Functional Analysis: March 2010

See discussions, stats, and author profiles for this publication at: https://www.researchgate.
net/publication/45904504
Functional Analysis
Article · March 2010

Source: arXiv
CITATION READS
1 10,713
2 authors, including:
Palle Jorgensen
University of Iowa
282 PUBLICATIONS 9,533 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
DVT Prophylaxis View project
Fractal Calculus View project
All content following this page was uploaded by Palle Jorgensen on 30 May 2014.
The user has requested enhancement of the downloaded file.

Functional Analysis
Feng Tian, and Palle Jorgensen
arXiv:1003.1117v1 [math.FA] 4 Mar 2010
Department of Mathematics
14 MLH
The University of Iowa
Iowa City, IA 52242-1419
USA.
1
2
Notes from a course taught by Palle Jorgensen in the fall semester of 2009.
The course covered central themes in functional analysis and operator theory, with
an emphasis on topics of special relevance to such applications as representation
theory, harmonic analysis, mathematical physics, and stochastic integration.
3
These are the lecture notes I took from a topic course taught by Professor
Jorgensen during the fall semester of 2009. The course started with elementary
Hilbert space theory, and moved very fast to spectral theory, completely positive
maps, Kadison-Singer conjecture, induced representations, self-adjoint extensions
of operators, etc. It contains a lot of motivations and illuminating examples.
I would like to thank Professor Jorgensen for teaching such a wonderful course.
I hope students in other areas of mathematics would benefit from these lecture
notes as well.
Unfortunately, I have not been able to fill in all the details. The notes are
undergoing editing. I take full responsibility for any errors and missing parts.
Feng Tian
03. 2010
Contents
Chapter 1. Elementary Facts 6

1.1. Transfinite induction 6
1.2. Dirac’s notation 9
1.3. Operators in Hilbert space 11
1.4. Lattice structure of projections 14
1.5. Ideas in the spectral theorem 17
1.6. Spectral theorem for compact operators 24
Chapter 2. GNS, Representations 25

2.1. Representations, GNS, primer of multiplicity 25
2.2. States, dual and pre-dual 28
2.3. Examples of representations, proof of GNS 31
2.4. GNS, spectral thoery 33
2.5. Choquet, Krein-Milman, decomposition of states 36
2.6. Beginning of multiplicity 38
2.7. Completely positive maps 40
2.8. Comments on Stinespring’s theorem 47
2.9. More on the CP maps 51
2.10. Krien-Milman revisited 52
2.11. States and representation 54
2.12. Normal states 58
2.13. Kadison-Singer conjecture 59
Chapter 3. Appliations to Groups 61

3.1. More on representations 61
3.2. Some examples 63
3.3. Induced representation 65
3.4. Example - Heisenberg group 72
3.5. Coadjoint orbits 76
3.6. Gaarding space 79
3.7. Decomposition of representation 82
3.8. Summary of induced reprep, d/dx example 84
3.9. Connection to Nelson’s spectral theory 87
Chapter 4. Unbounded Operators 92

4.1. Unbounded operators, definitions 92
4.2. Self-adjoint extensions 93
Appendix 96
semi-direct product 96
4
CONTENTS 5
Bibliography 98
CHAPTER 1
Elementary Facts
1.1. Transfinite induction

Let (X, ≤) be a paritially ordered set. A sebset C of X is said to be a chain,
or totally ordered, if x, y in C implies that either x ≤ y or y ≤ x. Zorn’s lemma
says that if every chain has a majorant then there exists a maximal element in X.
Theorem 1.1. (Zorn) Let (X, ≤) be a paritially ordered set. If every chain C
in X has a majorant (or upper bound), then there exists an element m in X so that
x ≥ m implies x = m, for all x in X.
An illuminating example of a partially ordered set is the binary tree model.
Another example is when X is a family of subsets of a given set, partially ordered
by inclusion. Zorn’s lemma lies at the foundation of set theory. It is in fact an axiom
and is equivalent to the axiom of choice and Hausdorff’s maximality principle.
Theorem 1.2. (Hausdorff Maximality Principle) Let (X, ≤) be a paritially
ordered set, then there exists a maximal totally ordered subset L in X.
The axiom of choice is equivalent to the following statement on infinite product,
which itself is extensively used in functional analysis.
Theorem 1.3. (axiom of choice) Let Aα be a family
Q of nonempty sets indexed
by α ∈ I. Then the infinite Cartesian product Ω = α Aα is nonempty.
Ω can be seen as the set of functions {(xα ) : xα ∈ Aα } from I to ∪Aα . The
point of using the axiom of choice is that if the index set is uncountable, there is no
way to verify whether (xα ) is in Ω or not. It is just impossible to check for each α
that xα in contained in Aα , for some coordinates will be unchecked. The power of
transfinite induction is that it applies to uncountable sets as well. In case the set
is countable, we simply apply the down to earth standard induction. The standard
mathematical induction is equivalent to the Peano’s axiom which states that every
nonempty subset of of the set of natural number has a unique smallest element.
The key idea in applications of the transfinite induction is to cook up in a
clear way a partially ordered set, so that the maximum element turns out to be
the object to be constructed. Examples include Hahn-Banach extension theorem,
Krein-Millman’s theorem on compact convex set, existance of orthonoral basis in
Hilbert space, Tychnoff’s theorem on infinite Cartesian product of compact spaces,
where the infinite product space is nonempty follows imediately from the axiom of
choice.
Theorem 1.4. (Tychonoff ) Let Aα be Q a family of compact sets indexed by
α ∈ I. Then the infinite Cartesian product α Aα in compact with respect to the
product topology.
6
1.1. TRANSFINITE INDUCTION 7
We will apply the transfinite induction to show that every infinite dimensional
Hilbert space has an orthonormal basis (ONB).
Classical functional analysis roughly divides into two branches
• study of function spaces (Banach space, Hilbert space)
• applications in physics and engineering
Within pure mathematics, it is manifested in
• representation theory of groups and algebras
• C ∗ -algebras, Von Neumann algebras
• wavelets theory
• harmonic analysis
• analytic number theory
Definition 1.5. Let X be a vector space over C. k·k is a norm on X if for all
x, y in X and c in C
• kcxk = ckxk
• kxk ≥ 0; kxk = 0 implies x = 0
• kx + yk ≤ kxk + kyk
X is a Banach space if it is complete with respect to the metric induced by k·k.
Definition 1.6. Let X be vector space over C. An inner product is a function
h·, ·i : X × X → C so that for all x, y in H and c in C,
• hx, ·i is linear (linearity)
• hx, yi = hy, xi (conjugation)
• hx, xi ≥ 0; and hx, xi = 0 implies x = 0 (positivity)
p
In that case hx, xi defines a norm on H and is denote by kxk. X is said to be
an inner product space is an inner product is defined. A Hilbert space is a complete
inner product space.
Remark 1.7. The abstract formulation of Hilbert was invented by Von Neu-
mann in 1925. It fits precisely with the axioms of quantum mechanics (spectral
lines, etc.) A few years before Von Neumann’s formulation, Heisenberg translated
Max Born’s quantum mechanics into mathematics.
For any inner product space H, observe that the matrix

hx, xi hx, yi
hy, xi hy, yi
is positive definite by the positivity axiom of the definition of an inner product.
Hence the matrix has positive determinant, which gives rises to the famous Cauchy-
Schwartz inequality
|hx, yi|2 ≤ hx, xihy, yi.
An extremely useful way to construct a Hilbert space is the GNS construction,
which starts with a semi-positive definite funciton defined on a set X. ϕ : X × X →
C is said to be semi-positive definite, if for finite collection of complex nubmers {cx },
X
c̄x cy ϕ(x, y) ≥ 0.
Let H0 be the span of δx where x ∈ X, and define a sesiquilinear form h·, ·i on H0
as X X X
h cx δx , cy δy i := c̄x cy ϕ(x, y).
1.1. TRANSFINITE INDUCTION 8
However, the positivity condition may not be satisfied. Hence one has to pass to a
quotient space by letting N = {f ∈ H0 , hf, f i = 0}, and H̃0 be the quotient space
H0 /N . The fact that N is really a subspace follows from the Cauchy-Schwartz
inequality above. Therefore, h·, ·i is an inner product on H̃0 . Finally, let H be the
completion of H̃0 under h·, ·i and H is a Hilbert space.
Definition 1.8. Let H be a Hilbert space. A family of vectors {uα } in H is
said to be an orthonormal basis of H if
(1) huα , uβ i = δαβ and
(2) span{uα } = H.
We are ready to prove the existance of an orthonormal basis of a Hilbert space,
using transfinite induction. Again, the key idea is to cook up a partially ordered
set satisfying all the requirments in the transfinite induction, so that the maximum
elements turns out to be an orthonormal basis. Notice that all we have at hands
are the abstract axioms of a Hilbert space, and nothing else. Everything will be
developed out of these axioms.
Theorem 1.9. Every Hilbert space H has an orthonormal basis.
To start out, we need the following lemmas.
Lemma 1.10. Let H be a Hilbert space and S ⊂ H. Then the following are
equivalent:
(1) x ⊥ S implies x = 0
(2) span{S} = H
Lemma 1.11. (Gram-Schmidt) Let {un } be a sequence of linearly independent
vectors in H then there exists a sequence {vn } of unit vectors so that hvi , vj i = δij .
Remark. The Gram-Schmidt orthogonalization process was developed a little
earlier than Von Neumann’s formuation of abstract Hilbert space.
Proof. we now prove theorem (1.9). If H is empty then we are finished.
Otherwise, let u1 ∈ H. If ku1 k 6= 1, we may consider u1 /ku1 k which is a normalized
vector. Hence we may assume ku1 k = 1. If span{u1 } = H we are finished again,
otherwise there exists u2 ∈/ span{u1}. By lemma (1.11), we may assume ku2 k = 1
and u1 ⊥ u2 . By induction, we get a collection S of orthonormal vectors in H.
Consider P(S) partially order by set inclusion. Let C ⊂ P(S) be a chain and let
M = ∪E∈C E. M is clearly a majorant of C. We claim that M is in the partially
ordered system. In fact, for all x, y ∈ M there exist Ex and Ey in C so that x ∈ Ex
and y ∈ Ey . Since C is a chain, we may assume Ex ≤ Ey . Hence x, y ∈ E2 and
x ⊥ y, which shows that M is in the partially ordered system.
By Zorn’s lemma, there exists a maximum element m ∈ S. It suffices to show
that the closed span of m is H. Suppose this is false, then by lemma (1.10) there
exists x ∈ H so that x ⊥ M . Since m ∪ {x} ≥ m and m is maximal, it follows that
x ∈ m, which implies x ⊥ x. By the positivity axiom of the definition of Hilbert
space, x = 0.
Corollary 1.12. Let H be a Hilbert space, then H is isomorphic to the l2
space of the index set of an ONB of H.
Remark 1.13. There seems to be just one Hilbert space, which is true in
terms of the Hilbert space structure. But this is misleading, because numerous
1.2. DIRAC’S NOTATION 9
interesting realizations of an abstract Hilbert space come in when we make a choice

of the ONB. The question as to which Hilbert space to use is equivalent to a good
choice of an ONB. This is simiar to the argument that there is just one set for
each given cardinality in terms of set structure, but there are numerous choices of
elements in the sets making questions interesting.
Suppose H is separable, for instance let H = L2 (R). Then H ∼ = l2 (N) ∼=
2
l (N × N). It follows that potentially we could choose a doublely indexed basis
{ψjk : j, k ∈ N} for L2 . It turns out that this is precisely the setting of wavelet
basis! What’s even better is that in l2 space, there are all kinds of diagonalized
operators, which correspond to self-adjoint (or normal) operators in L2 . Among
these operators in L2 , we single out the scaling (f (x) 7→ f (2j x)) and translation
(f (x) 7→ f (x − k)) operators, which are diagonalized, NOT simultaneously though.
Q
1.1.1. path space measures. Let Ω = ∞ k=1 {1, −1} be the infinite Cartesian
product of {1, −1} with the product topology. Ω is compact and Hausdorff by
Tychnoff’s theorem.
For each k ∈ N, let Xk : Ω → {1, −1} be the k th coordinate projection, and
assign probability measures µk on Ω so that µk ◦ Xk−1{1} = a and µk ◦ Xk−1 {−1} =
1 − a, where a ∈ (0, 1). The collection of measures {uk } satisfies the consistency
conditiond, i.e. µk is the restriction of µk+1 onto the k th coordinate space. By
Kolomogorov’s extension theorem, there exists a unique probability measure P on
Ω so that the restriction of P to the k th coordinate is equal to µk .
It follows that {Xk } is a sequence of independent identically distributed (i.i.d.)
random variables in L2 (Ω, P ) with E[Xk ] = 0 and V ar[Xk ] = 1; and L2 (Ω, P ) =
span{Xk }.
Let H be a separable Hilbert space with an orthonormal basis {uk }. The
map ϕ : uk 7→ Xk extends linearly to an isometric embedding of H into L2 (Ω, P ).
Moreover, let F+ (H) be the symmetric Fock space. F+ (H) is the closed span of
the the algebraic tensors uk1 ⊗ · · · ⊗ ukn , thus ϕ extends to an isomorphism from
F+ (H) to L2 (Ω, P ).
1.2. Dirac’s notation

P.A.M Dirac was every efficient with notations, and he introduced the “bra-ket”
vectors. Let H be a Hilbert space with inner product h·, ·i : H ×H → C. We denote
by “bra” for vectors hx| and “ket” for vectors |yi where x, y ∈ H.
With Dirac’s notation, our first observation is the followsing lemma.
Lemma 1.14. Let v ∈ H be a unit vector. The operator x 7→ hv, xiv can be
written as Pv = |vihv|. Pv is a rank-one self-adjoint projection.
Proof. Pv2 = (|vihv|)(|vihv|) = |vihv| = Pv . Since
hx, Pv yi = hx, vihv, yi = hhx, viv, yi = hhv, xiv, yi = hPv x, yi
so Pv = Pv∗ .
More generally, any rank-one operator can be wrritten as |uihv| sending x ∈
H to < v, x > u. With the bra-ket notation, it’s easy to verify that the set
of rank-one operators forms an algebra, which easily follows from the fact that
(|v1 ihv2 |)(|v3 ihv4 |) = |v1 ihv4 |. The moment that an orthonormal basis is selected,
the algebra of operators on H will be translated to the algebra of matrices (infinite).
1.2. DIRAC’S NOTATION 10
Every Hilbert space has an ONB, but it does not mean in pratice it is easy to select
one that works well for a particular problem.
It’s also easy to see that the operator
X
PF = |vi ihvi |
vi ∈F
where F is a finite set of orthonormal vectors in H, is a self-adjoint projection.

This follows, since
X X
PF2 = (|vi ihvi |)(|vj ihvj |) = |vi ihvi |
vi ,vj ∈F vi ∈F
and PF∗
= PF .
The Gram-Schmidt orthogonalization process may now be written in Dirac’s
notation so that the induction step is really just
x − PF x
kx − PF xk
which is a unit vector and orthogonal to PF H. Notice that if H is non separable,
the standard induction does not work, and the transfinite induction is needed.
1.2.1. connection to quamtum mechanics. Quamtum mechanics was born

during the years from 1900 to 1913. It was created to explain phenomena in black
body radiation, hydrogen atom, where a discrete pattern occurs in the frequences
of waves in the radiation. The radiation energy E = ν~, with ~ being the Plank’s
constant. Classical mechanics runs into trouble.
During the years of 1925~1926, Heisenberg found a way to represent the energy
E as a matrix, so that the matrix entries < vj , Evi > represents the transition prob-
ability from energy i to energy j. A foundamental relation in quantum mechenics is
the commutation relation satisfied by the momentum operator P and the position
operator Q, where
1
P Q − QP = I.
i
Heisernberg represented the operators P, Q by matrices, although his solution is
not real matrices. The reason is for matrices, there is a trace operation where
trace(AB) = trace(BA). This implies the trace on the left-hand-side is zero, while
the trace on the righ-hand-side is not. This suggests that there is no finite dimen-
sional solution to the commutation relation above, and one is forced to work with
infinite dimensional Hilbert space and operators on it. Notice also that P, Q do not
commute, and the above commutation relation leads to the uncertainty principle
(Hilbert, Max Born, Von Neumann worked out the mathematics), which says that
the statistical variance △P and △Q satisfy △P △Q ≥ ~/2 . We will come back to
this later.
However, Heisenberg found his “matrix” solutions, where
 
0 1 √
 1 0 2 √ 
 √ 
 2 0 3 
P = √

 .. 

 3 0 . 

.. ..
. .
1.3. OPERATORS IN HILBERT SPACE 11
and  
0 1 √
 −1 0 2 √ 
 √ 
1 − 2 0 3 
Q=  √

i .. 

 − 3 0 . 

.. ..
. .
the complex i in front of Q is to make it self-adjoint.
A selection of ONB makes a connection to the algebra operators acting on H
and infinite matrices. We check that using Dirac’s notation, the algebra of operators
really becomes the algebr of infinite matrices.
Pick an ONB {ui } in H, A, B ∈ B(H). We denote by MA = Aij := hui , Auj i
the matrix of A under the ONB. We compute hui , ABuj i.
X
(MA MB )ij = Aik Bkj
k
X
= hui , Auk ihuk , Buj i
k
X
= hA∗ ui , uk ihuk , Buj i
k
= hA∗ ui , Buj i
= hui , ABuj i
P
where I = |ui ihui |.
P Let w 2be a unit vector in H. w represents a quantum state. Since kwk2 =
2
|hui , wi| = 1, the numbers |hui , wi| represent a probability distribution over
the index set. If w and w′ are two states, then
X
hw, w′ i = hw, ui ihui , w′ i
which has the interpretation so that the transition from w′ to w may go through
all possible intermediate states ui . Two states are uncorrelated if and only if they
are orthogonal.
1.3. Operators in Hilbert space

Definition 1.15. Let A be a linear operator on a Hilbert space H.
(1) A is self-adjoint if A∗ = A
(2) A is normal if AA∗ = A∗ A
(3) A is unitary if AA∗ = A∗ A = I
(4) A is a self-adjoint projection if A = A∗ = A2
Let A be an operator, then we have R = (A + A∗ )/2, S = (A − A∗ )/2i which
are both self-adjoint, and A = R + iS. This is similar the to decomposition of a
complex nubmer into its real and imaginary parts. Notice also that A is normal if
and only if R and S commute. Thus the study of a family of normal operators is
equivalent to the study of a family of commuting self-adjoint operators.
Lemma 1.16. Let z be a complex number, and P be a self-adjoint projection.
Then U (z) = zP + (I − P ) is unitary if and only if |z| = 1.
Proof. Since P is a self-adjoint projection,

U (z)U (z)∗ = U (z)∗ U (z) = (zP + (I − P )) (z̄P + (I − P )) = |z|2 P + (I − P ).
If |z| = 1 then U (z)U (z)∗ = U (z)∗ U (z) = I and U (z) is unitary. Conversely, if
|z|2 P + (I − P ) = I then it follows that |z|2 − 1 P = 0. If we assume that P is
nondegenerate, then |z| = 1.
Definition 1.17. Let A be a linear operator on a Hilbert space H. The
resolvent R(A) is defined as
R(A) = {λ ∈ C : (λI − A)−1 exists}
and the spectrum of A is the complement of R(A), and it is denoted by sp(A) or
σ(A).
Definition 1.18. Let B(C) be the Borel σ-algebra of C. H is a Hilbert space.
P : B(C) → H is a projection-valued measure, if
(1) P (φ) = 0, P (C) = I
P∩ B) = P
(2) P (A P (A)P (B)
(3) P ( Ek ) = P (Ek ), Ek ∩ Ej = φ if k 6= j. The convergence is in terms
of the strong operator topology.
Von Neumann’s spectral theorem states that an operator A is ´normal if and
only if there exits a projection-valued measure on C so that A = sp(A) zP (dz),
i.e. A is represented as an integral again the projection-valued measure P over its
spectrum.
In quamtum mechanics, an observable is represented by a self-adjoint operator.
Functions of observables are again observables. This is reflected´ in the spectral
theorem as the functional calculus, where we may define f (A) = sp(A) f (z)P (dz)
using the spectral representation of A.
The stardard diagonalization of Hermitian matrix in linear algebra is a special
case
P of the spectral theorem. Recall that if A is a Hermitian matrix, then A =
′ ′
k λk Pk where λk s are the eigenvalues of A and Pk s are the self-adjoint projections
′
onto the eigenspace associatedP with λk s. The projection-valued measure in this case
can be written as P (E) = λk ∈E Pk , i.e. the counting measure supported on λ′k s.
Hersenberg’s commutation relation P Q − QP = −iI is an important example
of two non-commuting self-adjoint operators. When P is a self-adjoint projection
acting on a Hilbert space H, hf, P f i is a real number and it represents observation
of the observable P prepared in the state |f i. Quantum mechanics is stated using
an abstract Hilbert space as the state space. In practice, one has freedom to choose
exactly which Hilbert space to use for a particular problem. The physics remains to
same when choosing diffenent realizations of a Hilbert space. The concept needed
here is unitary equivalence.
Suppose U : H1 → H2 is a unitary operator, P : H1 → H1 is a self-adjoint
projection. Then U P U ∗ : H2 → H2 is a self-adjoint projection on H2 . In fact,
(U P U ∗ )(U P U ∗ ) = U P U ∗ where we used U U ∗ = U ∗ U = I, as U is unitary. Let
|f1 i be a state in H1 and |U f1 i be the corresponding state in H2 . Then
hf2 , U P U ∗ f2 i = hU ∗ f2 , P U ∗ f2 i = hf1 , P f1 i
i.e. the observable P has the same expectation value. Since every self-adjoint
operator is, by the spectral theorem, decomposed into self-adjoint projections, it
follows the expectation value of any observable remains unchanged under unitary
transformation.
Definition 1.19. Let A : H1 → H1 and B : H2 → H2 be operators. A is
unitarily equivalent to B is there exists a unitary operator U : H1 → H2 so that
B = U AU ∗ .
Example 1.20. Fourier transform U : L2 (R) → L2 (R),
1
ˆ
(U f )(t) = fˆ(t) = √ e−itx f (x)dx.
2π
The operators Q = Mx and P = −id/dx are both densely defined on the Schwartz
space S ⊂ L2 (R). P and Q are unitary equivalent via the Fourier transform,
P = F ∗ QF .
2
Apply Gram-Schmidt orthogonalization to polynomials against the measrue e−x /2 dx,
and get orthognoal polynomials. There are the Hermite polynomials (Hermit func-
tions). n
−x2 /2 −x2 d 2
hn = e Pn = e ex /2 .
dx
The Hermite functions form an orthonormal basis (normalize it) and transform P
and Q to Heisenberg’s infinite matrices. Some related operators: H := (Q2 + P 2 −
1)/2. It can be shown that
Hhn = nhn
or equivalently,
(P 2 + Q2 )hn = (2n + 1)hn
n = 0, 1, 2, . . .. H is called the energy operator in quantum mechanics. This explains
mathematically why the energy levels are discrete, being a multiple of ~.
A multiplication operator version is also available which works especially well
in physics. It says that A is a normal operator in H if and only if A is unitarily
equivalent to the operator of multiplication by a measurable function f on L2 (M, µ)
where M is compact and Hausdorff. We will see how the two versions of the spectral
theorem are related after first introduing the concept of transformation of measure.
1.3.1. Transformation of measure. Let (X, S) and (Y, T ) be two measur-
able spaces with σ-algebras S and T respectively. Let ϕ : X → Y be a measurable
function. Suppose there is a measure µ on (X, S). Then µϕ (·) := µ ◦ ϕ−1 (·) defines
a measure on Y . µϕ is the transformation measure of µ under ϕ.
Notice that if E ∈ T , then ϕ(x) ∈ E if and only if x ∈ ϕ−1 (E) ∈ S. Hence
χE ◦ ϕ(x) = χϕ−1 (E) .
P P
It follows that for simple function s(·) = ci χEi (·) = ci χϕ−1 (Ei ) , and
ˆ ˆ X ˆ X ˆ
s ◦ ϕdµ = ci χEi ◦ ϕ(x)dµ = ci χϕ−1 (Ei ) (·)dµ = sd(µ ◦ ϕ−1 ).
With a standard approximation of measurable functions by simple functions, we

have for any measurable function f : X → Y ,
ˆ ˆ
f (ϕ(·))dµ = f (·)d(µ ◦ ϕ−1 ).
The above equation is a generalization of the substitution formula in calculus.

1.4. LATTICE STRUCTURE OF PROJECTIONS 14
The multiplication version of the spectral theory states that every normal op-
erator A is unitarily equivalent to the operator of multiplication by a measurable
function Mf on L2 (M, µ) where M is compact and Hausdorff. With transforma-
tion of measure, we can go one step further and get that A is unitarily equivalent
to the operator of multiplication by the independent variable on some L2 space.
Notice that if f is nesty, even if µ is a nice measure (say the Lebesgue measure),
the transformation meaure µ ◦ f −1 can still be nesty, it could even be singular.
Let’s assume we have a normal operator Mϕ : L2 (M, µ) → L2 (M, µ) given by
multiplication by a measurable function ϕ. Define an operator U : L2 (ϕ(M ), µ ◦
ϕ−1 ) → L2 (M, µ) by
(U f )(·) = f (ϕ(·)).
U is unitary, since
ˆ ˆ ˆ
|f | d(µ ◦ ϕ ) = |f (ϕ)| dµ = |U f |2 dµ.
2 −1 2
Claim also that

Mϕ U = U Mt .
To see this, let f be a µϕ -measurable function. Then
Mϕ U f = ϕ(t)f (ϕ(t))
U Mt f = U (tf (t))
= ϕ(t)f (ϕ(t))
Recall we have stated two versions of the spectral theorem. (multiplication op-
erator and projection-valued measure) Consider the simplest case for the projection-
valued measure, where we work with L2 (X, µ). Claim that P (E) := M χE , i.e. the
operator of multiplication by χE on the Hilbert space L2 (X, µ), is a projection-
valued measure.
Apply this idea to P and Q, the momemtum and positon operators in quantum
mechanics. P = −id/dx, Q = Mx . As we discussed before,
P = F −1 QF
in other words, Q and P are unitarily equivalent via the Fourier transform, which
diagonalizs P . Now we get a projection-valued measure (PVM) for P by
E(·) := F −1 Mχ{·} F .
This can be seen as the convolution operator with respect to the inverse Fourier
transform of χ{·} .
1.4. Lattice structure of projections

We first show some examples of using Gram-Schmidt orthoganoliztion to obtain
orthonormal bases for a Hilbert space.
Example 1.21. H = L2 [0, 1]. The polynomials {1, x, x2 , . . .} are linearly inde-
pendent in H, since if X
ck xk = 0
then as an analytic function, the left-hand-side must be identically zero. By Stone-
Weierstrass theorem, span{1, x, x2 , . . .} is dense in C([0, 1]) under the k·k∞ norm.
Since k·kL2 ≤ k·k∞ , it follows that span{1, x, x2 , . . .} is also dense in H. By Gram-
Schmidt, we get a sequence {Vn } of finite dimensional subspaces in H, where Vn
has an orthonormal basis {h0 , . . . , hn−1 }, so that spanVn = span{1, x, . . . , xn−1 }.

Define
xn+1 − Pn xn+1
hn+1 = n+1 .
kx − Pn xn+1 k
The set {hn } is dense in H, since span{hn } = span{1, x, . . .} and the latter is dense
in H. Therefore, {hn } forms an orthonormal basis of H.
Example 1.22. H = L2 [0, 1]. Consider the set of complex exponentials {ei2πnx }∞
n=0 .
This is already an ONB for H and leads to Fourier series. Equivalently, may also
consider {cos 2πnx, sin 2πnx}∞
n=0 .
The next example constructs the Haar wavelet.

Example 1.23. H = L2 [0, 1]. Let ϕ0 be the characteristic function of [0, 1].
Define ϕ1 = ϕ0 (2x)−ϕ0 (2x−1) and ψjk = 2k/2 ϕ1 (2k x−l). For fixed k and j1 6= j2 ,
hψjk1 , ψjk2 i = 0 since they have disjoint support.
Exercise 1.24. Let Mt : L2 [0, 1] → L2 [0, 1] be the operator of multiplication
by t. Compute the matrix of Mt under wavelet basis. (this is taken from Joel
Anderson, who showed the A = A∗ implies that A = D + K where D is a diagonal
operator and K is a compact perturbation.d)
Theorem 1.25. Let H be a Hilbert space. There is a one-to-one correspondence
between self-adjoint projections and closed subspaces of H.
Proof. Let P be a self-adjoint projection in H. i.e. P 2 = P = P ∗ . Then
P H = {x ∈ H : P x = x} is a closed subspace. Denote by P ⊥ the completement of
P , i.e. P ⊥ = 1 − P . Then P ⊥ H = {x ∈ H : P ⊥ x = x} = {x ∈ H : P x = 0}. Since
P P ⊥ = P (1 − P ) = P − P 2 = P − P = 0, therefore P H ⊥ P ⊥ H.
Conversely, let W be a closed subspace in H. First notice that the parallelogram
law is satisfied in a Hilbert space, where for any x, y ∈ H, kx + yk2 + kx − yk2 =
2(kxk2 + kyk2 ). Let x ∈ H\W , define d = inf w∈W kx − wk. By definition, there
exists a sequence wn in W so that kwn −xk → 0 as n → ∞. Apply the parallelogram
law to x − wn and x − wm ,
k(x − wn ) + (x − wm )k2 + k(x − wn ) − (x − wm )k2 = 2(kx − wn k2 + kx − wm k2 )
which simplies to
wn + wm 2
4kx − k + kwn − wm k2 = 2(kx − wn k2 + kx − wm k2 ).
2
Notice here all we require is (wn + wm )/2 lying in the subspace W , hence it suffices
to require simply that W is a convex subset in H. see Rudin or Nelson page 62 for
more details.
Von Neumann invented the abstract Hilbert space in 1928 as shown in one of
the earliest papers. He work was greatly motivated by quantum mechanics. In order
to express quantum mechanics logic operations, he created lattices of projections,
so that everything we do in set theory with set operation has a counterpart in the
operations of projections.
SETS CHAR PROJECTIONS DEFINITIONS
A∩B χA χB P ∧Q PH ∩ PQ
A∪B χA∪B P ∨Q span{P H ∪ QH}
A⊂B χA χB = χA P ≤Q P H ⊂ QH
⊂ A2 ⊂ · · ·
A1 S χAi χAi+1 = χAi P1 ≤ P2 ≤ · · · ⊂ Pi+1 H
Pi H S
∞ ∞
Tk=1 Ak χ∪ k A ∨∞
k=1 Pk T∞ k=1 Pk H}
span{
∞
k=1 Ak χ ∩ k Ak ∧∞
k=1 Pk k=1 PK H
A×B (χA×X ) (χX×B ) P ⊗Q P ⊗ Q ∈ proj(H ⊗ K)
P H ⊂ QH ⇔ P = P Q. This is similar to set operation where A ∩ B =

A ⇔ A ⊂ B. In general, product and sum of projections are not projections. But
if P H ⊂ QH then the product is in fact a projection. Taking adjoint, one get
P ∗ = (P Q)∗ = Q∗ P ∗ = QP . It follows that P Q = QP = P . i.e. containment
implies the two projections commute.
During the same time period as Von Neumann developed his Hilbert space the-
ory, Lesbegue developed his integration theory which extends the classical Riemann
integral. The motone sequence of sets A1 ⊂ A2 ⊂ · · · in Lebesgue’s integration the-
ory also has a counterpart in the theory of Hilbert space. To see what happens here,
let P1 ≤ P2 and we show that this implies kP1 xk ≤ kP2 xk.
Lemma 1.26. P1 ≤ P2 ⇒ kP1 xk ≤ kP2 xk.
Proof. It follows from
kP1 xk2 = hP1 x, P1 xi = hx, P1 xi = hx, P2 P1 xi ≤ kP1 P2 xk2 ≤ kP2 xk2 .

As a consequence, P1 ≤ P2 ≤ · · · implies kPk xk forms a monotone increasing
sequence in R, and the sequence is bounded by kxk, since kPk xk ≤ kxk for all k.
Therefore the sequence Pk converges to P , in symbols
∨Pk = lim Pk = P
k
in the sense that (strongly convergent) for all x ∈ H, there exists a vector, which
we denote by P x so that
limkPk x − P xk = 0
k
and P really defines a self-adjoint projection.
The examples using Gram-Schmidt can now be formulated in the lattice of pro-
jections. We have Vn the n-dimensional subspaces and Pn the orthogonal projection
onto Vn , where
Vn ⊂ Vn+1 → ∪Vn ∼ Pn ≤ Pn+1 → P
Pn⊥ ≥ Pn+1
⊥
→ P ⊥.
Since ∪Vn is dense in H, it follows that P = I and P ⊥ = 0. We may express this
in the lattice notations by
∨Pn = sup Pn = I
∧Pn⊥ = inf Pn = 0.
The tensor product construction fits with composite system in quamtum me-
chanics.
1.5. IDEAS IN THE SPECTRAL THEOREM 17
Lemma 1.27. P, Q ∈ proj(H). Then P + Q ∈ proj(H) if and only if P Q =

QP = 0. i.e. P ⊥ Q.
Proof. Notice that

(P + Q)2 = P + Q + P Q + QP.
If P Q = QP = 0 then (P + Q)2 = P + Q = (P + Q)∗ , hence P + Q is a projection.
Conversely, if P +Q ∈ proj(H), then (P +Q)2 = (P +Q) implies that P Q+QP = 0.
Since (P Q)∗ = Q∗ P ∗ = QP it follows that 2QP = 0 hence P Q = QP = 0.
In terms of characteristic functions,

χA + χB = χA∪B − χA∩B
hence χA + χB is a characteristic function if and only if A ∩ B = φ.
The set of projections in a Hilbert space H is partially ordered according to the
corresponding closed subspaces paritially ordered by inclusion. Since containment
implies commuting, the chain of projections P1 ≤ P2 ≤ · · · is a family of commut-
ing self-adjoint operators. By the spectral theorem, {Pi } may be simultaneously
diagonalized, so that Pi is unitarily equivalent to the operator of multiplication by
χEi on the Hilbert space L2 (X, µ), where X is a compact and Hausdorff space.
Therefore the lattice structure of prjections in H is precisely the lattice structure
of χE , or equivalently, the lattice structure of measurable sets in X.
Lemma 1.28. Consider L2 (X, µ). The followsing are equivalent.
(1) E ⊂ F;
(2) χE χF = χF χE = χE ;
(3) kχE f k ≤ kχF f k, for any f ∈ L2 ;
(4) χE ≤ χF in the sense that hf, χE f i ≤ hf, χF f i, for any f ∈ L2 .
Proof. The proof is trivial. Notice that

ˆ ˆ
hf, χE f i = ¯ ¯
f χE f dµ = χE |f |2 dµ
ˆ ˆ
kχE f k2 = |χE f |2 dµ = χE |f |2 dµ
where we used that fact that

χE = χ̄E = χ2E .

1.5. Ideas in the spectral theorem

We show some main ideas in the spectral theorem. Since every normal operator
N can be written as N = A + iB where A, B are commuting self-adjoint operators,
the presentation will be focused on self-adjoint operators.
Let A be a self-adjoint operator acting on a Hilbert space H. There are two
versions of the spectral theorem. The projection-valued measure (PVM), and the
multiplication operator Mf .
1.5.1. Multiplication by Mf .
(1) In this version of the spectral theorem, A = A∗ implies that A is unitarily
equivalent to the operator Mf of multiplication by a measurable function
f on the Hilbert space L2 (X, µ), where X is a compact Hausdorff space,
and µ is a regular Borel measure.
A /H
HO O
U U
Mf
L2 (µ) / L2 (µ)
f induces a Borel measure µf (·) = µ ◦ f −1 (·) on R, supported on f (X).

Define the operator W : L2 (µf ) → L2 (µ) where
W : g → g ◦ f.
Then,
ˆ ˆ
kgk2L2 (µf ) = |g|2 dµf = |g ◦ f |dµ = kW gk2L2(µ)
X
hence W is unitary. Moreover,
Mf W g = f (x)g(f (x))
W Mt g = W (tg(t)) = f (x)g(f (x))
it follows that Mf is unitarily equivalent to Mt on L2 (µf ), and Mf =
W Mt W −1 . The combined transformation F := U W diagonalizes A as
A = F Mt F −1 .
It is seens as a vast extention of diagonalizing hermitian matrix in linear
algebra, or a generalization of Fourier transform.
A /H
? HO O _<

U U
2
)
Mf
F
2
L (µ) / L2 (µ) F
) O O

2 W W
<
Mt
L2 (µf ) / L2 (µf )
(2) What’s involved are two algebras: the algebra of measurable functions
on X, treated as multiplication operators, and the algebra of operators
generated by A (with identity). The two algebras are ∗-isomorphic. The
spectral theorem allows to represent the algebra of A by the algebra of
functions (in this direction, it helps to understand A); also represent the
algebra of functions by algebra of operators generated by A (in this di-
rection, it reveals properties of the function algebra and the underlying
space X. We will see this in a minute.)
(3) Let A be the algebra of functions. π ∈ Rep(A, H) is a representation,
where
π(ψ) = F Mψ F −1
and we may define operator

ψ(A) := π(ψ).
This is called the spectral representation. In particular, the spectral the-
orem of A implies the following substitution rule
X X
ck xk 7→ ck Ak
is well-defined, and it extends to all bounded measurable functions.
Remark 1.29. Notice that the map
ψ 7→ ψ(A) := π(ψ) = F Mψ F −1
is an algebra isomorphism. To check this,
(ψ1 ψ2 )(A) = F Mψ1 ψ2 F −1
= F Mψ1 Mψ2 F −1

= F Mψ1 F −1 F Mψ2 F −1
= ψ1 (A)ψ2 (A)
where we used the fact that
M ψ1 ψ2 = M ψ1 M ψ2
i.e. multiplication operators always commute.
1.5.2. Projection-valued measure. Alternatively, we have the PVM ver-
sion of the spectral theorem. ˆ
A = xP (dx)
where P is a projection-valued measure defined on the Borel σ-algebra of R. Notice

that the support of P P might be a proper subset of R. Recall that P is a PVM if
P (φ) = 0, P (∪Ek ) = P (E ) and Ek ∩ El = φ for k 6= l. By assumption, the
PN k
sequence of projections k=1 P (Ek ) (Ek′ s mutually disjoint) is monotone increas-
PN
ing, hence it has a limit, limN → k=1 P (Ek ) = P (∪Ek ). Convergence is in terms
of strong operator topology.
The standard Lebesgue integration extends to PVM.
hϕ, P (E)ϕi = hP (E)ϕ, P (E)ϕi = kP (E)ϕk2 ≥ 0
since P is countablely addative, the map E 7→ kP (E)ϕk2 is also countablely adda-
tive. Therefore, each ϕ ∈ H induces a regular Borel measure µϕ on the Borel
σ-algebra of R.
For a measurable function ψ,
ˆ ˆ
ψdµϕ = ψ(x)hϕ, P (dx)ϕi
ˆ
= hϕ, ψP (dx) ϕi
hence we may define ˆ

ψP (dx)
as the operator so that for all ϕ ∈ H,

ˆ
hϕ, ψP (dx) ϕi.
Remark 1.30. P (E) = F χE F −1 defines a PVM. In fact all PVMs come from
this way. In this sense, the Mt version of the spectral theorem is better, since it
implies the PVM version. However, the PVM version facilites some formuations in
quantum mechanics, so physicists usually prefer this version.
Remark 1.31. Suppose we start with the PVM version of the spectral theorem.
How to prove (ψ1 ψ2 )(A) = ψ1 (A)ψ2 (A)? i.e. how to chech we do have an algebra
isomorphism? Recall in the PVM version, ψ(A) is defined as the operator so that
for all ϕ ∈ H, we have ˆ
ψdµϕ = hϕ, ψ(A)ϕi.
As a stardard approximation technique, once starts with simple or even step func-
tions. Once it is worked out for simple functions, the extention to any measurable
functions is straightforward. Hence let’s suppose (WLOG) f
X
ψ1 = ψ1 (ti )χEi
X
ψ2 = ψ2 (tj )χEj
then
ˆ ˆ X
ψ1 P (dx) ψ2 P (dx) = ψ1 (ti )ψ2 (tj )P (Ei )P (Ej )
i,j
X
= ψ1 (ti )ψ2 (ti )P (Ei )
i
ˆ
= ψ1 ψ2 P (dx)
where we used the fact that P (A)P (B) = 0 if A ∩ B = φ.

As we delve into Nelson’s lecture notes, we notice that on page 69, there is
another unitary operator. By pieceing these operators together is precisely how we
get the spectral theorem. This “pieceing” is a vast generalization of Fourier series.
Lemma 1.32. pick ϕ ∈ H, get the measure µϕ where
µϕ (·) = hϕ, P (·)ϕi = kP (·)ϕk2
and we have the Hilbert space L2 (µϕ ). Take Hϕ := span{ψ(A)ϕ : ψ ∈ L∞ (µϕ )}.
Then the map
ψ(A)ϕ 7→ ψ
is an isometry, and it extends uniquely to a unitary operator from Hϕ to L2 (µϕ ).
To see this,
kψ(A)ϕk2 = hψ(A)ϕ, ψ(A)ϕi
= hϕ, ψ̄(A)ψ(A)ϕi
= hϕ, |ψ|2 (A)ϕi
ˆ
= |ψ|2 dµϕ .
Remark 1.33. Hϕ is called the cyclic space generated by ϕ. Before we can

construct Hϕ , we must make sense of ψ(A)ϕ.
Lemma 1.34. (Nelson p67) Let p = a0 + a1 x + · · · + an xn be a polynomial.
Then kp(A)uk ≤ max|p(t)|, where kuk = 1 i.e. u is a state.
Proof. M := span{u, Au, . . . , An u} is a finite dimensional subspace in H
(automatically closed). Let E be the orthogonal projection onto M . Then
p(A)u = Ep(A)Eu = p(EAE)u.
Since EAE is a Hermitian matrix on M , we may apply the spectral theorem for
finite dimensional space and get
X
EAE = λk Pλk
where λ′k s are eigenvalues associated with the projections Pλk . It follows that
X X
p(A)u = p( λk Pλk )u = p(λk )Pλk u
and
X 2
kp(A)uk2 = |p(λk )| kPλk uk2
X
≤ max|p(t)| kPλk uk2
= max|p(t)|
since X
kPλk uk2 = kuk2 = 1.
P
Notice that I = Pλk .
Remark 1.35. How to extend this? polynomials - continuous functions - mea-
surable functions. [−kAk, kAk] ⊂ R,
kEAEk ≤ kAk
is a uniform estimate for all truncations. Apply Stone-Weierstrass theorem to the
interval [−kAk, kAk] we get that any continuous function ψ is uniformly approxi-
mated by polynomials. i.e. ψ ∼ pn . Thus
kpn (A)u − pm (A)uk ≤ max|pn − pm |kuk = kpn − pm k∞ → 0
and pn (A)u is a Cauchy sequence, hence
lim pn (A)u =: ψ(A)u
n
where we may define the operator ψ(A) so that ψ(A)u is the limit of pn (A)u.
1.5.3. Convert Mf to PVM.
Theorem 1.36. Let A : H → H be a self-adjoint operator. Suppose A is
unitarily equivalent to the operator Mt of multiplication by the independent variable
on the Hilbert space L2 (µ). Then there exists a unique projection-valued measure
P so that ˆ
A = tP (dt)
i.e. for all h, k ∈ H, ˆ
hk, Ahi = thk, P (dt)ki.
Proof. The uniquess part follows from a standard argument. We will only
prove the existance of P . Let F : L2 (µ) → H be the unitary operator so that
A = F Mt F −1 . Define
P (E) = F χE F −1
for all E in the Borel σ-algebra B of R. Then P (φ) = 0, P (R) = I; for all
E1 , E2 ∈ B,
P (E1 ∩ E2 ) = F χE1 ∩E2 F −1
= F χE1 χE2 F −1
= F χE1 F −1 F χE2 F −1
= = P (E1 )P (E2 ).
Suppose {Ek } is a sequence of mutually disjoint elements in B. Let h ∈ H and
write h = F ĥ for some ĥ ∈ L2 (µ). Then
hh, P (∪Ek )hiH = hF ĥ, P (∪Ek )F ĥiH
= hĥ, F −1 P (∪Ek )F ĥiL2
= hĥ, χ∪Ek ĥiL2
ˆ
= |ĥ|2 dµ
∪Ek
Xˆ
= |ĥ|2 dµ
k Ek
X
= hh, P (Ek )hiH .
k
Therefore P is a projection-valued measure.

For any h, k ∈ H, write h = F ĥ and k = F k̂. Then
hk, AhiH = hF k̂, AF ĥiH
= hk̂, Mt ĥiL2
ˆ
= tk̂(t)ĥ(t)dµ(t)
ˆ
= thk, P (dt)hiH
´
Thus A = tP (dt).
Remark 1.37. In fact, A is in the closed (under norm or storng topology) span
of {P (E) : E ∈ B}. This is equivalent to say that t = F −1 AF is in the closed span
of the set of characteristic functions, the latter is again a standard approximation
in measure theory. It suffices to approximate tχ[0,∞] .
The wonderful idea of Lebesgue is not to partition the domain, as was the case
in Riemann integral over Rn , but instead the range. Therefore integration over an
arbitrary set is made possible. Important exmaples include analysis on groups.
Proposition 1.38. Let f : [0, ∞] → R, f (x) = x, i.e. f = xχ[0,∞] . Then there
exists a sequence of step functions s1 ≤ s2 ≤ · · · ≤ f (x) such that limn→∞ sn (x) =
f (x).
Proof. For n ∈ N, define

(
i2−n x ∈ [i2−n , (i + 1)2−n )
sn (x) =
n x ∈ [n, ∞]
where 0 ≤ i ≤ n2−n − 1. Equivalently, sn can be written using characteristic

functions as
n
n2
X −1
sn = i2−n χ[i2−n ,(i+1)2−n ) + nχ[n,∞] .
i=0
Notice that on each interval [i2−n , (i + 1)2−n ),

sn (x) ≡ i2−n ≤ x
sn (x) + 2−n ≡ (i + 1)2−n > x
sn (x) ≤ sn+1 (x).
Therefore, for all n ∈ N and x ∈ [0, ∞],
(1.5.1) x − 2−n < sn (x) ≤ x
and sn (x) ≤ sn+1 (x).
It follows from (1.5.1) that
lim sn (x) = f (x)
n→∞
for all x ∈ [0, ∞].
Corollary 1.39. Let f (x) = xχ[0,M] (x). Then there exists a sequence of step
functions sn such that 0 ≤ s1 ≤ s2 ≤ · · · ≤ f (x) and sn → f uniformly, as n → ∞.
Proof. Define sn as in proposition (1.38). Let n > M , then by construction
f (x) − 2−n < sn (x) ≤ f (x)
for all s ∈ [0, M ]. Hence sn → f uniformly as n → ∞.
Proposition (1.38) and its corollary immediate imply the following.

Corollary 1.40. Let (X, S, µ) be a measure space. A function (real-valued or
complex-valued) is measurable if and only if it is the pointwise limit of a sequence of
simple function. A function is bounded measurable if and only if it is the uniform
limit of a sequence of simple functions. Let {sn } be an approximation sequence
of simple functions. Then sn can be chosen such that |sn (x)| ≤ |f (x)| for all
n = 1, 2, 3 . . ..
Theorem 1.41. Let Mf : L2 (X, S, µ) → L2 (X, S, µ) be the operator of multi-
plication by f . Then,
(1) if f ∈ L∞ , Mf is a bounded operator, and Mf is in the closed span of the
set of self-adjoint projections under norm topology.
(2) if f is unbounded, Mf is an unbounded operator. Mf is in the closed span
of the set of self-adjoint projections under the strong operator topology.
Proof. If f ∈ L∞ , then there exists a sequence of simple functions sn so that

sn → f uniformly. Hence kf − sn k∞ → 0, as n → ∞.
1.6. SPECTRAL THEOREM FOR COMPACT OPERATORS 24
Suppose f is unbounded. By proposition () and its corollaries, there exists a

sequence of simple functions sn such that |sn (x)| ≤ |f (x)| and sn → f pointwisely,
as n → ∞. Let h be any element in the domain of Mf , i.e.
ˆ
|h| + |f h|2 dµ < ∞.
Then
2
lim |(f (x) − sn (x))h(x)| = 0
n→∞
and
|(f (x) − sn (x))h(x)|2 ≤ const|h(x)|2 .
Hence by the dominiated convergence theorem,
ˆ
2
lim |(f (x) − sn (x))h(x)| dµ = 0
n→∞
or equivalently,
k(f − sn )hk2 → 0
as n → ∞. i.e. Msn converges to Mf in the strong operator topology.
1.6. Spectral theorem for compact operators

CHAPTER 2
GNS, Representations
2.1. Representations, GNS, primer of multiplicity

2.1.1. Decomposition of Brownian motion. The integral kernel K : [0, 1]×
[0, 1] → R
K(s, t) = s ∧ t
is a compact operator on L2 [0, 1], where
ˆ
Kf (x) = (x ∧ y)f (y)dy.
Kf is a solution to the differential equation

d2
− u=f
dx2
with zero boundary conditions.
K is also seen as the covariance functions of Brownian motion process. A
stochastic process is a family of measurable functions {Xt } defined on some sample
probability space (Ω, B, P ), where the parameter t usually represents time. {Xt }
is a Brownian motion process if it is a mean zero Gaussian process such that
ˆ
E[Xs Xt ] = Xs Xt dP = s ∧ t.
Ω
It follows that the corresponding increament process {Xt − Xs } ∼ N (0, t − s). P

is called the Wiener measure.
Building (Ω, B, P ) is a fancy version of Riesz’s representation theorem as in
Theorem 2.14 of Rudin’s book. It turns out that
Y
Ω= R̄
t
which is a compact Hausdorff space.

Xt : Ω → R
is defined as
Xt (ω) = ω(t)
i.e. Xt is the continuous linear functional of evluation at t on Ω.
For Brownian motion, √ the increament of the process △Xt , in some statistical
sense, is proportional to △t. i.e.
p
△Xt ∼ △t.
It it this property that makes the set of differentiable functions have measure zero.
In this sense, the trajectory of Brownian motion is nowhere differentiable.
25
2.1. REPRESENTATIONS, GNS, PRIMER OF MULTIPLICITY 26
An very important application of the spectral theorem of compact operators in

to decompose the Brownian motion process.
X
Bt (ω) = λn sin(nπt)Zn (ω)
where X
s∧t = λn sin(nπt)
and Zn ∼ N (0, 1).
2.1.2. Idea of multiplicity. Recall that A = A∗ if and only if

ˆ
A= tP (dt)
sp(A)
where P is a projection-valued measure (PVM). The simplest example of a PVM

is P (E) = χE . The spectral theorem states that all PVMs come this way.
Let A be a positive compact operator on H. A is positive means hx, Axi ≥ 0
for all x ∈ H. The spectral theorem of A states that
X
A= λn Pn
where λ′n s are the eigenvalues of A, such that λ1 ≥ λ2 ≥ · · · λn → 0. and Pn′ s are
self-adjoin projections onto the corresponding finite dimensional eigenspace of λn .
P is a PVM supported on {1, 2, 3, . . .}, so that Pn = P ({n}).
Question: what does A look like if it is represented as the operator of multipli-
cation by the independent variable?
We may arrange the eigenvalues of A such that
s1 s2 sn
z }| { z }| { z }| {
λ1 = · · · = λ1 > λ2 = · · · = λ2 > · · · > λn = · · · = λn > · · · → 0.
We say that λi has multiplicity si . The dimension of the eigen space of λi is si ,
and X
dimH = si .
Example 2.1. We represent A as the operator Mf of multiplication by f on

L2 (X, µ). Let Ek = {xk,1 , . . . , xk,sk } ⊂ X, and let Hk = span{χ{xk,j } : j ∈
{1, 2, . . . , sk }}. Let
∞
X
f= λk χEk , (λ1 > λ2 > · · · > λn → 0)
k=1
Notice that χEk is a rank s1 projection. Mf is compact if and only if it is of the

given form.
Example 2.2. Follow the previous example, we represent A as the operator
Mt of multiplication by the independent variable on some Hilbert space L2 (µf ).
For simplicity, let λ > 0 and
f = λχ{x1 ,x2 } = λχ{x1 } + λχ{x2 }
i.e. f is compact since it is λ times a rank-2 projection; f is positive since λ > 0.
The eigenspace of λ has two dimension,
Mf χ{xi } = λχ{xi } , i = 1, 2.
2.1. REPRESENTATIONS, GNS, PRIMER OF MULTIPLICITY 27
Define µf (·) = µ ◦ f −1 (·), then

µf = µ({x1 })δλ ⊕ µ({x2 })δλ ⊕ cont. sp δ0
and
L2 (µf ) = L2 (µ({x1 })δλ ) ⊕ L2 (µ({x2 })δλ ) ⊕ L2 (cont. sp δ0 ).
Define U : L2 (µ) → L2 (µf ) by
(U g) = g ◦ f −1 .
U is unitary, and the following diagram commute.
Mf
L2 (X, µ) / L2 (X, µ)
U U
Mt

L2 (R, µf ) / L2 (R, µf )
To check U preserves the L2 -norm,

ˆ
kU gk2 = kg ◦ f −1 ({x})k2 dµf
= kg ◦ f −1 ({λ})k2 + kg ◦ f −1 ({0})k2
ˆ
= |g(x1 )|2 µ({x1 }) + |g(x2 )|2 µ({x2 }) + |g(x)|2 dµ
X\{x1 ,x2 }
ˆ
= |g(x)|2 dµ
X
To see U diagonalizes Mf ,
Mt U g = λg(x1 ) ⊕ λg(x2 ) ⊕ 0g(t)χX\{x1 ,x2 }
= λg(x1 ) ⊕ λg(x2 ) ⊕ 0
U Mf g = U (λg(x)χ{x1 ,x2 } )
= λg(x1 ) ⊕ λg(x2 ) ⊕ 0
Thus
Mt U = U Mf .
Remark 2.3. Notice that f should really be written as
f = λχ{x1 ,x2 } = λχ{x1 } + λχ{x2 } + 0χX\{x1 ,x2 }
since 0 is also an eigenvalue of Mf , and the corresponding eigenspace is the kernel
of Mf .
Example 2.4. diagonalize Mf on L2 (µ) where f = χ[0,1] and µ is the Lebesgue
measure on R.
Example 2.5. diagonalize Mf on L2 (µ) where
(
2x x ∈ [0, 1/2]
f (x) =
2 − 2x x ∈ [1/2, 1]
and µ is the Lebesgue measure on [0, 1].
Remark 2.6. see direc integral and disintegration of measures.
2.2. STATES, DUAL AND PRE-DUAL 28
In general, let A be a self-adjoint operator acting on H. Then there exists a

second Hilbert space K, a measure ν on R, and unitary transformation F : H →
L2K (R, ν) such that
Mt F = F A
for measurable function ϕ : R → K,
ˆ
kϕkL2K (ν) = kϕ(t)k2K dν(t) < ∞.
2.1.3. GNS. The multiplication version of the spectral theorm is an exmaple

of representation of the algebra of L∞ functions (or C(X)) as operators acting on
a Hilbert space.
π : L∞ → g(A) ∈ B(H)
where π(f g) = π(f )π(g) and π(f¯) = π(f )∗ .
The general question is representation of algebras B(H). The GNS construction
was developed about 60 years ago for getting representations from data in typical
applications, especially in quantum mechanics. It was developed independently by
I. Gelfand, M. Naimark, and I. Segal.
Let A be a ∗-algebra with identity. A representation of A is a map π on A, a
Hilbert space Hπ
π : A → B(Hπ )
so that
π(AB) = π(A)π(B)
π(A∗ ) = π(A)∗ .
The ∗ operation is given on A so that A∗∗ = A, (AB)∗ = B ∗ A∗ .
The question is given any ∗-algebra, where to get such a representation? The
answer is given by states. A state ω on A is a linear functional ω : A → C such that
ω(A∗ A) ≥ 0, ω(1A ) = 1.
For example, if A = C(X) where X is a compact Hausdorff space, then
ˆ
ω(f ) = f dµω
is a state, where µω is a Borel probability measure. In fact, in the abelian case,

states are Borel probability measures.
Theorem 2.7. (GNS) There is a bijection between states ω and representation
(π, H, Ω) where kΩk = 1, and
ω(A) = hΩ, π(A)Ωi.
2.2. States, dual and pre-dual

Banach space
• vector space over C
• norm k·k
• complete
Let V be a vector space.

Let V be a Banach space. l ∈ V ∗ if l : V → C such that
klk := sup |l(v)| < ∞.
kvk=1
Hahn-Banach theorem implies that for all v ∈ V , kvk 6= 0, there exists lv ∈ V ∗ such
that l(v) = kvk. The construction is to define lv on one vector, then use transfinite
induction to extend to all vectors in V . Notice that V ∗ is always complete, even V
is an incomplete normed space. i.e. V ∗ is always a Banach space.
V is embedded in to V ∗∗ (we always do this). The embedding is given by
v 7→ ψ(v) ∈ V ∗
where
ψ(v)(l) := l(v).
Example 2.8. Let X be a compact Hausdorff space. C(X) with the sup norm
is a Banach space. (lp )∗ = lq , (Lp )∗ = Lq , for 1/p + 1/q = 1 and p < ∞. If
1 < p < ∞, then (lp )∗∗ = lp , i.e. these spaces are reflexive. (l1 )∗ = l∞ , but (l∞ )∗
is much bigger than l1 . Also note that (lp )∗ 6= lq except for p = q = 2 where l2 is a
Hilbert space.
Hilbert space
• vector space over C
• norm forms an inner product h·, ·i
• complete with respect to k·k = h·, ·i1/2
• H∗ = H
• every Hilbert space has a basis (proved by Zorn’s lemma)
The identification H = H ∗ is due to Riesz, and the corresponding map is given by
h 7→ hh, ·i ∈ H ∗
This can also be seen by noting that H is unitarily equivalent to l2 (A) and the
latter is reflecxive.
Let H be a Hilbert space. The set of all bounded operators B(H) on H is a
Banach space. We ask two questions: What is B(H)∗ ? Is B(H) the dual space of
some Banach space?
The first question extremely difficult and we will discuss that later. We now
show that
B(H) = T1 (H)∗
where we denote by T1 (H) the trace class operators.
Let ρ : H → H be a compact self-adjoint operator. Assume ρ is positive, i.e.
hx, ρxi ≥ 0 for all x ∈ H. By the spectral theorem of compact operators,
X
ρ= λk Pk
where λ1 ≥ λ2 ≥ · · · → 0, and Pk is the finite dimensional eigenspace of λk . ρ is a

trace class operator, if
X
λk < ∞
P
in which case we may assume λk = 1.
In general, we want to get rid of the assumption that ρ ≥ 0. Hence we work,

instead, with ψ : H → H so that ψ is a trace class operator if
p
ψ∗ ψ
is a trace class operator.
Lemma 2.9. T1 (H) is a two-sided ideal in B(H).
(1) Let (en ) be an ONB in H. Define the trace of A ∈ T1 (H) as is
X
trace(A) = hen , Aen i
then trace(A) is independent of the choice of ONB in H.

(2) trace(AB) = trace(BA)
(3) T1 (H) is a Banach space with the trace norm kρk = trace(ρ).
Lemma 2.10. Let ρ ∈ T1 (H). Then
A 7→ trace(Aρ)
is a state on B(H).
Remark 2.11. Notice that Aρ ∈ T1 (H) for all A ∈ B(H). The map
A 7→ trace(Aρ)
is in B(H)∗ means that the dual pairing
hA, ρi := trace(Aρ)
satisfies
|hA, ρi| ≤ kAkoperator kρktrace .
Theorem 2.12. T1∗ (H) = B(H).
Let l ∈ T1∗ . How to get an operator A? It is supposed to be the case such that
l(ρ) = trace(ρA)
for all ρ ∈ T1 . How to pull an operator A out of the hat? The idea also goes back
to Paul Dirac. It is in fact not difficult to find A. Since A is determined by its
matrix, it suffices to fine
hf, Af i
which are the entries in the matrix of A. For any f1 , f2 ∈ H, the rank-one operator
|f1 ihf2 |
is in T1 , hence we know what l does to it, i.e. we know these numbers l(|f1 ihf2 |).
Since
l(|f1 ihf2 |)
is linear in f1 , and conjugate linear in f2 , by the Riesz theorem for Hilbert space,
there exists a unique operator A such that
l(|f1 ihf2 |) = hf2 , Af1 i.
This defines A.
2.3. EXAMPLES OF REPRESENTATIONS, PROOF OF GNS 31
Now we check that l(ρ) = trace(ρA) . Take ρ = |f1 ihf2 |. Then

trace(ρA) = trace(|f1 ihf2 | A)
= trace(A |f1 ihf2 |)
X
= hen , Af1 ihf2 , en i
n
= hf2 , Af1 i
where the last equality follows from Parseval’s indentity.
Remark 2.13. If B is the dual of a Banach space, then we say that B has a
predual. For example l∞ = (l1 )∗ , hence l1 is the predual of l∞ . Another example in
Rudin’s book, H 1 , hardy space of analytic functions on the disk. (H 1 )∗ = BM O,
where BM O referes to bounded mean oscillation. It was developed by Charles
Fefferman in 1974 who won the fields medal for this theory. Getting hands on a
specific dual space is often a big thing.
Let B be a Banach space and denote by B ∗ its dual space. B ∗ is a Banach
space as well, where the norm is defined by
kf kB ∗ = sup |f (x)|.
kxk=1
Let B1∗ = {f ∈ B : kf k ≤ 1} be the unit ball in B ∗ .

∗
Theorem 2.14. B1∗ is weak ∗ compact in B ∗ .

Q
Proof. This is proved by showing B1∗ is a closed subspace in kxk=1 C, where
the latter is given its product topology, and is compact and Hausdorff.
Corollary 2.15. Every bounded sequence in B ∗ has a convergent subsequence
in the W ∗ -topology.
Corollary 2.16. Every bounded sequence in B(H) contains a convergence
subsequence in the W ∗ -topology.
2.3. Examples of representations, proof of GNS

Example 2.17. Fourier algebra.
• discrete case: X
(a ∗ b)n = ak bn−k
k
involution
(a∗ )n = ā−n
Most abelian algebras can be thought of function algebras.
X
(an ) 7→ an z n
may specilize to z = eit , t ∈ R mod 2π. {F (z)} is an abelian algebra of
functions. X
F (z)G(z) = (a ∗ b)n z n
Homomorphism:
(l1 , ∗) → C(T 1 )
(an ) 7→ F (z).
2.3. EXAMPLES OF REPRESENTATIONS, PROOF OF GNS 32
If we want to write F (z) as power series, then we need to drop an for

n < 0. Then F (z) extends to an analytic function over the unit disk. The
sequence space
{a0 , a1 , . . .}
was suggested by Hardy. The Hardy space H2 is a Hilbert space. Rudin
has two beautiful chapters on H 2 . (see chapter 16)
• continuous case:
ˆ
(f ∗ g)(x) = f (s)g(t − s)ds.
Remark 2.18. C(T 1 ) is called the C ∗ -algebra completion of l1 . L∞ (X, B, µ) =

L (µ)∗ is also a C ∗ -algebra. It is called a W * algebra, or Von Neumann algebra.
1
The W * refers to the fact that its topology comes from the weak * topology. B(H),
for any Hilbert space, is a Von Neumann algebra.
Example 2.19. uf = eiθ f (θ), vf = f (θ − ϕ), restrict to [0, 2π], i.e. 2π periodic
functions.
vuv −1 = eiϕ u
vu = eiϕ uv
u, v generate C ∗ -algebra, noncummutative.
Example 2.20. (from quantum mechanics)
[p, q] = −iI
p, q generate an algebra. But they can not be represented by bounded operators.
May apply bounded functions to them and get a C ∗ -algebra.
Example 2.21. Let H be an infinite dimensional Hilbert space. H is iso-
metrically isomorphic to a subspace of itself. For example, let {en } be an ONB.
H1 = span{e2n }, H2 = span{e2n+1 }. Let
V1 (en ) = e2n
V2 (en ) = e2n+1
then we get two isometries.
V1 V1∗ + V2 V2∗ = I
Vi∗ Vi = I
Vi Vi∗ = Pi
where Pi is a self-adjoint projection, i = 1, 2. This is the Cuntz algebra O2 . More
general On . Cuntz showed that this is a simple algebra in 1977.
From algebras, get representations. For abelian algebras, we get measures. For non-
abelian algebras, we get representations, and the measures come out as a corallory
of representations.
Proof. (Sketch of the proof of GN S) Let w be a state on A. Need to construct
(π, H, Ω).
A is an algebra, and it is also a complex vector space. We pretend that A is a
Hilbert space, see what is needed for this.
We do get a homomorphism A → H which follows from the associative law of
A being an algebra, i.e. (AB)C = A(BC).
2.4. GNS, SPECTRAL THOERY 33
For Hilbert space H, need an inner product. Try

hA1 , A2 i = w(A∗1 A2 )
Then hA1 , A2 i is linear in A2 , conjugate linear in A1 . It also satisfies
hA1 , A1 i ≥ 0
which is built into the definition of a state. But it may not be positive definite.
Therefore we take
H := [H/{A : w(A∗ A) = 0}]cl .
Lemma 2.22. {A : w(A∗ A) = 0} is a closed subspace of A.
Proof. this follows from the Schwartz inequality.
Let π : A → H such that π(A) = A/ker. Let Ω = π(I). Therefore

H = span{π(A)Ω : A ∈ A}.
To see we do get a representation, take two typical vectors π(B)Ω and π(A)Ω in
H, then
hπ(B)Ω, π(C)π(A)Ωi = hΩ, π(B ∗ CA)Ωi
= w(B ∗ CA)
= w((C ∗ B)∗ A)
= hπ(C ∗ B)Ω, π(A)Ωi
= hπ(C ∗ )π(B)Ω, π(A)Ωi
it follows that
hv, π(C)ui = hπ(C ∗ )v, ui
for any u, v ∈ H. Therefore,
π(C)∗ = π(C ∗ ).

Example 2.23. C[0, 1], a 7→ a(0), a∗ a = |a|2 , hence w(a∗ a) = |a|2 (0) ≥ 0.
ker = {a : a(0) = 0}
and C/ker is one dimensional. The reason if ∀f ∈ C[0, 1] such that f (0) 6= 0, we
have
f (x) ∼ f (0)
because f (x) − f (0) ∈ ker, where f (0) represent the constant function f (0) over
[0, 1]. This shows that w is a pure state, since the representation has to be irre-
ducible.
2.4. GNS, spectral thoery

Remark 2.24. some dover books
• Stefan Banach, Theory of linear operators
• Howard Georgi, weak interactions and modern particle theory
• P.M. Prenter, splines and variational methods
2.4.1. GNS for C ∗ -algebras.

Definition 2.25. A representation π ∈ Rep(A, H) is cyclic if it has a cyclic
vector.
Theorem 2.26. Give a representation π ∈ Rep(A, H), there exists an index
set J, closed subspaces Hj ⊂ H such that
• (orthogonal) Hi ⊥ Hj , ∀i 6= j
P⊕
• (total) j∈J Hj = H, vj ∈ Hj such that the restriction of π to Hj is
cyclic with cyclic vector vj . Hence on each Hj , π is abelian.
Remark. This looks like the construction of orthonormal basis. But it’s a fam-
ily of mutually orthogonal subspaces. Of course, if Hj is one-dimensional for all j,
then it is a decomposition into an ONB. Not every representation is irreducible, but
every representation can be decompossed into direct sum of cyclic representations.
We use Zorn’s lemma to show total, exactly the same argument for the existance
of an ONB of any Hilbert space.
Theorem 2.27. (Gelfand-Naimark) Every C ∗ -algebra is isometrically isomor-
phic to a norm-closed sub-algebra of B(H), for some Hilbert space H.
Proof. Let A be any C ∗ -algebra, no Hilbert space H is given from outside.
Let S(A) be the states on A, which is a compact convex subset of A∗ . Compactness
refers to the weak * topology.
We use Hahn-Banach theorem to show that there are plenty of states. Specifi-
cally, ∀A ∈ A, ∃w such that w(A) > 0. This is done first on 1-dimensional subspace,
tA 7→ t ∈ R
then extend to A. This is also a consequence of Krein-Millman, i.e. S(A) =
cl(P ureStates). We will come back to this point later in the course.
For each state w, get a cyclic representation (πw , Hw , Ωw ), such that π = ⊕πw
is a representation on the Hilbert space H = ⊕Hw .
Theorem 2.28. A abelian C ∗ -algebra. Then A ∼
= C(X) where X is a compact
Hausdorff space.
Remark. Richard Kadison in 1950’s reduced the axioms of C ∗ -algebra from
about 6 down to just one on the C ∗ -norm,
kA∗ Ak = kAk2 .
2.4.2. Finishing spectral theorem on Nelson. We get a family of states
wj ∈ S, corresponding measures µj , and Hilbert spaces Hj = L2 (µj ). Note that all
the L2 spaces are on K = sp(A). So it’s the same underlying set, but with possibly
different measures.
To get a single measure space with µ, Nelson suggests taking the disjoint union
[
K DS = K × {j}
j
DS
and µ is the disjoint union of µ′j s. The existance of µ follows from Riesz. Then
we get
H = ⊕Hj 7→ L2 (K DS , µDS ).
Notice that this map is into but not onto.
The multiplicity theory starts with breaking up each Hj into irreducible com-
ponents.
2.4.3. Examples on disintegration.
Example 2.29. L2 (I) with Lebesgue measure. Let
(
1 t≥x
Fx (t) =
0 t<x
Fx is a monotone increasing function on R, hence by Riesz, we get the corresponding
Riemann-Stieljes measure dFx .
ˆ ⊕
dµ = dFx (t)dx.
i.e. ˆ ˆ ˆ
f dµ = dFx (f )dx = f (x)dx.
Equivalently, ˆ
dµ = δx dx
i.e. ˆ ˆ ˆ
f dµ = δx (f )dx = f (x)dx.
µ is a state, δx = dFx (t) is a pure state, ∀x ∈ I. This is a decomposition of state
into direct integral of pure states.
Q
Example 2.30. Ω = t≥0 R̄, Ωx = {w ∈ Ω : w(0) = x}. Kolmogorov gives
rise to Px by conditioning P with respect to “starting at x”.
ˆ ⊕
P = Px dx
i.e. ˆ
P () = P (·|start at x)dx.
Example 2.31. Harmonic function on D

ˆ
h 7→ h(z) = ()dµz
Poisson integration.
2.4.4. Noncommutative Radon-Nicodym derivative. Let w be a state,
K is an operator. √ √
w( KA K)
wk (A) =
w(K)
is a state, and wk ≪ w, i.e. w(A) = 0 ⇒ wK (A) = 0. K = dw/dwK .
Check:
wK (1) = 1
√ √
w( KA∗ A K)
wK (A∗ A) =
w(K)
√ ∗ √
w((A K) (A K))
= ≥0
w(K)
2.5. CHOQUET, KREIN-MILMAN, DECOMPOSITION OF STATES 36
see Sakai - C ∗ and W ∗ algebras.
2.5. Choquet, Krein-Milman, decomposition of states

The main question here is how to break up a representation into smaller ones.
The smallest representations are the irreducible ones. The next would be multi-
plicity free representation.
Let A be an algebra.
• commutative: e.g. function algebra
• non-commutative: matrix algebra, algebras generated by representation
of non-abelian groups
Smallest representation:
• irreducible - π ∈ Repirr (A, H) where H is 1-dimensional. This is the
starting point of further analysis
• multiplicity free - π ∈ Rep(A, H) assuming cyclic, since otherwise π =
⊕πcyc . π is multiplicity free if and only if A′ is abelian. In general A′ may
or may not be abelian.
Let C ⊂ B(H) be a ∗-algebra, C′ = {X : H → H|XC = CX, ∀C ∈ C}. C′ is also a
∗-algebra. It is obvious that C ⊂ C′ if and only if C is abelian.
Definition 2.32. Let π ∈ Rep(A, H). multi(π) = n ⇔ π(A)′ ≃ Mn (C).
Example. Let  
1
I2 0
A= 1 = .
0 2
2
Then AC = CA if and only if
 
a b
B 0
C = c d =
0 1
1
where B ∈ M2 (C).
2.5.1. Decomposition of states. Let A be a commutative C ∗ -algebra con-
taining identity. Gelfand-Naimark’s theorem for commutative C ∗ -algebra says that
A ≃ C(X), where X is a compact Hausdorff space. X is called the Gelfand space
or the spectrum of A. In general, for Banach ∗-algebra, also get X.
What is X?
Let A be a commutative Banach ∗-algebra. The states S(A) is a compact
∗
convec
Q non empty set in the dual A , which is embedded into a closed subset
of A R̄. w is a state if w(1) = 1; w(A∗ A) ≥ 0; and w is linear. Note that
convex linear combination of states are also states. i.e. if w1 , w2 are states, then
w = tw1 + (1 − t)w2 is also a state.
Note. dual of a normed vector space has its unit ball being weak ∗-compact.
Theorem 2.33. (Krein-Milman) Let K be a compact convec set in a locally
compac topological space. Then K is equal to the closed convex hull of its extreme
points. i.e.
K = conv(E(K)).
2.5. CHOQUET, KREIN-MILMAN, DECOMPOSITION OF STATES 37
Note. The dual of a normed vector space is always a Banach space, so the
theorem applies. The convex hull in an infinite dimensional space is not always
closed, so close it. A good reference to locally convex topological space is TVS by
F. Treves.
The decompsition of states into pure states was developed by R.Phelps for
representation theory. The idea goes back to Choquet.
Theorem 2.34. (Choquet) Let w ∈ K = S(A), there exists a measure µw
“concentrated” on E(K), such that for affine function f
ˆ
f (w) = f dµw .
”E(K)”
Note. E(K) may not be Borel. In this case replace E(K) be Borel set V ⊃
E(K) s.t.
µw (V \E(K)) = 0.
This is a theorem of Glimm. Examples for this case include the Cuntz algebra, free
group with 2 generators, uvu−1 = u2 for wavelets.
Note. µw may not be unique. If it is unique, K is called a simplex. The
unit disk has its boundary as extreme points. But representation of points in the
interior using points on the boundary is not unique. Therefore the unit disk is not
a simplex. A triangle is.
Proof. (Krein-Milman) If K % conv(E(K)), get a linear functional w, such
that w is zero on conv(E(K)) and not zero on w ∈ K\conv(E(K)). Extend by
Hahn-Banach theorem to a linear functional to the whole space, and get a contra-
diction.
2.5.2. The Gelfand space X. Back to the question of what the Gelfand
space X is.
Let A be a commutative Banach ∗-algebra. Consider the closed ideals in A
(since A is normed, so consider closed ideals) ordered by inclusion. By zorn’s
lemma, there exists maximal ideals M . A/M is 1-dimensional, hence A/M = {tv}
for some v ∈ A and t ∈ R. Therefore
A → A/M → C
the combined map ϕ : a 7→ a/M 7→ ta is a homomorphism. A ∋ 1 7→ v := 1/M ∈
A/M then ϕ(1) = 1.
Conversely, the kernel of a homomorphism ϕ is a maximal ideal in A. Therefore
there is a bijection between maximal ideas and homomorphisms. Note that if
ϕ : A → C is a homomorphism then it has to be a contraction.
Let X be the set of all maximal ideals ∼ all homomorphisms in A∗1 , where A∗1
is the unit ball in A∗ . Since A∗1 is compact, X is closed in it, X is also compact.
The Gelfand transform F : A → C(X) is defined by
F (a)(ϕ) = ϕ(a)
then
A/kerF ≃ C(X)
(mod the kernel for general Banach algebras)
2.6. BEGINNING OF MULTIPLICITY 38
Example 2.35. l1 (Z), the convolution algebra.

X
(ab)n = ak bn−k
k
a∗n = ā−n
X
kak = |an |
n
To identity X in practice, always start with a guess, and usually it turns out to be
correct. Since Fourier transform converts convolution to multiplication,
X
ϕz : a 7→ an z n
is a complext homormorphism. To see ϕz is multiplicative,
X
ϕz (ab) = (ab)n z n
X
= ak bn−k z n
n,k
X X
= ak z k bn−k z n−k
k n
! !
X X
k k
= ak z bk z .
k k
Thus {z : |z| = 1} is a subspace in the Gelfand space X. Note that we cannot

use |z| < 1 since we are dealing with two-sided l1 sequence. If the sequences were
trancated, so that an = 0 for n <P 0 then we Pallow |z| < 1.
ϕz is contractive: |ϕz (a)| = | an z n | ≤ n |an | = kak.
It turns out that every homorphism is obtained as ϕz for some |z| = 1, hence
X = {z : |z| = 1}.
Example 2.36. l∞ (Z), with kak = supn |an |. The Gelfand space in this case
is X = βZ, the Stone-Cech compactification of Z, which are the ultrafilters on
Z. Pure states on diagonal operators correspond to βZ. βZ is much bigger then
p−adic numbers.
2.6. Beginning of multiplicity

We study Nelson’s notes. Try to get the best generalization from finite dimen-
sional linear algebra. Nelson’s idea is to get from self-adjoint operators → cyclic
representation of function algebra → measure µ → L2 (µ).
Start with a single self-adjoint operator A acting on an abstract Hilbert space
H. H can be broken up into a direct sum of mutually disjoint cyclic spaces.
Let u ∈ H. {f (A)u}, f runs through some function algebra, generates a sub-
space Hu ⊂ H. The funciton algebra might be taken as the algebra of polynomials,
then later it is extended to a much bigger algebra contaning polynimials as a dense
sub-algebra.
The map f 7→ wu (f ) := hu, f (A)ui is a state on polynomials, and it extends to
a state on Cc (R). By Riesz, there exists a unique mesure µu such that
ˆ
wu (f ) = hu, f (A)ui = f dµ
R
2.6. BEGINNING OF MULTIPLICITY 39
It turns out that µw is supported on [0, kAk], assuming A ≥ 0. If A is not positive,

it can be written as the positive part and the negative part. Therefore we get L2 (µ),
a Hilbert space containing polyomials as a dense subspace. Let Hu = span{f (A)u}
for all polynomials f . Define W : Hu → L2 (µ), such that W f (A)u = f .
Lemma 2.37. (1) W is well-defined, isometric; (2) W A = Mt W , i.e. W
intertwines A and Mt , and W diagnolizes A.
Note. W A = Mt W ⇔ W AW ∗ = Mt . In finite dimension, it is less emphasized
that the adjoint W ∗ equals the inverse W −1 P. For finite dimensional case, Mt =
diag(λ1 , λ2 , . . . λn ) where the measure µ = δλi .
Proof. For (1),
ˆ
|f |2 dµ = hu, |f |2 (A)ui
= hu, f¯(A)f (A)ui

= hu, f (A)∗ f (A)ui
= hf (A)u, f (A)ui
= kf (A)uk2 .
Notice that strictly speaking, f (A∗ )∗ = f¯(A). Since A∗ = A, therefore f (A∗ )∗ =
f¯(A). π(f ) = f (A) is a representation. i.e. π(f¯) = f (A)∗ .
For (2), let f be a polynomial, where f (t) = a0 + a1 t + · · · an tn . Then
W Af (A) = W A(a0 + a1 A + a2 A2 + · · · + an An )
= W (a0 A + a1 A2 + a2 A3 + · · · + an An+1 )
= a0 t + a1 t2 + a2 t3 + · · · + an tn+1
= tf (t)
= Mt W f (A)
thus W A = Mt W .
The whole Hilbert space H then decomposes into a direct sum of mutually
orthogonal cyclic subspaces, each one is unitarily equivalent to L2 (µu ) for some
P⊕
cyclic vector u. This representation of L∞ onto H = Hu is highly non unique.
There we enter into the multiplicity theory.
Each cyclic representation is multiplicity free. i.e. π ∈ Rep(L∞ (µ), Hu ) is
multiplicity if and only if (π(L∞ ))′ is abelian. In general, (π(L∞ ))′ is not abelian,
and we say π has multiplicity equal to n if and only if (π(L∞ ))′ ≃ Mn (C). This
notation of multiplicity free generalizes the one in finite dimensional linear algebra.
Example 2.38. Mt : L2 [0, 1] → L2 [0, 1], Mt has no eigenvalues.
Example 2.39. Mϕ : L2 (µ) → L2 (µ), with ϕ ∈ L∞ (µ). Claim: π ∈ Rep(L∞ (µ), L2 (µ))
is multiplicity free, i.e. it is maximal abelian. Suppose B commutes with all Mϕ ,
define g = B1. Then
Bψ = Bψ1 = BMψ 1 = Mψ B1 = Mψ g = ψg = gψ = Mg ψ
thus B = Mg .
Examples that do have multiplicties in finite dimensional linear algebra:
2.7. COMPLETELY POSITIVE MAPS 40
Example 2.40. 2-d, λI, {λI}′ = M2 (C) which is not abelian. Hence mult(λ) =
2.
Example 2.41. 3-d,
 
λ1
λ1 I
 λ1 =
λ2
λ2
where λ1 6= λ2 . The commutatant is

B
b
where B ∈ M2 (C), and b ∈ C. Therefore the commutant is isomorphic to M2 (C),
and multiplicity is equal to 2.
Example 2.42. The example of Mϕ with repeteation.
Mϕ ⊕ Mϕ : L2 (µ) ⊕ L2 (µ) → L2 (µ) ⊕ L2 (µ)

Mϕ f1 ϕf1
=
Mϕ f2 ϕf2
the commutant is this case is isomorphic to M2 (C). If we introduces tensor product,
then representation space is also written asL2 (µ) ⊗ V2 , the multiplication operator
is amplified to Mϕ ⊗ I, whose commutant is represented as I ⊗ V2 . Hence it’s clear
that the commutatant is isomorphic to M2 (C). To check
(ϕ ⊗ I)(I ⊗ B) = ϕ⊗B
(I ⊗ B)(ϕ ⊗ I) = ϕ ⊗ B.
2.7. Completely positive maps

The GNS construction gives a bijection between states and cyclic representa-
tions. An extention to the GNS construction is Stinespring’s completely positive
map. It appeared in an early paper by Stinspring in 1956 (PAMS). Arveson in
1970’s (Arveson 1972) reformuated Stingspring’s result using tensor product. He
realized that complemetly positive maps are the key in multivariable operator the-
ory, and in noncommutative dynamics.
2.7.1. Motivations. Let A be a ∗-algebra with identity. w : A → C is a a

1, w(A∗ A) ≥ 0. If A was a C ∗ -algebra, A ≥ 0 ⇔ sp(A) ≥ 0, hence
state if w(1) = √
may take B = A and A = B ∗ B. Given a state w, the GNS construction gives a
Hilbert space K, a cyclic vector Ω ∈ K, and a representation π : A → B(K), such
that
w(A) = hΩ, π(A)Ωi
K = span{π(A)Ω : A ∈ A}
Moreover, the Hilbert space is unique up to unitary equivalence.
Stinespring modified a single axiom in the GNS construction. In stead of a
state w : A → C, Stinespring considered a positive map ϕ : A → B(H), i.e. ϕ maps
positive elements in A to positive operators in B(H). ϕ is a natural extention of
w, since C can be seen as a 1-dimensional Hilbert space, and w is a positive map
w : A → B(C). He soon realized that the ϕ being a positive map is not enough to
produce a Hilbert space and a representation. It turns out that the condition to
put on ϕ is complete positivity, in the sense that, for all n ∈ N
(2.7.1) ϕ ⊗ IMn : A ⊗ Mn → B(H ⊗ Cn )
maps positive elements in A ⊗ Mn to positive operators in B(H ⊗ Cn ). ϕ is called

a completely positive map, or a CP map. CP maps are developed primarily for
nonabelian algebras.
The algebra Mn of n × n matrices can be seen as an n2 -dimensional Hilbert
space with an ONB given by the matrix units {eij }ni,j=1 . It is also a ∗-algebra
generated by {eij }ni,j=1 such that
(
eil j = k
eij ekl =
0 j 6= k
Members of A ⊗ Mn are of the form

X
Aij ⊗ eij .
i,j
In other words, A ⊗ Mn consists of precisely the A-valued n × n matrices. Similarly,

members of H ⊗ Cn are the n-tuple column vectors with H-valued entries.
Let IMn : Mn → B(Cn ) be the identity representation of Mn onto B(Cn ).
Then,
ϕ ⊗ IMn : A ⊗ Mn → B(H) ⊗ B(Cn ) = B(H ⊗ Cn )

X X
ϕ ⊗ IMn ( Aij ⊗ eij ) = ϕ(Aij ) ⊗ eij
i,j i,j
where the right-hand-side is an n × n B(H)-valued matrix.
Note 2.43. The algebra B(Cn ) of bounded operators on Cn is generated by

the rank-one operators IMn (eij ) = |ei ihej |. Hence the eij on the left-hand-side is
seen as an element in the algebra Mn of n×n matrices, while on the right-hand-side,
eij is seen as the rank one operator |ei ihej | B(Cn ). With Dirac’s notation, when
we look at eij as operators,
(
|ei i j = k
ei,j (ek ) = |ei ihej | |ek i =
0 j 6= k
(
|ei ihel | j = k
ei,j ekl = |ei ihej | |ek ihel | =
0 j=6 k
which also proves that IMn is in fact an algebra isomorphism.

P
The complete positivity condition in (2.7.1) is saying that if i,j Aij ⊗ eij is
a positive element in the algebra A ⊗ Mn , then the n × n B(H)-valued matrix
P n
i,j ϕ(Aij ) ⊗ eij is a positive operator acting on the Hilbert space H ⊗ C , i.e.
P
take any v = k vk ⊗ ek in H ⊗ Cn ,
X X X
h vl ⊗ el , ( ϕ(Aij ) ⊗ eij )( vk ⊗ ek )i
l i,j k
X X
= h vl ⊗ el , ϕ(Aij )vk ⊗ eij (ek )i
l i,j,k
X X
= h vl ⊗ el , ϕ(Aij )vj ⊗ ei i
l i,j
X
= hvl , ϕ(Aij )vj ihel , ei i
i,j,l
X
= hvi , ϕ(Aij )vj i ≥ 0.
i,j
The CP condition is illustrated in the following diagram.

(
A → B(H) : A 7→ ϕ(A)
⊗
Mn → Mn : x 7→ IMn (X) = X (identity representation of Mn )
The CP condition can be formulated more conveniently using matrix notation: if
the operator matrix (Aij ) ∈ A ⊗ Mn is positive, the corresponding operator matrix
(ϕ(Aij )) is a positive operator in B(H ⊗ Cn ); i.e. for all v ∈ H ⊗ Cn ,
X X X
h vl ⊗ el , ( ϕ(Aij ) ⊗ eij )( vk ⊗ ek )i
l i,j k
  
ϕ(A11 ) ϕ(A12 ) ··· ϕ(A1n ) v1

 ϕ(A21 ) ϕ(A22 ) ··· ϕ(A2n ) 
 v2 

= v1 v2 · · · vn  .. .. .. ..  ..  ≥ 0.
 . . . .  . 
ϕ(An1 ) ϕ(An2 ) · · · ϕ(Ann ) vn
2.7.2. CP v.s. GNS. The GNS construction can be reformulated as a special
case of the Stinespring’s theorem. Let A be a ∗-algebra, given a state w : A → C,
there exists a triple (Kw , Ωw , πw ), such that
w(A) = hΩw , π(A)Ωw iw
Kw = span{π(A)Ωw : A ∈ A}.
The unit cyclic vector Ω ∈ Kw generates a one-dimensional subspace CΩw ⊂ Kw .
The 1-dimensional Hilbert space C is thought of being embedded into Kw (possibly
infinite dimensional) via the map V : C 7→ CΩw .
Lemma 2.44. V is an isometry. V ∗ V = IC : C → C, V V ∗ : Kw → CΩw is the
projection from Kw onto the 1-d subspace CΩ in Kw .
Proof. Let t ∈ C, then kV tkw = ktΩw kw = |t|. Hence V is an isometry. For
all ξ ∈ Kw ,
hξ, V tiw = hV ∗ ξ, tiC = tV ∗ ξ ⇐⇒ tV ∗ ξ = hξ, V tiw
by setting t = 1, we get
V ∗ ξ = hξ, V 1iw = hξ, Ωw iw = hΩw , ξiw ⇐⇒ V ∗ = hΩw , ·i
Therefore,
V ∗ V t = V ∗ (tΩw ) = hΩw , tΩw i = t, ∀t ∈ C ⇐⇒ V ∗ V = IC
V V ∗ ξ = V (hΩw , ξiw ) = hΩw , ξiw Ωw , ∀ξ ∈ Kw ⇐⇒ V V ∗ = |Ωw ihΩw | .

It follows that
w(A) = hΩw , π(A)Ωw iw
= hV 1, π(A)V 1iC
= h1, V ∗ π(A)V 1iC
= V ∗ π(A)V.
In other words, π(A) : Ωw 7→ π(A)Ωw sends the unit vector Ωw from the 1-
dimensional subspace CΩw to the vector π(A)Ωw ∈ Kw , and hΩw , π(A)Ωw iw cuts
off the resulting vector π(A)Ωw and only preserves component corresponding to the
1-d subspace CΩw . Notice that the unit vector Ωw is obtained from embedding the
constant 1 ∈ C via the map V , Ωw = V 1. In matrix notation, if we identify C with
its image CΩw in Kw , then w(A) is put into a matrix corner.

w(A) ∗
π(A) =
∗
so that when acting on vectors,

w(A) ∗ Ωw Ωw
w(A) = Ωw 0 = Ωw 0 π(A) .
∗ 0 0
Equivalently
w(A) = P1 π(A) : P1 Kw → C
∗
where P1 = V V is the rank-1 projection onto CΩw .
Stinespring’s construction is a generalization of the above formulation. Let A
be a ∗-algebra, given a CP map ϕ : A → B(H), there exists a Hilbert space Kϕ ,
an isometry V : H → Kϕ , and a representation πw : A → Kϕ , such that
ϕ(A) = V ∗ π(A)V, ∀A ∈ A.
Notice that this construction start with a possibly infinite dimensional Hilbert
space H (instead of the 1-dimensional Hilbert space C), the map V embeds H into
a bigger Hilbert space Kϕ . If H is identified with its image in K, then ϕ(A) is put
into a matrix corner,

π(A) ∗ ϕ(A) ∗
should this be ?
∗ ∗
so that when acting on vectors,

π(A) ∗ Vξ
ϕ(A)ξ = V ξ 0 .
∗ 0
Stinespring’s theorem can then be formulated alternatively as: for every CP map
ϕ : A → B(H), there is a dilated Hilbert space Kϕ ⊃ H, a representation πϕ : A →
B(Kϕ ), such that
ϕ(A) = PH π(A)
i.e. π(A) can be put into a matrix corner. Kϕ is be chosen as minimal in the sense
that
Kϕ = span{π(A)(V h) : A ∈ A, h ∈ H}.
V /K
H(A)
ϕ π(A)

H /K
V
Note 2.45. Kϕ ⊃ H comes after the identification of H with its image in Kϕ

under an isometric embedding V . We use ϕ(A) = PH π(A) instead of “ϕ(A) =
PH π(A)PH ”, since ϕ(A) only acts on the subspace H ⊂ Kϕ .
2.7.3. Stinespring’s theorem.
Theorem 2.46. (Stinespring) Let A be a ∗-algebra. The following are equiva-
lent:
(1) ϕ : A → B(H) is a completely positive map, and ϕ(1A ) = IH .
(2) There exists a Hilbert space K, an isometry V : H → K, and a represen-
tation π : A → B(K) such that
ϕ(A) = V ∗ π(A)V
for all A ∈ A.
(3) If the dilated Hilbert space is taken to be minimum, then it is unique up
to unitary equivalence. i.e. if there are Vi , Ki , πi such that
ϕ(A) = Vi∗ πi (A)Vi
Ki = span{πi (A)V h : A ∈ A, h ∈ H}
then there exists a unitary operator W : K1 → K2 so that
W π1 = π2 W
Proof. For the uniqueness: define
W π1 (A)V h = π2 (A)V h
then W is an isometry, which follows from the fact that
kπi (A)V hk2K = hπi (A)V h, πi (A)V hiK
= hh, V ∗ πi (A∗ A)V hiH
= hh, ϕ(A∗ A)hiH .
Hence W extends uniquely to a unitary operator from K1 to K2 . To show W
intertwines πi , notice that a typical vector in K1 is π1 (A)V h, hence
W π1 (B)π1 (A)V h = W π1 (BA)V h
= π2 (BA)V h
= π2 (B)π2 (A)V h
= π2 (B)W π1 (A)V h
therefore W π1 = π2 W .
Note 2.47. The norm of π1 (B)π1 (A)V h is given by
kπ1 (B)π1 (A)V hk2 = hπ1 (B)π1 (A)V h, π1 (B)π1 (A)V hi
= hh, V ∗ π1 (A∗ B ∗ BA)V hi
where
operator 7→ A∗ (operator)A
is an automorphism on B(H).
Suppose ϕ(A) = V ∗ π(A)V . Since positive elements in A ⊗ Mn are sums of the
operator matrix
 ∗ 
A1
X  A∗2 
A∗i Aj ⊗ eij =  .  A1 A2 · · · An
 
i,j
 .. 
A∗n
it suffices to show that
X X X
ϕ ⊗ IMn ( A∗i Aj ⊗ eij ) = ϕ(A∗i Aj ) ⊗ IMn (eij ) = ϕ(A∗i Aj ) ⊗ eij
i,j i,j i,j
is a positive operator in B(H ⊗ Cn ), i.e. to show that for all v ∈ H ⊗ Cn

  
ϕ(A∗1 A1 ) ϕ(A∗1 A2 ) · · · ϕ(A∗1 An ) v1
∗ ∗ ∗
 ϕ(A2 A1 ) ϕ(A2 A2 ) · · · ϕ(A2 An )   v2 

 
v1 v2 · · · vn  .. .. . .   .  ≥ 0.
 . . .. ..   .. 
ϕ(A∗n A1 ) ϕ(A∗n A2 ) · · · ϕ(A∗n An ) vn
This is true, since
  
ϕ(A∗1 A1 ) ϕ(A∗1 A2 ) ··· ϕ(A∗1 An ) v1

 ϕ(A∗2 A1 ) ϕ(A∗2 A2 ) ··· ϕ(A∗2 An ) 
 v2 

v1 v2 · · · vn  .. .. .. ..  .. 
 . . . .  . 
ϕ(A∗n A1 ) ϕ(A∗n A2 ) · · · ϕ(A∗n An ) vn
X
= hvi , ϕ(A∗i Aj )vj iH
i,j
X
= hvi , V ∗ π(A∗i Aj )V vj iH
i,j
X
= hπ(Ai )V vi , π(Aj )V vj iϕ
i,j
X
= k π(Ai )V vi kϕ ≥ 0.
i
Conversely, given a completely positive map ϕ, we construct Kϕ , Vϕ and πϕ

where the subscript indicates dependence on ϕ. Recall that ϕ : A → B(H) is a CP
map if for all n ∈ N,
ϕ ⊗ IMn : A ⊗ Mn → B(H ⊗ Mn )
is positive, and
ϕ ⊗ IMn (1A ⊗ IMn ) = IH ⊗ IMn .
The condition on the identity element can be stated using matrix notation as
    
ϕ 0 ··· 0 1A 0 · · · 0 IH 0 ··· 0
 0 ϕ ··· 0   0 1A · · · 0   0 IH · · · 0 
    
 .. .. . . ..   .. .. . . ..  =  .. .. .. ..  .
 . . . .   . . . .   . . . . 
0 0 ··· ϕ 0 0 · · · 1A 0 0 · · · IH
Let K0 be the algebraic tensor product A ⊗ H, i.e.

X
K0 = span{ Ai ⊗ ξi : A ∈ A, ξ ∈ H}.
f inite
Define a sesquilinear form h·, ·iϕ : K0 × K0 → C,

Xn n
X X
(2.7.2) h Ai ⊗ ξi , Bj ⊗ ηj iϕ := hξi , ϕ(A∗i Bj )ηj iH .
i=1 j=1 i,j
By the completely positivity condition (2.7.1),

Xn n
X X
h Ai ξi , Aj ξj iϕ = hξi , ϕ(A∗i Aj )ξj iH ≥ 0
i=1 j=1 i,j
hence h·, ·iϕ is positive semi-definite. Let ker := {v ∈ K0 : hv, viϕ = 0}. Since the
Schwartz inequality holds for any sesquilinear form, it follows that ker = {v ∈ K0 :
hs, viϕ = 0, for all s ∈ K0 }, thus ker is a closed subspace in K0 . Let Kϕ be the
1/2
Hilbert space by completing K0 under the norm k·kϕ := h·, ·iϕ .
Define V : H → K0 , where V ξ := 1A ⊗ ξ. Then
kV ξk2ϕ = h1A ⊗ ξ, 1A ⊗ ξiϕ
= hξ, ϕ(1∗A 1A )ξiH
= hξ, ξiH
= kξk2H
which implies that V is an isometry, and H is isometrically embeded into K0 .
Claim: V ∗ V = IH ; V V ∗ is a self-adjoint projection from K0 onto the subspace
1A ⊗ H. In fact, for any A ⊗ η ∈ K0 ,
hA ⊗ η, V ξiϕ = hA ⊗ η, 1A ⊗ ξiϕ
= hη, ϕ(A∗ )ξiH
= hϕ(A∗ )∗ η, ξiH
= hV ∗ (A ⊗ η), ξiϕ
therefore,
V ∗ (A ⊗ η) = ϕ(A∗ )∗ η.
It follows that
V ∗ V ξ = V ∗ (1A ⊗ ξ) = ϕ(1∗A )∗ ξ = ξ, ∀ξ ∈ H ⇔ V ∗ V = IH .
Moreover, for any A ⊗ η ∈ K0 ,
V V ∗ (A ⊗ η) = V (ϕ(A∗ )∗ η) = 1A ⊗ ϕ(A∗ )∗ η.
P P
For any A ∈ A, let πϕ (A)( j Bj ⊗ ηj ) := j ABj ⊗ ηj and extend to Kϕ . For
all ξ, η ∈ H,
hξ, V ∗ π(A)V ηiH = hV ξ, π(A)V ηiϕ
= h1A ⊗ ξ, π(A)1A ⊗ ηiϕ
= h1A ⊗ ξ, A ⊗ ηiϕ
= hξ, ϕ(1∗A A)ηiH
= hξ, ϕ(A)ηiH
hence ϕ(A) = V ∗ π(A)V for all A ∈ A.
2.8. COMMENTS ON STINESPRING’S THEOREM 47
2.8. Comments on Stinespring’s theorem

We study objects, usually the interest is not on the object itself but function
algebras on it. These functions reveal properties of the object.
Example 2.48. In linear algebra, there is a bijection between inner product
structions on Cn and positive-definite n×n matrices. Specifically, h·, ·i : Cn ×Cn →
C is an inner product if and only if there exists a positive definite matrix A such
that
hv, wi = v t Aw
  
a11 a12 ··· a1n v1

 a21 a22 ··· a2n 
 v2 

= v1 v2 · · · vn  .. .. .. ..  .. 
 . . . .  . 
an1 an2 · · · ann vn
n n
for all v, w ∈ C . We think of C as C-valuded functions defined on {1, 2, . . . , n},
then h·, ·iA is an inner product built on the function space.
This is then extended to infinite dimensional space.
Example 2.49. If F is a positive definite function on R, then on K0 = span{δx :
x ∈ R}, F defines a sesquilinear form h·, ·iF : R × R → C, where
X X X
h ci δx i , dj δxj iF = c̄i dj F (xi , xj ).
i j i,j
Let ker(F ) = {v ∈ K0 : hv, vi = 0}, and ker(F ) is a closed subspace in K0 . We get

1/2
a Hilbert space K as the completion of K0 /ker under the norm k·kF := h·, ·iF .
What if the index set is not {1, 2, . . . , n} or R, but a ∗-algebra?
Example 2.50. Let X be a locally compact Hausdorff space and M be a σ-
algebra in X. The space of continuous functions with compact support Cc (X) is an
abelian C ∗ -algebra, where the C ∗ -norm is given by kf k = max{|f (x)| : x ∈ X}. By
Riesz’s theorem, there is a bijection between positive linear functionals on Cc (X)
and Borel measures. A Borel measure µ is a state if and only it is a probability
measure.
M is an abelian algebra. The associative multiplication is defined as AB :=
A ∩ B. The identity element is X, since A ∩ X = X ∩ A = A, for all A ∈ M. Let
µ be a probability measure. Then µ(A ∩ B) ≥ 0, for all A, B ∈ M, and µ(X) = 1,
therefore µ is a state. As before, we consider functions
P on M, i.e. the index set
is the σ-algebra
P M. Then expressions such as i c i δ A i are precisely the simple
functions i ci χAi . The GNS construction starts with
K0 = span{δA : A ∈ M} = span{χA : A ∈ M}
and the sesquilinear form h·, ·iµ : K0 × K0 → C, where
X X X
h c i χ Ai , dj χBj i := ci dj µ(Ai ∩ Bj ).
i j i,j
h·, ·iµ is positive semi-definite, since

X X X X
h c i χ Ai , c i χ Ai i = ci cj µ(Ai ∩ Aj ) = |ci |2 µ(Ai ) ≥ 0.
i i i,j i
Let ker = {v ∈ K0 : hv, viµ = 0}, complete K0 /ker with the corresponding norm,
we actually get the Hilbert space L2 (µ).
Let A be a ∗-algebra. The set of C-valued functions on A is precisely A ⊗ C.
We may think of putting A on the horizontal axis, and at each point A ∈ A attach
a complex number to it i.e. building functions indexed by A. Then members of
A ⊗ C are of the form
X X X
Ai ⊗ ci 1C = ci Ai = c i δ Ai
i i i
with finite summation over i. Note that C is natually embedded into A ⊗ C as
1A ⊗ C, i.e. c 7→ cδ1A , and the latter is a 1-dimensional subspace. In order to build
a Hilbert space out of the function space, a quadradic form is required. A state on
A does exactly the job. Let w be a state on A. Then the sesquilinear form
X X X
h c i δ Ai , dj δBj iw := ci dj w(A∗i Bj )
i i i,j
is positive semi-definite, because

!
X X X X X
∗ ∗
h c i δ Ai , cj δAj iw = ci cj w(Ai Aj ) = w ( ci Ai ) ( ci Ai ) ≥ 0.
i i i,j i i
A Hilbert space Kw is obtained by taking completion of the quotient A⊗C/Ker(w).

A representation π, π(A)δB := δBA is in fact a “shift” in the index variable, and
extend linearly to Kw .
In Stinespring’s construction, A⊗C is replaced by A⊗H. i.e. in stead of working
with C-valued functions on A, one looks at H-valued functions on A. Hence we are
looking at functions of the form
X X
Ai ⊗ ξi = ξi δAi
i i
with finite summation over i. H is natually embedded into A ⊗ H as 1A ⊗ H, i.e.
ξ 7→ ξδ1A . Hδ1A is infinite dimensional, or we say that the function ξδ1A at δ1A
has infinite multiplicity. If H is separable, we are actually attaching an l2 sequence
at every point A ∈ A on the horizontal axis. How to build a Hilbert space out of
these H-valued functions? The question depends on the choice a quadratic form.
If ϕ : A → B(H) is positive, then
hξδA , ηδB iϕ := hξ, ϕ(A∗ B)ηiH
is positive semi-definite. When extend linearly, one is in trouble. Since
X X X
h ξi δAi , ηj δBj iϕ = hξi , ϕ(A∗i Bj )ηj iH
i j i,j
which is equal to
  
ϕ(A∗1 B1 ) ϕ(A∗1 B2 ) ··· ϕ(A∗1 Bn ) ξ1

 ϕ(A∗2 B1 ) ϕ(A∗2 B1 ) ··· ϕ(A∗2 B1 ) 
 ξ2 

ξ1 ξ2 · · · ξn  .. .. .. ..  .. 
 . . . .  . 
ϕ(A∗n B1 ) ϕ(A∗n B2 ) · · · ϕ(A∗n Bn ) ξn
it is not clear why the matrix (ϕ(A∗i Bj ))
should be a positive operator acting on
H ⊗ Cn . But we could very well put this extra requirement into an axiom, and
consider ϕ being a CP map!
• We only assume A is a ∗-algebra, may not be a C ∗ -algebra. ϕ : A →

B(H) is positive does not necessarily imply ϕ is completely positve. A
counterexample is taking A = M2 (C), and ϕ : A → B(C2 ) ≃ M2 (C) given
by taking transpose, i.e. A 7→ ϕ(A) = Atr . Then ϕ is positive, but ϕ⊗IM2
is not. P
• The operator matrix (A∗i Aj ), which is also written as i,j A∗i Aj ⊗ eij is
a positve element in A ⊗ Mn . All positive elements in A ⊗ Mn are in such
form. This notation goes back again to Dirac, for the rank-1 operators
|vihv| are positive and all positive operators are sums of these rank-1
operators.
• Given a CP map ϕ : A → B(H), we get a Hilbert space Kϕ , a represen-
tation π : A → B(Kϕ ) and an isometry V : H → Kϕ , such that
ϕ(A) = V ∗ π(A)V
for all A ∈ A. P = V V ∗ is a self-adjoint projection from Kϕ to the image

of H under the embedding. To check P is a projection,
P 2 = V V ∗ V V ∗ = V (V ∗ V )V ∗ = V V ∗ .
Exercise 2.51. Let A = B(H), ξ1 , . . . , ξn ∈ H. The map (ξ1 , . . . , ξn ) 7→

(Aξ1 , . . . , Aξn ) ∈ ⊕n H is a representation of A if and only if
n times n times
z }| { z }| {
idH ⊕ · · · ⊕ idH ∈ Rep(A, H ⊕ · · · ⊕ H)
where in matrix notation, we have

  
idA (A) 0 ··· 0 ξ1

 0 idA (A) ··· 0 
 ξ2 

 .. .. .. ..  .. 
 . . . .  . 
0 0 ··· idA (A) ξn
    
A 0 ··· 0 ξ1 Aξ1

 0 A ··· 0   ξ2

  Aξ2 
  
=  .. .. .. ..   ..  =  ..  .
 . . . .  .   . 
0 0 ··· A ξn Aξn
Note. In this case, we say the identity representation idA : A → H has

multiplicity n.
Exercise 2.52. Let Vi : H → H, and

 
V1
 V2 
V :=  .  : H → ⊕n1 H.
 
 .. 
Vn

Let V ∗ : ⊕n1 H → H be the adjoint of V . Prove that V ∗ = V1∗ V2∗ · · · Vn∗ .
Proof. Let ξ ∈ H, then

 
V1 ξ

 V2 ξ 

Vξ = .. 
 . 
Vn ξ
and
   
η1 V1 ξ

 η2  
  V2 ξ 
 X
h .. , .. i = hηi , Vi ξi
 .   .  i
ηn Vn ξ
X
= hVi∗ ηi , ξi
i
X
= h Vi∗ ηi , ξi
i
 
η1
h
∗
i η2 
= h V1∗ V2∗
 
· · · Vn  ..  , ξi
 . 
ηn
h i
∗
this shows that V ∗ = V1∗ V2∗ · · · Vn .
Let V as defined above.

Exercise 2.53. The following are equivalent:
2 2
P is ∗an isometry, i.e. kV ξk = kξk , for all ξ ∈ H;
(1) V
(2) Vi Vi = IH ;
(3) V ∗ V = IH .
Proof. Notice that

X X X
kV ξk2 = kVi ξk2 = hξ, Vi∗ Vi ξi = hξ, Vi∗ Vi ξi.
i i i
2 2
Hence kV ξk = kξk if and only if
X
hξ, Vi∗ Vi ξi = hξ, ξi
i
P ∗
for all ξ ∈ H. Equivalently, i Vi Vi = IH = V ∗ V .
More examples of tensor products.

Exercise 2.54. Prove the following.
• ⊕n1 H ≃ H ⊗ Cn
P⊕∞
• 1 H ≃ H ⊗ l2
• Given L2 (X, M, µ), L2 (X, H) ≃ H ⊗ L2 (µ) where L2 (X, H) consists of all
measurable functioins f : X → H such that
ˆ
kf (x)k2H dµ(x) < ∞
X
2.9. MORE ON THE CP MAPS 51
and ˆ
hf, gi = hf (x), g(x)iH dµ(x).
X
• Show that all the spaces above are Hilbert spaces.
Corollary 2.55. (Krauss, physicist) Let dimH = n. Then all the CP maps
are of the form X
ϕ(A) = Vi∗ AVi .
i
Remark. This was discovered in the physics literature by Kraus. The original
proof was very intracate, but it is a corollary of Stinespring’s theorem. When
dimH = n, let e1 , . . . en be an ONB. Vi : ei 7→ V ei ∈ K is an isometry, i =
1, 2, . . . , n. So we get a system of isometries, and
  
A V1
 A   V2 
ϕ(A) = V1∗ V2∗ · · · Vn∗ 
  
. ..   ..  .
  . 
A Vn
P ∗
Notice that ϕ(1) = 1 if and only if i Vi Vi = 1.
Using tensor product in representations.
Exercise 2.56. (Xi , Mi , µi ) i = 1, 2 are measure spaces. Let πi : L∞ (µi ) →
2
L (µi ) be the representation such that πi (f ) is the operator of multiplication by f
on L2 (µi ). Hence π ∈ Rep(L∞ (Xi ), L2 (µi )), and
π1 ⊗ π2 ∈ Rep(L∞ (X1 × X2 ), L2 (µ1 × µ2 ))
with
π1 ⊗ π2 (ϕ̃)f˜ = ϕ̃f˜
where ϕ̃ ∈ L∞ (X1 × X2 ) and f˜ ∈ L2 (µ1 × µ2 ).
More about multiplicaity
Exercise. dsf
2.9. More on the CP maps

Positive maps have been a recursive theme in functional analysis. An classical
example is A = Cc (X) with a positive linear functional Λ : A → C, mapping A into
a 1-d Hilbert space C.
In Stinespring’s formulation, ϕ : A → H is a CP map, then we may write
ϕ(A) = V ∗ π(A)V where π : A → K is a representation on a bigger Hilbert space
Kcontaining H. The containment is in the sense that V : H ֒→ K embeds H into
K. Notice that
V ϕ(A) = π(A)V =⇒ ϕ(A) = V ∗ π(A)V
but not the other way around. (???) In Nelson’s notes, we use the notation ϕ ⊂ π
for one representation being the subrepresentation of another representation. To
imitate the situation in linear algebra, we may want to split an operator T acting
on K into operators action on H and its complement in K. Let P : K → H be the
orthogonal projection. In matrix language,

PTP PTP⊥
.
P ⊥T P P ⊥T P ⊥
2.10. KRIEN-MILMAN REVISITED 52
A better looking would be

PTP 0 ϕ1 0
=
0 P ⊥T P ⊥ 0 ϕ2
hence
π = ϕ1 ⊕ ϕ2 .
Stinespring’s theorme is more general, where the off-diagonal entries may not be
zero.
2.10. Krien-Milman revisited

We study some examples of compact convex sets in locally convex topological
spaces. Some typical examples include the set of positive semi-definite functions,
taking values in C or B(H).
The context for Krein-Milman is locally convex topological spaces. Almost all
spaces one works with are locally convex. The Krein-Milman theorem is in all
functional analysis books. Choqute’s thoerem comes later, hence it’s not contained
in most books. A good reference is the book by R. Phelps. The proof of Choquet’s
theorem is not specially illuminating. It uses standard integration theory.
Theorem 2.57. (Krein-Milman) K is a compact convex set in a locally convex
topological space. Then K is equals to the closed convex hull of its extreme points.
K = cl(conv(E(K))).
P
P A convex combination of points (ξi ) in K has the form v = ci ξi , such that
ci = 1. Closure refers to taking limit, allow all limits of such convex combina-
tions. Such a v is obviously in K, since K was assumed to be convex. The point of
the Krein-Milman’s theorem is the converse. The Krein-Milman theorem is not as
useful as another version by Choquet.
Theorem 2.58. (Choquet) K is a compact convex set in a locally convex topo-
logical space. Let E(K) be the set of extreme points on K. Then for all p ∈ K, there
exists a Borel probability measure µp , supported on a Borel set bE(K) ⊃ E(K), such
that ˆ
p= ξdµ(ξ).
bE(X)
The expression in Choquet’s theorem is a generalization of convex combination.

In stead of summation, it is an integral against a measure. Since there are some
bazarre cases where the extreme points E(K) do not form a Borel set, the measure
µp is actually supported on bE(K), such that µp (bE(K) − E(K)) = 0. Examples of
such a decomposition include Fourier transform, Laplace transform, direct integrals.
Example 2.59. Let (X, M, µ) be a measure space, where X is compact and
Hausdorff. The set of all probability measures P(X) is a convex set. To see this,
let µ1 , µ2 ∈ P(X) and 0 ≤ t ≤ 1, then tµ1 + (1 − t)µ2 is a measure on X, moreover
(tµ1 + (1 − t)µ2 )(X) = t + 1 − t = 1, hence tµ1 + (1 − t)µ2 ∈ P(X). Usually we
don’t want all probability measures, but a closed subset.
We compute extreme points in the previous example.
2.10. KRIEN-MILMAN REVISITED 53
Example 2.60. K = P(X) is compact convex in C(X)∗ , which is identified

as the set of all measures due to Riesz. C(X)∗ is a Banach space hence is always
convex. The importance of being the dual of some Banach space is that the unit
ball is always weak ∗ compact. The weak ∗ topology is just the cylindar topol-
∗
Q The unit ball B1 sits inside the infinite product space (compact, Hausdorff)
ogy.
v∈B,kvk=1 D1 , where D1 = {z ∈ C : |z| =Q1}. The weak ∗ topology on B1∗ is just
the restriction of the product topology on D1 onto B1∗ .
Example 2.61. Claim: E(K) = {δx : x ∈ X}, where δx is the Dirac measure
supported
´ at x ∈ X. By Riesz, to know the measure is to know the linear functional.
f dδx = f (x). Hence we get a family of measures indexed by X. If X = [0, 1],
we get a continuous family of measures. To see there really are extreme points, we
do the GNS contructioin on the algebra A = C(X), with the state µ ∈ P(X). The
Hilbert space so constructed is simply L2 (µ). It’s clear that L2 (δx ) is 1-dimensional,
hence the representation is irreducible. Therefore δx is a pure state, for all x ∈ X.
Note 2.62. Extreme points: ν is an extreme point in P(X) if and only if
ν ∈ [µ1 , µ2 ] ⇒ ν = µ1 or ν = µ2 .
Example 2.63. Let A = B(H), and K = states. For each ξ ∈ H, the map
A 7→ wξ (A) := hξ, Aξi is a state, called vector state. Claim: E(K) = vector states.
To show this, suppose W is a subspace of H such that 0 W H, and suppose
W is invariant under the action of B(H). Then ∃h ∈ H, h ⊥ W . Choose ξ ∈ W .
The wonderful rank-1 operator (due to Dirac) T : ξ 7→ h given by T = |hihξ|, shows
that h ∈ W . Hence h ⊥ h and h = 0. Therefore W = H. We say B(H) acts
transitively on H.
Note 2.64. In general, any C ∗ -algebra is a closed subalgebra of B(H) for some
H. But if we choose B(H), then all the pure states are vector states.
Example 2.65. Let A be a ∗-algebra, S(A) be the set of states on A. w :
A → C is a state if w(1A ) = 1 and w(A) ≥ 0, whenever A ≥ 0. Let A be a
∗-algebra, then the set of completely postive maps is a compact convex set. CP
maps are generalizations of states. Back to the first example,
´ A = C(X) there is
a bijection between state ϕ µ and Borel measure µ. ϕ(a) = adµ. We check that
ϕ(1A ) = ϕ(1) = 1dµ = µ(X) = 1; and ϕ(f ) = g 2 dµ ≥ 0 for f ≥ 0, g 2 = f .
´ ´
The next example if taken from AMS as a homework exercise.

Q
Example 2.66. Take the two state sample space Ω = ∞ 1 {0, 1} with prod-
uct topology. Assign probability measure, so that we might favor one outcome
than the other. For example, let s = x1 + · · · xn , Pθ (Cx ) = θs (1 − θ)n−1 , i.e. s
heads, (n − s) tails. Notice that Pθ is invariant under permutation of coordinates.
x1 , x2 , . . . , xn 7→ xσ(1) xσ(2) . . . xσ(n) . Pθ is a member of the set of all such invariant
measures (invariant under permutation) Pinv (Ω). Prove that
E(Pinv (Ω)) = [0, 1]
i.e. Pθ are all the possible extreme points.
Remark 2.67. σ : X → X is a measurable transformation. µ is ergodic
(probability measure) if
[E ∈ M, σE = E] ⇒ µ(E) ∈ {0, 1}
2.11. STATES AND REPRESENTATION 54
which intuitively says that the whole space X can’t be divided into parts where µ
is invariant. It has to be mixed up by the transformation σ.
2.11. States and representation

The GNS construction gives rise to a bijection between states and representa-
tions. We consider decomposition of representations or equivalently states.
The smallest representations are the irreducible ones. A representation π : A →
B(H) is irreducible, if whenever H breaks up into two pieces H = H1 ⊕ H2 , where
Hi is invariant under π(A), one of them is zero (the other is H). Equivalently,
if π = π1 ⊕ π2 , where πi = π Hi , then one of them is zero. This is similar to
decomposition of natural nubmers into product of primes. For example, 6 = 2 × 3,
but 2 and 3 are primes and they do not dompose further.
Hilbert spaces are defined up to unitary equivalence. A state ϕ may have
equivalent representations on different Hilbert spaces (but unitarily equivalent),
however ϕ does not see the distinction, and it can only detect equivalent classes of
representations.
Example 2.68. Let A be a ∗-algebra. Given two states s1 and s2 , by the GNS
construction, we get cyclic vector ξi , and representation πi : A → B(Hi ), so that
si (A) = hξi , πi (A)ξi i, i = 1, 2. Suppose there is a unitary operator W : H1 → H2 ,
such that for all A ∈ A,
π1 (A) = W ∗ π2 (A)W.
Then
hξ2 , π2 (A)ξ2 i2 = hW ξ1 , π2 (A)W ξ1 i1
= hξ1 , W ∗ π2 (A)W ξ1 i1
= hξ1 , π1 (A)ξ1 i1
i.e. s2 (A) = s1 (A). Therefore the same state s = s1 = s2 has two distinct (unitarily
equivalent) representations.
Remark 2.69. A special case of states are measures. Two representations
are mutually singular π1 ⊥ π2 , if and only if two measures are mutually singular,
µ1 ⊥ µ2 . Later, we will follow Nelson’s notes to build Hilbert space out of equivalent
classes measures.
Lemma 2.70. (Schur) The following are equivalent.
(1) A representation π : A → B(H) is irreducible
(2) (π(A))′ is one-dimensional, i.e. (π(A))′ = cI.
Proof. (2)⇒(1). Suppose π is not irreducible, i.e. π = π1 ⊕ π2 . Claim that

(using tensor) M2 ⊂ (π(A))′ . Let

IH1 0 0 0
PH1 = , PH2 = 1 − PH1 =
0 0 0 IH2
then for all A ∈ A, PHi π(A) = π(A)PHi , i = 1, 2. Hence (π(A))′ has more than
one dimension.
(1)⇒(2). Suppose (π(A))′ has more than one dimension. Let X ∈ (π(A))′ , i.e.
Xπ(A) = π(A)X, ∀A ∈ A.
By taking adjoint, X ∗ ∈ (π(A))′ . Hence X + X ∗ is self-adjoint, and X + X ∗ 6= cI,

since by hypothesis (π(A))′ has more than one dimension. Therefore X + X ∗ has
non trivial spectral projection, by the spectral theorem, i.e. there is self-adjoint
projection P (E) ∈/ {0, I}. Let H1 = P (E)H, and H2 = (I − P (E))H. H1 and
H2 are both nonzero proper subspaces of H. Since P (E) commutes with π(A), it
follows that H1 and H2 are both invariant under π.
Corollary. To test invariant subspaces, one only needs to look at projections

in the commutant. π is irreducible if and only if the only projections in (π(A))′ are
0 or I.
Remark 2.71. In matrix notation, write

π1 (A) 0
π(A) = .
0 π2 (A)
If
X Y
∈ (π(A))′
U V
then

X Y π1 (A) 0 Xπ1 (A) Y π2 (A)
=
U V 0 π2 (A) U π1 (A) V π2 (A)

π1 (A) 0 X Y π1 (A)X π1 (A)Y
= .
0 π2 (A) U V π2 (A)U π2 (A)V
Hence
Xπ1 (A) = π1 (A)X
V π2 (A) = π2 (A)V
i.e.
X ∈ (π1 (A))′ , V ∈ (π2 (A))′
and
U π1 (A) = π2 (A)U
Y π2 (A) = π1 (A)Y.
i.e. U, Y ∈ int(π1 , π2 ), the set of intertwing operators of π1 and π2 .
π1 (A)
/ H1
B H1 \
Y U U Y
π2 (A)
H2 / H2
π1 and π2 are inequivalent if and only if int(π1 , π2 ) = 0. If π1 = π2 , then we say π

has multiplicity (multiplicity equals 2), which is equivalent to the commutant being
non-abelian (in the case where π1 = π2 , (π(A))′ ≃ M2 .)
Schur’s lemma addresses all representations. We characterise the relation be-
tween state and its GNS representation, i.e. specilize to the GNS representation.
Given a ∗ algebra A, the states S(A) forms a compact convex subset in the unit
ball of the dual A∗ .
Let A+ be the set of positive elements in A. Given s ∈ S(A), let t be a positive

linear functional. By t ≤ s, we means t(A) ≤ s(A) for all A ∈ A+ . We look for
relation between t and the commutant (π(A))′ .
Lemma 2.72. (Schur-Sakai-Nicodym) Let t be a positive linear functional, and
let s be a state. There is a bijection between t such that 0 ≤ t ≤ s, and self-adjoint
operator A in the commutant with 0 ≤ A ≤ I. The relation is given by
t(·) = hΩ, π(·)AΩi
Remark 2.73. This is an extention of the classical Radon-Nicodym derivative
theorem to the non-commutative setting. We may write A = dt/ds. The notation
0 ≤ A ≤ I refers to the partial order of self-adjoint operators. It means that for all
ξ ∈ H, 0 ≤ hξ, Aξi ≤ kξk2 . See [[3]], [[1]] and [[2]].
Proof. Easy direction, suppose A ∈ (π(A))′ and 0 ≤ A ≤ I. As in many
applications, the favorite√functions one usually
√ applies to self-adjoint operators
√ is
the squre root function ·. So let’s take A. Since A ∈ (π(A))′ , so is A. We
need to show t(a) = hΩ, π(a)AΩi ≤ s(a), for all a ≥ 0 in A. Let a = b2 , then
t(a) = hΩ, π(a)AΩi
= hΩ, π(b2 )AΩi
= hΩ, π(b)∗ π(b)AΩi
= hπ(b)Ω, Aπ(b)Ωi
≤ hπ(b)Ω, π(b)Ωi
= hΩ, π(a)Ωi
= s(a).
Conversely, suppose t ≤ s. Then for all a ≥ 0, t(a) ≤ s(a) = hΩ, π(a)Ωi. Again
write a = b2 . It follows that
t(b2 ) ≤ s(b2 ) = hΩ, π(a)Ωi = kπ(b)Ωk2 .
By Riesz’s theorem, there is a unique η, so that
t(a) = hπ(b)Ω, ηi.
Conversely, Let a = b2 , then
t(b2 ) ≤ s(b2 ) = hΩ, π(a)Ωi = kπ(b)Ωk2 .
i.e. π(b)Ω 7→ t(b2 ) is a bounded quadratic form. Therefore, there exists a unique
A ≥ 0 such that
t(b2 ) = hπ(b)Ω, Aπ(b)Ωi.
It is easy to see that 0 ≤ A ≤ I. A ∈ (π(A))′ ????
Corollary 2.74. Let s be a state. (π, Ω, H) is the corresponding GNS con-
struction. The following are equivalent.
(1) For all positive linear functional t, t ≤ s ⇒ t = λs for some λ ≥ 0.
(2) π is irreducible.
Proof. By Sakai-Nicodym derivative, t ≤ s if and only if there is a self-adjoint
operator A ∈ (π(A))′ so that
t(·) = hΩ, π(·)AΩi
Therefore t = λs if and only if A = λI.

Suppose t ≤ s ⇒ t = λs for some λ ≥ 0. Then π must be irreducible, since
otherwise there exists A ∈ (π(A))′ with A 6= cI, hence A ∋ a 7→ t(a) := hΩ, π(a)AΩi
defines a positive linear functional, and t ≤ s, however t 6= λs. Thus a contradiction
to the hypothesis.
Conversely, suppose π is irreducible. Then by Schur’s lemma, (π(A))′ is 1-
dimensional. i.e. for all A ∈ (π(A))′ , A = λI for some λ. Therefore if t ≤ s, by
Sakai’s theorem, t(·) = hΩ, π(·)AΩi. Thus t = λs for some λ ≥ 0.
Definition 2.75. A state s is pure if it cannot be broken up into a convex

combination of two distinct states. i.e. for all states s1 and s2 , s = λs1 +(1−λ)s2 ⇒
s = s1 or s = s2 .
The main theorem in this section is a corollary to Sakai’s theorem.
Corollary 2.76. Let s be a state. (π, Ω, H) is the corresponding GNS con-

struction. The following are equivalent.
(1) t ≤ s ⇒ t = λs for some λ ≥ 0.
(2) π is irreducible.
(3) s is a pure state.
Proof. By Sakai-Nicodym derivative, t ≤ s if and only if there is a self-adjoint

operator A ∈ (π(A))′ so that
t(a) = hΩ, π(a)AΩi, ∀a ∈ A.
Therefore t = λs if and only if A = λI.

We show that (1) ⇔ (2) and (1) ⇒ (3) ⇒ (2).
(1) ⇔ (2) Suppose t ≤ s ⇒ t = λs, then π must be irreducible, since otherwise
there exists A ∈ (π(A))′ with A 6= cI, hence t(·) := hΩ, π(·)AΩi defines a positive
linear functional with t ≤ s, however t 6= λs. Conversely, suppose π is irreducible.
If t ≤ s, then t(·) = hΩ, π(·)AΩi with A ∈ (π(A))′ . By Schur’s lemma, (π(A))′ =
{0, λI}. Therefore, A = λI and t = λs.
(1) ⇒ (3) Suppose t ≤ s ⇒ t = λs for some λ ≥ 0. If s is not pure, then
s = cs1 + (1 − c)s2 where s1 , s2 are states and c ∈ (0, 1). By hypothesis, s1 ≤ s
implies that s1 = λs. It follows that s = s1 = s2 .
(3) ⇒ (2) Suppose π is not irreducible, i.e. there is a non trivial projection
P ∈ (π(A))′ . Let Ω = Ω1 ⊕ Ω2 where Ω1 = P Ω and Ω2 = (I − P )Ω. Then
s(a) = hΩ, π(a)Ωi

= hΩ1 ⊕ Ω2 , π(a)Ω1 ⊕ Ω2 i
= hΩ1 , π(a)Ω1 i + hΩ2 , π(a)Ω2 i
Ω1 Ω1 Ω2 Ω2
= kΩ1 k2 h , π(a) i + kΩ2 k2 h , π(a) i
kΩ1 k kΩ1 k kΩ2 k kΩ2 k
Ω1 Ω1 Ω2 Ω2
= kΩ1 k2 h , π(a) i + (1 − kΩ1 k2 )h , π(a) i
kΩ1 k kΩ1 k kΩ2 k kΩ2 k
= λs1 (a) + (1 − λ)s2 (a).
Hence s is not a pure state.

2.12. NORMAL STATES 58
2.12. Normal states

The thing that we want to do with representations comes down to the smallest
ones, i.e. the irreducible representations. Let A be a ∗-algebra, a representation
π : A → B(H) generates a ∗-subalgebra π(A) in B(H). By taking norm closure,
one gets a C ∗ -algebra.
An abstract C ∗ -algebra is a Banach ∗-algebra with the axiom ka∗ ak = kak2 .
By Gelfand and Naimark’s theorem, all abstract C ∗ -algebras are isometrically iso-
morphic to closed subalgebras of B(H), for some Hilbert space H. The construction
of H comes down to states S(A) on A and the GNS construction. Let A+ be the
positive elements in A. s ∈ S(A), s : A √→ C and s(A+ ) ⊂ [0, ∞). For C ∗ -algebra,
positive elements can be written f = ( f )2 by the spectral theorem. In general,
positive elements have the form a∗ a. There is a bijection between states and GNS
representations Rep(A, H), where s(A) = hΩ, π(A)Ωi.
Example ´ 2.77. A = C(X) where X is a compact Hausdorff space. sµ given
by sµ (a) = adµ is a state. The GNS constructuion gives H = L2 (µ), π(f ) is the
operator of multiplication by f on L2 (µ). {ϕ1
´ : ϕ ∈ C(X)}
´ is dense in L2 , where 1
is the cyclic vector. sµ (f ) = hΩ, π(f )Ωi = 1f 1dµ = f dµ, which is also seen as
the expectation of f in case µ is a probability measure.
Breaking up representations corresponds to breaking up states. Irreducibe
representations correspond to pure states which are extreme points in the states.
Schur’s lemma addresses all representations. It says that a representation π :
A → B(H) is irreducible if and only if (π(A))′ is 1-dimensional. When specialize
to the GNS representation of a given state s, this is also equivalent to saying
that for all positive linear functional t, t ≤ s ⇒ t = λs for some λ ≥ 0. This
latter equivalence is obtained by using a more general result, which relates t and
self-adjoint operators in the commutant (π(A))′ . Specifically, there is a bijection
between t ≤ s and 0 ≤ A ≤ I for Xt ∈ (π(A))′ , so that
t(·) = hΩ, π(·)Xt Ωi
If instead of taking the norm closure, but using the strong operator topology,
ones gets a Von Neumann algebra. Von Neumann showed that the weak closure of
A is equal to A′′
Corollary 2.78. π is irreducible ⇐⇒ (π(A))′ is 1-dimensional ⇐⇒ (π(A))′′ =
B(H).
More general states in physics come from the mixture of particle states, which
correspond to composite system. There are called normal states in mathematics.
ρ : H → H where ρ ∈ T1 H (trace class operators) with ρ > 0 and tr(ρ) = 1.
Define state sρ (a) = tr(aρ). Since ρ is compact, by spectral theorem of compact
operators, X
ρ= λk Pk
k
P
such that λ1 > λ2 >→ 0; λk = 1 and Pk = |ξk ihξk | are the rank-1 projections.
• sρ (I) = tr(ρ) = 1 P P P
• sρ (a) = tr(aρ) = hξk , a k λk Pk ξk i = k hξk | λk (|ξk ihξk |) |aξk i = k λk hξk , aξk i,
i.e. X
sρ = λk sξk
k
2.13. KADISON-SINGER CONJECTURE 59
sρ is a convex combination of pure states sξk .

• oberserve that tr(|ξihη|) = hη, ξi. In fact, take any onb {en } then
X
tr(|ξihη|) = hen | (|ξihη|) |en i
n
X
= hen , ξihη, en i
n
= hη, ξi
where the last line comes from Parseval identity.
• If we drop the condition ρ ≥ 0 then we get the duality (T1 H)∗ = B(H).
2.13. Kadison-Singer conjecture

2.13.1. Dictionary of OP and QM.
• states - unit vectors ξ ∈ H. These are all the pure states on B(H).
• observable - self-adjoint operators A = A∗
• measurement - spectrum
The spectral theorem was developed by Von Neumann and later improved by Dirac
and others. A self-adjoint operator A corresponds to a quantum observable, and
result of a quantum measurement can be represented by the spectrum of A.
• simple eigenvalue: A = λ |ξλ ihξλ |,
sξλ (A) = hξλ , Aξλ i = λ ∈ sp(A) ⊂ R.
P
P operator: A = λ λ |ξλ ihξλ |, such that (ξλ ) is an ONB of H. If
• compact
ξ = cλ ξλ is a unit vector, then
X
sξ (A) = hξ, Aξi = λ|cλ |2
λ
2
where (cλ )λ is a probability distribution over the spectrum of A, and sξ
is the expectation value of A.
• more general, allow continuous spectrum:
ˆ
A = λE(dλ)
ˆ
Aξ = λE(dλ)ξ
and ˆ
sξ (A) = hξ, Aξi = λkE(dλ)ξk2 .
We may write the unit vector ξ as

ξλ
ˆ z }| {
ξ = E(dλ)ξ
so that ˆ ˆ
2 2
kξk = kE(dλ)ξk = c(dλ)2 = 1
2.13. KADISON-SINGER CONJECTURE 60
where c(dλ)2 = kE(dλ)ξk2 . It is clear that (c(dλ))λ is a probability

distribution on spectrum of A. sξ (A) is again seen as the expectation
value of A with respect (c(dλ))λ , since
ˆ
sξ (A) = hξ, Aξi = λc(dλ)2 .
2.13.2. Kadison-Singer conjecture (see Palle’s private conversation).

Dirac gave a lecture at Columbia university in the late 1950’s, in which he claimed
without proof that pure states on l∞ extends uniquely on B(H). Two students
Kadison and Singer sitting in the audience were skeptical about whether Dirac
knew what it meant to be an extension. They later formuated the conjecture in a
joint paper.
Remark. Isadon Singer. Atiyah(abel price)-Singer(nobel price)
Let H be a Hilbert space with an ONB {en }∞ n=1 . Pn = |en ihen | is a rank-one
projection. Denote by D the set of all diagonal operators, i.e. D is the span of
X
λn Pn , (λn ) ∈ l∞
and D is a subalgebra of B(H).
Conjecture 2.79. Does every pure state on the subalgebra D ≈ l∞ extend
uniquely to a pure state on B(H)?
The difficulty lies in the fact that it’s hard to find all states on l∞ . (the dual
of l ????) It is conceivable that a pure state on l∞ may have two unit vectors on
∞
B(H).
Lemma 2.80. Pure states on B(H) are unit vectors. Let u ∈ H such that
kuk = 1. Then
wu (A) = hu, Aui
is a pure state. All pure states on B(H) are of this form.
Remark. It is in fact the equivalent class of unit vectors that are the pure
states on B(H). Since
heiθ u, Aeiθ ui = hu, Aui.
Equivalently, pure states sit inside the projective vector space. In Cn+1 , this is
CP n .
Since l∞ is an abelian algebra, by Gelfand’s theorem, l∞ ≃ C(X) for some
compact Hausdorff space X. X = βN, the Stone-Cech compactification of N.
Points in βN are called ultrafilters. Pure states on l∞ correspond to pure states on
C(βN), i.e. Dirac measures on βN.
Let s be a pure state on l∞ . Use Hahn-Banach theorem to extend s, as a
linear functional, from l∞ to s̃ on the Banach space B(H). However, Hahn-Banach
theorem doesn’t gurantee the extension is a pure state. Let E(s) be the set of
all states on B(H) which extend s. Since s̃ ∈ E(s), E(s) is nonempty. E(s)
is compact convex in the weak ∗ topology. By Krein-Milman’s theorem, E(s) =
closure(Extreme Points). Any extreme point will then be a pure state extension of
s. But which one to choose? It’s the uniqueness part that is the famous conjecture.
CHAPTER 3
Appliations to Groups
3.1. More on representations

3.1.1. Motivations. Every group G is also a ∗ semigroup (Nagy), wherer
g ∗ := g −1 .
It is obvious that g ∗ ∗ = g. But G is not a complex ∗ algebra yet, in particular,
multiplication by a complex numbert is not defined. However, we may work with
C-valued functions on G, which is clearly a ∗ algebra.
• multiplication
(g ⊗ cg )(h ⊗ ch ) = gh ⊗ cg ch
• scalar multiplication
t(g ⊗ cg ) = g ⊗ tcg , ∀t ∈ C
• ∗ operation
(g ⊗ cg )∗ = g −1 ⊗ cg−1
Note 3.1. The ∗ operation so defined is the only choice in order to have the
properties (ab)∗ = b∗ a∗ and (ta)∗ = t̄a∗ , t ∈ C.
What if (g ⊗ cg )∗ := g ∗ ⊗cg∗ , without the complex conjugation? Then (t(g ⊗
cg )) 6= t̄(g ⊗ cg )∗ .
∗
What is (g ⊗ cg )∗ := g ⊗ cg , without taking g ∗ ? This also satisfies (g ⊗ cg )∗∗ =

(g ⊗ cg )∗ . But
((g ⊗ cg )(h ⊗ ch ))∗ = (gh ⊗ cg ch )∗ = gh ⊗ cg ch
(h ⊗ ch )∗ (g ⊗ cg )∗ = (h ⊗ ch )(g ⊗ cg ) = hg ⊗ ch cg .
So the idea is instead of working with G, look at the ∗-algebra A = span{g ⊗ cg ≃
cg : g ∈ G}. There is a bijection between representation of G and representation of
A. Suppose U ∈ Rep(G, H), then it extends to a representation of A as
U : g ⊗ cg 7→ Ug ⊗ cg = cg Ug .
Conversely, if T ∈ Rep(A, H), then restriction to g ⊗ 1 gives a representation of G.
3.1.2. Group - algebra - representation. Everything we say about alge-
bras is also true for groups. In physics, we are interested in representation of
symmetry groups, which preserve inner product or energy. We always want unitary
representations. Irreducible representation amounts to elementary particles which
can not be broken up further. In practice, quite a lot work goes into finding ir-
reducible representations of symmetry groups. The idea is to go from groups to
algebras and then to representations.
G→A→π
61
3.1. MORE ON REPRESENTATIONS 62
• πG ∈ Rep(G, H)

π(g1 g2 ) = π(g1 )π(g2 )

π(eG ) = IH
 ∗
= π(g −1 )

π(g)
• πA ∈ Rep(A, H)

π(A1 A2 ) = π(A1 )π(A)2

π(1A ) = IH

π(A)∗ = π(A∗ )

Case 1. G is discrete −→ A = G ⊗ l1
! !
X X X
a(g)g b(h)h = a(g)b(h)gh
g g g,h
XX
= a(g ′ h−1 )b(h)g ′
g′ h
!∗ !
X X
c(g)g = c(g −1 )g
g g
where c∗ (g) = c(g −1 ). The multiplication of functions in A is a generaliza-

tion of convolutions.
Case 2. G is locally compact −→ A = G ⊗ L1 (µ) ≃ L1 (G).
Existance of Haar measure: easy proof for compact groups, and extend
to locally compact cases. For non compact groups, the left / right Haar
measures could be different. If they are always equal, the group is called
unimodular. Many non compact groups have no Haar measure.
Let B(G) be the Borel σ-aglebra of G, E ∈ B(G). Left Haar: λL (gE) =

λL (E); right Haar: λR (Eg) = λR (Eg). Theorem: the two measures are
equivalent
λL ≪ λR ≪ λL
dλL
△G = :G→G
dλR
is called the modular function, △G is a homomorphism, △G (gh) = △G (g)△G (h).
L1 (G)
ˆ
(ϕ1 ⋆ ϕ2 )(g) = ϕ1 (gh−1 )ϕ2 (h)dλR (h)
ϕ∗ (g) = ϕ(g −1 )△(g −1 )
Take C ∗ completion in both cases!

3.2. SOME EXAMPLES 63
Note. (1) for the case l1 (G), the counting measure in unimodular,
hence △(g) does not appear; (2) In ϕ1 (gh−1 ), h−1 appears since operation
on points is dual to operation on functions. A change of variable shows
ˆ ˆ ˆ
ϕ(gh−1 )dλR (g) = ϕ(g ′ )dλR (g ′ h) = ϕ(g ′ )dλR (g ′ )
(3) L1 (G) is a Banach algebra. Fubini’s theorem shows that f ⋆ g ∈ L1 (G),

for all f, g ∈ L1 (G).
There is a bijection between representations of groups and representations of al-
gebras. Given a unitary representation π ∈ Rep(G, H), take a Haar measre dg in
L1 (G), then we get the group algebra representation
ˆ
πL1 (G) (ϕ) = ϕ(g)π(g)dg
G
πL1 (G) (ϕ1 ⋆ ϕ2 ) = πL1 (G) (ϕ1 )πL1 (G) (ϕ2 )
πL1 (G) (ϕ)∗ = πL1 (G) (ϕ∗ )
Conversely, given a representation of L1 (G), let (ϕi ) be a sequence in L1 such that
ϕi → δg . Then ˆ
ϕi (h)π(h)gdh → π(g)
i.e. the limit is a representation of G.
3.2. Some examples

3.2.1. ax+b group.

a b
a ∈ R+ , b ∈ R
0 1
′
a′ b′ a b a a a′ b + b ′
=
0 1 0 1 0 1
−1 1 b

a b −a
= a
0 1 0 1
′
The multiplication aa′ can be made into addition, by taking a = et , a′ = et so that
′ ′
et et = et+t . This is a transformation group
x 7→ ax + b
where composition gives
x 7→ ax + b 7→ a′ (ax + b) + b′ = aa′ x + (a′ b + b′ )
Left Haar measure:
ˆ ˆ ˆ
f (h g)g dg = f (g )(hg ) d(hg ) = f (g ′ )g ′−1 dg −1
−1 −1 ′ ′ −1 ′
1

− ab da db dadb
dλL = g −1 dg = a =
0 1 0 0 a2
check:
′ b−b′

a b a′ b′ 1
− ab ′ a b a
g= , h= , h−1 g = a′ = a′ a′
0 1 0 1 0 1 0 1 0 1
3.2. SOME EXAMPLES 64
a b − b′ dadb
ˆ ˆ
f (h−1 g)dλL (g) = ,
f( ) 2
a′ a′ a
d(a′ s)d(a′ t + b′ )
ˆ
= f (s, t)
(a′ s)(a′ s)
dsdt
ˆ
= f (s, t) 2
s
where with a change of variable
a
s= , da = a′ ds
a′
b − b′
t= , db = a′ dt
a′
dadb a′2 dsdt dsdt
2
= ′ 2
= 2
a (sa ) s
Right Haar:
ˆ ˆ
f (gh−1 )(dg)g −1 = f (g ′ )d(g ′ h)(g ′ h)−1
ˆ
= f (g ′ )(dg ′ )(hh−1 )g ′−1
ˆ
= f (g ′ )(dg ′ )g ′−1
1

da db − ab dadb
dλR = (dg)g −1 = a =
0 0 0 1 a
check:
′ ′
a b a′ b′ a b 1
− ab ′ a
− ab
a′ + b
g= ,h = , gh−1 = a′ = a′
0 1 0 1 0 1 0 1 0 1
a ab′ dadb
ˆ ˆ
−1
f (gh )dλR (g) = ,
f( − + b)
a′ a′ a
′
a dsdt
ˆ
= f (s, t) ′
as
dsdt
ˆ
= f (s, t)
s
where with a change of variable
a
s= , da = a′ ds
a′
ab′
t=− + b, db = dt
a′
dadb a′ dsdt dsdt
= =
a a′ s s
3.3. INDUCED REPRESENTATION 65
3.3. Induced representation

Two questions involved: (1) How to get a representation of a group G from a
representation of the a subgroup Γ ⊂ G? (2) Given a representation of a group G,
how to test whether it is induced from a representation of a subgroup Γ? The main
examples we have looked at so far are
• ax + b
• Heisenberg
• SL2 (R)
• Lorents
• Poincare
Among these, the ax + b, Heisenberg and Poincare groups are semi-direct product
groups. Their representations are induced from a smaller normal subgroup. It is
extremely easy to find representations of abelian subgroups. Unitary representation
of abelian subgroups are one-dimensional, but the induced representation on an
enlarged Hilbert space is infinite dimensional. See the appendix for a quick review
of semi-direct product.

a b
Example 3.2. The ax+b group (a > 0). G = {(a, b)} where (a, b) = .
0 1
The multiplication rule is given by
(a, b)(a′ , b′ ) = (aa′ , b + ab′ )

1 b
(a, b)−1 = ( , − ).
a a
The subgroup Γ = {(1, b)} is one-dimensional, normal and abelian.
• abelian: (1, b)(1, c) = (1, c + b)

• normal: (x, y)(1, b)(x, y)−1 = (1, xb), note that this is also Adg acting on
the normal subgrup Γ
• The other subgroup {(a, 0)} is isomorphic to the multicative group (R+ , ×).
Because we have
(a, 0)(a′ , 0) = (aa′ , 0)
by the group multiplication rule above.

• Notice that (R+ , ×) is not a normal subgroup, since (a, b)(x, 0)( 1a , − ab ) =
(ax, b)( 1a , − ab ) = (x − bx + b).
The multiplicative group (R+ , ×) acts on the additive group (R, +) by
ϕ : (R+ , ×) 7→ Aut((R, +))

ϕa (b) = ab
check:
(a, b)(a′ , b′ ) = (aa′ , b + ϕa (b′ ))

= (aa′ , b + ab′ )
(a, b)−1 = (a−1 , ϕa−1 (b−1 ))
1 b
= (a−1 , a−1 (−b)) = ( , − )
a a
(a, b)(1, x)(a, b−1 ) = (a, b + ϕa (x))(a, b−1 )
1 b
= (a, b + ax)( , − )
a a
= (1, b + ax − b)
= (1, ax)
= ϕa (x)

1 0 0 1
Example 3.3. The Lie algebra of G is given by X = ,= .
0 0 0 0
We check that

tX et 0
e =
0 1
which is subgroup (R+ , ×); and

sY 1 s
e = I + sY + 0 + · · · + 0 =
0 1
which is subgroup (R, +). We also have [X, Y ] = Y .

Form L2 (µL ) where µL is the left Haar measure. Then π : g → π(g)f (x) =
−1
f (g x) is a unitary representation. Specifically, if g = (a, b) then
x y−b
f (g −1 x) = f ( , ).
a a
Differentiate along the a direction we get
d x y−b ∂ ∂
X̃f = f( , ) = (−x − y )f (x, y)
da a=1,b=0 a a ∂x ∂y
d x y−b ∂
Ỹ f = a=1,b=0
f( , ) = − f (x, y)
db a a ∂y
therefore we have the vector field
∂ ∂
X̃ = −x −y
∂x ∂y
∂
Ỹ = −
∂y
or equivalently we get the Lie algebra representation dπ on L2 (µL ). Notice that
[X̃, Ỹ ] = X̃ Ỹ − Ỹ X̃
∂ ∂ ∂ ∂ ∂ ∂
= (−x − y )(− ) − (− )(−x −y )
∂x ∂y ∂y ∂y ∂x ∂y
∂2 ∂2 ∂2 ∂ ∂2
= x + y 2 − (x + + y 2)
∂x∂y ∂y ∂x∂y ∂y ∂y
∂
= −
∂y
= Ỹ .
Notice that X̃ and Ỹ can be obtained by the exponential map as well.
d
X̃f = f (e−tX x)
dt t=0
d
= f ((e−t , 1)(x, y))
dt t=0
d
= f (e−t x, e−t y + 1)
dt t=0
∂ ∂
= (−x − y )f (x, y)
∂x ∂y
d
Y˜f = f (e−tY x)
dt t=0
d
= f ((1, −t)(x, y))
dt t=0
d
= f (x, y − t)
dt t=0
∂
= − f (x, y)
∂y
Example 3.4. We may parametrize the Lie algebra of the ax+b group us-
ing (x, y) variables. Build the Hilbert space L2 (µL ). The unitary representation
π(g)f (σ) = f (g −1 σ) induces the follows representations of the Lie algebra
d
dπ(s)f (σ) = f (e−sX σ) = X̃f (σ)
dx s=0
d
dπ(t)f (σ) = f (e−tY σ) = Ỹ f (σ).
dy t=0
Hence in the paramter space (s, t) ∈ R2 we have two usual derivative operators
∂/∂s and ∂/∂t, where on the manifold we have
∂ ∂ ∂
= −x −y
∂s ∂x ∂y
∂ ∂
= −y
∂t ∂y
The usual positive Laplacian on R2 translates to

∂ 2 ∂
−△ = ( ) + ( )2
∂s ∂t
= −[(X̃)2 + (Ỹ )2 ]
∂ ∂ ∂ ∂ ∂
= (−x − y )(−x − y ) + (−y )2
∂x ∂y ∂x ∂y ∂y
2 2 2
∂ ∂ ∂ ∂ ∂ ∂2
= −[(x2 2 + 2xy + y2 2 + x + y ) + y2 2 ]
∂x ∂x∂y ∂y ∂x ∂y ∂y
2 2 2
∂ ∂ ∂ ∂ ∂
= −(x2 2 + 2xy + (y 2 + 1) 2 + x + y ).
∂x ∂x∂y ∂y ∂x ∂y
This is in fact an elliptic operator, since the matrix
2
x xy
xy y 2 + 1
has trace trace = x2 + y 2 + 1 ≥ 1, and det = x2 ≥ 0. If instead we have “y 2 ” then

the determinant is the constant zero.
Notice the term “y 2 + 1” is essential for △ being elliptic. Also notice that all
the coefficients are analytic functions in the (x, y) variables.
Note 3.5. Notice the Γ is unimodular, hence it is just a copy of R. Its invariant
measure is the Lebesgue measure on R.
Example 3.6. Heisenberg group G = {a, b, c} where
 
1 a c
(a, b, c) =  0 1 b 
0 0 1
(a, b, c)(a′ , b′ , c′ ) = (a + a′ , b + b′ , c + ab′ + c′ )
(a, b, c)−1 = (−a, −b, −c + ab)
The subgroup Γ = {(0, b, c)} where
 
1 0 c
(1, b, c) =  0 1 b 
0 0 1
is two dimensional, abelian and normal.
• abelian: (0, b, c)(0, b′ , c′ ) = (0, b + b′ , c + c′ )
• normal:
(a, b, c)(0, x, y)(a, b, c)−1 = (a, b, c)(0, x, y)(−a, −b, −c + ab)
= (a, b + x, c + y + ax)(−a, −b, −c + ab)
= (0, x, y + ax + ab − ab)
= (0, x, ax + y)
Note that this is also Adg acting on the Lie algebra of Γ.
The additive group (R, +) acts on Γ = {(0, b, c)} ≃ (R2 , +) by

ϕ : (R, +) → Aut(Γ)

c 1 a c
ϕ(a) =
b 0 1 b

c + ab
=
b
check:
(a, (b, c))(a′ , (b′ , c′ )) = (a + a′ , (b, c) + ϕ(a)(b′ , c′ ))
= (a + a′ , (b, c) + (b′ , c′ + ab′ ))
= (a + a′ , b + b′ , c + c′ + ab)
(a, (b, c))−1 = (−a, ϕa−1 (−b, −c))
= (−a, (−b, −c + ab))
= (−a, −b, −c + ab)
′ ′ −1
(a, b, c)(0, b , c )(a, b, c) = (a, b + b′ , c + c′ + ab′ )(−a, −b, −c + ab)
= (0, b′ , c′ + ab′ )
′
c
= ϕa
b′
3.3.1. Induced representation. This also goes under the name of “Mackey
machine”. Its modern formulation is in the context of completely positive map.
Let G be a locally compact group, and Γ ⊂ G is closed subgroup.
group right Haar measure modular function
G dg △
Γ dξ δ
Recall the modular functions come in when the translation was put on the
wrong side, i.e. ˆ ˆ
f (gx)dx = △(g −1 ) f (x)dx
G G
or equivalently, ˆ ˆ
△(g) f (gx)dx = f (x)dx
G G
similar for dξ on the subgroup Γ.
Form the quotient M = Γ\G. Let π : G → Γ\G be the quotient map or the
covering map. M carries a transitive G action.
Note 3.7. M is called fundamental domain or homogeneous space. M is a
group if and only if Γ is a normal subgroup in G. In general, M may not be a
group, but it is still a very important manifold.
Note 3.8. µ is called an invariant measure on M , if µ(Eg) = µ(E), ∀g ∈ G.
µ is said to be quasi-invariant, if µ(E) = 0 ⇔ µ(Eg) = 0, ∀g. In general there is
no invariant measure on M , but only quasi-invariant measures. M has an invariant
measure if and only if M is unimodular (Heisenberg group). Not all groups are
unimodular, a typical example is the ax + b group.
Define τ : Cc (G) → Cc (M ) by
ˆ
(τ ϕ)(π(x)) = ϕ(ξx)dξ.
Γ
Lemma 3.9. τ is surjective.
Note 3.10. Since ϕ has compact support, the integral is well-defined. τ is
called conditional expectation. It is simply the summation of ϕ over the orbit Γx.
This is because if ξ runs over Γ, ξx runs over Γx. τ ϕ may also be interpreted as
taking average, only it does not divide out the total mass, but that only differs by
a constant.
Note 3.11. We may also say τ ϕ is a Γ-periodic extention, by looking at it as
a function defined on G. Then we check that
ˆ
τ ϕ(ξ1 x) = ϕ(ξξ1 x)dξ = τ ϕ(x)
Γ
because dξ is a right Haar measure. Thus τ ϕ is Γ-periodic in the sense that
τ ϕ(ξx) = τ ϕ(x), ∀ξ ∈ Γ.
Example 3.12. G = R, Γ = Z with dξ being the counting measure on Z.
ˆ X
(τ ϕ)(π(x)) = ϕ(ξx)dξ = ϕ(z + x)
Γ z∈Z
As a consequence, τ ϕ is left translation invariant by integers, i.e. τ ϕ is Z-periodic,
X
(τ ϕ)(π(z0 + x)) = ϕ(z0 + z + x)
z∈Z
X
= ϕ(z + x)
z∈Z
Since ϕ has compact support, ϕ(z + x) vanishes for all but a finite number of z.
Hence it is a finite summation, and it is well-defined.
Let L be a unitary representation of Γ on a Hilbert space V . The task now is
to get a unitary representaton U ind of G on an enlarged Hilbert space H. i.e. given
L ∈ Rep(Γ, V ), get U ind ∈ Rep(G, H) with H ⊃ V .
Let F∗ be the set of function f : G → V so that
f (ξg) = ρ(ξ)1/2 Lξ (f (g))
where ρ = δ/△.
Note 3.13. If f ∈ F∗ , then f (·g ′ ) ∈ F∗ , as
f (ξgg ′ ) = ρ(ξ)1/2 Lξ (f (gg ′ )).
It follows that F∗ is invariant under right translation by g ∈ G, i.e. (Rg f )(·) =
f (·g) ∈ F∗ , ∀f ∈ F∗ . Eventually, we want to define (Ugind f )(·) := f (·g), not on F∗
but pass to a subspace.
Note 3.14. The factor ρ(ξ)1/2 comes in, since later we will defined an inner
product h·, ·inew on F∗ so that kf (ξg)knew = kf (g)knew . Let’s ignore ρ(ξ)1/2 for a
moment.
Lξ is unitary implies that kf (ξg)kV = kLξ f (g)kV = kf (g)kV . Notice that
Hilbert spaces exist up to unitary equivalence, Lξ f (g) and f (g) really are the same
function. As ξ running throught Γ, kf (ξg)k is a constant on the orbit Γg. It follows

that f (ξg) is in fact a V -valued function defined on the quotient M = Γ\G.
We will later use these functions as multiplication operators.
Example 3.15. The Heisenberg group is unimodular, ρ = 1.
Example 3.16. For the ax + b group,
dadb
dλR =
a
dadb
dλL =
a2
dλL 1
△ = =
dλR a
On the abelian normal subgroup Γ = {(1, b)}, where a = 1, △(ξ) = 1. Γ is
unimodular, hence δ(ξ) = 1. Therefore, ρ(ξ) = δ(ξ)/△(ξ) = 1, ∀ξ ∈ Γ.
For all f ∈ F∗ , the map µf,f : Cc (M ) → C given by
ˆ
µf,f : τ ϕ 7→ kf (g)k2V ϕ(g)dg
G
is a positive linear functional. By Riesz, there exists a unique Radon measure µf,f
on M , such that ˆ ˆ
kf (g)k2V ϕ(g)dg = (τ ϕ)dµf,f .
G M
Note 3.17. Recall that given a measure space (X, M, µ), let f : X → Y .
Define a linear functional Λ : Cc (Y ) → C by
ˆ
Λϕ := ϕ(f (x))dµ(x)
Λ is positive, hence by Riesz’s theorem, there exists a unique regular Borel measure
µf on Y so that ˆ ˆ
Λϕ = ϕdµf = ϕ(f (x))dµ(x).
Y X
It follows that µf = µ ◦ f −1 .
Note 3.18. Under current setting, we have a covering map π : G → Γ\G =: M ,
and the right Haar measure µ on G. Thus we may define a measure µ ◦ π −1 .
However, given ϕ ∈ Cc (M ), ϕ(π(x)) may not have compact support, or equivalently,
π −1 (E) is Γ periodic. For example, take G = R, Γ = Z, M = Z\R. Then
π −1 ([0, 1/2)) is Z-periodic, which has infinite Lebesgue measure. What we really
need is some map so that the inverse of a subset of M is restricted to a single Γ
period. This is essentially what τ does. Taking τ ϕ ∈ Cc (Γ), get the inverse image
ϕ ∈ Cc (G). Even if ϕ is not restricted in a single Γ period, ϕ always has compact
support.
Hence we get a family of measures indexed by elements in F∗ . If choosing
f, g ∈ F∗ then we get complex measures µf,g . (using polarization identity)
• Define kf k2 := µf,f (M ), hf, gi := µf,g (M )
• Complete F∗ with respect to this norm to get an enlarged Hilbert space
H.
3.4. EXAMPLE - HEISENBERG GROUP 72
• Define induced representation U ind on H as

Ugind f (x) = f (xg)
Ugind is unitary.
kU ind f kH =???
´
Note 3.19. µf,g (M ) = M τ ϕdξ, where τ ϕ ≡ 1. What is ϕ then? It turns out
that ϕ could be constant 1 over a single Γ-period, or ϕ could spread out to a finite
number of Γ-periods. Draw a picture here!! Therefore, in this case
ˆ
kf k2 = kf (g)k2V ϕ(g)dg
G
ˆ
= kf (g)k2V ϕ(g)dg
1−period
ˆ
= kf (g)k2V dg
1−period
ˆ
= kf (g)k2V dg
M
Define P (ψ)f (x) := ψ(π(x))f (x), for ψ ∈ Cc (M ), f ∈ H, x ∈ G. P (ψ) is the

abelian algebra of multiplication operators. Observe that
Ugind P (ψ)Ugind
−1 = P (ψ(·g))
check:
Ugind P (ψ)f (x) = Ugind ψ(π(x))f (x)
= ψ(π(xg))f (xg)
P (ψ(·g))Ugind f (x) = P (ψ(·g))f (xg)
= ψ(π(xg))f (xg)
Conversely, how to recognize induced representation?
Theorem 3.20. (imprimitivity) Let G be a locally compact group with a closed
subgroup Γ. Let M = Γ\G. Suppose the system (U, P ) satisfies the covariance
relation,
Ug P (ψ)Ug−1 = P (ψ(·g)).
Then, there exists a unitary representation L ∈ Rep(Γ, V ) such that U ≃ indG
Γ (L).
3.4. Example - Heisenberg group

Heisenberg group G = {(a, b, c)} where
 
1 a c
(a, b, c) =  0 1 b 
0 0 1
(a, b, c)(a′ , b′ , c′ ) = (a + a′ , b + b′ , c + c′ + ab′ )
(a, b, c)−1 = (−a, −b, −c + ab)
The subgroup Γ = {(0, b, c)} where

 
1 0 c
(1, b, c) =  0 1 b 
0 0 1
is two dimensional, abelian and normal.
• abelian: (0, b, c)(0, b′ , c′ ) = (0, b + b′ , c + c′ )
• normal:
(a, b, c)(0, x, y)(a, b, c)−1 = (a, b, c)(0, x, y)(−a, −b, −c + ab)
= (a, b + x, c + y + ax)(−a, −b, −c + ab)
= (0, x, y + ax + ab − ab)
= (0, x, ax + y)
i.e. Ad : G → GL(n), as
Ad(g)(n) : = gng −1
(x, y) 7 → (ax + y)
the orbit is a 2-d transformation.
Fix h ∈ R\{0}. Recall the Schrodinger representation of G on L2 (R)
Ug f (x) = eih(c+bx) f (x + a)
We show that the Schrodinger representation is induced from a unitary represen-
tation L on the subgroup Γ.
(1) Let L ∈ Rep(Γ, V ) where Γ = {(0, b, c)}, V = C,
Lξ(b,c) = eihc .
The complex exponential comes in since we want a unitary representation.
The subgroup {(0, 0, c)} is the center of G. What is the induced represen-
tation? Is it unitarily equivalent to the Schrodinger representation?
(2) Look for the family F∗ of functions f : G → C (V is the 1-d Hilbert space
C), such that
f (ξ(b, c)g) = Lξ f (g).
Since
f (ξ(b, c)g) = f ((0, b, c)(x, y, z)) = f (x, b + y, c + z)
Lξ(b,c) f (g) = eihc f (x, y, z)
f (x, y, z) satisfies
f (x, b + y, c + z) = eihc f (x, y, z).
i.e. we may translate the y, z variables by arbitrary amount, and the only
price to pay is multiplicative factor eihc . Therefore f is really a function
defined on the quotient
M = Γ\G ≃ R.
M = {(x, 0, 0)} is identified with R, and the invariant measure on the
homogeneous space M is simply the Lebesgue measure. It is almost clear
at this point why the induced representation is unitarily equivalent to the
Schrodinger representation on L2 (R).
(3) τ ϕ 7→ G kf (g)k2V ϕ(g)dg induces a measure µf,f on M . This can be seen

´
as follows.
ˆ ˆ
2
kf (g)kV ϕ(g)dg = |f (x, y, z)|2 ϕ(x, y, z)dxdydz
G G≃R3
ˆ ˆ
2
= |f (x, y, z)| ϕ(x, y, z)dydz dx
M≃R Γ≃R2
ˆ ˆ
2
= |f (x, y, z)| ϕ(x, y, z)dydz dx
R2
ˆR
= |f (x, y, z)|2 (τ ϕ)(π(g))dx
ˆR
= |f (x, y, z)|2 (τ ϕ)(x)dx
R
ˆ
= |f (x, 0, 0)|2 (τ ϕ)(x)dx
R
where
ˆ
(τ ϕ)(π(g)) = ϕ(ξg)dξ
ˆΓ
= ϕ((0, b, c)(x, y, z))dbdc
2
ˆR
= ϕ(x, b + y, c + z)dbdc
2
ˆR
= ϕ(x, b, c)dbdc
2
ˆR
= ϕ(x, y, z)dydz
R2
= (τ ϕ)(x).
Hence Λ : Cc (M ) → C given by
ˆ
Λ : τ ϕ 7→ kf (g)k2V ϕ(g)dg
G
is a positive linear functional, therefore
Λ = µf,f
i.e.
ˆ ˆ
2
|f (x, y, z)| ϕ(x, y, z)dxdydz = (τ ϕ)(x)dµf,f (x).
R3 R
(4) Define
ˆ ˆ ˆ
kf k2ind := µf,f (M ) = 2
|f | dξ = 2
|f (x)| dx = |f (x, 0, 0)|2 dx
M R R
Ugind f (g ′ ) := f (g ′ g)
By definition, if g = g(a, b, c), g ′ = g ′ (x, y, z) then
Ugind f (g ′ ) = f (g ′ g)
= f ((x, y, z)(a, b, c))
= f (x + a, y + b, z + c + xb)
and U ind is a unitary representation by the definition of kf kind .

(5) To see U ind is unitarily equivalent to the Schrodinger representation on
L2 (R), define
W : H ind → L2 (R)
(W f )(x) = f (x, 0, 0)
If put other numbers into f , as f (x, y, z), the result is the same, since
f ∈ H ind is really defined on the quotient M = Γ\G ≃ R.
W is unitary:
ˆ ˆ ˆ
kW f k2L2 = |W f |2 dx = |f (x, 0, 0)|2 dx = |f |2 dξ = kf k2ind
R R Γ\G
Intertwining: let Ug be the Schrodinger representation.

Ug (W f ) = eih(c+bx) f (x + a, 0, 0)
W Ugind f = W (f ((x, y, z)(a, b, c)))
= W (f (x + a, y + b, z + c + xb))

= W eih(c+bx) f (x + a, y, z)
= eih(c+bx) f (x + a, 0, 0)
(6) Since {U, L}′ ⊂ {L}′, then the system {U, L} is reducible implies L is
reducible. Equivalent, {L} is irreducible implies {U, L} is irreducible.
Consequently, Ug is irrducible. Since Ug is the induced representation, if
it is reducible, L would also be reducible, but L is 1-dimensional.
Note 3.21. The Heisenberg group is a non abelian unimodular Lie group, so
the Haar measure on G is just the product measure dxdydz on R3 . Conditional
expectation becomes integrating out the variables correspond to subgroup. For
example, given f (x, y, z) conditioning with respect to the subgroup (0, b, c) amounts
to integrating out the y, z variable and get a function f˜(x), where
¨
f˜(x) = f (x, y, z)dydz.

a b
3.4.1. ax+b group. a ∈ R+ , b ∈ R, g = (a, b) = .
0 1
Ug f (x) = eiax f (x + b)
could also write a = et , then
t
Ug f (x) = eie x f (x + b)
x
Ug(a,b) f (x) = eiae f (x+b)
d
[ , iex ] = iex
dx
[A, B] = B
or
x
U(et ,b) f = eite f (x + b)
3.5. COADJOINT ORBITS 76

0 b
0 1
1-d representation. L = eib . Induce indG
L ≃ Schrodinger.
3.4.2. ax + b gruop.
3.5. Coadjoint orbits

It turns out that only a small family of representations are induced. The
question is how to detect whether a representation is induced. The whole theory
is also under the name of “Mackey machine”. The notion of “machine” refers to
something that one can actually compute in practice. Two main examples are the
Heisenberg group and the ax + b group.
What is the mysteries parameter h that comes into the Schrodinger represen-
tation? It is a physical constant, but how to explain it in mathematical theory?
3.5.1. review of some Lie theory.

Theorem 3.22. Every Lie group is diffeomorphic to a matrix group.
The exponential function maps a neighborhood of 0 into a connected component
of G containing the identity element. For example, the Lie algebra of the Heisenberg
group is
 
0 ∗ ∗
 0 0 ∗ 
0 0 0
All the Lie groups the we will ever encounter come from a quadratic form. Given
a quadratic form
ϕ:V ×V →C
there is an associated group that fixes ϕ, i.e. we consider elements g such that
ϕ(gx, gy) = ϕ(x, y)
and define G(ϕ) as the collection of these elements. G(ϕ) is clearly a group. Apply
the exponential map and the product rule,
d
ϕ(etX x, etX y) = 0 ⇐⇒ ϕ(Xx, y) + ϕ(x, Xy) = 0
dt t=0
hence
X + X tr = 0
The determinant and trace are related so that
det(etX ) = et·trace(X)
thus det = 1 if and only if trace = 0. It is often stated in differential geometry that
the derivative of the determinant is equal to the trace.
P
Example 3.23. Rn , ϕ(x, y) = xi yi . The associated group is the orthogonal
group On .
There is a famous cute little trick to make On−1 into a subgroup of On . On−1
is not normal in On . We may split the quadratic form into
Xn
x2i + 1
i=1
where 1 corresponds to the last coordinate in On . Then we may identity On−1 as
a subgroup of On
g 0
g 7→
0 I
where I is the identity operator.
Claim: On /On−1 ≃ S n−1 . How to see this? Let u be the unit vector corre-
sponding to the last dimension, look for g that fixes u i.e. gu = u. Such g forms a
subgroup of On , and it is called isotropy group.
In = {g : gu = u} ≃ On−1
For any g ∈ On , gu =?. Notice that for all v ∈ S n−1 , there exists g ∈ On such that
gu = v. Hence
g 7→ gu
n−1
in onto S . The kernel of this map is In ≃ On−1 , thus
On /On−1 ≃ Sn
Such spaces are called homogeneous spaces.
Example 3.24. visualize this with O3 and O2 .
Other examples of homogeneous spaces show up in number theory all the time.
For example, the Poincare group G/discrete subgroup.
G, N ⊂ G nornal subgroup. The map g · g −1 : G → G is an automorphism
sending identity to identity, hence if we differentiate it, we get a transformation
in GL(g). i.e. we get a family of maps Adg ∈ GL(g) indexed by elements in G.
g 7→ Adg ∈ GL(g) is a representation of G, hence if it is differentiated, we get a
representation of g, adg : g 7→ End(g) acting on the vector space g.
gng −1 ∈ N . ∀g, g · g −1 is a transformation from N to N , define Adg (n) =
−1
gng . Differentiate to get ad : n → n. n is a vector space, has a dual. Linear
transformation on vector space passes to the dual space.
ϕ∗ (v ∗ )(u) = v ∗ (ϕ(u))
m
∗ ∗
hΛ v , ui = hv ∗ , Λui.
In order to get the transformation rules work out, have to pass to the adjoint or
the dual space.
Ad∗g : n∗ → n∗
the coadjoint representation of n.
Orbits of co-adjoint representation acounts precisely to equivalence classes of
irrducible representations.
Example 3.25. Heisenberg group G = {(a, b, c)} with
 
1 a c
(a, b, c) =  0 1 b 
0 0 1
normal subgroup N = {(0, b, c)}

 
1 0 c
(0, b, c) =  0 1 b 
0 0 1
with Lie algebra n = {(b, c)}
 
1 0 c
(0, ξ, η) =  0 1 b 
0 0 1
Adg : n → n given by
gng −1 = (a, b, c)(0, y, x)(−a, −b, −c + ab)
= (a, b + y, c + x + ay)(−a, −b, −c + ab)
= (0, y, x + ay)
hence Adg : R2 → R2

x x + ay
Adg : 7→ .
y y
The matrix of Adg is (before taking adjoint) is

1 a
Adg = .
0 1
The matrix for Ad∗g is

1 0
Ad∗g = .
a 1
We use [ξ, η]T for the dual n∗ ; and use [x, y]T for n. Then

∗ ξ ξ
Adg : 7→
η aξ + η
What about the orbit? In the example of On /On−1 , the orbit is S n−1 .
For ξ ∈ R\{0}, the orbit of Ad∗g is

ξ ξ
7→
0 R
i.e. vertical lines with x-coordinate ξ. ξ = 0 amounts to fixed point, i.e. the orbit
is a fixed point.
The simplest orbit is when the orbit is a fixed point. i.e.

∗ ξ ξ
Adg : 7→ ∈V∗
η η
where if we choose
ξ 0
=
η 1
it is a fixed point.
The other extreme is to take any ξ 6= 0, then

∗ ξ ξ
Adg : 7→
0 R
3.6. GAARDING SPACE 79
i.e. get vertical lines indexed by the x-coordinate ξ. In this example, a cross section
is a subset of R2 that intersects each orbit at precisely one point. Every cross section
in this example is a Borel set in R2 .
We don’t always get measurable cross sections. An example is the construction
of non-measurable set as was given in Rudin’s book. Cross section is a Borel set
that intersects each coset at precisely one point.
Why does it give all the equivalent classes of irreducible representations? Since
we have a unitary representation Ln ∈ Rep(N, V ), Ln : V → V and by construction
of the induced representation Ug ∈ Rep(G, H), N ⊂ G normal such that
Ug Ln Ug−1 = Lgng−1
i.e.
Lg ≃ Lgng−1
now pass to the Lie algebra and its dual
Ln → LA → LA∗ .
3.6. Gaarding space

We talked about how to detect whether a representation is induced. Given
a group G with a subgroup Γ let M := Γ\G. The map π : G → M is called a
covering map, which sends g to its equivalent class or the coset Γg. M is given its
projective topology, so π is coninuous. When G is compact, many things simplify.
For example, if G is compact, any irreducible representation is fnite dimensional.
But many groups are not compact, only locally compact. For exmaple, ax+b, H3 ,
SLn .
Specialize to Lie groups. G and subgroup H have Lie algebras g and h respec-
tively.
g = {X : etx ∈ G, ∀t ∈ R}
Almost all Lie algebras we will encounter come from specifying a quadratic form
ϕ : G × G → C. ϕ is then uniquely determined by a Hermitian matrix A so that
ϕ(x, y) = xtr · Ay
Let G = G(ϕ) = {g : ϕ(gx, gy) = ϕ(x, y)}, then
d
ϕ(etX x, etX y) = 0
dt t=1
and with an application of the product rule,
ϕ(Xx, y) + ϕ(x, Xy) = 0
tr tr
(Xx) · Ay + x · AXy = 0
tr
X A + AX = 0
hence
g = {X : X tr A + AX = 0}.
Let U ∈ Rep(G, H), for X ∈ g, U (etX ) is a one parameter continuous group of
unitary operator, hence by Stone’s theorem,
U (etX ) = eitHX
for some self-adjoint operator HX (possibly unbounded). We often write
dU (X) := iHX
to indicate that dU (X) is the directional derivative along the direction X. Notice
∗
that HX = HX but
(iHX )∗ = −(iHX )
i.e. dU (X) is skew adjoint.
Example 3.26. G = {(a, b, c)} Heisenberg group. g = {X1 ∼ a, X2 ∼ b, X3 ∼
c}. Take the Schrodinger representation Ug f (x) = eih(c+bx) f (x + a), f ∈ L2 (R).
• U (etX1 )f (x) = f (x + t)
d d
U (etX1 )f (x) = f (x)
dt t=0 dx
d
dU (X1 ) =
dx
• U (etX2 )f (x) = eih(tx) f (x)
d
U (etX2 )f (x) = ihxf (x)
dt t=0
dU (X2 ) = ihx
• U (etX3 )f (x) = eiht f (x)
d
U (etX3 )f (x) = ihf (x)
dt t=0
dU (X2 ) = ihI
Notice that dU (Xi ) are all skew adjoint.
d
[dU (X1 ), dU (X2 )] = [ , ihx]
dx
d
= ih[ , x]
dx
= ih
In case we want self-adjoint operators, replace dU (Xi ) by−idU (Xi ) and
get
1 d
−idU (X1 ) =
i dx
−idU (X2 ) = hx
−idU (X1 ) = hI
1 d h
[ , hx] = .
i dx i
What is the space of functions that Ug acts on? L. Gaarding /gor-ding/ (Sweed-
ish mathematician) looked for one space that always works. It’s now called the
Gaarding space.
Start with Cc (G), every ϕ ∈ Cc (G) can be approximated by the so called
Gaarding functions, using the convolution argument. Define convolution as
ˆ
ϕ ⋆ ψ(g) = ϕ(gh)ψ(h)dR h
ˆ
ϕ ⋆ ψ(g) = ϕ(h)ψ(g −1 h)dL h
Take an approximation of identity ζj , so that

ϕ ⋆ ζj → ϕ, j → 0.
Define Gaarding space as functions given by
ˆ
U (ϕ)v = ϕ(h)U (h)vdL h
where ϕ ∈ Cc (G), v ∈ H, or we say

ˆ
U (ϕ) := ϕ(h)U (h)vdL h.
Since ϕ vanishes outside a compact set, and since U (h)v is continuous and bounded
in k·k, it follows that U (ϕ) is well-defined.
Lemma 3.27. U (ϕ1 ⋆ ϕ2 ) = U (ϕ1 )U (ϕ2 ) (U is a representation of the group
algebra)
Proof. Use Fubini,
ˆ ¨
ϕ1 ⋆ ϕ2 (g)U (g)dg = ϕ1 (h)ϕ(h−1 g)U (g)dhdg
¨
= ϕ1 (h)ϕ(g)U (hg)dhdg (dg is r-Haarg 7→ hg)
¨
= ϕ1 (h)ϕ(g)U (h)U (g)dhdg
ˆ ˆ
= ϕ1 (h)U (h)dh ϕ2 (g)U (g)dg
Choose ϕ to be an approximation of identity, then

ˆ
ϕ(g)U (g)vdg → U (e)v = v
i.e. any vector v ∈ H can be approximated by functions in the Gaarding space. It

follows that
{U (ϕ)v}
is dense in H.
Lemma 3.28. U (ϕ) can be differented, in the sense that
dU (X)U (ϕ)v = U (X̃ϕ)v
where we use X̃ to denote the vector field.
Proof. need to prove
1
lim (U (etX ) − I)U (ϕ)v = U (X̃ϕ)v.
t→0 t
Let vϕ := U (ϕ)v, need to look at in general U (g)vϕ .

ˆ
U (g)vϕ = U (g) ϕ(h)U (h)vdh
ˆ
= ϕ(h)U (gh)dh
ˆ
= △(g)ϕ(g −1 h)U (h)dh
3.7. DECOMPOSITION OF REPRESENTATION 82
set g = etX .
Note 3.29. If assuming unimodular, △ does not show up. Otherwise, △ is
some correction term which is also differnetiable. X̃ acts on ϕ as X̃ϕ. X̃ is called
the derivative of the translation operator etX .
Note 3.30. Schwartz space is the Gaarding space for the Schrodinger repre-
sentation.
3.7. Decomposition of representation

We study some examples of duality.
• G = T , Ĝ = Z
χn (z) = zn
χn (zw) = z n wn = χn (z)χn (w)
• G = R, Ĝ = R
χt (x) = eitx
• G = Z/nZ ≃ {0, 1, · · · , n − 1}. Ĝ = G.
This is another example where Ĝ = G.
Let ζ = ei2π/n be the primitive nth -root of unity. k ∈ Zn , l = {0, 1, . . . , n−
1}
2πkl
χl (k) = ei n
If G is a locally compact abelian group, Ĝ is the set of 1-dimensional representations.
Ĝ = {χ : g 7→ χ(g) ∈ T, χ(gh) = χ(g)χ(h)}.
Ĝ is also a group, with group operation defined by (χ1 χ2 )(g) := χ1 (g)χ2 (g). Ĝ is
called the group characters.
Theorem 3.31. (Pontryagin) If G is a locally compact abelian group, then
ˆ
G = Ĝ.
Note 3.32. This result first appears in 1930s in the annals of math, when
John Von Neumann was the editor of the journal at the time. The original paper
was hand written. Von Neumann rewrote it, since then the theorem became very
popular.
There are many groups that are not locally compact abelian. We want to study
the duality question in general. Examples are
• compact group
• fintie group (abelian, or not)
• H3 locally compact, nonabelian, unimodular
• ax+b locally compact, nonabelian, non-unimodular
If G is not abelian, Ĝ is not a group. We would like to decompose Ĝ into irreducible
representations. The big names under this development are Krein (died 7 years
ago), Peter-Weyl, Weil, Segal.
Let G be a group (may not be abelian). The right regular representation is
defined as
Rg f (·) = f (·g)
3.7. DECOMPOSITION OF REPRESENTATION 83
Rg is a unitary operator acting on L2 (µR ), where µR is the right invariant Haar

measure.
Theorem 3.33. (Krein, Weil, Segal) Let G be locally compact unimodular
(abelian or not). Then the right regular representation decomposes into a direct
integral of irreducible representations
ˆ ⊕
Rg = irrep dµ
Ĝ
where µ is called the Plancherel measure.

Example 3.34. G = T , Ĝ = Z. Irreducible representations {ein(·) }n ∼ Z
(Uy f )(x) = f (x + y)
X
= fˆ(n)χn (x + y)
n
X
= fˆ(n)ei2πn(x+y)
n
X
(Uy f )(0) = f (y) = fˆ(n)ei2πny
n
The Plancherel measure in this case is the counting measure.
Example 3.35. G = R, Ĝ = R. Irreducible representations {eit(·) }t∈R ∼ R.
(Uy f )(x) = f (x + y)
ˆ
= fˆ(t)χt (x + y)dt
ˆR
= fˆ(t)eit(x+y) dt
R
ˆ
(Uy f )(0) = f (y) = fˆ(t)eity dt
R
where the Plancherel measure is the Lebesgue measure on R.
As can be seen that Fourier series and Fourier integrals are special cases of
the decomposition of the right regular representation Rg of a unimodular locally
´⊕
compact group. =⇒ kf k = kfˆk. This is a result that was done 30 years
earlier before the non abelian case. Classical function theory studies other types of
convergence, pointwise, uniform, etc.
Example 3.36. G = H3 . G is unimodular, non abelian. Ĝ is not a group.
Irreducible representations: R\{0} Schrodinger representation, {0} 1-d trivial
representation
Decomposition:
ˆ ⊕
h
Rg = Uirrep hdh
R\{0}
2
For all F ∈ L (G),
ˆ ⊕
(Ug f )(e) = U h f hdh, U h irrep
3.8. SUMMARY OF INDUCED REPREP, d/dx EXAMPLE 84
F (g) = (Rg F )(e)

ˆ ⊕
= eih(c+bx) f (x + a)(U h F ) hdh
R\{0}
ˆ
F̂ (h) = (Ugh F )dg
G
Plancherel measure: hdh and the point measure δ0 at zero.
Example 3.37. G ax + b group, non abelian. Ĝ not a group. 3 irreducible
representations: +, −, 0 but G is not unimodular.
The duality question may also be asked for discrete subgroups. This leads
to remarkable applications in automorphic functions, automorphic forms, p-adic
nubmers, compact Riemann surface, hypobolic geometry, etc.
Example 3.38. Cyclic group of order n. G = Z/nZ ≃ {0, 1, · · · , n−1}. Ĝ = G.
This is another example where the dual group is identical to the group itself. Let
ζ = ei2π/n be the primitive nth -root of unity. k ∈ Zn , l = {0, 1, . . . , n − 1}
2πkl
χl (k) = ei n
In this case, Segal’s theorem gives finite Fourier transform. U : l2 (Z) → l2 (Ẑ) where
1 X kl
U f (l) = √ ζ f (k)
N k
3.8. Summary of induced reprep, d/dx example

We study decomposition of group representations. Two cases: abelian and non
abelian. The non abelian case may be induced from the abelian ones.
non abelian
• semi product G = HN oftne H, N are normal.
• G simple. G does not have normal subgroups. Lie algebra does not have
any ideals.
Example 3.39. SL2 (R) (non compact)

a b
, ad − bc = 1
c d
with Lie algebra
sl2 (R) = {X : tr(X) = 0}
sl2 is generated by

0 1 0 −1 1 0
.
1 0 1 0 0 −1

0 −1 cos t − sin t
generates the one-parameter group ≃ T whose dual
1 0 sin t cos t
group is Z, where
χn (g(t)) = g(t)n = eitn .
May use this to induce a representation of G. This is called principle series. Need
to do something else to get all irreducible representations.
A theorem by Iwasawa states that simple matrix group (Lie group) can be
decomposed into
G = KAN
where K is compact, A is abelian and N is nilpotent. For example, in the SL2
case,
s
cos t − sin t e 0 1 u
SL2 (R) = .
sin t cos t 0 e−s 0 1
The simple groups do not have normal subgroups. The representations are much
more difficult.
3.8.1. Induced representation. Suppose from now on that G has a normal
abelian subgrup N ⊳ G, and G = H ⋉ N The N ≃ Rd and N ∗ ≃ (Rd )∗ = Rd . In
this case
χt (ν) = eitν
for ν ∈ N and t ∈ N̂ = N ∗ . Notice that χt is a 1-d irreducible representation on C.
Let Ht be the space of functions f : G → C so that
f (νg) = χt (ν)f (g).
On Ht , define inner product so that
ˆ ˆ
kf k2Ht := |f (g)|2 = kf (g)k2 dm
G G/N
where dm is the invariant measure on N \G ≃ H.

Define Ut = indG N (χt ) ∈ Rep(G, Ht ). Define Ut (g)f (x) = f (xg), for f ∈ Ht .
Notice that the representation space of χt is C, 1-d Hilbert space, however, the
representation space of Ut is Ht which is infinite dimensional. Ut is a family of
irreducible representations indexed by t ∈ N ≃ N̂ ≃ Rd .
Note 3.40. Another way to recognize induced representations is to see these
functions are defined on H, not really on G.
Define the unitary transformation Wt : Ht → L2 (H). Notice that H ≃ N \G is
a group, and it has an invariant Haar measure. By uniqueness on the Haar measure,
this has to be dm. It would be nice to cook up the same space L2 (H) so that all
induced representations indexed by t act on it. In other words, this Hilbert space
L2 (H) does not depend on t. Wt is defined as
Wt Ft (h) = Ft (h).
So what does the induced representation look like in L2 (H) then? Recall by
definition that
Ut (g) := Wt indG ∗
χt (g)Wt
and the following diagram commutes.
indG
χt
Ht / Ht
Wt Wt
Ut

L2 (H) / L2 (H)
Let f ∈ L2 (H).
Ut (g)f (h) = Wt indG ∗
χt (g)Wt f (h)
= (indG ∗
χt (g)Wt f )(h)
= (Wt∗ f )(hg).
Since G = H ⋉ N , g is uniquely decomposed into g = gN gH . Hence hg = hgN gH =
−1
gN gN hgN gH = gN h̃gH and
Ut (g)f (h) = (Wt∗ f )(hg)
= (Wt∗ f )(gN h̃gH )
= χt (gN )(Wt∗ f )(h̃gH )
−1
= χt (gN )(Wt∗ f )(gN hgN gH )
This last formula is called the Mackey machine.
The Mackey machine does not cover many important symmetry groups in
physics. Actually most of these are simple groups. However it can still be ap-
plied. For example, in special relativity theory, we have the Poincare group L ⋉ R4
where R4 is the normal subgroup. The baby version of this is when L = SL2 (R).
V. Bargman fomulated this baby version. Wigner poineered the Mackey machine,
long before Mackey was around.
Once we get unitary representations, differentiate it and get self-adjoint al-
gebra of operators (possibly unbounded). These are the observables in quantum
mechanics.
Example 3.41. Z ⊂ R, Ẑ = T . χt ∈ T , χt (n) = eitn . Let Ht be the space of
functions f : R → C so that
f (n + x) = χt (n)f (x) = eint f (x).
Define inner product on Ht so that
ˆ 1
kf k2Ht := |f (x)|2 dx.
0
2
Define indR
χt (y)f (x) = f (x + y). Claim that Ht ≃ L [0, 1]. The unitary transfor-
2
mation is given by Wt : Ht → L [0, 1]
(Wt Ft )(x) = Ft (x).
Let’s see what indR
χt (y) looks like on L2 [0, 1]. For any f ∈ L2 [0, 1],
(Wt indG ∗
χt (y)Wt f )(x) = (indG ∗
χt (y)Wt f )(x)
= (Wt∗ f )(x + y)
Since y ∈ R is uniquely decomposed as y = n + x′ for some x′ ∈ [0, 1), therefore
(Wt indG ∗
χt (y)Wt f )(x) = (Wt∗ f )(x + y)
= (Wt∗ f )(x + n + x′ )
= (Wt∗ f )(n + (−n + x + n) + x′ )
= χt (n)(Wt∗ f )((−n + x + n) + x′ )
= χt (n)(Wt∗ f )(x + x′ )
= eitn (Wt∗ f )(x + x′ )
3.9. CONNECTION TO NELSON’S SPECTRAL THEORY 87
Note 3.42. Are there any functions in Ht ? Yes, for example, f (x) = eitx . If
f ∈ Ht , |f | is 1-periodic. Therefore f is really a function defined on Z\R ≃ [0, 1].
Such a function has the form
X X
f (x) = ( cn ei2πnx )eitx = cn ei(2πn+t)x .
Any 1-periodic function g satisfies the boundary condition g(0) = g(1). f ∈ Ht has
a modified boundary condition where f (1) = eit f (0).
3.9. Connection to Nelson’s spectral theory

In Nelson’s notes, a normal representation has the form (counting multiplicity)
⊕
X
ρ= nπ Hn , Hn ⊥ Hm
where
nπ = π ⊕ · · · ⊕ π (n times)
is a representation acting on the Hilbert space
⊕
X
K = lZ2 n ⊗ K.
In matrix form, this is a diagonal matrix with π repeated on the diagonal n times.
n could be 1, 2, . . . , ∞. We apply this to group representations.
Locally compact group can be divided into the following types.
• abelian
• non-abelian: unimodular, non-unimodular
• non-abelian: Mackey machine, semidirect product e.g. H3 , ax + b; simple
group SL2 (R). Even it’s called simple, ironically its representation is
much more difficult than the semidirect product case.
We want to apply these to group representations.
Spectral theorem says that given a normal operator A, we may define f (A)
for quite a large class of functions, actually all measurable functions. One way to
define f (A) is to use the multiplication version of the spectral theorem, and let
f (A) = F f (Â)F −1 .
The other way is to use the projection-valued meaure version of the spectral theo-
rem, write
ˆ
A = λP (dλ)
ˆ
f (A) = f (λ)P (dλ).
The effect is ρ is a representation of the abelian algebra of measurable functions

onto operators action on some Hilbert space.
ρ : f 7→ ρ(f ) = f (A)
ρ(f g) = ρ(f )ρ(g)
To imitate Fourier transform, let’s call fˆ := ρ(f ). Notice that fˆ is the multiplication
operator.
Example 3.43. G = (R, +), group algebra L1 (R). Define Fourier transform
ˆ
ˆ
f (t) = f (x)e−itx dx.
{eitx }t is a family of 1-dimensional irreducible representation of (R, +).

Example 3.44. Fix t, H = C, ρ(·) = eit(·) ∈ Rep(G, H). From the group
representation ρ, we get a group algebra representation ρ̃ ∈ Rep(L1 (R), H) defined
by
ˆ ˆ
ρ̃(f ) = f (x)ρ(x)dx = f (x)eitx dx
It follows that
fˆ(ρ) := ρ̃(f )
f[ ⋆g = f[ ⋆ g = fˆĝ
i.e. Fourier transform of f ∈ L1 (R) is a representation of the group algebra L1 (R)
on to the 1-dimensional Hilbert space C. The range of Fourier transform in this
case is 1-d abelian algebra of multiplication operators, multiplication by complex
numbers.
Example 3.45. H = L2 (R), ρ ∈ Rep(G, H) so that
ρ(y)f (x) := f (x + y)
i.e. ρ is the right regular representation. The representation space H in this
case is infinite dimensional. From ρ, we get a group algebra representation ρ̃ ∈
Rep(L1 (R), H) where
ˆ
ρ̃(f ) = f (y)ρ(y)dy.
Define
fˆ(ρ) := ρ̂(f )
then fˆ(ρ) is an operator acting on H.
ˆ
fˆ(ρ)g = ρ̃(f )g = f (y)ρ(y)g(·)dy
ˆ
= f (y)(Ry g)(·)dy
ˆ
= f (y)g(· + y)dy.
If we have used the left regular representation, instead of the right, then
ˆ
fˆ(ρ)g = ρ̃(f )g = f (y)ρ(y)g(·)dy
ˆ
= f (y)(Ly g)(·)dy
ˆ
= f (y)g(· − y)dy.
Hence fˆ(ρ) is the left or right convolution operator.

Back to the general case. Given a locally compact group G, form the group
algebra L1 (G), and define the left and right convolutions as
ˆ ˆ
(ϕ ⋆ ψ)(x) = ϕ(g)ψ(g −1 x)dL g = ϕ(g)(Lg ψ)dL g
ˆ ˆ
(ϕ ⋆ ψ)(x) = ϕ(xg)ψ(g)dR g = (Rg ϕ)ψ(g)dR g
Let ρ(g) ∈ Rep(G, H), define ρ̃ ∈ Rep(L1 (G), H) given by

ˆ
ρ̃(ψ) := ψ(g)ρ(g)dg
G
and write
ψ̂(ρ) := ρ̃(ψ).
ψ̂ is an analog of Fourier transform. If ρ is irreducible, the operators ψ̂ forms an

abelian algebra. In general, the range of this generalized Fourier transform gives
rise to a non abelian algebra of operators.
For example, if ρ(g) = Rg and H = L2 (G, dR ), then
ˆ ˆ
ρ̃(ψ) = ψ(g)ρ(g)dg = ψ(g)Rg dg
G G
and
ˆ ˆ
ρ̃(ψ)ϕ = ψ(g)ρ(g)ϕdg = ψ(g)(Rg ϕ)dg
ˆG G
= ψ(g)ϕ(xg)dg
G
= (ϕ ⋆ ψ)(x)
Example 3.46. G = H3 ∼ R3 . Ĝ = {R\{0}} ∪ {0}. 0 ∈ Ĝ corresponds to the

trivial representation, i.e. g 7→ Id for all g ∈ G.
ρh : G → L2 (R)
ρh (g)f (x) = eih(c+bx) f (x + a) ≃ indG

H (χh )
where H is the normal subgroup {b, c}. It is not so nice to work with indG H (χh )
directly, so instead, we work with the equivalent representations, i.e. Schrodinger
representation. See Folland’s book on abstract harmonic analysis.
ˆ
ψ̂(h) = ψ(g)ρh (g)dg
G
Notice that ψ̂(h) is an operator acting on L2 (R). Specifically,

ˆ
ψ̂(h) = ψ(g)ρh (g)dg
G
˚
= ψ(a, b, c)eih(c+bx) f (x + a)dadbdc
¨ ˆ
= ψ(a, b, c)eihc dc f (x + a)eihbx dadb
¨
= ψ̂(a, b, h)f (x + a)eihbx dadb
ˆ ˆ
ihbx
= ψ̂(a, b, h)e db f (x + a)da
ˆ
= ψ̂(a, hx, h)f (x + a)da

= ψ̂(·, h·, h) ⋆ f (x)
Here the ψ̂ on the right hand side in the Fourier transform of ψ in the usual sense.
Therefore the operator ψ̂(h) is the one so that

L2 (R) ∋ f 7→ ψ̂(·, h·, h) ⋆ f (x).
If ψ ∈ L1 (G), ψ̂ is not of trace class. But if ψ ∈ L1 ∩ L2 , then ψ̂ is of trace class.

ˆ ⊕ ˆ ˆ
tr ψ̂ ∗ (h)ψ̂(h) dµ = |ψ|2 dg = ψ̄ψdg
R\{0}
where µ is the Plancherel measure.

If the group G is non unimoduler, the direct integral is lost (not orthogonal).
These are related to coherent states from physics, which is about decomposing
Hilbert into non orthogonal pieces.
Important observables in QM come in pairs (dual pairs). For example, position
- momentum; energy - time etc. The Schwartz space S(R) has the property that
[ = S(R). We look at the analog of the Schwartz space. h 7→ ψ̂(h) should
S(R)
decrease faster than any polynomials. P
Take ψ ∈ L1 (G), Xi in the Lie algebra, form △ = Xi2 . Require that
△n ψ ∈ L1 (G), ψ ∈ C ∞ (G).
For △n , see what happens in the transformed domain. Notice that
d
(R tX ψ) = X̃ψ
dt t=0 e
where X 7→ X̃ represents the direction vector X as a vector field.
Let G be any Lie group. ϕ ∈ Cc∞ (G), ρ ∈ Rep(G, H).
ˆ
dρ(X)v = (X̃ϕ)(g)ρ(g)vdg
where ˆ
v= ϕ(g)ρ(g)wdg = ρ(ϕ)w. generalized convolution
If ρ = R, the n ˆ
v= ϕ(g)R(g)
X̃(ϕ ⋆ w) = (Xϕ) ⋆ w.
Example 3.47. H3
∂
a →
∂a
∂
b 7→
∂b
∂
c 7→
∂c
get standard Laplace operator. {ρh (ϕ)w} ⊂ L2 (R) . ” = ” due to Dixmier.
{ρh (ϕ)w} is the Schwartz space.
2 2
d d
+ (ihx)2 + (ih)2 = − (hx)2 − h2
dx dx
Notice that 2
d
− + (hx)2 + h2
dx
is the Harmonic oscilator. Spectrum = hZ+ .
CHAPTER 4
Unbounded Operators
4.1. Unbounded operators, definitions

We study unitary representation of Lie groups, one there is a representation, it
can be differentiated and get a representation of the Lie algebra. These operators
are the observables in quantum mechanics. They are possibly unbounded.
4.1.1. Domain.
Example 4.1. d/dx and Mx in QM, acting on L2 with dense domain the
Schwartz space.
An alternative way to get a dense domain, a way that works for all representa-
tions, is to use Garding space, or C ∞ vectors. Let u ∈ H and define
ˆ
uϕ := ϕ(g)Ug udg
where ϕ ∈ Cc∞ , and U ∈ Rep(G, H). Let ϕǫ be an approximation of identity. Then

for functions on Rd , ϕǫ ⋆ ψ → ψ as ǫ → 0; and for C ∞ vectors, uϕǫ → u, as ǫ → 0
in H i.e. in the k·kH - norm. The set {uϕ } is dense in H. It is called the Garding
space, or C ∞ vectors, or Schwartz space. Differentiate Ug and get a Lie algebra
representation
d
ρ(X) := t=0 U (etX ) = dU (X).
dt
Lemma 4.2. kuϕǫ − uk → 0, as ǫ → 0.
´
Proof. Since uϕǫ − u = ϕǫ (g)(u − Ug u)dg, we have
ˆ
kuϕǫ − uk = k ϕǫ (g)(u − Ug u)dgk
ˆ
≤ ϕǫ (g)ku − Ug ukdg
´
where we used the fact that ϕǫ = 1. Notice that we always assume the represen-
tations are norm continuous in the g variable, otherwise it is almost impossible to
get anything interesting. i.e. asuume U being strongly continuous. So for all δ > 0,
there is a neigborhood O of e ∈ G so that u − Ug u < δ for all g ∈ O. Choose ǫδ so
that ϕǫ is upported in O for all ǫ < ǫδ . Then the statement is proved.
Note 4.3. There are about 7 popular kernels in probability theory. Look at
1
any book on probability theory. One of them is the Cauchy kernel πx 1+x 2.
Note 4.4. Notice that not only uϕ is dense in H, their derivatives are also
dense in H.
92
4.2. SELF-ADJOINT EXTENSIONS 93
4.2. Self-adjoint extensions

In order to apply spectral theorem, one must work with self adjoint operators
including the unbounded ones. Some examples first.
In quantum mechanics, to understand energy levels of atoms and radiation,
the engegy level comes from discrete packages. The interactions are givn by Colum
Law where
cjk
H = −△~r +
krj − rk k
and Laplacian has dimension 3 × #(electrons).
In Schrodinger’s wave mechanics, one needs to solve for ψ(r, t) from the equa-
tion
1 ∂
Hψ = ψ.
i ∂t
If we apply spectral theorem, then ψ(t) = eitH ψ(r, t = 0). This shows that motion
in qm is governed by unitary operators. The two parts in Shrodinger equation are
separately self-adjoint, but justification of the sum being self-adjoint wasn’t made
rigorous until 1957. Kato wrote a book called perturbation theory. It is a summary
of the sum of self-adjoint operators.
In Heisenberg’s matrix mechanics, he suggested that one should look at two
states and the transition probability between them.
hψ1 , Aψ2 i = hψ1 (t), Aψ2 (t)i, ∀t.
If ψ(t) = eitH ψ, then it works. In Heisenberg’s picture, one looks at evolution
of the observables e−itH AeitH . In Shrodinger’s picture, one looks at evolution of
states. The two point of views are equivalent.
Everything so far is based on application of the spectral theorm, which requires
the operators being self-adjoint in the first place.
4.2.1. Self-adjoint extensions.

Lemma 4.5. R(A)⊥ = N (A∗ ). N (A∗ )⊥ = cl(R(A)).
Proof. y ∈ R(A)⊥ ⇐⇒ hAx, yi = 0, ∀x ∈ D(A) ⇐⇒ x 7→ hAx, yi being the
zero linear functional. Therefore by Riesz’s theorem, there exists unique y ∗ = A∗ y
so that hAx, yi = hx, A∗ yi = 0, i.e. y ∈ N (A∗ ).
Note 4.6. In general, (set)⊥⊥ = cl(span(set)).
Definition 4.7. Define the defeciency spaces and defeciency indecies as
D+ := R(A + iI)⊥ = N (A∗ − iI) = {v ∈ dom(A∗ ) : A∗ v = +iv}
D− := R(A − iI)⊥ = N (A∗ + iI) = {v ∈ dom(A∗ ) : A∗ v = −iv}
d+ := dim(D+ )
d− := dim(D− )
Von Neumann’s notion of defeciency space expresses the extent to which A∗
is bigger than A. One wouldn’t expect to get a complex eigenvalue for a self
adjoint operator. We understand A not being self adjoint by looking at its “wrong”
eigenvalues. This reveals that A∗ is defined on a bigger domain. The extend that
A∗ is defined on a bigger domain is refleced on the “wrong” eigenvalues.
D+ D−
R(A + i) R(A − i)
Definition 4.8. Define the Caley transform
CA : R(A + i) → R(A − i)
(A + i)x 7→ (A − i)x
i.e.
A−i
CA = (A − i)(A + i)−1 = ” ”
A+i
The inverse map is given by
1 + CA
A = i(1 + CA )(1 − CA )−1 = ”i ”.
1 − CA
Lemma 4.9. CA is a partial isometry.
Proof. We check that

k(A + i)xk2 = hAx + ix, Ax + ixi
= kAxk2 + kxk2 + hAx, ixi + hix, Axi
= kAxk2 + kxk2 + ihAx, xi − ihx, Axi
= kAxk2 + kxk2 + ihx, Axi − ihx, Axi
= kAxk2 + kxk2
where the last two equations follows from A being Hermitian (symmetric). This
shows that CA is a partial isometry.
CA is only well-defined on R(A + i). Since CA is an isometry from R(A + i)

onto R(A − i), it follows that if R(A + i) is dense in H, then CA extends uniquely
to a unitary operator C̃A on H, and thus D+ = 0. The only way that D+ is non
zero is that R(A + i) is not dense in H.
Failure of A being self adjoint ⇐⇒ failure of CA being everywhere defined; A
Hermitian ⇐⇒ CA is a partial isometry (by lemma above); Von Neumann’s method:
look at extensions of CA , and transform the result back to A.
Theorem 4.10. (VN) CA extends to H if and only if d+ = d− .
Example 4.11. d+ = d− = 1. Let e± be corresponding eigenvalues. e+ 7→ ze−
is the unitary operator sending one to the other eigenvalue. It is clear that |z| = 1.
Hence the self adjoint extension is indexed by U1 (C).
Example 4.12. d+ = d− = 2, get a family of extensions indexed by U2 (C).
Remark 4.13. M. Stone and Von Neumann are the two pioneers who worked
at the same period. They were born at about the same time. Stone died at 1970’s
and Von Neumann died in the 1950’s.
Definition 4.14. A conjugation is an operator J so that J 2 = 1 and it is
conjugate linear i.e. J(cx) = c̄Jx.
Theorem 4.15. (VN) If there exists a conjugation J such that AJ = JA, then
d+ = d− .
Proof. Claim that if A commutes with J, so does A∗ . Assuming this is true,

then we claim that J : D+ → D− is a bijection. Suppose A∗ v+ = iv+ . Then
A∗ (Jv+ ) = JA∗ v+ = J(iv+ ) = −iJv+ .

Theorem 4.16. (VN) A ⊂ A∗ , A closed. Then D(A∗ ) = D(A) ⊕ D+ ⊕ D− .
Proof. It is clear that D(A), D+ and D− are subspaces of D(A∗ ). D+ ∩D− =
0, since if Ax = ix and Ax = −ix implies x = 0. D(A) ∩ D+ = 0 as well, since if x
is in the intersection then
hx, Axi = hx, A∗ xi = ikxk2
but hx, Axi being a real number implies that x = 0. Similarly, D(A) ∩ D− = 0.
Therefore, D(A) ⊕ D+ ⊕ D− ⊂ D(A∗ ). To show the two sides are equal, use the
graph norm of A∗ k·kG , show that x ⊥ D(A)⊕ D+ ⊕ D− = 0 implies that kxkG = 0.
Another proof: let f ∈ D(A∗ ) and we will decompose f into the direct sum of
three parts. Since H = R(A + i) ⊕ D+ , therefore (A∗ + i)f decomposes into
(A∗ + i)f = (A + i)f0 + s
where f0 ∈ D(A) and s ∈ D+ . Write s = 2if+ , for some f+ ∈ D+ . (For example,
take f+ = s/2i.) Thus we have
(A∗ + i)f = (A + i)f0 + 2if+
m
∗
A (f − f0 − f+ ) = −i(f − f0 − f+ )
Define f− = f − f0 − f+ . Then f− ∈ D− , and
f = f0 + f+ + f− .
It remains to show the decomposition is unique. Suppose f0 + f+ + f− = 0. Then
A∗ (f0 + f+ + f− ) = Af0 + if+ − if−
i(f0 + f+ + f− ) = if0 + if+ + if−
Notice that the left hand side of the above equations are zero. Add the two equations
together, we get (A + i)f0 + 2if+ = 0. Since H = R(A + i) ⊕ D+ , f+ = 0. Similarly,
f− = 0 and f0 = 0 as well.
Example 4.17. A = d/dx on L2 [0, 1]. Integration by parts shows that A ⊂ A∗ .
Appendix
semi-direct product

G = HK

Theorem 4.18. H, K commute ⇐⇒ G ≃ H × K.


H ∩K =1
Proof. =⇒ Define ϕ : H × K → G by (h, k) 7→ hk. Then ϕ is a homomor-

phism, because H, K commute; ϕis 1-1, since if hk = 1, then k = h−1 ∈ H ∩ K,
therefore k = 1. This implies that h = 1. Moreover, ϕ is onto, since G = HK. ⇐=
is trivial.
We modify the above theorem.

G = HK

Suppose H ⊳ G . Define H ⋊ K = {(h, k) : h ∈ H, k ∈ K} with multi-


H ∩K =1
plication given by
(h, k1 )(h2 , k2 ) := (h1 k1 h2 k1−1 , k1 k2 ).
This turns H ⋊ K into a group, so that G ≃ H ⋊ K. If H, K commute, then
k1 h2 k1−1 = h2 and we are back to the product group. The direct product is always
abelian. Here K acts on H by conjugation in this case.
Turn it around, start with two groups H, K with a map ϕ : K → Aut(H),
build a bigger group G = H̃ K̃.
Theorem 4.19. The following are equivalent.

(
H, K groups
(1)
K acts on H by ϕ


 G = H̃ K̃

H̃ ⊳ G
(2) Morever, K acts on H via ϕ translates to K̃ acts on


 H̃ ≃ H, K̃ ≃ K

H̃ ∩ K̃ = 1
H̃ by conjugation.
Once we get G = H̃ K̃,from previous discussion, we have G ≃ H̃ ⋊ K̃. H̃ ⋊ K̃ is

unique up to isomorphism. (consider the ax+b group embeded into a matrix group
with dimension greater than 2 × 2.)
How does it work?
96
SEMI-DIRECT PRODUCT 97
Define G = {(h, k) : h ∈ H, k ∈ K} with multiplication and inverse given by

(h1 , k1 )(h2 , k2 ) = (h1 ϕk1 (h2 ), k1 k2 )
−1
(h, k) = (ϕk−1 (h−1 ), k −1 )
Then G is a group.
check:
• the inverse is correct
(a, b)(ϕb−1 (a−1 ), b−1 ) = (aϕb (ϕb−1 (a−1 )), bb−1 )
= (aϕb ◦ ϕb−1 (a−1 ), bb−1 )
= (aϕ1 (a−1 ), 1)
= (aa−1 , 1)
= (1, 1)
• H ≃ H̃ = {(h, 1) : h ∈ H} is a subgroup in G
(a, 1)(b, 1) = (ab, 1)
(a, 1)−1 = (ϕ1−1 (a−1 ), 1−1 )
= (a−1 , 1)
• K ≃ K̃ = {(1, k) : k ∈ K} is a subgroup in G
(1, a)(1, b) = (1, ab)
(1, b)−1 = (1, b−1 )
• H̃ ∩ K̃ = {1} obvious
• H̃ is normal in G
(x, y)(h, 1)(x, y)−1 = (xϕy (h), y)(ϕy−1 (x−1 ), y −1 )
= (xϕy (h)ϕy (ϕy−1 (x−1 )), yy −1 )
= (xϕy (h)x−1 , 1)
• K̃ acts on H̃ by conjugation
(1, y)(h, 1)(1, y)−1 = (ϕy (h), y)(1, y −1 )
= (ϕy (h), 1)
Bibliography
[1] William Arveson. An invitation to C ∗ -algebras. Springer-Verlag, New York, 1976. Graduate
Texts in Mathematics, No. 39.
[2] Richard V. Kadison and John R. Ringrose. Fundamentals of the theory of operator algebras.
Vol. II, volume 16 of Graduate Studies in Mathematics. American Mathematical Society,
Providence, RI, 1997. Advanced theory, Corrected reprint of the 1986 original.
[3] Shôichirô Sakai. C ∗ -algebras and W ∗ -algebras. Springer-Verlag, New York, 1971. Ergebnisse
der Mathematik und ihrer Grenzgebiete, Band 60.
98
View publication stats

Functional Analysis: March 2010

Uploaded by

Copyright:

Available Formats

Functional Analysis: March 2010

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Functional Analysis: March 2010

Uploaded by

Copyright:

Available Formats

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

Article · March 2010

DVT Prophylaxis View project

Fractal Calculus View project

The user has requested enhancement of the downloaded file.

Chapter 1. Elementary Facts 6

Chapter 2. GNS, Representations 25

Chapter 3. Appliations to Groups 61

Chapter 4. Unbounded Operators 92

1.1. Transfinite induction

interesting realizations of an abstract Hilbert space come in when we make a choice

1.2. Dirac’s notation

where F is a finite set of orthonormal vectors in H, is a self-adjoint projection.

1.2.1. connection to quamtum mechanics. Quamtum mechanics was born

1.3. Operators in Hilbert space

Proof. Since P is a self-adjoint projection,

With a standard approximation of measurable functions by simple functions, we

The above equation is a generalization of the substitution formula in calculus.

Claim also that

1.4. Lattice structure of projections

has an orthonormal basis {h0 , . . . , hn−1 }, so that spanVn = span{1, x, . . . , xn−1 }.

The next example constructs the Haar wavelet.

P H ⊂ QH ⇔ P = P Q. This is similar to set operation where A ∩ B =

Lemma 1.27. P, Q ∈ proj(H). Then P + Q ∈ proj(H) if and only if P Q =

Proof. Notice that

In terms of characteristic functions,

Proof. The proof is trivial. Notice that

where we used that fact that

1.5. Ideas in the spectral theorem

f induces a Borel measure µf (·) = µ ◦ f −1 (·) on R, supported on f (X).

and we may define operator

where P is a projection-valued measure defined on the Borel σ-algebra of R. Notice

hence we may define ˆ

as the operator so that for all ϕ ∈ H,

where we used the fact that P (A)P (B) = 0 if A ∩ B = φ.

Remark 1.33. Hϕ is called the cyclic space generated by ϕ. Before we can

Therefore P is a projection-valued measure.

Proof. For n ∈ N, define

where 0 ≤ i ≤ n2−n − 1. Equivalently, sn can be written using characteristic

Notice that on each interval [i2−n , (i + 1)2−n ),

for all x ∈ [0, ∞].

Proposition (1.38) and its corollary immediate imply the following.

Proof. If f ∈ L∞ , then there exists a sequence of simple functions sn so that

Suppose f is unbounded. By proposition () and its corollaries, there exists a

1.6. Spectral theorem for compact operators

2.1. Representations, GNS, primer of multiplicity

Kf is a solution to the differential equation

It follows that the corresponding increament process {Xt − Xs } ∼ N (0, t − s). P

which is a compact Hausdorff space.

An very important application of the spectral theorem of compact operators in

2.1.2. Idea of multiplicity. Recall that A = A∗ if and only if

where P is a projection-valued measure (PVM). The simplest example of a PVM

Example 2.1. We represent A as the operator Mf of multiplication by f on

Notice that χEk is a rank s1 projection. Mf is compact if and only if it is of the

Define µf (·) = µ ◦ f −1 (·), then

To check U preserves the L2 -norm,

In general, let A be a self-adjoint operator acting on H. Then there exists a

2.1.3. GNS. The multiplication version of the spectral theorm is an exmaple