0% found this document useful (0 votes)
50 views13 pages

Spectral Review

This document provides definitions and properties related to spectral theory of linear operators on Hilbert and Banach spaces. It defines self-adjoint, unitary, normal, and compact operators. It discusses properties of eigenvalues and eigenvectors for self-adjoint operators. It also defines resolvent sets, point spectra, and residual spectra of operators. Examples of multiplication and shift operators on function spaces are given to illustrate the definitions.

Uploaded by

German Chiappe
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
50 views13 pages

Spectral Review

This document provides definitions and properties related to spectral theory of linear operators on Hilbert and Banach spaces. It defines self-adjoint, unitary, normal, and compact operators. It discusses properties of eigenvalues and eigenvectors for self-adjoint operators. It also defines resolvent sets, point spectra, and residual spectra of operators. Examples of multiplication and shift operators on function spaces are given to illustrate the definitions.

Uploaded by

German Chiappe
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 13

Review of Spectral Theory

Definition 1 Let H be a Hilbert space and A ∈ L(H).

(a) A is called self–adjoint if A = A∗ .

(b) A is called unitary if A∗ A = AA∗ = 1l. Equivalently, A is unitary if it is bijective (i.e.


1–1 and onto) and preserves inner products.

(c) A is called normal if A∗ A = AA∗ . That is, if A commutes with its adjoint.

(d) Let X and Y be Banach spaces. A linear operator C : X → Y is said to be compact


if for each bounded sequence {xi }i∈IN ⊂ X , there is a subsequence of {Cxi }i∈IN that is
convergent.

Remark 2 Let H be a Hilbert space and A ∈ L(H). In this remark, we assume that H
is finite dimensional, so that A is multiplication by a matrix, that we also denote A.

(a) If A is self–adjoint, we have:


(i) All eigenvalues of A are real, since
0 6= x ∈ H, Ax = λx =⇒ λ hx, xi = hx, Axi = hAx, xi = λ hx, xi
(ii) Eigenvectors of A that correspond to different eigenvalues are perpendicular, since
0 6= x ∈ H, Ax = λx, 0 6= y ∈ H, Ay = µy, λ 6= µ
=⇒ λ hx, yi = hAx, yi = hx, Ayi = µ hx, yi
(iii) A is diagonalizable. That is, there is a basis of H consisting of eigenvectors of A.
See the notes “Families of Commuting Normal Matrices”.
As a consequence of (ii) and (iii), there is an orthonormal basis of H consisting of eigenvec-
tors of A. Call the basis vectors x1 , · · ·, xn and the corresponding eigenvalues λ1 , · · ·, λn .
 
Write U = x1 · · · xn (that is, the columns of U are eigenvectors of A) and denote by D
the diagonal matrix whose diagonal entries are the λj ’s. Then U is unitary (the condition
that U ∗ U = 1l is the same as the condition that the columns are orthonormal) and
AU = Ax1 · · · Axn = λx1 · · · λn xn = U D or U −1 AU = D or U ∗ AU = D
   

(b) This and similar arguments give that


A is self–adjoint ⇐⇒ H has an orthonormal basis of eigenvectors of A and
all eigenvalues of A are real
A is unitary ⇐⇒ H has an orthonormal basis of eigenvectors of A and
all eigenvalues λ of A obey |λ| = 1
A is normal ⇐⇒ H has an orthonormal basis of eigenvectors of A


c Joel Feldman. 2011. All rights reserved. October 28, 2011 Review of Spectral Theory 1
Remark 3

(a) When H is finite dimensional, A is normal if and only if there is a unitary matrix U
and a diagonal matrix D such that U ∗ AU = D. That is, A is diagonalizable by a unitary
matrix. The corresponding statement when H is infinite dimensional, is that A is normal
if and only if there is a unitary operator U such that U ∗ AU is a multiplication operator.
We shall prove this.

(b) If an operator is either self–adjoint or unitary, it is also normal.

(c) An operator A is normal if and only if it can be written in the form A = B + iC with
B and C self–adjoint and commuting. (Take B = 21 (A + A∗ ) and C = 2i 1
(A − A∗ ).)

Remark 4 In quantum mechanics, “physical observables” tend to be self–adjoint opera-


tors — energy, momentum, etc are real quantities. In quantum mechanics, time evolution
is by a unitary operator. “Total probability is preserved.”

Remark 5 When H is finite dimensional, all operators are compact. Even when H is
infinite dimensional, compact operators behave a lot like finite dimensional matrices. See
Theorem 18. They tend to be easier to work with than other operators.

Remark 6 In the finite dimensional case, we say that A is diagonalizable if there is an


invertible (but not necessarily unitary) matrix V such that V AV −1 is diagonal. Then H
has a basis of eigenvectors of A, but the basis vectors need not be mutually perpendicular.
A natural extension of this finite dimensional definition to infinite dimensions would
be that A is diagonalizable if there is a bounded linear bijection V : H → H′ = L2 (M, µ)
(then it has a bounded inverse by the inverse mapping theorem) such that V AV −1 is a
multiplication operator. Again, V need not be unitary. But we can make V unitary by
changing the inner product on H (thereby changing the meaning of “orthogonal”). Define

hx, yiH = hV x, V yiH′
′  ′ 
Then V is unitary as a map from H, h · , · iH to H′ . Here H, h · , · iH is the vector

space H equipped with the inner product h · , · iH instead of the inner product h · , · iH .
Changing the inner product in this way changes the lengths of vectors in H and also the
angles between vectors in H, but does not change the topology since
−1
kV −1 k kxkH ≤ kxk′H ≤ kV k kxkH

As A is diagonalizable by a unitary operator if and only if it is normal, we have that A


is diagonalizable if and only if A is normal with respect to some inner product on H that


c Joel Feldman. 2011. All rights reserved. October 28, 2011 Review of Spectral Theory 2
gives a topology equivalent to the original topology. (All such inner products are of the
form = hV x, V yiH for some bijection V on H that is bounded with bounded inverse. See
Corollary 35 in the notes “Review of Hilbert and Banach Spaces”.)

Definition 7 Let B be a Banach space and T ∈ L(B).

(a) The resolvent set, ρ(T ), of T is the set of all complex numbers λ such that λ1l − T is a
bijection with bounded inverse.
−1
(b) The resolvent of T at λ ∈ ρ(T ) ⊂ C is Rλ (T ) = λ1l − T .

(c) The spectrum of T is σ(T ) = C \ ρ(T ).

(d) The complex number λ is in the point spectrum, σp (T ), of T if λ1l − T is not injective.
That is, if there is a nonzero vector x ∈ H such that T x = λx. Then x is said to be an
eigenvector of T with eigenvalue λ.

(e) The complex number λ is in the residual spectrum, σr (T ), of T if λ1l − T is injective


but the range of T is not dense in B.

Remark 8 Let M be an n × n matrix. By definition, M is diagonalizable if there is an


invertible matrix V and a diagonal matrix D such that V −1 M V = D. In this case the
columns of V are eigenvectors of M that form a basis for Cn , the diagonal elements of D
are the eigenvalues of M and σ(M ) = σp (M ) is the set of all eigenvalues of M .
If M is not diagonalizable, there still exists an invertible matrix V such that V −1 M V
is in Jordan form. This means that it is of the form

J1 0 ··· ··· 0
 
 0 J2 0 ··· 0 
 . .. .. 
 .. . . 
0 ··· 0 Jm

with each diagonal block being a Jordan block. A 4 × 4 Jordan block is of the form

λ 1 0 0
 
0 λ 1 0
Bλ,4 =
0 0 λ 1

0 0 0 λ

Observe that λ1l − Bλ,4 has range (z1 , z2 , z3 , 0) z1 , z2 , z3 ∈ C which is not dense. In
this finite dimensional world, λ1l − Bλ,4 must also have a kernel (since the dimesion of the
range plus the dimension of the kernel must be the dimension of the world). That is, λ


c Joel Feldman. 2011. All rights reserved. October 28, 2011 Review of Spectral Theory 3
must be an eigenvalue. For an infinite dimensional world that is no longer the case. The
existence of residual spectrum signals the failure of diagonalizability.

Example 9

Multiplication Operators: Let


◦ (X, M, µ) be a semifinite measure space,
◦ 1 ≤ p ≤ ∞ and
◦ a : X → C be a bounded measurable function on X.
Define the bounded operator A : Lp (X, M, µ) → Lp (X, M, µ) by

(Aϕ)(x) = a(x)ϕ(x)

Then

ρ(A) = λ ∈ C ∃ ε > 0 such that |λ − a(x)| ≥ ε a.e.
 
σp (A) = λ ∈ C µ {x ∈ X | a(x) = λ} > 0
∅ if 1 ≤ p < ∞

σr (A) = 
λ ∈ C 6 ∃ ε > 0 such that |λ − a(x)| ≥ ε a.e. \ σp (A) if p = ∞

Shift Operators: Define the right and left shift operators acting on ℓ2 by

L(α1 , α2 , α3 , · · ·) = (α2 , α3 , · · ·)
R(α1 , α2 , α3 , · · ·) = (0, α1 , α2 , α3 , · · ·)

Then  
ρ(L) = λ ∈ C |λ| > 1 ρ(R) = λ ∈ C |λ| > 1

σp (L) = λ ∈ C |λ| < 1 σp (R) = ∅

σr (L) = ∅ σr (R) = λ ∈ C |λ| < 1

See the notes “Spectral Theory Examples” for derivations and other examples.

Definition 10 Let X be a Banach space and D an open subset of C. A function x : D → X


is analytic at z0 ∈ D if
0)
lim x(z)−x(z
z−z0 z→z0

exists.


c Joel Feldman. 2011. All rights reserved. October 28, 2011 Review of Spectral Theory 4
Lemma 11 Let B be a bounded linear operator on the Banach space B. Assume that B
−1
has a bounded inverse and that C is a bounded operator on B with kCk < B −1 . Then
B + C is 1–1 and onto and has a bounded inverse and furthermore
−1 2
(B + C)−1 ≤ kB −1 k (B + C)−1 − B −1 ≤ kB −1 k kCk

−1
1−kB k kCk 1−kB k kCk

Theorem 12 Let B be a Banach space and T ∈ L(B). Then

(a) ρ(T ) is an open subset of C. Furthermore, if λ ∈ ρ(T ), then



lim Rµ (T ) − Rλ (T ) = 0
µ→λ

(b) Rλ (T ) is an analytic L(B)–valued function of λ on ρ(T ).

(c) First resolvent formula: If λ, µ ∈ ρ(T ), then Rλ (T ) and Rµ (T ) commute and

Rλ (T ) − Rµ (T ) = (µ − λ)Rµ (T )Rλ (T )

Second resolvent formula: If S ∈ L(B) and λ ∈ ρ(S) ∩ ρ(T ), then

Rλ (S) − Rλ (T ) = Rλ (S) (S − T ) Rλ (T ) = Rλ (T ) (S − T ) Rλ (S)

(d) If |λ| > kT k, then λ ∈ ρ(T ).

(e) σ(T ) 6= ∅.

Lemma 13 Let B be a bounded linear operator on the Banach space B. If P (z) is a



polynomial and λ ∈ σ(B) then P (λ) ∈ σ P (B) .

Theorem 14 Let B be a Banach space and T ∈ L(B). The spectral radius of T is defined
to be
r(T ) = sup |λ|
λ∈σ(T )

We have
1
r(T ) = lim T n n
n→∞
If B is a Hilbert space and T is self–adjoint, then

r(T ) = kT k


c Joel Feldman. 2011. All rights reserved. October 28, 2011 Review of Spectral Theory 5
Theorem 15 Let H be a Hilbert space and A ∈ L(H).
(a) σ(A∗ ) = σ(A).
(b) λ ∈ σr (A) =⇒ λ ∈ σp (A∗ )
λ ∈ σp (A) =⇒ λ ∈ σp (A∗ ) ∪ σr (A∗ )

Lemma 16 Let H be a Hilbert space and A ∈ L(H) be normal.


(a) If ϕ is an eigenvector of A of eigenvalue λ, then ϕ is an eigenvector of A∗ of eigenvalue
λ̄.
(b) Eigenvectors of A with different eigenvalues are orthogonal.
(c) A has no residual spectrum.
(d) If A = A∗ , then σ(A) ⊂ IR.
(e) λ ∈ σ(A) if and only if, for each ε > 0, there exists a ϕ ∈ H with kϕk = 1 and
k(λ1l − A)ϕk < ε.

Theorem 17 Let H and H′ be Hilbert spaces. Let U ∈ L(H, H′ ) be bijective and A ∈ L(H)
and set A′ = U AU −1 ∈ L(H′ ). Then

ρ(A′ ) = ρ(A) σ(A′ ) = σ(A) σp (A′ ) = σp (A) σr (A′ ) = σr (A)

Theorem 18 (The Spectrum of Compact Operators) Let C : X → X be a compact


operator on the Banach space X . The spectrum of C consists of at most countably many

points. For any ε > 0, λ ∈ σ(C) |λ| > ε is finite. If 0 6= λ ∈ σ(C), then λ is an
eigenvalue of C of finite multiplicity.

Theorem 19 (Spectral Theorem - Multiplication Operator Version)


Let A be a bounded self–adjoint operator on a Hilbert space H. There exist
◦ a measure space hM, Σ, µi,
◦ a bounded measurable function a : M → IR, and
◦ a unitary operator U : H → L2 (M, Σ, µ)
such that
U AU −1 ϕ (m) = a(m) ϕ(m)


for all ϕ ∈ L2 (M, Σ, µ). If H is separable, µ can be chosen to be a finite measure.


Example 20 Let A be a self–adjoint, compact operator and let ϕn n∈I be a complete
orthonormal basis for H consisting of eigenvectors of A. Denote by λn the eigenvalue of A
for the eigenvector ϕn .


c Joel Feldman. 2011. All rights reserved. October 28, 2011 Review of Spectral Theory 6
Think of ℓ2 (I) as L2 of the measure space I, equipped with the counting measure.
So, think of an element of ℓ2 (I) as a function on I rather than a sequence. Define the
unitary operator U : H → ℓ2 (I) by
 P 
U xn ϕn (m) = xm
n∈I

The inverse operator U −1 = U ∗ : ℓ2 (I) → H is given by


X
U −1 v = v(n)ϕn
n∈I

For each m ∈ IN, denote by em the element of ℓ2 (I) all of whose components are zero
except for the mth , which is 1. That is

1 if n = m
em (n) =
0 6 m
if n =

Observe that U ϕm = em , for each m ∈ I. Also define the function a : IN → IR by


a(m) = λm . Then, for each v ∈ ℓ2 (I) and each m ∈ I,

(U AU −1 v)(m) = em , U AU −1 v ℓ2 = U −1 em , AU −1 v H = ϕm , AU −1 v H



= Aϕm , U −1 v H = λm ϕm , U −1 v H = λm hU ϕm , viℓ2


= λm hem , viℓ2 = a(m)v(m)

If, as will often be the case, I = IN, the counting measure on I is not finite. But it is easy
to rework the above construction so as to use a finite measure space. Define the measure µ
1 2 m/2
 P
on IN by µ {m} = 2m . Then define U : H → L (IN, µ) by U n∈I xn ϕn (m) = 2 xm
and the function a by a(m) = λm again.

Theorem 21 (Spectral Theorem - Multiplication Operator, with Multiplicity)


Let A be a bounded self–adjoint operator on a separable Hilbert space H. There exist
N ∈ IN ∪ {∞} and measures µn , 1 ≤ n ≤ N , on the spectrum σ(A) ⊂ IR, of A, and a
unitary operator U : H → ⊕N 2
n=1 L (σ(A), µn ) such that

U AU −1 ϕ n (λ) = λ ϕn (λ)


for all ϕ ∈ ⊕N 2
n=1 L (σ(A), µn ).


c Joel Feldman. 2011. All rights reserved. October 28, 2011 Review of Spectral Theory 7

Example 22 Let A be a self–adjoint, compact operator and let ϕn n∈I be a complete
orthonormal basis for H consisting of eigenvectors of A. Denote by λn the eigenvalue of A
for the eigenvector ϕn . Denote by 1 ≤ N ≤ ∞ the supremum of the multiplicities of the
eigenvalues of A. Define, for each 1 ≤ n ≤ N , the Borel measure µn on IR by
X n 1 if λ ∈ B and λ has multiplicity at least n
n n
µn (B) =
n∈I
0 otherwise

Then ⊕N 2
n=1 L (IN, µn ) is the set of all N –vector valued functions on IR with

N Z
X 2
~ 2=

kψk dµn (x) ψn (x)
n=1

Define the unitary operator U H → ⊕N 2


n=1 L (IN, µn ) by

∞ ∞

1 if ϕm is the nth eigenvector of eigenvalue λ
 P  P
U xm ϕm (λ) = xm
m=1 n m=1 0 otherwise

Theorem 23 (Spectral Theorem - Commuting Operators Version)


Let A1 , · · · , An be a finite set of commuting, bounded, self–adjoint operators on a Hilbert
space H. There exist
◦ a measure space hM, µi,
◦ bounded measurable functions aℓ : M → IR, 1 ≤ ℓ ≤ n, and
◦ a unitary operator U : H → L2 (M, µ)
such that, for each 1 ≤ ℓ ≤ n,

U Aℓ U −1 ϕ (m) = aℓ (m) ϕ(m)




for all ϕ ∈ L2 (M, µ). If H is separable, µ can be chosen to be a finite measure.

Corollary 24 Let A be a bounded normal operator on a Hilbert space H. There exist a


measure space hM, µi, a bounded measurable function a : M → C, and a unitary operator
U : H → L2 (M, µ) such that

U AU −1 ϕ (m) = a(m) ϕ(m)




for all ϕ ∈ L2 (M, µ).


c Joel Feldman. 2011. All rights reserved. October 28, 2011 Review of Spectral Theory 8
Theorem 25 (Spectral Theorem - Functional Calculus Version)

Let A be a bounded self–adjoint operator on a Hilbert space H. Let B = B [−kAk, kAk]
denote the set of all bounded Borel functions on [−kAk, kAk]. There exists a unique map
Φ : B → L(H) such that
(a) Φ is an algebra ∗–homomorphism. That is

Φ(αf + βg) = αΦ(f ) + βΦ(g) Φ(f¯) = Φ(f )∗


Φ(f g) = Φ(f )Φ(g) Φ(1) = 1l
for all f, g ∈ B and α, β ∈ C.

(b) Φ(f ) L(H) ≤ kf kL∞ for all f ∈ B.
(c) Φ(x) = A

(d) If the sequence fn n∈IN ⊂ B converges pointwise to f and is uniformly bounded, then
Φ(f ) = s-lim n→∞ Φ(fn ).
Furthermore, Φ has the following properties.
(e) If, for some λ ∈ IR and ψ ∈ H, we have Aψ = λψ, then Φ(f )ψ = f (λ)ψ, for all
f ∈ B.
(f ) If f ∈ B is nonnegative, then Φ(f ) is a nonnegative operator.
(g) If A, B ∈ L(H) commute, then so do Φ(f ) and B, for all f ∈ B. That is, if AB = BA,
then Φ(f )B = BΦ(f ).

Remark 26 It is common to write f (A) in place of Φ(f ). In this notation, the above
conclusions are
(a)
(αf + βg)(A) = αf (A) + βg(A) f¯(A) = f (A)∗
(f g)(A) = f (A)g(A) 1(A) = 1l

(b) f (A) L(H) ≤ kf kL∞
(c) x(A) = A
(d) If fn → f pointwise and is uniformly bounded, then f (A) = s-lim fn (A).
n→∞
(e) If Aψ = λψ, then f (A)ψ = f (λ)ψ.
(f) If f ≥ 0, then f (A) ≥ 0.
(g) If AB = BA, then f (A)B = Bf (A).

Example 27 Let A be a self–adjoint, compact operator and let ϕn n∈I be a complete
orthonormal basis for H consisting of eigenvectors of A. Denote by λn the eigenvalue of A
for the eigenvector ϕn . We just have to define, for each f ∈ B(IR)
P  P
f (A) xn ϕn = f (λn ) xn ϕn
n∈I n∈I


c Joel Feldman. 2011. All rights reserved. October 28, 2011 Review of Spectral Theory 9
Definition 28 (Projection Valued Measure) Denote by BIR the σ–algebra of Borel
subsets of IR and by L(H) the set of a bounded operators on H. A projection valued
measure is a map E : BIR → L(H) that obeys the following conditions.
(i) For each B ∈ BIR , the operator E(B) is an orthogonal projection on some closed
subspace of H. That is, E(B)2 = E(B) and E(B) = E(B)∗ .
(ii) E(∅) = 0 and E(IR) = 1l

(iii) If Bn n∈IN is a countable family of disjoint Borel subsets of IR, then


 S  N
X
E Bn = s-lim E(Bn)
n=1 N→∞
n=1

A projection valued measure is said to be bounded if, in addition,



(iv) There is an a > 0 such that E (−a, a) = 1l.
A projection valued measure automatically also obeys
(v) E(B1 ∩ B2 ) = E(B1 )E(B2 ) for all B1 , B2 ∈ BIR .
(vi) E(B1 ) and E(B2 ) commute for all B1 , B2 ∈ BIR .
R
Definition 29 ( f (λ) dE(λ)) Let B 7→ E(B) be a bounded projection valued measure,
and ϕ, ψ ∈ H. Observe that
◦ B 7→ hϕ, E(B)ϕi is an ordinary finite Borel measure on IR.
◦ B 7→ hϕ, E(B)ψi is an ordinary complex measure on IR. That is, there are (positive)
measures µ1 , µ2 , ν1 and ν2 such that

hϕ, E(B)ψi = µ1 (B) − µ2 (B) + iν1 (B) − iν2 (B)

By the polarization identity, we can take


1 1
µ1 (B) = 4
hϕ + ψ, E(B) (ϕ + ψ)i µ2 (B) = 4
hϕ − ψ, E(B) (ϕ − ψ)i
1 1
ν1 (B) = 4 hϕ + iψ, E(B) (ϕ + iψ)i ν2 (B) = 4 hϕ − iψ, E(B) (ϕ − iψ)i

◦ Use B(IR) to denote the set of bounded, Borel measurable functions on IR. If
R
f ∈ B(IR), then the map (ϕ, ψ) 7→ f (λ) d hϕ, E(λ)ψi is well–defined, bounded
and sesquilinear. So, by (a corollary to) the Riesz representation theorem, there is a
unique F ∈ L(H) such that
Z
hϕ, F ψi = f (λ) d hϕ, E(λ)ψi

for all ϕ, ψ ∈ H.
We define Z
f (λ) dE(λ) = F


c Joel Feldman. 2011. All rights reserved. October 28, 2011 Review of Spectral Theory 10
Example 30 Let H = L2 (M, µ) for some measure space (M, µ) and let a : M → IR be
bounded and measurable. Define, for each B ∈ BIR

E(B) = multiplication by χa−1 (B) (m) = χB a(m)

This is a bounded projection valued measure and


Z Z
f (λ) d hϕ, E(λ)ϕi = f a(m) |ϕ(m)|2 dµ(m)


Z Z

f (λ) d hϕ, E(λ)ψi = f a(m) ϕ(m)ψ(m) dµ(m)

Theorem 31 (Spectral Theorem - Projection–valued Measure Version)


There is a 1–1 correspondence between bounded self–adjoint operators and bounded
projection–valued measures A ↔ EA such that
Z
A = λ dEA (λ)


Example 32 Let A be a self–adjoint, compact operator and let ϕn n∈I be a complete
orthonormal basis for H consisting of eigenvectors of A. Denote by λn the eigenvalue of
A for the eigenvector ϕn . Define, for each m ∈ I, Pm to be the orthogonal projector onto
the linear subspace of H consisting of all scalar multiples of ϕm . That is
 P 
Pm xn ϕn = xm ϕm
n∈I

Set, for each Borel subset B of IR


X
EA (B) = Pm
m∈I
with λm ∈B

Then X X Z

A= λm Pm = λm EA {λm } = λ dEA (λ)
m∈I m∈I

Corollary 33

λ ∈ σ(A) ⇐⇒ EA (λ − ε, λ + ε) =6 0 for all ε > 0

λ ∈ σp (A) ⇐⇒ EA {λ} 6= 0

EA ρ(A) ∩ IR = 0
range EA ({0}) = ker A


c Joel Feldman. 2011. All rights reserved. October 28, 2011 Review of Spectral Theory 11
Corollary 34 Let −∞ < a < b < ∞.

(a) Then
Z b
1 1
(A − λ − iε)−1 − (A − λ + iε)−1 dλ
   
2
E [a, b] + E (a, b) = s-lim 2πi
ε→0 a

(b) Let a and b be in the resolvent set of A. Then


Z
1
  
EA [a, b] = EA [a, b) = EA (a, b] = EA (a, b)) = 2πi
Rζ (A) dζ
Ca,b

for any simple closed curve Ca,b in the complex plane with counterclockwise orientation

that contains λ + i0 a ≤ λ ≤ b in its interior.

Definition 35 A Banach Algebra A is a set which


(a) is a complex algebra, i.e. a set equipped with three operations

(A, B) ∈ A×A 7→ A+B ∈ A (α, A) ∈ C×A 7→ αA ∈ A (A, B) ∈ A×A 7→ AB ∈ A

called addition, scalar multiplication and multiplication, that obey the usual vector
space axioms and

A(BC) = (AB)C (A + B)C = AB + BC A(B + C) = AB + AC


α(BC) = (αB)C = B(αC)

for all A, B, C ∈ A and α ∈ C (so multiplication is associative, but not necessarily


commutative) and
(b) is normed, with the usual norm axioms, and also obeys kABk ≤ kAk kBk for all
A, B ∈ A and
(c) and is complete.
If, in addition,
(d) A contains an identity element 1l that obeys k1lk = 1 and A1l = 1lA = A for all A ∈ A
then A is called a unital Banach algebra. Any Banach algebra can be easily extended to

a unital Banach algebra A + 1l. (Define A1l = (α, A) α ∈ C, A ∈ A , the algebraic
operations by thinking of (α, A) as α1l + A, and k(α, A)k = |α| + kAkA .)

Definition 36 A C ∗ –algebra is a Banach algebra A together with a map ∗ : A → A that


obeys
(A + B)∗ = A∗ + B ∗ (αA)∗ = ᾱA∗ (AB)∗ = B ∗ A∗
A∗∗ = A kA∗ Ak = kAk2
for all A, B ∈ A and α ∈ C. (This used to be called a B ∗ –algebra.)


c Joel Feldman. 2011. All rights reserved. October 28, 2011 Review of Spectral Theory 12
Remark 37 The condition kA∗ Ak = kAk2 implies that kA∗ k = kAk since

kAk2 = kA∗ Ak ≤ kA∗ k kAk =⇒ kAk ≤ kA∗ k =⇒ kA∗ k ≤ kA∗∗ k = kAk

Example 38

(a) Let A = A∗ ∈ L(H). Then A = polynomials in A with complex coefficients , with




the overbar denoting norm closure, is a commutative C ∗ –algebra contained in L(H).

(b) Let A1 , · · · , An be self–adjoint, commuting, bounded operators on H. Then A =


polynomials in A1 , · · · , An with complex coefficients is a commutative C ∗ –algebra in


L(H).

(c) Any closed subalgebra of L(H) that is closed under the taking of adjoints is a C ∗ –
algebra. Conversely, any C ∗ –algebra is isomorphic to a subalgebra of L(H).

(d) Let X be a Hausdorff space (that is, a set equipped with open sets such that distinct
points have disjoint open neighbourhoods). Then A = C(X), the set of all continuous
functions on X, with the supremum norm, is a commutative C ∗ –algebra .

Theorem 39 (Gelfand–Naimark)
Let A be a commutative C ∗ –algebra with identity. Then there is a compact Hausdorff
space, X, (unique up to homeomorphism) such that A is ∗–isomorphic to C(X). That is,
there is a 1–1, onto map Ψ : A → C(X) such that

Ψ(A + B) = Ψ(A) + Ψ(B) Ψ(αA) = α Ψ(A) Ψ(AB) = Ψ(A) Ψ(B)


Ψ(A∗ ) = Ψ(A)

Ψ(A)
C(X)
= kAkA

for all A, B ∈ A and α ∈ C.


c Joel Feldman. 2011. All rights reserved. October 28, 2011 Review of Spectral Theory 13

You might also like