Group C*-algebras A.K. Recall 1 that if G is a nonempty set, the linear space

Similar documents
, the. L and the L. x x. max. i n. It is easy to show that these two norms satisfy the following relation: x x n x = (17.3) max

Generalized Orlicz Spaces and Wasserstein Distances for Convex-Concave Scale Functions

EECE 301 Signals & Systems Prof. Mark Fowler

Laplace Transform. Inverse Laplace Transform. e st f(t)dt. (2)

Introduction to SLE Lecture Notes

Question 1: Question 2: Topology Exercise Sheet 3

6.8 Laplace Transform: General Formulas

Randomized Perfect Bipartite Matching

Mathematische Annalen

A Guided Tour in the Topos of Graphs

Analysis of a Non-Autonomous Non-Linear Operator-Valued Evolution Equation to Diagonalize Quadratic Operators in Boson Quantum Field Theory

4 Sequences of measurable functions

To become more mathematically correct, Circuit equations are Algebraic Differential equations. from KVL, KCL from the constitutive relationship

Matrix Versions of Some Refinements of the Arithmetic-Geometric Mean Inequality

Lower and Upper Approximation of Fuzzy Ideals in a Semiring

Fréchet derivatives and Gâteaux derivatives

Optimality Conditions for Unconstrained Problems

18.03SC Unit 3 Practice Exam and Solutions

Finish reading Chapter 2 of Spivak, rereading earlier sections as necessary. handout and fill in some missing details!

The multisubset sum problem for finite abelian groups

Fractional Ornstein-Uhlenbeck Bridge

On the Exponential Operator Functions on Time Scales

NECESSARY AND SUFFICIENT CONDITIONS FOR LATENT SEPARABILITY

EE Control Systems LECTURE 2

A Logic of Orthogonality

Example on p. 157

CHAPTER 7. Definition and Properties. of Laplace Transforms

Exponential Sawtooth

1 Solutions to selected problems

An Introduction to Malliavin calculus and its applications

Springer theory and the geometry of quiver ag varieties

Lecture 3: Fourier transforms and Poisson summation

FREE ODD PERIODIC ACTIONS ON THE SOLID KLEIN BOTTLE

18 Extensions of Maximum Flow

Lie Derivatives operator vector field flow push back Lie derivative of

GCD AND LCM-LIKE IDENTITIES FOR IDEALS IN COMMUTATIVE RINGS

U T,0. t = X t t T X T. (1)

A LOGIC OF ORTHOGONALITY

u(t) Figure 1. Open loop control system

FUZZY n-inner PRODUCT SPACE

ū(e )(1 γ 5 )γ α v( ν e ) v( ν e )γ β (1 + γ 5 )u(e ) tr (1 γ 5 )γ α ( p ν m ν )γ β (1 + γ 5 )( p e + m e ).

Flow Networks. Ma/CS 6a. Class 14: Flow Exercises

1 Motivation and Basic Definitions

Math 315: Linear Algebra Solutions to Assignment 6

Chapter 6. Laplace Transforms

t is a basis for the solution space to this system, then the matrix having these solutions as columns, t x 1 t, x 2 t,... x n t x 2 t...

Bilal Khan Department of Mathematics and Computer Science, John Jay College of Criminal Justice, City University of New York, New York, NY 10019, USA.

Lecture Notes 2. The Hilbert Space Approach to Time Series

T-Rough Fuzzy Subgroups of Groups

Groupoidification in Physics

Classification of 3-Dimensional Complex Diassociative Algebras

Chapter 2. First Order Scalar Equations

arxiv: v1 [math.fa] 9 Dec 2018

ESTIMATES FOR THE DERIVATIVE OF DIFFUSION SEMIGROUPS

Flow networks. Flow Networks. A flow on a network. Flow networks. The maximum-flow problem. Introduction to Algorithms, Lecture 22 December 5, 2001

CONTROL SYSTEMS. Chapter 10 : State Space Response

Problem Set If all directed edges in a network have distinct capacities, then there is a unique maximum flow.

Chapter 6. Laplace Transforms

CHAPTER. Forced Equations and Systems { } ( ) ( ) 8.1 The Laplace Transform and Its Inverse. Transforms from the Definition.

Syntactic Complexity of Suffix-Free Languages. Marek Szykuła

Sample Final Exam (finals03) Covering Chapters 1-9 of Fundamentals of Signals & Systems

The Residual Graph. 12 Augmenting Path Algorithms. Augmenting Path Algorithm. Augmenting Path Algorithm

THE GENERALIZED PASCAL MATRIX VIA THE GENERALIZED FIBONACCI MATRIX AND THE GENERALIZED PELL MATRIX

Introduction to Congestion Games

2. VECTORS. R Vectors are denoted by bold-face characters such as R, V, etc. The magnitude of a vector, such as R, is denoted as R, R, V

SOLUTIONS TO ECE 3084

FIXED POINTS AND STABILITY IN NEUTRAL DIFFERENTIAL EQUATIONS WITH VARIABLE DELAYS

23.2. Representing Periodic Functions by Fourier Series. Introduction. Prerequisites. Learning Outcomes

The Residual Graph. 11 Augmenting Path Algorithms. Augmenting Path Algorithm. Augmenting Path Algorithm

ANSWERS TO ODD NUMBERED EXERCISES IN CHAPTER

Math 2142 Exam 1 Review Problems. x 2 + f (0) 3! for the 3rd Taylor polynomial at x = 0. To calculate the various quantities:

Essential Maps and Coincidence Principles for General Classes of Maps

A remark on the H -calculus

Reminder: Flow Networks

CHAPTER 7: SECOND-ORDER CIRCUITS

Bernoulli numbers. Francesco Chiatti, Matteo Pintonello. December 5, 2016

NEW EXAMPLES OF CONVOLUTIONS AND NON-COMMUTATIVE CENTRAL LIMIT THEOREMS

Solutions from Chapter 9.1 and 9.2

Convergence of the Neumann series in higher norms

Linear Algebra Primer

Note on Matuzsewska-Orlich indices and Zygmund inequalities

6.003: Signals and Systems. Relations among Fourier Representations

SOLUTIONS TO ALGEBRAIC GEOMETRY AND ARITHMETIC CURVES BY QING LIU. I will collect my solutions to some of the exercises in this book in this document.

Heat kernel and Harnack inequality on Riemannian manifolds

CSC 364S Notes University of Toronto, Spring, The networks we will consider are directed graphs, where each edge has associated with it

FLAT CYCLOTOMIC POLYNOMIALS OF ORDER FOUR AND HIGHER

Algorithmic Discrete Mathematics 6. Exercise Sheet

arxiv:math/ v1 [math.oa] 12 Oct 2004

t )? How would you have tried to solve this problem in Chapter 3?

First variation formula in Wasserstein spaces over compact Alexandrov spaces

THE MATRIX-TREE THEOREM

Energy Equality and Uniqueness of Weak Solutions to MHD Equations in L (0,T; L n (Ω))

23.5. Half-Range Series. Introduction. Prerequisites. Learning Outcomes

String diagrams. a functorial semantics of proofs and programs. Paul-André Melliès. CNRS, Université Paris Denis Diderot.

CONTROL SYSTEMS. Chapter 3 Mathematical Modelling of Physical Systems-Laplace Transforms. Prof.Dr. Fatih Mehmet Botsalı

The Asymptotic Behavior of Nonoscillatory Solutions of Some Nonlinear Dynamic Equations on Time Scales

Piecewise-Defined Functions and Periodic Functions

On Line Supplement to Strategic Customers in a Transportation Station When is it Optimal to Wait? A. Manou, A. Economou, and F.

Notes on cointegration of real interest rates and real exchange rates. ρ (2)

Admin MAX FLOW APPLICATIONS. Flow graph/networks. Flow constraints 4/30/13. CS lunch today Grading. in-flow = out-flow for every vertex (except s, t)

Transcription:

Group C*-algebra A.K. Recall 1 ha if G i a nonempy e, he linear pace c oo (G) = {f : G C : upp f finie } ha a Hamel bai coniing of he funcion {δ : G} where { 1, = δ () = 0, Thu every f c oo (G) i a finie um f = G f()δ. The Hilber pace l 2 (G) i he compleion of c oo (G) wih repec o he calar produc f, g = f()g() G and hen {δ : G} become an orhonormal bai of l 2 (G). In cae G i a group, he group operaion (, ) and 1 exend linearly o make c oo (G) ino a *-algebra: we define δ δ = δ and (δ ) = δ 1, o ha ( ) ( ) f g = f()δ g()δ = f()g()δ, and 1 gpar, 16/1/07 ( ) f = f()δ = f()δ 1 1

in oher word (eing r = ) f g = ( ) f()g( 1 r) δ r = r r ( ) f(r 1 )g() δ r and (changing o r = 1 ) f = r f(r 1 )δ r. Thu (f g)(r) = f()g( 1 r) = f(r 1 )g() (r G) and (f )(r) = f(r 1 ) (r G). We may alo complee c oo (G) wih repec o he l 1 norm f 1 = f() o obain he Banach pace l 1 (G). Noe ha becaue of he relaion f g 1 f 1 g 1 and f 1 = f 1 (he proof of he fir one i eay 2 and he econd one i obviou) he muliplicaion and he involuion exend coninuouly o l 1 (G), which become a Banach algebra wih iomeric involuion, alhough rarely a C*-algebra. For example if e, and 2 are differen elemen of G and f = δ 1 +δ e δ hen f 1 = 1 + 1 + 1 and f f = (δ + δ e δ 1)(δ 1 + δ e δ ) = δ 2 + 3δ e δ 2 hence f f 1 = 1 + 3 + 1. In order o equip c oo (G) wih a uiable C*-norm, we udy i *-repreenaion on Hilber pace. 2 r f()g( 1 r) r f(). g( 1 r) = r f(). g( 1 r) = f(). r g( 1 r) = f() g() 2

The lef regular repreenaion Le H = l 2 (G). Each G define a uniary operaor λ on H by he formula λ ( ξ()δ ) = ξ()δ (ξ = ξ()δ l 2 (G)). For example, if G = Z hen λ n = U n where U i he bilaeral hif, U(δ n ) = δ n+1, on l 2 (Z). Making he change of variable r =, we find (λ ξ)(r) = ξ( 1 r) (r G). Noe ha λ i a well-defined linear iomery, becaue λ (ξ) 2 2 = r ξ( 1 r) 2 = ξ() 2 = ξ 2 2. Alo λ e = I (he ideniy operaor) and becaue λ λ = λ λ (λ δ r ) = λ (δ r ) = δ r = δ ()r = λ (δ r ) for each r G. Since he operaor involved are bounded and linear and he {δ } pan l 2 (G) he claim follow. In paricular i follow ha each λ i inverible wih invere (λ ) 1 = λ 1 and o i i an ono iomery, i.e. a uniary, wih (λ ) = λ 1. Thu we have a group homomorphim G U(B(H)) ino he group of uniary operaor on H = l 2 (G). Thi i called a uniary repreenaion of G on H. The uniary repreenaion λ immediaely exend o a *-repreenaion, alo denoed by λ, of he *-algebra c oo (G) on l 2 (G). More preciely, given f = f()δ c oo (G) we define λ(f) = i.e. (λ(f)ξ)(r) = f()λ f()ξ( 1 r) (ξ l 2 (G)). 3

Thi i a bounded operaor becaue λ(f) = f()λ f() λ = f() = f 1 ince each λ i uniary. In fac hi inequaliy how ha λ exend o a (conracive) map l 1 (G) B(l 2 (G)). The fac ha λ i a *-repreenaion immediaely follow from he properie of i rericion o G: (( ) ( )) ( ) λ f()δ g()δ = λ f()g()δ = f()g()λ,, = ( ) ( ) f()g()λ λ = f()λ g()λ = λ(f)λ(g), (( ) ) ( ) and λ f()δ = λ f()δ 1 = f()λ 1 = ( ) f()λ = f()λ. The above calculaion can be carried ou for any uniary repreenaion of G. The deail are lef a an exercie. Propoiion 1 There i a bijecive correpondence beween uniary repreenaion of G and *-repreenaion of c oo (G): If π : G U(B(H)) i any uniary repreenaion of he group G, he formula ( ) π f()δ = f()π() define a unial *-repreenaion of c oo (G) (and of l 1 (G)) on he ame Hilber pace H which i 1 -conracive. Converely, every unial 1 -conracive *-repreenaion ρ of c oo (G) (or of l 1 (G)) define a uniary repreenaion π by rericion : π() = ρ(δ ) aifying π = ρ. We uually ue he ymbol π for π. 4

Definiion 1 Le Σ be he e of all 1 -conracive *-repreenaion (π, H π ) of c oo (G) (equivalenly, of l 1 (G)). The C*-norm on c oo (G) (or l 1 (G)) i defined by he formula f = up{ π(f) : π Σ}. The group C*-algebra C (G) i defined o be he compleion of c oo (G) (or equivalenly of l 1 (G)) wih repec o hi norm. Remark 2 Fir of all, he e Σ i non-empy: i conain he lef regular repreenaion. Clearly i a eminorm on c oo (G), being he upremum of eminorm, all of which are (by definiion) bounded by 1, hence o i. Alo, aifie he C*-ideniy, becaue all he eminorm f π(f) aify i. 3 Bu why i a norm? In oher word, why i i rue ha f > 0 whenever f c oo (G) i nonzero? The reaon i ha he lef regular repreenaion i faihful on c oo (G) and l 1 (G); hu if f l 1 (G) i nonzero hen λ(f) 0 and o f λ(f) > 0. Indeed if f = f()δ l 1 (G) i nonzero hen here exi G wih f() 0 and hen 4 λ(f)δ e, δ l 2 (G) = f()λ (δ e ), δ = f() δ, δ = f() becaue he δ are orhonormal in l 2 (G). Thu λ(f) 0. The uefulne of C (G) come from he following propery, whoe proof i an immediae conequence of he previou propoiion and he fac ha c oo (G) i a dene *-ubalgebra of C (G). Propoiion 3 There i a bijecive correpondence beween uniary repreenaion of G and unial *-repreenaion of he group C*-algebra C (G). In paricular, he lef regular repreenaion λ exend o a *-repreenaion of C (G) on l 2 (G). However, he fac ha λ i faihful on c oo (G) doe NOT mean ha i exenion remain faihful on C (G)! 3 π(f f) = π(f) π(f) = π(f) 2. 4 ince f() <, he um f()λ converge (aboluely) in he norm of B(l 2 (G)). 5

The image λ(c (G)) in B(l 2 (G)) i a C*-algebra; i equal he cloure of λ(c oo (G)) in he norm of B(l 2 (G)) and i called he reduced C*-algebra Cr (G) of G. In many cae, for example when G i abelian, λ i faihful on C (G), o ha Cr (G) C (G) (iomerically and *-iomorphically). In general, however, Cr (G) i a quoien of C (G) and doe no conain all uniary repreenaion of G. Example 4 Le G = F 2 be he free group in wo generaor a and b; ha i, any elemen of G i a (finie) word of he form a n b m a k b j where n, m, k, j Z and here are no relaion beween a and b. I i known ha he reduced C*-algebra C r (F 2 ) i imple, i.e. i ha no nonrivial cloed wo-ided ideal. Thu all of i repreenaion are iomorphim; Since C r (F 2 ) i obviouly infinie-dimenional, i canno have finie dimenional repreenaion. On he oher hand, he group F 2 doe have uniary repreenaion on finiedimenional pace: Ju ake any wo uniary n n marice U and V and define π(a) = U and π(b) = V. Since here are no relaion beween a and b, hi exend o a uniary repreenaion of F 2 on C n ; for example π(a n b m a k b j ) = U n V m U k V j. Hence C (F 2 ) doe have nonrivial finiedimenional repreenaion: herefore i canno be iomorphic o C (F 2 ). Thu, C r (F 2 ) i a proper quoien of C (F 2 ). Example 5 Le G = Z. If we repreen each n Z by he funcion ζ n (z) = z n, z T, he convoluion produc ζ n ζ m = ζ n+m become poinwie produc, involuion become complex conjugaion, and he elemen of c oo (G) become rigonomeric polynomial. Hence if P C(T) i he e of rigonomeric polynomial we have a *-iomorphim c oo (Z) P : n f(n)δ n p f, where p f (z) n f(n)z n. Noe ha, a oberved earlier, he lef regular repreenaion λ i generaed by λ(1) = U, he bilaeral hif on l 2 (Z), which i uniarily equivalen o muliplicaion by ζ on L 2 (T). Therefore, for each f c oo (Z), λ(f) = p f (U) i uniarily equivalen o he muliplicaion operaor M pf acing on L 2 (T) and o λ(f) = M pf = p f = up{ p f (z) : z T}. I follow ha he cloure C r (Z) of λ(c oo (Z) i iomerically iomorphic o he up-norm cloure of he rigonomeric polynomial, namely C(T). 6

We will how ha C (Z) i iomerically *-iomorphic wih C(T). Since c oo (Z) i -dene in C (Z) and P i -dene in C(T) (Sone- Weierra) i uffice o how ha he norm f on c oo (Z) coincide wih he up norm p f of C(T). For hi, ince we ju proved ha p f = λ(f) f, i i enough o prove he revere inequaliy, namely ha if π i any uniary repreenaion of Z on ome Hilber pace, hen π(f) p f for any f = n f(n)δ n c oo (Z). Indeed le V = π(1); hi i a uniary operaor and π(f) = n f(n)π(n) = n f(n)v n = p f (V ). Now p f (V ) i a normal operaor and hence i norm equal i pecral radiu. By he pecral mapping heorem, σ(p f (V )) = {p f (z) : z σ(v )} {p f (z) : z T} becaue V i uniary and o σ(v ) T. Thu π(f) = p f (V ) up{ p f (z) : z T} = p f. Abelian group The iuaion of hi la example generalize o arbirary abelian group. Briefly, if G i an abelian group, hen of coure c oo (G) i abelian, and hence o i C (G). Thu C (G) C(K), where K i he compac pace of muliplicaive linear funcional on C (G) wih he weak* opology. We idenify he pace K: Define he e of characer of G Ĝ = = {γ : G T : homomorphim}. Wih he opology of poinwie convergence, i i no hard o ee ha hi i a compac pace (a cloed ubpace of he Careian produc T G ) and i i a group wih poinwie operaion. In fac i can be hown o be a opological group (he group operaion are coninuou). I i called he dual group of G. Any γ i a *-repreenaion of G on he Hilber pace C (ince γ()γ() = γ() and γ( 1 ) = (γ()) 1 = γ()) and hu exend (Propoiion 7

3) o a *-repreenaion γ of C (G) on C, i.e. a muliplicaive linear funcional on C (G). Converely, any muliplicaive linear funcional on C (G) reric o a characer on G. Thu here i a bijecion beween he e of characer of G and he e K of muliplicaive linear funcional on C (G). We claim ha hi bijecion i a homeomorphim; ince boh pace are compac and Haudorff, i uffice o prove ha i i coninuou. Le γ i γ in ; hi mean γ i () γ() for each G. To prove ha γ i γ in K, we need o prove ha γ i (a) γ(a) for all a C (G). Fix uch an a. Since c oo (G) i dene in C (G), given ɛ > 0 here exi f c oo (G) wih a f < ɛ. Now each γ i and γ ha norm 1, and o γ i (a f) γ(a f) 2 a f < 2ɛ. On he oher hand, if f i a finie um f()δ, we have γ i (f) γ(f) = f()(γ i () γ()) f(). γ i () γ(). Now ince γ i () γ() for each G, here i i o uch ha γ i () γ() < ɛ for each i i o and each in he finie uppor of f. Combining wih he previou inequaliy we conclude ha γ i (a) γ(a) < (2 + f 1 )ɛ whenever i i o ; hu γ i γ in he weak*-opology. Thi conclude he proof ha and K are homeomorphic; we henceforh idenify K wih and now we can conclude by Gelfand heory ha C (G) C(). In fac he *-iomorphim i given by a â, where â(γ) = γ(a), a C (G) and in paricular ˆf(γ) = f()γ(), f c oo (G). Haar meaure on We now wih o equip wih a uiable probabiliy meaure µ and form L 2 (, µ). We fir define a ae: ω : c oo (G) C : f f(e) 8

Clearly hi i linear 5 and ω(1) = ω(δ e ) = δ e (e) = 1. We check poiiviy: ω(f f) =(f f)(e) = f( 1 ) 2 = f ()f( 1 e) = f( 1 )f( 1 ) = f() 2 0 (1) for all f = f()δ c oo (G). Noe alo ha ω i coninuou in he norm of C (G) C(): Indeed ω(f) = f(e) = λ(f)δ e, δ e λ(f) δ e 2 2 = λ(f) f = ˆf when f c oo (G). Therefore ω exend o a coninuou linear form on he compleion C() and he exenion i a ae. By he Riez repreenaion heorem, here exi a unique Borel probabiliy meaure µ on he compac pace uch ha ω(a) = â(γ)dµ(γ) for all a C (G). (2) Lemma 6 The meaure µ (i) i lef invarian, i.e. µ(γe) = µ(e) for every Borel ube of and any g (where γe = {γγ : γ E}), and (ii) ha full uppor, i.e. µ(u) > 0 for every nonempy open e U. Proof (i) Fix γ. We claim ha ˆf(γ 1 γ )dµ(γ ) = ˆf(γ )dµ(γ ) for all f c oo (G). Indeed, eing g() = γ()f() we eaily find ha ĝ(γ ) = ˆf(γ 1 γ ˆf(γ ) and o 1 γ )dµ(γ ) = ĝ(γ )dµ(γ ) = g(e) = γ(e)f(e) = f(e). Since c oo (G) i dene in C() i follow ha a(γ 1 γ )dµ(γ ) = a(γ )dµ(γ ) for all a C(). 5 Thi i no muliplicaive: he produc on c oo (G) i no poinwie muliplicaion, i i convoluion 9

By uniquene of µ hi implie χ E (γ 1 γ )dµ(γ ) = χ E (γ )dµ(γ ) for every Borel e E. Bu ince χ E (γ 1 γ ) = χ γe (γ ), claim (i) follow. (ii) Le U be a nonempy open e. Oberve ha {γu : γ } i an open cover of (he map γ γγ i a homeomorphim) and o here i a finie ubcover {γ i U : i = 1,..., n}. Now µ(γ i U) = µ(u) by lef invariance, hence ( n ) n µ() = µ γ i U µ(γ i U) = nµ(u). i=1 Since µ() > 0 i follow ha µ(u) > 0. i=1 The Fourier ranform I follow from (2) ha for f c oo (G) (remembering ha he Gelfand ranform i a *-morphim, o ha ĝ f = ĝ ˆf) we have ω(f f) = f fdµ = f ˆfdµ = ˆf ˆfdµ = ˆf(γ) 2 dµ(γ). (3) Combine hi wih (1) o conclude ha ˆf(γ) 2 dµ(γ) = f() 2 for all f c oo (G). Thi equaliy how (if we wrie L 2 () for L 2 (, µ)) ha he linear map (c oo (G), l 2 (G) ) (C(), L 2 () ) : f ˆf i iomeric and ha dene range, and hu exend o a uniary bijecion F : l 2 (G) L 2 () which i called he Fourier ranform. Finally, if f c oo (G) and ξ c oo (G) l 2 (G) we have F (λ(f)ξ) = F (f ξ) = f ξ = ˆf ˆξ = M ˆf ˆξ = M ˆfF ξ 10

where M g denoe muliplicaion by g on L 2 (). The operaor F λ(f) and M ˆfF are boh bounded operaor on l 2 (G) and coincide on he dene ubpace c oo (G); herefore hey are equal: F λ(f) = M ˆfF or F λ(f)f = M ˆf (F i uniary). I follow ha λ(f) = M ˆf. Bu, ince µ ha full uppor, 6 M ˆf = ˆf. Thu finally λ(f) = ˆf for all f c oo (G) o ha he lef regular repreenaion i iomeric on c oo (G). Therefore i exenion o C (G) C() i alo iomeric, hence injecive, and implemen a *-iomorphim beween C (G) and C r (G). We ummarize: Theorem 7 If G i an abelian group and = Ĝ, hen C (G) C() and he Fourier ranform F : l 2 (G) L 2 () implemen a uniary equivalence beween he lef regular repreenaion λ of C (G) on l 2 (G) and he muliplicaion repreenaion g M g of C() on L 2 (). Hence λ i iomeric and o C (G) C r (G). 6 If U i an open e on which ˆf ˆf ɛ, hen ξ = χ U i a nonzero elemen of L 2 () and M ˆf ξ 2 M ˆf ξ 2 ( ˆf ɛ) ξ 2. 11