dx dt . it is a Cauchy sequence and so there is an i0 such that for i > i0 . We have already seen that operators of the above type are continuous with K ≤ M (Recall that k(x. Then (ui (x)uj (t))∞ is an orthonormal i. for each i. Note that if k(x. The operator K on L2 [a. t)ui (x) dx uj (t) dt . Let (ui ) be an orthonormal basis of L2 [a. suppose φ(x. in which case the integral is 1. is compact. j > i0 we have Kn zi − Kn zj < 2 . since (Kn zi ) converges. b] × [a.j=1 basis of L2 (S). b] defined by (Kf )(x) = where b b a a b a k(x. j. Indeed. the function a φ(x. Then 0= b a b a φ(x. Choose one fixed such n. Let S be the square [a. b]. Example. t)|2 dx dt = M 2 < ∞. v u = (v ⊗ u)f φ(x. t) is called the kernel of the integral operator K). uk (x)ul (t) = = b a b a b a ui (x)uj (t)uk (x)ul (t) dx dt b a ui (x)uk (x) dx uj (t)ul (t) dt = 0 unless i = k and j = l. b This shows that. t)ui (x)uj (t) dx dt = b a b a φ(x. To show that it is a basis. j.

Now. Lemma 4.] Using the fact that K is compact. δ Kxnj − y < 8C . Let K be compact and suppose that T K − Tn K → 0. Tn ≤ C for some constant C. t) − kn (x. (T − Tnj )y < 4 . the above can be rephrased as : If K ∈ K(H) and Tn x − T x → 0 for all x ∈ H then Tn K − T K → 0. where T ∈ B(H). Since (Tn ) is a bounded sequence. Then for all j δ < (T K − Tnj K)xnj ≤ (T − Tnj )(Kxnj − y) + (T − Tnj )y . t) = i. for j > max[n. Therefore.3 Let K be a compact operator on H and suppose (Tn ) is a bounded sequence in B(H) such that. Then for all x ∈ H. t)f (t) dt . Note that K − Kn is an integral operator (of the same type as K) with kernel k(x. we have that Kn is a finite rank operator (of rank at most n2 ). Thus K − Kn 2 ≤ b a b a |k(x. In words : multiplying by a compact operator on the right converts a pointwise convergent sequence of operators into a norm convergent one. we can find a subsequence (xnj ) of (xni ) such that (Kxnj ) is convergent. for each x ∈ H the sequence (Tn x) converges to T x. note that k ∈ L2 (S). . t). Then (Tn K) converges to T K in norm.j=1 αij ui (x)uj (t) and b a (Kn f )(x) = kn (x. t)|2 dx dt = k − kn 2 L2 (S) and the right hand side → 0. Let the limit of this sequence be y. Briefly. using the convergence of (Kxnj ) to y. and this contradiction shows that the supposition that T K − Tn K → 0 is false. by Theorem 3.j=1 αij ui (x)uj (t) . m] the right hand side of the δ displayed inequality is less than 2 . Then. [That this can be done follows directly from the definition of the norm of an operator. Proof.3 (iii) it has a fourier expansion using the basis (ui uj ) of the type ∞ k(x.24 Operators on Hilbert space Returning to the operator K. Then there exists some δ > 0 and a subsequence (Tni K) such that T K − Tni K > δ. writing kn (x. δ for nj > m. Therefore Theorem 4.2 shows that K is compact. t) = n i. Choose unit vectors (xni ) of H such that (T K − Tni K)xni > δ. for nj > n. T x = limn Tn x ≤ C x and so T ≤ C. there exists m so that. Thus. t) − kn (x. there exists n so that. Also. using the convergence of (Tnj ) to T .

for all x ∈ H. Let xi be an orthonormal basis of H.3 (Pn K) converges to K in norm.] From Theorem 3. Also. Proof. but we shall only prove it for the case when the space is separable. Pn x converges to x i=1 (that is. Pn K is of finite rank and. Pn converges pointwise to the identity operator I). Now. · · · . Define Pn by n Pn h = i=1 h.3(iii). xn . xi xi .John Erdos 25 The theorem below is true for all Hilbert spaces. . if K is any compact operator. x2 .4 Every compact operator on H is a norm limit of a sequence of finite rank operators. Theorem 4. from Theorem 4. [Note that Pn is the projection onto span x1 . Pn could be written as Pn h = n xi ⊗ xi .

xn xn is bounded with A ≤ sup |αn | . Given that K ∗ K is compact.] Apply this. Show that. Hence prove that. and show that no subsequence of (Kxn ) can converge to a non-zero vector. prove that K is compact. conversely if lim (αn ) = 0 then A is not compact. [Use the ideas in the proof of Lemma 4. Let K be a compact operator. limn→∞ An x. prove that (Kxn ) is a Cauchy sequence. y ∈ H. y = 0. Let An be a bounded sequence in B(H) such that. 3.] 4. xn . Axm − Axn n→∞ 2 = |αm |2 + |αn |2 . Prove that. with z = K ∗ y for any y.26 Operators on Hilbert space Exercises 4 1. for all x. [A result of the course states that | xn .3. Show that T 2 = 0 and that T is not compact. prove that for any orthonormal sequence {xn }. when m = n. n→∞ lim KAn K = 0 . z → 0. Let T be the operator on l2 ⊕ l2 defined by T (x. 2. n n→∞ Hence prove that if lim (αn ) = 0 then A is compact. 5. x). Let (xn ) be an orthonormal sequence in a Hilbert space H and let (αn ) be a bounded sequence of complex numbers. (Kxn ) → 0 as n → ∞ Hints: Observe that. y) = (0. Prove that the operator A defined by ∞ Ax = n=1 αn x. [Hint: if (K ∗ Kxn ) is convergent. z |2 is convergent. Using the hints below.] . for any compact operator K. for any vector z.

Since x ≤1 x ≤1 y ≤1 for all x and. Tx ≥ c x . since if | Ax. xn − xm ≤ 1 c 1 c T (xn − xm ) = yn − ym ) so it follows easily that (xn ) is a Cauchy sequence and so convergent to some element x. Lemma 5.2 If T is an in the lemma. For the two classes of operators that we shall be concerned with (compact operators and selfadjoint operators) the required facts about the spectrum will be established by simple methods. y is real. Proof.3 A = sup | Ax. since T is continuous. Ax = A∗ x. y = lim yn = lim T xn = lim T x. is a positive constant. y | . x ≤1 2 Proof. from the A = sup Ax = sup sup | Ax. Note that every eigenvalue of an operator T is in the spectrum of T . y | = eiθ Ax. A will denote a selfadjoint operator. Now. Then. is called the resolvent set of T and is denoted by ρ(T ). x . x | . Let (yn ) be a convergent sequence of elements of ran(T ) converging to y. y . We shall not need this general fact in this course. if the K is a compact operator on an infinite-dimensional Hilbert space then 0 ∈ σ(K) (this merely repeats the fact that K does not have an inverse). We now derive some simple properties of the spectrum of a selfadjoint operator. The spectrum of T is denoted by σ(T ). Lemma 5. it is a fact that σ(T ) is a non-empty compact subset of C. Then | Ax. T n x ≥ cn x for all x ∈ H. x | ≤ k x Cauchy-Schwatrz in equality. as required. x = x. Definition. For any element T of B(H).1 Let T be an operator such that for all x ∈ H. y | ≤ k whenever x ≤ 1 and y ≤ 1. x = Ax. For the rest of this section. we need to show that | Ax. k ≤ A . Then yn = T xn for some sequence (xn ) and we need to show that y = T x for some x. The complement (in C) of σ(T ). Since (yn ) is convergent it is a Cauchy sequence. z |. Recall that Ax. Let k = sup z ≤1 | Az. x is real for all x since Ax. Corollary 5. where c Proof. Also. The spectrum of an operator T is the set of all complex numbers λ such that λI − T has no inverse in B(H).John Erdos 27 5 The Spectrum. Then the range of T is closed. to show that A ≤ k. the set of all complex numbers λ such that λI − T has an inverse in B(H). the range of T n is closed for each positive integer n. It is sufficient to prove this when Ax. that is.

However. M ]. Note that. (x + iy) − A(x − iy). M ] . (ii) m ∈ σ(A) and M ∈ σ(A) .e. x . x =1 Corollary 5. Thus. This is done by a single calculation in the body of the proof. (λI − A)x = sup | (λI − A)x.4 For all T ∈ B(H) T ∗T = T Proof. x | = sup T x x ≤1 x ≤1 2 2 . x | = sup | Ax. x ≥ λ − M y ≤1 so that (dividing by x ) we have (λI − A)x ≥ (λ − M ) x for all x. A similar proof holds when λ < m . 4 Ax. = T 2 . using the above result for X = 1 (A − ξI).4). x x =1 and M = sup Ax. η = 0). if X is selfadjoint then (iI − X)x 2 = = = (iI − X)x. proves the general result. y where x = eiθ x. then. y A(x + y). If λ is real with λ > M . (x − iy) ] ≤ k{ x + y 2 + x − y 2 } = k(2 x 2 + 2 y 2 ) ≤ 4k . Since T ∗ T is selfadjoint.3) and the paralellogram law (Lemma 1. x Xx 2 + x 2 ≥ x 2 .28 Operators on Hilbert space then applying the result for the real case for Ax . T ∗ T = sup | T ∗ T x. = (the expression in square brackets being zero since Ax. x |. y | ≥ (λI − A)x. y is real). (x + y) − A(x − y). Xx + i Xx.5 (i) σ(A) ⊆ [m. Note that sup | Ax. we have that for x = 1. (iI − X)x x 2 + Xx 2 − i x. we have η (λI − A)x = η(iI − X)x ≥ |η| x . (x − y) +i[ A(x + iy). We write x ≤1 x =1 m = inf Ax. . Theorem 5. using the polarization identity (Lemma 2. if λ = ξ + iη is not real (i. Now. it can also be established by a sequence of simpler proofs as follows. The key to the next result is proving that (λI −A)x ≥ c x whenever λ ∈ [m.

z = x. (M − m)I − (M I − A) = A − mI has no inverse. It follows that (λI − A)x ≥ d x [apply the above for x ]. . Applying the result just proved to the operator M I − A shows that M − m ∈ σ(M I − A). for any y ∈ H. X(A − M I)xn → 0] and so M ∈ σ(A). Hence m ∈ σ(A). M ]). noting that d = dist(λ. xn → 0 . 1 = xn = X(A − M I)xn ≤ X . [m. from above. (αI − A)x = 0 and (λI − A)x 2 = [λI − αI + (αI − A)]x 2 = [λI − αI + (αI − A)]x. by Lemma 5. Thus we have shown that the spectrum of a selfadjoint operator is non-empty and real and its norm is equal to its spectral radius. [m. x = x. Then (αI − A)x. A is either M or −m. For m.e. x . Hence A − M I has no inverse in B(H) [since if X were such an operator. it is continuous). Further. Therefore. Hence λI − A is x injective and. x . (λI − A)z for all x and so (λI − A)z = 0. x = M − m = M I − A x =1 since inf x =1 (M I − A)x. xn ) → M . (ii) From Lemma 5. proving (i). (being both dense and we have λI closed). But this ¯ is impossible. it has closed range. applied to −A. x = 0. there is a unique x ∈ H such that y = (λI − A)x. if 0 = z ⊥ ran(λI − A) ¯ ¯ then 0 = (λI − A)x. Therefore. that is. Then y ≥ d x so (λI − A)−1 y = x ≤ 1 d y showing that (λI − A)−1 ∈ B(H) (i.John Erdos 29 Proof. If A = M = sup x =1 Ax. ¯ − A)z ≥ d z . (i) Suppose λ ∈ [m. Thus λ ∈ σ(A).1. ν(T ). ran(λI − A) = H. Then (A − M I)xn 2 = Axn 2 + M 2 − 2M Axn . M ]) = dist(λ. Let x ∈ H be any unit vector and write α = Ax. [λI − αI + (αI − A)]x = |λ − α|2 x 2 + (¯ − λ) (αI − A)x. [m.3. (if A = −m the same proofs. M ] and let d = dist(λ. (αI − A)x + (αI − A)x 2 ≥ |λ − α|2 ≥ d2 . x α ¯ +(α − λ) x. hold) there exists a sequence (xn ) of unit vectors such that ( Axn . of an operator T is defined to be ν(T ) = sup{|λ| : λ ∈ σ(T )} . The spectral radius. note that sup (M I − A)x. Define (λI − A)−1 y = x. xn ≤ 2M 2 − 2M Axn . since. M ]).

T ∈ B(H) and suppose that X is invertible. Prove that σ(T ) = σ(X −1 T X). Let A ∈ B(H) be a selfadjoint operator. 2. Show that U = (A − iI)(A + iI)−1 is a unitary operator. Let X. .30 Operators on Hilbert space Exercises 5 1.

1 Suppose this is false. Also (λI − K)Hn = Hn+1 H0 ⊇ H1 ⊇ H2 ⊇ H3 ⊇ · · · .1 that (Hn ) is a sequence of closed subspaces. Then the inequality fails for c = k for k = 1. Applying the condition that K is compact. if y ∈ Hn then Ky = ((K − λI)y + λy) ∈ Hn so that K(Hn ) ⊆ Hn . We show that (Kxn ) cannot have a Cauchy subsequence. Using Lemma 1. The proof of this is in several stages. Let Hn = ran(λI − K)n and write H0 = H. i→∞ λ This contradicts the fact that λ is not an eigenvalue. It follows from (a) using Lemma 5. We now use the compactness of K to show that the inclusion Hn ⊆ Hn+1 is not always proper. if m>n Kxn − Kxm = (K − λI)xn + λxn − Kxm = λxn + [(K − λI)xn − Kxm ] = λxn + z where z ∈ Hn+1 [Kxm ∈ Hm ⊆ Hn+1 follows from m > n]. so (a) is established. Call its limit y. we have that that (λI − K)x ≥ c x for all x ∈ H. Since (xki ) is a sequence of unit vectors. In this section K will always denote a compact operator. (b) ran(λI − K) = H. Theorem 6. Then xki = 1 ((λI − K)xki + Kxki ) λ y and so (xki ) → λ . Suppose. Proof. y = 0. Therefore there is a sequence of unit vectors such that (λI − K)xk ≤ 1 . Suppose that λ = 0 is not an eigenvalue of K. y (λI − K)y = lim (λI − K)xki = λ − y = 0 .7. on the contrary that H0 ⊃ H1 ⊃ H2 ⊃ H3 ⊃ · · · . (a) For some c > 0. We show that λ ∈ ρ(K). Indeed. for each n we can find a unit vector xn such that xn ∈ Hn and xn ⊥ Hn+1 . there is a subsequence (xki ) such that (Kxki ) is convergent. Thus Kxn − Kxm 2 = |λ|2 + z 2 ≥ |λ|2 .John Erdos 31 6 The Spectral analysis of compact operators. ((λI − K)xk ) → 0.1 If λ = 0 then either λ is an eigenvalue of K or λ ∈ ρ(K). k that is. 2. But then. · · ·. Note that.

the inclusion is not always proper. Now x ∈ Hk so x − z = 0 and (λI − K)(x − z) = 0 contradicting the fact that λ is not an eigenvalue. we show that {λ ∈ C : |λ| > δ} ∩ σ(K) is finite for any δ > 0. x2 . · · · .2 If {xn } are eigenvectors of K corresponding to different eigenvalues {xn }. xk is linearly dependent.32 Operators on Hilbert space and so (Kxn ) has no convergent Cauchy subsequence. Lemma 6. The only possible point of accumulation of σ(K) is 0.4) to any linearly independent sequence]. we can find an orthonormal sequence (xn ) of elements of N [apply the Gram-Schmidt process (Theorem 3. there is a unique x ∈ H such that y = (λI − K)x. (c) Completing the proof. This is done exactly as in Theorem 5. Thus λ ∈ σ(K). by hypothesis Kxi = λi xi with the λi ’s all i=1 different. If N is not finite-dimensional. Then we have vectors xi with Kxi = λi xi . Let Hn = span{x1 . Proof. that is ran(λI − K) = H1 = H0 = H. we have the proper inclusions H1 ⊂ H2 ⊂ H3 ⊂ H4 ⊂ · · · . Suppose the statement is false and k is the first integer such that x1 . (λI − K)x = (λI − K)k+1 y = (λI − K)z where z = (λI − K)k y ∈ Hk . Therefore k = 0.5 (i). To show that σ(K) has no points of accumulation other than (possibly) 0. for some y. Now xk = k−1 βi xi (where βi = −αi /αk ) and so i=1 k−1 0 = (λk I − K)xk = i=1 (λk − λi )βi xi showing that x1 . since K is compact. Also. xn }. If k = 0 then choose x ∈ Hk−1 \ Hk . · · · . · · · . For any y ∈ H. xk−1 is linearly dependent. Let k be the smallest integer such that Hk = Hk+1 . Then k αi xi = 0 and αk = 0. Then. Then Kxn − Kxm 2 = λxn − λxm 2 = 2|λ|2 which is impossible. contradicting the definition of k. since {xn } is a linearly independent set. Theorem 6. x2 . Then (λI − K)x ∈ Hk = Hk+1 and so. This is exactly as in an elementary linear algebra course. Suppose this is false and there is a sequence of distinct eigenvalues (λi ) with |λi | > δ for all i. Let λ be any non-zero eigenvalue and let N = {x : Kx = λx} be the eigenspace of λ. x2 . then {xn } is a linearly independent set.3 σ(K)\{0} consists of eigenvalues with finite-dimensional eigenspaces. Therefore. Proof. .e. Define (λI − K)−1 y = x. Then y ≥ c x so (λI − K)−1 y = x ≤ 1 y c showing that (λI − K)−1 ∈ B(H) (i. it is continuous).

1 a sequence of unit vectors (yn ) with yn ∈ Hn and yn ⊥ Hn−1 . Therefore. Proof. This is immediate from Theorem 5. This is immediate from Theorem 5. Choose. since yn ⊥ Hn−1 . For the rest of this section A will denote a compact selfadjoint operator. x = x. as in Theorem 6. x = Ax.5 If A is a compact selfadjoint operator then A equals its eigenvalue of largest modulus. t)f (t) dt = 0 has a finite number of linearly independent solutions. either (I − µK)−1 exists or the equation (I − µK)x = 0 b a has a finite number of linearly independent solutions. OR the associated homogeneous equation f (x) − µ b a k(x. (Fredholm formulated this result for the specific operator (Kf )(x) = In fact. t)f (t) dt = g(x) has a unique solution. he said : EITHER the integral equation f (x) − µ b a k(x. The Fredholm alternative. If > 0 is given then. [The set of all eigenvalues is (possibly) 0 together with the countable union 1 of the finite sets of eigenvalues > n . x = λ . · · ·). Corollary 6. Then. Kyn − Kym = λn yn − [(λn I − K)yn − Kym ] . where x is a unit eigenvector.) We now turn to compact selfadjoint operators. k(x. ¯ λ = Ax. Note that every eigenvalue of of A is real. Since (λn I − K)yn ∈ Hn−1 and Kym ∈ Hm ⊆ Hn−1 . 2.John Erdos 33 It is easy to see that K(Hn ) ⊆ Hn and (λn I − K)Hn ⊆ Hn−1 . |λ| ≥ is finite.6 Distinct eigenspaces of A are mutually orthogonal. . we have that |λi | < for all but a finite number of values of i. Hence (λi ) → 0. ] Corollary 6. the vector in square brackets is in Hn−1 . since λ : λ an eigenvalue of K. Lemma 6. Proof. (n = 1.4 The eigenvalues of K are countable and whenever they are put into a sequence (λi ) we have that lim→∞ λi = 0. but can be proved much more simply since if Ax = λx. t)f (t) dt . for n > m.5. Kyn − Kym > |λn | > δ showing that (Kyn ) has no convergent subsequence.5 (ii). For any scalar µ. Ax = A∗ x.

Let M = {z : z ⊥ xn for all n}. Theorem 6.34 Operators on Hilbert space Proof.3 (iii). Proof. Clearly this operator is selfadjoint [ Ax. for all h. if y ∈ M we have that xn . λ x. it is zero. Theorem 6. It follows from Lemma 6. Therefore A with its domain restricted to M is a compact selfadjoint operator on the Hilbert space M .5. y = x. Proof. Let (xn ) be the basis found in the Theorem and let λn = Axn . Let x and y be eigenvectors corresponding to distinct eigenvalues λ and µ.··· be the sequence of all the non-zero eigenvalues of A and let Ni be the eigenspace of λi .7 If A is a compact selfadjoint operator on a Hilbert space H then H has an orthonormal basis consisting of eigenvectors of A. xn xn . Then.6 that this sequence is orthonormal. by Corollary 6. y = x. A∗ y = x. . Let (λi )i=1. xn xn . Let (xn ) be the union of all these. Also it cannot be have a no-zero eigenvector [for then M ∩ Nk = (0) for some k > 0]. ∞ Ah = n=1 λn h.2. Ay = Axn . xn xn . y = 0. Ay for all x.9 If A is a compact selfadjoint operator on a Hilbert space H then there is an orthonormal basis {xn } of H such that ∞ A= n=1 λn (xn ⊗ xn ) where the series is convergent in norm. for any h ∈ H.8 Then there is an orthonormal basis {xn } of H such that. Therefore. xn xn . y = 0 and so A(M ) ⊆ M . xn (this is merely re-labeling the eigenvalues. Ay = µ x. ∞ h= n=1 h. y ∈ H so certainly for all x. But then M ⊆ N0 . Then. y ¯ (since µ is real) and so x. Therefore (xn ) is a basis. Acting on this by A. Proof. y = λn xn . Let {xn } be the basis found as above so that Axn = λn xn and ∞ Ah = n=1 λn h. Take an orthonormal basis of each Ni and an orthonormal basis of N0 = ker A. y = µ x. y = Ax. put into a sequence. since A is continuous and Axn = λn xn we have that ∞ Ah = n=1 λn h. But also M ⊥ N0 and so M = (0). The from Theorem 3. y ∈ M ]. Corollary 6.

Let k 35 Ak = n=1 λn (xn ⊗ xn ) . Theorem 4. xn |2 2 n≥k+1 n≥k+1 sup |λn |2 sup |λn | h . Now (A − Ak )h = n=k+1 ∞ λn h. xn xn . xn |2 | h. since {xn } is a basis. and. Since Ak = APk . Theorem 4. We need to show that A − Ak → 0 as k → ∞. n=1 Then.4 shows that (Ak ) converges to A in norm.3 (v) (A − Ak )h 2 ∞ = n=k+1 |λn h.4 may be used to prove the above result. xn |2 ∞ n=k+1 ∞ n=1 2 ≤ ≤ = n≥k+1 sup |λn |2 | h.3 (iii) shows that (Pk ) converges pointwise to the identity operator I. Let {xn } and Ak and A be as above and let k Pk = (xn ⊗ xn ) .John Erdos Note that (λn ) → 0. using Theorem 3. we have that A − Ak → 0 as k → ∞. and so since (λn ) → 0. Thus (A − Ak ) ≤ supn≥k+1 |λn |. . Alternatively. Theorem 3.

Define i=1 Tn h = n αi T xi and show that i=1  (T − Tn )h 2 ∞ n+1   ≤ |αi |. . ∞ T xn 2 converges then n=1 ∞ T xn n=1 2 ∞ = n=1 T T xn . Prove that n=1 (i) T is compact.4 and the fact that the norm of the compact selfadjoint operator V ∗ V is given by its largest eigenvalue.  ∞  T xi 2  . T xi  ≤ h 2 .] 2. Now use the spectral theorem for T ∗ T to prove that if for any orthonormal basis {xn }. 1] by (V f )(x) = x 0 f (t) dt . the order of summation may be interchanged. xi . Note : an operator satisfying the above is called a Hilbert-Schmidt operator. [You may assume that any vector in the range of V ∗ V (being in the range of two integrations) is twice differentiable (almost everywhere). Hints: (i) write h ∈ H as a Fourier series. Find the norm of the compact operator V defined on L2 [0. [Hint : let N = ker(λI − K) and let M = N ⊥ . 3. φk = T φk 2 ≥ 0. show that y = limn→∞ (λI − K)zn with zn ∈ M . Show that λI − K has closed range. h = ∞ αi xi where αi = h. If y ∈ ran(λI − K). Now use the result of Exercises 2 Question 6 to show that if f satisfies V ∗ V f = λf then it satisfies λf + f = 0 f (1) = 0. n+1 (ii) Take an orthonormal basis φk of H consisting of eigenvectors of the compact operator T ∗ T .] Note that a direct approach to evaluating V seems to be very difficult (try it !). ∞ 2 (ii) converges for every orthonormal basis {yn } of H and for the sum n=1 T yn is the same for every orthonormal basis. xn = k=1 ∗ ∞ µk . Let K be a compact operator on a Hilbert space H and let λ = 0 be an eigenvalue of K. Note that for a double infinite series with all terms positive. Observe that if T ∗ T φk = µk φk then µk = T ∗ T φk . f (0) = 0.36 Operators on Hilbert space Exercises 6 1. Now imitate the proof for the case when λ is not an eigenvalue. Let {xn } be an orthonormal basis of H and suppose that T ∈ B(H) is such that the series ∞ T xn 2 converges. Hints: Use Corollary 5.

This is quite a reasonable assumption. However. for a bounded operator. as the integrals on the right of each line are 0. z + y. subject to (†) does not have a unique solution [an arbitrary multiple of the eigenfunction corresponding to λ = 0 could be added to any solution to obtain another solution]. Specifically we need L to act on functions that are twice differentiable and whose second derivatives are in L2 [a. β1 and β2 are real and α1 α2 = 0. The following calculation is of interest because it shows that L satisfies a symmetry condition that would.John Erdos 37 7 The Sturm-Liouville problem. for scalars λ and corresponding functions f that satisfy (†) and the equation Lf = λf . We make the additional assumption that λ = 0 is not an eigenvalue of the system. it will not be used in the sequel. α2 ) when y and z satisfy the boundary conditions (†). η) = (α1 . −Ly. α2 . by saying that “everything makes sense”. b]. b]. We shall assume that p(x) > 0 and q(x) real for a ≤ x ≤ b. broadly speaking. Indeed. So z (a)y(a)− z (a)y (a) = 0 and similarly z (b)y(b)− z (b)y (b) = 0. that is. Lz . since if it fails then the problem (*). dy d p(x) + q(x)y dx dx In this section we shall discuss the differential operator Ly = acting on functions y defined on a closed bounded interval [a. β1 β2 = 0. which has the non-trivial solution (ξ. We also need to have that p is differentiable with p continuous on [a. b]. ¯ z ξy(a) + ηy (a) = 0 . dt dt dt dt dt dt dt a a b (†) and. . then Ly. Note. Lz = = b d d¯(t) z d dy p(t) y− p(t) z dt + ¯ (−qy¯ + yq¯) dt z z dt dt dt dt a a b b d¯ z dy dy d¯ d¯ dy z z p(t) y(t) − p(t) z (t) + ¯ p(t) − . z = y. We make further assumptions that may be summarized. We shall be looking for eigenvalues and eigenfunctions of L that satisfy the conditions (†). Therefore ¯ ¯ ¯ ¯ Ly. If L is restricted to act on the set of functions that satisfy the boundary conditions. ¯ ¯ ¯ ¯ But z (a)y(a) − z (a)y (a) is the determinant of the 2 × 2 system ¯ ¯ ξ z (a) + η¯ (a) = 0 . z = y. make it self-adjoint. Lz . this will vanish if p(b) z (b)y(b) − z (b)y (b) = p(a) z (a)y(a) − z (a)y (a) . We shall be concerned with solving the problem Ly = d dy p(x) + q(x)y = f (x) dx dx (∗) subject to boundary conditions α1 y(a) + α2 y (a) = 0 β1 y(b) + β2 y (b) = 0 Where α1 .

u(a) = −α2 .) Under the assumptions stated above.v (t) − l. (also from Questions 1. m are constants to be chosen.2 and 3 of Exercises 6) we have that there is a unique function u such that Lu = 0. dt so ∆ is a constant (i. b]. t) to have a unit discontinuity at x = t. t) = u(x). t) f (t) dt where k(x. Solving for l.v(x) ∆ a ≤ x ≤ t. u (a) = α1 . ∂x k(x. ξ. the problem (*).38 Operators on Hilbert space Theorem 7. . α1 y(a) + α2 y (a) = 0 is a scalar multiple of u. at some point t0 ξ.u + η. Let k(x. b] × [a. Also.] Choose l. t) is real-valued and continuous on the square [a. contradicting the linear independence of u and v. Then y = ξ.u+η. t) = l u(x) m v(x) a≤x≤t t≤x≤b where (for fixed t) l.1 (Existence of the Green’s function. [Our motivational work suggests that ∂ we require k(x. subject to (†) has the solution y(x) = b a k(x. It follows easily that every solution of Ly = 0. m gives l= v(t) ∆ m= u(t) ∆ where ∆ = p(t)(v u − u v) = pJ(u. t) to be continuous and p(x).v(t0 ) = 0.v (t0 ) = 0 has a non-trivial solution (ξ. p(t)[m. Proof. also independent of t). d∆ = u(pv ) + u (pv ) − v(pu ) − v (pu ) = −quv + vqu = 0 . independently of the theory of Jacobeans. v(b) = −β2 . η). The assumption that 0 is not an eigenvalue implies that u and v are linearly independent [if u were a multiple of v then it would be an eigenfunction]. m such that m. Similarly we have a unique v such that Lv = 0.u (t0 ) + η.u(t) = 0.u (t)] = 1. From the elementary theory of the initial value problem for linear differential equations. y(t0 ) = y (t0 ) = 0 and so ξ. t ≤ x ≤ b.] Hence we have that k(x. [One can see. v).e. where J is the Jacobean and hence non-zero [since u and v are independent].v is a solution of Ly = 0. that ∆ = 0 since otherwise. v (b) = β1 .v(t) ∆ u(t).v is identically 0.u(t0 ) + η.v(t) − l.

Similarly β1 y(b) + β2 y (b) = 0. For notational convenience we substitute ∆ y(x) (remembering that ∆ is constant). we just verify directly that 39 y(x) = b a k(x. Therefore x a (p(x)(∆ y(x)) ) + q∆y = [(pv ) + qv] u(t) f (t) dt b x +[(pu ) + qu] = f.v(t) f (t) dt and.∆ v(t) f (t) dt + pf (v u − u v) since u and v are solutions of Ly = 0. Therefore α1 y(a) + α2 y (a) = 0 since u satisfies the boundary condition at x = a. First note that when x = a we have x ≤ t throughout the range of integration and so y(a) = 1 ∆ b a u(a). since y (x) = b a ∂k(x. v(x) can be taken outside the integration. t) f (t) dt = v(x) x a u(t) f (t) dt + u(x) v(t) f (t) dt = v(x)u(x)f (x) + v (x) u(t) f (t) dt b x −u(x)v(x)f (x) + u (x) (p(x)(∆ y(x)) ) = (pv ) x a v(t) f (t) dt b x u(t) f (t) dt + pv uf + (pu ) v(t) f (t) dt − pu vf . Since u(x). t) f (t) dt is the required solution. we obtain b a x a b x ∆ y(x) = ∆ (∆ y(x)) k(x. t) f (t) dt. ∂x y (a) = 1 ∆ b a u (a). .John Erdos To complete the proof. We now substitute into the equation.v(t) f (t) dt .

(†) for some right hand side. We use these notations and observations in the statements a proofs below. 3. from Theorem 7. Define the operator K by (Kf )(x) = b a k(x.) Also. x for the most general f . that for every f ∈ L2 [a. Thus LK = I. that is KL acts like the identity on D. b] Kf ∈ D. has a countable sequence (µi ) of real eigenvalues such that (|µi |) → ∞ . Theorem 7. b] × [a. 1 (ii) λ is an eigenvalue of K if an only if µ = λ is an eigenvalue of the system (*). |k|2 < ∞ and so. Informally D is the all y ∈ L2 [a.(y) has a derivative almost everywhere such that (p. as shown in Section Let D be the set of functions y such that Ly exists as a function in L2 [a. K is compact. Note that L fails to be an inverse of K since it is not defined on the whole of L2 [a. t) f (t) dt . 2. has eigenfunctions which form an orthonormal basis of L2 [a.(†). Proof. b] and satisfies the boundary conditions (†). the system (*).1. b]. it cannot be invertible. b] that qualify as solutions of (*). Naturally. b]. A more precise statement is: y ∈ D (which is an equivalence class of functions) if there is a representative y such that y is differentiable and p. has finite-dimensional eigenspaces.1 shows that Kf = K(Ly) = y. L is defined on a dense subset. the Hilbert space in question. (†). b] it is clear that 4. . LKf = f . it follows from the proof of Theorem 7.2 (i) The operator K does not have λ = 0 as an eigenvalue. Consequently.) If y ∈ D and f = Ly then Theorem 7. (There is a little technical hand waving here.(†) 1. b]. In the other order. the differentiation of expressions like a u(t) f (t) dt one requires the relevant background from Lebesgue integration. Indeed. (i) If f = 0 the solution of Ly = f is Kf and cannot be y = 0.1. However.(y) ) ∈ L2 [a.40 Operators on Hilbert space We can now apply the results of Section 6 to draw conclusions about the eigenfunctions and eigenvalues of the Sturm-Liouville system (*). Since k is continuous on [a. (The verification that Kf is a solution of Ly = f explicitly shows this. since K is compact. Therefore 0 is not an eigenvalue of K.

y(a) = 0. Note. The consequences are immediate deductions from the results of Section 6 (principally 6. infinite-dimensional) subspace. Let u be the (unique) solution of Lc y = dy d p(t) dt dt + [c + q(t)]y = 0. [The trigonometric functions are actually the eigenfunctions of y = 0 subject to periodic boundary conditions y(−π) = y(π).4 and 6.John Erdos 41 (ii) If λ is an eigenvalue of K the Kφ = λφ and since φ is in the range of K. π]. But pu is strictly positive at a so it is strictly positive in (a.7). π]. b]. 3. π]. Similar. · · ·}. y (−π) = y (π). Conversely. Then u must vanish between a and z. 3. y (0) = y (π) = 0 and so also form an orthonormal basis of L2 [0. 2. “Foundations of modern analysis”. = 0. Similarly the functions {cos nt : n = 0. λ is an eigenvalue of K. Then KLφ = φ = µKφ and so Kφ = λφ where λ = 1 µ 1 φ = µφ. it follows that u is strictly positive in some interval (a. y ∈ D. For example.] The eigenfunctions of the system y = 0. from Theorem 7. For any constant c. Since u (a) > 0 and u(a) = 0. We then have Lφ = µφ and φ ∈ D. z). z) contradicting that u vanishes between a and z. Suppose z is the smallest zero of u that is > a (if any). and such systems are not covered by our work. this is what justifies the expansions that are required in solving partial differential equations by the method of separation of variables. from the discussion above. z) and so pu is increasing in (a. Note that in this case the set of eigenvalues of K cannot be finite because this would imply (by Corollary 6. y(a) = 0 is a multiple of u so to show that 0 is not an eigenvalue we must show that u(b) = 0. Section 7).3. If z ≤ b then (pu ) = −[c + q]u is positive in (a. The trigonometric functions form an orthonormal basis of L2 [−π.2 these form an orthonormal basis of L2 [0.(†). (It is true that . The assumption that 0 is not an eigenvalue of the system is not an essential restriction. if µ is an eigenvalue of (*). 2. φ ∈ D. Any solution of Lc y = 0. a + δ). for example. 6. It is a fact that. by assumption µ = 0. since. assume c + q(t) < 0 for a ≤ t ≤ b. choose c so that c + q does not change sign in [a. y(0) = y(π) = 0 are the functions {sin nt : n = 1. y (a) = 1 . Therefore. if the boundary conditions are y(a) = y(b) = 0. 1. · · ·} are the eigenfunctions of the system y . however the device below gives us the result. This fact can be deduced from the work of the present section.(†).8) that K vanishes on a non-zero (in fact. Hence 0 is not an eigenvalue of Lc y = 0. for definiteness. by adding a suitable constant to q we can always ensure that λ = 0 is not an eigenvalue of the system. but more complicated arguments can be used for other boundary conditions (see Dieudonne. The most important result arising from this is consequence 2. the eigenfunctions of L and L + c are the same and the eigenvalues of L + c are λ + c whenever λ is an eigenvalue L. Chapter XI. Then λLφ = LKφ = φ so that Lφ = and µ is an eigenvalue of (*).

A similar calculation with cosines shows that f is also an odd function on [−π. y (0) = y (π) = 0 for a suitable constant k – any non-integral k will do). Then a simple change of variable shows that for each integer n. but this is covered by the note above. π] is orthogonal to all the trigonometric functions.42 Operators on Hilbert space 0 is an eigenvalue of the latter system. Thus f = 0 (almost everywhere) and the fact is proved. Therefore f (t) = f (−t) (almost everywhere) for 0 ≤ t ≤ π. + ky = 0. 0= π −π f (t) sin nt dt = 0 −π f (t) sin nt dt + π 0 f (t) sin nt dt = π 0 [f (t) − f (−t)] sin nt dt . Now suppose that f ∈ L2 [−π. π]. . on can consider the system y . π]. showing that f is an even function on [−π. Alternatively.

t)| ≤ M n x a kn (x. b] × [a. . Let X be a Banach space (that is. . = y (n−1) (0) = 0 where pi and f are continuous. By putting u(x) = y (n) (x). t) ds. (n − 1)! Using this result and the formal binomial expansion of (I − K)−1 . |x − t|n−1 . . b] × [a. b]). Prove that K ≤ (b − a)M where M is a bound for k in [a. .] Prove that (†) has a unique solution in L2 [0. Prove that (K n f )(x) = Show by induction that |kn (x.] For each λ = 0. t) = tx k(x. Let k(x. Prove that an absolutely convergent series in a Banach space is convergent. By quoting appropriate theorems show that this solution is continuous (strictly. b]. b] by (Kf )(x) = x a k(x. Let kn be defined inductively by k1 = k and kn (x. . deduce from Question 1 that (I − K) has an inverse in B(H). verify by multiplication that the sum of the formal expansion is the inverse of (I − K). b] → L2 [a. 3. this problem reduces to (†) (I − K)u = f x (x−t)r−1 0 (r−1)! where K is an operator on L2 [0. (in particular this will hold if k is continuous on [a. = y (n−1) (0) = an−1 . b] × [a. . b] of the type considered in Question 6. Note that the general initial value problem  (n)  y (x) + p1 (x)y (n−1) (x) + . [Hint: prove that the sequence of partial sums is Cauchy. t)f (t) dt. that the equivalence class of this solution contains a continuous function). [Hint : show that y (n−r) (x) = u(t) dt. observe that K is an operator of the same type as K and deduce that (λ − K) has λ an inverse in B(H). for any b > 0. + pn (x)y(x) = f (x) (∗)  y(0) = y (0) = . ∞). Deduce that there is a unique continuous function with n continuous derivatives that satisfies (∗) in [0. b]. 2. Consider the initial value problem  (n)  y (x) + p1 (x)y (n−1) (x) + . [Hint : after proving absolute convergence. show that. A series n xn in X is said to be absolutely convergent if the series n xn of real numbers is converegent.] Existence theory for linear initial value problems using operator theory. y (0) = a1 . t) be bounded and square integrable over the square [a. . s)kn−1 (s. t)f (t) dt. a normed linear space that is complete). + pn (x)y(x) = f (x) (∗∗)  y(0) = a0 . . b] for each b > 0. Define K : L2 [a. .John Erdos 43 Exercises 7 1.

5. Find a Green’s function for the system y = f. y(0) = y(1) = 0 . Check your answer by verifying that it gives x(x − 1) as the solution when f = 2.44 Operators on Hilbert space can be transformed into the form (*) by changing the dependent variable from y to z where n−1 z(x) = y(x) − k=0 ak xk and thus (∗∗) also has a unique solution. 1]. and consequently find an orthonormal basis of L2 [0. Repeat the above question with the system y = f. 4. y(0) = y(1) = 0 . . y(0) + y (0) = y(1) − y (1) = 0 . Evaluate the eigenvalues and eigenfunctions of y = λy.

q(T ).2 If A is a selfadjoint operator then. Let p be any polynomial. if λ ∈ σ(T ) then p(λ) ∈ σ(p(T )). using Corollary 5. In this section A will always denote a selfadjoint operator with spectrum σ = σ(A) where σ(A) ⊆ [m. For the general case. p(x) − p(λ) = (x − λ)q(x) for some polynomial q. The functional calculus. We know from Theorem 5. Then p(T ) − kI = (T − λ1 I)(T − λ2 I)(T − λ3 I) · · · (T − λn I) . the norm equals the spectral radius.X = (T − λI). since (¯. Theorem 8. Proof. If each T − λi I has an inverse then the product of all these would be an inverse of p(T ) − kI. σ(p(T )) = {p(λ) : λ ∈ σ(T )} . M ]. for any polynomial p. p(A) = sup{|p(λ)| : λ ∈ σ(A)} Proof. That is.p(A) = sup{|¯p(λ)| : λ ∈ σ(A)} p = sup{|p(λ)|2 : λ ∈ σ(A)} = (sup{|p(λ)| : λ ∈ σ(A)})2 . Therefore T − λi I fails to have an inverse for at least one root λi of p(x) = k. M ] as defined in Section 6.John Erdos 45 8 The Spectral analysis of selfadjoint operators.p)(A) = p(A). Therefore. That is k = p(λi ) for some λi ∈ σ(T ). if k ∈ σ(p(T )) then the polynomial p(x) − k factors over the complex field into linear factors: p(x) − k = (x − λ1 )(x − λ2 )(x − λ3 ) · · · (x − λn ) .p(A) we have. x − λ is a factor of p(x) − p(λ).X . λ2 .(T − λI) = [p(T ) − p(λ)I].p(A) = p(A)∗ . [If f = g + ih is complex.X). that is. M ].] The operator f (A) is defined by f (A) = lim pn (A) . and these factors clearly commute. Therefore (T − λI) has both a left inverse (namely X. p(T ) − p(λ)I = (T − λI)q(T ) . Therefore.q(T )) and a right inverse (q(T ). If p is real. For any polynomial p.q(T ). From Weierstrass’ approximation theorem there is a sequence (pn ) of polynomials converging to f uniformly on [m. Corollary 8.[p(T ) − p(λ)I] = X. p(A) = sup{|k| : k ∈ σ(p(A))} = sup{|p(λ)| : λ ∈ σ(A)} . λ ∈ σ(T ). Where λ1 . (pn ) is found by combining sequences approximating g and h.1 (Spectral Mapping Theorem) Let T ∈ B(H). Now if p(λ) ∈ σ(p(T )) then p(T ) − p(λ)I has an inverse X and so I = X. Definition. n→∞ . p(A) is selfadjoint. Let f ∈ C[m. using the result of the theorem. λ3 · · · λn are the roots of p(x) = k. By the elementary scalar remainder theorem. An easy algebraic argument shows that these are equal and (T − λI) has an inverse. that is. Conversely.5 that for selfadjoint operators.4 p ¯ p(A) 2 = p(A)∗ .

any positive operator is selfadjoint. it follows from Theorem 5. Suppose that (pn ) and (qn ) are two sequence of polynomials converging to f uniformly on [m. M ].5 that A ≥ 0 if and only if σ(A) ⊆ R+ . it is clear that (pn (A)) is a Cauchy sequence in B(H) and so is convergent. M ].t] (A) will satisfy Et = Et = Et . Proof. Also. x ≥ 0 for all x ∈ H. Theorem 2. since X − Y is independent of n and the right hand side → 0 as n → ∞. We first show that the above definition determines an operator f (A).g(A) ¯ f (A) = f (A)∗ . Note that χ[−∞. we need to show that. Definition.t] is a real function ∗ 2 which is equal to its square. The map f → f (A) is a *-algebra homomorphism of C[m. Our hope is that Et = χ[−∞. . this implies that X = Y = f (A) and that the limit depends only on the function f . that Et is a projection. M ] into B(H) and f (A) = sup{|f (λ)| : λ ∈ σ(A)} . Then X −Y = X − pn (A) + pn (A) − qn (A) + qn (A) − Y ≤ X − pn (A) + pn (A) − qn (A) + qn (A) − Y ≤ X − pn (A) + qn (A) − Y + sup |pn (t) − qn (t)| m≤t≤M and. Let pn be a sequence of polynomials converging to f uniformly on [m. (The immediate goal is to attach a meaning to χ[−∞. if (pn ) and (qn ) are sequences of polynomials converging uniformly on [m. These follow easily from the definitions since. To demonstrate the *-algebra homomorphism statement. M ] to f and g respectively. (αf + βg)(A) = αf (A) + βg(A) (f. We wish to extend the functional calculus to limits of pointwise monotonically convergent sequences of continuous functions. To show that this defines a unique operator we must show that it is independent of the choice of the sequence. We write A ≥ B to mean A − B ≥ 0.t] (A). that is. Write lim pn (A) = X and lim qn (A) = Y .3 and one of the exercises. An operator A is said to be positive (in symbols A ≥ 0) if Ax. n→∞ n→∞ λ∈σ λ∈σ The extension of the functional calculus. Then Corollary 8. Since (pn ) is a Cauchy sequence in C[m.) We first need a definition some technical results. M ].g)(A) = f (A). Note that. from the polarization identity. M ]} = pn − pm . Finally.3 The operator f (A) ∈ B(H) and is well defined. then (αf + βg)(A) = lim (αpn + βqn )(A) = α lim pn (A) + β lim qn (A) = αf (A) + βg(A) n→∞ n→∞ n→∞ and the other statements are easily proved in the same way.46 Operators on Hilbert space Theorem 8.2 shows that pn (A) − pm (A) = sup{|pn (λ) − pm (λ)| : λ ∈ σ(A)} ≤ sup{|pn (λ) − pm (λ)| : λ ∈ [m. in the norm of C[m. M ]. that is. f (A) = n→∞ lim pn (A) = lim pn (A) = lim sup |pn (λ)| = sup |f (λ)| .

Ax Ax. (i) This is proved in exactly the same way as Theorem 1. it depends only on φ and not on the approximating sequence). . If Ax = 0. hence the alternative terminology. It is routine to show that A is linear.1 from the fact that A(λx + y). Call the limit Ax. [Note that pointwise convergence is convergence in a topology on B(H) called the “strong operator topology”. An = sup x =1 An x. we write Ax. It is a fact that φ(A) is well defined (that is. Definition. Ay . Ay. Theorem 8. y (i) (ii) Ax ≤ A Ax. B respectively (and if An is bounded) then (An . (Am − An )x. Proof. [E.John Erdos Lemma 8. Ax Ax. 47 Proof. x . M ]. for each x ∈ H. The operator φ(A) is defined as the pointwise limit of the sequence (fn (A)).Bn ) → AB pointwise.6 If the sequences (An ). y = lim An x. to show that A is selfadjoint. Note also that by a theorem of Banach space theory (the Uniform Boundedness Theorem) the condition in brackets is implied by the convergence of (An ). x . x ) is a decreasing sequence of positive real numbers and hence is convergent.5 Let (An ) be a decreasing sequence of positive operators. Note that for each n. Let φ be a positive function on [m. x = A1 . The hypothesis shows that.g.4 Let A be a positive operator. (Am − An )x 2 ≤ Am − An .4 (ii). n→∞ n→∞ Lemma 8. An y = x. (ii) The result is obvious if Ax = 0. x ] and this shows that (An ) is a Cauchy sequence. The extended functional calculus. using (i) with y = Ax we have Ax 4 = | Ax. (λx + y) ≥ 0 for all λ.[ Am x. (Bn ) converge pointwise (strongly) to an operators A.] Proof. x − An x. Then | Ax. x ≤ sup x =1 A1 x. M ] that is the pointwise limit of a decreasing sequence (fn ) of positive functions ∈ C[m. Ax |2 ≤ ≤ ≤ Ax. A2 x . If m > n then Am − An ≥ 0 and from Lemma 8. A . the sequence ( An x. A2 x. A1 . Ax 2 and the result follows on dividing by Ax 2 . x . x . Then (An ) converges pointwise (strongly) to an operator A such that 0 ≤ A ≤ An for all n. bounded. selfadjoint and 0 ≤ A ≤ An for all n. x ≤ 2. and so convergent. (Bn − B)x + (An − A)Bx and the right hand side → 0 from the pointwise convergence of (An ) and (Bn ) [at the points Bx and x respectively]. y | ≤ Ax. using Lemma normsa. y = lim x. For any x ∈ H we have (An Bn − AB)x = (An Bn − An B + An B − AB)x ≤ An . x .

M ] since if (pn ) is a sequence of polynomials converging uniformly to f on [m.7 Let φ. X.λ] . n→∞ n→∞ Now if (fn ) is a decreasing sequences of functions in C[m. lim gn (A) = φ(A).gn ) is a decreasing sequence of continuous functions converging pointwise to φ.λ] (A).X . M ] that are the pointwise limits of a decreasing sequences of positive functions ∈ C[m. Now let Eλ = χ[−∞.ψ(A). It is fact that it is (essentially) uniquely determined by A. M ] and converge pointwise [m. We call the family {Eλ : −∞ < λ < ∞} as obtained above the bounded resolution of the identity for the operator A. (ii) φ(A).φ(A) = lim X. Then (i) φ(A) + ψ(A) = (φ + ψ)(A) . Similarly.pn (A) = lim pn (A). (ii) As in (i) (fn . Also Eλ ≥ 0 and so Eλ = Eλ .48 Operators on Hilbert space Lemma 8. where the limits indicate pointwise convergence in H. M ] to φ then.6.ψ(A) = (φ. µ] . If λ < m then for all t ∈ [m. We say that the integral M m f (λ) dEλ of a function with respect to a bounded resolution of the identity exists and is equal to T if. n→∞ n→∞ n→∞ using Lemma 8. M ].fn (A) = lim fn (A). M ] that converge pointwise [m. 2 ∗ It follows that Eλ = Eλ . Proof. (i) Let (fn ) and (gn ) be decreasing sequences of functions that are continuous on [m. using Lemma 8.ψ)(A) = lim (fn .Eµ = Eν where ν = min[λ. A family of projections with these properties is called a bounded resolution of the identity.λ (t) = 0 for all sufficiently large n and so Eλ = 0.X = φ(A). ψ be a positive functions on [m. M ] we have fn. Eλ = I for λ ≥ M .ψ)(A) . Note that Eλ .λ ) is is a decreasing sequence of continuous functions converging pointwise to χ[−∞. (iii) If X commutes with A then X commutes with f (A) for every f ∈ C[m.ψ and (φ.gn )(A) = lim fn (A). (iii) If X commutes with A then X commutes with φ(A) .6.X . M ].f (A) = lim X. given any > 0 there exists a partition λ0 < λ1 < λ2 < · · · λn−1 < λn . n→∞ n→∞ For every real λ we define the operator Eλ as follows: let fn. M ] to φ and ψ respectively.λ = 1 − n(t − λ)  0  1 t<λ 1 λ≤ t≤λ+ n 1 t>λ+ n Then (fn.X = f (A). X. Then (fn +gn ) is a decreasing sequence of continuous functions converging pointwise to φ + ψ and (φ + ψ)(A) = lim (fn + gn )(A) = lim fn (A) + lim gn (A) = φ(A) + ψ(A) n→∞ n→∞ n→∞ where the limits indicate pointwise convergence in H.

m≤t≤M Since the integral is approximated arbitrarily closely in norm by these sums. From the lemma.9 M m f (λ) dEλ ≤ sup |f (t)| . ∆i h 2 2 n ≤ = 1≤i≤n 1≤i≤n max ∆i X∆i max ∆i X∆i .8 Let {∆i : 1 ≤ i ≤ n} be orthogonal projections such that I = ∆i . If X ∈ B(H) commutes with each ∆i then X = max ∆i X∆i . Lemma 8. Therefore. i=0 ∆i h 2 2 . n f (ξi )(Eλi − Eλi−1 ) = max f (ξi )(Eλi − Eλi−1 ) = max |f (ξi )| ≤ i=1 1≤i≤n 1≤i≤n sup |f (t)| . h . Let h ∈ H. j=0 ∆j h = i=0 ∆i h 2 since the cross terms are zero. since X∆i = X∆2 = ∆i X∆i = ∆i X i Xh 2 n = i=0 n ∆i Xh 2 n = i=0 ∆i X∆i . X . Lemma 8.10 For λ > µ. 1≤i≤n n i=0 ∆i and Proof. since for each i we have ∆i X∆i ≤ ∆i .∆i h 2 2 ≤ i=0 ∆i X∆i 2 . n 49 T− i=1 f (ξi )(Eλi − Eλi−1 ) < . . λi ]. µ(Eλ − Eµ ) ≤ (Eλ − Eµ )A(Eλ − Eµ ) ≤ λ(Eλ − Eµ ) . ∆i = X . for any ξ ∈ (λi−1 . the result follows. For the proof of the Spectral Theorem. Thus X ≤ max1≤i≤n ∆i X∆i . we need the following lemmas.John Erdos with λ0 < m and λn > M such that. But the opposite inequality is clear. m≤t≤M Proof.∆j = 0 when i = j. Then h 2 n = i=0 ∆i h 2 n n n = i=0 ∆i h. Corollary 8.

From Lemma 8. h = ∆i h Therefore. 2 Note that Eλ commutes with A and AEλ = AEλ . This is clear since for any h ∈ H. . Let Eλ be the resolution of the identity as found above. for then fλ. [This is obvious when t ≤ λ. We now use that (I − Eµ ) = (I − Eµ )∗ and the general fact from the start of the proof to conclude that (Eλ − Eµ )A(Eλ − Eµ ) = (I − Eµ )Eλ AEλ (I − Eµ ) ≤ (I − Eµ )λEλ (I − Eµ ) = λ(Eλ − Eµ ) . Xh ≤ T Xh.11 (The Spectral Theorem for bounded selfadjoint operators. Choose λ0 < λ1 < λ2 < · · · λn with λ0 < m. Note that when h = 1. λi ].3 ∆i A∆i − ξi ∆i ≤ max[|λi − ξi |. λ](t).n + n .] Also (gn ) is decreasing (this is an elementary but tedious exercise) and so the pointwise (strong) limit of (gn (A)) exists and is positive. (λi−1 − ξi )∆i ≤ ∆i A∆i − ξi ∆i ≤ (λi − ξi )∆i .8 we have that i=1 n 2 ≤ 1. since ∆i is an orthogonal projection. h . λ + n ) we have λ − t > − n so. Therefore. since 0 ≤ fλ. The fact that µ(Eλ − Eµ ) ≤ (Eλ − Eµ )A(Eλ − Eµ ) is proved in an exactly similar way. Let fλ. ∆i . It is easy to see that gn ≥ 0. Proof. h = SXh. for any ξi ∈ [λi−1 . applying Lemma 8.) For any bounded selfadjoint operator A there exists a bounded resolution of the identity Eλ such that A= M m λ dEλ . Hence. (A − i=1 ξi ∆i ) = max ∆i A∆i − ξi ∆i < 1≤i≤n and this is exactly what is required.50 Operators on Hilbert space Proof.n be as in the definition of Eλ and define gn (t) = 1 (λ − t)fλ.n (t) ≤ 1. Let > 0 be given. Therefore λEλ − AEλ ≥ 0 proving the claim. and also 1 1 1 when t ≥ λ + n . λn > M such that 0 ≤ λi − λi−1 < . using Theorem 5. λ](t) − tχ(−∞. But (gn (t)) is pointwise convergent to λχ(−∞.∆j = 0 when i = j. Observe that {∆i : 1 ≤ i ≤ n} satisfies the hypotheses of Lemma 8. Theorem 8. X ∗ SXh. Write ∆i = Eλi − Eλi−1 . it follows that gn (t) ≥ 0. Therefore we have that 2 Eλ AEλ = AEλ ≤ λEλ . 0 ≤ ∆i h. for t ∈ (λ.8 and that X = (A − n ξi ∆i ) commutes with each ∆i . Xh = X ∗ T Xh.10 we have that λi−1 ∆i ≤ ∆i A∆i ≤ λi ∆i . First note the general fact that if S ≤ T then for any operator X we have X ∗ SX ≤ X ∗ T X. |λi−1 − ξi |] < .n (t) = 0. Next we claim that AEλ ≤ λEλ .

by taking linear combinations. For f ∈ C[m. . choose a polynomial p such that sup |f (λ) − p(λ)| < .12 If f is continuous on [m. m≤λ≤M Then. ξ2 · · · ξn .3 f (A) − p(A) < f (A) − M m and p(λ) dEλ + M m M m f (λ) dEλ ≤ = f (A) − M m p(λ) dEλ − M m f (λ) dEλ f (A) − p(A) + p(λ) − f (λ) dEλ < 2 . Therefore λr dEλ . λ1 . . . Since f (A) = is arbitrary. choose n. Therefore. M ]. p(A) = for all polynomials p. M m and the right hand side is the approximating sum to Ar = and.John Erdos Corollary 8. k→∞ r But r Ik = n n ξi ∆i i=1 = i=1 r ξi ∆i .9. M ] then f (A) = M m 51 f (λ) dEλ . . from Theorem 8. Then limk→∞ Ik = A. the last step using Corollary 8. for any integer r. ξ1 . it follows that M m f (λ) dEλ . . Proof. p(λ) dEλ > 0. λ0 . For any integer k. given any M m M m λr dEλ . λn as in the Theorem so that n (A − i=1 ξi ∆i ) < 1 k and write Ik = n i=1 ξi ∆i . r Ar = lim Ik .

[Note that A is selfadjoint. Deduce that the only positive isometry is I. prove that ker A = (ran A)⊥ .] . λ ≥ k. since Ax. [Use Question 1. if A and B are similar then they are unitarily equivalent). Let A and B be selfadjoint operators.] 6. Deduce that any operator T can be written as a linear combination of at most 4 unitary operators. [First write T = X + iY with X.] Use the spectral theorem to find an orthogonal projection E such that U = 2E −I. x is real – make sure you know how this follows!] Deduce that if A ≥ I then An ≥ I for every positive integer n. (alternatively. For any selfadjoint operator A. [Definition: V is an isometry if V h = h for all h ∈ H.) If X ∈ B(H) show that ¯ (i) σ(X ∗ ) = {λ : λ ∈ σ(X)}. Show that if there exists an invertible operator T such that T −1 AT = B then there exists a unitary operator U such that U ∗ AU = B (that is. if you are given the result it is trivial to verify that E = 1 (U +I) 2 is a suitable E).52 Operators on Hilbert space Exercises 8 1. [Hint: use the polar decomposition of T . Let U be both selfadjoint and unitary.] 5. 1} (unless U = ±I). [Alternatively factorise An − I. 2. Deduce that every member of the spectrum of a unitary operator has modulus 1. Prove that σ(U ) = {−1. Note that a self adjoint isometry must be unitary [use Question 2].] (ii) If B and C commute and B ≥ C ≥ 0 then prove that B n ≥ C n for every positive integer n. (i) Use results on the spectrum show that A ≥ kI (where k ∈ R) if and only if for all λ ∈ σ(A). Suppose X ∈ B(H) is selfadjoint and X ≤ 1. 3.] √ 4. (ii) if X is invertible then σ(X −1 ) = {λ−1 : λ ∈ σ(X)}. Y selfadjoint. Observe that X + i I − X 2 can be defined and prove that it is unitary. [Factorise B n − C n . (More spectral mapping results.

Sign up to vote on this title
UsefulNot useful