C*-ALGEBRAS

JACQUES nIXMIER
University of Paris VI

1977

NORTH-HOLLAND PUBLISHING COMPANY AMSTERDAM NEW YORK OXFORD

' NORTH-HOLLAND PUBLISHING COMPANY -1977

All rights reserved. No part of this publication may be reproduced. stored in a retrieval system, or transmitted. in any [orm or by any means, electronic, mechanical. photocopying. recording or otherwise, without the prior permission of the copyright owner,

North-Holland ISBN for the series: 0 7204 2450 X North-Holland ISBN for this llolume: 0 7204 0762 1 A translation of:

LES C*-ALGEBRES ET LEURS REPRESENTATIONS

' BORDAS (Gauthier-Villars), Paris. 1969

Translated by:

Dr. Francis Jellett, Chelsea College. University of London

Published by:

NORTH-HOLLAND PUBLISHING COMPANY AMSTERDAM NEW YORK OXFORD

Sole distributors for the U.S.A. and Canada: ELSEVIER NORTH-HOLLAND. INC. 52 Vanderbilt Avenue New York. N.Y. 10017

Library of Congress Cataloging in Publication Data

Dixmier, Jacques. C*-algebras (North-Holland mathematical library; v. 15) Translation of Les C*-algebres et leurs representations. Bibliography: p. 410 I. C*-algebras. 2. Representations of algebras. 3. Representations of groups. I. Title. QA326.D5213 512.55 77-7133 ISBN 0-7204-0762-1

PRINTED IN THE NETHERLANDS

INTRODUCTION

Let H be a Hilbert space and :(H) the set of bounded linear operators on H. Let A be a subset of :(H) which is closed under addition, multiplication, multiplication by scalars and under the adjoint operation, and which is also a closed subset of :(H) in the operator norm topology. A is then a special type of involutive Banach algebra; such an algebra is called a C*-algebra. The theory of C*-algebras had its origins in 1943 after Gelfand and Naimark showed that among all involutive Banach algebras, C*-algebras could be characterized by a few simple axioms. It was subsequently seen that C*-algebras play a basic role in the study of the represen tations of a very extensive class of involutive Banach algebras; for each algebra B of this class, a C*-algebra A can be constructed such that the representations of B in a Hilbert space can be identified with those of A. For many questions (notably those involving ideals), A is easier to handle than B. In particular, this construction is applicable when B is the algebra of integrable functions on a locally compact group G, so that the study of the unitary representations of G reduces to that of the representations of a certain C*-algebra, called the C*-algebra of G. The study of C*-algebras takes up almost four fifths of this book, where the main results due, among others, to Fell, Glimm, Kadison, Kaplansky, Mackey and Segal are expounded. It seemed to me a pity not to make use of the material thus accumulated together with that contained in my book on von Neumann algebras (Cahiers Scientifiques, fasc. XXV; referred to as [A v N]) to say a little about unitary represen tations of groups. All the more so as the theory of groups provides some of the most interesting examples of C*-algebras. However, the latter pages of the book do not in any way constitute a treatise on group representations. To get an idea of the questions which would have to be tackled in such a work, the reader should consult Mackeys article "Infinite dimensional group representations" (Bull. Amer. Math. Soc., 69, 1963, p. 629-686); only the first dozen or so pages of this sixty page

xii

INTRODUCTION

article are concerned with questions treated in the present work. Fur thermore, Gelfand and the Russian School on (he one hand, and Harish Chandra on the other are hardly mentioned here; in other words my exposition is very incomplete as far as groups are concerned. Although von Neumann algebras are examples of C*-algebras, the theory of C*-algebras in fact depends on that of von Neumann algebras, as will be seen time and again. To save the reader having to refer constantly to [A v N], the essential results concerning von Neumann algebras are assembled in Appendix A. Furthermore, a mixed bag of results, used in the book, is grouped in Appendix B, and the results presented here follow no particular plan. It seemed to me, while writing [A v N], that most of the theorems were already in a more or less definitive form. By contrast, the theory of C*-algebras appeared to me to have a long way to go before a stable state is reached. Each chapter finishes with some additional results, given either without proof or with just a sketch of the proof. Some of the easier of these results can serve as exercises, although it would have been difficult to distinguish these clearly from the others. I have merely indicated, with an asterisk, those results whose proof is really lengthy (e.g. more than three pages), although one should not attach too much importance to this classification. The additional results are not used in the sub sequent chapters. The bibliography includes the essential references for the theory of C*-algebras, while being far from complete in the area of groups. The reader of this book, as with [A v N] is assumed to be well acquainted with general topology, topological vector spaces and in tegration theory. The proofs of results in C*-algebras are presented in reasonable detail, while those dealing with groups are more condensed and assume that the reader is familiar with the properties of the convolution product. On the other hand, the reader is assumed to have a knowledge of the theory of locally compact commutative groups; for, while no call is, in principle, made on this theory, apart from one or two points of detail, reference to the commutative case often sheds light on the problems studied. The results are numbered lexicographically. A reference such as 4.7.5 is self-explanatory. References of the form A 8, B 15 refer to Appendices A and B. In conclusion, a few words on the connection that this book has with

INTRODUCTION

xiii

those of Naimark and Rickart. The books of Naimark and Rickart include, among other things, the general theory of normed algebras, which is not to be found here. On the other hand, in the special case of C*-algebras, the present book is more detailed. There are, of course, some themes common to all three books, but they are relatively few. The small number of general results on normed algebras which are used here are quoted in Appendix B. Concerning the English edition, a few corrections have been made. The bibliography has been completed up to 1975. I thank S. Berberian and J. Brohan for their help in this connection.

CHAPTER 1

NORMED INVOLUTIVE

ALGEBRAS

1.1. Involutive algebras

1.1.1. Let A be an algebra over the field C of complex numbers. An involution in A is a map x ~ x* of A into itself such that
(i) (x*)* = x (ii) (x + y)* = x* + y* (iii) (Ax)* = AX* (iv) (xy)* = y*x*

for any x, yEA and A E C. An algebra over C endowed with an involution is called an involutive algebra. x* is often called the adjoint of x. A subset of A which is closed under the involution operation is said to be self-adjoint. Property (i) implies that an involution in A is necessarily a bijection of A onto itself.
1.1.2. Examples (1) On A = C, the map z ~ z (where z is the complex conjugate of z) is an involution with which A becomes a commutative involutive alge bra. (2) Let X be a locally compact space, and A the algebra of complex valued continuous functions vanishing at infinity on X. Endowed with the map f ~ 1, A is a commutative involutive algebra. When X is a single point, this reduces to example (1). (3) Let H be a Hilbert space and A = 2(H) the algebra of continuous endomorphisms of H. Furnished with the usual adjoint operation, A is an involutive algebra. Examples (2) and (3) will playa fundamental role. (Throughout the book, "Hilbert space" means "complex Hilbert space".) (4) Let G be a unimodular locally compact group and A the con volution algebra L l( G). For each f E L l( G) put f*(s) = f(s -1)(S E G). With the map f ~ A is an involutive algebra.

r.

NORMED INVOLUTIVE ALGEBRAS

[CH. I, I

1.1.3. We now introduce some terminology suggested by example (3) above. Let A be an involutive algebra. An element x E A is said to be hermitian if x* = x and normal if xx" = x*x. An idempotent hermitian element is called a projection. Each hermitian element is normal and the set of hermitian elements is a real vector subspace of A. If x and yare hermitian, we have (xy)* = y*x* = yx, and so xy is hermitian if x and y commute. For every x E A, xx* and x*x are hermitian, although a general hermitian element cannot be so represented as example (1) above shows. 1.1.4. Each x E A can be written uniquely in the form XI + iX 2 with XI X2 hermitian. (In example (1) above, this expression is just the decom position of a complex number into its real and imaginary parts.) In fact, if we put
x
I

=(x + x*) 2

1 x =-(x-x*) 2 2i

then XI and X2 are hermitian and we have x =XI + iX2 Conversely, if x = XI + iX2 with XI and X2 hermitian, we have x* = x l - ix 2 and so
x
I

=2 (x + x*)

1 x = -(x - x*) 2 2i

which proves our assertion. Note that

so that x is normal if and only if

XI

and

X2

commute.

1.1.5. If A possesses a left identity 1, we have, for each x E A,

x . 1* = (l . x*)* = x** = x,

so that 1* is a right identity, and thus 1 = 1* is the identity for A. If X is an invertible element of A, then
(x-I)*x* x**

= (xx- I )* = 1* = 1,

x*(x- I)* = (x-Ix)*

= 1* = 1,

so that x* is invertible, and (X*fl = (x ")"; conversely, if x* is invertible, = x is also invertible. Since (x - A 1)* = x* - A. 1 for each A E C, it follows that
SpAx*

=SPAx.

(In every unital algebra E, the spectrum of an element x, denoted by

CH. I, I)

INVOLUTIVE ALGEBRAS

SP&X or simply Sp x, is the set of scalars A such that x - A 1 is not invertible.) An element x E A is said to be unitary if xx* = x*x = 1, or in other words if x is invertible and x = X*-l. [In example (1) above, the unitary elements are the complex numbers of absolute value 1.] The unitary elements of A constitute a group under multiplication, the unitary group of A; in fact, if x and yare unitary elements of A, then
(xy)*-l = (y*X*)-1 = X*-ly*-l = xy,

so that xy is unitary, and (X-1)*-1 = (x*-Ir l =

X-I,

so that

X-I

is unitary.

1. 1.6. Let A be an involutive algebra. If A is the algebra obtained from A by adjoining an identity to A, it is easy to see that the involution on A can be extended to A in a unique way: we define (A, x)* =(A, x*) for A E C, x E A. For each x E A, we have
Sp~x*

= Sp~x

(For any algebra B, unital or not, SPBX or Spx denotes the set SPBX, where B is the algebra obtained from B by the adjunction of an identity. Clearly, 0 E SPBx for each x E B.) 1.1.7. Let A and B be two involutive algebras. A morphism (resp. isomorphism) of A into B is a map (resp. a bijection) q> of A into B such that q>(x + y) =q>(x) + q>(y), q>(Ax) =Aq>(x), q>(xy) = q>(x)q>(y), q>(x*) = q>(x)* for any x, yEA, A E C. In particular, q> is a morphism of the underlying algebra of A into the underlying algebra of B. In cases of possible confusion, we say more precisely "morphism for the involutive algebra structure" or "morphism for the algebra structure" as the case may be.
1.1 .8. Let A be an involutive algebra. An involutive subalgebra of A is a subalgebra of A which is closed under the involution. The intersection of any family of involutive subalgebras is again an involutive subalgebra; thus, if M is any subset of A, there is a smallest involutive subalgebra of A containing M, namely the intersection of all the involutive subalgebras of A which contain M, and this is called the involutive subalgebra of A generated by M; it is the set of linear combinations of elements of the , x, EMU M*. When M consists of a form XIX2 x, where XI X2" single element x, this subalgebra is commutative if and only if X is normal.

NORMED INVOLUTIVE ALGEBRAS

[CH. I, 2

Let A be an involutive algebra and B a self-adjoint two-sided ideal of A. The involution on A induces an involution on the quotient algebra AIB, and the canonical map of A onto AlB is a morphism. The product of any family of involutive algebras is itself an involutive algebra in a natural way. The reversed algebra of an involutive algebra is itself an involutive algebra when endowed with the same involution. 1.1.9. Let A be an involutive algebra and M a self-adjoint subset of A. Then the commutant M of M in A is an involutive subalgebra of A, and the bicommutant Mil of M in A is thus an involutive subalgebra of A containing M; it is in general different from the involutive subalgebra of A generated by M (for example when A is commutative). If the elements of M commute pairwise, then Me M, so that M:>Mil and Mil is commutative. If x E A and M = {x, x*}, then Mil is commutative if and only if x is normal. 1.1.10. Let A be an involutive algebra. If I is a linear form on A, the function x ....I(x*) on A is also a linear form on A, denoted by f* and called the adjoint of f. Plainly, f** =I, (f + 1)* =f* +1*, (Af)* = A/* if A E C, and I is said to be hermitian if 1= f*. Every linear form I on A has a unique expression of the form II + i/2with 11,12 hermitian; indeed, we have II

="2Y+ f*),

A linear form I on A is hermitian, if and only if I takes real values on the set A h of hermitian elements of A. The map I .... I IA h is an isomorphism of the real vector space of hermitian forms onto the dual space of the real vector space A h If A is commutative and X is a character of A, then X* is also a character of A.

References: [1101], [1323].

1.2. Normed involutive algebras

1.2.1. DEFINITION. A normed involutive algebra is a normed algebra A together with an involution x ....x* such that Ilx*11 = Ilxll for each x E A. If, in addition, A is complete, A is called an involutive Banach algebra.

CR. 1. 2j

NORMED INVOLUTIVE ALGEBRAS

1.2.2. Examples. The four examples of 1.1.2 are examples of involutive Banach algebras if the norms are defined as follows: in example (1) set [z] = [z] for each z E C; in example (2) set IIIII= SUPtEX I/(t)1 for each I E A; in example (3) take the usual norm of ,;t(H); and in example (4), set 11/11 =fa I/(g)ldg for each IE L1(G).

1.2.3. Let A be a normed involutive algebra, and A the involutive algebra obtained from A by the adjunction of an identity. The norm on A can be extended to A in such a way as to make A a normed involutive algebra (for example, one can put II(A, x)11 =IA 1+ [x] for AE C, x E A). Any normal involutive algebra so obtained is called a normed involutive algebra obtained from A by the adjunction of an identity. 1.2.4. Let A and B be two normed involutive algebras. A morphism of A into B will simply mean a morphism of the underlying involutive algebras, without any condition on the norms. On the other hand, an isomorphism will mean a norm-preserving isomorphism of the underly ing involutive algebras. 1.2.5. The closure of an involutive sub algebra of a normed involutive algebra A is itself an involutive subalgebra. If MeA, the smallest closed involutive sub algebra B containing M is called the closed involu tive subalgebra generated by M and is the closure of the involutive sub algebra generated by M. If M consists of a single normal element, then B is commutative. The quotient of a normed involutive algebra by a closed self-adjoint two-sided ideal, the product of a finite number of normed involutive algebras, the reversed algebra of a normed involutive algebra and the completion of a normed involutive algebra are all normed involutive algebras in a natural way. 1.2.6. Let A be a normed involutive algebra. If f is a continuous linear form on A, then 1* is also continuous and 111*11 = Ilfll as the unit ball of A is self-adjoint. The set A h of hermitian elements of A is a real normed vector space. Now let I be a continuous hermitian linear form on A and let g =I IA h Then 11/11 = IIgll; in fact, it is clear that Ilfll~ JlglI; on the other hand, for each E > 0 there exists an x E A such that Ilxll,;;;;1 and I/(x)1 ~ 11/11- E and multiplying x by a scalar of absolute value 1 if necessary, we can assume that I(x) ~ O. Then
Ig(!<x + x*1

=~1/(x)

+l(x*)1 =I(x) ~ 11/11- E

NORMED INVOLUTIVE ALGEBRAS

[CH. I, 3

and since IH(x + x*)11 ,;;;; I, we see that Ilgll ~ 11/11- E and the assertion follows. The continuous hermitian linear forms on A may thus be identified with the continuous real linear forms on A h References: [1101], [1323].

1.3. C*-algebras

1.3.1. DEFINITION. A C*-algebra is an involutive Banach algebra A such that IIxl1 2 =Ilx*xll for every x E A. 1.3.2. Examples (1), (2) and (3) of 1.1.2, 1.2.2 are examples of C* algebras. Example (4), however, is not in general an example of a C*-algebra. 1.3.3. If A is a C*-algebra, so is every closed involutive subalgebra of A. In particular, if H is a Hilbert space, every closed involutive subalgebra of 2(H) is a C*-algebra; it will later be seen (2.6.1) that every C*-algebra is isomorphic to a C*-algebra of this type, and it is this example that has given rise to the theory of C*-algebras. Let (A;)iEI be a family of C*-algebras. Let A be the set of (X;)iEI such that Xi E Ai for each i E I and such that SUPiEI Ilxill < + 00. With the algebraic operations
(X;) + (Yi) = (Xi + Yi), (X;)(Yi) = (XiY;),
A (x;) = (AXi),

(x;)* = (x1),

and the norm we immediately see that A is a C*-algebra, called the product C* algebra of the A;s. It should be noted that the set A is not the product (set) of the A/s. Let A be a C*-algebra, and consider the algebra obtained from A by replacing the multiplication (x, y) ~ xy by the multiplication (x, y) ~ yx, while all other algebraic operations and the norm are the same as those of A. Consider, in other words, the reversed normed involutive algebra A 0 of A. Then it is immediate that A 0 is a C*-algebra. The question of quotient C*-algebras is a more delicate matter (cf. 1.8.2).

CH. 1, 3l

C-ALGEBRAS

1.3.4. Let A be a Banach algebra endowed with an involution such that 2 IIxl1 os; Ilx* x]. It follows that IlxW OS; IIx*IIllxll, hence that IIxll OS; Ilx*11 and interchanging x and x*, we see that Ilx*11 = IIxll. The above hypothesis thus implies that

so that A is a C*-algebra. 1.3.5. Let A be a C*-algebra. For each x E A, we have [z] = sup [xx[,
IlxlI"l

In fact, it is clear that Ilxllos; 1 implies Ilxxll OS; [x], To show that Ilxllos; sUPllx1<1 [xx[, we can assume that [x] = 1; then Ilx*11 = 1 and 2 sup [xx] ~ Ilxx*11 = IIxl1 = 1.
IIxlI"l

1.3.6. Let A be a unital C*-algebra. Then 2 11111 = 111*111 = [I], so that 11111 = 0 or 1. We thus see that, unless A = 0, 11111 = I, and it follows that if A,t 0 then 1/2 Ilull =Ilu*uI1 = 1 for each unitary element u of A. 1.3.7. We recall that no continuity condition appeared in the definition of morphisms of normed involutive algebras. We shall see that, for C* algebras, a morphism is automatically continuous, or, more precisely, that we have:
PROPOSITION. Let A be an involutive Banach algebra, B a C*-algebra and 1T a morphism of A into B. Then 1I1T(x)1I os; [x] for every x E A.

For each hermitian element y of B, we have Illll = Ily*yll = IIyl12 and n n hence by induction IIl ll2- = Ilyli. As n --,> + 00 the left hand side of this equation tends to the spectral radius p(y) of y (cf. Bl), so that
(1)
p(y)

= Ilyll
p( 1T(X os; p(x) os; [x],

N ow for each x E A, we have Sp ~1T(X)

~

Sp ~x

so that

10

NORMED INVOLUTIVE ALGEBRAS

[CR. I, 3

and hence, using (1), we have

1/1T(X)W =111T(x*x)1I = p(1T(X*X)) ~ lIx*xll ~ Ilx*IIllxll = IlxW

1.3.8. The following proposition will enable us almost always to confine our attention to unital C*-algebras without any loss of generality.
PROPOSITION. Let A be a C*-algebra and A the involutive algebra obtained from A by the adjunction of an identity. Then the norm on A can be extended to A in exactly one way that makes A a C*-algebra.

The uniqueness follows from 1.3.7. We prove the existence. First suppose that A has an identity e and let 1 denote the identity of A. In A, A and C(1- e) are complementary self-adjoint two-sided ideals, and so there is an isomorphism of the involutive algebra A onto the involutive algebra C x A which maps A onto {OJ x A; C x A can then be given the product C*-algebra structure. Now suppose that A does not possess an identity. Each element x of A defines an operator of left multiplication L, in the two-sided ideal A of A; put [x] = IILxll. If x E A, this coincides with the original norm on A, by 1.3.5. Moreover, it is clear that x ~ Ilxll is a seminorm on A and that Ilxyll ~ Ilxllllyll. This seminorm is in fact a norm, for let x = ,\ - x (,\ E C, x E A) be an element of A such that lILx l1 = 0, i.e. such that xy = 0 for each yEA. We show that then x = O. If ,\ 0, we have for each yEA, 0 = ,\ -Ixy = Y -,\ -lxy, ,\ -IX is a left identity for A, and A possesses an identity (1.1.5) contrary to hypo thesis. Hence ,\ = 0, so that x E A, and Ilxll = 0 implies that x = 0; it follows that x ~ Ilxll is a norm on A. Since A is complete and of co dimension 1 in A, A is also complete. It remains to show (1.3.4) that IIxl1 2 ~ IIx*xll for each x E A and it is enough to do this for IIxll = 1. For each r < 1, there is yEA such that Ilyll ~ 1 and IIxyll2 ~ r; then, as xy E A, we have Ilx*xll ~ Ily*(x*x)YII =II(xy)*xYII = IIxyl12 ~ r and therefore Ilx*xll ~ 1. 1.3.9. PROPOSITION. Let A be a C*-algebra. (i) If h is a hermitian element of A, each element of Sph is real. (ii) If A has an identity, and u is a unitary element of A, each element of Sp u has absolute value 1. In proving both parts of this result we may assume, by 1.3.8, that A has an identiy (and that A 0). We have Ilull = Ilu-111 = 1 by 1.3.6, so that p(u) ~ 1, p(u- l) ~ 1 and Sp u and Sp(u- l ) = (Sp U)-I are both contained in the unit disc of the complex plane C, from which (ii) follows

CH. 1, 41

COMMUTATIVE

C-ALGEBRAS

11

immediately. For (i)

(exp(ih*=
=

( l+ih+-+"
2!
exp(- ih),

eh Z

)*=I+(-i)h+--+" (_ i)Zh Z
2!

so that exp(ih) is unitary; thus if z E Sp h, we have [expfizj] = 1 (B 4) and so z E R. (Throughout the book, R denotes the set of real numbers.) 1.3.10. The following proposition shows that, for C*-algebras, it does not matter which algebra containing a particular element we take in the definition of the spectrum of that element.
PROPOSITION. Let A be a C*-algebra, B a sub-C*-algebra and x an element of B. Then

(i) Sp~x = Sp~x. (ii) If A has an identity which also lies in B, then SpAx = SPBx.

We see that (i) follows from (ii) on adjoining an identity. We therefore prove (ii). If x is hermitian, we have SPBX C R (1.3.9), so that SpBx = SpAx (B 2). In the general case, if x E B is invertible in A, xx* is also invertible in A, therefore in B by the above, and so x has a right inverse in B; similarly x has a left inverse in B and is therefore invertible in B. Applying this result to x - A . 1 where A E C, we obtain (ii). C*-algebras are called completely regular algebras in [108] and B* algebras by numerous authors. References: [604], [1101], [1320], [1322]: [1323].

1.4. Commutative C*-algebras 1.4.1. The theory of the above is summed up in the following theorem:
THEOREM. Let A be a commutative C*-algebra, S its spectrum (which is a locally compact space), and B the C*-algebra of continuous complex-valued functions on S which vanish at infinity. Then (i) Every character of A is hermitian. (ii) The Gelfand map is an isomorphism of the C*-algebra A onto the C*-algebra B.

Let X be a character of A. If y is an hermitian element of A, we have X(y) E Spy C R (1.3.9). For an arbitrary element x of A write x = XI + ixz

12

NORMED INVOLUTIVE ALGEBRAS

[CH. 1, 5

with

XI>

X2 hermitian; we have x(x*) = X(x\ - iX2) = X(x\) - iX(x2) = X(x)

which proves (i). In other words, if fJ denotes the Gelfand map, then fJ(x*) = fJ(x) for each X E A. Moreover, fJ(A) separates the points of S and for each point of S there is at least one function belonging to g;(A) which does not vanish there (B 3). The Stone-Weierstrass theorem then shows that fJ(A) is dense in B. The proof will be complete if we now show that fJis isometric. Now, 1IfJ(Y)!1 = !Iy!l for any hermitian y [1.3.7, formula (1)], and so for each x E A, IIxl1 = Ilx*xll = 11fJ(x*x)11 = 11fJ(x)fJ(x)11 = IIfJxI1
2 2

1.4.2. Retaining the above notation, if fEB, the symbol g(f) is well defined for each continuous complex-valued function g on f(S) U {O} = Spf such that g(O) = 0, and represents an element of B. Thus, if x E A, it is possible to define hex) E A for any continuous complex valued h defined on Spx and satisfying h(O) = O. We shall see in 1.5 that one can define such a "functional calculus" for the normal elements of any, riot necessarily commutative, C*-algebra, and this turns out to be a very useful tool in the sequel. 1.4.3. PROPOSITION. Let A be a commutative unital C*-algebra, S its spectrum and let x E A. Suppose that the sub-C*-algebra of A generated by 1 and x is equal to A. Then X ~ X(x) is a homeomorphism of Santo SPAx. This map is continuous and its range is SPAX (B 3). Moreover, let X,
X E S be such that X(x) = x(x). Since X and X are continuous and

hermitian, the set A of those yEA for which X(y) = x(y) is a sub-C* algebra of A containing x. Hence A = A and X = X. The map under consideration is injective, and is therefore a homeomorphism as S is compact. References: [100], [618], [1101], [1323].
1.5. Functional calculus in C*-algebras

1.5.1. THEOREM. Let A be a unital C*-algebra, x a normal element of A, S = SpAx and A the C*-algebra of continuous complex-valued func tions on S. Then there is a unique morphism f/> of A into A such that

CH. 1, 5l

FUNCTIONAL CALCULUS IN C*-ALGEBRAS

13

</>(1) = 1, and </>(t) =x where t denotes the function z ~ z on S. Moreover, this morphism is isometric and its image </>(A) is the sub-C*-algebra of A generated by 1 and x and therefore consists entirely of normal elements.

The polynomials in z and z are dense in A and each morphism of A into A is continuous (1.3.7) which establishes the uniqueness of </>. Let B be the commutative sub-C*-algebra of A generated by 1 and x, Tits spectrum, C the C*-algebra of continuous complex-valued functions on T and :JiB the Gelfand map for B, which is an isomorphism of B onto C (1.4.1). Proposition 1.4.3 furnishes a homeomorphism of Tonto SPBX = S (1.3.10) which induces an isomorphism ljJ: A ~ C that maps t to :JiBx, since we have, for each X E T, (:JiBx)(X) = X(x) = t(X(x)). Now consider the composite isomorphism
~

A'~C~B

Composing this with the canonical injection of B into A, we obtain a morphism of A into A with the properties required by the theorem. 1.5.2. DEFINITION. Let A be a unital C*-algebra. If x is a normal element of A and if f is a continuous complex valued function on Sp AX, the element </>(f) of theorem 1.5.1 is denoted by I(x). The fact that </> is an isometric isomorphism is expressed in the following formulae in which / and g denote continuous complex-valued functions on Sp AX:
(1)
(2)

(f + g)(x) = f(x) + g(x), (fg)(x) = f(x)g(x), lex) = f(x)*,

(3)

(4)

Ilf(x)11 = II/II

If f is the restriction to Sp AX of a polynomial z ~ P (z, z) in z and Z, then f(x) = P(x, x*) where Ptx, x*) has its usual algebraic interpretation (remember that xx* = x*x). With the above notation, we have

or in other words,
(5)

14

NORM ED INVOLUTIVE ALGEBRAS

[CH. I, 5

1.5.3. PROPOSITION. Let A and B be unital C*-algebras, 4> a morphism ofA into B mapping 1 to 1 and x a normal element of A so that q,(x) is a normal element of B. Let f be a continuous complex-valued function on SPAx. Then if the restriction of f to SPB4>(X) is again denoted by f, we have 4>(f(x = f(4)(x)). Let C be the C*-algebra of continuous complex-valued functions on Sp AX. The maps f ~ 4>(f(x and f ~ f( q,(x are morphisms of C into B which take the same value when f is anyone of the functions z ~ I, z ~ z, z ~ i. Since the sub-C*-algebra of C generated by these functions is equal to the whole of C, the two morphisms are identical. 1.5.4. COROLLARY. Let A be a commutative unital C*-algebra, x an element of A, [fI the Gelfand map for A, and f a continuous complex valued function on SPAX. Then [fI(f(x = f 0 [fI(x). This follows, for instance, from 1.5.3 applied to q, =[fl. 1.5.5. COROLLARY. Let A be a unital C*-algebra, x a normal element of A, C the C*-algebra of continuous complex-valued functions on Sp x, f an element of C, C the C*-algebra of continuous complex-valued functions on Sp f(x) = f(Sp x), and g an element of C. Then g 0 fEe and (g f)(x) = g(f(x.
0

The map g ~ (g f)(x) is a morphism of C into A which maps 1 to 1 and the function z ~ z to f(x). From the uniqueness statement of theorem 1.5.1, it follows that (g 0 f)(x) = g(f(x.
0

1.5.6. PROPOSITION. Let A be a C*-algebra, x a normal element of A, S = Spx and A the C*-algebra of continuous complex-valued functions on S which vanish at o. Then there is exactly one morphism q, of A into A such that q,(L) = x where Lis the function z ~ z on S. This morphism is isometric and its image 4>(A) is the sub-C*-algebra of A generated by x which therefore consists entirely of normal elements. Since the polynomials in z and z without constant term are dense in A, the uniqueness of q, is immediate. The existence follows from theorem 1.5.1 on adjoining an identity to A. 1.5.7. By the adjunction of an identity, all the results of this section can be extended immediately, with obvious modifications, to the case of non-unital C*-algebras. We shall thus make use of these results even when an identity is not assumed to be present.

CH. I, 6]

POSITIVE ELEMENTS IN C*-ALGEBRAS

15

In particular, let x be an hermitian element of the C*-algebra A, so that its spectrum is real. Consider the continuous functions of a real variable

We A, (and, indeed, of the sub-C*-algebra of A generated by x), because II =t.. lz =fz, 13 =!J. Since i.. t-. f3 take non-negative values only, we have
(1)
Sp'(x+)~O,

= sup(t, 0), t ~ Ut) = sup( - t, 0), t ~ Ut) = Itl. put x" = fl(X), x- = fz(x), Ixl =!J(x). These are hermitian elements of
t ~ ft(t)

Sp(lxl) ~
Nt) + fz(t) = It I and

0.

Since
Nt) - fz(t) = t, Nt)fzCt) = 0

we have
(2) x

=x -

x",

[x] = x" + x",

x+x-

= x-x+ =O.

The norm of Ixl is the same as that of x, while the norms of x+ and x are less than or equal to that of x. We note, with a view to no. 1.6.4, that x" and x- are the squares of two hermitian elements whose product is zero: consider the functions V I, and V fz. 1.5.8. Let A be a C*-algebra. For each positive integer n, we have A = An, the set of linear combinations of products of n elements of A; it is enough to show that each hermitian element x of A is a product of n elements of A. Now, if t.. .. .i, are continuous real-valued functions of a real variable such that
Nt)fz(t) ... fn(t)

= t,

NO)

=... = fn(O) =0,

then
x

= fl(x)fz(x) ... fn(x).

References: [618], [1101], [1323].

1.6. Positive elements in C*-algebras

1.6.1. PROPOSITION. Let A be a C*-algebra and x an hermitian element of A. Then the following conditions are equivalent: (i) Sp~~ O. (ii) x is of the [orm yy* for some yEA. (iii) x is of the [orm h Z for some hermitian h E A.

16

NORMED INVOLUTIVE ALGEBRAS

[CH. I, 6

Furthermore, the set P of those elements which satisfy these conditions is a closed convex cone such that P n (- P) = {O}.
To prove this, we first denote by P the set of those hermitian elements of A which satisfy condition (i). (i) ~ (iii): if Sp~x "" 0, h = X 1/2 exists as a hermitian element of A and we have x = h 2 (iii) ~ (i): if x = h 2 with h hermitian, we have Sp ~x = (Sp ~h)2 ""0 since Sp ~h is real. (iii):::}(ii): obvious. To prove the implication (ii) ~ (iii) and the last part of the proposition we need the following lemmas: 1.6.2.
LEMMA.

lit - x] ~

1 then x E P. If x E P and [x] ~

Suppose that A is unital. If x E A is hermitian and 1 then 111- x] ~ 1.

By passing to the sub-C*-algebra of A generated by x, it is enough to consider the case of commutative A which can in turn, thanks to 1.4.1, be reduced to the case where A is the C*-algebra of continuous complex-valued functions on a compact space. The lemma is then clear. 1.6.3. LEMMA. Suppose that A is unital and let x be an hermitian element of A. In order that x E P, it is necessary and sufficient that II(IIxll . 1) - xII ~ [x]. We can assume that x =I- 0, and then, multiplying by an appropriate scalar. that Ilxll = 1. Lemma 1.6.3 then follows at once from Lemma 1.6.2. 1.6.4. We now return to the situation of 1.6.1. In showing that P is a closed convex cone such that P n (- P) = {O} we can assume. thanks to 1.3.8, that A is unital. Lemma 1.6.3 implies that P is closed. It is clear that x E P and A,"" 0 .mply that Ax E P. We now show that x + yEP whenever x, yEP. We can assume that Ilxll ~ I and Ilyll ~ 1. Then III - x] ~ 1 and 111- yll ~ 1 (1.6.2), and so 111- ~(x + y)II = ~1I1- x + 1 - yll ~ !(111- xII + 111- yll) ~ 1 so that !(x + y) E P (1.6.2) and thus x + yEP. If x E P n (-P) then Sp x = {O}, p(x) = 0 and so x = 0 [1.3.7, formula (1)]. We finally prove the implication (ii):::}(iii) of 1.6.1 without assuming that A is unital. Let yEA, and write (yy*t =: u 2 , (yy*)- = v 2 with U, v hermitian elements of A such that uv =: 0 (1.5.7). Then

(vy)(vy)*

= v(yy*)v

vu 2v - v 4 =: -v 4 E -P,

CH. 1, 6j

POSITIVE ELEMENTS IN C*-ALGEBRAS

17

since (iii) ~ (i). Let vy =s + it with s, t hermitian. We have

(vy)*(vy) = -(vy)(vy)* + (s

+ it)(s 2 2

it) + (s - it)(s

+ it)

= -(vy)(vy)* + 2s + 2t

EP

as -(vy)(vy)* E P, S2 E P, t 2 E P and P is a convex cone. Hence (vy)(vy)* E P because, in any algebra, the spectrum of a product is independent of the order of its factors (B 26). Hence (vy)(vy)* E P n (- P) = O. Thus v 4 = 0 and so v =0 and yy* = u, Proposition 1.6.1 is now proved. 1.6.5. DEFINITION. Let A be a C*-algebra. An element x E A is said to be positive and we write x ~ 0 if it is hermitian and satisfies the three equivalent conditions of proposition 1.6.1. The set of positive elements of A is denoted by A ".
If B is a sub-C*-algebra of A and x E B, it follows from 1.3.10 that the statement x ~ 0 has the same meaning whether it is interpreted in A or in

Since A + is a convex cone and A + n (- A +) ={O}, the relation x - y ~ 0 is a partial ordering in A which is compatible with the real vector space structure of A; we write this relation x ~ y or y ... x. If z is an hermitian element of A then z ~ 0, z: ~ 0 and [z] ~ 0 by 1.5.7, formulae (I); z" (resp. z ") is called the positive (resp. negative) part of z. By 1.5.7, formulae (2), each hermitian element of A is the difference of two elements of A ", 1.6.6. Let A be the C*-algebra of continuous complex-valued functions vanishing at infinity on a locally compact space T and suppose f E A. It follows from condition (I) of proposition 1.6.1 and the fact that Sp ~f = f( T) U {O} that the relation f ~ 0 has its usual meaning in the algebra A. 1.6.7. Let H be a Hilbert space, A the C*-algebra 2(H), and x an element of A. We show that the condition x ~ 0 is equivalent to the condition (xg Ig) ~ 0 for every g E H, i.e. to the usual definition of positive operators. If x = y*y for some yEA, we have
(xg

B.

Ig) = (y*yg Ig) = Ilyglf ~

for each

gE

H.

18

NORMED INVOLUTIVE ALGEBRAS

[CH. 1, 7

Conversely, suppose that (x~ I~) ~ 0 for each ~ E H. For every TJ E H, we have O:s;:; (x(x-TJ) IX-TJ)

= (x+ -x-)(x-TJ) IX-TJ)

= -(x-x TJ

(X-)3

Since (X-)3 ~ 0 we also have = 0, X- = 0 and finally x = ~+ ~ O.

IX-TJ) = _((X-)3 TJ ITJ) ((X-)3 TJ ITJ) ~ 0 and thus

((X-)3 TJ

ITJ) = 0,

1.6.8. Let A be a C-algebra, and a, b, x elements of A. If a :s;:; b, then such that b - a = y*y from which it follows that x*bx - x*ax = x*(y*y)x = (yx)*(yx) ~ O.
x*ax:s;:; x*bx, as there is yEA

N ow suppose that A is unital and let b be an element of A with b ~ 1 (hence b is invertible). Applying the above with x = b- l12 we see that 1 ~ b -I. More generally, let a and b be invertible elements of A + such that O:s;:; a :s;:; b. Then a-I ~ s:, indeed, from the above, we have that 1:s;:;a- 1I2ba- 1I2 , l~aIl2b-IaI12 and so a-I~b-I. 1.6.9. Let A be a C*-algebra, and let x, yEA + be such that y :s;:; x. Then assume A to be unital and we then have 1.4.1. Hence O:s;:; y :s;:; Ilxli1 and so lIyll:s;:;IIxll, again by 1.4.1.

Ilyll :s;:;[x]. In fact, we can x :s;:; IIxli1, for example by

1.6.10. Let A and B be two C*-algebras and </> a morphism of A into B. It is plain that c/>(A+) C </>(A) n B ". Suppose conversely that y E </>(A)n B+; there exists x E A with y = </>(x) and we have y = (y*y)1I 2 = c/>X*X)112), so that y E </>(A"), Hence </>(A+) = </>(A)n B+. References: [604], [918], [1101], [1323], [1477].
1. 7. Approximate identities in C*-algebras

1.7.1. Let A be a C*-algebra. We say that an approximate identity (uJ of A (B 29) is increasing if UA ~ 0 for every A and if A :s;:; JL implies
uA:s;:; U w

1.7.2. PROPOSITION. Let A be a C*-algebra, and m a two-sided ideal of A which is dense in A. Then there is an increasing approximate identity of A consisting of elements of m. If A is separable, this approximate identity can be taken to be indexed by {l, 2, ...}.

CH. 1, 7l

APPROXIMATE IDENTITIES IN C*-ALGEBRAS

19

Let A be the C*-algebra obtained by adjoining an identity to A. Let A be the set of finite subsets of m ordered by inclusion. For A = {x., ... ,x.} E A, put

(the element UA is computed in A, but in fact UA Em). Since the function of a real variable t ~ t(lln + t)-I only takes values between and 1 for t ;;?;0, we have o:s; UA :s; 1. Furthermore,

[(U -l)x ][(UAL

A i
;=1

I)x;]* = (UA -1)V A(UA -

1)

=2

1 (1 -+ )-2
VA VA

Now the function of a real variable t ~ t(lln + t)-2 is always

:s; ~n.

Thus

L [(u
i~1

1 -1)x;][(uA -l)x;l* :S;-.

4n

For i

= 1,2, ... ,n

we deduce that
[(u A
-

I 1)x;][(uA - l)x;]*: S ; - ,

4n

from which it follows that 1/(uA - l)x i I 1:s; ~n (1.6.9). Thus II(uA - l)x] ~ for each x E m and therefore for each x E A as m = A and IluJ:s; 1; hence
2

(uA ) is thus an approximate identity. Now let A, p., E A be such that A :s; u: We have A ={x., ... ,x.}, p., ={XI, ... , xp } with p;;?; n, so that VA :s; VI" and (lIn + VA)-I;;?; (lIn + VIL)-I by 1.6.8. For any real number t 0, we have

;; ;;+t
so that

I(I ) 1;;?;p 1(I )-1 p+t ,

p p "

- -+v"
n n

I(1 )-1 ;;;,:1(1 -I -+V) ,

VI

all of which implies that

1(1 1(1 1-;; ;;+VA )-1 :s; 1-;; ;;+

i.e.
UA

)-1 :s; 1- p 1(1 .: )-1

VI

:s; ulL Hence

the approximate identity (u.) is increasing.

20

NORMED INVOLUTIVE

ALGEBRAS

[CH. 1, 8

Now suppose A is separable, so that there is a sequence (YI Yb ...) which is dense in m. Put u; = u!y\..... y.}. The above argument shows that, for each i, Iluny; - };I!~ 0 as n ~ +00. Since Ilunll ~ 1, we deduce that u.x ~ x for each x E A and the proof is now concluded in the same way. 1.7.3. Let A be a C*-algebra and I a right ideal of A. Then there exists a family (uJ in I n A + indexed by a directed set such that (I) IluA11 ~ 1; (2) A ~ /..L implies UA ~ u/L; (3) for each x E 1..IIuAx - xll~O. The proof of 1.7.2 applies here unchanged. References: [452], [1455].
1.8. Quotient of a C*-algebra

1.8.1. PROPOSITION. Let A be a C*-algebra, B a normed involutive algebra and 1J an injective morphism of A into B. Then 111J(x)1I ~ IIxll for each x E A. Let x E A. If we can show that 111J(x*x)11 ~ Ilx*xll, we can deduce that IIxl1
2

= Ilx*xll

~ 111J(x*x)11

= 111J(x*)1J(x)11

~ 111J(x)11

from which the proposition follows. We can thus assume in addition that x is hermitian, and replacing 1J by its restriction to the sub-C*-algebra of A generated by x, we can assume that A is commutative. Replacing B by 1J(A) we can assume B is commutative, and we then replace B by its completion. Furthermore, we can adjoin identities to A and B. In short we may confine attention to the case in which A and B are commutative, complete and unital. Now let Sand T be the spectra of A and B, which are compact spaces. For each X E T, X 1J is a character of A, i.e. an element of S. which we denote by 1J(X); if x E A. then 1J(X)(x) = X(1J(x)) is a continuous function of x; 1J is thus a continuous map of T into Sand 1J(T) is a compact subset of S. If 1J(T) # S, there exists a continuous complex-valued function f on S such that f# 0 and f 11J(T) = O. By 1.4.1. f is the Gelfand transform of an x E A. We have x# O. X(1J(x)) = 0 for every X E T, hence 1J(x) = 0, and this is absurd.
0

1.8.2. PROPOSITION. Let A be a C*-algebra and I a closed two-sided ideal of A. Then I is self-adjoint and AII. endowed with the natural inuolutioe algebra structure and the quotient norm, is a C*-algebra.

CH. 1, 8]

QUOTIENT OF A C*-ALGEBRA

21

Let (UA) be a family with the properties of 1.7.3. If x E I, we have

and xr u, E I, so that x* E I = I and hence 1=1*. We know that AII is an involutive algebra and satisfies the axioms for a Banach algebra. Denote by x ~ x the canonical map of A onto AII. To show that AII is a C*-algebra, it is enough to show that Ilxll 2 ,,;::;; Ilxx*11 (1.3.4). We have
(1)

In fact, if Y E I we have /lAY - Y ~ 0, and so lim Ilx - uAxil

= =

lim [x - UAX + Y - uAyl1 lim 11(1-/lA)(X + Y)II,,;::;; [x + yll

(we are working in ,4). Thus Ilxll?; lim [x - uAxll?; lim [x - uAxll?; inf [x + yll
yEI

Ilxll,

which proves (1). This established, we have for each

E I,

IIxW = lim Ilx = = =

/lAx/1 2 = lim II(x - uAx)(x - uAx)*11

lim Ilxx* - u.xx" - xxr u, + UAXX*uAII lim Ilxx* + Z - UAZ - u.xx" - xx* U A- ZU A + UAZU A + u.xx" uJ lim 11(1 - uA)(xx* + z)(1 - uJII,,;::;;Ilxx* + z],

and thus Ilx11 2 ,,;::;; Ilxx*ll. 1.8.3. COROLLARY. Let A and B be C*-algebras, cP a morphism of A into B and I the kenzel of cPo Consider the canonical decomposition of cP:
A~AII

"

cP(A)~B.

Then I is closed in A, cP(A) is closed in Band IjJ is an (isometric) isomorphism of the C*-algebra AII onto the C*-algebra cP(A).

Since cP is continuous (1.3.7), I is closed and AII is a C*-algebra (1.8.2). The morphism AI I ~ B obtained from cP by passing to the

22

NORMED INVOLUTIVE ALGEBRAS

[CH. 1, 9

quotient is injective and therefore isometric (1.3.7 and 1.8.1). Hence et>(A) is complete and consequently closed in B. 1.8.4. COROLLARY. Let A be a C*-algebra, B a sub-C*-algebra of A and I a closed two-sided ideal of A. Then B + I is a sub-C*-algebra of A and the C*-algebras (B + 1)1 I and BIB n I are canonically isomorphic. Let et> be the canonical morphism A ~ AI I, and let I/! be the restriction of et> to B. Then I/!(B) = (B + 1)11 is closed in All (1.8.3), and so B + I is closed in A. Consider the canonical decomposition of I/!:
B ~ BI(B

n I)~
n
I)~

I/!(B)~

All.

By (1.8.3), the morphism BI(B isomorphism.

l/J(B) = (B

+1)11 is a C*-algebra

1.8.5. PROPOSITION. Let A be a C*-algebra, I a closed two-sided ideal of A and J a closed two-sided ideal of I. Then J is a closed two-sided ideal of A. We have J

= J3 (1.5.8),

and so AJA = AJ3 A

III

c J.

References: [618], [893], [1101], [1323], [1456]. The proofs of 1.8.2 and 1.7.3 were communicated to me orally by F. Combes.

1.9. Addenda 1.9.1. Let A be a C*-algebra. If the norm and all the algebraic operations of A are retained with the exception of (A, x)~Ax which is changed to (A, x) ~ Ax, a C*-algebra A called the conjugate of A is obtained. 1.9.2. Let A be a C*-algebra. If x E A is normal, then [x] = p(x). (Use 1.4.1) [1323]. 1.9.3. Let A be a C*-algebra. (a) If every maximal commutative sub-C*-algebra of A has an iden tity, so does A. (b) If every maximal commutative sub-C*-algebra of A is finite di mensional, so is A. [1159]. 1.9.4. Let A be a C*-algebra. If the conditions xEA+, yEA+, imply x 2 ~ y2, then A is commutative. [1160].
x~y

CH. 1, 9j

ALlDENDA

23

*1.9.5. Let A be a unital Banach algebra endowed with an involution such that Ilx*xll = Ilx*IIllxll for each x E A. Then A is a C*-algebra. [633], [1187]. 1.9.6. Problem: does a unital C*-algebra A of dimension > 1, whose only closed two-sided ideals are 0 and. A possess any projections other than 0 and I? [903]. 1.9.7. Let A be a unital C*-algebra, and x an element of A which does not have a left inverse. Then x*x is not invertible in the sub-C*-algebra B generated by I and x*x, so there exist Yt, Y2, ... E B such that IIYnl1 = 1 and IIYnx* x] ~ 0 (use 1.4.1). Hence if x E A is not invertible, x is a topological divisor of zero. [1323]. 1.9.8. Let A be a unital C*-algebra, N the set of normal elements of A,

x E N and V a neighbourhood of 0 in C. There exists a neighbourhood U of x in N such that for every Y E U, Sp Y C Sp x + V and Sp x C Sp Y + V. [1323].
*1.9.9. Let A be a unital C*-algebra, H a Hilbert space and c/J a linear map of A into 2(H) such that c/J(A+) C 2(Ht. Then, for every her mitian x in A, we have c/J(x 2) ;a. c/J(X)2. [839]. *1.9.10. Let A and B be unital C*-algebras. (a) Let p: A~B be a bijective linear map such that p(x*) = p(x)* and p(x n ) =p(x)" for X hermitian and n an integer >0. Then p(A+) =B+, p is isometric, p maps a pair of commuting elements into a pair of commuting elements, and pO) = 1. (b) Let (T: A ~ B be an isometric bijective linear map. Then (T is the composition of a map having the properties of (a) and multiplication on the left by the element (T(l) of B. (c) Let T: A ~ B be an isometric linear map such that T(l) = 1. Then r(x*) = T(X)* for each x E A. [179], [837]. *1.9.11. Let A be a C*-algebra. (a) Every derivation of A is continuous. (b) Let D be a derivation of A and x a normal element of A. If x(Dx) = (Dx)x, then Dx = O. In particular, the only derivation of a commutative C*-algebra is the zero operator. It follows from this that a commutative closed two-sided ideal in a C*-algebra is central. (c) Let D be a derivation of A and x a normal element of A such that Dx = O. Then Dx* = O. [903], [1402], [1403].

24

NORMED INVOLUTIVE

ALGEBRAS

[CH. I, 9

(d) Let D be a derivation of A and I a closed two-sided ideal of A. Then D(I) C I (use 1.5.8). (e) Let H be an infinite-dimensional Hilbert space, A the C*-algebra of all compact operators in H and x an element of .2(H) which does not belong to A + C . 1. Then the derivation y ~ xy - yx of A is not inner. [530J, [532], [535J, [538], [1416J, [1417], [1428], [1429J, [1433]. 1.9.12. Let A be a C*-algebra. (a) Let I and I be closed two-sided ideals of A. Then the product ideal IJ is equal to In 1. (If x E (I n It, then X 1/2 E (I nIt.) [508]. Further, (I +I t = I" + I", [1239], [1515]. (b) If I n 1= 0, the canonical map I + I ~ I x I is a C*-algebra isomorphism. (c) Let (Ia) be a family of closed two-sided ideals of A whose intersection is I. Let W a (resp. w) be the canonical morphism of A onto All; (resp. A/I). For each x E A, Ilw(x)11 = sUPallwa(x)lI. [584). 1.9.13. Let A be a C*-algebra. A C*-semi-norm on A is a semi-norm N such that
N(x):s;;

Ilxll,

N(xy):s;; N(x)N(y),

N(x*x) = N(X)2

for any x, yEA. The set X(A) of all C*-semi-norms on A is compact for the topology of pointwise convergence on A. If I is a closed two-sided ideal of A and x E A, let N1(x) be the norm of the canonical image of x inA/I. Then I~NI is a bijection of the set ~(A) of all closed two-sided we have N InJ = sup(Nb N J ) [use ideals of A onto X(A). If I, I E ~(A), 1.9.12.b]. A C*-semi-norm N is said to be extremal if N cannot be the upper bound of two C*-semi-norms without being equal to one of them. For a C*-semi-norm N to be non-zero and extremal, it is necessary and sufficient that I be prime. [In any algebra R, a two-sided ideal I is said to be prime if liR and if the relation J J" c I, where J, J" are two-sided ideals of R, implies that Ic I or J" C I.] If A is separable, the set E of extremal C*-semi-norms is a Os in X(A). [The set of (N, N) E X(A) x X(A) such that N:s;;N or N:s;;N is compact. Its complement is a countable union of compact sets, and its image in X(A) under the continuous map (N, N')~sup(N, N) is the complement of E.) [509], [584]. 1.9.14. Let (Ai)iEI be a family of C*-algebras. Let A be the set of x = (Xi) E II iE1Ai such that for each e > 0, Ilxill:s;; for all but a finite

CH. I, 9]

ADDENDA

25

number of indices i. If we put [x] = supllx], A becomes a C*-algebra in a natural way, and this algebra is called the restricted product of the Ai 1.9.15. The radical R of a C*-algebra is zero. [If x E R, then 1 + Ax*x is invertible in A for each A E C, thus Sp(x*x) = 0 and x = 0.] [1454].

CHAPTER 2

POSITIVE FORMS AND REPRESENTATIONS

One of the most important concepts of this book is that of a represen tation. Given an involutive Banach algebra A, it would be a difficult task to establish the existence of representations of A directly. However, we shall set up a correspondence between representations of A and positive forms on A, and in particular between irreducible representations and pure positive forms. Moreover, the classical tools of functional analysis, namely the Hahn-Banach and Krein-Milman theorems, enable us to prove the existence of positive forms and indeed of pure positive forms as well. This is the basic idea of this chapter. In the various results which follow, the involutive algebras being studied are subjected to a variety of conditions. At a first reading, however, it may be assumed that we are concerned exclusively with C*-algebras. 2.1. Positive forms 2.1.1. DEFINITION. Let A be an involutive algebra. A linear form f on A is said to be positive if f(x*x) ~ 0 for each x E A. If A is a normed involutive algebra, a state of A is a continuous positive linear form f on A such that 11111 = 1. Let A be an involutive algebra and f and g linear forms on A. We say that f dominates g and we write f ~ g or g :s;; f if f - g is positive. This defines a preorder in the dual of A which is compatible with the real vector space structure. If A is a C*-algebra this relation is in fact a partial ordering, for if f ~ 0 and f:S;; 0, then f vanishes on A +, therefore on the set of hermitian elements of A (1.6.5) and hence f = O. Let a be a locally compact space and A the C*-algebra of continuous complex-valued functions on a which vanish at infinity. A continuous linear form on A is simply a bounded measure f..t on a, and to say that this linear form is positive is exactly the same as saying that the measure f..t is positive.

CH. 2. lj

POSITIVE FORMS

27

x, yEA put (x I y) = f(y*x). This scalar product is linear in x, anti-linear in y and (x Ix) ~ 0 for each x. A is thus endowed with a pre-Hilbert

2.1.2. Let A be an involutive algebra and f a positive form on A. For

space structure. In particular, we have

(1)

f(y*x) = f(x*y)
If(y*xW~f(x*x)f(y*y)

(x E A, yEA), (x E A, yEA).

(2)

If A is unital, we deduce, putting y

(3)

= 1 in

(1) and (2), that

f(x*)

= f(x),

(4)

If(xW ~ f(1)f(x*x).

Let H be the Hausdorff pre-Hilbert space constructed canonicaIly from the pre-Hilbert space A, so that H = AIN where N is the set of those x E A for which f(x*x) = O. By (2), N is equaIly the set of x E A such that f(yx) = 0 for all yEA, so that N is a left ideal of A. [Putting (x I y) = f(xy*) we would obtain another pre-Hilbert space structure on A with analogous properties.} 2.1.3.
LEMMA. Let A be a unital Banach algebra, x an element of A with Ilxll ~ 1 and x = I + x. The series

= x. If A has an isometric converges to an element y of A such that involution and if x is hermitian, then so is y.

The series

l+-IIxII+--

1 2

1(1) 1/1(1)1 11 --1 . IIxil +... +- --1 ... 2! 2 2 n! 2 2

2

G - + 1)
n

"lI x ll

+...

is convergent from which it foIlows that the element y exists. If we compute i, we obtain a power series in x whose coefficients we recognize from the classical situation where A = C; indeed, we have

28

POSITIVE FORMS AND REPRESENTATIONS

[CH. 2, 1

l = 1 + x = x. If A hasan isometric involution and if x is hermitian, y is seen to be a limit of hermitian elements and is therefore hermitian because the involution is continuous.
2.1.4. PROPOSITION. Let A be an involutive Banach algebra having an identity 1 such that /11/1 = 1. If f is a positive linear form on A, then f is continuous and IIfli = f(1).

If x E A is hermitian and Ilxll:;;; 1, then 1- x can be written in the form y*y (lemma 2.1.3), from which it follows that f(1- x) ~ 0 and so tu: f(I). If x E A and IIxll:;;;1, then Ilx*xll:;;; 1, and hence, using 2.1.2, formula (4), If(x )12 :;;; f(1)f(x*x):;;;f(If

i: 11111,

This shows that f is continuous and that and hence 11111 = f(1)

Ilfll:;;; f (I).

It is plain that

1(1)

2.1.5. PROPOSITION. Let A be an involutive Banach algebra having an approximate identity (B 29), A the involutive algebra obtained from A by adjoining an identity to it, and f a continuous positive linear form on A. Then: (i) f(x*) = f(x) for every x E A. 2 If(x)1 :;;; It11l f(x*x) for every x E A. (ii) If(y*xY)I:;;;llxllf(y*y) for all x,yEA. (iii) 11111 = sUPxEA.114<;1 f(x*x). (iv) If (Xi)iEI is a family of elements of A indexed by a directed set, such that Ilxill:o;:;1 and f(x i ) --,) 11111, then f(x~ x;) ~ Ilf1I. (v) If (Uj)jEJ is an approximate identity for A, then f(uj) ~ 11111, and f(u~u)~II11I (vi) f can be uniquely extended to a positive form on A such that = llill Every positive form on A which extends f dominates f. (vii) In the notation of (iv), we have Xi ~ 1 for the pre-Hilbert space structure defined on A by f, so that A is dense in A for this structure.

We have, for each x E A,

f(x*) = lim f(x*u j) = lim f(ujx)

= lim fx*uj)*) = f(x**) = f(x),

If(x)12 = lim If(xuj)12 :;;; f(x*x) lim f(uju):;;; 1111lf(x*x).

This proves (i) and it is plain that (i) :? (iv) :? (iii). We prove (vi). The uniqueness of 1 is immediate, so we just have to establish the existence of f. For each element (A, x) = A + x of A (A E C, x E A) put /(A + x) =

CH. 2,

sn

POSITIVE FORMS

29

,\11111 +f(x): Then

have
f,\

f is a linear form on
=

A extending

f and using (i), we

+x)*(,\ +x)

and so f is positive. form on A which extends f. A can be given an involutive Banach algebra structure extending that of A and such that 11111 = 1. By (2.1.4), we have g(1) = IIgll~ 11111, from which we obtain
g\ +x)*(,\

+Ax +h*)+1,\1211111 = f(x*x) + 2 Re Af(x) + l,\flll1l ~ f(x*x) - 21\1111111/2f(x*x)I2+1,\1 2 1 111 = [f(X*X)1/2 _1\111111 2]2 ~ 0, Clearly, j(1) = 11111. Now let g be any positive
f(x*x
1 /

+x)

~
=

2 +Ax +\x*) +1,\ 1 g (1) 211111 f(x* X +Ax +h*) +1,\1

f(x*x f,\

+ x)*(,\ + xj),

so that g ~ f, and (vi) is proved. If yEA, the form x --+ j(y*xy) on A is positive, because j(y*(x*x)y)=fxy)*(xy~O; by (2.1.4), its norm is f(y*y) and (ii) now follows. In the notation of (iv), we have
fx; - 1)*(Xi -I) = f(xtx i) - f(x;) - f(Xi) +11111
--+ 11111 -11111-11111 +11111 = 0, from which (vii) follows. For every xEA, (ujX)=f(x*uJ--+ We have (ujluj)=f(u~uj),,;;;III1I. f(x*) = (11 x); since A is dense in A (d. (vii, we see that u, converges weakly to 1. Hence f(u;) = (u, 11)--+(1, 1) = /(1) = 11111. From (iv), f(u1 Uj) --+ 11111.

2.1.6.
A.

In the notation of

2.1.5, f is called the

canonical extension of f to

COROLLARY. Let f, g be continuous positive forms on A and f, g their canonical extension to A. Then

II!+ gil = 11111 + IlglI,

(f+ g( =f+ g.

The first relation follows from 2.1.5 (v), while the second follows from the first. It follows that the set of states of A is a convex subset of the dual of A.

30

POSITIVE FORMS AND REPRESENTATIONS

[CH. 2, l

2.1.7: We retain the notation of 2.1.5. Now let Q be the set of con tinuous positive forms on A and Q the set of positive forms on A such that g(1) = Ilg IA]. Then 1....0, 1 is a bijection of Q onto Q which preserves both the additive and order structures. If a positive form on A is dominated by an element of Q, it must itself belong to Q; in fact let g = gl + g2 where g E Q and gl, s. are positive forms on A, and let I, II, 12 be the restrictions of g, gt, g2 to A. We have

s, + g2 = g =1=I. +12
and
gl(l) ~ 11(1),

from which it follows that

and thus gl E Q, g2 E Q. 2.1.8. Let A be a C*-algebra and I a positive form on A. Then I is continuous. Indeed, let (XI X2, ) be a sequence of elements of A + such that Ilx;11 ~ I. We show that the numbers I(x;) form a bounded set. For ) of non-negative real numbers satisfying each sequence (AI> A 2 , L7~1 Ai < + 00, the series L7~1 Ajx i converges to an element X of A. For every integer n ~ 1, we have L7=n+1 AjXj ~o by 1.6.1, and so

Hence L7~1 AJ(x;) < + 00. Since this holds for any sequence (A;) of non-negative reals with L7~1 Aj < + 00, it follows that the set of the numbers I(xj) is bounded. It now follows from this that M=
If x is hermitian and of norm

sup
XEA+.llxll"l
~

I(x)<+oo.

1, we have

and if x is any element of norm

1, then

Hence

I is continuous.

CH. 2. 2j

REPRESENTATIONS

31

In the sequel, positive forms on C*-algebras will be mentioned over and over again, and the reader should never forget that these forms are automatically continuous. 2.1.9. PROPOSITION. Let A be a unital C*-algebra and I a continuous linear [orm on A. For I to be positive, it is necessary and sufficient that

Ilfll=1(1)
If I ;if: 0, then Ilfll =1(1) (2.1.4). Now suppose Ilfll =I(l). Let x E A + and suppose that I(x) ;;!: 0. (We can, of course, assume that 1(1) = 1.) There is a closed disc [z - zol-s; p in C which contains Sp x but not I(x). Now the spectrum of the normal element x - Zo is contained in the disc Izl-s; p, from which it follows that Ilx - zoll-s; p. Hence
/(x) -

zoll = I/(x) -

zof(1)1

= I/(x - zo)l-s; IIfllllx- zJI ;if: p,

which is a contradiction. (An alternative proof is based on B 28.) The work "state" is borrowed from physics. References: [174], [618], [619], [638], [1097], [1101], [1323], [1455]. The result of 2.1.8 can be generalized to involutive Banach algebras with an approximate identity [1755]. 2.2. Representations 2.2.1. DEFINITION. Let A be an involutive algebra and H a Hilbert space. A representation of A in H is a morphism of the involutive algebra A into the involutive algebra X(H). In other words, a representation of A in H is a map TT of A into 2(H) such that
TT(x + y) = TT(x) + TT(y), TT(xy) = TT(x)TT(y), TT(Ax) = ATT(x), TT(x*) = TT(x)*

for x, yEA, A EC. The (Hilbert) dimension of H is called the dimension of TT and is denoted by dim TT. The space H is called the space of TT and is denoted by H". Two representations TT and TT" of A in Hand H are said to be equivalent, and we write TT = TT" if there is an isomorphism U of the Hilbert space H onto the Hilbert space H which transforms TT(x) into TT"(x) for each x E A. Hence the definition of a class 01 representations.

32

POSITIVE FORMS AND REPRESENTATIONS

[CR. 2, 2

(For convenience, we will often not distinguish between representations and classes of representations.) 2.2.2. An intertwining operator for TT and TT is any continuous linear operator T: H ~H' such that TTT(x) = 7T(x)T for any x E A. The operator U of the previous definition is an intertwining operator. The set of all intertwining operators for 7T and 7T is a vector space whose dimension is called the intertwining number of 7T and 7T, and is the same as the intertwining number of 7T and 7T. For let T: H ~ H be an intertwining operator for TT and 7T. Then T*: H'~H is an intertwining operator for 7T and 7T, because
T*7T(X) = (7T(X*)T)* = (T7T(X*))* = 7T(x)T*.
It follows from this, that

T*T7T(X) = T*7T(x)T =7T(x)T*T,

and thus ITI = (T* T)1/2 commutes with 7T(A). Let T = UI TI be the polar decomposition of T. We have, for each x E A,
(1)

U7T(x)IT! = UITI7T(X) = T7T(X) = TT(x)T= 7T(x)UITI.

If Ker T
(2)

= 0, then

ITi(H) is dense in Hand (1) implies U7T(X)

If moreover T(H)

=H, i.e. if H onto H, and (2) shows that

=TT(x)U. Ker T* =0, then

7T

and

7T

U is an isomorphism of are equivalent.

2.2.3. Let (7TJiEI be a family of representations of A in Hilbert spaces Hi Let H be the Hilbert sum of the Hi If the set of numbers IITTi(x)1I is bounded for each x E A (which is the case, by 1.3.7, if A is an involutive Banach algebra), we can construct the continuous linear operator TT(x) in H which induces 7Ti(X) in each Hi Then x ~ TT(x) is a representation of A in H, known as the Hilbert sum of the 7Ti and denoted by E9iEr7Ti or 7T1 E9 7TzE9 E9 7Tn in the case of a (finite) family (7Tl 7Tz, ... , 7Tn ) of representations. If (7Ti)iEI is a family of representations of A each of which is equal to some representation 7T, and if Card 1= c, the representation E97Ti is denoted by C7T. Every representation equivalent to a representation of this type is called a multiple of TT.

CH. 2, 21

REPRESENT A nONS

33

2.2.4. Let 7T be a representation of A in H. If a closed vector subspace K of H is invariant under 7T(A), the restrictions of the 7T(X)S to K define a subrepresentation tr of A in K denoted by TTK or 7TE if E = P K In this situation H K is also invariant under 7T(A), since if g E K and "1 E H K, we have that 7T(X*)g E K for each x E A, and therefore

(7T(x)T/1 g) = (T/I7T(X)*g)

= (T/I7T(X*)g) = 0,

so that 7T(X)T/ E K. Hence P K commutes with 7T(A). If tr" is the subrepresentation of 7T defined by K, then 7T = 7T E!1Tr". Let p, p be two representations of A. If p is equivalent to a subrepresentation of p, we say that p is contained in p or that p contains o, and we write p ~ p or p.~ p,

He

He

2.2.5. Let 7T be a representation of A in H and let g E H. The closure of 7T(A)g is a closed subspace of H, invariant under 7T(A). If this subspace is equal to H, we say that g is a cyclic vector for 7T.
2.2.6. PROPOSITION. Let Tr be a representation of A in H, and let K be the closed subspace of H generated by the 7T(X)g (x E A, g E H). Let K be the closed subspace of H consisting of those g E H such that 7T(X)g = o for every x E A. Then K and K are invariant under 7T(A) and 7T(A) [the commutant of 7T(A) in .P(H)], are mutually orthogonal and their direct sum is equal to H.

Clearly, K and K are invariant under 7T(A). Any operator that commutes with 7T(X) leaves the kernel and range of 7T(X) invariant so that K and K are invariant under 7T(A). Let g E H. We have

g E K ~

(7T(A)g IH)

= O~ (g ITr(A)H) = 0

~gEHeK,

hence K=HeK. K is called the essential subspace of 7T. 7T is said to be non-degenerate if K = H. The above argument shows that every representation of A can be uniquely expressed as the direct sum of a trivial representation and a non-degenerate representation. 2.2.7. It is plain that a direct sum of non-degenerate representations is non-degenerate and that a representation that admits a cyclic vector is non-degenerate. Conversely:

34

POSITIVE FORMS AND REPRESENTATIONS

[CH. 2, 3

PROPOSITION. Every non-degenerate representation of A is a direct sum of representations that admit cyclic vectors.

He

Let tt be a non-degenerate representation of A in H with H - O. It is enough to show that there is a sub-representation of Tr in a non-trivial space which admits a cyclic vector (for the result then follows from an application of Zorns lemma). Let g be a non-zero element of H. The closure K of Tr(A)g is non-trivial and invariant under 1r. Let L = K, and put g = g! + g2 with gl E K, g2 E L. If x E A, then 1r(x)gt E K, 1r(X)g2 ELand 1r(x)gl + Tr(X)g2 =1r(x)g E K, so that Tr(X)g2 =0; con sequently Tr(x)g = 1r(x)g!. Thus K is the closure of 1r(A)g!.

2.2.8. Let tt be a representation of A in H and let ii be the Hilbert space conjugate to H (ii = H, but when passing from H to ii, mul tiplication by A E C is replaced by multiplication by A and scalar products are changed into their complex conjugates). Let AO be the reversed involutive algebra of A. It is easily checked that the map x ~ 1r(x*) is a representation of AO in tt, which is denoted by iTo. 2.2.9. Let A be the involutive algebra obtained from A by adjoining an identity and let 1r be a representation of A in H. Then tt has a unique extension (known as the canonical extension) to a representation ii of A in H such that ii(l) = 1. 2.2.10. Let A be an involutive Banach algebra and tt a representation of A in H. Recall (1.3.7) that 1I1r(x)11 ~ Ilxll for every x E A. If A has an approximate identity (u i ) and if tt is nondegenerate, then 1r(uJ tends strongly to 1. Indeed, for each yEA and each g E H, Tr(uJ( Tr(Y )g) = Tr(UiY)g tends strongly to 1r(y)g since

IITr(u;y) -

Tr(Y )11 ~

IluiY - yll ~

0;

now, the Tr(y)g constitute a total subset of H and moreover lIu;11 ~ I, so that 1r(uJ1/ tends strongly to 1/ for each 1/ E H. References: [618], [1101], [1323].

111r(uJII ~

2.3. Topologically irreducible representations

2.3.1. PROPOSITION. Let A be an involutive algebra, H a Hilbert space and tt a representation of A in H. Then the following conditions are equivalent:

CH. 2, 3]

TOPOLOGICALLY IRREDUCIBLE REPRESENTATIONS

35

(i) The only closed subspaces of H invariant under 7T(A) are 0 and H; (ii) The com mutant of 7T(A) in 2(H) is just the set of scalar operators; (iii) Every non-zero vector of H is a cyclic vector for 7T, or 7T is I-dimensional.

(i)::} (iii): suppose condition (i) is satisfied. Let ~ E H, f- O. If 7T(A)~ is not dense in H, then 7T(A)~ =0 by (i). Thus C~ is invariant under 7T(A), and so H = C~ and 7T is I-dimensional. (iii)::} (i): suppose condition (iii) is satisfied. Let K - 0 be a closed subspace of H invariant under 7T(A). We must show that K = H, which is plain if dim 7T = 1. Now suppose that every non-zero vector of H is a C K and 7T(A)~= cyclic vector for 7T, and let ~ E K, ~- O. Then 7T(A)~ H, so that K = H. (ii)::} (i): suppose condition (ii) is satisfied. Let K be a closed sub space of H invariant under 7T(A). P K then commutes with 7T(A) (2.2.4), and is therefore a scalar operator so that either P K = 0 or P K = 1, i.e. K=OorH. (i)::} (ii): suppose condition (i) is satisfied. Let T be an element of 2(H) that commutes with 7T(A); we show that T is scalar. Since T + T* and T - T* commute with 7T(A), we need only consider the case of hermitian T. The spectral projections of T then commute with 7T(A) and are all therefore equal to either 0 or 1, by (i). Hence T is scalar. 2.3.2. DEFINITION. Let A be an involutive algebra, H a Hilbert space and 7T a representation of A in H. 7T is said to be topologically irreducible if H - 0 and if 7T satisfies the equivalent conditions of 2.3.1. Such a representation is either trivial, l-dirnensional or else non trivial and non-degenerate. We denote by A the set of classes of non-trivial topologically irreducible representations of A. Remember that irreducibility of 7T in the algebraic sense means that the only subspaces of H invariant under 7T(A) are {O} and H. If dim H = +00, this is a condition far more restrictive than topological irreducibility. We shall presently see (2.8.4), however, that if A is a C*-algebrathen the two notions are equivalent. A finite-dimensional representation of A is a direct sum of topologi cally irreducible representations (2.3.5). We shall see (8.5.2) that this result also holds to a certain extent for infinite-dimensional represen tations, albeit in a rather more subtle form, which is one of the reasons why we shall be particularly concerned with topologically irreducible

36

POSITIVE FORMS AND REPRESENTATIONS

[CH. 2, 3

representations in what follows. A fundamental problem associated with any given involutive algebra A is that of determining, up to equivalence, all the topologically irreducible representations of A. It should be noted that theremay well not exist any non-zero topologically irreducible representations(or even any non-zero representationsat all) of A; we shall nevertheless see (2.7.3) that every C*-algebra has "enough" topologically irreducible representations. 2.3.3. Let A be a separable normed involutive algebra and 7T a to pologically irreducible representation of A in a Hilbert space H. Then H is separable: this is obvious if 7T is zero l-dimensional; otherwise, let ~ be a non-zero vector of H; if (x n ) is a dense sequence in A, then the 7T(Xn)~ are dense in H. The same argument shows that, for any normed involutive algebra B, the dimensions of the topologically irreducible representations of B are bounded above by some fixed cardinal. 2.3.4. Let A be an involutive algebra and 7T, 7T two topologically irreducible representations of A with intertwining number n. If n = 0, then 7T and 7T are not equivalent. If n > 0, let T: H",~ H"" be a non-zero intertwining operator. By 2.2.2, T*T and TT* are non-zero scalar operators and 7T and 7T are equivalent. 2.3.5. Let A be an involutive algebra. We study the finite-dimensional representations of A, and while what we say here can be regarded as a special case of certain theorems which we shall meet later on, or equally as a special case of some purely algebraic results, we think it better, for the convenience of the reader, to present a direct treatment here and now. Let 7T be a finite-dimensional representation of A. Then
7T = 7TI

EB ... EB 7Tn ,

where the 7Tj are irreducible. This is obvious if dim 7T = 0 with of course, n = O. Now suppose that dim 7T = q and that the assertion has been proved for dim 7T < q. If 7T is irreducible, there is nothing to prove. Otherwise, 7T = 7T EB 7T" where dim 7T < q, dim 7T" < q and we merely have to make use of the inductive hypothesis. The decomposition 7T = 7T1 EB ... EB 7Tn is not unique; for example, the representation A ~ A . I of e in en for n > I admits infinitely many different decompositions into l-dimensional representations. We shall, nevertheless, establish a uniqueness result. Let PI and P2 be two irreducible subrepresentations of 7T, and PI and P 2 the projections of H", onto H p 1 and H!2 These

en. 2. 4l

POSITIVE FORMS AND REPRESENTATIONS

37

projections commute with IT(A), and so the restriction of P2 to HpJ is an intertwining operator for PI and P2 Hence, unless H p1 and Hpz are orthogonal, PI =" P2 (2.3.4). This proves that every irreducible subre presentation of IT is equivalent to one of the IT/s, and thus, rearranging the IT/s, we see that IT = VI EB ... EB Vm, where each Vi is a multiple Pivi of an irreducible representation vi and the vis are mutually inequivalent. If P is an irreducible sub-representation of IT, the above discussion shows that H; is orthogonal to all but one of the Hvjs and so H; is contained in one of the Hvjs. This proves that every subspace H is uniquely determined, namely, it is the subspace of H", generated by the spaces of the subrepresentations of IT equivalent to vi. Thus, in the decomposition IT = PI v; EB ... EB Pmv;" of IT(v;, ... , v;,. irreducible and inequivalent), the integers Pi and the classes of the vi are uniquely determined, just as are the spaces of the Pivi.
Vj

References: [1101], [1323].

2.4. Positive forms and representations
2.4.1.
PROPOSITION. Let A be an involutive algebra.

(i) If IT is a representation of A in Hand

a positive form on A. (ii) Let IT and ITbe representations

g E H,

then x ~ (IT(x)g

Ig)

is

of A in Hand H, and let

g (resp.

be a cyclic vector for IT (resp. IT).If (IT(x)g g) = (IT(x)f f) for every x E A, there is a unique isomorphism of H onto H mapping IT to ITand g to f.

f)

We have
(IT (x x)g

Ig) = (IT(x)*IT(x)g Ig) = 1IIT(x)gI12 ~

0,

which gives (i). Now suppose the conditions of (ii) are satisfied. For any x, yEA we have
(IT(x)g !IT(y)g) = (IT(y*x)g ~)

I = (IT(y*x)fIf)

= (IT(x)f 1IT(y)f).

(resp. IT(x)f) are dense in H (resp. H) it follows that Since the 'IT(x)~ there is an isomorphism U of H onto H such that U( 'IT(x)~) = IT(x)f for any x E A. We show that U transforms IT into IT,i.e. that UIT(x) = IT(x)U for each x E A; for every yEA, we have
(UIT(x(IT(y)g) = UIT(xy)g = IT(xY)f
= IT(x)(IT(y)f)= (IT(x)U)(IT(y)g).

38

POSITIVE FORMS AND REPRESENTATIONS

[CR. 2, 4

Since the 7T(Y)~ are dense in H, it follows that U7T(X) Moreover, for each x E A,

= 7T(X)U.

(f I7T(x)f) =

(~I7T(X)~)

= (U~ I U7T(X)~)

=(U~

17T(X)f),

which implies that f = U~. Finally, the uniqueness of U is immediate since the values that U takes on the (dense) set of the 7T(A)~ are predetermined. 2.4.2. In the above notation, the form x ~ (7T(X)~ I~) on A is called the [orm defined by 7T and ~. If 7T is fixed while ~ varies in H, we obtain the [orms associated with 7T. If S is a set of representations of A, the forms associated with S are just the forms associated with the various ele ments of S. Let H be a Hilbert space, B an involutive subalgebra of .2(H) and ~ an element of H. We denote by we the positive form on B defined by the identical representation of B and ~, i.e. the form x ~ (x~ I~). A positive form on B is said to be a vector form if it is equal to we for a suitable choice of ~ in H. 2.4.3. Let A be an involutive Banach algebra with an approximate identity (u;), 7T a non-degenerate representation of A in H, ~ an element of Hand f the positive form defined by 7T and f Then Iitli = (~ Ig). In fact, by 2.1.5 (v),
1It11 = lim f(uj) = lim( 7T(U;)g I g),

and 7T(U;) tends strongly to I (2.2.10). It follows that if A is the involutive algebra obtained by adjoining an identity to A, and j and iT are the canonical extensions of f and 7T to A, then
j(x) = (iT(x)~

Ig)

for each x

EA.

In particular, if 7T is the identical representation, assumed to be non degenerate, of a sub-C*-algebra A of 2(H) which does not contain I, then the canonical extension to A = A + C . I of we IA is WE IA. 2.4.4.
PROPOSITION. Let A be an involutive Banach algebra with an approximate identity, A the involutive algebra obtained by adjoining an identity to A, f a continuous positive fonn on A, j its canonical extension to A, N the left ideal of A consisting of those x E A such that j(x*x) = 0, Ai the (Hausdorff) pre-Hilbert space AIN and AI the Hilbert space which is the completion of Ai. For each x E A, let 7T(X) be the operator

CR. 2, 4]

POSITIVE FORMS AND REPRESENTATIONS

39

in AI N obtained from left multiplication by x in A by passage to the quotient. Let g be the canonical image of 1 in A f. (i) Each 7T(X) has a unique extension to a continuous linear operator 7T(X) in At (ii) The map x ~ 7T(X)(X E A) is a representation of A in At. (iii) g is a cyclic vector for 7T(A). (iv) f(x) = (7T(X)g g) for each x E A.

By 2.1.5 (ii), we have, for x, yEA,

(7T(X )7T(y)g I 7T(X )7T(y )g)

=/(y*x*xy) ~ Ilx* xll/{Y*y) = Ilx* xl/( 7T(y)g 17T(y )g),

from which (i) follows. It is plain that 7T and then 7T also, are represen tations in the sense of the algebra structure. For x, Y, z E A, we have
(7T(X)7T(y)g !7T(Z)g)

= j{z*(xy = j{(x*z)*y) = (7T(y)g 17T(X*)7T(Z)g),

which implies that 7T(X)* = 7T(X*), and (ii) is proved. The set 7T(A)g is the canonical image of A in A f and is therefore dense in At by 2.1.5 (vii); this proves (iii). Finally, we have for each x E A,
(7T(X)g

Ig) =fO*xl} =j{x).

g are defined
by f, and

We say that the representation 7T and the vector we denote them by 7Tt and gt.

2.4.5. With the above notation, let M be the left ideal of A consisting of those x E A such that f(x*x) = O. The canonical image of A in At can be identified with AIM, and is moreover dense in At. At can thus be defined as the completion of the Hausdorff pre-Hilbert space AIM, and for each x E A, 7T(X) can be defined as the continuous extension of the operator in AIM of left multiplication by x. This avoids the introduction of A and 1. although, in this approach, the definition of g would be less simple. 2.4.6. Let A be an involutive Banach algebra with an approximate identity, f a continuous positive form on A, and 7T and g the represen tation and vector defined by f. Then by 2.4.4 (iv) f is precisely the positive form defined by 7T and g. Conversely, starting with a representation 7T of A in a Hilbert space and a cyclic vector g for 7T, let f be the positive form defined by 7T and g, which is continuous as 7T is. Let 7T and f be the representation and

40

POSITIVE FORMS AND REPRESENTATIONS

[CH. 2, 4

vector defined by f. Then

(7i(x)g 1g) = f(x) = (7i(x)f 1f)

for each x E A,

and f is a cyclic vector for 7i. By 2.4.1 (ii), there is a unique isomor phism of H", onto H"" mapping 7i to tt and g to f. In particular, let tt be a non-trivial topologically irreducible represen tation of A. Every non-zero vector of H", is a cyclic vector for tt, and so tt is defined up to equivalence by any non-zero form associated with tr. 2.4.7.
PROPOSITION. Let A be an involutive Banach algebra with an approximate identity, and tt a representation of A. For each x E A,

117i(x)!f = sup f(x*x),

where f varies over the set of positive forms associated with 11111::;;; 1.

7i such that

Thanks to 2.2.6, we need only consider the case of non-degenerate tt, By 2.4.3, the positive forms associated with 7i of norm ::;;;1 are just the forms we 0 7i where g E H, IIgll::;;;1. Thus

117T(X)11 2 =
=

SUp(7T(X)~ Ilell"

17T(X)g)
0

sup(we
lIell")

7i)(x*x).

2.4.8.

PROPOSITION. Let A be an involutive Banach algebra with an approximate identity, f a continuous positive form on A and tt and g the representation and the vector defined by f. (i) If Xo E A, the form x ~ f(x~xxo) is associated with at, (ii) Iff is a positive from associated with tt, f is the limit (in the norm topology) of forms x ~ f(x~xxo), where Xo E A.

If Xo E A, then

and we have (i).

CH. 2, 5)

PURE FORMS AND IRREDUCIBLE REPRESENTATIONS

41

Now let ~' E H 7T and f = wf 0 7T. For every E > 0 there is an Xo E A such that 117T(Xo)~ - fll :os;; E, and so for each x E A,

If(x) - f(x~xxo)1

= 1(7T(X)f If) :os;;

(7T(X)7T(XO)~

17T(Xo)~)1

117T(X)fll11f - 7T(xo)~11 + 117T(x)f - 7T(X)7T(Xo)~II'II7T(Xo)~11 :os;; Ilxli. Ilfll. E+ Ilxll. E(11f11 + E) 2 ), =Ilxll(2Ellfll+ E
and

2Ellfll + E is
2

arbitrarily small.

2.4.9. Let A be a C*-algebra, I a closed two-sided ideal of A, B the C*-algebra AI I and w: A ~ B the canonical morphism. If f is a positive form on A such that f(l) =0, then f defines a positive form g on B by passage to the quotient. For each x E A, we have

so that 7Tg (2.4.1).

and

~g

may be identified with 7Tf and

~f

respectively

PROPOSITION. Let A be a C*-algebra, f a positive form on A and I a closed two-sided ideal of A. Then f(l) = 0 if and only if 7Tf(l) = O.

Ker 7Tf C Ker f, so that 7Tf(I) = 0 implies f(l) = O. If f(l) = 0, then 7Tf = 7Tg 0 W in our previous notation, so that 7Tf(I) = O. 2.4.10. COROLLARY. Ker 7Tf is the largest closed two-sided ideal of A contained in Ker f. 2.4.11. COROLLARY. Let A be a C*-algebra and f and g positive forms on A. Then the following conditions are equivalent: (i) Ker 7Tf C Ker 7Tg (ii) g vanishes on Ker 7Tt References: [618], [619], [638], [1097], [1101], [1323], [1455]. 2.5. Pure forms and irreducible representations We know how to associate representations with positive forms, and we now settle the question of when this procedure leads to irreducible representations.

42

POSITIVE FORMS AND REPRESENTATIONS

[CH. 2, 5

2.5.1. PROPOSITION. Let A be an involutive algebra, 7T a representation of A in H, g an element of Hand f the positive form defined on A by 7T and g. (i) If T is an hermitian operator on H which commutes with 7T(A) and satisfies 0 ~ T ~ I, the form X~(7T(x)Tg I Tg) = (7T(X)g I T 2g)

on A is a positive form fT which is dominated by f. (ii) If g is a cyclic vector for 7T, the map T ~ [r is injective. (iii) If A is an involutive Banach algebra with an approximate identity, every continuous positive form on A which is dominated by f can be written fT for some T.
Since fT = CUT{
0

7T,

fT ;;;0. If x E A,

fT(x*x) = (7T(x*x)Tg I Tg) = 117T(X) Tgl1 2 = IIT7T(X)gW ~ 117T(X)gI12 = f(x*x),

so that fT ~

f,

and (i) is proved.

If fT =t-. then (7T(X)g I T2~) = (7T(X)g I T2g) for any x E A, so that T 2g = T,2g if g is a cyclic vector for 7T(A). Since g is then a separating vector for the commutant of 7T(A) (A 14) it follows that T 2= T2, and hence T = T as T;;;. 0, T;;;. O. This proves (ii). Let g be a positive form on A which is dominated by f. For x, yEA Ig(y*x)12~ g(x*x)g(y*y) ~ f(x*x)f(y*y) = 117T(x)gW 1I7T(y)gW
Hence the relation defines a unique continuous sesquilinear form on the subspace 7T(A)g of H which is clearly positive and hermitian. There then exists an her mitian operator To on X = 7T(A)g such that 0 ~ To ~ I and

g(y*x) = (7T(X)g I T 0 7T(y )g)

For x, y, z E A, we have

(7T(y)g I T o7T(Z)7T(X)g) = g((zx)*y) = g(x*(z*y = (7T(Z*y)g I T 07T(X)g) = (7T(Z)*7T(y)g I To7T(X)g) = (7T(y)g !7T(z)T o7T(X)g),

CH. 2, 5j

PURE FORMS AND IRREDUCIBLE REPRESENTATIONS

43

from which it follows that T o7T(Z) =7T(z)To on X. Moreover, X is invariant under 7T(A) and so P x commutes with 7T(A). Hence ToP x is an hermitian operator in H, lying between 0 and I, which commutes with the 1T(Z)S (z E A). Let T be its positive square root, which again commutes with the 7T(Z)S. Then O~ T ~ 1, and
g(y*x)

I T27T(Y)~) = (7T(x)T~ = (7T(Y*X)n In) = fT(Y*X).

= (7T(X)~

17T(y)TD

Finally, suppose that A is an involutive Banach algebra with an approximate identity (uJ and that g is continuous. Then
g(y*)
=

lim g(Y*Uj),

and thus g

fT

2.5.2. DEFINITION. Let A be a normed involutive algebra and f a continuous positive form on A. f is said to be pure if f- 0 and if every continuous positive form on A which is dominated by f is of the form >..f(0 ~ A ~ 1). We denote by P (A) the set of pure states of A. Let n be a locally compact space and A the C*-algebra of continuous complex-valued functions on n that vanish at infinity. The pure positive forms on A may be identified with the positive measures on n whose support consists of a single point, i.e. with measures of the form f ~ Af(w) where A> 0 and co is a fixed point of n: it follows from this that the pure states of a commutative C*-algebra are just the characters of the algebra. 2.5.3. Let A be an involutive Banach algebra with an approximate identity, A the involutive Banach algebra obtained by adjoining an identity to A, f a continuous positive form on A and f its canonical extension to A. Then f is pure if and only if f is pure. Indeed, to begin with, the conditions f = 0 and f = 0 are equivalent. Moreover, as g runs through the set of continuous positive forms on A dominated by f, g runs through the set of continuous positive forms on A dominated by f (2.1.7). Finally, for g to be of the form Af where 0 ~ A ~ I it is necessary and sufficient that g = At 2.5.4. PROPOSITION. Let A be an involutive Banach algebra with an approximate identity, f a continuous positive form on A and 1T the representation of A defined by f Then 1T is non-trivial and topologically irreducible if and only if f is pure.

44

POSITIVE FORMS AND REPRESENTATIONS

[CR. 2, 5

Let ~ be the vector of H", defined by f. Suppose f is pure, and let E be a projection in H", which commutes IE~) on A is continuous and positive with 7T(A). The form x ~ (7T(x)E~ and is dominated by f [2.5.1 (i)], and is therefore equal to Af for some o A~ 1. Thus
(7T(x)E~

IE~) = (7T(X)A l/2~ IA 1/2~)

for each x E A.

By 2.5.1 (ii), E = 11. 1/ 2 1 and so E = 0 or 1. Furthermore, there exist x E A such that f(x):j; 0 and hence such that (7T(X)~ I~):j; 0, which shows that 7T is non-trivial and topologically irreducible. N ow suppose 7T is non-trivial and topologically irreducible. There I~) :j; 0 and hence f:j; O. Let g be a con exist x E A such that (7T(X)~ tinuous positive form on A dominated by f. By 2.5.1 (iii), there is a hermitian operator T E 7T(A) such that 0 ~ T ~ I and g(x) = (7T(X)n I for every x E A. Since 7T is topologically irreducible, T = p., . 1 with 0 ~ p., ~ 1 and so g = p., t. Hence f is pure, and the proposition is proved. Proposition 2.5.4 allows us to define a canonical map

n)

P(A)~A.

This map is surjective by 2.4.6 and 2.5.4, and the inverse image in P (A) of 7T E A is the set of states associated with 7T (on this subject, d. 2.5.7).
2.5.5. PROPOSITION. Let A be an involutive Banach algebra with an approximate identity and B the set of continuous positive forms on A of norm ~ I. (i) B is convex and compact in the weak*-topology (T(A, A) of the dual A of A. (ii) The extreme points of P consist of 0 and the pure states. (iii) B is the weak*-closed convex hull of 0 and the set of pure states.

B is a weak*-closed convex subset of the unit ball of A. This ball is weak*-compact and so we have (i). We next show that 0 is an extreme point of B. If fEB and - fEB, then f(x*x) = 0 for each x E A, so that If(xW ~ Ilfllf(x*x) = 0 (2.1.5); hence f= O. Now let f be a pure state, and suppose f = Afl + (1 - A)f2 with 0 < 11.< 1, fl f2 E B. Then Afl is dominated by f so that Afl = p.,f with 0 ~ p., ~ 1. Since

CR. 2, 6j

EXISTENCE OF REPRESENTATIONS

OF C"-ALGEBRAS

45

and IlfIII, IIf211:o;:; 1 we must have Ilf,11 = IIf211 = 1, and therefore A = I.t and f. = f = f2 We have thus proved that f is an extreme point of B. Conversely, let f be a .non-zero extreme point of B. Clearly Iitli = 1. Now let f = I, + f2 with I, f2 continuous positive and non-zero. Put 11f,1I = A, so that IIf211 = 1 - A, and let g, = A"t; g2 = (1- A)-f2 Then f = Agl + (1- A)g2, with gl, g2 E B. Since f is extreme, f = gl = g2 Hence t. = Af, f2 = (1- A)f from which it follows that f is a pure state. This proves (ii). Finally (iii) follows from (i) .while (ii) follows from the Krein-Milman theorem. 2.5.6. We retain the above notation and assume in addition that A is unital. B is then the set of positive forms f on A such that f(l) :0;:; 1, and the set E(A) of states of A is the set of positive forms f on A such that f(1) = 1. It follows that E(A) is convex and weak*-compact, and that the set of extreme points of E(A) is the set of those extreme points of B which belong to E(A), i.e. the set P (A). E(A) is therefore the closed convex hull of P (A). 2.5.7. PROPOSITION. Let A be an involutive algebra, 7T a non-trivial topologically irreducible representation of A, ~l and ~2 two vectors of H", and fI and f2 the positive forms defined by (7T, ~,) and (7T, ~2)' Then f, = f2 if and only if there is a complex number A of absolute value I such that
~2

= A~,.

If ~2 = A~, with IA I= 1, then clearly I, = f2 On the other hand, suppose f, = f2 Since ~l and ~2 are cyclic vectors for 7T (2.3.1), there is an automorphism U of H commuting with the 7T(X)S such that U~I = ~2 (2.4.1 (iij). Now U is a scalar operator (2.3.1), and so ~2 = A~, with IAI = 1. In particular, the canonical map P (A) ~ A is bijective if and only if every topologically irreducible representation of A is one-dimensional. When A is a C*-algebra, theorem 2.7.3 implies that this condition is fulfilled if and only if A is commutative. References: [618], [619], [6381, [l097], [1101], [1323], [1455].

2.6. Existence of representations of C*-algebras

2.6.1. THEOREM. Any C*-algebra A has an isometric representation on a Hilbert space.

46

POSITIVE FORMS AND REPRESENTATIONS

[CH. 2. 6

Let A h be the real Banach space consisting of the hermitian elements of A, and let x be a non-zero element of A. By (1.6.1), -x*xe A+, and since A + is a closed convex cone (1.6.1), there is a continuous linear form t, on A h such that fAy p~ 0 if yEA + and Ix(- x* x) < 0 (B 5). Identifying fx with an hermitian form on A, we see that t, is a positive form on A and fx(x*x) > 0, so that the representation 7Tx defined by i, following 2.4.4 satisfies 7Tx(X) # O. Let the representation 7T be the direct sum of the 7TxS for x E A, x:;t: O. Then 7T is injective and therefore isometric (1.3.7 and 1.8.1) and the theorem is proved. Thus, as we previously asserted, the closed involutive sub-algebras of Ie(H) for H a Hilbert space are indeed the most general examples of C*-algebras. 2.6.2. PROPOSITION. Let A be a C*-algebra and let x following conditions are equivalent: (i) x;;;;. O. (ii) The operator 7T(X) is ;;;;. 0 for every representation (iii) f(x);;;;. 0 for every positive form f on A.
E

A. Then the

7T

of A.

(i) ~ (iii): obvious. (iii) ~ (ii): let 7T be a representation of A and ~ E H",; the form y ~ (7T(Y)~ I~) on A is positive, and so (7T(X)~ I~) ;;;;. 0 if (iii) holds. The operator 7T(X) is thus > O. (ii) ~ (i): let 7T be an isometric representation of A (2.6.1); if the operator 7T(X) is ;;;;. 0, then 7T(X) is positive in 7T(A) (1.6.5) and so x is positive in A. 2.6.3. COROLLARY. Let A be a C*-algebra, A h the set of hermitian elements of A, B the set of positive forms on A of norm ~ 1 and PCB the set of pure states of A, where Band P are endowed with the weak*-topology. Let cg(B), cg(P) be the sets of continuous real-valued functions on Band P respectively, and for each x E A h let F, be the continuous real-valued function f -+ <I, x) on B and Ox its restriction to P. Then the map x ~ F, (resp. x -+ Ox) is an isometric isomorphism of the ordered Banach space A h onto a subspace of the ordered Banach space cg(B) [resp. cg(P)].
It is clear that the maps are linear and that
x ;;;;. 0 ~ F, ;;;;. 0 ~ Ox ;;;;. O.

Suppose conversely that Ox;;;; 0, i.e. that f(x);;;;. 0 for each f E P. By

CH. 2. 7]

INVOLUTIVE BANACH ALGEBRA

47

2.5.5, f(x) ~ 0 for each fEB, and so x ~ 0 (2.6.2). Now suppose A is represented in a Hilbert space H (2.6.1) and let x E A h with IIxll = 1. There exist g E H of norm 1 with I(xg Ig)1 arbitrarily close to 1, hence there exist fEB with If(x)1 arbitrarily close to 1 and therefore also f E P with If(x)! arbitrarily close to 1. Consequently, 1 ~ IIOJ ~ Fxll ~ 1 and the corollary follows. When A is commutative, the map x ~ Ox is simply the Gelfand isomorphism restricted to A h An alternative generalization of the Gel fand isomorphism to the non-commutative case, more satisfactory in some ways, will be studied later on (10.5.4).
2.6.4. COROLLARY. Let A be a C*-algebra and g a continuous hermitian linear form on A. Then there exist positive forms f, I on A such that
g

=f - 1, Ilgll = 11111 + Ilfll

We keep the notation of 2.6.3, and identify A h with a subspace of Cf5(B). By the Hahn-Banach theorem, g can be extended to a linear form p. on Cf5(B) with 11p.11 = IIgli (d. 1.2.6). The measure p. on B can be written p. = p. +- P.- where p. +, P.- are positive measures such that lip.II = lip.+11 + lip.-II Let f = p. +!A h , I = p. -I A h Then g = f - I and

IlglI ~ 11111 + 11111 ~ lip.+11 + IIJL-II =IIJLII =Ilgll,

so that

Ilgll = 11111 + 11f11

We shall see later (12.3.4) that the decomposition of 2.6.4 is unique. References: [618], [619], [681], [848], [1097], [1101], [1323], [1455], [1589]. 2.7. The enveloping C*algebra of an involutive Banach algebra 2.7.1. PROPOSITION. Let A be an involutive Banach algebra with an approximate identity. Let R be the set of representations of A, R the set of topologically irreducible representations of A, B the set of continuous positive forms of norm ~ 1 on A, and P the set of pure states of A. For each x E A, wehave (1) sup
1TER

117T(x)1I

sup
1TER

117T(x)11

sup f(X*X)1/2
fEB

sup f(x*x)1/2,
fEP

and denoting the common value of these four expressions by IIx II, we

48

POSITIVE FORMS AND REPRESENTATIONS

[CH. 2, 7

have Ilxll",;;;[x]. The map x ~ Ilxll is a seminorm on A such that

Ilxyll",;;;llxllllyll, Ilx*II=llxll, Ilx*x] = IIxll,2

for any x, yEA.

We denote the four numbers in the order in which they appear in (1) by a, b, c, d. d e: b: let f E P; then f is associated with a representation 1T E R (2.5.4) and f(x*x)",;;; 111T(X)W (2.4.7); b ",;;; a: obvious; a ",;;; c: this follows from 2.4.7; c";;; d: let fEB; by 2.5.5, f(x*x) lies in the closed interval whose left-hand end-point is 0 and whose right-hand end-point is SUPgEP g(x*x), i.e. f(x*x)";;;SUPgEPg(X*X). By 1.3.7, 111T(X)II",;;; Ilxll for every 1T E R, so that IIxll",;;;[r]. Moreover, each function x ~ 1I1T(x)1I is a seminorm on A, hence x ~ IIxll is a seminorm on A. We have

for each 1T E R, and so IIx*II = IIxll and IIx*xll = IIxll,2.

Finally, if x, yEA, we have for each 1T E R, 111T(XY)II",;;; 111T(x)III/1T(y)ll",;;; Ilxll llyll, hence Ilxyll",;;;Ilxll llyll

2.7.2. Let I be the set of x E A such that Ilxll = 0, which is a closed self-adjoint two-sided ideal of A. The map x ~ IIxll defines a norm on the quotient AlI. Endowed with this norm, All satisfies all the C*-algebra axioms except that AI I is not complete in general. The completion B of AI I is a C*-algebra called the enveloping C*-algebra of A. The canonical map of A into B is a norm-reducing involutive algebra morphism whose image is dense in B. 2.7.3. When A is a C*-algebra, IIxl/ = IIxll by 2.6.1, and A may be identified with its enveloping C*-algebra. In this case, proposition 2.7.1 shows the existence of "sufficiently many" topologically irreducible re presentations, as we had already claimed, and considerably strengthens theorem 2.6.1:

CH. 2, 7j

INVOLUTIVE

BANACH ALGEBRA

49

THEOREM. For any C*-algebra A, there is a family (7T;) of topologi cally irreducible representations of A such that Ilxll = sup; II7Tj(x)11 for every x E A.

2.7.4. PROPOSITION. Let A be an involutive Banach algebra with an approximate identity, B the enveloping C*-algebra of A and T the canonical map of A into B. (i) If 7T is a representation of A, there is exactly one representation p of B such that 7T = pOT, and pCB) is the C*-algebra generated by 7T(A). (ii) The map 7T ~ P is a bijection of the set of representations of A onto the set of representations of B. (iii) p is non-degenerate (resp. topologically irreducible) if and only if 7T is non-degenerate (resp. topologically irreducible).

Let 7T be a representation of A. In the notation of 2.7.2, 7T vanishes on I, and defines, by passing to the quotient, a representation 7T of All such that 117T(z)11 ~ Ilzll for each zEAl I, where II II denotes the norm of B. 7T therefore extends to a representation p of B which satisfies 7T = pOT. The uniqueness statement of (i) follows from the fact that T(A) is dense in B. This also implies that 7T(A) is dense in pCB) in the operator-norm topology. Since pCB) is a C*-algebra (1.8.3), it is the C*-algebra generated by 7T(A). Statements (ii) and (iii) are immediate. Proposition 2.7.4 shows that B is the solution of a universal problem. It also shows that in the majority of questions concerning represen tations of involutive Banach algebras with an approximate identity, it is enough to deal only with the C*-algebra case.
With A, B, T as in 2.7.4: a continuous positive [orm on A, there is exactly one positive form g on B such that f = gOT. Moreover Ilgll = 11!l1. (ii) The map f ~ g is a bijection of the set of continuous positive forms on A onto the set of positive forms on B. (iii) Let M be a bounded set of continuous positve forms on A. Then the map f ~ g, restricted to M, is bicontinuous for the weak*-topologies u(A, A), at B, B), where A, B denote the duals of A and B respec tively.
PROPOSITION.

2.7.5.

(i) If f

is

Let f be a continuous positive form on A. Then for each x E A we have, using 2.1.5 (i),

50

POSITIVE FORMS AND REPRESENTATIONS

[CH. 2, 8

Ilgll ~ IlflI. If y E

so that there is a continuous linear form g on B such that f = gOT and B there is a sequence (x n ) in A such that T(Xn ) ~ Y which implies that
g(y*y)

=lim f(x~xn) ~ 0, and so g is positive. If x E A and Ilxll ~ 1, then If(x)1 = Ig( T(x1 ~ IIgllIIT(x)1I~ IIglllIxll~ IIgll,

and hence

1If11 ~ "gil

The uniqueness of g follows from the fact that T(A) is dense in B. We have now proved (i), and (ii) is immediate. Let MeA be a set of continuous positive forms on A and NcB its image under the map f ~ g. The map f ~ g of M onto N is plainly bicontinuous for the weak*-topologies T(A, A), T(B, T(A. If M is bounded, N is bounded by (i), so that (T(B, T(A and (T(B, B) coincide on N as T(A) is dense in B. 2.7.6. Let A be an involutive Banach algebra with an approximate identity, and let Q be the set of continuous positive forms on A. For each f E Q, let TTf be the representation defined by f. Then 7T = E9 f E Q TTf is called the universal representation of A, and is non-degenerate. By (2.2.7) and (2.4.6), every non-degenerate representation of A is a direct sum of representations of the form 7Tf so that IIxll= 117T(x)1I for every x E A. Hence if B is the enveloping C*-algebra of A, the representation of B corresponding to 7T is an isomorphism of B onto 7T(A). References: [582], [619], [638], [1097], [1101], [1323], [1455].

2.8. A theorem on transitivity

2.8.1. LEMMA. Let H be a Hilbert space, (gl, ... , gn) an orthonormal system in H, and vectors in H of norm ~ r. Then there is a bgn = no If there is b E .P(H) of norm ~ (2n)I/2r such that bg l = an hermitian element h of .P(H) such that hg. = hgn = b can also be chosen to be hermitian.

I",.,n

I,...I,... ,

n,

Let K be the subspace of H generated by g., ... , gm no We shall define b as an operator which leaves K invariant and which vanishes on He K. This reduces us to the case where K = H. Now let (gl, ... , gn, gn+(,... , gm) be an orthonormal basis for H, and let b be the operator in H whose matrix with respect to (gl, .. , gm) is as follows:

I" .. ,

CH. 2. 8l

A THEOREM ON TRANSITIVITY

51

(a) the first n columns are the coordinates of with respect to the ~i; (b) if h exists, the first n rows are the complex conjugates of the first n columns (this is possible since the existence of h implies that the square matrix comprising the first n rows and the first n columns is hermitian). (c) all the entries not already defined are zero. We have and

I... ,n

lib liz ,,;;; Tr(b*b)";;; 2(11lllz+... + "n"z) ,,;;; 2nr z.

If h exists, b is hermitian by construction.

2.8.2.

LEMMA. Let HI, ... , H p be Hilbert spaces, H = HI EB ... EB H p , and A a C*-algebra of operators on H which is strongly dense in the von Neumann algebra .2(H1) x x .2(Hp ) . Let (~I"'" ~n) be an orthonormal system in H and Tit, , TJn be vectors in H of norm ,,;;; r, where we assume that for i = 1, ... , n, ~i and TJi both lie in the same subspace Hj(i). Then there exists abE A of norm ,,;;; 3m I/Z such that b~1 = TJl ... , b~n =TJn If there is an hermitian element h of .2(H) such that h~1 = TJI , h~n = TJ.. then b can in addition be chosen to be an hermitian element of A.

By Kaplanskys density theorem, the unit ball of A (respectively of the hermitian part of A) is strongly dense in the unit ball of .2(H I) x ... x .2(Hp ) (respectively of the hermitian part of .2(H1) x ... x .2(Hp . N ow by 2.8.1 applied to the spaces HI>"" H; in turn, there is a Yo E .2(H I ) x ... x .2(Hp ) such that

and then by the preceding remark there is

Xo

E A such that

Next, again by 2.8.1, there exists

i: Yl!ol i: i: = TJI- x O!ol Yl!on

Yl

E .2(H1) x ... x .2(Hp ) such that

t: XO!on

= TJn -

IIYIII,,;;; -zl(2n)I/Z r

and then XI E A such that

52

POSITIVE FORMS AND REPRESENTATIONS

[CH. 2, 8

By induction, we construct sequences .2(H]) x ... x .2(Hp ) and x, E A such that

of

elements

Yk

(If h exists, the Yk and Xk can be chosen to be hermitian.) x o+ XI +... then converges in norm to an element b E A such that

bg] = TIl.

, bgn = TIm

Ilbll:s;;;(l +1+!+. )(2n)I/2r :s;;; 3m 1/2.

2.8.3. THEOREM. Let A be a C*-algebra, A the C*-algebra obtained by adjoining an identity to A, 7T., , 7Tp representations of A in Hilbert spaces H., ... , H p , and iT], ... , iT p their canonical extensions to A, where the 7Tj are assumed to be topologically irreducible, non-zero and mutually inequivalent. (i) Let T) E .2(H1) , , T p E .2(Hp ) and K), . . . .K; be finite-dimen sional subspaces of HI>... , H; respectively. Then there exists x E A such that 7Tj(x)IKj = T j IKj forj=l, ... ,p.

(ii) Let T] E .2(H]), ... , Tp E .2(Hp ) be hermitian operators and K., ... , K, finite-dimensional subspaces of HI ... , H p Then there is an
hermitian element x of A such that
7Tj(X)

Ix, =

T j I x, for j = 1, ... ,p.

(iii) Let T, E .2(H,), ... , Tp E .2(Hp ) be unitary operators and K., ... , K p finite-dimensional subspaces of H), ... , H p Then there is a unitary element x of A such that

iT/x) IK,

Let H = HI EB ... EB H p , 7T(A) is a sub C*-algebra of .2(H) which commutes with the projections E j = PHi Now if B is the von Neumann algebra generated by 7T(A), we have
7T(A) C .2(H 1) x ... x .2(Hp ) ,

= Tj IK, for j = 1, ... ,p. 7T = 7T1 EB ... EB 7Tp By 1.8.3,

and hence Be .:t(H) x ... x 5(Hp ) .

7T

Since each

7Tj

is non-degenerate,

is non-degenerate and so 7T(A) is

CH. 2, 81

A THEOREM ON TRANSITIVITY

53

strongly dense in B. Since TTj is topologically irreducible, E j is a minimal projection of TT(A) = B and the von Neumann algebra induced on H, by B is ;t(Hj ) . Let Fj E B n B be the central support of E j (A 10), and let j and k be distinct indices. If Fj and F k are not orthogonal, there exist non-zero projections Ej, E k of B dominated by E j and E, respectively and equivalent relative to B (A 44); since E j , E, are minimal projections of B, we have E, = E; E k= E b and so there is a partial isometry of B with E j and E, as initial and final projections respectively; but then TTj and TTk are equivalent which is a contradiction. Hence the Fjs are mutually orthogonal. Now Fj ~ E j for each j, and so Ej = Fj E B. Since BEl =!(H), we see that B:2 ;t(HI) X . x ;t(Hp ) and finally B = ;t(HI ) X 5(Hp ) . Let T I E 5(H I) , . . . , T p E ;t(Hp ) and Kj, . . . , K; be finite-dimensional subspaces of HI"",H p By 2.8.2, there is an x E A such that ITj(x) I K, = Tj I K, for j = 1, ... , p. If the T/s are hermitian, x can be chosen so that TT(x) is hermitian (2.8.2), and as TT(x) = TT(!(x + x*, x can itself be chosen to be hermitian. Now suppose the T/s are unitary. For j = 1, ... ,p there is a finite-dimensional subspace Kj:2 K, of H, and a unitary operator T, E 5(H j ) which leaves K, invariant and is such that t; Itc, = t, IK j ; there then exists an hermitian operator Tj E 5(H j ) , again leaving Kj invariant, such that exp(iTj) IKj = Tj IKj. By the foregoing work, there is an hermitian element y of A such that TT(y) IKj = TI I Kj for each j. Then x = exp(iy) is a unitary element of A and iTj(x) I K, = T, IK, for each j. 2.8.4. COROLLARY. Every topologically irreducible representation of a C*-algebra is algebraically irreducible. We have merely to apply theorem 2.8.3 with p = 1 and dim K I = 1. Thus we will henceforth speak of irreducible representations of a C*-algebra without further qualification. 2.8.5. COROLLARY. Let A be a C*-algebra, f a pure positive form on A and N the left ideal of those x E A such that f(x*x) =O. Then AIN, with the scalar product derived from f(y*x) is complete and therefore coin cides with the space of the representation defined by f. Indeed, let TT be this representation, topologically irreducible by (2.5.4). From the construction of TT, AI N is a subspace of H 1T invariant under TT. Now AIN,t 0 and so AIN = H 1T (2.8.4).

54

POSITIVE FORMS AND REPRESENTATIONS

[CH. 2, 9

2.8.6.

COROLLARY. Let A be a C*-algebra, A the C*-algebra obtained by adjoining an identity to A, and t, and f2 two pure states of A. Then t. and f2 define equivalent representations 7TI and 7T2 if and only if there is a unitary element u of A such that f2(x) = iJ(u*xu) for each x E A.

Let ~I be the vector of H", defined by fl The states of A that define representations equivalent to 7TI are just the states associated with 7TI (2.4.6), i.e. the forms x ~ (l1ix)~ I~) where ~ is a unit vector of H"I (2.4.3). Now the unit vectors of H"I are just the vectors iil(u)~1 where u is unitary in A (2.8.3) and iii is the canonical extension of 7TI to A. Finally,
('7Tb)iil(u)~11 iil(U)~I)

= ('7TI(u*xu)~11

~l)

= iJ(u*xu).

References: [633], [849], [1323].

2.9. Ideals in C*-algebras
2.9.1.
PROPOSITION. Let A be a C*-algebra, f a positive form on A, M its kernel and N the set of x E A such that f(x*x) = O. Then M ~ N + N*, and if f is pure M = N + N*.

Since If(x)1 2 :os:: Ilfllf(x*x), N ~ M, hence N* ~ M* = M and so N + N* ~ M. Now suppose f is pure, and let A be the C*-algebra obtained by adjoining an identity to A, and! the canonical extension of f to A Let 7T and ~ be the irreducible representation and vector defined by f. Let b EM and let 11 be its canonical image in H",. By the construction of 7T and ~, (11 I~) = !( 1* b) = 0, and hence there is an hermitian operator in H", which maps { to zero and leaves 1/ fixed. By 2.8.3, there is an hermitian element a of A such that 7T(a){ = 0 and 7T(a)1/ = 11, i.e, such that a E Nand b - ab E N. Thus b* and so
2.9.2.
M=M*~N*+N.

= (b - ab)* + b*a E N* + N,

LEMMA. Let A be a unital C*-algebra, L a closed left ideal of A and x E A ". If, for every E > 0, there is a positive element u. of L such that x :os:: u. + E, then x E L.

If t

u;/2 then t. is a limit of polynomials in u, with no constant term

CH. 2, 9l

IDEALS IN C*-ALGEBRAS

55

and therefore belongs to L. Now

IIx I/2(t. + Ye)-I t

- x l /2W= Ilx1/2Ye(t. + -.;e)-111 2 = ell(t. + Ye)*-I(xI/ 2)*x l/2(t. + Yerlll

= ell(t. + Ye)-Ix(t. + Ye)-Ill

and since 0 ~ x ~

t; + e, we have (1.6.8):
(t. + Ye)-l(t; + e)(t. + Ye)-I

o~ (t. + Yer l x (r, + Y e)-I ~ =(t;+ e)(t;+2Yet. + ef ~

so that

I,

!lx1/2(t. + Yerlt. - xl/ 21 f ~ e. I 2 Therefore x1/2(t. + Ye)-I t E L, hence X / E L as L is closed and so xEL.

2.9.3. LEMMA. Let A be a C*-algebra and L a closed left ideal of A. Then A is the closed left ideal generated by L n A+. This follows immediately from 1.7.3. 2.9.4. LEMMA. Let A be a C*-algebra and L, L two closed left ideals of A such that L C L. Suppose every positive form on A that vanishes on L also vanishes on L. Then L =L. We may assume A to be unital. Let a E L n A + and let e be >0. The set S. of positive forms f on A such that f(l) = !lfli = 1 and f(a) ~ e is weak*-compact. If f E S., f is not identically zero on L and therefore nor on L, so that there exists an xf E L such that f(xf) 0; consequently there is a weak*-neighbourhood U, of f in S. such that g(xf) 0 if g E Uf , and, by the compactness of S., there are a finite open covering (U;),.. j<;;m of S. and elements a l , ,am of L such that
0<lf(a;)12~f(aia;)

for fE U;.

Hence f(atal + ... + a:a m ) > 0 for every f E S., and multiplying the a, by sufficiently large scalars, we have f(atal + ... + a:a m ) ~ f(a) for every f E Se- Thus f(atal + ... + a:a m + E - a) ~ 0 for every positive form f, from which it follows that a ~ atal + ... + a:a m + e (2.6.2). Hence a E L (2.9.2). Land L thus contain the same positive elements and are therefore equal (2.9.3).

56

POSITIVE FORMS AND REPRESENTATIONS

[CH. 2, 9

2.9.S.THEOREM. Let A be a C*-algebra. (i) Let f be a positive form on A and N, the left ideal of those x E A such that f(x*x) = O. Then N, is maximal regular if and only if f is pure. (ii) The map f ~ N, is a bijection from the set of pure states onto the
set of maximal regular left ideals. (iii) Every closed left ideaL of A is the intersection of the maximal regular left ideaLs containing it.

(a) Let f be a positive form on A. By 2.8.4 and 2.8.5, the canonical representation 1T of A in AIN, is non-trivial and algebraically ir reducible. There exists ~ E H = AI N, such that N, is the set of x E A for which 1T'(x)~ = 0 and thus N, is a maximal regular left ideal of A. (b) Let L be a closed left ideal of A, and let 5 be the set of positive forms on A of norm ~1 which vanish on L. If f E 5, then u,:: L, because for each x E L, x*x ELand so f(x*x) = O. By 2.9.4 we have L = feS N f Moreover 5 is convex and weak*-compact and is therefore the weak*-closed convex hull of its set 5 of extreme points. It follows that if x E A satisfies f(x* x) = 0 for every f E 5, then also f(x* x) = 0 for each f E 5, so that teS N, = teS N; Finally, if f E 5 can be written in the form fl + f2, for positive forms t.. f2, we have, for each x E L,
7T

o~Ux*x)

+ fz(x*x) = f(x*x) = 0,

and so !J(x*x) = fz(x*x) = 0, hence Ux) = fz(x) = 0 and hence fl E 5 and f2 E 5. This shows that the non-zero elements of 5 are pure forms, and by virtue of (a), the equality L = fes N, proves (iii). (c) We have just seen that if L is maximal regular, and thus closed (B 1), then L is of the form N, with f pure. Hence the map described in (ii) is surjective. If f and / are pure states such that N, = N f , proposition 2.9.1 shows that f and / have the same kernel, so that f = A/ for some A ~ O. Since Ilfll = 11/11 = 1, we have f = I, which proves (ii). (d) Finally, let f be a positive form such that N, is maximal regular. Then N, = N, for some pure /, and the kernel of /, which is N f + N";, = N, + N";, is contained in the kernel of f (2.9.1). Hence f and f are proportional and so f is pure. This, together with (a), proves (i).

2.9.6. Recall that two representations 1T1,1T2 of an algebra in vector spaces E I , E 2 are said to be algebraically equivalent if there is an isomorphism of E I onto E 2 mapping 1T1 to 1T2

CH. 2. 9j
COROLLARY.

IDEALS IN C*-ALGEBRAS

57

Let A be a C*-algebra.

(i) Every algebraically irreducible representation of the (non-in volutive) algebra A in a complex vector space is algebraically equivalent to a representation of the C*-algebra A (in a Hilbert space). (ii) Let 17, 17" be two irreducible representations of the C*-algebra A in

Hilbert spaces H, H. If 17 and 17" are algebraically equivalent, then 17" are equivalent in the sense of 2.2.1.

17 and

If 17 is a non-zero algebraically irreducible representation of A in a complex vector space, there is a maximal regular left ideal L of A such that 17 can be identified with the regular representation of A in AIL. There then exists (2.9.5) a pure positive form f on A such that L = N r and AIL can be given a Hilbert space structure such that 17 is a representation of the C*-algebra A in this Hilbert space. Hence (i). Now let 17, 17" be two non-zero irreducible representations of the C*-algebra A in the Hilbert spaces H, H. Let {, f be unit vectors in H, H respectively and L, L be the sets of x E A such that 17(x){= 0 and 17"(x)f = 0 respectively. Put

f(x)

= (17(x){Ig),

!,(x) = (17"(x)f If).

Then Land L are the sets of x E A such that f(x*x) = 0 and f(x*x) = 0 respectively. If 17 and 17" are algebraically equivalent, { and f can be chosen so that L = L. Then f = f (2.9.5), so that 17 = 17" (2.4.1) and (ii) ollows, It thus makes sense to speak henceforth of classes of irreducible representations of a C*-algebra, without specifying whether the equivalence is purely algebraic or unitary. 2.9.7. Recall that a two-sided ideal of an algebra A is said to be primitive if it is the kernel of a non-zero algebraically irreducible representation of A in a vector space. We denote by Prim(A) the set of all primitive two-sided ideals of A.
Let A be a C*-algebra. (i) The primitive two-sided ideals of A are just the kernels of the non-zero topologically irreducible representations of A in a Hilbert space. (ii) Every closed two-sided ideal of A is the intersection of the primitive two-sided ideals containing it.
THEOREM.

(i) follows from 2.8.4 and 2.9.6 (i).

58

POSITIVE FORMS AND REPRESENTATIONS

[CH. 2, 1O

Let I be a closed two-sided ideal of A, so that AI I is a C*-algebra (1.8.2). There therefore exists a family (1TJ of non-zero irreducible representations of A, vanishing on I, such that the intersection of the kernels, Ker 1Tj, is exactly I (2.7.3). Moreover, for each i, Ker 1Tj is primitive. Theorem 2.9.7 defines a canonical mapping
A~Prim(A)

which is surjective. We shall encounter (4.3.7) a great variety of cases where this map is in fact bijective. References: [604], [849], [893], [1323], [1456]. 2.10. Extension of representations of C*-algebras 2.10.1. LEMMA. Let A be a unital C*-algebra, and B a self-adjoint subspace of A such that 1 E B. Let F be the set of linear forms g on B such that g(x*) = g(x) for x E B, g(x);;. 0 for x E B n A + and g(1) = 1. Then (i) Every element of F can be extended to a state of A. (ii) Every extreme point of F can be extended to a pure state of A. (iii) Let g be an extreme point of F. If there is exactly one pure state f of A extending g, then f is the only state of A which extends g, and, lor every hermitian element x E A, we have
f(x) =

sup
y=y"'EB,y~x

g(y).

Let A h , B h be the sets of hermitian elements in A, B respectively. Let g E F and let g be its restriction to B h which is real-valued and ;;.0 on B n A+. The element I of B; is an interior point of the cone A+, and so (B 6) g can be extended to a real linear form f on A h which is ;;.0 on A +. Then f extends to a hermitian linear form I on A, and I is a state of A. The forms g and fiB coincide on Bi; and therefore on B; + iB h = B. This proves (i). N ow let g be an extreme point of F, and let K be the set of states of A that extend g. K is non-empty by (i), and is clearly convex and weak*-compact. We can therefore take f to be an extreme point of K and we shall see that f is then a pure state. Indeed, suppose fl f2 are states of A such that f =!(fl+ f2); then g =!fl IB) + (/21 B and as g is extreme in F, we have II IB = f2/ B = g. Hence II, f2 E K and as I is extreme in K, f = fl = f2 This proves (ii).

CH. 2. 10l

EXTENSION OF REPRESENTATIONS

OF C*-ALGEBRAS

59

N ow let g be an extreme point of F and suppose there is exactly one pure state f of A which extends g. By the above argument, f is the only extreme point of K and so K = {fl. Let x E A h , x~ B and put
a =

sup
yEBh.y",x

g(y).

There is exactly one linear form g, on B + Cx extending g such that g.(x) = a. Clearly, g)(y*) = g.(y) for y E B + Cx, and if y + Ax ~ 0 (y E B h , A E R) then gl(y + Ax) ~ 0: this is obvious if A =0; if A > 0, we have -A-)Y~x, so that g(-A-ty)~a, and Aa+g(y)~O, while if A<O, -A-y~x, and so g(-A-ly)~a and again Aa+g(y)~O. By the above, gt extends to a state of A which must coincide with f. Thus f(x) =
gt(x)

= a.

2.10.2. PROPOSITION. Let A be a C*-algebra, B a sub-C*-algebra of A and p a representation of B in a Hilbert space K. Then there is a Hilbert space H containing K as a subspace such that the norm of H induces the original norm on K, and a representation 7T of A in H such that p(x) = 7T(X) IK for each x E B. If p is irreducible, 7T can be chosen to be irreducible. Let A be the C*-algebra obtained by adjoining an identity to A. Let jj be the sub-C*-algebra B + C . 1 of A. The representation p has a unique extension to a representation p of jj such that p(l) = 1, and p is irreducible if and only if p is irreducible. We can thus confine attention to the case where A is unital and I E B. Furthermore, we can assume by 2.2.6 and 2.2.7 that p admits a cyclic vector f If we put g(x) = (p(x)g Ig) for x E B, then g is a state of R which is pure if p is irreducible. Let f be a state of A extending g, pure if p is irreducible (2.10.1). Let 7T = 7Tt, 1/ = gt and let H o =7T(B)1/ C Hi; For each x E B, let 7Tr(X) be the restriction of 7T(X) to H o Then 7T is a representation of B in H o, 1/ is a cyclic vector for 7T and for each x E B, we have
(7T(X)1/ 11/) = f(x)

= g(x) = (p(x)g I g).

There thus exists an isomorphism from K onto H o which maps g to 1/ and p to 7T (2.4.1), and so we may identify K with H o and p with 7T. Finally, if p is irreducible. f is a pure state by construction, and so 7T is irreducible. 2.10.3. LEMMA. Let A be a C*-algebra, I a closed two-sided ideal of A, 7T a representation of A in H, and H a closed (vector) subspace of H invariant under 7T(I). For each x E I, let p(x) = 7T(X) I H, and suppose

60

POSITIVE FORMS AND REPRESENTATIONS

[CH. 2, II

that the representation p of I in H is non-degenerate. Then (i) H is invariant under 1T(A), and if 1T denotes the subrepresentation of 1T defined by H, then (ii) 1T(1) is strongly dense in 1T(A).

Let (uJ be an approximate identity in the C*-algebra I. Since p is non-degenerate, p(u A ) tends strongly to 1 (2.2.10). If x E A, xU A E I, hence for each ~ E H,
1T(X)P(UA)~

= 1T(X)1T(UA)~

= 1T(XUA)~

= P(XUA)~

E H,

so that, in the limit, 1T(X)~ E H. This proves (i). Moreover, we see that 1T(X) is the strong limit of 1T(XUJ from which (ii) follows. 2.10.4. PROPOSITION. Let A be a C*-algebra, I a closed two-sided ideal of A and p a non-degenerate representation of I in H. Then (i) There is exactly one representation 1T of A in H which extends p. (ii) p (1) is strongly dense in 1T(A). By (2.10.2) there is a Hilbert space H ~ H and a representation 1T! of A in H such that p(x) = 1T!(X) IH for each x E I. By 2.10.3, H is invariant under 1T!(A), and taking for 1T the subrepresentation of 1T! defined by H, we have the existence statement of (i) while (ii) follows from 2.1O.3(ii). Finally let v be another representation of A in H which extends p, and let (uJ be an approximate identity of I. For each x E A, p(xu A) tends strongly to 1T(X) and to v(x), hence 1T(X) = v(x). References: [590], [630], [1101], [1455].

2.11. Passage to an ideal and to a quotient algebra Let A be a C*-algebra. We shall see that any closed two-sided ideal of A leads to a decomposition of the sets P (A), A and Prim(A).
2.11.1. LEMMA. Let A be a C*-algebra, I a closed two-sided ideal of A, 1T a representation of A in a Hilbert space H, K the essential subspace of 1T II and B the strong closure of 1T(A). Then (i) P K E B n B. (ii) The subrepresentation of 1T defined by H K vanishes on I. (iii) If 1T is the subrepresentation of 1T defined by K, then 1T(1) is strongly dense in 1T(A).

CH. 2. 11]

PASSAGE TO AN IDEAL AND TO A QUOTIENT ALGEBRA

61

The operator P K is in the strong closure of 1T(l), so that P K E B. By 2.10.3, K is invariant under 1T(A), and so P K E B. Statement (ii) is obvious and (iii) follows from 2.10.3. 2.11.2. PROPOSITION. Let A be a C*-algebra, I a closed two-sided ideal of A and AI and Al the sets of 1T E A such that 1T(I) = 0 and 1T(l) " 0 respectively. Then (i) for each 1T E A, let 1T be the quotient representation of1T of AI1. Then 1T .....;; 1T" is a bijection from Al onto (AI I) . (ii) The map 1T"";; 1T II is a bijection from A I onto 1. Statement (i) is clear, and we pass to (ii). Let 1T E AI. In the notation of 2.11.1, we have K" {O} and so K = H as 1T is irreducible. 2.11.1(iii) then shows that 1T II is irreducible (d. also 2.11.3), and the map 1T - 1T II of AI into i is bijective by 2.10.4. 2.11.3. LEMMA. Let A be an algebra over a (commutative) field, and 1T an irreducible representation of A in a vector space E. Then (i) If I is a two-sided ideal of A and 1T(l) " 0, then 1T II is irreducible. (ii) If II> 12 are two-sided ideals of A with 1T(lI) " 0 and 1T(l2) " 0, then 1T(I1 . 10" O. The set of elements of E on which 1T(I) vanishes is invariant under 1T(A) and is not the whole of E, and therefore consists of the zero vector alone. Hence if ~ is any non-zero element of E, then 1T'(lK~ 0, and since 1T'(I)~ is invariant under 1T(A), we have 1T(I)~ = E, which proves (i). Moreover, the preceding work shows that 1T(l0E =E and 1T(II) 1T(l0E = E, hence 1T(l1 . 12) " O. 2.11.4. LEMMA. Let II 12 be two-sided ideals of the algebra A and I a primitive ideal of A. If I ~ II . 12 (in particular if I ~ II n 12), then either I ~ II or I ~ 12, Suppose ItJ II and ItJ 12, By 2.11.3(ii) applied to an irreducible re presentation 1T of A with kernel I, we have 1T(II 10 " 0 and hence ItJ 111 2, 2.11.5. PROPOSITION. Let A be a C*-algebra, I a closed two-sided ideal of A and PrimI(A) and Prim(A) the sets ofprimitive two-sided ideals of A containing I and not containing I respectively. Then (i) The map I - II I is a bijection of PrimI(A) onto Prim(AIl). (ii) The map I - I n I is a bijection of Prim(A) onto Prim(l).

62

POSITIVE FORMS AND REPRESENTATIONS

[CH. 2, ll

Statement (i) is obvious, so we prove (ii). If 1 E Prim! (A), then 1 is the kernel of an irreducible representation 11" of A such that 11"(1) : 0, hence 11" It e i (2.11.2) and J n I = Ker(11" II) E Prim(1). If J E Prim(1), there exists 11" E i such that J = Ker 11" and then there is 11" E A such that 11" = 11" II (2.10.4), from which it follows that J = (Ker 11") n I and the map described in (ii) is surjective. Let 1 1,12 E Prim! (A) and suppose that 1 I n I = 12 n I; then h:2 1 1n I and 1 212 I, so that 12 :2J I (2.11.4) and similarly 1 1:2 12, Thus J I = J 2 and the map given in (ii) is injective. 2.11.6. Before obtaining a result analogous to propositions 2.11.2 and 2.11.5 for pure states, we establish several properties of general positive forms. Let A be a C*-algebra, I a closed two-sided ideal of A and w: A ~ AI I the canonical morphism. Since w(A +) = (AI It, the map k ~ k w is a bijection of the set of positive forms on AII onto the set of positive forms on A that vanish on 1. It follows from general properties of Banach spaces, that this bijection is norm-preserving and is bicontinuous for the weak*-topologies.
0

2.11.7. PROPOSITION. Let A be a C*-algebra, I a closed two-sided ideal of A and f a positive form on A. Then (i) There exists a unique decomposition f = fl + f2 where t.. f2 are positive forms on A such that Ilflll= IlflIIII and fiI) = O. (ii) The pair (11"f, ~f) may be identified with ( 7Th EB 7Th ~h + ~h); 7Th vanishes on I and 7Th II is non-degenerate. the essential subspace of 7T II, K 2 = H", KI> 11"1 and 7T2 the subrepresentations of 7T defined by K I and K 2 [2.11.1(i)]. Then ~j is a cyclic vector for 11"jo and so if Ii = W~i 0 11"j, (7Tj,~) may be identified with (7Tfi ~f) (2.4.6). 11"2(1) = 0, and so fiI) = O. The representations 11"1 and 11"1 II are non-degenerate, and so Put 7Tf
~I

= PKI~f'

= 7T. Let K I be ~2 = PKi!, and

by 2.4.3. The relation ~f

= ~I + ~2

implies f

= fl + f2 In

particular, f2 = 0 if

It remains to prove the uniqueness statement in (i), so let f;, f~ be positive forms on A such that f = J; + f~, Ilf:ll= IIf; IIII and f~(1) = O. The above argument, applied to instead of f, shows that 7T/j II is non degenerate; thus if (u A ) is an approximate identity of I we have J;(x) =

Ilill=Ilf III/.

CH. 2. t1j

PASSAGE TO AN IDEAL AND TO A QUOTIENT ALGEBRA

63

lim f;(xu A ) for every x E A and similarly f.(x) = lim f.(xuJ. /1 and t; therefore coincide on I since f2(1) =f2(l) = 0 and thus f; = I, and /2 = f2 2.11.8. PROPOSITION. Let A be a C*-algebra, I a closed two-sided ideal 0/ A, and PI (A) and P I(A) the sets of pure states of A which vanish on I and which do not vanish on I respectively. Then (i) If, for each / E PI(A), f denotes the positive [orm on AI I obtained from / by passing to the quotient, then f ~ I is a bijection of PI (A) onto p(AII). (ii) The map f ~ /11 is a bijection of P I(A) onto P (I). Let f E PI(A), and let w denote the canonical morphism of A onto AI I. If I dominates a positive form g on AI I, then f dominates g =g 0 wand so g =Af with 0.;:;A .;:; 1, s = AI from which it follows that I E P (AI I). It is clear that the map f ~ I is injective, so we have only to show that it is surjective. Thus, let h E P (All) and f = how. If f dominates a positive form g, then g vanishes on t: and therefore on I, and so g can be written k w where k is a positive form dominated by h. Hence k=Ah, with O.;:;A;:;I, so that g=Af and fEPI(A). Since h =1, we have indeed shown that the map f ~ I is surjective. Now let fE pI(A). In the decomposition f= /1 + f2 of 2.11.7, I, and f2 are proportional to f. Now f2 vanishes on I and f does not vanish on I so that f2 = 0 and f =t.. Hence IlfIIII = Iitil = I, from which it follows that f II is a state of I. By 2.10.3 and 2.11.7(ii), 7Tfl I is irreducible and so f II E P (I). Let f, I E P I(A); if f II =11 I, there is an isomorphism of H"I onto H"f which maps ~flI (i.e. ~f) into ~flI (i.e. ~f) and 7TflI into 7TrII and therefore 7Tf into 7Tr (2.10.4). Thus f = f, and the map f ~ f II of (ii) is injective. Finally, if g E P (I), the irreducible representation 7Tg of I can be extended to an irreducible representation 7T of A (2.10.4). Since II~gll = 1, f = we. 0 7T is a pure state of A extending g, and thus f E P I (A) with f II = g, which shows that the map f ~ / II of (ii) is surjective.
0

2.11.9. The six bijections of propositions 2.11.2, 2.11.5 and 2.11.8 are said to be canonical. To sum up, we have the following diagram of canonical maps:

P(AIl) (All)

P(A)

++-

P(I)
j

i i i
~

i i i
Prim(AII) ~ Prim(A)
+- Prim(l)

64

POSITIVE FORMS AND REPRESENTATIONS

[CR. 2, 12

The vertical arrows represent surjective maps, and on each line, the two horizontal arrows represent injections of complementary subsets into the middle set. It is easily checked that this diagram is commutative. References: [508], [819], [896], [1456]. 2.12. Addenda 2.12.1. Let A be a unital C*-algebra and III - gil < 2. Then 7Tt = 7Tg- [633].

I,

g pure states of A such that

2.12.2. Let A be a unital C*-algebra, E(A) the set of states of A, and IE E(A). Then the following conditions are equivalent: (i) for every neighbourhood U of I in E(A), there is a I) > 0 and a E A with 00;.;; a 0;.;; 1, I(a) = 1 and g(a)< I-I) for g EE(A)- U; (ii) for every closed G, set S of E(A) containing I, there is an a E A such that lIall = I/(a)1 and S contains the set of g E E(A) for which

[c] = Ig(a)l;
(iii)

is a pure state. [631].

2.12.3. Let A be a C*-algebra and K; the set of states I of A such that dim 7Tt =n + 00). Then P (A) n K; is an open subset of K n [638]. 2.12.4. Let A be a von Neumann algebra, ~ its centre and a the spectrum of ~. For each (J) E a, let I., be the norm-closed two-sided ideal of A generated by Ker (J), and for each TEA, let T., be the canonical image of T in Al l.; Then (a) If I is a pure state of A, then I I~ is an element (J) of a and I vanishes on I.,. (b) For every TEA, the function (J) ~ II T.,II is continuous on a. (c) In AII." the product of two non-zero two-sided ideals is non-zero. [630]. 2.12.5. Let A be an involutive unital Banach algebra. Suppose that for every continuous hermitian linear form I on A, there exist two positive forms g, h such that 1= g - hand II.f1I =Ilgll + Ilhll. Then A is a C* algebra. [681]. *2.12.6. Let AI A 2 be unital C*-algebras, and P(A;) the weak*-c1osure of P(A j ) in the dual of A j Let 5j be the set of functions I~ I(x) on P(A j ) (x E A;). If there is a homeomorphism of P(A I ) onto P(AJ which maps 2 1 into 2 2, then there is a bijective linear isometry p: AI ~ A 2 such

CH. 2, 12]

ADDENDA

65

that p(x*) = p(x)* for each x E Al and p(x n ) in AI (d. 1.9.10). [839].

= p(xt for

each hermitian x

2.12.7. Let A be a unital C*-algebra, P(A) the weak*-closure of P(A) in the dual of A and .2 the set of functions f ~ f(x) on P (A) (x E A). If the product of two elements of .2 is again in .2, A is commutative. [839]. 2.12.8. Let A be a unital C*-algebra, f a state of A and N the set of x E A such that f(x*x) = O. If Ker f = N + N*, then f is pure. [849]. *2.12.9. For a unital C*-algebra A to be isomorphic to a von Neumann algebra, it is necessary and sufficient that it satisfies the following conditions: (i) every increasing family of hermitian elements of A that is bounded above has a least upper bound; (ii) for every non-zero x E A +, there is a state f of A with f(x) # 0, such that for every increasing family (y;) in A + with least upper bound yEA +, f(y) = sup f(y;). [849], [1590], [1591]. *2.12.10. Let A be a unital C*-algebra and X the compact space P(A). For each x E A, let cPx be the function f ~ f(x) on P (A) and let .2 be the set of all the cPxs for x E A. Each state g of A goes over, via x ~ cPx to a positive linear form on .2, and this form can be extended, in infinitely many ways in general, to a positive measure on X; the collection of those subsets of X which are of measure zero for all these measures will be denoted by N g Now let 7T be a representation of A, and denote by N" the intersection of the N h o" as h varies through the set of normal states in the weak* closure of 7T(A). Let H be a separable Hilbert space, A a sub-C*-algebra of .2(H) containing 1, B the weak closure of A and 7T a representation of A in a separable Hilbert space. Then 7T can be extended to an uItraweakly continuous representation of B in H" if and only if N,,:2 N" where, denotes the identical representation of A. [848]. 2.12.11. Let A be a C*-algebra and G a topological group. For each g E G, let {g be an automorphism of A. Suppose that g ~ {g is a representation of G and that {ix) is a continuous function of g for each x E A. A state f of A is said to be stationary for { if f({ix = f(x) for each x E A and g E G. Suppose f satisfies this. Then there exists a unique continuous unitary representation p of G in H", such that: (1)

66

POSITIVE FORMS AND REPRESENTATIONS

[CH. 2. 12

7T/(g(X = p(g)7T/(X)p(g-I) for each xEA and gEG; (2) if." denotes the canonical map of A into H"f then "(g(x = p(g).,,(x) for each x E A and g E G. [1460].

2.12.12. Let A be a unital C*-algebra, x a normal element of A and A E Sp x. Then there is an irreducible representation of A and a non-zero gin H" such that 7T(X)g = Ag, (Apply 2.10.2 to the sub-C*-algebra B of A generated by I and x and to the character of B which takes the value A at x.) [1455]. 2.12.13. Let A be a non-unital C*-algebra. Then 0 is in the weak" closure of P (A). [Let XI .. , x n E A + and > 0; we have to construct an IE P (A) such that l(x I ) , , I(x n ) :,,;; ; replacing XI . , X n by XI +... + x, we can assume n = 1 and using 2.5.5, it is enough to construct a state g of A such that g(x l ) :,,;; e ; realizing A in a Hilbert space H by means of a non-degenerate representation, XI is not invertible in A + C . 1 and therefore not invertible in ::t(H); hence there is a unit vector g of H such that (xjg Ig):,,;; e: take g = we] This was communicated to me by J. Glimm. 2.12.14. Let A be a C*-algebra, I and J closed two-sided ideals of A and h a state of A that vanishes on I n J. Then h = AI + ug where A ~ O. /L ;;:.0, A + /L = 1 and where I, g are states of A which vanish on I and J respectively. (It is enough to consider the case where In J = O. Then 7Th II and 7Th IJ have orthogonal essential subspaces. Apply 2.11.7.) [1456]. 2.12.15. Let A and B be C*-algebras, 7T, 7T injective representations of A and o, p injective representations of B. Let D and D be the respectively C*-algebras of operators on H" 0 n, and H", 0 generated by the 7T(X) 0 p(y) and 7T(X) 0 p(y) respectively where X E A, y E B. There is a unique isomorphism of D onto D which maps 7T(X) 0 p(y) into 7T(X) 0 p(y) for any x E A, y E B. Thus D, regarded as an abstract C*-algebra, depends only on A and B and is called the C*-tensor product of A and B. [692], [701], [1613], [1614], [1620], [1728], [1729], [1731], [1847].

n,

*2.12.16. (a) Let A and B be C*-algebras and /L a linear map from A into B. /L is said to be positive if /L(A+)C B+. Now let A(n) be the C*-algebra of (n x n) matrices with entries in A. Applying /L to each element of such a matrix we obtain a map /L(n) from A (n) into e, and /L

CH. 2. 12j

ADDENDA

67

is said to be completely positive if p.. (n) is positive for each n. In general, complete positivity is a more restrictive condition than positivity, al though the two notions are equivalent if A is commutative. (b) Let A be a unital C*-algebra, H a Hilbert space and p.. a linear map from A into 5E(H) such that p..(1) = 1. Then there is a Hilbert space K containing H as a subspace and a morphism p from A into 5E(K) such that p..(x) = PHP(x) IH for each x E A if and only if u. is completely positive. [1506]. 2.12.17. Let A be a C*-algebra, A the Banach dual of A and A h and A~ the hermitian parts of A and A, which are partially ordered vector spaces. Then the following conditions are equivalent: (1) A h is a lattice; (2) A~ is a lattice; (3) A is commutative. [606]. 2.12.18. Let A be a C*-algebra, f a pure state of A and N, the set of x E A such that f(x*x) = 0, so that AINf endowed with the scalar product derived from f(y*x) is a Hilbert space. Then the norm on this space is equally the quotient norm of the Banach space AIN f [1612]. *2.12.19. Let A be a unital C*-algebra and f a positive form on A. (a) Let E f be the set of linear forms on A of the type x ~ f(xoXx~) where X o E A. Let Ff be the closure (in the norm topology) of E, in the dual A of A. Then if g is a positive form on A, the following conditions are equivalent: (i) g E F f; (ii) 1Tg os;; 1Tf; (iii) there is a ~ in the space of 1Tf such that g(x) = (1T'f(x)~ I0 for each x E A. (b) If f is pure, then F, = Ef . (c) Let Ei be the set of positive forms on A dominated by a multiple AI of f (where A ~ 0). Let Fi be the closure (in the norm topology) of Ei in A. Then F, is the set of forms
x~(1T'f(x)TJ

ITJ)

on A,

where TJ E 1T'f(A)'~f

[852].

2.12.20. Let A be an involutive Banach algebra with an approximate identity and f a continuous positive form on A. Let L 2(f) denote the set of g E A such that sup /g(x)lf(x*X)-1/2 < + 00.
xEA,f(xx);<O

Let H, be the space of 1Tf and TJf the canonical map of A into H f Then taking the transpose (f of TJf we have a map of H, into A such that (f(g), x) = (TJf(x) Ig) for any g E H, and x E A. The map (f is a conjugate linear bijection of H, onto L 2(f) which maps ~f into f. For each x E A, let

68

POSITIVE FORMS AND REPRESENTATIONS

[CH. 2, 12

7T(X) be the transpose of left-multiplication by x in A, which is a continuous linear operator on A. Then ~f maps 7Tf(x) into 7T(X*) IL 2(f) . Finally, (f is isometric when L 2(f) is endowed with the norm IIgIIL2(f)

sup
xE:A, f(xx);<,O

Ig(x )If(x*x

l/2

2.12.21. If A is a C*-algebra in which the only nilpotent element is 0, then A is commutative. [Let x, yEA with x = x* and f, g be continuous functions on Sp x with fg = 0; then (f(x)yg(X))2 = 0, hence f(x)yg(x) = 0 and so f(7T(X)) Tg(7T(X)) =0 for every irreducible representation 7T of A and every T E .2(H7T ) , therefore 7T(A) is scalar.] (I. Kaplansky, un published; communicated to me by R. V. Kadison.)

CHAPTER 3

THE SPECTRUM

OF A C*ALGEBRA

From now on, our two principle goals will be the following: (1) the decomposition of a representation into irreducible representations; (2) the structure of C*-algebras. For both problems, it is important to generalize to arbitrary C*-algebras the spectrum of a commutative C*-algebra. We are already acquainted with three candidates: the sets P (A), A and Prim(A). For commutative A, all these sets reduce to the set of characters of A: this is obvious for A, we have observed it for P (A) (2.5.2), and for Prim(A) it follows from the fact that the characters of a commutative C*-algebra are in one-to-one correspondence with their kernels. Only P (A) yet has a topology, namely the weak*-topology (the norm topology on the dual of A is hardly ever of any use). Unfortunately, this space is too large (it tends not to be finite-dimen sional if A is the C*-algebra of a Lie group) and is not locally compact. We shall define certain topologies on A and on Prini(A) in this chapter. -In "good" cases, these two spaces are homeomorphic. [In bad cases, Prim(A) can reduce to a single point while A is highly complicated, thus reflecting the complexity of A; it is for this reason, however uncon vincing it may seem, that preference is given to A over Prim(A).) These spaces are almost locally compact. Although they are not Hausdorff in general, we shall see that this is of only limited inconvenience. Moreover, in particular examples, this non-Hausdorff property ac curately reflects certain special features of the structure of the algebras being studied.

3.1. The Jacobson topology

3.1.1. Let A be an algebra over a (commutative) field. For each subset T of Prim(A), let I(T) be the intersection of the elements of T, which is a two-sided ideal of A. Let f denote the set of all primitive ideals of A containing I(T).

70 LEMMA.

THE SPECTRUM OF A C*-ALGEBRA

[CH. 3, l

(i) (iii)

0 = 0;

(ii) T

= T for T ~ Prim(A); (iv) T( U T 2 = T( U T 2 for T I , T 2 c Prim(A).

c T for T c Prim(A);

(i) and (ii) are obvious. Clearly I(T) = I(T) so that t = T. If I( = I(T() and 12 = I(T2) then I(T I U T 2) = I( n 12, and so T 1 U T 2 is the set of primitive ideals of A containing 1( n 12 , i.e. (2.11.4) the set of primitive ideals of A which contain either II or 12 , (iv) now follows. It follows that there is a unique topology on Prim(A) such that for each T ~ Prim(A), T is the closure of T in this topology. DEFINITION. This Prim(A). topology is called the Jacobson topology on

3.1.2. PROPOSITION. Let A be an algebra and let T be a subset of Prim(A). Then, T is closed if and only if T is exactly the set of primitive ideals containing some (fixed) subset of A.

If T is closed, T = T and so T is the set of primitive ideals of A containing I(T). Now let M ~ A. If T is the set of primitive ideals of A containing M, then I(T)::2 M so that T ~ T and T = T.
3.1.3. Recall that a topological space is called a To-space if for any two (distinct) points of the space there is a neighbourhood of one of the points which does not contain the other. PROPOSITION. The space Prim(A) is a To-space. Let II> 12 be two distinct points of Prim(A) so that, say, II g 12 , Then the set of those IE Prim(A) which contain II is a closed subset T (3.1.2), such that I, E T and 12 ~ T. 3.1.4. PROPOSITION. Let IE Prim(A). Then {I} is closed in Prim(A) if and only if I is maximal among primitive ideals. Indeed, the closure of {I} is just the set of primitive ideals of A containing 1. 3.1.5. Let A be an algebra and A the set of classes of non-zero irreducible representations of A. The map 7T -+ Ker 7T is a canonical surjection of A onto Prim(A).

CH. 3. I]

THE JACOBSON TOPOLOGY

71

DEFINITION. The spectrum of A is the set A endowed with the inverse image of the Jacobson topology under the canonical map
A~Prim(A).

3.1.6. PROPOSITION. The following three conditions are equivalent: (i) A is a To-space. (ii) Two irreducible representations of A with the same kernel are equivalent. (iii) The canonical map A ~ Prim A is a homeomorphism. (ii) ~ (iii): obvious; (iii) ~ (i): follows from 3.1.3; (i) ~ (ii): let lT1. 7Tz be two representations in A with the same kernel; every open subset of A that contains lT1 also contains lTz, and so IT) = 7Tz if A is a To-space. 3.1.7. LEMMA. Suppose A is unital, and that I is a maximal two-sided ideal of A. Then I is a primitive ideal. Let M be a maximal left ideal containing I and let 7T be the canonical representation of A in AIM, which is irreducible. Since IA ~ I ~ M, the kernel I of 7T contains I and does not contain 1, so that I = I and I is primitive. 3.1.8. PROPOSITION. If A is unital, Prim (A) and A are quasi-compact (i.e. they satisfy the Borel-Lebesgue axiom without necessarily being
Hausdorff.t

It is enough to prove this for Prim(A), so let (T j ) be a family of closed subsets of Prim(A) with empty intersection. If ~ I(Tj ) - A, the ideal ~ I(Tj ) is contained in a maximal two-sided ideal I, and, by 3.1.7, I is primitive. Thus lET; for each i as each T; is closed and we have a contradiction. Hence ~ leT;) = A. and so there exist XI E I(T;,), . . . x, E I(T;) such that I = Xl + ... + X n This implies that l(T;,) +, .. + l(T;) = A from which it follows that Til n ... n T;o = 0.

3.1.9. Let A be an algebra and let q, be an automorphism of A. For every irreducible representation IT of A, 7T 0 q, -I is an irreducible representation of A whose kernel and class depend only on the kernel and class of IT. We thus have bijections Primro): Primt Aj-e PrimrA) and cf>: A ~ A which are clearly homeomorphisms. If q,l q,z are two automorphisms of A, then Prim(q"q,z) = Prim(q,l) Prim(q,z) and

(q,Iq,Z( = cf>1cf>Z

72

THE SPECTRUM OF A C*-ALGEBRA

[CH. 3. 2

Many authors use the term "dual" instead of "spectrum", but this leads to confusion with the dual of the vector space A or of the Banach space A. References: [819]. 3.2. The spectrum of an ideal and of a quotient algebra We restrict attention to the C*-algebra case, although the following proposition can be extended to the general case. 3.2.1. PROPOSITION. Let A be a C*-algebra and I a closed two-sided ideal of A. Then the canonical bijections (cf. 2.11.2 and 2.11.5)
Priml(A) ~ Prim(AI I), Al ~ (All) A,

Prim(A) ~ Prim(l),

AI ~ j

are homeomorphisms. The sets Prirn.tA), Al are closed subsets of Prim(A), A, while the sets Prim(A), Al are open subsets of Prim(A), A.
By 3.1.2, Prim.IA) is closed in Prim(A) and so Prim{A) is open; hence in A, Al is closed and AI is open. The fact that the bijection Priml(A)~Prim(AII) is a homeomorphism follows immediately from the definition of the Jacobson topology and hence the bijection Al ~ (All) is a homeomorphism. To show that the bijection Prim! (A) ~ Prim(l) is a homeomorphism, it is necessary to prove the following: if (lJ is any family of elements of Prim(A) and 1 is a single element of Prim(A), then 1 :?n1 n I :?n(l A ~1 A n I).
A

The implication ~ is clear. :? fA) n I and 112 I so equivalence is proved. The homeomorphism follows from

(n

Now suppose 1 n I:? (JA n 1). Then that 1:? s, (2.11.4). and the above fact that the bijection Al ~ j is a this at once.

3.2.2. Let A be a C*-algebra. If I is a closed two-sided ideal of A, we may identify (AI I) with a closed subset of A and j with the comple mentary open set, thanks to proposition 3.2.1.
A

PROPOSITION. Let A be a C*-algebra. Then I ~ j is a bijection from the set of closed two-sided ideals of A onto the set of open subsets of A. Moreover I. c 12 ~ II h I2

CR. 3, 2] THE SPECTRUM OF AN IDEAL AND OF A QUOTIENT ALGEBRA

73

Let U be an open subset of A and F =A\U. Since each primitive ideal of A is closed (2.9.7) the intersection of the kernels of the tr E F is a closed two-sided ideal I of A and since F is closed in A every irreducible representation of A whose kernel contains I is an element of F. Therefore 1 = A\F = U, and so the map mentioned in the proposition is surjective. Now let I( and 12 be two closed two-sided ideals of A. If 11 =12, the primitive ideals of A containing I( are the same as those which contain 12 , Now each closed two-sided ideal of A is the in tersection of the primitive ideals which contain it (2.9.7). Hence II = 12 and the map I ~ 1 is injective. The assertion that 11 ~ 12 :> 11 ~ 1 2 is obvious.
3.2.3. Recall that in a C*-algebra, the sum of two closed two-sided ideals is itself a closed two-sided ideal (1.8.4).
PROPOSITION. Let A be a C*-algebra and I and J closed two-sided ideals of A. Then (I + J) ~ = 1 u j and (I n J) = 1 n 1. In particular I n J = 0 if and only if 1 n j = 0.
A

This follows from proposition 3.2.2, since 1+ J and I n J are the least upper bound and greatest lower bound respectively of the set {I, J} in the partially ordered set of closed two-sided ideals of A. 3.2.4. Let A be a C*-algebra, and B the C*-algebra obtained by ad joining an identity to A. Then A is a closed two-sided ideal of codimen sion I in B, namely the kernel of the one-dimensional representation w of B defined by
w(A . 1 + x)

=A

(A E C, x E A).

The only primitive ideal of B containing A is A itself, and the only irreducible representation of B whose kernel is A is t, Hence {A} is closed in Prim(B) and {ei} is closed in B. By 3.2.1, the spaces Prim(A) and A may be identified with Prim(B)\{A} and B\{w}. Now suppose that A is unital with identity e. For each 7r E B\{w}, 7r(e) is a non-zero idempotent commuting with 7r(A), so that 7r(e) = 1 and hence 7r(e - 1) = O. Consequently, for each IE Prim(B)\{A}, we have e - I E I and then e - I $. A. Thus A is not in the closure of Prim(B)\{A} in Prim(B). In other words, {A} is not only closed but is also open in Prim(B) and therefore [or] is not only closed but is also open in B. Conversely, suppose that {w} is open in B. By 3.2.2 and 3.2.3, B is the

74

THE SPECTRUM OF A C-ALGEBRA

[CH. 3. 3

direct sum of two closed two-sided ideals, one of which is A, and since B possesses an identity, the same is true of A. References: [896].

3.3. Norm and topology

3.3.1. LEMMA. Let A be a C*-algebra, a an hermitian element of A and L a closed set of real numbers. Then (i) The set Z of those 7T E A for which Sp7T(a) ~ {O} U L is closed in

A.
(ii) If A is unital, the set Z of those 7T E closed in A.

A for

which Sp 7T(a) ~ L is

Let p E i, and suppose that Spp(a) contains a point o:~ {O} U L. Let f: R~ R be a continuous function vanishing on {O} U L with f(o:) of O. Then 7T(f(a = f(7T(a vanishes for 7T E Z and is non-zero at 7T = p, which contradicts the hypothesis that p E Z: Hence we have (i). (ii) is proved in a similar way. 3.3.2. PROPOSITION. Let A be a C*-algebra and let x E A. Then the function 7T ~ 117T(X)11 is lower semicontinuous on A. We may at once confine attention to the case of unital A. Moreover, since 117T(X)112 = 117T(X*x)ll, we can assume that x E A ". Let k ~ 0 and let Z be the set of 7T E A such that 117T(x)II ::5;; k. Then the set Z is equally the set of 7T E A. such that Sp 7T(xH: [0, k]. By 3.3.1, Z is closed and the proposition is proved. 3.3.3. LEMMA. Let A be a C*-algebra, (x.) a collection of elements which is dense in A, and Z, the set of 7T E A such that 117T(xj)ll> 1. Then the Z, constitute a base for the topology of A. Let U be an open subset of A and let 7T E U. Let I ~ A be the intersection of the kernels of the representations belonging to A\ U. There exists an x E I such that 117T(X)11 = 2 and we have p(x) = 0 for p E A\ U. Let i be an index such that IIx - Xiii < 1. Then 117T(XJII> 1 and IIp(xj)11 < 1 for p E A\ U, and so 7T E Z, k U. As Z, is open (3.3.2), the lemma is proved. 3.3.4. PROPOSITION. If A is separable, the topology of table base. This follows immediately from 3.3.3.

A.

has a coun

CH. 3. 3j

NORM AND TOPOLOGY

75

3.3.5. LEMMA. Let A be a unital involutive algebra, x an hermitian element of A and A E R. Then A E SpAx if and only if there is a primitive ideal I of A such that A E Sp AlIX, where x denotes the canonical image of x in AlI.

If x - A is invertible in A, i-A is invertible in AI I for every primitive ideal 1. Suppose x - A is not invertible. Then x - A is not left-invertible, because a left-invertible hermitian element is also right-invertible, and is therefore contained in a maximal left ideal L. The kernel I of the canonical representation of A in AlLis a primitive ideal of A, and the image of x - A in AII is not invertible.
3.3.6. LEMMA. Let A be a C*-algebra. Then for each x E A, the function at ~ IITr(x)11 on A attains its least upper bound "xii. Replacing x by x*x, we are reduced to the case where x E A+, and we may further assume that A is unital. Then [x] E Sp AX. Therefore, (3.3.5), there exists a 7T E A such that IIxil lies in the spectrum of the operator Tr(x), from which it follows that [x] 0;;;; IITr(x)II 0;;;; [x]. 3.3.7. PROPOSITION. Let A be a C*-algebra, x an element of A and a> O. Then the set Z of those 7T E A such that 117T(x)II ~ a is quasi compact. Let (Zi) be a decreasing filtering family of relatively closed non-empty subsets of Z, and for each i. let I, be the intersection of the kernels of the elements of Z; The canonical image of x modulo I, is of norm ~a. The I, constitute an increasing filtering family; let I be the closure of their union, which is a closed two-sided ideal of A. The definition of the norm in the quotient of a normed space shows that the canonical image of x modulo I is of norm ~a. By 3.3.6, there is a tt E A whose kernel contains I and is such that IITr(x)11 ~ a. We thus have 7T E Z and 7T is in Z, and the proposition is the closure of Z, for each i; hence 7T E proved.

3.3.8. Recall that a topological space is said to be locally quasi-compact if each point has a base of quasi-compact neighbourhoods.
COROLLARY.

The spectrum

A of a C*-algebra A is locally quasi

compact.
Let 7T E A, and let U be an open neighbourhood of 7T in A. Since A\U is closed, there is an x E A such that 7T(X) - 0 and p(x) = 0 for p E A\u.

76

THE SPECTRUM OF A C*-ALGEBRA

[CH. 3. 4

Let V and W be the sets of pEA such that

IIp(x)/1 > ~11'7T(x )11

and

IIp(x)11 ~ !117T(x )11 respectively.

By 3.3.2, V is an open neighbourhood of 7T. Hence W is a neigh bourhood of 7T contained in U and W is quasi-compact by 3.3.7. 3.3.9. COROLLARY. Let A be a C*-algebra such that A is Hausdorff. Then for each x E A, the function 7T ~ 1117"(x)11 is continuous on A. In this case, the set Z of 3.3.7 is closed, and so the function 7T ~ 1117"(x)11 is upper semi-continuous. Now apply 3.3.2. References: [582], [583], [584], [896], [1323].

3.4. Second definition of the topology on the spectrum

3.4.1. LEMMA. Let A be a unital C*-algebra, A h the set of hermitian elements of A, E(A) the set of states of A and Q a subset of E(A). Suppose that if x E A h satisfies f(x) ~ 0 for each f E Q, then x E A ", Then under these conditions: (i) the weakr-closure of Q contains P (A); (ii) the weak*-closed convex hull of Q is E(A). Let x E A h and a E R be such that f(x) ~ a for every f E Q. Then x - a . I E A + by hypothesis, hence f(x) ~ a for every f E E(A). This proves that the weak*-closed convex hull C of Q contains E(A). Conversely, it is clear that C C E(A). This proves (ii), and (ii) implies (B 14)that (j contains the set of extreme points of E(A), i.e. peA) (2.5.6). 3.4.2. PROPOSITION. Let A be a C*-algebra and (17"JjEI a family of representations of A in spaces Hi Then (i) Each state of A which vanishes on Ker tr, is a weak*-limit of states of the [orm W~I 0 7Tjl +... + w~. 0 17"j. where ii, ... , in E I, and g\ E Hj" , gn E Hi. (ii) Each pure state of A which vanishes on Ker 7Tj is a weak*-limit of states of the [orm W{ 0 17"i where i E I, g E H;

EB Hi Passing to the quotient by

By 2.2.6, we can assume that each 17"j is non-degenerate. Let H = Ker 17"j we can assume that

CH. 3. 4j

SECOND DEFINITION OF THE TOPOLOGY ON THE SPECTRUM

77

Ker TTj = O. The representation p = EB TTj is then injective, and we can identify A with the sub-C*-algebra p(A) of 2(H), so we can now say that the identical representation of A is non-degenerate. By 2.4.3, we may replace A by A + C . 1. Let Q be the set of states of A of the form r)~ TTj (i E I, ~ E H; II~II = 1). If x is an hermitian element of A such that r)~( TT;(x));;;, 0 for each i E I and each unit vector ~ of H;, then TTj(x);;; 0 for each i and so x;;;. O. All that is now necessary is to apply 3.4.1.
0

COROLLARY. Let A be a C*-algebra and f a state of A. The states of A that vanish on Ker TTf are weak*-limits of states of the form where Yl ... Yn E A. The pure states of A that vanish on x ~ ~ f(y~xy;) Ker TTf are weak*-limits of states of the form x ~ f(y*xy) where yEA.

3.4.3.

This follows from 3.4.2 and 2.4.8 (ii).

Let A be a C*-algebra, TT a representation of A and S a set of representations of A. Then the following conditions are equivalent: (i) The kernel of TT contains the intersection of the kernels of the elements of S. (ii) Every positive form on A associated with TT is a weak*-limit of linear combinations of positive forms associated with S. (iii) Every state of A associated with TT is a weak*-limit of states which are sums of positive forms associated with S.
3.4.4.
THEOREM.

(iii) ~ (ii): obvious. pES Ker p, each (ii) ~ (i): suppose condition (ii) is satisfied. If a E positive form associated with S vanishes at a*a, so that every positive form associated with TT vanishes at a*a and hence TT(a) =O. (i) ~ (iii): if Ker TT d pES Ker p, each state of A associated with TT vanishes on pES Ker p and we just apply 3.4.2 (i).

3.4.5. It is clear that if TT is contained in one of the elements of S, the conditions of theorem 3.4.4 are fulfilled. The converse fails even when S consists of a single element. We are thus led to the following definition:
DEFINITION. If TT and S satisfy the equivalent conditions (i), (ii), (iii) of theorem 3.4.4, TT is said to be weakly contained in S.

If Sand T are two sets of representations of A, T is said to be weakly contained in S if each element of T is weakly contained in S. Also, S

78

THE SPECTRUM OF A

C*-ALGEBRA

[CH. 3, 4

and T are said to be weakly equivalent if T is weakly contained in Sand S is weakly contained in T. 3.4.6. DEFINITION. Let A be a C*-algebra and 1T a representation of A. We call the set of pEA which are weakly contained in 1T the support of
1T.

Making use of condition (i) of 3.4.4, we see that this support S is a closed subset of A. Since Ker 1T is the intersection of the primitive ideals containing it, 1T is weakly contained in S. 3.4.7. Let A be a C*-algebra, 1T a representation of A, S a set of representations of A, A the C*-algebra obtained by adjoining an identity to A, iT the canonical extension of 1T to A and S the set of canonical extensions to A of the elements of S. If iT is weakly contained in S, it is plain that 1T is weakly contained in S [condition (i) of theorem 3.4.4]. The converse is true if 1T is non-degenerate, for let x E A, A E C be such that p(x) = A . 1 for each pES. Then p(xy - Ay) = 0 for each yEA, and each pES, so that 1T(XY - Ay) = 0 for each yEA, hence (1T(X) - A . l)1T(Y) = 0 for each yEA and hence 1T(X) = A . 1 since 7T is non-degenerate. 3.4.8. Let A be a C*-algebra. With each set S of representations of A is associated the set Q(S) of positive forms on A which are weak*-limits of sums of positive forms associated with S. Clearly, Q(S) is a weak* closed convex cone in the dual of A. Further, T is weakly contained in S if and only if Q(T) C Q(S).
1T

3.4.9. PROPOSITION. We go back to the notation of theorem 3.4.4. When has a cyclic ector t, the conditions of theorem 3.4.4 are equivalent to the following: (ii) The positive form x ~ f(x) = (7T(X)~ I~) on A is a weak*-limit of linear combinations of positive forms associated with S.
It is clear that (ii) implies (ii). We now suppose that condition (ii) holds, and we show that condition (ii) is satisfied. By 2.4.8 (ii), it is enough to show that for each yEA the form x ~ g(x) = f(y*xy) is a weak*-limit of linear combinations of positive forms associated with S. Now f is the weak*-limit of forms fj which are linear combinations of positive forms associated with S. Hence g is the weak*-limit of the forms x ~ gj(x) =/i(y*xy), and moreover every g is, by 2.4.8 (i), a linear combination of positive forms associated with S.

CH. 3, 4]

SECOND DEFINITION OF THE TOPOLOGY ON THE SPECTRUM

79

3.4.10. THEOREM. Let A be a C*-algebra, following conditions are equivalent:

(i)
7T

7T

A and SeA. Then the

S;

(ii) 7T is weakly contained in S; (iii) at least one of the non-zero positive fonns associated with 7T is a weak*-limit of positive fonns associated with S; (iv) every state associated with 7T is a weak*-limit of states associated with S. (i) > (ii): immediate by condition (i) of theorem 3.4.4; (iv)::} (iii): obvious; (iii) ::} (ii): follows from 3.4.9; (ii)::} (iv): follows from 3.4.2(ii). 3.4.11. THEOREM. For any C*-algebra A, the canonical map P (A) ~ is continuous and open.

Let S be a subset of A, Q its inverse image in P (A), f a point of P (A) and 7T its image in A. For 7T to belong to the closure of S, it is necessary and sufficient that f should belong to the closure of Q (3.4.10). Hence S is closed in A if and only if Q is closed in P (A). This shows that the canonical map P (A) ~ A is continuous. Now let U be an open subset of P (A) and V its image in A. We show that V is open. Let p E V, and let g be a point of U whose image in A is p. Then g is not in the closure of P (A)\ U and a fortiori, g is not in the closure of the inverse image of A\ V in P(A). Hence p is not in the closure of A\ V, and consequently V is open. This shows that the canonical map P (A) ~ A is open. 3.4.12. The second definition to be given of the topology of A is therefore the following: it is the quotient topology of the topology of P (A) for the canonical map P (A) ~ A. We see that for commutative A, the topological space A is just the usual spectrum of A. 3.4.13. COROLLARY.

For any Ct-algebra, the space

A is

a Baire space.

Let (VI V 2 ) be a decreasing sequence of dense open subsets of A, and let U; be the inverse image of V n in P(A). By 3.4.11, U; is a dense open subset of P (A). Since P (A) is a Baire space (B 14), U; is dense in P(A), and by 3.4.11, V", which is the image of U; in A is dense in A.

References: [174], [443], [582], [896], [1607].

80

THE SPECTRUM OF A C-ALGEBRA

[CH. 3, S

3.5. Third definition of the topology on the spectrum

3.5.1. For every cardinal number n, the Hilbert space of families ({;) of complex numbers indexed by a set of cardinality n, for example the smallest ordinal segment of cardinality n, such that ~ l~d2 < + 00 is called the standard Hilbert space of dimension n. This choice is of no impor tance in what follows: its purpose is merely to fix ideas. Let n be a cardinal, H; the standard Hilbert space of dimension n, A a C*-algebra, Repn(A) the set of representations of A in H; and Irrn(A) the set of non-zero irreducible representations of A in H n There is an obvious canonical map of Irrn(A) into A, namely that which maps each element of Irrn(A) to its class. The canonical image of Irr.tA) in A is the set An of classes of non-zero irreducible representations of dimen

sion n.
3.5.2. We endow Repn(A) with the topology of weak pointwise con vergence over A. Then 1TA ~ 1T, where 1TA 1T E Rep.I A), means that (1TA(a){I71)~(1T(a){I71) for any aEA, {, 71EHn. This topology is the same as that of strong pointwise convergence over A: indeed, if 1TA ~ 1T in the above sense, then for each a E A and {E H n , we have: II1TA (a)~
- 1T(a)~lf
= (1T A(a)~

I1TA (a )~) - 2 Re( 1TA (a)~

11T(a)~)
I ~) -

11T(a )~) 11T(a )~)

+ (1T(a)~
= (1T A(a* a)~

2 Re( 1TA (a)~

2 Re(1T(a)~
=

+ (1T(a* a)~

1 ~)

~(1T(a*a)~

I~)

I~)

11T(a)~)

+ (1T(a*a)~

O.

3.5.3. Let 1TA 1T E Repn(A) and suppose 1T admits a cyclic vector f Then 1TA~1T if and only if 1TA(a)~~1T(a)~ for each a EA; indeed, we then have, for any a, bE A:
II1TA(a)1T(b)~ -1T(a)1T(b)~11 ~ II1TA(a)1T(b)~ -1TA(a)1TA(b)~1I

+ II1TA (a )1T A(b)~

~ IlallII1T(b)~II

- 1T(a )1T(b )gll

+ II1TA(ab)

- 1T(ab )~II

~ 0;

and as the 1T(b )~'s are dense in H; and II1TA(a)11 ~ Iiall, it follows that 1TA(a)71 ~ 1T(a)71 for every 71 E H n. 3.5.4. Let 1TA 1T E Repn(A) and let At be a dense subset of A. Then 1TA~ 1T if (1TA(a)~ 171)~(1T(a)~ 171) for any a E At,~, 71 E H n: this follows at once from

CH. 3. 5]

THIRD DEFINITION OF THE TOPOLOGY ON THE SPECTRUM

81

3.5.5. In Repn(A), equivalence of representations is an open equivalence relation. In fact, let G be the group of unitary operators on H.. and for U E G and 7T E Repn(A) let U7T E Repn(A) be the re presentation defined by (U7T)(X) = U7T(X)U I for each x E A We see immediately that every map 7T ~ U7T is a homeomorphism of Repn(A) onto itself. Hence the saturation of an open subset n of Rep.I A), which is the union of the Utis for U E G, is open in Repn(A). 3.5.6.
E

Let H be a Hilbert space, ~I>"" ~n vectors of Hand E > 0 with the following property: if 111" .. 11n E H satisfy 1(11i l11j) - (~i I)I,,;;; E for any i and j, then there is a unitary operator U on H such that II U11i - ~il , ;;; E for each i.
LEMMA.

> O. Then there is an

It is enough to prove the following: let (l1f, T/f, ... , 11~), where p = I, 2, ... , be a sequence of systems of vectors such that (11f I T/n ~ (~i I~j) for all iand j. Then if G denotes the unitary group of H, we have
UEG

inf

(\1 Ul1f -

~ll

+ ... + II UT/~ -

~nll)

~ 0

as p ~

co.

This is obvious for n = O. Suppose that it is proved in the case where we start with n - 1 vectors. Let H be the subspace of H generated by ~I"'" ~n-I and let H" = H 8 H. By the inductive hypothesis, there exist UI U 2 E G such that II UpT/f - ~II + ... + II Up11~-1 - ~n-III~O. Let ~~ = P H{ .. ~~ = pw~ .. i,,;;;n-I,wehave
I(~~
T/~P

= PH,UpT/~,

T/~P

= PH"Up11~.

Then for 1",;;

- 11 ~p I~;)I

= I(~n

, ;;; I(~n , ;;; I(~n

so that
11~P ~ ~~

- Up11~ I~i)1 I~i) - (Upl1~ IUpl1DI + I( Up11~ I Up11D I~i) - (11~ I11DI + 1I11~II'IIUpl1f - ~il ~ 0

(Up11~

I ~i)1

in the finite-dimensional space H. At the same time,

Now there exists a unitary operator V p on H. coinciding with the identity operator on H, such that Vpl1~P ~ ~~. For 1,,;;; i,,;;; n - I, we have
IIPwUpT/fll ~ IIpH"~il
=

0,

hence

l/VpUpT/f - UpT/fll ~ O.

Therefore

82

THE SPECTRUM OF A C*-ALGEBRA

[CH. 3, 5

Moreover and hence inf

UEG

L IIUl1fi~1

~JI~O

as p ~ + 00.

3.5.7. LEMMA. Let H be a Hilbert space, A a Ct-algebra, 1To a non-zero irreducible representation of A in a closed subspace H o of H, and S the set of elements of Ii whose dimension is ~ dim H. Let al, .. , ap E A, ~l' , ~n E H o and e > O. Let TO E S be the class of 1To. Then there is a neighbourhood V of To in S with the following property: if T E V, there is a representation 1T of A in a closed subspace of H such that: (1) the class of 1T is T; (2) 111T(a;)PH,Ji -1To(a;)~ill ~ for any i and j; (3) if dim T = dim H, then H" = H.

We can at once restrict attention to the case of unital A, where a(

Ilaill~ 1 and ~! ~ O.

= 1,

(1) Suppose first that n = 1, and let ( > O. Then there is a neigh bourhood W of To in Ii with the following property: if T E W, there is a representation p of A of class T and a ~ E H; such that
(1)
I(p(a~a)~ I~)

- (1To(a~aj)~!1

~I)I

~ !

for any i and j (this follows from theorem 3.4.11). If T E V = W n S, we may assume that H p is a closed subspace of H, equal to H if dim T = dim H. For x E A, we extend p(x) to an operator p(x) E .<(H) which vanishes on H H p The inequality (1) may be written

I(p(a)~

Ip(a;)~)

- (1To(aj)~II1To(ai)~I)1

~ !.

If

( has been suitably chosen, this implies (3.5.6) the existence of a unitary operator U on H such that

IIUp(a;)~

- 1To(a;)~III

~ /2

for each i.

In particular, taking i = 1, we have

IIUp'(a;)U-!~l7To(a;)~!11

II U~

- ~!I ~ /2. Therefore

~ IIUp'(a;)U-!~(-

Up(a;)~11

+ IIUp(a;)~
- 7To(a;)~!11

- 7TO<a;)~(11
~ e

1 U~!~! lI aill 1

- ~II

+ II Up(ai)~

for each i. We can then take for 7T the representation x~ Up(x)U- I (2) Now suppose n is arbitrary. There exist hi, .. , bn E A such that

CH. 3. 5j

THIRD DEFINITION OF THE TOPOLOGY ON THE SPECTRUM

83

Ilgj - 7To(b)glll,,;;; E/4 and by (1), there is a neighbourhood V of TO in S with the following property: if l E V, there is a representation 7T of A of class l in a closed subspace of H (equal to H if dim l =dim H), such that 117T(a ibj)PH,JI- 7TO(aibj)glll,,;;; 4
for any i and j. Then
E

117T(aJPH,Jj - 7To(aJ~jll ,,;;; 117T(a;)PH,,<~j - 7To(b)~I)1I + 117T(ai bj)P H,JI- 7To(ajb)~111

E E E E

+117T(a;)PHw7To(b)~1
+ II7To(aJ7To(b)~1

- 7T(aJ7T(bj)PHJIII - 7To(aJ~jl

, ;;;4Jla;11 +41Ia;11 +4+4 Jla;ll,,;;;E.

3.5.8. THEOREM. Let A be a C*-algebra and n a cardinal number. Then the canonical map of Irrn(A) onto An is continuous and open. Let Q be a subset of Irrn(A), 7To a point of Irrn(A), and S and TO their images in An. If 7To is in the closure of Q, every positive form associated with 7To is a weak*-limit of positive forms associated with Q and so TO is in the closure of S. This shows that the map Irrn(A)~ An is continuous. Furthermore, evey neighbourhood of 7To in Irrn(A) contains a neigh bourhood of the following type: if a b " " an E A, ~b'" ,~m E H; and E > 0, we take for our neighbourhood the set of 7T E Irrn(A) such that 117T(a;)~j - 7To(aJ~jl , ;;; E for any i and j. By 3.5.7, the image of such a neighbourhood in An is a neighbourhood of To in Am and this shows that the map Irrn(A)~An is open. The topology of An is thus the quotient topology of that of Irrn(A) for the canonical map Irrn(A)~ An.
3.5.9. PROPOSITION. Let A be a C*-algebra and let x E A ", Then for each 7T E A, Tr 7T(X) is a (possibly infinite) number> 0, and the function 7T ~ Tr 7T(X) is lower semi-continuous on A.

Let H be a Hilbert space such that every irreducible representation of A can be realized in a closed subspace of H (2.3.3). Let 7To E A be realized in a closed subspace H o of H. Let a be < Tr 7To(x) and (~I' ... , ~n) be an orthonormal system in Hi; such that
n

~ (7T o(x )gj I~j) - a

j=1

={3 > O.

84

THE SPECTRUM OF A C*ALGEBRA

[CH. 3. 6

By 3.5.7. there is a neighbourhood V of 1To in A such that every 1T E V is realized in a closed subspace of H in such a way that

Then Tr 1T(X);;

L (1T(X)PHJj
j=1

Ie j);.;. L (1T o(x )ej Ie) j=1

f3
n .-

a.

References: [447], [582], [583]. 3.6. Finite-dimensional representations 3.6.1. Let k be a (commutative) field. n an integer>0, and Mn(k) the algebra of (n x n) matrices over k. This algebra is of dimension n 2 and hence. if r is an integer> n 2 then
(I)

L
uES r

EaX"CT(l)XCT(2)"

XCT(r)

=0

for any Xl. .. , X, E Mn(k). where Sr denotes the symmetric group on {I, 2, ... , r} and E CT is the signature of the permutation o: Let r(n) be the smallest integer r such that the identity (1) holds in Mn(k). 3.6.2.
~UE31
LEMMA.

r(n);.;. r(n - I)

+ 2.

1) - 1, and let Xl"" X, E Mn-l(k) be such that Y = # O. Putting Y = (Yi), there exist hand L such that Yhl # O. Let Xi. Y be the matrices obtained from Xi and Y respectively by adjoining an nth row and an nth column of zeros. Let X;+I and X;+2 be the matrices of Mn(k) all of whose entries are zero apart from the (I, n)th and (n, n)th entries respectively which are equal to I. In the sum Let t
EaX"u(I) XCT(tl

= r(n -

(1)

L
TE2: t +2

ETX~(l)X~(2)'"

X~'+2)

all the terms in which X;+2 does not appear as the final factor vanish, because X;+2Xj = 0 for j < t + 2, and as regards those which do ter minate in X;+2 they too are zero unless X;+2 is preceded by X;+l. because XjX;+2 = 0 for j < t + 1. The sum (1) is therefore equal to YX;+\X;+2 = YX;+I and this matrix is non-zero as Yhl# O. Hence r(n) > t + 2 = r(n - 1) + 1.

CH. 3. 6]

FINITE-DIMENSIONAL REPRESENTATIONS

85

3.6.3. PROPOSITION. Let A be a C*-algebra, n an integer >0, nA the set of 7T E A such that dim 7T ~ n, and An the set of 7T E A such that dim 7T = n. Then A is closed in A and An is open in nA. (i) n (ii) Let In be the intersection of the kernels of the 7T E nA. Then nA is canonically homeomorphic to the spectrum of AIIn and An is canonically homeomorphic to the spectrum of In~dI nLet t = r(n) and X l " " X, E A. The element ~()"E3 belongs to the kernel of each of the representations of to In. Let 7T E A be such that Ker 7T :2 In. Then
uEa t

nl. and therefore

=0

EaX"()"(l)

X()"(t)

E u 7T (Xu (l ) "

7T(Xu ( t)

= 7T (L

EuXu(l)"

XU(t)

aE6 t

for any XI .. ,X, E A. Since 7T(A) is strongly dense in .2(H,,), it follows that ~UE3, EuUu(l) uu(t) = 0 for any u!, , U, E .:(H,,). By 3.6.2, H" does not contain any finite dimensional subspace of dimension> n, and so dim H ~ nand 7T E nA. Thus nA is closed in A, hence An = nA\n-IA is open in nA. The assertion (ii) follows from (i) and 3.2.1.
7T

3.6.4. PROPOSITION. (i) The topology of An is locally compact. (ii) This topology is the coarsest for which the functions 7T ~ Tr 7T(X) on Am X E A, are continuous. Let H; be the standard Hilbert space of dimension n. For each x E A, the map 7T ~ 7T(X) is continuous on Irrn(A), and since n < + 00, it follows that the function 7T ~ Tr 7T(X) is continuous on Irrn(A). Since this func tion takes the same value on any two equivalent representations, it induces, passing to the quotient, a continuous function cPx on An (3.5.8). Moreover if 7T and 7T are two inequivalent irreducible representations of A in H., there is an X E A such that Tr 7T(X) - Tr 7T(X) (2.8.3 or B 27), and so the functions 4Jx separate the points of An. This shows, inciden tally, that the topology of An is Hausdorff. Now An can be identified with the spectrum of a C*-algebra [3.6.3(ii)), hence is locally compact (3.3.8). Let T be its topology, T the coarsest topology on An for which the cps are continuous. We have proved that T is coarser than T; let us prove that it is finer. Replacing A by In-II In (d. 3.6.3) we may assume that A = An. Let Z = An U {w} be the Alexandroff compactification of An. By 3.3.7, the functions 4Jx can be extended to continuous functions rjlx on Z which take the value 0 at w. The functions l/Jx separate the points of Z; indeed, by the above, it is enough to show that, for every 7T E An, there is an X E A such

86

THE SPECTRUM OF A C*-ALGEBRA

ICH. 3. 7
~

that I/JA 7T)

0; however if yEA is such that 7T(Y)

0, then

I/Jy.y( 7T) = Tr 7T(y* y) = Tr 7T(y)*7T(Y) > O.

Thus the topology defined on Z by the functions I/Jx is Hausdorff and coarser than the original topology of Z, and so the two topologies coincide. This proves (ii). References: [582], [894], [896].
3.7. More about the spaces Rep.(A)

3.7.1.
~

PROPOSITION.

If A is a separable C*-algebra and n a cardinal

?\o, then the topological space Repn(A) is polish.

The standard n-dimensional Hilbert space H; is separable. Let 5,(Hn) be the set 5(Hn) endowed with the strong topology which is quasi complete (B 11). Let 5,(A,5,(Hn be the space of continuous linear maps of A into 5,(Hn), endowed with the topology of pointwise con vergence. Each closed equicontinuous subset of 5s(A,5,(Hn is a complete uniform subspace of 5,(A, 5s(Hn (B 12). In particular,the set of linear maps 7T of A into 5s(Hn) such that 117T(X)11 ~ IIxlIfor every x E A is a complete uniform subspace B of 5,(A,5,(Hn Let (Xi)iEI be a fundamental sequence in A such that Ilxill~ 1 for each i, and let ({j)jEJ be a fundamental sequence in H; with II{jll~ 1 for each j. The uniform structure of B admits as a fundamental system of entourages, the set of entourages defined by the inequalities

117T(X;){j - 7T(x;){jll~

I k

(i E I, j E J; k = 1,2, ...),

and so this uniform structure is metrizable. The mapping

7T ~ (7T(X i ) { )
iEI, jEJ

of B onto a subspace of H~xJ is bicontinuous, and so B is separable. In short, B endowed with the topology of strong pointwise convergence is a polish space. Finally, Repn(A) is the set of 7T E B such that 7T(XY) = 7T(X)7T(Y) and 7T(X*) = 7T(X)* for any x, yEA. The second condition is equivalent to the condition that (7T(X*){ 17]) =({ 17T(X)7]) for any x E A, {, 7] E Hi. Repn(A) is thus a closed subspace of B, and is consequently a polish space.
3.7.2.
LEMMA.

Let E, F be two complex Banach spaces with F se-

CH. 3, 7]

MORE ABOUT THE SPACES Rep.(A)

87

parable, M a topological space and m ~ U; a continuous map from M into ::s(E, F), the set of continuous linear maps from E into F endowed with the topology of pointwise convergence. Let n be an integer. Then the set of m EM such that codim Um(E)"; n is a Ga set in M.
Let t.. fz, ... .t. be elements of F, and let R be the set of sequences r =(rl rz, .. , rk) of complex numbers such that

!",; Irllz+ . . . + Ird2 ",; 1.

For each r E R, e E E and integer p > 0, the set M(r, e, p) of mE M such that 1 Ilrdl +... + rJk - Um(e)11 <
p

is open in M. Now f., ... .t, are linearly dependent modulo Um(E) if and only if, for every p, there is r E Rand e E E such that

Ilrdl + ... + rJk - Um(e)11 <-;

p

in other words, if and only if, mE

p=1

U U

00

M(r,e,p).

rER,eEE

[Indeed, if this is so, we can assume that (rio .. , rk) tends to a limit as p ~ +00 and then Um(e) has a limit point, thus an element of Um(E), which is a non-zero linear combination of fl>"" k Conversely, if t..... ,A are linearly dependent modulo Um(E), there is, for each p, an r =(rio .. , rk) and an e E E such that

IIrdl +... + rJk - Um(e)11 <-,

p

from which the assertion follows.] The set of points mE M such that t.. .. ,fk are linearly dependent modulo Um(E) is therefore a Ga in M. Now let (gl gz, ..) be a dense sequence of elements in F. Codim Um(E)"; n if and only if gilg~, . . . ,gi are linearly dependent modulo Um(E) for any choice of indices it, , in +l . As the family of systems ( i l " " in +l ) is countable, the set of lemma 3.7.2 can be exhibited as a countable intersection of Gas, and is therefore itself a Ga. 3.7.3. LEMMA. Let A be a C*-algebra, n a cardinal ";?\o, M a topologi cal space, m ~ 7Tm and m ~ 7T ~ continuous maps from Minto Rep, (A),

88

THE SPECTRUM OF A C-ALGEBRA

[CH. 3. 7

and p an integer of 7Tm and 7T~ is

~O.
~p

The set of m EM for which the intertwining number is a Os in M.

Let H; be the standard Hilbert space of dimension n, K the unit ball of H; and B the unit ball of A. We denote by Y the Banach space of families (JLijdiEB.iEK.kEK of complex numbers such that
II(JLiik)11

iEB.iEK. kEK

IJLiikl < +00.

Its dual is the Banach space X of families (Aiik)iEB.iEK.kEK of complex numbers such that II(Aiidll = sup IAiid < +00. We denote by P the predual of the von Neumann algebra 5t(H n ) (A 23), which is separable. For each m EM, let V m be the continuous linear map from 5t(H n ) into X defined by
Vm(T)

= (7Tm(i)T -

T7T~(i))j

I k)iEB.iEK.kEK

The intertwining number of 7Tm and 7T~ is then the dimension of Ker V m We have for each T E 5t(H n ) and (JLiik) E Y, writing Xi ~; instead of i, j,
JLiik), V m(T = ~ JLiik 7Tm(x;) T - T7T ~(Xj~i' iik
=

~k)

~ JLiikni 17Tm(X~)~k)
iik

- (T7T~(xi)~i

I~k'

Given two vectors Tl and ~ of H n , we denote by <I>(Tl, ~) the element of P and we have II</>(Tl, l)11 = IITlIIlllll. Then defined by (</>(Tl, ~), T) = (TTl I

JLiik), V m(T =

(~k JLiji</>(~i'
I}

7T

m(X~)~k)

</>(7T~(Xi)~i'

~k,

T)

so that V m is the transpose of the continuous linear map Ui; of Y into P defined by
(1) UmfLiik

= ~ JLiik(</>(~i'
iik

7Tm(x~)~d

</>(7T~(x;)~i'

~k,

and the intertwining number of 7Tm and 7T~ is codim Um(Y). To prove 3.7.3, it is enough, by 3.7.2, to establish the continuity of the map m - U m from Minto 5t(Y, P) endowed with the topology of pointwise convergence. We therefore show that for a fixed (JLijk) E Y the element (1) of P depends continuously on m, in the sense of the norm in P. Since
IIJLiik(</>(~i' 7Tm(X~)~k)
-

</>(7T~(Xi)~i'

~k11

~ 2lJLiid,

CH. 3. 8)

THE MACKEY BOREL STRUCTURE

89

it is enough to show that each term </>(t </>(1T :"(Xj)tj , tk) j , 1Tm(xt)td, depends continuously on m. Now the maps m ~ 1Tm (x t;)t j and m ~ 1T :"(Xi)tj from Minto H; are continuous by definition of the topology of Repn(A) and the map (T/,~)~</>(T/,n from Hi x H; into P is con tinuous by virtue of the relation II</>(T/, nil =liT/II . I ~I 3.7.4.
~('\o.

PROPOSITION. Let A be a separable C*-algebra and n a cardinal Then the topological space Irrn(A) is polish.

Apply 3.7.3 with M = Repn(A), 1Tp = 1T~ = P for each p E Repn(A), and p = 1. The resulting set in M = Repn(A) is, if n > 0, the set Irrn(A) possibly augmented by the, trivial representation (if n = 1). Irrn(A) is thus a 0 6 in the space Rep.j A) which is polish (3.7.1), and therefore Irrn(A) is itself polish (B 15). 3.7.5. If we apply 3.7.3 with M = Repn(A), 1Tp = p, 1T~ = the trivial re presentation of A in H; and p =0, we obtain the set of non-degenerate representations of A in H n If A is a separable C*-algebra, we see that this set is a 0 6 in Repn(A) and is consequently a polish space. References: [444]" [1005].

\.8. The Mackey Borel structure 3.8.1. Let A be a separable C*-algebra. For every cardinal n ~ (\o, we endow Repn(A) and Irrn(A) with their respective topological Borel structures (B 19), which are the coarsest Borel structures for which the functions 1T~(1T(X)tlT/) (xEA, t and T/ in the standard Hilbert space of dimension n) are Borel measurable. Let Rep(A), Irr(A) be the union of the Repn(A) and Irrn(A) respectively for n = 1,2, ... , (\o endowed with the sum-Borel-structure of the structures of the RePn(A) and Irrn(A) respectively. The Borel space Irr(A) is a Borel subset of Rep(A), by the proof of 3.7.4, and the structure induced on Irr(A) by the Borel structure of Rep(A) is precisely the Borel structure of Irr(A). Moreover, by 3.7.1 and 3.7.4, the Borel spaces Rep(A) and Irr(A) are standard
(B 20).

3.8.2. Since A is separable, the canonical map Irr(A) ~ each element of Irr(A) to its class, is surjective (2.3.3).

A which

maps

90
DEFINITION.

THE SPECTRUM

OF A C*-ALGEBRA

[CH. 3. 9

on

For a separable C*-algebra A, the Mackey Borel structure is the quotient of the structure of Irr(A) for the canonical map

Irr(A)~A.

This Borel structure is not standard, in general (cf. however, 4.6.1). 3.8.3. PROPOSITION. For a separable C*-algebra A, the Mackey Borel structure on A is finer than the topological Borel structure of A. Let [l be an open subset of A. For n = 1,2, ... , ?\o the inverse image of [l in Irrn(A) is open (3.5.8) and therefore a Borel set. The inverse image of [l in Irr(A) is thus a Borel set, so that [l is a Borel set for the Mackey structure. It follows from this that every Borel subset of A for the topological structure of A is also a Borel subset for the Mackey structure. 3.8.4. PROPOSITION. Let A be a separable C*-algebra. Then each point of A is a Borel set for the Mackey structure. Let 7To E Irrn(A) and TO be its image in A. The set of 7T E Irrn(A) for which the intertwining number of 7To and 7T is :e;;O is a Os in Irrn(A) (3.7.3). Hence by 2.3.4, the set of 7T E Irrn(A) inequivalent to 7To is a Os in Irrn(A). The equivalence class of 7To in Irrn(A) is thus a Borel set, and so {To} is a Borel set in A for the Mackey structure. References: [583], [1007].

3.9. Addenda

3.9.1. Let A be a C*-algebra. (a) If {O} is a primitive ideal, A contains a point which is dense in A. In particular, any two non-empty open subsets of A have non-empty intersection. (b) If {O} is not a primitive ideal and if A is separable, A contains a disjoint pair of non-empty open sets. (Use 3.3.2 and 3.4.13). (c) If A is separable and I is a closed two-sided ideal of A, then I is primitive if and only if I is prime (1.9.13). Problem: can the se parability condition in band c be suppressed? *(d) There exist separable C*-algebras A such that {O} is a primitive ideal and such that every non-empty open subset of A is non-Hausdorff.
[443].

CH. 3. 9l

ADDENDA

91

3.9.2. Let A be a C*-algebra. (a) Let X(A) be the compact space of C*-seminorms (1.9.13), let ep be the canonical map I ~Nl (1.9.13), and let Prim(A) be endowed with its topological Borel structure. Then ep is a Borel isomorphism from Prim(A) onto a subset of X(A). (b) If A is separable, Prim(A) is a standard Borel space. (Use 1.9.13 and 3.9.1). (c) If A is separable, the Borel structure of Prim(A) is the quotient structure of that of Irr(A) for the canonical map Irr(A) ~ Prim(A). (Use band B 22.) We therefore speak of the Borel structure of Prim(A) without further qualification. [509]. 3.9.3. Let X be a locally quasi-compact space, and let .o/(X) be the set of closed subsets of X. For each quasi-compact subset C of X and each finite set (J of non-empty open subsets of X, let U(C, (J) be the set of Y E .o/(X) such that Y n C = 0 and y n B #- 0 for each B E (J. Then the set of all the U (C, 0) is a base for a topology on .o/(X) in which .o/(X) is a compact space. N ow let A be a C*-algebra, and for each F E .o/(A), let N F be the function x ~ SUP1TEFII7T(x)11 on A. Then N F is a C*-seminorm on A (1.9.13), and the map F ~ N F is a homeomorphism from g;(A) onto the compact space X(A) of 1.9.13. [584]. 3.9.4. Let X be a topological space. A point b of X is said to be separated in X if, for each point b of X which is not in the closure of {b}, the points b and b admit a pair of disjoint neighbourhoods. Now let A be a C*-algebra. Then (a) '7T~ E 4. is separated in A if and only if for each x E A, the function 7T ~ 111T(X)11 on A is continuous at 7To. (b) If A is separable, the set of separated points of A is a dense G8 in A. (Use a and B 18.) [447]. *3.9.5. Let A be a C*-algebra, A the C*-algebra obtained by adjoining an identity 1 to A and for each representation 7T of A, let ii be the canonical extension of 7T to A. If I is a closed two-sided ideal of A, let I) be the set of elements x + A . 1, x E A, A E C such that xy + Ay E I for every yEA. Then I] is a closed two-sided ideal of A such that II n A = 1. If I = Ker 7T and 7T is non-degenerate, then I. = Ker ii. Fur ther, let (7Ta) be a family of non-degenerate representations of A. Then 7T is weakly contained in the set of 7T a S if and only if ii is weakly contained in the set of iias. [582].

92

THE SPECTRUM OF A C-ALGEBRA

[CH. 3. 9

*3.9.6. Let A be a C*-algebra, n an integer >0 and (7TJiEI a family of n-dimensional elements of A indexed by a directed set. Let {31, ... , {3, be points of A, and suppose that {{31, ... ,{3,} is the set of all limits of subfamilies of (7TJ. Then there is a directed subset J of I and integers mt, ... , m, > 0 such that L~~I mk dim 13k os:; n and lim] Tr 7Ti(X) = L~=I m, Tr (3k(X) for each x E A. [582]. 3.9.7. Let B be a C*-algebra and A a sub-C*-algebra of B such that, for each 7T E 13, 7T IA is irreducible. Then the relation 7T1 IA = 7T21 A in 13 is an equivalence relation R, and the space A can be canonically identified with the space 131 R. [584]. 3.9.8. Let A be a C*-algebra, H a Hilbert space and Rep(A, H) the set of representations of A in H. For 7T E Rep(A, H), E > 0, at, ... , ap E A and gl"" gn in the essential subspace of 7T let V(E, ai ... , ap , ~I" , gn, 7T) be the set of 7T E Rep(A, H) such that 117T(aJg j - 7T(ai)gjll os:; E for every i and j. Let g be the topology on Rep(A, H) for which V(E, al" . ,ap , gl, ... , gO 7T) constitute a base of neighbourhoods of 7T. This topology is not Hausdorff, but induces a Hausdorff topology on the set of non-degenerate representations of A in H. We say that 1T and 7T E Rep(A, H) are almost equivalent if there is a partial isometry U, with initial and final subspaces the essential subspaces of 7T and 1T respectively, such that U7T(X) = 7T(X)U for every x E A. Let Q be the quotient space of Rep(A, H) for this equivalence relation. Then the set Q can be canonically identified with the set of classes of non-degenerate representations of A of dimension os:;dim H. Let 7T E Q, S ~ Q. Then 1T is weakly contained in S if and only if 7T is in the closure of the set of direct sums of finite subfamilies of S. Provided dim H is large enough, we have A ~ Q and the topology induced on A by that of Q is just the topology studied in this chapter. [585]. 3.9.9. Let A be a C*-algebra, 0 a topological group and w a homomor phism of 0 into the group of automorphisms of A, which is continuous for the topology of pointwise convergence. For each 7T E A and s E 0, let s . 7T E A be the representation x ~ 1T(W(S)-I X) of A. Then 7T ~ s . 1T is a homeomorphism of A onto itself and the map (s, 7T) ~ S 1T of 0 x A into A is continuous. [632]. 3.9.10. Let H = L 2([O, I]), A be the C*-algebra of continuous complex valued functions on [0, 1] and for each f E A, let 7T(f) be the operator in

CH. 3, 9]

ADDENDA

93

H of multiplication by f. Then 7T is a representation of A in H, every character of A is weakly contained in 7T and no character of A is contained in 7T.

3.9.11. Let A be a C*-algebra such that Prim(A) is Hausdorff. Then, for each x E A, the function 7T ~ 117T(X)11 is continuous on A. [896]. 3.9.12. Let A be a C*-algebra, Z its centre and suppose that whenever two primitive ideals I and J satisfy I n Z = J n Z, then I = J. Then, under these conditions, I ~ I n Z is a homeomorphism of Prim(A) onto Prim(Z). [893]. 3.9.13. Let A be a separable C*-algebra and B a sub-C*-algebra of A. Then (\0 card A ~ card E. (For each 7T E A, let E" be the set of pEE contained in 1T IB. Then E" is countable and U "EA E" = E.) 3.9.14. Let A be an algebra over a (commutative) field, Prim(A) the set of primitive ideals of A, endowed with the Jacobson topology, and A the set of classes of irreducible representations of A endowed with the inverse image of the topology of Prim(A) for the canonical map A ~ Prim(A). Let I be a two-sided ideal of A. Then, with the notation of 3.2.1, there exist canonical homeomorphisms Prim.fA) ~ Prim(AI I), AI ~ (AII)A, Prirrr{Ar-ePrimff ) and AI~i. [896].

CHAPTER 4

LIMINAL

C*ALGEBRAS

The two problems that we mentioned at the beginning of Chapter 3 are a long way off being solved in general. On the other hand, for certain special classes of C*-algebras, a more or less satisfactory answer is available. These are the classes of C* algebras that we are now going to introduce. 4.1. The algebra of compact operators
4.1.1. Let H be a Hilbert space. The compact operators on H constitute a self-adjoint two-sided ideal of 5t(H), and a norm-limit of compact operators is itself compact. The set of compact operators on H is thus a C*-algebra that we denote by ~(H). In the study of algebras of finite dimension over the complex field, we know that the algebras 5t(H) (for H a finite-dimensional vector space) playa basic part. In the study of C*-algebras, this part is played not by the algebras 5t(H) (for H a Hilbert space), but by the algebras g;cg(H). (Remember that 5t(H) = n(H) if H is finite-dimensional). A C*-algebra A is said to be elementary if there is a Hilbert space H such that the C*-algebras A and .;eeg(H) are isomorphic. 4.1.2. PROPOSITION. Let H be a Hilbert space, A = n(H), and let T be the vector space of trace-class operators on H. For each t E T, let ft be the linear form x ~ Tr(tx) on A. Then t ~ ft is a linear bijection of Tonto the dual of the Banach space A. The form t. is hermitian (resp. positive) if and only if t is hermitian (resp. positive).

We have ITr(tx)1 =s;; (Trltl>lIxll (A 32), so that f, is a continuous linear form on A. The mapping t ~ t, is clearly linear. If ft = 0, we have 0= f,(t*) = Tr(tt*), and so t = 0, and the mapping t ~ f, is injective. We now show that if f is a continuous linear form on A, then there exists a t E T such that f = ft. It is enough to consider hermitian f only. Let A 0 ~ A be the set of Hilbet-Schmidt operators; it is endowed with

CH.4.1]

THE ALGEBRA OF COMPACT OPERATORS

95

a natural Hilbert space structure, and the canonical injection of AO into A is continuous. Hence the restriction I of I to A is a continuous linear form on the Hilbert space A 0. Consequently, there exists a to E A 0 such that that
f(x) =(x Ito)

=Tr(xt~)

for each x E A 0.

We have, for each x E A O, Tr(xt o) =Tr(t~x*)

= Tr(x*t~)

=f(x*) = f(x) =Tr(xt~),

so that to = t~. Now perform the spectral decomposition of to: there exist mutually orthogonal projections e, of rank 1 on H and real numbers Aj such that

Putting

E;

= 1 if

Ai;a= 0 and

E;

= -1 if

Ai

< 0, we have

IAII + ... + IAnl =TrEt e, +

~ 1Ii1llk,e, +

+ Enen)i o) =I(Elel + ... + Enen) + Enen II = 1Ii11,

so that
0

L IAil < +00.

Thus to E T. The forms I and Ito coincide on A by construction of to, therefore I = Ito since A 0 is dense in A. This proves that t - It is a bijection of T onto the dual of the Banach space A For t E T and x E A, we have
j,.(x)

=Tr(t*x) =Tr(tx*) = ft(x*),

It = (/t)*

so that
(1)

This proves that It is hermitian if and only if t is hermitian. If t ;a= 0, tl/ 2 is a Hilbert-Schmidt operator, and we have, for each x E A,
j,(x*x)

= Tr(t/2x*xt/2) = Trxt /2)*(xt1/2 ;a= 0, so that i O.

Finally, let t E T be such that It ;a= O. Let E be a unit vector and let e be the orthogonal projection on CE; we have

o ~ft(e)
This proves that t 0:

= Tf(re) = Tr(ete) =tte IE).

96

LIMINAL C*-ALGEBRAS

[CH. 4, l

4.1.3.

COROLLARY.

The positive forms on A are just those of the type

x ~ Abgll gl) + Az(xgzl gz)

+.. "

where (gl> gz, ... ) is an orthonormal system in H and the A/ s are non-negative real numbers such that L Aj < +00.

The operators with positive trace on H are the operators of the form + AzP cQ + .. " where (gl gz, ... ) is an orthonormal system in H and the A/s are non-negative real numbers such that L Aj < +00. Since
A1PCI;! Tr(PCl;jx)

=Tr(PqjxPcl;) = (xg Igj),

j

the corollary follows immediately from 4.1.2. 4.1.4.

COROLLARY.

x ~ (xg

Ig)

where

g is

The pure positive forms on A are just the forms a non-zero vector in H.

Since the identity mapping of A is irreducible (if Hi 0), the forms x ~ (xg I g), where g is a non-zero vector of H, are pure (2.5.4). Conversely, if f is a pure positive form on A, all the positive forms dominated by fare proportional to f, so that f is of the type x ~ (xg Ig) by 4.1.3. 4.1.5. COROLLARY. Every non-trivial irreducible representation of A is equivalent to the identical representation. This follows from 2.5.4, 4.1.4, and 2.4.1 (ii). 4.1.6. COROLLARY. Let B( i 0) be a sub-C*-algebra of A and suppose that the identical representation of B is irreducible. Then B = A. The commutant of B in X(H) is just the set of scalar operators. Since B i 0, we see (A 12) that B is ultra-strongly dense in X(H), and a fortiori in A. With the notation of 4.1.3, the form
x~

L Ai(xgj Ig;) = L WAI2I;j(x)

is ultra-strongly continuous on A; hence each continuous linear form on A is ultra-strongly continuous on A. The preceding work proves that every continuous linear form on A which vanishes on B is identically zero. It then follows from the Hahn-Banach theorem that B = A. 4.1.7.
COROLLARY.

The only closed two-sided ideals of A are 0 and A.

By 4.1.5. the only primitive ideal of A is O. Moreover, every closed two-sided ideal of A is the intersection of primitive ideals [2.9.7 (ii)].

CH. 4, Ij

THE ALGEBRA OF COMPACT OPERATORS

97

4.1.8. COROLLARY. Let Hand H be two Hilbert spaces, and ip an isomorphism of the C*-algebra .Te(H) onto the C*-algebra .Te(H). (i) There exists an isomorphism U of H onto H that defines sp. (ii) Let U be an isomorphism of H onto H. Then U defines cp if and only if U = 11: U, where A is a complex number of modulus 1.

Since cp is a non-trivial irreducible representation of .Te(H) (provided H, H 10), (i) follows from 4.1.5. Now U defines ip if and only if U,-I U defines the identical automorphism of .2Cfd(H), in other words, if and only if it commutes with .Te(H), in other words, if and only if it is a scalar operator; hence (ii) follows.
4.1.9. We retain the notation of 4.1.8. We see that tp transforms the set of Hilbert-Schmidt elements (resp. the set of elements of finite rank, of rank I, ...) of .Te(H) into the set of Hilbert-Schmidt elements (resp. the set of elements of finite rank, of rank 1, ...) of .2Cfd(H). Given an elementary C*-algebra A, we can therefore speak of the Hilbert-Schmidt elements, ... , of A independently of the realisation of A as an algebra of compact operators. Similarly, the Hilbert dimension a of H is intrinsic to the C*-algebra .2Cfd(H) which is said to be of rank a 2 For finite a, we recover the usual notion of rank of an algebra. Note that for infinite a, we have a 2 = a. 4.1.10. COROLLARY. Let A be a C*-algebra and 7T an irreducible re presentation of A. If 7T(A) n .Te(HTr ) I 0, we have 7T(A):2 .Te(HTr ) , and every irreducible representation of A with the same kernel as 7T is equivalent to 7T.

Let A = 7T- 1(.Te (H . Then A is a closed two-sided ideal of A. We have 7T(A) I0, so that 7T(A) is irreducible in H (2.11.2 (iij). Hence 7T(A) = .2Cfd(H (4.1.6). Every representation of A with the same kernel as 7T is the composition of 7T and an injective representation of 7T(A). To prove the second assertion of 4.1.10, it is therefore enough to establish the following: let H be a Hilbert space, and B a sub-C*-algebra of .2(H) containing .Te(H); then every injective irreducible representation p of B is equivalent to the ide~tical representation. Since p 1.Te(H) is equivalent to the identical representation of .2Cfd(H) (4.1.5), we can suppose that H; = H and that p 1.2Cfd(H) is the identity mapping of .2Cfd(H). Then 2.10.4 (i) proves that p is the identity mapping of B.
Tr Tr

Tr )

4.1.11. COROLLARY. Let A be a C*-algebra, 7T a non-trivial irreducible representation of A. Suppose that 7T(A) k .2Cfd(H
Tr ) .

98

LlM[NAL C*-ALGEBRAS

[CH. 4, 2

(i) 7T(A) = ~(H,,). (ii) The kernel of

7T

is a maximal closed two-sided ideal of A.

(i) follows from 4.1.10, and (ii) follows from (i) and from 4.1.7. References: [893], [896], [1097], [1099], [1101], [1323]. 4.2. Liminal C*algebras 4.2.1. DEFINITION. A C*-algebra A is said to be liminal if, for every irreducible representation 7T of A and each x E A, 7T(X) is compact. 4.2.2. Let H be a Hilbert space. The C*-algebra ~(H) is liminal (4.1.5), and the zero ideal of this algebra is primitive (if H. 0). Con versely, if A is a liminal C*-algebra in which the zero ideal is primitive, A is elementary by 4.1.11 (i). Every commutative C*-algebra is liminal. Later on (10.4.5, 13.11.12, 15.5.6) we will see other examples of liminal C*-algebras. 4.2.3. Let A be a liminal C*-algebra. Each primitive ideal of A is a maximal closed two-sided ideal of A (4.1.11 (iij), If 7T is a non-trivial irreducible representation of A, we have 7T(A) = ~(H,,) (4.1.11 (ij), 4.2.4. PROPOSITION. Let A be a liminal C*-algebra. Then every sub-C* algebra of A and every quotient C*-algebra of A is liminal. This is clear for a quotient C*-algebra AII of A, because every irreducible representation of All may be identified with an irreducible representation of A. Now let B be a sub-C*-algebra of A, and let 7T be an irreducible representation of B. There exist a Hilbert space K :2 H" and an irreducible representation p of A in K such that 7T(X) = p(x) IH" for each x E B (2.10.2). Since A is liminal, the operators p(x) are compact, and therefore the operators 7T(X) are also compact. Thus B is liminal. 4.2.5. PROPOSITION. Let A be a liminal C*-algebra, 7T., , 7Tn mutually inequivalent elements of A, and T[, ... , T; compact operators in H"I... ,H". Then there exists x E A such that 7T)(X) = T), ... , 7Tn(X) = Tn. This is clear for n = 1, so suppose that it has been proved for n - 1. Then there exist x, yEA such that 7Tl(X) = T), ... , 7Tn-l(X) = Tn-I, 7Tn (Y ) = T; Let I, = Ker 7Tk. By 4.1.10 and 4.2.3, we have I n 1J I k for

CH. 4, 3]

POSTLIMINAL

C*-ALGEBRAS

99

y E In such that 1, ... , n.

= 1, ... , n -1; hence [,,121 = II nn [,,_I (2.11.4). There exist x E I, x + x = y + y = Z, and we have 7Tk(Z) = T, for k =

4.2.6. PROPOSITION. Let A be a C*-algebra. Let I be the set of x E A such that 7T(X) is compact for every irreducible representation 7T of A. Then I is the largest liminal closed two-sided ideal of A.
It is clear that I is a closed two-sided ideal of A. Let p be an irreducible representation of 1. Then there exists an irreducible re presentation 7T of A in H p which extends p (2.10.4). Hence p(x) = 7T(X) is compact for each x E I, so that I is liminal. Finally, let J be a liminal closed two-sided ideal of A. We shall show that J ~ I, i.e. that for every irreducible representation a of A and every x E J, a(x) is compact. This is clear if a(J) = O. Otherwise, a IJ is irreducible (2.11.2 (iij), and a(x) is compact because J is liminal. Liminal C*-algebras were introduced in 1951 by Kaplansky under the name of CCR-algebras (CCR = completely continuous representations; "completely continuous" is synonymous with "compact"). Kaplansky and Glimm also introduced the notation GCR and NGCR for C* algebras later called postliminal and antiliminal. These notions are fun damental, but the accepted terminology seemed somewhat inelegant to me.

References: [896]. The proof of 4.2.5 was communicated to me by J. Tomiyama.

4.3. Postliminal C*-algebras

4.3.1. DEFINITION. A C*-algebra A is said to be postliminal if every non-zero quotient C*-algebra of A possesses a non-zero liminal closed two-sided ideal.

These are the most important C*-algebras for what follows. Every liminal C*-algebra is postliminal (4.2.4) but the converse is false.
DEFINITION. A C*-algebra is said to be antiliminal if the zero ideal is its only liminal closed two-sided ideal.

By 1.8.5, such a C*-algebra does not possess any postliminal closed two-sided ideal either, apart from the zero ideal.

100

LIMINAL C*-ALGEBRAS

[CH. 4, 3

For examples of postliminal or antiliminal C*-algebras, see the ad denda (4, 5, 9, 10, 13, 17).
DEFINITION. Let A be a C*-algebra. A composition sequence for A is an increasing family (lp)O"P"a of closed two-sided ideals of A, indexed by the ordinals lying between and some fixed ordinal a, and possessing the following properties: (i) 1o = 0, L, = A; (ii) if p:s;; a is a limit ordinal, Ip is the closure of U p<p Ip"

4.3.2.

PROPOSITION. Let A be a C*-algebra. Then there exists an ordinal a and an increasing family (lp)O",p",a of closed two-sided ideals of A, possessing the following properties: (i) 1o = 0, All; is antilim inal ; (ii) if p :s;; a is a limit ordinal, I, is the closure of U p<p I p; (iii) if p < a, Ip+11Ip is the largest liminal two-sided ideal of Al I; and is non-zero. Furthermore, a and the family (lp) are uniquely determined by the above conditions.

4.3.3.

The Ips are constructed using transfinite induction: if p is a limit ordinal, condition (ii) defines the construction of I p ; otherwise, we have p = pi + I, and condition (iii) defines the construction of I p , unless AII p is antiliminal, in which case the construction comes to a halt with a = p,
4.3.4.
PROPOSITION. Let A be a C*-algebra. Then the following con ditions are equivalent: (i) A is postliminal; (ii) A possesses a composition sequence (Jp)o"P" (3 such that all the Jp+lIJ/s are postliminal; (iii) A possesses a composition sequence (lp)O",p",a such that all the Ip+dI ps are liminal.

(i):;(iii): suppose that A is postliminal. With the notation of 4.3.3, we cannot have AIIa,t. 0, so that Ia = A, and (lp)O",p",a is a composition sequence for A possessing the required properties. (iii):; (ii): clear; (ii):;(i): suppose that there exists a composition sequence (J p )o"P "(3 such that the Jp+lIJp are postliminal. Let JI- A by a closed two-sided ideal of A. We have to show that AIJ possesses a non-zero liminal two-sided ideal. There exists a least ordinal o P such that J p g J. If pi < p, we have Jp ~ J; hence the closure of U p<p Jp is contained in J and so p is not a limit ordinal. Let p = pi + 1, and let I = Jp n J. Then Jrl I may be

CH. 4. 3j

POSTLIMINAL

C*-ALGEBRAS

101

identified with a closed two-sided ideal of All (1.8.4). On the other hand, I)I is a non-zero quotient of the postliminal C*-algebra lp,+dl p" and therefore possesses a non-zero liminal two-sided ideal. This ideal is also a two-sided ideal of All by 1.8.5. 4.3.5. PROPOSITION. Let A be a postliminal C*-algebra. Then every sub-C*-algebra of A and every quotient C*-algebra of A is postliminal. This is clear for quotient C*-algebras. Now let B be a sub-C*-algebra of A, and let (Ip)O";p,;;a be a composition sequence of A such that the Ip+dlp are liminal (4.3.4). Then (Ip+1 n B)I(Ip n B) may be identified with a sub-C*-algebra of Ip+Illp (1.8.4) and is therefore liminal (4.2.4). To finish the proof, it is enough, by 4.3.4, to show that (Ip n B)o.,;p.,;a is a composition sequence of B. It suffices to show that, if p is a limit ordinal, and if x E Ip n B, then x is a limit of points of U p<p Ip n B. Let e > O. Then there exists a p < P such that the distance of x from Ip is <e. Since BI(Ipn B) may be identified with (B + Ip,)IIp" there is a point of I p n B whose distance from x is <e, and the assertion follows. 4.3.6. PROPOSITION. Let A be a C*-algebra. Then the ideal Ia of 4.3.3 is the largest postliminal two-sided ideal of A, and the smallest closed two-sided ideal such that the corresponding quotient of A is antiliminal. The ideal I; is postliminal by 4.3.4. Let I be a postiiminal two-sided ideal of A. Then I(I n Ia) may be identified with a postiiminal two-sided ideal of Al I.; Since Al I; is antiliminal, we have III n I; = 0, so that I.~ I a. Finally, let I be a closed two-sided ideal of A such that All is antiliminal. Then Ial(l n I a) may be identified with a postliminal two sided ideal of All, so that IaI(1 n Ia) = 0 and so I;;;) Ia. Proposition 4.3.6 to some extent reduces the study of general C* algebras to the study of postliminal C*-algebras on the one hand together with that of antiliminal C*-algebras on the other. Not much is known about antiliminal C*-algebras. In contrast, fairly detailed in formation concerning postiiminal C*-algebras is available. 4.3.7.
THEOREM. Let A be a postliminal C*-algebra (i) For every non-trivial irreducible representation TT of A, TT(A) includes all the compact operators on H (ii) IfTTl and TT2 are two irreducible representations of A with the same kernel, then TTl and TT2 are equivalent.
7T

Let TT be a nontrivial irreducible representation of A. Then AIKer TT

102

LIMINAL C"-ALGEBRAS

[CH. 4, 4

contains a non-zero liminal two-sided ideal IIKer 1T. The restriction of 7T to I is nontrivial and irreducible (2.11.2 (iij), so that 1T(I) = .T (H,,) because IIKer 7T is liminal (4.2.3). This proves (i); (ii) follows from (i) and 4.1.10. Theorem 4.3.7 admits a converse (9.1). 4.3.8. Let A be a separable C*-algebra and (Ip)O<p<a a compositton sequence for A. If the I p +1/lp are non-zero, then the family (Ip) is countable. This is because, for each ordinal p < a, we can choose an x p E I p+1 whose distance from I p is ~ 1. The distance between any two distinct xps is thus ~ I and so the family of the xps is countable. 4.3.9. Let A be a C*-algebra and A the C*-algebra obtained by ad joining an identity element to A. If A is postliminal, A is postliminal (4.3.4). If A is postliminal, A is postliminal (4.3.5). If A is antiliminal, A is clearly antiliminal. Now suppose that A is antiliminal. If A is unital, then A possesses a I-dimensional ideal and so it is not antiliminal. But if A is not unital, A is antiliminal, for if A possessed a non-zero liminal two-sided ideal I, we would have I n A = 0, so that I would be com plementary to A in A, and thus A would be a quotient of A and would therefore be unital. References: [443], [583], [631], [896].

4.4. The spectrum of a postliminal algebra

4.4.1. If A is a postliminal C*-algebra, Theorem 4.3.7 proves that the canonical mapping A ~ Prim(A) is a homeomorphism. Hence A is a To-space (3.13). If, in addition, A is liminal, the primitive ideals of A are all maximal among the closed two-sided ideals (4.2.3), so that the points of A are closed (3.1.4). 4.4.2. LEMMA. Let A be a C*-algebra, 7To a point of A and x an element of A +, such that the function 7T ~ Tr 7T(X) on A is finite and continuous at 7To (i) If yEA + is dominated by x, then the function 7T ~ Tr 7T(y) on A is continuous at 7To. (ii) If 7To(x) ;z!: 0, there exists a z E A + such that 1T(Z) is a projection of rank I for every 1T in some neighbourhood of 7To. Let y = x - yEA +. The functions 7T ~ Tr 1T(y), 1T ~ Tr 1T(y) are lower

CH. 4, 4]

THE SPECTRUM OF A POSTLIMINAL ALGEBRA

103

semicontinuous (3.5.9), and their sum is finite and continuous at xo, so that they are both continuous at xo To prove (ii), we can assume that II7To(x)11 = 1. Let K be a I-dimen sional eigenspace of 7To(x) corresponding to the eigenvalue 1. The irreducible algebra 7To(A) contains the non-zero compact operator 7To(x), and therefore contains 2ce(H"o) (4.1.10); hence there exists z, E A+ such that 7TO(ZI) = P K. Let Z2 = f(z.) where f(t) = t for o~ t ~ 1 and f(t) = 1 for t~I; then Ilz211~I and 7To(z2)=f(PK)=PK. Let Z3=Xl/2Z2XI/2~X; we have 7TO(Z3) = 7TO(X)I/2PK7TO(X)I/2 = P K By (i), we have Tr 7T(Z3) ~ i in some neighbourhood of 7To; moreover, 117T(z3)11 ~ ~ in some neighbourhood of 7To (3.3.2); hence, in some neigh bourhood V of 7To, the largest eigenvalue of 7T(Z3) is ~~ and of mul tiplicity 1, while the others are ~i - ~ =!. Let g: R ~ R be a non-negative continuous function equal to 0 for t ~! and to 1 for t ~~; let z = g(Z3) E A +; then 7T(Z) = g( 7T(Z3 is a projection of rank 1 for every 7T E V.
4.4.3.
LEMMA. Let A be a liminal C*-algebra. Let I be the set of x E A such that the operator 7T(X) is of finite rank for every 7T E A. Then I is a dense self-adjoint two-sided ideal of A.

It is clear that I is a self-adjoint two-sided ideal of A. For n = 1, 2, ... , let t.. R ~ R be a continuous function which vanishes in a neighbourhood of 0 (depending on n), such that fn(t) converges uni formly to t as n ~ + 00. For every hermitian element x of A, fn(x) ~ x as n ~ +00. Moreover, for each 7T E A, 7T(X) is compact, so that its eigen values (counted with their multiplicities) converge to zero, and hence 7T(fn(X = fn(7T(X only has a finite number of non-zero eigenvalues of finite multiplicity; hence 7T(fn(X is of finite rank. Thus, I = A.
LEMMA. Let A be a non-zero postliminal C*-algebra. Then A possesses a non-zero closed two-sided ideal I with the following pro perty: for any pair of distinct points 7T1 and 7T2 in the open subset i of A, there exists an x e t: such that: (a) 7T(X) is of rank ~I for every 7TEi; (b) the function 7T~Tr7T(x) is continuous on t, (c) Tr 7T.(X) = 1, Tr 7T2(X) = O.

4.4.4.

We can assume that A is liminal. By 4.4.3, there exists a non-zero Xl in A + such that 7T(Xl) is of finite rank for each 7T E A; hence the function 7T ~ f( 7T) = Tr 7T(Xl) on A is everywhere finite. This function is lower

104

LIMINAL C*-ALGEBRAS

[CH. 4, 5

semi-continuous and not identically zero, and therefore is ~ a>0 throughout some non-empty open subset U of A. Because U is a Baire space (3.2.2 and 3.4.13), f admits at least one point of continuity 7To in U (B 18). By 4.4.2 (ii), there exists an X z E A + and an open neighbourhood V of 7To such that 7T(Xz) is a projection of rank 1 for each 7T E V. Let I be the closed two-sided ideal of A such that j = V, and let 7Th 7Tz be two such that 7T.(X3) = 7T1(Xz), distinct points of V. There exists an X3 E 7TZ(X3) =0 (4.2.4 and 4.2.5). Let x =X~/2X3X~/2. We have x ~ IIx311xz, so that 7T ~ Tr 7T(X) is continuous on V (4.4.2 (ij). Moreover,

Tr 7T\(X)

= Tr 7T\(XZ)Z = I,

Tr 7T2(X)

= O.

Finally, the rank of 7T(X) is less than or equal to that of 7T(X2) and therefore less than or equal to I, for every 7T E V. 4.4.5. THEOREM. Let A be a postliminal C*-algebra. Then dense, locally compact open subset.

Ii possesses a

With the notation of 4.4.4, the points of j are separated by continuous real-valued functions, and so j is Hausdorff. The family of Hausdorff open subsets of Ii is clearly inductive with respect to inclusion. Let S C Ii be a maximal Hausdorff open subset, and suppose that there exists a non-empty open subset U of Ii disjoint from S. Then U may be identified with the spectrum of a non-zero closed two-sided ideal of A, which is itself a postliminal C*-algebra. By the first part of the proof, U contains a non-empty Hausdorff open subset S. It is clear that SUS is a Hausdorff open subset of A, which contradicts the maximality of S. Hence = Ii. By 3.3.8, S is locally compact. References: [452], [896].
4.5. C*-algebras with continuous trace

4.5.1. LEMMA. Let A be a C*-algebra, and P a subset of A+ such that: (a) P+ PC P; (b) if x E A is such that xx* E p, then x*x E p; (c) each element of A + dominated by an element of p belongs to p, (i) The set n of x E A such that xx* E p is a self-adjoint two-sided ideal of A. (ii) The two-sided ideal m = nZ is the set of linear combinations of elements of m" = m n A+. Its closure in A is the same as that of n. (iii) m" = p.

CH. 4, 5l

C*-ALGEBRAS

WITH CONTINUOUS TRACE

105

It is clear that x E n implies that x* E n. Suppose that x, yEn. Then we have

2xx*

+ 2yy* -

(x

+ y)(x + y)* =

xx*

+ yy* -

xy* - yx*

= (x - y)(x - y)*;a. 0,

and so
(x

+ y)(x + y)* ";;;2xx* +2yy* E

xzz*x*,,;;; IIzz*IIxx* E V,

V,

so that (x + y)(x + y)* E V, hence x + yEn. If x E nand z E A, we have

hence (xz)(xz)* E V, hence xz E n. Thus n is a right ideal of A and consequently a two-sided ideal since n = n*, and (i) follows. Each element of 11 is a linear combination of products of pairs of elements of 11 (1.5.8), and therefore a limit of elements of m from which it follows that 11 = 1ft. Let x E m. We have x = L ajbj with aj, b, E n. Hence
4x = ~ (aj

+ b)(aj + b)* - ~
(aj

(a, - bj)(aj - b j)* (aj - ibj)(aj - ibj )*

+i ~
and
(a j + b)(aj

+ ibj)(a j + ib)* - i ~

+ b)*, ...

, (a j - ib)(aj - ib j)* E

m n A +.

Whence (ii). If the above element x is hermitian, we have

4x

=~

(a j + b)(aj

+ b j)* - ~
V.

(aj - bj)(aj - b j)*

,,;;; ~ (aj

+ bj)(aj + b)* E

If x;a. 0, we deduce from this that x E V. Hence rn" ~ V. Conversely, let xEV; we have xl/ 2En, hence x=xl/2xI/2Em+, and (iii) follows.

4.5.2. Let A be a C*-algebra. Let V be the set of x E A + such that the function 1T ~ Tr 1T(X) is finite and continuous on A. It is clear that V+ V~ V If x E A is such that xx* E V, then x*x E V since Tr 1T(X*X) = Tr 1T(XX*) for every 1T E A. If x E V and if yEA + is dominated by x, we have y E V by 4.4.2 (i). Then, lemma 4.5.1 proves that V is the positive part of a self-adjoint two-sided ideal m of A, namely the set of linear combinations of elements of V. If x E m, the operator 1T(X) is a trace-

106

LIMINAL CALGEBRAS

[CH. 4, 5

class operator for every TT E A, and the function TT -4 Tr TT(x) is con tinuous on A. We will put m = m(A).
DEFINITION. A C*-algebra is said to have continuous trace if the ideal m(A) is dense in A.

Since m(A) is the intersection of the primitive ideals of A containing it, the above definition is equivalent to the statement that for each TT E A, there exists an x E m(A) such that TT(x) f:. O. By 4.5. 1(ii), it is equivalent again to the statement that the set of x E A such that the function TT ~ Tr TT(x)TT(x)* is finite and continuous on A (a set which is a self-adjoint two-sided ideal n such that n Z =m(A is dense in A. Later on (10.5.8 and 10.9.4) we will give some examples of C* algebras with continuous trace. 4.5.3.
PROPOSITION.

Let A be a C*-algebra with continuous trace. Then

(i) A is liminal

(ii) A is Hausdorff (iii) For each TTo E A there exists ayE A + such that TT(y) is a projection of rank 1 for every TT in some neighbourhood of TTo. Let TT E A. For each x E meA), TT(x) is a trace-class operator. Hence for each x E meA) =A, TT(x) is a limit in norm of trace-class operators, and is therefore compact. Hence A is liminal. Condition (iii) follows from 4.4.2(ii). Let TTl and TTz be two distinct points of A, I the kernel of TTl and x an element of meA) such that TTz{x) f:. O. By 4.2.5, there exists Xl E I such that TTz{XI) = TTz(x)*. Then XIX E meA), so that the function 'TT~Tr TT(xlx) is finite and continuous on A; we have Tr TT.(x.x) =Tr(O) = 0, hence and Tr TTz(x.x) = Tr TTz(x)*TTz(x) f:. 0;

A is

Hausdorff.

4.5.4. PROPOSITION. Let A be a C*-algebra satisfying conditions (ii) and (iii) of 4.5.3. Then A is a C*-algebra with continuous trace. Let TTo E A. There is an open neighbourhood V of TTo and ayE A + such that, for each TT E V, TT(y) is a projection of rank 1. Since A is Hausdorff, A is locally compact (3.3.8), and there is an open neigh bourhood W of TTo such that W ~ V. Let I be the closed two-sided ideal

CH. 4, 5j

C*-ALGEBRAS

WITH CONTINUOUS TRACE

107

of A such that i = W. Since 1To E W, 1To II is irreducible, and so there exists an hermitian element x of I such that 1To(x) restricted to the l-dimensional subspace 1To(y)(H".o) is the identity operator (2.8.3); re placing x by x", we can assume that x E r . Let z = yllZxyl/Z E r, The function 1T ~ f( 1T) =Tr 1T(Z) on A vanishes on A\ W, and is therefore continuous at each point of A\ V; moreover, z e: Ilxlly, so f is continuous at each point of V (4.4.2(i)). Hence Z E m(Ar. Since 1To(z) = 1To(Y) 0, we have proved that A is a C*-algebra with continuous trace. 4.5.5. THEOREM. Let Abe a postliminal C*-algebra. Then A admits a composition sequence (Ip)O,,;;p,,;;a such that the quotients Ip+1/Ip are C* aLgebras with continuous trace.
If A 0, A has a closed two-sided ideal I( 0 which is a C*-algebra with continuous trace (4.4.4). Then, if I( A, AI II has a closed two-sided ideal IziII 0 which is a C*-algebra with continuous trace. Continuing transfinitely, we obtain the composition sequence (Ip). To some extent, this theorem reduces the study of the structure of postliminal C*-algebras to that of the structure of C*-algebras with continuous trace. We will go into this more thoroughly in due course (Ch. 10).

4.5.6. If a C*-algebra A admits a composition sequence (Ip)O,,;;p""a, the are an increasing family of open subsets of A, and for a limit ordinal p, we have i, = U p<p i p: this follows immediately from 3.2.2. Hence A is the union of the sets ip+l\ip. If A is postliminal and if we choose the Ips so that the Ip+dI p have continuous trace, the spaces i p+ 1\i p are locally compact (4.5.3(ii). The existence of such a family of open subsets of A also follows easily from theorem 4.4.5. 4.5.7. Let A be a separable postliminal C*-algebra. Let (Ip)O""p""a be a composition sequence for A such that the Ip+dlp are non-zero C* algebras with continuous trace. Then the family (Ip) is countable (4.3.8). Hence A is the union of the countable family of subsets ip+l\ip Each of these subsets is the intersection of an open and a closed subset of A. Moreover, each space ip+l\ip is locally compact with a count able base since ip+1\ip =(Ip+lIlp)A and Ip+dlp is a separable C*-algebra (3.3.4). References: [452], [584].

t,

108

LIMINAL C*-ALGEBRAS

[CH. 4, 6

4.6. Borel structure on the spectrum of a postliminal C*algebra

4.6.1. PROPOSITION. Let A be a separable postliminal C*-algebra. The Mackey Borel structure on A is the Borel structure given by its topology, with which A is a standard Borel space. Let SI be the Mackey Borel structure on A and S2 the Borel structure given by its topology. Let AI> A2 be the corresponding Borel spaces. By 4.5.7, A2 is the union of a sequence (X n ) of disjoint Borel subsets, each member of which, endowed with the structure induced by S2, is a standard Borel space. Hence A2, which is the sum of the Borel spaces X m is standard. Moreover, SI is finer than S2 (3.8.3). Since AI is a quotient of the standard Borel space Irr(A) (3.8.1), we have 51 = S2 (B 22). When A is a separable postliminal C*-algebra, one can therefore speak of the Borel space A with no risk of confusion. 4.6.2. PROPOSITION. Let A be a separable postliminal C*-algebra. Then there exists a Borel mapping (~ 7T() of A into Irr(A) such that, for each ( E A, 7T() is in the class (. Let Ap be the canonical image of Irrp(A) in A. The topological space Irrp(A) is polish (3.7.4), and the canonical mapping Irrp(A)~Ap is continuous and open (3.5.8), Further, A is the union of a sequence (X n ) of disjoint subsets, each of which is the intersection of an open and a closed subset of A, and is Hausdorff (4.5.7). Let Irrp.n(A) be the inverse image of X; in Irrp(A). Then in Irrp(A), Irrp.n(A) is the intersection of an open and a closed subset, and is therefore a 0 8 ; hence Irrp.n (A) is a polish space (B 15). The saturation in Irrp.n(A) of an open subset is open since equivalence in Repp(A) is defined by a group of homeomorphisms (3.5.5). Finally, since X; is Hausdorff, the equivalence classes are closed in Irrp,n(A). Hence (B 17) there is a Borel subset B p.n of Irrp,n(A) which meets each equivalence class of Irrp,n(A) in exactly one point. The union B of the Bp,n is a Borel subset of Irr(A) which meets each equiv~lence class of Irr(A) in exactly one point. Let 7T be the mapping of A into Irr(A) such that 7T() is the unique representative of ( belonging to B. Let l/J be the canonical mapping Irr(A) ~ A. Since l/J IB is a Borel bijection of the standard Borel space B onto the standard Borel space A (4.6.1), l/J IB is a Borel isomorphism of B onto A (B 21). Hence 7T is a Borel mapping of A into Irr(A). References: [444], [449], [583].

CH. 4, 7]

ADDENDA

109

4.7. Addenda

4.7.1. Let H be a Hilbert space, t a trace-class operator on H, and ft the linear form x ~ Tr(tx) on ~(H). Then Ilftll =Tr [r]. *4.7.2. Let A be a separable C*-algebra and 7T an irreducible represen tation of A. If 7T(A) n :;e<g(H,,) = 0, there exists a family of mutually inequivalent irreducible representations of A with the cardinality of the continuum and with the same kernel as 7T [454]. 4.7.3. If a separable C*-algebra A is such that Ii reduces to a single point, A is isomorphic to ~(H) for some Hilbert space H. (Use 4.7.2.) Problem: can the separability condition be suppressed? 4.7.4. (a) Let A be a C*-algebra and I a closed two-sided ideal of A. If I and A/I are antiliminal, then A is antiliminal. (b) A sub-Ct-algebra of an antiliminal C*-algebra A is not antiliminal in general (it can be commutative!). However, a closed two-sided ideal of A is antiliminal. (c) A quotient C*-algebra of an antiliminal C*-algebra is not an tiliminal in general. (By 4.3.9, it suffices to construct a non-unital antiliminal C*-algebra. For example, if B is an antiliminal C*-algebra, one can form the restricted product of the sequence (B, B, B, ... [631].

*4.7.5. There exist antiliminal C*-algebras A possessing a family (7T) of finite-dimensional irreducible representations such that Ker 7Tj =O. [689].

be Hilbert spaces of dimension 1, 2, .... Then the 4.7.6. Let HI H 2 , C*-algebra product of the 5(H) is neither postliminal nor antiliminal. [896]. 4.7.7. (a) Let PI < P2 < ... be a sequence of integers such that Pi divides Pi+I Let H be an infinite-dimensional Hilbert space. Then there exists a sequence of sub-C*-algebras AI A 2 , of 5(h) such that: (i) Ai is a factor of type Ip ; ; (ii) Ai k A i+ 1 Let A be the closure of U A j Then A is a separable, antiliminal, unital C*-algebra whose only closed two-sided ideals are 0 and A. The given sequence of p;s defines the C*-algebra A up to isomorphism. *(b) There exist sequences (Pi), (p;) giving rise to non-isomorphic C*-algebras. [629].

110

LIMINAL C*-ALGEBRAS

[CH. 4, 7

4.7.8. Let A be a postliminal C*-algebra, and I the largest liminal closed two-sided ideal of A. Then i is dense in A. (Use 1.9.12b). Each 1T E i has as its kernel a minimal primitive ideal of A. [443]. A minimal primitive ideal of A is not always the kernel of some 1T E i. (P. Green, unpublished.) 4.7.9. There exist separable liminal C*-algebras A such that the set of non-separated points of A (3.9.4) is dense in A. There exist liminal C*-algebras A such that A does not possess any separated points [447]. If A is postliminal, the set of separated points of A is not necessarily contained in a separated open subset of A. (C. Delaroche, unpublished.) 4.7.10. Let A be a separable liminal C*-algebra and X a quasi-compact subset of A. Then X is a 0 8 in A. (Let (U 1, U2 , ) be a base for the topology of A; for each 1T E A\X there is a finite cover of X by a finite number of U;s which do not contain 1T). On the other hand, if A is separable postliminal, a quasi-compact subset of A is not necessarily even a Borel subset of A. *4.7.11. Let A be a C*-algebra. The set I of x E A such that the rank of 1T(X) is finite and bounded as 1T varies in A is a self-adjoint two-sided ideal of A. Suppose I is dense in A, and let (1T;) be a family of elements of A indexed by a directed set. Let PI"" p, E A be such that lim Tr 1Tj(X) = L~=1 Tr Pk(X) for each x E 1. Then {Ph ,Pr} is the set of limits of (1T;) in A. [582]. The C*-algebra A is liminal and A admits an open cover (U;) such that, for each i, every family of elements of A indexed by a directed set possesses an at most finite number of distinct limits in U, (Fell, unpublished.) 4.7.12. (a) For each C*-algebra A, put J(A) = m(A) (4.5.2). There exists an ordinal a and an increasing family (1,,)O,,;;,,""-a of closed two-sided ideals of A possessing the following properties: (a) Jo = 0, J(AIJ a ) = 0; (b) if P ~ a is a limit ordinal, Jp is the closure of U 1"<1 J p ; (c) if P < a, Jp+dJp = J(A/Jp) - O. Moreover, a and the family (J p) are uniquely determined by the above conditions. Put Ja = K(A). (b) We have K(A) = A if and only if J(AII) - 0 for each closed two-sided ideal I of A not equal to A. When this is the case, A is said to be a C*-algebra with generalised continuous trace. (c) Suppose that A has generalised continuous trace. For each P < a, 11+1\11 is an open dense subset of A\1", and each point of 11+1\11 has a base of closed neighbourhoods in A\1". The algebra A is liminal (use

CH. 4. 7l

ADDENDA

III

4.7.15), but a liminal algebra does not always have generalised con tinuous trace. (4.7.9). (d) Let A be a postliminal C*-algebra, let a be an ordinal and let (Vp)OEp,.;a be an increasing family of open subsets of A with the follow ing properties: (10) V o = 0, Va = A; (2) if p;;;a is a limit ordinal, V p = U p<p V p. ; A(3) each point of VP+l\ V p has a base of closed neigh bourhoods in A \ v; Then A is a C*-algebra with generalised continuous trace. In par ticular, if A is Hausdorff, A has generalised continuous trace. [452]. 4.7.13. Let A be a C*-algebra with continuous trace, and let x be an element of A such that the rank of 1T(x) is finite and bounded as 1T varies in A. Then 1T --+Tr 1T(x) is continuous on A. [584]. 4.7.14. (a) If A is a liminal C*-algebra, the C*-algebra A obtained by adjoining an identity to A is not always liminal. (Example: A =~(H) with H an infinite-dimensional Hilbert space; in this example, the spectrum of A consists of two points 1T1, 1T2, the closed sets being 0, {1T[}, {1Tt, 1TJ.) (b) If a liminal C*-algebra A is unital, all its irreducible represen tations are finite-dimensional. [896]. 4.7.15. Let A be a postliminal C*-algebra. Then A is liminal if and only if each point of A is closed. (Use 4.3.7(i. [631]. 4.7.16. Let A be a C*-algebra. (a) Each weak*-limit of pure states is proportional to a pure state if and only if the following conditions are satisfied: (i) A is liminal; (ii) A is Hausdorff; (iii) AIJ(A) (4.7.12) is commutative. [631]. (b) Suppose that A is unital and does not have any non-trivial, irreducible l-dirnensional representations. Then P (A) is weak*-closed in the dual of A, if and only if A is the product of a finite number of , An with the following property: for each 1T E Ai C*-algebras A lo dim 1T is finite and independent of 1T. [1708]. 4.7.17. Let a be an ordinal and let E be the set of mappings f: [0, a]--+N such that f(p) = for all but a finite number of p. Let H be the Hilbert space L~ which admits a canonical orthonormal basis (E/)/EE If S ~ [0, a], let E(S) be the set of fEE which vanish outside S. Let p .;;; a. We identify E with E([O, pD x E[p, aD. For g E E([O, pD, let be the Hilbert subspace of H with orthonormal basis (E(R,hl)hEE([p.a)); for g,

n,

112

LIMINAL C*-ALGEBRAS

lCH. 4, 7

g E E([O, p[), let Ug,g. be the isomorphism of HK onto H g. which maps E(K,h) to E(g,h) for each hE E([p, a]). Let B p be the set of x E .P(H) such that: (10) for each g E E([O, p[), x(HK ) HK ; (2) the x I are compact

n,

and are mapped to the corresponding operators on H g. by Ug,g" Let C; be the norm closure of Lu<pBu. Then the C, are postliminal. For Ce" the sequence of ideals of 4.3.3 is (Cp)o,""p,""". 4.7.18. Let H be a Hilbert space, and let A be a commutative von Neumann algebra on H not contained in ::eee(H). Then B = ::eee(H) + A is postliminal and not liminal. 4.7.19. Let A be the C*-algebra of complex 2 x2 matrices, and let B be the C*-algebra of sequences x = (x" X2 ) of elements of A such that x, converges to a diagonal matrix
A(X) (

/L(x)

0)

as n ~ + 00. Then 13 is the set of mappings x ~ x n , X ~ A(x), x ~ /L(x). The topological space B is not Hausdorff. The algebra B has generalised continuous trace (4.7.12). [893]. 4.7.20. Let A be a C*-algebra. For every subset M of A, we denote by R(M) (resp. L(M the set of x E A such that Mx = 0 (resp. xM = 0). The following conditions are equivalent: (i) (resp. (i): For every closed left (resp. right) ideal I of A, we have L(R(I = I (resp. R(L(I = 1); (ii) (resp. (iij):The sum of the minimal left (resp. right) ideals of A is dense in A; (iii) A is isomorphic to a sub-C*-algebra of an algebra ::eee(H); (iv) There exists a family (AJ of elementary C*-algebras such that A is isomorphic to the restricted product of the Ai; (v) (resp. (vj): For each x E A, left (resp. right) multiplication by x is a weakly compact operator on A; (vi) The spectrum of every maximal commutative sub-C*-algebra of A is discrete; (vii) For each hermitian x in A, every non-zero point of Spx is isolated in Spx. Such a C*-algebra is said to be dual. A dual C*-algebra is liminal. [165], [896], [1164], [1165]. 4.7.21. Let A be a C*-algebra. Left and right multiplication by the

CH. 4. 7j

ADDENDA

113

elements of A are all compact operators on A if and only if there exists a family (Ai) of finite-dimensional C*-algebras such that A is isomorphic to the restricted product of the A j [893]. 4.7.22. Let H be a separable infinite-dimensional Hilbert space. (a) The only closed two-sided ideals of .:(H) are 0, .Te(H), .:(H). (b) The C*-algebra .:(H)/.Te(H) is antiliminal, non-separable, non isomorphic to a von Neumann algebra, and contains a family of c orthogonal projections, where c is the power of the continuum. (c) Every representation of .:(H) may be written in the form 1f EB (EB 1fJ where tt vanishes on .Te(H) and each 1fj is equivalent to the identical representation of .:(H). Moreover, either H" = 0 or H" is non-separable. Consequently, there is no sub-C*-algebra of .:(H) complementary to .:cg(H). (In fact, .:cg(H) does not even have a complementary closed subspace in .:(H. (d) Card (.:(H() = 2 (use 2.10.1, 2.8.6 and a commutative sub-C* algebra A of .:(H) with card A = 2). [578], [581], [876], [900], [1097], [1617]. 4.7.23. Let A be a unital C*-algebra with minimal right ideals and a smallest non-zero closed two-sided ideal. For every subset P of A, let P be the set of x E A such that xP = O. Let 513 be the set of P as P runs over the set of subsets of A. Suppose that, if it, 1 2 E 513 are such that 1 111 = 0, then J 1 + J 2 E 513. Under these conditions, A is isomorphic to a C*-algebra .:(H). [1833].
I I

4.7.24. Let A be a C*-algebra with continuous trace. The set of those x E A such that 1f(x) is of bounded rank over A and vanishes outside some compact subset of A is not necessarily the smallest self-adjoint two-sided ideal m of A such that 1ft = A. [1256]. 4.7.25. Problem: construct a liminal C*-algebra A which does not admit any finite composition sequence (Ip) such that the (Ip+.t I p) are Haus dorff.

CHAPTER 5

THE TYPE OF_A REPRESENTATION

5.1. Comparison of representations and comparison of projections

5.1.1. Given a family (7TJ of representations of A, t we can always regard the 11j as subrepresentations of a single representation p of A (for example p = EB 11;). Let E, be the projection of p(A) corresponding to 11j. As we shall see, the relations between the 7Tj are often interpreted as relations between the projections E, in the von Neumann algebra
p(A).

5.1.2. PROPOSITION. Let p be a representation of A, ~ the von Neumann algebra generated by p(A) and 11 and 11" two subrepresentations of p, corresponding to two projections E, E of ~'. Let T be a continuous linear mapping of E(Hp ) into E(Hp ) . Then T is an intertwining operator for 11 and 11" if and only if T . E E ~'. Suppose that T is an intertwining operator for 11 and 11". Let x E A and ~ E H p We will show that TEp(x)~ =p(x)TE~, from which it will follow that TE E f!JJ. If ~ E E(Hp ) , we have p(x)~ E E(Hp ) , hence
TEp(x)~

= Tp(x)~

= T7T(X)~

= 11"(x)n = p(x)T~

= p(x)TEf

If ~ E (1- E)(Hp ) we have p(x)~ E (1- E)(Hp ) and so Ep(x)~ = E~ = 0, hence TEp(x)~ = 0 = p(x)TE~. Conversely, if TE E ::;~', we have, for each x E A and each g E E(H... ),
T7T(X)~

= TEp(x)~

= p(x)TE~
11

11"(x)n,
11".

so that T is an intertwining operator for

and

5.1.3. COROLLARY. We retain the notation p, f!JJ, 11, 11", E, E of 5.1.2. Then 11 = 11" if and only if E - E relative to the von Neumann algebra f!JJ1 (A 41). In fact,
11

= 11" means that there exists an intertwining operator for

11

tThroughout this chapter, A denotes an involutive algebra.

CH. 5. 2]

DISJUNCTION

115

and 11" which is an isomorphism of E(Hp ) onto E(Hp ) , hence (5.1.2) there exists a partial isometry of ~' whose initial and final projections are E and E. 5.1.4. COROLLARY. We again retain the notation p, ~, 11, 11", E, E of 5.1.2. Then 11 ~ 11" if and only if E < E relative to the von Neumann algebra ~' (A 41). In fact, 11 ~ 11" means that there exists a projection E; E ~' dominated by E such that 11 is equivalent to the subrepresentation of p defined by E;, i.e. (5.1.3) such that E ~ E; relative to ~'. 5.1.5. COROLLARY. Let 11 and 11( be two representations of A. If 11 ~ and 111 ~ 11 then we have 11 = 111 This follows from 5.1.1, 5.1.3, 5.1.4 and A42. References: [1005], [1006]. 5.2. Disjunction 5.2.1. PROf OSITION. Let 11 and 111 be two representations of A. Then the following conditions are equivalent: (i) The only intertwining operator for 11 and 11( is 0; (ii) If T (resp. Tl) is a nontrivial sub-representation of 11 (resp. 111) then T and T\ are inequivalent. Suppose further that 11 and 111 are both subrepresentations of the same representation p of A. Let ~ be the von Neumann algebra generated by p(A), and E and E( the projections of~' corresponding to 11 and 11]. Then conditions (i) and (ii) are again equivalent to the following: (iii) The central supports of E and E] in ~' are orthogonal. We can suppose from the outset that 11 and 111 are subrepresentations of p (5.1.1). We will prove that (i) (ii) (iii) (i). (i) (ii): if 11 and 111 possess equivalent non-trivial subrepresentations 0" and 0"\, there exists a non-zero intertwining operator for 11 and 111 that is to say a partial isometry with H; and H U J as initial and final subspaces. (ii) (iii): if the central supports of E and E( are not orthogonal, there exist non-zero projections F, F I E ~', equivalent relative to ~', with F ~ E, F I ~ E 1 (A 44). The corresponding subrepresentations of 11 and 11( are equivalent (5.1.3).
11(

=>

=> =>

=>

=>

116

THE TYPE OF A REPRESENTATION

[CH. 5, 2

(iii) =} (i): suppose that the central supports G, 0 1 of E, E 1 are orthogonal. Let T be an intertwining operator for 7T and 7TI, and let S = TE E 00 Ker S and on S(Hp ) belong to 00and (5.1.2). The projections on H p are equivalent relative to 00 (A 45), and therefore have the same central support (A 43). Now the projection on H p Ker S is dominated by E, and therefore by G, and the projection on S(Hp ) is dominated by E 1, and therefore by G I. Since GG I = 0, we see that S(Hp ) = 0, S = 0, T = 0.

5.2.2. DEFINITION. If two representations 7T, 7TI of A satisfy the equivalent conditions of 5.2.1, we say that 7T and 7T I are disjoint. We then write 7TO 7TI This relation is symmetric. If 7T and 7TI are topologically irreducible, to say that 7T and 7TI are disjoint simply means that they are inequivalent (2.3.4). If 7T and 7TI are finite-dimensional representations, to say that 7T and 7TI are disjoint means that the same irreducible representation cannot appear in the decompositions into irreducible representations of both 7T and 7TI (this follows from condition (ii) of 5.2.1). 5.2.3. PROPOSITION. Let 7T be a representation of A, and (7T;)iEI a family of representations of A. If, for each i E I, 7T and 7Ti are disjoint, then 7T and EEl 7Ti are disjoint.

In fact, let H, be the space of 7Tj, H the direct sum of the Hi and P, the projection of H onto Hi Let T be an intertwining operator for 7T and EEl 7Ti For each i, PiT is an intertwining operator for 7T and 7Tj, and hence is zero. Hence T = O. Hence 7T and EEl 7Tj are disjoint.
5.2.4. PROPOSITION. Let p be a representation of A, 00 the von Neumann algebra generated by p(A), and E a projection of 00. Let 7T, 7T1 be the sub representations of p corresponding to E, 1 - E. Then 7T and 7TI are disjoint if and only if E belongs to the centre of 00. Let G, G I be the central supports of E, 1- E. We have G 1 ~ 1- E, and therefore:
7T07TI~GGI=O~O=E
G ~

E,

and

G I= l-E~E

is in the centre of 00.

5.2.5. COROLLARY. Let p be a representation of A, and 00 the von Neumann algebra generated by p(A). Then the following conditions are equivalent: (i) It is not possible to express p as the direct sum of two non-trivial disjoint sub representations ;

CH. 5. 2j

DISJUNCTION

117

(ii) (iii)

is a factor; =p(A) is a factor.

By 5.2.4, condition (i) is equivalent to saying that the only projections in the common centre of @ and @ are 0 and 1. i.e. that this centre is just the set of scalar operators. 5.2.6. DEFINITION. A representation of A which satisfies the conditions of 5.2.5 is said to be a factor representation. By 2.2.6. such a represen tation is either trivial or else non-degenerate.
If p is topologically irreducible, p is clearly a factor representation. The converse is not true: for example, a direct sum of equivalent irreducible representations is a factor representation (5.3.4); in the case of finite dimensional representations, this is moreover, the most general example of a factor representation (this follows from condition (i) of 5.2.5).

5.2.7. PROPOSITION. Let p be a representation of A, !YJ the von Neumann algebra generated by p(A), q the centre of !YJ, E a non-zero projection in @, 7T the subrepresentation corresponding to p, and F the central support of E. Then 7T is a factor representation, if and only if F is a minimal projection of q. The von Neumann algebra generated by 7T(A) is the induced von Neumann algebra !YJE (A 15), hence is isomorphic to !YJ p (A 20), and the centre of !YJ p is qp (A 15). Hence 7T is a factor representation if and only if qp is just the scalar operators, i.e. (A 34) if and only if F is a minimal projection of q. 5.2.8. Let p be a representation of A, !YJ the von Neumann algebra generated by p(A) and q the centre of @. To decompose p into a direct sum of topologically irreducible representations amounts to writing the operator I on H; as the sum of mutually orthogonal minimal projections of @. To decompose p into a direct sum of mutually disjoint factor representations amounts to writing 1 as the sum of mutually orthogonal minimal projections of q, by 5.2.4 and 5.2.7. The second problem. involving q, is, in principle, easier to solve than the first, which involves
@.

Let (Fj) be the family of minimal projections (mutually distinct and therefore mutually orthogonal) of q. If I = L F, we have a decom position of p into a direct sum of mutually disjoint factor represen tations. The problem of writing p as a direct sum of topologically

118

THE TYPE OF A REPRESENTATION

[CH. 5, 3

irreducible representations is thus reduced to the case where p is a factor representation. We shall see (5.4.11) that, in the important special case of representations of type I, this problem is then quite easy to solve. (In the case of representations which are not of type I, the problem, even so reduced, is, on the contrary, very difficult). In general, however, we have F = ~ Fj:f. 1. The above decomposition only yields a subrepresentation of p; and it can happen, unfortunately, that F = O. We will in due course obtain a satisfactory theory by replacing the decomposition of p into a direct sum of mutually disjoint factor representations with a decomposition into a direct integral.
5.2.9 PROPOSITION. Let 1T and 1T1 be factor representations of A. Then one of the relations 1T01TJ, 1T ~ 1T1, 1T. ~ 1T holds.

We can regard 1T and 1T1 as subrepresentations of a single represen tation p of A. Let 00 be the von Neumann algebra generated by p(A), E and E J the projections of 00 corresponding to 1T and 1T1, and F and F I their central supports. These are minimal or zero projections of 00 n 00 (5.2.7), and so F and F I are either orthogonal (in which case 1TO 1TJ) or else equal. In this case, we have, relative to the factor oo~, either E < E. or else E( < E (A 46); hence 1T ~ 1T1 or 1TJ ~ 1T (5.1.4). References: [1005], [1006].

5.3. Quasi-equivalence

5.3.1. PROPOSITION. Let 1T and 1TJ be two representations of A, and 00 and 00\ the von Neumann algebras generated by 1T(A) and 1T1(A). Then the following conditions are equivalent: (i) No non-trivial subrepresentation of 1T is disjoint from 1T1; no non-trivial subrepresentation of 1T1 is disjoint from 1T; (ii) There exists an isomorphism <P of 00 onto 00 1 such that 1Tlx) = <P(1T(x for each x E A; (iii) There exists a multiple 1T2 of 1T and, in 1T2(A), a projection of central support 1, such that the corresponding subrepresentation of 1T2 is equivalent to 1T.; (iv) There exist multiples of 1T and of 7T1 which are equivalent. Suppose further, that 1T and 1TJ are subrepresentations of the same representation p of A. Let Cf5 be the von Neumann algebra generated by p(A), and E and E I the projections of Cf5 corresponding to 1T and 1T1

CH. 5. 3]

QUASI-EQUIV ALENCE

119

Then conditions (i)-(iv) are again equivalent to the following condition: (v) The central supports F, F, of E, E 1 are equal.
(ii) ~ (iii): by A 22, <P is the composite of an ampliation 1/ of gJ onto a von Neumann algebra gJ2, an induction 1f!" of gJ2 onto 00 1 defined by a projection E E gJ2 of central support 1, and a spatial isomorphism 1Jr". Put 7T2(X) = 1/(7T(X for each x E A. Then 7T2 is a multiple of the representation 7T of A. In addition,

is the operator induced on E(H"2) by 7T2(X) and hence 7T, is equivalent to the subrepresentation of 7T2 defined by E. (iii) ~ (ii): suppose condition (iii) is satisfied. Let gJ2 be the von Neumann algebra generated by 7T2(A). There clearly exists an isomor phism of gJ onto gJ2 mapping 7T(X) to 7T2(X) for each x E A. Further more, if E E gJ2 is a projection of central support I, the induction of gJ2 onto (gJ~E is an isomorphism (A 20). Hence there exists an isomorphism of gJ2 onto gJ .. which maps 7T2(X) to 7T.(X) for each x E A. (iii) ~ (iv): again suppose condition (iii) is satisfied; there exists a cardinal number n such that 7T1 ~ n7T and we can assume that n is infinite. Since (iii) ::> (ii), condition (iii) is symmetric with respect to 7T and 7T1 and so there also exists an infinite cardinal nl such that 7T ~ nl7T,. Let p = sup(n, n,). We have p7T ~ pnvtr, = p7T1 ~ pntr = p7T. and therefore
P7T. = p7T (5.1.5).

(iv) ~ (i): suppose that p7T = P7TJ. Let T be a non-trivial subrepresen tation of -n:. Then T is not disjoint from p7T = P7TI, hence 7" is not disjoint from 7T, (5.2.3). Similarly, a non-trivial subrepresentation of 7Tt is not disjoint from 7T. Now suppose that 7T and 7T1 are subrepresentations of p. (i) ~ (v): if F,t- F,. there is a non-zero projection G in the centre of C(5 which is dominated by F and orthogonal to F I (interchanging F and F I if necessary). We have GE,t- 0 (since otherwise. the central support of E would be orthogonal to G). and GE is orthogonal to F I , so that the central support of GE is orthogonal to F,. Hence there exists a non-trivial subrepresentation of 7T which is disjoint from 7T,. (v) ~ (ii): the mapping S ~ SE (resp. T ~ T E) of C(5F onto C(5E (resp. of C(5F, onto C(5E) is an isomorphism (A 20). Hence, if F = F" there exists an isomorphism of C(5E = gJ onto C(5EI = gJ. which maps P(X)E = 7T(X) to P(X)E, = 7T,(X) for each x E A.

120

THE TYPE OF A REPRESENTATION

[CH. 5, 3
7T

5.3.2. DEFINITION. If the equivalent conditions of 5.3.1 are satisfied, and 7TI are said to be quasi-equivalent. We then write 7T =7TI

In view of condition (ii) of 5.3.1, it is clear that we are dealing with an equivalence relation. If 7T and 7TI are finite-dimensional, 7T = 7TI means that the same irreducible representations occur in the decomposition of 7T and 7TI into their irreducible components, although not necessarily with the same multiplicity (this follows from condition (i) of 5.3.1). 5.3.3. PROPOSITION. (i) If 1T = 7TI, then 7T = 7Tl (ii) If 7T and 7TI are topologically irreducible, and if
7T

7T1

then

7T

=7TI

(i) is clear. If 7T and 7TI are topologically irreducible and quasi equivalent, then 7T and 7TI are non-disjoint (condition (i) of 5.3.1), and therefore equivalent (condition (ii) of 5.2.1). 5.3.4. PROPOSITION. Let 7T and 7T1 be two quasi-equivalent represen tations of A. Then if 7T is a factor representation, so is 7T1 This follows immediately from condition (ii) of 5.3.1 and condition (ii) of 5.2.2. For irreducible representations, quasi-equivalence and equivalence coincide (5.3.3). In the class of factor representations, quasi-equivalence seems to be the appropriate equivalence relation. 5.3.5. PROPOSITION. Let 1T be a factor representation of A. Then every non-trivial subrepresentation of A is quasi-equivalent to 7T. Since the only non-zero projection in the centre of 7T(A) is 1, this follows from condition (iii) of 5.3.1. 5.3.6. COROLLARY. Two factor representations of A are either quasi equivalent or else disjoint. This follows from 5.2.9 and 5.3.5. 5.3.7. COROLLARY. Let A be a separable involutive Banach algebra. Then every factor representation of A is quasi-equivalent to a factor representation in a separable space. Let 7T be a factor representation of A, and ~ a non-zero vector of H". Then the closure K of 7T(A)~ is separable and invariant under 7T(A). The subrepresentation of 7T defined by K is quasi-equivalent to 7T (5.3.5). 5.3.8. Let
1T,

1T1

be two quasi-equivalent representations of A in a

CH. 5, 4j

REPRESENTATIONS

OF TYPE I

121

separable space. We have: (\0 7T =(\0 7T1: consider the proof of 5.3.1, (ii) =?(iii) =?(iv), and use A 22. The current notation for quasi-equivalence is 7T ~ 7Tl Since this conflicts with the notation E ~ E I for projections, it seemed preferable to alter it slightly. References: [l005], [1006].

5.4. Representations of type I 5.4.1. PROPOSITION. Let 7T be a representation of A. Then the following conditions are equivalent: (i) The von Neumann algebra generated by 7T(A) is of type I; (ii) The von Neumann algebra 7T(A) is of type I; (iii) 7T is quasi-equivalent to a representation 7Tl such that 7Tl(A) is commutative (iii) =?(i): suppose condition (iii) is satisfied. Since 7Tl(A) is com mutative, 1T(A) generates a von Neumann algebra of type I (A 35), and hence 1T(A) generates a von Neumann of type I (condition (ii) of 5.3.1). (i) =?(ii): follows from A 52. (ii) =?(iii): suppose condition (ii) is satisfied. Then there exists a projection E E 1T(A), with central support 1, such that 1T(A)1 is com mutative (A 35). The subrepresentation 7TI of 1T defined by E is quasi equivalent to 1T (condition (iii) of 5.3.1), and 1T1(A)= 1T(A)1 is com mutative. 5.4.2. DEFINITION. If the representation 7T satisfies the equivalent con ditions of 5.4.1, then 7T is said to be of type I. Every finite-dimensional representation is of type I. Indeed, in a finite-dimensional Hilbert space, every von Neumann algebra is of type I. 5.4.3. Every representation quasi-equivalent to a representation of type I is of type I. Every subrepresentation of a representation of type I is of type I (because every von Neumann algebra induced by a von Neumann algebraof type I is of type I (A 53. Every direct sum p of representations 1Ti of type I is of type I; in fact, let ~ be the von Neumann algebragenerated by peA), E, E ~' the projection corresponding to 7Ti and F, its central support; then ~Ei is of type I, hence ~Fi' isomorphic to ~Ei (A 20), is of type

122

THE TYPE OF A REPRESENTATION

[CH. 5. 4

I, so that the greatestprojection F of i?lJ n i?lJ such that i?lJ F is of type I (A 39) is equal to 1 since sup F; = 1. 5.4.4. PROPOSITION. Let 7T be a representation of A. The following conditions are equivalent: (i) 7T(A) is commutative; (ii) For every decomposition 7T = 7TI E9 7T2 7T1 and 7T2 are disjoint. By 5.2.4, condition (ii) means that every projection of 7T(A) belongs to the centre of 7T(A). and hence 7T(A) is commutative. 5.4.5. DEFINITION. If 1T satisfies the equivalent conditions of 5.4.4, 1T is said to be multiplicity-free. Let
Pt1T. EB 1T

be

finite-dimensional

... E9 Pn1Tn the decomposition of

representation, and 1T = 1T into multiples of inequivalent

irreducible representations 7T1, , 7Tn Then 7T is multiplicity-free if and only if PI = ... = Pn = I (by condition (ii) of 5.4.4). 5.4.6. A multiplicity-free representation is of type I. A representation of type I is quasi-equivalent to a multiplicity-free representation 1T (con dition (iii) of 5.4.1); moreover, 1T is uniquely determined by 1T up to equivalence; in fact. we have the following result which generalises 5.3.3 (ii):
PROPOSITION. Let 1T, P be two multiplicty-free representations of A. If 1T = o, then we have 1T = p.

Let (E; F;)jEl be a maximal family among families with the following properties: (10) E, E 1T(A), F; E p(A); (2) the E, are mutually or thogonal non-zero projections and the F, are also mutually orthogonal non-zero projections; (30) the subrepresentation of 1T defined by E, is equivalent to the subrepresentation of p defined by F; Let E =~ E j , F = ~ F; Then the subrepresentation 1T) of 1T defined by E is equivalent to the subrepresentation PI of p defined by F. Suppose E" 1. Since 1T = p, there is a non-zero projection E E 1T(A) dominated by 1 - E and a non-zero projection F E p(A) such that the subrepresentations 1T/, p of 1T and p defined by E and F are equivalent (condition (i) of 5.3.1). Since p(A) is commutative, we have F = F; + F 2, F; and F 2 being projections of p(A) such that F;,,;;; F, F 2,,;;; I-F. Hence there exist projections E;, E 2 of 1T(A) such that E = E; + E 2 and such that the subrepresentations of 1T and p defined by E; and F;

CH. 5. 4j

REPRESENTATIONS

OF TYPE I

123

are equivalent (j == 1,2). The maximality of the family (E; F;) implies that F~ == E~ == O. Then F ~ F, and so 1T" is equivalent to a subrepresen tation of PI and hence to a subrepresentation of 1T,. Now this con tradicts condition (ii) of 5.4.4. Hence E == 1, and similarly F == l. Hence 1T "" p. 5.4.7. PROPOSITION. Let 1T be a representation of A and n a cardinal number. Then the following conditions are equivalent: (i) 1T(A) is a von Neumann algebra of type In (A 47); (ii) 1T is the direct sum of n equivalent multiplicity-free representations. Suppose condition (ii) is satisfied: then there exists a multiplicity-free representation P of A such that 1T == np. Then H 1T may be identified with the Hilbert space H; l8l K (K an n-dimensional Hilbert space), in such a way that 1T(x) is identified with p(x) l8l1 for each x E A (A 17). Then 1T(A) == p(A) l8l 2(K) (A 18), and p(A) is commutative, so that 1T(A) is of type In (A 47). The converse is proved in an analogous way. 5.4.8. DEFINITION. If 1T satisfies the equivalent conditions of 5.4.7, 1T is said to be of multiplicity n. A representation of multiplicity n is of type I. A multiplicity-free representation is just a representation of multiplicity l. The cardinal n of 5.4.7 is uniquely determined by 1T (A 47). Moreover, if 1T "" n1T" with 1T" multiplicity-free, the class of 1T" is uniquely determined by 1T, because. if 1T "" n1Tj, with 1Tj multiplicity-free, we have 1T" =1T; (condition (iv) of 5.3.1), hence 1T" "" 1Tj (5.4.6). Let 1T be a finite-dimensional representation, and 1T == PI1T1 ffi ... ffi P,1T, the decomposition of 1T into multiples of inequivalent irreducible representations 1T1, ... , 1Tr Then 1T is of multiplicity n, if and only if PI == . . == p, == n (by condition (ii) of 5.4.7). 5.4.9. PROPOSITION. Let 1T be a representation of A of type I. Then there exists a family (Ej)jEJ of mutually orthogonal, non-zero projections of 1T(A), with sum 1, such that: (a) the subrepresentations 1Tj of 1T defined by the E, are mutually disjoint; (b) 1Tj is of multiplicity nj; (c) the nj are mutually distinct. The family (Ej)jEJ is unique up to a permutation of the set of indices. The E j belong to the centre of 1T(A). Let :!l be the centre of 1T(A). In view of 5.2.4, the proposition means

124

THE TYPE OF A REPRESENTATION

[CH. 5, 4

that there exists a family (Ej)jEJ of mutually orthogonal, non-zero projections of ~, with sum 1, such that the von Neumann algebras 1T(A)Ej are of type I n j with the nj mutually distinct; and that this family (E) is unique up to a permutation of the set of indices. Since 1T(A)is of type I (condition (ii) of 5.4.1), this follows from A 50. The above completely reduces the study of representations of type I to that of multiplicity-free representations. 5.4.10. We retain the notation of 5.4.9. Let U be a unitary operator on H", such that U7T(A) U- I = 7T(A). Then U~U-l =~, and so the Ej = UEjU- 1 are mutually orthogonal non-zero projections of ~, with sum 1, such that '1T(A)~ is of type I n r Hence Ej = E; In other words, U commutes with the E j 5.4.11.
PROPOSITION. Let 7T be a representation of A. Then the following conditions are equivalent: (i) 7T is a factor representation of type I; (ii) 7T is quasi-equivalent to a topologically irreducible representation; (iii) 7T is of the form n7T with 7T topologically irreducible. Under these conditions, 7T is of multiplicity n.

(iii) ~ (ii): follows from condition (iv) of 5.3.1. (ii) ~ (i): follows from 5.3.4 and condition (iii) of 5.4.1. (i) ~ (iii): if 7T is a factor representation of type I, 5.4.9 proves that 7T is of multiplicity n for a certain cardinal n. Hence 7T = np with p multiplicity-free (condition (ii) of 5.4.7), and is a factor representation, being quasi-equivalent to 7T. Then p(A) is a commutative factor, and is therefore just the scalar operators, from which it follows that p is topologically irreducible. 5.4.12. Let 7T be a representation of A and ~ the centre of 7T(A). We have seen that 7T is the direct sum of mutually disjoint factor represen tations if and only if the operator 1 is the sum of minimal projections of ~. Although this is far from being the general situation, it is nevertheless worthwhile to interpret definitions 5.4.2, 5.4.5 and 5.4.8 in this case.
PROPOSITION. Let (7Ti) be a family of mutually disjoint factor re presentations of A, and let 7T =EB 7Ti (i) 7T is of type I if and only if each 7Ti is of type 1. (ii) 7T is of multiplicity n if and only if each 7Ti is of multiplicity n.

CH. 5, 4j

REPRESENTATIONS

OF TYPE I

125
11j

(iii) 11 is multiplicity-free if and only if each reducible

is topologically ir

(i) follows from 5.4.3; (ii) follows from the fact that a product of von Neumann algebras 9)lj is of type In if and only if each 9)lj is of type In (A 50), (iii) is a special case of (ii).
5.4.13. We have seen (2.3.5) that a finite-dimensional representation is the direct sum of irreducible representations. More generally we have:
PROPOSITION. Let 11 be a representation of A such that 11(x) is compact for every x E A. Then (i) The representation 11 is the direct sum of a family (11j)jEI of topologically irreducible representations. (ii) Let i E I be such that 11j(A) T- O. Then there is only a finite number of indices j E I such that 11j== 11j.

We show that, if H : 0,11 possesses an irreducible subrepresentation. We can suppose that 11: O. Let 00 be the von Neumann algebra generated by 11(A), and let f!l be its centre. There exists an hermitian x in A such that 11(x) : O. Since 11(x) is compact, 11(x) has a non-zero eigenvalue and the corresponding spectral projection E is non-zero and of finite rank. We have E E 00. Let F be a projection of minimal rank among the non-zero projections of 00 dominated by E. Then F is a minimal projection of 00, and its central support F is a minimal projection of ~. The induced von Neumann algebra 9)lp is then a factor which possesses a minimal projection, and is therefore a factor of type I
7T

(A 36).

Thus, 11 possesses a non-zero factor subrepresentation of type I, and consequently (5.4.11) a non-zero irreducible subrepresentation. This established, let (E;)jEI be a maximal family of mutually or thogonal non-zero projections of 11(A), such that the corresponding subrepresentations of 11 are irreducible. If ~ E j : 1, let 11" be the subre presentation of 11 corresponding to 1 - ~ E; For each x E A, 11"(x) is compact. Then the preceding result, applied to 11", contradicts the maximality of (E j ) . Hence ~ E, = 1, which proves (i). If x E A and i E I are such that 11j(x) ;f; 0, and if there are infinitely many indices j E I such that 11j== 11j, then it is clear that 11j(x) is non-compact. This proves (ii). References: [896], [1005], [1006].

126

THE TYPE OF A REPRESENTATION

lCH. 5, 5

5.5. Involutive algebras of type I

5.5.1. We will see that the decomposition of representations is more easily accomplished in the case of representations of type I. This explains the interest of the following notion. DEFINITION. An involutive algebra is said to be of type I if all its representations are of type I. For algebras of type I, the notion of factor representation can there fore be set aside in favour of the more natural idea of topologically irreducible representation (5.4.11). Unfortunately, there exist C*-alge bras which are not of type I, and, for these, the study of factor representations sometimes leads to simpler theorems than that of to pologically irreducible representations. 5.5.2. THEOREM. Every post/iminal C*-algebra is of type 1. Let A be a non-zero postliminal C*-algebra. By 4.4.4, there exists a non-zero x E A + such that 1T(X) is either zero or of rank 1 for each 1T E A. Then 1T(xAx) = 1T(X)1T(A)1T(X) is commutative for each 1T E A, and therefore xAx is commutative. This established, let 1T be a representation of A. Let (EJ be a maximal family of mutually orthogonal non-zero projections in the centre of 1T(A) such that the corresponding subrepresentations of 1T are of type I. By 5.4.3, it is enough to prove that L E, = 1. Suppose that E = 1- L E, and let p be the subrepresentation of 1T defined by E. If we had that p(A) = 0, we could adjoin E to the family (E j ) , contradicting the maximality of this family. Hence p(A) f. 0. Moreover, p(A) is postli minal (4.3.5). From the first part of the proof, there exists an x E A + such that p(x) is non-zero and p(x)p(A)p(x) is commutative. If <t6 denotes the von Neumann algebra generated by p(A), p(x)<t6p(x) is oo commutative. Let p(x) = f o + A de, be the spectral decomposition of p(x). We have eA E <t6 for each A, and hence Y. = A -I de, E <t6 for each > 0. Provided is chosen small enough, p(x)y. = Y.P(x) is a non-zero projection P of <t6. We have FV6P ~ p(x)<t6p(x), and so FV6P is com mutative. Let G E <:g n <t6 be the central support of F. The von Neumann algebra <t6o is of type I (A 35). Then the projection on H which vanishes on (1- E)(H and coincides with G on E(H is a non-zero element of the centre of 1T(A) orthogonal to the E j , and the cor responding subrepresentation is of type 1. This contradicts the maxi mality of (Ej ) . Hence L E, = 1, and 1T is of type I.

roo

1T

1T

1T

CH. 5. 6l

REPRESENTATIONS

OF TYPES II AND III

127

In due course (9.1) we shall present a converse to theorem 5.5.2. 5.5.3. COROLLARY. Let A be a postliminal C*-algebra. Then every fac tor representation of A is quasi-equivalent to an irreducible represen tation. This follows from 5.5.2 and 5.4.11. References: [896], [897].
5.6. Representations of types II and III

5.6.1. As well as von Neumann algebras of type I, there exist the more complicated von Neumann algebras of type II or of type III (A 38). A representation 7T of A is said to be of type II (resp. of type III) if the von Neumann algebra generated by 7T(A) is of type II (resp. III). This is equivalent to saying that 7T(A) is a von Neumann algebra of type II (resp. III) (A 52). 5.6.2. Every representation quasi-equivalent to a representation of type II (resp. III) is of type II (resp. III). Every subrepresentation of a representation of type II (resp. III) is of type II (resp. III). Every direct sum of representations of type II (resp. III) is of type II (resp. III). These facts follow using the same arguments that were used for re presentations of type I. 5.6.3. A non-trivial representation cannot be simultaneously of two different types (A 38). It follows from this that two representations of two distinct types are disjoint. 5.6.4. PROPOSITION. Let 7T be a representation of A. Then there exist projections E b E lb E Ill E 7T(A) uniquely determined by the following conditions: (1) EI> Ell E Ill are mutually orthogonal and with sum 1; (2) the subrepresentations of 7T defined by EI> E lb Em are respectively of types I, II, III. These projections belong to the centre of 7T(A). This follows from the corresponding property of von Neumann alge bras (A 39). The study of arbitrary representations of A is thus reduced to that of representations of type I, of representations of type II, and of represen tations of type III. However, this does not lead to any striking sim plification.

128

THE TYPE OF A REPRESENTATION

[CH. 5, 7

5.6.5. PROPOSITION. Let 7T be a representation of A. Then 7T is the direct sum of a representation of type II and a representation of type III if and only if every subrepresentation of 7T is the direct sum of two equivalent representations. Now 7T is the direct sum of a representation of type II and a representation of type III if and only if the projection E, of 5.6.4 is non-zero, in other words, if and only if the von Neumann algebra 7T(A) is continuous (A 39). This is the case if and only if every projection of 7T(A) is the sum of two equivalent orthogonal projections of 7T(A) (A 48). It is then enough to apply 5.1.3. 5.6.6. Let 7T be a representationof A of type I, and C the quasi-equivalence class of 7T. The different equivalence classes contained in C may be distinguished using multiplicities (5.4.9). For representations not of type I, there is an analogous although more complicated theory that we will not go into here. Nevertheless we do have the result, simple enough to be given here:
Let 7TI 7T2 be two representations of A of type III in separable spaces. Then 7TI = 7T2 implies 7T, = 7T2
PROPOSITION.

In fact, let ~I' ~2 be the von Neumann algebras generated by 7TI(A) and 7T2(A). If 1T, =1T2 there is an isomorphism 4J of ~, onto ~2 which maps 7T,(X) to 7T2(X) for each x E A. Now ~" ~2 are of type III and operate in separable spaces. Hence 4J is defined by an isomorphism of H fr 1 onto H"2
(A 51).

5.6.7. The classification of von Neumann algebras includes the definition of von Neumann algebras of type II" of type II"", semi-finite von Neumann algebras, etc. A representation 7T of A is said to be of type II" of type II"", of finite type, etc. if 7T(A) generates a von Neumann algebra of type II" of type II"", a finite von Neumann algebra, etc. References: [1005], [1006]. 5.7. Addenda 5.7.1. A finite-dimensional C*-algebra A is the product of a finite number of C*-algebras .:(HI ) (with HI finite-dimensional Hilbert spaces). (The irreducible representations of A are finite-dimensional, and

CH. 5. 7)

ADDENDA

129

A admits an injective finite-dimensional representation, which is the direct sum of mutually disjoint irreducible representations.)

5.7.2. Let A be an involutive algebra, and tt and p representations of A. Then there exists a projection E in the centre of Tr(A) and a projection F in the centre of p(A), such that TrE =pp, Trl-EO p, TrO PI-F- [1006]. 5.7.3. Let A be an involutive algebra, and tt a factor representation of A in a separable space. Then there exists a function m which associates, with each representation P in a separable space quasi-equivalent to tt, a number m(p) E [0, +00] with the following properties: (i) m(Pl) = m(P2) if and only if PI = P2; (ii) for every sequence PI P2... of representations in separable spaces quasi-equivalent to tt, m(EB pJ = ~ m(pJ. Furthermore, the function m is unique up to multiplication by a constant >0. [1006]. 5.7.4. Let A be a C*-algebra, and] a closed two-sided ideal of A. Then A is of type I if and only if] and AI] are of type I. (Use 2.11.2.) [1847]. 5.7.5. Let d be a continuous von Neumann algebra, and B a C*-algebra weakly dense in d. Then B is antiliminal. (Use 2.11.1 and 5.5.2). A von Neumann algebra of type I can have weakly dense sub-C*-algebras which are antiliminal (consider an irreducible representation of a factor of type III) [630]. 5.7.6. Let A be a C*-algebra, and Tr a representation of A. (a) tt is said to be homogeneous if, for every projection E E Tr(A), TrE has the same kernel as it. We then have IITr(x)EII = 111T(X)11 for each
xEA.

(b) A factor representation is homogeneous. The converse is false. (c) For every closed two-sided ideal I of A, let Q(l) be the projection of H", onto the essential subspace of 1T I1. The Q(1), as I varies, generate a von Neumann sub-algebra ~ of the centre of 1T(A), called the ideal centre associated with 1T. Then tt is homogeneous if and only if ~ is just the scalar operators. (d) The kernel of an homogeneous representation is prime. [509]. *(e) For each open subset U of A, let I( U) be the closed two-sided ideal of A whose spectrum is U and let Q(U) = Q(1(U. Then U ~ Q( U) can be extended to a mapping Q" of the family of Borel subsets of A into gJ such that Q"(X I U X 2 u .. ) = ~ Q"(XJ if XI X 2, are mutually disjoint Borel subsets of A. [632]. *5.7.7. Let A be a unital C*-algebra. We use the notation of 2.12.19. Let

130

THE TYPE OF A REPRESENTATION

[CH. 5. 7

be positive forms on A; we write g < I if F g S Ff or, which amounts to the same thing, if g E Ff . The relation F, = Ff is an equivalence relation. which leads to a notion of a class of positive forms. (a) TTf is a factor representation if and only if the relation < totally orders Ff . Then I is said to be primary. (b) Suppose that f is primary. Then TTf is of type I if and only if there exists a pure state g such that g < f. Also TTf is of type III if and only if the class of f is minimal and maximal in the set of classes of primary positive forms and f is not a morphism of A into C [852].

I, g

CHAPTER 6

TRACES

AND REPRESENTATIONS

Let A be a C*-algebra. In Chapter 2, we associated a representation 17"/ with each positive form f on A. All representations may be obtained in this way, or at least all those that admit a cyclic vector (remember that every non-degenerate representation is the sum of representations ad mitting cyclic vectors). However the mapping f ~ 17"/ is by no means injective. We would now like to parametrise representations bijectively. Let 17" and 17" be two finite-dimensional representations of A. If Tr 7T(X) =Tr 11"(x) for each x E A, we know that 11" and 7T are equivalent. It is this fact that we are going to generalise to a certain extent. In fact, it is only the traceable representations, or rather their quasi-equivalence classes, that we will be able to parametrise.
6.1. Traces

6.1.1. DEFINITION. Let A be a C*-algebra. A trace on A + is a function [: A + ~ [0, +00] satisfying the following axioms: (i) If x, yEA +, we have f(x + y) = f(x) + f(y); (ii) If x E A + and if A is a non-negative scalar, then we have [(Ax) = Af(x) (with the convention that 0 + 00 = 0); (iii) If z E A, we have f(zz*) = [(z*z). f is said to be finite if f(x) < +00 for each x E A+. f is said to be semi-finite if, for each x E A+, f(x) is the supremum of the numbers f(y) for those yEA + such that y ~ x and f(y) < +00.
[In the definition of semi-finite traces, it is clearly enough to consider only those x such that f(x) = + 00]. Our interest will be confined almost exclusively to those traces on A + which are lower semi-continuous. If A is a von Neumann algebra, we recover the familiar definitions (cf., however A 28).

6.1.2. PROPOSITION. Let A be a C*-algebra, and f a trace on A ". (i) Let n be the set of x E A such that f(xx*) < +00. Then n is a

132

TRACES AND REPRESENTATIONS

[CR. 6, l

self-adjoint two-sided ideal of A; the two-sided ideal m = n2 is the set of linear combinations of elements of m+ = m n A +, and m+ is the set of x E A + such that f(x) < +00. The ideals m and n have the same closure in A. (ii) There exists a unique linear form f on m which coincides with f on

m".
(iii) We have

f(x*) = f(x), f(zx) = f(XZ), f(uv) = f(vu),

for each x E m; for any x E m, Z E A; for any u, v E n.

Let p be the set of x E A + such that f(x) < +00. Applying lemma 4.5.1 to p, we obtain (i). Assertion (ii) follows immediately from (i). The form I is real on the hermitian elements of m, and so f(x*) = f(x) for each xEm. If uEn, we have uU*Em and u*uEm, and therefore
f(uu*) = f(uu*)

= f(u*u) =f(u*u);

applying this to u + v, u - v, u + iv, u - iv, we deduce that f(uv*) = f(v*u) for u, v En. Finally, let z E A and x E m. We have x = ~ UjVj with Uj, Vj E n, and so

1(z( L UjVj)) = L fzu)v) =L f(Vj(ZUj

=

L fVjz)uj = L f(uj(Vjz

f (( L UjVj)Z),

from which the proposition follows. The ideal m will be called, by abuse of language, the ideal of definition of f and denoted by mf . 6.1.3. If tn, =A and if f is lower semi-continuous, f is semi-finite. In fact, let (u,\) be an increasing approximate identity of A consisting of elements of mf (1.7.2). Let x E A+. The X1/2U,\X1/2 constitute an increasing family of elements of A + dominated by x (1.6.8) and converging to x, so that lim f(X I/2U,\XI/2) ~ f(x), hence f(xI/ 2u,\XI/2) --+- f(x) since f(x 1/2 U,\X 1/2) :so, f(x). Now
X1/2 U,\X 1/2 E mi,

hence f(xI/2U,\XI/2) < +00,

and the assertion follows.

CR. 6, 2j

BITRACES

133

6.1.4. Let.A and B be two C*-algebras, p a morphism of A into B, and f a trace on B+. Then (f 0 p) IA+ is a trace on A+. If f is lower semi continuous, then (f 0 p) IA + is lower semi-continuous. If f is finite, (f 0 p) IA + is finite. If f is semi-finite, (f 0 p) IA + is not semi-finite in general. Nevertheless, if A is a closed two-sided ideal of B and if p is the canonical injection of A into B, then (f 0 p) IA + =f IA + is semi

finite; for if x E A+, and if y E B+ is dominated by x, we have y E A+ as can be seen by considering the canonical mapping of B onto BI A.
the von Neumann algebra generated by 7T(A), t a normal trace on and m a self-adjoint two-sided ideal of A, such that 7T(m) is contained in m, and strongly dense in Then f = (t 0 7T) A + is semi-finite and lower semi continuous, and, if (u,\) is an increasing approximate identity of m, we have f(x) = lim f(xl/Zu,\X I/Z) for each x E A ",

6.1.5. LEMMA. Let A be a C*-algebra, 7T a representation of A,

sr,

au

au.

Since 7T(m) is strongly dense in

au, 7T Im is

7T(U,\) converges strongly to 1 (2.2.10). Let 7T(X)I/Z7T(U A)7T(X)I/Z constitute an increasing family in verges strongly to 7T(X). Since t is normal, t(7T(X)I/Z 7T(U,\)7T(X )1/Z) -+ t(7T(X

non-degenerate, and so x E A ", The u: which con

in other words, f(xI/ Zu,\xI/2)-+f(x). We can take the UA to be in m (1.7.2); then 7T(XI/ZUAXI/Z) E 7T(m) ~ m, and so f(xI/ZuAx 1/ z) < +00; hence f is semi-finite. Since t is lower semi-continuous in the weak topology and a fortiori for the norm topology, f is lower semi-con tinuous. Reference: [452]. 6.2. Bitraces 6.2.1. DEFINITION. Let A be a C*-algebra. A bitrace of A is a function s: n x n -+ C, where n is a self -adjoint two-sided ideal of A, satisfying the following axioms: (i) s(x, y) is linear in x, conjugate-linear in y, s(y, x) =s(x, y), and s(x, x) ~ 0 (i.e. s is a sesquilinear positive hermitian form); (ii) s(y, x) =s(x*, y*) for x, yEn; (iii) s(zx, y) =s(x, z"y) for x, yEn, z E A;

134

TRACES AND REPRESENTATIONS

[CH. 6, 2

(iv) For each z E A, the mapping x ~ zx of n into n is continuous for the pre-Hilbert structure defined on n by s; (v) The set of elements xy, for x, yEn, is dense in n for the pre-Hilbert structure defined by s. The ideal n is called the ideal of definition of s and is denoted by n; We easily deduce the additional properties that s(xz, y) = s(x, yz*) for x, yEn, z E A and that, for each x E A, the mapping x ~ xz of n into n is continuous for the pre-Hilbert structure defined by s. 6.2.2. Let N, be the set of x E ns such that s(x, x) = O. By (i), this is a vector subspace of n., and is self-adjoint by (ii); it is a left ideal of A by (iii), and therefore a two-sided ideal of A. Let As be the canonical mapping of ns onto nsfNs. The form s defines, by passage to the quotient, an Hausdorff pre-Hilbert structure on nsf N s; we denote the completion of this pre-Hilbert space by H, and the scalar product in the Hilbert space H, by ( I). By (ii), the mapping x ~ x* defines, by passage to the quotient, an isometric mapping of nsfNs onto itself; extending by continuity, we obtain a mapping J s : H, -+ H, which is conjugate-linear, satisfies J; = 1, and is such that U, IJsb) = (b Ia) for any a, b E H; Let z E A. The mappings x -+ zx, x ~ xz in n s go over, by passage to the quotient, into mappings in nsf N s, and extend, by continuity, to con tinuous linear operators As(z), Ps(z) in H; It is easy to see that As is a representation of A in H s ; for example, we have by (iii), for x, yEns
(As(z*)A"x IAsY) = (As(z*x) IAsY) = s(z*x, y) = s(x, zy)
= (i\"x

IAs(zy

= (A"x

IAs(z)AsY) = (As(z)* A"x IAs),

whence As(z*) = As(z)*. Similarly, Ps is a representation in H, of the reversed C*-algebra AO of A. The representations As, o, are non-de generate by virtue of (v). We have, for x E n, and z E A,
Ps(z*)A"x = As(xz*) = As(zx*)* = Js"-s(z)J,A"x,

whence Since I, is an isomorphism of H, onto the conjugate Hilbert space, we see that the representations ~ (2.2.8) and Ps are equivalent. 6.2.3. Axioms (i) to (v) imply that the involutive algebra nsfN, endowed with the scalar product (I), is a Hilbert algebra (A 54). The von Neumann algebras au s "If s left- and right-associated with nsf N" are the

CR. 6, 3]

MAXIMAL

BITRACES

135

von Neumann algebras generated by A.(n s) and ps(ns); each is the commutant of the other (A 54). The von Neumann algebras generated by As(A) and Ps(A) are a priori larger than the above; however, since it is obvious that they commute, they are in fact equal to OU s and V s respectively. 6.2.4. The Hilbert algebra nJ N, defines a trace on 011; in a natural way, which we denote by ts (A 60). This trace is normal. For each x E n; we have ts(As(x)As(x)*) = (Asx IA.x) (A 60), and therefore
(1)

ts(A.(xx*

s(x, x) < +00.

If t~ denotes the linear extension of t s

Im7,

to m., , we have, for all x,

(A 60),

yEnS
t~(As(x)As(Y)*)
=

(A.x IAsY)

hence
(2)
t~(As(xy*

s(x, y).

We can similarly define a trace on V;, obtained from ts using JS but we will not have any need for it. References: [641], [689]. 6.3. Maximal bitraces 6.3.1. LEMMA. We retain the above notation. Let q = m., and q = ql/2 (A 9). If, for x, y E A;I(q), we put s(x, y) = t~(As(xy*, the~ s is a bitrace extending s, and the pre-Hilbert space n, is dense in the pre-Hilbert space n;
If x Ens, we have As(x, x*) E mt" hence As(x) E q, hence x E A;I(q). This, together with formula (2) of 6.2.4, prove that s extends s. Since q is the full Hilbert algebra corresponding to the Hilbert algebra As(ns) (A 60 and A 61), the image under As of A;I(q) is a Hilbert algebra in which As(ns) is dense. It follows from this that s is a bitrace, and that the pre-Hilbert space n, is dense in the pre-Hilbert space ns

6.3.2.

LEMMA.

Let s be a bitrace extending s. Then we have and s(x, x);;;. s(x, x)

for every x Ens"

136

TRACES AND REPRESENTATIONS

[CR. 6, 3

We clearly have N, = N s n n., and therefore an isometric linear map ping nslN, ~ nA N s" This mapping extends to an isometric linear mapping T: H, ~ H; Let K = T(H s ) Since n, is a two-sided ideal of A, K is invariant under the operators A..{z) , Ps,(z) for any z E A, and so the orthogonal projection of H; onto K belongs to the centre of OU s" Moreover, if we identify H, with K by means of T, A..{z) induces As(z) on H; Let B C H s be the Hilbert algebra of elements bounded relative to the Hilbert algebra nANs" and let B be the orthogonal projection of B onto H; Then the von Neumann algebra left-associated with B is the von Neumann algebra induced by 6f1 s on H s and the natural trace defined by B is the trace induced by t, (A 65). Now nslN, is contained in B and is dense in H s hence contained in B and dense in B; hence nslNs defines the same von Neumann algebras as B as well as the same natural trace. In other words, 6f1 s and t, are the von Neumann algebras induced by 6f1 s on H, and the trace induced by t s on 6f1: respectively. This established, let x En . . ,. We have, by the above,
ts(As(xx*
~

ts.(A..{xx* < +00,

and so x E n.. and S(x, x) ~ s(x, x). 6.3.3. DEFINITION. A bitrace is said to be maximal if there is no other bitrace which properly extends it. 6.3.4.
PROPOSITION. Let s be a bitrace. Then (i) s is maximal if and only if s = s; (ii) s is maximal

If s is a bitrace extending s and therefore s as well, we have n; Cit.. (6.3.2), and hence s = s, so that s is maximal. Consequently, if s = s then s is maximal. The converse is clear.

6.3.5. DEFINITION. Let s be a bitrace. Then s is called the maximal canonical extension of s. In general, s admits maximal extensions other than s: consider, for example, the case where A is the product of Its and another C*-algebra. 6.3.6. Let s be a bitrace. Endow n, with the norm induced by that of A. The mapping x ~ A.x of n s into H, is not continuous in general. However, the mapping
(x,
y)~As(xy)

As(x)AsY = Ps(y)A.x

CH. 6, 4)

RELATIONS BETWEEN TRACES AND BITRACES

137

of n, x ns into H, is separately continuous, because the representations As and p, of A are continuous. It follows from this and axiom (v) for bitraces, that, if E ~ n, is dense in ns for the norm topology of A, then the set of As(xy), (x, y E E), is dense in H; In particular, if A is separable, then H, is separable. References: [641], [689]. 6.4. Relations between traces and bitraces 6.4.1. LEMMA. Let A be a C*-algebra and f a lower semi-continuous trace on A +. In the notation of 6.1.2, we have that if for x, yEn we put s(x, y) =f(xy*) = f(y*x), then s is a bitrace of A.
It is clear that s is sesquilinear. For x, yEn and z E A, we have, in view of 6.1.2 (iii),
s(y, x)

= f(yx*) = f(xy*) = s(x, y)

~

s(x, x) = f(xx*) s(x*, y*) s(zx, y)

=f(x*y) = s(y, x)

= f(y*(zx = fz*y)*x) = s(x, z*y) s(zx, zx) = f(x*z*zx) = f(x*z*zx) ~ Ilz*zllf(x*x) =Ilz*zlls(x, x).
Axioms (i) to (iv) for bitraces are thus satisfied. Now let (u,\) be an increasing approximate identity of the C*-algebra n. Then, for each x E n, we have

stu, - x, u,\x - x) = stu.x, u,\x) - s(u,\x, x) - s(x, u,\x) + s(x, x)

~

lIu ! u, lis (x, x) -

f(x*u,\x) - f(x*u,\x)

+ s(x, x)

~2(f(x*x)-

f(x*u,\x.

N ow the x* u,\x constiute an increasing family of elements of A + con verging to x*x, and so f(x*u,\x)- f(x*x) since f is lower semi-con tinuous. Thus, u,\x converges to x in the pre-Hilbert structure defined by s. Moreover, we can take the u,\ to be in n by 1.7.2. We will say that s is the bitrace associated with f. 6.4.2. Now, let S be a C*-algebra, and let s be a bitrace of A. For each x E A +, put f(x) = ts(As(x. We have As(n;) ~ m and As(n;) is strongly ", dense in OU s Thus f is a lower semi-continuous semi-finite trace on A + (6.1.5) which will be said to be associated with s.

138

TRACES AND REPRESENTATIONS

[CH, 6, 4

6.4.3. LEMMA. Let A be a C*-algebra, s a bitrace of A and f the associated trace. Then the bitrace s associated with f is the maximal canonical extension of s. Let q = m,s and q= q1/2, For each x E A, we have
xE
"s'~

f(xx*) < + 00 ~ ts(As(xx* < + 00 ~ As(x) E q~ x E A;I(q).

os"~

Hence

Os = os,

and, for x E
s(x, x)

=f(xx*) =ts(As(xx* = S(x, x).

6.4.4. LEMMA. Let A be a C*-algebra, f a lower semi-continuous semi finite trace on A+, and s the bitrace associated with f. Then s is maximal, and the trace associated with s is f. Let [be the trace associated with s, and s the bitrace associated with [. By 6.4.3, s is the maximal canonical extension of s. For yEn, we have [,(yy*) = s(y, y) = s(y, y) = f(yy*). Hence, if x E A + is such that f(x) < +00, we have [(x) = f(x). Let x E A + be such that f(x) = +00. For every finite real number a, there exists an XI E A + such that XI :s;; X and a :s;; f(xl) < +00, since f is semi-finite. Then
[(x) ~ [(XI) = f(x I) ~ a,

and hence [(x) = +00.

This shows that f = f. Therefore s = s, from which it follows that s is maximal. 6.4.5. PROPOSITION. Let A be a C*-algebra. Let E be the set of lower semi-continuous semi-finite traces on A +, and let E be the set of maximal bitraces of A. The mappings "trace - associate bitrace" and "bitrace - associated trace", define bijections E ~ E, E ~ E, each being the inverse 0/ the other. This follows from 6.4.1, 6.4.3, 6.4.4. Let / be a lower semi-continuous semi-finite trace on A +, and s the associated maximal bitrace. We will put N, = N f , n, = of ... , t, = If; and, conversely, m f = m; Remember that
Iftf = Ilf (6.1.2

(i.

Reference: [452].

CH. 6. 5j

THE SUM OF TWO TRACES

139

6.5. The sum of two traces

1+ I

6.5.1. Let A be a C*-algebra, and 1 and I two traces on A +. Clearly, is a trace. If 1 and I are lower semi-continuous, then so is 1+1. We now suppose that 1 and I are semi-finite, and we show that 1 +I is then also semi-finite. Let x E A + be such that (f + f)(x) = +00, and let a be a finite real number; we have, say, that I(x) =+00; there exists a yEA+ such that y.s:;x and a.s:;/(y)<+oo; if I(y) < +00, we have a.s:; (f + f)(y) < + 00; if I(y) =+ 00, there exists a z E A + such that z .s:; y and a .s:; I(z) < + 00, whence z .s:; x and a .s:; (f + I)(x) < + 00; in both cases, we see that 1 + I is semi-finite.

6.5.2. The set of traces on A + is, evidently, partially ordered by the relation I.s:; f, i.e. the relation
I(x).s:; I(x)

for each x E A ".

Suppose that 1 and I are lower semi-continuous and semi-finite, and let s, s be the associated bitraces; then the relation I.s:; I is equivalent to the relation

n; ~ n,

and

s(x, x).s:; s(x, x)

for each x E n;

6.5.3. LEMMA. Let A be a C*-algebra, and sand s two maximal bitraces 01 A. Suppose that there exists an involutive sub-algebra B 01 A, which is dense in A, contained in both "5 and "5" and such that sand s coincide on B x B. Then s = s, By 6.3.6, B is dense in n, (resp. "5) for the pre-Hilbert structure of (resp. n.). Since sand s coincide on B x B, there is an isomorphism 8 of H, onto H; which, for each x E B, maps Asx to Asx. For x, Y E B, we have
8A s(x)AsY = 8A.(xy) = As(xy) = A.{x)Asy = As.(x)8AsY

"s

so that 8 maps As(x) to A.{x) for each x E B and consequently for each

x E A. Moreover, since 8 defines an isomorphism of the Hilbert algebra

As(B) (dense in "s! N s) onto the Hilbert algebra A.{B) (dense in "s!N s) 8 maps I1J1 s to I1J1 s and t s to t s" Let I, I be the traces associated with s, s, For each x E A+, we have I(x)

= ts(As(x = ts.(A.{x = I(x).

Since sand s are maximal, s = s (6.4.3).

140

TRACES AND REPRESENTATIONS

[CH. 6, 5

LEMMA. Let A be a C*-algebra, / and f two lower semi-con tinuous semi-finite traces on A +, and sand s the associated bitraces. Suppose that f .... f on (ftst. Then (i) There exists a unique T E :(Hs) such that s(x, y) = (TAsx IAS) for x, y E n.. (ii) We have 0 .... T .... 1 and T E UU s n Vs (iii) If we put ft(x) = ts(TAs(x) for each x E A+, then t, is a lower semi-continuous semi-finite trace on A +, which is dominated by f and coincides with f on ftst.

6.5.4.

For xE"s we have s(x,x) .... s(x,x); hence, for x,yEn s, s(x,y) depends only on Asx and AS, and is a continuous function of Asx, AS for the Hilbert space structure of H; This implies the existence of T E :(Hs ) satisfying (i). The uniqueness of T follows from the fact that As("s) is dense in H; We have 0 ~ s(x, x) .... sex, x) for each x E "so and therefore

If z E A, x E n; Y E

"so we
= =

have

(TAs(z)Asx IAsY) = (TAs(zx) IAsY) = s(zx, y) = s(x, z*y) (TAsx IAs(z*y = (TA,x IAs(z*)AsY) (As(z)TAsx IAsY),

and hence TAs(z)

= A,(z)T.

We see, in a similar way, that Tp,(z) = Ps(z)T, whence T E UU, nVs The function S ~ t,(TS) is a normal trace t; we have A,(m,) ~ rn, (for, if x Em;, we have t,(A,(x < +00, and so

and As(m s) is strongly dense in UU s ; hence t. is a lower semi-continuous semi-finite trace on A + (6.1.5). For x E n., we have ft(xx*) = ts(TAs(x)As(x*; since T E UU" the element TA,x is bounded relative to the Hilbert algebra As("s) (A 59), and
ts(TAs(x)As(x* = (TAsx IAsx) = s(x, x) = f(xx*).

Then I, I(ftst and f I(ftst are lower semi-continuous semi-finite traces on (ftst (6.1.4), finite and equal on (n;t by the above, and therefore equal (6.5.3). Moreover, let (u ... ) be an increasing approximate identity of fts; by

CH. 6. 6]

TRACES AND REPRESENTATIONS

141

6.1.5, we have, for each x E

ft(X)

A\
f(x)

= lim ft(x 1/2 UAX 1/2) = lim f(x 1/2 UAX 1/2) ~

because XI/2UAXI/2 E (fist and X1/2 UAX 1/2 ~ x. 6.5.5. PROPOSITION. Let A be a C*-algebra, and { and f two lower semi-continuous semi-finite traces on A ". Then f ~ { i{ and only i{ there exists a lower semi-continuous semi-finite trace {" on A + such that
{= f

+{".

The condition is clearly sufficient. Now suppose that f ~f. We make use of lemma 6.5.4. and its notation. Put {"(x) = ts((l- T)As(x)) for each x E A +. By 6.1.5, {" is a lower semi-continuous semi-finite trace on A ". For each x E A +, we have
ft(x)

+{"(x) =

ts(TAs(x)) + ts((l- T)As(x)) = ts(As(x)) = {(x).

In view of 6.5.4, f + f" coincides with { on (fist. If x E A + while x ~ (fist we have

f(x) + {"(x)
;:?;

ft(x)

+ {"(x) ={(x) = +00,

so again, f(x) + {"(x)

= {(x).

6.5.6. PROPOSITION. Let A be a C*-algebra, and { and f two lower semi-continuous traces on A + such that f ~ {. We suppose that mt is dense in A, {rom which it follows that { and f are semi-finite (6.1.3). Then there exists a unique lower semi-continuous trace {" 011 A + such that {=I +{". The existence of {" follows from 6.5.5. Let {= I +{" and {= t + f{ be two decompositions of {. For x E m/, the numbers {(x), f(x), {"(x) and f{(x) are finite, so that = f{(x). Hence {" = f{ by 6.5.3.

rex)

6.5.7. We no longer have uniqueness of semi-finite (6.9.3.). References: [452], [689].

6.6. Traces and representations

f" when

mt;t

A, even if { is

6.6.1. DEFINITION. Let A be a C*-algebra. A trace representation of A is an ordered pair (7T, t) with the following properties: (i) 7T is a non-degenerate representation of A in a Hilbert space;

142

TRACES AND REPRESENTATIONS

[CH. 6, 6

(ii) t is a faithful normal trace on OU+ (OU denotes the von Neumann algebra generated by 17(A)); (iii) 17(A) n n, is weakly dense in the von Neumann algebra au. These conditions imply that t is semi-finite, so that au is semi-finite. Thus, 17 is the direct sum of a representation of type I and a represen tation of type II. 6.6.2. Let (17, t), (77", t) be two trace representations of A, and au and OU the von Neumann algebras generated by 17(A) and 17"(A). We say that (77, t) and (77", t) are quasiequivalent (resp. equivalent) if there exists an isomorphism (resp. a spatial isomorphism) of au onto au which trans forms 17 into 17" and t into t. Then, 77 and IT are quasi-equivalent (resp. equivalent). 6.6.3. Let A be a C*-algebra, and (17, t) a trace representation of A. Then f = (t 17) IA + is a lower semi-continuous semi-finite trace on A + (6.1.5). We will say that I. and the corresponding maximal bitrace, are associated with (IT, t). If we replace (77, t) by a quasi-equivalent trace representation, f is unaltered.
0

6.6.4. Conversely, let f be a lower semi-continuous semi-finite trace on A +, and s the corresponding bitrace. Then (As ts) is a trace represen tation of A which will be said to be associated with I. or with s. If As is of type I, or of type II, ... , then f and s are said to be of type I, or of type II, ....
6.6.5. PROPOSITION. (i) Let f be a lower semi-continuous semi-finite trace on A +, and (IT, t) the trace representation associated with f. Then the trace associated with (IT, t) is f. (ii) Let (17, t) be a trace representation of A, and f the trace associated with (77, t). Then the trace representation associated with f is quasi equivalent to (17, t).

Let f, 17, t be as in (i), s the bitrace associated with f, and f the trace associated with (77, t). Then f =(Is As) IA+, and so f = f by 6.4.4. Let (17, t) be a trace representation of A, and sf the von Neumann algebra generated by 17(A). Recall the following facts, which have nothing to do with 77: endowed with the scalar product (S, T) ~ t(ST*), n, is a full Hilbert algebra (I denotes the linear extension to m, of t Im); At is an isomorphism of the von Neumann algebra sf onto the von Neumann algebra OUt on H,; At maps t to the natural trace () defined
0

CH. 6. 7]

CHARACTERS AND TRACEABLE

FACTOR REPRESENTATIONS

143

on 6117 by the Hilbert algebra n, (A 61). This established, let s be the bitrace of A associated with (7T, t); we have n, = 7T- 1(n,), and, for x,
yEns

s(x, y) = t(7T(XY*.

Hence 7T Ins defines, by passage to the quotient, an isomorphism ep of the Hilbert algebra nsf N, onto the Hilbert algebra 7T(n s ) ~ n, Since 7T(n s ) generates the von Neumann algebra d by the definition of trace re presentations, At (7T(n s generates the von Neumann algebra 6JL t , and hence (A 55) 7T(ns ) is dense in the full Hilbert algebra n, Hence ep extends to an isomorphism (again denoted by ep) of H, onto H,. It is easy to see that, for each z E A, ep maps As(z) to A,( 7T(Z. Moreover, ep transforms ts into O. Now (AS t s ) is the trace representation associated with s (or with f). We see that it is equivalent to (At 0 7T, 0). Since A, is an isomorphism of d onto 611, which transforms t into 0, (As t.) is quasi-equivalent to (7T, t). 6.6.6. COROLLARY. There exists a canonical bijection of the set of lower semi-continuous semi-infinite traces on A +, onto the set of quasi equivalence classes of trace representations of A. 6.6.7. In fact, one is not really interested in the trace representations of A, but in representations of A. We accordingly make the following definition:
DEFINITION. Let A be a C*-algebra, 7T a representation of A, and 611 the von Neumann algebra generated by 7T(A). Then 7T is said to be traceable if there exists a trace T on 611 + such that (7T, t) is a trace representation.

A representation quasi-equivalent to a traceable representation IS traceable. We are now going to study a case where there is little difference between the notions of trace representation and traceable represen tation. References: [452], [641], [689].
6.7. Characters and traceable factor representations

6.7.1. DEFINITION. Let A be a C*-algebra. A character of A is a non-zero, lower semi-continuous semi-finite trace f on A ". which

144

TRACES AND REPRESENTATIONS

[CH. 6, 7

satisfies the following condition: every lower semi-continuous semi-finite trace on A + dominated by f is proportional to f on (lftf)+ 6.7.2. Recall that if 3 is a semi-finite factor, there exist faithful, normal, semi-finite traces t on 3+ and that these traces are all proportional. Hence the ideal n, of 3 is uniquely determined; its elements are called the Hilbert-Schmidt operators relative to 3 (A 32).
PROPOSITION. Let A be a C*-algebra, 7T a non-trivial factor represen tation of A, and 3 the factor generated by 7T(A). Then the following conditions are equivalent: (i) 7T is traceable; (ii) 3 is semi-finite and 7T(A) contains a non-zero Hilbert-Schmidt operator relative to 3.

(i) ~ (ii) is clear. Now suppose that condition (ii) is satisfied, and let t be a faithful normal semi-finite trace on 3+. By hypothesis, n = 7T(A) n n, is a non-zero two-sided ideal of 7T(A). Hence the weak closure B of n is a non-zero weakly closed two-sided ideal of the weak closure of7T(A), i.e, of 3. Since 3 is a factor, we have B = 3 (A 9). 6.7.3.
THEOREM. Let A be a C*-algebra, f a lower semi-continuous semi-finite trace on A +, and (7T, t) the trace representation associated with f. Then 7T is a non-trivial factor representation if and only if f is a character.

Suppose that 7T is a non-trivial factor representation. Then f is non-zero. Let f be a lower semi-continuous semi-finite trace on A +, dominated by f on (mft. Apply lemma 6.5.4. With the notation of this lemma, T is scalar, since 7T is a factor representation. By 6.5.4 (iii), f is proportional to f on (mft. Now suppose that f is a character. Then 7T is non-trivial. Let T be an element of OUf n V f such that 0 ~ T ~ 1. If we put f(x) = t(T7T(X)) for x E A +, then by 6.1.5, f is a lower semi-continuous semi-finite trace on A +; it is clear that f is dominated by f. Hence there exists A E [0, 1] such that f = Af on (lftfr. For each x E"f we have (TAtx, Atx) = t(T7T(X)7T(X*)) = f(xx*) = Af(xx*) = A(Atx, Atx) and so T = A. This proves that OUf is a factor. 6.7.4.
COROLLARY. There exists a canonical bijective correspondence between: (a) The set of quasi-equivalence classes of non-trivial traceable factor representations of A;

CH. 6, 7)

CHARACTERS

AND TRACEABLE

FACTOR REPRESENTATIONS

145

(b) The set of characters of A, where two characters which only differ by a positive scalar factor are identified.

For each character f of A, Af is a non-trivial traceable factor re presentation (6.7.3). Moreover, every non-trivial traceable factor re presentation of A can be obtained in this way up to quasi-equivalence (6.6.5 and 6.7.3). Finally, let f, f be two characters of A such that Af and Ar are quasi-equivalent; there exists an isomorphism rp of 6l1f onto 6l1r which transforms Af into A,; since 6l1f and 6l1r are factors, rp transforms tf into atf" for some a E ]0, + 00[. Then, for each x E A +, we have
f(x) = tf(Af(x = atf(Ar(x = af(x).

Thanks to this result, we can speak, with a slight abuse of language, of the character of a non-trivial traceable factor representation, this character being defined up to a positive scalar multiple. 6.7.5. Let A be a C*-algebra, and 7T a non-trivial irreducible (and therefore factor) representation of A. Then 7T is traceable if and only if 7T(A) ~ ~(Hff); in fact, this condition is clearly sufficient, and is neces sary by 4.1.1O. (We can normalise the character of 7T by considering the usual trace on .2(H fft .) In particular, every irreducible representation of a postliminal C*-algebra is traceable (4.3.7), and hence every factor representation of a postliminal C*-algebra is traceable (5.5.3). 6.7.6. Let A be a postliminal C*-algebra, and 7T a non-trivial irreducible representation of A. For each x E A . put fff(X) = Tr 7T(X). Then [; is a character of A (6.7.3 and 6.7.5). If 7T is another irreducible represen tation of A, and if t, = fff" then 7T and 7T are quasi-equivalent (6.6.6), and therefore equivalent (5.3.3). Finally, if g is a character of A, Ag is a non-trivial factor representation (6.7.3), and therefore quasi-equivalent to a non-trivial irreducible representation 7T (5.5.3), so that g is propor tional to I, (6.7.4). 6.7.7. Retain the notation of 6.7.6, and let f = fff Since 7T(A) ~ ~(Hff)' Af defines, by passage to the quotient, an isomorphism of nfl N, onto the Hilbert algebra B of Hilbert-Schmidt operators on H". Identify nfl N, with B by means of this isomorphism. Then, for each x E A, Af(x) is the operator in B of left-multiplication by 7T(X), and pj{x) is the operator of right-multiplication by 7T(X)* in the space conjugate to B. If we now identify B canonically with u, 0 Hm Aix) becomes 7T(X) 01 and P}(x) becomes 10 7T(X). Hence 6l1f is identified with .2(H,,) 0 C, and V f is

146

TRACES AND REPRESENTATIONS

[CH. 6, 8

identified with C @ 5([[.,); also, Jj is identified with the canonical involution of H" @ H". Finally, (j(T @ 1) =Tr T for every T E 5(H"t. 6.7.8. Let A be a commutative C*-algebra. By 6.7.6, the characters of A in the sense of 6.7.1 are just the functions x ~ ag(x) on A . where a > 0 and g is a character of A in the classical sense. We therefore have a slight inconsistency in the terminology. 6.7.9. The reader will have noticed an analogy between the work of this chapter and that of chapter 2; the correspondence is as follows: Chapter 2 Positive forms. Pure positive forms. Representations. Irreducible representations. Chapter6 Lower semi-continuous semi-finite traces. Characters. Traceable representations. Traceable factor representations.

Unfortunately, the analogy breaks down over a crucial point. While in Chapter 2 we were able to prove the existence of pure positive forms thanks to the Krein-Milman theorem, it is not always the case that characters exist (cf, 6.9.2,6.9.10 and 9.5.7). This question is only settled for postliminal C*-algebras, since in that case irreducible represen tations (whose existence we have proved) are automatically traceable. Moreover, the theories of Chapters 2 and 6 both embrace a certain special case, which we are now going to study. References: [641], [689].
6.8. Finite traces

6.8.1. Let A be a C*-algebra, and g a linear form on A. Then g is said to be central if g(xy) = g(yx) for any x, yEA. Suppose that g is positive and central. Then the restriction f of g to A + is clearly a finite trace. Conversely, let f be a finite trace on A+. We make use of lemma 6.1.2 and its notation. We have m = A (1.5.8), so that f is a linear form on A, extending f, and therefore positive, and central (6.2.2(iii.We have thus established a bijective correspondence between central positive forms g on A and finite traces f on A +; we will identify f and g from time to time. By 2.1.8, finite traces are automatically continuous.

CH. 6, 8]

FINITE TRACES

147

6.8.2. Let f be a lower semi-continuous trace on A + and s the associated bitrace. We have mf = n; ~ n; By 1.5.8, the conditions mf = A and n, = A are equivalent. The condition mf = A clearly means that f is finite. 6.8.3. PROPOSITION. Let A be a C*-algebra, f a positive central form on A, and g = f IA +.
(i) We have 7Tf = Ag

(ii) The trace tl? on au; is finite; let t ~ be its linear extension to aug; we have t~(7Tf(z)) = (7Tf(z){f I{f) for every Z E A. (iii) aug and "VI? are finite von Neumann algebras. (iv) Ag(z){f =PI?(Z){f = AgZ for every z E A, and ll?{f = {J(v) The element {f is a separating and cyclic vector for aug and "Vg.

Let s be the bitrace associated with g. The Hilbert space H(7TJ) is the completion of A with respect to the scalar product f(y*x) = f(xy*) = s(x, y), the elements x such that s(x, x) = 0 having been factored out. We see that H(7Tf) = H, = H f. The representations TTJ and Ag are both ob tained from the left regular representation of A, and so 7Tf = AI? For x, yEA, we have, denoting the linear extension of tg to mt by t~,
(1)
t~(AI?(x)Ag(Y)*)

= s(x, y) = f(xy*)

= (7Tf(xY*){f I{f) = (7TJ(X)7TJ(Y)*{J I{f)

We at once deduce from this, by passing to the weak limit, that
(STg J gJ) = (TSg J gf)

for any S, T E aug.

Hence S ~ (sgJ I{J) is a finite normal trace of au;. Moreover, (1) proves that t~(7Tf(z)) = (7TJ(z){f I{f) for each z E A. Let A be the C*-algebra obtained by adjoining an identity element 1 to A; let 1 be the canonical extension of f to A and g = 11 A +. The Hilbert space H(7Tf) is equally the completion of A with respect to the scalar product /(y*x) =/(xy*), the elements x E A such that /(x*x) = 0 having been factored out; further, {f is the canonical image of 1. Let z E A. Since the operators of left- and right-multiplication by z are continuous on the pre-Hilbert space A, and the same is true of the involution, we have

and
19{J = A g 1* = A g 1 = {f.

We deduce from this that {f is a cyclic vector for aul? and "Vg , and

148

TRACES AND REPRESENTATIONS

[CH. 6, 8

therefore a separating vector for OU ~ = "fig and "fI~ = OUg- The normal trace S ~ (S~f I~f) on OU; is thus faithful, and coincides on lTf(A+) with the faithful normal trace tg , and hence these two traces are equal (A 29). Hence tg is a finite trace, and the von Neumann algebras OU g , "fig are finite. 6.8.4.
COROLLARY. Let A be a C*-algebra, and f a lower semi-con tinuous semi-finite trace on A+. Then t f is finite on OUt if and only if f is finite.

The necessity is clear since f(x) sufficiency follows from 6.8.3. 6.8.5.

= tf(Af(x for each

x E A ". The

COROLLARY. Let A be a C*-algebra, and f a character of A. Then f is of finite type if and only if it is finite.

This is a special case of 6.8.4. 6.8.6.

COROLLARY. There exists a canonical bijective correspondence between (a) the set of quasi-equivalence classes of non-trivial factor representations of A of finite type; (b) the set of finite characters of A of norm I.

This follows from 6.7.4 and 6.8.5. We can thus normalise the character of a non-trivial factor represen tition of finite type by imposing the condition that it be of norm I. This normalisation is occasionally at odds with that of 6.7.5.
PROPOSITION. Let A be a C*-algebra, and C the set of central positive forms on A of norm ",;; 1. (i) C is convex and compact for the weak*-topology of the dual of A. (ii) The extreme points of Care 0 together with the finite characters of norm l. (iii) C is the weak*-closed convex hull of 0 and the set of finite characters of norm 1.

6.8.7.

(i) is immediate; (iii) follows from (i), (ii) and the Krein-Milman theorem. Moreover, by 6.7.1, a finite character of A is a non-zero finite trace f on A + such that every finite trace dominated by f is proportional to f. Having said this, the proof of (ii) is exactly the same as the corresponding proof of 2.5.5.

References: [641], [689].

CH, 6. 91

ADDENDA

149

6.9. Addenda

6.9.1. Let A be a C*-algebra, I a lower semi-continuous semi-finite trace on A +, and s the bitrace associated with I. We are going to see that any given closed two-sided ideal ] of A defines a decomposition of f. (a) Let I =I II", and let s be the bitrace of J associated with 1. Then s is a bitrace of A. Let SI be its canonical maximal extension, and 11 the associated trace on A ". (b) We have

HII = n, C u,

OUII = OUr,

tii = tr

For each x E A, Afl(x) is therestriction of Af(x) to Hr. If x E A +, we have II(x)::;:: tf(PAf(x, where P denotes the orthogonal projection of H, on Hr. (c) Let (Uj) be an increasing approximate identity of Iftr . If x E A . we have ft(x) =lim I(XI/2U;X1/2). (d) For each xEA+, put g(x)=tj (I - P )1Tj(X . Then g is a lower semi-continuous semi-finite trace on A + vanishing on r. and I = 11 + g. [452]. 6.9.2. Let H be a separable infinite-dimensional Hilbert space, A the C*-algebra 2(H)/2Cf5(H), and t a trace on A + such that mt;t O. Then we have t = O. (Every two-sided ideal of A is equal to A, and so m, = A. Let el e2 be two orthogonal projections in H of infinite rank and with sum 1; let 110 12 be their canonical images in A; then we have t(fl) = t(1), t(fz} = t(1), hence t(l) = 2t(l), and t(l) = 0.) [631]. 6.9.3. Let A be the C*-algebra of continuous complex-valued functions on [0,1]. For x E A ". put I(x) = fol x(t)r l dt. Then I is a lower semi continuous semi-finite trace, but Iftf;t A. If, for x E A +, we put I(x) = x(O), then I is a continuous finite trace and I + I = I. [452]. *6.9.4. Let A be a C*-algebra. Two traceable irreducible representations of A with the same kernel are equivalent (6.7.5), but two traceable factor representations of A with the same kernel can be disjoint. [689]. 6.9.5. Consider again the C*-algebra B of 4.7.18. The identical re presentation of B is irreducible; let I be its character. Let f be a character obtained by composing the canonical mapping B ~ B/ fe (H) with a character of the commutative C*-algebra B/fe (H). Then I dominates I without being proportional to it.

150

TRACES AND REPRESENTATIONS

[CH. 6, 9

6.9.6. Let A be a C*-algebra, and f a character of A. Then there exists a family (j;) of pure positive forms on A such that f = L t. if and only if f is of type I. [326]. 6.9.7. Let A be a separable C*-algebra. (a) The set E of tt E Rep(A) such that 7T is traceable and irreducible is a Borel subset of Rep(A). (b) Let E, E" be the canonical images of E in A and Prim(A). Then E is a Borel set for the Mackey structure, and E" is a Borel set in Prim(A) (d. 3.9.2). The set E is the largest subset of A such that the restriction of the mapping A ~ Prim(A) to E is injective (4.7.2). In view of 3.9.2, we see that E is a standard Borel subset of A on which the Mackey Borel structure is the same as the topological Borel structure. [689]. 6.9.8. Let A be a separable C*-algebra. The set of 7T E Rep(A) such that Tr(A) = .;ecg(H rr ) is a Borel subset of Rep(A) (Guichardet, unpublished). 6.9.9. Every representation of a postliminal C*-algebra is traceable. (Use the proof of 5.5.2.) 6.9.10. Let E, be the class of separable C*-algebras all of whose representations are of type I. Let E; be the class of separable C* algebras admitting "sufficiently many" representations of type I (i.e. such that, for each non-zero x in the algebra, there exists a represen tation of type I, tt, of the algebra such that 7T(X) rf 0). Replacing "of type I" successively by "of semi-finite type", "of finite type", "traceable",we obtain the classes E 2 , E 2 , E 3 , E 3, E 4 , E~ of C*-algebras. What can be said about these classes? The class E, is well-known (d. 9.1). We have E, = E 2 = E 4 by 6.9.9., 9.1, 9.5.4. The class E; (and a fortiori E2) is the class of all separable C*-algebras, since an irreducible representation is of type I. Next, E 3 ~ E 2 = E, is the class of separable C*-algebras all of whose ir reducible representations are finite-dimensional. The classes E 3, E 4 are distinct from the class of all separable C*-algebras, but not much is known about them. [453]. 6.9.11. Let A be a C*-algebra. The set of all positive central forms on A is a Choquet simplex. (See [1629] for a more general result.)

CHAPTER 7

THE QUASI-SPECTRUM

Let A be a separable C*-algebra. In chapter 3, we defined the Mackey Borel structure on the set of irreducible representations of A (or rather on the set of their equivalence classes). We are now going to define a Borel structure on the set of factor representations of A (or rather on the set of their quasi-equivalence classes). This will be useful in Chapter

8.
7.1. The space of factor representations

7.1.1. Let n be a cardinal os:; ~o. Let H; be the standard n-dimensional Hilbert space (3.5.1). For each r;;a:0, we denote by Ir(Hn) the closed ball of centre 0 and radius r in I(Hn).
LEMMA. Let d be a metric on I.(Hn) compatible with the weak topology (B 9), S an element of I\(Hn), a and b non-negative real numbers, and x\, . . . ,X n E A. Let Y be the set of 7T E Repn(A) with the following property: there exist T\, ... , T m E Ib(Hn) n 7T(A) such that

7T(x\)T\ +... + 7T(xm)Tm E I\(Hn) and d(7T(x\)T\ +... Then Y is closed in Repn(A).

+ 7T(xm)Tm , S)

a.

Let (7T i ) be a net of elements of Y, and let 7T be an element of Rep.j A), such that 7T; -+ 7T in Repn(A). We show that 7T E Y. For each i, there exist such that and

152

THE QUASI-SPECTRUM

[CH. 7, 1

Since 2!b(Hn) is weakly compact, we can suppose, considering subnets of (7T i ) if necessary, that T;, ... , T~ converge weakly to T I, ... , T m E
2!b(Hn) Let x E A and tp, I/J E H n. We have (T;7T
i(X)<p

II/J) = (7T i(X)<p I T;*I/J)--;>(7T(X)<p ITN) =(T

I7T(X)<p

II/J)
(3.5.2).

and
(7T
i(X)T:<p

II/J) = (T;<p I7T i(X*)I/J)--;>(Tt<p 17T(X*)I/J) = (7T(x)T t<p II/J)

(T I7T(X)<p

Hence

II/J) =(7T(x)T I<P I I/J),

,
--;>

and so T. E 7T(A),

and similarly T 2 ,

T m E 7T(A). Furthermore, we have that

7T

i(XI)T;

+... + 7Ti(xm)T~<p

II/J) 7T(x.)T 1 +... + 7T(X m)Tm)<p II/J),

+ 7Ti(xm)T~

from which it follows that

7T i(X\)T; +...

converges weakly to whence and Thus,

7T

E Y.

7.1.2. LEMMA. Let d be a metric on 2!1(Hn ) compatible with the weak

topology, (Xl X 2, ... ) a dense sequence in A, and (S], S2"") a weakly dense sequence in 2!1(Hn ) (B 9). Let k, m, p be integers. Let Y k. m. p be the set of 7T E Repn(A) possessing the following property: there exist T I, T.2, , T m in 2!m(H n) n 7T(A) such that 7T(x l)Tt +... and

+ 7T(x m)T m E 2!1(Hn )

Then (i) Yk,m,p is closed in Repn(A). (ii) Y = k.p U m Yk.m.p is the set of non-trivial factor representations 7T E Repn(A).

CH. 7, I]

THE SPACE OF FACTOR REPRESENTATIONS

153

Assertion (i) follows from 7.1.1. For each 1T E Rep.(A), let N(1T) be the set of linear combinations of operators of the form RR, where R E 1T(A), R E 1T(A). Let N(1T) be the weak closure of N(7T-). Then 1T is a non-trivial factor representation if and only if N(1T) = It(H.). Suppose, to begin with, that 1T E Y. We will show that, for each k, Sk E N(1T), and this will imply that N(1T) = It(H.). Let p be a positive integer. There exists m such that 1T E Yk,m,p. Hence there exist T t, . . . , T m E Itm(Hn) n 1T(A) such that
1T(xl)Tt +...

+ 1T(xm)Tm E

ItI(H.)
1

and
d(1T(Xt)T I +...
,

+ 1T(xm)Tm , Sk) =S:;-. p

Since 1T(xt)Tt +... + 1T(xm)Tm E N(1T), we have

d(N(1T)

n Itt(H.), Sk) =S:;-.

P

Since p is arbitrary, we have just proved that Sk E N( 1T). Hence 1T is a non-trivial factor representation. Conversely, suppose that 1T E Repn(A) and that N(1T) = It(Hn). Let k and p be positive integers. We will prove that there exists an m such that 1T E Yk.m,P; this will complete the proof. By A 13, there exists a P EN(1T) such that IIPII<1 and d(P,Sk)<l!P. We have
P

= 1T(Yt)T, +... + 1T(yq)Tq

and
T t, ... , T q E 1T(A).

with
Yto . , Y q E A

Then there exist XiI .. ,xiq sufficiently close to Yt, ... ,Yq for
111T(X;)Tt + ...

+ 1T(Xi) Tqll =s:; 1

I

and
d(1T(X;)T I +...

+ 1T(Xiq )Tq, Sk) =S:;-. p

we will certainly have 1T E

If we take m > ii... ,iq ,

Yk.m,p

IITtll,... , IITqll,

7.1.3. Let A be a separable C*-algebra. We denote by Facn(A) the set of non-trivial factor representations of A in H m and by Fac(A) the union of the Fac.(A) for n = 1,2, ... ,

"0

154

THE QUASI-SPECTRUM

ten.

7, 2

7.1.4. THEOREM. Let A be a C*-algebra. Then Fac(A) is a Borel set in Rep(A). This follows immediately from 7.1.2. Thus, Fac(A), endowed with the Borel structure induced by that of Rep(A), is a standard Borel space since Rep(A) is standard (3.8.1). Reference: [559].
7.2. Definition of the quasi-spectrum

7.2.1. We will denote by A the set of quasi-equivalence classes of non-trivial factor representations of A. If A is separable, every factor representation is quasi-equivalent to a factor representation in a se parable space (5.3.7), and so the obvious canonical mapping Fac(A)-+ A is surjective. 7.2.2. DEFINITION. Let A be a separable C*-algebra. The Mackey Borel structure on A is the quotient structure of that of Fac(A) for the canonical mapping Fac(A) ~ A. The Borel space A is called the quasi spectrum of A. 7.2.3. PROPOSITION. Let A be a separable C*-algebra. Let B be a Borel subset of Fac(A) which meets each quasi-equivalence class in at most one point. Let t/J be the canonical mapping Fac(A) ~ A. Then t/J(B) is a Borel set in A and t/J I B is an isomorphism of the Borel space B onto the Borel space t/J(B). Let fl;(71, 71") be the intertwining number of two representations 71,71" of A. By 3.7.3, the set of pairs (71,71") E (Fac A) x (Fac A) such that ~(71', 71") ~ 0 is a Borel set in (Fac A) x (Fac A). Its complement R is the set of pairs (71,71") E (Fac A) x (Fac A) such that 71 = 71" (5.3.6). We see that S = R n (Fac A x B) is a Borel set in (Fac A) x (Fac A). Let p be the first coordinate projection (Fac A) x (Fac A) ~ Fac A. If (71, 71") E S and (711,71J)ES are such that p(71,71")=P(71l,71J), we have 71"=71= 711 = 71;, and so 71" = 71; by the condition imposed on B. Hence piS is injective. Hence p(S) is a Borel subset of Fac A (B 21). Now p(S) is the saturation of B in Fac A with respect to quasi-equivalence. Hence t/J(B) is a Borel set in A. The mapping t/J I B of B onto t/J(B) is a Borel bijection. To show that it is a Borel isomorphism, it suffices to show that the image t/J(B) of every Borel subset B of B is Borel. Now this follows from the above reasoning applied to B in place of B.

CR. 7, 3]

RELATIONS

BETWEEN THE SPECTRUM AND THE QUASI-SPECTRUM

155

7.2.4.

COROLLARY.

Each point of

A is

a Borel set in

A.

The quasi-spectrum has hitherto been known as the quasi-dual. Reference: [559].
7.3. Relations between the spectrum and the quasi-spectrum

7.3.1. Let A be a separable C*-algebra. For p = 1, 2, ... , let Fac Ip(A) be the set of representations 7T E Fac(A) such that 7T(A) is a factor of type Ip. We have Fac It(A) = Irr(A). Let Fac I(A) be the union of the Fac Ip(A). Then Fac I(A) is the set of 7T E Fac(A) such that 7T(A) is of type I, i.e. the set of 7T E Fac(A) which are of type I. The sets Fac Ip(A) are saturated with respect to equivalence of representations, but not with respect to quasi-equivalence. On the other hand, Fac I(A) is saturatedwith respect to quasi-equivalence. A representation of Rep(A) belongs to Fac Ip(A) if and only if it is equivalent to a representation of the form P7T where 7T E Irr(A) (5.4.11). All the sets Fac Ip(A) therefore have the same canonical image in A as Irr(A); we will denote this image by AI; it is the set of quasi-equivalence classes of factor representations of type I. 7.3.2. Recall that, on the set of unitary operators on a Hilbert space, the weak and strong topologies coincide and are compatible with the group structure. Having said this, we now have:
LEMMA. Let H be a separable Hilbert space, and C(j the group of unitary operators on H. Then for the strong (or weak) topology, C(j is a polish group

"0

In fact, the unit ball 2 t(H) of 2(H) is compact and metrisable for the weak topology, and is therefore a polish space. It is thus enough to show that C(j is a Os in 2(H). Now, if (~i) is a dense sequence in the unit sphere of H, then C(j is the set of T E 2 t(H) such that

for any integers i and j. Moreover, each of these inequalities defines a weakly open subset of 2(H) (because for each ~ E H, the mapping T ~ IInil is weakly lower semi-continuous; while T ~ T* is a homeomorphism for the weak topology).

156

THE QUASI-SPECTRUM

[CH. 7, 3

7.3.3. LEMMA. Let A be a separable C*-algebra, p a cardinal :S;;"o and B a Borel subset of Irr(A) which is saturated with respect to equivalence. Let C be the set of elements of Fac Ip(A) quasi-equivalent to an element of B. Then C is a Borel set in Rep(A). Let H, be the standard n-dimensional Hilbert space. The sets B n Repn(A) are saturated with respect to equivalence, and are Borel sets. It is enough to consider each of these sets separately. In other words, we will suppose from now on that the elements of B all operate on the same H n Then the elements of C operate on Hnp. We give the proof for n =p = (the other cases are dealt with in an analogous way and are somewhat simpler). We thus put H = H; = H np Let (K" K 2, ) be an infinite sequence of mutually orthogonal infinite-dimensional closed subspaces of H, whose direct sum is H. Let I, be an isomorphism of H onto K; For each T E .2(H), the operators JjTJ;1 define, by linearity and continuity, an operator f on H, and T ~ f is an isomorphism of .2(H) onto a factor ~ of type L." namely the set of bounded linear operators on H which leave the K;s invariant, and whose restrictions to the Kjs are interchanged by the isomorphisms JJ;. The commutant ~' of ~ is also a factor of type I",. Let 7T be a representation of A in H. We will denote the representation x ~ 7T(X)- of A in H by 7T ": If 7T E Irr(A), then 1T - (A) generates the factor ~. The mapping 1T ~ 1T- is clearly injective. For every ~ belonging to one of the K, and each x E A, the mapping 1T ~ 7T-(X)~ is continuous. Hence the mapping 7T ~ 7T- is Borel. By B 21, the image B- of B under the mapping 7T ~ 7T - is a Borel subset of Rep(A), contained in Fac Ioo(A). The set C of the statement of the lemma, is the union of the VB - V-I as V runs over the group C of unitary operators on H (5.3.3. and 5.4.11). In other words, C is the image of C x B- under the mapping

"0

8: (V, u)~

VuV~

of

x B- into Rep(A).

For each x E A and each

(V,u)~VU(X)V-l{

E H, the mapping of CxRep",(A) into H.

is continuous (CfJ being endowed with the strong topology), and hence the mapping 8 is Borel. We are going to show that, for an element V of C the following conditions are equivalent (provided that B;I=. cP, which case we need clearly only consider):

CH. 7. 3]

RELATIONS BETWEEN THE SPECTRUM AND THE QUASI-SPECTRUM

157

(1) U~U-l

=s .
with V, W unitary, V
E~,

(2) U is of the form VW (3) UB~ U- 1 = B-;

W E ~~;

(4) There exist 7T1, 7T2 E B - such that U7TI U- I = 7T2: (4) ~ (1): let 7T1, 7T2 E B - with U7TI U- I = 7T2 Then U7TI(A)U- 1 = 7T2(A), and hence, taking weak closures, U~U-I = ~. (1) ~ (2): if U~U-I = ~, then U defines an automorphism of ~. Now every automorphism of a factor of type I is defined by a unitary operator belonging to this factor (A 37). There therefore exists a unitary operator V of ~ such that W = V-I U defines the identical automorphism of ~, i.e. belongs to ~', and we have U = VW. (2) ~ (3): let U = VW, with V, W unitary, V E~, W E ~'. Let 7T E B. We have W7T(Xr W- I = 7T(Xr for every x E A. Moreover, V is of the form T- with T a unitary operator on H. Then
V7T(X)- V-I

= (T7T(X)T- I)-,
lr.

hence
U7T - U- I =(T7TT-

Now T7TT- I E B since B is saturated with respect to equivalence. Hence U7T-U- IEB- and finally UB-U-I=B-. (3) ~ (4): obvious. (Conditions (1)-(4) define a subgroup Cfj of Cfi. Concerning condition (1), it is clear that Cfj is closed in Cfj. By 7.3.2 and B 17, there exists a Borel subset I of Cfj which meets each left coset of Cfj in Cfj in exactly one point. We show that e(Cfj x B -) = e(I x B "). It is clear that B(CfjxB-)=:2e(IxB-). Conversely, let UECfj, uEB-. We have U= U IU2 with UIEI, U 2ECfj. Then B(U,u)= U I(U2UU;I)Ut lE UIB- Uti \; B(I x B-) by condition (3) above, whence the stated equality. We show that the restriction of e to I x B - is injective. Let 7T1, 7T2 E B- and VI V 2 E I be such that V I7TI VII = V27T2 yZl; then V2"1 VI satisfies condition (4) above, hence VI and V 2 belong to the same left coset of Cfj in Cfj, hence VI = V 2 and then 7TI = 7T2 This established, C = e(I x B-) is a Borel set in Rep(A) (B 21).. 7.3.4. PROPOSITION. Let A be a separable C*-algebra. Then each set Fac lp(A) is a Borel set in Rep(A). This follows from 7.3.3 applied to B

= Irr(A). = Up Fac lp(A) with

7.3.5. We endow the sets Fac lp(A) and Fac leA)

158

THE QUASI-SPECTRUM

[CH. 7. 3

the Borel structure induced by that of Rep(A) (or of Fac(A. The Borel spaces thus obtained are standard. 7.3.6. The equivalence relation induced on Irr(A) by quasi-equivalence is just equivalence of representations (5.3.3). Hence the canonical injec tion i of Irr(A) into Fac I(A) defines, by passage to the quotient, a bijection (/J of A onto AI (7.3.1).
THEOREM.

Let A be a separable C*-algebra. Then

(i) AI is a Borel set in A. (ii) The bijection (/J is an isomorphism of A onto AI for the Mackey

Borel structures.

Assertion (i) follows from 7.3.4. We prove (ii). Let

q: Irr(A)~A

and

p: FacI(A)~A[,

be the canonical mappings. Let S be a Borel subset of AI Then p -I(S) is a Borel subset of Fac I(A) which is saturated with respect to quasi equivalence.
i

Irr(A) -

Fac I(A)

.j
A -

j.
AI

Hence i-(p-(S =q-I(C/>-I(S is a Borel subset of Irr(A), saturated with respect to equivalence. Hence C/>-I(S) is a Borel set in A. Con versely, let T be a Borel subset of A. Then q-(T) is a Borel subset of Irr(A), saturated with respect to equivalence. The set p-I(/J(T is the saturation of i(q-I(T with respect to quasi-equivalence. Applying 7.3.3 with B = q-l(T), we see that p-(C/>(T is a Borel set in Fac I(A). Hence C/>(T) is a Borel set in A[, and the theorem follows. We will henceforth identify the Borel space A with the Borel subset AI of A. 7.3.7. In particular, if A is a separable postliminal C*-algebra, the Borel spaces A and A may be identified (5.5.3). Reference: [449].

CH. 7, 4]

THE FINITE PART OF THE QUASI-SPECTRUM

159

7.4. The finite part of the quasi-spectrum

7.4.1. We do not really know which topology is most suitable for A. However, we are going to define a subset Af of A on which there exists a sufficiently satisfactory topology. Let A be a separable C*-algebra, H; the standard n-dimensional Hilbert space, Facfn(A) the set of non-trivial factor representations of A in H; of finite type, and Facf(A) the union of the Facfn(A) for n = 1,2, ... , ~o. (We have, incidentally, that Facfn(A) = Facn(A) for n < ~o.) We will denote the canonical image of Facf(A) in A by Af , i.e. Af is the set of quasi-equivalence classes of non-trivial factor representationsof A of finite type. 7.4.2. Now let C be the convex set of central positive forms on A of norm ~1. Then C, endowed with the weak*-topology, is a compact metrisable space. The set D of non-zero extreme points of C is a 0 8 in the space C (B 13), and is therefore a polish space for the weak* topology. Moreover, D is the set of finite characters of A of norm 1 (6.8,7). By 6.8.6, the mapping
g~

quasi-equivalence class of Ag

is a bijection of D onto Af . We will endow Af with the topology obtained from the weak*-topology on D by means of the above bijection. Thus, Af becomes a polish space. 7.4.3.
PROPOSITION. Let A be a separable C*-algebra. Then (i) Af is a Borel subset of A. (ii) The Borel structure induced on Af by that of A is the Borel structure given by the topology of A f . (iii) The Borel space Af is standard.

We are going to construct a Borel mapping Q of D into Rep(A) such that, for each g E D, Q(g) is a factor representation of finite type with character g. For g, g E D, g-:l: g, the representations Q(g), Q(g) will then be non-quasi-equivalent. Since D and Rep(A) are standard Borel spaces, Q(D) will be a Borel subset of Rep(A) meeting each quasi equivalence class in at most one point, and Q will be a Borel isomor phism of D onto Q will be a Borel isomorphism of D onto Q(D) (B 21). By 7.2.3, if ljJ denotes the canonical mapping of Fac(A) onto A, ljJ(Q(D))

160

THE QUASI-SPECTRUM

[CH. 7, 5

will be a Borel subset of A and ljJ Ifl(D) will be a Borel isomorphism of fl(D) onto ljJ(fl(D)). Now ljJ(fl(D)) = At by 6.8.6, and proposition 7.4.8 will therefore be proved. Let (XI> X2 . ) be a dense sequence in A. We adopt again the notation of 6.2.2. For each g E D, the AgXn are dense in H g, because g is continuous. The functions
g ~ (AgXn

IAgXp)

= g(xnx~)

are continuous on D. Hence (A 98) there exists a unique Borel field structure on the field g ~ H g such that the vector fields g ~ AgXn are Borel. Let D, be the set of g E D such that dim H, = pCP = 1,2, ... ,"0); this is a Borel subset of D (A 96). It is enough to construct fl on each D p. Now there exists an isomorphism g ~ V(g) of the field g ~ Hg(g E D p ) onto the constant field defined by the standard Hilbert space H, (A 96). It then suffices to put
fl(g)(x)

= V(g)Ag(x)V(g)-I

for each X E A. Since g ~ Ag(x) is a Borel field of operators, the mapping g ~ fl(g)(x) is Borel, hence fl is a Borel mapping of D into Rep(A). 7.4.4. It follows from 7.4.3 that Facf(A) and Facfn(A) are Borel subsets of Fac(A). Reference: [689].

7.5. Addenda

7.5.1. Let A be a C*-algebra, I a closed two-sided ideal of A and lJ) the canonical morphism of A onto AI1. Let E be the set of pEA which vanish on 1. Then 7T ~ 7T ta defines a bijection of (AI 1) ~ onto E, and 7T ~ 7T II defines a bijection of A\E onto 1. (U se 2.11.1.) If A is separable, then E is a Borel subset of A and the above bijections are Borel isomorphisms. (Use the existence of an approximate identity in 1.) In particular, A may be identified with the Borel space which is the sum of a standard space and a space E, where B is a separable antiliminal C*-algebra. [451].
0

7.5.2. Let A be a separable C*-algebra. Let M be the set of 7T E Rep(A) which are homogeneous (5.7.6). Then M is a Borel subset of Rep(A). [509].

CH. 7. 5]

ADDENDA

161

7.5.3. Let A be a separable C*-algebra. On induced by the topology of A and by that of coincide (take for A the example of 4.7.19).
1T

A n AI the

topologies

AI do not necessarily

7.5.4. Let A be a separable C*-algebra. The set of 1T E Fac(A) such that is traceable (resp. of type II) is a Borel subset of Rep(A) [752] (resp. [1152]). The set E of traceable elements of A is a standard Borel space. [752].

CHAPTER 8

INTEGRATION AND DISINTEGRATION OF REPRESENTATIONS

We now have enough tools to get to grips with one of the two problems mentioned at the beginning of Chapter 3: namely, the decom position of a representation into irreducible representations. As we have said before, we will not be concerned with a decomposition into a direct sum of irreducible representations, but with a subtler decomposition (that we will call disintegration). We can get some idea of it by con sidering the very simple example that follows. Let A be the C*-algebra of continuous complex-valued functions on [0, 1]; for each f E A, let 7T(j) be the operator of multiplication by f in the Hilbert space H = L 2 ([0, 1]). Then 7T is a representation of A in H. The irreducible represen tations of A are its characters f ~ f(to) (where to E [0, 1]). Decomposing 7T into a direct sum of irreducible representations thus amounts to finding an orthonormal basis of H, each element of which is a common eigenvector of the 7T(f). Now the 7T(f) have no non-zero common eigenvectors. Nevertheless, loosely speaking, the Dirac functions of the various points of [0, 1] can be regarded as mutually orthogonal common eigenvectors of the 7T(j), and each element of H can be written as an integral linear combination of these Dirac functions. Throughout this chapter, A denotes a separable C*-algebra.

8.1. Integration of representations

8.1.1. Let Z be a Borel space, p, a positive measure (B 30) on Z, and (~H() a p,-measurable field of Hilbert spaces over Z (A 69). For each (E Z, let 7T() be a representation of A in H(): we say that (~7T'() is a field of representations of A. The field of representations (~7T'(0 is said to be if, for each x E A, the field of operators (~7T')(x) is measurable (A 77).
DEFINITION.

measurable,

CH. 8, 1]

INTEGRATION OF REPRESENTATIONS

163

8.1.2. We have 117T(l)xll 0;;;; Ilxll for every x E A and every lE Z. If the field l~ 7T(l) is measurable, we can therefore construct, for each x E A, the continuous operator 7T(X) = 7T(l)(X) df..t(l) on the Hilbert space
H

f:

f:

H(n df..t(l). The mapping x ~ 7T(X) is a representation of A in H.

PROPOSITION.

We have, for x, yEA,

7T(X

+ y) =

J
z

Ell

7T(()(X

+ y)

df..t(()

J
z

Ell

(7T(()(X)

+ 7T(()(Y

df..t(()

Ell

Ell

7T(l)(X) df..t(()

J
z

7T(()(Y) df..t(l) = 7T(X)

+ 7T(Y)

z We see similarly that

7T(Ax) = A7T(x), 7T(Xy)

= 7T(X)7T(y),
7T

7T(X*) = 7T(X)*.

8.1.3.

DEFINITION.

With the above notation,

integral of the 7T(l) and we write

f:

7T is said to be the

direct

7T(l) df..t(n

8.1.4. Every operator 7T(X) is decomposable, and therefore commutes with the algebra of diagonalisable operators (A 80). In other words,
~

C 7T(A).

8.1.5. Let K be the essential subspace of 7T (2.2.6). Then P K commutes with 7T(A), and therefore with ~; hence there exists a measurable field l~ E(l) E 2(H(l of projections such that P K = fEll E(l) df..t((). Let (UI, Uz, ... ) be an approximate identity of A (1.7.2). We have that 7T(U n) converges strongly to P K (2.2.10); hence, taking a subsequence of (un) if necessary, 7T(l')(Un)~ E(() strongly almost everywhere (A 79). Hence, almost everywhere, E(()(H(l is the essential subspace of 7T((). In particular, 7T is non-degenerate if and only if almost all the 7T(() are non-degenerate. 8.1.6.
PROPOSITION. Let (~7T(l'),

representations

l~ 7T(l) be two measurable fields of of A (relative to the same measurable field (~H((. Let
=

7T
Then

Ell

Ell

7T(l) df..t(l),

7T =

7Ta) df..t(l).

7T = 7T

if and only if 7T(l)

= 7T(l)

almost everywhere.

164

INTEGRATION AND DISINTEGRATION

OF REPRESENTATIONS

[CH. 8, 1

The condition is obviously sufficient. Suppose then that 7T = 7T. Let (Xj) be a dense sequence in A. We have 7T({)(X j) = 7TU)(X;) except for a negligible set N, of {s. Then N = U N, is negligible, and, for {e N, we have 7T({)(X;) = 7T({)(Xj) for each i, and hence 7T({) = 7T({). It follows from 8.1.6 that the 7T({) are determined by 7T to within negligible sets. 8.1.7.
PROPOSITION. Let {~7T({) tations of A in the H({). Let

be a measurable field of represen

H =

ffi

H({) df-L({),

Suppose that there exists a representation 7To in a Hilbert space H o such that 7T({) = 7To for every {. Suppose also that f-L is standard. Then there exists an isomorphism of H onto L~(Z, f-L) Q9 Hi; which maps 7T(X) to I Q9 7To(x) for every x E A. The representation 7T is a multiple of 7To.

This is a special case of A 83. 8.1.8. Let HIH 2 , 2, ... ,?\o.

.H; be the standard Hilbert spaces of dimension I,

PROPOSITION. Suppose that Z is the union of mutually disjoint Borel sets ZI> Z2, .. ,Z", and that {~H({) reduces on Z; to the constant field {~Hn. Let {~7T({) be a field of representations of A in the H({). Then {~7T({) is measurable if and only if {~7T({) is equal almost everywhere to a Borel mapping of Z into Rep(A).

The condition is obviously sufficient. For the necessity, let (gj) be an orthonormal basis of H n Let (x;) be a dense sequence in A. If the field {~H({) is measurable, the functions {~( 7T({)(Xj)gj I gk) are measurable on Zn and therefore equal almost everywhere to Borel functions. Hence, altering {~7T({) on a negligible subset of Zn we can ensure that all the mappings {~( 7T({)(Xj) Igk) are Borel. Then, for each x E A and all ~, T/ E H n , the mapping {~( 7T({)(X)g I T/) is Borel; hence {~ 7T({) is a Borel mapping of Z; into Repn(A). References: [509], [638], [1004], [1007], [1036], [1037], [1041], [llOI], [1105], [1458], [1680].

eH. 8, 2j

EQUIVALENCE OF TWO DIRECT INTEGRALS OF REPRESENTATION

165

8.2. Equivalence of two direct integrals of representation

8.2.1.
{~H

PROPOSITION. Let Z be a Borel space, /L a positive measure on Z, ({) a measurable field of Hilbert spaces on Z, {~7T(t) a measurable field of representations of A in the H(t),

f
f
$

H({) d/L({),

7T

f f
$

7T() d/L(),

and ~ the algebra of diagonalisable operators on H. Let ,l be a positive measure on Z equivalent to /L,

fI =

H(n d,l(),

iT =

7T(n

d,l(),

and ii the algebra of diagonalisable operators on if. Then the canonical isomorphism of H onto fI (A 75) transforms 7T into iT and ~ into ii.

Let pen = dji(nld/L(n. The canonical isomorphism U of H onto fI is the isomorphism which maps x E f$ x() d/L() E H to Ux = ~ p-1/2()X() d,l(). It is clear that U transforms ~ into ii. Let a E A. Then we have
iT(a)Ux

7T(t)(a)p-1/2()x() dji()

=U
and so U transforms
7T

7T(n(a)x() d/L({)

= U7T(a)x

into iT.

8.2.2. PROPOSITION. Let Z be a Borel space, /L a positive measure on Z, t ~ H() a u-measurable field of Hilbert spaces on Z, {~7T() a u-measurable field of representations of A in the H(),
H and
~

H() d/L),

7T

7T) d/L(t),

the algebra of diagonalisable operators on H. The entities

z;

/Lt, (t ~ HI(t),

t, ~

7Tt({t),

n;

7T1>

s;

are correspondingly defined.

166

INTEGRATION AND DISINTEGRATION

OF REPRESENTATIONS

[CH. 8, 2

Suppose that there exist: (1) a u-negliglble Borel subset N of Z, and a ""l-negligible subset N I of ZI; (2) a Borel isomorphism 1/ of Z\N onto ZI\N I transforming u. into ""1; (3) an n-isomorphism ~ ~ V(~) of the field ~ ~ H () (~E Z\N) onto the field (I ~ HMI)(I E ZI\N I) (A 70, A 74), such that V() transforms 7T) into 7T1(1/() for every (. Then V = fEB V() d,.,,() transforms ~ into ~I and 7T into 7TI

Let [mo, I H M j ) denote the identity operators on H() and HI(I) If C(Z, ,.,,), and if the function I, on ZI\N I is the transform by 1/ of f , Z\N, then V maps fEB f()IH ({) d,.,,() to fEB U(I)[H1({ , ) d,."Ml), and so V transforms ~ into ~I' Moreover, let a E A and x = fEB x() d,.,,() E H. We have
E

EB
V7T(a)x

=V
EB

(f 7T()(a)x() d,."U)
V( 1/-1(I)7T(1/ -I(l(a )x( 1/-1(1 d,."MI)

=
=

J
EB

J7TMI)(a)V(1/-I(tx(1/-

tUt

d""l(l)

= 7Tl(a)Vx

and so V transforms 7T into 7Tt.

8.2.3. PROPOSITION. We retain the notation Z, u; H(), 7T), H, 7T, ~, Zt> ... , ~I of 8.2.2. Suppose that there exist N, Nt> 1/ with properties (1) and (2) of 8.2.2. Suppose, in addition, that, for every (E Z\N, the representations 7T() and 7Tt(1/( are equivalent. Then if, further, ,." is standard, there exists an isomorphism V of H onto HI transforming ~ into ~t and 7T into 7Tl
There exists (A 82) an 1/-isomorphism (V(n,EZ\N of the field (H(,EZ\N onto the field (HMItEZjIN, such that V() maps 7T()(X) to 7Tt(1/((x) for every (E Z\N and every xED (where D is a countable dense subset of A), and therefore transforms 7T(n into 7Tt(1/(). 8.2.3 then follows from 8.2.2.

CH. 8. 3j

DISINTEGRATION

OF REPRESENTATIONS

167

8.2.4. Propositions 8.2.1 and 8.2.2 admit a converse:

PROPOSITION. We retain the notation Z, JL, H({), 1T({), H, 1T, !l, Zt, ... ,!ll of 8.2.2. Suppose that there exists an isomorphism U of H onto HI transforming !l into !It and 1T into 1Tt. Then if JL and JLI are

standard, there exist: (1) a u-negligible Borel set N in Z, and a JLI-negligible Borel set N I in
ZI;

(2) a Borel isomorphism T/ of Z\N onto Zl\N I transforming JL into a measure ill equivalent to JLI; (3) an n-isomorphism {~V({) of the field {~H ({)({ E Z\N) onto the field {I ~ HMI)({I E ZI\N I), such that V(O transforms 1T({) into 1TI(T/({ for every {EZ\N, with the following property: U is the com position of f\fJ V({) dJL({) and the canonical isomorphism of f\fJ HM) dilM.) onto HI Since U transforms !l into !It, there exist N, NI> T/, ilt, {~ VU) with all the properties of 8.2.4, except for the property that VU) transforms 1T({) into 1TI(T/(m for {E Z\N (A 85). Since U is the composition of f\fJ V({) dJL({) and the canonical isomorphism of f\fJ HMI) dilMI) onto HI> U transforms 1T into f\fJ 1T;({I) dJLI({I), where 1T;({I) is the transform of 1T( T/ -l({I by V( T/ -I({t (8.2.1 and 8.2.2). Since U transforms 1T into 1TI we have 1TI({I) = 1T;({I) almost everywhere (8.1.6). Adjoining a suitable negligible subset to N, we thus see that V({) transforms 1T({) into 1TI(T/({ for {EZ\N. References: [638], [1004], [1007], [1036], [1041], [1101], [1105], [1458], [1680]. 8.3. Disintegration of representations 8.3.1.
{~H

LEMMA. Let Z be a Borel space, p, a positive measure on Z, ({) a u-measurable field of Hilbert spaces on Z, H = j\J H(O dJL({), and 1T a representation of A in H. Suppose that each operator 1T(X)(X E A) is decomposable. Then there exists, for each {E Z, a representation 1T({) of A in H({), such that 1T = f\fJ 1T(O dJL(n

Let (XI X2 ) be a dense sequence of distinct points in A. We can suppose, adding to this sequence if necessary, that the set of x, is an involutive subalgebra B of A over the (countable) field of rational complex numbers. For each i, choose any measurable field {~Tj({)

168

INTEGRATION AND DISINTEGRATION OF REPRESENTATIONS

[CH. 8, 3

such that
1T(X/)

T/({) dJL({).

Let N/ be the negligible set of the CE Z such that IIT/({)II> 1/1T(X/)II. The union N of the N/ is negligible. Moreover, between the Xi there hold a countable infinity of relations of the form Xk = AIXi + A2x j (AI A 2 rational complex numbers), or of the form x, =xix j, or of the form Xj = x1. These relations imply that, for almost every C,
Tk(C) = At TM)

+ A2 1j (C), or

Tk(C) = TM)Tj(C),

or

Tj(C) = TM)*.

There therefore exists a negligible subset N of Z such that, for C~ N, the mapping XI ~ TI(C) is a morphism of B into 5(H(C (regarded as an involutive algebra over the field of rational complex numbers). Replace the TI(C)s by 0 for CE NUN. Then, for every CE Z, we have IITM)II ~ 117T(xJII ~ Ilxill for any i and the mapping XI ~ T/(C) is a morphism of B into 5(H(C. This morphism extends uniquely to a representation 7T(C) of A in H(n. Let X E A, and write 7T(X) = f$ T(C) dJL(C). There exists a sequence (x.) of elements of B converging to x. We have

IIT(C) - 7T(C)(x n)11

IIx -

x.ill

almost everywhere; hence, almost everywhere, 7T(C)(X.) converges to T(n. Hence T(() = 7T(()(X) almost everywhere, from which it follows that 1T(X) = f$ 7T({)(X) dJL().
THEOREM. Let H be a separable Hilbert space, 11 a representation of A in H, and ~ a commutative von Neumann subalgebra of 7T(A). Then there exist a standard Borel space Z, a bounded positive measure JL on Z, a measurable field (~H(C) of Hilbert spaces on Z, a measurable field (~7T(C) of representations of A in the H({), and an isomorphism of H onto f$ H({) dJL({) which transforms ~ into the algebra of di agonalisable operators and 11 into f$ 1T({) dJL(C).

8.3.2.

There exist (A 84) a standard Borel space Z, a bounded positive measure JL on Z, a measurable field C~ H (C) of Hilbert spaces on Z, and an isomorphism of H onto f tfJ H(C) dJL(C) which transforms ~ into the algebra of diagonalisable operators. Since ~ ~ 7T(A), every operator 7T(X)(X E A) is decomposable. It then suffices to apply 8.3.1. References: [638], [1004], [1007], [1036], [1037], [1041], [1101], [1105], [1458], [1680].

CH. 8, 4j

CENTRAL DISINTEGRATION

169

8.4. Central disintegration

8.4.1.
(~H()

Let Z be a Borel space, J.t a positive measure on Z, a measurable field of Hilbert spaces on Z, (~7T() a measurable field of representations of A in the H(), d() the strong closure of 7T()(A),
LEMMA.

f!l the algebra of diagonalisable operators, and d the strong closure of 7T(A). Suppose that f!l ~ d. (i) We have d = I6:J d() dJ.t(). (ii) If f!l is the centre of d, then 7T() is almost everywhere a factor

representation. (iii) If J.t is standard, there exists a negligible subset N of Z such that, for (, r E Z\N, (;t. (, we have 7T() 6 7T(().

Since f!l \: d, 7T is non-degenerate, and so 7T() is non-degenerate almost everywhere (8.1.5), hence d (resp. d( is also the von Neumann algebra generated by 7T(A) (resp. 7T)(A. Note that f!l \: 7T(A) =si, and hence f!l is contained in the centre of d. Let (x;) be a dense sequence in A. Then d() is the von Neumann algebra generated by the 7T()(X;). Moreover, d can be regarded as the von Neumann algebra generated by the 7T(Xj) and by f!l. Now 7T(Xi ) = I6:J 7T()(X) dJ.t(). Hence d =I6:J d() dJ.t() by A 88. If f!l is the centre of d, then d() is a factor almost everywhere (A 89), and so 7T() is a factor representation almost everywhere. Let T = I6:J T() dJ.t() Ed. Since T is a strong limit of elements of 7T(A), there exists a negligible subset N(T) of Z and a sequence (XI> Xz, ...) in A such that, for (E Z\N(T), 7T()(X) converges strongly to T() (A 13 and A 79). Let, then, (, E Z\N(T), and let R be an intertwining operator for 7T() and 7T(). The equality

R7T()(X;) = 7T()(x;)R,

which holds for every i, implies that RT() = T()R. This established, suppose that Z is standard and let (Y" Y z, ) be a sequence of Borel subsets of Z separating the points of Z. Let ({)j be the characteristicfunction of Y j ; <Pi defines an element E, of f!l. The union N of the N(E) is negligible. Let (, E Z\N be such that (;t. Let R be an intertwining operatorfor 7T(n and 7T(). By what we have seen above,

r.

170

INTEGRATION AND DISINTEGRATION OF REPRESENTATIONS

[CH. 8, 4

we have, for every i,

REM) = EM)R,

i.e. <f>M)R = cPM)R.

=

Now there exists i such that cPM) = 1, cPM) = 0, whence R 7T(l) 0 7T({). This proves (iii). 8.4.2.

O. Hence

THEOREM. Let H be a separable Hilbert space, 7T a non-de generate representation of A in H, and :!l the centre of the von Neumann algebra generated by 17"(A). (i) There exist a standard measure IL on A, a u-rneasurable field ~ ~ H(~) of Hilbert spaces on A, a u-measurable field ~ --; 7T(~) of non-trivial factor representations of A in the H(l) such that 7T({) E~, and an isomorphism Wof H onto PI! uu: dlL(~) which transforms s: into the algebra of diagonalisable operators and 7T into fEB 7T(t) dl-(n (ii) Let ILl ~ ~ Ht), 7T ~ 7Tt), WI have the same properties. Then I- and ILl are equivalent. After modification of the HI(~) and the 7TI(~) on a negligible set, there exists an isomorphism ~ ~ V(~) of ~ --; H(~) onto ~ --;Ht) such that V(~) transforms 7T(~) into 7Tt) for every ~.

To 17" and !l we apply theorem 8.3.2 and make use of its notation Z, ~ ~ H(o, ~ ~ 7T(C); there exists an isomorphism of H onto fEB H(~) dl-'(~) which transforms !l into the algebra of diagonalisable operators and 7T into fEB 7T(~) dlL{~). Since 7T is non-degenerate, :!l is contained in the strong closure of 7T(A). By 8.4.1, almost all the 7T() are mutually disjoint factor representations. By 8.1.5, almost all the 7T(~) are non-trivial. After subtracting from Z a negligible set, we can therefore suppose that the 7T(C) are mutually disjoint non-trivial factor represen tations. Then, by the isomorphisms on the H(~), we can suppose that ~ ~ 7T(t) is a Borel mapping of Z into Fac(A) (8.1.8). Since the 17"(~) are mutually disjoint, this mapping is injective; it is therefore (B 21) a Borel isomorphism 8 of Z onto a Borel subset B of Fac(A), which meets each quasi-equivalence class in at most one point. Let rjJ be the restriction to B of the canonical mapping Fac(A) ~ A; this is a Borel isomorphism of B onto a Borel subset C of A (7.2.3). Then e 8 is a Borel isomorphism of Z onto C. We can therefore replace Z by C and make the hypothesis that the quasi-equivalence class of 7T(~) is ~ for every ~ E C. We now have (i), extending IL to A by IL{A\C) = O. We introduce the notation ILr- ~ ~ Ht), ~ ~ 7T1(C), WI of (ii). We apply 8.2.4, with Z = z, = A, which yields N, N I , iii T/ and the V(~). For ~ E Z\N, V(~) transforms 7T(~) into 7Tt(T/(m. Hence the quasiIL,
0

CH. 8, 4l

CENTRAL DISINTEGRATION

171

equivalence classes (, 1/() of 1T(), 1T\(1/U are equal. Thus, 1/ is the identity mapping of Z\N. Hence JL and JLI are equivalent, and we have proved (ii). 8.4.3. The class of the measure JL on by 1T.
DEFINITION.

A is therefore uniquely determined

This is said to be the class of measures associated with

1T.

The disintegration
EB

1T =

J1T(n dJL(}
A

of theorem 8.4.2 is called, by a slight abuse of language, the central disintegration of 1T. 8.4.4. PROPOSITION. Let 1T, p be two non-degenerate representations of A in separable Hilbert spaces. Then 1T = P if and only if the classes of measures on A associated with 1T and p are equal. Suppose that the classes of measures on A associated with 1T and p are equal. Let 6,!J (resp. ~) be the centre of the von Neumann algebra d (resp. @) generated by 1T(A) (resp. p(A. There exist a standard measure JL on A, a measurable field (~H() (resp. (~K( of Hilbert spaces on A, a measurable field (~1T() (resp. (~p( of factor representations of A in the H() (resp. K(, such that 1T() E ( (resp. p() E (), 1T = fEB 7T() dJL() (resp. p = fEB p() dJL(m, 6,!J (resp. ~) being the algebra of diagonalisable operators (all this up to spatial isomor phism). Since 1T() =p() for every (, there exists an isomorphism ~() of the von Neumann algebra d() generated by 1T()(A) onto the von Neumann algebra @() generated by p()(A) such that
~()(

1T()(X

= p(}(x)

for every x E A.

We have (8.4.1) We show that

(~~()

is a measurable field of isomorphisms (A 91). Let

T = fEB T() dJL() E d. There exists a sequence (xJ in A such that (Xj)

172

INTEGRATION AND DISINTEGRATION OF REPRESENTATIONS

[CH. 8, 4

converges strongly to T. By taking a subsequence, we can suppose that 1T({)(X;) converges strongly to T({) almost everywhere. Then cP({) (1T({)(X;) converges strongly to <P({)(T({) (A 1 and A 27). Since the fields (~cP({)(1T(O(X;)) = p({)(x;) are measurable, we see that the field (~cP({)(T({) is measurable, which latter proves our assertion. Then cP =I fIJ eJJ(O dp,({) is an isomorphism of si onto {3 (A 91) which, for every x E A, transforms
1T(X) =

fIJ

7T(O(X) dp,({)

into

fIJ

p(O(x) dp,(O = p(x).

Hence 1T "" p. Suppose that 1T "" p. It is clear that, if 1T = I fIJ 1T({) dp,({), and if n is a cardinal ~ we have n1T = I fIJ n1T({) dp,({). Hence the classes of measures associated with 1T and n1T are equal. Now "01T = "oP (5.3.8), and so the classes of measures associated with 1T and P are equal.

"0

8.4.5. PROPOSITION. Let 1T, P be two non-degenerate representations of A in separable Hilbert spaces. Let C and D be the classes of measures on A associated with 1T and p. Then p is quasi-equivalent to a sub representation of 1T if and only if D is absolutely continuous with respect to C. Suppose that p is quasi-equivalent to a sub-representation pi of 1T. Let E E 1T(A) be the projection corresponding to pi, :!l the centre of 1T(A), E I E:!l the central support of E, and PI the sub-representation of 1T corresponding to E t- We have p = p "" PI (condition (v) of 5.3.1), and so D is also the class of measures associated with PI (8.4.4). Now let 1T = If 1T() dp: () be the central disintegration of 1T; we have p, E C. Let Y be the p,-measurable subset of A whose characteristic function 4>y corresponds to E I We have
PI

f
A

fIJ

1T({) d(jy . p,)(),

hence D is the class of 4>y p" and hence D is absolutely continuous with respect to C. Suppose that D is absolutely continuous with respect to C. Let p, E C. There exists a wmeasurable subset Y of A such that 4>yp, E D, where

CH. 8, 5]

DISINTEGRATION

INTO IRREDUCIBLE REPRESENTATIONS

173

<py denotes the characteristic function of Y. Let 7T = If 7T(n dp.,(t) be the central disintegration of 7T. Let E I be the diagonalisable projection corresponding to <py, and PI the sub-representation of 7T corresponding to E I . The class of measures associated with PI is the class of <py p., and is therefore D. Hence P = PI (8.4.4),
8.4.6. The second part of the above proof proves that, if a class of measures C on A is associated with a representation of A (in a separable Hilbert space), every class of measures absolutely continuous with respect to C is also associated with a representation of A. 8.4.7. PROPOSITION. Let 7T, p, C, D be as in 8.4.5. Then disjoint if and only if C and D are mutually singular.
7T

and pare

In fact, to say that 7T and P are non-disjoint means that there exists a non-trivial representation quasi-equivalent to a sub-representation of 7T and to a sub-representation of p. For this to be so, it is necessary and sufficient that there exist a non-empty class of measures absolutely continuous with respect to C and with respect to D (8.4.5 and 8.4.6), i.e. that C and D mutually singular.
8.4.8. PROPOSITION. Let 7T be a non-degenerate representation of A in a separable Hilbert space, and 7T = 7T(t) dp.,(t) its central disintegration. Then 7T is of type I, if and only if almost all the 7T(t) are of type I, or, in other words, if and only if p., is concentrated on A.

I1

Let d, d(n be the von Neumann algebras generated by 7T(A), 7T(t)(A). We have d = I EB d(t) dp.,(t) (8.4.1) (i, and so d is of type I if and only if all the d(t) are of type I (A 90). References: [559], [685], [1102], [1103].
8.5. Disintegration into irreducible representations

8.5.1. LEMMA. Let Z be a Borel space, p., a positive measure on Z, t ~ H(t) a measurable field of Hilbert spaces on Z, t ~ 7T(t) a measurable field of representations of A in the H(t),
EB EB
7T =

H = and
~

JH(t) dp.,(t),

J7T(t) dp.,(t),

the algebra of diagonalisable operators. Then the following two

174

INTEGRATION AND DISINTEGRATION OF REPRESENTATIONS

[CH. 8, 6

conditions are equivalent: (i) f!l is a maximal commutative von Neumann subalgebra of 7T(A); (ii) almost all the 7T({) are irreducible.

Let d, d() be the von Neumann algebras generated by 7T(A), 7T()(A). Let (xJ be a dense sequence in A. Then the 7T(Xj) generate the von Neumann algebra d, the 7T()(Xj) generate the von Neumann algebra d((), and 7T(Xj) = fEIJ 7T()(XJ dJL(() Hence (A 86), f!l is a maximal commutative subalgebra of 7T(A) = d, if and only if d() = .2(H({)) almost everywhere, i.e. if and only if 7T({) is irreducible almost every where. 8.5.2. THEOREM. Let H be a separable Hilbert space, 7T a representation of A in H, and f!l a maximal commutative von Neumann subalgebra of 7T(A). Then there exist a standard Borel space Z, a bounded positive measure JL on Z, a measurable field { .... H({) of Hilbert spaces on Z, a measurable field { .... 7T({) of irreducible representations of A in the H({), and an isomorphism of H onto fEIJ H ({) dJL({) which transforms f!l into the algebra of diagonalisable operators and 7T into fEIJ 7T({) dJL(n This follows from 8.3.2 and 8.5.1. 8.5.3. Recall that, when 7T and f!l are given, the disintegration of 7T given in theorem 8.5.2 exhibits a certain degree of uniqueness (8.2.4). But when, 7T being given, we change fIl, this disintegration can change completely to the point where not one irreducible representation appears simultaneously in both disintegrations (ct. 18.9.8). (In this respect, the central disintegration behaves in a much more satisfactory way.) This pathology will nevertheless be ruled out in the case of postliminal C*-algebras. References: [638], [1004], [1005], [1007], [1036], [1037], [1041], [1101], [1105], [1458], [1680]. 8.6. The case of postliminal C*-algebras 8.6.1. Let A be a separable postliminal C*-algebra. Then A = A is a standard Borel space (4.6.1 and 7.3.7). Let An (n = I, 2, ... ,00) be the set of elements of A of dimension 1, 2, ... ,00; it is a Borel subset of A (3.6.3(i)). Let H; be the standard n-dimensional Hilbert space. There exists a unique Borel field {~ H ({) of Hilbert spaces on A (A 96) which

CR. 8, 6]

THE CASE OF POSTLIMINAL

C*-ALGEBRAS

175

. We will call it the reduces on An to the constant field defined by H n canonical field. For every positive measure /.L on A, the canonical field ? ~ H(?) is automatically endowed with a /.L-measurable field structure (A 97).
8.6.2.
LEMMA.

We retain the above notation

(i) There exists a field ? ~ 7T(?) (? E A) of representations of A in the H(n, such that 7T(n belongs to the equivalence class? for every? E A,

and which is measurable for every positive measure on A. (ii) Let ? ~ 7T\(() be another field of representations having the properties of (i). Then, for every positive measure /.L on A, we have

f
Z
=

EB

7T(() d/.L(?) =

EB

7Tt(?) dJL(n

There exists a Borel map ? ~ 7T(?) of A into Irr(A) such that, for every? E A, 7T(?) is in the class? (4.6.2). For every positive measure on Z, the field ? ~ 7T(() is measurable (8.1.8). This proves (i). Assertion (ii) follows from proposition 8.2.3, applied with Z\
=

A,

/.L

/.Ll>

H(?) = HM),

T/ being the identity mapping of

A.

8.6.3. Lemma 8.6.2 (ii) shows that, with each positive measure /.L on A, there is associated a well-defined equivalence class of representations of A (in a separable space: d. A 73).

f! ? d/.L(?).

DEFINITION.

This equivalence class of representations is denoted by

8.6.4. PROPOSITION. Let A be a separable postliminal C*-algebra, and /.L a positive measure on A. Let (~7T(()(( E A) be a u-rneasurabie field of representations of A in the spaces H(() introduced in 8.6.1, such that 7T(() belongs to the class? for every (E A. Let 7T = 7T(?) d/.L((), and let ~ be the algebra of diagonalisable operators. Then 7T(A) = ~.

f!

For each measurable subset Y of A, we will denote by E; the diagonalisable projection corresponding to the characteristic function of Y. Every open subset of A may be identified with the spectrum i of a closed two-sided ideal I of A (3.2.2). Let (x\, x 2 , ) be an approximate identity of the C*-algebra I (1.7.2). If ? E A\i, we have 7T(OXn = 0 for

176

INTEGRATION AND DISINTEGRATION OF REPRESENTATIONS

[CH. 8, 6

every n. If (E i, 7T() II is a non-degenerate representation of I, and so 7T()(Xn ) converges strongly to 1. Hence 7T(Xn ) converges strongly to Er (A 79). Thus, Ej belongs to the von Neumann algebra st generated by 7T(A). Taking limits, we deduce from this that E y E sti for every Borel subset Y of A. Hence ~ ~ st and, consequently, ~' d st. Moreover, by 8.5.1, ~ is a maximal commutative subalgebra of sd, in other words, s: n st ~~. Since ~' d ; we conclude from this that st ~~. However, the inclusion ~ ~ sti is obvious; hence sti =~. 8.6.5. THEOREM. Let A be a separable postliminal C*-algebra (i) For each positive measure JL on A, If (dJL() is a non-degenerate
multiplicity-free class of representations and the class of measures on A = A associated with this class of representations is the class of J.L. (ii) Let JL and J.Ll be two positive measures on A. Then

f
A

Ell

(dJL()

f
A

Ell

(dJLM)

if and only if JL and JLI are equivalent. (iii) Every non-degenerate multiplicity-free representation of A in a separable Hilbe~ space is of class If (dJL() for some bounded positive measure JL on A.

Let 7T = {dJL(n. By 8.6.4, 7T(A) is commutative, and so 7T is multiplicity-free. By 8.1.5, 7T is non-degenerate. Since ~ is the centre of the von Neumann algebra generated by 7T(A) (8.6.4), the class of JL is the class of measures associated with 7T. If JL and JLI are equivalent, we have

It

f
A

Ell

(dJL() =

f
A

Ell

(dJLM)

by 8.2.1. If

f
A

Ell

(dJL({)

f
A

Ell

(dJLM),

JL and JLI are equivalent by 8.4.2. (ii). Lastly, let 7T be a non-degenerate multiplicity-free representation of A in a separable Hilbert space. Let JL be a bounded element of the class of

CH. 8, 6j

THE CASE OF POSTLIMINAL C"-ALGEBRAS

177

measures associated with 7T, and p = J~ ~ dl--(O. Then, by (i), p is a multiplicity-free representation, and the class of measures associated with p is the class of 1--. Hence P"" 7T (8.4.4), and hence P = 7T(5.4.6). 8.6.6.
THEOREM.

Let A be a separable postliminal C*-algebra.

(i) Let 7T be a non-degenerate representation of A in a separable

Hilbert space. Then there exist mutually singular positive measures 1--1,1--2... , 1--00 on A, such that
7T

Ell

t dl--I(t) EB 2

Ell

t dI--2(C) EB

... EB "0 Cdl--oo(t).

Ell

(ii) If 1--;, I-'-~, ... , I-'-~ possess the same properties, I--n and IL~ are equivalent for every n. (iii) The class of measures on A = A associated with 7T is the class of 1--1 + 1--2+ .. + 1--00

Since H 7T is separable, we have 7T = 171EB 172EB ... EB 1700where 17 n is of multiplicity n (with the convention that 00 = "0)and where the 17 n are mutually disjoint (5.4.9 and 5.5.2). Hence 7T = PI EB 2P2 EB ... EB "oPoo, where the Pn are mutually disjoint and multiplicity free. By 8.6.5, p; = fEll t dILn(t), where u is a positive measure on A. Since the p; are mutually disjoint, the I--n are mutually singular (8.4.7). Let 1--;, IL~, .. . ,IL~ have the properties of (i), and P~ =fEll t dIL~(t). We have 17 = P; EB 2p~EB ... EB "'oP~, and the P~ are mutually disjoint (8.4.7). By 5.4.9, np; = np~. Hence I--n and I-- ~ are equivalent for every 11. Let I-- = ILl + IL2 + ... + ILoo. Then 17 "" fEll CdIL(C), hence the class of measures on A associated with 7T is the class of IL. 8.6.7. We see that, in order to know a representation of A up to equivalence, we have to know a class of measures together with certain "multiplicities". In the central disintegration, only the class of measures is relevant. This is the result of studying representations up to quasi equivalence, about which, questions of multiplicity do not come up (remember that a representation is quasi-equivalent to all its multiples). 8.6.8. PROPOSITION. Let A be a separable postliminal C*-algebra. Let 17 be a non-degenerate representation of A in a separable Hilbert space, I- an element of the class of measures associated with 17, and S the support

178

INTEGRATION

AND DISINTEGRATION

OF REPRESENTATIONS

[CH. 8, 7

of p. (i.e. the smallest closed subset of A whose complement is negligible). Let pEA. (i) p is weakly contained in 1T if and only if pES. (ii) p is contained in 1T if and only if p.({p}) > O.

We write 1T =If 1T() dp.(), where 1T(~) is of class t. Suppose that S. As S is closed, there exists an x E A such that p(x) -:I 0 and O(x) = 0 for 0 E S. Then 1T()(X) =0 for almost every (, and so 1T(X) =O. Hence p is not weakly contained in 1T. Now suppose that pES. Let x E A be such that 1T(X) =O. Let U be the open subset of A consisting of the (E A such that 1T(~)(X) -:I O. Since 1T()(X) = 0 almost everywhere, U is negligible, and therefore disjoint from S. Hence p(x) =0 and p is weakly contained in 1T. We have therefore proved (i). Let Ep be the Dirac measure at p. To say that p.({p}) > 0 means that any p.-negligible set is Ep-negligible, in other words that Ep is absolutely continuous with respect to p., in other words, that p is quasi-equivalent to a sub-representation pi of 1T (8.4.5). However, every representation quasi-equivalent to an irreducible representation contains it (5.2.1 and 5.3.1). The above condition also means, therefore, that p is equivalent to a sub-representation of 1T.
p~

8.6.9. We retain the notation of 8.6.8. We see that the support of 1T(3.4.6) is just the support of u, We will say that a subset P of A carries 1T if P carries u, in other words, if p.(A\P) = O. Naturally, S carries 1T but, in general, there exist sets strictly smaller than S which also carry 1T. However, if pEA is such that p.({p}) > 0, every set carrying 1T necessarily contains p. References: [585], [685], [1007].

8.7. An interlude

Since the decompositon of representations into irreducible represen tations was one of our major objectives, we are going to pause for a moment to sum up the main results obtained. For finite-dimensional representations, we have the ideal situation: a finite-dimensional representation can be decomposed into a finite sum of irreducible representations, and two such decompositions are isomorphic. We now pass on to infinite-dimensional representations, restricting

CH. 8, 7j

AN INTERLUDE

179

ourselves (for this is the case we can best deal with) to separable C*-algebras and to representations in separable spaces. The simplest (even commutative) examples show that it is necessary to replace the idea of direct sum by that of direct integral. This corresponds to the well-known fact that the decomposition of hermitian operators, in infinite dimensions, brings in "infinitely small" spectral projections. If A is a separable postliminal C*-algebra (an important case for applications to the theory of topological groups), we have a quite satisfactory theory. The set A of irreducible representations is a locally quasi-compact space with a countable base, which is almost locally compact (it is locally compact if A is a C*-algebra with continuous trace). The Borel structure defined by the topology, which coincides with the Mackey structure, is standard. Every representation is, uniquely, the sum of representations of multiplicity 1,2, ... ,00. Every representation of multiplicity n is the sum of n equivalent multiplicity free representations, whose class is uniquely determined. Also, every non-degenerate multiplicity-free representation 7T may be written as an where the class of IL" is uniquely determined by 7T. integral fA ~ dlL,,(~) The mapping 7T ~ class of IL" defines a bijection of the set of classes of non-degenerate multiplicity-free representations onto the set of classes of measures on A. If A is a separable non-postliminal C*-algebra, things are not so good. The topological space A, the Borel structure given by its topology, and the Mackey Borel structure (now distinct from the topological Borel structure) have pathological properties (d. 9.5.6). Every representation can indeed be written as an integral of irreducible representations, but the uniqueness completely disappears. Moreover, since there exist re presentations of types II and III (9.5.4), it is not a matter of having to consider only multiplicity-free representations. We introduce the quasi spectrum A, which is not a topological space, but a Borel space (and which reduces to the Borel space A when A is postliminal). This Borel space is at least as pathological as A, since it contains A as a Borel subset. However, every representation 7T can be written canonically as an integral over A with respect to a standard measure whose properties adequately reflect those of 7T. This defines a bijective correspondence between the quasi-equivalence classes of representations of A and certain classes (not all, unfortunately) of standard measures on A. We reiterate that 7T admits a canonical decomposition on A, but not a unique decomposition, since uniqueness is already absent on A. For

180

INTEGRATION

AND DISINTEGRATION

OF REPRESENTATIONS

[CH. 8, 8

example, if 7T is a factor representation, its canonical representation may be written 7T = 7T; but sometimes 7T may be written as an integral of irreducible representations in many different ways. It remains to be seen whether or not a more satisfactory theory would be obtained by restricting attention to traceable representations; see [752] for recent results in this direction. 8.8. Disintegration of a positive form and of a trace

8.8.1.

THEOREM. Let f be a positive form on A, H the space of 7Tp Z a Borel space, p. a positive measure on Z, (~H() a measurable field of Hilbert spaces on Z, and (~7T() a measurable field of irreducible representations of A in the H() such that

EB

EB

JH() dJL(),

7Tf

J7T(0 dp.()

Let (~~(O be a field of vectors such that ~f = fEB ~() dJL(O (i) For almost every (, we have ~() # 0; let f(O be the positive form (pure for almost every 0 defined by 7T() and ~(n (ii) For every x E A, the function (~f(')(x) is integrable, and f(x)

J!()(x) dJL().

Let Y ~ Z be the measurable set of (E Z such that ~() # O. Let E y be the corresponding diagonalisable projection. For every x E A, we have

Ey7Tf(x)~f

Jq,y()7Ta)(X)~a)
I~(),

EB

dJL(),

where q,y denotes the characteristic function of Y. Since q,y() = 0 implies that ~() = 0, we see that Ey7Tf(x)~f = 7Tf(x)~f' But ~f is a cyclic vector for 7Tp hence By = I, and so ~a) # 0 almost everywhere. We have
f()(x) = (7T()(X)~()

and hence the function value

(~f()x

is integrable and its integral has the

CH. 8, 8]

DISINTEGRATION

OF A POSITIVE FORM AND OF A TRACE

181

8.8.2.

THEOREM. Let f be a lower semi-continuous semi-finite trace on i; Af, Pf OUf, Vf, tf A ". defining the entities (i) There exist: (1) a standard positive measure f.t on A; (2) for each A, a character f(); (3) a structure of measurable field of Hilbert spaces on the field, -+ H fm with the following properties: (a) H, may be identified with fEB H fm df.t(); the fields

u;

, -+ J f(), , -+ OUf(),

, ~ Af (), (), , ~ Vf

, ~ Pf} , ~ tfm

are measurable

and

t, =
=

f f
EB

EB

Jf) df.t(O,

Af =

EB

EB

Afm dp:(),

Pf =

f Pfm df.t(),

OUf

OUfW d/L(),

r. = f Vfm d/L(),

EB

tf

EB

tf() d/L().

The algebra OUf n Vf is the algebra of diagonalisable operators. (b) A/(/;) E , and ;fJ() E , almost everywhere. (c) For every x E A +, the function -+ f()(x) is measurable and f(x)

f o

df.t().

(ii) Let a measure f.t, characters /(,) and a structure of measurable field of Hilbert spaces on the field ,~HfW with the same properties as in (i), be given. Then f.t and u are equivalent; let f.t = d . f.t with d({) > 0 almost everywhere. We have f() = d(,)/,(,) almost everywhere. After altering the f(,r s on a negligible set, there exists an isomorphism of the measurable field '~Hf) onto the measurable field -+ Hfm which trans fonns Jf(() into JfU;) Af Ul into Af), Pfm into Pru OUf(() into OUf(), Vf(() into "Vf), and tf) into d()tf(()
Proof of (i). Let B be the Hilbert algebra ufl N, which is dense in H f, and let B be the corresponding full Hilbert algebra. There exist (A 84, A 94 and A 95): (1) a standard Borel space S; (2) a positive measure f.t on S; (3) a measurable field -+ H(,) of non-zero Hilbert spaces on S; (4) a measurable field '~B(') of full Hilbert algebras in the H(,), such that H, may be identified with fEB H(O df.t(O,

182
~

INTEGRATION AND DISINTEGRATION

OF REPRESENTATIONS

[CH. 8, 8

V/ with the algebra of B() d/L(n, 611/ and V/ with

diagonalisable operators, B with

EfJ

6fl() d/L()

and

EfJ

V(n du. ()

(where 6U() and V() denote the von Neumann algebras left- and right-associated with B(D), t, with f EfJ J() d/L) (where J(D denotes the involution defined by B(, t/ with I EfJ teD d/L(O (where teD denotes the natural trace on 6fl(t defined by B(. Since the algebra of diagonalisable operators is the common centre of 611/ and of V/, 6fl() and V(C) are factors for almost every ( (A 89). By 8.3.1, there exists a measurable field (--+A() (resp. (--+p of representations of A (resp. A~ in the H(D such that
EfJ EfJ

11./ =

A.() d/L(),

p/ =

p() d/L().

Let (x;) be a dense sequence in A. Since the A./(x;) generate the von Neumann algebra 611/, the A.(D(x;) generate, for almost every (, the von Neumann algebra 6fl() (A 88). Hence 6fl() is almost everywhere the von Neumann algebra generated by A.()(A). Similarly, V(D is almost everywhere the von Neumann algebra generated by p(O(A). If x E A . the function (--+ t()(A()(x is measurable and we have
(An)

lex)

= t/(A/x =
!()(x)

t(D(A.()(x d/L()

fo

d/L(),

putting

=t()(A.()(x

for every x E A+.

Let (y;) be a dense sequence in "/ in the sense of its pre-Hilbert structure (6.3.6). Since the A./(y;) generate the von Neumann algebra 6111 the A. ()(y;) generate, for almost every (, the von Neumann algebra 6fl(). Moreover,

!()(y;y";) d/L()

=!(YS";) < +00,

and hence !()(YiY";) < +00 almost everywhere. Hence (A.(), t( is a non-trivial trace factor representation of A for almost every (. Hence

CH. 8, 8j

DISINTEGRATION

OF A POSITIVE FORM AND OF A TRACE

183

almost all the fen are characters (6.7.3). Put AfYi = fIB Yi({) dp,(n. For almost every {, the yM) are dense in H({) (A 73). We have, for any i and j,

IB

A(n(Yi)Yj(n dp,(n = Af(yJAfYj = (AfyJ(AfYj)

IB

Yi(nYj({) dp,(n,

and so
A ({)(Yi)Yj(n

= Yi({)Yj({)

almost everywhere. Hence A({)(Yi) is almost everywhere the continuous operator extending left-multiplication by Yi({) and we have a cor responding result for p(O(YJ. Consequently, for almost all {, we have, for any i,
(yM) Yj({ = t({)(A({)(YJA({)(yj

and the latter expression is equal to

t({)(A({)(yiY; = f({)(YiYj).

If the Yi have been so chosen that they make up an involutive algebra over the field of rational complex numbers, we deduce, from the equality (Yi({) I yM = f({)(YiYj), that H({) may be identified with a closed subspace of H f ((). We can also impose on the Yi the condition that every product xiYj be a Yk Then the subspace H({) of H f(() is almost every where invariant under Af (() and Pf(l); since f({) is a character, Af(()(A) and pf(()(A) generate the von Neumann algebra .;E(Hf((), from which it follows that H, = H f (() (almost everywhere). We have
l({)Yi({) = (ltAfyJ({)

= y,\,({) = lfwyM)

almost everywhere, and hence l({) = lfw almost everywhere. Similarly,

A({)(YJYj({)

= Yi({)Yj({) = AfW(Yi)Yj({),

and so A({) and Afl coincide on Itf (almost everywhere), and con sequently, coincide on A, since A({) and p(O are factor representations and non-trivial on Itf (2.10.4 and 2.11.1). Similarly, p({) = Pf((). It follows from this that OiL(l.) = OiLf cel , Lastly, for every xEA+, we have

184

INTEGRATION AND DiSINTEGRATION

OF REPRESENTATIONS

[CH. 8, 8

almost everywhere. Writing this for x = yyt, x = Y2Y~' ... , and applying A 29, we conclude that t({) = tf m almost everywhere, After subtracting from Z a negligible set, the A({) are mutually disjoint (8.4.1). Just as in the proof of 8.4.2, we can replace S by a standard Borel subset of A, and arrange that Af m = A({) E { almost everywhere. Then pJW = Am) implies pJm E {, and all the assertions of (i) are proved.
Proof of (ii). The fact that p., and p., are equivalent follows from 8.4.2(ii). Therefore, let p., = d . p." with d({) > 0 almost everywhere. For almost every {, f({) and !({) are the characters of two quasi-equivalent representations, and so there exists a number k({) > 0 such that !({)= k({)f({). For any x E A +, we have

f({)(x) dp.,({)

= f(x) =

!,({)(x) dp/({)

k({)f({)(x) d({) dp.,({),

whence, for x

= YiYi, we

get

N ow the equalities
yM) = Y2({) = ... = 0

define a p.,-negligible subset of

{~(y;(()

A. Since

the functions

Iy;({))k({) d({)

are p.,-measurable, we see immediately that kd is measurable. The function kd defines, by multiplication in H, = f EfJ H({) dp.,({), a self adjoint positive operator T. From (1), we have II T 1/ 2 ydl = IIYili for every i, whence T I / 2 = 1 and kd = 1 almost everywhere. Hence f({) = d({)!({) almost everywhere. Then the mapping x ~ d({)1/2 X of "fW into itself defines an isomorphism V({) of H fm onto Hfm almost everywhere, which gives rise to an isomorphism of the field {~Hfm onto the field { ~ H rm which clearly possesses the properties of (ii). 8.8.3. We retain the notation of 8.8.2. Let (z.) be a sequence of elements of mI The functions {~f({)(z;) are p.,-integrable, and therefore finite almost everywhere. Hence there exists a negligible subset N of A such that z, E {EAIN mfW for every i.

CH. 8, 8l

DISINTEGRATION

OF A POSITIVE FORM AND OF A TRACE

185

(n teA mfW)

In particular, after altering the f() on a negligible set, mf n is dense in m f, in the norm-topology of A.

8.8.4. The measure JL and the characters f() constructed in 8.8.2. possess the following additional property: let x E nf h a bounded JL-measurable function on A, and ~h the diagonalisable operator cor responding to h. Then

(~hAtx

IAtx) =

J
A

h()f()(x*x) df..L().

In fact, put Atx = f$ x() dJL). We then have

(.1 hA tx IAtx) =

(h()x() Ix()) df..L().

Now, we saw, during the proof of 8.8.2 (i), that

(x() Ix())

= f()(x*x) almost everywhere.

8.8.5. Let A be a separable postliminal C*-algebra. Let ( - K() be the canonical field of Hilbert spaces on A (8.6.1). Choose a field ( - 7T() of representations of A in the K(), such that 7T() belongs to the class ( for every, and which is measurable for every positive measure on A (8.6.2). For the sake of simplification, we will write ( instead of 7T(). Let K() be the Hilbert space conjugate to K(). We will denote by ?01 the representation x - nx)01 of A in K()0K(), and by 10(the representation x - 10 (x) of A in K() 0 K(). Let J, be the canonical involution of K(,) 0 K(), and t, the trace T 01- Tr T on
((K()) 0

ct.

THEOREM. Let f be a lower semi-continuous trace on A + such that mf is dense in A. Then there exist a positive measure JL on A and an isomorphism W of onto f$ (K() 0 K()) df..L() with the following properties:

n,
$

(i) W transforms
E9

Jf
-0

into
$

JJ, dJL(),

Af into
OUf

J 0
$

1) dJL()

Pf

into

J00 ,) df..L(),

into

J((K())0C)df..L(),

186

INTEGRATION

AND DISINTEGRATION OF REPRESENTATIONS

[CH. 8, 8

v,

into

EB

(C

::t(K() dJL(),

tf

into

EB

t, dJL(?),

and OUf n V f into the algebra of diagonalisable operators. (ii) If x E A +, the function (~Tr (x) on A is lower semi-continuous, and we have f(x) =

JTr ?(x) dJ,L(?).

We bring in all the notation of 8.8.2. Since A is postliminal, the Borel space A may be identified with the Borel space A (7.3.7). For x E A+ and ( E A, put !,()(x) = Tr (x). By 8.8.2 (i) b, there exists, for almost every (, a number k(?) > 0 such that !() = k()f(?). For any x E A ". the function (~f()(x) is measurable (8.8.2 (i)c) and the function (~ !U)(x) is lower semi-continuous (3.5.9). For every (0 E A, there exists an x E m, such that f(n(xx*) > 0 at (0 and therefore throughout an open is in neighbourhood VO of (0; moreover, the function (~f()(xx*) tegrable (8.8.2 (i)c) and therefore finite almost everywhere; hence the function (~ k() is measurable in V0; covering A by a sequence of sets VO (3.3.4), we see that the function k is measurable. Replacing JL by k-JL, we can suppose, from now on, that !(?) = fU) for every (. There then exists an isomorphism of Hfm onto K(n K() which transforms Af m into (I, pJ({) into I ~, OU f ({) into ::t(K(C, V m ) into C::t(K(?, Jf ({) into J, and tf ({) into t, (6.7.7). It follows immediately from the above that dim Hf ({) =dim K(O K(?) for every t,

and hence (A 72) that the measurable fields ( ~ H f m and ( ~ K(O K() are isomorphic. For each (E A, then, let lJ() be the set of isomorphisms of the Hilbert space H f ({) onto K(?) K() which transform Af({) into ( I, pJw into 1 (, Jm ) into J, (and hence OUf W into ::t(K( C, Vj({) into C ::t(K (m, tf ({) into t,). We have seen that lJ()": 0. There therefore exists (A 82) an isomorphism (~ W() between the two fields such that and W() E lJ() for every (. The identification of the field (~Hf({) (~ K(?) K() can therefore be made in such a way that
Af(C)

= (

I,

pJUJ = I

c.

Jf(C)

=J,.

The theorem is then a consequence of 8.8.2 (i).

CH. 8, 8l

DISINTEGRATION

OF A POSITIVE FORM AND OF A TRACE

187

8.8.6.

THEOREM. We retain the hypotheses and notation of 8.8.5. Let JL be a positive measure on A such that, for each x E A +, we have

f(x) Then JL = JL.

Tr

dJL(C).

Let G be an open subset of A, CPG its characteristic function, I the closed two-sided ideal of A such that i = G, and (Ul U2, ... ) an in creasing approximate identity of the C*-algebra I (1.7.2). Let x E A ". If CE A\G, we have C(xI/2unxI/2) = 0 for every n; if CE G,
C(xI/2unxI/2) = C(X)1/2C(U n)C(X)1/2

converges strongly, in an increasing fashion, to C(x), and hence Tr C(Xl/2UnXI/2) converges, in an increasing fashion, to Tr C(x). Now

Tr C(x 1/2 UnX 1/2) dJL(C)

=f(x 1/2 UnX 1/2) =

Tr C(XI/2UnXI/2) dJL(C),

and hence, in the limit,

cPG(C) Tr C(x) dJL(C) = {cPG(C) Tr C(x) dJL(C)

If x E m/, this formula extends to the case where G is a closed subset of A, and then, taking limits, to the case where G is an arbitary Borel subset of A. Hence, if x E m/, the measures Tr C(x) dJL(C) and Tr C(x) dJL (C) are equal. Let (x IX2, ... J be a dense sequence in Let Me) =Tr C(x i ) For every Co E A, there exists an i such that hi(e) > 0 at Co and hence throughout some neighbourhood of Co. Since hiJL = hiP. for every t, p. and p. are equivalent. Let u = ku. with 0 < h (C) < +00 almost every where. We have hi = hik almost everywhere, whence k = 1 almost everywhere, and p. = u.,

mI

8.8.7. We retain the hypotheses and notation of 8.8.5. Let x E"p let h be a bounded JL-measurable function on A, and .::1 h the corresponding diagonalisable operator. By 8.8.4, we have
(.::1 hA tx IAtx)

J
A

h(C) Tr (C(x)*C(x)) dp.(C).

References: [638], [639], [689], [1033], [1036], [1037], [1041], [1101], [1105], [1458], [1522], [1680], [1681], [1682], [1683], [1722], [1736], [1869].

188

INTEGRATION AND DISINTEGRATION OF REPRESENTATIONS

[CH. 8, 9

8.9. Addenda

*8.9.1. Let A be a separable C*-algebra, Z a standard Borel space, u. a positive measure on Z, ~ ~ H(~) a t-t-measurable field of Hilbert spaces on Z, t ~ '7T(~) a t-t-measurable field of representations of A in the H(t), ffi 7T =I '7T(~) dt-t(~) and ~ the algebra of diagonalisable operators. Then the 7T(n are homogeneous almost everywhere if and only if ~ contains the ideal centre qy associated with 7T (5.7.6). Further, the 7T(t) are homogeneous and have mutually distinct kernels (after removal of a negligible set), if and only if ~ =1iJJ. [509]. 8.9.2. Let A be a separable C*-algebra. If A is not postliminal, there exist standard measures on A whose class is not associated with any representation of A in a separable space. [454], [559]. *8.9.3. Theorems 8.8.5 and 8.8.6 can be generalised to non-separable postliminal C*-algebras. [452]. 8.9.4. (a) Let A be a separable C*-algebra. The canonical Borel field of Hilbert spaces on A is defined as in 8.6.1. For each standard positive measure t-t on A, there exists a Borel mapping ~ ~ 7T(~) of A into Irr(A) such that 7T(t) E ~ almost everywhere. The class of I ffi 7T(t) dt-t(l) depends only on the class C of u ; we denote it by 9ll(C). (b) For every class R of representations of A of type I, we denote by <(R) the associated class of standard measures on A. If R is a class of multiplicity-free representations of A, we have R = 9ll(<(R. (c) If C is a class of standard measures on A, and if 9ll(C) is of type 1 then 9ll(C) is multiplicity-free and <(911(C = c. *(d) If A is not postliminal, there exist two distinct classes C, C of standard measures on A such that 9ll(C) = 9ll(C) is a factor represen tation of type II. [454], [1003], [1007], [1868]. Problems: What are the classes C of standard measures on A such that 9ll(C) is of type I? What are the equivalence classes of represen tations of the form 9ll(C), where C is a class of standard measures? Is every representation quasi-equivalent to an element of a class ~(C), where C is a class of standard measures? *8.9.5. Let A be a separable C*-algebra. We define the canonical Borel field of Hilbert spaces on A to be equal to the canonical field on A, and equal to the constant field, defined by a Hilbert space of dimension ~o, on A\A. If t-t is a standard positive measure on A, there exists a Borel

CH. 8, 9l

ADDENDA

189

mapping {~7T({) of A into Fac(A) such that 1T({) E { almost every where. The quasi-equivalence class of fEB 7T({) dfJ-({) depends .only on the class C of fJ-; we denote it by 9ll(C). For each quasi-equivalence class R of representations of A, we denote by cg(R) the associated class of standard measures on A. We have 9ll(C(J(R = R. If A is not postliminal, there exist classes C of standard measures on A such that C(J(911(C # c. [454], [509], [559], [1007]. 8.9.6. Let A be a separable C*-algebra, 7T a representation of A in a separable Hilbert space, and 7T = fEB 7T({) dfJ-() a disintegration of 7T into irreducible representations. If 7T is a factor representation of type I, almost all the 7T() are equivalent. [1006].

CHAPTER 9

C*-ALGEBRAS

OF TYPE I

9.1. Statement of the theorem, start of the proof

THEOREM. Let A be a separable C*-algebra. Then the following conditions are equivalent: (i) A is of type I; (ii) Every factor representation of A is of type I; (iii) A is postliminal; (iv) If 7T is an irreducible representation of A, 7T(A) contains the set of compact operators on H.,,; (v) If 7Tl and 7T2 are two irreducible representations of A with the same kernel, then 7Tl and 7T2 are equivalent.

(For the case where A is not separable, cf. 9.5.9.) The pattern of the proof is as follows: (iii) ~ (iv) ~ (v) ~ (ii) ~ (iii) ~ (i) ~ (ii). (iii) ~ (iv): d. 4.3.7. (iv)~(v): d. 4.1.10. (iii) ~ (i): d. 5.5.2. (i) ~ (ii): Obvious. (v) ~ (ii): Suppose that (v) is satisfied. Let 7T be a factor represen tation of A; we show that 7T is of type I. We can assume that H." = His separable (5.3.7). We apply theorem 8.5.2, whose notation we adopt. For every IL-measurable subset Y of Z, let E y be the corresponding di agonalisable projection; it commutes with 7T(A), and therefore defines a subrepresentation 7Ty of 7T. If /L(Y);c 0, we have Ey;c 0, hence 7Ty = 7T (5.3.5), and so
117T(X)11

= II7Ty(x)11 = IIEy7T(x)1I

for every x E A

(condition (ii) of 5.3.1). For n

= 1,2,

.. . , let Y. x be the set of ~ E Z such

CH. 9, 2]

PRELIMINARIES CONCERNING SYSTEMS OF MATRIX UNITS

191

that 117T({)(x)ll,,:;; 117T(X)II- tln. We have

IIE yn.x7T(x)ll,,:;; 117T(X)II- n

(A 77),

and so Yn,x is negligible by the above. This being true for every n, we conclude that /I7T()(x)11 = /I7T(X)/1 except for a negligible set NX Let D be a countable dense set in A. Then N = U xED N, is negligible. If C EN, we have /I7T({)(X)/1 = /I7T(X)/1 for every xED, and therefore for every x E A; hence Ker 7T(C) = Ker 7T. By (v), almost aII the 7T() are equivalent to a single irreducible representation 7To. Hence 7T is a multiple of 7To (8.1.7), and is consequently of type I (5.4.11). It remains to prove (ii) ~ (iii), and this requires a considerable amount of preparation. References: [443], [583], [631], [896], [897].
9,2. Preliminaries concerning systems of matrix units

We set out these preliminaries in greater detail than is necessary for the sequel, in order to throw some light on a subsequent proof of a very technical nature. 9.2.1. We endow CP with its canonical p-dimensional Hilbert space structure:
ZI"

,zp) I(z], ... ,z~

= ZI; +... + zp~.

The algebra Mp(C) of complex p x p matrices may then be identified with the C*-algebra 5t(C P ) , which is simple and of dimension p2. Let (eii) be the canonical basis of Mp(C). We have ek/eji = il!eki eJ = .eii where ill is the Kronecker delta. 9.2.2. More generaIIy, let A be a C*-algebra, and (fi)""i,i<p a family of elements of A. If fkJij = ilJki and fJ =fji this family is caIIed a system of matrix units of order p, The linear combinations of the fii then form a sub-C*-algebra B of A. The linear mapping of Mp(C) onto B which maps eij to fij for 1,,:;; i, j,,:;;p is a morphism; since Mp(C) is simple, this morphism is an isomorphism if B ~ O. Suppose that A is realised as a C*-algebra of operators in a Hilbert space H. Since [ = ft =f~, the fii are projections. It can be verified that fii is a partial isometry with initial projection f ii and final projection fii

192

C-ALGEBRAS OF TYPE]

ICH. 9, 2

Put Vj = fj,(j = 1, ... , p). We have fij = ViV~, The Vj are partial isometries all having v, = f" as initial projection; their final projections are mutually , vp be orthogonal, i.e. we have ViVi = 0 for j:;6 i. Conversely, let V" partial isometries of A, all having v. as initial projection, such that VjVi = 0 for ;:;6 j. We can easily check that, if we put fij = ViV~, the fij form a system of matrix units of order p. Geometrically, there exist mutually orthogonal closed subspaces H" ... , H; of H, Vj is a partial isometry whose initial subspace is H, and whose final subspace is H j, and v, is the projection on H,. 9.2.3. Let a ... . . ,an denote variables each taking either the value 0 or the value 1. Let (V(aI, ... , ana) ..... anE{O, 1\ be a system of partial isometries of A, all having v(On) as initial projection (where On denotes the n-tuple consisting entirely of zeros), such that

Then the v(b., . . , bn)v(a., ... ,a n)* form a system of matrix units of order 2". If Hta., ... , an) is the final subspace of v(a .. . . , an), then the H(a], ... , an)s are mutually orthogonal, v(a l , , an) has H(On) as initial subspace, and v(On) is the projection on H(On) 9.2.4. Let

be two systems of elements of A possessing the above properties. Suppose that: (1) the projection v(On) is the sum of the projection v(On+!) and the final projection of v(On 1);
(2) v(a" , a n+,) = v(a" , an)v(On a n+,). Hta, , an), Hia., , an+!) are the final subspaces of v(a., ... ,an), v(a" . , an+!) this implies that H(a), ... , an) is the direct sum of Hta., , an> 0) and Hta, .. , an 1). We will say that the system of the v(a" , a n+.) is deduced from the system of the v(al" .. , an) by

If

bisection. 9.2.5. For n = 0, 1, 2, ... , let (v (a" ... , anat..... anE{O.I\ be a system of elements of A having the properties of 9.2.3, each system being deduced from the preceding one by bisection. (For n = 0, the system reduces to

CH. 9, 2j

PRELIMINARIES CONCERNING SYSTEMS OF MATRIX UNITS

193

the single element v(q, ).) Among others, we have the following proper ties: (i) if j"" k and (a ..... , aj) :;6 (b l , . . . ,b j), then v(al"" a)* x otb-, ... , bk ) = 0; (ii) if k > 1, then v(al"" ak) = v(al, . , ak_l)v(Ok_l, ak); (iii) if j < k, then v(a l, .. . , aj)*v(al ... , aj)v(Ok_I ak) = V(Ok_l, ak); (iv) V(Ok);;; O. If H(al,"" an) is the final subspace of v(al . . , an), (i) means that Hta., , aj ) and H(b l , , bd are orthogonal if j",. k and (ai , aj):;6 (b h . , bj); (iii) means that H(Oj) contains H(Ok) and H(Ok-h 1) for j < k. 9.2.6. In an arbitrary C*-algebra, we do not know how to construct non-zero systems of elements having all the properties of 9.2.5. Ne vertheless, under certain fairly wide conditions (cf. 9.3.7), we will be able to bring off the construction in A of non-zero systems of elements (v(al"" anal..... a.E{O,I) (n = 0, 1, 2, ...) satisfying conditions (i), (ii), (iii), (iv). For the present, we are going to indicate several consequences of properties (i), (ii), (iii). We will suppose that A is realised as a C*-algebra of operators in a Hilbert space H. We will put
e(n) =

~
al ... a.E{O. I}

v(a ..... , an)v(al"" a n)*,

which is an element of A ". Let

E(n)

=e(n)H =He

Ker e(n).

By (i), the v(al" .. , an)(H) (for n fixed) are mutually orthogonal, and so PE(n) is the sum of the supports of the v(a ... . . ,an)v(al"" a n)*.
9.2.7. The E(n) form a decreasing sequence. In fact, if that e(n)g = 0, we have

g E H is such

for any al"" an and hence v(al"" an)*g = 0 for any a l , hence v(al" . , an)*~ =0 by (ii), and so e(n + 1)g =O.

... ,

an

9.2.8. The operators v(ah ... , an)v(b .. . , bn)* leave E(n + 1) invariant.

194

C-ALGEBRAS OF TYPE

[CH. 9. 2

In fact, let TI E H and

(1)

=v(CI> ... , Cn+l)TI.

We have

v(a), ... , an)v(b., ... , bn)*~

= v(al> .

, an)v(b), . . . , bn)*V(CI ... ,cn)v(On Cn+I)TI c5 b:v(a),

c5~:v(a

= c5~:
= 8~:

, an)v(On, cn+)TI . , an cn+))TI

(,

and this belongs to E(n + 1) by 9.2.6. 9.2.9. The v(a(, ... , an)v(bl>" . ,b n)* induce in E(n + 1) a system of matrix units of order 2n (hence e(n) IE(n + 1) is a projection, hence e(n) IE(n + 1) = 1 by 9.2.7). For we have (9.2.8, formula (1)):
v(a), ... , an)v(b), . . , bn)*v(a;, , a:Jv(b;, , bn)*v(a;, , an)v(On cn+))TI , b:J*g , a:Jv(Om cn+))TI

= c5:i.. c5~~v(al"'"
=

an)v(b

c5~i ... c5:~c5:i

... c5:Zv(a l,

= 8:1c5:Zv(al>"
and so

, an)v(b;, .. . ,b:J*g

(v(al ... , an)v(b l, (v(a;,

=

, b n)* IE(n , a:Jv(b;, ,an)v(b;,

+ 1
, b:J* IE(n+ 1)) , b:J* IE(n

c5:1.c5:Zv(a),

+ 1).

Moreover, for the operators

v(a), ... , an)v(b l, ... , b n)*

and

v(b), ... , bn)v(a), ... , a n)*,

each is the adjoint of the other and both leave E(n + 1) invariant, and hence the same is true of their restrictions to E(n + 1). 9.2.10. We have
v(a), ... , an_l)v(b), ... , b n - I )* IE(n
= (v(a),

+ 1)
),

, an-I, O)v(b), , an-I 1)v(b l,

, bn-I> 0)* , bn-

+ v(al>

1)*) IE(n

+ 1)

In fact, in view of formula (1) of 9.2.8,

v(al>"" an-I, O)v(bl>" ,b n - I , O)*~+v(al"'" an-I l)v(b), ... , bn-I> 1)*g

= c5~: ... a~:=: c5~n v(a), , an-I 0, Cn+l)TI + c5 b :.. c5 b := :c5fn v(al> , an-I 1, cn+))TI = c5 b := :v(al, . . , a n- .. Cn cn+))TI :... 8 b

CR. 9. 2)

PRELIMINARIES CONCERNING SYSTEMS OF MATRIX UNITS

195

and
v(at . . . an-t)v(bl> ... , bn-t)*g , an-I)v(b t, ... , bn_I)*v(CI ... , Cn-t)v(On-1> cn)v(On c n+I)7] SZ:=: v(a t, ... , an-t)v(On-1> cn)v(0n> Cn+l)7] S~:=: v(al> .. , an-I> Cn> Cn+t )7].

= v(al>
= s~: = S~:

9.2.11. The v(a t, ... , an)v(b., ... ,b n)* leave E(r) invariant for r > n (this generalises 9.2.8). In fact, we know this is true for r - n = 1. Suppose that it is established for r - n =s ;;:. I, and consider the case where r = n + s + I (;;:. n + 2). By the induction hypothesis, the v(al> ,an+t)v(bl> .. ,b n+t)* leave E(r) invariant. By 9.2.10, the , an)v(bl>"" b n)* operate in E(n +2) as linear combinations of v(a t, operators v(al>"" an+l)v(b t, ... , b n+t)*, and therefore leave E(r) C E(n + 2) invariant. 9.2.12. Let the set of linear combinations of the and N(n) the set of linear combinations of elements of M(O) U M(l) U .. U M(n). All the elements of N(n) leave E(n + 1) invariant (9.2.11). Let tPn be the mapping x ~ x IE(n + 1), where x runs through N(n). By 9.2.10, we have
M(n)
lo

be

v(a., ... , an)v(b

,b n)*

tPn(N(n

= tPn(M(n.
,an

Suppose, further, that v(a l, . . . , an) :f. 0 for any n, aI

v(al, ... , an)v(b l, ... , bn)*v(bl> ... , b n+t)

Since

= v(al

, an b n+l)

(by formula (I) of 9.2.8 applied for

v(at, ... , an)v(b t, ...

Ct

=b ; ... , c, =b n ) , , b n)* IE(n + 1) :f. O.

we see that

In view of 9.2.9, there exists an isomorphism 1JIn of the involutive algebra tPn(M(n onto M 2"(C). For x E N(n), put In(x) = Tn Tr (1Jtn(tPn (x ). Then In is a linear form on N(n). We have
In(x*)

=In(x),

In (x) ;;:.0

for

x;;:. 0,

I.(v(at, ... , an)v(b t, . . , b n)*) = TnS:: ... S::.

Also 9.2.10, applied successively, proves that

In(v(al> ... , ap)v(b t, , b p)*) = T
P5::

... S:;

for p ~ n.

We have N(O) C N(l) C

, and the above proves that each of the

196

C*-ALGEBRAS

OF TYPE

[CH. 9, 2

forms In is an extension of the preceding one, and hence they collec tively define a linear form I on N = U n"O N(n) which possesses the following properties: (1) I(x*) =I(x); (2) !(x) ~ 0 for x ~ 0; (3) !(v(a., ... , an)v(bt> ... , b n)*) = Tn8~: ... 8~: for any n, at> ... , an b-, ... , bn 9.2.13. For n =0,1, ... , r, let (v(a l, ... , a.))alo ....a.E{O.I) be a system of elements of A satisfying conditions (i), (ii), (iii), (iv) of 9.2.5, and let V(Or+l) and v(O,. 1) be arbitrarily chosen elements of A. For aI ... , ar+1 E {O, 1} and (at> ... , ar) 0" put
v(a., ... , ar+l) = v(al, ,ar)v(O,. ar+l)

We have thus defined a system (v(at> , ar+!al ..... ar+1E{O.I). We check whether conditions (i), (ii), (iii), (iv) are still satisfied. It is necessary that
(1) v(at> ... , ar)*v(a l, ... , ar)v(O" ar+l) = v(O,. ar+!)

by (iii), that

(2)
by (i), and that
(3)

by (iv). We show that relations (1), (2), (3) are sufficient. Suppose that they hold. Let j < k os;;; r + 1, and let us prove (iii). If k os;;; r, (iii) is already assumed to hold. If k = r + 1, j = r, (iii) holds by (1). Suppose that k = r + 1 and j < r. We have, by (1) and (3),
v(O" ar+l) = v(Or)*v(Or)v(O,. ar+!) = V(Or)2 V(0" a r+!).

Using (iii) for systems with at most r indices, this yields

v(at> ... , a)*v(a l, ... , aj)v(O" ar+l) = v(a 1, , a)*v(a l, ... , aj)v(Or)2 v (0" ar+!)

= v(Or)(v(Or)v(O"

ar+!) = v(O" a r+ 1) .

We now prove (ii). It is enough to do this for k

v(at> ... , ar+l) = v(al

=r + 1, i.e. to prove that

, ar)v(O" ar+I) , ar+l) if (aI... , ar) 0,. Also,

Now, this is just the definition of v(at> the relation already noted,

v(O" ar+l) = v(Or)2 v(Or ar+l)

CR. 9, 3)

SOME LEMMAS

197

implies
v(O" ar+!) = v(Or)v(O" a r+ l)

since v(Or) ~ O. Hence (ii) is proved in all cases. We prove (i). It is enough to deal with the case where k = r + 1. If j < r + 1, we have
v(a" ... , aj)*v(b" ... , b r+,) = v(al, ... , aj)*v(b ... , br)v(O" b r+ ,) "

since (ii) has already been proved, and it is enough to apply (i) for systems with at most r indices. Suppose j = r + 1. We have, by (ii),
(4) v(a l, ... , ar+,)*v(b"
= v(O" ar+I)*v(a l,

, b r+ ,) , ar)*v(b" ... , br)v(On b r+, ) . a r+ 1 - b r+ ,.

This vanishes if (ai ... , ar) - (b" ... ,br). Lastly, suppose that
(a l, . . . , ar) =(b" ... , b r)

and

By (4), and (iii) which has already been proved,

and this vanishes by (2). Reference: [631]. 9.3. Some lemmas 9.3.1.
LEMMA. Let A be a unital C*-algebra, 17 a non-trivial irreducible representation of A, H a finite-dimensional subspace of Hi; L the set of states x ~ (17(X)g Ig) of A, where g runs over the unit sphere of H, and K the weak*-closed convex hull of L. Let K , be the set of states of A which take the value 1 for every x E A such that 0 OS;; x os;; 1 and 17(X) is the identity on H, Then K = K,.

lt is clear that K , is convex and weak*-c1osed, and that L ~ K, so that K ~ K ,. Let I be an extreme point of K ,. We will show that IE K. This, together with the Krein-Milman theorem, will imply that K I ~ K. Suppose that I =!(I, +,2) , I and 2 being states of A. If X is an element of A such that 0 os;; x os;; 1 and 1T(X) is the identity on H, we have
l(x) os;; I, llx) os;; I,

1 = I(x) = !(II(X) + Mx,

and so I(x) =2(x ) = 1; hence I" 12 E K I, and soil =2 = I; in short, I is a pure state. Now suppose that 171 is not equivalent to 17. There then exists

198

C*-ALGEBRAS OF TYPE

[CR. 9, 3

an hermitian element y of A such that 1T(y) reduces to the identity on H and 7T1(y)~, = 0 (2.8.3). Let f: R 4 R be the function such that f(t) = 0, t or 1 according as t e 0, O:s:; t :s:; 1, t ~ 1. Replacing y by f(y), we can ~I) = 0 suppose that O:s:; y:s:;1. But then the equality I(y) = (7TI(Y)~I' proves that l~ K I , which is a contradiction. Hence 7T1 = 7T, and I is the state defined by 7T and a unit vector e of H7f Let = 7] with 7] E H, ,E HeH . There exists an hermitian element x of A such that 7T(X) reduces to the identity on H and 1T(X), = 0 (2.8.3). Reasoning as before, we can suppose that O::s:;x:S:; 1. Since 1 E K I , we have

+,

But whence
, =

0,

= 7]

E H and

I E K.

9.3.2. LEMMA. We retain the notation of lemma 9.3.1. Let E be the set of states of A. For every 8 > 0, and every x E A such that O:s:; x :s:; 1 and 7T(X) is the identity on H, let VX.8 denote the set of fEE such that f(x) ~ 1 - 8. Then the V x 8 s form a base of neighbourhoods of K in E for the weak*-topology.
It is clear that V x 8 is a closed neighbourhood of K in E. Moreover, the intersection of all the Vx./s is K by 9.3.1. Since E is compact, it is then enough to prove that the family of the Vx.8 s is decreasing filtering. Now let XI X2 E A be such that O:s:; xl>X2:S:; 1 and 7T(X I), 7T(X2) are both the identity on H; let 81,82 > O. Put

x = ~(XI + xJ E
We have O:s:; x:S:; 1, and f(x) ~ 1- 8/2, we have
f(xi
7T(X)

A,

is the identity on H. If fEE is such that

f(xI) :s:; 1, :s:; 1;

+ xJ ~

2 - 8,

f(x~

9.3.3.

LEMMA.

Let H be a Hilbert space, T E .:(H),

eE Hand e > 0 be

CR. 9, 3)

SOME LEMMAS

199

such that
E

IITII:S;;I,
Then II T~ - ~112 :s;; e.

11~11:S; 1,

Re(T~I~);a.1-2

For
IIT~

- ~112

=(n In) - 2 Re(T~

I~)

+ (~I~):s;;

1- 2 + e + 1 = e.

9.3.4.

LEMMA. Let e > 0, and n be a positive integer. There exists a number 8(E, n) > 0 possessing the following property: if H is a Hilbert space, if T I, ... , T; E 2(H) and ~ E H are such that O:s;; T I, ... , T; :s;; 1, 11~11:s; 1 and

Re(TnT n-I .. . TI~ then (Ti~

Ia

1- 8(E, n),

I t);a. 1- e and II Tit - ~II:s;

e for i = 1, ... , n.

Put e2) 8(e, 1) = inf ( 1"2 and 8(e, n) We have 8(e,n):s;; 8(e, n-1) , 2

si n - 1)2
16

The lemma is true for n = 1 by 9.3.3. Let n ;a. 2, and suppose the lemma true for n - 1 (with the chosen values of the 8s). If
Re(TnT n- 1 Tit

Io

1- 8(E, n),

We have
1- 8(e, n):S;; Re(T n . .. TI~ It) :s;; (TnT n- 1 Tit I Tn-I .. TI~)IJ2(Tn~

I~)1/2:s;

(Tn~

I~)1/2,

whence

(Tn~

I~)

1- 28(e, n);a. 1- e, and, in view of 9.3.3, IITn~ - ~112:s; 48(e, n):S;; e 2.

TIt I~") TIt I~)

Then
Re(T n_ l
TI~

I~) = Re(T n
;a. Re(T n

+ Re(T n_ 1

TI~

I(1- Tn)t)
n -1) - ~8(e, n - 1).

-lit-

Tn~11

;a. 1- 8(e, n) -

28(e, n)l/2;a. 1- ~8(e,

200

C*-ALGEBRAS OF TYPE

[CH. 9, 3

By the induction hypothesis, (Ti~ 1, ... , n - 1.

I~);;;.

1-

and IITi~

- ~II

for i =

9.3.5. For each E ]0, 1], we will denote by f. the function equal to 0 on ]-00,1- e], to 1 on [1- /2, +00[ and linear on [1 - , 1 - /2]. For e E ]0, D, we have fJ2. = I eLEMMA. Let A be an antiliminal C*-algebra, d an element of A + such that IIdll = 1, and T E ]0, 1]. Then there exist w, w, d E A such that: (i) Ilwll = Ilwll= Ildll= 1, w;;;. 0, d;;;. 0, w*w = 0; (ii) fT(d)w = w, fT(d)w = w; (iii) w 2d = d, w*wd= d,

Put

0"

= 718. Choose u E A, c E A such that Ilull = I, 0 ~ c ~ 1. Put

do = f2u(d)cf2u(d)

and
d,
~

d,

= h,.(d) -

do.

We have 0 ~ do ~ I, and so -1

1. Moreover,

h,.(d)do = do =doh,.(d),

hence do and hT(d) generate a commutative sub-C*-algebra B of A which contains d.. If P is any character of B such that p(do) =0, we have p(h,(d = I, hence p(d!) = 1 - p(do); also, 0 ~ p(do) ~ 1. Hence, if g: R vanishes on [0, !], we have
g(p(dog(p(d!
--+R

is a continuous function that

=0,

i.e. p(g(do)g(d!

=o.

It follows that g(do)g(d!) = O. Put v = lu(d!)ufu(d o). We have

[e] = 1 and v*v = fu(d o)u*fu(d ,)2Ufu(d o).

Hence
f2u(d o)v*v

= v*v,

and so Furthermore,
fsu(d)d o = do

and

fsu(d)d,

= d.;

hence h,(d)p(do) = p(d o) if p is a polynomial with no constant term, hence also if p is a continuous function vanishing at the origin, and similarly

CH. 9, 3l

SOME LEMMAS

201

fgq(d)p(d.) = p(d l); hence fgq(d)v = v and fsu(d)v* = v*. Finally, put w = vk(v*v),

where k: R~R is the function equal to (f1/2(t)t-)/2 if t- 0, and to 0 if t =O. We have O:s;; d: 1, O:S;; w:s;; 1. Since v*(v*v) =0, we have w*w = O. Since f8u(d)v = v and f8u(d)v* = v*, we have
fsu(d)w

= w,

fgq(d)w = w.

Further,
w*w = k 2(v*v)v*v

= fl/2(V*V),

hence

and similarly w*w = d; moreover

IlwlI= Ilw*wIII/2:s;;1.
We are going to prove that, provided c and u are suitably chosen, we have IIdll;;.:1. This, together with (iii), will imply that

Ilwll,llwll;;.: 1,
whence

Ilwll = Ilwll= Ildll= 1

and the proof will be complete. We have fu(d) - O. Since A is an tiliminal, there exists an irreducible representation 7T of A such that 7T(fu(d is not compact (4.2.6). We can therefore find two orthogonal unit vectors in the set of values of 1T(fu(d. There exists an hermitian element c of A such that 1T(C)~ = g, 1T(C)17 = 0 (2.8.3) and, reasoning as in 9.3.1, we can suppose that O:S;; c:S;; 1. There exists a unitary element u of A such that 1T(U)g = 17 (2.8.3). Since

and

g, 17 E

1T(fu(d)(H,,), we have

202

C*-ALGEBRAS OF TYPE

[CH. 9, 3

and similarly 1T(j4q(dT/ = T/. We deduce from this, successively, that

1T(do)~

= ~,

1T(do)T/

= 0,

1T(d ,)T/ = T/,

1T(V*V)~

= fu( 1T(do1T(U )-Ifu(1T(dl21T(U )fu( 1T(do~

= fu( 1T(d o1T(u)-fu(1T(dl21T(U)~

= fu( 1T(do1T(U rfu(1T(d l2T/

= fu(1T(do1T(u)-IT/ =fu(1T(do~
1T(d')~

=~,

=~,

111T(d)11 ~ 1,

Ildll ~ 1.

9.3.6. LEMMA. Let A be a unital C*-algebra, and VI" .. ,vp , c elements of norm 1 of A such that c ~ 0, V~VjC = c, V~Vj =0 for i# j. Let s be an hermitian element of A, and E > O. Then there exist an hermitian element t of A, a linear combination of the vtVj, an element b of A with O:s;; b :s;; 1, and a Y > 0 such that: (i) If f is a state of A such that feb) ~ 1- Y, we have If(s - t)l:s;; e ; (ii) lie llf=l v~ bvdl = 1. There exists a pure state of the commutative sub-C*-algebra of A generated by 1 and c which takes the value 1 at c. This pure state can be extended to a pure state m of A (2.10.l). Put
aij = m(v~sv)

and

=L ajjvjv) E A.
ij

We have ajj =aji, and so t is hermitian. Moreover,

and so 1Tm (c )~m

= ~m,

hence

1Tm(V~Vi)~m
1Tm(Vi)~m'

= 1Tm(V~ViC)~m

=1Tm(C)~m

= ~m'
Trm

Let H be the finite-dimensional subspace of H We have

(1Tm(t)1Tm(Vk)~m l1Tm(VI)~m)

generated by the

= L ajj(1Tm(v7vk)~m
ij

l1Tm(V~V')~m)

=a'k(~m

I~m) = ak

hence

CH. 9, 3]

SOME LEMMAS

203

Let E be the set of states of A and W the set of fEE such that If(s - 01 ~ E. Let K be the weak*-closed convex hull of the set of states of A of the form x -+ (7T m (x)1/ 11/), where 1/ runs over the unit sphere of H. By the above, W is a neighbourhood of K in E for the weak* topology. By 9.3.2, there exists abE A such that ~ b ~ 1 and 7Tm(b ) reduces to the identity on H, and a Y > 0, such that W contains the set of fEE for which f(b) ~ 1- Y. Moreover, it is clear that

Lastly,
7Tm(C)

n
p

7Tm(vj)7Tm(b)7Tm(V;)~m

= 7Tm(C)

;=(

n
p
j~(

7Tm(vjV;)~m

7Tm(C)~m

~m'

and hence

9.3.7. LEMMA. Let A be a unital antiliminal C*-algebra. Let (so, Sl>") be a sequence of hermitian elements of A. There exist non-zero elements v(a(, ,an) in the unit ball of A (where al>"" an E {O, I}, n = 0,1,2, ) with the following properties: (i) if j ~ k, and (al"" aj) - (b l , ... ,bj ), then v(a 1, , aj)* , bk) = 0; v(b 1, (ii) if k ~ 1, then v(al> .. , ak) = v(a l, , ak-l)v(Ok-l, ak); (iii) if j < k, then v(al> ... , a)*v(a l, , a)v(Ok_1> ak) = V(Ok-l, ak); (iv) v(0) = 1, V(Ok)~O; (v) putting e(j) = La\, ... ,ajE{O.I} v(al"" a)v(a(, , a)*, there exists, for every j ~ 0, a linear combination tj of the v(a l, , aj)v(bl> ... , bj )* such that

Ile(j + 1)(Sj -

t)e(j + 1)11 ~ j + 1.

We shall construct, not only v(al> ... , an)s with these properties, but also elements b(n) E A + such that Ilb(n)1I = 1 and (vi) v(a l, ... , an)*v(a(, .. . ,an)b(n) = b(n). For n = 0, put v(0) = b(O) = 1. Now suppose that non-zero elements v(a l, ... , a) in the unit ball of A and elements b(j) in A + of norm 1 have

204

C*-ALGEBRAS OF TYPE

[CH. 9, 3

been constructed for j:s:; n, in such a way that properties (i) to (vi) are satisfied (property (v), only for j ~ n - 1, of course). By 9.3.6, there exist an hermitian element tn of A, a linear combination of the v(a l , , an)v(b], ... , b n)*, an element b of A with 0 ~ b :s:; 1, and a y > 0 such that: (1) if f is a state of A such that f(b);;. 1 - y, we have If(sn - tn)l:s:; tln + 1; (2) IIail = 1, where

(the order in which the v(a 1, , an) are arranged is immaterial). Now apply 9.3.5 with d = aa* and T for the moment arbitrary within 1]. Denote by V(On+l), v(Om 1), b(n + 1) the elements w, w, d constructed in 9.3.5. By 9.3.5 (ii), .

]0:

v(al ... , an)*v(a 1, ... , an)v(On, an+l)

= v(a[, ...
Now (vi) implies that whence hence

, an)*v(a], ... , an)fT(aa*)V(On, a n+]).

v(a], ... , an)*v(al> ... , an)v(On, a n+])= fT(aa*)V(On, an+l) = v(On, a n+]),

again using 9.3.5 (ii). Moreover, v(On, 1)*v(On+l) = 0 by 9.3.5 (i), and V(On+l);; O. By 9.2.13, if we put
v(a., ... , a n+[)= v(a!, . . . , an)v(Om a n+ l )

for (al> ... , an);t Om

conditions (i), (ii), (iii), (iv) of 9.3.7 are still satisfied. Moreover, ben + 1) E A + and IIb(n + oil = 1 (9.3.5 (il). We have
v(On, an+l)*v(Om an+])b(n

+ 1) = b(n + 1)

by 9.3.5 (iii); and, for (ai, ... , an) ;t On, we have, in view of condition (iii) of 9.3.7 (already established)
v(a[, ... , an+l)*v(al>.. , an+l)b(n

+0 + 1)

= v(On, an+l)*v(a], ... , an)*v(al> ... , an)v(On, an+l)b(n

= v(Om

an+1)*v(Om an+l)b(n

+ 1) = b(n + 1),

CR. 9, 3j

SOME LEMMAS

205

hence condition (vi) of 9.3.7 is still satisfied, which implies, in addition, that the v(al>" ,a n+l) are non-zero. It remains to prove (v) for j = n. Suppose that A is realised as a C*-algebra of operators in a Hilbert space H. Let g E H, and TJ = v(al" .. , a n+1)g be a vector of norm 1. We have, by (ii) and (iii),
v(a., ... , an)*TJ

= v(al .

, an)*v(al ... , an)v(0n> an+l)g = v(On> a n+1)g

and, since t ~ [,(t) -!T for every t ~ 0,

(aa*v(a l, ... , an)*TJ

Iv(al> . . , an)*TJ) =(aa*v(On an+l)g Iv(0n> a n+l)g)

~ (f,(aa*)v(On, an+l)g I v(On> an+.)g) -

2;

in view of 9.3.5 (ii), this is equal to

(v(On> an+l)g Iv(On> a n+1)g)-2
T

~ Ilv(al" .. ,an)v(On> an+.)gll2-

T II 1/2 T T 2 = TJ - 2 = 1- 2

Apply lemma 9.3.4, and suppose

T ..:: ~('Y 2 ~ 2
u

chosen so that

2n +3

2 +. 2)
n +1

We see that
(v(al> .. , an)*bv(al" ., an)v(a], ... , an)*TJ

Iv(al> ..

,an)*TJ)~

1- 22n +3

Since
v(a., , an)v(al" , an)*TJ

= v(al ...

, an)v(0n> a n+1)g

= TJ,

this may be written

Let TJ be a vector of norm 1 in the set of values of v(b], ... , b n+ 1) ; we have similarly

206

C*-ALGEBRAS

OF TYPE

[CH. 9, 4

if (a l, ... , an+l) ;e (b l , hence

,b n +I ) , T/ and T/ are orthogonal by (i), and

2

l(bT/ I T/W = 1(1- b)T/ I T/)1

:!i;;

1- b)T/ I T/)(1- b)T/ I T/):!i;;C2~+3r

~
u(al ... , a n+!),

This established, let a E e(n + I)H be a vector of norm 1; by definition of e(n + 1) we have
a =

a.....a.+lE{O. I}

where u(a), . . , a n+)) belongs to the set of values of v(a), ... , a n+)) whence
iba Iu) ~

~
at._a n +l

(baia-; ... , an+!) Iu(a), ... , an+l))

~
(a ... a.+l)"(b, .... b.+ 1)

I(bu(a l ,

a n +l ) Iu(b l, ... , bn+I))1

1- - - -

2 2

1- y.

It follows from this that Isn - tn)u 1u)l:!i;;1/(n + 1). Hence

1 - --:!i;;e(n

n+l

+ 1)(sn

- tn)e(n

+ 1) :!i;;-n+l
1 :!i;;- - .

and
Ile(n

+ 1)(sn -

tn)e(n

+ 1)11

n+l

Reference: [631]. 9.4. The proof of the theorem, concluded

In order to prove (ii)::? (iii) in theorem 9.1 we are going to show that if A is a separable non-postliminal C*-algebra, then A admits factor

CH. 9, 4]

THE PROOF OF THE THEOREM, CONCLUDED

207

representations which are not of type I. Since A admits a non-zero antiliminal quotient (4.3.6), this will be a consequence of the sharper result that we now give:
PROPOSITION. Let A be a non-zero

Then A possesses

separable antiliminal C*-aLgebra. a factor representation of type II.

By 4.3.9, we can suppose that A is unital. Let (so, s. S2, ) be a dense sequence in the hermitian part of A. Let v(a .. . . . , an) be elements with the properties of 9.3.7. Let M(n) be the set of linear combinations of the v(a t , , an)v(b 1, , b n)*, N(n) the set of linear combinations of elements of M(O) U M(l) U ... U M(n), and
N=UN(n).
n~O

By 9.2.12, there exists a linear form f on N, such that f(x*) = f(x),

f(x);;;. 0 for x;;; 0, and f(v(a
1, ,

an)v(b

1,

b n)*) = 2-n8~:

... 8~:,

whence f(l) = 1.

By 2.10.1 (i), there exists a state g of A which extends f. We are going to show that 1Tg is a factor representation of type II, i.e. that the von Neumann algebra glJ generated by 1Tg (A ) is a factor of type II. The 1Tg (v (a ., ... , an possess properties (i), (ii), (iii) of 9.2.5. Put F(n) = 1Tg(e(n(H). By 9.2.7, the F(n) form a decreasing family. It is clear that F(n) is invariant under every operator of 1Tg (A ). Hence the intersection F of the F(n) is invariant under every operator of 1T g (A ), from which it follows that P F E 1T g(A)" = glJ. Furthermore, F is invariant under every operator of 1T/N) by 9.2.11. We have, for every n,
II1Tg(e(n~gllll~gll;;;.
=

(1Tg(e(n~g

I~g) = g(e(n = f(e(n

2" . Tn = 1 ;;;.1I1Tg(e(n~glllI~gll,

hence 1Tg(e(n~g = ~g, and hence ~g E F(n). Hence ~g E F. Now ~g is a cyclic vector for 1Tg(A) and, a fortiori, for glJ. Hence PF E glJ has central support 1, from which it follows that glJ is isomorphic to glJp (A 20). It suffices to prove that glJp is a factor of type II, and hence that glJF is a factor of type II (A 52). By 9.2.9 and 9.2.11, the 1Tg(v(a 1 , , an)v(b t , , b,,)*) induce in F a system of matrix units of order 2 n. Since 1Tg(e(n~g = ~g, this system is

208

C*-ALGEBRAS OF TYPE

[CH. 9, 4

non-trivial. Hence there exists an isomorphism IJIn of the involutive algebra 7Tg(M(n IF = 7Tg(N(n IF onto M 2n(C). We easily see that 97J p is infinite-dimensional as a vector space. We shall prove: (1) that (aa'~g I~g) = (a a~g I~g) for any a, a E 97Jp ; (2) that 97J p is a factor. This factor will be finite (A 33), and cannot be of type I q , for q < +00 (A 49); it will therefore be of type IIIand the proof will be at an end. Let s be an hermitian element of A. For every positive integer n, there exists a j;a. 2n such that lis - sAl ~ 1/(2n), and then a tj E M(j) such that
lieU + I)(sj - tj)e(j

+ 1)11 ~-.

2n

Then

which shows that 7Tg(N )P p is norm-dense in Pp7Tg(A)Ppo If x E 7Tg(N(n, we have

(x~g

I~g)

Tr (lJln(x IF;

it is enough to check this for x of the form 7Tg ( v (a l" " an) v(b h . , b n )*), and we then have
(x~g

I~g) = g(v(ah

... , an)v(b 1, ... , b )* ) =

Tn.s:: ... 8::.

We deduce from this that, for x, x E 7TiN(n,

(1)

because Tr (1Jn(x IF)1Jn(x IF

=Tr (1Jn(x IF)lJln(x IF.

Equality (I), being established for arbitrary n, holds for x, x E 7Tg (N ), and therefore for x, x E Pp7Tg(A)P p since 7Tg(N )P p is norm-dense in Pp7Tg(A)Pp, and therefore for x, x E PlJ p since Pp7Tg(A)P p is strongly dense in P p97JP po Finally, let r be a projection belonging to the centre of PlJpo Let .,., = r(~g), (= (1- r)(~g). Since a ~(agg I~g) is a trace on PlJp, we see immediately that a ~ (a.,., /.,.,) and a ~ (a(, () are traces t I and t 2 on PlJp. Since 7Tg(N(n IF is isomorphic to M 2n(C), the restrictions of t; t 2 to

CH. 9, 5j
7r g(N 7r g(N

ADDENDA

209

(n p (the set of operators induced on F by the operators of (n ) are proportional. Since t 1 , t 2 are weakly continuous, t 1" and t 2 are proportional. We have 717=0 or ~7'= O. If, for example, 1/7=0we have (rT/ I1/) 7= 0, but then (r, I~) = 0, hence t 2 = 0, and hence

0= t 2(1 - r) = (1- r)~

I~)
gJp,

=II~W

and

gg = rg

Since gg is a cyclic vector for This proves that gJp is a factor. Reference: [631].
9.5. Addenda

we conclude from this that r

= 1.

9.5.1. Let A be a separable C*-algebra. If, for every primitive ideal I of A, AI I is postliminal, then A is postliminal. (Apply 9.1). 9.5.2. Let A be a separable C*-algebra. Then A is postliminal if and only if A is a To-space. (Apply 9.1). 9.5.3. Let A be a separable C*-algebra. Then A is liminal if and only if every point of A is closed. (Apply 9.5.2 and 4.7.15). *9.5.4. Let A be a separable C*-algebra. The following conditions are equivalent: (i) A is antiliminal; (ii) A possesses an injective represen tation of type II; (iii) A possesses an injective representation of type III; (iv) there exists a family (7r;) of irreducible representations of A such that EB7ri is injective and 7ri(A) n .Te(H = 0; (v) there exist families (7r;)iEI and (P;)iEI of irreducible representations of A such that EB 7ri and EB Pi are injective, Ker 7rj = Ker Pi (for every i), and Pi is non-equivalent to 7ri (for every i). [631].
1T)

*9.5.5. (a) Let G be the group of order 2, X the compact group G x G x G x ... , X the dense subgroup of X consisting of those elements (gl> g2 ... ) of G x G x G x ... all but a finite number of whose components are equal to e. Let Y be the quotient Borel space XIX. Then there does not exist any sequence of Borel subsets of Y separating the points of Y. (b) Let A be a separable non-postliminal C*-algebra. Then there exists a Borel injection of Y into A (A being endowed with the Mackey structure). [631]. 9.5.6. Let A be a separable C*-algebra. The following conditions are

210

C*ALGEBRAS OF TYPE

[CH. 9, 5

equivalent: (i) A is post/iminal; (ii) for the Borel structure given by the topology, A is standard; (iii) for the Mackey structure, there exists a sequence of Borel subsets of A separating the points of A; (iv) the Mackey structure of A is the same as the Borel structure given by the topology of A. (We have (i) ~ (ii) and (iv) by 4.6.1, (ii) ~ (iii) by 3.8.3, (iv) ~ (iii) by 3.3.4 and 3.8.4, (iii) ~ (i) by 9.5.5.) [444], [631]. *9.5.7. Let H, H be two infinite-dimensional real vector spaces, and B a bilinear form on H x H with respect to which Hand H constitute a dual pair. Let K be a Hilbert space. For every x E H (resp. x E H) let p(x) (resp. q(x be a (not necessarily bounded) self-adjoint operator on K. Suppose that the p(x) commute pairwise, that p(.AIXI + A2X2) is the least closed extension of AIP(X 1) + A2P(X2), and that the analogous properties hold for q. Suppose that

If M, M are subspaces of H, H, let .stJ(M, M ) be the von Neumann algebra generated by the eip(x) and the /q(x) (x E M, x EM). Let A be the C*-algebra generated by the .stJ(M, M) as M (resp. M ) runs through the set of finite-dimensional subspaces of H (resp. H). Then A only depends, up to isomorphism, on (H, H, B). The C*-algebra A is an tiliminal, has no closed two-sided ideals other than A and 0, and does not admit a traceable representation. If we take for H a separable real Hilbert space, take H = H, and B(x, x) = (x Ix), A admits factor re presentations of type II"" and factor representations of type III. Some irreducible representations of A can be constructed explicitly. [611], [631], [1466]. The above relates to the problem of the representation of the physi cists commutation relations. The representation of the anticommutation relations ([610) has been translated by Mackey ([1008])into a problem of group representations.

9.5.8. Let A be a separable C*-algebra. If there exists a Borel subset of Irr(A) which meets each equivalence class in exactly one point, then A is postliminal. (Use 7.2.3 and 9.5.6). [1007]. 9.5.9. Let A be a C*-algebra. The conditions (i), (ii), (iii), (iv) of 9.1 are all equivalent, but we do not know if they are equivalent to (v). [1411], [1412], [1415].

CHAPTER 10

CONTINUOUS FIELDS OF C*-ALGEBRAS

Let A be a C*-algebra, T a topological space, and B the set of continuous mappings f: T ~ A such that SUPtET IIf(t)11 < +00. If f, g E B and A E C define f + g, AI, fg, f* E Band Ilfll by
(f + g)(t) = f(t) (fg)(t)

+ g(t),

(Af)(t) = A/(t),

=f(t)g(t),

f*(t)

=f(t)*

for every t E T, and

1IIII = SUPtETllf(t)II
It is easy to see that B thus becomes a C*-algebra. If T is discrete, B is simply the product C*-algebra IItETA t where At = A for every t. We have already introduced the general idea of the product IItETA t of C*-algebras, where T is a set and At is a C*-algebra which varies with t. We intend to define a way of constructing C* algebras which includes the above procedures: T will be a (not neces sarily discrete) topological space, and the At will vary with t. The main interest of this construction is that very broad classes of C*-algebras can be obtained by such a method, using only very simple Ats. It is by this method that we will reach a solution, albeit a very partial one, of the second problem mentioned at the beginning of Chapter 3.
10.1. Continuous fields of Banach spaces

10.1.1. Let T be a topological space, and (E(ttET a family of complex Banach spaces. Every element of IItETE(t), i.e. every function x defined on T such that x(t) E E(t) for each t E T, is called a vector field. More generally, if Y k T, an element of IItEYE(t) is called a vector field over

Y.
10.1.2. DEFINITION. Let T be a topological space. A continuous field {; of Banach spaces over T is a family (E(ttET of Banach spaces, with a

212

CONTINUOUS FIELDS OF C*-ALGEBRAS

[CH. 10, 1

r k IIET E(t) of vector fields such that: (i) r is a (complex linear subspace of II,ETE(t); (ii) For every t E T, the set of x(t) for x E r is dense in E(t); (iii) For every x E r the function t ~ IIx(t)11 is continuous; (iv) Let x E II,ETE(t) be a vector field; if, for every t E T and every E > 0, there exists an x E r such that IIx(t) - x(t)II:s;;E throughout some neighbourhood of t, then x E r.
set The elements of r are called the continuous vector fields of ~. Definition 10.1.2 is alogous with the definition of measurable fields. 10.1.3. Let
l)

= E(tEn n,

l)= E(tET, T)

be two continuous fields of Banach spaces over T. An isomorphism of ~ onto ~' is a family cP = (CP)ET such that: (1) each CPt is an (isometric) isomorphism of E(t) onto E(t); (2) <p transforms r into T.
10.1.4. Example. Let E be a Banach space, T a topological space, and r the set of continuous mappings of T into E. For every t E T, put E(t) = E. Then ~ =E(t/En is a continuous field of Banach spaces over T, called the constant field over T defined by E. (The reader would do well always to have this example in mind in what follows.) A field isomorphic to a constant field is said to be trivial.

10.1.5. Example. Suppose, in definition 10.1.2, that T is discrete. Axioms (ii) and (iv) then prove that r must be equal to IIET E(t).

10.1.6. Let T be a topological space, and ~ = E(t/En a continuous field of Banach spaces over T. Let Y k T and to E Y. A vector field x over Y is said to be continuous at to if, for every E > 0, there exists an x E r such that Ilx(t)- x(t)II:s;;e in the neighbourhood of to. It is said to be continuous on Y if it is continuous at every point of Y. For Y = T, this agrees with the definition of continuous vector fields already adop ted, by axiom (iv). 10.1.7. Let r y be the set of continuous vector fields over Y. It is immediate that E(tEY r y ) is a continuous field of Banach spaces over Y, which is called the field induced by ~ on Y, and which is denoted by ~ I Y. 10.1.8. If every point of T possesses a neighbourhood V such that ~ is trivial, then l) is said to be locally trivial.

IV

CH. 10, 11

CONTINUOUS FIELDS OF BANACH SPACES

213

10.1.9. PROPOSITION. Let E(tteTl n be a continuous field of Banach spaces. (i) Let to E T, x be a vector field over T continuous at to, and f: T ~ C a function, continuous at to. Then the vector field
t ~ (fx)(t) is continuous at to. (ii) If Y E and if g: T

= f(t)x(t)

C is continuous on T, we have gy E

r.

Let E > O. There exists x E T such that Ilx(t) - x(t)11 :;,:;;; E in the neigh bourhood of to. We have

If(t)I:;,:;;;If(to)1 + 1,
in the neighbourhood of to. Hence
Ilf(t)x(t) - f(to)x(t)lI:;,:;;; If(t)Illx(t) - x(t)11 + If(t) :;,:;;; E(lf(to)1 +Ilx(to)11 + 2)

f(to)I/Ix(t)11

in the neighbourhood of to. This proves (i), and (ii) follows from (i). 10.1.10. PROPOSITION. For every to E T and every an x E T such that x(to) = f
~

E E(to), there exists

In view of axiom (ii) of 10.1.2, there exists a sequence (~n) of non-zero vectors of E(to) possessing the following properties: (a) ~ I ~nll < +00; (b) ~ ~n = ~; (c) for every n, there exists an x, E r such that xn(to) = ~n' Let i, be the function on T equal to 1 when Ilxn(t)II:;,:;;;I ~nll and to II~nll/l xn(t)11 when Ilxn(t)11 > I ~nll. This function is continuous, and so Yn = t,, E r (10.1.9 (ii.We have Yn(tO) =~"' IIYn(t)/I:;,:;;;lI~nll for every t E T, hence for every t E T, ~ Yn(t) converges to an element x(t) of E(t). We have x(to) =~ ~n =~. For every E > 0, there exists an integer p such that Ix(t) - ~~ Yn(t)l:;,:;;; E throughout T, whence x E r by axioms (i) and (iv) of 10.1.2. 10.1.11. LEMMA. Let T be a paracompact space, Ya closed subset of T, E(t, 1) a continuous field of Banach spaces over T, x (resp. y) a continuous vector field over T (resp. Y), and a a positive number such that Ilx(t) - y(t)II:;,:;;;a on Y. Then there exists an x E T such that Ilx(t) y(t)11 :;,:;;; al2 on Y, and [x(r)- x(t)1I :;,:; ; 2a on T. There exists a cover (Vi) of Y by open subsets of T and, for every i, an Xi E T such that Ilxi(t)- y(t)!I:;,:;;;a/2 on Vi Replacing (V;) by a finer

214

CONTINUOUS AELDS OF C*ALGEBRAS

[CH. 10, I

cover if necessary, we can suppose that the cover (Vi) is locally finite. Let (T/J be a family of non-negative continuous functions on T, su bordinate to (V;), such that L T/i =: 1 on Y. Let w =: L T/jXj, a function which is continuous at each point of T, and therefore belongs to r. Let t E Y. In the sum L T/iCt)XjCt), counting only those indices i such that T/i(t) > 0, we have
II

xiCt ) - y(t)11

~~

hence

II ~

T/iCt)Xi(t) - yCt)1I

~~,

for those indices. Consequently, !Iw(t) - xCt)1I ~ 2a on a neighbourhood Z of Y. Let f: T ~ [0, 1] be a continuous function equal to 1 on Y and to 0 on T\Z. Put x = fw + (1- f)x. On Y, we have
[x(r)- y(t)11
=:

JlwCt) - YCt)ll ~-.

a
2

On T, we have

[x(r)- x(t)11 = If(t)!llw(t) - xCt)11 ~ 2a.

10.1.12. PROPOSITION. Let T be a paracompact space, Y a closed subset of T, ((E(ttET> r) a continuous field of Banach spaces, and y a continuous vector field on Y. Then there exists an x E r which extends y. Reasoning as for lemma 1O.1.11,we at first construct an xoEr such that Ilxo(t) - y(t)11 ~ 1 on Y. Then, applying lemma 10.1.11, we obtain inductively a sequence (x n ) of elements of T such that and The x, converge uniformly to an x E

r. We

have x(t) = yCt) on Y.

10.1.13. PROPOSITION. Let T bea completely regular space, and (Yi)iElan open cover of T; put Y ij = Y j n Y j. For every i E I, let ~i = ((Ei(t, r;) be a continuous field of Banach spaces over Y i. For every i, j E I, let gij be an isomorphism of ~j I Y ij onto ~i I Y ij Suppose that, for all i, j, k E I, we have gijgjk = gil< Then there exists a continuous field ~ of Banach spaces over T, unique up to isomorphism, possessing the following property: for every i E I, there exists an isomorphism hiof ~i onto ~ I Y i, such that gij = hit hdor all i, j E 1. The uniqueness is easy: we prove the existence. For every t E T, let ICt) - 0 be the set of i E I such that t E Y i. If i, j E I(t), then gijCt) is an

CH. 10, 2j

TOTAL SUBSETS

215

isomorphism of Ej(t) onto Ei(t) and

gjj(t)gjk(t) = gik(t)

if i, j, k E I(t).

There therefore exists a Banach space E(t), and, for every i E I(t), an isomorphism h;(t) of Ei(t) onto E(t) such that
gjj(t) = hj(tr1hj(t)

for

i, j E l(t)

(take, for example, E(t) = Eio(t), where ioE I(t) and h;(t) = gioi(t. There exists exactly one set .1j of vector fields defined over Y j and taking values in the E(t) such that (hi(ttEY is an isomorphism of (;i onto fFi =E(ttEY i , .1J. Also, we have Sj I Y jj = Sj I Y ij for any i, j E 1. This established, let T be the set of x E TI t E T E(t) such that x I Y j E .1j for every i E 1. It is immediate that T satisfies axioms (i), (iii) and (iv) of 10.1.2. We show that r satisfies axiom (ii). Let to E T and ~ E E(to). Let i E I be such that to E Y j Let V!: Y i be a closed neighbourhood of to in T. Let 1 : T ~ R be a continuous function equal to 1 at to, and to 0 on T\ V. Let y E.1 j be such that y(to) =~. Let x be the element of TItETE(t) equal to 0 on T\ V and to 1y on V. We have x(to) = ~. We will show that x E r. Let j E I and Jet us show that x I Y j E .1j It is enough to show that, for every t. E }j, x coincides, in a neighbourhood of t l , with an element of L1 j. This is immediate if t 1 ~ V. Suppose that tiE V. Then t, E Y ij. Since Sj I Y ij = ~ I Y jj and x coincides, in a neighbourhood of t; with an element of .1j, our assertion is proved. is a continuous field of We have thus proved that (;= E(t), Banach spaces ovc. T. \1Ta; easily see that (; I Y j = Sj and the proposition follows. We say that (;is obtained by gluing together the (;jS using the gj/s.
j

References: [463], [584], [638], [896], [1101], [1680], [1708].

10.2. Total subsets

10.2.1. DEFINITION. Let (;= E(t)tET> r) be a continuous field of Banach spaces over T. Let A !: r. Then A is said to be total if, for every t E T, the set of x(t), as x runs through A, is total in E(t). (; is said to be separable if r has a countable total subset.
If A is a countable total subset of T, the set A of linear combinations of elements of A with complex rational coefficients is again countable, and A (t) is dense in E(t) for every t.
I

216

CONTINUOUS FIELDS OF C*-ALGEBRAS

[CH. 10, 2

10.2.2. PROPOSITION. Let E(t, be a continuous field of Banach spaces over T. Let A C be a total subset, and .If the.linear subspace of generated by A. The following are equivalent properties for a vector field x E ll/ETE(t): (i) x E (ii) For every to E T and every E > 0, there exists x E such that IIx(t) - x(t)II ~ E in a neighbourhood of to; (ii) For every to E T and every E > 0, there exists x E.If such that IIx(t) - x(t)II ~ E in a neighbourhood of to; (iii) For every x Er, the function t -+ Ilx(t) - x(t)11 is continuous; (iii) For every x E.If, the function t -+ Ilx(t) - x(t)II is continuous.

(ii)=? (ii) =? (i) =? (iii) =? (iii): Obvious. (iii) =? (ii): Suppose condition (iii) is satisfied. Let to E T and E > O. Since .If(to) is dense in E(to), there exists x E.If such that Ilx(to) x(to)II < E. By (iii), we have IIx(t) - x(t)II < E on a neighbourhood of to, whence (ii). 10.2.3. PROPOSITION. Let T be a topological space, (E(t/ET a family of Banach spaces, and A a subset of ll/ETE(t) satisfying axioms (i), (ii), (iii) of 10.1.2 (with T replaced by A). Then there exists a unique subset r of ll/ETE(t) containing A and satisfying axioms (i) to (iv) of 10.1.2, namely the set of vector fields x satisfying condition (ii) of 10.2.2. The uniqueness follows from 10.2.2: if r exists, its elements are characterised by property (ii). Conversely, the set of vector fields x satisfying this property (ii) clearly satisfies axioms (i) to (iv) of 10.1.2, and contains A. 10.2.4. PROPOSITION. Let ee; = E(t, 0, ee; = E(t, F) be two con tinuous fields of Banach spaces over T, and let A be a total subset of r. For every t E T, let (()/ be an (isometric) isomorphism of E(t) onto E(t). Then (() = (()/)/ET is an isomorphism of ee; onto ee; if and only if (()(A) C F. The necessity is clear. We therefore suppose that (()(A) C I" and prove that tp is an isomorphism. We can suppose that A is a linear subspace of If x E T, x is a limit of elements of A for local uniform convergence (10.2.2 (iiI), hence (()(x) is a limit of elements of F for local uniform convergence, and hence (()(x)EF. Conversely, let xEF, x = (()-I(X), to E T and E > O. There exists ayE r such that y(to) =x(to). We have

r.

(()(y) E F and

(()(y)(t o) = (()to(x(to = x(to),

CH. 10, 2l

TOTAL SUBSETS

217

whence 1Iq>(y )(t) - x(t)11 ~ E in the neighbourhood of to, whence lIy(t) E in the neighbourhood of to. Hence x E T, from which itfollows that q> -I(T) C r.
x(t)1I ~

10.2.5. PROPOSITION. Let T be a locally compact space, E(t, a continuous field of Banach spaces over T, x E T, K a compact subset of T, E > 0, and A a total subset of r. Then there exist continuous complex valued functions t.. ... , t, on T with compact support and elements Xt, ... , x, of A such that
IIx(t) - iJ(t)Xt(t) - ... - fn(t)x n(t)1I ~
E

on K.

Let A be the linear subspace of r generated by A. By 10.2.2 (ii), there exists a cover (V j ) of K by relatively compact open subsets of T and x; E A such that Ilx(t) - x;(t)11 ~ E in Vi Since K is compact, we can suppose that (Vi) is a finite cover (VI" ., V p). Let (1/1>.. , 1/p+t) be a continuous partition of unity in T subordinate to the open cover (VI"" v; T\K) of T. We have
IIx(t) - 11t(t)X;(t) - ... -11p(t)x~(t)1I
~ E

throughout K. Expressing the x;s as linear combinations of elements of A, we obtain the proposition. a continuous 10.2.6. Let T be a locally compact space, [;=E(t, field of Banach spaces over T, and T the compact space obtained by -uljcining a point at infinity w to T. Put E(t) = E(t) for t E T, E(w) =O. Let T be the set of x E IlEr E(t) such that x IT E rand Ilx(t)11 vanishes at infinity on T.
PROPOSITION. [;= E(t))tET, T) is a continuous field of Banach spaces over T, and [;I T =[;.

It is immediate that F satisfies axioms (i), (iii), (iv) of 10.1.2. Let ~ E E(to). There exists x E r such that x(to) = f Multiplying x by a function equal to 1 at to and with compact support, we can suppose that Ilx(t)11 vanishes at infinity on T, which leads at once to the existence of an x E T such that x(to) = f Let r o be the set of restrictions to T of the elements of T, i.e. the set of elements of r which vanish at infinity; by the above, r o is a total subset of r. For every t E T, let q>, be the identity mapping of E(t); let q> = (q>I),ET For every x E ro, q>(x) is a vector field, continuous relative to [; IT. Hence q> is an isomorphism of [;onto [; I T (10.2.4).
to E T and

218

CONTINUOUS FIELDS OF C*-ALGEBRAS

[CH. 10, 3

10.2.7. PROPOSITION. Let T be a separable metrisable space, and l:= E(t)), a locally trivial continuous field of Banach spaces over T. If each E(t) is separable, then l: is separable.

There exists a countable open cover (Un) of T such that each s ] U; is trivial, and then an open cover (V n ) of T such that Vn C U; for every n. For every n, there exists a sequence (x nm) of vector fields, defined on U; and bounded and continuous there, such that, for every t E U; the xnm(t) are dense in E(t) (in fact, l:I U; may be identified with the constant field defined by a separable space). Let l1n: T ~ [0, 1] be a continuous func tion equal to 1 on V~ and to 0 on T\ U; Let Ynm be the vector field on T equal to T/nXnm on U; and to 0 on T\ U; We have Ynm E F: in fact, if t E Ui; then Ynm is continuous at t; if t E T\ Ui; and if E > 0, there exists a neighbourhood of t in which IIYnm(t)11 ~ E. Let t E T and g E E(t). There exists an n such that t E V n Then g is a limit of the vectors xnm(t) = Ynm(t). Hence the Ynm constitute a total subset of r. References: [463], [584], [638], [896], [1101], [1680], [1708].

10.3. Continuous fields of C*-algebras

10.3.1. DEFINITION. Let T be a topological space. A continuous field of C*-algebras over T is a continuous field A(t)), e) on Banach spaces over T, each A(t) being endowed with a multiplication and an involution with which it is a C*-algebra, and 8 being closed under multiplication and involution. When we speak of isomorphisms of continuous fields of C*-algebras, we will be dealing with isomorphisms compatible with the involutive algebra structure on the fibres. Similarly for triviality and local triviality. A continuous field of elementary C*-algebras will mean a continuous field A(t)), 8 ) of C*-algebras such that each A(t) is an elementary C*-algebra (4.1.1). 10.3.2. PROPOSITION. Let .sIl = A(t)), 8) be a continuous field of Banach spaces over T, each A(t) being a C*-algebra. Then .sIl is a continuous field of C*-algebras if and only if there exists a total subset A of 61 which is closed under multiplication and involution. In fact, let X be the linear subspace of 8 generated by A; then X is closed under multiplication and involution. Moreover, every element of

CH. 10, 4]

THE C"-ALGEBRA DEFINED BY A CONTINUOUS FIELD

219

8 is a limit of elements of .if under local uniform convergence. Hence, if x, y E 8, the vector field xy is a limit, under local uniform convergence, of elements of .if, hence xy E 8. We see similarly that x* E 8.
10.3.3. PROPOSITION. Let A(t)), 8) be a continuous field of C*-alge bras over T, x an element of 8 such that xx* =x* x and f: C ~ C a continuous function such that f(O) = O. Then t ~ f(x(t)) is an eLement of

8.
Let to E T. There exist a neighbourhood V of to and a constant M > 0 such that Ilx(t)1I ~ M on V. Let (Pn) be a sequence of polynomials in z and z without constant terms which converges uniformly to f on the disc [z] ~ M. Then Pn(xx*) E 8, and Pn(x(t), x*(t)) converges uniformly on V to f(x(t)). Whence the proposition. The field t ~ f(x(t)) is denoted by f(x).
10.3.4. COROLLARY. If x E 8 is such that x(t);;:. 0 for every t, the field t ~ X(t)1/2 is an element y of 8, such that y = y*, y2 = x. 10.3.5. lf each A(t) is unital, with identity element 1/, and if t ~ It is an element of 8, we can suppress the hypothesis in 10.3.3 that f(O) =O. 10.3.6. PROPOSITION. Let A(t)),8) be a continuous field of C* algebras over T, x an element of 8 such that xx* =x*x, toE T, and U a neighbourhood of Spx(to) in C. Then there exists a neighbourhood Vof to in T such that Spx(t) C U for every t E V.

Let f: C~C be a continuous function equal to 0 on Spx(to) and to 1 on C\ U. Then f(x)(to) = 0, hence 11!(x)(t)1I < 1 in a neighbourhood V of to. lf t E V, then Spx(t) cannot intersect C\ U. References: [463], [584], [896J, [1101], [1708].

10.4. The C*-algebra defined by a continuous field of C*-algebras

10.4.1. Let T be a locally compact space, d = A(t)), e) a continuous field of C*-algebras over T, A the set of x E e such that Ilx(t)11 vanishes at infinity on T. Then A is an involutive subalgebra of 8. For x E A, put

IIxll

= sup IIx(t)1I < +00.

tET

It is immediate that, for this norm, A is a C*-algebra which we will call

220

CONTINUOUS FIELDS OF C*-ALGEBRAS

[CR. 10. 4

the C*-algebra defined by d. For every t E T, x A onto A(t).

x(t) is a morphism of

10.4.2. LEMMA. Let I be a closed (left or right) ideal of A. For every t E T, let I(t) be the set of x(t), as x runs through I. Then I is the set of x E A such that x(t) E I(t) for every t E T.

We deal with the case of a right ideal. If x E I, we have x(t) E I(t) for every t. Conversely, let x E A be such that x(t) E I(t) for every t. Let IE > O. For every T E T, there exists a YT E I such that x( T) = YT( T), and a zTEA such that IIYT(T)-YT(T)zT(T)II<IE (1.7.2), whence IIx(t) )T(t)ZT(t)11 :s;; E in a neighbourhood of T. Let K be the compact set of t E T such that IIx(t)II;3 E. There exists a cover (U I , U 2 , , Un) of K by open subsets of T, and, for every i, a Yi E I and a z, E A, such that Ilx(t) - Yi(t)Zi(t)II:s;; IE in U; We put U o = T\K, Yo = Zo = O. Let (f0,[1, ... ,fn) be a continuous partition of unity subordinate to the cover (Uo, U I , , Un) of T. Since fizi E A (10.1.9 (ii)), we have ~ fiYizi E 1. Let t E T. Let ii, i2 , ,ip be the indices i such that t E U; We have

and
Ilx(t) - Yi/t)Zil(t)II, ... , IIx(t) - Yip (t)Zi p(t)11 :s;; IE;

since fil(t) +... + fi p(t) = 1 we see that

II x(t) -

L fi(t)Y;(t)Zi(t)1l :s;;

IE.

Hence [x - ~ fiYizill :s;; E. Since I is closed, x E 1.

10.4.3. THEOREM. Let T be a locally compact space, d = A(t)), B) a continuous field of C*-algebras over T, and A the C*-algebra defined by d. For every t E T and every 7T E A(t)", let p; be the irreducible re presentation x ~ 7T(X(t)) of A. Then 7T ~ P", is a bijection of the union of the A(t)~ onto A.

Let 7T E A(t)~, 7T E A(t)". If ti: t, there exists a continuous complex valued function f on T such that f(t) = 1, f(t) = 0; let x E A be such that 7T(X(t)) 0; then
P"(fx) = 7T(X(t)) 0, P",,(fx) = 7T(O) = 0,

and so P", and P", are non-equivalent. It follows from this that the

CH. 10, 4)

THE C*-ALGEBRA DEFINED BY A CONTINUOUS FIELD

221

mapping 7T -) P", is injective. Now let pEA. Let I be the kernel of p. Let I(t) be the set of x(t) as x runs through 1. Let Y be the set of t E T such that I(t) # A(t). We have Y# 0 by 10.4.2. Suppose that Y contains two distinct points t\ and t 2 Let U\, U 2 be disjoint neighbourhoods of t.; t 2 Let K, be the closed two-sided ideal of A consisting of those x E A which vanish outside U; We have I ~ 0 = K, . K 2 , and therefore 1;2 K\, say (2.11.4). Hence I(tl) ~ K\(t\) = A(t.) which is a contradiction since tIE Y. Hence Y consists of a single point to E T. By 10.4.2, I is the set of x E A such that x(to) E I(to). Hence p is the composition of the morphism x -)x (to) and a (necessarily irreducible) representation of A(to). 10.4.4. COROLLARY. Let T be a locally compact space, d = ((A(t, @) a continuous field of non-zero elementary C*-algebras over T, and A the C*-algebra defined by d. For every t E T, let PI be the irreducible representation of A which is the composition of the morphism x -)x(t) and of the irreducible representation, unique up to equivalence, of A(t). Then t -)PI is a homeomorphism of T onto A. This mapping is bijective (10.4.3). Now, let YeT and toE T. The intersection of the kernels of the PI(t E Y) is the set I y of the x E A such that x(t) = 0 on Y. Now Pto is in the closure of the set of the Pt(t E Y) if and only if I{lo};2 I y If to E Y, it is clear that I{lo};2 I y If to E Y, there exists a continuous complex-valued function on T equal to 1 at toand to 0 on Y. and therefore an x E A, non-zero at to and vanishing on Y, from which it follows that I{to} ~ I y . The corollary follows. We identify T and A by means of the above homeomorphism. 10.4.5. COROLLARY. With the notation of corollary 10.4.4, A is a liminal C*-algebra with spectrum T. In fact, for every t E T = because A(t) is elementary.

A, pt(A) consists of compact operators

10.4.6. PROPOSITION. Let T be a compact space, d = ((A(t, @) a continuous field of C*-algebras over T, and A = @ the C*-algebra defined by d. Suppose that each A(t) is unital with identity It and that the field t -) 11 (denoted by 1) is an element of A. Let B be a sub-C* algebra of A. Suppose that, for every pair (tl t 2) of distinct elements of T, there exists an x E B such that x(t\) = 111 x(tz} =0 (and, if T reduces to a single point, that 1 E B). Then, for every continuous complex-valued function f on T, the field t -)f(t)1 1 is an element of B.

222

CONTINUOUS FIELDS OF C*-ALGEBRAS

[CH. 10, 4

Let t l , t 2 E T with t\ oF- t 2 There exists x E B with x(t.) = 111 x(t0 = O. Replacing x by xx*, we can suppose that x,:: B+.. Let p :R~R be a continuous function equal to 0 in a neighbourhood of 0 and to 1 in [1, +00[. Put Y = p(x). Then y(tl) = 1 and y(t) vanishes in a neigh bourhood of t 2 Let K be a non-empty compact subset of T not containing t.. There exist open subsets V ... . . , U; of T covering K and elements Yl ... , Yn of B such that Yi(t\) = Ill yj(t) = 0 in U; Forming the product YI .. YnY~ . . . Yr, we see that there exists a Z E B+. such that z(t\) = 111 z(t) = 0 on K. This conclusion plainly still holds if K = 0. In some neighbourhood of t .. we have Ilz(t)-IIII~~, hence z(t)~~ I,. Now let L be a compact subset of T disjoint from K. There exist open of B+ such subsets VI"" V p of T covering L and elements z ... . . that Zj(t) = 0 on K, Zi(t) ~ ~. lIon Vi Then u = 2(zl +... + zp) E B+, u(t) = 0 on K, u(t) ~ I, on L. Let f be a continuous real-valued function on T, and let E > O. Let (WI> .. , W q ) be an open cover of T such that the oscillation of f in each lVi is ~E; let Ai be a value that f takes in lVi. Let (W;, . . . , W~) be an open cover of T such that Wik lVi for each i. There exists Vj E B+ such that Vj(t) ~ lIon W; and Vj(t) = 0 on T\ lVi. Let v = V\ +... + Vqo We have v(t);;. 11 for every t, and hence, by the functional calculus (10.3.3), there exists a v E B+ such that (V2V )(t ) = 1 for every t. Let Wj = VUiV E B+, so that L Wj = vvv = 1. Let t E T. We have, say,

.s,

Then
q

f(t) . 11 -

~ AjWj(t)
j=\

=f(t) ~
i=1

Wi(t) -

~ AjWj(t)
i=l

=~
j=\

[f(t) - A;]Wi(t),

whence
m m m

- E 11 =

-E~
j=l

Wj(t)~~
j=\

[f(t)-AdWj(t)~E~
j=1

W;(t) =

E 11;

hence

II f (t ) . 1,- ~ Aj W;(t )1I

~E

for every t.

Since E > 0 was arbitrary, the field t ~ f(t) . I, is an element of therefore of B.

B,

and

CH. 10. Sj

THE CONTINUOUS FIELD OF C*-ALGEBRAS

223

10.4.7. PROPOSITION. Let T be a locally compact space with a coun table base, d a separable continuous field of C*-algebras over T, and A the C*-algebra defined by d. Then A is separable. Thanks to 10.2.6, we need only consider the case where T is compact and has a countable base. Let (fl f2 ...) be a dense sequence in the Banach space of continuous complex-valued functions on T. Put d = A(t, 8), and let (x., X 2, ) be a total sequence of elements of 8. The f;Xj are elements of A. We are going to prove that the set of linear combinations of them is dense in A, which will establish the proposition. Let X E A and E > O. By 10.2.5, there exist continuous complex-valued functions gl" .. ,gp on T such that [x - glXl - ... - gpxpll ~ E. Choosing the indices i l , , ip suitably, we will have /Ix - fi1x I - fjpx pII ~ 2E. References: [584], [896], [1101], [1696], [1708].

10.5. The continuous field of C*.algebras defined by certain C*-algebras 10.5.1. Let A be a C*-algebra whose spectrum T is assumed Hausdorff. This spectrum is then a locally compact space (3.3.8). Two irreducdible representations with the same kernel are equivalent (3.1.6). Hence the elements of T may be identified with the primitive ideals of A. For every t E T, let A(t) = A/t, which is a non-zero C*-algebra. Every x E A defines a vector field t ....x(t) E A(t) over T, x(t) being the canonical image of x in A(t). Let A be the set of all the t ....x(t), for X E A. The set A is an involutive subalgebra in the set of all vector fields. For every x E A, the function t ....IIx(t)II is continuous (3.3.9). For every t E T, the set of the x(t), where x runs through A, is the whole of A(t). Consequently (10.2.3), there exists a unique subset r of TI1ETA(t) containing A such that d = A(t/ET, r) is a continuous field of Banach spaces; r is the set of vector fields which are limits of elements of A for local uniform convergence. By 10.3.2, d is a continuous field of C*-algebras. We will say that d is the continuous field of C*-algebras defined by A. If A is liminal, d is a continuous field of elementary C*-algebras. If A is separable, d is separable. 10.5.2. THEOREM. Let T be a locally compact space, d = A(t, 8) a continuous field of non-zero elementary C*-algebras over T, and A the liminal C*-algebra defined by d, whose spectrum is T. Let d = A(t, 8) be the continuous field of non-zero elementary C*-algebras

224

CONTINUOUS FIELDS OF C*-ALGEBRAS

[CH. 10, 5

defined by A. For every t E T, let cp(t) be the canonical isomorphism of A(t) onto A(t). Then (CPCttET is an isomorphism of.s4 onto si:

The elements of A constitute a total family of continuous vector fields relative to .sIl, and their images under (CPCttET are elements of e by definition of sd, It then suffices to apply 10.2.4. 10.5.3. The following lemma will be considerably generalised later on (11.5.3).
LEMMA. Let T be a locally compact space, .sIl = ((ACt, e) a con tinuous field of elementary C*-algebras over T, A the C*-algebra defined by .sIl, and B a sub-C*-algebra of A such that, for all t l E T, tzE T, tl E ACtI), tz E A(tz} (with tl = tz if t l = t z), there exists x E B with x(t l ) = tl, xCtz) = tz Then B = A.

Adding a point at infinity to T, it is enough to deal only with the case of compact T (10.2.6). Then A = e, and B is a total subset of e. In view of 10.2.5, it is enough to prove that, for every x E B and every con tinuous complex-valued function cP on T, we have a: E B. Let E > O. We are going to construct an element of B whose distance from tpx does not exceed 2E[sup Icplz] and this will complete the proof. Put y = x*x. Let to E T. There exist numbers a, {3 such that 0 < a < (3 < E and such that the spectrum of YCto) does not intersect [a, {3] (because A(to) is elementary). Let a, {3 satisfy a < a < {3< (3. There exists a neigh bourhood of to, such that, for t in this neighbourhood, Spy(t) does not meet [a, {3] (10.3.6). Let p :R~R be a non-negative continuous func tion equal to 0 in ]-00, a] and to 1 in [(3, +00[. Let e = p(y) E B+. Then, for every t E T, e(t) is an element of A(tt with finite spectrum; and, for every t in a closed neighbourhood U of to, e(t) is idempotent and Ile(t)y(t) - y(t)II,,;;; E. Let ((A(ttEU..1) be the continuous field of C*-algebras induced by .sIl on U. For every z E..1, the following conditions are equivalent:
(1)

z(t) E ACt) = e(t)A(t)e(t)

for every t E U;
=

(2) there exists z E..1 such that z(t)

e(t)z(t)e(t) for every t E U.

Let ..1 be the set of the z E..1 satisfying these conditions. It is immediate that .sIl = ((A(ttEU, ..1) is a continuous field of C*-algebras over U; for every t E U, e(t) is the identity element of ACt). Let A be the C*-algebra defined by si, Let B be the set of elements of ..1 which

CH. 10, 5l

THE CONTINUOUS FIELD OF C*-ALGEBRAS

225

are restrictions to U of elements of B. The mapping x ~ x I U is a morphism of the C*-algebra A = e onto the C*-algebra ..1 (10.1.12); the image of B under this morphism is a sub-C*-algebra of ..1, and so B is a sub-C*-algebra of A. By the hypothesis made concerning B, for all t l, t 2 E U, gl E A(tl) gz E A(tz) (with gl =gz if t l =tz), there exists an XI E B such that XI(t l) = gl xz(tz) = gz; the restriction of exle to U is then an element of B which takes the values gl gz at t., t z. We can therefore apply 10.4.6: for every continuous complex-valued function if! on U, the field t ~ /I(t)e(t) on U is an element of B. Hence, for every continuous complex-valued function /I on T, the field /Ie coincides on U with an element of B. Let V be a closed subset of T contained in the interior 0 of U. We show that there exists an fEB such that f(t) =e(t) on V, f(t) =0 on T\ O. Let tiE T\ 0, t 2 E V. Since e(t I) has finite spectrum, the hypothesis made about B proves that there exists an hermitian element d of B such that d(tl) is an idempotent,
d(tz)

= o.

Replacing d by qed), where q is a function vanishing in a neighbourhood of 0 and equal to 1 at 1, we can suppose that d(t) vanishes in a neighbourhood of t z. Covering V with a finite number of such neigh bourhoods, and multiplying together the corresponding elements d, we obtain a d E B such that
det I)e(t I) = e(t I), det)

=0

on V. Then d" = e - de E B coincides with e on V and vanishes at t l U sing the functional calculus, we obtain a d" E B which coincides with e on V and vanishes in a neighbourhood of t l Covering T\O with a finite number of such neighbourhoods and multiplying together the corresponding elements d" we obtain f. Let (} be a continuous complex-valued function on T, vanishing outside V. We have seen that there exists an element of B which takes the value O(t)e(t) on U. Multiplying it by the above element [, we obtain an element of B which takes the value (}(t )e(t) throughout T. Summing up, and changing the notation a little, we know how to associate, with every t E T, an e, E B+ and a closed neighbourhood U, of t possessing the following properties:
(1)

IIe,(t)y(t)- y(t)11 ~ e

in U,;

226

CONTINUOUS FIELDS OF C*-ALGEBRAS

[CH. 10, 5

(2) If (J is a continuous complex-valued function on T whose support is contained in Ot, then (Jet E B. Let, then, (UI... , Ut ) be a finite cover of T. Let (VI , V r ) be an , (Jr) be a open cover of T such that VI ~ 01 ... , Vr ~ 0". Let (J continuous partition of unity subordinate to (VI>"" V r ) . We have P8jejE B.Let Then

(z - Px)*(z - cpx) = 1P12

Since

L (Jj(Jj(etjyetj - etjy j.j

yetj + y).

Iletj(t)y(t)- y(t)11 ",;; E we have hence

in Uj

liz -

PxIl2:s;; 2E[SUp 1P12l.

10.5.4. THEOREM. Let A be a liminal C*-algebra whose spectrum Tis Hausdorff. Let .stJ = A(t, 8) be the continuous field of non-zero elementary C*-algebras over T defined by A. Let A be the C*-algebra defined by d. For every x E A, let x be the element of 8 defined by x. Then x E A and x ~ x is an isomorphism of A onto A.
If x E A, the function t ~ IIx(t)11 vanishes at infinity on T (3.3.7), and hence x E A. It is clear that x ~ x is a morphism TT of A into A. By 2.7.3, this morphism is isometric. If t E T, the set of x(t), where x E A, is the whole of A(t) by definition of A(t). If t l , t 2 are two distinct points of T, if ~I E A(tt) and ~2 E A(t2), there exists x E A such that x(r.) = ~I' x(tJ = ~2 (4.2.5). Hence TT(A) = A (10.5.3).

10.5.5. Theorems 10.5.2 and 10.5.4 show that there exists a canonical bijective correspondence between the liminal C*-algebras with Haus dorff spectrum (taken to within isomorphism) and the continuous fields of non-zero elementary C*-algebras over locally compact spaces (taken to within isomorphism, the notion of isomorphism employed here meaning the possible existence of a homeomorphism of the underlying topological spaces). Let A be a liminal C*-algebra with Hausdorff spectrum, and .stJ the field defined by A. If A is separable, A has a countable base (3.3.4) and .stJ is separable. The converse also holds by 10.4.7.

CH. 10. 6] SOME REMARKS CONCERNING ELEMENTARY C*-ALGEBRAS

227

10.5.6. COROLLARY. Let A be a liminal C*-algebra such that A is Hausdorff, x an element of A, and f a bounded continuous complex valued function on A. Then there exists ayE A such that 1T(Y) = f( 1T )1T(X) for every 1T E A. With the notation of 10.5.4, we can replace A by A, and the corollary then follows from 10.1.9 (ii). 10.5.7. DEFINITION. Let T be a topological space, and .st1 = A(t, 6) a continuous field of elementary C*-algebras over T. .st1 is said to satisfy Fells condition if, for every to E T, there exists a neighbourhood V of to and a vector field p of .st1, defined and continuous in V, such that, for every t E V, p(t) is a projection of rank 1 (cf. 4.1.9). It is clear that a locally trivial field of non-zero elementary C* algebras satisfies Fells condition. However, the converse is false. 10.5.8. PROPOSITION. Let A be a liminal C*-algebra with Hausdorff spectrum, and .st1 the continuous field of non-zero elementary C*-alge bras defined by A. Then A has continuous trace if and only if.st1 satisfies Fells condition.
If A has continuous trace, .st1 satisfies Fells condition (4.5.3 (iii. Suppose that .st1 satisfies Fells condition. Let 1To EA. There exists a vector field p of .st1 defined and continuous in a compact neighbourhood V of 1To, such that, for every t E V, pet) is a projection of rank 1. Let f: T ~ R be a non-negative continuous function equal to 1 in a neigh bourhood of 1To and to 0 in a neighbourhood of T\ V. Then the vector field equal to 0 on T\ V and to fp on V is continuous and vanishes at infinity. It arises from an element y of A+ such that 1T(y) is a projection of rank 1 in the neighbourhood of 1To. Hence A has continuous trace (4.5.4).

References: [463], [584], [896], [1101], [1694], [1708].

10.6. Some remarks concerning elementary C*algebras

10.6.1. Let H be a Hilbert space. For ~,TJ E H, we will, in the rest of this chapter, denote by 8f ."I the operator ~~(~ I TJ)~ on H. It is of rank ";:;1, and every operator of rank ,,;:;1 is of this type. If f, TJ E H, we have
8f."I8~."I,a)

= 8f."I~

ITJ')~')

= (~I

TJ)(f ITJ)~

=(f ITJ)8~."I.(n,

228

CONTINUOUS FIELDS OF C*-ALGEBRAS

[CR. 10, 6

and so
(hry8 e,ry = (~' I 7))ef,ry

Moreover, if (E H, we have
(8t,.,(~)
1

= (~ = (~I

8e ,,., ())
( 17))~)

= (~I

~)(7)

In = ~ 1~)7) In

= (8,."e(~)

1n,

hence

8t,., =8,."e
We have 8u = 11~llzPce' and so every finite-rank operator on H is a linear combination of operators 8~,,., (and even of operators 8u ).
~Zn be variable vectors of H. If ~I converges to converges to 7)Zm then it is clear that 8~"Q +... + 8Q n_ I,Qn converges uniformly to 8"1.1J2 + .. , + 8TJ2n - IIJ2n This established, lemma 3.5.6 proves that, if (~; I~;) converges to (11; 111) for any i and i. then 118 e"Q + + 8Q n_ I,QJ converges to 118"11J2 +... + 81J2n-lolJ2J. In other words, 118 e"Q + + 8Q n- 1,QJ is a continuous function of the scalar products (~i I~i)'

10.6.2. Let
7)1... , ~Zn

~l"'"

10.6.3. If H is a Hilbert space with a unit vector " we will put =(A, p) where A is the elementary C*-algebra ,Tg(H) and where pEA is the projection on C~. Similarly, let (A, p) = a(H, f). Every isomorphism U: (H, ~)-+ (H, f) (i.e. every isomorphism of H onto H mapping ~ to f) defines an isomorphism a (U): (A, p) -+ (A, pi) (i.e. an isomorphism of A onto A mapping p to pi). If U 1, U z are two such isomorphisms, and if a(U\) = a( U2), we have U I = U z; in fact, U, and U z differ by a scalar factor by 4.1.8, and this factor has the value 1 since U1~ = f = Uzf
a(H,~)

10.6.4. We retain the notation H,

~,

a*a E pAp, and so a*a [eaves

hence

C~

A, p above. If a E Ap, we have invariant and vanishes on H C~,

Ila*all = (a*a~

I~) = Ila~lIz,

whence Iiall = lIa~lI.

The mapping a -+ at is therefore an isometric isomorphism of the normed vector space Ap onto the normed vector space H. It follows from this that Ap can be regarded as a Hilbert subspace of A, which is, moreover, uniquely determined, because, in a Hilbert space, scalar products can be calculated using only the norms of elements. If a, bE A p , we have b*a E pAp, hence Tr (b*a) = (b*a, I~) = (a~ I bt);

CH. 10, 61

SOME REMARKS CONCERNING ELEMENTARY

C*-ALGEBRAS

229

hence the scalar product in the Hilbert space Ap is given by (a Ib) = Tr(b*a). It should be noted, moreover, that ApA is the set of elements of A of rank ~ 1; this follows from the formula (J"TI = (Ju.(J",(J,TI (where II{II=1). 10.6.5. If A is an elementary C*-algebra with a projection p of rank 1, we will put ~(A, p) = (K, 7/) where K is the Hilbert space Ap (10,6.4), and where 7/ is the unit vector p of K. Similarly, let (K',7/')=~(A',p'). Every isomorphism V:(A,p)~ (A,p) defines an isomorphism ~(V):(K, 7/)~(K', 7/). If V .. V 2 are two such isomorphisms, and if ~(VI) = ~(V2)' we have VI = V 2; in fact, VI and V 2 coincide on Ap and therefore on pA = (Ap)*, hence on ApA = (Ap )(pA), hence on A by 10.6.4. 10.6.6. LEMMA. (i) Let H be a Hilbert space, and g a unit vector of H. Put a(H, g) = (A, p) from which it follows that ~(a(H, ~ = (Ap, p). For every a E Ap, put ~(a) = af Then ~ is an isomorphism of the Hilbert space Ap onto the Hilbert space H, such that ~(p) = ~. (ii) Let A be an elementary C*-algebra, and p a projection of A of rank 1. Put ~(A, p) = (H, ~), a(H,~) = (A, p). For every a E A, let l/!(a) be the linear mapping of H = Ap into itself defined by l/!(a)y = ay. Then l/! is an isomorphism of the C*-algebra A onto the C*-algebra A, such that l/!(p) = p, (i) has been proved in 10.6.4. We adopt the notation of (ii). There exists a Hilbert space H o and a unit vector ~o of H o such that (A, p) may be identified with a (Ho, ~o)' Let ~o be the isomorphism of (H, g) onto (Ho, ~o) defined in (i). For every a E A and every b E H, we have
~o(l/!(a)'

b)

= ~o(ab)

= (ab)(~o)

= a(~o(b,

hence l/!(a) = q;ola~o, and l/! is surely an isomorphism of A onto A which maps PCf.o = P to PC{ = p. Whence the lemma. The isomorphisms q; and l/! will be said to be canonical. 10.6.7. Use the notation of 10.6.6. (ii). If a, a E A, we have ap E H, a*p E H, hence we can form (Jap,a"p which is an element of A. We have l/!(apa) = (Jap,a"p. In fact, for every bE A, we have
l/!(apa)(bp) = apabp = a(Tr(pabpp

230

CONTINUOUS FIELDS OF C*-ALGEBRAS

[CH. 10, 7

and by 10.6.4, this is equal to

(bp

Ia*p)ap =

8ap a"p(bp).
~'~I

10.6.8. Let H, HI be Hilbert spaces, and

unit vectors of H, HI

= a(H, ~), (AI, PI) = a(H, ~I)'

(A, p)

(H, f) = (3(A, p),

(Hi, ~D

= (3(A I, PI)

Let
cp:(H',f)~(H,~)

and

cPI:(H;,~D~(HI>~I)

(HI, ~I)'

be the canonical isomorphisms. Let U be an isomorphism of (H, ~) onto Then U = (3(a(U is an isomorphism of (H, f) onto (HI ~D. We have Ucp = CPI U. In fact, let a E H = Ap. We have
(Ucp)(a)

= U(a~)

= UaU-I~1

= (a(U)a)~1

= cpl(a(U)(a = CPIU(a).

An analogous result holds if we start with (A, p), (AI PI) and an isomorphism of (A,p) onto (AI,PI) Reference: [463].

10.7. The continuous field of elementary C*-algebras defined by a con tinuous field of Hilbert spaces

10.7.1. Let T be a topological space, and if{ = H(t, n a continuous field of Hilbert spaces over T. If x, Y E r, the function t ~ (x(t) I yet~ is continuous because of the equality
4(x(t) I yet~

= Ilx(t) + y(t)W -llx(t) -

+ illx(t) +iy(t)IIz -

y(t)W illx(t) - iy(t)IIz.

10.7.2. For every t E T, let A(t) = iN6(H(t. For x, y E T, define 8x,y E llIErA(t) by the formula 8,jt) = 8x ( I) ,Y( I) By 10.6.1, we have

with z(t) fields

=(x(t) Iy(tx(t).

It follows from this that the set A of the

where XI Xz," ., XZn E is an involutive subalgebra of "IETA(t). By 10.6,1, A(t) is the set of operators of finite rank in H(t), and is therefore

CH. 10. 7j

THE CONTINUOUS FIELD OF ELEMENTARY

C*-ALGEBRAS

231

dense in A(t). By 10.6.2,

II(;Ixl.x2(t) +... + (;IX2n_I,x2.<t)!1

is a continuous function of the (Xj(t) Ix/t, and therefore of t. There then exists (10.2.3) a unique set e k TItETA(t), containing A, such that si =A(t, e) is a continuous field of Banach spaces. Since A is an involutive algebra, si is, in fact, a continuous field of elementary C*-algebras (10.3.2), said to be associated with It. It will be denoted by si(It). Thanks to 10.2.4, it is immediate that, if Y k T, we have
si(It I Y)

= si(It) I Y.

10.7.3. If a E e and x E F, the field ax: t ~ a(t)x(t) is an element of r. In fact, since every element of e is a limit, for local uniform con vergence, of elements of A (10.2.3), it is enough to verify it for a E A, and therefore for a = (;Iy,Z (y, z E Ty; however, (;Iy,zX is the field t ~ (x(t) Iz(ty(t). 10.7.4. Let (x",) be a family of elements of r such that, for every t E T, the set of the x",(t) is dense in H(t); then, for every t E T, the set of the (;Ix (t) is dense in the set of operators on H(t) of rank ~I; it follows from this that, if It is separable, then si(It) is separable.
a~~

10.7.5. In order to study the correspondence between continuous fields of Hilbert spaces and continuous fields of elementary C*-a1gebras, it will be convenient to make the following definitions: Let It be a continuous field of Hilbert spaces over T, with a con tinuous vector field x such that Ilx(t)11 = 1 for every t. We will put a (It, x) = (si, p) where si = si(It) and p = (;Ix,x (from which it follows that p is a continuous field relative to si of projections of rank I). Every isomorphism P : (It, x) ~ (It,x) defines an isomorphism a(P): a (It, x) ~ a (It, x} in an obvious way. It follows from 10.6.3 that a(PI) = a(P2) implies that PI = P2 Let si be a continuous field of elementary C*-algebras over T, with a continuous field p of projections of rank I. We will put {3(si, p) = (It, x), where It and x are defined in the following way. Put si = A(t), e). Let H(t) be the subspace A(t)p(t) of A(t); this is a Hilbert space (10.6.4). Let r be the set of x E 8 such that x(t) E H(t) for every t E T. It is clear that It = H(t, r) satisfies axioms (i), (iii), (iv) of 10.1.2. Furthermore, if a E e, then ap is an element of T, from which it follows that r(t) = A(t)p(t) = H(t). Hence It is a continuous field of Hilbert

232

CONTINUOUS FIELDS OF C*-ALGEBRAS

[CH. 10, 7

spaces over T. Moreover, put x = p, so that x is an element of r such that Ilx(t)11 = I for every t. If sd is separable, Je is separable: for if (aI a2 ...) is a sequence of elements of @ such that the set of the aj(t) is dense in A(t) for every t, then (alP, alP, ...) is a sequence of elements of r such that the set of the (ap )(t) is dense in H (t) for every t, Every isomorphism P :(sd, p) -+ (.sti,p) defines an isomorphism
(3(P):(3(sd, p) -+ (3(sd, p)

in an obvious way. It follows from 10.6.5 that (3(PI) = (3(P2) implies that PI= P2
LEMMA. (i) Let Jebe a continuous field of Hilbert spaces over T, with a continuous vector field x such that Ilx(t)11 = I for every t. Let (Je,x) =(3(a(Je, x. Put Je=H(t, Fv, Je=H(t, T). For every t E T, let Pt be the canonical isomorphism of H(t) onto H(t) (10.6.6). Then P = (Pt)tET is an isomorphism of Je onto Je such that P(x) = x. (ii) Let sd be a continuous field of elementary C*-algebras over T, with a continuous field p of projections of rank 1. Let (.sti,p)=a({3(sd,p. Put sd = A(t, e), sd = A(t, B). For every t E T, let IjJt be the canonical isomorphism of A(t) onto A(t) (10.6.6). Then IjJ =(IjJt)tET is an isomorphism of sd onto sd such that ljJ(p) = p.

10.7.6.

We adopt the notation of (i). It is clear that P(x) =x. Let y E rand y = Oy,x E I", We have
Pt(y(t

= Oy,x<t)x(t) = y(t).

Thus, P-I(y) = y E T, which proves that P-I is an isomorphism of Je onto Je (10.2.4). Now adopt the notation of (ii). It is clear that ljJ(p) = p. To show that IjJ is an isomorphism of sd onto sd; it is enough to show that ljJ(a) E @ for every a of the form apa" (a, a" E B): in fact, the fields of this form constitute a total subset of B (10.6.4). But if a =apa", we have ljJ(a) = Oap.a"*p E B by 10.6.7. 10.7.7. PROPOSITION. Let T be a topological space, and sd a continuous field of elementary C*-algebras over T. The following conditions are equivalent: (i) For every to E T, there exists a neighbourhood V of to and a continuous field Jeof non-zero Hilbert spaces over V, such that sd I V is isomorphic to .sti(Je); (ii) sd satisfies Fells condition.

CR. 10, 7]

THE CONTINUOUS FIELD OF ELEMENTARY

C*-ALGEBRAS

233

(i) ~ (ii). Suppose that condition (i) is satisfied. Let to E T. Let V o be a be a continuous field of non neighbourhood of to and Je= H(t)), zero Hilbert spaces over V o such that sil V o is isomorphic to d(Je). Let ~ be a non-zero element of H(to). Let x be a continuous vector field of le such that x(to) = f The set V of the t E V o such that x(t) ~ 0 is a neighbourhood of to; put y(t) = Ilx(t)11-1x(t) for t E V. The vector field ()y,y of d(le) is defined and continuous on V, and ()y,y(t) is a projection of rank 1 for every t E V. There therefore exists a vector field p of d, defined and continuous on V, such that the p(t) are projections of rank

1.

(ii) ~ (i). Suppose that si satisfies Fells condition. Let to E T. There exists a neighbourhood V of to and a continuous field p of projections of d of rank 1 defined on V. Put {3(d I V, p) = (le, x). Then, by 10.7.6, d I V is isomorphic to d(le). 10.7.8. Proposition 10.7.7 is a strictly local result. Even if d satisfies Fells condition, there does not, in general, exist any continuous field le of Hilbert spaces over T such that si is isomorphic to d(le). The existence of le comes up against a co homological obstacle that we will describe in detail. 10.7.9. LEMMA. Let le and Je be two continuous fields of non-zero Hilbert spaces over T, d and d the associated fields of C*-algebras, and f an isomorphism of d onto sl, For every t E T, there exists a neighbourhood V of t and an isomorphism of le I V onto le I V which defines f I

v.

There exists a neighbourhood V of t and a continuous field p of projections of sil V of rank 1 (10.7.7). Then p = f(p) is a continuous field of projections of sil V of rank 1. Let x be a continuous vector field of Je such that x(t) = p(t)x(t) ~ O. We have p(t)x(t)~ 0 in a neighbourhood of t. Replacing x by the field t ~ lip (t)x(t)1I-1p(t)x(t) we obtain, in a neighbourhood of t, a continuous vector field y of le such that lIy(t)1I = 1 and such that p is the field of projections defined by y (in a neighbourhood of t). Similarly, there exists a continuous vector field y of le in a neighbourhood of t such that lIy(t)11 = 1 and such that p is the field of projections defined by y (in a neighbourhood of t). Since the result to be proved is of a local nature, we can suppose, from now on, that y, y, P, p are defined on the whole of T. Then f3(f) is an isomorphism of {3(si,p) = {3(a(Je, y onto {3(si,p) = f3(a(Je, y, By

234

CONTINUOUS FIELDS OF C*ALGEBRAS

[CH. 10, 7

10.7.6 and 10.6.8, there exists an isomorphism g of (Je,y) onto such that f3(j) = f3(a(g. Hence f = a(g) (10.7.5). 10.7.10.
LEMMA.

oc, y)

Let Je= (t, be a continuous field of Hilbert spaces over T, and d = d(Je). An automorphism cP of Je defines the identical automorphism of d if and only if it is of the form (u(t) . I t ) t ET, where I, denotes the identity mapping of H(t), and where u is a continuous mapping of T into U (the set of complex numbers of modulus
1).

The condition is clearly sufficient. Conversely, if cP defines the iden tical automorphism of d, each CPt is a homothetic mapping of ratio u (t) E U (4.1.8), and it is therefore a matter of showing that u is a continuous function. For every to E T, there exists a continuous vector field x of Je such that x(t) ;tf 0 in a neighbourhood V of to. On V, we have
u(t)

=(cp,x(t) Ix(tlIx(t)1I- 2,

and so u is indeed continuous at to. 10.7.11. LEMMA. Let T be a paracompact space, and d a continuous field of elementary C*-algebras over T satisfying Fells condition. There exist: (i) An open cover (Tj)jEI of T; (ii) For every i E I, a continuous field :lej of Hilbert spaces over T, and an isomorphism h j of d(Je I t.. j ) onto d (iii) For all i, j E I, an isomorphism gjj of Je I which j I T ij onto defines the isomorphism hi1h j of d(Je j I Tj). j I T ij) onto d(Je

x; r,

(We put T, n T j = Tjj ; similarly, further on, T, n T, n T, = T jjh etc.) There exists an open cover (Ua)aEA of T and, for every a E A, a continuous field J{a of Hilbert spaces over U a and an isomorphism k a of d(J{a) onto d IUa (10.7.7). Since Tis paracompact, we can suppose that (Ua) is locally finite. There exists an open cover (Va)aEA of T such that Va ~ U a for every a. Let a E A and t E Va be fixed for the time being. There exists an open neighbourhood W ~ Va of t and a finite subset A of A such that W n V/l = cP for f3~ A. Let f3 E A. There exists an open neighbourhood W of t and an isomorphism of I W/l n onto J{a I W/l n Vil which defines k~ I kll on w, n V Il ; this is obvious if t V Il ; if t E V/l we have t E U/l n U; and our assertion follows from 10.7.9. Finally, since A is finite, we obtain an open neighbourhood W a .t ~ Va of

w, c

x,

v, e

CH. 10, 7j

THE CONTINUOUS FIELD OF ELEMENTARY

C"ALGEBRAS

235

t and, for every f3 E A, an isomorphism of J{fJ I W a t n VfJ onto x, I Wa t n VfJ which defines k~lkfJ on Wa,t n V(J- This established, the set I of the lemma will be the set of i = (a, t) where a E A and t EVa; we will put t, = Wa t, lei = x, I W a t, hi = ka I Wa t The existence of the gij follows at once from the above construction. In the last part of this chapter, we will make use of the notion of sheaf. On this subject, the reader could consult the book of R. Gode ment: Theory of sheaves (Act. Sc. Ind., no. 1252, Hermann, Paris, 1958).
10.7.12. LEMMA. We retain the notation of 10.7.11. (i) For all i, j, k E I there exists a unique continuous function Uijk : Tijk ~ U such that gi~jk = Uij~ik. (ii) The Uijk define a 2-cocycle of (T;) with values in the sheaf au of germs of continuous mappings of T into U. (iii) The image l of this cocycle in the cohomology group H 2(T, au) depends only on .sIl.
If i, j, k E I, then gi~j~il/ defines the isomorphism hilhjhjlhkhJ;lhi = 1 of .sIl(le I T onto itself; hence (i) follows from 10.7.10. i ijk) If i, j, k, I E I, we have, on T jjkl :

whence (ii). We prove (iii). Let (TA)AEA leA hA , gAIL have the properties analogous to those of 10.7.11, and let (UAILJ be the corresponding cocycle. To prove that (Uijk) and (UAILJ define the same element of H 2(T, au), we can replace (Tj) by a finer cover, the objects lei hi, gij being replaced by their restrictions to the sets of the new cover, and similarly for (TA ) . This allows us to suppose, in reasoning as for lemma 10.7.11, that, for every i E I and every A E A, there exists an isomorphism giA of leA I T iA onto lej I TiA which defines the isomorphism hi1h A of the associated fields of C*-algebras. Let the set A be the (disjoint) union of I and A. The above data define a cover (Ta)aEA lea ha(a E A) and gab(a, b E A) which satisfy the properties of 10.7.11, and therefore elements Uabc(a, b, c E A). The element of H 2(T, au) defined by (Uabc) is equal, on the one hand, to that defined by (uijd, on the other hand, to that defined by (UAILJ. The lemma now follows.

236

CONTINUOUS FIELDS OF C*-ALGEBRAS

[CH. 10, 7

10.7.13. Let 9lt be the sheaf of germs of continuous mappings of T into R. The mapping t ~ exp(2i1Tt) of R onto U defines a morphism 9lt ~ 011. Its kernel is the constant sheaf Z. Since every continuous function g: T ~ U can be expressed locally in the form exp(2i1Tf), where f is a continuous mapping of T into R, we have an exact sequence of sheaves
O~Z~9lt~OU~l,

whence an exact cohomology sequence

... ~ Hn(T, Z)
~

Hn(T, 9lt) ~ Hn(T, 011) ~ H n+1(T, Z)

...

Now the existence of continuous partitions of unity on T proves that 9lt is a sheaf, whence Hn(T, 9lt) = 0 for n > O. We deduce from this in particular that the canonical morphism H 2(T, OU) ~ H 3(T, Z) is an isomorphism. 10.7.14. DEFINITION. Let T be a paracompact space, and d a con tinuous field of elementary C*-algebras over T, satisfying Fells con dition. Let Y be the element of H 2(T, OU) defined in lemma 10.7.12. Its canonical image in H 3(T, Z) will be denoted by ~(d). 10.7.15. THEOREM. Let T be a paracompact space, and d a continuous field of elementary C*-algebras over T, satisfying Fells condition. Then there exists a continuous field ;Jt of Hilbert spaces over T such that d is isomorphic to d(;Jt), if and only if ~(d) =O. Suppose that d is isomorphic to d(;Jt). In the construction of ~(d), we can take for the cover of T the cover consisting of the single open set T. Every 2-cocycle relative to this cover is cohomologous to 0, hence ~(d) = O. Suppose ~(d) = O. There exist Ti, ;Jti hj, gjj, Ujjk (i, j, k E I) with the properties of 10.7.11, 10.7.12 and continuous functions Vij: Tij ~ U such that for all i, j, k E I, we have Ujjk =VjkVij/Vjj If we put gij =vijlgij the isomorphisms gij:;Jtj 1 Tij ~ ;Jtj I Tij satisfy
g i~ jk
" -I = Vij gi~jkVjk -I -I -I, = Uijk Vii< Uij~iI<

= g ,ik

We can then define a continuous field ;Jt of Hilbert spaces over T by gluing together the ~ using the gij (lO.1.l3). In other words, we can henceforth suppose that ;Jtj = ;JtI T, and that the gi j are the identical automorphisms. Then h j : d(;Jt) I ~ d I t; and hj: d(;Jt) I t, ~ d 111 coincide on 1/j. Hence the hi define an isomorphism of d(;Jt) onto d.

r,

References: [463], [584].

CH. 10, 8]

LOCALLY TRIVIAL FIELDS OF ELEMENTARY

C*-ALGEBRAS

237

10.8. Locally trivial fields of elementary C*algebras 10.8.1. LEMMA. Let H be an infinite-dimensional Hilbert space. There exist: (a) a unit vector ~ E H; (b) for every t E [0, 1], a closed subspace H, of H; (c) for every t E ]0, 1] a linear isometry U, of H, onto H, with the following properties: (i) The projection P, = PH, is a strongly continuous function of t E [0,1]; HI = H, n, ={O}. (ii) The operators U,P" U; are strongly continuous functions of t E ]0, 1]; U I = 1. (iii) [P~W = t. Suppose first of all that dim H =?\o, so that H may be identified with L 2([0, 1]). Let H, be the set of f E L 2([0, I]) such that f(x) =0 for x ~ t. If t E ]0,1] and f E H" we define UJ E H by (UJ)(x) = Vtf(tx). Lastly, take for the constant function equal to 1 for all t E [0, I]. Properties (i) and (iii) are immediate. We have, for f E H"

IIUJI1 2 =

f
o

tlf(tx)1 dx

f
o

If(x)1 dx =11f11

2 ,

and it is clear that U, is a linear isometry of H, onto H; if g E H, we have, for O:s;; x :s;; t,

Let

fE

H. The expression

tends to zero as t ~ t. Moreover,

238

CONTINUOUS FIELDS OF C"-ALGEBRAS

[CH. 10, 8

if, say, t

t, we therefore have
Ilu,-lf -

U~lfW

2//fW- 2Re

f JI!..
t
o

f(X)f(x

~)dX' t

and this tends to zero as t ~ t. Whence property (ii) follows. If, now, dim H is an arbitraryinfinite cardinal, we can identify H with a Hilbert space K K, where K and K are Hilbert spaces such that dim K = ?\o. There exist operators P t , Ut on K and a vector g E K with properties (i), (ii), (iii)and it is enough to take in H, the operators P, I, U, 1 and the vector g f, f being a unit vector of K. 10.8.2. LEMMA. Let H be an infinite-dimensional Hilbert space, G the unitary group of H endowed with the strong topology, and S the unit sphere of H endowed with the strong topology. Then G and S are contractible. Let H" P" t E [0, 1], put
U/, g have the properties of 10.8.1. For U E G and

c1J(U, t)

=(1- P,) + U;I UU,P t.

The operator c1J( U, t) induces the identity mapping in H, and induces in H, the unitary operator which is the restriction of U;I UU,. Hence c1J(U, t) E G and c1J(U, 1) = U. Put c1J(U, 0) = 1. We show that the mapping c1J of G x [0, 1] into G is continuous. On G x ]0, I], this follows immediately from lemma 10.8.1. Moreover, suppose that (U, t) E G x ]0, I] converges to (U, 0) E G x [0, I]; then P; converges strongly to zero and, for every f E H, II U,-;-I UUt,p,lll =IIPdl1 ~ 0, hence c1J( U, t) con verges strongly to 1 = c1J( U, 0). Thus, G is contractible. For f E Sand t E ]0, 1], put
1f(f, t)

He

= vt U,-If + (1- P,)g.

= t + 1 - t = l.

We have
//1/(f, t)/1 2 = tIIU;lf//2 + 11(1- P,)g1l2

Hence 1/(f, t) E S, and 1/(f, l) = f. Put 1/(f, 0) =g. The mapping 1/ of S x [0,1] into S is continuous on S x ]0, I]; moreover if (f, t) E S x and hence ]0,1] converges to (f,O)ESX[O,I], then IlvtU;If'II~O, 1/(f, t) converges to g = 1f(f,0). Thus, S is contractible.

CH. 10, 8j LOCALLY TRIVIAL FIELDS OF ELEMENTARY C*ALGEBRAS

239

10.8.3. A continuous field A(t)), B) of elementary C*-algebras is said to be of rank a if each A(t) is of rank a (4.1.9).
10.8.4. THEOREM. Let T be a paracompact space, and a an infinite cardinal. The mapping .s4 ~ 5(.s4) defines a bijective correspondence between H 3(T, Z) and the locally trivial continuous field of elementary C*-algebras of rank a over T (taken up to isomorphism).

We select a Hilbert space H of dimension a. Let G be its unitary group, C the sheaf of germs of continuous mappings of T into G, and au the sheaf of germs of continuous mappings of T into U. (I) Let .s4,.s4 be two locally trivial continuous fields of elementary C*-algebras over T, of rank a. Suppose that 5(.s4) = 5(.s4). We show that .s4 and .s4 are isomorphic. There exist: (a) a locally finite open cover (Tj)iEI of T; (b) for every i E I, continuous fields lei le: of Hilbert spaces over T, and isomor phisms hj: .s4('le;)~.s4ITj, h::.s4('le:)~.s4'ITi; (c) for all i,jEI, isomorphisms
gij : ~

I T;j ~

I r;

g:j : lej T jj ~ le: T jj

which define the isomorphisms hi1hj h:-1hj. If we define Uijb U:jk: Tijk ~ U by

then (Uijk) and (U:jk) both define 5(.s4). Replacing (Tj) by a finer cover, we can therefore suppose that there exists a cochain (Vi) of (T i ) with values in au such that U:jk = UijkVijVjkVii/. Replacing the g:j by the vijlgij, we are led to the case where Uijk = U:jk Since .s4 and .s4 are locally trivial, we can suppose, again replacing (Ti) by a finer cover, that lei = le:is the constant field over T, defined by H. Then gij, g:j are continuous mappings of T ij into G, endowed with the strong topology, and therefore define continuous sections Yij, Y:j of C(j over Tij. Let (Si)iEI be an open cover of T such that !{ ~ T, for every i E 1. Let L be the set of pairs (J, (3), where J ~ I, where f3 = (f3;)iEJ and where, for every i E J, f3i is a continuous section of C(j over Si, such that Y:j = f3iYijf3jl on Si n Sk for i, j E J. The set L is inductively ordered. We accordingly denote by (J, (3) a maximal element of L, and prove that J = 1. Suppose that there exists an i E 1\1. Let R =i n U jEJ Sj: Let t E R. If i, and j2 are two indices of J such that t E Sh and t E Sh we

240

CONTINUOUS FIELDS OF C-ALGEBRAS

[CH. 10, 8

have
Y1h(t ){3j/t)Y ih(t )-1 =Ylh(t )(3h(thith(t)- IYiit(t

r:1Uiith(t)

= Ylh(t hjlh(t)-t{3h(thh(t )-1Uihh(t) = Yiit(t)Uiith(t)-I(3it(t)Yiit(t)-I Uiith(t) = Y;it(t){3j.(thijt(t)-I.

It follows from this that we can define a continuous section { of f} over R such that Ylj = {Yij{3jl for any j E J. Since G is contractible (10.8.2), f} is weak, hence { extends to a continuous section {3i of f} over 5i, which contradicts the maximality of (J, (3). We have thus proved that J = 1. This established, each (3i defines an automorphism ai of .sd(J(i lSi) such that 1h j = a, h-1h-1 h ,i i jaj
I

in other words, such that

h Iiaih-I i

I hjajhj

Thus, the h ;aihil:.sd IS, ~.stt' IS, define an isomorphism of .stt onto .stt. (2) Let (Ti)iEI be a locally finite open cover of T, (Uijk) an alternating 2-cocycle of this cover with values in 611. We are going to construct a locally trivial continuous field of elementary C*-algebras over T of rank a, such that 8(.stt) is the element of H 3( T , Z) defined by (Uijk) This will complete the proof of the theorem. Let (Tf) be an open cover such that t; ~ T; Let L be the set of pairs (J, y), where J ~ I, Y = (Yij)ijEJ> and where, for all i, j E J, Yij is a continuous section of f) over t; n t;, such that YijYjk = YikUijk for i, j, k E J. The set L is inductively ordered. We accordingly denote by (J, y) a maximal element of L. We are going to show that J = 1. Suppose that there exists an i E I\l. Let M be the set of pairs (K, y), where K ~ J, Y = (Yij)jEK and where, for every j E K, Yij is a continuous section of f) over t; n tj, such that YijYjk = YikUijk for j, k E K. Let (K, y) now denote a maximal element of M. We show that K = J. If not, let j E J\K. We can define a section {3 of f} over t; n t; n U kEK t k such that {3Yjk = YikUijk for any k E K; in fact, if k, I E K, define 13k on t; n tj n t k, {3, on tin tin t; by from which we deduce that on t; n tj n t k n ti,
{3/Yjk Yk'U~:

= Yik YIdUiiJ Ujjl>

CH. 10, 8)

LOCALLY TRIVIAL FIELDS OF ELEMENTARY

C*-ALGEBRAS

241

whence

Since Cf} Cf} over fi n fi and this contradicts the maximality of (K, Y). Hence K =J. Consequently, there exists a family (Yij)jEJ> where, for every j E J, Yij is a continuous section of Cf) over fi n fi, with YijYjk = YjkUijk for j, k E J. The Yij and the Yij then contradict the maximality of (J, Y). Hence J = 1. Let lei be the constant field over Ti defined by H, and sIli = sIl(le i). For i, j E I, Yij defines an isomorphism fij of d j I Ti j onto d i I Ti j, and we have ftljk = fik Hence there exists a continuous field sIl of elementary C* algebras over T and, for every i E I, an isomorphism hi of sIli onto sill Ti, such that fij = hi1hj (10.1.13). To calculate 8(d), we can use the cover (Ti), the lei the hi and the restrictions of the Yij to the Ti j. We then see that 8(sIl) is the element of H 3(T, Z) defined by (U;jk) 10.8.5. We are now going to see that the hypothesis of local triviality made in theorem 10.8.4 is sometimes satisfied automatically.
LEMMA. Let T be a topological space, le = H(t, n a separable continuous field of Hilbert spaces over T, and H o a Hilbert space of dimension There exist, for every t E T, a closed subspace K(t) of H; and an isomorphism U(t) of the Hilbert space H(t) onto the Hilbert space K(t), with the following properties: (a) if.1 denotes the set of continuous mappings x: T ~ H o such that x(t) E K(t) for every t E T, then K(t,.1) is a continuous field J{ of Hilbert spaces over T; (b) (U(ttET is an isomorphism of le onto J{.

-I -I -I f3 f3 I = YikYjk UijlUikl Ujkl = YikYjk Uijk = k is weak, we can extend f3 to a continuous section of
~

"0

Let (XIX2, ..) be a total subset of r. Multiplying the Xi by suitable continuous functions, we can suppose that IIxi(t)11 ~ 1/i for every i and every t. Let (EI E2 ...) be an orthonormal basis of H o. For every t E T, there exists a continuous linear mapping S(t) of H o into H(t) such that S(t)E j = Xj(t) for every i. We have

IIS(t)11

~Tr

S(t)*S(t) ~

L r,
;=1

00

and S(t)(Ho) is dense in H(t). Let t~x(t)=~~=IA.j(t)Ej mapping of T into Hs: Then t ~ S(t)x(t) = ~~=I

be a continuous A.i(t)Xi(t) belongs to r

242

CONTINUOUS FIELDS OF C*-ALGEBRAS

[CH. 10, 8

because the Aj are continuous functions and

II n+l
every
E

Aj(t)Xi(t)11

~~
n+l

IAi(t)/lI xi(t) 1 1 Ilx(t)11

(~
n+l

i_2 )1 /2 .

We show that the mapping t ~ S(t)*S(t)x(t) of T into H o is continuous. It suffices to establish it in the case when x(t) = E; for every t. Now, for E H o, we have
2

IIS(t)*S(t)Ei - EII2 = IIS(t)*S(t)EdI

co

+ IIElf- 2Re (S(t)*S(t)Ei IE)

L I(S(t)* S(t)E;
j~1

IEj)1 2 + IIEI12 -

2Re (S(t)Ei I S(t)E)

L I(x;(t) IXj(t)W + liEW- 2Re (x;(t) IS(t)E);

j=\

this is a continuous function of t by the above, and it certainly follows from this that t ~ S(t)* S(t)Ei is a continuous mapping. Let, then, z E r. For every to E T and every E > 0, there exists a continuous mapping x: T ~ H o such that
IIS(to)x(to) - z(to)11

< E,

whence
cc

IIS(t)*S(t)x(t) - S(t)*z(t)11

< ( ~ C2

)1/2
E

in a neighbourhood of to. This proves that the mapping t ~ S(t)*z(t) of T into H o is continuous. Let K(t) be the closure of S(t)*(H(t)) in H o The above proves that the set .::1 of the lemma satisfies axiom (ii) of 10.1.2; axioms (i), (iii), (iv) are obviously satisfied, and we have therefore shown that the K(t) possess property (a). The mapping t~S(t)*S(t) of T into 2(H o) is strongly continuous, and hence the same is true of the mapping t ~ [S(t)* S(t)]1/2 = IS(t)I. Let S(t) = V(t)IS(t)1 be the polar decomposition of S(t). Since IS(t)I(Ho) and S(t)(H o) are dense in K(t) and H(t) respectively, we see that (V(t maps a total family in .::1 to a total family in r. Now V(t) IK(t) is an isomorphism of K(t) onto H(t), whence (b). 10.8.6. Recall, now, a result of E. Michael (Ann. of Math., v. 64, 1956, p. 562-580, tho 1.2.). Let T be a paracompact space of finite dimension (T is said to be of dimension ~n if, for every finite open cover of T, there exists a finer finite open cover of order ~n). Let A be a closed subset of

CH. 10, 8j

LOCALLY TRIVIAL FIELDS OF ELEMENTARY

C*-ALGEBRAS

243

T, Y a complete metric space, and cp a mapping of T into the set of non empty closed subsets of Y. Suppose that cp is lower semi-continuous, i.e. that if cp(to) meets an open subset Y of Y, then cp(t) also meets Y for all t sufficiently close to to. Suppose that, for every t E T, cp(t) is contractible. Lastly, suppose that there exists an Eo> 0 such that, for every t E T and for every open ball f3 of radius :0;;;;(0 in Y, cp(t) n f3 is contractible. Then, if f is a continuous mapping of A into Y such that f(t) E cp(t) for every tEA, f extends to a continuous mapping f of T into Y such that f(t) E cp(t) for every t E T.

10.8.7.

LEMMA.

Let T be a paracompact space of finite dimension, and

Ye = H(t), T) a separable continuous field of Hilbert spaces over T, each H(t) being of dimension ?\o. Then Ye is trivial.
Let H o be a Hilbert space of dimension ?\o. By 10.8.5, we can suppose that each H(t) is a closed subspace of H o and that T is the set of continuous mappings x of T into H o such that xU) E H(t) for every t. Let (YI> Y2, ...) be a sequence of elements of T such that, for every t E T, the set of the Yi(t) is dense in H(t). We are going to construct a sequence (XI> X2, ...) of elements of T such that, for every t E T, (Xj(t)) is an orthonormal basis of H(t). Then, if U(t) is the isomorphism of H(t) onto H o which transforms (Xi(t into a fixed orthonormal basis of Hi; (U(t)tET will be an isomorphism of Ye onto the constant field defined by u, (10.2.4). Suppose that we have already constructed elements Xl> Xn of T possessing the following properties: (1) for every t E T, (x\(t), X2(t), ... ,xn(t) is an orthonormal system; (2) there exist con tinuous complex-valued functions fil,fi2," .t (i = 1,2, ... , n) on T such that
1 IIYM) - fi,(t)X1(t) - ... - /;1(t)xj(t)II:o;;;;-;I

(i = 1,2, ... , n).

We are going to construct X n +l E T such that (XI> ... , x n +l ) possesses the same properties with n replaced by n + 1. This will enable us to construct the sequence (x., X 2, ) by induction. (The system (x\(t), X2(t), .. .) will certainly be total in H(t), since one can get within l/i of Yi(t) with a linear combination of elements of this system, and the set of the Yi(t) is dense in H(t). Let K(t) be the closed subspace of H(t) orthogonal to Xt(t), ... ,xn(t)

244

CONTINUOUS FIELDS OF C*-ALGEBRAS

[CH. 10, 8

and let cp(t) be the unit sphere of K(t). Since dim K(t) = l\o, cp(t) is contractible (10.8.2). Moreover, the intersection of P(t) with every open ball of radius <1 in H o is contractible (because it is homeomorphic, by stereographic projection, with an open ball of H o). We show that the mapping t ~ cp(t) of T into the set of non-empty closed subsets of H o is lower semi-continuous. Let U be an open subset of H o meeting cp(t o). Let ~ E cp(to) n U. There exists an x E r such that x(to) = f Let x be the element of r defined by
n

x(t) = x(t) - ~ (x(t) x;(tx;(t) E K(t).

i=1

We have x(to) =~,

and so x(t) ~ 0 for t sufficiently close to to. Then

t ~ x"(t) = x(t)/llx(t)II

is a continuous mapping of a neighbourhood V of to into H o, such that E cp(t) for t E V, and x"(to) = ~. We have x"(t) E U for t sufficiently close to to, hence cp(t) n U ~ 0 for t sufficiently close to to. We will therefore be able to apply 10.8.6. Define z E r by the equality
x"(t) Yn+I(t)

=L (Yn+1(t) Ix;(tXj(t) + z(t).

;=1

We have z(t) E K(t) for every t. Let T be the closed set of t E T such that Ilz(t)11 ~ (2(n + 1)-1. By 10.8.6, the mapping t ~ z(t)/IIz(t)II of T into H o extends to a continuous mapping Xn+l: T ~ H o such that Xn+l(t) E cp(t) for every t E T. Let f: T ~ R be the continuous function equal to IIz(t)II for every t E T and to (2(n + 1))-1 for tEnT. We have

II Yn+I(t) - ~
T\T.

(Yn+I(t) xiUXj(t) - f(t)X n+1(t)/1 = IIz(t) - f(t)x n+1(t)11

and this vanishes on T and is dominated by 2/2(n + 1) = 1/(n + 1) on 10.8.8. THEOREM. Let T be a paracompact space of finite dimension, and .st1 a separable continuous field of elementary C*-algebras over T, of rank l\o, satisfying Fells condition. Then d is locally trivial. Let to E T. There exist a closed neighbourhood V of to, and a con tinuous field p over V, relative to d, such that, throughout V, p(t) is a projection of rank 1. Put f3(dl V, p) = (n, x) and n = H(t, I). Then n

CH. 10. 9] APPLICATION TO C*-ALGEBRAS WITH CONTINUOUS TRACE

245

is separable (10.7.5), each H(t) is of dimension ?\o and V is paracompact of finite dimension, hence Je is trivial (10.8.7). Now d I V is isomorphic to .sI1(Je) (10.7.6), and is therefore trivial. Reference: [463].
10.9. Application to C*-algebras with continuous trace

10.9.1. Let A be a separable C*-algebra with continuous trace, and d the continuous field of elementary C*-algebras over A defined by A, which satisfies Fells condition (10.5.8). Moreover, A is locally compact with countable base (3.3.4, 3.3.8 and 4.5.3), and therefore paracompact. We can therefore form the element 8(d) of H 3(.A, Z) (10.7.14), which will be denoted by c5(A). 10.9.2. Let T be a locally compact space, and Je a continuous field of non-zero Hilbert spaces over T. Then d(Je) satisfies Fells condition (10.7.7). Let A be the C*-algebra defined by .sI1(Je) (10.4.1). Then A has T for spectrum (10.4.4) and has continuous trace (10.5.2 and 10.5.8). We will say that A is defined by Je. 10.9.3. THEOREM. Let A be a separable C*-algebra with continuous trace. Then A is isomorphic to the C*-algebra defined by a continuous field of non-zero Hilbert spaces over .A, if and only if 8(A) = O. Let Je be a continuous field of non-zero Hilbert spaces over A such that A is defined by .sI1(Je). Then c5(A) = c5(d(Je)) 00.5.2), and hence 8(A) = 0 (10.7.15). Suppose that c5(A) = O. Let .sI1 be the continuous field of C*-algebras defined by A. There exists a continuous field Je of Hilbert spaces over A such that d is isomorphic to d(Je) (10.7.15). Then A is isomorphic to the C*-algebra defined by d(Je) (10.5.4). 10.9.4. DEFINITION. Let a be a cardinal. A C*-algebra A is said to be homogeneous of degree a if every irreducible representation of A is of dimension a. 10.9.5. THEOREM. Let T be a locally compact space with countable base and of finite dimension. (i) Let A be a separable C*-algebra with continuous trace, homo geneous of degree ?\o, with spectrum T. Then the continuous field of elementary C*-algebras over T defined by A is locally trivial.

246

CONTINUOUS FIELDS OF C*-ALGEBRAS

[CH. 10, IO

(ii) The mapping A ~ 8(A) defines a bijection of the set of separable C*-aLgebras with continuous trace, homogeneous of degree ?\O and with spectrum T (taken up to isomorphism) onto H 3(T, Z). Assertion (i) follows from 10.8.8. Let A, A be two separable C*-algebras with continuous trace, homogeneous of degree ?\O and with spectrum T. Let .sIl,.sIl be the continuous fields of C*-algebras over T defined by A, A. Suppose that 15(A) = 8(A ), i.e. that 15(SIl) = 15(.sIl ). The fields .sIl and SIl are of rank ?\o and locally trivial by (i). Then SIl and SIl are isomorphic (10.804), hence A and A are isomorphic. Let 15 E H 3(T, Z). There exists a locally trivial continuous field .sIl of elementary C*-algebras over T of rank ?\O such that 8(.sIl) = 15 (10.8.4). This field satisfies Fells condition, and therefore defines a C*-algebra A with continuous trace, with spectrum T, and homogeneous of degree ?\O (10.5.8). Moreover, .sIl is separable (10.2.7), hence A is separable (10.4.7). The continuous field of C*-algebras defined by A is isomorphic to .sIl (10.5.2), and therefore 8(A) = 8. 10.9.6. COROLLARY. Let T be a locally compact space with countable base, of finite dimension, such that H 3(T, Z) = O. Let A o be an elementary C*-algebra of rank ?\O, and A the C*-algebra of continuous mappings t~x(t) of T into A o such that Ilx(t)" vanishes at infinity. Then every separable C*-algebra with continuous trace, homogeneous of degree ?\o and with spectrum T, is isomorphic to A. In fact, A is separable (1004.7), with continuous trace (10.5.8), with spectrum T (100404), and homogeneous of degree ?\O (10.404). It is then enough to apply 10.9.4 (ii). References: [463], [584].

10.10. Addenda

10.10.1. Let (Aj)jEI be a family of C*-algebras. The restricted product of the A j is the C*-algebra defined by a continuous field of C*-algebras over the discrete space 1. 10.10.2. Let T be a locally compact space, B a C*-algebra, and A the C*-algebra of continuous mappings of T into B vanishing at infinity. Then the space .Ii may be canonically identified with the space T x B.

CH. 10, 101

ADDENDA

247

More generally, in the situation of 10.4.3, one knows how to describe the topology of A. [584]. 10.10.3. Let A be a C*-algebra. Form the compact space H(A) and the canonical injection of Prim(A) into H(A) (1.9.13). Identify the set Prim(A) with a subset of H(A) (in general, the topology induced on Prim(A) by that of H(A) is not the Jacobson topology of Prim(A) since this latter is not always Hausdorff). The closure T of Prim(A) in H(A) is called the regularised spectrum of A. For every t E T, let l(t) be the closed two-sided ideal of A consisting of those x E A on which the semi-norm t vanishes; let A(t) = A/I(t); let x(t) be the canonical image of x E A in A(t). The set T of the vector fields t ~ x(t)(x E A) possesses the following properties: (1) T is an involutive algebra of vector fields; (2) the set of x(t), for x E A, is the whole of A(t); (3) for every x E A, the function t ~ Ilx(t)1I is continuous on T; (4) the set of these fields is complete for uniform convergence. A(t, n is a continuous field of C*-algebras if and only if Prim(A) is Hausdorff for the Jacobson topology. [584]. 10.10.4. Let T ={O, 1, 2, 3, ... , oo}, endowed with the usual topology. For t E T, let A(t) be the C*-algebra defined as follows: if t =00, A(t) =C; if t~ 00, A(t) is the C*-algebra of complex 2 x 2 matrices. Let 8 be the set of vector fields t ~ x(t) E A(t) possessing the following properties (we put x(t) = (Xij(ti,I~I.2 for t~ 00): lim xnCn)
n~

= lim x21(n) = lim x22(2n) = 0, n-+OO

n~

lim xll(n) = lim x22(2n + 1) = x(oo).

n-J>OO

n-tolXl

Then d = A(tteT> 8) is a continuous field of C*-algebras. Let A = 8 be the C*-algebra defined by d. We have A = T. The ideal meA) (4.5.2) is the set of x E A such that x(oo) = 0, A is a liminal algebra, with generalised continuous trace (4.7.12), and does not have continuous trace; d does not satisfy Fells condition at the point 00. [582]. 10.10.5. Let H =L~([O, where n = 1, 2, ... , Si
=

1]). For every finite sequence 0 or 1, put

k,
=

=(SI, .. , sn)

L
i~l

sJi,

248

CONTINUOUS FIELDS OF C*-ALGEBRAS

[CH. 10, IO

If sand t are two such sequences of n terms, denote by Qs,t the operator on H which maps f E H to g E H, with
g(x)

=0

for xe Is,

g(ks

+ u)

= [tk,

+ u)

if O::s:;; u ::s:;;2-n

Let A(n) be the C*-algebra of operators generated by the Qs" (s, t sequences with n terms). Then A(n) is isomorphic to the C*-algebra of complex 2n x 2n matrices. We have A(n) k A(n + 1). Let A be the sub-C*-algebra of 2(H) generated by the A(n). Let A(oo) be the com mutative C*-algebra of multiplication operators by continuous complex valued functions on [0, 1]. We have A (00) k A. Let T ={O, 1, ... ,oo} endowed with the usual topology. The A(t), where t E T, are already defined. Let 8 be the set of continuous mappings x: T ~ A such that x(t) E A(t) for every t E T. Then d = A(t,8) is a continuous field of C*-algebras. Let A = 8 be the C*-algebra defined by d. Then the elements of A are the lTn: x ~ x(n)(n < 00), and the lTx : x ~ X(x(oo, where X E A (00) The C*-algebra A is liminal, with generalised continuous trace. The sequence (lT n ) (n = 1,2, ...) converges to all the lTx simultaneously (Fell, unpublished.)
A

10.10.6. Let A be a C*-algebra.

(a) Prim(A) is discrete if and only if A is the restricted product of a family of C*-algebras with no non-trivial closed two-sided ideals. (b) A is dual if and only if A is liminal and has a discrete spectrum.
[893].

10.10.7. Let A be a liminal (resp. postliminal) C*-algebra, T a compact space, and B the C*-algebra of continuous mappings of T into A. Then B is liminal (resp. postliminal). (If A is liminal, apply 10.4.3. If A is postliminal, let Up) be a composition sequence of A such that the Ip+1/I p are liminal. Let Jp be the set of continuous mappings of T into Ip Then the Jp form a composition sequence of B, and Jp+dJp is isomorphic to the C*-algebra of continuous mappings of T into Ip+dIp.) [896].
10.10.8. Let A be a liminal C*-algebra, U a dense Hausdorff open subset of A, and I the closed two-sided ideal of A such that i = U. Let d = A(t, 8) be the continuous field of elementary C*-algebras over U defined by i. For every x E A, and every t E U, let x(t) be the canonical image of x in A(t) (the image of A under the representation t is the same as that of I, namely ~(H,. For every x E A, let x be the field t -+ x(t) over U. Then the mapping x -+ x is injective. The field x is

CH. 10, lOj

ADDENDA

249

not, in general, an element of e, but, for every y E I, the fields xy and yx are in 8. Suppose that A has generalised continuous trace and 1= l(A) (4.7.12). Then, for every x E A, we have x E 8. (For every y E I, the function t ~ IIx(t) - y(t)1I is continuous on U because the points of U are separated in A (3.9.4. *10.10.9. Let S be the two-dimensional sphere, and T the compact space S x S x S x .. . There exists over T a continuous field Jt = H(t), r) of Hilbert spaces of dimension ?\o, which is not locally trivial at any point of T. Let .sII be the C*-algebra with continuous trace defined by .sII(Jt). Then A does not possess any closed two-sided ideal I#-O such that I is isomorphic to the C*-algebra of continuous mappings X ~ A o which vanish at infinity (X, a locally compact space, A o, an elementary C*-algebra). [463]. 10.10.10. Let T be a topological space, and

two continuous fields of elementary C*-algebras over T, satisfying Fells condition. For every t E T, let A(t) be the elementary C*-algebra A.(t)@A 2(t ) (2.12.15). Let r be the set of vector fields of the form
t ~ al(t) a2(t)

+ bl(t)

b 2(t)

+ ... + kl(t)

k 2(t),

where all b., ... , k, E 81> a2 b 2, ... , k 2 E 8 2, There exists exactly one continuous field of elementary C*-algebras .sII = A(t)), 8) over T such that 8:2 r. We put .sII = .sill .s11 2 This field satisfies Fells condition. If T is paracompact, we have

If .sill is separable and if .s112 is locally trivial and of infinite rank, then .sill @.sII 2 is locally trivial. [450].

10.10.11. Problem: let (A(t, 8 ) be a continuous field of postliminal C*-algebras over T. For every t E T, let B(t) be the largest liminal ideal of A(t). Let 8 be the set of x E 8 such that x(t) E B(t) for every t. Is B(t, 8) a continuous field of C*-algebras?

CHAPTER 11

EXTENSION TO C*-ALGEBRAS OF THE STONE WEIERSTRASS THEOREM

Let A be the C*-algebra of continuous complex-valued functions vanishing at infinity on a locally compact space. The Stone-Weierstrass theorem gives conditions which ensure that a sub-C*-algebra B of A is equal to A. Now the C*-algebras of type A are just, up to isomorphism, the commutative C*-algebras. In this chapter, we are going to extend the Stone-Weierstrass theorem to arbitrary C*-algebras. 11.1. The case of postliminal C*algebras 11.1.1. DEFINITION. Let A be a C*-algebra, and B a sub-C*-algebra of A. Then B is said to be a rich sub-C*-algebra of A if the following conditions are satisfied. (i) For every irreducible representation TT of A, TT IB is irreducible; (ii) If TT and TT are inequivalent irreducible representations of A, then TT IBand TT IB are inequivalent.
It follows from this that if TT is a non-trivial irreducible representation

of A, then TT I B is non-trivial. 11.1.2. LEMMA. Let A be a C*-algebra, B a rich sub-C*-algebra of A, and 7T an irreducible representation of B. Then there exists an irreducible representation TT of A in H" such that TT = TT IB. There exists a Hilbert space H :2 H", and an irreducible represen tation TT of A in H such that TT(x) = TT(x) IH" for every x E B (2.10.2). Then TT I R is irreducible since B is rich, and H" is invariant under TT(B), hence H" = H. 11.1.3. LEMMA. Let A be a C*-algebra, B a rich sub-C*-algebra of A, and I a closed two-sided ideal of A. (i) B/(B n I) is a rich sub-C*-algebra of A/I. (ii) B n I is a rich sub-C*-algebra of I.

CH. II.

su

THE CASE OF POSTLIMINAL C*-ALGEBRAS

251

Assertion (i) follows immediately from the definitions. We prove (ii). Let 7f be an irreducible representation of I and let us prove that tt IBn I is irreducible. We can suppose that 7f- 0. There exists an irreducible representation p of A in H" extending tr (2.10.4). Since B is rich, p IB is irreducible. Since B n I is a closed two-sided ideal of B, tr IBn I is trivial or irreducible (2.11.2). If tt IBn I = 0, then p IB defines an irreducible representation of BI(B n I) which, by (i) and 11.1.2, extends to an irreducible representation of AI I; in other words, p IB extends to an irreducible representation p of A which vanishes on I; since B is rich, p is equivalent to p, hence tt = P II = contrary to hypothesis. Hence 1T IBn I is irreducible. Now, let 1T,7f be two ir reducible representations of I and suppose that 7f IBn I = 7f IBn I. Let p, pi be representations of A in H", H"" extending 7f, tr (2.10.4). If tt IBn I and 7f IBn I = 0, we see, as above, that tt and tt are trivial. If not, pi B = p I B (2.11.2) and so p = pi as B is rich, and -tt = 1T. 11.1.4. LEMMA. Let A be a liminal C*-algebra with Hausdorff spectrum, and B a rich sub-C*-algebra of A. Then B = A. Let d be the continuous field of elementary C*-algebras over A defined by A. Then A may be identified with the C*-algebra defined by d (10.5.4). If tr E A, we have -tr I B E B, hence 7f(A) = 7f(B) = n(H",). If 7f and at are two inequivalent irreducible representations of A, then tt IBand tt IB are inequivalent; hence, if T E n(H",) and T E n(H",,), there exists x E B such that 7f(x) = T, 7f1(X) = T (4.2.5). Then B = A by 10.5.3. 11.1.5. LEMMA. Let A be a C*-algebra, I its largest postliminal ideal, and B a rich sub-C*-algebra of A. Then B :1 I. There exists a composition sequence (Ip)O",p"a of I such that the IpHllp are C*-algebras with continuous trace (4.5.5), and therefore liminal C*-algebras with Hausdorff spectrum (4.5.3). We have 10 = ~ B. Sup pose that there exists a least ordinal p such that Ipt, B. We have I p ~ B for p < P so that p is not a limit ordinal. Let p = p + I. By 1I. 1.3(i), BlIp is a rich sub-C*-algebra of A/lp" By 11.1.3(ii), (B n Ip)lIp is a rich sub-C*-algebra of ijt, By 11.1.4, (B n Ip)lIp = whence B:1 t; which is a contradiction. Hence I = la ~ B.

ut,

252

C*-ALGEBRAS

OF THE STONE-WEIERSTRASS

THEOREM

[CH. II, I

11.1.6.

PROPOSITION.

In a postliminal C*-algebra A, every rich sub-C*

algebra is equal to A.

This is a special case of 11.1.5.

11.1.7. Let E, F be two vector spaces in duality, A c E, B c F. We will say that A separates B if, for every pair (x, y) of distinct points of B, there exists an a E A such that (a, x),= (a, y).
LEMMA. Let A be a C*-algebra, and B a sub-C*-algebra which separates peA) U {O}. Then B is a rich sub-C*-algebra of A.

Let 1T E A. Suppose that H Tr is the direct sum of two non-zero subspaces HI, Hz invariant under 1T IB. Let gl E HI>gz E Hz, be non-zero vectors such that Ilgl + gzll = 1. Then
x ~ (1T(x )(gl

+ gz) Igl + gz}

and

x ~ (1T(x )(gl - ~z)

Igi -

gz)

are pure states t.. fz of A. If x E B, we have

!J(x)

= (1T(x )~I I~I) +(1T(x )~zl

= (1T'(X)~1

~z)

1~I) +(1T(x)(-

gz) 1- ~z}

=fz(x).

Since B separates P (A), we have t, = fz, and so ~I + ~z is proportional to ~I - gz (2.5.7), hence gl and ~z are proportional, which is absurd. Hence 1T IB is irreducible. Let 1T1, 1Tz be two irreducible representations of A, and suppose that there exists an isomorphism U of H TrI onto H Tr2 mapping 1T1 I B to 1Tzl B. Let ~I be a unit vector of H Tr" ~2 = U(gl), and t, the element of peA) U {O} defined by 1Tj and ~i' We have !J(x) = fz(x) for x E B, and so t, = fz Hence, either 1T1 and 1Tz are both trivial, or else 1TI and 1Tz are both non-trivial; in this latter case, ~i is a cyclic vector for 1Tj, and the equality t, = fz implies that 1T1 = 1Tz (2.4.1).
11.1.8. THEOREM. Let A be a postliminal C*-algebra, and B a sub-C* algebra of A. If B separates P (A) U {O}, then B = A.

This follows from 11.1.6 and 11.1.7. We do not know if the hypothesis that A be postliminal is essential. For general C*-algebras, we will later obtain (11.5.2) a result less precise than 11.1.8. References: [630], [896]. This section is from some unpublished work of Fell.

CH. 11. 2j

ABUNDANCE OF PURE STATES IN CERTAIN C*-ALGEBRAS

253

11.2. Abundance of pure states in certain C*algebras

A a sub-C" 11.2.1. LEMMA. Let H be a Hilbert space, C = ~(H), algebra of Y!(H) containing 1, and f a state of A vanishing on A n c. (i) f is a weak*-limit of vector states of A (2.4.2). (ii) If A is irreducible in H, I is a weak*-limit of pure states of A.
(ii) follows immediately from (i); we prove (i). Suppose, to start with, that A :2 C. Then I defines a state f of AIC which is a weak*-limit of convex combinations of pure states of AIC (2.5.6).

Hence f is a weak*-limit of states of the form Adl + ... + A,jn, where AI, .. , An ~ 0, Al +... + An = 1, and where II" .. .I, are pure states of A vanishing on C. It is then enough to study the case where f is itself of the form Ad, +... + A,jn with the above properties. Let XI" Xs be hermitian elements of A such that XI = 1. We are going to construct unit vectors g., ... , gn in H such that (Xjgj J gk) = 0 for j < k and such that INx;} - wfj(x;}1 ~ 1 for any i and j. Suppose that the gj have been constructed for j < m. Let K be the linear subspace of H generated by the xgj(l ~i ~s,j<m), and K = H K. We have PKE C~ A hence PK,AP K, is a subalgebra of A, and IPK,AP K, is a state of PK,AP K, because

t;

Im{PK.) = I;(PK) + fm{PK)

= fm(l) = 1.

If 1m IPK,AP K, = Ag, + (1- A)g2 where 0 < A < 1 and g" g2 are states of P K,AP K, denote by gi the state X ~ gj(PKXP d of A; since f m vanishes on C, we have 1m = Ag i + (1- A)g2 hence g, = g2 since 1m is pure, hence s, = g2 and so 1m IPK,AP K, is a pure state of PK,AP K ,. By 3.4.1, t; IPK,AP K , is a weak*-limit of forms wa where the ?a are unit vectors in K. Hence, for 1 ~ i ~ s,

is the limit of

We can therefore find

gm E

K such that
for 1 ~ i ~ s.

Ifm(Xi) - WgJx;}J ~ 1

The construction of the gj can therefore proceed by induction. This established, let g = yAlg l + ... + YAng n. Since x, = 1, we have (gj Igk) = 0 for j ~ k, and so g is a unit vector. Since Xj is hermitian, we have (xjgj Igk) = 0

254

C*-ALGEBRAS

OF THE STONEWEIERSTRASS

THEOREM

[CH. 11. 2

for

k, hence

This proves that ~j~1 Ajj is a weak*-limit of vector states of A. Consider now the general case. Let AI be the C*-algebra A + C. Then AdC may be identified with AlA n C, and so / defines a state of AdC, i.e. / extends to a state f of Al vanishing on C. By the above work, f is a weak*-limit of vector states of AI and so / is a weak*-limit of vector states of A. 11.2.2. Let A be a C*-algebra. For every representation (J of A, we will provisionally denote by C6 the set of x E A such that (J(x) is compact. We have C6 = (J-l(.;ee(H 6 hence C6 is a closed two-sided ideal of A.

0/ states

Let A be a unital antiliminal C*-algebra, and P(A) the set which vanish on at least one C 6 (8 E A). Then P (A) and P (A) have the same weak*-c1osure.
LEMMA.

Let x be an hermitian element of A such that /(x) ~ 0 for every A, and let TT6 be a representation of A with kernel C 6 Every state associated with 7T6 is an element of P (A), and is therefore ~o at x; hence 7Tix) ~ o. Now 6E A C6 = 0 because A is antiliminal (4.2.6). Hence EB6EA 7T6 is isometric, from which it follows that x ~ O. By 3.4.1, the weak*-c1osure of P (A) contains P (A). If g E P (A), there exists a 8 E A such that g( C 6 ) = O. Hence g defines a state g of 8(A) vanishing on 8(A) n .;ee(H 6 ) . By 11.2.1, g is a weak*-limit of pure states of (J(A), and so g is a weak*-limit of pure states of A. The weak*-c1osure of P (A) thus contains P (A).
/ E P (A). Let () E

11.2.3. Let A be a C*-algebra. Throughout the rest of this chapter, P (A) will denote the weak*-c1osure of P(A) in the dual of A, and E(A) the set of states of A.
LEMMA.

Every state

Let A be a unital antiliminal C*-algebra, and / E P(A). 0/ A vanishing on Ker TTl belongs to P (A).

CH. 11, 3j

STATEMENT

OF THE THEOREM

255

By 3.4.3, it is enough to prove that the state

x ~ AI!(ytXYl) +...

+ AJ(y~xYn)

(AI... ,An ;;. 0, Yl> ... , Yn EA, ~

AJ(Y~y;)

= 1)

belongs to P (A). Now f is a weak*-limit of states fa E P (A) (11.2.2), and the state x ~ Al!a(ytXYl) +... + AJa(y~xYn) clearly belongs to P (A), and therefore to P (A) (11.2.2). 11.2.4. LEMMA. Let A be a unital antiliminal C*-algebra. Suppose that two non-zero closed two-sided ideals of A always have a non-zero intersection. Then E(A) = P (A). The condition on the ideals of A means that, in A, two non-empty open subsets have a non-empty intersection (3.2.2). Let R be the equivalence relation defined in P (A) by the canonical mapping P (A) ~ A. If f E P (A) the elements of P (A) congruent to f mod R are the states x ~ f(u*xu), where u belongs to the unitary group G of A (2.8.6). Since A is the quotient space of P (A) by R, two non-empty open subsets of P (A), saturated for R, have a non-empty intersection. This established, let f), f2 E P(A), A E [0,1]. By the above, Afl + (1- A)f2 is a weak*-limit of states
(1)

x ~ Af(x) + (1- A)f(u*xu),

where f E P(A) and u E G. Now the state (1) vanishes on Ker rr, and therefore belongs to P (A) (11.2.3). Hence Afl + (1- A)f2 E P (A). By induction, every convex combination of pure states of A belongs to P (A). Hence E(A) c P (A) (2.5.6). Reference: [630). 11.3. Statement of the theorem 11.3.1. THEOREM. Let A be a unital C*-algebra, and B a sub-C* algebra of A containing 1. If B separates P (A), then B = A. The proof will be given in 11.5.1. 11.3.2. B=A.
LEMMA.

Let A and B be as in 11.3.1. If B separates E(A), then

256

C*-ALGEBRAS

OF THE STONE-WEIERSTRASS

THEOREM

[CH. II. 4

Let A h (resp. B h ) be the set of hermitian elements of A (resp. B). Suppose that A;;t B, so that Ah;;t B h. By the Hahn-Banach theorem, there exists a non-zero continuous hermitian linear form f on A such that f(B) = O. Let f = g - g, where g, s are positive forms on A (2.6.4). We have
0= f(1)
=

g(1) - g(1),

hence g(1)

g(l) ;;t O.

Multiplying f, g, g by the same scalar, we can suppose that g(l) = g(l) = 1, and hence g, g E E(A). Then B does not separate g, g, nevertheless g;;t s. 11.3.3. By 11.2.4 and 11.3.2, theorem 11.3.1 is already proved if A is an antiliminal C*-algebra in which the intersection of two non-zero closed two-sided ideals is non-zero. We shall reduce the general case to this special case. References: [15], [630].

11.4. Several lemmas (For a simplification of this section, see [15].) 11.4.1. LEMMA. Let A be a unital C*-algebra, and B a sub-C*-algebra of A containing 1 and separating P(A). Let K]o K 2 be two closed two-sided ideals of A, K = B n (K 1 + K0, K = B n K[ + B n K 2 Then K=K.
It is clear that K:2 K. Let f be a pure state of B such that f(K) =o. We will show that f(K) =0, which will complete the proof (2.7.3). There exists an !E P(A) extending f (2.1O.1(ii)); since B separates P (A), this extension is unique. By 2.1O.1(iii) we have, for every her mitian element x of A,
(1)
!,(x) =

sup
y=y*EB,y ..x

f(y).

Let x E If y = y* E Band y ~ x, we have, denoting the canonical morphism of A onto A/K 1, by tp, that cp(y) ~ cp(x) = 0, hence cp(y+) = cp(yt = 0, hence y+ E K!, hence f(y+) = 0, and f(y) ~ O. Then (1) proves

x;

CH. 11, 4]

SEVERAL LEMMAS

257

that !(x) = O. Thus, f(K) = O.

t vanishes on K!, and similarly on K 2 , and so

11.4.2. LEMMA. Let A and B be as in 11.4.1 and 7Tl and 7T2 be represen tations of A such that 7T\(A) = 7Tl(B), 7T2(A) = 7T2(B). Then (7Tl EB 7T0 (A) = (7T\ EB 7T2)(B).

Let K, = Ker 7Tj. We have A mediately have that

= B + K 1 = B +K 2,

whence, we im

In view of 11.4.1,
K\

(B

n K\)+(B nK0+K 2=(B n K\)+K2

From this, we deduce that and consequently which proves the lemma.
11.4.3. LEMMA. Let A be a unital antiliminal C*-algebra, and~a sub-C*-algebra of A containing 1 and separating peA). If fl,f2EP(A), we have (7Th EB 7Th)(A) =(7Th EB 7Th)(B). If f E P (A), every state of A vanishing on Ker 7Tf belongs to P (A) (11.2.3), and so 7Tf(B) separates the states of 7Tf(A), hence 7Tf(B) = 7Tf(A) (11.3.2). It is then enough to apply lemma 11.4.2.
Let H be a Hilbert space, T E 2(H) be such that a unit vector of H such that (n I~);;:. 1- e. (i) If f: R ~ R is a non-negative Borel function such that 1 in [1- e1/2, 1], then we have (f(T)~ I~);;:. 1- e 1/2 (ii) If T E 2(H), we have I(rg I (Tn I~)l =;;; 2e\/~IT'II (iii) We have T + 3e \/2;;:. P q. (iv) If S is an hermitian operator on H such that S;;: T, and if g: R~R is a non-negative Borel function such that g;;:1 in [e, +00[, then (g(S)~ I~);;:. 1 - 4e 1/2.
LEMMA.

o=;;; T =;;; 1, e E [0,1], and ~

11.4.4.

e) -

258

C*-ALGEBRAS

OF THE STONE-WEIERSTRASS

THEOREM

[CH.

II, 4

Let T =fol A dE). be the spectral decomposition of T. Put 1- EI/Z We have

1 - El/z(Ea~

=a.

I~) = (1- EI/2)(Ea~

I~) + (1- Ea)~ I~)

~
whence (Ea~
I~) ~
E

J d(E).~
A
o

I~) +

J d(E).~
A

I~) = (n I~) ~

1-

E,

I/Z,

and

(f(T)~

I~)

~ (1- Ea)~

I~)

~ 1- fEI/Z.

Whence (i). If T E .:t(H), we have, in view of 9.3.3,

I(T'~

I~) -

(T'T~

I~)I = I(T'(~ - n) I~)I ~ IITIII ~ - T~II ~

(2fE)

IIZ

IIT II,

whence (ii). Let (E H be a unit vector. We have ( = A(I + JL~ with A, JL E C and (I a unit vector of H orthogonal to ~. Let P be the orthogonal projection onto

cr, + C~

in H. Let

(~

the basis

(.. ~).

q) be the matrix of PTP with respect to

We have
r

=(PTP~

I~) =(n I~)

E [1- E, 1].

Since O~PTP~P, Hence

T - P ct )(

we have pE[O,I] and Iqlz~(I-p)(1-r)~E.

I() =(pA + qJL)A + (ijA + (r-l)JL)}:l: ~ PIAl z + (r -l)IJLlz - 21qAJLI ~ 0 -

E -

2El/2~

-3E

I/Z,

whence (iii). Let S = fo" A dF). be the spectral decomposition of S. We have ~ = P~I + lT~2 with p, 0 E C, ~l and ~z unit vectors of (1 - FE)(H) and of FE(H) respectively. Let Q be the orthogonal projection onto C~I + C~z in H, We have QSQ =alPc~1 + a2PC~' with al ~ E ~ az ~ O. Hence, in view of (iii),
4E /
1Z

a2

+ 3E I /2 = QSQ + 3EI/Z)~zl
~ (P c~zl ~2)

~z) ~ QTQ

= ~zl

~)~

I ~2)

lulz.

+ 3El/2)~21

~z)

CH. II. 4]

SEVERAL LEMMAS

259

Then
(g(S)~

I~);;':

1 - F.)~

I~)

Ipl2 = 1 -luj2;;.: 1- 4112.

ll.4.5. LEMMA. Let A and B be as in 11.4.3, K a closed two-sided ideal of A, f E P (A) and e E [0, ~[. Suppose that there exists x E K such that o E; x E; I and f(x);;.: 1- e. Then there exists ayE B n K such that OE; y E; 1 and f(y);;.: 1- l/2.
~x'

By 11.4.3, there exists x E B such that 7rf(x) = 7rf(x); replacing x by + x*), we can suppose that x is hermitian. If x - x E;!, let p : R ~ R be the function equal to 0 on ]- 00, !], to 1 on [1_ 1I2,+00[, and linear in [!, 1_1/2]; let y = p(x)EB; we have OE;y E; 1; moreover, if ta is the canonical morphism of A onto AIK, we have w(x) = w(x- x) E;!, hence w(y) = p(w(x)) = 0, and hence y E K; lastly,
7rf(Y)

=P( 7rf (x )) =P( 7rf (x )) ,

hence by 11.4.4(i). If nothing is known about x- x, we shall modify x in a way that reduces us to the above case. If g E P (A), there exists tg E B such that
7rf(tg) = 7rfx - x)")

=0

and

7rg(tg) = 7rgx - xt) (11.4.3);

replaci~ by !(tg + t:t, we can suppose that t g ;;.: O. Let Wg be the set of h E P (A) such that h(tg) > h(x - x) -!; this is a neighbourhood of g in P(A), since

g(t g) = gx - xt);;.: g(x - x)

> g(x -

x)

-!.

Let (W g 1 , x" = x - ~7=1

, Wg) be a cover of P (A) (which is compact), and let tg; E B; we have 7rf(x") = 7rf(x) = 7rf(x), Moreover, if h E

P (A), we have h E W g ; for some index i, and therefore

h(x"- x) = h(x - x) - h(t gl ) - ... - h(t 8" ) E; h(x - x) - h(t 8,.) <
1 -2

hence x" - x E;! and the proof is complete. 11.4.6. LEMMA. Let A and B be as in 11.4.3, K be a closed two-sided ideal of A, XI" .. .x; elements of K such that 0 E; Xj E; 1, and E ]0, H. Let Q be the set of f E P (A) such that f(xj);;: 1 - e for at least one i. Then there exists ayE B n K such that 0 E; Y E; 1 and f(y);;.: 1- 61/4 for every f E Q.

260

C*-ALGEBRAS

OF THE STONE-WEIERSTRASS

THEOREM

[CH. 11, 4

Let f E Q. There exists an if such that f(Xif ) ~ 1- E, then a Yf E B n K such that 0 os;; Yf os;; 1, f(Yf) ~ 1 - E 1/2 (11.4.5). Let "1 be the set of g E Q such that g(Yf) ~ 1 - 2E 1/ 2 ; this is a neighbourhood of f in Q; let (Wft, .. , "1) be a finite cover of Q. Let x = Yft + ... + Yf, Let p: R~ R be the function equal to 0 in ]- 00, 0], to 1 in [2e 1/2, +oo[ and linear in [0,2E 1/2] . Let Y = p(x). We have Y E B n K and 0 OS;; Y OS;; 1. If f E Q, we hence f(y)~1-4(2el/2)1/2~ have fEWft say, hence f(Yft)~1-2E1/2, 1-6e l /4 by 11.4.4 (iv).
11.4.7. LEMMA. Let A and B be as in 11.4.3, K be a closed two-sided ideal of A,!J, ... .t, states of A such that each t, IK is a state of K, and E > O. Then there exists ayE B n K such that 0 OS;; Y os;; 1 and Ny) ~ 1 - E, .. .t, (y) ~ 1 - e.

We can suppose that (E/12)1/4 E ]0, ~[. For lOS;; i OS;; n, there exists Xi E K such that 0 os;; Xj os;; 1 and fi(Xj) > 1- (E/12)3. Let Q be the set of f E peA) such that f(Xi) ~ 1- (E/12)4 for at least one i. There exists ayE B n K such that 0 os;; Y os;; 1 and f(y) ~ 1- e/2 for every f E Q 01.4.6). There exist convex combinations Ail4 +... + A~.ti; of pure states of A which converge weak*- to Ii Let J a be the set of indices j E {I, 2, ... , p} such that lij(xi) ~ 1- (E/12t; let 13a = ~iEJa Aij; since
(A4f4 +...

+ A~f~)(xJ

os;; 13a

(1 - C~) )0- 13a),

13,,),

we have, for sufficiently large a,

1-

C~y
4

OS;; 13" + (14

C~))O-

C~)
Now, for j E J a , for sufficiently large a,

13a

~ C~)

c~r
~

e 1--. 12 we have fij E Q, hence fij(y)

13" ~

1- E/2. Consequently,

(~Aijfij)(y)
1

13a( 1-~)

(1-

1~)( 1-~) ~

1- e

and, in the limit, My) ~ 1- e. References: [15], [630].

CH. \ J. 5j

PROOF OF THE THEOREM

26\

11.5. Proof of the theorem 11.5.1. Let A be a unital C*-algebra, and B a sub-C*-algebra containing 1 and separating P (A). Suppose that B,," A. The aim is to arrive at a contradiction. Let M be the set of continuous hermitian linear forms on A of norm ,,;;; 1 which vanish on B. By the Hahn-Banach theorem, M,," O. Moreover, M is convex and weak*-compact; by the Krein-Milman theorem, M admits at least one extreme point m,," O. It is obvious that 11m II = 1. Let 1 be the largest two-sided ideal of A contained in Ker m; this ideal is closed. We have B + I ~ Ker m, and hence (B + 1)1I,," AI1. Every state of AI I which is a weak*-limit of pure states of AI I defines a state of A which is a weak*-limit of pure states of A. Hence (B + 1)1 I separates P(AII). We show that All is antiliminal. Let J be a closed two-sided ideal of A containing I such that JII is postliminal. By 11.1.5 and 11.1.7, we have JII c (B +1)11, whence J c B + I c Ker m, whence J c 1 by definition of I; hence JII =0 and we have indeed proved that AI I is antiliminal. This at once shows that it is enough to establish theorem 11.3.1 for antiliminal C*-algebras. We therefore add to our previous hypotheses the hypothesis that A be antiliminal. We are going to show that, in All, the intersection of two non-zero closed two-sided ideals is non-zero. Thanks to 11.3.3, we will thus have obtained the looked-for contradiction. Let K, L be two closed two-sided ideals of A not contained in I; we show that K n Lft, 1. There exist two positive forms /t, /2 on A such that /1 - /2 = m, II/III+ 11/211 =Ilmil = 1 (2.6.4). Moreover, for i = 1,2, there exist two positive forms gj, hi on A such that /i = gi + hi! MK) =0, Ilgi IKII = lid (2.11.7). Let x E B; for every e > 0, there exists y E B n K such that Igj(x) - gj(xy)l,,;;; e for i = 1,2 (11.4.7 and 11.4.4 (iil), Since xy E K n B, we then have
Ig\(x) - g2(x)l,,;;; Ig\(xy) - g2(xy)1

= lfI(xy) Hence

f2(xy)1

+ 2e + 2e =2e.

hence, Kft, I, we have Kft, Ker m, and so g.- g2 cannot be proportional to

(gl - g0(B) =0 and, consequently, (hi - h0(B) =O. Moreover, 1 = Ilmil =IlgI- g2+ h 1 - h21 1,,;;; IIgl-gdl +Ilh l- hdl , ;;; IIglli+ IIhlll + Ilgdl + IIhdl =1I!I11 + 11f2/1 = 1; if we put A =Ilg, - gdl, JL = Ilh t - h21 1 . we have A + iL = 1. Since

262

C*-ALGEBRAS OF THE STONE-WEIERSTRASS

THEOREM

[CH. II. 5

hi - h 2 (which vanishes on K). This at once proves that A;c O. If JL;C 0, the equality m = A(A -I(g, - gJ) + JL(JL -(hi - hJ)

contradicts the fact that m is extreme in M. Hence JL =0 and m = gl - g2 Let Z E L be such that m(z);c O. There exists z E K such that
Igl(Z) - gt(zz)

and

Igiz) - g2(zz)1

are arbitrarily small (1104.7 and 11.404 (iij), hence such that
m(zz) =gl(zz) - g(zz) ;c O.

Now zz E K

L, and therefore K

Lg: 1.

11.5.2. COROLLARY. Let A be a C*-algebra, and B a sub-C*-algebra separating P(A) U {O}. Then B = A. Let A = A + Ce be the C*-algebra obtained by adjoining an identity element e to A. Let 13 = B + Ceo Let IE P (A "), Then I is a weak*-limit of pure states t, of A. If I IA is non-zero, we can suppose that the la IA are non-zero; they are then pure states of A (2.11.8), hence I IA E P (A) in this case. Hence I IA E P(A) U {O} in all cases. Then, if I and g are two distinct elements of P(A -), I IA and g IA are distinct and belong to P (A) U {O}, and are therefore separated by B. Thus, 13 separates P (A -). Theorem 11.3.1 proves that 13 = A, whence B = A. 11.5.3. COROLLARY. Let T be a locally compact space, d = A(t), e) a continuous field 01 C*-algebras over T, A the C*-algebra defined by d, and B a sub-C*-algebra 01 A such that, lor all t l E T, t z E T, ~I E A(t l ) , ~z E A(tz) (with ~I = ~z il t} = tJ, there exists x E B with x(tt) =~ .. x(tz) = ~z. Then B = A. Adjoining a point at infinity to T, we can suppose that T is compact (10.2.6). Let IE P (A). Then I is a weak*-limit of pure states la of A. By 10.4.3. there exist a ta E T and a pure state ga of A(ta) such that la(x) =ga(x(ta for every x E A. We can suppose that the ta converge to a point t of T. Let x E A be such that x(t) = 0; we have Ilx(ta)II ~ IIx(t)II = 0, hence ga(x(ta~O, hence I(x) =O. This proves that there exists a positive form g on A(t) such that I(x) = g(x(t for every x E A. This established, let J., Iz be two distinct points of P (A) U {O} and let us show that B separates /1 and /z (then. 11.5.3 will be a consequence of 11.5.2). Let t" t z E T, and let g; be a positive form on A(t;), such that

CH. 11. 6]

ADDENDA

263

!j(X) == gj(x(t j for every x E A. At least one of the two forms gl, g2 is non-zero, gl, say. If t l == t 2, we have gl #- g2 hence there exists ~ E A(t l) such that gM) == g2(~); then there exists x E B such that X(tl) == ~, and then fI(x) #- fix). If t( #- t 2, there exists y E B such that y(t 2) == 0, gl(y(tt #- 0, whence fI(y) #- f2(y).

References: [15], [584], [630], [1694]. 11.6. Addenda *11.6.1. Let H be a Hilbert space, A a sub-C*-algebra of :(H) such that 1 E A (resp. 1 ~ A), and V the weak*-closure of the set of vector states of A. Then V is the set of a(w IA) + bg where a, bE [0, 1], a + b == 1 (resp. ~ 1), ~ is a unit vector of A(H), and g is a state of A such that g(A n .;;ew(H == O. [630], [631]. *11.6.2. Let A be a unital C*-algebra, K its largest postliminal two-sided ideal, B a sub-C*-algebra of A containing 1, S the set of pure states of A such that f IK is a (pure) state of K, and T the set of states that are weak*-limits of pure states of A vanishing on K. If B separates S U T, then B == A. [630]. *11.6.3. Let A be a von Neumann algebra, and B a sub-C*-algebra of A which is antiliminal, weakly dense in A, and which contains the centre ~ of A. Then P (B) is the set of states f of B such that! I~ is a character of ~. In particular, if F is a continuous factor, then P (F) = E(F). [630]. *11.6.4. Let A be a von Neumann algebra, and ~ its centre. Then P (A) is the set of ag + (1- a)f, where a E [0, 1], g, hE E(A), (ag + (1- a)f) I~ is a character of ~,ag(e) = a for any abelian projection e of A, and (1- a)h annihilates all the abelian projections of A. [630]. *11.6.5. Let A be a von Neumann algebra, and B a sub-C*-algebra of A containing 1 and weakly dense in A. Then PCB) is the set of restrictions to B of elements of P (A). [630]. 11.6.6. Let A ~ unital C*-algebra. The following conditions are equivalent: (i) P (A) = E(A); (ii) A is antiliminal and the ideal 0 is prime. (We have (ii) ~ (i) by 11.2.4. Let I and J be non-zero closed two-sided ideals of A such that In J = O. Let f (resp. g) be a pure state of A whose restriction to I (resp. J) is non-zero. Then the restriction of g (resp. f) to I (resp. J) is zero, and ~(f + g) ~ P (A. [1708].

CHAPTER 12

THE ENVELOPING VON NEUMANN ALGEBRA OF A C*ALGEBRA

12.1. The second dual of a C*-algebra 12.1.1. PROPOSITION. Let H be a Hilbert space, A a sub-C*-algebra of 5(H), and B the weak closure of A. Suppose that 1 E B and that every norm-continuous linear form f on A is ultra-strongly continuous, and therefore admits a unique ultra-strongly continuous extension 1 to B. (i) The mapping f ~ 1 is an isometric isomorphism of the Banach dual A of A onto the predual of B. (ii) For every y E B, let y be the linear form f ~ f(y) on A. Then y --+y is an isometric isomorphism of B onto the Banach dual A" of A, whose restriction to A is the canonical injection of A into A".

If f E A, then 1 belongs to the predual of B by hypothesis. Let AI, B I be the unit balls of A and B. Since Al is ultra-strongly dense in B 1 (A 13), we have SUPYEB 1 If(y)1 = SUPYEA t If(y)l, i.e. IIfll = 11111. Moreover, every element of the predual of B is the ultra-strongly continuous extension of its restriction to A. This proves (i). In view of (i) and A 23, y --+y is an isometric isomorphism of B onto A"; if yEA, then y is the linear form f ~ f(y) on A, i.e. the canonical image of y in A".
12.1.2. COROLLARY. Let H be a Hilbert space, and A the C*-algebra of compact operators on H. (i) Every linear form f E A is ultra-strongly continuous on A, and therefore admits a unique ultra-strongly continuous extension 1 to 5(H). (ii) The mapping f ~ 1 is an isometric isomorphism of A onto the predual of 5(H). (iii) For every y E 5(H), let y be the linear form f ~ f(y) on A. Then y ~ y is an isometric isomorphism of 5(H) onto A", whose restriction to A is the canonical injection of A into A".
(i) follows from 2.6.4 and 4.1.3, and (ii), (iii) are then special cases of 12.1.1.

CH. 12.

sn

THE ~ECOND

DUAL OF A C*-ALGEBRA

265

12.1.3. COROLLARY. Let A be a C*-algebra, 7T the universal represen tation of A (2.7.6), and B the weak closure of 7T(A), which is a von Neumann algebra on H r (i) Every normal positive form on B is a form W~ with ~ E H",_ Every ultra-strongly continuous linear form on B is weakly continuous. (ii) If f E A, there exists a unique weakly continuous linear form f on B such that f(7T(X = f(x) for every x E A. (iii) The mapping f ~ f is an isometric isomorphism of the dual A of A onto the predual of B which transforms the set of positive forms on A into the set of normal positive forms on B. We have (f*)- = (f-)* for every f E A. (iv) For every y E B, let y be the linear form f ~ f(y) on A. Then y ~ y is an isometric isomorphism of B onto A", whose composition with 7T is the canonical injection of A into A". (v) This isomorphism is bicontinuous for the weak operator topology on B and the topology u(A", A) on A". Let f be a normal positive form on B. Then f /7T(A) is a positive form on 7T(A). By construction of 7T, there exists ~ E H", such that f( 7T(X = (7T(X)g Ig) for every x E A. The equality f(y) = Wt(y), valid for y E 7T(A), is still true, by continuity, for y E B. Since every ultra-strongly con tinuous linear form on B is a linear combination of normal positive forms on B (A 25), we have proved (i), We have just observed that every positive form on 7T(A) is weakly continuous. The same is therefore true of every continuous linear form on 7T(A) (2.6.4), whence (ii). By 12.1.1, f ~ f is an isometric isomorphism of A onto the predual of B. We have seen that if f ~ 0, then f is a form Wt, and is therefore normal positive, and that we thus obtain all normal positive forms on B. If f E A, we have
(f*)-(7T(X

= f*(x) = f(x*) =r(7T(X)*) for

x E A,

hence, by weak continuity, (f*)-(y) = f(y*) for y E B, i.e. (f*)- = (f-)*. We have thus proved (iii), and (iv) is a consequence of 12.1.1. By (i) and (iii), the weak operator topology on B is defined by the linear forms y ~ f(y), where f E A; the isomorphism y ~ y transforms this topology into u(A", A) since f(y) = Y(f). Whence (v), 12.1.4. DEFINITION. B is said to be the enveloping von Neumann alge bra of A and 7T is the canonical morphism of A into B. We sometimes identify A with 7T(A) and B with A".

266

THE ENVELOPING VON NEUMANN

ALGEBRA

OF A C*-ALGEBRA

[CH. 12, 1

12.1.5. The pair (B,

7T)

is the solution of a universal problem. In fact:

PROPOSITION. Let A be a C*-algebra, B the enveloping von Neumann algebra of A, and 7T: A ~ B the canonical morphism. Let p be a re presentation of A. There exists exactly one normal representation Ii of B in H p such that p(7T(x = p(x) for every x E A. The image pCB) is the weak closure of peA).

A normal representation is ultra-weakly continuous (A 27); since 7T(A) is ultra-weakly dense in B, this proves the uniqueness of p. To prove the existence of p, we can suppose that p admits a cyclic vector, since every non-degenerate representation of A is the direct sum of representations admitting cyclic vectors (2.2.7). But then there exists a positive form f on A such that p may be identified with 7Tf (2.4.6); by construction of 7T, there exists a closed subspace K of H" invariant under 7T{A) such that p may be identified with the representation x ~ 7T{x) IK. We can then take for p the representation y ~ y IK of B in K. Lastly, the image of B under a normal representation is weakly closed (A 27), whence the last assertion of the proposition. 12.1.6. Recall that if E is a Banach space, and if E , E", ... denote its successive duals, the adjoint of the canonical injection of E into E" is a projection of norm 1, called the canonical projection, of E" onto E. The kernel of this projection is the set of continuous linear forms on E" which vanishes on E. Having said this, let A be a von Neumann algebra, A its dual, and A" its second dual, identified with the enveloping von Neumann algebra of A. Recall that A may be identified with the dual P of its predual P, a closed subspace of A.
I

PROPOSITION. Let p be the normal representation of A" on A which extends the identical mapping of A (12.1.5). If we identify A and A" with pi and P" then p is the canonical projection of P" onto P. The kernel of this projection is a weakly closed two-sided ideal of A".

The canonical projection

7T

of P" onto P I is continuous for and u{P , P) = u{A, P).

u{P"" P) = u{A", A)

In view of 12.1.3 (i) and (v), p is continuous for u(A", A) and u(A, P). Now 7T and p reduce on A to the identical mapping and A is dense in A" for u{A", A). Hence 7T = p. The kernel of p is a two-sided ideal of A", which is ultra-weakly and therefore weakly closed.

CH. 12. 2]

POLAR DECOMPOSITION

OF A LINEAR FORM

267

12.1.7. Corollary 12.1.3 reduces problems about continuous linear forms on a C*-algebra to problems about ultra-weakly continuous linear forms on a von Neumann algebra. These latter problems are even more general, in certain respects, because the predual of a von Neumann algebra A is not always the dual of a sub-C*-algebra of A. References: [681], [1476], [1589].
12.2. Polar decomposition of a linear form

12.2.1. Let A be a C*-algebra, A its dual, I an element of A, and x an element of A. We denote by x . I and I x the linear forms Y ~ I(xy), Y ~ I(Yx) on A, which are elements of A. We have

[x . III~ JIxII . 11111,

III xII ~ IIxlllltll

The operators I ~ I . x (resp. I ~ X f) define a representation of A (resp. of the reversed algebra) in the (non-Hilbert) vector space A. For x, yEA and I E A, we have
(x : f)*(y)

=(x f)(y*) =I(xy*) = f*(yx*) = (f* . x*)(y)

hence (x f)* = f* . x*. If A is a von Neumann algebra and if I is ultra-weakly continuous (resp. weakly continuous), the same is true of x ] and I x. 12.2.2. LEMMA. Let A be a unital C*-algebra, A I the ball {x: IIxll:s;; I} in A, and t an extreme point 01 AI Then t*t is idempotent. We have 0 ~ t*t ~ 1, IIt*tll = 1. Let B be the commutative sub-C* algebra of A generated by 1 and t*t, n its (compact) spectrum, B the algebra of continuous complex-valued functions on Il, cp the Gelfand isomorphism of B onto B , and 1= cp(t*t). We have that 0 ~ I ~ 1, and I takes the value 1 at least once. Suppose that there exists an w E {}, such that 0 < I(w) < 1. There then exists a continuous non-negative function g : {}, .... R, such that

Ig- 0,

sup 1(1 + gi = sup 1(1 - g)2 = I

(just take the support of g to be in a sufficiently small neighbourhood of w, and Igi to be sufficiently small). There therefore exists u E B+ such that Ilt*t(1 + uill = Ilt*t(1- u)211 = 1.

268

THE ENVELOPING VON NEUMANN ALGEBRA OF A C*-ALGEBRA

[CR. 12, 2

Then t(1 + u) E AI t(l- u) E AI The equality t = !(t(l+ u) + t(1- u implies, since t is extreme, that t =t(1 + u) = t(1- u), whence tu =0, fg =0 which is absurd. Hence f only takes the values 0 and 1, from which it follows that t*t is a projection.
12.2.3. LEMMA. Let A be a von Neumann algebra, e a projection of A, and f an element of the predual of A. z z (i) IIfll ;a. lie . fll + II(1 - e) . flf (ii) If Ilfll = lie fll, we have f = e . f. Assertion (ii) follows from (i). We prove (i). Let us assume that A is a sub-C*-algebra of 2(H), H a Hilbert space. For every x E A and g E H, we have IIxgliZ= IIexgW + 11(1- e)xgllZ, hence IIx11 2:5; II ex liz +11(1- e)xW. Let A;a. 0, /.L ;a. 0, E > O. There exist y, z E A such that Ilyll :5; I, IIzll:5; 1, and

Aile fll + /.L11(1- e) . fll :5;A(e . f)(y) + /.L(1- e) . /)(z) + E = f(Aey + /.L(l- e)z) + E :5; IlfllIIAey + /.L(1- ejz] + E ~ 11/11[lle(Aey + /.L(1- e)zW + 11(1- e)(Aey + /.L(1- e)z)lfl llz + E =II/II[/1AeyliZ + 1I/.L(1- e)zWl l12 + E :5; IIfll(A Z + /.L z)I/Z + E. Therefore, Aile fll+ /.L1I(l- e) fll :5;11/1I(A z + /.L 2) l/z. Taking A= lie fll /.L =11(1- e) . fll, we obtain (i).

and

12.2.4. THEOREM. Let A be a von Neumann algebra, and f an element of the predual of A. (i) There exists a pair (p, u) with the following properties: (a) p is a normal positive form on A, and Ilpll =Ilfll; (b) u is a partial isometry of A whose final projection is equal to the support of p (A 26); (c) f = u . p, p = u* . f. (ii) Let p be a normal positive form on A, and u a partial isometry of A whose final projection is dominated by the support of p, with f = u . p, p = u* . f. Then p = p, u = u.

We can suppose that II/II = 1. Let B be the ball {x : Ilxll ~ I} in A, and B the set of x E B such that f(x) = 1. Since B is ultra-weakly compact and f is ultra-weakly continuous, there exist x E B such that If(x)1 = 1. Multiplying x by a suitable scalar, we see that B,e. 0; furthermore, B is convex and ultra-weakly compact. There therefore exists an extreme point v in B. This point is also extreme in B, because if v = !(s + t) with

CH. 12. 2j

POLAR DECOMPOSITION OF A LiNEAR FORM

269

s, t E B we have
If(t)1 ~ 1

and 1 = f(v)

=~(f(s)

+ f(t,

whence f(s)

= f(t) = 1,

hence s, t E B, and s = t = v since v is extreme in B. By 12.2.2, v*v is a projection, and so v is a partial isometry. Put p = v . f. We have lipII ~ IIvllllfll = 1. Since p(1) = f(v) = 1, p is positive (2.1.9), of norm 1, and normal because ultra-weakly continuous. Let e be the support of p. We have p(v*v) = f(vv*v) = f(v) = I, and so v*v ~ e. Put u = ev*. Then uu* = e(v*v)e = e, hence u is a partial isometry with final projection e. For every x E A, we have
(u* f)(x)

= f(u*x) = f(vex) = p(ex) = p(x),

=

and hence p

u* . f. Put u*u

e, We have

lie . fll ~

1, and

(e . f)(u*)

whence lie" xEA, (u . p)(x) = p(ux) = f(u*ux)

= f(u*uu*) = f(u*) = f(ve) = p(e) = I, fll = 1 =IIfll, and so ef = f (12.2.3 (iij), Then, for every = f(ex) = (e f)(x) = f(x),

and hence f = u . p. We have therefore proved (i). 12.2.5. In order to prove (ii), we will first of all establish the following proposition:
PROPOSITION. Let A be a unital C*-algebra, f a continuous linear fonn on A, p a positive fonn on A, and a an element of A such that lIall~l, a p > ] , lipII =IIfll If bEA satisfies IIbll~l, b'f~O and lib . fll =Ilfll. then b . f =p.
r

We can suppose that lip II = IIfll

(~p

= 1. We have

l7Tp(b*a*)gp)

=(7Tp(ab)~p

I~p)
: f)(1)

= p(ab) = f(b) = tb

=lib fll = 1.

Since Ilb*a*1I~ 1, the converse of the Cauchy-Schwarz inequality then proves that 7Tp(b*a*)~p =~p. Hence, for every x E A,
(b f)(x)

= f(bx) = p(abx) = (7Tp(abx)~p I~p) = (7Tp(X)~p l7Tp(b*a*)~p) = (7Tp(X)~p Igp) = p(x),

whence b . f

= p.

270

THE ENVELOPING VON NEUMANN ALGEBRA OF A C-ALGEBRA

[CH. 12, 2

12.2.6. Let us return to the proof of 12.2.4. Let (p, u) be a pair having the properties of (ii). We have Ilu*II::s;1, u* . f;;:. 0, and IIplI = Ilu*. fll::s; IIfll = Ilu" pll::s;IIplI, hence lIu*. fll = IVII Then 12.2.5 (applied with a = u, b = u*) gives p = u* . f = p. We show that u = u, Let x = uu*. Then
p(x) = (u . p)(u*) = f(u*) = (u * f)(1) = p(1) = I,

hence p(x*) = I, 1= p(X)2::s; p(xx*)::s; I, hence p(xx*) = I, and

pe - x)(e - x)*)

= 1- 1- I + I = O.

Now exe = x since the final projections of u and u are dominated bye. Since p is faithful on eAe, we see that x = e. Let H be the Hilbert space in which A operates, and e 1 the initial projection of u. For ~ E e(H), we have lIu(u'*~)1I = lIe~1I =I ~I , and therefore lIu(u'*~)1I ;;:'lIu'*~II, hence u'*~ E e1(H); since u maps e1(H) isometrically onto e(H), the equality uu'*~ = ~ = uu*~ implies that 'u'*~ = u*~. Furthermore, u* and u* vanish on He e(H); hence u* = u*, u = u. 12.2.7.
DEFINITION.

absolute value of f, and is denoted by If I The equality f the polar decomposition of f.

With the notation of 12.2.4, p is said to be the =u . If I is called

12.2.8. DEFINITION. Let A be a C*-algebra, E the enveloping von Neumann algebra of A, and f a continuous linear form on A. The polar decomposition f = u . p of f, regarded as an element of the predual of E, is called the enveloping polar decomposition of f. Regarded as a positive form on A, p is said to be the absolute value of f and is denoted by Ifl.
If A is a von Neumann algebra and if f belongs to the predual of A, then f admits both a polar decomposition and an enveloping polar decomposition, in general distinct (12.5.7). Nevertheless, we shall see that the two possible definitions of If I coincide. Let f = u . p be the polar decomposition of f, and let p be identified with a normal positive form on E (12.1.3 (iii)). We have
u*EA~E,

Ilu*ll::s; I,

Ilpll=llfll

The equality p(x) = f(u*x), valid for x E A, extends by continuity to every x E E (we identify f with a weakly continuous linear form on E).

CH. 12. 3]

DECOMPOSITION OF AN HERMITIAN FORM

271

Then proposition 12.2.5, applied to E and to b absolute value of f in the sense of 12.2.8.

= u*, proves that p

is the

12.2.9. PROPOSITION. Let A be a C*-algebra, f a continuous linear form on A, and p a positive form on A. Then p = IfI if and only if we have lipII = IIfll and If(xW:o;;; Ilfllp(x*x) for every x E A. Let f = u . IfI be the enveloping polar decomposition of we have Ilpll = Ilfll, and, for every x E A,

f. If

p =

Ifl,

If(xW = Ip(uxW", p(u*u)p(x*x):o;;; Ilpllp(x*x) = Ilfllp(x*x).

Suppose, conversely, that every x E A. We have

lipII = IIfll, and

that If(xW"llfllp(x*x) for

If(xW "llfll(1Tp(X*X){p I{p) =

IlfllII1Tp(x){pW

for every x E A, and therefore for every x in the enveloping von Neumann algebra E of A (suppose that 1Tp is extended to a normal representation of E; x ~ II1Tp(X){pll is a continuous function for the ultra-strong topology, and A is ultra-strongly dense in E). Applying this inequality to u * x yields

p(x){pll2 Ilfl(x)12:o;;; 11f11II1Tp(u*x){pll2:o;;;Ilfllll1T

for every x E E. Hence there exists a g in the space of 1Tp such that II{II... IIfW /2 and Ifl(x) = (1T p(X)gp I{). Then ({p, {) = Ifl(1) = Ilfll = IIpW/211f111/2;:. II{pll . II{II, hence ~ = ~p and Ifl(x) = (1Tp(X)~p I~p) = p(x). References: [508], [837], [1401], [1404], [1682]. The proof of 12.2.3 given here was indicated to me by G. Zeller-Meier. 12.3. Decomposition of an hermitian form into positive and negative parts 12.3.1. PROPOSITION. Let A be a von Neumann algebra, f and f two normal positive fonns on A, e and e their supports, and g = f - f. Then (i) IIgil =Ilfll+ IIfII if and only if ee =O. (ii) Suppose that this is so. Then 19l = f +f, and the support of Igi is e + e; g =(e - e) Igi is the polar decomposition of g; we have f = e . g. and f = - e . g.

272 THE ENVELOPING VON NEUMANN ALGEBRA OF A C*-ALGEBRA

[CH. 12, 3

Suppose that ee = O. A projection d of A satisfies (j +1) (d) =0 if and only if I(d) =I(d) =0, i.e. de = de =O. Hence e + e is the support of 1+1. Moreover, e - e is an hermitian partial isometry, with e + e as initial and final projections. We have, for every x E A,
e - e) . (j + I))(x) =(j +I)(ex - ex)

= I(ex) -

I(ex) = I(x) - I(x) = g(x)

e - e)* . (j - I))(x) = (j - I)(ex - ex)

= [tex) + [(ex) = I(x) + I(x). All this proves (12.2.4 (ii) that Igl = I + I and that g = (e - e) . Igi is the polar decomposition of g. We have e . g = e . I =I, e . g =- e . I =- 1.
Lastly,

Ilgll = Ilf +Ill = Ilfll+ IlfllIIxil

Suppose that Ilgll = 11/11 + 11/11- The set B of x E A such that x = x*, :!$; 1 is ultra-weakly compact; also, g is ultra-weakly continuous. Hence there exists x E B such that g(x) =Ilgl/. We deduce from this that
I(x+) + I(x-) - I(x-) - I(x+) = g(x)

= IIgll =Ilfll+ IlfII

Now

whence

Let s, s be the supports of x" and

11/11

X-.

We have

=I(x+):!$; I(s):!$; 11/11

= O.
I and I

whence 1(1- s)

= 0,

S ~

e; similarly, s ~ e. Now ss = 0, and so ee

12.3.2. DEFINITION. If the conditions of 12.3.1 (i) are satisfied, are said to be mutually singular.

12.3.3. THEOREM. Let A be a von Neumann algebra, and g an ultra weakly continuous hermitian linear [orm on A. There exists a unique pair (j,I) 01 mutually singular normal positive [orms on A such that g =

i : 1.
Let P be the predual of A, A its dual, and A" its second dual, identified with its enveloping von Neumann algebra. By 2.6.4, there exist

CH. 12. 4]

THE POSITIVE PART OF AN IDEAL IN A C*-ALGEBRA

273

two positive forms t. I on A (identifiable with normal positive forms on A") such that g = f - 1,IIgll = Ilfll+ lit/II. Applying 12.3.1 to A", we see that Igi =f +1. Now Igi E P (12.2.8). Hence f E P, IE P from which it follows that f and I are normal. Suppose that g =t, - fi, with t.. t; normal, positive and mutually singular. Let e. e, e., e; be the supports of f, 1, t.. fi. By 12.3.1 (ii), we have

I, +t; = 19l

=f +1,

whence

f =t.. I =J;.

12.3.4. COROLLARY. Let A be a C*-algebra and g a continuous her mitian linear form on A. There exists a unique pair (f, f) of positive forms on A such that g = f - 1, Ilgll = Iltll+ 11111 The existence is known already. The uniqueness follows from 12.3.3 and 12.1.3 (iii). 12.3.5. The forms f and I of 12.3.4 are called the positive part and negative part of g, and are denoted by s and g", We have Igl = s + g by 12.3.1. 12.3.6. Let A be a C*-algebra, E the enveloping von Neumann algebra of A, and f a positive form on A. Then f, regarded as a normal positive form on E, admits a support e which is a projection of E. It is called the enveloping support of f. (If A is a von Neumann algebra and if f is normal, f admits both a support and an enveloping support which are distinct in general (12.5.7).) If f and I are two positive forms on the C*-algebra A, it comes to the same thing to say either that Ilf- Ill = Ilfll+ litll. or to say that the enveloping supports of f and I are or thogonal (12.3.1 (i)); we then say that f and I are mutually singular, which generalises definition 12.3.2. References: [681], [848], [1404], [1589], [1590], [1591]. 12.4. The positive part of an ideal in a C*-algebra 12.4.1. PROPOSITION. Let A be a C*-algebra, g the set of closed left ideals of A, and ,1 the set of closed subsets J of A + which possess the following properties: (i) x, y E J => x + Y E J; (ii) x E J, yEA +, Y ~ x => y E J. Then the mapping I ~ r = I n A + is a bijective mapping of g onto ,1.

274

THE ENVELOPING VON NEUMANN

ALGEBRA

OF A C*-ALGEBRA

[CR. 12, 4

We observe, to start with, that, if [E.f; and x E I", we have X I/ 2 E I", for t n t is a sub-C*-algebra of A, and r k [ n I. We observe next, that, if I E J and x E I, we have x 1/2 E I, because there exists a sequence of non-negative continuous functions t, :Spx ~ R, converging uniformly to the function t ~ Vt, and vanishing in a neighbourhood of O. There then exist constants An;;:: 0 such that fn(t) ~ Ant on Spx. Whence we have O~fn(x)~Anx, fn(x)EI and x l / 2EI. We interrupt the proof to prove the following lemma:
12.4.2. LEMMA. Regard A as embedded in its enveloping von Neumann algebra E. Let I E,j, and let El" be the set of products ax(a E E, x E r). Then t = (Er) n A.

Let yE1. We have y=u Iy! for some uEE. Now IYI=(y*y)I/2 E r by an earlier remark. Hence y E El", whence I k (Er) n A. Conversely, let a E E, x E r, and suppose that ax E A. We have to prove that ax E 1. Now a is in the u(E, A)-closure of A, and so ax is in the u(A, A)-closure of Ax, and Ax k 1. Since I is a norm-closed subspace of A, it is u(A, A)-closed. Hence ax E 1.
12.4.3. We return now to the proof of 12.4.1. Let [E,j. If x, Y E r, we have x + y E r, Let z E A + be such that z ~ x. There exists sEE such that Zl/2 = sx" (A 8); since X 1/2 E by an earlier remark, we have z = (sx 1/2)*(SX 1/2) = (X I /2S*S)X I /2 E hence z E t: by 12.4.2. Hence I" E J. The mapping I ~ of.f; into J is injective by 12.4.2. We show that it I. Let is surjective. Let IE J and put [= (EI) n A. Obviously, a E E, x E J be such that ax E A +; we show that ax E I. We have

sr,

r:

(ax)2 = xa*ax ~ IIaWx 2 ~ IlaWllxllx,

hence (ax)2 E I, and so ax E I by an earlier remark. We have therefore proved that t: = J. We show that I is a left ideal of A. It is clear that AI k 1. Let x, y E 1. Then x*x, y*y E I" = I, and so x*x + y*y E I; since
(x

+ y)*(x + y) ~ 2(x*x + y*y), + y)*(x + y) E I,

hence [x + yl E I

we deduce from this that

Ix + yl2 = (x

by an earlier remark; now x + y = ulx + yj for some u E E, and so

x + Y E (EI) n A = 1. We show, lastly, that I is closed. If (x n ) is a

CH. 12. 5j

ADDENDA

275

sequence of points of I converging to x E A, then x~xn E t: = 1 con verges to x*x, whence Ixl2 = x*x E I, [x] E I, and x E (El) n A = I. 12.4.4. Proposition 12.4.1 still holds if left ideals are replaced by right ideals in it. Furthermore, if I is a left ideal of A, then l* is a right ideal and t: = (I*t. Reference: [508].

12.5. Addenda

12.5.1. Let A be a C*-algebra. In the dual A of A, let (fa) be a family, indexed by a directed set, weak*-converging to f. If Ilfall-lIfll,then If aI weak*-converges to IfI [508]. 12.5.2. Let H be a Hilbert space, A a von Neumann algebra on H, and f an ultra-weakly continuous linear from on A. There exists an ultra weakly continuous linear form g on 5(H), extending f, and such that Ilfll = IlglI (If f ~ 0, there exist gj E H such that L IIgjl12 < +00 and such that f(x) = L (xgj Igj) for every x E A. We put g(x) = L (xgj I0 for every x E 5(H), whence Ilgll = g(l) = f(l) = Ilfll. In the general case, use the polar decomposition of f). [1401], [1404]. 12.5.3. Let A be a C*-algebra, E the enveloping von Neumann algebra of A, and f and g two pure states of A. Then 7Tf = 7Tg if and only if the supports of f and g, regarded as normal positive forms on E, are equivalent relative to E. [1695]. *12.5.4. Let A be a C*-algebra. The dual of A is separable for the norm topology, if and only if A admits a countable composition sequence (Ip) such that the quotients Ip+tllp are separable dual C*-algebras. [1695]. rcr. also, Math. Reviews, 1963, p. 808.) 12.5.5. A factor, regarded as a Banach space, is the second dual of a C*-algebra if and only if it is of type I. [681], [1591]. 12.5.6. Let A be a von Neumann algebra, A" its enveloping von Neumann algebra, P the predual of A, and f an element of P identified with a linear form on A". If x E A", we have x . f E P and f . x E P. (Let

276

THE ENVELOPING VON NEUMANN

ALGEBRA

OF A C*-ALGEBRA

[CH. 12. 5

p be the canonical projection of A" onto A; if yEA, we have, since f vanishes on Ker p, f(xy)

= f(p(xy = f(p(x)y),
E P.) (Grothendieck, unpublished.)

and therefore x . f

= p(x) .

12.5.7. Let H be a Hilbert space, (e., E2, . . ) an orthonormal basis of H, A the von Neumann algebra on H generated by the Pi = P CE" and f the positive linear form ~ AiPj ~ ~ i- 2A i. This form is normal and has support 1. Let g be the positive linear form I AiP i ~ lim Ai (take a generalised Banach limit). Regard f and g as normal positive forms on the en veloping algebra E of A. Let Qn = P; + P n +1 +... E A ~ E. The Qn have a strong limit Q in E which is a projection of E; we have g(Qn) = 1 for every n, hence g(Q) = 1 and Q,t O. Moreover, f(Qn)~O and hence f( Q) = O. Hence the support of f relative to E is ,t 1. The polar decomposition of f therefore differs from its enveloping polar decomposition. 12.5.8. Let A be a von Neumann algebra, f and t two mutually singular normal positive forms on A, and g = f - [, We have

Ilfll=

sup
XEA+,lIxll"l

g(x),

Iltll=

sup
xEA+,lIxH""

- g(x).

(If x E A + and Ilxll ~ 1, we have g(x) ~ f(x) ~ Ilfll; denoting the support of f bye, we have, moreover, gee) =fee) = IIfll.)

12.5.9. Let A be a von Neumann algebra. Let E be the set of weakly closed left ideals of A. Let E be the set of weakly closed subsets P of A + such that: (1) x, yEP:::;> x + yEP; (2) x E P and A ~ 0:::;>Ax E P ; (3) x E P, yEA + and y ~ x :::;> yEP. Then L ~ L n A + is a bijection of E onto E. [508]. 12.5.10. Let A be a C*-algebra. Let E be the set of closed two-sided ideals of A. Let E be the set of closed subsets P of A + such that: (1) x, yEP:::;> x + yEP; (2) x E P and A ~ 0:::;>Ax E P; (3) x E P and yEA:::;> y*xy E P. Then I ~ I n A + is a bijection of E onto E. [508]. 12.5.11. Let A be a C*-algebra, B its enveloping von Neumann algebra, the canonical morphism. Define A b B., 1T1 similarly. Let and 1T:A~B P: A ~ Al be a morphism. Then there exists a unique normal morphism l/J: B ~ B I such that 1T I P = l/J 1T. If we identify Band B I with A" and A!, then l/J is the second adjoint of P.
0 0

CHAPTER 13

UNITARY REPRESENTATIONS OF LOCALLY COMPACT GROUPS

13.1. Elementary definitions concerning representations 13.1.1. DEFINITION. Let G be a topological group and H a Hilbert space. A continuous unitary representation of G in H is a morphism of the group G into the unitary group of H which is continuous for the strong topology. In other words, a continuous unitary representation of G in H is a mapping 71 of G into the unitary group of H such that 71(st) = 71(s)71(t) and such that, for every ~ E H, the mapping s ~ 71'(S)~ of G into H is continuous (for the norm topology of H). The condition 71(St) = 71(S)71(t) implies that 71(e) = 1 (the neutral element of the group being studied will always be denoted bye) and 71(S-I)= 71(S)-1 = 71(S)*. The functions s .... qlE.1(s) = (71'(S)~ ITJ) on G (~, TJ being.fixed elements of H) are called the coefficients of 71. We have qlE,1(s) = cP'I~(S-I). 13.1.2. The strong and weak topologies coincide on the unitary group of H. In the above, we can therefore replace the strong topology by the weak topology throughout. On the other hand, if we were to stipulate continuity for the norm topology, we would obtain a more restrictive definition, of no interest for what follows. (Of course, if dim H < +00, all these notions of continuity are the same). 13.1.3. The Hilbert dimension of H is called the dimension of 71 and is denoted by dim 71. The space H is called the space of 71, and is denoted by H1/" By analogy with the case of involutive algebras, the following notions are also easily defined: equivalent representations, class of represen tations, intertwining operator, intertwining number, direct sum of re presentations, multiple of a representation, subrepresentation, represen tation contained in a representation, cyclic vector (~ is said to be a cyclic

280

UNITARY

REPRESENTATIONS

OF LOCALLY

COMPACT GROUPS

[CH. 13. !

vector for TT if 'TT(G)~ is total in H 7T ) . We employ the same notation as in the algebra case, and we have the same elementary properties; we leave it to the reader to convince himself of this. There is no need to define here the essential subspace of a represen tation. Everything proceeds as if continuous unitary representations were automatically non-degenerate. 13.1.4. Let 7T be a continuous unitary representation of G. The following conditions are equivalent: (i) the only closed subspaces of H 7T invariant under TT(G) are 0 and H; (ii) the commutant of TT(G) in 5t(H7T ) is just the scalar operators; (iii) every non-zero vector of H 7T is a cyclic vector for TT. If these conditions are satisfied, and, in addition, H 7T ,e. 0, then TT is said to be topologically irreducible or simply irreducible. (We will never encounter the notion of algebraic irreducibility, except when dim TT < +00, in which case it is equivalent to topological irreducibility). We denote by G the set of classes of irreducible representations of G. Just as in Chapter 5, we define the terms disjoint, factor, quasi equivalent, of type I, II, ... , multiplicity-free, of multiplicity n as applied to representations. All the arguments and all the results of Chapter 5 extend immediately to group representations. (We can also make use of the correspondence between representations of G and representations of L I( G) which will be established in 13.3, at least if g is locally compact). 13.1.5. We are now going to define, for group representations, opera tions which have no meaning for representations of involutive algebras. Let TT, TT be continuous unitary representations of G. For every s E G, we form, in the Hilbert tensor product H 7T H 7T" the (unitary) operator TT(s) TT(s). It is immediate that s ~ TT(s) TT(s) is a continuous unitary representation of G in H 7T H 7T" called the tensor product of TT and TT, and denoted by TT sr, Let 7r be a continuous unitary representation of G in H, and let ii be the Hilbert space conjugate to H. Each TT(s) is a unitary operator in H, and it is immediate that s ~ 7r(s) is also a continuous unitary represen tation of G in if, called the conjugate representation of TT, and denoted by iT. 13.1.6. Let G be a locally compact group endowed with a left Haar measure ds. For every s E G, let \(s) be the operator in L\G) defined by (\(s)f)(x) = !(S-I X) (f E L 2(G), x E G).

CH. 13. I]

ELEMEN1ARY

DEFINITIONS CONCERNING REPRESENTATIONS

281

We immediately verify that A is a continuous unitary representation of G in L 2(G), called the left regular representation. Let ..1 : G ~ R be the modular function of G. For every s E G, let p(s) be the operator in L 2(G) defined by (p(s)f)(x) = ..1 (s)1/2f(xs) (f E L 2 (G ), x E G). We immediately verify that p is a continuous unitary representation of G in L\G), called the right regular representation. The representations A and p are injective. For every f E L 2(G ), define f E L 2 (G ) by f(x) = ..1 (X)-1/2f(x- 1). Then f ~ f is an isomorphism of the Hilbert space e(G) onto itself, and we have, for every s E G,
(A (s)f),(x)

=..1 (X)-/2f(s-IX-I) = ..1 (S)I12..1(XS)-/2fxS)-I) = (p(s)f)(x).

Hence the isomorphism f ~ f transforms A into p, from which it follows that A = p. We thus speak sometimes of the "regular representation" of G, without specifying left or right. For every f E L 2(G), let 1 E L 2(G) be the complex conjugate function. Then f ~ 1 is an isomorphism of the Hilbert space L 2(G) onto the conjugate Hilbert space which transforms A into A. Hence A = A and similarly p = p. 13.1.7. Let a be a cardinal and H a Hilbert space of (Hilbert) dimension a. The representation s ~ I of G in H is called the trivial representation of G in H, or the trivial representation of G of dimension a. It is denoted by I H , or just by 1 when there is no uncertainty as to what His. 13.1.8. PROPOSITION. Let G 1, G 2 be topological groups, G = G, X G 2, and a an irreducible continuous unitary representation of G. Suppose that every continuous unitary factor representation of G j is of type 1. Then there exist, for i = 1,2, an irreducible continuous unitary represen tation tr, of G, such that U is equivalent to the representation

(Sl>

s2)~uj(sj)0u2(S2)

in the Hilbert space H U 1 0 H lT2

Let .st1j be the von Neumann algebra generated by u(G;). Then .st1 j,.st1 2 commute with each other and generate the von Neumann algebra :(HO). Every element of the centre of .st1 j commutes with .st1 j and with .st1 2 , hence with :(HO), and is therefore a scalar operator. Hence .st1 1 is a factor, of type I by the hypothesis made about G j There then exist Hilbert spaces

282

UNITARY

REPRESENTATIONS

OF LOCALLY

COMPACT GROUPS

[eH. 13, 2

HI>H 2 such that H; may be identified with HI H 2 and s1 1 with .P(H1)C (A 36). For every SI E G I , u(St) may be written UI(SI) 1, where UI(SI) is a unitary operator on HI> and 001 is a continuous unitary representation of G. in HI The UI(SI) generate the von Neumann algebra .P(HI ) , and so 001 is irreducible. We have s12 ~ C .P(H2) (A 18), hence, for every S2 E G 2, U(S2) may be written I U2(S2), where 002 is a con tinuous unitary representation of G 2 in H 2 If T E .P(H~ commutes with the UiS2), then 1 T commutes with s11 and s12, and is therefore a

scalar operator; hence T is a scalar operator and 002 is irreducible. References: [995], [1108].

13.2. The involutive algebra L1(G) 13.2.1. Let G be a locally compact group and M1(G) the algebra (under convolution) of bounded complex measures on G. This is a Banach algebra, having as identity the Dirac measure E. at the point e. If, for every f.t E M1(G), we define f.t * by the relation du. *(s) = df.t(s I), it can be checked that M I ( G) becomes an involutive algebra, and that 1If.t *11 = 11f.t11. Hence MI(G) is an involutive Banach algebra. 13.2.2. Choose, once and for all, a left Haar measure ds on G. If, with every f E L1(G), we associate the measure df.t(s) =f(s) ds E MI(G), we obtain an isometric morphism <P of the Banach algebra L I( G) into the Banach algebra MI(G). For g E X(G) (the set of continuous complex valued functions on G with compact support), we have, denoting the modular function of G by ..::1,

g(s)

dis. *(s) = g(S-I) df.t(s)

g(s-I)f(s) ds

g(S)!(S-I)..::1(S-I) ds.

For every complex-valued function f on G, define f* by

f*(s)

The above then shows that Hence f ~ f* is an isometric involution on L and <P is a morphism of the involutive algebra LI(G) into the involutive algebra M(G). We identify Lt(G) with <P(LI(G)). In general, LI(G) is not a C*-algebra.

=!(S-I)..::1(S-I). du. *(s) =f*(s) ds.

13.2.3. For every complex-valued function f on G, we define

and f by

CH. 13. 3]

REPRESENTATIONS

OF

AND REPRESENTATIONS

OF

L(O)

283

I(s)

= f(S-I),

l(s)

= f(s-I). We thus have f* = 1,1-1. If a E G, put

J(s)

=f(as),
f

fa(s)

=f(sa).
Ea ) :

We have (denoting the Dirac measure at the point a by

Ea

* f = a-If,
=f*
*
E:

* E a = J.(a-J)fa- "
gives

The formula (Ea * f)*

=f* * Ea-I then (a-.j)* = J.(a)(f*)a

13.2.4. If G is discrete, then L 1(G) has an identity. If G is separable, then LJ(G) is separable. 13.2.5. Let s E G. Let I be the family of neighbourhoods of s, ordered by reverse inclusion. For every i E I, let u, be a positive function of 1(G), L vanishing on G\i, with integral equal to 1. We have, for every
fEL (G ),
1

f E J{(G), and this is then immediate by uniform continuity.] Applying this to s = e, we see that L1(G) possesses an approximate identity.
[It is enough to verify this for

References: [995], [1108]. 13.3. Representations of G and representations of L 1(G). 13.3.1. PROPOSITION. Let 7T be a continuous unitary representation of G. 1 For every IL E M ( G ), PUt 7T (IL) = f 7T(S) dIL(s) E I(H1r ) . Thenc u. ~ 7T(IL) is a representation of the involutive algebra M 1( G) in H 1r , whose restriction to L1(G) is non-degenerate. The fact that IL ~
7T(/L)

is a representation of the involutive algebra

M1(G) follows from easy calculations; for example, for IL E M1(G), ~ E H 1r , 7] E H1f) we have
(7] 17T(IL

*)~) = (7T(IL *)~

=

17]) =

(7T(S

i)~

I 1/)dIL(s)

(7T(S)1/

I~)

dIL(s)

= (7T(IL)7] I~) = (7] I 7T(IL)*~),

~

whence 7T(IL *) = 7T(IL)*. Moreover, let s E G,

EH

1r

and

> O. With the

284

UNITARY

REPRESENTATIONS

OF LOCALLY

COMPACT GROUPS

[CH. 13, 3

notation of 13.2.5, there exists an i E I such that 1I1T(s')~ s E i, whence 1I1T(u;)~ - 1T(s)~11 ~ E. Hence
(1)
1T(U;)~

-1T(s)~11

for

1T(S)

for the strong operator topology. Applying this to s = e, we see that the representation 1T of L I(G) is non-degenerate. 13.3.2. The representations of M1(G) and LI(G) obtained above are said to be associated with the given representation of G. 13.3.3. For s E G, we have 1T(Es ) = 1T(S). Hence, if f E L1(G),
1T(J) = 1T(Ea- 1T(fa) = 1T(L1(a-l )f

* f) = 1T(a- I)1T(f),

* Ea-I) = L1(a- I)1T(f)1T(a- 1).

13.3.4. PROPOSITION. Let H be a Hilbert space, and 1T a non-degenerate representation of the involutive algebra L I( G) in H. Then 1T is associated with exactly one continuous unitary representation of G. The uniqueness follows from formula (I) of 13.3.1. We prove the existence. Let H be the subspace of H generated by the 1T(f)H. By hypothesis, H is dense in H. Let s E G. With the notation of 13.2.5, we have
Ilu; * f

e,

* III~

0,

hence II 1T (u;) 1T(f) - 1T(Es

* f)11 ~ o.

Hence 1T(Uj) converges, for the topology of pointwise convergence over H, to an operator 1T(S) on H, such that
(1)

1T(S)1T(f) = 1T(Es

* f).

Since 111T(Uj)II:os;; Ilu;II = 1, 1T(S) has a unique extension to a continuous linear operator on H, which we again denote by 1T(S), such that 111T(S)II:os;; l. For s, t E G and f E L1(G) we have, by (I),
1T(St)1T(f)

= 1T(Est * f) = 1T(E. * e, * f) = 7T(S)7T(Et * f) =

7T(S)7T(t)7T(f)

and so 7T(St) = 7T(S)1T(t) on H and consequently on H. Equality (I) shows that 7T(e) = I, and that 1T(S)~ depends continuously on S for ~ E H, and therefore for ~ E H. Finally, since 7T(S) and 1T(S-I) = 7T(S)-1 are both norm-reducing, 7T(S) is unitary, which proves that 7T is a continuous unitary representation of G. We now show that the associated representation of L1(G) is just the

CH. 13. 3j

REPRESENTATIONS

OF

AND REPRESENTATIONS

OF L(G)

285

original representation with which we started. Let f, g E L1(G). Since

f

*g =

f(S)(E s

* g) ds * g) ds

in L1(G), we have
7T(f)7T(g) =7T(f

* g) =

f(S)7T(Es

f(s )7T(S )7T(g) ds =

(J f(s )7T(S) dS) 7T(g),

hence 7T(f) = f f(s )7T(S) ds. 13.3.5. Propositions 13.3.1 and 13.3.4 establish a bijective correspon dence between the continuous unitary representations of G and the non-degenerate representations of L1(G). Let 7T be a continuous unitary representation of G and 7T the associated representation of L I( G). The closed subspaces invariant under 7T(G) and 7T(L1(G are the same and the cyclic vectors for 7T( G) and 7T(LI( G are the same (in fact, by formula (1) of 13.3.1). Hence 7T(G) and 7T(L1(G have the same com mutant and generate the same von Neumann algebra. In particular, the topological irreducibility of 7T is equivalent to that of 7T. We have thus established a bijective correspondence between the irreducible con tinuous unitary representations of G and the topologically irreducible representations of L1(G). Similarly, to say that 7T is a factor represen tation, is of type I, ... is equivalent to saying that 7T is a factor representation, is of type 1, .... If 7Tl,7T2 are continuous unitary re presentations of G and 7T;, 7T~ the associated representations of L l( G), then the intertwining operators for 7Tl and 7T2 are the same as those for 7T; and 7T~, we have 7T; 7T~ = (7T1 EB 7TJ etc. Henceforth, we will denote by the same letter a continuous unitary representation of G and the associated representation of L I( G).

13.3.6. Let A be the left regular representation of G in L 2(G). If fEL1(G), then A(f) is the operator g~f * gin L\G). This represen tation A of L1(G) is called the left regular representation of L1(G) in L 2(G). If A(f) = 0, we have f * g = 0 for every g E J(G), and hence f = 0 by 13.2.5. Hence the left regular representation of L1(G) in L 2(G) is injective. If G is commutative, and if (] denotes the dual group of G, the Plancherel isomorphism of L 2(G) onto L 2(] ) transforms A(f) into the

286

UNITARY

REPRESENTATIONS

OF LOCALLY

COMPACT GROUPS

[CH. 13, 4

operator of multiplication by :Jif (the Fourier transform of f) in L 2(G ). We therefore have II,\.(/)11 = sup 1(:Jif)(t)I
IE6

In general, f If(s)1 dss sup l:Jifl, and so ,\. is not isometric (from which it follows that L(G) is not a C*-algebra). References: [995], [1108]. 13.4. Positive forms on L I( G) and positive-definite functions 13.4.1. Section 13.3 leads us to the study of the representations of
L1(G), or, which amounts to the same thing, the continuous positive forms on L1(G). Now a continuous linear form on L1(G) is defined by an

element of L OO( G). This leads to the introduction of certain bounded functions on G.
DEFINITION. A continuous complex-valued function qJ on G is said to be positive-definite if, for any elements Sl>"" s; of G, the matrix l (qJ(si S)h.;;ijE;;n is positive hermitian.

In other words, for any

(1)

SI""
n

s; E G and
1

al," , an

E C, we have

~
j,j=\

ajOj<p(si

Sj);;: O.

13.4.2. The sum of two positive-definite continuous functions is posi tive-definite. If cp is positive-definite continuous and if ,\. ;;:. 0, then ,\.cp is positive-definite. For examples of positive-definite continuous functions, d. 13.6.3. 13.4.3. In (I), put n

=2,

s\

=e,

S2

=s E

G. The matrix

must be positive hermitian. This implies that

qJ(S-l) = cp(s)

IqJ(s)1 ~ qJ(e)

for every s E G. In particular, qJ is bounded, and IlqJlloo = qJ(e). 13.4.4. PROPOSITION. Let qJ be a continuous complex-valued function on G. Then the following conditions are equivalent:

CH. 13. 41

POSITIVE FORMS ON

L(G) AND POSITIVE-DEFINITE FUNCTIONS

287

(i) q; is positive-definite; (ii) q; is bounded and, for every bounded complex-valued on G, we have (qJ, p, * * p,);; 0, in other words

measure p,

ff ff
al""

q;(y-I Z ) dp,(y) dp,(z);;, 0;

(iii) for every f E X( G), we have (q;, f* * t: 0, ill other words

q;(y-Iz)f<y)f(z) dy dz

O.

(ii) ~ (i): Suppose condition (ii) is satisfied. Let Sl"" s; E G, an E C, and p, be the measure on G defined by masses a, at the points Si We have
0.:::;

ff

q;(y-I Z ) dp,(y) dp,(z) = itl a,-CljP(si ls),

and so q; is positive-definite. (i) ~ (iii): Suppose q; is positive definite. Let f E X( G). The function on G x G is continuous and of compact support, S; S is contained in a set K x K, for K some compact subset of G. The measure induced on K by the Haar measure is the weak*-limit of positive measures Vi of finite support which are norm-bounded; hence the measure induced on K x K by the Haar measure of G x G is the weak*-limit of the Vi Vi- Now, if Vi is defined by masses (31" ~n at the points Slo __ ,Sn we have
(y, z)~q;(y-Iz)f(y)!(z)

ff

qJ(y-Iz)f(y)f(z) dVj(Y) dVi(Z) =

q;(si ISj)f(Sj)!(S)(3i(3j;; 0

and, in the limit,

ff

qJ(y-lz)f(y)f(z) dy dz

O.

(iii) ~ (ii): Suppose condition (iii) is satisfied. Let p, be a complex valued measure on G with compact support. For every! E X(G), p, * f is an element of X(G), hence
O:::;(q;,!* * p,*

fff J

* JL * f)

q;(xyzl)!*(x)!(t) dxdtdp,*(y) dp,(z)

JJ

dp, *(y) dp,(z)

JJ

q;(xyzt)!*(x)!(t) dx dt.

288

UNITARY

REPRESENTATIONS

OF LOCALLY

COMPACT GROUPS

[CH. 13, 4

Taking f ~ 0, with integral equal to 1, and with support contained in a smaller and smaller neighbourhood of e, f f P(xyzt)f*(x)fCt) dx dt con verges to P(Yz) uniformly over every compact subset of G x G. We thus find that, in the limit, (P, JL * * JL) ~ O. Reasoning as in (ii) ~ (i), we at once see that P is positive-definite, and therefore bounded. Now let JL be an arbitrary bounded measure on G. Then JL is the norm-limit of measures with compact support v. Since P is bounded, (P, JL * * JL) is the limit of (P, v* * v), and is therefore ~ 0, by the above. 13.4.5. THEOREM. (i) Let P E L roc G), and w be the continuous linear form on L1(G) defined by P. Then ta is positive if and only ifP is equal, locally almost everywhere, to a continuous positive-definite function. (ii) A complex-valued function I/J on G is continuous positive-definite if and only if there exists a continuous unitary representation 7T of G and an ~ E H (which we can suppose is a cyclic vector for 7T) such that
1r

I/J(s)

=(7T(S)~

I~).

We then have IIPII",= (~ I~). (iii) Let 7T and 7T be continuous unitary representations of G, and ~ (resp. f) a cyclic vector for 7T (resp. 7T). If (7T(S){ I~) = (7T(S){ If) for any s E G, there exists an isomorphism of H onto H which transforms 7T into 7T and { into f.
1r 1r ,

Suppose that P is equal, locally almost everywhere, to a continuous positive-definite function. Then, for every f E L I( G), we have

JP(s)(f* * f)(s) ds ~ 0
by 13.4.4. Hence eo is a positive form on LI(G). Conversely, suppose that w ~ O. Form the representation 7T., of L1(G) and the vector ~.,. By 13.3.4, 7T., is associated with a continuous unitary representation of G in the space of 7T." which we again denote by 7T.,. Then, for any f E L1(G), we have

JP(s)f(s) ds = w(f) =(7T, !(f)~.,

and so
(1)

I~.,)

J(7T.,(S)~.,

I{.,)f(s) ds,

P(s) = (7T.,(S)~w

I~.,)

locally almost everywhere. If P is continuous, equality (1) holds every-

CH. 13. 4]

POSITIVE FORMS ON L(G) AND POSITIVE-DEFINITE FUNCTIONS

289

where, since the right-hand side is a continuous function of s. To accomplish the proof of (i) and (ii), it is enough to prove that, if 1T is a E H", then s ~ continuous unitary representation of G, and if (1T(S)e Ie) is positive-definite. Now, if /-L E M 1(G), we have

o~ 1I1T(/-L)eIl 2=(1T(/-L * * /-L)e Ie) =

(1T(S)e

Ie)

d(/-L * * /-L)(s),

whence our assertion, by 13.4.4. We now adopt the hypothesis and notation of (iii). We have

Hence there exists an isomorphism tP of H" onto H", which transforms into g, and 1T(f) into 1T(f) for every f E L1(G) (2.4.1 (iij). By 13.3.4, tP transforms 1T(S) into 1T(S) for every s E G.

13.4.6. Let 1T be a continuous unitary representation of G, and let We say that the function s ~ (1T(S)e Ie) is the positive-definite function defined by 1T and For 1T fixed and varying in H", we obtain the positive-definite functions associated with 1T. If S is a set of representations of G, the positive-definite functions associated with S are just the positive-definite functions associated with the various ele ments of S. Let <p be a continuous positive-definite function on G. It defines a positive form co on L 1(G), and therefore a pair (1T w, ew), where 1Tw can be regarded as a continuous unitary representation of G. This pair is also denoted by (7Ttp etp) and is said to be defined by <p. It is characterised, up to isomorphism, by the fact that <p(s) = (7Ttp(S)etp Ietp) for every s E G and that ~tp is a cyclic vector for 1Ttp

eE H".

e.

13.4.7. PROPOSITION. Let <p be a continuous on G. If s, t E G, we have

positive-definite function

In fact, let

7T =

7Ttp

e = etp

We have

l<p(s) - <p(tW = 11T(S) -1T(te

Ie)1 2~ 11~112117T(S)~ -1T(t)~112 2 = <p(e)(II1T(s )~1I2 + 117T(t)eI1 - 2 Re( 1T(S)e I7T(t)e = <p(e)(2I1eW-2Re(7T(s-lt)e Ie = 2<p(e)(<p(e)-

Re <p(s-It.

290

UNITARY

REPRESENTATIONS

OF LOCALLY

COMPACT GROUPS

[CH. 13, 4

13.4.8.

PROPOSITION. Every translate of a continuous positive-definite function is a linear combination of four continuous positive-definite functions.
7T

Let ~ be a continuous positive-definite function on G, a, bEG. Put 7T(a)~ = 1/, 7T(b)-t~ = t. Then
4~(bsa)
= 4(7T(bsa)~

= 7T"" ~ = ~""

I~)

4(7T(S)1/

1()
()

= (7T(S)(1/ +() 11/ +() - (7T(S)(1/ - () 11/ .

+ i(7T(S)( 1/ + in 11/ + in - i(7T(S)( 1/ - i() 11/ - i().

13.4.9.

be two continuous positive-definite functions on G. Then ~~' is positive-definite. More precisely, let 7T = 7T",. ~ = ~<p, 7T = 7T.", ~' = ~<p" Then
PROPOSITION.

Let

~,~'

~(s)~'(s)

7T 0 7T')(S)(~
f) = (7T(S)~
= (7T(S)~

00 I~ 0 0
7T(S)g I~

g).

In fact,

7T 0 7T')(S)(~

f) I~

g)

I~)(7T'(S)g'

Ig) = ~(s)~'(s).

13.4.10. PROPOSITION. Let ~ be a continuous positive-definite function on G, and 7T = 7T",. Then every continuous positive-definite function associated with 7T is the uniform limit over G of functions of the form

where, St, . . . Sn E G, AI> .. , An E C.

Let ~ = ~"" 1/ E H 1T, and At, ... ,An E C such that
E

> O. There exist

St, ... ,

s; E G

and

Then, for every

G, we have

Now the left-hand side may be written

CH. 13. 5j

WEAK"-CONVERGENCE

AND COMPACT CONVERGENCE

291

13.4.11. Let we have

fE

L 2(G). If A denotes the left regular representation of G,

(A(s)f If) =

f(S-lt)f(t) dt =

f(t)f(t-IS) dt = (f

* f)(s).

Hence f * f is a continuous positive-definite function associated with A. If f E X( G), then f * f E X( G); passing to the limit, if f E L 2(G), then f * f is the uniform limit of functions of X(G), i.e, it vanishes at infinity. References: [620], [635], [818], [HOI], [1454], [1455]. 13.5. Weak*-convergence and compact convergence of continuous positive-definite functions
13.5.1. LEMMA. Let A be a bounded set in L""(G). Let f E L1(G). If cP E A weak*-converges to CPo E A, then f * cP converges to f * Po for the

topology of compact convergence.

In fact,
(f

* cP )(s) =

Jf(t)cp(r ls) dt = Jf(st)cp(t-

dt = (q,,.f).

When cp weak*-converges to CPo while remaining in A, (cp, g) converges to (CPo, g) uniformly over every compact subset of LI(G). Now, when s runs through a compact subset of G, the set of the J is norm-compact in L1(G), since the mapping s ~ J of G into Ll(G) is norm-continuous. 13.5.2. THEOREM. Let PI be the set of continuous positive-definite func tions cp on G such that cp(e) = I. On PI the weak*-topology u(L""(G), LI(G coincides with the topology of compact convergence. Since IIcpll"" = I for every cp E PI it is clear that, on PI the topology of compact convergence is finer than the weak*-topology. Now let PoE PI K be a compact subset of G, and E > O. We are going to prove that, if cp E PI is in a suitable weak*-neighbourhood of CPo, we have Icp(s) - CPo(s)l:s;; e + 4vE for every s E K. This will finish off the proof. There exists a compact neighbourhood V of e in G such that

Icpo(s) -II = lcpo(s) - Po(e)!:s;; E

for every s E V.

Let X be the characteristic function of V, and a > 0 the measure of V. Let V be the weak*-neighbourhood of CPo in P I defined by the condition

292

UNITARY

REPRE~ENTATIONS

OF LOCALLY

COMPACT GROUPS

[CR. 13, 6

I(lp - lpo, X)! :!S; ea, i.e. by the condition

IJ
v

(lp(s) -lpo(s d+;;;ea.

For

tp

E V, we have
(I - <p(s dsl

IJ

IJ

(I - <Po(s dsj +

IJ

(lp(s) -lpo(s dsl:!S; 2ea.

v v v Moreover, for lp E V and s E G, we have

I(a-Ix

* lp)(S)-lp(s)I=la-f X(t)lp(t-IS)dt-lp(S)/
=

a-I f lp(t-IS) dt - a-I

J
v

lp(S) dtl

:!S; a-I f IIp(t-I s) -lp(s)1 dt.

In view of 13.4.7, this is dominated by

a-I

f
v

y2(I - Re lp(tI/2 dt ~ y2a- 1

(J
v

(I - Re lp(t dt

/2 (J 1 . )1
v

dt

)1/2

:!S; y2a-

Iy(2ea)ya

= 2ye.

Now, there exists (13.5.1) a weak*-neighbourhood V of lpo in PI such that <p E V implies
!(a-Ix * lpo)(s)-(a-x * lp)(S)!:!S;E

for every sEK.

Then, for lp E V n V, we have

IIp(s)-lpo(s)l:s=e+4YE

for every sEK.

References: [1309], [1867].

13.6. Pure positive-definite functions 13.6.1. DEFINITION. A continuous positive-definite function lp on G is said to be pure if 7T", is irreducible.

CH. 13. 6j

PURE POSITIVE-DEFINITE FUNCTIONS

293

13.6.2. This amounts to saying that the positive form defined on L1(G) by P is pure (2.5.4). In view of 13.4.5, this again amounts to saying that, in every decomposition P = PI + P2 of P into a sum of two continuous positive-definite functions, PI and P2 are proportional to P. 13.6.3. If G is commutative, the irreducible continuous unitary re presentations of G are the characters of G. The pure continuous positive-definite functions on G are therefore the functions of the form AX (A ;;;. 0, X a character of G). 13.6.4. THEOREM. Let P be a continuous positive-definite function on G such that P(e) = 1. Then P is the limit, for the topology of compact convergence, of functions of the form AlcPl +... + AncPn, where PI, ... , Pn are pure continuous positive-definite functions equal to 1 at e, and where AI" .. ,An are non-negative numbers such that A\ +... + An = 1. The positive form !/J on LI(G) defined by P is the weak*-limit of convex combinations !/Ji of pure states of A and of 0 (2.5.5). Since lim "!/Ji" ;;;!/J = 1 and "!/Ji"";; 1, we can even suppose, multiplying the !/Ji by suitable scalars, that II !/Jill = 1. Then !/Ji is the positive form on L \(G) defined by a function A\P\ + ... + AnPn (Ph"" Pn pure continuous positive-definite functions, P\(e) =... = Pn(e) = 1, AI;;; 0, ... ,An;;; 0, A\ + ... + An = 1). It is now enough to apply 13.5.2. 13.6.5. COROLLARY. Every continuous complex-valued function on G is the limit, for the topology of compact convergence, of linear com binations of pure positive-definite functions.
It is clearly enough to prove the corollary for a function f of J( G). Now such a function is the uniform limit over G of functions f * g, with g E J(G). Moreover,

4f * g

= (f + g) * (f + gr -

(f - g)

* (f -

gf

+i(f+ig) * (f+ig)--i(f-ig) * (f-igf. Lastly, each function of the form h * h -, where h E J( G), is continuous positive-definite (13.4.11), and 13.6.4 can be applied to it. 13.6.6. COROLLARY. For every s E G different from e, there exists an irreducible continuous unitary representation 1T of G such that 1T(S) #- L There exists a continuous complex-valued function on G taking different values at s and at e, hence (13.6.5) a pure continuous positive-

294

UNITARY

REPRESENTATIONS

OF LOCALLY

COMPACT GROUPS

[CH. 13, 6

definite function

tp

on G taking different values at s and at e. We have

(7T'tp(s)~tp

I~tp) = cp(s);e

cp(e)

=u; I~tp)'

hence 11tp(s);e 1, and 11tp is irreducible. 13.6.7. Corollary 13.6.6, due to Gelfand and Raikov, expresses the fact that a locally compact group admits "enough" irreducible continuous unitary representations. This corollary would have been false, had we been limited to finite-dimensional continuous unitary representations. This justifies the study of infinite-dimensional continuous unitary re presentations. 13.6.8. We saw in 13.6.4 how to recover all continuous positive-definite functions using the pure continuous positive-definite functions. We are going to give a result in the same spirit, but in an integral form. Suppose that G is separable. Let B be the convex set of continuous positive-definite functions on G whose value at e is ~ 1. This is a compact set for the weak*-topology, separable, because L1(G) is se parable (B 7). For every s E G, the function cp ~ cp(s) on B, which is bounded in absolute value by 1, is Borel, being the limit of a sequence of continuous functions cp ~ J cp(t)/n(t) dt (take the In in JC( G), ;;;.0, with integral equal to 1, and with supports lying within smaller and smaller neighbourhoods of s). Moreover, the set P of pure continuous positive definite functions equal to 1 at e is the set of extreme points of B, less 0 (2.5.5), and is therefore a Ga set in B (B 13). This established, we have:
PROPOSITION. Let cp be a continuous positive-definite function on G such that cp(e) = 1. Then there exist a positive measure of norm 1 on B, concentrated on P, such that cp(s) = Jp ~(s) dp,(~) for every s E G.

There exists a positive measure p, of norm 1 on B, concentrated on the integral being taken in the weak P U {O}, such that cp =JpU{O} ~ dp,(~), sense (B 13). We can clearly suppose that p,({0}) = 0, hence that p, is concentrated on P. Let f E L1(G). We have
(1)

J
u.
G

cp(s)f(s) ds

= dp,(~)
B

J J~(s)f(s)
G

ds.

The function s)~ ((s) is continuous on P x G by 13.5.2, and is therefore measurable on B x G for the product measure dp,(~) ds; it is, on the other hand, bounded in absolute value by 1. Hence (~, s) ~

CH. 13, 7j

POSITIVE-DEFINITE MEASURES

295

ns)f(s) is integrable for dJL() ds and (1) can be written

f
G

cp(s)f(s) ds

f
G

f(s) ds

f
B

l(s) dJL({).

Hence
(2)

cp(s) =

f
B

C(s) dJL(C)

almost everywhere on G. Moreover, if (s., S2, ...) is a sequence of elements of G converging to s, we have l(sn) -+ l(s) for every l E B, and so

by Lebesgues theorem. Hence f B ns) dJL({) depends continuously on s, from which it follows that (2) holds everywhere on G. References: [620], [635], [638], [1101], [1455], [1868].

13.7. Positive-definite measures

13.7.1. DEFINITION. A complex-valued measure JL on G is said to be positive-definite if
(1)

for every

f E X( G). We then write

~

JL

O.

13.7,2. If JL

0, we have JL * = JL. In fact,

(JL*,f

* 1> = (JL, (f * f)-) = (JL,f * 1> = (JL,f * 1>

----::-

for every fEX(G); hence, by polarisation, (JL*,f * g)=(JL,f * g) for any f,gEX(G); hence, passing to the limit, (JL*,h)=(JL,h) for any
hE X(G),

13.7.3. Relation (1) of 13.7.1 may be written

ff

f(s

v: s) ds dJL(t) ~ 0

2%

UNITARY

REPRESENTATIONS

OF LOCALLY

COMPACT GROUPS

[CR. 13, 7

for every f E JC(G), or again, changing f into

I,

J
for every

(f.L

* f)(s )f(s) ds ;?: 0

f E JC(G).

13.7.4. Suppose that f.L is bounded. If A is the left regular representation of G, we can form A(f.L). By 13.7.3, to say that f.L pO means that (A(f.L)f If);?: 0 for every f E JC(G). Since JC(G) is dense in L 2(G), we have
f.LpO~A(f.L);?:O.

13.7.5. Let .d be the modular function of G, and

(f

f E JC(G). We have

* f*)(S).d(S)I/2 =
=

J J

f(t)I(s-lt).d(S-l t).d(S)I/2 dt

f(t).d(t)I/2f(s-lt).d(S-lt)I/2 dt

= (f.d 1/2) * (f.d 1/2n(s)

Condition (1) of 13.7.1 can thus also be expressed, changing f into f.d 1/2, by (.d 1/2f.L f * f*);?: 0 for every f E JC(G).
13.7.6. DEFINITION. A locally integrable function <p on G is said to be positive-definite if the measure .d -1/2(S)<p(S) ds on G is positive-definite. We then write <p pO.

By 13.7.5, this may be expressed by

(<p, f

* f*);?: 0

for every f E JC(G).

For continuous sp, we just recover the previous notion (13.4.4). If <p and " are two locally integrable functions on G such that " - <p p 0, we write " P <po
13.7.7. If <p pO, we have the following equality of measures:
.d -1/2(S)<p(S) ds = (.d -1/2(S)<p(S) ds)* =.d1/2(S)cP(S-I) d(S-I) =.d-1/2(S)cP(S-I) ds,

and so
<p

= cP locally almost everywhere.

13.7.8. Let <p be a complex-valued function on G such that .d -1/2<p E

CH. 13. 7j

POSITIVE-DEFINITE MEASURES

297

L1(G). By 13.7.4, cp ~o

if and only if A(.1-1/2cp)~O, left regular representation of G.

where A denotes the

13.7.9. Let JL be a positive-definite measure on G. For f, g E X(G), put

(j Ig)JL

=(JL, g * f) =JJf(t-lS)g(C I) dt dJL(s).

X( G) then becomes a pre-Hilbert space. Let HI be the (Hausdorff)

s E G,

Hilbert space which is the completion of this pre-Hilbert space. For f E X(G), put

i.e.
(u(s)f)(x)

=f(S-IX).1 1/2(s).

We have

Ilu(s)fl ~

JJ(u(S)f)(t-I)(U(S)f)(t-lU) dt dJL(u) = JJ {(S-l t-I).1 1/2(S)f(s-lt-IU).11/2(S) dt dJL(u)

=
=

JJ{(t-l)f(t-I

U).1(S).d(S)-1

dt dJL(u)

JJ{(t-I)f(t-I

U)

dt dJL(u) =

Ilf l~

Hence u(s) defines a unitary operator on HI" that we again denote by u(s). It is clear that a is a unitary representation of G in HI-Moreover, when s ~ So, u(s)f converges uniformly to u(so)f, and its support remains within a fixed compact set, from which we easily deduce that the unitary representation a is continuous. 13.7.10. PROPOSITION. Let V be a neighbourhood of e, and JL a positive definite measure on G such that
(1)
(JL,f

* f) ~

(J

f(x) dx

for all non-negative f in X(G) which vanish outside V (K denotes a /2(s) ds, where number that does not depend on f). Then dJL(s) = cp(s).1- 1 cp is a continuous positive-definite function.

298

UNITARY

REPRESENTATIONS

OF LOCALLY

COMPACT GROUPS

[CH. 13, 7

There exists a net (f;) of non-negative functions of X(G) vanishing outside V such that J fi(X) dx = 1, and such that

Ji(x)g(x) dx -+ g(e)

for every g E X(G).

By (1), we have IlJil ~ have

(fi Is); =
=

~ K for every i. Moreover, for every g E X(G), we

II I I
d/L(s)

g(t-I)Ji(CIS) dt d/L(s) g(t-IS-I)Ji(t- l) dt -+

g(S-I) d/L(s).

Hence, if A denotes the canonical mapping of X( G) into HI" Afi converges weakly to an element e of HI" such that
(2)

(e lAg) =

g(S-I) d/L(s)

for every g E X(G). Put fP(s) = (u(s)e Ie) for every S E G. Then fP is a continuous positive-definite function on G (13.4.5). Moreover, (2) im plies, for every t E G, that
(u(t)e lAg)

= (e Iu(C1)Ag)
=

(u(t l)g)(S-I) d/L(s) =

g(tS-l).1-1/2(t) d/L(s)

hence, for every f E X( G),

f(t)(u(t)eIAg)dt= =

II

f(t)g(tS-l).1-1/2(t)dtd/L(s)

JJf(t-l)g(t-lS-l)Ll- 1/2(t) dt d/L(s) = JJg(t-I)f(t-I S)Ll- I12(S-l t) dt d/L(s)

= (/L, g

* (Ll I/2f) =(Ll 1/2f Ig),..

From this we deduce that

f(t)(u(t)e

Ie) dt = (Ll 1/2f Ie),

ell. 13. 8j

SQUARE-INTEGRABLE

POSITIVE-DEFINITE FUNCTIONS

299

i.e., in view of (2) and the fact that /-t * =u,

Since this holds for every

X(G), we conclude that

d/-t(t) = .1- 1/2(t)<p(t) dt.

13.7.11.

COROLLARY. Let <p be a continuous positive-definite function, and " a locally integrable positive-definite function such that <p ~ ". Then " is equal locally almost everywhere to a continuous positive definite function.

In fact, for every

E X( G), we have

f "(s).1- 1/ 2(S)(j

* f)(s) ds :s;; f

<p(s ).1-1/2(S)(j

* f)(s) ds.

If f E X( G) is non-negative and vanishes outside a compact symme tric neighbourhood V of e, the right-hand side is dominated by

sup 1<p(s)..:1-1/2(S)/ fff(t-l)j(t-1S) dt ds

sEV2

= sup
sEV2

1<p(S).1-1/2(S)/ sup .1(t)(Jf(t) dt)2

lEV

and it suffices to apply 13.7.10. References: [620], [635].

13.8. Square-integrable positive-definite functions we will denote by A the left regular represen 13.8.1. Throughout 13.~, tation of G. Let f E L 2(G). For every g E X(G), we have A (g)f E L 2(G). If there exists a finite constant M such that IIA(g)fllz:s;; Mllgllz for every g E X(G), we will say that f is a moderated element of L 2(G). The mapping g ~ A (g)f then extends uniquely to a continuous linear operator

300

UNITARY

REPRESENTATIONS

OF LOCALLY

COMPACT GROUPS

[CR. 13, 8

on L 2(G), which we will denote by p(j). For every g E X(G), we have

(p(j)g Ig) = (A(g)! Ig) =

(g

* f)(s)g(s) ds
* g).

=f

f g(t)f(r1s)g(s) ds dt = (f, g*
~

Hence f s- 0 if and only if p(j)

13.8.2. Now let

0 (13.7.6).

L 2(G), positive-definite but not necessarily moderated. The operator g ~ A(g)f, defined on X(G) and with values in L 2(G), is ~ 0 by the same calculation as in 13.8.1. We will denote by p(j) its Friedrichs extension (B 23), which is self-adjoint and ~ O. When f is moderated, we recover the operator p(j) defined in 13.8.1.

f be an element of

13.8.3.

Let f be a positive-definite element of L 2(G). (i) For every s E G, p(j) commutes with A(s). (ii) If h belongs to the domain of definition D of p(j), we have p(j)h = h * f.
LEMMA.

For g E X(G) and s E G, we have

e,

* gEX(G)
=

and

e,

* (g

*f)=(Es

* g) t.

in other words A(s)p(j)g

p(j)A(s)g. Hence

p(f) 1 X(G)

= A(s)p(j)A(s)-1 1 X(G)

and, consequently, p(j) = A(s)p(j)A(s)-1 (B 23). Moreover, p(j) coincides on D with the adjoint of p(f) I X(G) (B 23). Hence, if hE D and g E X(G), we have
(1)
(p(f)h Ig)

= (h Ip(f)g) = f

h(s) ds f g(t)!(t-I S) dt.

Now h(s)g(t)f(t-s) is ds dt-measurable on G x G, and zero outside a countable union of ds dt-integrable sets; moreover,
f* ds f* Ih(s)g(t)f(t-lS)1 dt

f* Ih(s)l<jgl

* Ifl(s ds < +00,

because Ihl E L 2(G) and

Igj * IfIE L 2(G).

Hence h(s)g(t)!(t-lS) is ds dt-

CH. 13. 8j

SQUARE-INTEGRABLE

POSITIVE-DEFINITE FUNCTIONS

301

integrable, and (I) may be written

(p(f)h

Ig) = J[J h(S)!(S-It)

= h

dS] g(t) dt =

(h

* f)(t)g(t)

dt,

whence p(f)h

*f

almost everywhere.

13.8.4. LEMMA. Let a, b be two moderated positive-definite elements of L 2(G) such that pea) and pCb) commute. Then a * b is a continuous positive-definite function, and is a moderated element of L 2(G). We have p(a * b) = p(a)p(b), and (a I b) ~ O. If, further, a <t, b, we have

lib - alg.:;; Ilbl ~ -

I al ~

The convolution product of two functions of L 2(G) is a continuous function on G. Let (an) be a sequence of functions of X(G) converging to a in L 2(G). Then an * b = p(b)a n converges to p(b)a in L 2(G). Moreover,
IIan

*b -

* bll",~O,

and so

* b = p(b)a

E L

2(G).

Let f E X(G). We have

f * an E
and f

X(G),

converges to f * a in L 2(G), hence (f * an) * b = p(b)(f converges to p(b)(f * a) = p(b)p(a)f in L 2(G). Moreover, 2(G). f * (an * b) converges to f * (a * b) in L Hence

* an * an)

* (a * b) = p(b)p(a)f=p(a)p(b)f

for every fEX(G),

which proves that a * b is moderated and that p(a * b) = p(a)p(b). Since pea), p(b) are positive and commute, p(a)p(b) is positive, and so a * b is positive-definite. We have
(a

I b)
(a

a(s)b(sI) ds = (a
~

* b)(e) ~

O.

Finally, if a hence

<t, b,

let c

b- a

0; c is moderated, and, by the above

Ia).:;; (a I a) +(a Ic) = (a I b),

e( G),

Let aI a-, ... be moderated positive-definite elements of such that a I <t, a2 <t, " and the pea;) commute pairwise. If sup Ilailiz < +00, the a, have a norm-Limit in e(G).
LEMMA.

13.8.5.

302

UNITARY

REPRESENTATIONS

OF LOCALLY

COMPACT GROUPS

[CH. 13, 8

By 13.8.4, the Ilajllz form an increasing, and therefore convergent, sequence. Applying 13.8.4 again, we then see that the a, form a Cauchy sequence. 13.8.6. THEOREM. Let ({I be a square-integrable continuous definite function on G. Then there exists a square-integrable definite function 1/1 on G such that ({I = 1/1 * 1/1 = 1/1 * ~'
positive positive

Suppose, to begin with, that ({I is moderated. We can suppose, mul tiplying ({I by a constant >0, that 0 ~ p({I) ~ 1. Let (Pl(t), P2(t), , .. ) be an increasing sequence of non-negative polynomials on [0, 1], vanishing at over [0, 1]. Thanks to 13.8.4, we can 0, which converges uniformly to form 1/11 = PI({I), 1/12 = P2({I), . . (using convolution for the multipli cation); these elements are in L 2(G) and are moderated, and we have

vt

p(l/Iz} = P2(P({I

, ....

We see that 0 ~ P(I/II) ~ p(l/Iz} ~ . " hence 04i 1/11 4i 1/124i .. ; also, the P(I/In) commute pairwise. Since p~(t) ~ t on [0, 1], we have P(l/In)2:o;;; p({I), hence I/In * I/In 4i ({I, and so (I/In * I/In)(e) = II I/In 1/2 ~ ((I(e). Hence the I/In converge in norm to an element 1/1 of L 2(G ) (13.8.5). At the same time, P(I/In) = Pn(p({I converges in norm to p({I)1/2. For f E JC(G), f * I/In = P(I/In)f converges in L 2(G ) to f * 1/1 on the one hand, and to p({I)1/2f on the other hand. Hence 1/1 is moderated and positive-definite, and p(l/I) = p({I )1/2, This implies that p({I) = p(I/I)2 = p(1/I * 1/1), whence ({I = 1/1 * 1/1. We now pass to the general case. Let p({I) =J "(dE, be the spectral o decomposition of p( ({I). The projections E, commute with the A (s )(s E G) by 13.8.3 (i). Let ({I, = Ee({!. For every g E JC(G), we have
(g

* cj5,)(t) =

g(s)cj5,(S-l t) ds

g(s)cj5,(r

ls)

ds
1s)

= (g A(t)Ee({!) = (E,g IA (t)({I) = = (E,g)(S)({I(S~lt)

(E,g)(s)cj5(r

ds

ds

=(E,g * ({I)(t)

and this is equal to (p({I )E,g)(t) by 13.8.3 (ii). Hence cj5, is moderated and p(cj5,) = p({I )E, ~ 0, hence cj5, = ({I, ~ O. Besides, p({I,) ~ p({I), and so ({I,4i ({I, hence ({I, is continuous (13.7.11). By the first part of the proof, there exists a moderated positive-definite element w, such that ({I, = w, * we Take, in particular, (= 1,2, ... , n, .. . . The P({In) commute pairwise and

CH. 13. 9]

THE C*-ALGEBRA

OF A LOCALLY

COMPACT GROUP

303

WI

increase with n, and hence the same is true of the p(w n) = P(Pn)I/2. Hence 2 ~ W2 ~ . Moreover IIwnl1 = Pn(e)"" P(e), and hence the W n have a norm-limit W ;;;. 0 (13.8.5). Then Pn i.e. ta; * w" converges uniformly on G to W * W; on the other hand, Pn = EnP converges in norm to P in L 2(G). Hence P = W * w. Reference: [635].
13.9. The C*-algebra of a locally compact group

13.9.1. Since LI(G) is an involutive Banach algebra with an approximate identity, we can form its enveloping C*-algebra (2.7.2). This C*-algebra is called the C*-algebra of G and is denoted by C*(G). For IE LI(G), put 11111 = sup 111T(f)11 ""11/111 where 1T runs through the set of non-degenerate representations of LI(G), or, which amounts to the same thing, the set of continuous unitary representations of G. Then I ~ 11111 is a seminorm on LI(G) (2.7.1), and, indeed, a norm, since LI(G) admits an injective representation (13.3.6). The C*-algebra of G is just the completion of LJ(G) for this norm. 13.9.2. If G is discrete, C*(G) admits an identity element. If G is separable, C*(G) is separable (13.2.4). 13.9.3. By 2.7.4 and 13.3.5, there exists a bijective correspondence between continuous unitary representations of G and non-degenerate representations of C*(G). Everything that was said in 13.3.5 is still valid when C(G) is replaced by C*(G). To the left regular representation of G, corresponds a representation of C*(G) called the left regular representation of C*(G) in e(G). 13.9.4. The group G is said to be liminal, postliminal, antiliminal, of type I, if C*( G) is liminal, postliminal, antiliminal, of type I. G is liminal if and only if, for every irreducible continuous unitary representation 1T of G and every IE L I( G), 1T(f) is compact. Suppose that G is postliminal; then, for every irreducible continuous unitary representation 1T of G, the norm-closure of 1T(L I( G contains .:ee(Hrr) (4.3.7). The converse is true if G is separable (9.1) (and even in general: cf. 9.5.9). G is of type 1 if and only if, for every continuous unitary represen tation of G, the von Neumann algebra generated by 1T(G) is of type I. If G is postliminal, G is of type 1 (5.5.2). If G is separable (and even in

304

UNITARY

REPRESENTATIONS

OF LOCALLY

COMPACT GROUPS

[CH. 13, IO

general: cf. 9.5.9), the following conditions are equivalent: (J) G is of type I; (2) for every continuous unitary factor representation 1T of G, the factor generated by 1T(G) is of type I; (3) G is postliminal (9.1). References: [582], [896].

13.10. The Hilbert algebra of a unimodular locally compact group

13.10.1. For the convolution product and the involution I ~ f*, X( G)t is an involutive algebra. We endow :JC(G) with the scalar product (f Ig) = f I(s)g(s) ds. It is easy to see that X(G) then becomes a Hilbert algebra; the completed Hilbert space is just L 2(G). The full Hilbert algebra A of bounded elements (A 57) is called the Hilbert algebra 01 G; we have :JC(G) CAe L 2(G). It is clear that A is closed under the mapping I ~ f, and also, therefore, for the mapping I ~ ! = (1)*. 13.10.2. A continuous linear operator on L 2(G) commutes with the left-translation operators if and only if it commutes with the operators of left-convolution by the elements of X(G) (this follows, for example, from 13.3.5 applied to the left regular representation of G). Hence OU(A) is the von Neumann algebra generated by the left-translation operators on L 2(G). Similarly, V(A) is the von Neumann algebra generated by the right-translation operators on L 2(G). 13.10.3. If I E A, recall that Uf , Vf denote the continuous linear opera tors on L 2(G) which extend left- and right-multiplication by I in A. If f E A and g E L 2(G), we have Ufg =I * g (and similarly Vfg =g * f). In fact, suppose to start with that g E X(G); let (fn) be a sequence of elements of X(G) converging to I in the e-sense; then VIn = In * g converges to V J = Ufg in the e-sense, and also to I * g, so that Ufg = f * g. In the general case, let (gn) be a sequence of elements of X(G) converging to g in the L 2-sense; then Ufg n converges to Ufg in the L 2-sense, and I * gn converges to I * g uniformly over G; since Ufg n = I * gn by the first part of the proof, we have Ufg =I * g. 13.10.4. Recall that the mapping I ~ 1* commutes with every hermitian element of OU(A) n V(A) (the common centre of OU(A) and V(A)) (A 54). Recall also that each of OU(A), V(A) is the commutant of the other, and that each is a semi-finite von Neumann algebra (A 60).
tThroughout this section, G denotes a unimodular locally compacl group.

CH. 13. 11]

ADDENDA

305

13.10.5. PROPOSITION. Let G be a unimodular locally compact group, and A its Hilbert algebra. Then the following conditions are equivalent: (i) (resp. (i)) The von Neumann algebra OU(A) (resp. V(A is a finite von Neumann algebra; (ii) There exists in G a base of compact neighbourhoods of e invariant under the inner automorphisms of G.

Conditions (i) and (i) are equivalent for every Hilbert algebra (A 63). Suppose that there exists a base of compact neighbourhoods (VJiEI of e invariant under the inner automorphisms of G. Let I, be the charac teristic function of Vi The I, are central elements of A (they belong to the centre of L1(G. Moreover, every f E L 2(G) is in the norm-closure of the set of the fi * f. Hence the characteristic projection of A is 1 (A 62). Hence OU(A) and V(A) are finite von Neumann algebras (A 63). Suppose that OU(A) and V(A) are finite von Neumann algebras. The characteristic projection of A is 1 (A 63). Let s be an element of G different from e. The operator f --7 Jon L 2(G) is not the identity; now it commutes with right-translations, and the set of right-translations by the elements of L 2(G) central relative to A is total in L 2(G) (A 62). Hence there exists an f E L 2(G) central relative to A and such that J is distinct from f. Since f is central, we have J = t. for every t E G (A 62), and so f is invariant under the inner automorphisms of G. Consider the function t --7 g(t) = (tf If) on G. It is continuous, invariant under inner automor phisms, vanishes at infinity, and g(s) ~ g(e). If W is a compact neigh bourhood of gee) in C containing neither 0 nor g(s), the relation g(t) E W defines a compact neighbhourhood of e, invariant under inner automorphisms, and not containing s. Hence the intersection of all the compact neighbourhoods of e invariant under inner automorphisms is just {e}. Every neighbourhood of e therefore contains a compact neigh bourhood invariant under inner automorphisms. References: [639], [641], [1036], [1037], [1457], [1458].

13.11. Addenda
13.11.1. Let G be a locally compact group, and f E L I(G). If f(s) ~ 0 for every s E G, the norm of f in L I( G) and in C*( G) is the same. (Consider the trivial I-dimensional representation of G). [582].

306

UNITARY

REPRESENTATIONS

OF LOCALLY

COMPACT GROUPS

[CR. 13,

su

13.11.2. Let 0 be a locally compact group, 7T a continuous unitary representation of 0, S a set of continuous unitary representations of 0, and K the set of ~ E HIT such that the function s ~ (7T(S)~ I~) is the uniform limit over every compact set of sums of positive-definite func tions associated with S. Then K is a closed subspace of HIT invariant under 7T(0). [587]. 13.11.3. Let 0 be a locally compact group, A the left regular represen tation of 0, and 7T a continuous unitary representation of O. Then A 0 7T = (dim 7T) A. (Consider the isomorphism of L Z( 0) 0 H", onto L 2( 0) 0 H", = L;, (0) which transforms f 0 ~ into the function s ~ f(S)7T(S-I)~). [585j. 13.11.4. Let 0 be the locally compact group of affine transformations of R, 0 the ~ normal closed subgroup of translations. If 7T is a continuous unitary representation of 0 and if X is a non-trivial character of G. then 7T I 0 does not contain X. (If 7T I 0 contained X, then 7T I 0 would contain all the non-trivial characters of 0 by the action in 0 of the inner automorphisms of 0; the corresponding subspaces of H", would be mutually orthogonal). Deduce from this that a non-trivial characterof 0 cannot be extended to a continuous positive-definite function on O. (Douady, unpublished.) 13.11.5. Let 0 be a locally compact group, and J its modular function. If a continuous positive-definite function on 0 is integrable for J (s )-1/2 ds, it is square-integrable for ds. [635].

*13.11.6. Let 0 be a locally compact group, and E the set of linear combinations of continuous positive-definite functions. For cp E E, let K~(resp. K~) be the closed convex hull of the set of left- (resp. right-) translates of cp in the space of continuous complex-valued functions on o endowed with the uniform norm. Then K~ and K~ both contain the same single constant M(cp). M(cpcp) = 0 for a continuous positive-definite function cp if and only if 7TI does not admit any non-zero finite dimensional subrepresentation. [635].
13.11.7. Let 0., O 2 be two topological groups. (a) Let 7T1 be an irreducible continuous unitary representation of OJ> and 7Tz and 7T~ continuous unitary representations of O 2 If the re-

CH. 13. llj

ADDENDA

307

presentations (s., s0 ~ 7T((SI) 7Tisz) and

(s; sz)~
7T1(SI)

7T2(SZ)

are equivalent, then TTZ and 7T2 are equivalent. (b) Let 7T be a continuous unitary representation of G( x G z If 7T is a factor representation, then 7T I G. and 7T IG z are factor representations. (c) Let 7T be a continuous unitary representation of G 1 x G 2 , 7T( = 7T IG) and 7T2 = 7T IG 2 Suppose that 7T( is a factor representation of type I. Then there exist 7T; "" 7T(, 7T2"" 7T2 such that 7T is equivalent to the representation (s\> s0~ 7T;(St) 7T2(S2) (d) If G 1 and G, are of type I, then G 1 x G 2 is of type I. [1005]. 13.11.8. Let G be a locally compact group, and 7T a unitary represen tation of G. Suppose that, for all g, T/ E H = Hi; the function S ~ (7T(S)g IT/) is measurable for the Haar measure on G. We can define 7T(j) = f !(S)7T(S) ds for every! E L1(G). Let K ~ H be the essential subspace for the representation 7T of L1(G). Then K and K are invariant under 7T(G). The representation S ~ 7T(S) IK of G is con K, we have (7T(S)g IT/) = 0 locally almost tinuous. For g, T/ E H everywhere on G. The space K is either zero or non-separable. It can happen that it is non-zero. [1473].

He

He

13.11.9. Let G be a topological group. Let 7T be a continuous unitary representation of G. In H 1T H 1T ... H 1T (n factors), the symmetric tensors generate a closed subspace K invariant under p = 7T 7T ... 7T. The subrepresentation PK of P is called the nth symmetric tensor power of 7T. The antisymmetric tensor powers are defined analogously. *13.11.10. Let G be a connected real Lie group, and 7T a unitary representation of G, continuous for the norm topology of operators and irreducible. Then 7T is finite-dimensional. [1484]. 13.11.11. Let G be a topological group. A unitary representation 7T of G is said to be real if there exists a closed real subspace K of H 1T such that H 1T is the direct sum of K and of iK, the scalar product is real on K, and 7T(G) leaves K invariant (in other words, H", is the Hilbert space which is the complexification of a real Hilbert space K, and 7T is the com plexification of a representation of G in K). It comes to the same thing to say that there exists an involution J of H commuting with 7T( G). If 7T is real, we have 7T = iT, but the converse is not true.

308

UNITARY

REPRESENTATIONS

OF LOCALLY

COMPACT GROUPS

[CH. 13, ll

Semis imp Ie connected real Lie groups and nilpotent con nected real Lie groups are liminal (cf. 15.5.6). A real algebraic linear group is postliminal. Large classes of postliminal solvable Lie groups are known, but Mautner has given an example of a non-postlirninal solvable Lie group. The non-commutative, 2-dimensional, solvable, connected, real Lie group is postliminal but not liminal. A countable discrete group is of type I if and only if it is the extension of a finite group by a commutative group. Problems: is a p-adic algebraic group postliminal? Does the C*-algebra of a liminal real Lie group have generalised continuous trace? [439],[441],[448],[586],[758],[903],[922],[1507],[1608], [1774].

*13.11.12.

*13.11.13.

Let G be a locally compact group, and 1T an irreducible continuous unitary representation of G. Then 1T(L1(G is not algebraically irreducible in general.

13.11.14. (a) Let G be a discrete group, A its Hilbert algebra, 611

6lL(A), which is a finite von Neumann algebra, and 1 the natural trace on 611+ defined by A. Since the operator 1 E 611 corresponds to the charac

teristic function of e, we have nt t =nt = 611, and every element of 611 is of the form Ux , where x E L 2(G). We will still denote by 1 the linear extension of 1 to 611. If U, E 611 with x E e(G), we have I(Ux ) = x(e). (b) Let E[. (resp. Ell) be the greatest projection of the centre of 611 such that the corresponding algebra induced by 611 is of type I, (resp, of type II) (i takes the values 1, 2, ...). We have E IT + ~ Eli = 1. Put 1(Ell ) = r I(EI ) = r.; whence + ~ = 1. (c) Let C be the group of commutators of G. Let rpc be the charac teristic function of C. If C is infinite, then = O. If C is finite, then 1 = (Card C)-I, and e, is the element of 611 defined by the function (Card C)-Irpc on G. (d) If G is infinite, and if C coincides with the centre of G and is of prime order p, then 611 is the product of a commutative von Neumann algebra and p - 1 factors of type III We have 1 = lip, r = (p - l)/p. (e) Let (G j ) be an infinite family of non-commutative finite groups. Let G be the set of elements of Il G, all but a finite number of whose components are equal to e. Regard G as a discrete group. Then 611 is of type 11\. (f) Let G be a discrete group, and Go the subgroup of G which is the union of the finite classes of G. If GIGo is infinite, then 611 is of type III If Go = {e}, and G is infinite, then 611 is a factor of type III [897], [1006], [1035], [1039].

CHAPTER 14

SQUARE-INTEGRABLE IRREDUCIBLE REPRESENTATIONS

14.1. Definition of square-integrable representations

14.1.1. LEMMA. Let (() be a square-integrable continuous positive-definite function on G. t Then the representation 1fl is contained in the left regular representation. There exists l/J E L 2(G) such that ({)
(()(s)

= J; * ;fr

(13.8.6). In other words

J;(t)</J(s-lt) dt

= (A(s)l/J Il/J)

where A denotes the left-regular representation of G. Hence TTI is equivalent to the subrepresentation of A for which l/J is a cyclic vector (13.4.5(iii)). 14.1.2. LEMMA. Let A be the left-regular representation of G, and tt a representation contained in A, with H",.= O. Then there exists a non-zero, square-integrable positive-definite function associated with tr. Let A be the Hilbert algebra of G. Let K be a closed subspace of
L 2(G) invariant under A(G). We have P K E OU(A) = V(A). Hence PK(A) k A (A 59). If K = 0, we see that there exists a non-zero l/J in

K n A. The subrepresentation of A corresponding to K has, as as Il/J) = J; * ;fr. sociated positive-definite function, the function s ~('TT'(s)l/J Now J; E A, and hence J; * ;fr E L 2(G).

14.1.3. DEFINITION. An irreducible continuous unitary representation of G is said to be square-integrable if there exists an g E H",., g= 0, such that the coefficient s ~ (1f(s)g Ig) of 1f is square-integrable over G. [We shall see that all the coefficients of tt are then square-integrable (14.3.2)]. By 14.1.1 and 14.1.2, these representations are the irreducible re presentations contained in the regular representation.
tThroughout this chapter, G denotes a unimodular locally compact group.

310

SQUARE-INTEGRABLE

IRREDUCIBLE REPRESENTATIONS

[CH. 14, 2

X of G, and we have

14.1.4. Suppose that G is commutative. An element of G is a character Ixl2 = 1. Hence, if G is compact, every element of G is square-integrable (further, d. chapter 15); and, if G is not compact, no element of G is square-integrable.

References: [634], [759]. The results of chapter 14 can also be ob tained without reference to 13.8; see [194], [889].

14.2. Square-integrable representations and minimal biinvariant subspaces

of L 2(G).

14.2.1. Let A be the Hilbert algebra of G. Let K be a closed subspace of L 2(G). To say that K is invariant for the left- (resp. right-) regular representation of G is equivalent to saying that P K E V(A) (resp. P K E OU(A. To say that K is invariant for both the left- and right-regular representations, or, as we shall say, biinvariant, is equivalent to saying that P K belongs to OU(A) n V(A), the common centre of OU(A) and V(A). We are going to establish a bijective correspondence between the classes of square-integrable irreducible representations of G and certain minimal biinvariant subspaces of L 2(G), in other words certain minimal projections of OU(A) n V(A). 14.2.2. PROPOSITION. Let A be the left-regular representation of G, and H a closed subspace of L 2(G) which is a minimal invariant subspace for A, so that the subrepresentation (T of A in H is irreducible. Let E = PH E V(A), and F be the central support of E in OU(A) n V(A). (i) F is a minimal projection of OU(A) n V(A). (ii) OU(A)F andV(A)F are mutually commuting factors of type 1. (iii) Every element of F (e(G belongs to A. (i) follows from 5.2.7. By condition (v) of 5.3.1, the subrepresentation of A defined by F is quasi-equivalent to (T, and therefore a factor representation of type I (5.4.11), whence (ii). Lastly, (iii) follows from A 65 and A 67. 14.2.3. PROPOSITION. We retain the notation of 14.2.2. Let (T be ano ther irreducible subrepresentation of A, and let F be the corresponding minimal projection of OU(A) n V(A). Then (T and (T are equivalent (resp. inequivalent) if and only if F and F are equal (resp. orthogonal).

CH. 14, 3]

COEFFICIENTS OF SQUARE.INTEGRABLE

REPRESENTATIONS

311

If F = F , then a and a are quasi-equivalent (condition (v) of 5.3.1), and therefore equivalent (5.3.3(ii)). If F and F are orthogonal, a and a are disjoint (condition (iii) of 5.2.1).

14.2.4. We say that F (resp. F(L 2(G))) is the central projection (resp. the biinvariant subspace) associated with a, or with every equivalent representation. 14.2.5. PROPOSITION. Let F be a minimal projection of OU(A) n V(A) such that OU(A)F and V(A)F (which are factors) are of type 1. There exists at least one minimal projection E of V(A) dominated by F. The subrepresentation tr of A defined by E is irreducible and square-in tegrable, its class does not depend on the choice of E, and the associated central projection is F. Every type I factor possesses minimal projections (A 36), whence the existence of E. Since a is an irreducible subrepresentation of A, it is square-integrable (14.1.3); it does not depend on the choice of E (14.2.3); the associated central projection is clearly F. References: [634], [759]. 14.3. Coefficients of square-integrable representations 14.3.1. PROPOSITION. Let a be an irreducible subrepresentation of the e:T)(s) = left-regular representation of G. For g. 7/ E H a let s -+ P (u(s)g 17/) be the corresponding coefficient of a. We have ihT) = g * 7/* E L 2(G). The closed subspace of L 2(G) generated by the </> 1 (g,7/ E H a ) is the biinvariant subspace associated with o: We have
cP. T)(s)=(glu(s)7/)=

g(t)ii(s-lt)dt=

g(t)7/*(r 1s)dt=(g

* 7/*)(s).

Let K be the biinvariant subspace associated with a. Then g and 7/, and therefore 7/*, belong to K, hence to the Hilbert algebra A of G (14.2.2(iii)), and so g * 7/* = VT).g E K. Moreover, since H; is invariant under OU(A), we see that the set of the g * 7/* (g, 7/ E H a ) is closed under left- or right-multiplication by elements of A. Hence if K denotes the closed subspace generated by this set, then K is a closed biinvariant subspace of L 2(G) contained in K. Since K is a minimal closed bi invariant subspace (14.2.2 (i), we have K = K.

312

SQUARE-INTEGRABLE

IRREDUCIBLE REPRESENTATIONS

[CH. 14. 3

14.3.2. COROLLARY. Let a be a square-integrable irreducible represen tation of G. Then all the coefficients of a are square-integrable. 14.3.3. THEOREM. Let a be a square-integrable irreducible represen tation of G. For g, TJ E HU put <P~,1j(s) = (7T(S)g ITJ) (s E G). There exists a unique constant d; (0 < d; < + (0) such that
(1) (2)

for any g, TJ, g, TJ E Her In particular, if <p is a positive-definite function associated with a which equals 1 at e, we have
(3)

(4)
Let A be the Hilbert algebra of G. We can suppose that a is the subrepresentation of the left-regular representation defined by a minimal projection E of Y(A). Using the equality q;~,1j = g * TJ * = V1j.g of 14.3.1, we have

J<P~.1j(s)<P~'.1j.(s)

ds

J<P~,1j(S-l)<Pf.1j'(S-l)

ds

= (V1j.g I V1j"g)

= (V1j v~g

If) = (EV1j V~Eg If).

Now, since E is a minimal projection of Y(A), there exists a A1j.1j E C, such that EV1j V~E = A1j.1j.E. Hence

The integral is also equal, by 13.1.1, to

Hence A1j,1j = Au( TJ ITJ), where Au is a finite constant. Putting g = f = TJ = TJ - 0, we see that Au > 0. It then suffices to put A~l = d u to obtain (1).

CH. 14. 3]

COEFFICIENTS OF SQUARE-INTEGRABLE

REPRESENTATIONS

313

In particular.
(P<,l * Pf.l)(s)=

J(u(t)~!

r/)(U(t-Is)g ITj) dt

= (u(t)~ = d~I(~
14.3.4. a.
DEFINITION.

ITj)(u(t)Tj Iu(s)g) dt

ITj')(u(s)~'1

n)

= d~I(~

ITj)Pf.l(s).

The constant d; is called the formal dimension of

Its value depends on the choice of the Haar measure of G. We will see in 15.2.3 that if G is compact and if the chosen Haar measure is of total mass 1, then d; is the dimension of a in the usual sense. This will explain the terminology employed.
14.3.5. PROPOSITION. Let A, P be the left- and right-regular represen tations of G, tr an irreducible subrepresentation of A, K ~ H(J" the bi invariant subspace of L 2 (G ) associated with tr, Al and PI the subre presentations of A and P defined by K, and the conjugate Hilbert space of H(J" (i) There exists exactly one isomorphism cP of the Hilbert space H(J" 0 onto K which transforms ~ 0 Tj and d~2eP<.'l for any ~, Tj E H(T (ii) Let I H ,Iii be the trivial representations of G in H(J" and H(T Then " o cP transforms tr 0 Iii" into Al and IH " 0 ii into PI (iii) Al = (dim u)u, PI = (dim u)ii.

it

it

The mapping (~, Tj) ~ d~2eP<.'l is linear in ~, and conjugate-linear in Tj. Hence there exists a linear mapping of the algebraic tensor product of H(J" and H(J" onto a dense subspace of K which transforms ~ 0 Tj into d~2eP<.'l (14.3.1). If ~, g, Tj, Tj E Hen we have, by 14.3.3,
(~0
Tj

Ie 0

Tj) =(g If)(Tj

ITj) =(d~2eP<.'l

Id~2ePf.'l.)'

and so the above linear mapping extends to an isomorphism cP of the Hilbert space H(J" 0 onto the Hilbert space K. The uniqueness property of (i) is immediate since the set of the ~ 0 Tj is total in H(J" 0 H(J" Let s E G, ~, Tj E H eT ; we have

it

cP(u(s)g 0 Tj) = ePeT(S)<.l

314

SQUARE-INTEGRABLE

IRREDUCIBLE REPRESENTATIONS

[CH. 14, 3

and

hence <P transforms (J" 0 hiu into AI> and we see similarly that <P transforms I Hu 0 ii into PI Lastly, (iii) follows from (ii).

He, 0 it canonically
we have

14.3.6.

PROPOSITION.

We retain the notation of 14.3.5. If we identify with the set of Hilbert-Schmidt operators on Hen

<P(u*) = <P(u)*,

for any u, v E H;

0 HU"

It is enough to verify this for u =

g0

TJ, v = g 0 TJ (g, TJ, g, TJ E H u )

Now and
<P((g

TJ)(g

TJ) = <P((g ITJ)g

= d~2ducP<'TI

0 TJ) = d~\g' * cPf,TI = d~2<p(g

ITJ)cPe,TI 0 TJ) * <P(g 0

TJ)

14.3.7. THEOREM. Let (J", (J" be inequivalent square-integrable irreducible representations of G. For g, 1/ E H u , g, TJ E H,; we have

J(/<,TI(s)(/e.TI.(s) ds
(J"

0,

The biinvariant subspaces K and K of e(O) associated with (J" and are orthogonal (14.2.3). Moreover, cPe,TI E K and cPe,TI E K (14.3.1), whence

Finally, K and K are invariant under OU(A) and Y(A) (A denotes the Hilbert algebra of 0), whence
cPf.TI

* cP~,TI

K n K =

and

(/'~,TI

* (/f.TI =

0.

References: [634], [759].

CH. 14, 4j

FORMAL DIMENSION AND TRACE

315

14.4. Formal dimension and trace 14.4.1. LEMMA. Let H be a Hilbert space, B the Hilbert algebra of Hilbert-Schmidt operators on H, b ~ U; the canonical mapping of B into Oll(B), E a minimal projection of V(B), and u(b) the operator induced by U; in E(B). For every bE B, we have Ilblf = Tr(u(b)u(b)*).
Oll(B) is a type I factor, and b ~ U b is an isomorphism of the Hilbert algebra B onto the Hilbert algebra of Hilbert-Schmidt operators relative to Oll(B) (A 61). Moreover, T ~ T E is an isomorphism of Oll(B) onto !(E(B. Hence b ~ u(b) is an isomorphism of the Hilbert algebra B onto the Hilbert algebra of Hilbert-Schmidt operators on E(B). Hence

Tr(u(b )u(b )*) = Tr(bb*)

= IIbW.

14.4.2. PROPOSITION. Let a be a square-integrable irreducible rep resentation of G, and F the central projection associated with a, For every f E L1(G) n L 2(G ), we have

!(Ff)(s)!2 ds

= du Tr(u(f)u(f)*)

Let A be the Hilbert algebra of G, and K = F(L 2(G . Then PK(A) = A n K = K (14.2.2), and K is a Hilbert algebra whose associated von Neumann algebras are Oll(A)F and V(A)p Let E be a minimal projection of V(A) dominated by F, so that tr may be identified with the subre presentation of the left-regular representation defined by E. Every f E A defines an operator U, E Oll(A), and U, induces on H; = E(L 2(G an operator that we will denote by u(f); this is consistent with the usual notation if f E L I(G) n L 2(G). In fact, we are going to prove the equality of the proposition for every I E A. We have UFf = FUf , hence u(Ff) = u(f). Neither side of the equality to be established is therefore altered if f is replaced by Ff. We will suppose from now on that f E K. We return to the notation of 14.3.5. There exists a u E H; ~ Hu such that f = l/>(u), and we have 1I/Ib =Iluli. Moreover, if v E H; ~ Hen we have l/>(u)l/>(v) = d~1/2l/>(UV) (14.3.6), and so l/> -I transforms the operator on K of left-multiplication by f into the operator on H; ~ Hu of left-multiplication by d~1/2U. In view of 14.4.1, we have

316

SQUARE-INTEGRABLE

IRREDUCIBLE REPRESENTATIONS

[CH. 14. 5

14.4.3. PROPOSITION. Let f E L1(G) n e(G), f E L(G) n tsov. and g = f * f. For every t E G, denote by s the [unction s ~ g(tst -I) on G. Let a be a square-integrable irreducible representation of G. (i) The operator u(g) is a trace-class operator. (ii) For every unit vector ~ of H", the function t ~ (U(gl)~ I~) on G is integrable, and we have

(u(gt){

I{) dt = d~I

Tr u(g).

By linearity, we are led to the case where f = f*. Since u(f) is a Hilbert-Schmidt operator (14.4.2), u(g) = u(f)u(f)* is a positive trace class operator. Let (e.) be an orthonormal basis of H; consisting of eigenvectors of u(g). Let U(g)Ei = A,oEi, so that L Ai < +00. For every t E G, we have
(U(gl){

I{) = (u(t-I)u(g)u(t){ I{) = (u(g)u(t){ Iu(t){) =}: (u(g)u(t){ IEi)(U(t){ lEi) =}: Ai I(u(t){ I EiW,
i
i

whence, in view of 14.3.3

f (U(gl)~

I~)

dt

= ~ Ai I(u(t)~

I EiW dt

=}: Aid~I

= d~I Tr u(g).

Reference: [759].
14.5. Integrable representations
14.5.1. PROPOSITION. Let a be an irreducible continuous unitary re presentation of G. The following conditions are equivalent: (i) There exists {EH", {O, such that the function s~(u(s)~I~) is integrable over G; (ii) There exists a dense subspace H of H such that for 7/, (E H, the function s ~ (u(s)7/ I() is integrable over G.

Suppose condition (i) is satisfied. Let g, hE X(G). For every s E G, we have

(u(g){ Iu(s)u(h){)

ff =Jf
=

g(t)h(u)(u(t){ Iu(s)u(u){) dt du g(t)h(u)({ IU(t-I S U ) { ) dt duo

CR. 14, 6]

ADDENDA

317

Hence
f* I(u(g)g Iu(S)u(h)g)1 ds

~
=

f J J* Ig(t)h(u)(g IU(r1su)g)1 ds dt du

JI

Ig(t)h(u)(g I u(s)g)1 ds dt du

= J Ig(t)1 dt J Ih(u)1 du JI(g

I u(s )g)1 ds < + 00.

Hence condition (ii) is satisfied, taking H = u(JC(Gg. (This subspace is non-zero and invariant under u(G), hence H = H since a is ir reducible).
14.5.2. DEFINITION. An irreducible continuous unitary representation of G is said to be integrable if it satisfies the equivalent conditions of 14.5.1.

With the notation of 14.5.1, the function s ~(u(s)g Ig) on G, being bounded and integrable, is square-integrable. Hence an integrable ir reducible representation is square-integrable. However, the converse is false. Reference: [759].
14.6. Addenda

14.6.1. Let G be a locally compact group, 7T a continuous unitary representation of G, and p and q numbers in [1, +00] such that l/p + 1/q = 1. The following conditions are equivalent: (a) there exists a finite constant M such that

117T(f)11 ~

Mllfllp

for / E L 1(G) n U(G)

Tr ,

(II/lip
s ~(7T(S)g

denotes the norm in U(G; (b) for all g,1/ E H 11/) belongs to Lq(G). [950].

the function

14.6.2. Let G be a unimodular locally compact group, a a square integrable, irreducible, continuous unitary representation of G, and K; the associated biinvariant subspace. Then if is square-integrable and irreducible, and the associated biinvariant subspace is the image of K; under the mapping / ~ J and also under the mapping / -r! of e(G) onto
L 2(G).

318

SQUARE-INTEGRABLE

IRREDUCIBLE REPRESENTATIONS

[CR. 14, 6

*14.6.3. Semi-simple connected complex Lie groups and nilpotent simply connected real Lie groups have no square-integrable, irreducible, continuous unitary representations. For the case of real semi-simple connected Lie groups, see a series of important papers by Harish Chandra, beginning with [759] and [760]; a lot of work has been done recently on this subject. The notion of square-integrable representations admits of several generalizations (representations "square-integrable relative to the cen tre", case of non-unimodular groups). See, for example, C. C. Moore and J. A. Wolf, Square-integrable representations of nilpotent Lie groups, Trans. Amer. Math. Soc., 185 (1973) 445-462, and C. C. Moore and M. Duflo, On the regular representation of a non-unimodular locally compact group, J. Functional Anal., 21 (1976) 209-243.

CHAPTER 15

REPRESENTATIONS OF COMPACT GROUPS

15.1. Complete reducibility

15.1.1. LEMMA. There exists in Gt a base of neighbourhoods of e which are invariant under inner automorphisms. We have to show that, if s E G and S,p e. there exists a neighbourhood V of e, compact and invariant under inner automorphisms, such that s ~ V. Let W be a compact neighbourhood of s such that e ~ W. The image of G x W under the mapping (t, w) ~ twt- 1 is a compact neigh bourhood W of s, invariant under inner automorphisms, such that e~ W. The complement of W is therefore an open neighbourhood of e, invariant under inner automorphisms, and its closure does not contain s. 15.1.2. LEMMA. The Hilbert algebra A of G is equal to L 2(G). Let (1;) be the family of (mutually orthogonal) minimal projections of OU(A) n V(A). Then A is the direct sum of the Ai = F/(A), the Ai are self-adjoint two-sided ideals of A, which mutually annihilate one another, and each Ai is isomorphic to the Hilbert algebra of Hilbert-Schmidt operators on a finite-dimensional Hilbert space. We have
A;;;2L
1(G)nL 2(G)=e(G),

hence A=L2(G).

By 15.1.1 and 13.10.5, OU(A) and V(A) are finite von Neumann algebras. The lemma then follows from A 65 and A 68. 15.1.3. THEOREM. Let 1T be a continuous unitary representation of G. Then 1T is the direct sum of finite-dimensional irreducible representations.
tThroughout this chapter, G denotes a compact group. The chosen Haar measure on G will always be that of total mass I. We have L (G):J U(G), and

llflb;;.Ilfll,

for every f

E U{G).

320

REPRESENTATIONS

OF COMPACT GROUPS

[CH. 15, I

He

We retain the notation 15.1.2. Let H, be the closed subspace of H" generated by the 1T(j)~(f E Ai>~ E H",). Since A = L 2(G) is dense in L1(G), the sum of the H, is dense in Hi; If H #- 0, there exists an i such that Hi~ 0. For j~ i, we have AjA i = 0, hence 1T(A j)1T(A i )(H = 0, and so 1T(A j )(H i ) = 0. Let ~ be a non-zero element of Hi The finite-dimen sional subspace 1T(Ai)~ is equal to 1T(A)~, and therefore to 1T(LI(G~; it is non-zero and invariant under 1T( G). This subspace contains a non-zero subspace, invariant under 1T( G), such that the corresponding subre presentation of G is irreducible (2.3.5). Let, then, (Ka ) be a maximal family of mutually orthogonal finite dimensional subspaces of H"" which are invariant under 1T(G) and such that the corresponding subrepresentations of G are irreducible. Then (ffia K a) is invariant under 1T(G), hence zero by the above and the maximality of the family (Ka ). Hence H = ffia K a.
Tr

Tr )

15.1.4. COROLLARY. Every irreducible continuous unitary representation of G is finite-dimensional. 15.1.5.

COROLLARY.

The C*-algebra of G is liminal.

This follows from 15.1.4, since, in a finite-dimensional Hilbert space, every linear operator is compact. 15.1.6. COROLLARY. Let (u a ) be the family of finite-dimensional ir reducible continuous unitary representations of G. Then a Ker (Ja = {e}.

This follows from 13.6.6 and 15.1.4. 15.1.7.

COROLLARY.

G is a projective limit of compact Lie groups.

Let (Pa) be the family of finite-dimensional continuous unitary re presentations of G. The family (Ker Pa) is decreasing filtering, and Ker o; ={e} (15.1.6), hence G is the projective limit of the G/Ker Pa isomorphic themselves to the Pa(G). Since dim Pa < +00, the unitary group of H pm is a Lie group; since Pa(G) is a closed subgroup of this group, Pa(G) is a Lie group.

15.1.8.

COROLLARY. Let 1T be a continuous unitary representation of G. (i) There exist cardinals nO(u E 0) such that 1T = ffiOEG nuo-. (ii) The subspace Hn,p of H", is independent of the decomposition of 1T

chosen: it is the closed subspace of H", generated by the spaces of the subrepresentations equivalent to a. (iii) The cardinals n; are independent of the decomposition of 1T chosen.

CH. 15. 2]

IRREDUCIBLE REPRESENTATIONS

OF A COMPACT GROUP

321

Let (I be a subrepresentation of 7T equivalent to (I (I E G). Then (I is disjoint from T for T E G, T=I- (I (5.2.2), and so H", is orthogonal to H n TT for T=I- (I (condition (iii) of 5.2.1), hence H". r;;, Hn"cn whence (ii). We have dim Hn.,u = ncr dim CT, so that (iii) is a consequence of (ii). The intrinsic nature of the Hn.,u will be revealed in a different way in 15.3.12. 15.1.9. With the notation of 15.1.8, we say that References: [710], [933], [995], [1095].
(I

occurs ncr times in

7T.

15.2. Irreducible representations of a compact group

15.2.1. By the foregoing, the study of the continuous unitary represen tations of G leads to that of the irreducible representations. These latter are clearly square-integrable, and we can therefore apply to them the results of Chapter 14. It is also to be noted that they are all subrepresen tations of the regular representation. 15.2.2. THEOREM. Let A = L 2(G) be the Hilbert algebra of G, (F;) the family of minimal projections of OU(A) n Y(A), A and P the left- and right-regular representations of G, Ai and Pi the subrepresentations of A and P defined by F;, a, the class of irreducible representations whose associated central projection is F;, and 5j = dim (Ij. (i) A is the direct sum of the Ai, P is the direct sum of the pj (ii) Ai is the direct sum of 5j representations of class (Ij, Pi is the direct sum of 5i representations of class ih We know that e(G) is the direct sum of the F;(L 2(G )) (15.1.2), whence (i), and (ii) is a special case of 14.3.5 (iii). 15.2.3. PROPOSITION. Let (I be an irreducible continuous unitary re presentation of G. Then the formal dimension d.; of CT is equal to dim (I. The algebra CT(X(G)) is irreducible in Hen and therefore equal to .P(H,,) since dim H; < +00. Let f E X(G) be such that CT(f) = 1. Let
g =f have

* f.

For every t E G, let g be the function s ~ g(tst-I) on G. We

CT(g) = CT(t)-ICT(f

* f)CT(t) =

1.

322

REPRESENTATIONS

OF COMPACT GROUPS

[CH. 15, 3

Let { be a unit vector of H(T By 14.4.3, we have

d~l(dim
whence d u 15.2.4.
=

u) =

d~1

Tr u(g) = (u(gt){

I~)

dt = 1,

dim a,
For every f E L 2(G), we have If(sW ds

THEOREM.

J
G

= L. (dim u) Tr(u(f)u(f)*).
uEG

For every a E 6, let F; be the central projection in L 2( G) associated with a. By 15.1.2, we have
L
2(G)

=E9uEG Fu(L 2(G ,

=

hence Ilf"~

= L "Fufl ~
uEG

By 14.4.2 and 15.2.3, IIFJII~

(dim u) . Tr(u(f)u(f)*).

15.2.5. For every a E G, choose an orthonormal basis ({I, {2, .. , {<timu) of HU and put CP;j,fs) = (u(s){; I{j) Then, by 15.1.2 and 14.3.5 (i), the (dim U) 1I2Pjj,u(u E G, i, j, = 1,2, ... ,dim rr) constitute an orthonormal basis of L 2(G). We have, by 14.3.3 and 14.3.7,
T",}.u

m..

* m.,., ..,.....).0" =

(dim

u)-ISim.,. I..,... ,J,U

Pjj,u

* Pj,j.u= 0

if u-l- a,

References: [995], [1101], [1790].

15.3. Characters of compact groups

15.3.1. DEFINITION. Let a be an irreducible continuous unitary re presentation of G. The character of a, denoted by XU is the continuous function s --+Tr u(s) on G. The normalised character is the function
(dimur1xu

For compact commutative G, we recover the usual notion of charac ter. 15.3.2. The characterof a depends only on the class of a, If ({I... , {.) is an orthonormal basis of HU we have

CH. 15, 3j

CHARACTERS

OF COMPACT GROUPS

323

which proves that X" is a positive-definite function (13.4.5 (ii) equal to dim a at e. We will say a function on a group is central if it is invariant under the inner automorphisms of the group. It follows immediately from de finition 15.3.1 that X" is central. 15.3.3. By 15.2.5, if a E
(I)

and a E

are inequivalent, we have

(Xu IXu) = 0

(2)

x" * X,, = 0,
(X"

and moreover, keeping the notation of 15.3.2,

* X,,)(s) = n-I[(u(S)~1

I~I) +... + (U(S)~n

I~n)]'

whence
(3)

in particular

(4)

Ilx"lb= I.

15.3.4. Let K; be the biinvariant subspace of L 2(G) associated with a, As a Hilbert algebra, K" is isomorphic to the algebra of Hilbert-Schmidt operators on a finite-dimensional Hilbert space (15.1.2). This algebra has a l-dimensional centre. Now X" = = Xu is a central function on G belonging to K; (14.3.1). Hence the centre of K; is CX .... By 15.1.2, the centre of the Hilbert algebra L 2(G) is the direct sum of the CXu In view of formula (4) of 15.3.3, the X" form all orthonormat basis of the centre

x"

of L 2(G).

15.3.5. By formula (3) of 15.3.3, the operator of convolution by (dim u)Xu in L 2(G) is a non-zero projection. Since Xu = X;, this pro jection is hermitian. It belongs to OU(A) n V(A) (A the Hilbert algebra of G), and its set of values is contained in K o- Since P Ka is a minimal projection of OU(A) n V(A), the operator of convolution by (dim u)Xu is P Ka For every f E L 2(G), we therefore have

f = L (dim
uEO 2(G).

u)(f

* X,,),

the series converging in L If f is central, this expansion reduces to the expansion of f with respect to the orthonormal basis (X,,) of the

324

REPRESENTATIONS

OF COMPACT GROUPS

[CH. 15. 3

centre of L 2(G):
f

=L
uEG

(f IXu)Xu-

Let f be a central continuous complex-valued function on G. In view of 15.1.1, f is the uniform limit over G of functions of the form f * g, where g is central and in L 2(G). Since f and g are limits in L 2(G) of finite linear combinations of characters of G, we see that every central continuous complex-valued function on G is the uniform limit over G of finite linear combinations of characters of G. If we take for G the group of real numbers modulo I, these facts reduce to well-known facts about Fourier series. 15.3.6. For every finite-dimensional continuous unitary representation 17 of G, we can define the character X7T of 7T as in 15.3.1. Let 17 = n JO"I EB ... EB n.o; be the decomposition of 17 into irreducible represen tations, the 0"; being mutually inequivalent, and the n; being integers. We have
(X7T

IXu) = ni(Xuj IXu) = n;

Knowledge of X7T thus allows us to recover the ni In other words, two finite-dimensional continuous unitary representations of G are equivalent if and only if they have the same character. If 7T and 17 are two finite-dimensional continuous unitary represen tations of G, we immediately have

Henceforth, when we speak of characters of a compact group, we will only mean characters of irreducible representations. 15.3.7. PROPOSITION. Let /.L be a central measure on G (i.e. /.L com mutes, in the sense of convolution, with every measure on G). Then /.L ;p 0 if and only if /.L(X) ~ 0 for every character X of G.
If /.L;P 0, we have /.L(X)

presentation of Conversely, suppose that /.L(X) ~ 0 for every character X. Let A be the left-regular representation of G. To prove that /.L;P 0, it is enough to prove that A(/.L) ~ 0 (13.7.4). Let A be the Hilbert algebra of G, (FJ the family of minimal projections of OU(A) n V(A), K, = Fj(L2(G. Since /.L is central, A(/.L) E OU(A) n V(A), and A(/.L) reduces on each K, to a scalar,

= (dim O")/.L(X * X), where (T denotes character X. Now /.L(X * X) = /.L(X * X*) ~ O.

a re

CH. 15. 3j

CHARACTERS OF COMPACT GROUPS

325

Ai Let XI be the character belonging to K/. We have (denoting by a, a representation of character xJ:
A/(x/ IXi) = (A(J.L)X/ IXi) = (J.L

= (dim

Uir(J.L

* Xi IXi) = (J.L * Xi * xf)(e) * Xi)(e) =(dim UirJ.L<Xi) ~ o.

15.3.8. PROPOSITION. Let f E L 2(G) be a central function on G. Let f = ~uE6 Aux", with ~ A~ < +00 (15.3.4). Then f is continuous positive definite if and only if Au ~ 0 for every U E G and ~uE6 Au(dim u) < +00. The condition is sufficient because Xu is positive-definite and dominated in absolute value by dim o: Suppose that f is continuous positive-definite. We have f = g * g* for some g E L 2(G) (13.8.6). For every U E G, let K; be the biinvariant subspace of L 2(G) associated with a. We have

with uEG Since gO * g:. = 0 for U~ a and Ilu * vii.,..,; Ilulbllvib for any u, v E L 2(G ), f is the sum of the uniformly convergent series ~uE61J.LuI2(gu * g:). Furthermore, gO * g: E K", so that lJ.LuI 2(gu * g:) is the orthogonal projection of f on K"., and is therefore central. We can suppose that gO is central if 1--0 = O. Hence gO * g: is central, and
(gO

lJ.LuI < +00.

* g:)(e) = I gull~

= 1.

By 15.3.4, gO * g: cludes the proof.

=(dim

urX", so that Au

= lJ.LuI 2(dim U)-I,

which con

15.3.9. COROLLARY. In the convex set of central continuous positive definite functions f on G such that f(e)..,; 1, the extreme elements are 0 and the normalised characters. 15.3.10. PROPOSITION. Let !/J be a continuous complex-valued function on G. Then !/J is a normalised character if and only if !/J~ 0 and (1)

!/J(xsx-1t) dx

=!/J(s)!/J(t)

for any s, t E G.

326

REPRESENTATIONS OF COMPACT GROUPS

[CH. 15, 3

Suppose that there exists a E (] such that l/J = (dim U)-IXo- Let

u(xsx-

1)

dx E 5t(Hu )

We have, for every t E G,

u(t)S

u(txsx-

1 )

dx

u(xs(C1xr

1)

dx

= Su(t).

Since

tr

is irreducible, S is scalar. Moreover, Tr S

Xu(xsx-

1)

dx =

X,,(s) dx

=X,,(s).

Hence
(2)

u(xsx-

1)

dx

= (dim

urIXu(s) 1.

It follows from this that

u(xsx-1t) dx

(J

u(xsx- I) dx )u(t) = (dim U)-IXu(S)U(t).

Taking the trace of each side, we obtain (I). Now suppose that l/Jyf 0 and that l/J satisfies (I). Since l/Jyf 0, there exists an irreducible continuous unitary representation a of G such that u(l/J) yf O. We have
l/J(s)ii(l/J)

=
=

J JJ

l/J(s)l/J(t)ii(t) dt =

JJ

l/J(xsx-1t)u(t) dxdt

l/J(t)u(XS-1x-1t) dx dt

=
By (2),

l/J(t)(J ii(XS-1X- I ) dx )u(t) dt.

l/J(s)ii(l/J)

Jl/J(t)(dim

ur1)X..{s-l)ii(t) dt

= (dim u)-IX,,(s)ii(l/J),
whence l/J = (dim U)-I Xo:

CR. 15, 4j

REPRESENTATIONS

OF FINITE GROUPS

327

15.3.11. LEMMA. Let tr and a be irreducible continuous unitary re presentations of G. If o: and a are inequivalent, then u(Xa) =O. If a and a are equivalent, then u(xa) = (dim U)-I 1.

Let A be the left-regular representation of G. Then Adim u)Xa) is the projection on the biinvariant subspace K" of L 2(G) associated with a (15.3.5). Moreover, we can suppose that a is a subrepresentation of A (15.2.1), and R", is orthogonal to K" or contained in K" according as a is inequivalent to a or is equivalent to a (14.2.3).
15.3.12. THEOREM. Let 7r be a continuous unitary representation of G, and tt = EB"EO np its decomposition into irreducible representations. For every a E G, the orthogonal projection of RTf onto Rn.,u is 7rdim u)Xa).

This follows at once from 15.3.11. References: [635], [995], [llOl], [1790]. 15.4. Representations of finite groups
15.4.1. PROPOSITION. Let G be a finite group of order n. Let r be the number of conjugacy classes in G. Then G possesses r elements, and ~qE6 (dim ul = n.

For every a E G, let K" be the biinvariant subspace of L 2(G) as sociated with a. We have dim K q

=(dim ul

and

n = dim

2(G)

=L
<TEO

dim K"

by 15.2.1 and 15.2.2. Hence ~qEO (dim U)2 = n. Moreover, the number of elements of G is equal to the number of characters of G, i.e. to the dimension of the centre Z of L 2(G) (15.3.4); the characteristic functions of the conjugacy classes in G form a basis of the vector space Z; hence dim Z= r.
Ut,

15.4.2. Let C" C z, . ,C r be the distinct conjugacy classes in G, and Uz, ... , o; the distinct elements of G. The character table of G is the matrix (xOt"i,j,,;;;n where xl is the value that the character of Uj takes on C; Let h j be the number of elements of C j By formulas (1) and (4) of 15.3.3, we have

(1)

1~ -;;f

hiXix =

" {O1

if j~ if j = j.

r.

328

REPRESENTATIONS

OF COMPACT GROUPS

[CR. 15, 4

In other words, the matrix (v(h;/n)xf) is unitary. Consequently

or

(2)

..!. ""

~ XiXi l/h if n i l

j_j _

{O

if

I -

.- I .,.

i- r,

15.4.3. The following proposition calls for very different methods.

PROPOSITION. The dimension of an irreducible representation of G divides the order of G.

We retain the notation of 15.4.2. Let d j = dim a). (a) Let s E C; Then x! = Tr 7Tj(S) is the sum of the eigenvalues of 7Tj(S). Since 7Tj(S) is of finite order, these eigenvalues are roots of unity. Hence x! is an algebraic integer. (b) Let t. be the characteristic function of C; We have fi * fi = ~~'=. cin,fi where Ciii is the number of ways an element of Ci, can be written as the product of an element of Ci and an element of Ci ,. Hence the CWi are rational integers. Moreover, for every a E G, we have
u(/;)u(fi) =

;"=]

L CWi,U(fi)

Identifying U(fi), U(fi), U(fi) with scalars, we see that u(f;) is an eigenvalue of the matrix (CWi)."i,iand is therefore an algebraic integer. Now
Uj(fi)

"" =-1 Tr Uj(f;) =-1 ~

dj dj
sEC I

Tr

Uj(s)

hi . =xi.
dj

Hence (hddj)x! is an algebraic integer. (c) By formula (1) of 15.4.2,

n "" (hi .) ._j d. = ~i d. xi Xi
} }

is an algebraic integer. However, it is also a rational number and is therefore a rational integer. There is a considerable literature on the representations of finite groups. See, for example, M. Hall, The theory of groups, The MacMillan Company, New York, 1959.

CH. 15, 5]

USE OF COMPACT SUBGROUPS OF ARBITRARY

GROUPS

329

15.5. Use of compact subgroups of arbitrary groups

15.5.1. Let G be a locally compact group, and G a compact subgroup of G, If f E L(G), we can identify f with a measure on G, and therefore with a measure of compact support on G. This is what we will do in this section.
LEMMA. Let 7T be an irreducible continuous unitary representation of G, p = 7T IG, and p = EBuEG np the decomposition of p into irreducible representations. Suppose that nu is finite for every a E G. (i) If X is a character of G, and if f E LI(G), then 7T(f * X) is an operator of finite rank. (ii) The set of linear combinations of the f * x(f E L 1(G), X a character of G) is dense in L 1(G). (iii) We have 7T(C*(G))= ~(H1T)'

Let a E G. By 15.3.12, p(Xu) is of rank nu(dim u). Hence, for every f E L 1(G), 7T(f * Xu) = 7T(f)P(Xa) is of finite rank. Whence (i). There exist in G arbitrarily small neighbourhoods Vi of e which are invariant under the inner automorphisms of G. Let ({)j be the charac teristic function of Vi in G. By 15.3.5, ({)i is a limit in L 2(G), and therefore in M(G), of finite linear combinations of characters of G. Now, if f E LI(G), f * (() can be made arbitrarily close to f in L 1(G). Whence (ii). Lastly, (iii) follows from (i), (ii) and 4.1.11. 15.5.2. THEOREM. Let G be a locally compact group, and G a compact subgroup of G. Suppose that, for every 7T E G and every a E G, a only occurs in 7T I G a finite number of times. Then G is liminal. This follows at once from 15.5.1 (iii). 15.5.3. LEMMA. Let r be a topological group, a a (finite)-d-dimensional representation of T, and n an integer >d. Then na does not admit a cyclic vector. In fact, we have no = a Q9 T, where T denotes the trivial n-dimensional representation of r. Every vector of H; Q9 H T may be written C= ~I Q9 Til +... + ~d Q9 Tid where (~I' ... , gd) is a basis of Her Let H be the subspace of H generated by Til" .. .iu. We have H # H and [(u Q9 T) (r)]() c Q9 H, and so is not a cyclic vector.

n,

T,

330

REPRESENTATIONS OF COMPACT GROUPS

[CH. 15, 5

15.5.4. We say that a topological group is linear if it admits a continuous injective linear representation (not necessarily unitary) in a finite-di mensional complex vector space. 15.5.5. Let n be an integer >0, and M; the algebra of complex n x n matrices. Denote by r(n) the least integer r such that we have the identity
(1)

L
uEiE,

e"-xu(I)"

Xu(r)

in the algebra M n (ct. 3.6.1). Recall that r(n + 1) > r(n).

LEMMA. Let G be a semisimple linear connected real Lie group, G a maximal compact subgroup of G, a E G, and d = dim CT. Let X be the character of tr and " = dX, so that " is an idempotent of the algebra JC(G), and w* JC(G) * "is a subalgebra A of JC(G). Then the identity (1) is satisfied in A, with r = r(d 2) .

Let T be a linear (not necessarily unitary) representation of G in a complex finite-dimensional vector space V; if ~ E V and f E V (the dual of V), the function s ~ (T(S )~, f) is called a coefficient of T. Let C be the set of linear combinations of the coefficients of finite-dimensional continuous linear representations of G. By considering the conjugate representation of such a representation and the tensor product of two such representations, we see that C is an involutive algebra of con tinuous functions on G (for the product (f, g)~ fg and the involution f ~ 1). Since G is linear, this algebra separates the points of G. By the Stone-Weierstrass theorem, every continuous complex-valued function on G is the uniform limit, over every compact subset, of elements of C. Hence, if f E JC( G) and f # 0, there exists gEe such that f f(s )g(s )ds # O. Consequently, if f = JC(G) and f# 0, there exists a finite-dimensional continuous linear (not necessarily unitary) representation T of G such that T(f) !0; and, since G is connected semisimple, we can suppose that T is irreducible. To prove the lemma, it is therefore enough to establish the following: let T be a finite-dimensional irreducible (not necessarily unitary) representation of G, and let r = r(d 2) ; then the identity (1) is satisfied in T(A). We can endow the space V of T with a Hilbert space structure invariant with respect to T I G (B 34). Let T I G =E9 pE6 np/J be the decom position of T I G into irreducible representations. Then T(") is the projection on the space of np (15.3.12), and is therefore of rank n"d.

CH. 15, 61

ADDENDA

331

Consequently, T(A) may be identified with a subalgebra of the algebra of nad x nad matrices, and the problem amounts to showing that niT";; d. There exists a connected soluble closed subgroup N of G such that every element of G is the product of an element of G and an element of N. By Lies theorem, there exists a non-zero ~ in V such that TIN leaves C~ invariant. Since ~ is a cyclic vector for T, we see that ~ is a cyclic vector for T IG. Hence np, which is a subrepresentation of T IG, admits a cyclic vector. Hence nrJ";; dim ii = d (15.5.3). 15.5.6. THEOREM. Let G be a semisimple linear connected real Lie group, and G a maximal compact subgroup of G. (i) If 7T E G and o E G, then r occurs at most (dim u) times in 7T I G. (ii) G is liminal. Let 7T E G, and let 7T IG =EBuEG nuU be the decomposition of 7T IG. Let a E G, d = dim tr, and E ~ H" be the space of nuU. By 15.5.5, the identity (1) is satisfied in the algebra P1T(J{(G))P E, with r = r(d 2) . Now, since 7T is irreducible, 7T(J{(G)) is strongly dense in 2(H7I). Hence P1T(J{(G))P E is strongly dense in P~(H7I)PE' from which it follows that the identity (1) is satisfied in 2(E), with r = r(d 2) . Since rtd? + 1) > r(d 2) , E cannot contain any subspace of dimension d 2 + 1. Hence dim E,,;; d 2 and, consequently, nu ";; d. This proves (i), and (ii) follows from (i) and 15.5.2. References: [641], [758], [1507].
15.6. Addenda

15.6.1. Every continuous unitary representation of Gt that admits a cyclic vector is a subrepresentation of the regular representation. 15.6.2. A continuous complex-valued function f on G is said to be almost invariant if the linear combinations of its translates Jb form a finite-dimensional vector space. The coefficients of finite-dimensional representations are such functions. Every continuous complex-valued function on G is a uniform limit of linear combinations of these coefficients. 15.6.3. Let H be a Hilbert space, and 7T a representation in H of the involutive subalgebra J{(G) of L1(G). There exists exactly one contThe letter G denotes a compact group.

332

REPRESENTATIONS

OF COMPACT GROUPS

[CR. IS, 6

tinuous unitary representation 1T of 1T(S)!(S) ds for every! E X(G).

G in H

such that 1T(j) =

15.6.4. Let G be a compact group, and P a continuous positive-definite function on G. There exists a sequence PIP2,... of pure continuous positive-definite functions on G, and constants AI \2, ... ~ 0 such that

(where the series converges uniformly over G). 15.6.5. All the compact connected real Lie groups are known. For such a group, the list of irreducible continuous unitary representations is known as well: these representations may be classified using a finite number of integers. The simplest of the non-commutative compact connected real Lie groups is the group G of rotations in three-dimen sional space; if (] is its universal covering (with two sheets), then (] admits a 2-dimensional irreducible continuous unitary representation 1T, and the other irreducible continuous unitary representations of (] are the symmetric tensor powers of 1T. We do not have anything like a complete list of all finite groups, so that the results are less complete in this case. However, many particular results are known.

CHAPTER 16

ALMOST -PERIODIC FUNCTIONS

In this chapter, G denotes a topological group. We shall see that investigation of the finite-dimensional continuous unitary represen tations of G leads to that of the finite-dimensional continuous unitary representations of a compact group L associated with G. In truth, even for very simple groups G, the group L is often complicated, so that this will not give us a practical method for the determination of the finite dimensional continuous unitary representations of G. However, the properties of Chapter 15, applied to L, will imply remarkable properties of a certain class of functions on G, called almost-periodic functions.
16.1. The compact group associated with a topological group

16.1.1. THEOREM. Let G be a topological group. There exists a compact group L and a continuous morphism a: G ~ L possessing the following property: for every compact group I and every continuous morphism a':G~L', there exists a unique continuous morphism {3:I~I' such that a = {3 0 a. Furthe-rmore, the pair (I, a) is determined up to isomorphism by this property. Let (lTj) be the family of finite-dimensional continuous unitary re presentations of G. Let U, be the unitary group of H",. V the (compact) product group of the Ui a the continuous morphism s ~('lTi(S of G into U, and I the closure of a(G) in U, which is a compact subgroup of U. We show that (I, a) possesses the property of the theorem. Let I be a compact group and a: G ~ I a continuous morphism. The uniqueness of the continuous morphism {3 : L ~ I such that a = (3 a is immediate since a(G) is dense in I. We prove the existence of {3. The intersection of the kernels of the finite-dimensional unitary representations of I reduces to e (15.1.6); hence I may be identified with a subgroup of a group IIiVi where each Vi is the unitary group of a finite-dimensional Hilbert space; then a may be identified with a morphism s ~(Pi(s,
0

334

ALMOSTPERIODIC FUNCTIONS

[CH. 16, I

where each Pi is equivalent to one morphism {3i of ~ tinuous morphism

~'.

a continuous morphism of G into Vi Since Pi is of the representations 7Tj, there exists a continuous into Vi such that Pi = {3i a. The {3j define a con {3 of ~ into II Vj such that a =(3 0 a, whence /3(~) ~
0

The uniqueness of the pair (~, a) up to isomorphism is an easy general property of universal problems: let (~\, a\) be another solution; there exist continuous morphisms /3 :~~~\, /3\:~\~~, such that a\ = {3 0 a, a = /3\ a\, whence a = ({3\ /3) a; hence {3\ {3 is the identity map ping of ~ (by the uniqueness of (3 when a and a are given); similarly, /3 {3\ is the identity mapping of ~'; hence (3 is an isomorphism of ~ onto ~\, which transforms a into a\.
0 0 0 0 0

16.1.2. DEFINITION. The group ~ is called the compact group associated with G, and a is called the canonical morphism of G into ~. We thus have a(G) =~. With the notation of 16.1.1, (3(~) is the closure of a(G). If G is compact, we can take for ~ the group G and for a the identity mapping of G. 16.1.3. PROPOSITION. For every finite-dimensional continuous unitary representation P of ~, let P =P 0 a, which is a finite-dimensional con tinuous unitary representation of G. We thus define a bijection of the set of classes of finite-dimensional continuous unitary representations of L onto the set of classes of finite-dimensional continuous unitary re presentations of G.
If P is equivalent to pi, then P is equivalent to PI (because a(G) = ~), and so the mapping considered is injective. Moreover, if 7T is a finite dimensional continuous unitary representation of G, there exists (l6.1.1) a continuous unitary representation P of ~ in H." such that 7T = P 0 a, and so the considered mapping is surjective.

16.1.4. When G is a commutative locally compact group, the pair (~, a) is given by the following proposition:
PROPOSITION. Let G be a commutative locally compact group, and G its dual locally compact group. Let G be the commutative compact group whose dual G is the group G endowed with the discrete topology. Let lp be the continuous morphism of G into G whose dual cP is the identity morphism of G into G. Then G may be identified with the

CH. 16, 2j

ALMOSTPERIODIC FUNCTIONS

335

compact group associated with G, and P with the canonical morphism of G into G.

If X E G is trivial on P(G), then 4>(x) is trivial on G, and so 4>(X) = 1 and X = 1. This proves that P(G) = G. This established, let H be a compact group and !/J a continuous morphism of G into H. We are going to prove that there exists a unique continuous morphism {3 of G into H such that !/J = {3 tp ; this will confirm the proposition. Replacing H by !/J(H), we are led to the case where H is commutative and compact. Then the dual (f, of !/J is a morphism of the discrete group fI into G. We are reduced to proving that there exists a unique morphism ~ of fI into G such that (f, = 4> ~; now this is plain since 4> is the identity mapping of G onto G.
0 0

References: [995], [1790].

16.2. Almost-periodic functions

16.2.1. THEOREM. Let G be a topological group, ~ the associated com pact group, a the canonical morphism of G into ~, E(G) the vector space of bounded continuous complex-valued functions on G endowed with the uniform norm, and f E E(G). Then the following conditions are equivalent: (i) The set of the J (s E G) is relatively compact in E(G); (ii) The set of the t, (t E G) is relatively compact in E(O); (iii) The set of the Jr (s, t E G) is relatively compact in E(G); (iv) There exists a continuous complex-valued function g on ~ such that f = goa; (v) f is the uniform limit over G of linear combinations of coefficients of finite-dimensional irreducible continuous unitary representations of G. (Since E(G) is complete, we can replace "relatively compact" by "precornpact" in (i), (ii) and (iii. (iv) (iii): Let g E ,'J{(~) and f = goa. The mapping (CT, r) ~.,.gT of ~ x ~ into the Banach space ,'J{(~) is continuous, and so the set of the UgT is compact. The image of this set in E( G) under the isometric mapping h ~ h a is compact and contains the set of the Jr (s, t E G). (iii) (i) and (ii): Obvious. (i) => (iv) (the proof of (ii) => (iv) is analogous): For every g E E( G) and every s E G, put !/J(s)g = s-Ig. Then !/J(s) is a linear isometry of E(G)

=>
0

=>

336

ALMOST-PERIODIC FUNCTIONS

[CH. 16, 2

onto B(G), and "(ss) = "(s)"(s). Let A be the set of the J (s E G). It is clear that each "(s) induces a bijection of A onto A, and hence also a bijection q>(s) of A onto A. The q>(s) are isometric, and A is compact if we suppose that condition (i) is satisfied. Hence q> (G) is relatively compact in the set ceCA, A) of continuous mappings of A into itself, endowed with the topology of uniform convergence (Ascolis theorem); hence the closure of q>(G) in ceCA, A) is a compact group r of homeomorphisms of A (B 16). The mapping q> is a morphism of G into r. We show that it is continuous. It is enough to prove that, when s ~ e, q>(s) converges to q>(e) for the topology of uniform convergence over A, or even just for the topology of pointwise convergence over A (which comes to the same thing, because cp(G) is equicontinuous). Therefore, let g E A, and let V be an open neighbourhood of g in A. Then A\ V is compact. If h E A\ V, we have Ig(u) - h(u)1 > 0 at at least one point u of G, and so there exists an open neighbourhood V h of h in A and a neighbourhood W h of e in G such that cp(s)g~ V h for s E Who We can cover A\ V with a finite number of neighbourhoods V h I , , V h ; then, for n S E W hl n .. n Whn we have q>(s)g~ V hl U .. U Vhn, and so q>(s)g E V, and we have indeed proved that q> is a continuous morphism of G into r. By the definition of ~ and a, there exists a continuous morphism {3 of ~ into r such that cp = (3 0 a. The function u~({3(u)f)(e) on ~ is con tinuous, and
I(s-I)

= (q>(s)f)(e) = ({3(a(sf)(e),

hence condition (iv) is satisfied. (iv) =} (v): Suppose that 1= goa, where g E J(~). Then g is the uniform limit over ~ of linear combinations of coefficients of (finite dimensional) irreducible continuous unitary representations of ~ (13.6.5 and 15.1.4). Hence I is the uniform limit over G of linear combinations of coefficients of finite-dimensional irreducible continuous unitary re presentations of G. (v) =} (iv): Suppose that, for every n, there exists a linear combination In of coefficients of finite-dimensional irreducible continuous unitary representations of G, such that III- in II", :s;; iln. There exists a function gn on ~, a linear combination of coefficients of irreducible continuous unitary representations of ~, such that in = gn 0 a (16.1.3). We have Ilgm - gnll", =Ilim - inll",~O as m and n tend to +00, and so the gn converge uniformly to a continuous function g on ~, and we have i = goa. 16.2.2.
DEFINITION.

An almost-periodic

function on G is a bounded

CH. 16, 3j

THE MEAN OF AN ALMOST-PERIODIC FUNCTION

337

continuous complex-valued function on G which satisfies the equivalent conditions of 16.2.1. We denote the set of these functions by AP(G). 16.2.3. By condition (iv) of 16.2.1, a function of AP(G) is uniformly continuous for the right and left uniform structures of G. The sum, product, etc. of two almost-periodic functions is almost-periodic; AP(G) is closed in E( G). References: [995], [1790].
16.3. The mean of an almost-periodic function

16.3.1. THEOREM. Let f E AP(G). Let K be the closed convex hull in E(G) (with the notation of 16.2.1) of the set of the J, where s runs through G. Then K contains exactly one constant M(f). If f is the function on ~ corresponding to f, we have M(f) = f};f(s) ds (Haar measure of total mass 1 being chosen on ~). Thanks to the canonical isomorphism of AP(G) onto J{(~), we are led to the case where G = ~. Therefore, let f E J{(~). Let E > O. There exists a neighbourhood V of e in ~ such that st" E V implies If(s) - f(t)1 ~ E. Let (Vsj)t",j"n be a finite covering of S, and let (h;)l",j"nbe a partition of unity subordinate to this covering. Putting c, = f};hj(s) ds, we have

If

f(s) ds - }: C!(Sjt)

I= I~ f

h;(s)U(st) - f(sl)] dsl

and this is dominated by E because If(st) - f(sl)1 ~ E when h;(s) -# O. Moreover, ~ Cj = 1. Hence the constant f f(s) ds belongs to K. Lastly, all the translates of f have the same integral over ~, and hence all the functions of K have the same integral over ~; if a constant is in K, its value must, therefore, necessarily be f f(s) ds. 16.3.2. DEFINITION. The constant M(f) of 16.3.1 is called the mean of f. 16.3.3. The mapping f ~ M(f) is a linear form on AP(G). We have M(/) ~ 0 for f ~ 0, M(1) = I, M(J) = M(f,) = M(f) for every s E G. All this follows from the equality M(f) = J!f(s) ds of 16.3.1. 16.3.4. For I. g E AP(G), put (f Ig) = M(fg). Then AP(G) becomes a Hausdorff pre-Hilbert space, canonically isomorphic to J{(~) regarded as a subspace of L2(~). Let C be the set of classes of finite-dimensional

338

ALMOST-PERIODIC FUNCTIONS

[CR. 16, 4

irreducible continuous unitary representations of G; for every a E C, choose an orthonormal basis (EI."" Edimu) in H u ; put Pi,j,u(S)= (u(s)ei IEj) Then the (dim U)!/2Pij,u (u E C, I ~ i, j ~ dim u) constitute an orthonormal basis in the pre-Hilbert space AP(G): this follows from 15.2.5 and 16.1.3. References: [995], [1790].

16.4. Groups injectable in a compact group

16.4.1. LEMMA. The following three subgroups of G are equal: (1) the kernel of the canonical morphism G ~ L; (2) the intersection of the kernels of the continuous morphisms of G into all compact groups; (3) the intersection of the kernels of the finite-dimensional irreducible continuous unitary representations of G. Denote these three sets by N I , N 2 , N 3 , We have N 2 :1 N, by definition of L and of the morphism G ~ L. It is clear that N 3 :1 N 2 Finally, if s E N 3 , its canonical image in L belongs to the kernel of every ir reducible continuous unitary representation of L (16.1.3), and is there fore equal to e (13.6.6); hence N 3 k Nt. 16.4.2. DEFINITION. A topological group is said to be injectable in a compact group if the subgroup of 16.4.1 reduces to e. 16.4.3. A subgroup of a group injectable in a compact group is itself injectable in a compact group. A product of groups injectable in a compact group is injectable in a compact group. Hence a projective limit of groups injectable in a compact group is injectable in a compact group. The quotient of any topological group by the normal subgroup of lemma 16.4.1 is injectable in a compact group. A commutative locally compact group is injectable in a compact group since the subgroup defined by condition (3) of 16.4.1 reduces to e. 16.4.4. PROPOSITION. Let G be a topological group generated by a compact neighbourhood of e, and injectable in a compact group. Then G is the projective limit of Lie groups locally isomorphic to compact groups. Let [Ji be the filter base on G consisting of the kernels of the finite-dimensional continuous unitary representations of G. Since G is

CH. 16. 4]

GROUPS INJECTABLE IN A COMPACT GROUP

339

injectable in a compact group, the intersection of the elements of fiP is {e}. By hypothesis, there exists a compact symmetric neighbourhood U of e generating G. The trace of fiP on the compact set U 3 converges to e. There therefore exists No E fiP such that U 3 n No U. For N E N ~ No, put

s.

K N = U 3nN

c U.

The K N possess the following properties: (a) K N is a compact subgroup of G, because KNK;/ ~ U 2 n N ~ K N. (b) K N is normal in U" = G, because, for every s E U, we have
sKNs
t

cU

nN

K N;

since n K N = {e}, G is the projective limit of the GIKN. (c) K N is open in N, hence NIK N is a discrete subgroup of G/KN; consequently, GIK N is locally isomorphic to GIN; now GIN admits an injective finite-dimensional continuous unitary representation, and is therefore a Lie group. Furthermore, the Lie algebra 9 of G/N is isomorphic to a subalgebra of a compact Lie algebra; consequently, 9 is the product of a commutative Lie algebra and a compact Lie algebra, and is therefore the Lie algebra of a compact group C; and GIK N is locally isomorphic to C. 16.4.5. LEMMA. Let G be a connected locally compact group having a base of neighbourhoods of e invariant under inner automorphisms. Let A and B be closed subgroups of G such that: (1) A is contained in the centre of G; (2) A ~ B, and BIA is contained in the centre of GIA. Then B is contained in the centre of G. Let x E B, y E G, let X be a character of the commutative group A, and N be the kernel of X. We have xyx-1y-1 E A. We are going to show that X(xyx-1y-l) = 1. Since X is arbitrary, this will prove that xyx-1y-t = e, and the lemma will be proved. There exists a neighbourhood V of e in G such that Ix(u) - 11 ~ 1 for u E V n A, and a neighbourhood V of e in G, invariant under inner automorphisms, such that V V,-I ~ V. For any z E V, xzxlzI is an element a of A, and the relation xzr " = az leads, by recurrence, to xnzx- n = at z. Whence an = (xnzx- n) . Z-I E V . V,-I ~ V, and consequently an E V n A, Ix(a n ) - 11 ~ 1. Since this is true for any n, we deduce from it that x(a) = 1, i.e. that x and z commute modulo N.

340

ALMOST-PERIODIC FUNCTIONS

[CH. 16. 4

Since G is connected, y is a product of elements of V, and so x and y commute modulo N.

16.4.6. THEOREM. Let G be a connected locally compact group, and Z its centre. The following properties are equivalent: (i) G is injectable in a compact group; (ii) G is the product of a compact group and a group R"; (iii) G is the projective limit of Lie groups locally isomorphic to compact groups; (iv) G/Z is compact; (v) There exists in G a base of neighbourhoods of e invariant under inner automorphisms. Furthermore, if G possesses these properties, every connected group locally isomorphic to G also possesses them.
(i) ~ (iii): This follows from 16.4.4. (iii) ~ (ii): Suppose condition (iii) is satisfied. There exists in G a filter base converging to e, consisting of closed normal subgroups N such that G/N = GN are Lie groups locally isomorphic to compact groups. Fur ther, G N is connected. Let G~, G~ be the connected normal closed subgroups of G N corresponding to the centre and to the derived algebra of the (reductive) Lie algebra of G N ; then G N = G~. G~, G~ and G~ commute, and G~ is compact. If N C N, the canonical mapping of GN , onto G N maps G~, onto G~, and G~, onto G~. Let, then, A C G be the projective limit of the G~, and KeG be the projective limit of the G~; A is connected and commutative, K is compact, A and K commute, and G = AK. We have A = R" x K], with K] compact (B 25). Hence G = R nK 2, with a compact subgroup K 2 commuting with R". Necessarily, K 2 n R" = {e}, and so G is isomorphic to R" x K 2 (ii) ~ (v): This follows from 15.1.1. (v) ~ (iv): Suppose condition (v) is satisfied. For every s E G, let j(s) be the inner automorphism of G defined by s. The hypothesis implies that the right and left uniform structures of G coincide and that j( G) is an equicontinuous group of homeomorphisms of G. There exists a compact neighbourhood U of e invariant under inner automorphisms. For every s E G, there exists an integer n > 0 such that s E U" since G is connected; the conjugacy class of s is then contained in U", and is therefore relatively compact. Let ;y be the topology of compact con vergence on the set Ci(G, G) of continuous mappings of G into itself. By Ascolis theorem, j(G) is relatively compact in Ci(G, G) for ;Yo Hence

CH. 16, 5]

ADDENDA

341

j(O) is a group r of homeomorphisms of 0, compact for ff (B 16). Finally, j is continuous as a mapping of 0 into T, and so the injective morphism of 0/ Z into r obtained from j by passage to the quotient is

injective. Thus, O/Z is injectable in a compact group. By the already established implication (i):::}(ii), we conclude from this that 0/ Z = A x K with A commutative and K co~pact. By 16.4.5, A reduces to {e}, and so 0/ Z is compact. (iv):::} (i): If O/Z is compact, Z is generated by a compact neigh bourhood of e (B 24), and is therefore of the form K x L with compact and L = R" x ZP (B 25). Let s be an element of 0 distinct from e. We are going to prove, and this will establish (iv):::}(i), that there exists a continuous morphism rp of 0 into a compact group such that rp(s) 1- e. There exists a closed subgroup M of L such that s M and LIM is compact. Then 0/ M, which is the extension of 0/ Z by K x (LIM), is compact, and it suffices to take for rp the canonical morphism of 0 onto

O/M.
Lastly, if 0 possesses property (v), every connected group locally isomorphic to 0 also possesses it. Instead of "injectable in a compact group", one also finds "represen table in a compact group", or "maximally almost-periodic". References: [119], [1790]. 16.5. Addenda 16.5.1. A group injectable in a compact group can possess infinite dimensional irreducible continuous unitary representations. [635]. *16.5.2. Let 0 be a locally compact group. We reintroduce the notation of 13.11.6. (a) For rp, IjJ E E and x EO, the function t ~ 1jJ(t)rp(t-I X ) is an element W x of E. Put M(w x ) = (rp x 1jJ)(x). Then rp x IjJ E AP(O). (b) Every continuous positive-definite function may be written rpl + rp2 with rpl almost-periodic positive-definite, and rp2 positive-definite such that M(rp2cP2)=0. (c) Every almost-periodic positive-definite function may be written L A;rp; with Aj > 0, L Ai < + 00, and rpj a pure positive-definite function corresponding to a finite-dimensional representation. [635]. 16.5.3. (a) Let 0 be a locally compact group, and 0 0 the connected component of e. Suppose that 0/0 0 is compact. Then the following

342

ALMOST-PERIODIC FUNCTIONS

[CH. 16, 5

conditions are equivalent: (i) G is injectable in a compact group; (ii) Go is injectable in a compact group; (iii) G is the semi-direct product of a compact subgroup K and a normal subgroup V isomorphic to R", every element of V commuting with every element of the component of K containing the identity; (iv) G admits a base of compact neighbourhoods of e invariant under inner automorphisms. (b) There exist totally discontinuous locally compact groups, in jectable in a compact group, and not satisfying condition (iv) of (a). [951], [1086]. 16.5.4. Problem: prove the equivalence of conditions (i). (ii), (iv), (v) of 16.4.6 without using Lie groups in passing.

CHAPTER 17

CHARACTERS OF A LOCALLY COMPACT GROUP

17.1. Definitions 17.1.1. Let G be a locally compact group. A trace (resp. bitrace, charac ter) of G is a lower-semicontinuous semi-finite trace (resp. a maximal bitrace, a character) of C*(G). 17.1.2. A trace representation of G is a pair (7T, t) with the following properties: (i) 7T is a continuous unitary representation of G; (ii) t is a faithful normal trace on CU+ (CU denotes the von Neumann algebra generated by 7T(G; (iii) 7T(C*(G n m, generates the von Neumann algebra CU. The trace representations of G may be identified with the trace representations of C*(G). We define, in an obvious way, quasi equivalence of two trace representations (d. 6.6.2). A trace of G is canonically associated with every trace representation of G. 17.1.3. A trace f of G defines the objects mt, nt, At, H t, it, At, Pt CUt, "lit tt (6.2 and 6.4.5). The representation At of C*(G) may be identified with a representation of G that we will also denote by At. The represen tation p~ of C*( G) in H t is the transform of At by the isomorphism it of H, onto Ht (6.2.2); it may be identified with a representation of G, that we will also denote by p~, and which is the transform by it of the representation At of G; we therefore have p~ = At. A trace of G is said to be of type I, of type II, ... if At is of type I, of type II, etc. The above and 17.1.2 define a bijection of the set of traces of G onto the set of quasi-equivalence classes of trace representations of G. 17.1.4. Let 7T be a continuous unitary representation of G, and CU the von Neumann algebra generated by 7T(G). Then 7T is said to be traceable if there exists a trace t on CU + such that (7T, t) is a trace representation. The traceable representations of G may be identified with the traceable representations of C*( G).

u;

344

CHARACTERS OF A LOCALLY

COMP ACT GROUP

[CH. 17. 2

17.1.5. Let 7T be a continuous unitary factor representation of G, and [Ji the factor generated by 7T(G). The following conditions are equivalent: (i) 7T is traceable, (ii) f1Jf is semi-finite and 7T( C*( G contains a non-zero Hilbert-Schmidt operator relative to [Ji (6.7.2). Let f be a trace of G. The corresponding representation of G is a factor representation if and only if f is a character (6.7.3). This yields a canonical bijective correspondence between: (a) the set of characters of G defined to within a positive scalar factor; (b) the set of quasi equivalence classes of traceable factor representations of G (6.7.4). 17.1.6. Suppose that G is postliminal. Every irreducible continuous unitary representation 7T of G is traceable (6.7.5) and admits a nor malised character f." defined by f.,,(x) =Tr 7T(X) for x E C*(Gt. Let 7T be another irreducible continuous unitary representation of G; then 7T and 7T are equivalent, if and only if f." =f.", (6.7.6). Every characterof G is of the form Af." with A> 0 and 7T an irreducible continuous unitary representation of G (6.7.6). Let f = f.". Then At p~, OUt, may be canonically identified with H." 0 H."., 7T 0hi, I H _ 0 7T, 5E(H.".) 0 C, C 05E(H.,,); It may be identified with the ~ano~ical involution of H." 0 H.,,; tt may be identified with the trace T 0 1 ~ Tr T on

n;

v,

(5E(H.".)

cr (6.7.7).

Reference: [641].

17.2. The character defined by a measure, and by a distribution

17.2.1. LEMMA. Let A be a C*-algebra, A a dense involutive subalgebra of A, and s a complex-valued function on A x A satisfying the following conditions: (i) s is a positive hermitian sesquilinear form; (ii) s(y, x) = s(x*, y*) for x, y E A; (iii) s(zx, y) = s(x, z*y) for x, y, ZEA; (iv) s(zx, zx) :0;;; IIzI1 2s(x, x) for x, zEA; (v) The set of elements xy (where x, y E A) is total in A for the pre-Hilbert structure defined by s. Then there exists a unique maximal bitrace of A which extends s.
The uniqueness follows from 6.5.3. We prove the existence. Let N be the set of x E A such that s(x, x) = O. We verify as in 6.2.2 that N is a self-adjoint two-sided ideal of A and that A/N is a Hilbert algebra for the scalar product (I) obtained from s. Let the Hilbert space H be the

CH. 17. 2] THE CHARACTER DEFINED BY A MEASURE AND BY A DISTRIBUTION 345

completion of A/N. Let A be the canonical mapping of A into H. The hypothesis (iv) implies that there exists, for every zEA, a unique continuous linear operator A(z) on H, such that A(z)Ax = A (zx) for every x E A; furthermore, IIA(z)ll,,;; Ilz/I. It is immediate that A is a representation of A in H; it extends uniquely to a representation, again denoted by A, of A in H. Then A(A) and A(A) generate the same von Neumann algebra, namely OU(A/N). Let 8 be the natural trace on OU(A/Nr defined by the Hilbert algebra A/N. Then "e:2 A(A) and hence (A,8) is a trace representation of A; let SI be the corresponding maximal bitrace of A. For x E A, we have A(x) E "e, hence x E "SI and
SI(X,

x)

= 8(A(x)A(x)*) = sex,

x);

thus, s I extends s. 17.2.2. If G is a locally compact group, JC(G) is an involutive subalgebra of L 1(G), dense in LI(G) and therefore in C*(G).
PROPOSITION. Let G be a unimodular locally compact group, and JL a measure Oil G. For I, g E JC(G), put s(f, g) = JL(g* * f). Theil s satisfies conditions (i) to (v) 01 17.2.1, il and oniy il JL is a central positive definite measure.

It is clear that s is a sesquilinear form, and so condition (i) of 17.2.1 is satisfied if and only if JL(f* * f) ~ 0 for every I E JC(G), i.e. if and only if JL ~ O. Suppose from now on that this is the case. Condition (ii) is satisfied if and only if JL(g* * f) = JL(f * g*) for any I, g E JC(G), i.e. if and only if JL is central. Condition (iii) may be written JL(g* * (h * f) = JLh* * g)* * f) for any I, g, hE JC(G); it is automatically satisfied. Let IE JC(G) and let K be a compact neighbourhood of the support of I. There exists g E JC(G) such that I * g has support contained in K and III- I * glloo is arbitrarily small; it immediately follows from this that condition (v) is satisfied. We saw in 13.7.9 that JL defines a unitary representation o: of G. The Hilbert space H" is obtained precisely by endowing JC(G) with the scalar product s(f, g) and then passing to the completion of JC(G) for this scalar product; also, denoting the canonical mapping of JC(G) into H, by A, we have u(x)AI =A (Ex * f) for every x E G and every IE JC(G). It follows from this that u(g)AI = A(g * f) for I, g E JC(G) and therefore
s (g

* I, g * f) = (u(g )AI 1aig )Af) :::;;; Ilu(g )11 2(AI

I Af)

= lIu(g )11 2S (f, f)

and condition (iv) is satisfied.

346

CHARACTERS OF A LOCALLY COMPACT GROUP

[CH. 17, 2

17.2.3. Let p., be a central positive-definite measure on G (assumed unimodular). By 17.2.1 and 17.2.2, the form (j, g)~ p.,(g* * f) on X(G) x X(G) extends to a unique maximal bitrace p., of C*(G). Let a be the representation of C*(G) (or of G) defined by p., and keep the notation tr of 17.2.2. By 6.3.6, we have Her = Her; moreover, for f E X(G), u(j) is given by the operator of convolution on the left by f in X(G), and so u(j) = atf). Consequently a = a. In fact, p., defines a trace represen tation [rr, t) of G; for f E X( G), we have u(j) E", and
t(u(j)u(j)*) = p.,(j,f) = p.,(j*

* f).

The mapping p., ~ u is injective since knowledge of p., on functions of completely determines u: the form g * f (j, g E X( We sometimes identify p." p., and the associated trace, and it therefore makes sense to say that a central positive-definite measure p., on G is a character of G. By the above, this is so if and only if the continuous unitary representation of G obtained from u. by the procedure of 13.7.9 is a factor representation. In particular, this gives a meaning to the statement that a central continuous positive-definite function cp on G is a character (we identify cp with the measure cp(x) dx).

17.2.4. We detail a special case which is important for applications.

PROPOSITION. Let G be a unimodular postliminal group, tr an ir reducible continuous unitary representation of G, and p., a central positive-definite measure on G. Then a admits the normalised character p." if and only if u(j) is a Hilbert-Schmidt operator for every f E X(G), and for f, g E X(G), we have

Tr(u(g )*u(j = p.,(g* * f). If this is so, a is quasi-equivalent to the representation of G obtained from p., by the procedure of 13.7.9.

17.2.5. Suppose that G is unimodular. The Dirac measure

E.

at the point

e is central and positive-definite. It defines a maximal bitrace s of

C*(G). Then X(G) is dense in n, for the pre-Hilbert structure of n.: for f, g E X(G), we have s(j, g) =f f(x)g(x) dx. The space H, is just L 2(G ), As is the left-regular representation, p~ is the conjugate of the right regular representation, and J s is the mapping f ~ f* of e(G) onto L 2(G); the full Hilbert algebra of nsf N, is the Hilbert algebra A of G; the von Neumann algebra OU s is OU(A); it is the von Neumann algebra generated

CH. 17. 2]

THE CHARACTER DEFINED BY A MEASURE AND BY A DISTRIBUTION

347

by the left-translation operators on L 2(G); 611; is endowed with a natural trace t; and (Ast s) is a trace representation of G. If f E L 1(G) n L 2(G), we have
ts(A.(f)As(f)*) < + 00;

and, for f,gEL 1(G)nL 2(G),

ts(As(f)As(g)*)

f(x)g(x) dx.

17.2.6. Let G be a real Lie group. Let f0(G) denote the set of infinitely differentiable complex-valued functions with compact support on G. This set is dense in L1(G), and therefore in C*(G). A distribution p, on G is said to be central if it is invariant under the inner automorphisms of G, or, which comes to the same thing, if p,(f * g) = p,(g * f) for any f, g E f0(G). A distribution p, on G is said to be positive-definite if f.L (f * f) ~ 0 for any f E f0 (G).
Let G be a unimodular real Lie group, and p, a distri bution on G. For t. g E f0(G), put s(f, g) = p,(g* * f). Then s satisfies conditions (i) to (v) of 17.2.1, if and only if p, is a central positive definite distribution.
PROPOSITION.

The proof, entirely analogous to that of 17.2.2, is left to the reader. 17.2.7. We retain the hypotheses of proposition 17.2.6. By 17.2.1, the form s extends to a unique maximal bitrace p, of C*(G), which defines a trace representation (0", t) of G. The representation 0" is obtained by endowing f0(G) with the scalar product s(f, g), completing with respect to s to obtain a Hilbert space, and making G act in 0)(G) by left translation. For f E 0)(G), we have O"(f) E n, and
t(u(f)u(f)*)
=

u(], f)

p,(f*

* f).

The mapping p, -,) u is injective. We sometimes identify p" p, and the associated trace and it therefore makes sense to say that a central positive-definite distribution on G is a character of G. In particular: 17.2.8. PROPOSITION. Let G be a unimodular postliminal Lie group, 0" an irreducible continuous unitary representation of G, and u. a central positive-definite distribution on G. Then 0" admits the normalised character p" if and only if O"(f) is a Hilbert-Schmidt operator for every

348

CHARACTERS OF A LOCALLY COMPACT GROUP

[CH. 17, 3

f E

~(G),

and, for

t. g E

~(G),

we have

Tr(u(g)*u(f

= JL(g* * f).

If this is so, a is quasi-equivalent to the representation of G obtained from JL by the procedure of 17.2.7.

References: [641], [1144].

17.3. Characters of finite type

17.3.1. PROPOSITION. Let G be a unimodular locally compact group, sa character of G, and (1T, t) a trace factor representation of G with which s is associated. (i) s is of finite type if and only if s is defined by a continuous function X on G which is central and positive-definite. (ii) If this is so, we have x(x) = t(1T(X, where t denotes the linear extension of t to the von Neumann algebra generated by 1T(G).

Now s is of finite type if and only if (6.8.1 and 6.8.5) s extends to a positive central linear form on C*(G). The positive central linear forms on C*(G) correspond bijectively with the positive central continuous linear forms on L1(G) (2.7.5), and thus also to the positive-definite central continuous functions on G. Hence, if s is of finite type, there exists a positive-definite central continuous function X on G such that s(f * /*) = tx.! * f*) for every f E X(G); thus, s is defined by x. The converse is plain. Suppose that s is defined by X. The function x .... t(1T(X on G is continuous. (In fact, t is normal, hence ultra-strongly continuous (A 25), hence strongly continuous on the unit ball (A I). For any t. g E X( G), we have
J t(1T(X(g*

* f)(x)

dx

= t(J 1T(X)(g* * f)(x) = t(1T(g* * f)

=

dX)

J X(x)(g*

* f)(x) dx,

and so the measures t(1T(X dx and X(x) dx are equal, and consequently t(1T(X = X(x) for every x E G. We will identify the characters of G of finite type with the central positive-definite continuous functions on G.

CH. 17, 3)

CHARACTERS OF FINITE TYPE

349

17.3.2. Suppose that G is compact. Every factor representation of G is a multiple of an irreducible representation 1T (15.1.8), and dim 1T < +00. By 17.3.1, every characterof G in the sense of 17.1.1 may be identified with a function s ~ A Tr 1T(s) on G, where A > 0, and where 1T is an ir reducible representation of G. To within the factor A, we recover definition 15.3.1. 17.3.3. Suppose that G is commutative. Every factor representation of G is a multiple of a I-dimensional representation. The characters of G in the sense of 17.1.1 may therefore be identified with the functions A . X, where A > 0 and where X is a character of G in the usual sense. 17.3.4. PROPOSITION. Let G be a unimodular locally compact group. There exists a canonical bijective correspondence between (a) The set of quasi-equivalence classes of continuous unitary factor representations of G of finite type; (b) The set of characters of G of finite type equal to 1 at e. This follows from 6.8.6. 17.3.5. PROPOSITION. Let G be a unimodular locally compact group, and C the set of central positive-definite continuous functions on G such that cp(e) ~ 1. (i) C is compact convex for the weak*-topology u(C(G), L 1(G )). (ii) The extreme points of Care 0 and the characters of G of finite type equal to 1 at e. (iii) C is the weak*-closed convex hull of 0 and the set of characters of finite type equal to 1 at e. This follows from 6.8.7. 17.3.6. PROPOSITION. Suppose that there exists in G a base of compact neighbourhoods of e invariant under inner automorphisms. (i) For every s E G different from e, there exists a character X of G of finite type such that x(s) ~ x(e). (ii) For every s E G different from e, there exists a continuous unitary factor representation 7T of G of finite type such that 1T(s) ~ I. Let s E G, s ~ e. There exists a symmetric compact neighbourhood Y of e, invariant under inner automorphisms, such that sg v. Let f be the characteristicfunction of Y. Then g = f * f* is a positive-definite central continuous function on G, equal to 0 at s and to A = I !I ~ > 0 at e. By

350

CHARACTERS

OF A LOCALLY

COMPACT GROUP

[CH. 17. 3

17.3.5, A" is the weak*-limit of linear combinations, with positive coefficients, hi of characters of finite type such that Ilhili oo ~ 1. Since IIA-lglloo~limllhjlla" we can suppose that Ilhill oo = 1 for every i; then hi convergesuniformly to A" on every compact set (13.5.2). We therefore have hj(e) hi(s) for some i, whence (i). Let X be a character of G of finite type such that X(s) X(e). Let (1T, t) be the corresponding trace factor representation of finite type. Let t denote the linear extension of t to the von Neumann algebra generated by 1T(G). We have

t(1T(S)) = X(s) X(e)

17.3.7. We However:

t(1T(e)), hence 1T(S) 1T(e) = 1.

do not know whether the converse of 17.3.6 is true.

PROPOSITION. (i) Let G be a locally compact group. Suppose that, for every s E G different from e, there exists a continuous unitary factor representation 1T of G of finite type such that 1T(S) 1. Then: (*) if K is a compact subset of G such that e~ K, there exists a neighbourhood of e not meeting K and invariant under inner automor phisms. (ii) If a locally compact group G satisfies (*) and is generated by a compact set, there exists in G a base of compact neighbourhoods of e invariant under inner automorphisms.

Let s be a point of G other than e. There exists a continuous unitary factor representation 1T of G of finite type such that 1T(S) 1. Let t be the canonical trace on the factor generated by 1T(G). The function

x ~ cp(x) = t[(1T(X)* - 1)(1T(X) - 1)] = t(2 - 1T(X) - 1T(X- I ) )

on G is continuous (cf, proof of 17.3.1) and central; we have cp(e) = 0, cp(s) O. There therefore exists a closed neighbourhood V s of s invariant under inner automorphisms and such that e~ V s If K is a compact subset of G such that e~ K, we can cover K by a finite number of sets V S I " " V s" and the complement of V S I U ... U V s" is a neigh bourhood of e not meeting K and invariant under inner automorphisms. Whence (i). Assertion (ii) will follow immediately from the following lemma: 17.3.8. LEMMA. Let G be a locally compact group satisfying condition (*) of 17.3.7. Let V be a neighbourhood of e and K a compact subset of

CH. 17. 4]

ADDENDA

351

G. There exists a neighbourhood of e contained in V and invariant under the inner automorphisms defined by the elements of K.

Making V smaller if necessary, we can suppose it to be open and relatively compact. Then, enlarging K, we can suppose that V ~ K and K = K- 1 Let K = K 3 n (G\ V), which is compact. We have e~ K. By ( *), there exists a neighbourhood W of e not meeting K and invariant under inner automorphisms. We are going to show, and this will finish the proof, that, if s E K, then V n W is invariant under the automor phism g ~ sgs" of G. Let x E V n w. We have s x S-l E W, and s x 1 3 3 S-I E KVK- ~ K ; if s x s :~ V, we have s x S-l E K n (G\ V) = K, which is impossible; hence s x S-I E V n w. 17.3.9. A connected group is generated by every neighbourhood of the neutral element. Hence, among the locally compact groups possessing "enough" continuous unitary factor representations of finite type, the only connected groups are, by 17.3.7 and 16.4.6, the products of groups R" with compact groups. These are not very interesting examples, since, in fact, the irreducible continuous unitary representations of these groups are finite-dimensional (13.1.8). However, we observe that, by 17.3.6, every discrete group possesses "enough" continuous unitary factor representations of finite type. We will in due course (18.7.9) obtain some results concerning the existence of characters not necessarily of finite type. References: [639], [641].
17.4. Addenda

17.4.1. (a) Let G be a locally compact group admitting a base of compact neighbourhoods (Vi) of e invariant under inner automorphisms. Let f be a character of G such that mt is dense in C*(G). Then f is of finite type. (Let Pi be the characteristic function of Vi The operator Af(Pi) is scalar and non-zero for some i. Since Pi is the limit in C*(G) of elements of mf the operator 1 is the norm-limit of trace-class operators relative to 6I1f Hence 6I1 f is a finite factor.) *(b) A countable discrete group can possess characters which are not of finite type. [689]. *17.4.2. There exist antiliminal countable discrete groups such that the space of characters of finite type and of norm 1 is not locally compact. [689].

352

CHARACTERS OF A LOCALLY COMPACT GROUP

[CH. 17. 4

17.4.3. Let G be a connected locally compact group, 7r a continuous unitary representation of G, and A the von Neumann algebra generated by 7r(G). Then the greatest projection E of the centre of A such that A E is of type III> is O. [874J, [1473J. 17.4.4. Let G be a connected locally compact group admitting a family of unitary representations (7r;) possessing the following properties: (1) for every s E G, there exists an i such that 7r;(s);c e; (2) the mappings S ~ 7rj(s) are norm-continuous. Then G is the product of a compact group and a group R". (Use 16.4.6 and 17.3.8). [874], [1484]. 17.4.5. Let G be a non-compact simple connected real Lie group. The only continuous unitary representations tr of G such that 7r(G) is contained in a finite factor are the trivial representations. This applies in particular to the finite-dimensional representations. [1473]. *17.4.6. Let G be a semisimple connected real Lie group. Let tt be an irreducible continuous unitary representation of G. If f is an element of L 2(G) with compact support, then 7r(f) is a Hilbert-Schmidt operator. If f is infinitely differentiable with compact support on G, then 7r(f) is a trace-class operator; and f ~ Tr 7r(f) is a distribution x". of G. Thus, 7r admits a character which is the distribution x".. This distribution x.. is defined by a locally integrable function on G, analytic on the set of regular elements of G. For the classical complex simple groups and some tt, X.. has been computed by Gelfand and Naimark. [758], [759], [760], [1144] and more recent papers by Harish-Chandra, *17.4.7. Let G be a nilpotent simply-connected real Lie group. Let tt be an irreducible continuous unitary representation of G. If f is infinitely differentiable on G and decreases rapidly at infinity, then 7r(f) is a trace-class operator; and f ~ Tr 7r(f) is a tempered distribution x". on G. Thus, 7r admits a characterwhich is the distribution x".. This distribution is not, in general, defined by a function nor even by a measure. [440], [441], [442], [639], [922]. 17.4.8. Let G be a locally compact group in which e admits a base of compact neighbourhood invariant under inner automorphisms. If JL is a central positive-definite measure on G, the corresponding unitary re presentation of G is of finite type. [639].

CHAPTER 18

THE DUAL OF A LOCALLY COMPACT GROUP

18.1. Definition of the dual

18.1.1. Let G be a locally compact group. There exists a canonical bijection of C*(G)" onto 6 (13.9.3). Transporting the topology of C*(Gr using this bijection, we obtain a topology on 6.

The topological space thus obtained is called the dual of 6 will denote this topological space. We will often identify the spaces 6 and C*(G)~.
DEFINITION.

G. From now on,

18.1.2. The space 6 is a locally quasi-compact Baire space (3.4.13 and 3.3.8). If G is discrete, C*(G) admits an identity element, and so 6 is quasi-compact (3.1.8). If G is separable, then 6 is separable (3.3.4). If G

is postliminal, there exists a dense locally compact open subset in

(4.4.5). 18.1.3. Let

71 be a continuous unitary representation of G, and S a set of .ontinuous unitary representations of G. We say that 71 is weakly contained in S if 71, regarded as a representation of C*(G), is weakly contained in S, regarded as a set of representations of C*(G). Having said this, we are going to translate 3.4.4, 3.4.9, 3.4.10 into a group context. To bring off this translation, we will use: (I) proposition 2.7.5 which enables us to pass from the positive forms on C*(G) to the continuous positive forms on L\G); (2) theorems 13.4.5 and 13.5.2 which enable us to pass from the positive forms on L1(G) to the continuous positive-definite functions on G. We then obtain this:

18.1.4. PROPOSITION. Let 7T be a continuous unitary representation of G, and S a set of continuous unitary representations of G. The following

conditions are equivalent: (i) 7T is weakly contained in S; (ii) Every positive-definite function associated with 71 is the uniform limit over every compact set of sums of positive-definite functions associated with S.

354

THE DUAL OF A LOCALLY

COMPACT GROUP

ten.

18, I

When Tf admits a cyclic oector E; these conditions are again equivalent to the following: (ii) The function s ~ ('Tf(s)~ I~) is the uniform limit over every com pact set of positive-definite functions associated with S.
18.1.5. PROPOSITION. Let tt E are equivalent:

G and S c 6. The following conditions

(i) tr E S; (ii) Tf is weakly contained in S; (iii) One of the non-zero positive-definite functions associated with tr

is the uniform limit over every compact set of positive-definite functions associated with S; (iv) Every positive-definite function associated with Tf is the uniform limit over every compact set of positive-definite functions associated with S. 18.1.6. The equivalence (i) ~ (iii) of 18.1.5 proves in particular that if G is commutative, the topology defined in 18.1.1 is the usual topology of the dual group of G. 18.1.7. DEFINITION. If tt is a continuous unitary representation of G, the support of tt is the set of a E G which are weakly contained in Tf. If cp is a continuous positive-definite function on G, the analyser of cp is the support of Tfl"

If S is the support of ar, then S contained in S (3.4.6).

IS

closed

III

6,

and

tr

is weakly

18.1.8. PROPOSITION. Let cp be a continuous positive-definite function on G, and A ~ 6 its analyser. (i) cp is the uniform limit over every compact set of sums of (pure) positive-definite functions associated with the elements of A. (ii) Every positive-definite function associated with an element of A is the uniform limit over every compact set of functions s ~ ~?j~1 AjA/"P(stsSj) where Sh"" s; E G, AI" .. , An E C. Since Tfl is weakly contained in A, (i) follows from 18.1.4. Let Tf E A. If we regard tt and TfP as representations of C*(G), we have Ker Tf ~ Ker m, and so every state associated with tt is a weakr-limit of states associated with ttP (3.4.2 (iij), Then (ii) follows from 13.4.10.

CH. 18. Il

DEFINITION OF THE DUAL

355

18.1.9. Let n be a cardinal, H; the standard n-dimensional Hilbert space, and Repn(G) the set of continuous unitary representations of G in H; There exists a canonical bijection of Repn(G) onto Rep~(C*(G (the set of non-degenerate representations of C*(G) in H n ) . The inverse (3.5.2) is a image, under this bijection, of the topology of Rep~(C*(G certain topology on Repn(G), which we are going to describe directly. In view of 3.5.8, this will furnish a new definition of the topology of On (the subspace of consisting of the a E such that dim a = n).

PROPOSITION. Let TrA tt E Repn(G), and Tr~, tt the corresponding elements of Repn(C*(G. The following conditions are equivalent: (i) Tr~~Tr'; (ii) For every ~ E H; and every compact subset K of G,

IITrA(s)~ (iii) For all

~,

- Tr(S )~I

~ 0

uniformly over

K;

1/ E H, and every compact subset K of G, /1/) - (Tr(s)~ /1/)1 ~ 0

uniformly over K.

/(Tr). (s)~

(ii) ~ (iii): Obvious. (iii) ~ (i): Suppose that condition (iii) is satisfied. For every we have

E X(G),

and the right hand side tends to zero. Since X(G) is dense
C*(G), Tr~ ~ tr (3.5.4).

In

(i) ~ (ii): Suppose that condition (i) is satisfied, and let us prove (ii). It is enough to do it for ~ of the form Tr(f)1/ (f E L 1(G), 1/ E H n) since the set of vectors of this form is dense in H n Now IITr). (s )Tr(f)1/ - Tr(s )Tr(f)1/II,,;;;117T). (S)TrA (f) 1/ - 7T(S )7T(f)1/11 + IITrA (s )Tr(f)1/ - TrA (s )Tr). (f) 1/II = IITr).(.-tf)1/ - Tr(slf)1/11 + IITr(f)1/ - Tr).(f)1/II The set of mappings g~Tr).(g)1/ of Lt(G) into H; is, for fixed 1/, equicontinuous. On the set of the s-If (s E K), which is compact, point wise convergence of these mappings is therefore equivalent to uniform convergence. Hence IITr).(s-tf)1/ - Tr(.-If)1/II~O uniformly over K. Moreover, IITr(f)1/ - Tr). (f)1/11 ~ O. 18.1.10. We will denote by Irrn(G) the set of irreducible continuous

356

THE DUAL OF A LOCALLY

COMPACT GROUP

[CH. 18. 2

unitary representations of G in Hi; We endow it with the topology induced by that of Repn(G). If G is separable, then Repn(G) and Irrn(G) are polish spaces (3.7.4 and 3.7.5). References: [582], [635], [1007].

18.2. The Fourier transformation 18.2.1. DEFINITION. Let /L E MI(G). For every 1T E 6, put (;ji/L)(1T) = 1T(/L) (which is a continuous linear operator on H TT) . The function 1T ~ (;ji/L)( 1T) on 6 is called the Fourier transform of /L. Remember that the space of 1T is only defined up to isomorphism, which is rather troublesome. If G is commutative, then (;ji/L)(1T) is the scalar operator f 1T(S) d/L(s) in the I-dimensional space H TT, and we recover the usual definition of the Fourier transformation by identifying the scalar operator and its ratio. 18.2.2. We have immediately

= a;ji/L ;ji(/L + /L) = ;jilL + ;jilL ;ji(/L * /L) =(;ji/L) . (;ji/L) ;ji(/L *) = (;ji/L)*
;ji(a/L)

(a E C, IL E MI(G

(/L, ILE MI(G)) (/L,/L E M\G (IL E M1(G (IL E M I ( G).

sup II;jiILII:s;; IIILII

18.2.3. The Fourier transformation is injective. In fact, let /L be a non-zero element of M1(G). There exists an f E X(G) such that g = f * IL is a non-zero element of LI(G). There therefore exists a 1T E 6 such that 1T(f)1T(/L)=1T(f * IL);t:O (2.7.3), whence (;jiIL)(1T)=1T(IL);t:O. 18.2.4. If f E L1(G), we denote by ;jif the Fourier transform of the measure f(s) ds. In other words, we put (;jif)(1T) = 1T(f). If E > 0, the function 1T ~ 11(;jif)(1T )11 is < E outside a quasi-compact subset of 6 (3.3.7). References: [520], [521], [522], [639], [949], [1041], [1112], [1458], [1508], [1509], [1522], [1681], [1722].

CH. 18. 3l

THE REDUCED DUAL

357

18.3. The reduced dual

18.3.1. DEFINITION. The reduced dual of G is the support of the regular representation of G. This reduced dual is a closed subset

Gr

of

G.

18.3.2. Let A be the left-regular representation of G. Regarded as a representation of C*(G), it has a certain kernel N. Then Gr is the set of the tr E G = (C*(G" whose kernels contain N. In other words, is the spectrum of the C*-algebra C*(G)/N, itself isomorphic to the C* algebra A(C*(G. This C*-algebra A(C*(G is the norm-closure, in 5(L 2(G of the set of left convolution operators by the elements of L1(G). Recall that N n L1(G) = 0 (13.3.6).

a,

18.3.3. Suppose that G is unimodular. Every square-integrable ir reducible representation 7T of G is contained in the regular represen tation, and therefore belongs to In particular, if G is compact,

G=Gr

o,

18.3.4. Suppose that G is commutative. Let f E L I(G), let X be a character of G, and A the regular representation of G. We have lx(f)/:!S; IIA(f)1I (13.3.6), and this inequality extends to the case where f E C*(G) by continuity. Hence X is weakly contained in A, so that =G. For other examples, d. 18.3.9, 18.9.8 and 18.9.11.

o.

18.3.5. PROPOSITION. (a) Let q; be a continuous positive-definite func tion on G. Then the following conditions are equivalent: (i) q; is the uniform limit on every compact set of functions of the form f *1(f E J{(G; (ii) (f) is the uniform limit on every compact set of continuous positive definite functions of compact support; (iii) q; is the uniform limit on every compact set of square-integrable continuous positive-definite functions; (iv) q; is associated with a representation weakly contained in the regular representation. (b) The functions satisfying (i), (ii), (iii), (iv) form a weak*-closed convex cone Q in C(G). (c) If q; E Q and if ljJ is a continuous positive-definite function, we have q;rjJ E Q.

358

THE DUAL OF A LOCALLY

COMPACT GROUP

[CH. 18. 3

(i) :?(ii):? (iii): Obvious. (iii):? (i): It suffices to show that, if q> is a square-integrable con tinuous positive-definite function, then q> is the uniform limit over G of functions f * I where f E X(G). Now there exists e E L 2(G) such that (() =" * .fr (13.8.6). Let (In) be a sequence of elements of X(G) con verging to " in L 2(G). Then fn * In converges uniformly to " * .fr = q>. (i):? (iv): Let A be the left regular representation of G. If f E X( G), then f * I is a positive-definite function associated with A (13.4.11). If q> satisfies (i), then 711p is weakly contained in A (18.1.4). (iv):? (ii): If q> satisfies (iv), then q> is the uniform limit over every compact set of sums of positive-definite functions associated with A (18.1.4). Now a positive-definite function associated with A is of the form", * .fr, where e E L 2(G), and is therefore the uniform limit over G of functions of the form f * where f E X(G); a sum of such functions is the uniform limit over G of continuous positive-definite functions of compact support. We have thus proved (a). It is clear that the functions satisfying (i), (ii), (iii), (iv) form a convex cone Q in L""(G). Let Q\ be the intersection of Q with the unit ball of L""(G). To prove that Q is weak*-closed, it is enough to prove that Q\ is weak*-closed (B 8). Let (() be a function of L ""(G) in the weak*-closure of Q\; let w be the positive form on C*(G) defined by tp, Then w is the weak*-Iimit of positive forms on C*(G) defined by elements of Q\ (2.7.5 (iii)). Hence 71., is weakly contained in A (3.4.9), so that cp E Q iLastly, (c) follows from the characterisation (ii) of the elements of Q and from the fact that the product of two continuous positive-definite functions is positive-definite (13.4.9).

I,

18.3.6. PROPOSITION. Let G be a locally compact group. Then the following conditions are equivalent: (i) The dual of G is equal to the reduced dual; (i) The trivial t-dimenslonal representation of G belongs to the reduced dual; (ii) Every continuous positive-definite function on G is the uniform limit over every compact set of functions of the form f * 1. where
E X(G);

(ii) The function 1 is the uniform limit over every compact set of functions of the form f * 1. where f E X(G); (iii) For every bounded positive-definite measure p., on G and every

CR. 18, 3]

THE REDUCED DUAL

359

continuous positive-definite function cp on G, we have I cp(x) d/L(x) ~ 0; (iii) For every bounded positive-definite measure /L on G, we have I d/L(x) ~ O.

The equivalence (i) :> (ii) follows from 18.3.5. The definition of positive-definite measures shows that (ii) ~ (iii). Suppose condition (iii) is satisfied. Let tp be a continuous positive-definite function on G. We bring in the notation Q of 18.3.5. To show that cp E Q, it suffices to prove that cp belongs to the bipolar of Q. Let, therefore, f E L1(G) be an element of the polar of Q. Since Q is a cone, we have Re(j, q) ,,;;; 0 for every q E Q, and in particular, Re(j, g * g),,;;; 0 for every g E J(G), so that the measure -if + f*)(s) ds is positive-definite. By condition (iii), we have ReI f(s)cp(s) ds ";;;0, which indeed proves that tp belongs to the bipolar of Q. Hence (iii) ~ (ii). The equivalences (i) :> (ii) :> (iii) are established in an analogous way. It is clear that (ii) ~ (ii). Lastly, if 1 E Q, every continuous positive definite function belongs to Q by 18.3.5 (c). 18.3.7. LEMMA. If G is the semidirect product of two groups H, K satisfying the conditions of 18.3.6, then G also satisfies these conditions. Let C be a compact subset of G. There exist compact subsets C, C" of H, K such that C ~ CC", and we can suppose that e E C. Let E > O. There exists an x E X(H) such that [x * i-II,,;;; E on C. Let Cb be the support of x. The set C~ of the h-1kh(h E Ci; k E C") is compact in the normal subgroup K. There exists y E X(K) such that Iy * y - 11,,;;; E on C~. Define z E X( G) by
z(hk) = x(h)y(k) (h E H, k E K).

We have, for h E Hand k E K

(z

* z)(hk) =
=

f ff =f *
HxK

z(hkg)i(g)dg x(hh)yh-1kh)k)x(h)y(k)dh dk

(y

y)(h-1kh)x(hh)x(h)dh.

360

THE DUAL OF A LOCALLY

COMPACT GROUP

[CH. 18, 3

Hence

I(z * z)(hk)-ll=s;;

If
H

(y

* y)(h-1kh)x(hh)i(h)dh x(hh)i(h) dh -

f
H

x(hh)i(h) dhl

+
=s;;

IJ
H

11
dh

J I(y * y)(h-

1kh)-IIlx(hh)Ili(h)1

co + I(x * x)(h) - II.

If hkE C, we have h E C, k E C", and so h-Ikh E C~ for hE q. Then
I(z

* z)(hk)- 1/ =s;;

elx(hh)I/x(h)1 dh Ix(h Wdh

+e

=s;; e

+E

=e[(x

* x)(e)+I]=S;;E(2+e).

18.3.8. LEMMA. Let G be a locally compact group satisfying the con ditions of 18.3.6, and N a closed normal subgroup of G. Then GIN satisfies the conditions of 18.3.6.

Let G* = GIN, and let ~ be the canonical morphism of G onto G*. Let C* be a compact subset of GIN and E > O. There exists a compact subset C of G such that ~(C) ~ C*, and we can suppose that e E C. There then exists f E X(G) such that If * ! - 11 =s;; eon C and (f * !)(e) = 1, in other words Ilflb = 1. Let F be the function defined on G* by
(1)

F(s*)

= (f If(sn)1 2 dn) 112,

N

where dn denotes a left Haar measure on N, and where s* = ~(s) (the right-hand side of (1) plainly depends only on s*). Then FE X(G*) and

(2)

I FI ~

f f
ds*
~

If(snW dn

J
G

If(sW ds

= 1.

CH. 18. 3j

THE REDUCED DUAL

361

In view of (2), we have, for every s EO,

I(P

* F)(s*) - 11 2 =

If f ,; ; f
G*

F(s*s*)F(s*) ds* -

11

= /

(P(S*S*)-F(S*))F(S*)ds*/2

IP(s*s*)-F(s*)12ds*

Applying the inequality pression is dominated by

(llall-llbll)2.,;Iia - bl1 2 in

L 2(N), this last ex

f f f
ds*
N

If(ssn) - f(snW dn

=
=

If(ss) - f(s)jZ ds

2- 2Re

f(ss)f(s) ds

2[1- Re(f

Hence, if s E C, we have 2 on C*.

* j)(s)] I(F * F)(s*) - W.,; 2.

Hence IF

* F - W,,;;;

18.3.9. PROPOSITION. Let 0 be a connected real Lie group, and R its radical. If OIR is compact, we have 0 = Or Let 0 be the universal covering of 0, which is the semidirect product of a compact group C and a simply-connected soluble group S. Since 0 is a quotient of 0, it is enough to prove that 0 =O~ (18.3.8). Now C = (18.3.3), and, by 18.3.7, we are reduced to showing that a We argue by induc simply-connected soluble group S satisfies S = tion on the dimension n of S. The assertion is obvious if n = O. Let n > 0 and suppose the assertion established for dim S < n. Then S is the semi-direct product of a group isomorphic to R and a simply-connected

c,

s,

362

THE DUAL OF A LOCALLY

COMPACT GROUP

[CH. 18, 4

soluble group of dimension n - I; it now suffices to apply 18.3.4, 18.3.7 and the induction hypothesis. References: [382], [582], [1607]. The groups considered in 18.3.6 are called amenable groups. They can be characterized by a lot of equivalent properties. See, for example, F. P. Greenleaf, Invariant means on
topological groups and their applications, Van Nostrand-Reinhold, New York,I%9.

18.4. The reduced dual and integrable representations

18.4.1. PROPOSITION. Let G be a unimodular locally compact group, and a a square-integrable irreducible representation of G. Then {u} is
closed in

Or

For f E X(G), u(f) is a Hilbert-Schmidt operator (14.4.2), and so u(C*(G r:;;, ~(Hu)' Consequently, the kernel of a in C*(G) is a maximal closed two-sided ideal I of C*(G) and every irreducible re presentation of C*(G) of kernel I is equivalent to a (4.1.10 and 4.1.11); hence {I} is closed in Prim C*(G) (3.1.4), and {o-} is closed in (C*(G)f = O. Furthermore, a E G r (18.3.3). 18.4.2. PROPOSITION. Let G be a unimodular locally compact group, and a an integrable irreducible representation of G. Then {c} is open and closed in Or We already know that {u} is closed in Or (18.4.1). Let A be the left-regular representation of G; then A(C*(G = B is a C*-algebra of operators in L 2(G) whose spectrum is Or (18.3.2). We can identify tr with a subrepresentation of A since tr is square-integrable. Let K ~ H; be the biinvariant subspace of L 2(G) associated with rr; let A, be the subrepresentation of A defined by K. For every f E L(G), AM) is the restriction of A(f) to K; hence, if we identify At with a representation of B, A( is just the mapping T ~ T IK of B into .2(K). The kernel of a in B is equal to that of Alo since A. = (dim o-) . a (14.3.5 (iii; this kernel is therefore the set I of the T E B such that T I K = O. Let J be the set of the T E B such that T I L 2(G ) 8 K= O. Let ~ be a non-zero vector of H; such that the function s ~ cp(s) =(A(s)~ I~) on G is integrable. We have if; E K (14.3.1). Since if; E L I( G), we can form A (if;) which is a non-zero element of B. For every g E L 2(G), we have A(if;)g = if; * g E K since K

CH. 18, 5l

THE MACKEY BOREL STRUCTURE

363

is biinvariant: hence A(ci) I e(G)8 K = O. We conclude from this that 1#0. Let, then, T be an element of Or which we can regard as an irreducible representation of B. Its kernel Ker T in B is a primitive ideal which contains 0 = I n 1, and which therefore contains I or 1 (2.4.4). If Ker T :2 I, then T is adherent to {IT} and therefore equal to IT. This proves that TE G T;iuKerT:21#0. Hence Or\{IT}is not dense in Or Hence {IT} contains a"non-empty open subset of 0,., and {IT} is open in Or

18.4.3.

COROLLARY.

If G is compact, then

is discrete.

Since every irreducible representation of G is integrable, this follows from 18.4.2. References: [446], [635].

18.5. The Mackey Borel structure 18.5.1. Let G be a separable locally compact group. For every cardinal n -s; we will endow Repn(G) and Irrn(G) with their topological Borel structures. With the notation of 18.1.9, the bijections Repn(G)-+ Rep~(C*(G, Irrn(G)-+Irrn(C*(G are Borel isomorphisms. We then define Rep(G), Irr(G) as the Borel sums of the Repn(G) and of the Irrn(G); all these spaces are standard.

"0

18.5.2. The functions 'IT-+('lT(s)~I1/) on Repn(G) (where sEG, ~,1/E H n ) define a Borel structure on Repn(G) identical to that of 18.5.1. In fact, the functions IT-+(lT(s){11/) are continuous (18.1.9) and so this structure is less fine than that of 18.5.1. It then suffices to observe that the functions IT-+(lT(Sk){p l1/ q ) sd, a dense sequence in G, ({p), (T/q) dense sequences in H n ) separate the points of Repn(G), and to apply B21. 18.5.3. The Mackey Borel structure on 0 is the quotient structure of that of Irr(G) for the canonical mapping Irr(G)-+O. This structure may be identified with the Mackey Borel structure of C*(G)". It is finer than the topological Borel structure of 0 (3.8.3), and is identical to it if G is postliminal (4.6.1). The points of 0 are Borel sets for the Mackey structure (3.8.4).

Reference: [1007].

364

THE DUAL OF A LOCALLY

COMPACT GROUP

[CH. 18, 7

18.6. The quasi-dual

18.6.1. Let G be a separable locally compact group, n a cardinal, and H; the standard n-dimensional Hilbert space. Facn(G) will denote the set of continuous unitary factor representations of G in Hi, Fac(G) the union of the Facn(G) for n = 1,2, ... ,l\o, Fac Ip(G) the set of the 7T E Fac(G) such that 7T(G) is a factor of type I p , and Fac I(G) the set of the 7T E Fac(G) which are of type I. All these sets are Borel subsets of Rep(G), and are therefore standard Borel spaces (7.1.4 and 7.3.4). They may be identified with Facn(C*(G, etc. 18.6.2. We will denote by 6 the set of quasi-equivalence classes of continuous unitary factor representations of G. The Mackey Borel structure on 6 is the quotient structure of that of Fac(G) for the canonical mapping Fac(G)~ 6. The Borel space 6 is called the quasi dual of G. It may be identified with C*(Gr. Every point of 6 is a Borel subset of 6 (7.2.4). 18.6.3. Let 6[ be the set of quasi-equivalence classes of the 7T E Fac(G) which are of type I. Then 6[ may be identified with C*(G)i. It is a Borel subset of 6, and the canonical bijection of 6 onto 6[ is a Borel isomorphism (7.3.6). 18.6.4. Let Facfn(G) be the set of the 7T E Facn(G) which are of finite type, Facf(G) the union of the Facfn(G), and 6, the canonical image of Facf(G) in 6. The sets Facf(G), Facfn(G) are Borel sets in Fac(G), 6, is a Borel set in 6, all these spaces are standard (7.4.3 and 7.4.4) and may be identified with Facf(C*(G, Facfn(C*(G, C*(G); Let C be the set of central continuous positive-definite functions cp on G such that <p(e) ~ 1. Then C is compact metrisable for the weak* topology. The set D of the characters of G of finite type equal to 1 at e is a G8 in C (17.3.5 and B 13) and is therefore a polish space. Endow with the topology obtained from that of D by the canonical bijection of D onto 6 f (17.3.4). Then 61 becomes a polish space, and the topological Borel structure is the structure induced by that of 6 (7.4.3).

o,

References: [559], [689].

18.7. Integration and disintegration of representations

18.7.1. Let Z be a Borel space, IL a positive measure on Z, and l: ~ H a measurable field of -Hilbert spaces on Z. For every l: E Z, let 7T(l:) be a

(n

CH. 18, 7j

INTEGRATION

AND DISINTEGRATION

OF REPRESENTATIONS

365

continuous unitary representation of G in H(,): we say that '~7T(') is a field of continuous unitary representations of G. This field is said to be is measurable if. for every s E G, the field of operators '~7T(n(S) measurable. 18.7.2. Suppose that G is separable. Suppose that Z is the union of mutually disjoint Borel sets ZI Z2 . . . , Z"" and that ,~ H (,) reduces on Z; to the constant field '~Hn (Hno the standard n -dimensional Hilbert space). Let '~7T(') be a field of continuous unitary representations of is measurable if and only if '~7T(') is G in the H (n Then '~7T(n equal almost everywhere to a Borel map of Z into Rep(G). This IS proved in exactly the same way as 8.1.8.
be a field of continuous unitary 18.7.3. PROPOSITION. Let '~7T(') representations of G. For every E Z, let 7T() be the representation of C*(G) corresponding to 7T(n Suppose that G is separable. Then the field '~7T(') is measurable if and only if the field '~7T'(') is measurable.

We are led, to begin with, to the case where the field '~H({) is the constant field defined by a fixed Hilbert space H (A 72). We can then suppose that H is a standard Hilbert space Hi; To say that the field '~7T(') (resp. '~7T'(,n is measurable means that it is equal almost everywhere to a Borel mapping of Z into Repn(G) (resp. into Repn(C*(G), by 18.7.2 (resp. 8.1.8). The proposition then follows from 18.5.1. 18.7.4. PROPOSITION. We keep the hypotheses of 18.7.3, and suppose that the fields ,~ 7T(n ,~ 7T() are measurable. Let 7T = fEB 7T() djL(n Let 7T be the continuous unitary representation of G corresponding to 7T. Then, for every s E G, we have 7T(S) = 7T()(S) djL(). Let (VI V 2 , ) be a sequence of compact neighbourhoods of s whose intersection is {s}. Let fn be a continuous positive function, with support contained in V n, and with integral equal to 1. Then 7T(fn) converges strongly to 7T(S), 7T,(,)(fn) converges strongly to 7T()(S) for every, and 7T(fn) = fEB 7T()(S) djL(). Hence 7T(S) = fEB 7T(n(S) djL() (A 79). 18.7.5. With the above notation, we say that 7T is the direct integral of the 7T(n and we write 7T =fEB 7T(n djL(). 18.7.6. This established, we can repeat the whole of chapter 8, from proposition 8.1.6 to no. 8.7, replacing A by G (separable locally compact

366

THE DUAL OF A LOCALLY COMPACT GROUP

[CH. 18, 7

group) throughout and "representation" by "continuous unitary re presentation". 18.7.7. We will not translate 8.8.1. Applying 8.8.2, we obtain, in par ticular, the following:
PROPOSITION. Let G be a unimodular separable locally compact group, A and p the left and right regular representations of G, au and V the von Neumann algebras on L 2(G) generated by A(G) and p(G), J the mapping f ~ f* of L 2(G) into itself, and t the natural trace on u: defined by E. (17.2.5). (i) There exist: (1) a standard positive measure JL on 6; (2) for every {E 6, a character f(t); (3) a structure of measurable field of Hilbert - with the following properties: spaces on the field {~HIW (a) e(G) may be identified with j$ H m l dJL(t); the fields

{~Jm)'
{~

{~A/W'

{~Plw,
"Vj(C)
{~tf(Cl

aum ),

{~

are measurable and

(B (B

Ell

J
(B

JIW dJL({),

= JAm)
EI;)

dJL ({),

P=
t=

JPlw dJL(z),
Ell

au = au m)dJL({),

"V =

"V m ) dJL({),

J m)
t

dJL({).

The algebra au n "V is the algebra of diagonalisable operators. (b) Am) E { and p~w E { almost everywhere. (c) For every x E C*(Gt, the function {~f(t)(x) is measurable and f(x) =f f({)(x) dJL({). (ii) Let a measure JL, characters f({), and a structure of measurable field of Hilbert spaces on the field {~Hrw, with the same properties as in (i) be given. Then JL and JL are equivalent; let JL = d . JL, with d(t) > 0 almost everywhere. We have f({) = d({)f({) almost everywhere. After alteration of the f(t) on a negligible set, there exists an isomorphism of the measurable field {~Hm) onto the measurable field {~Hrw which transforms JIW into J rw , Am> into Am), Plw into Pl,w. au lW into au rw , "V IW into "V rw , and tlW into trw

18.7.8.

COROLLARY.

We

keep

the

above

notation.

Let

uE

CH. 18. 8j

THE PLANCHEREL MEASURE

367

L1(G)

n L 2(G).

For almost every (E

6,

we have u E n fW; the function

(-, tf(CP"f(()(U)Af(Clu)*)
is u-integrabte, and we have

J
6

lu(s)jZ ds =

J
6

tfw(Afw(U)Aml u)*) dj.t().

We have u E C*(G), u

* u* E C*(Gt.

Hence the function

(-, f({)(u

* u*) = tfW(Af(()(U)AfW(u)*)
j.t

is measurable, and the value of its integral with respect to

f(u

is

* u*) =

e.(u

* u*) =

lu(s)jZ ds

< +00.
u E nf ).

Then, for almost every {, we have

tfdAf)(U)Af({)(U)*)

< +00, and therefore

18.7.9. COROLLARY. Let G be a unimodular separable locally compact group. Let s be an element of G different from e. Then there exists a traceable continuous unitary factor representation 7" of G such that 7"(s) ;c 1. With the notation of 18.7.7, we have A(s);c 1, and therefore AfW(s);c 1 on a non-negligible set of values of (. Now the Af(C) are traceable factor representations. 18.7.10. Let (uJ be a sequence of elements of X(G). By 8.8.3, we can choose the representation 7" of 18.7.9 in such a way that 7"(uJ is, for every i, a Hilbert-Schmidt operator relative to the factor generated by
7"(G).

References: [559], [638], [639], [689], [1003], [1004], [1005], [1006], [1007], [1036], [1037], [1100], [1101], [1102], [1103], [1105], [1458], [1681].
18.8. The Plancherel measure

18.8.1. Let G be a postllminal separable locally compact group. Let (-, K () be the canonical field of Hilbert spaces on 6 (8.6.1). Choose a field (-, 7T() of continuous unitary representations of G in the KC(), such that 7T() belongs to the class ( for every (, and the field is

368

THE DUAL OF A LOCALLY

COMPACT GROUP

[CH. 18, 8

measurable for every positive measure on G (8.6.2). For simplicity, we will write ( instead of 1T({). Let K({) be the conjugate Hilbert space of K({). Let J, be the canonical involution of K(n @ K(o, and t, the trace T @ I ~ Tr T on (.2(K({ @

ct.

THEOREM. Suppose, further, that G is unimodular. Let A and p be the left- and right-regular representations of G, 6/1 and V the von Neumann algebras on L 2(G) generated by A(G) and p(G), J the mapping f ~ f* on L 2(G), and t the natural trace on 6/1+ defined by Ee (17.2.5). Then there exist a positive measure JL on G and an isomorphism W of L 2(G) onto J(f) (K({) @ K({ dJL({) with the following properties: (i) W transforms

J into

pinto

f f
(f) (f)

(f)

J, dJL({),

A into

(1 @ [) dJL({),

6/1 into

f({ f
(f) (f)

(f)

1) dJL({),

(.2(K({ @ C) dJL({),

V into

f (C @ .2CK({) dJL({),

t into

t, dJL({),

and 6/1 n "V into the algebra of diagonalisable operators. (ii) If x E C*(Gt, the function {~Tr ((x) on G is lower semi-con tinuous, and we have Ee(X) = Ja Tr ((x) dJL({). In particular, if u E L1(G) n e(G), we have

(1)

f
G

lu(sW ds =

Tr({(u){(u)*) dJL({).

This follows from 8.8.5 applied to the trace the Plancherel Formula.

Ee

Formula (1) is called

18.8.2. THEOREM. Let G be a postliminal unimodular separable locally compact group. Then there exists a unique measure JL on G such that we have, for every u E LI(G) n L 2(G),

f
G

lu(sW ds =

f
a

Tr({(u){(u)*) dJL({).

CH. 18. 8]

THE PLANCHEREL MEASURE

369

The measure IL of 18.8.1 possesses this property. Let IL be another positive measure on 6 having this property. For every x E C*(Gt, put f(x) = fa Tr ~(x) dlL'(~). It is clear that f is a trace on C*(Gt. If (x., X2" ) is a sequence of elements of C*(Gt converging to x, we have 11~(xn) - ~(x)11 ~ 0 for every ~ E 6, hence, by Fatous lemma, lim f(x n) ~

f ~f

lim Tr

~(xn)

dlL(n
= f(x);

Tr ((x) dlL(O

and so f is lower semi-continuous. We have f(u * u*) < +00 for every u E L(G) n L 2(G), and so fit is dense in C*(G). The bitraces cor responding to f and to Ee are equal on L \ G) n L 2( G), and are therefore equal (6.5.3). We consequently have
Ee(X) =

Tr

~(x) dlL'(~)

for every

X E

C*(Gt, whence IL = IL (8.8.6).

18.8.3. DEFINITION. The measure IL of 18.8.2 is called the Plancherel 6 associated with the Haar measure of G. Suppose that we multiply the Haar measure by a scalar k > O. Let u E L(G) n L 2(G). For every (E 6, ~(u) is multiplied by k, and hence Tr(~(u)~(u)*) is multiplied bye. We therefore see that the Plancherel measure is multiplied by k- l
measure on

18.8.4. The support of the Plancherel measure IL is equal to the support of the regular representation of G (8.6.9 and 18.8.1), i.e. to the reduced dual of G (18.3.1). If K is a quasi-compact subset of 6, we have IL(K) < + 00. In fact, for every 7T o E K, there exists a u E L(G) n L 2(G) such that 7To(u) - 0; there therefore exists an a> 0 such that Tr(7T(u)7T(U)*) ~ a in an open neigh bourhood V"o of 7To. By the Plancherel formula, we have IL(V"O> < +00. We can now cover K by a finite number of sets V"o 18.8.5. PROPOSITION. Let G be a postliminal unimodular separable lo cally compact group, endowed with a Haar measure, and IL the cor responding Plancherel measure- on 6. Let ~o E 6. Then ~o is square integrable if and only if IL(Uo}) > 0, and IL(Uo}) is then equal to the [ormal dimension of ~o.

370

THE DUAL OF A LOCALLY

COMPACT GROUP

[CH. 18. 9

If p,({{o}) > 0, then {o 1 is contained in the regular representation A (8.6.8 (ii) and 18.8.1), hence {o is contained in A. and {o is therefore square-integrable (14.1.3). Suppose that {o is square-integrable. Use the notation of 18.8.1. Let F be the central projection associated with {o (14.2.4). It is diagonalisable; let Y be the p,-measurable subset of 6 such that F corresponds to the characteristic function cpy of Y. Then, the subrepresentation of A defined by F is on the one hand quasi-equivalent to {o (14.3.5), and on the other hand equivalent to ({ l)cpy({) dp,({) (18.8.1). Hence cpy p, is equivalent to the Dirac measure at {o (8.4.4). Hence p,({{o}) > O. Furthermore, for f E L1(G) n L 2(G), we have, by 8.8.7,

It

f
G

IFf(sW ds =
=

f
(]

<Py(C) Tr({(f)*C(f dp,(C)

p,({{o}) Tr({o(f)*Co(f.

Moreover, denoting the formal dimension of Co by d, we have

f
G

IFf(sW ds

d Tr(Co(f)*{o(f

(14.4.2).

We can choose p,({{o}) = d.

E L1(G)

n L 2(G)

such that Co(f);e 0, whence

18.8.6. If G is separable and commutative, the classical Plancherel theorem shows that the Plancherel measure is just the (suitably nor malised) Haar measure of the dual group 6. References: [452], [639], [759], [782], [922], [1033], [1036], [1037], [1041], [1100], [1458], [1522], [1681], [1722], [1736].

18.9. Addenda
18.9.1. Let G be a postliminal unimodular separable locally compact group, 6 r its reduced dual; and 7T E 6. (a) If {7T} is open in 6" then 7T is square-integrable. (Because 7T belongs to the support of the regular representation; hence, if p, denotes the Plancherel measure, we have p,({7T}) > 0.) *(b) It can happen that 7T E 6 r is square-integrable, that {7T} is open in

CH. 18, 9l

ADDENDA

371

0, and that 7T is not integrable. It can happen that 7T is square-integrable, and that {7T} is not open in O. It can happen that {7T} is open in 0 and that 7T{;! Or [446], [878]. (c) If 7T is integrable, bTl is open in O. [1786]. (d) Problem: If 7T is square-integrable, is {7T} open in Or?
*18.9.2. The Plancherel formula can be generalised to non-separable postliminal unimodular locally compact groups. [452]. 18.9.3. Let G be a locally compact group. Every continuous complex valued function on G is the uniform limit on every compact set of linear combinations of functions! * f (f E X(G; these functions are asso ciated with the regular representation. This is comparable with 18.1.4 and with the fact that, for certain G, some elements of 0 are not weakly contained in the regular representation. [582]. *18.9.4. Let G be a semisimple connected real Lie group. The set of the ! E L(G) such that the rank of 7T(f) is finite and bounded for 7T running through 0 is a self-adjoint two-sided ideal of L1(G) dense in L1(G), We can therefore apply 4.7.11. [582]. 18.9.5. Let G be a locally compact group. If the regular representation of G weakly contains a finite-dimensional continuous unitary represen tation, we have 0 = Or [587]. 18.9.6. Let G be a postliminal separable unimodular locally compact group. Let p. be the Plancherel measure on G. Let E be the set of the 7T E 0 such that 7T(C*(G =T (H.,,), i.e. such that {7T} is closed in O. Then p.(O\E) = O. What can be said about E from a topological point of view? 18.9.7. Let G be a separable locally compact group in which e possesses a base of compact neighbourhoods invariant under inner automor phisms. In 18.7.7, almost all the characters !() are of finite type. (Use 8.8.3 and 17.4.1). *18.9.8. Let G be the free group on two generators, endowed with the discrete topology. Then Or = 6, the set of pure positive-definite func tions is dense in the set of positive-definite functions for the topology of compact convergence, and the disintegration of the regular represen tation A into irreducible representations can be accomplished in two completely different ways. The representation A, which is a factor

372

THE DUAL OF A LOCALLY

COMPACT GROUP

[CH. 18, 9

representation, admits a disintegration (on a standard Borel space) into representations which, almost everywhere, are not factor represen tations. (We can observe that the von Neumann algebras associated with the Hilbert algebra of G are factors of type III, and apply 5.7.5, 5.7.6, 11.2.4). [382], [454], [1868]. 18.9.9. Let G be a locally compact group, H a closed normal subgroup of G, 7T a continuous unitary factor representation of G, and p = 7T IH. Then p is of type I, or of type II., or of type II"" or of type III. If p is of type I, then p is of multiplicity n for a uniquely determined n. The class of measures on If associated with p is invariant and ergodic for the action of G on If (G operates on H as inner automorphisms, and therefore on If by transfer of structure). [1008]. 18.9.10. Let G be a locally compact group, and H an open subgroup of G. If 6 = 6" then If = tt; [1607].

*18.9.11. Let Then 6#6,.

G be a non-compact simple connected real Lie group. [1607].

7T ~

18.9.12. Let G be a locally compact group. Then the mapping homeomorphism of 6 onto 6.

iT is a

18.9.13. The topological space 6 has only been explicitly worked out in a small number of cases. Here are some examples. If G is the non commutative 3-dimensional simply-connected nilpotent real Lie group, we can identify 6 with (RHO}) U R 2 ; the topology is that of the topologi cal sum of R\{O} and R 2, except that a point of R\{O} which converges to zero in the ordinary sense, converges in 6 to every point of R 2 If G = SL(2, C), we can identify 6 with the subset of R 2 consisting of the pairs (n, y)(n = 1,2, ... , y E R), the pairs (0, y)(y;;;. 0), and the pairs (x, 0)(- 1 ~ x ~ 0); the topology is that induced by the topology of R 2 except that a point of 6 which converges to (-1,0) in the ordinary sense converges in 6 to the points (- I, 0) and (2,0). [582]. In the case of SL(2, C), the structure of C*(G) has been completely determined. [584]. For the majority of groups, we have not got beyond the determination of the set 6. This work is almost finished for the complex semisimple Lie groups, and we then have the Plancherel measure; in the case of real semi-simple Lie groups, the work is not at such an advanced stage. The

CH. 18, 9j

ADDENDA

373

fundamental work here Naimark.

IS

that of Gelfand, Graev, Harish-Chandra,

18.9.14. If G is commutative, G is endowed with a group structure. The extension of this fact to the non-commutative case is the existence, for every pair (IT, IT) of elements of G, of a class of measures and of multiplicities on G, namely the class of measures and the multiplicities associated with IT IT (if G is separable postliminal). This class of measures and these multiplicities have only been calculated in a small number of cases. 18.9.15. Let G be a locally compact group, IT], lT2 continuous unitary representations of G, and 51> 52 sets of continuous unitary represen tations of G. If lTk is weakly contained in S, (k = 1,2), then lT1 lTz is weakly contained in the set of the PI P2 (PI E 51> P2 E 52) [587].

*18.9.16. Let G be a locally compact group, IT a finite-dimensional irreducible continuous unitary renresentation of G, 5 a set of continuous unitary representations of G, and T the set of the P iT where p runs through 5. Then 5 weakly contains IT, if and only if T weakly contains the trivial l-dimensional representation. The two parts of this assertion can fail to be equivalent if dim IT = +00, [587].

APPENDIX A

VON NEUMANN ALGEBRAS

This appendix is a list of results, used in the book, concerning von Neumann algebras. We do not give proofs, but only references to Les algebres doperateurs dans lespace hilbertien, 2nd edition, Gauthier Villars, Paris, 1969 (referred to as [I]). H denotes a Hilbert space.
A 1. The norm topology on 2(H) is the topology given by the norm

T-IITII. The strong, weak. ultra-strong, and ultra-weak topologies are the four topologies defined respectively by the following seminorms: T-IIT~II (~ an element of H). for the strong topology; T -I(T~ 111)1 (~, 11 elements of H), for the weak topology; T - [~~=llIniIl2]1/2 (~I' ~2" .. elements of H such that ~ lI~iW < +00), for the ultra-strong topology; T -I~~=t (ni 111;)1 (~t, ~2' ... ,111,112 ... , elements of H such that ~ I ~i112 < + 00, ~ II11iW < + (0), for the ultra-weak topology. On a bounded subset of 5t(H), the strong and ultra-strong topologies coincide. and the weak and ultra-weak topologies coincide ([1]. p. 30-34).
A 2. Let M be a subset of 5t(H). The commutant of M. denoted by M, is the set of elements of 5t(H) which commute with the elements of M. We put (M)= Mil (the bicommutant of M). The set M is a subalgebra of 5t(H) containing 1. The inclusion M ~ N implies that M::2 N. We have M = M", and M k Mil.
A 3. Let A be an involutive subalgebra of 5t(H), and A I the unit ball of A. The following eight conditions are equivalent: A (resp. At) is weakly closed, A (resp. AI) is strongly closed, A (resp. AI) is ultra-strongly closed, A (resp. At) is ultra-weakly closed. If this is so, there exists in A a projection E greater than all the others. For every TEA, we have ET = TE = T. The elements of A" are the elements T + A . I, where TEA ([1], p. 41. Th. 2).

A 4. Let A be an involutive subalgebra of 5t(H). The following nine

VON NEUMANN ALGEBRAS

375

conditions are equivalent: (1) the eight conditions of A 3, to each of which we add the hypothesis 1 E A; (2) A = A". When these conditions are satisfied, we say that A is a von Neumann algebra on H ([1], p. 42). The set 5(H) is a von Neumann algebra on H. The set of scalar operators on H is a von Neumann algebra on H denoted by C H or just by C. A 5. If M is a self-adjoint subset of 5(H), then M is a von Neumann algebra ([1], p. 2). A 6. If d is a von Neumann algebra, then each of d and d is the commutant of the other. The common centre of d and d is d n d. If this centre reduces to just the scalar operators, then d is said to be a factor. d is a factor if and only if d is a factor ([I), p. 3). A 7. Let d be a von Neumann algebra on H, and T an hermitian element of 5(H). Then T belongs to d, if and only if the spectral projections of T belongs to d ([I], p. 3). A 8. Let d be a von Neumann algebra, and Sand T elements of d+ such that S ~ T. There exists an REst such that SI/~ = RTl/~ ([1], p. 1l , Lemma 2). A 9. Let d be a von Neumann algebra, and m a two-sided ideal of d. Then m is self-adjoint and is the set of linear combinations of elements of m" = m n si": The set of TEd such that TT* E m is a two-sided ideal" of d such that the ideal "2 is equal to m. We denote it by m ll2 ([1], p. 10--12). Every weakly closed two-sided ideal of d is of the form Ed, where E is a projection of d n d ([1], p. 42, Cor. 3). A 10. Let d be a von Neumann algebra, and E a projection of d. Among the projections of d n d dominating E, there exists one smaller than all the others. It is called the central support of E ([I], p. 6). A 11. Let M be a subset of 5(H). There exists a smallest von Neumann algebra d on H containing M; it is called the von Neumann algebra generated by M. If M = M*, we have d = M" ([1], p. 2). A 12. Let A be an involutive subalgebra of 5(H). The weak, strong, ultra-weak, and ultra-strong closures of A all coincide; let qJ be this

376

APPENDIX A

closure. Let K be the closed subspace of H generated by the T~(T E A, ~ E H). Then P K is the greatest projection of 00. If K = H, then 00 = A" is the von Neumann algebra generated by A ([1], p. 42). A 13. Let A and B be involutive subalgebras of 2?(H), with A c B. If A is strongly dense in B, then the unit ball AI of A is strongly dense in the unit ball B I of B ([1], p. 43, Th. 3). If, further, H is separable, then every element of B I is the strong limit of a sequence of elements of AI ([1], p. 32). A 14. Let A be an involutive subalgebra of :(H) and ~ E H. We say = 0 imply that ~ is a separating vector for A if the conditions TEA, that T = O. We say that ~ is a cyclic vector for A if A~ is dense in H. If ~ is a cyclic vector for A, then ~ is a separating vector for A. The converse is true if A is a von Neumann algebra ([1], p. 5-6).

A 15. Let d be a von Neumann algebra on H, and E = P K a projection of d. Let 00 be the set of TEd such that ET = TE = T. The operators T IK, where T runs through 00, form a von Neumann algebra on K, called a reduced algebra of d, and denoted by dE or d K Let E = P K , be a projection of d. The operators Til K, where T runs through d, form a von Neumann algebra on K, said to be induced by d on K, and denoted by dE or d K ,. The two notations dE, dE are consistent in the case of a projection belonging. to d n sd, The mapping T ~ T IK (T E d) is called the induction of d onto dE" The von Neumann algebras (dE), (d)E are equal and are denoted simply by d~ ([1], p. 16-18). If ~ is the centre of d, the centre of dE is ~E ([1], p. 18, COL). A 16. Let (Hj ) be a family Of Hilbert spaces, H their direct sum, and d j a von Neumann algebra on H; For every family (Tj ) such that T, E d j for every i and sup IITjll < +00, form the operator (xj)~(TrXj) on H. The set of the operators so obtained is a von Neumann algebra on H called the product of the d j and denoted by nd j ([1], p. 19). A 17. Let HI H 2 be two Hilbert spaces. There exists, on the algebraic tensor product H o of HI and H 2 , a unique pre-Hilbert structure such that (~I @ ~217)1 @ 7)2) = (~I 17)2)(617)2) for ~,,7)1 E H" ~2, 7)2 E H 2 This structure is Hausdorff. The Hilbert space H, the completion of Hi; is called the Hilbert tensor product of HI and H 2 and is denoted by HI @ H 2 Let T I E 2?(HI ) , T 2 E :(H2) . The algebraic tensor product of

VON NEUMANN ALGEBRAS

377

T 1 and T 2 acts on H o and is continuous. Its continuous extension to 2(H) is denoted by T I T 2 ([1], p. 21). There exists a family (Kj)jEI of closed subspaces of HI H 2 whose direct sum is HI H 2 , and isomorphisms Ui: HI ~ K; with the following properties: (1) for every T E 2(H I), T 1 is the operator on HI H 2 , which leaves each K, invariant and is such that (T 1) IK, = UjTU;I; (2) Card I = dim H 2 ([1], p. 22-24).

A 18. Let d, be a von Neumann algebra on HI> and d 2 a von Neumann algebra on H 2 The operators of the form R, R 2 + 5, 52 + + T I T 2 , where R" ... , T] E d l , R 2 , , T 2 E d 2 , form an involutive subalgebra A o of 2(H 1 H 2) . The von Neumann algebra on H, H 2 generated by A o is called the tensor product von Neumann algebra of d, and d 2 and is denoted by d, d 2 ([1], p. 24). If A is an involutive subalgebra of 2(H,), the commutant of the von Neumann algebra generated by A C Hz is A 2(H 2) ([1], p. 24).
0

A 19. Let d" d 2 be von Neumann algebras on H" H 2 An isomorphism of d, onto d 2 is just an isomorphism of the underlying involutive algebras; such an isomorphism is automatically isometric. Of course, a Hilbert space isomorphism HI ~ H 2 defines an isomorphism (said to be spatial) of every von Neumann algebra on HI onto a von Neumann algebra on H 2 However, an isomorphism is not in general spatial ([1], p. 9).
o

A 20. Let d be a von Neumann algebra, and E a projection of d. Then the induction T ~ T E, of d onto dE is an isomorphism, if and only if the central support of e is 1 ([1], p. 18, Prop. 2). A 21. Let d be a von Neumann algebra on H, and H a Hilbert space. The mapping T ~ T 1 is an isomorphism of d onto the von Neumann algebra d C H , \: 2(H H). Such an isomorphism is called an am pliation ([1], p. 24). A 22. Let d" d 2 be von Neumann algebras on HI H 2 , and ep an isomorphism of d l onto d 2 There exists a Hilbert space K and a projection EE (d l C K ) of central support 1 such that ep is the composition of the amplition T ~ T lK the induction T l K ~ (T l K )E, and a spatial isomorphism. If HI and H 2 are separable, we can suppose K to be separable ([1], p. 55-56). A 23. Let d be a von Neumann algebra on H, and d the Banach dual

378

APPENDIX A

of the Banach space stl. If f E stl, it comes to the same thing to say that f is ultra-weakly continuous or that f is ultra-strongly continuous; the set of these forms is a subspace P of stl which is closed for the norm of si, Similarly, if f E stl, it comes to the same thing to say that f is weakly continuous or that f is strongly continuous; the set of these forms is a subspace of P which is dense in P for the norm of stl. The canonical bilinear form on stl x stl induces a bilinear form on stl x P for which stl with its norm is the Banach dual space of P. P is said to be the predual of stl ([1], p. 34-40). A 24. Let f be a weakly continuous linear form on 2(H). There exist , E~) in H, and non-negative two orthonormal systems (E), ... , En), (E;, + An and f(T) = ~ A;(TEj IE;) numbers AI>" ,An such that Ilfll = A) + for every T E 2(H) ([1], p. 36-37). A 25. Let .stl be a von Neumann algebra. Every increasing filtering subset of .stl+ which is bounded above has a least upper bound in .stl+ ([1], p. 4). For a positive form f on stl, the following conditions are equivalent: (1) for every increasing filtering subset g; C stl+ with least upper bound T E st", f(T) is the least upper bound of f(:!F); (2) f belongs to the predual of stl. f is then said to be normal. Every element of the predual of stl is a linear combination of normal positive forms ([1], p.50-51). A 26. Let stl be a von Neumann algebra, and f a normal positive form on stl. There exists a greatest projection F of stl such that f(F) = O. We have f(TF) = f(FT) = 0 for every T E stl. The projection E = 1 - F is called the support of f. We have f(T) = f(ET) = f(TE) = f(ETE) for every T E.stl ([1], p. 57). A 27. Let.stl and ~ be two von Neumann algebras, and cp a morphism of stl into ~. We say that cp is normal if, for every increasing filtering family g; C stl+ with least upper bound T E , cp(T) is the least upper bound of cp(:!F). cp is then ultra-strongly and ultra-weakly continuous, and cp(stl) is weakly closed ([1], p. 53, Th. 2, and p. 54, Cor. 2). A 28. Let .stl be a von Neumann algebra. A trace on : is a function cp defined on : with non-negative finite or infinite values, possessing the following properties: (i) If S E stl+ and T E sd", we have cp(S + T) = cp(S) + cp(T);

VON NEUMANN

ALGEBRAS

379

(ii) If S E.s!J+ and if A is a non-negative number, we have q;(AS) = Aq;(S) (with the convention that 0 . + 00 = 0); (iii) If S E .s!J+, and if U is a unitary operator of .s!J, we have q;( USU- I ) = q;(S). q; is said to be finite if q;(S) < + 00 for every S E d+. q; is said to be semi-finite if, for every S E .s!J+, q;(S) is the least upper bound of the numbers q;(T) for the T E.s!J+ such that T::s;; Sand q;(T)< +00.
= 0, imply that S=O. q; is said to be normal if, for every increasing filtering subset [iF C with least upper bound S E , q;(S) is the least upper bound of q;(5i).

(All the above is a special case of 6.1.1.) q; is said to be faithful if the conditions s E , q;(S)

A 29. Let.s!J be a von Neumann algebra, wand eo normal traces defined on : and B an involutive subalgebra of .s!J which is weakly dense in the von Neumann algebra d. Suppose that w(TT*) = w(TT*) < + 00 for every T E B. Then w = eo ([I], p. 203, Lemma I).
Tr(T) =};;(Te,

A 30. Let (e;) be an orthonormal basis of H. For every T E !t(Ht, put Ie;). Then Tr is a semi-finite faithful normal trace on !t(H+), which is independent of the orthonormal basis chosen ([1], p. 94, Th.5).

A 31. A von Neumann algebra d is said to be finite (resp. semi-finite) if, for every non-zero T of , there exists a finite (resp. semi-finite) normal trace I on .s!J+ such that I(T) # O. It is said to be properly infinite (resp. purely infinite) if the only finite (resp. semi-finite) normal trace on d+ is 0 ([1], p. 97, Def. 5). On a semi-finite von Neumann algebra, there exists a semi-finite faithful normal trace ([1], p. 99, Prop. 9).
A 32. Let [iF be a semi-finite factor, and I a semi-finite faithful normal trace on [iF+. Every normal trace on :!F+ is of the form AI where A E [0, +00] (with the convention that O +00 = 0) ([1], p. 81, Cor. 2 and p. 90, COL). The ideals m/ n, are independent of the choice of I. The elements of m, (resp. n/) are called the trace-class elements (resp. the Hilbert-Schmidt elements) relative to the semi-finite factor g; ([1], p. 101). In the case where g; = !t(H), we speak simply of trace-class operators, or of Hilbert-Schmidt operators. If T, T E !t(H) and if T is a trace-class operator, then we have ITr(TT)I::s;;(Tr ITI) IITII(fl], p. 106).

380

APPENDIX A

A 33. Let f!li bea factor. If there exists a non-zero finite normal trace f on f!li+, then f!li is finite (because f is faithful by [1], p. 83, Cor. 2). A 34. Let.sli be a von Neumann algebra. A projection E of d is said to be minimal if E= 0 and if the only projections of d dominated by E are o and E. It comes to the same thing to say that E= 0 and that the reduced algebra dE does not contain the scalar operators ([1], p. 123). A projection E of d is said to be abelian if the reduced algebra .sIiE is commutative ([1], p. 123, Def. 3). A 35. Let.sli be a von Neumann algebra. The following conditions on d are equivalent: (i) .sIi is isomorphic to a von Neumann algebra gj3 such that gj3 is commutative; (ii) every non-zero projection of .sIi dominates a non-zero abelian projection; (iii) there exists an abelian projection of d whose central support is 1. If these conditions are satisfied, then d is said to be of type I. An algebra of type I is semi-finite ([1], p. 122, Prop. 2; p. 123, Th. 1; p. 126, no. 4). A 36. Let.sli be a von Neumann algebra on H. The following conditions are equivalent: (i) d is a factor of type I; (ii) d is a factor and possesses minimal projections; (iii) there exist Hilbert spaces HI and Hz, and an isomorphism of H onto HI Hz, which transforms d into .2(HI ) C H2 ; (iv) there exists a family (K;) of closed subspaces of H whose direct sum is H, and isomorphisms U, of a Hilbert space K onto the K j , with the following property: .sIi is the set of the T E .2(H) reduced by the K j , for which the T IK, are obtained by the U, from a single element of .2(K) ([1], p. 124, Cor. 3, and p. 21-23). A 37. Every automorphism of a factor d of type I is inner and is defined by a unitary element of d ([I], p. 241, Cor. 2; this can also be regarded as a consequence of 4.1.5). A 38. A von Neumann algebra d is said to be continuous if there is no projection E= 0 in d n d such that dE is of type I. It is said to be of type II if it is semi-finite and continuous. A purely infinite algebra is continuous; such an algebra is also said to be of type III. If the space on which d acts is non-zero, then d cannot be of two different types ([1], p. 121, 122, 126). A 39. Let.sli be a von Neumann algebra. There exist projections E l , Ell Em of d n sd, characterisedby the following properties: E b E l b Em are

VON NEUMANN ALGEBRAS

381

mutually orthogonal, their sum is 1, .sIlEI is of type I, .sIlEII is of type II, and .sIlEIII is of type III. E I (resp. E lb Em) is the greatest projection of .sIl n .sil such that .sIlE is of type I (resp. II, III); E I + Ell is the greatest projection E of .sIl n.sll such that .sIlE is continuous ([1], p. 98 and 122). A 40. A finite von Neumann algebra of type II is said to be of type III> and a properly infinite algebra of type II is said to be of type II"". If SIl is a von Neumann algebra, there exist projections E 11I, E11~ characterised by the following properties: E ll1, ElL. are orthogonal with sum Ell and the corresponding algebras induced by SIl are of type IIIII~ ([1], p. 98 and 122). A 41. Let .sIl be a von Neumann algebra, and E and F two projections of .sIl. We say that E and F are equivalent relative to .sIl, and we write E - F, if there exists a U E SIl such that U* U = E, UU* = F. We write E -c F, or F>- E, if there exists a projection of .sIl equivalent to E and dominated by F ([1], p. 215, Def. 1). A 42. The relation E - F is an equivalence relation. The relation E < F is a pre-order relation. If we have E < F and F -c E, then E - F ([1], p. 215-216). A 43. Two equivalent projections have the same central support ([1], p. 215). A 44. Let E, F be two projections of si ; let E 1, F I be their central supports. If E I and F I are non-orthogonal, there exist two equivalent projections E , F of SIl dominated respectively by E, F. ([1], p. 217, Lemma 1). A 45. Let H be the space on which .sIl acts. Let T E SIl. The projections onto H E9 Ker T and onto T(H) belong to SIl and are equivalent ([1], p. 216, Prop. 2). A 46. If .sIl is a factor, and if E, F are projections of .sIl, we have either E < F for F < E ([I], p. 218, Cor. 1). A 47. A von Neumann algebra SIl is said to be homogeneous if there exists in .sIl a family (EJ of mutually orthogonal, equivalent abelian projections whose sum is 1. Such an algebra is of type I. The cardinal n of the family (EJ depends only on SIl. We say that .<t1 is of type In. It comes to the same thing to say that .sIl = 00 It(K) where 00 is a

382

APPENDIX A

commutative von Neumann algebra and where K is an n-dimensional Hilbert space ([1], p. 239-240). A 48. A von Neumann algebra .sIl is continuous if and only if every projection of .sIl is the sum of two orthogonal equivalent projections of .sIl ([1], p. 219, Cor. 3). A 49. A factor is of one of the types I, III, Il.; III, these cases being mutually exclusive (except for the space H = 0). If it is of type I, it is of type In for a uniquely determined n. The only factors which are finite-dimensional vector spaces are those of type In with n < +00. (This follows, for example, from A 48.) A 50. Let.sll be a von Neumann algebra of type I. There exists a family n .sil, which are mutually or thogonal, have sum 1, and are such that the .sIlE ; are of type In; with the n, mutually distinct. The E, are well determined to within a permutation of the index set ([1], p. 240).
(Ej)iEI of non-zero projections of .sIl

A 51. Every isomorphism between von Neumann algebras of type III acting on separable spaces is spatial ([1], p. 301, Cor. 7). A 52. Let .sIl be a von Neumann algebra. The projections E I (resp. E l b E Ill) relative to .sIl and .sil are equal. .sIl is of type I (resp. of type II, of type III, semi-finite, continuous), if and only if .sil is of type I (resp. of type II, of type III, semi-finite, continuous) ([1], p. 104, Cor. 1, 2, 3; p. 123, Th. 1; p. 124, Cor. I). A 53. Let.sll be a von Neumann algebra, and E a projection of .sIl or of .sil. If .sIl is of type I (resp. of type II, of type III, semi-finite, con tinuous), then .sIlE is of type I (resp. of type II, of type III, semi-finite, continuous) ([1], p. 103, Prop. 11; p. 104, Cor. 4; p. 125, Prop. 4). A 54. A Hilbert algebra is an involutive algebra A endowed with a scalar product which makes A a Hausdorff pre-Hilbert space, the following axioms being satisfied: (i) (x Iy) = (Y* Ix*) for x, yEA; (ii) (xy Iz) = (y Ix*z) for x, y, z E A; (iii) for every x E A, the mapping y ~ xy of A into A is continuous; (iv) the set of the xy (x E A, yEA) is dense in A. We have that (yx Iz) =(y Izx*) for x, y, z E A, and the mapling y ~ yx of A into A is continuous. Let the Hilbert space H be the completion of A, J the involution of H which extends the mapping x ~ x* of A into A, and U, and V x the

VON NEUMANN ALGEBRAS

383

elements of 2(H) which extend left- and right-multiplication by x in A. Then x ~ U, (resp. x ~ V x ) is a non-degenerate representation of A (resp, of the reversed algebra) in H. The U, commute with the V y We have fUx:! = V x for every x E A. The weak closure of the set of the U, (resp. V x ) is a von Neumann algebra OU(A) (resp. V(A said to be left (resp. right-) associated with A. We have V(A) = OU(A) = fOU(A)f, OU(A) = V(A) = fV(A)f. The mappings x ~ ux, x ~ V x are called the canonical mappings of A into OU(A), V(A). If C E OU(A) n V(A), we have ICI = C* ([1], p. 69-73).
A 55. Let A be a Hilbert algebra, and A an involutive sub algebra of A. Then A is a Hilbert algebra. If the set of the Ux, for x E A, is strongly dense in OU(A), then A is dense in A. If A is dense in A, we have OU(A ) = OU(A), V(A ) = V(A) ([1], p. 72, Prop. I).

A 56. We retain the above notation. For every a E H, the following conditions are equivalent: (i) the mapping x ~ Vxa of A into H is continuous; (ii) the mapping x ~ U,a of A into H is continuous. We then say that a is bounded (relative to A). Let Ua (resp. Va) be the continuous extension to H of the mapping x ~ Vxa (resp. x ~ Uia) of A into H. If a E A, then a is bounded and the notation Ua Va is consistent with the previous notation. The element a E H is bounded if and only if fa is bounded. We say that A is full if every bounded element belongs to A ([1], p. 72-74). A 57. Let A be a Hilbert algebra which is dense in the Hilbert space H. If a, b E H are bounded, then we have Uab = Vba; put Uab = Vba = abo Put, also, Ia = a *. Then the set of bounded elements becomes a Hilbert algebra A containing A as a dense Hilbert subalgebra. The Hilbert algebra A is full. We have OU(A ) = OU(A), V(A ) = V(A) ([I], p. 74). A 58. The mappings a ~ Ua a ~ Va (a E A) are injective ([1], p. 70). A 59. The U; (resp. Va), where a E A, form a self-adjoint two-sided ideal of OU(A) (resp. V(A. For a E A, T E OU(A), T E V(A), we have Ta E A, T E A, UTa = TUa, VTa = TVa ([1], p. 72, Prop. 2). A 60. For S E OU(At, put f(S) =(a I a) if Sl/2 = u, with an a E A, and f(S) = + 00 otherwise. Then f is a semi-finite faithful normal trace on OU(At. The two-sided ideal of OU(A) formed by the U, (a E A) is the set

384

APPENDIX A

of the T E OU(A) such that f(T*T) < + 00 i.e. nf . If a, bE A, we have Ib) =i( ut Ua ) , where i denotes the linear extension of f to m, Analogous results hold for V(A). The von Neumann algebras OU(A) and V(A) are semi-finite. The trace f on OU(A)+ and the analogous trace on V(At are called the natural traces on OU(At, V(At defined by A. These traces are unaltered if we replace A by A ([1], p. 85-88).
(a

A 61. Let.s!i be a von Neumann algebra, f a semi-finite faithful normal trace on A +, and i the linear extension of f to mf Endowed with the scalar product (S I T) = i(T*S), nf is a full Hilbert algebra. Let the Hilbert space H, be the completion of nf and Jf the continuous involu tion of H, extending that of nf . For REd, left- and right-multiplication by R in n f extend to continuous linear operators Af(R), pf(R) = JfA-rCR)Jf. The mapping Af is an isomorphism of A onto OU(n f) which extends the canonical mapping of n f into OU(n f) and transforms f into the natural trace defined by n f on OU(nft ([1], p. 88, Th. 2). A 62. Let A be a Hilbert algebra which is dense in a Hilbert space H. Let a E H. The following conditions are equivalent: (i) (a Ixy) = (a Iyx) for any x, yEA; (ii) Ta = JT*Ja for every T E OU(A). If a satisfies these conditions, then a is said to be central relative to A. The elements of A which commute with every element of A are central relative to A. Let Z be the set of central elements relative to A. The transforms by OU(A) (resp. rCA~ of the elements of Z generate a closed subspace Y (resp. Y) of H. We have Y = Y, and P y E OU(A) n V(A) is called the characteristic projection of A ([1], p. 75-76). A 63. With the notation of A 62, the following conditions are equivalent: (i) OU(A) is a finite von Neumann algebra; (ii) V(A) is a finite von Neumann algebra; (iii) P y = 1 ([1], p. 100, Th. 6). A 64. Let (A;) be a family of Hilbert algebras. Let the Hilbert space Hi be the completion of Ai Let A be the direct sum of the Ai which, in an obvious way, is an involutive algebra and a pre-Hilbert space. Then A is a Hilbert algebra called the direct sum Hilbert algebra of the AI The Hilbert space H which is the completion of A is the direct sum of the Hi We have OU(A) = IIiOU(A;), V(A) = IIir(Ai). If E, is the projection of H onto Hi we have E, E OU(A) n rCA), OU(A I ) = OU(A)Ej , rCA;) = V(A)Ej ([1], p. 76).

VON NEUMANN ALGEBRAS

385

A 65. Let A be a full Hilbert algebra, H the completion of A, (Ei ) a family of mutually orthogonal projections of O1L(A) n Y(A) whose sum is 1, and Hi == Ei(H). Let Ai == Ei(A) == Hi n A Then Ai is an involutive subalgebra of A, and is a Hilbert algebra dense in Hi The Ai mutually annihilate one another and are therefore two-sided ideals of A We have O1L(AJ == O1L(A)E, Y(AJ == Y(A)E The direct sum Hilbert algebra of the Ai is a dense subalgebra of A The natural trace of O1L(At induces on OU(A)~. == O1L(A it the natural trace of O1L(A;t ([1], p. 77).
I

A 66. Let A be the involutive algebra of the Hilbert-Schmidt operators on the Hilbert space H. Endowed with the scalar product (T IT) = Tr(T*T), this is a complete Hilbert algebra. Let If be the Hilbert space conjugate to H. There exists a unique isomorphism 4J of the Hilbert space A onto the Hilbert space H If which transforms the operator {~({ 11})~ (~E H, 1} E H) on H into ~ 1}. If we identify A with H If by means of 4J, we have, for ~, TJ E Hand S E 5f(H),
(~

1})* == (1}

~),

(~

1}) . S == ~

S*1}

([1], p. 95-97). A 67. Let A be a full Hilbert algebra such that O1L(A) is a factor. The following conditions are equivalent: (i) A is complete; (ii) O1L(A) is of type I; (iii) after multiplication of the norm of A by a constant, A is isomor phic to the Hilbert algebra of the Hilbert-Schmidt operators on a suitable Hilbert space. The above Hilbert space is uniquely determined by A up to isomor phism ([1], p. 127). A 68. Let A be a complete Hilbert algebra. Let (E;) be the family of minimal projections of O1L(A) n Y(A). (i) The Hilbert space A is the direct sum of the Ai == Ej(A). (ii) The A; are self-adjoint two-sided ideals of A mutually anni hilating each other. (iii) After multiplying the norm of A; by a constant, Ai is isomorphic to the Hilbert algebra of the Hilbert-Schmidt operators on a suitable Hilbert space. (iv) If OU(A) and Y(A) are finite von Neumann algebras, each Ai is finite-dimensional ([1], p. 127, Prop. 7).

386

APPENDIX A

A 69. Let Z be a Borel space (B 19) and JL a positive measure on Z (B 30). A JL-measurable field of Hilbert spaces over Z is a pair ~ = H(,ez, F), where (H(,ez is a family of Hilbert spaces indexed by Z, and where is a set of vector fields satisfying the following conditions: (i) is a vector subspace of TI,ez H(); (ii) there exists a sequence (XI X2, ) of elements of r such that, for every (E Z, the x n() form a total sequence in H(); (iii) for every X E T, the function ( ....llx()/1 is JL-measurable; (iv) let X be a vector field; if, for every y E r, the function ( .... (x(n I is JL-measurable, then X E Under these conditions, the elements of are called the measurable vector fields of ~. If x E rand y E T, then the function ( ....(x(D Iy( is measurable ([1], p. 142).

y(n

r.

A 70. Let Z, Z be Borel spaces; JL, JL positive measures on Z, Z, ~ = H(), r), ~' = H(t), F) JL-measurable and JL-measurable fields of Hilbert spaces. Let TJ: Z ....Z be a Borel isomorphism transforming JL into JL. An TJ-isomorphism of ~ onto ~' is a family (V(,ez possessing the following properties: (i) for every (E Z, V() is an isomorphism of H() onto H(TJ(; (ii) a vector field ( .... x(D E H() over Z is JL measurable, if and only if the field TJ(D.... V(Dx(D E H(TJ(n) over Z is JL-measurable ([1], p. 143). If Z =Z and JL = JL, an isomorphism of ~ onto ~' is just an TJ-isomorphism, where TJ is the identity mapping of Z into Z ([I], p. 143). A 71. Let Z be a Borel space, JL a positive measure on Z, and H o a separable Hilbert space. For every (E Z, put H() = R o Let r be the set of measurable mappings of Z into Hi; i.e. of the mappings x: Z .... H o such that ( ....(x() I a) is measurable for every a E H o Then ~ = H(D), n is a measurable field of Hilbert spaces over Z called the constant field defined by H o ([1], p. 143). A field isomorphic to a constant field (cf. A 70) is said to be trivial ([1], p. 143). A 72. We retain the notation of A 69. Let p = 1,2, ... , ~o. The set Z, of the (E Z such that dim H() = p is measurable. Let ~p be the restriction (defined in the obvious way) of ~ to Zp. Then each ~p is trivial ([1], p. 144-145, Prop. 1 and 3). A 73. Let
~ =

R(, r) be a JL-measurable field of Hilbert spaces over

VON NEUMANN ALGEBRAS

387

f Ilx)I/2 dJL({) < +00. If x,

(x

Z. A vector field x is said to be square-integrable if x E rand yare square-integrable, so are x + y and Ax, and {~(x({) Iy({)) is integrable; put

Iy)

(x({) Iy({)) dJL({).

Then the square-integrable vector fields constitute a Hilbert space H called the direct integral of the H) and denoted by fffi H({) dJL({). (We identify two fields which are equal almost everywhere.) An element x of this space is also written fffi x) dJL({). If JL is standard, then H is separable ([I], p. 145-146, and p. 149, Cor.). Let XI X2 E r. Let.Jl be the set of square-integrable vector fields obtained from multiplying the x;s by measurable complex-valued functions. If .Jl is total in H, then, almost everywhere in Z, (xM), xi{), . . .) is total in H({). ([1],p. 150, Prop.
8).

A 74. Go back to the notation of A 70. Then the mapping which trans forms the field x E H = fffi H({) dJL({) into the field

T/({)~

V({)x({) E H =

ffi

H({) dJL({)

is an isomorphism of H onto H denoted by fffi V) dJL({) ([1], p. 159, Def. 2). A 75. Let t; = H(n), F) be a JL-measurable field of Hilbert spaces over Z, {i a positive measure on Z equivalent to JL, and pee) = d{i({)/dJL(C). The mapping which, with every x E H =Jffi H ({) dJL makes the cor

(n.

responding field

(~p({)-1/2x(n

E fl =

ffi

H({) d{i({),

is an isomorphism of H onto fl, said to be canonical ([1], p. 148). A 76. Let Z be a Borel space, JL a positive measure on Z, H o a separable Hilbert space, t; = H({)), n the constant field on Z defined by H o There exists a unique isomorphism (said to be canonical) of L~(Z, JL)@H o onto fffi H({) dJL) which, for every f E L~(Z, JL) and every ~ E Hi; transforms f @ ~ into the field (~f({)~ ([1], p. 153, Cor.). A 77. Let t; = H({)), n be a measurable field of Hilbert spaces over Z.

388

APPENDIX A

For every {E Z, let T({) E 2(H(0). We say that {--,; T({) is a measurable field of operators if, for every x E T, the field {~ T({)x({) is measurable. If this is so, the function {~IIT(Oli is measurable. Suppose further that this function is essentially bounded, in which case the field {~T({) is said to be essentially bounded. Let H = f(f) H({) dj.L({). Then, for every x E H, the field Tx: {~ T(Ox({) belongs to H. We have T E 2(H), and IITII is the essential supremum of {~IIT(OII. We put T = f(f) T({) dj.L({). The T({) are determined by T to within negligible sets. The operators of the form f(f) T({) dj.L({) on H are said to be decomposable ([I], p. 156-159).
A 78. If S

=f(f)

S({) dj.L() and T

=f(f)

T(,) dj.L() are decomposable,

EEl

then we have

+T

f
EEl

(f)

(S()

+ T(,

dj.L(),

ST

AS =

f AS(O dj.L(O,

S*

f =f
EEl

S(,)T(,) dj.L(),

S()* dj.L().

([1], p. 159, Prop. 3).

A 79. Let T, =fEEl T;(n dj.L() (i = 1,2, ...) and T = fEEl T(O dj.L({) be decomposable operators. If T; converges strongly to T, there exists a subsequence (Tnt) such that T nt({) converges strongly to T(O almost everywhere. If T;(n converges strongly to T(,) almost everywhere and (if sup IIT;II < + 00, then T; converges strongly to T ([1], p. 160, Prop. 4).
A 80. The operators of the form fEEl T(,) dj.L(), where T({) is scalar for every, are called diagonalisable operators. Let fl be the set of diagon alisable operators. Then fl is a commutative von Neumann algebra, and fl is the set of decomposable operators ([1], p. 162, Def. 3; p. 163, Prop. 7; p. 164, Cor.).

A 81. Suppose that [;= n is the constant field defined by H o. Let S E 2(Ho). Put S({) = S for every," The canonical isomorphism of L~(Z, j.L)Q!)Ho onto fEEl H(,) dj.L(O transforms I Q!) S into f(f) S({) dj.L() ([1], p. 169-170). A 82. Let Z and Z be Borel spaces, and j.L and j.L positive measures on Z and Z.

H(,,

VON NEUMANN ALGEBRAS

389

Let ~ = H({EZ 0 be a IJ--measurable field of Hilbert spaces over Z. Let if}= H(C"EZ,I") be a IL-measurable field of Hilbert spaces over Z. Let I be a countable set. Let 1/ be a Borel isomorphism of Z onto Z transforming IL into IL. For every i E I, let {~ Ti ( { ) be a measurable field of operators relative to ~; let C ~ T;({) be a measurable field of operators relative to ~'. Suppose that, for every {E Z, there exists an isomorphism of H({) onto H(1/({ which, for every i E I, transforms Tj ( { ) into Ti(1/({. Suppose that IL and IL are standard. Under these hypotheses, there exists an 1/-isomorphism (A 70) of if} onto if}which, for every i E I, transforms the field {~ Ti(n into the field
{~

Ti(1/({

(We are at once led to the case where Z =Z, IL =IL, and where 1/ is the identity mapping of Z. Then, d. [1], p. 166, Lemma 2.)
A 83. Let if}= H({), 0, be a IJ--measurable field of Hilbert spaces over Z, T, =Jffi TM) dlL({) (i E 1) a family of decomposable operators on Jffi H({) dlL(n, H o a separable Hilbert space, and (S;)iEI a family of elements of 2(Ho). Suppose that, for every {E Z, there exists an isomorphism of H({) onto H o which, for every i, transforms Tj ( { ) into S; Suppose that IL is standard. Then there exists an isomorphism of Jffi HM) dlL({) onto L~(Z, 1J-)Ig)Ho which, for every t. transforms t, into 10 s, ([1], p. 167, Th. 2).

A 84. Let H be a separable Hilbert space, and 11: k 2(H) a commutative von Neumann algebra. There exist a standard Borel space Z, a bounded positive measure IL on Z, a measurable field {~H ({) of non-zero Hilbert spaces over Z, and an isomorphism of H onto Jffi H({) dlL({) which transforms 11: into the algebra of diagonalisable operators ([1], p. 210, Th. 2). A 85. Let Z be a Borel space, IJ- a standard positive measure on Z, if}= (H({ a IL-measurable field of non-zero Hilbert spaces over Z, H = Jffi H({) dlL({), and 11: the algebra of diagonalisable operators on H. Similarly, define ZI> ILl> ~I =(HMI HI>11:1, Let U be an isomorphism of H onto HI transforming 11: into [1 There exist: (i) a u-negligible Borel subset N of Z, and a ILrnegligible Borel subset N I of ZI; (ii) a Borel isomorphism 1/ of Z\N onto ZI\N I transforming IL into a measure iii equivalent to IJ-l; (iii) an 1/~isomorphism (V({ of if} restricted to Z\N

390

APPENDIX A

onto ~I restricted to ZI\N I, - such that U is the composition of fEf! V(~) d/L(~) and the canonical isomorphism of fEf! HM) dJi.M) onto HI ([1], p. 212, Th. 4). A 86. Let ~ = CCH(~, T) be a measurable field of Hilbert spaces, H = fEf! H(n d/Lcn, T, = fEf! Ti(n d/LCn a sequence of decomposable opera tors, d the von Neumann algebra generated by the T;, d(~) the von Neumann algebra generated by the Ti(~)' and :!l the algebra of di agonalisable operators. Then :!l is a maximal commutative von Neu mann subalgebra of d, if and only if d(~) = .xCH(~)) almost every where ([1], p. 172, Cor. I). A 87. Let ~ = CCH(~)), T) be a measurable field of Hilbert spaces over Z. For every ~ E Z, let d(~) be a von Neumann algebra on H(~). We say that ~ -+ dC~) is a measurable field of von Neumann algebras over Z if ... of measurable fields of there exists a sequence ~ -+ TM), ~ -+ T2C~), operators such that, for every ~ E Z, dC~) is the von Neumann algebra generated by the T;Cn. The set of the decomposable operators T = fEf! TC~) dJ.LC~) such that TC~) E dC~) for every ~ is a von Neumann algebra d on H = tIl HCn d/L(~), denoted by JEf! d(n d/L(~). The d(n are determined by d to within negligible sets. The von Neumann algebras of the form fEf! dC~) d/LC~) are said to be decomposable. Let :!l be the algebra of diagonalisable operators. Suppose that /L is standard. A ([1], von Neumann algebra d is decomposable if and only if :!l ~ d ~:!l' p. 173-174). A 88. Let

f lfJ TiC~)

~ = CCH(~)), T) be a measurable field of Hilbert spaces, T, = d/LCn a sequence of decomposable operators, :!l the algebra of

diagonalisable operators, d(~) the von Neumann algebra generated by the TjC~), and d the von Neumann algebra generated by :!l and the T; Then d =fEf! d(~) d/LC() ([1], p. 171, Th. 1). A 89. Let d = f lfJ d(~) d/L(~) be a decomposable von Neumann algebra. Let ~ be the algebra of diagonalisable operators. If :!l is the centre of d, then dC~) is a factor almost everywhere ([I], p. 174, Th. 3). A 90. Let d = f lfJ dCn d/LCn be a decomposable von Neumann algebra. Suppose that /L is standard. Then d is of type I, if and only if d(~) is of type I almost everywhere ([1], p. 183, Cor. 2 of Prop. 7). A 91. Let d = f lfJ d(~) d/L(~), and d 1 = f lfJ dtC{) d/L(~) be two decom posable von Neumann algebras. For every ~ E Z, let cP(~) be an

VON NEUMANN ALGEBRAS

391

isomorphism of d() onto d)(). The field of isomorphisms (--+ cP({) is said to be measurable if, for every T =fEB T() df.L() E d, the field (~cP()(T( is measurable. Then, if we put cP(T) = fEB cP()(T( df.L(), cP is an isomorphism of .st1 onto .st1!, denoted by fEB cP()df.L() ([I], p. 183 and 185). A 92. Let .st1 =fEB .st1() be a decomposable von Neumann algebra. For every (E Z, let f() be a trace on .st1(t. The field of traces (~f({) is said to be measurable if, for every measurable field of operators (~ T() E d(t, the function (~f()(T( is measurable. If T = fEB T() dJL() E : put f(T) ~ f f()(T( dJL(). Then f is a trace on d+, denoted by fEB f() dJL() ([1], p. 198). A 93. Let ~ = H(, n be a measurable field of Hilbert spaces over Z. For every (E Z, let A() be a Hilbert algebra such that the Hilbert space H() is its completion. The field (~A() of Hilbert algebras is said to be JL-measurable when the following conditions are satisfied: (i) if x E T is such that x() E A() for every (, the field x*: (~x()* is measurable; (ii) if x, y E r are such that x(), y() E A() for every (, the field xy : (~x()y() is measurable; (iii) there exists a sequence (XI> x 2, ) of elements of r such that x;() E A() for every i and every (, and such that the sequence (xM), x 2(), ...) is total in H() for every ( ([1], p. 187, Def. 1). A 94. We keep the notation of A 93. Let H = fEB H({) dJL({). Let A be the set of the x E H such that x() E A() for every ( and such that the field of operators (~UX({) (which is automatically measurable) is essentially bounded. Then, for the operations x~x*, (x, y)~xy and the scalar product induced by that of H, A is a Hilbert algebra dense in H, denoted by fEB A() dJL(). We then have

EB
OU(A) =

EB
V(A) =

JOU(A(dJL(),
ffi

JV(A( dJL().

If J (resp. J({ is the involution of H (resp. H(m defined by A (resp. A( we have

J

JJ() df.L().
fEB
t() df.L(), where t()

The natural trace on OU(At defined by A is

392

APPENDIX A

denotes the natural trace on OU(A()t defined by A() ([I], p. 188-190, and p. 199, Th. I). A 95. Let if = H(, n be a measurable field of Hilbert spaces over Z, H = fEB H() dJL(), !I the algebra of diagonalisable operators, and A a full Hilbert algebra whose completion is H. If OU(A) ~ !I, then there exists a measurable field (~A) of full Hilbert algebras, whose com pletions are the H(n, such that A = fEB A(,) dJL() ([I], p. 194, Th. 1). A %. Let Z be a Borel space. A Borel field if of Hilbert spaces over Z is a family (H(,Ez of Hilbert spaces together with a set T of vector fields satisfying the conditions (i)-(iv) of A 69, where we replace the phrase "u-measurable" by the word "Borel" throughout ([1], p. 143 and 144). We have properties analogous to those of measurable fields, which are established in the same way. In particular, we can rewrite A 71 and A 72, replacing "measurable" by "Borel" throughout. A 97. Let if = H(n), n be a Borel field of Hilbert spaces over Z. Let JL be a positive measure on Z. There exists exactly one set T:2 T of vector fields such that H(n), T) is a JL-measurable field of Hilbert spaces; T is the set of vector fields x such that, for every y E T, the is JL-measurable. The /-t-measurable field function ,~(x()y( H(, T) is said to be deduced from the Borel field H(n), T). ([I], p. 144). A 98. Let Z be a Borel space, (~H(,) a field of Hilbert spaces over Z, and (~Xj() (i = 1.2, ...) a sequence of vector fields having the follow ing properties: (i) for every (E Z, the sequence of the Xj() is total in H (); (ii) the functions (~(Xj() IXj( are Borel. Then there exists on the field (~H () a unique Borel field structure such that the (~x;) are Borel vector fields ([I], p. 145, Prop. 4 and p. 146, Remark).

APPENDIX B

MISCELLANEOUS

RESULTS

This Appendix is a collection, in no special order, of a wide variety of results used in the course of the book. The references are to the special bibliography, to be found at the end of the Appendix. B 1. Let A be a complex Banach algebra. For every x E A, IlxW I tends to a limit p(x) as n -+ +00, called the spectral radius of x. We have p(x):so; [x], and p(x) is the supremum of the set of [z] as z runs through Sp~x ([11], Th. 1.4.1 and 1.6.4). Every maximal regular left ideal of A is closed ([11], Cor. 2.1.4). B 2. Let A be a unital Banach algebra, B a closed subalgebra containing 1, and x E B. We have SPsX:1 SpAx, and every frontier point of SPBX relative to C belongs to SpAx ([11], Th. 1.6.12). B 3. Let A be a commutative Banach algebra. Every character of A is of norm :so; 1. The set of characters of A, endowed with the weak* topology, is a locally compact space S called the spectrum of A. For every x E A, the function X -+ X(x) on S is called the Gelfand transform of x and is denoted by fJix. The functions fJix (x E A) are continuous. separate the points of S, and, for every XES, there exists an x E A such that (fJix)(X) - O. If A is unital, the set of values of fJix is SPAX ([11], p. 110, Th. 3.1.6, Cor. 3.1.7. Th. 3.1.20). B 4. Let A be a unital Banach algebra, x E A, and I a function holomorphic in a neighbourhood of SPAX. There exists an element I(x) E A which possesses the following properties: (i) SpA!(x) =I(SPAX); (ii) if I(z) = ~~=o c.z" is an entire function, we have I(x) = ~~=o c.x", the series being norm-convergent ([11], p. 157-158 and references there cited. The hypothesis, made in [11], that A is commutative, is not really used. However, in the only place in the present book (1.3.9) where these results are applied, it would be easy to reduce matters to the case of commutative algebras).

394

APPENDIX B

B 5. Let E be a real locally convex space, C a closed convex cone with vertex 0, and x a point of E not belonging to C. There exists a continuous linear form on E which is ~o on C and <0 at x. (In fact, there exist a continuous linear form 1 on E and a real number a such that I(Y) ~ a on C and I(x) < a. We have 0 = 1(0);;;:a, hence I(x) < O. If I(Y) < 0 at any point of C, we have I(Ay) < a for A> 0 large enough, which is a contradiction.) B 6. Let E be a Hausdorff locally convex space, P a convex cone in E such that P n (- P) = 0, M a linear subspace of E, and 1 a linear form on M, ~o on M n P. Suppose that M contains an interior point of P. Then 1 extends to a linear form on E, ;;;:.0 on P ([4], p. 64, Cor.). B 7. Let E be a separable Banach space. The unit ball of the dual of E is compact metrisable for the weak*-topology ([5], p. 112, Prop. 2). B 8. Let E be a Banach space, E its dual, and K a convex subset of E. Then K is weak*-c1osed, if and only if the intersection of K with every closed ball of E is weak*-c1osed ([5], p. 74, Th. 5). B 9. Let H be a separable Hilbert space, and B the unit ball of 2(H) endowed with the weak topology. Then B is compact metrisable ([5], p. 25, Prop. 6 and p. 65, Cor. 3). B 10. Let H be a separable Hilbert space. The set of unitary operators on H is a polish space for the weak topology ([7], p. 280, Lemma 4). B 11. Let E, F be two locally convex spaces. Suppose that E is bar relled, and that F is Hausdorff and quasi-complete. Then 2(E, F) endowed with the topology of pointwise convergence, is Hausdorff and quasi-complete ([5], p. 31, Cor. 2). B 12. Let E, F be two topological vector spaces. Suppose that F is Hausdorff and quasi-complete. Then. in 2(E. F) endowed with the topology of pointwise convergence, every equicontinuous closed subset is a complete uniform subspace ([5], p. 30, Th, 4). B 13. Let E be a Hausdorff locally convex space, and C a metrisable compact convex subset of E. The set A of the extreme points of C is a OlJ of C ([8], p. 118-119). Let x E C. There exists a positive measure JL on C of total mass 1, concentrated on A (i.e. such that C\A is JL-negligible). and such that x = Ie y dJL(Y), in other words such that

MISCELLANEOUS

RESULTS

395

f(x) = Jc f(y) df.L(Y) for every continuous linear form f on E. (Choquets theorem: [9], Th.)

B 14. Let E be a Hausdorff locally convex space, C a compact convex subset of E, and A the set of extreme points of C. Then C is the closed convex hull of A (Krein-Milman theorem). If a subset A of C is such that the closed convex hull of A is equal to C, then A :2 A. We show that A is a Baire space. (This result, unpublished, is due to G. Choquet, as is the proof which follows.) We can suppose that E is real. For every continuous linear form f on E and every real number ex, denote by U"Cl (resp. F,.Cl) the set of the x E C such that f(x) < a (resp.
f(x)::;;; a). Let x E A. We first show that the set of the F,.a such that x E U,.Cl is a

base of neighbourhoods of x. By the Hahn-Banach theorem, the intersection of these Fi.Cl is {x}; since the F,.Cl are compact, it is enough to show that the family of the Fi,Cl such that x E U,.Cl is decreasing filtering. Let fl al h az be such that x E UIi .Cl I n Uf2, a2 Let C\. C z be the complements of UJ,Cl\ Uh,a2 in C. Since C I and C z are compact, the convex hull C3 of C I and C z is compact. It does not contain x because x is extreme. Hence there exists a continuous linear form f on E and a real number a such that f(x) < a, and f(y) > a for y E C 3 Then F"Cl meets neither C I nor C z i.e. is contained in UI1al n U2.a2 and x E U,.a Let (V n ) be a sequence of dense open subsets of A. We have to prove V n is dense in A, i.e. meets every non-empty open subset V of A. that Let Un U be open subsets of C, with U; dense in C, such that U; n A = V n , unA = V. We can suppose that the U; and the V n are decreasing. We can also suppose that U is of the form U,..al This established, we are going to prove the existence of (fl> al) (fz, az), ... with the following properties: F.+I.a.+, ~ U.a. n U n+ 1 and U,. a. n A# 0 for 11 = 1, 2..... We already have (fl. al) Suppose that (fl> al), ... , (fn, an) have already been constructed. There exists an x E U,.a. nAn Un+ 1 Then U.a. n Un+! is a neighbourhood of x in C, and there therefore exists (fn+1> an+!) such that x E U.+I.a.+1~ F.+I.a.+l ~ U,.a. n Un+!. Since x E A, we have Un+I.a.+1 n A # 0, and the inductive step is established. The F,. Cl. are decreasing, non-empty, and compact, and therefore have a non-empty intersection F. We have F ~ UI and F c u; Finally, F is compact convex, and its complement in C is convex. An easy lemma then proves that F contains at least one extreme point y of C. (The set F

3%

APPENDIX B

possesses an extreme point x. If x is extremal in C, we are done. If not, let j) be a line passing through x and such that x is an interior point of the segment C n B; we then show that one of the end-points of F n j) is an extreme point of C.) The point y belongs to A n U; = V, for every n, and to A n U/i.al = V. B IS. Every 0 8 in a polish space is a polish space ([2], p. 123, Th. 1). B 16. Let E be a locally compact space, 0 a group of homeomorphisms of E onto E, and C(E, E) the set of continuous mappings of E onto E, endowed with the topology of compact convergence. Suppose that 0 is relatively compact in C(E, E). Then the closure of 0 in C(E, E) is a compact group of homeomorphisms of E onto E ([3], p. 52. COL). B 17. Let E be a polish space, and R an equivalence relation on E. Suppose that the saturation in E of every open subset is Borel, and that the equivalence classes are closed. Then there exists a Borel subset of E which meets each equivalence class in exactly one point ([7], p. 279, Lemma 2). B 18. Let E be a Baire space, and [: E ~ R a lower semi-continuous function. Then / has a point of continuity. In fact, replacing t by /(l + 1/1)-1, we can suppose that / is bounded. For every x E E. let w(x) be the oscillation of / at x. Then oa is upper semi-continuous. and so the set En of the points where w(x) ~ lIn is closed. Suppose that En contains a non-empty open set U. Let a = SUPXEU f(x). There exists an Xo E Y such that /(x o) > a - 1/2n. Throughout a neighbourhood of xo, we have a - 1/2n < /(x) ~ a, whence w(x o) ~ 1/2n, which is a contradiction. Hence E\E n is open and dense, and so n(E\E n ) is non-empty. If x E n(E\E n ) , then w(x) = 0, and so / is continuous at x. B 19. A Borel space is a set E together with a family ga of subsets of E possessing the following properties: E E ga, 0 E ga, and ga is closed under countable unions, countable intersections, and the taking of complements. The elements of ga are called the Borel subsets of E. Let E be a set, and sti a collection of subsets of E. Among the families ga of subsets of E such that ga :1 sti and (E, ga) is a Borel space, there exists one gao, which is smaller than all the others. We say that the Borel structure of (E, gao) is generated by sti. Let (E, gal), (E, ga2) be two Borel spaces. If ga2:1 gat, we say that the Borel structure of (E, f1J 2) is finer than that of (E, ~I)'

MISCELLANEOUS

RESULTS

397

Let (EJiEl be a family of Borel spaces, E a set. and t.. E ~ E, mappings. The Borel structure on E defined by the Ii is the structure generated by the li 1(A ), where i runs through I and A runs through the set of Borel subsets of E; Let E, F be two Borel spaces. A mapping I: E ~ F is said to be Borel if the inverse image under I of every Borel subset of F is a Borel subset of E. Let E be a Borel space, and E a subset of E. The intersections with E of the Borel subsets of E define a Borel structure on E. The Borel space E is said to be a Borel subspace of E. Let E be a Borel space, and R an equivalence relation on E. The subsets of EIR whose inverse images in E are Borel define a Borel structure on EI R. The Borel space EIR is said to be the quotient Borel space of E by R. Let (E a ) be a family of Borel spaces. Let E be the disjoint union of the Ea. The subsets of E whose intersections with the E; are Borel define a Borel structure on E. The Borel space E is said to be the Borel sum of the Ea. Let E be a topological space. The Borel subsets of E for the topology define a Borel structure on E said to be subordinate to the topology. If E k; E, the Borel structure induced on E by that of E is the structure subordinate to the topology induced on E by the topology of E. (Concerning all the above, d. [10], p. 136-137].) B 20. A Borel space E is said to be standard if its Borel structure is subordinate to a polish space topology. If E is standard and countable, every subset of E is Borel. If E is standard and uncountable, then E is isomorphic to the Borel space [0, I] (endowed with the Borel structure subordinate to the usual topology). Every Borel subset of a standard space is standard. The sum of a sequence of standard spaces is standard ([10], p. 138). B 21. Let E" E 2 be two Borel spaces, and I: E 1 ~ E 2 an injective Borel mapping. Suppose that: (i) E) is standard; (ii) there exists a sequence of Borel subsets of E 2 separating the points of E 2 , and generating the Borel structure of E 2 (which is the case if E 2 is standard). Then !(E.) is Borel in E 2 and! is an isomorphism of E. onto I(E,) ([ 10], p. 139, Th. 3.2). B 22. Let E" E 2 be two Borel spaces, and [: E. ~ E 2 a bijective Borel mapping. Suppose that: (i) E 2 is standard; (ii) E, is the quotient of a

398

APPENDIX B

standard Borel space. Then I is an isomorphism of E I onto E 2 ([10], p. 149, Th. 4.2 and p. 141. Cor.). B 23. Let H be a Hilbert space, and S a linear operator defined on a dense linear subspace D s of H, with values in H, such that (SI If) ~ 0 for every IE o; For I, g E o; put (f Ig)s = (SI Ig) + (f Ig). Every Cauchy sequence for ( I)s is a Cauchy sequence in H, and two Cauchy sequences equivalent for ( I>s are equivalent in H. Hence the completion K of o, for ( , )s may be identified with a linear subspace of H. Let D be the intersection of K and the domain of definition of S*. Let S be the restriction of S* to D. Then S is self-adjoint and non-negative. We say that S is the Friedrichs extension of S. If a unitary operator U on H is such that USU- I = S, we have US U- 1 = S ([12], p. 35-36). B 24. Let G be a locally compact group, and H a closed subgroup such that the homogeneous space GI H is compact. If G is generated by a compact neighbourhood of e, the same is true of H ([6], p. 78, Lemma
3).

B 25. Let G be a commutative locally compact group generated by a compact neighbourhood of e. Then G is isomorphic to a group K x R" x ZP, where K is a commutative compact group ([13], p. 110). B 26. Let A be an algebra, and a and b elements of A. We have Sp~(ab) = Sp~(ba) (this is due to N. Jacobson). In fact, let A be the algebra obtained by adjoining an identity element e to A. It is enough to prove that, if A:j; 0 is such that ab - Ae has an inverse u in A, then ba - Ae is invertible in A. Now
(ba - Ae)(bua - e) = b(abu)a - ba - sbua
= b(Au

+ e)a

- ba -

+ Ae Xbua + Ae = Ae,

and similarly (bua - e)(ba - Ae) = Xe. Since A:j; 0, we see that ba - Ae is indeed invertible. B 27. Let A be an algebra over a field k of characteristic O. Let 7T, 7T be irreducible representations of A in finite-dimensional vector spaces over k. If Tr 7T(X) = Tr 7T(X) for every x E A, then 7T and 7T are equivalent ([I], p. 136, Prop. 3). B 28. Let T be a locally compact space, and IL a bounded complex valued measure on T. If IIILII =IL(I), we have IL ~ O. In fact, let 1I =IlLI

MISCELLANEOUS

RESULTS

399

We have II~II = Ilvll = v(1), and ~ = f v, where f is a ~-measurable complex-valued function such that If(t)1 = I for every t. Let a < 1 and let T; be the set of the t E T such that a :s;; Re f(t) (:s;; I). We have Ilvll
=

~(1)

Re

f(t) dv(t) + Re

:os:; a

dv(t)+

f f =f
dv(t)
T

f(t) dv(t)

dv(t)-(l-a)

dv(t)

= Ilvll-(1- a)

f
TIT.

dv(t).

Hence T\ T a is v-negligible. This being true for every a < 1, we have Re f(t) ~ 1 almost everywhere, f(t) = 1 almost everywhere, and ~ = v ~

o.
B 29. Let A be a normed algebra. An approximate identity of A is a net of elements of A, possessing the following properties: (a) Ilu;ll:s;; 1 for every i; (b) Ilu;x - xII ~ 0 and Ilxu; - xII ~ 0 for every x E A.
(U;);El

B 30. Let X be a Borel space. Let 973 be the set of Borel subsets of X. A positive measure on X is a mapping ~: 973 ~ [0, +00] such that: (I) if Xl X z, are mutually disjoint elements of 973,
~(Xl

U X,

u.. ) = ~(Xl)

+ ~(Xz)

+... ;

(2) X is the union of a sequence of Borel subsets Y\, Y z, such that ~(YJ < +00 for every i. A positive measure ~ on X is said to be

standard if there exists a Borel subset N of X such that X\N is a standard Borel space.

~(N)

0 and

B 31. A topological space is said to be separable if there is a countable base for its topology. When a space is metrisable, to say that it is separable is equivalent to saying that the space has a countable dense subset. A compact metrisable space is separable. B 32. In a topological space, a set is said to be a G8 if it is the intersection of a countable family of open sets. B 33. A character of a commutative algebra is a non-zero morphism of the algebra into the underlying field.

400

APPENDIX B

B 34. Let V be a finite-dimensional complex vector space, G a compact group, and 7T a continuous morphism of G into the linear group of V. There exists on Va scalar product invariant under 7T(G) which makes V a Hilbert space ([6], p. 70, Prop. 1).
Special bibliography for Appendix B
N. BOURBAKI, Algebra, chap. VIII (Act. Sc. Ind., no. 1261, Hermann, Paris, 1958). N. BOURBAKI. General topology. chap. IX. 3rd ed. (Hermann, Paris. 1974). N. BOURBAKI, General topology. chap. X, 3rd ed. (Hermann, Paris, 1974). N. BOURBAKI, Topological vector spaces, chap. I-II (2nd ed.) (Act. Sc. Ind. no. 1189. Hermann. Paris. 19(6). [5] N. BOURBAKI. Topological vector spaces, chap. III-IV-V (Act. Sc. Ind., no. 1229. Hermann, Paris, 1955). [6] N. BOURBAKI. Integration. chap. VII-VIII (Act. Se. Ind., no. 1306, Hermann, Paris, 19(3). [7] J. DIXMIER, Dual et quasi-dual dune algebre de Banach involutive (Trans. Amer. Math. Soc., 104, 1962, p. 278-283). [8] R. GODEMENT, Sur la theorie des representations unitaires (Ann. of Math., 53, 1951, p.68-124). [9] M. HERVE, Sur les representations integrates a Iaide des points extremaux dans un ensemble compact convex metrisable (C. R. Acad. Sc., 253, 1961, p. 366-368). [10] G. W. MACKEY, Borel structure in groups and their duals (Trans. Amer. Math. Soc., 85, 1957, p. 134-165). [11] C. E. RrCKART, General theory of Banach algebras (The University series in higher mathematics, D. Van Nostrand, Princeton, 19(0). [121 B. Sz. NAGY, Spektraldarstellung linearer transformationen des Hilbertschen Raumes (Erg. der Math., Springer-Verlag, Berlin, 1942). [13] A. WElL, Lintegration dans les groupes topologiques et ses applications, 2nd ed. (Act. Sc. Ind.. no. 1145, Hermann, Paris, 1953). [1] [21 [31 [41

INDEX OF NOTATION

c: 1.1.1.
Z (z, complex number): 1.1.2. .:(H) (H, Hilbert space): 1.1.2. SPAX. Sp x (x, element of a unital algebra A). 1.1.5. Sp,, Spx (x, element of an algebra A): 1.1.6.

M (M, subset of an algebra): 1.1.9. (f, linear form on an involutive algebra): 1.1.10. R: 1.3.9. f(x) (x, normal element of a C*-algebra; [, complex function): 1.5.2, 1.5.7. x+,x-,Ixl (x, hermitian element of a C*-algebra): 1.5.7. A + (A, C*-algebra): 1.6.5. x;;;" y, Y "" x (x, y, elements of a C*-algebra): 1.6.5. f;;;" R. R "" f <t. g. linear forms): 2.1.1. H~ (17", representation): 2.2.1, 13.1.3. 17" = 17". (17",17" representations): 2.2.1, 13.1.3. 8117""17", 81 . ffi17"n (17";, representations): 2.2.3, 13.1.3. C17" (c, cardinal; 17", representation): 2.2.3, 13.1.3. 17"K. 17" (17". representation; K. subspace of H~ invariant under 17"; E. projection on K): 2.2.4. 13.1.3. 17" "" 17", 17";;;" 17" (17", 17", representations): 2.2.4, 13.1.3. fTo (17", representation): 2.2.8. A (A, involutive algebra): 2.3.2. w( (~, vector of a Hilbert space): 2.4.2. 17"/> ~f (f, positive form): 2.4.4. P(A) (A, normed involutive algebra): 2.5.2. Prim(A) (A, algebra): 2.9.7. A" A (A, C*-algebra; I, closed two-sided ideal): 2.11.2. Prim,(A), Prim(Al.{A, C*-algebra; I, closed two-sided ideal): 2.11.5. P, (A), P I (A) (A, C*-algebra; I, closed two-sided ideal): 2.11.8. Prim(cp), q, (cp, automorphism of an algebra): 3.1.9. Rep.(A), Irr.(A), An (A, C*-algebra): 3.5.1. Rep(A), Irr(A) (A, separable C*-algebra): 3.8.1. ~(H) (H, Hilbert space): 4.1.1. m(A) (A, C*-algebra): 4.5.2. J(A) (A, C*-algebra): 4.7.12. 17" 017"(17",17", representations): 5.2.2. 17" = 17" (17",17", representations): 5.3.2.

f*

402

INDEX OF NOTATION

1, t,

(j, trace): 6.1.2,17.1.3. n, (s, bitrace): 6.2. I. N .. An H.. J.. A.. o, (s, bitrace): 6.2.2. "Il.. Y, (s, bitrace): 6.2.3. I, (s, bitrace): 6.2.4. n; A f, n; J{o x; Pf u; J[, If (j, trace): 6.4.5, 17.1.3. m, (s, bitrace): 6.4.5. Fac.(A), Fac(A) (A, separable C*-algebra): 7.1.3. A (A, separable C*algebra): 7.2.1. Fac Ip(A), Fac I(A), A, (A, separable C*-algebra): 7.3. I. Facf.(A), Facf(A), Af (A, separable C*-algebra): 7.4.1. Iff! IT(n d/.L(n (c -> IT(C), measurable field of representations): 8.1.3. It Cd/J.(n (A, separable postliminal C*-algebra. JL positive measure on A): 8.6.3. 8f.~ (~, 1/, vectors of a Hilbert space): 10.6.I. a(H,~) (H, Hilbert space; ~ E H): 10.6.3. (3(A, p) (A, elementary C*-algebra; pEA): 10.6.5. d(K) (:rt, continuous field of Hilbert spaces): 10.7.2. a(:rt, x) (:rt, continuous field of Hilbert spaces; x, continuous vector field): 10.7.5. (3(d, p) (d, continuous field of elementary C*-algebras;p, continuous field of projections): 10.7.5. U: 10.7.10. Sed) (d, continuous field of elementary C*-a1gebras satisfying Fells condition): 10.7.14. SeA) (A, separable C*algebra with continuous trace): 10.9.1. E(A) (A, C*algebra): 11.2.3. e, g- (g, hermitian linear form): 12.3.5. IT IT (?T, IT, representations): 13.1.5. iT (IT, representation): 13.1.5. t; (H, Hilbert space): 13.1.7. M(G) (G, locally compact group): 13.2.1. :K(G) (G, locally compact group): 13.2.2. V(G) (G, locally compact group): 13.2.2. f*, t; .f (j, function on a group): 13.2.3. IT., ~. (cp, continuous positive-definite function): 13.4.6. :1>0: 13.7.1, 13.7.6. C*(G) (G, locally compact group): 13.9.1. AP(G) (G, topological group): 16.2.2. (G, topological group): 13.1.3, 18.1.1. Rep.(G) (G, topological group): 18.1.9. Irr.(G) (G, topological group): 18.LlO. A, (G, locally compact group): 18.3.1. Rep(G), Irr(G) (G, separable locally compact group): 18.5.1. Fac.(G), Fac(G), Fac Ip(G), Fac uo: (G, separable locally compact group): 18.6.1. (G, separable locally compact group): 18.6.2. (G, separable locally compact group): 18.6.3. Facf.(G), Facf(G), Of (G, separable locally compact group): 18.6.4. m /2 (m, ideal): A 9.

m,

a,

INDEX OF NOTATION

403

.54E (.54, von Neumann algebra: E, projection): A 15. II.54, (.54" von Neumann algebras): A 16. .54, .54 2 (.54" .542 , von Neumann algebras): A 18. Tr: A30. E - F (E, F, projections): A 41. E < F, E> F (E, F, projections): A 41. OU(A), Y(A) (A, Hilbert algebra): A54. Iff> H() df.LW U H({), measurable field of Hilbert spaces): A 73. I$ V() df.LW U V(), isomorphism): A 74. I ffi T() df.LW ( T(), measurable field of operators): A 77. J$ .54() dIL({) ({ .54({), measurable field of von Neumann algebras): A 87. J$ <P({) dILW ({ <P({), measurable field of isomorphisms): A 91. I$ IW dILW ({ IW, measurable field of traces): A 92. I$ AW dILW ({ A({), measurable field of Hilbert algebras): A 94. p(x) (x, element of a Banach algebra): B I. q;x (x, element of a commutative Banach algebra): B 3. G.: B 32.

INDEX OF TERMINOLOGY
Abelian projection A 35 Adjoint I. 1.1 Algebra decomposable von Neumann - A 87 direct sum Hilbert - A 64 enveloping von Neumann - 12.1.4 Hilbert- A 54 Hilbert - of a group 14.1.1 induced. reduced von Neumann - A 15 involutive - I. 1.1 involutive Banach normed involutive - 1.2.1 normed involutive - obtained by ad junction of an identity 1.2.3 von Neumann A 4 von Neumann - generated by a set A
II

Algebraically equivalent representations 2.9.6 Ampliation A 21 Antiliminal - C*algebra 4.3.1 - group 13.9.4 Bi-invariant subspace 14.3.1 Bitrace 6.2.1, 17.1.1 - associated with a trace 6.4.1 - of type I,II. . .. 6.6.4 maximal- 6.).3 Bounded element relative to a Hilbert space A 56 C*-algebra 1.3.1 conjugate - of a - 1.9.1 enveloping - 2.7.7 - defined by a continuous field of - s 10.4.1 - defined by a continuous field of Hilbert spaces 10.9.1 - of a group 13.9.1 - with continuous trace 4.5.2 - with generalised continuous trace 4.7.12

Central - element relative to a Hilbert algebra A 62 -linear form 6.8.1 - function 15.3.2 Centre, ideal. - associated with a representation 5.6.7 Character normalised - of a compact group -defined by a distribution 17.2.7 - defined by a function. by a measure 17.2.3 -of a C*-algebra 6.7.1 - of a commutative algebra B 33 -of a compact group 15.3.1, 15.3.6 -of a locally compact group 17.1.1 - of a traceable factor representation 6.7.4 Class - of measures associated with a represent" ation 8.4.3 - of representations 2.2.1. 13.1.3 Coefficient of a representation 13.1.1 Completely positive mapping 2.12.16 Condition. Fells - 10.5.7 Conjugate - Hilbert space 2.2.8 - representation 13.1.5 Contained representation 2.2.4. 13.1.3 Continuous von Neumann algebra A 38 C*-tensor product 2.12.15 C*-semi-norm 1.9.13 Cyclic vector - for an algebra A 14 - for a representation 2.2.5. 13.1.3 Decomposition polar-of a linear form 12.2.7 enveloping polar- of a linear form 12.2.8 Dimension formal- of a representation 14.4.4 - of a representation 2.2.1, 13.1.3 - of a topological space 10.8.6

INDEX OF TERMINOLOGY

405

Disintegration, central-of a represen tation 8.4.3 Disjoint representations 5.2.2, 13.1.4 Dual reduced - 18.3.1 -of a locally compact group 18.1.1 - C*-algebra 4.7.20 Element Hilbert-Schmidt - in an elementary C* algebra 4.1.9 Hilbert-Schmidt - relative to a factor A 32 - with trace relative to a factor A 32 Elementary C*-algebra 4.1.1 Equivalent - projections A 41 - representations 2.2.1, 13.1.3 Extension canonical-of a positive form 2.1.5 canoniCal-of a representation 2.2.9 maximal canonical - of a bitrace 6.3.9 Factor A 6 - representation 5.2.6, 13.1.4 Faithful trace A 28 Field Borel- of Hilbert spaces A 96 canonical Borel- of Hilbert spaces over A 8.6.1 constant- 10.1.4, A 71 continuous - of Banach spaces 10.1.2 continuous - of C*-algebras 10.3.1 continuous - of C*-algebras defined by a C*-algebra 10.5.1 continuous - of elementary C*-algebras 10.3.1 continuous - of. elementary C* algebrasassociated with a continuous - of Hilbert spaces 10.7.2 continuous vector - 10.1.2, 10.1.6 induced - 10.1.7 locally trivial- 10.1.8 measurable- of Hilbert algebras A 93 measurable- of Hilbert spaces A 69 measurable- of isomorphisms A 91 measurable - of operators A 77 measurable - of representations 8.1.1, 18.7.1 measurable - of traces A 92 measurable - of von Neumann alge bras A87 measurable vector - A 69

- obtained by joining together 10.1.13 square-integrable vector - A 73 trivial- 10.1.4, A 71 Finite von Neumann algebra A 31 Form positive - associated with a represen tation 2.4.2 positive - associated with a set of representations 2.4.2 positive - defined by a representation and a vector 2.4.2 positive vector - 2.4.2 Friedrichs extension B 23 Full Hilbert algebra A 56 Function almost periodic - 16.2.2. positive-definite - 13.4.1, 13.7.6 positive-definite - associated with a representation 13.4.6 positive-definite - associated with a set of representations 13.4.6 positive-definite - defined by a repres entation and a vector 13.4.6 pure positive-definite - 13.6.1 Group compact - associated ogical - 16.1.2 unitary - I. 1.6 Hermitian -element 1.1.3 - linearform 1.1.\0 Homogeneous - C*-algebra 10.9.3 - representation 5.7.6 Identity approximate - B 29 increasing approximate - I. 7.1 Induction of a von Neumann algebra A 15 Injectable, group - in a compact group 16.4.2 Integral direct-of Hilbert spaces A 73 direct - of representations 8.1.3, 18.7.5 Involution 1.1.1 Irreducible representation 2.3.2, 13.1.4 Isomorphism 1.1.7, 1.2.4, 10.1.3, A 19, A 70 Liminal - C*algebra 4.2.1 with a topol

406
- group 13.9.4 Linear topological group

INDEX OF TERMINOLOGY

tensor - of von Neumann algebras A 15.5.1

18

Mapping Borel- B 19 canonical- A -+ Prim(A) 2.9.7 canonical- P(A)-+A 2.5.4 Mean of an almost-periodic function 16.3.2 Measure Plancherel- 18.8.3 positive-definite - 13.7.1 positive - on a Borel space B 30 standard- B 30 . Minimal projection A 34 Moderated 13.8.1 Morphism 1.1.7. 1.2.4 Multiple of a representation 2.2.3. 13.1.3 Multiplicity representation of - n 5.4.8. 13.1.4 - free representation 5.4.5. 13.1.4 Non-degenerate representation 2.2.6 Normal Norm topology A I - element 1.1.3 - morphism A 27 - positive form A 25 -trace A 28 Number, intertwining - 2.2.2. 13.1.3 Operator decomposable - A 77 diagonalisable- A 80 Hilbert-Schmidt - A 32 intertwining - 2.2.2, 13.1.3 trace-class- A 32 Positive - and negative parts of a hermitian linear form 12.3.5 - element of a C*-algebra 1.6.5 -linear form 2.1.1 Postliminal -C*-algebra 4.3.1 - group 13.9.4 Power, symmetric tensor - of a representation 13.10.11 Predual of a von Neumann algebra A 23 Prime. two-sided ideal- 1.9.13 Primitive. two-sided ideal- 2.9.7 Product restricted - of C*-algebras 1.9.14

tensor - of two representations 13.1.5 - of C*-algebras 1.3.3 - of von Neumann algebras A 16 Projection 1.1.3 characteristic- of a Hilbert algebra A 62 central- associated with a represen tation 14.3.4 Properly infinite von Neumann algebra A 31 Pure positive form 2.5.2 Purely infinite von Neumann algebra A 31 Quasi-dual 18.6.2 Quasi-equivalent representations 5.3.2, 13.1.4 Quasi-equivalent trace representations 6.6.2, 17.1.2 Quasi-spectrum 7.2.2 Radius. spectral- B I Rank - of a continuous field of elementary C*algebras 10.8.3 - of an elementary C*-algebra 4.1.9 Real representation 13.10.13 Regular right - representation 13.1.6 left - representation 13.1.6. 13.9.3 - representation 13.1.6 Representation continuous unitary - 13.1.1 integrable - 14.6.2 square-integrable- 14.2.3 trace- 6.6.1. 17.1.2 universal- 2.7.6 - defined by a positive-definite function 13.4.6 - defined by a positive form 2.4.4 - of an involutive algebra 2.2.1 Rich sub-C*-algebra 11.1.1 Self-adjoint set 1.1.1 Semi-finite - trace 6.1.1 - von Neumann algebra A 31 Separable - continuous field 10.2.1 - topological space B 31

INDEX OF TERMINOLOGY

407

Separated point 3.9.4 Separation of one set by another 11.1.7 Separator A 14 Sequence, composition - of a C*-algebra 4.3.2 Singular, mutually - positive forms 12.3.2 Space Borel- B 19 quotient Borel- B 19 sum Borel- B 19 standard Hilbert - 3.5.1 - of a representation 2.2.1, 13.1.3 Spectrum 3.5.1 - of a commutative Banach algebra B 3 Standard Borel space B 19 State 2.1.1 Strong topology A I Structure BoreI- defined by functions B 19 Borel- subordinate to a topology B 19 finer Borel- B 19 generated Borel - B 19 Mackey Borel- 3.8.2, 7.2.2. 18.5.3. 18.6.2 Sub-algebra generated closed involutive - 1.2.5 generated involutive - 1.1.8 Subrepresentation 2.2.4 Subspace bi-invariant- associated with a repre sentation 14.3.4 Borel- B 19 essential- 22.2.6 Sum, direct - of representations 2.2.3, 13.1.3 Support central- A 10 enveloping - of a positive form 12.3.6 - of a normal positive form A 26 - of a representation 3.4.6

To-space 3.1.3 Table, character- 15.4.2 Topology, Jacobson - 3.1.1 Topologically irreducible representation 2.3.2, 13.1.4 Total subset 10.2.1 Trace 6.1.1, 17.1.1 natural - A 60 - associated with a bitrace 6.4.2 - of type 1, II, . .. 6.6.4, 17.1.3 - representation 6.6.1 Traceable representation 6.6.7, 17.1.4 Transformation. Gelfand - B 3 Trivial representation 13.1.7 Type -I group 13.9.4 - I involutive algebra 5.5.1 - I representation 5.4.1, 13.1.4 - I von Neumann algebra A 35 - In von Neumann algebra A 47 - II, III representation 5.6.1, 13.1.4 - II, III von Neumann algebra A 38 - IIIIt. finite type representation 5.6.7. 13.1.4 - IllII.. von Neumann algebra A 40 Ultra-weak topology A 1 Ultra-strong topology A 1 Unitary element 1.1.5 Value. absolute-of a linear form 12.2.7. 12.2.8 Vector - defined by a positive-definite func tion 13.4.6 - defined by a positive form 2.4.4 Weakly contained 3.4.5. 18.1.3 Weak topology A I

REFERENCES
J. F. AARNES, The Vitali-Hahn-Saks theorem for Von Neumann algebras, Math. Scand. 18 (1966) 87-92. (2) J. F. AARNES, On the continuity of automorphic representations of groups, Commun. Math. Phys.7 (1%8) 332-336. [3] J. F. AARNES, Physical states on C*-a1gebras, Acta Math. 122 (1969) 161-172. [4] J. F. AARNES, Full sets of states on a C*-algebra, Math. Scand. 26 (1970) 141-148. [5) J. F. AARNES, Quasi-states on C*-a1gebras, Trans. Amer. Math. Soc. 149 (1970) 601-625. (6) J. F. AARNES, Continuity of group representations with applications to C* algebras, J. Functional Anal. 5 (1970) 14-36. [7] J. F. AARNES, E. G. EFFROS and D. A. NIELSEN, Locally compact spaces and two classes of C*-algebras, Pacific J. Math. 34 (1970) 1-16. [8] J. F. AARNES and R. V. KADISON, Pure states and approximate identities, Proc. Amer. Math. Soc. 21 (1969) 749-752. [9] S. M. ABDALLA and J. Szucs, On an ergodic-type theorem for Von Neumann algebras, Acta Sci. Math. 36 (1974) 167-172. (10) L. ACCARDI, The noncommutative Markovian property, Funkcion. Anal. Priloi: 9 (1975) 1-8 (in Russian). [II] C. A. AKEMANN, The dual space of an operator algebra, Trans. Amer. Math. Soc. 126, (1967) 286-302. [12] C. A. AKEMANN, Projections onto separable C*-subalgebras of a W*-algebra, Bull. Amer. Math. Soc. 73 (1967) 925. [13) C. A. AKEMANN, Sequential convergence in the dual of a W*-a1gebra, Commun. Math. Phys. 7 (1%8) 222-224. [14) C. A. AKEMANN, Interpolation in W*-algebras, Duke Math. J. 3S (1968) 525-533. [l5) C. A. AKEMANN, The general Stone-Weierstrass problem, J. Functional Anal. 4 (1969) 277-294. [16) C. A. AKEMANN, Approximate units and maximal abelian C*-subalgebras, Pacific J. Math. 33 (1970) 543-550. [17] C. A. AKEMANN, Separable representations of a W*-a1gebra, Proc. Amer. Math. Soc. 24 (1970) 354-355. [18] C. A. AKEMANN, Left ideal structure of C*-algebras, J. Functional Anal. 6 (1970) 305-317. [l9] C. A. AKEMANN, A Gelfand representation for C*-a1gebras, Pacific J. Math. 39 (1971) I-II. [20] C. A. AKEMANN, P. G. DoDDS and J. L. B. GAMLEN, Weak compactness in the dual space of a C*-algebra, J. Functional Anal. 10 (1972) 446-450. [21] C. A. AKEMANN, G. A. ELLIOTT, G. K. PEDERSEN and J. TOMIYAMA, Derivations and multipliers of C*-algebras, Amer. J. Math. 98 (1976) 679-708.
(1)

REFERENCES

409

[22) C. A. AKEMANN and S. M. NEWBERGER, Physical states on a C.-algebra, Proc. Amer. Math. Soc. 40 (1973) 500. [23] C. A. AKEMANN and P. A. OSTRAND, On a tensor product C-algebraassociated with the free group on two generators, J. Math. Soc. Japan 27 (1975) 589-599. [24] C. A. AKEMANN and P. A. OSTRAND, The spectrum of a derivation of a C*-algebra, J. London Math. Soc. 13 (1976) 525-530. [25] C. A. AKEMANN and G. R. PEDERSEN, Complications of sernicontinuity in C*-algebra theory, Duke Math. J. 40 (1974) 785-795. (26) C. A. AKEMANN, G. R. PEDERSEN and J. TOMIYAMA, Multipliers of C*-a1gebras, J. Functional Anal. 13 (1973) 277-301. [27] C. A. AKEMANN and M. ROSENFELD, Maximal one-sided ideals in operator algebras, Amer. J. Math. 94 (1972) 723-728. [28] C. A. AKEMANN and B. Russo, Geometry of the unit sphere of a C-algebraand its dual, Pacific J. Math. 32 (1970) 575-585. [29] C. A. AKEMANN and M. E. WALTER, Nonabelian Pontriagin duality Duke Math. J. 39 (1972) 451-463. [30) C. A. AKEMANN and M. E. WALTER, The Riemann-Lebesgue property for arbitrary locally compact groups, to appear. [31] F. E. ALEXANDER and B. J. TOMIUK, Complemented B*-algebras, Trans. Amer. Math. Soc. 137 (1969) 459-480. [32] E. M. ALFSEN, Facial structure of compact convex sets, Proc. London Math. Soc. 18 (1968) 385-404. [33] E. M. ALFSEN and T. B. ANDERSEN, Split faces of compact convex sets, Proc. London Math. Soc. 21 (1970) 415-442. [34] E. M. ALFSEN and F. W. SHULTZ, Non-commutative spectral theory for affine function spaces on convex sets, I, II, Mem. Amer. Math. Soc. 6 (1976) 172. [35] G. R. ALLAN, A note on B-algebras, Proc. Cambridge Phi/os. Soc. 61 (1965) 29-32. [36] R. A. ALO, C. CHENEY and A. DE KORVIN, An abstract Radon-Nikodyrn theorem, J. Math. Anal. Appl. 42 (1973) 110-118. [37] R. A. ALO, C. A. CHENEY and A. DE KORVIN, Martingales in Banach *-algebras, J. Reine Angew. Math. 273 (1975) 49-56. [38] W. AMBROSE, Structure theorems for a special class of Banach algebras, Trans. Amer. Math. Soc. 57 (1945) 364-386. [39] W. AMBROSE, The V-system of a unimodular group, I, Trans. Amer. Math. Soc. 65 (1949) 27-48. [40] I. AMEMIY A and H. ARAKI, A remark on Pironspaper, Publ. Res. Inst. Math. Sci. Kyoto Univ. 2 (1967) 423-427. [41] S. ANASTASIO, Maximal abelian subalgebras in hyperfinite factors, Amer. J. Math. 87 (1965) 955-971. (42) S. ANASTASIO, Non-normal abelian subalgebras, T6hoku Math. J. 20 (1968) 26-37. [43] S. ANASTASIO. Some recent results in W*-algebras, Trans. New York Acad. Sci. 33 (1971) 405-410. [44] S. ANASTASIO and H. WILUG, The Structure of Factors (Algorithmics Press, New York, 1974). [45) T. B. ANDERSEN, On multipliers and order bounded operators in C-algebras, Proc. Amer. Math. Soc. 25 (1970) 896-899.

410

REFERENCES

[46] T. B. ANDERSEN, Linear extensions projections and split faces, J. Functional Anal. 19 (1974) 161-173. [47] J. ANDERSON, Derivation ranges and the identity, Bull. Amer. Math. Soc. 79 (1973) 705-708. [48] J. ANDERSON, Extreme points in sets of positive linear maps on 9Il(m. to appear. [49] J. ANDERSON, Extensions, restrictions, and representations of states on C* algebras, to appear. [50] J. ANDERSON, On pure states of C*-algebras, Proc. Amer. Math. Soc. 55 (1976) 471. [51] J. ANDERSON and J. BUNCE, A type II~ factor representation of the Calkin algebra, to appear. [52] V. V. ANSELEVIC, The central limit theorem in "non commutative" probability theory, Dokl. Akad. Nauk SSSR 208 (1973) 1265-1267. [53] C. ApOSTOL, Sur les prolongements des representations des ideaux dune algebre, Mathematica 9 (1967) 5-7. [54] C. ApOSTOL and L. ZSIDO, Ideals in W*-algebras and the function I) of A. Brown and Pearcy, Rev. Roumaine Math. Pures Appl. 18 (1973) 1151-1170. [55] H. ARAKI, A generalization of Borchers theorem, Helv. Phys. Acta 36 (1963) 132-139. [56] H. ARAKI, A lattice of Von Neumann algebras associated with the quantum theory of a free bose field, J. Math. Phys. 4 (1963) 1343-1362. [57] H. ARAKI, Von Neumann algebras of local observables for free scalar field, J. Math. Phys. 5 (1964) 1-13. [58] H. ARAKI, On the algebra of all local observables, Progr. Theor. Phys. 32 (1964) 844-854. [59] H. ARAKI, Type of Von Neumann algebra associated with free field, Prog. Theor. Phys. 32 (1964) 956-965. [60] H, ARAKI, Local quantum theory, I, Proc. Intem. School Phys. E. Fermi 15 (1968) 65-96. [61] H. ARAKI, Multiple time analyticity of a quantum statistical state satisfying the KMS boundary condition, Publ. Res. lnst. Math. Sci. Kyoto Univ. 4 (1968) 361-371. [62] H. ARAKI, A classification of factors, II, Publ. Res. Inst. Math. Sci. Kyoto Univ. 4 (1968) 585-593. [63] H. ARAKI, On the diagonalization of a bilinear Hamiltonian by a Bogoliubov transformation, Publ. Res. Inst. Math. Sci. Kyoto Univ. 4 (1968) 387-412. [64] H. ARAKI, Systeme a nombre infini de degres de liberte, Colloq. Intern. C.N.R.S. 181 (1969) 75-86. [65] H. ARAKI, Gibbs states of a one-dimensional quantum lattice, Commun. Math. Phys. 14 (1969) 120-157. [66] H. ARAKI, A remark on Bures distance function for normal states, Publ. Res. Inst. Math. Sci. Kyoto Univ. 6 (1970) 477-482. [67] H. ARAKI, Product states, Cargese Lectures Phys. 4 (1970) 1-30. [68] H. ARAKI, On quasi-free states of CAR and Bogoliubov automorphisms, Publ. Res. Inst. Math. Sci. Kyoto Univ. 6 (1970) 385-442. [69] H. ARAKI, On representations of the canonical commutation relations, Commun. Math. Phys. 20 (1971) 9-25. [70] H. ARAKI, Asymptotic ratio set and property L~, Publ. Res. Inst. Math. Sci. Kyoto Univ. 6 (1970) 443-460.

REFERENCES

411

[71] H. ARAKI, On quasi-free states of the canonical commutation relations I, Publ. Res. Inst. Math. Sci. Kyoto Univ. 7 (1971) 105-120. [72] H. ARAKI, On quasi-free states of the canonical commutation relations II, Pub/. Res. Inst. Math. Sci. Kyoto Univ. 7 (1971) 121-152. [73] H. ARAKI, Bures distance function and a generalization of Sakais noncommutative Radon-Nikodym theorem, Publ. Res. Inst. Math. Sci. Kyoto Univ. 8 (1972) 335-362. [74] H. ARAKI, Remarks on spectra of modular operators of Von Neumann algebras, Commun. Math. Phys. 28 (1972) 267-277. [75] H. ARAKI, Normal positive linear mappings of norm 1 from a Von Neumann algebra into its commutant, and its application, Publ. Res. Inst. Math. Sci. Kyoto Univ. 8, (1973) 439-470. [76] H. ARAKI, Structure of some Von Neumann algebras with isolated discrete modular spectrum, Pub/. Res. Inst. Math. Sci. Kyoto Univ. 8 (1975) 1-44. [77] H. ARAKI, Relative hamiltonian for faithful normal states of a Von Neumann algebra, Pub/. Res. Inst. Math. Sci. Kyoto Univ. 9 (1973) 165-209. [78] H. ARAKI, Golden-Thompson and Peierls-Bogoliubov inequalities for a general Von Neumann algebra, Commun. Math. Phys. 34 (1973) 167-178. [79] H. ARAKI, On the equivalence of the KMS condition and the variational principle for quantum lattice systems, Commun. Math. Phys. 38 (1974) 1-10. [80] H. ARAKI, One-parameter family of Radon-Nikodym theorems for states cf a Von Neumann algebra, Pub/. Res. Inst. Math. Sci. Kyoto Univ. 10 (1974) 1-10. [81] H. ARAKI, Some properties of modular conjugation operator of Von Neumann algebras and a noncommutative Radon-Nikodym theorem with a chain rule, Pacific J. Math. 50 (1974) 309-354. [82] H. ARAKI, On uniqueness of KMS states of one-dimensional quantum lattice systems, Commun. Math. Phys. 44 (1975) 1-7. [83] H. ARAKI, Inequalities in Von Neumann algebra, to appear. [84] H. ARAKI, Relative entropy of states of Von Neumann algebras, Publ. Res. Inst. Math. Sci. Univ. 11 (1976) 809-834. [85] H. ARAKI, Relative entropy and its applications, to appear. [86] H. ARAKI, Positive cone, Radon-Nikodym theorems, relative Hamiltonian and the Gibbs condition in statistical mechanics - an application of the Tomita-Takesaki theory, to appear. [87] H. ARAKI and G. A. ELLIOTT, On the definition of C*-algebras, Pub/. Res. Inst. Math. Sci. Kyoto Univ. 9 (1973) 93-112. [88] H. ARAKI and P. D. F. ION, On the equivalence of KMS and Gibbs conditions for states of quantum lattice systems, Commun. Math. Phys. 35 (1974) 1-12. [89] H. ARAKI and H. MIYATA, On KMS boundary condition, Publ. Res. Inst. Math. Sci. Kyoto Univ. 4 (1968) 373-385. [90] H. ARAKI and Y. NAKAGAMI, A remark on an infinite tensor product of Von Neumann algebras, Pubi. Res. Inst. Math. Sci. Kyoto Univ. 8 (1972) 363-374. [91] H. ARAKI, M. S. B. SMITH and L. SMITH, On the homo topical significance of the type of Von Neumann algebra factors, Commun. Math. Phys. 22 (1971) 71-88. [92] H. ARAKI and E. 1. WOODS, Representation of the canonical commutation relations describing a nonrelativistic infinite free Bose gas, J. Math. Phys, 4 (1963) 637-662. [93] H. ARAKI -md E. J. WOODS, Complete boolean algebras of type I factors, Publ. Res. Inst. Math. Sci. Kyoto Univ. 2 (1966) 157-242,451-452.

412

REFERENCES

[94] H. ARAKI and E. J. WOODS, A classification of factors, Publ. Res. Inst. Math. Kyoto Univ. 4 (1968) 51-130. [95] H. ARAKI and W. WYSS, Representations of canonical anticommutation relations, Helv. Phys. Acta 37 (1964) 136-159. [96] R. J. ARCHBOLD, Prime C*-algebras and antilattices, Proc. London Math. Soc. 24 (1972) 669-680. [97] R. J. ARCHBOLD, Order and commutativity in C*-algebras, Proc. Cambridge Phi/os. Soc. 76 (1974) 153-155. [98] R. J. ARCHBOLD, Density theorems for the centre of a C*-algebra, J. London Math. Soc. 10 (1975) 189-197. [99] R. J. ARCHBOLD, On the centre of a tensor product of C*-a1gebras, J. London Math. Soc. 10 (1975) 257-262. [100] R. ARENS, On a theorem of Gelfand and Neumark, Proc. Natl. Acad. Sci. U.S.A. 32 (1946) 237-239. [101] G. ARSENE and L. ZSIDO, Une propriete de type de Darboux dans les algebres de Von Neumann, Act. Sci. Math. Szeged 30 (1969) 195-198. [102] W. B. ARVESON, A theorem on the action of abelian unitary groups, Pacific J. Math. 16 (1966) 205-212. [103] W. B. ARVESON, Analyticity in operator algebras, Amer. J. Math. 89 (1967) 578-642. [104] W. B. ARVESON, An algebraic conjugacy invariant for measure-preserving trans formations, Bull. Amer. Math. Soc. 73 (1967) 121-125. [105] W. B. ARVESON, Operator algebras and measure-preserving automorphisms, Acta Math. 118 (1967) 95-109. [106] W. B. ARVESON, A density theorem for operator algebras, Duke Math. J. 34 (1967) 635-647. [107] W. B. ARVESON, On subalgebras of C*-algebras, Bull. Amer. Math. Soc. 75 (1969) 790-794. [108] W. B. ARVESON, Subalgebras of C*-algebras, Acta Math. 123 (1969) 141-224. [109] W. B. ARVESON, Unitary invariants for compact operators, Bull. Amer. Math. Soc. 76 (1970) 88-91 [110] W. B. ARVESON, Operators with compact imaginary parts, Indiana Univ. Math. J. 20 (1971) 877-878 [111] W. B. ARVESON, Subalgebras of C*-algebras 11, Acta Math. 128 (1972) 271-308 [112] W. B. ARVESON, On groups of automorphisms of operator algebras, J. Functional Anal. 15 (1974) 217-243. [113] W. B. ARVESON, Operator algebras and invariant subspaces, Ann. Math. 100 (1974) 433-532. (114] W. B. ARVESON, A note on essentially normal operators, Proc. Roy. Irish. Acad. (1974) 143-146. [115] W. B. ARVESON, An Invitation to C*-Algebras, Graduatetexts in mathematics, no. 39 (Springer, Berlin, 1976). [116] W. B. ARVESON and K. B. JOSEPHSON, Operator algebras and measure-preserving automorphisms II, J. Functional Anal. 4 (1969) 100-134. [117] L. A. ASIMOV and A. J. ELLIS, On hermitian functionals on unital Banach algebras, Bull. London Math. Soc. 4 (1972) 333-336. [118] E. A. AzOFF, Borel measurability in linear algebra, Proc. Amer. Math. Soc. 42 (1974) 346-350.

REFERENCES

413

[119] A. BADRIKIAN, Structure de certains groupes localement compacts, Cahiers Rhodaniens 5 (1953) 27-51. [120] E. BALSLEV and A. VERBEURE, States on Clifford algebras, Commun. Math. Phys. 7 (1%8) 55-76. [121] E. BALSLEV, J. MANUCEAU and A. VERBEURE, Representations of anticom mutation relations and Bogoliubov transformations, Commun. Math. Phys. 8 (1968) 315-326. [122] B. A. BARNES, On the existence of minimal ideals in a Banach algebra, Trans. Amer. Math. Soc. 133 (1%8) 511-517. [123] B. A. BARNES, Banach algebras which are ideals in a Banach algebra, Pacific J. Math. 38 (1971) 1-7. [124] B. A. BARNES, Irreducible algebras of operators which contain a minimal idem potent, Proc. Amer. Math. Soc. 30 (1971) 337-342. [125] B. A. BARNES, Closed one-sided ideals in certain B*-algebras, Illinois J. Math. 15 (1971) 515-524. [126] B. A. BARNES, Locally R*-equivalent algebras, Trans. Amer. Math. Soc. 167 (1972) 435-442. [127] B. A. BARNES, Pure states with the restriction property, Proc. Amer. Math. Soc. 33 (1972) 492-494. [128] B. A. BARNES, Locally B*-equivalent algebras, II, Trans. Amer. Math. Soc. 176 (1973) 297-303. [129] B. A. BARNES, Representations of B*-algebras on Banach spaces, Pacific J. Math. SO (1974) 7-18. [130] B. A. BARNES, Linear functionals continuous on abelian *-subalgebras of a B*-algebra, J. London Math. Soc. 10 (1975) 320-328. [131] H. BEHNCKE, Structure of certain non-normal operators, J. Math. Mech. 18 (1968) 103-107. [132] H. BEHNCKE, Automorphisms of crossed products, Tohoku Math. J. 21 (1%9) 580-600. [133] H. BEHNCKE, A remark on C*-algebras, Commun. Math. Phys. 12 (1%9) 142-144. [134] H. BEHNCKE, Projections in Hilbert space, Tohoku Math. J. 22 (1970) 181-183. [135] H. BEHNCKE, Generators of W*-algebras, Tohoku Math. J. 22 (1970) 541-546. [136] H. BEHNCKE, A central decomposition of automorphisms, Math. Ann. 189 (1970) 308-310. [137] H. BEHNCKE, A note on the Gelfand-Naimark conjecture, Commun. Pure Appl. Math. 23 (1970) 189-200. [138] H. BEHNCKE, Topics in C*- and Von Neumann algebras, in: Lecture Notes in Math. 247 (1972). [139] H. BEHNCKE, Generators of finite W*-algebras, Tohoku Math. J. 24 (1972) 401408. [140) H. BEHNCKE, Generators of W*-algebras, II, Tohoku Math. J. 24 (1972) 371-381. [141] H. BEHNCKE, Generators of W*-algebras, III, Tiihoku Math. J. 24 (1972). [142] H. BEHNCKE, Structure of certain non-normal operators, II, Indiana Math. J. 22 (1972) 301-308. [143] H. BEHNCKE, Nilpotent elements in Banach algebras, Proc. Amer. Math. Soc. 37 (1973) 137-141. I [144] H. BEHNCKE, The description of transformations between superselection sectors

414

REFERENCES

of algebraic field thecries by means of Banach *-algebraic bundles over discrete groups (crossed products), to appear. [145] H. BEHNCKE, A class of non-normal operators, to appear. [146] H. BEHNCKE and W. Bos, A class of C*-algebras, Proc. Amer. Math. Soc. 40 (1973) 128-134. [147] H. BEHNCKE, F. KRAUSS and H. LEPTlN, C*-algebren mit geordneter Idealfolgen, J. Functional Anal. 10 (1972) 204-211. [148] H. BEHNCKE and H. LEPTlN, C*-algebras with a two-point dual, J. Functional Anal. 10 (1972) 330-335. [149] H. BEHNCKE and H. LEPTlN, C*-algebras with finite duals, J. Functional Anal. 14 (1973) 253-268. [150] H. BEHNCKE and H. LEPTIN, Classification of C*-algebras with a finite dual, J. Functional Anal. 16 (1974) 241-257. [151] D. C. BENSON and K. KREITH, On abstract Pruefer transformations, Proc. Amer. Math. Soc. 26 (1970) 137-140. [152] S. K. BERBERIAN, On the projection geometry of a finite AW*-algebra, Trans. Amer. Math. Soc. 83 (1956) 493-509. [153] S. K. BERBERIAN, The regular ring of a finite AW*-algebra, Ann. Math. 6S (1957) 224-240. [154] S. K. BERBERIAN, N X N matrices over an A W*-algebra, Amer. J. Math. 80 (1958) 27-44. [155] S. K. BERBERIAN, Note on a theorem of Fuglede and Putnam, Proc. Amer. Math. Soc. 10 (1959) 175-182. [156] S. K. BERBERIAN, Trace and the convex hull of the spectrum in a Von Neumann algebra of finite class, Proc. Amer. Math. Soc. 23 (1969) 211-212. [157] S. K. BERBERIAN, A note on the algebra of measurable operators of an AW* algebra, T6hoku Math. J. 22 (1970) 613-618. [158] S. K. BERBERIAN, Equivalence of projections, Proc. Amer. Math. Soc. 33 (1972) 485-490. [159] S. K. BERBERIAN, Baer *-rings, Grundl. Math. Wiss. Vol. 19S (Springer, Berlin, 1972). [160] S. K. BERBERIAN and G. H. ORLAND, On the closure of the numerical range of an operator, Proc. Amer. Math. Soc. 18 (1967) 499-503. [161] F. A. BEREZIN, Some remarks on representations of communication relations, Uspehi Mat. Nauk 24 (1969) 65-88 (in Russian; English Transl.: Russian Math.
Sum).

[162] C. A. BERGER and L. A. COBURN, C*-algebras of transformations and multipliers, Bull. Amer. Math. Soc. 74 (1968) 1008-1012; Erratum Ibid. 7S (1969) 468. [163] C. A. BERGER and L. A. COBURN, One-parameter semi-groups of isometries, Bull. Amer. Math. Soc. 76 (1970) 1115-1129. [164].c. A. BERGER, L. A. COBURN and A. LEBOW, C*-algebras generated by commuting isornetries, I, Bull. Amer. Math. Soc. 81 (1975) 747-749. [165] M. C. F. BERGLUND, Ideal C*-algebras, Duke Math. J. 40 (1973) 241-257. [166] E. BERKSON, Some characterizations of C*-algebras, Illinois J. Math. 10 (1966) 1-8. [167] E. BERKSON, Action of W*-algebr\ls in Banach spaces, Math. Ann. 189 (1970) 211-271.

REFERENCES

415

[168] E. BERKSON, Prehermitian elements and B*-algebras, Math. Ann. 195 (1972) 192-198. [169] K. BICHTELER, A generalization to the non-separable case of Takesakis-duality theorem for C*-algebras, Invent. Math. 9 (1969) 89-98. [170] M. BILLIK, Idempotent Reynolds operators, J. Math. Anal. Appl. 18 (1967) 135 144. [171] I. D. BIRRELL, Maximal simple C*-algebras, I, Bull. London Math. Soc. 6 (1974) 141-143. [172] I. D. BIRRELL, Maximal simple C*-algebras, II, to appear. [173] R. J. BLATTNER, Automorphic group representations, Pacific J. Math. 8 (1958) 665-677. [174] H. F. BOHNENBLUST and S. KARLIN, Geometrical properties of the unit sphere of Banach algebras, Ann. Math. 62 (1955) 217-229. [175] F. F. BONSALL, Extreme maximal ideals of a partially ordered vector space, Proc. Amer. Math. Soc. 7 (1956) 831-837. [176] F. F. BONSALL, Operators that act compactly on an algebra of operators, Bull. London Math. Soc. 1 (1969) 163-170. [177] F. F. BONSALL and J. DUNCAN, Numerical Ranges of Operators on Normed Spaces and of Elements of Normed Algebras, London Math. Soc. Lecture Notes Ser. Vol. 2 (Cambridge Univ. Press, London, 1971). [178] F. F. BONSALL and M. J. CRABB, The spectral radius of a hermitian element of a Banach algebra, Bull. London Math. Soc. 2 (1970) 178-180. [179] F. F. BONSALL, J. LINDENSTRAUSS and R. R. PHELPS, Extreme positive operators on algebras of functions, Math. Scand. 18 (1966) 161-182. [180] H. J. BORCHERS, On the structure of the algebra of field operators, I, Nuovo Cimento 24 (1962) 214-236. [181] H. J. BORCHERS, On the structure of the algebra of field operators, II, Commun. Math. Phys. 1 (1965) 57-79. [182] H. J. BORCHERS, On the vacuum state in quantum field theory, II, Commun. Math. Phys. 1 (1965) 57-79. [183] H. J. BORCHERS, Local rings and the connection of spin with statistics, Commun. Math. Phys. 1 (1965) 281-307. [184] H. J. BORCHERS, Energy and momentum as observables in quantum field theory, Commun. Math. Phys. 2 (1966) 49-54. [185] H. J. BORCHERS, A remark on a theorem of B. Misra, Commun. Math. Phys. 4 (1967) 315-323. [186] H. J. BORCHERS, On the converse of the Reeh-Schlieder theorem, Commun. Math. . Phys. 10 (1968) 269-273. [187] H. J. BORCHERS, On the implementability of automorphism groups, Commun. Math. Phys. 149 (1969) 305-314. [188] H. J. BORCHERS, Strongly continuous automorphism groups on C*-algebras, Car gese Lectures in Phys. 4 (1970) 31-59. [189] H. J. BORCHERS, On groups of automorphisms with semi-bounded spectrum, in: Colloq. sur les systeme a un Nombre Infini de Degres de Liberte (CNRS, Paris, 1970) 125-142. _ , [190] H. J. BORCHERS, Uber C*-A1gebren mit lokalkompakten Symmetriegruppen, Nachr. Akad. Wiss. Gottingen (1973) 1-18.

416

REFERENCES

[191] H. J. BORCHERS, Uber Ableitungen von C*-Algebren, Nachr. Akad. Wiss. Got tingen (1973) 19-35. [192] H. J. BORCHERS, Characterizationof inner *-automorphisms of W*-algebras, Pubi. Res. Inst. Math. Sci. Kyoto Univ. 10 (1974) II-50. [193] H. J. BORCHERS, On fields of C*-algebras, to appear. [194] A. BOREL, Representations de groupes localement compacts, Lecture Notes in Math. 276 (Springer, Berlin, 1972). [195] O. BRATIELI, Conservation of estimates in quantum field theory, Commun. Pure Appl. Math. 25 (1972) 759-779. [196] O. BRATIELI, Inductive limits of finite-dimensional C*-algebras, Trans. Amer. Math. Soc. 17 (1972) 195-234. [197] O. BRATIELI, Structure spaces of approximately finite-dimensional C*-algebras, J. Functional Anal. 16 (1974) 192-204. [198] O. BRATIELI, Local norm convergence of states on the zero time Bose fields, to appear. [199] O. BRATIELI, The center of approximately finite-dimensional C*-algebras, J. Functional Anal. 21 (1976) 195-202. [200] O. BRATIELI, Unbounded derivations of C*-algebras, to appear. [201] O. BRATIELI, Self-adjointness of unbounded derivations on C*-algebras, to ap pear. [202] O. BRATIELI and D. KASTLER, Relaxing the clustering condition in the derivation of the K.M.S. property, Commun. Math. Phys. 46 (1976) 37-42. [203] O. BRATIELI and D. ROBINSON, Unbounded derivations of C"-algebras, Commun. Math. Phys. 42 (1975) 253-268. [204] O. BRATIELI and D. W, ROBINSON, Unbounded derivations of C*-algebras, II, Commun. Math. Phys. 46 (1976) 11-30. [205] O. BRATIELI and D. W. ROBINSON, Unbounded derivations and invariant trace states, Commun. Math. Phys. 46 (1976) 31-36. [206] O. BRATIELI and D. W. ROBINSON, Unbounded derivations of Von Neumann algebras, to appear. [207] M. BREUER, A generalization of Kuipers theorem to factors of type IL, J. Math. Mech. 16 (1967) 917-926. [208] M. BREUER, Fredholm theories in Von Neumann algebras, I, Math. Ann. 178 (1968) 243-254. [209] M. BREUER, Fredholm theories in Von Neumann algebras, II, Math. Ann. 180 (1969) 313-325. [210] M. BREUER, On the homotopy type of the group of regular elements of semi-finite Von Neumann algebras, Math. Ann. 185 (1970) 61-74. [211] M. BREUER, Theory of Fredholm operators and vector bundles relative to a Von Neumann algebra, Rocky Mountain J. Math. 3 (1973) 383-429. [212] M. BREUER and R. S. BUTCHER, A generalized Riesz-Schauder decomposition theorem, Math. Ann. 203 (1973) 221-230. [213] O. L. BRITION, Primitive ideals of twisted group algebras, Trans. Amer. Math. Soc. 202 (1975) 221-241. [214] M. BROISE, Sur certaines aijplications unitaires de Iespace des operateurs de Hilbert-Schmidt, c.R. Acad. Sci. Paris 263 (1966) 722-725. [215] M. BROIsE, Sur les isomorphismes de certaines algebres de Von Neumann, Ann.

REFERENCES

417

Sci. Ecole Norm. Sup. 83 (1966) 91-111. [216) M. BROISE, Sur les vecteurs separateurs des algebres de Von Neumann com mutatives, c.s. Acad. Sci. Paris 264 (1967) 937-940. [217) M. BROISE, Sur les vecteurs separateurs des algebres de Von Neumann, J. Functional Anal. 1 (1967) 281-289. [218) M. BROISE, Commutateurs dans le groupe unitaire dun facteur, J. Math. Pures Appl. 46 (1967) 299-312. [219) A. BROWN, The unitary equivalence of binormal operators, Amer. J. Math. 76 (1954) 414-434. [220) A. BROWN and C. PEARCY, Structure of commutators of operators, Ann. Math. 86 (1965) 112-127. [221) A. BROWN and C. PEARCY, Multiplicative commutators of operators, Can. J. Math. 18 (1966) 737-749. [222) A. BROWN and C. PEARCY, Commutators in factors of type III, Can. J. Math. 18 (1966) 1152-1160. [223] A. BROWN, C. PEARCY and N. SAUNAS, Ideals of compact operators on Hilbert space, Michigan Math. J. 18 (1971) 373-384. [224} A. BROWN, C. PEARCY and D. TOPPING, Commutators and the strong radical, Duke Math. J. 3S (1968) 853-860. [225) L. G. BROWN, R. G. DoUGLAS and P. A. FILLMORE, Unitary equivalence modulo the compact operators and extensions of C*algebras, in: Lecture Notes in Math. 34S (Springer, Berlin, 1973) 58-128. [226} L. G. BROWN, R. G. DoUGLAS and P. A. FILLMORE, Extensions of C*-algebras, operators with compact self-commutators, and K-homo1ogy, Bull. Amer. Math. Soc. 79 (1973) 973-978. [227} J. BROMING, Indextheorie fur eine C*-Algebra von Toeplitzoperatoren, Math. Ann. 216 (1975) 113-121. [228} D. BUCHHOLZ, Product states for local algebras, Commun. Math. Phys. 36 (1974) 287-304. [229} J. W. BUNCE, Characters on singly generated C*-a1gebras, Proc. Amer. Math. Soc. 2S (1970) 297-303. [230} J. W. BUNCE, A note on two-sided ideals in C*-algebras, Proc. Amer. Math. Soc. 28 (1971) 635. [231) J. W. BUNCE, The joint spectrum of commuting non normal operators, Proc. Amer. Math. Soc. (1971) 499-505. [232) J. W. BUNCE, Representations of strongly amenable C*-a1gebras, Proc. Amer. Math. Soc. (1972) 241-246. [233) J. W. BUNCE, Irreducible representations of the C*-algebra generated by a quasi normal operator, Trans: Amer. Math. Soc. 183 (1973) 487-494. [234] J. W. BUNCE, Characterizations of amenable and strongly amenable C*algebras, Pacific J. Math. 43 (1973) 563-572. [235} J. W. BUNCE and J. A. DEDDENS, Irreducible representations of the C*algebra generated by an n-normal operator, Trans. Amer. Math. Soc. 171 (1972) 301-307. [236} J. W. BUNCE and J. A. DEDDENS, C*-algebras generated by weighted shifts, Indiana Univ. Math. J. 23 (1973) 257-272. [237} J. W. BUNCE and J. A. DEDDENS, Subspace approximants and GCR operators, Indiana Univ. Math. J. 24 (1974) 341-349.

418

REFERENCES

[238] J. W. BUNCE and J. A. DEDDENS, A family of simple C*-algebras related to weighted shift operators, J. Functional Anal. 19 (1975) 13-24. [239] J. W. BUNCE and J. A. DEDDENS, C*-algebras with Hausdorff spectrum, Trans. Amer. Math. Soc. 212 (1975) 199-217. [240] J. W. BUNCE and N. SALINAS, Completely positive maps on C*-algebras and the left matricial spectra of an operator, Duke Math. J. 43 (1976) 747-774. [241] D. BURES, Certain factors constructed as infinite tensor products, Compositio. Math. IS (1963) 169-191. [242) D. BURES, Tensor products of W*-algebras, Pacific J. Math. 27 (1968) 13-37. [243) D. BURES, An extension of Kakutanis theorem on infinite product measures to the tensor product of semifinite W*-algebras, Trans. Amer. Math. Soc. 135 (1969) 199-212.

[244] D. BURES, Representations of infinite weak product groups, Compositio. Math. 22 (1970) 7-18. [245] D. BURES, Abelian subalgebras of Von Neumann algebras, Mem. Amer. Math. Soc. 110 (1971). [246] R. C. BUSBY, On structure spaces and extensions of C*-algebras, J. Functional Anal. I (1967) 370-377. [247) R. C. BUSBY, Double centralizers and extensions of C*-algebras, Trans. Amer. Math. Soc. 132 (1968) 79-99. [248] R. C. BUSBY, On a theorem of Fell, Proc. Amer. Math. Soc. 30 (1971) 133-140. [249] R. C. BUSBY, On the equivalence of twisted group algebras and Banach *-algebraic bundles, Proc. Amer. Math. Soc. 37 (1973) 142-148. [250) R. C. BUSBY and H. A. SMITH, Representations of twisted group algebras, Trans. Amer. Math. Soc. 149 (1970) 503-537. [251] R. B. BUTTS and P. PROCELLI, Multiplicity theory, Studia Math. 38 (1970) 391-405. [252] J. W. CALKIN, Two-sided ideals and congruences in the ring of bounded operators in Hilbert space, Ann. Math. 42 (1941) 839-873. [253) R. W. CAREY and J. D. PINCUS, Mosaics, principal functions, and mean motion in Von Neumann algebras, to appear. [254) J. M. CHAIKEN, Finite-particle representations and states of the canonical com mutation relations, Ann. Phys. 42 (1967) 23-80. [255) J. M. CHAIKEN, Number operators for representations of the canonical com mutation relations, Commun. Math. Phys. 8 (1968) 164-184. [256) E. CHEN, Continuity of automorphisms of quasi-local algebras, J. Math. Phys. 12 (1971) 591-593. [257) E. CHEN, Operator algebras and axioms of measurements, J. Math. Phys. 12 (1971) 2364-2371. [258) E. CHEN, Structure of simplexes of equilibrium states in quantum statistical mechanics, J. Math. Phys. 13 (1972) 1130-1135. [259] E. CHEN, Choquet simplexes of states in physical systems, J. Math. Phys. 15 (1974) 2022-2025. [260) W. M. CHING, ~on-isomorphic non-hyperfinite factors, Can. J. Math. 21 (1969) 1293-1308. [261) W. M. CHING, A continuum of non-isomorphic non-hyperfinite factors, Commun. Pure Appl. Math. 23 (1970) 927-938. [262] W. M. CHING, Free products of Von Neumann algebras, Trans. Amer. Math. Soc.

REFERENCES

419

178 (1973) 147-163. [263] W. M. CHING, Topologies on the quasi-spectrum of a C*-algebra, Proc. Amer. Math. Soc. 46 (1974) 273-276. [264] W. M. CHING and P. WILLIG, Non-asymptotically abelian factors of type III, Proc. Amer. Math. Soc. 34 (1972) 102-104. [265] H. CHODA, On automorphisms of abelian Von Neumann algebras, Proc. Japan Acad. 41 (1%5) 280-283. [266] H. CHODA, On Maharam subfactors of finite factors, t, II, Proc. Japan Acad. 43 (1%7) 451-455, 937-940. [267] H. CHODA, On the crossed product of abelian Von Neumann algebras, I, II, Proc. Japan Acad. 43 (1%7) 111-116, 198-201. [268] H. CHODA, On Maharam subfactors of finite factors, II, Proc. Japan Acad. 43 (1%7) 937-940. [269] H. CHODA, On extremal property of the polar decomposition in Von Neumann algebras, Proc. Japan Acad. 46 (1970) 341-344. [270] H. CHODA, On a simple representation of groups as outer automorphism groups of the hyperfinite II,-factor, Mem. Osaka Kyoiku Uni. 24 (1971) 113-115. [271] H. CHODA, On ergodic and abelian automorphism group of Von Neumann alge bras, Proc. Japan Acad. 47 (1971) 982-985. [272] H. CHODA, On a decomposition of automorphisms of Von Neumann algebras, Proc. Japan Acad. 49 (1973) 809-811. [273] H. CHODA, A comment on the Galois theory for finite factors, Proc. Japan Acad. SO (1974) 619-622. [274] H. CHODA, On freely acting automorphisms of operator algebras, Kodai Math. Sem. Rep. 26 (1974) 1-21. [275] H. CHODA and M. CHODA, Some remarks on Von Neumann algebras with an algebraic property, Proc. Japan Acad. 41 (1%5) 71-72. [276] H. CHODA and M. ECHIGO, On theorems of Korovkin, Proc. Japan Acad. 39 (1%3) 107:-108. [277] H. CHODA and M. ECHIGO, A new algebraical property of certain Von Neumann algebras, Proc. Japan Acad. 39 (1%3) 651-655. [278] H. CHODA and M. ECHIGO, Some remarks on Von Neumann algebras with the property Q, Mem. Osaka Gakugei Univ. 13 (1964) 13-21. [279] H. CHODA and M. ECHIGO, A remark on a construction of finite factors, I, II, Proc. Japan Acad. 40 (1%4) 474-478, 479-481. [280] H. CHODA, Y. KUIMA and Y. NAKAGAMI, Some extremal properties in the unit ball of Von Neumann algebras, Kodai Math. Sem. Rep. 21 (1969) 175-181. [281] H. CHODA and M. NAKAMURA, Elementary proofs of Gleason-Kahane-Zelaskos theorem for B*-algebras, Mem. Osaka Kyoiku Univ. 20 (1971) 111-112. [282] M. CHODA, On the conditional expectation of a partial isometry in a certain Von Neumann algebra, ~oc. Japan Acad. 41 (1%5) 277-279. [283) M. CHODA, A remark on the normal expectations, Proc. Japan Acad. 44 (1%8) 462-466. [284] M. CHODA, A normal expectation and the tensor product of Von Neumann algebras-on Takesakis theorem. Meln. Osaka Kyoiku Unit. 20 (1971) 117 120. [285] M. CHODA, Abelian projection over a Von Neumann subalgebra, Proc. Jap. Acad. 48 (1972) 384-388.

420

REFERENCES

[286] M. CHODA, A Von Neumann algebra continuous over a Von Neumann algebra, Proc. Japan Acad. 49 (1973) 174-178. [287) M. CHODA, A remark on the normal expectations, II, Proc. Japan Acad. 49 (1973) 252-256. [288] M. CHODA, On types over Von Neumann subalgebras and Dye correspondence, Publ. Res. lnst, Math. Sci. Kyoto Univ. 9 (1973) 45-60. [289) M. CHODA, Shift automorphism groups of Von Neumann algebras, Proc. Japan. Acad. 50 (1974) 470-475. [290) M. CHODA, Normal expectations and crossed products of Von Neumann algebras, Proc. Japan Acad. 50 (1974) 738-742. [291) M. CHODA, The fixed algebra of a Von Neumann algebra under an automorphism group, Tohoku Math. J. 28 (1976) 227-234. [292) M. CHODA and H. CHODA, On extensions of automorphisms of abelian Von Neumann algebras, Proc. Japan Acad. 43 (1967) 295-299. [293) M. CHODA and H. CHODA, On the minimality of the polar decomposition in finite factors, Proc. Japan Acad. 44 (1968) 798-800. [294) M. CHODA and H. CHODA, Some characterizations of certain Von Neumann algebras, Proc. Japan Acad. 46 (1970) 1086-1090. [295) M. CHODA, I, KASAHARA and R. NAKAMOTO, Dependent elements of an au tomorphism of a C*-algebra, Proc. Japan Acad. 48 (1972) 561-565. [296) M. CHODA and M. NAKAMURA, A remark on the concept of channels, II, Proc. Japan Acad. 46 (1970) 932-935. [297) M. CHODA and M. NAKAMURA, A remark on the concept of channels, III: An algebraic theory of extended Toeplitz operators, Proc. Japan Acad. 47 (1971) 464-469. [298] C. CHo-Ho, Prime faces in C*-algebras, J. London Math. Soc. 7 (1973) 175-180. [299] M. D. CHOI, Positive linear maps on C*-algebras, Can. J. Math. 24 (1972) 510-529. [300] M. D. CHOI, A Schwarz inequality for positive linear maps on C*-algebras, Illinois J. Math. 18 (1974) 565-574. [301) M. D. CHOI and E. G. EFFROS, Injectivity and operatorspaces,!. Functional Anal. 24 (1977) 156-209. [302] M. D. CHOI and E. G. EFFROS, Nuclear C*-algebras and the approximation property, to appear. [303] M. D. CHOI and E. G. EFFROS, Separable nuclear C*-algebras and injectivity, Duke Math. J. 43 (1976) 309-322. [304] E. CHRISTENSEN, Noncommutative integration for monotone sequentially closed C*-algebras, Math. Scand. 31 (1972) 171-190. [305] E. CHRISTENSEN, Perturbations of type I Von Neumann algebras, J. London Math. Soc. 9 (1975) 395-405. [306] E. CHRISTENSEN, Perturbations of operator algebras, I, II, to appear. [307] C. CHo-H<", On convexity theory and C*-algebras, Proc. London Math. Soc. 31 (1975) 257-288. [308] R. CIRELLI, F. GALLONE and B. GUBBA Y, An algebraic representation of con tinuous superselection rules, J. Math. Phys. 16 (1975) 201-213. [309] S. B. CLEVELAND, Homomorphisms of non-commutative *-algebras, Pacific J. Math. 13 (1963) 1097-1109. [310] L. A. COBURN, The C*-algebra generated by an isometry, Bull. Amer. Math. Soc. 73 (1967) 722-726.

REFERENCES

421

[3Il] L. A. COBURN, The C*-algebra generated by an isometry, II, Trans. Amer. Math. Soc. 137 (1969) 211-217. [312] L. A. COBURN, C*-a1gebras generated by semi-groups of isometries, in: Hilbert Space Operators, Colloq. Math. Soc. Janes Bolyai Vol. 5 (North-Holland, Am sterdam, 1970) 99-105. [313] L. A. COBURN and R. G. DoUGLAS, Translation operators on the half line, Proc. Natl. Acad. Sci. U.S.A. 62 (1969) 1010-1013. [314] L.. A. COBURN and R. G. DoUGLAS, C*-a1gebrasof operators on a half space, I, Pub/. Math. I.H.E.S. 40 (1971) 59-67. [315] L. A. COBURN, R. G. DoUGLAS, D. G. SCHAEFFER and I. M. SINGER, C* algebras of operators on a half space, II: Index theory, Publ. Math. I.H.E.S. 40 (1971) 69-79. [316] L. A. COBURN, R. G. DoUGLAS and I. M. SINGER, An index theorem for Wiener-Hopf operators on the discrete quarter-plane, J. Differential Geometry 6 (1972) 587-593. [317] L. A. COBURN and A. LEBOW, Components of invertible elements in quotient algebras of operators, Trans. Amer. Math. Soc. 130 (1968) 359-366. [318] L. A. COBURN, R. D. MOYER and I. M. SINGER, C*-a1gebrasof almost periodic pseudo-differential operators, Acta Math. 130 (1973) 279-307. [319] L. A. COBURN and M. SCHECHTER, Joint spectra and interpolation of operators, J. Functional Anal. 2 (1968) 226-237. [320] H. S. COLLINS and R. A. FONTENOL, Approximate identities and the strict topology, Pacific J. Math. 43 (1972) 63-79. [321] F. COMBES, Relations entre formes positives sur une C*-a1gebre, c.R. Acad. Sci. Paris 260 (1965) 5435-5438. [322] F. COMBES, Representations dune C*-algebre et formes lineaires positives, c.R. Acad. Sci. Paris 260 (1965) 5993-5996. [323] F. COMBES, Etude des representations tracees dune C*algebre, C. R. Acad. Sci. Paris 262 (1966) 114-117. [324] F. COMBES, Sur les etats factoriels dune C*-a1gebre, c.R. Acad. Sci. Paris US (1967) 736-739. [325] F. COMBES, Etudes des poids definis sur une C*a1gebre, c.R. Acad. Sci. Paris US (1967) 340-343. [326] F. COMBES, Poids sur une C*-a1gebre, J. Math. Pures Appl. 47 (1968) 57-100. [327] F. COMBES, Elements semi-continus associes a une C*-algebre, CiR. Acad. Sci. Paris 267 (1968) 986-989. [328] F. COMBES, Sur les faces dune C*-algebre, Bull. Sci. Math. 93 (1969) 37~2. [329] F. COMBES, Poids assode a une algebre hibertienne generalises, c.R. Acad. Sci. Paris 270 (IWO) 33-36. [330] F. COMBES, Quelques proprietes des C*-a1gebres, Bull. Sci. Math. 94 (1970) 165-192. [331] F. COMBES, Poids et esperances conditionnelles dans les algebres de Von Neumann, Bull. Soc. Math. France. 99 (1971) 73-112. [332] F. COMBES, Poids assode a une algebre hilbertienne a gauche, Compositio. Math. 23 (1971) 49-77. [333] F. COMBES and C. DELAROCHE, Groups modulaire dune esperance condition nelle dans une algebre de von Neumann, Bull. Soc. Math. France 103 (1975) 385-426. [334] F. COMBES and F. PERDRIZET, Certains ideaux dans les espaces vectoriels

422

REFERENCES

ordonnes, CiR, Acad. Sci. Paris 268 (1969) 1552-1555. [335] F. COMBES and F. PERDRIZET, Certains ideaux dans les espaces vectoriels ordonnes, J. Math. Pures Appl. 49 (1970) 29-59. [336] A. CONNES, Un nouvel invariant pour les algebres de Von Neumann, C.R. Acad, Sci. Paris 273 (1971) 900-903. [337] A. CONNES, Calcul des deux invariants dAraki et Woods par la theorie de Tomita et Takesaki, c.n. Acad. Sci. Paris 274 (1972) 175-177. [338] A. CONNES, Etats presque-periodiques sur une algebre de Von Neumann, c.R. Acad. Sci. Paris 274 (1972) 1402-1405. [339] A. CONNES, Groupe modulaire dune algebre de Von Neumann, CiR. Acad. Sci. Paris 274 (1972) 1923-1926. [340] A. CONNES, Une classification des facteurs de type III, C.R. Acad. Sci. Paris 275 (1972) 523-525. [341] A. CONNES, Une classification des facteurs de type III, Ann. Ecole Norm. Sup. 6 (1973) 133-252. [342] A. CONNES, Sur Ie theoreme de Radon-Nikodym pour les poids normaux fideles semi-finis, Bull. Sci. Math. 97 (1973) 253-258. [343] A. CONNES, Almost periodic states and factors of type III" J. Functional Anal. 16 (1974) 415-445. [344] A. CONNES, Caracterisation des espaces vectoriels ordonnes sous-jacents aux algebres de Von Neumann, Ann. Inst. Fourier 24 (1974) 121-155. [345] A. CONNES, On hyperfinite factors of type IlL, and Kriegers factors, J. Functional Anal. 18 (1975) 318-327. [346] A. CONNES, A factor not anti-isomorphic to itself, Ann. Math. lOt (1975) 536-554. [347] A. CONNES, A factor not anti-isomorphic to itself, Bull. London Math. Soc. 7 (1975) 171-174. [348] A. CONNES, Sur la classification des facteurs de type II, C.R. Acad. Sci. Paris 281 (1975) 13-15. [349] A. CONNES, On the hierarchy of W. Krieger, fllinois J. Math. 19 (1975) 428-432. [350] A. CONNES, Periodic automorphisms of the hyperfinite factor of type 11" to appear. [351] A. CONNES, Classification of automorphisms of hyperfinite factors of type II. and IL, and application to type III factors, Bull. Amer. Math. Soc. 81 (1975) 1090-1095. [352] A. CONNES. Outer conjugacy classes of automorphisms of factors, Ann. Sci. Ec. Norm. Sup. 8 (1975) 383-420. [353] A. CONNES, Classification of injective factors; cases 11" II~, III }d- 1, Ann. Math. 104 (1976) 73-115. [354] A. CONNES, P. GHEZ, R. LIMA, D. TESTARD and E. J. WOODS, Review of "Crosjed product of Von Neumann algebras", mimeographed. [355] A. CONNES and E. STORMER, Etnropy for automorphisms of II. Von Neumann algebras, Acta Math. 134 (1975) 289-306. [356] A. CONNES and M. T AKESAKI, Flot des poids sur les facteurs de type III, c.R. Acad. Sci. Paris 278 (1974) 945-948. [357] A. CONNES and M. T AKESAKI, The flow of weights on factors of type III, to appear. [358] A. CONNES and A. VAN DAELE, The group property of the invariant S of Von Neumann algebras, Math. Scand. 32 (1973) 187-192. [359] A. CONNES and E. J. WOODS, Existence de facteurs infinis asymptotiquement

REFERENCES

423

abeliens, C.R. Acad. Sci. Paris 279 (1974) 189-191. [360) J. B. CONWAY, The numerical range and a certain convex set in an infinite factor, J. Functional Anal. 5 (1970) 428-435. [361) J. B. CONWAY, On the Calkin algebra and the covering homotopy property, Trans. Amer. Math. Soc. 211 (1975) 135-142. [362) J. B. CONWAY, Multipliers of matroid C*-algebras, to appear. [363) J. B. CONWAY and J. Szucs, The weak sequential closure of certain sets of extreme points in a Von Neumann algebra, Indiana Univ. Math. J. 22 (1973) 763-768. [364) H. O. CORDES, On a class of C*-algebras, Math. Ann. 170 (1967) 283-313. [365) H. O. CORDES, On a generalized Fredholm theory, J. Reine Angew. Math. 227 (1967) 121-149. [366) H. O. CORDES, An algebra of singular integral operators with two symbol homomorphisms, Bull. Amer. Math. Soc. 75 (1969) 37-42. [367) A. CORDESSE and G. RIDEAU, On some representations of anti-commutation relations, I, II, III, Nuovo Cimento 4S (1966) 1-14, 46 (1966) 624-636, SO (1967) 244-255. [368) M. COURBAGE, S. MIRACLE-SOLE and D. W. ROBINSON, Normal states and representations of the canonical commutation relations, Ann. lnst. H. Poincare (A) 14 (1971) 171-178. [369] I. G. CRAW, Axiomatic cohomology of operator algebras, Bull. Soc. Math. France 100 (1972) 449-460. [370] I. Cuctn.ascu, A proof of (A@B)=A@B for Von Neumann algebras, Rev. Roumaine Math. Pures Appl. 16 (1971) 665-670. [371] I. Cuctn.sscu, Martingales on Von Neumann algebras, J. Multivariate Anal. I (1971) 17-27. [372) J. CUNTZ, Locally C*-equivalent algebras, J. Functional Anal. 23 (1976) 95-106. [373) J. CUNTZ, On the continuity of semi-norms on operator algebras, Math. Ann. 220 (1976) 171-183. [374) DANG Nooc NGHIEM, ~*-a1gebres, probabilites non commutatives et applications, Mem. Soc. Math. France 3S (1973) 145-189. [375] DANG NGOC NGHIEM, Sur la classification des systernes dynamiques non com mutatifs, J. Functional Anal. IS (1974) 188-201. [376] DANG Nooc NGHIEM, On the integral representation of states on a C*-algebra, Commun. Math. Phys. 40 (1975) 223-233. [377] DANG NGOC NGHlEM, Decomposition et classification des systemes dynamiques, Bull. Soc. Math. France 103 (1975) 149-175. [378) DANG NGOC NGHIEM, Plancherel measure for type I traces, to appear. [379] DANG NGOC NGHIEM, Integral of semi-finite traces and o-finite measures, to appear. [380) DANG Nooc NGHIEM, Produits croises restreints et extensions de groupes, to appear. [381] DANG Nooe NGHIEM and F. LEDRAPPIER, Sur les systemes dynamiques sim pliciaux, C.R. Acad. Sci. Paris 277 (1973) 777-779. [382] W. F. DARSOW, Positive definite functions and states, Ann. Math. 60 (1954) 447-453. [383] R. B. DARST, On a theorem of Nikodym with applications to weak convergence and Von Neumann algebras, Pacific J. Math. 23 (1967) 473-477.

424

REFERENCES

[384] R B. DARST, On a theorem of Nikodym with applications to weak convergence and Von Neumann algebras, Bull. Amer. Math. Soc. 74 (1968) 283-284. [385] J. DAUNS, Multipliers rings and primitive ideals, Trans. Amer. Math. Soc. 145 (1969) 125-158. [386] J. DAUNS, Categorical W*-tensor product, Trans. Amer. Math. Soc. 166 (1972) 439-456. [387] J. DAUNS, The primitive ideal space of a C*-algebra, Can. J. Math. 26 (1974) 42-49. [388] J. DAUNS and K. H. HOFMANN, Representations of rings by section, Mem. Amer. Math. Soc. 83 (1968). [389] J. DAUNS and K. H. HOFMANN, Spectral theory of algebras and adjunction of identity, Math. Ann. 179 (1969) 175-202. [390] E. B. DAVIES, On the Borel structure of C*-algebras (with an appendix by RV. Kadison), Commun. Math. Phys. 8 (1968) 147-163. [391] E. B. DAVIES, Decomposition of traces on separable C*-algebras, Quarterly J. Math. 20 (1969) 97-111. Quarterly J. Math. 20 (1969) 351-366. [392] E. B. DAVIES, The structure of ~*-algebras, [393] E. B. DAVIES, Involutory automorphisms of operator algebras, Trans. Amer. Math. Soc. 158 (1971) 115-142. [394] C. DAVIS, Generators of the ring of bounded operators, Proc. Amer. Math. Soc. 6 (1955) 970-972. [395] C. DAVIS, A. Schwartz inequality for convex operatorfunctions, Proc. Amer. Math. Soc. 8 (1957) 42-44. [396] C. DAVIS, Various averaging operations onto subalgebras, Illinois J. Math. 3 (1959) 538-553. [397] C. DAVIS, Operator-valued entropy of a quantum-mechanical measurement, Proc. Japan Acad. 37 (1961) 533-538.

[398] G. DE BARRA, An inner measure associated with a Von Neumann algebra, Proc. Cambridge Phi/os. Soc. 64 (1968) 645-650. [399] G. DE BARRA, A class of measurable sets associated with a Von Neumann algebra, J. London Math. Soc. 5 (1972) 123-126. [400] J. C. DEEL, Derivations of AW*-algebras, Proc. Amer. Math. Soc. 42 (1974) 85-95. [401] A. DE KORVIN, Normal expectations in Von Neumann algebras, Pacific J. Math.
27 (1968) 333-338. [402] A. DE KORVIN, Expectations in Von Neumann algebras, Bull. Amer. Math. Soc. 74 (1968) 912-914. [403] A. DE KORVIN, Stable maps and Schwartz maps, Trans. Amer. Math. Soc. 148 (1970) 283-291.

[404] A. DE KORVIN, Complete sets of expectation on Von Neumann algebra, Quart. J. Math. 22 (1971) 135-142. [405] A. DE KORVIN and R. J. EASTON, Expectations on B*-algebras, J. Reine Angew. Math. 251 (1971) 1-6. [406] P. DE LA HARPE and M. KAROUBI, Perturbations compactes des representations dun groupe dans un espace de Hilbert, C.R. Acad. Sci. Paris 281 (1975) 901-904. [407] C. DELAROCHE, Sur les centre des C*-algebres, C.R. Acad. Sci. Paris 255 (1967)
465-466. [408] C. DELAROCHE, Sur les centres des C*-algebres, Bull. Sci. Math. 91 (1967) 105-112.

REFERENCES

425

[409] C. DELAROCHE, Sur les centres des C*-algebres, II, Bull. Sci. Math. 92 (1968) 111-128. [410] C. DELAROCHE, Spectres des extensions de C*-algebres, C.R. Acad. Sci. Paris 269 (1969) 1003-1005. [411] C. DELAROCHE, Limites et valeurs dadherence dun filtre sur Ie spectre dune C*-a1gebre, c.R. Acad. Sci. Paris 271 (1970) 434-437. [412] C. DELAROCHE, Sur 1es extensions des C*-algebres, C.R. Acad. Sci. Paris 272 (1971) 727-730. [413] C. DELAROCHE, Extensions des C*-algebres, Bull. Soc. Math. France Suppl, 29 (1972). [414] G. F. DELLANTONIO, On some groups of automorphisms of physical observables, Commun. Math. Phys. 2 (1966) 384-397. [415] G. F. DELLANTONIO, On the limits of sequences of normal states, Commun. Pure Appl. Math. 20 (1967) 413-429. [416] G. F. DELLANTONIO, Structure of the algebras of some free systems, Commun. Math. Phvs. 9 (1968) 81-117. [417] G. F. DELLANTONIO, Role of automorphisms in abstract field theory, in: Proc. Intern. Sem. on Elementary Particle Theory, Varna (1968) 44-67. [418] G. F. DELLANTONIO, Can local gauge transformations be implemented? J. Math. Phys. 12 (1974) 148-156. [419] G. F. DELLANTONIO and S. DoPLlCHER, Total number of particles and Fock representation, J. Math. Phys. 8 (1967) 663-666. [420] G; F. DELLANTONIO, S. DoPLlCHER and D. RUELLE, A theorem on canonical commutation and anticommutation relations, Commun. Math. Phys. 2 (1966) 223 230. [421] A. DERIGHETfI, Some results on the Fourier-Stieltjes algebra of a locally compact group, Comment. Math. Helv. 45 (1970) 219-228. [422] T. DIGERNES, On classes of projections in a Von Neumann algebra, Math. Scand. 31 (1972) 191-200. [423] T. DIGERNES, A new characterisation of separable GCR-algebras, Proc. Amer. Math. Soc. 36 (1972) 448-450. [424] T. DIGERNES, Poids dual sur un produit croise, C.R. Acad. Sci. Paris 278 (1974) 937-940. [425] T. DIGERNES and H. HALPERN, On open projections of GCR.algebras, Can. J. Math. 24 (1972) 978-982. [426] J. DIXMIER, Mesure de Haar et trace dun operateur, c.R. Acad. Sci. Paris 228 (1949) 152-154. [427] J. DIXMIER, Les anneaux doperateursde c1asse linie, Ann. Sci. Ecole Norm. Sup. 66 (1949) 209-261. [428] J. DIXMIER, Les fonctionnelles lineaires sur lensemble des operateurs homes dun espace de Hilbert, Ann. Math. 51 (1950) 387-408. [429] J. DIXMIER, Application dans les anneaux doperateurs, CiR, Acad. Sci. Paris 230 (1950) 607-608. [430] J. DIXMIER, Sur certains espaces consideres par M. H. Stone, Summa Brasil. Math. 2 (1951) 151-182. [431] J. DIXMIER, Sur la reduction des anneaux doperateurs, Ann. Sci. Ecole Norm. Sup. 68 (1951) 185-202.

426

REFERENCES

[432] J. DIXMIER, Algebres quasi-unitaires, C.R. Acad. Sci. Paris 233 (1951) 837-839. [433] J. DIXMIER, Applications dans les anneaux doperateurs,Compositio Math. 10 (1952) I-55. [434] J. DIXMIER, Algebres quasi-unitaires, Comment. Math. Helv. 26 (1952) 275-322. [435] J. DIXMIER, Remarques sur les applications, Arch. Math. (Basel) 3 (1952) 290-297. [436] J. DIXMIER, Formes lineaires sur un anneau doperateurs, Bull. Soc. Math. France 81 (1953) 9-39. [437] J. DIXMIER, Sous-anneaux abeliens maximaux dans les facteurs de type fini, Ann. Math. 59 (1954) 279-296. [438] J. DIXMIER, Sur les anneaux doperateurs dans les espaces hilbertiens, C.R. Acad. Sci. Paris 238 (1954) 439-441. [439] J. DIXMIER, Sur les representations unitaires des groupes de Lie algebriques, Ann. Inst. Fourier 7 (1957) 315-328. [440] J. DIXMIER, Sur les representations unitaires des groupes de Lie nilpotents, IV, Can. J. Math. 11 (1959) 321-344. [441] J. DIXMIER, Sur les representations unitaires des groupes de Lie nilpotents, V, Bull. Soc. Math. France 87 (1959) 65-79. [442] J. DIXMIER, Sur les representations unitaires des groupes de Lie nilpotents. VI, Can. J. Math. 12 (1960) 324-352. [443] J. DIXMIER, Sur les C*-algebres, Bull. Soc. Math. France 88 (1960) 95-112. [444] J. DIXMIER, Sur les structures boreliennes du spectre dune C*-algebre, Publ. Inst. Hautes Etudes Sci. 6 (1960) 297-303. [445] J. DIXMIER, Representations integrables du groupe de De Sitter, Bull. Soc. Math. France 89 (1961) 9--41. [446] J. DIXMIER, Points isoles dans Ie dual dun groupe localement compact, Bull. Sci. Math. 85 (1961) 91-96. [447] J. DIXMIER, Points separes dans le spectre dune C*-algebre, Acta Sci. Math. Szeged 22 (1961) 115-128. [448] J. DIXMIER, Sur les representations unitaires des groupes de Lie resolubles, Math. J. Okayama Univ. 11 (1962) 1-18. [449] J. DIXMIER, Dual et quasi-dual dune algebre de Banach involutie, Trans. Amer. Math. Soc. 104 (1962) 278-283. [450] J. DIXMIER, Champs continus despaces hilbertiens et de C*-algebres, J. Math. Pures Appl. 42 (1963) 1-20. [451] J. DIXMIER, Quasi-dual dun ideal dans une C*-algebre, Bull. Sci. Math. 87 (1963) 7-11. [452] J. DIXMIER, Traces sur les C*-algebres, Ann. Inst. Fourier 13 (1963) 219-262. [453] J. DIXMIER, Traces sur les C*-algebres. II, Bull. Sci. Math. 88 (1964) 39-57. [454] J. DIXMIER, Utilisation des facteurs hyperfinis dans la theorie des C*-algebres, c.s. Acad. Sci. Paris 258 (1964) 4184--4187. [455] J. DIXMIER, Existence de traces non norrnales, C.R. Acad. Sci. Paris 262 (1966) 1107-1108. [456] J. DIXMIER, On some C*-algebras considered by Glimm, J. Functional Anal. 1 (1967) 182-203. [457] J. DIXMIER, Sur les automorphismes des algebres de Banach, C.R. Acad. Sci. Paris 264 (1967) 729-731. [458] J. DIXMIER, Ideal center of a C*-algebra, Duke Math. J. 35 (1968) 375-382.

REFERENCES

427

[459] J. DIXMIER, Localization of a theorem of Glimm, Proc. Amer. Math. Soc. 19 (1968) 364--366. [460] J. DIXMIER, Quelques proprietes des suites centrales dans les facteurs de type II.. Invent. Math. 7 (1969) 215-225. [461] J. DIXMIER, Sur les groupes dautomorphismes normiquement continus des C* algebres, C.R. Acad. Sci. Paris 269 (1969) 643--644. [462] J. DIXMIER, us Algebres dOperateurs dans IEspace Hilbertien, 2nd ed. (Gau thier-Villars, Paris, 1969). [463] J. DIXMIER and A. DoUADY, Champs continus despaces hilbertiens et de C* algebres, Bull. Soc. Math. France 91 (1963) 227-284. [464] J. DIXMIER and E. C. LANCE, Deux nouveaux facteurs de type Il., Invent. Math. 7 (1969) 226-234. [465] J. DIXMIER and O. MARECHAL, Vecteurs totalisateurs dune algebre de Von Neumann, Commun. Math. Phys. 22 (1971) 44-50. [466] P. G. DIXON, Generalized B*-a1gebras, Proc. London Math. Soc. 21 (1970) 693 715. [467] P. G. DIXON, Unbounded operator algebras, Proc. London Math. Soc. 23 (1971) 53-69. [468] DoN DECKARD and C. PEARCY, On matrices over the ring of continuous complex valued functions on a Stonian space, Proc. Amer. Math. Soc. 14 (1963) 322-328. [469] DoN DECKARD and C. PEARCY, On continuous matrix-valued functions on a Stonian space, Pacific J. Math. 14 (1964) 857-869. [470] A. H. DoOLEY, The spectral theory of posets and its applications to C*-algebras, Trans. Amer. Math. Soc. 224 (1976) 143-156. [471] S. DoPLICHER, An algebraic spectrum condition, Commun. Math. Phys. 1 (1965) 1-5. [472] S. DoPLICHER, Superselection rules and fields from local observables, in: Cargese Lectures in Phys. 4 (1970) 79-93. [473] S. DoPLICHER, A remark on a theorem of Powers and Sakai, Commun. Math. Phys. 45 (1975) 59-61. [474] S. DoPLICHER, R. HAAG and J. E. ROBERTS, Fields, observables and gauge transformations, I, II, Commun. Math. Phys. 13 (1969) 1-23, 15 (1969) 173-200. [475] S. DoPLICHER, R. HAAG and J. E. ROBERTS, Local observables and particle statistics, I, II, Commun. Math. Phys. 23 (1971) 199-230, 35 (1974) 48-85. [476] S. DoPLICHER, R. V. KADISON, D. KASTLER and D. W. ROBINSON, Asymp totically abelian systems, Commun. Math. Phys. 6 (1967) 101-120. [477] S. DoPLICHER and D. KASTLER, Ergodic states in a non-commutative ergodic theory, Commun. Math. Phys. 7 (1968) 1-20. [478] S. DoPLICHER, D. KASTLER and D. W. ROBINSON, Covariance algebras in field theory and statistical mechanics, Commun. Math. Phys. 3 (1966) 1-28. [479] S. DoPLICHER, D. KASTLER and E. ST0RMER, Invariant states and asymptotic abelianness, J. Functional Anal. 3 (1969) 419-434. [480] S. DoPLICHER and R. T. POWERS, On the simplicity of the even CAR algebra and free field models, Commun. Math. Phys. 7 (1968) 77-79. [481] S. DoPLICHER and J. E. ROBERTS, Fields, statistics and non-abelian gauge groups, Commun. Math. Phys. 28 (1972) 331-348.

428

REFERENCES

[482] D. DoPPING, Vector lattices of self-adjoint operators, Bull. Amer. Math. Soc. 69 (1963) 251-255. [483] R. G. DoUGLAS, On operators similar to normal operators, Rev. Roumaine Math. Pures Appl. 14 (1969) 193-197. [484] R. G. DOUGLAS. On the C*-algebra of a one-parameter semi-group of isometries, Acta Math. 128 (1972) 143-151. [485] R. G. DoUGLAS and R. HOWE, On the C*-algebra of Toeplitz operators on the quarter plane, Trans. Amer. Math. Soc. 158 (1971) 203-217. [486] R. G. DoUGLAS and C. PEARCY, Von Neumann algebras with a single generator, Michigan Math. J. 16 (1969) 21-26. [487] R. G. DoUGLAS and D. M. TOPPING, Operators whose squares are zero, Rev. Roumaine Math. Pures Appl. 12 (1967) 647-652. [488] H. R. DOWSON, On an un starred operator algebra, J. London Math. Soc. 5 (1972) 489-492. [489] W. DRIESSLER, Comments on lightlike translations and applications in relativistic quantum field theory, Commun. Math. Phys. 44 (1975) 133-141. [490] M. DUBOIS-VIOLETIE, A generalization of the classical moment problem on *-algebras with applications to relativistic quantum theory, I, Commun. Math. Phvs. 43 (1975) 225-254. [491] 1. DUNCAN. The continuity of the involution in Banach *-algebras, J. London Math. Soc. 41 (1966) 701-706. [492] N. DUNFORD, Resolution of the identity for commutative B*-algebras of operators, Acta. Sci. Math. Szeged 12 (1950) 51-56. [493] C. F. DUNKL and D. E. RAMIREZ, C*-a1gebras generated by measures, Bull. Amer. Math. Soc. 77 (1971) 411-412. [494] M. J. DUPRE, Classifying Hilbert bundles, J. Functional Anal. 15 (1974) 244-278. [495] M. J. DUPRE, Hilbert bundles with infinite-dimensional fibres, Mem. Amer. Math. Soc. 148 (1974) 165-176. [496] M. J. DUPRE, Classifying Hilbert bundles, II, to appear. [497] M. 1. DUPRE, Cohomology for classifying categories of Banach bundles, to appear. [498] H. A. DYE, The Radon-Nikodym theorem for finite rings of operators, Trans. Amer. Math. Soc. 72 (1952) 243-280. [499] H. A. DYE, The unitary structure in finite rings of operators, Duke Math. J. 20 (1953) 55-69. [500] H. A. DYE, On the geometry of projections in certain operator algebras, Ann. Math. 61 (1955) 73-89. [501] H. A. DYE, On groups of measure-preserving transformations, 1, Amer. J. Math. 81 (1959) 119-159. [502] H. A. DYE, On groups of measure-preserving transformations, II, A mer. J. Math. 85 (1963) 551-576. [503] M. ECHIGO and M. NAKAMURA, A remark on the concept of channels, Proc. Japan Acad. 38 (1962) 307-309. [504] J. P. ECKMANN and K. OSTERWALDER, An application of Tomitas theory of modular Hilbert algebras: duality for free Bose fields, J. Functional Anal. 13 (1973) 1-12. [505] G. EDGAR, J. ERNEST and S. G. LEE, Weighting operator spectre, to appear. [506] C. M. EDWARDS, C*-algebras of central group extensions, I, Ann. Inst. H. Poincare (A) 10 (1969) 229-246.

REFERENCES

429

[507) C. M. EDWARDS, The operational approach to algebraic quantum theory, I, Commun. Math. Phys. 16 (1970) 207-230. [508) E. G. EFFROS, Order ideals in a C*-algebra, Duke Math. J. 30 (1963) 391-412. [509) E. G. EFFROS, A decomposition theory for representations of C*-a1gebras, Trans. Amer. Math. Soc. !07 (1963) 83-106. [510) E. G. EFFROS, Order ideals in a C*-algebra and its dual, Duke Math. J. 30 (1963) 391-412. [511) E. G. EFFROS, Transformation groups and C*-algebras, Ann. Math. 81 (1965) 38-55. [512) E. G. EFFROS, The Borel space of Von Neumann algebras on a separable Hilbert space, Pacific J. Math. 15 (1965) 1153-1164. [513) E. G. EFFROS, Global structure in Von Neumann algebras, Trans. Amer. Math. Soc. 21 (1966) 434-454. [514) E. G. EFFROS, The canonical measures for separable C*-algebras, Amer. J. Math. 92 (1970) 56-60. [515) E. G. EFFROS, Property r and inner amenability, Proc. Amer. Math. Soc. 47 (1975) 483-486. [516) E. G. EFFROS and F. HAHN, Locally compact transformation groups and C* algebras, Bull. Amer. Math. Soc. 73 (1967) 222-226. [517) E. G. EFFROS and F. HAHN, Locally compact transformation groups and C* algebras, Mem. Amer. Math. Soc. 75 (1967). [518) E. G. EFFROS and E. C. LANCE, Tensor product of operators algebras, to appear. [519) E. G. EFFROS and S. ST0RMER, Jordan algebras of self-adjoint operators, Trans. Amer. Math. Soc. 127 (1967) 313-316. [520) L. EHRENPREIS and F. I. MAUTNER, Some properties of the Fourier transform on semi-simple Lie groups, Ann. Math. 61 (1955) 406-439. [521) L. EHRENPREIS and F. I. MAUTNER, Some properties of the Fourier transform on semi-simple Lie groups, II, Trans. Amer. Math. Soc. 84 (1957) 1-55. [522) L. EHRENPREIS and F. I. MAUTNER, Some properties of the Fourier transform on semi-simple Lie groups, III, Trans. Amer. Math. Soc. 90 (1959) 431-484. [523) K. E. EKMAN, Indices on C*-algebras through representations in the Calkin algebra, Duke Math. J. 41 (1974) 413-432. [524) G. A. ELLIOTT, Derivations of matroid C*-algebras, Invent. Math. 9 (1970) 253 269. [525] G. A. ELLIOTT, A weakening of the axioms for a C*-algebra, Math. Ann. 189 (1970) 257-260. [526) G. A. ELLIOTT, An extension of some results of Takesaki in the reduction theory of Von Neumann algebras, Pacific J. Math. 39 (1972) 145-148. [527] G. A. ELLIOTT, Ideal-preserving automorphisms of postliminary C*-algebras, Proc. Amer. Math. Soc. 27 (1971) 107-109. [528] G. A. ELLIOTT. Another weak Stone-Weierstrass theorem for C*-algebras, Can. Math. Bull. 15 (1972) 355-358. [529] G. A. ELLIOTT. Convergence of automorphisms in certain C*-algebras. J. Func tional Anal. 11 (1972) 204-206. [530) G. A. ELLIOTT. Some C*-algebras with outer derivations, Rocky Mountains J. Math. 3 (1973) 501-506. [531] G. A. ELLIOTT, Perspectivity in the projection lattice of an AW*-algebra. Bull. Amer. Math. Soc. 38 (1973) 367-368.

430

REFERENCES

[532) G. A. ELLIOTT, Derivations of matroid C*algebras, II, Ann. Math. 100 (1974) 407-422. [533] G. A. ELLIOTT, Some C*-algebras with outer derivations. II, Can. J. Math. 26 (1974) 185-189. [534) G. A. ELLIOTT, On lifting and extending derivations of approximately finite dimensional C*-algebras, J. Functional Anal. 17 (1974) 395-408. [535] G. A. ELLIOTT, Derivations determined by multipliers on ideals of a C*-algebra, Publ. Res. lnst, Math. Sci. Kyoto Univ. 10 (1975) 721-728. [536) G. A. ELLIOTT, Finite projections in tensor product Von Neumann algebras, Trans. Amer. Math. Soc. 212 (1975) 47--{j(). [537] G. A. ELLIOTT, On the classification of inductive limits of sequences of semi simple finite-dimensional algebras, to appear. [538) G. A. ELLIOTT, On derivations of AW*-algebras. to appear. [539) G. A. ELLIOTT, Automorphisms determined by multipliers on ideals of a C* algebra, J. Functional Anal. 23 (1976) 1-10. [540) G. A. ELLIOTT, On the Radon-Nikodym derivative with a chain rule, to appear. [541) G. A. ELLIOTT, An abstractDauns-Hofmann-Kaplansky multiplier theorem. Can. J. Math. 27 (1975) 827-836. [542) G. A. ELLIOTT, The Mackey-Bore! structure on the spectrum of an approximately finite-dimensional separable C*-algebra, to appear. [543) G. A. ELLIOTT and D. OLESEN, A simple proof of the Dauns-Hofmann theorem, Math. Scand. 34 (1974) 231-234. [544) A. J. ELLIS. Some applications of convexity theory to Banach algebras, Math. Scand. 33 (1973) 23-30. [545) A. J. ELLIS, Central decomposition for compact convex sets, Compositio Math. 30 (1975) 211-219. [546) M. R. EMBRY, Self-adjoint strictly cyclic operator algebras, Pacific J. Math. 52 (1974) 53-57. [547) G. G. EMCH, Algebraic Methods in Statistical Mechanics and Quantum Field Theory (Wiley-Interscience, New York, 1972). [548) M. ENOCK. Produit croise dune algebre de Von Neumann par une algebre de Kac, to appear. [549) M. ENOCK and J. M. SCHWARTZ, Une dualite dans les algebres de Von Neumann. C. R. Acad. Sci. Paris 277 (1973) 683-685. [550) M. ENOCK and J. M. SCHWARTZ. Une categoric dalgebres de Kac, C. R. Acad. Sci. Paris 279 (1974) 643-645. [551) M. ENOCK and J. M. SCHWARTZ, Une dualite dans les algebres de Von Neumann. Bull. Soc. Math. France, Suppl . Mern. 44 (1975). 1-144. [552) M. ENOCK and J. M. SCHWARTZ, Une nouvelle construction du poids dual sur Ie produit croise dune algebre de Von Neumann par un groupe localement compact, cu. Acad. Sci. Paris 282 (1976) 415-418. [553) M. ENOMOTO and K. TAMAKI, On a theorem of Pearcy and Ringrose, Math. Japon. 18 (1973) 253-256. [554) M. ENOMOTO and K. T AMAKI. Freely acting automorphisms of abelian C* algebras, Nagoya Math. J. 56 (1975) 7-11. [555) H. EpSTEIN, Remarks on two theorems of E. Lieb, Commun. Math. Phys. 31 (1973) 317-325.

REFERENCES

431

[556] J. ERDOS, On certain elements of C*-algebras, Illinois J. Math. 15 (1971) 682--693. [557] J. A. ERDOS, An abstract characterization of nest algebras, Quart. J. Math. 22 (1971) 47--63. [558] J. ERNEST, A decomposition theory for unitary representations of locally compact groups, Bull. Amer. Math. Soc. 67 (1961) 385-388. [599] J. ERNEST, A decomposition theory for unitary representations of locally compact groups, Trans. Amer. Math. Soc. 104 (1%2) 252-277. [560] J. ERNEST, A new group algebra for locally compact groups, I, Amer. J. Math. 86 (1964) 467-492. [561] J. ERNEST, Notes on the duality theorem of non-commutative, non-compact topological groups, T6/lOku Math. J. 16 (1964) 291-296. [562] J. ERNEST, A new group algebra for locally compact groups, II, Can. J. Math. 17 (1965) 604--615. [563] J. ERNEST, The representation lattice of a locally compact group, Illinois J. Math. 10 (1966) 127-135. [564] J. ERNEST, Hopf-Von Neumann algebras, in: Functional Analysis (Academic Press, New York, 1967) 195-215. [565] J. ERNEST, The enveloping algebra of a covariant system, Commun. Math. Phys. 17 (1970) 61-74. [566] J. ERNEST, A duality theorem for the automorphism group of a covariant system, Comml/n. Math. Phys. 17 (1970) 75-90. [567] J. ERNEST, A strong duality theorem for separable locally compact groups, Trans. Amer. Math Soc. 156 (1971) 287-307. [568] J. ERNEST, On the topology of the spectrum of a C*-algebra, Math. Ann. 216 (1975) 149-153. [569] J. ERNEST, A classification, decomposition and spectral multiplicity theory for bounded operators on a separable Hilbert space, to appear. [570] T. FACK, Construction de nouveaux facteurs de type II" c.R. Acad. Sci. Paris 280 (1975) 1369-1372. [571] A. FEINTUCH and P. ROSENTHAL, Remarks on reductive operator algebras, Israel J. Math. 15 (1973) 130-136. [572] J. FELDMAN, Isomorphisms of finite type II rings of operators. Ann. Math. 63 (1956) 565-571. [573] J. FELDMAN, Embedding of AW*-algebras, Duke Math. J. 23 (1956) 303-308. [574] J. FELDMAN, Some connections between topological and algebraical properties in rings of operators, Duke Math. J. 23 (1956) 365-370. [575] J. FELDMAN, Non-separability of certain finite factors, Proc. Amer. Math. Soc. 7 (1956) 23-26. [576] J. FELDMAN, The uniformly closed ideals in an A W*-algebra, Bull. Amer. Math. Soc. 62 (1958) 245-246. [577] J. FELDMAN, Borel sets of states and of representations, Michigan Math. J. 12 (1%5) 363-366. (578) J. FELDMAN and J. M. G. FELL, Separable representations of rings of operators, Ann. Math. 65 (1957) 241-249. [579] J. FELDMAN and R. V. KADISON, The closure of the regular operators in a ring of operators, Proc. Amer. Math. Soc. 5 (1954) 909-916. [580] J. FELDMAN and C. C. MOORE, Ergodic equivalence relations, cohomology, and

432

REFERENCES

Von Neumann algebras, Bull. Amer. Math. Soc. 81 (1975) 921-924. [581] J. M. G. FELL. Representations of weakly closed algebras. Math. Ann. 133 (1957) 118-126. [582] J. M. G. FELL, The dual spaces of C*-algebras, Trans. Amer. Math. Soc. 94 (1960) 365-403. [583] J. M. G. FELL, C*-algebras with smooth dual. Illinois J. Math. 4 (1960) 221-230. [584] J. M. G. FELL. The structure of algebras of operator fields, Acta Math. 106 (1%1) 233-280. [585] J. M. G. FELL. Weak containment and induced representations of groups. Can. J. Math. 14 (1962) 237-268. [586] J. M. G. FELL, A new proof that nilpotent groups are CCR, Proc. Amer. Math. Soc. 13 (1%2) 93-99. [587] J. M. G. FELL, Weak containment and Kronecker product of group representations, Pacific J. Math. 13 (1%3) 503-510. [588] J. M. G. FELL. Weak containment and induced representations of groups. II. Trans. Amer. Math. Soc. 110 (1%4) 424-447. [589] J. M. G. FELL, An extension of Mackeys method to Banach *-a1gebraic bundles, Mem. Amer. Math. Soc. 90 (1969). [590] J. M. G. FELL and E. THOMA, Einige Bemerkungen iiber vollsymmetrische Banachsche Algebren, Arch. Math. 12 (1%1) 69-70. [591] P. A. FILLMORE, Perspectivity in projection lattices, Proc. Amer. Math. Soc. 16 (1965) 383-387. [592] P. A. FILLMORE, On products of symmetries, Can. J. Math. 18 (1%6) 897-900 [593] P. A. FILLMORE and D. M. TOPPING. Operator algebras generated by projections, Duke Math. J. 34 (1967) 333-336. [594] P. A. FILLMORE and D. M. TOPPING, A direct integral decomposition for certain operator algebras, Amer. J. Math. 91 (1%9) 11-17. [595] P. A. FILLMORE and D. M. TOPPING. Sums of irreducible operators, Proc. Amer. Math. Soc. 20 (1%9) 131-133. [5%] M. FLENSTED-JENSEN, A note on desintegration, type and global type of Von Neumann algebras, Math. Scand. 24 (1969) 232-238. [597] C. FOIAS and I. KOVACS, Une caracterisation nouvelle des algebres de Von Neumann finies, Acta Sci. Math. Szeged. 23 (1%2) 274-278. [598] R. A. FONTENOT, The double centralizer algebra as a linear space, Proc. Amer. Math. Soc. 53 (1975) 99-103. [599] B. FUGLEDE and R. V. KADISON. On a conjecture of Murray and Von Neumann, Proc. Nat/. Acad. Sci. U.S.A. 37 (1951) 420-425. [600] B. FUGLEDE and R. V. KADISON. On determinants and a property of the trace in finite factors, Proc. Natl. Acad. Sci. U.S.A. 37 (1951) 425-431. [601] B. FuGLEDE and R. V. KADISON, Determinant theory in finite factors, Ann. Math. 55 (1952) 520-530. [602] M. FUJII, On hypercentral states of C*-algebras, Math. Japan. 18 (1973) 215-219. [603] M. FUJII and M. NAKAMURA, On normal approximate spectrum, VI, Proc. Japan Acad. 49 (1973) 596-600. [604] M. FuKAMIYA, On B*-a1gebras, Proc. Japan Acad. 27 (1951) 321-327. [605] M. FUKAMIYA, On a theorem of Gelfand and Neumark and the B*-algebra, Kumamoto J. Sci. 1 (1952) 17-22.

REFERENCES

433

[606] M. FUKAMIY A, M. MISONOU and Z. TAKEDA, On order and commutativity of B*-algebras, T6hoku Math. J. 6 (l954) 89-93. [607] I. FURUTA and R. NAKAMOTO, A metric characterization of cartesian decomposition in a *-a1gebra, Proc. Japan Acad. 44 (1%8) 801-804. [608] P. GAJENDRAGADKAR, Norm of a derivation of a Von Neumann algebra, Trans Amer. Math. Soc. 170 (1972) 165-170. [609] W. D. GARBER, The connexion of duality and causal properties for generalized free fields, Commun. Math. Phys. 42 (l975) 195-208. [610] L. GARDING and A. WIGHTMAN, Representations of the anticommutation relations, Proc. Natl. Acad. Sci. U.S.A. 40 (1954) 617-621. [611] L. GARDING and A. WIGHTMAN, Representations of the commutation relations, Proc. Natl. Acad. Sci. U.S.A. 40 (1954) 622-626. [612] L. T. GARDNER, A note on isomorphisms of C*-algebras, Bull. Amer. Math. Soc. 70 (I%4) 788-791. [613] L. T. GARDNER, On isomorphisms of C*-algebras, Amer. J. Math. 87 (l965) 384-3%. [614] L. T. GARDNER, On the Mackey-Borel structure, Can. J. Math. 23 (1971) 674-678. [615] L. T. GARDNER, On the "third definition" of the topology of the spectrum of a C*-algebra, Can. J. Math. 23 (1971) 445-450. [616] D. J. H. GARLING, On ideals of operators in Hilbert space, Proc. London Math. Soc. 17 (1%7) 115-138. [617) M. GARTENBERG, Extensions of the index in factors of type II-, Proc. Amer. Math. Soc. 43 (1974) 163-168. [618] I. GELFAND and M. NAIMARK, On the imbedding of normed rings into the ring of operators in Hilbert space, Mat. Sb. 12 (l943) 197-213 (in Russian). [619] I. M. GELFAND and M. A. NAlMARK, Normed rings with involution and their representations, Izv. Akad. Nauk SSR 12 (1948) 445-480 (in Russian). [620] I. M. GELFAND and D. RAlKOV, Irreducible unitary representations of locally compact groups, Mat. Sb. 13 (l943) 301-316 (in Russian). [621] P. GHEZ, R. LIMA and D. TESTARD, Une extension dun theorerne de A. Connes sur les facteurs constructibles, Commun. Math. Phys. 32 (1973) 305-311. [622] R. GILES and H. KUMMER, A matrix representation of a pair of projections in a . Hilbert space, Can. Math. Bull. 14 (1971) 35-44. [623] R. GILES and H. KUMMER, A non-eommutative generalization of topology, Indiana Univ. Math. J. 21 (1971) 91-102. [624] F. GILFEATHER, On the Suzuki structure theory for non self-adjoint operators on Hilbert space, Act. Sci. Math. 32 (1971) 239-249. [625] J. F. GILLE, An exponentiation theorem for unbounded derivations, Ann. Inst. H. Poincare (A) 13 (1970) 215-220. [626] M. S. GLASER, Asymptotic abelianness of infinite factors, Trans. Amer. Math. Soc. 178 (1973) 41-56. [627] A. M. GLEASON, Measures on the closed subspaces of a Hilbert space, J. Rational Mech. Anal. 6 (1957) 885-894. [628] B. W. GLICKFELD, A metric characterization of C(X) and its generalization to C*-algebras, Illinois J. Math. 10 (1966) 547-556. [629] J. GLIMM, On a certain class of operator algebras, Trans. Amer. Math. Soc. 95 (l96O) 318-340.

434

REFERENCES

[630] J. GLIMM, A Stone-Weierstrass theorem for C*-algebras, Ann. Math. 72 (1960) 216-244. [631] r. GLIMM, Type I C*-algebras, Ann. Math. 73 (1961) 572-612. [632] J. GLIMM, Families of induced representations, Pacific J. Math. 12 (1%2) 885-911. [633] J. GLIMM and R. V. KADISON, Unitary operators in C*-algebras, Pacific J. Math. 10 (1%0) 547-556. [634] R. GODEMENT, Sur les relations dorthogonalitede V. Bargmann, C.R. Acad. Sci. Paris 225 (1947) 521-523, 657-659. [635] R. GODEMENT, Les fonctions de type positif et la theorie des groupes, Trans. Amer. Math. Soc. 63 (1948) 1-84. [636] R. GODEMENT, Theorie generale des sommes continues despaces de Banach, c.R. Acad. Sci. Paris 228 (1949) 1321-1323. [637] R. GODEMENT, Sur la theorie des caracteres, I: Definition et classification des caracteres, C.R. Acad. Sci. Paris 229 (1949) 967-%9. [638] R. GODEMENT, Sur la theorie des representations unitaires, Ann. Math. 53 (1951) 68-124. [639] R. GODEMENT, Memoire sur la theorie des caracteres dans les groupes localement compacts unimodulaires, J. Math. Pures Appl. 30 (1951) 1-110. [640] R. GODEMENT, Theorie des caracteres, I: Algebres unitaires, Ann. Math. 59 (1954) 47-62. [641] R. GODEMENT. Theone des caracteres, II: Definition et proprietes generales des caracteres, Ann. Math. 59 (1954) 63-85. [642] M. GOLDMAN, Structure of AW*-algebras, I, Duke Math. J. 23 (1956) 23-34. [643] M. GOLDMAN, On subfactors of factors of type II" Michigan Math. J. 6 (1959) 167-172. [644] V. lA. GOLODEC, On factor representations of type II for the commutation and anticommutation relations, Uspehi Mat. Nauk 20 (1%5) 68-72 (in Russian). [645] V. lA. GOLODEC, On factor representations of anticommutation relations, Dokl. Akad. Nauk SSSR 167 (1966) 19-22 (in Russian). [646] V. lA. GOLODEC, Factor representations of type II, for a Clifford algebra, Dok/. Akad. Nauk SSSR 173 (1%7) 745-747 (in Russian). [647] V. lA. GOLODEC, Certain properties of representations of an infinite-dimensional Clifford algebra, Dokl. Akad. Nauk SSSR 178 (1968) 771-773 (in Russian; English Trans!.: Soviet Math. 9 (1%8) 184-187). [648] V. lA. GOLODEC, A study of approximately finite Von Neumann algebras with finite trace, Dokl. Akad. Nauk SSSR 181 (1968) 1307-1310 (in Russian; English Trans!.: Soviet Math. I) (1968) 1016-1019. [649] V. lA. GOLODEC, The finite groups of automorphisms of hyperfinite factors, Dokl. Akad. Nauk SSSR 188 (1%9) 1214-1216 (in Russian; English Trans!. (with an addendum): Soviet Math. 10 (1969) 1266-1269). [650] V. lA. GOLODEC, The factor representations of type II and III for the commutation relations, Mat. Sb. 78 (1969) 501-511 (in Russian). [651] V. lA. GOLODEC, On hyperfinite factors of type III, Math. Phys. Functional Anal. I (1969) 220-228 (in Russian). [652] V. lA. GOLODEC, Classification of representations of the anticommutation relations, Uspehi. Mat. Nauk 24 (1%9) 3-64 (in Russian; English Trans!.: Russian Math. Surv. 24 (4), 1-63).

REFERENCES

435

[653] V. JA. GOLODEC, On groups of transformations leaving a measure quasi-invariant, Mat. Zametki 7 (1970) 223-227 (in Russian). [654] V. JA. GOLODEC, The classes of approximately finite factors, Funkcion. Ana/. Priloien. 4 (1970) 14-20 (in Russian). [655] V. JA. GOLODEC, Factor representations of the anticommutation relations, Tr. Moskov. Mat. ObIt. 22 (1970) 3-62 (in Russian). [656] V. JA. GOLODEC, Hyperfinite factors of types II., and III, Funkcion. Anal. Priloien 5 (1971) 34-44 (in Russian) .. [657] V. JA. GOLODEC, Tensor and cross products of type III factors, Funkcion. Anal. Priloien. 5 (1971) 96-97 (in Russian). [658] V. JA. GOLODEC, Crossed products of Von Neumann algebras, Uspehi Mat. Nauk 26 (1971) 3-50 (in Russian; English Transl.: Russian Math. Surv. 26, 1-50). [659] V. JA. GOLODEC, Conditional expectation and modular automorphisms of Von Neumann algebras, Funkcion. Ana/. Priloien. 6 (1972) 68-69 (in Russian). [660] V. JA. GOLODEC, Spectral properties of modular operators, Funkcion. Anal. Priloien. 6 (1972) 70-72 (in Russian). [661] V. JA. GOLODEC, Modular operators and the set of asymptomatic ratios, Funkcion. Anal. Priloien. 7 (1973) 77-78 (in Russian). [662] V. JA. GOLODEC, Asymptotic algebra, its applications to the study of modular operators and their spectral properties, Dok/. Akad. Nauk SSSR 220 (1975) 15-18 (in Russian). [663] V. JA. GOLODEC and A. M. STEPIN, On the factors of type II with the property I, Tr. Akad. Nauk SSSR 1 (1969) 229-249 (in Russian). [664] H. GONSHOR, Spectral theory for a class of non-normal operators, Can. J. Math. 8 (1956) 449-461. [665] H. GONSHOR, Spectral theory for a class of non-normal operators, II, Can. J. Math. 10 (1958) 97-102. [666] H. GONSHOR, Injective hulls of C*-algebras, Trans. Amer. Math. Soc. 131 (1968) 315-322. [667] H. GONSHOR, Remarks on the algebra of bounded functions, Math. Z. 188 (1969) 325-328. [668] H. GONSHOR, Injective hulls of C*-algebras, II, Proc. Amer. Math. Soc. 24 (1970) 486-491. [669] E. C. GOOTMAN, Primitive ideals of C*-algebras associated with transformation groups, Trans. Amer. Math. Soc. 170 (1972) 97-108. [670] E. C. GOOTMAN, The type of some C*- and W*-algebras associated with transformation groups, Pacific J. Math. 48 (1973) 93-106. [671] A. GOULLET DE RUGY, Faces parallelisableset topologies faciales sur lespace des etats dune algebre stellaire, C.R. Acad. Sci. Paris 270 (1970) 376-379. [672] E. GRANINER, Exposed points of convex sets and weak sequential convergence, to appear. [673] W. L GREEN, Toplogical dynamics and C*-algebras, Trans. Amer. Math. Soc. 210 (1975) 107-121. [674] F. GREENLEAF and M. MOSKOWITZ, Cyclic vectors for representations of locally compact groups, Math. Ann. 190 (1971) 265-288. [675] E. L. GRIFFIN, Some contributions to the theory of rings of operators, Trans. Amer. Math. Soc. 75 (1953) 471-504.

436

REFERENCES

[676] E. L. GRIFFIN, Some contributions to the theory of rings of operators, II, Trans. Amer. Math. Soc. 79 (1955) 389-400. [677] E. L. GRIFFIN, Everywhere defined linear transformations affiliated with rings of operators, Pacific J. Math. 18 (1968) 489-493. [678] L. GROSS, A noncommutative extension of the Perron Frobenius theorem, Bull. Amer. Math. Soc. 77 (1971) 343-347. [679] L. GROSS, Existence and uniqueness of physical ground states, J. Functional Anal. 10 (1972) 52-109. [680] A. GROTHENDIECK, Rearrangements de fonctions et inegalites de convexite dans les algebres de Von Neumann munies dune trace, Sem. Bourbaki (March 1955) 16 pp. [681] A. GROTHENDIECK, Un resultat sur Ie dual dune C*-algebre, J. Math. Pures Appl. 36 (1957) 97-108. [682] S. GUDDER and J. P. MARCHAUD, Non-commutative probability and Von Neumann algebras, J. Math. Phys. 13 (1972) 799-806. [683] M. GUENIN and B. MISRA, On the Von Neumann algebras generated by the field operators, Nuovo Cimento 30 (1963) 1272-1290. [684] A. GUICHARDET, Une caracterisation des algebres de Von Neumann de type I, cu. Acad. Sci. Paris 248 (1959) 3398-3400. [685] A. GUICHARDET, Sur un problerne pose par G. W. Mackey, C.R. Acad. Sci. Paris 250 (1960) 962-963. [686] A. GUICHARDET, Sur les representations factorielles des C*-algebres, C.R. Acad. Sci. Paris 252 (1961) 1088-1089. [687] A. GUICHARDET, Sur les structures boreliennes du dual et du quasi-dual dune C*-algebre, c.s. Acad. Sci. Paris 253 (1961) 2030-2032. [688] A. GUICHARDET, Une caracterisation des algebres de Von Neumann discretes, Bull. Soc. Math. France 89 (1961) 77-101. [689] A. GUICHARDET, Caracteres des algebres de Banach involutives, Ann. Inst. Fourier 13 (1962) 1-81. [690] A. GUICHARDET, Sur la decomposition des representations des C*-algebres, C.R. Acad. Sci. Paris 258 (1964) 768-770. [691] A. GUICHARDET, Caracteres et representations des produits tensoriels de C* algebres, Ann. Sci. Ecole Norm. Sup. 81 (1964) 189-206. [692] A. GUICHARDET, Sur les produits tensoriels de C*-algebres, Dokl. Akad. Nauk. SSSR 106 (1065) 986-989, 168 (1965) 1231. [693] A. GUICHARDET, Utilisation des sous-groupes distingues ouverts dans letude des representations unitaires des groupes localement compacts, Compositio Math. 17 (1965) 1-35. [694] A. GUICHARDET, Produits tensoriels infinis et representations des relations danti commutation, Ann. Sci. Ecole Norm. Sup. 83 (1966) 1-52. [695] A. GUICHARDET, Sur lhomologie et la cohomologie des algebres de Banach, C.R. Acad. Sci. Paris 262 (1966) 38-41. [696] A. GUICHARDET, Sur la categorie des algebres de Von Neumann, Bull. Sci. Math. 90 (1966) 41-64. [697] A. GUICHARDET, Lecons sur Certaines Algebres Topologiques: Algebre de Von Neumann, algebres topologiques et fonctions holomorphes, algebres de Banach commutatives (Gordon and Breach, New York, 1967).

REFERENCES

437

[698] A. GUiCHARDET, Produits tensoriels continus despaces et dalgebres de Banach, Commun. Math. Phvs. 5 (1967) 262-287. [699] A. GUICHARDET, Sur un theoreme de Sakai, c.R. Acad. Sci. Paris 266 (1968) 974-975. [700] A. GUICHARDET, Algebres dObservables Associies aux Relations de Commutation, (Colin, Paris, 1968). [701] A. GUICHARDET, Tensor products of C"-algebras, I, II, Aarhus Univ. (1969). [702] A. GUICHARDET, Symmetric Hilbert spaces and related topics, Lectures Notes in Math. 261 (Springer, Berlin, 1972). [703] A. GUICHARDET, Sur les systemes dynamiques non commutatifs, C.R. Acad. Sci. Paris 277 (1973) 289-290. [704] A. GUICHARDET, Systemes dynamiques non commutatifs, Asterisque 13, 14 (1974). [705] A. GUiCHARDET, Automorphismes et type de certaines algebres de Von Neumann, to appear. [706] A. GUICHARDET and D. KASTLER, Desintegration des etats quasi-invariants des C"-algebres, J. Math. Pures Appl. 19 (1970) 349-380. [707] A. GUICHARDET and A. WULFSOHN, Sur les produits tensoriels continus despaces hilbertiens, J. Functional Anal. 2 (1968) 371-377. [708] A. GUICHARDET and A. WULFSOHN, Continuous tensor product states which are translation invariant but not quasi-free, Commun. Math. Phys. 17 (1970) 133-142. [709] J. GUNSON, Physical states on quantum logics, I, Ann. Inst. H. Poincare (A) 17 (1972) 295-311. [710] A. GUREVIC, Unitary representations in Hilbert space of a compact topological group, Mat. Sb. 13 (1943) 79-86. [711] R. HAAG, N. M. HUGENHOLTZ and M. WINNINK, On the equilibrium states in quantum statistical mechanics, Commun. Math. Phys. 5 (1967) 215-236. [712] R. HAAG, R. V. KADISON and D. KASTLER, Nets of C"-algebrasand classification of states, Commun. Math. Phys. 16 (1970) 81-104. [713] R. HAAG, R. V. KADISON and D. KASTLER, Asymptotic orbits in a free Fermi gas, Commun. Math. Phys. 33 (1973) 1-22. [714] R. HAAG and D. KASTLER, An algebraic approach to quantum field theory, J. Math. Phys. 5 (1964) 848-861. [715] R. HAAG, D. KASTLER and E. B. TRYCH-POHLMEYER, Stability and equilibrium states, Commun. Math. Phys. 38 (1974) 173-193. [716] R. HAAG and B. SCHROER, Postulates of quantum field theory, J. Math. Phys. 3 (1962) 248-256. [717] R. HAAG and J. A. SWIECA, When does a quantum field theory describe particles, Commun. Math. Phys. 1 (1965) 308-320. [718] U. HAAGERUP, Normal weights on W"-algebras, J. Functional Anal. 19 (1975) 302-317. [719] U HAAGERUP, Operator-valued weights in Von Neumann algebras, to appear. [720J U. HAAGERUP, On the dual weights for crossed products of Von Neumann algebras, I, II, to appear. [721] U. HAAGERUP, Tomitas theory for Von Neumann algebras with a cyclic separating vector, to appear. [722] U. HAAGERUP, L"-spaces associated with arbitrary Von Neumann algebras, to appear.

438

REFERENCES

[723] U. HAAGERUP. The standardform of Von Neumann algebras, Math. Scand. 37 (1975) 271-283. [724] U. HAAGERUP, The centralizer of a weight can be of type III, to appear. [725] Y. HAGA, A note on traces of Von Neumann algebras, Proc. Japan Acad. 44 (1968) 933-938. [726] Y. HAGA, On subalgebras of a cross product Von Neumann algebra, T6hoku Math. J. 25 (1973) 291-305. [727] Y. HAGA, On approximately finite algebras, T6hoku Math. J. 26 (1974) 325-332. [728] Y. HAGA, Crossed products of Von Neumann algebras by compact groups, to appear. [729] Y. HAGA and Z. TAKEDA, Correspondence between subgroups and subalgebras in a cross product Von Neumann algebra. Tohoku Math. J. 24 (1972) 167-190. [730] J. HAKEDA, On property P of Von Neumann algebras, T6hoku Math. J. 19 (1967) 238-242. [731] J. HAKEDA, On property P of Von Neumann algebras, II, Bull. Yamagata Univ. Natur. Sci. 7 (1968) 5-6. [732] J. HAKEDA and J. TOMIYAMA, On some extension properties of Von Neumann algebras, Tohoku Math. J. 19 (1967) 315-323. [733] A. A. HALL, Derivations of certain C*-algebras, J. London Math. Soc. 5 (1972) 321-329. [734] I. HALPERIN, A remark on a preceding paper by J. von Neumann, Ann. Math. 41 (1940) 554-555. [735] I. HALPERIN, Introduction to Von Neumann algebras and continuous geometry, Can. Math. Bull. 3 (1960) 273-288, S (1962) 59. [736] H. HALPERN, The maximal GCR ideal in an AW*-algebra, Proc. Amer. Math. Soc. 17 (1966) 906-914. [737] H. HALPERN, An integral representation of a normal functional on a Von Neumann algebra, Trans. Amer. Math. Soc. 125 (1966) 32-46. [738] H. HALPERN, Finite sums of irreducible functionals on C*-algebras, Proc. Amer. Math. Soc. 18 (1967) 352-358. [739] H. HALPERN, A spectral decomposition for self-adjoint elements in the maximum OCR ideal of a Von Neumann algebra with applications to non-eommutative integration theory, Trans. Amer. Math. Soc. 133 (1968) 281-306. [740] H. HALPERN, Commutators in properly infinite Von Neumann algebras, Trans. Amer. Math. Soc. 139 (1969) 55-73. [741] H. HALPERN, Module homomorphisms of a Von Neumann algebra into its center, Trans. Amer. Math. Soc. 140 (1969) 183-194. [742] H. HALPERN, Irreducible module homomorphisms of a Von Neumann algebra into its center, Trans. Amer. Math. Soc. 140 (1969) 195-222. [743] H. HALPERN, Embedding as a double commutator in a type I AW*-algebra, Trans. Amer. Math. Soc. 148 (1970) 85-98. [744] H. HALPERN, Commutators mudulo the center in a properly infinite Von Neumann algebra, Trans. Amer. Math. Soc. ISO (1970) 55-68. [745] H. HALPERN, A generalized dual for a C*-algebra, Trans. Amer. Math. Soc. 153 (1971) 139-156. [746] H. HALPERN, Unitary implementation of automorphism groups on Von Neumann algebras, Commun. Math. Phys. 2S (1972) 253-275. [747] H. HALPERN, Integral decomposition of functionals on C*-algebras. Trans. Amer.

REFERENCES

439

Math. Soc. 168 (1972) 371-385. [748] H. HALPERN, Essential central spectrum and range for elements of a Von Neumann algebra, Pacific J. Math. 43 (1972) 349-380. [749] H. HALPERN, Primitive ideals in Von Neumann algebras, Proc. Amer. Math. Soc. 39 (1973) 563-566. [750] H. HALPERN, Mackey Borel structure for the quasi-dual of a separable C*-algebra, Can. J. Math. 26 (1974) 621-628. [751] H. HALPERN, Open projections and Borel structures for C*-algebras, Pacific J. Math. 50 (1974) 81-98. [752] H. HALPERN, Quasi-equivalence classes of normal representations for a separable C*-a1gebra, Trans. Amer. Math. Soc. 203 (1975) 129-140. [753] H. HALPERN, Essential central range and self-adjoint commutators in properly infinite Von Neumann algebras, to appear. [754] H. HALPERN, Proper values in the maximal GCR ideal of a Von Neumann algebra, to appear. [755] H. HALPERN, The norm of an inner derivation of an AW*-algebra, to appear. [756] H. HALPERN, One-parameter automorphism group of generalized KH-algebras, to appear. [757] H. HALPERN, Spectra of operator equations of Von Neumann algebras, to appear. [758] HARISH-CHANDRA, Representations of semi-simple Lie groups, III, Trans. Amer. Math. Soc. 76 (1954) 234-253. [759] HARISH-CHANDRA, Representations of semi-simple Lie groups, VI, Amer. J. Math. 78 (1956) 564-628. [760] HARISH-CHANDRA, Invariant eigendistributions on semi-simple Lie groups, Bull. Amer. Math. Soc. 69 (1963) 117-123. [761] L. A. HARRIS, Banach algebras with involution and Mobius transformations, J. Functional Anal. 11 (1972) 1-16. [762] R. T. HARRIS, A direct integral construction, Duke Math. J. 33 (1966) 535-537. [763] R. HAYDON and S. WASSERMAN, A commutation result for tensor products of C*-a1gebras, Bull. London Math. Soc. 5 (1973) 283-287. [764] A. JA. HELEMSKII, Strong decomposability of finite-dimensional extensions of C*-algebras, Vestnik Mosko.Univ. Mat. Meh. 22 (1967) 50-52 (in Russian). [765] A. JA. HELEMSKII, On a method for calculating and estimating the global homological dimension of Banach algebras, Mat. Sb. 87 (1972) 122-135 (in Russian; English trans!.: Math. USSR Sb. 16 (1972) 125-138. [766] A. HELEMSKII and JA. SINAI. A description of differentiations in algebras of the type of algebras of local observables of spin systems, Functional Anal. Appl. 6 (1973) 343-344 (in Russian). [767] M. HENLE, Spatial representation of groups of automorphisms of Von Neumann algebras with properly infinite commutant, Commun. Math. Phys. 19 (1970) 273 275. [768] M. HENLE, A Lebesgue decomposition theorem for C*-algebras, Can. Math. Bull. 15 (1972) 87-91. [769] M. HENLE, Galois theory of W*-algebras, to appear. [770] J. N. HENRY and D. C. TAYLOR, The l3-topology for W*-a1gebras, Pacific J. Math. 60 (1975) 123-139. [771] R. H. HERMAN, Some remarks on expectations, Acta Sci. Math. Szeged 31 (1970) 125-128.

440

REFERENCES

[772] R. H. HERMAN, Invariant states, Trans. Amer. Math. Soc. 158 (1971) 503-512. [773] R. H. HERMAN, Perturbations of the modular automorphism group, Commun. Math. Phys. 28 (1972) 237-243. [774] R. H. HERMAN, Centralizers and ordering for faithful normal states, J. Functional Anal. 13 (1973) 317-323. [775] R. H. HERMAN, Spectra of automorphism groups of operator algebras, Duke Math. J. 4 (1974) 667-673. [776] R. H. HERMAN, Automorphism groups of operator algebras, Pacific J. Math. SO (1974) 91-99. [777] R. H. HERMAN, The centralizer under tensor products, Math. Scand. 36 (1975) 286-290. [778] R. H. HERMAN, Unbounded derivations, J. Functional Anal. 20 (1975) 234-239. [779] R. H. HERMAN and M. C. REED, Covariant representations of infinite tensor product algebras, Pacific J. Math. 40 (1972) 311-326. [780] R. H. HERMAN and M. T AKESAKI, States and automorphism groups of operator algebras, Commun. Math. Phys. 19 (1970) 142-160. [781] E. HEWIlT and D. McDuFF, Some pathological maximal ideals in operator algebras and measure algebras on groups, Mat. Sb. 83 (1970) 527-546 (in Russian). . [782] R. HIRSCHFELD, Sur lanalyse harmonique dans les groupes localement compacts, Acad. Sci. Paris 246 (1958) 1138-1140. [783] K. H. HOFMANN, Extending C*-algebras by adjoining an identity, in: Contributions to Extension Theory of Topological Structures, Proc. Symp. Berlin., 1967 (VEB Deutscher Verlag der Wiss., Berlin, 1969) 119-125. [784] K. H. HOFMANN, The duality of compact semi-groups and Cr-bigebras, Lecture Notes in Math. 129 (Springer, Berlin, 1970). [785] K. H. HOFMANN, Representations of algebras by continuous sections, Bull. Amer. Math. Soc. 78 (1972) 291-373. [786] A. S. HOLEVO, On the mathematical theory of quantum communication channels, Problemy Peredaci Injormacli 8 (1972) 62-71 (in Russian). [787] A. S. HOLEVO, The theory of statistical decisions on an operator algebra, (Dokl, Akad. Nauk SSSR 218 (1974) 54-57 (in Russian). [788] S. S. HOLLAND, Uniform angle pairs of subspaces in Hilbert space, to appear. [789] E. HONGO, A note on the commutor of certain operator algebras, Bull. Kyushu lnst. Technol. I (1955) 19--22. [790] E. HONGO, On left rings of certain *-algebras, Bull. Kyushu Inst. Technol. 2 (1956) 1-15. [791] E. HONGO, On quasi-unitary algebras with semi-finite left rings, Bull. Kyushu Inst. Technol. 3 (1957) 1-10. [792] E. HONGO, On some properties of quasi-unitary algebras, Bull. Khushu Inst. Technol. 4 (1958) 1-6. [793] E. HONGO, On left multiplicative operators on a quasi-unitary algebra, Bull. Kyushu Inst. Technol. 5 (1959) 19-22. [794] E. HONGO, A structure theory for semi-finite quasi-unitary algebras, Bull. Kyushu Inst. Technol. 7 (1961) 1-17. [795] E. HONGO and M. ORIHARA, A remark on a quasi-unitary algebra, Yokohama Math. J. 2 (1954) 69-72. [796] T. B. HOOVER, Operator algebras with reducing invariant subspaces, Pacific J. Math. 44 (1973) 173-179.

c.s.

REFERENCES

441

[797] A. HOPERWASSER, Boundary representations on C*-algebras with matrix units, Trans. Amer. Math. Soc. 177 (1973) 483-490. [798] R. HOWE, The Brauer group of a compact Hausdorff space and n-homogeneous C*-algebras, Proc. Amer. Math. Soc. 34 (1972) 209-214. [799] R. HOWE, A functional calculus for hyponormal operators, Indiana Univ. Math. J. 23 (1973) 631-644. [800] T. T. HSIEH, Invariant normal positive functionals, J. Math. Anal. Appl. 45 (1974) 556-562. [801] T. T. HSIEH, Invariant normal positive functionals, II, J. Math. Anal. Appl. 47 (1974) 554-562. [802] N. M. HUGENHOLTZ, On the factor type of equilibrium states in quantum statistical mechanics, Commun. Math. Phys. 6 (1%7) 189-193. [803] N. M. HUGENHOLTZ and R. V. KADISON, Automorphisms and quasi-free states of the CAR-algebra, Commun. Math. Phys. 43 (1975) 181-197. [804] N. M. HUGENHOLTZ, M. TAKESAKI and M. WINNINK, Local normality of the KMS states in quantum statistical mechanics, to appear. [805] N. M. HUGENHOLTZ and J. D. WIERINGA, On locally normal states in quantum statistical mechanics, Commun. Math. Phys. 11 (1969) 183-197. [806] A. HULANICKI and R. R. PHELPS, Some applications of tensor products of partially ordered linear spaces, J. Functional Anal. 2 (1%8) 177-20\. [807] T. HURUYA, On the structure of certain C*-algebras, Proc. Japan Acad. 45 (1%9) 354-358. [808] T. HURUYA, C*-algebras having the property (T), Sci. Rept. Niigata Univ. (A) 8 (1971) 7-12. [809] T. HURUYA, The second dual of a tensor product of C*algebras, I, II, Sci. Rept. Niigata Univ. 9 (1972) 35-38, 11 (1974) 21-23. [SIll] T. HURUYA, Boundary representations of a tensor product of C*-algebras, Sci. Rept. Niigata Univ. (A) 11 (1974) 55-59. [1] A. IKUNISHI and Y. NAKAGAMI, Automorphism groups of Von Neumann algebras and semi-finiteness of an infinite tensor product of Von Neumann algebras, to appear. [812] A. IKUNISHI and Y. NAKAGAMI, On invariants G(u) and nu) for an automor phism group of a Von Neumann algebra, Publ. Res. Inst. Math. Sci. Kyoto Uni. 12 (1976) 1-30. [813] A. INOUE, V-space associated with a weight, Mem. Fac. Sci. Kyushu Uni. (A) 27 (1973) 291-308. [814] A. INOUE, Conditional expectations in Von Neumann algebras with respect to weight, Mem. Fac. Sci. Kyushu Univ. (A) 27 (1973) 309-32\. [815] C. T. IONESCU TULCEA, Deux theoremes concernant certains espaces de champs de vecteurs, Bull. Sci. Math. 79 (1955) 1-6. [816] E. J. IRWIN, Concerning time-like asymptotic abelianness, to appear. [817] V. ISTRATESCU and I. ISTRATEscu, On characters of singly generated Von Neumann algebras, Proc. Japan Acad. 47 (1971) 42-43. [818] S. ITO, Positive definite functions on homogeneous spaces, Proc. Jap. Acad. 26 (1950) 17-28. [819] N. JACOBSON, A topology for the set of primitive ideals in an arbitrary ring, Proc. Nat!. Acad. Sci. U.S.A. 31 (1945) 333-338. [820] A. Z. JADCZIK, On some groups of automorphisms of Von Neumann algebras with

442

REFERENCES

cyclic and separating vector, Commun. Math. Phys. 13 (1%9) 142-153. [821] J. JANAS, Toeplitz operators related to certain domains in en, Studia Math. S4 (1975) 73-79. [822] G. JANSEN, Bemerkungen zum Dichtigheitsatz Von Kaplansky, Math. Z. 139 (1974) 159-164. [823) G. JANSEN, Reelle Jordanalgebren mit endlicher Spur, Manuscripta Math. 13 (1974) 237-273. [824] G. JANSEN, Die Strukturendlicher schwach abgeschlossener Jordanalgebren, I, II, Manuscripta Math. 16 (1975) 277-305, 307-332. [825) J. W. JENKINS, On the characterizationof abelian W*-algebras, Proc. Amer. Math, Soc. 3S (1972) 436-438. [826] B. E. JOHNSON, AW*-algebras are QW*-algebras, Pacific J. Math. 23 (967) 97-99. [827] B. E. JOHNSON, Continuity of homomorphisms of algebras of operators, J. London Math. Soc. 42 0%7) 537-541. [828] B. E. JOHNSON, Continuity of homomorphisms of algebras of operators, II, J. London Math. Soc. 1 0%9) 81-84. (829) B. E. JOHNSON, Cohomology in Banach algebras, to appear. [830) B. E. JOHNSON, Perturbations of Banach algebras, to appear. [831] B. E. JOHNSON, R. V. KADISON and J. R. RINGROSE, Cohomology of operator algebras, III: Reduction to normal cohomology, Bull. Soc. Math. France 100 (1972) 73-%. [832] B. E. JOHNSON and S. K. PARROTT, Operators commuting with a Von Neumann algebra modulo the set of compact operators, J. Functional Anal. 11 (1972) 39-61. [833] B. E. JOHNSON and J. R. RINGROSE, Derivation of operator algebras and discrete group algebras, Bull. London Math. Soc. 1 0%9) 70-74. (834) J. P. JURZAK, Poids strictement semi-finis et groupes dautomorphismes modu laires sur une algebre de Von Neumann decomposable, c.R. Acad. Sci. Paris. 279 (974) 221-224. [835] J. P. JURZAK, Decomposable operators. Application to KMS weights in a decom posable Von Neumann algebra, to appear. (836) G. I. KAC, Certain arithmetic properties of ring groups, Funkcion. Anal. Priloien.B (1972) 88-90. [837) R. V. KADISON, Isometries of operator algebras, Ann. Math. 54 (1951) 325-338. [838) R. V. KADISON, Order properties of bounded self-adjoint operators, Proc. Amer. Math. Soc. 2 (951) 505-510. [839) R. V. KADISON, A generalised Schwarz inequality and algebraic invariants for operator algebras, Ann. Math. 56 (1952) 494-503. [840) R. V. KADISON, Infinite unitary groups, Trans. Amer. Math. Soc. 72 (1952) 386-399. (841) R. V. KADISON, Infinite general linear groups, Trans. Amer. Math. Soc. 76 (1954) 66-91. (842) R. V. KADISON, The general linear group of infinite factors, Duke Math. J. 22 (1955) 119-122. (843) R. V. KADISON, On the orthogonalization of operator representations, Amer. J. Math. 77 (1955) 600-621.

REFERENCES

443

[844] R. V. KADISON, Multiplicity theory for operator algebras, Proc. Nail. Acad. Sci. U.S.A. 41 (1955) 169-173. [845] R. V. KADISON, Isomorphisms of factors of infinite type, Can. J. Math. 7 (1955) 322-327. [846] R. V. KADISON, On the additivity of the trace in finite factors, Proc. Natl. Acad. Sci. U.S.A. 41 (1955) 385-387. [847] R. V. KADISON, Operator algebras with a faithful weakly closed representation, Ann. Math. 64 (1956) 175-181. [848] R. V. KADISON, Unitary invariants for representations of operator algebras, Ann. Math. 66 (1957) 304-379. [849] R. V. KADISON, Irreducible operator algebras, Proc. Natl. Acad. Sci. U.S.A. 43 (1957) 273-276. [850] R. V. KADISON, Theory of operators, Part II: operator algebras, Bull. Amer. Math. Soc. 64 (1958) 61-85. [851] R. V. KADISON, The trace in finite operator algebras, Proc. Amer. Math. Soc. 12 (1961) 973-977. [852] R. V. KADISON, States and representations, Trans. Amer. Math. Soc. 103 (1962) 304-319. [853] R. V. KADISON, Normalcy in operator algebras, Duke Math. J. 29 (1962) 459-464. [854] R. V. KADISON, Remarks on the type of Von Neumann algebras of local ob servables in quantum field theory, J. Math. Phvs. 4 (1963) 1511-1516. [855] R. V. KADISON. The trace in finite operator algebras, Mimeograph, Columbia Univ., New York (1963). [856] R. V. KADISON, Transformations of states in operator theory and dynamics, Topology 3 (1965) 177-198. [857] R. V. KADISON, Derivations of operator algebras, Ann. Math. 83 (1966) 280-293. [858] R. V. KADISON, The energy momentum spectrum of quantum field, Commun. Math. Phys. 4 (1967) 258-260. [859] R. V. KADISON, Lectures on Operator Algebras, Cargese Lectures in Theor. Phys. (Gordon and Breach, New York, 1967). [860] R. V. KADISON, Strong continuity of operator functions, Pacific J. Math. 26 (1968) 121-129. [861] R. V. KADISON, Mappings of operator algebras, Actes Congres Intern. Math., 1970. Vol. 2, pp. 389-393. [862] R. V. KADISON, Some analytic methods in the theory of operator algebras. in: Lectures in Modem Analysis and Applications, II, Lecture Notes in Math. 140 (Springer, Berlin, 1970) 8-29. [863] R. V. KADISON, Normal states and unitary equivalence of Von Neumann algebras, Lectures at Univ. of Pennsylvania, Philadelphia, Pa. (1972). [864] R. V. KADISON, A note on derivations of operator algebras, Bull. London Math. Soc. 7 (1975) 41-44. [865] R. V. KADISON and D. KASTLER, Perturbations of Von Neumann algebras, I: Stability of type, Amer. J. Math. 94 (1972) 38-54. [866] R. V. KADISON, E. C. LANCE and J. R. RINGROSE, Derivations and automor phisms of operator algebras, II, J. Functional Anal. 1 (1967) 204-221. [867] R. V. KADISON and G. K. PEDERSEN, Equivalence in operator algebras, Math. Scand. 27 (1970) 205-222.

444

REFERENCES

[868] R. V. KADISON and J. R. RINGROSE, Derivations of operator group algebras, Amer. J. Math. 88 (1966) 562-576. [869] R. V. KADISON and J. R. RINGROSE, Automorphisms of operator algebras, Bull. Amer. Math. Soc. 72 (1%6) 1059-1063. [870] R. V. KADISON and J. R. RINGROSE, Derivations and automorphisms of operator algebras, Commun. Math. Phys. 4 (1%7) 32-63. [871] R. V. KADISON and J. R. RINGROSE, Cohomology of operator algebras, I: Type 1 Von Neumann algebras, Acta Math. 126 (1971) 227-243. [872] R. V. KADISON and J. R. RrNGROSE, Cohomology of operator algebras, II: Extended cobounding and the hyperfinite case, Arch. Math. (Basel) 9 (1971) 55-63. [873] R. V. KADISON and J. R. RINGROSE, Algebraic automorphisms of operator algebras, J. London Math. Soc. 8 (1974) 329-334. [874] R. V. KADISON and I. M. SINGER, Some remarks on representations of connected groups, Proc. Natl. Acad. Sci. U.S.A. 38 (1952) 419-423. [875] R. V. KADISON and I. M. SINGER, Three test problems in operator theory, Pacific J. Math. 7 (1957) 1101-1106. [876] R. V. KADISON and I. M. SINGER, Extensions of pure states, Amer. J. Math. 81 (1959) 383-400. [877] R. V. KADISON and I. M. SINGER, Triangular operator algebras. Fundamentals and hyperreducible theory, Amer. J. Math. 82 (1960) 227-259. [878] D. A. KAJDAN, On the relations between the dual space of a group and the structure of its closed subgroups, Funkcion. Anal. Priloien. 1 (1967) 71-74 (in Russian). [879] R. R. KALLMAN, Uniform continuity, unitary groups and compact operators, J. Functional Anal. 1 (1967) 245-253. [880] R. R. KALLMAN, The strong bounded topology on groups of automorphisms of a Von Neumann algebra, Proc. Amer. Math. Soc. 23 (1969) 367-372. [881] R. R. KALLMAN, A remark on a paper of J. F. Aarnes, Commun. Math. Phys. 14 (1969) 13-14. [882] R. R. KALLMAN, Unitary groups and automorphisms of operator algebras. Amer. J. Math. 91 (1969) 785-806. [883] R. R. KALLMAN, A generalization of free action, Duke Math. J. 36 (1969) 781-789. [884] R. R. KALLMAN, A generalization of a theorem of Berger and Coburn, J. Math. Mech. 19 (1%9) 1005-1010. [885] R. R. KALLMAN, One-parameter groups of automorphisms of II, Von Neumann algebras, Proc. Amer. Math. Soc. 24 (1970) 336-340. [886] R. R. KALLMAN, Groups of imler automorphisms of Von Neumann algebras, J. Functional Anal. 7 (1971) 43-60. [887] R. R. KALLMAN, Spatially induced groups of automorphisms of certain Von Neumann algebras, Trans. Amer. Math. Soc. 156 (1971) 505-515. [888] R. R. KALLMAN, A theorem on the restriction of type I representations of a group to certain of its subgroups, Proc. Amer. Math. Soc. 40 (1973) 291-296. [889] R. R. KALLMAN, Square-integrable representations are direct summands of the left regular representations, to appear. [890] J. KAMINKER and C. SCHOCHET, Ext(X) from a homological point of view, in: Lecture Notes in Math. 34S (1973) 129-142. [891] J. KAMINKER and C. SCHOCHET, Steenrod homology and operator algebras, Buli. Amer. Math. Soc. 81 (1975) 431-437.

REFERENCES

445

[892J E. KANIUTH, Der Type der regularen Darstellung diskreter Gruppen, Math. Ann. 182 (1969) 334-339. [893J I. KAPLANSKY, Normed algebras, Duke Math. J. 6 (1949) 399-418. [894J I. KAPLAN SKY , Groups with representations of bounded degree, Can. J. Math. 1 (1949) 105-112. [895J I. KAPLANSKY, Quelques resultats sur les anneaux doperateurs, CiR. Acad. Sci. Paris 231 (1950) 485-486. [896] I. KAPLANSKY, The structure of certain operator algebras, Trans. Amer. Math. Soc. 70 (1951) 219--255. [897J I. KAPLANSKY, Group algebras in the large, Tohoku Math. J. 3 (1951) 249-256. [898J I. KAPLANSKY, A theorem on rings of operators, Pacific J. Math. 1 (1951) 227-232. [899] I. KAPLANSKY, Projections in Banach algebras, Ann. Math. 53 (1951) 235-249. [900] I. KAPLANSKY, Representations of separable algebras, Duke Math. J. 19 (1952) 219--222. [901] I. KAPLANSKY, Algebras of type I, Ann. Math. 56 (1952) 460-472. [902] I. KAPLANSKY, Modules over operator algebras, Amer. J. Math. 75 (1953) 839-853. [903] I. KAPLANSKY, Functional analysis, Some Aspects of Analysis and Probability (Wiley, New York, 1958) 1-34. [904J I. KAPLANSKY, Rings of Operators (Benjamin, New York, 1968). [905] I. KAPLANSKY, Algebraic and analytic aspects of operator algebras, Regional Conf. Ser. in Math. (Amer. Math. Soc., Providence, R.I., 1970).

[906) M. KAROUBI, Matrices de Jacobi, periodicite de Bott et C*-algebres, C.R. Acad. Sci. Paris 268 (1969) 1091-1094. [907] I. KASAHARA and H. TAKAI, Approximate proper values and characters of Calgebras, Proc. Japan Acad. 48 (1972) 91-93. [908] D. KASTLER, The C-algebra of a free boson field, Commun. Math. Phys. 1
(1965) 14-48. [909J D. KASTLER, Topics in the algebraic approach to field theory, Cargese Lectures (Gordon and Breach, New York, 1967) 289-302. [910] D. KASTLER, G. LOUPIAS, M. MEBKHOUT and L. MICHEL, Central decom

position of invariant states. Applications to the group of time translations and of euclidean transformations in algebraic field theory, Commun. Math. Phys. 27 (1972)
195-222. [911] D. KASTLER and J. C. T. POOL, Quasi-unitary algebras attached to temperature

[912J [913J [914J [915J

states in statistical mechanics; a comment on the work of Haag, Hugenholtz and Winnink, Commun. Math. Phys. 12 (1969) 175-192. D. KASTLER and D. W. ROBINSON, Invariant states in statistical mechanics, Commun. Math. Phys. 3 (1966) 151-180. D. KASTLER, D. W. ROBINSON and A. SWIECA. Conserved currents and as sociated symmetries; Goldstones theorem, Commun. Math. Phys. 2 (1966) 108-120. D. KASTLER, M. SIRUGUE and J. C. TROTIN, Commutants of certain operator algebras on Fock space Ann. Inst. H. Poincare (A) 4 (1966) 77-82. Y. KATAYAMA, The tensor product of weights, Proc. Japan Acad. 50 (1974)

430-432. [916] E. T. KEHLET, On the monotone sequential closure of a C-algebra, Math. Scand. 25 (1969) 59--70. [917) J. L. KELLEY, Commutative operator algebras, Proc. Nat!. Acad. Sci. U.S.A. 38 (1952) 598-{)05.

446

REFERENCES

[918] J. L. KELLY and R. L. VAUGHT, The positive cone in Banach algebras, Trans. Amer. Math. Soc. 74 (1953) 44-55. [919] M. R. W. KERVIN, The trace-class of a full Hilbert algebra, Trans. Amer. Math. Soc. 178 (1973) 259-270. [920] S. S. KHORUZY, The extended locality and other properties of local W*-algebras, Teoret. Mat. Fiz. 1 (1%9) 95-100 (in Russian). (921] S. S. KHORUZY, The structure of the local W*-algebras in quantum field theory, Teoret. Mat. Fiz. 2 (1970) 350-360 (in Russian). [922] A. A. KIRILLOV, Unitary representations of nilpotent Lie groups, Uspehi Mat. Nauk 17 (1962) 57-110 (in Russian). [923] A. A. KIRILLOV, Dynamical systems, factors and representations of groups, Russian Math. Surv. 22 (1967) 63-75 (in Russian). [924] E. V. KISSIN, C*-algebras generated by dynamical systems and weighted shifts, Dokl. Akad. Nauk SSSR 216 (1974) 1215-1218 (in Russian). [925] E. V. KISSIN, C*-algebras generated by dynamical systems and N-dimensional weighted shifts, Dokl. Akad. Nauk. SSSR 219 (1974) 1061-1064 (in Russian). [926] C. W. KOHLS and L. J. LARDY, Extensions and retractions of algebras of continuous functions, TIlinois J. Math. 12 (1968) 539-549. [927] M. KONDO, Sur la notion de dimension, Proc. Imper. Acad. Tokyo 19 (1943) 215-223. [928] M. KONDO, Les anneaux des operateurs et les dimensions, Proc. Imper. Acad. Tokyo 20 (1944) 389-398. [929] M. KONDO, Sur les sommes directes des espaces lineaires, Proc. Imper. Acad. Tokyo 20 (1944) 425-431. [930] M. KONDO, Sur la reductibilite des anneaux des operateurs, Proc. Imper. Acad. Tokyo 20 (1944) 432-438. [931] M. KONDO, Le produit kroneckerien infini des espaces lineaires, Proc. lmper. Acad. Tokyo 20 (1944) 569-579. [932] M. KOND6, Les anneaux des operateurs et les dimensions, II, Proc. lmper. Acad. Tokyo 20 (1944) 689-693. [933] P. KOOSIS, An irreducible unitary representation of a compact group is finite dimensional, Proc. Amer. Math. Soc. 8 (1957) 712-715. [934] I. KOVACS, Un complement a la theorie de "integration non commutative, Acta Sci. Math. Szeged 21 (1960) 7-11. [935] I. KOVACS, Theoremes ergodiques non commutatifs, CiR. Acad. Sci. Paris 253 (1961) 770-771. [936] I. KOVACS, Sur certains automorphismes des algebres hilbertiennes, Acta. Sci. Math. Szeged 22 (1961) 234-242. [937] I. KOVACS, Ergodic theorems for gages, Acta. Sci. Math. Szeged 24 (1963) 103-118. [938] I. KOVACS, Power bounded operators and finite type Von Neumann algebras, in: Hilbert Space Operators, Colloq. Math. Soc. Janes Bolyai (North-Holland, Am sterdam 1970) 351-352. [939] I. KOVACS and J. Szucs, Theorernes de type ergodique dans les algebres de Von Neumann, CiR. Acad. Sci. Paris 262 (1966) 341-344. [940] I. KOVACS and J. Szucs, Ergodic type theorem in Von Neumann algebras, Acta Sci. Math. Szeged 27 (1966) 233-246. [941] L. KOVACS and J. Szucs, A note. on invariant linear forms on Von Neumann algebras, Acta Sci. Math. Szeged 30 (1969) 35-37.

REFERENCES

447

[942) F. KRAUSS, Structure theory of C*-algebras, Dissertation, Tulane Univ. New Orleans, La., (1973). [943) F. KRAUSS and T. C. LAWSON, Examples of homogeneous C*-algebras, Mem. Amer. Math. Soc. 148 (1974) 153-164. [944] W. KRIEGER, On constructing non-t-isomcrphic hyperfinite factors of type III, J. Functional Anal. 6 (1970) 97-109. [945) W. KRIEGER, On the Araki-Woods asymptotic ratio set and non-singular trans formations of a measure space, in: Contributions to Ergodic Theory and Pro bability, Lecture Notes in Math. 160 (Springer, Berlin, 1970). [946) W. KRIEGER, On a class of hyperfinite factors that arise from null-recurrent Markov chains, J. Functional Anal. 7 (1971) 27-42. [947) W. KRIEGER, On hyperfinite factors and non-singular transformations of a measure space, to appear. [948) W. KRIEGER, On ergodic flows and the isomorphism of factors. Math. Ann. 223 (1976) 19-70. [949) R. A. KUNZE, LP-Fourier transforms on locally compact unimodular groups, Trans. Amer. Math. Soc. 89 (1958) 519-540. [950) R. A. KUNZE and E. M. STEIN, Uniformly bounded representations and harmonic analysis of the 2 x 2 real unimodular group, Amer. J. Math. 82 (1960) 1-62. [951] M. KURANISHI, On non-connected maximally almost periodic groups, Tohoku Math. J. 2 (1950) 40-46. [952) D. LAISON, On W*-imbedding of AW*-algebras. Proc. Amer. Math. Soc. 35 (1972) 499-502. [953) D. LAISON and G. LAISON, Topological dynamics on C*-algebras, Trans. Amer. Math. Soc. 204 (1975) 197-205. [954] E. C. LANCE, Automorphisms of postliminal C*-algebras, Pacific J. Math. 23 (1967) 547-555. [955) E. C. LANCE, Inner automorphisms of UHF algebras, J. London Math. Soc. 43 (1968) 681-688. [956) E. C. LANCE. Automorphisms of certain operator algebras, A mer. J. Math. 91 (1969) 160--174. [957) E. C. LANCE, Notes on the Glimm-Sakai theorem, Lecture Notes, Univ. of Newcastle upon Tyne (1970--71). [958] E. C. LANCE, On nuclear C*-algebras, J. Functional Anal. 12 (1973) 157-176. [959) E. C. LANCE, Direct integrals of left Hilbert algebras, Math. Ann. 216 (1975) 11-28. [960] L. J. LANDAU, A note on extended locality, Commun. Math. Phys, 13 (1969) 246--253. [961] M. B. LANDSTAD, Covariant systems over a compact group, to appear. [962] M. B. LANDSTAD, Duality theory for covariant systems, to appear. [963] O. E. LANFORD, The KMS states of a quantum spin system, in: Colloq. Sur les Systemes Ii un Nombre Infini de Degres de Liberte (CNRS, Paris, 1970) 146---154. [964) O. LANFORD and D. RUELLE, Integral representations of invariant states on B*-algebras, J. Math. Phys. 8 (1967) 1460--1463. [965] O. LANFORD and D. W. ROBINSON, Statistical mechanics of quantum spin systems. III, Commun. Math. Phys. 9 (1968) 327-338. [966] O. E. LANFORD and D. W. ROBINSON, Mean entropy of states in quantum statistical mechanics, J. Math. Phys. 9 (1968) 1120--1125.

448

REFERENCES

[%7] O. E. LANFORD, and D. RUELLE. Observables at infinity and states with short range correlations in statistical mechanics, Commun. Math. Phys. 13 (1%9) 194 215. [%8] J. LANGERHOLC and B. SCHROER, On the structure of the Von Neumann algebras generated by local functions of the free Base field, Commun. Math. Phys. I (1965) 215-239. [%9] C. LANSKY, The group of units of a simple ring, II, J. Algebra 16 (1970) 108-128. [970] J. M. LASRY, Un theoreme de convexite dans les algebres de Von Neumann, CiR. A cad. Sci. Paris. 281 (1975) 153-154. [971] K. B. LAURSEN, Tensor products of Banach *-algebras. Trans. A mer. Math. Soc. 136 (1%9) 467-487. [972] K. B. LAURSEN, A note on lifting of matrix units in C*-algebras, Math. Scand. 33 (1973) 338-342. [973] K. B. LAURSEN and A. M. SINCLAIR, Lifting matrix units in C*-algebras, II, Math. Scand. 37 (1975) 167-172. [974] K. B. LAURSEN and J. D. STEIN, Automatic continuity in Banach spaces and algebras, Amer. J. Math. 95 (1973) 495-506. [975] A. J. LAZAR and D. C. TAYLOR. Double centralizers of Pedersens ideal of a C*-algebra, Bull. Amer. Math. Soc. 78 (1972) 992-997. [976] A. J. LAZAR and D. C. TAYLOR, Double centralizer of Perdersens ideal of a C*-algebra, II, Bull. Amer. Math. Soc. 79 (1973) 361-366. [977] A. J. LAZAR and D. C. TAYLOR, A. Dauns-Hofmann theorem for ras. Mem. Amer. Math. Soc. 148 (1974) 135-144. [978] A. LEBOW, Spatial homomorphisms of operator algebras, Indiana Univ. Math. J.24 (1975) 865-873. [979] A. LEBOW and A. SCHROEDER, Bernstein theorem for projections. Proc. Amer. Math. Soc. 19 (1%8) 144-145. [980] M. J. LENNON. Direct integrals of locally measurable operators. Math. Seand. 32 (1973) 123-132. [981] H. LEPTIN, Zur Reduktionstheorie Hilbertscher Raiime. Math. Z. 69 (1958) 4()....58. [982] H. LEPTIN, Reduktion linearer Funktionale auf Operatorringen, Abh. Math. Sem. Univ. Hamburg 22 (1958) 98-113. [983] H. LEPTIN. Verallgemeinerte L -algebren und projektive Darstellungen lokal Kompakter Gruppen, I, II. lncent. Math. 3 (1967) 257-28t. 4 (1%7) 68-86. [984] H. LEPTIN, Darstellungen verallgemeinerter L-Algebren. Invent. Math. 5 (1%8) 192-215. [985] H. LEPTIN, A separable postliminal C*-algebra without maximal closed ideals, Trans. Amer. Math. Soc. 159 (1971) 489-496. [986] H. LEPTIN, Darstellungen verallgemeinerter L-Algebren, II, in: Lecture Notes in Math. 247 (Springer, Berlin, 1972) 25t-307. [987] J. T. LEWIS and P. N. M. SISSON, A C*-algebra of the two-dimensional Ising model, Commun. Math. Phvs. 44 (1975) 279-292. [988] E. LIEB and M. RUSKAI. Some operator inequalities of the Schwarz type, to appear. [989] A. LIEBERMAN, Measurable vectors for Von Neumann algebras. J. Functional Anal. 18 (1975) 191-212. [990] G. LINDBLAD, Expectations and entropy inequalities for finite quantum systems.

REFERENCES

449

Commun. Math. Phys. 39 (1974) 111-119. [991] A. A. LODKIN, Every measure on the projectors of a W*-algebra can be extended to a state, Funkcion. Anal. Priloi. 8 (1974) 54-58 (in Russian). [992] R. I. LOEBL, Injective Von Neumann algebras, Proc. Amer. Math. Soc. 44 (1974) 46-48. [993] A. I. LOGINOV and V. S. SULMAN, The Sarason theorem and the Radjavi Rosenthal conjecture, Dokl. Akad. Nauk SSSR 205 (1972) 284-285 (in Russian). (994) A. I. LOGiNOV and V. S. SULMAN. The hereditary and intermediate reflexivity of W*-algebras, Dok/. Akad. Nauk. SSSR 212 (1973) 810-812 (in Russian). [995] L. H. LOOMIS, An Introduction to Abstract Harmonic Analysis, (Van Nostrand, New York, 1953). [996] L. H. LOOMIS, The lattice-theoretic background of the dimension theory of operator algebras, Mem. Amer. Math. Soc. 18 (1955). [997] L. H. LOOMIS, Unique direct integral decomposition of convex sets, Amer. J. Math. 84 (1962) 509-526. [998) F. LORENZ, Die Epimorphismen der Ringe von Operatoren, Arch. Math. (Basel) 20 (1969) 48-53. [999] G. LOUPIAS and S. MIRACLE-SOLE, C*algebres des systernes canoniques, I, Commun. Math. Phys. 2 (1966) 31-48. [1000] G. LOUPIAS and S. MIRACLE-SOLE, C*algebres des systernes canoniques, Il, Ann. Inst. H. Poincare 6 (1966) 39-58. (1001) E. LUFf, The two-sided closed ideals of the algebra of bounded linear operators of a Hilbert space, Czech. Math. J. 18 (1968) 595-605. [1002] G. LUMER, Etats, algebres quotients et sous-espaces invariants, CiR. Acad. Sci. Paris 274 (1972) 1308-1311. [1003] G. W. MACKEY, Induced representations of groups, Amer. J. Math. 73 (1951) 576-592. [1004] G. W. MACKEY, Induced representations of locally compact groups. I, Ann. Math. 5S (1952) 101-139. [1005] G. W. MACKEY, Induced representations of locally compact groups, II: The Frobenius reciprocity theorem, Ann. Math. 58 (1953) 193-221. (1006) G. W. MACKEY, The theory of group representations, Mimeographed Notes, Univ. of Chicago, Chicago. Ill. (1955). [1007) G. W. MACKEY, Borel structure in groups and their duals, Trans. Amer. Math. Soc. 85 (1957) 134-165. [1008] G. W. MACKEY, Unitary representations of group extensions, Acta Math. 99 (1958) 265-311. (1009) G. W. MACKEY. Induced representations and normal subgroups, in: Proc. Intern. Symp. on Linear Spaces, Jerusalem (1960). [1010] S. MAEDA, Relative dimensionality in operator rings. J. Sci. Hiroshima Unit.. II (1941) 1-6. [1010] S. MAEDA, Relative dimensionality in operator rings, J. Sci. Hiroshima Univ. 11 19 (1955) 211-237. [1012] S. MAEDA, Lengths of projections in rings of operators, J. Sci. Hiroshima Univ. 20 (1956) 5-11. [1013] S. MAEDA, On the lattice of projections of a Baer *-ring, J. Sci. Hiroshima Univ.22 (1958) 75-88.

450

REFERENCES

[1014] B. D. MALVIYA, A problem concerning weakly completely continuous A*-alge. bras, Amer. Math. Monthly 81 (1974) 267-268. [1015] B. D. MALVIYA and B. J. TOMIUK, Multiplier operators on B*-algebras, Proc, Amer. Math. Soc. 31 (1972) 505-510. [1016] J. MANUCEAU, Etude de quelques automorphismes de la C*-algebre du champ de bosons libres, Ann. Inst. H. Poincare (A) 8 (1968) 117-138. [1017] J. MANUCEAU, C*algebre des relations de commutation, Ann. Inst. H. Poincare (A) 8 (1968) 139-161. [1018] J. MANUCEAU, Etude algebrique des etats quasi-libres, Cargese Lectures in Phys. 4 (1970) 303-322. [1019] J. MANUCEAU, J. NAUDTS and A. VERBEURRE, Entropy and normal states, Commun. Math. Phys. 27 (1972) 327-338. [1020] J. MANUCEAU, F. Roccx and D. TESTARD, On the product form of quasi-free states, Commun. Math. Phys. 12 (1969) 43-57. [1021] J. MANUCEAU, M. SIRUGUE, D. TESTARD and A. VERBEURRE, The smallest C*-algebra for canonical commutation relations, Cammun. Math. Phys. 32 (1973) 231-243. [1022] J. MANUCEAU and J. C. TROTIN, On lattice spin systems, to appear. [1023] J. MANUCEAU and A. VERBEURRE, Quasi-free states of the CCR algebra and Bogoliubov transformations, Cammun. Math. Phys. 9 (1968) 293-302. [1024] J. MANUCEAU and A. VERBEURRE, Non factor quasi-free states of the CAR algebra, Cammun. Math. Phys. 18 (1970) 319-329. [1025] O. MARECHAL, Operateurs decomposables dans les champs mesurables despaces de Hilbert, C.R. Acad. Sci. Paris 266 (1968) 710-713. [1026] O. MARECHAL, Operateurs decomposables dans les espaces L'~, CiR. Acad. Sci. Paris 266 (1968) 1043-1045. [1027] O. MARECHAL, Decomposition des operateurs dans les champs mesurables despaces de Hilbert, C.R. Acad. Sci. Paris 267 (1968) 636-639. [1028] O. MARECHAL, Champs mesurables despaces hilbertiens, Bull. Sci. Math. 93 (1969) 113-143. [1029] O. MARECHAL, Champs mesurables despaces hilbertiens, C.R. Acad. Sci. Paris 270 (1970) 1316-1319. [1030] O. MARECHAL, Topologie et structure borelienne sur lensemble des algebres de Von Neumann, C.R. Acad. Sci. Paris 276 (1973) 847-850. [1031] O. MARECHAL, Une remarque sur un theoreme de Glimm, Bull. Sci. Math. 99 (1975) 41-44. [1032] K. MATOBA, A remark on a theorem of H. Araki. Proc. Japan Acad. 42 (1966) 344-346. [1033] S. MATSUSHITA, Sur Ie theoreme de Plancherel, Proc. Japan Acad. 30 (1954) 557-561. [1034] F. I. MAUTNER, The completeness of the irreducible unitary representations of a locally compact group, Proc. Natl. Acad. Sci. U.S.A. 34 (1948) 52-54. [1035] F. I. MAUTNER, The structure of the regular representation of certain discrete groups, Duke Math. J. 17 (1950) 437-441. [1036] F. I. MAUTNER, Unitary representations of locally compact groups, 1, Ann. Math. 51 (1950) 1-25. [1037] F. I. MAUTNER, Unitary representations of locally compact groups, 11. Ann. Math. 52 (1950) 528-556.

REFERENCES

451

[1038] F. 1. MAUTNER, Infinite-dimensional representations of certain groups, Proc. Amer. Math. Soc. 1 (1950) 582-584. [1039] F. 1. MAUTNER, The regular representation of a restricted direct product of finite groups, Trans. Amer. Math. Soc. 70 (1951) 531-548. [1040] F. 1. MAUTNER, Induced representations, Amer. J. Math. 74 (1952) 737-758. [1041] F. 1. MAUTNER, Note on the Fourier inversion formula on groups, Trans. Amer. Math. Soc. 78 (1955) 371-384. [1042] M. E. MAYER, Automorphism groups of C*-algebras, Fell bundles, W*-bigebras and the description of internal symmetries in algebraic quantum theory, Acta Phys. Austriaca, Suppl, 8 (1971) 117-226. [1043] C. MCCARTHY, Optimal conditioning of operators on Hilbert space, in: Proc. Symp. on Functional Analysis (Academic Press, New York, 1970) 107-126. [1044] C. A. MCCARTHY, The norm of a certain derivation, Pacific J. Math. S3 (1974) 515-518. [1045] E. A. MCCHAREN, A characterizationof dual B*-algebras, Proc. Amer. Math. Soc. 37 (1973) 84. [1046] E. A. MCCHAREN. A new topology on B*-algebras arising from the Arens products. Proc. Amer. Math. Soc. 37 (1973) 77-83. [1047] D. McDuFF, A countable infinity of Il,-factors, Ann. Math. 90 (1969) 361-371. [I048J D. McDuFF, Uncountable many Il--factors, Ann. Math. 90 (1969) 372-377_ [1049] D. McDuFF, On the structure of Il,-factors, Russian Math. Surv. 2S (1970) 29-50. [1050] D. McDuFF, Central sequences and the hyperfinite factor, Proc. London Math. Soc. 21 (1970) 443-461. [1051] D. McDuFF, On residual sequences in a Il,-factor, J. London Math. Soc. 3 (1971) 273-280. [1052J K. McKENNON, The strict topology and the Cauchy structure of the spectrum of a C*-algebra, Gen. Topology Appl. S (1975) 249-262. [1053] S. A. McKILLlGAN, Duality in B*-algebras, Proc. Amer. Math. Soc. 38 (1973) 86-88. [1054] E. J. MCSHANE, Families of measures and representation of algebras of operators, Trans. Amer. Math. Soc. 102 (1962) 328-345. [1055] C. R. MIERS, Lie isomorphisms of factors, Trans. Amer. Math. Soc. 147 (1970) 55-63. [1056] C. R. MIERS, Lie homomorphisms of operator algebras, Pacific J. Math. 38 (1971) 717-735. [1057] C. R. MIERS, Derived ring isomorphisms of Von Neumann algebras, Can. J. Math. 2S (1973) 1254-1268. [1058J C. R. MIERS, Lie derivations of Von Neumann algebras, Duke Math. J. 40 (1973) 403-409. [1059] C. R. MIERS, Polynomially ideal C*-algebras. Amer. J. Math. 98 (1976) 165-170. [1060] P. E. MILES, Order isomorphisms of B*-algebras. Trans. Amer. Math. Soc. 107 (1963) 217-236. [1061] P. MILES, B*-algebras. Pacific J. Math. 14 (1964) 627-637. [1062] P. MILES, Derivations on B*-algebras, Pacific J. Math. 14 (1964) 1359-1366. [1063] D. MILICIC, A note on invariant orderideals in the predual ofa Von Neumann algebra, Glasnik Mat. 6 (1971) 71-72. [1064] D. MILICIC, On C*-algebras with bounded trace. Glasnik Mat. 8 (1973) 7-22. [1065] D. MILMAN, On the theory of rings with involution. Dokl. Akad. Nauk SSSR 76 (1951) 349-352.

452

REFERENCES

[1066] S. MIRACLE-SOLE, Traitement de la convolution gauche pour les systemes infinis, Ann. Inst. H. Poincare (A) 6 (1%7) 59-71. (1067) S. MIRACLE-SOLE and D. W. ROBINSON, The physical states of Fermi systems, Commun. Math. Phys. 14 (1%9) 235-270. [1068) S. MIRACLE-SOLE and D. W. ROBINSON, Statistical mechanics of quantum mechanical particles with hard cores, II: The equilibrium states, Commun. Math. Phys. 19 (1970) 20+-218. (1069) Y. MISONOU, On a weakly central operator algebra, Tohoku Math. J. 4 (1952) 194-202. [1070] Y. MISONOU, Operator algebras of type I, Kodai Math. Sem. Rept. 5 (1953) 87-90. [1071) Y. MISONOU, Unitary equivalence of factors of type III, Proc. Japan. Acad. 29 (1953) 482-485. (1072) Y. MISONOU, On the direct product of W*algebras, Tohoku Math. J. 6 (1954) 189-204. [1073) Y. MISONOU, Generali~ed approximately finite operator algebras, Ttihoku Math. J. 7 (1955) 192-205. [1074) Y. MISONOU, On divisors of factors, Tohoku Math. J. 8 (1956) 63-69. (1075) Y. MISONOU and M. NAKAMURA, Centering of an operator algebra, Tohoku Math. J. 3 (1951) 243-248. [1076] H. MIYATA, On clustering states, Commun. Math. Phys. 34 (1973) 1-6. (1077) GH. MOCANU, The joint approximate spectrum of a finite system of elements of a C*-algebra, Studia Math. 49 (1973) 252-262. (1078) J. MOFFAT, Continuity of automorphic representations, Proc. Cambridge Phi/os. Soc. 74 (1973) 461-465. . [1079) J. MOFFAT, On groups of automorphisms of the tensor product of Von Neumann algebras, Math. Scand. 34 (1974) 226-230. [1080) J. MOFFAT, Connected topological groups acting on Von Neumann algebras, J. London Math. Soc. 9 (1975) 4\1-417. (1081) M. S. MONTGOMERY, Left and right inverses in group algebras, Bull. Amer. Math. Soc. 75 (1%9) 539-540. [1082) C. C. MOORE, Invariant measures on product spaces, in: Proc. 5" Berkeley Symp. on Mathematical Statistics and Probability, Vol. 2, part 2 (1%7) 447-459. (1083) R. T. MOORE, Hermitian functionals on B-algebras and duality characterizationsof C*-algebras, Trans. Amer. Math. Soc. 162 (1971) 253-265. [1084) R. D. MOYER, Computation of symbols on C*-algebras of singular integral operators, Bull. Amer. Math. Soc. 77 (1971) 615-620. [1085) P. MUHL Y, A structure theory for isometric representations of a class of semi groups, to appear. (1086) S. MURAKAMI, Remarks on the structure of maximally almost periodic groups, Osaka Math. J. 2 (1950) 119-129. [1087) G. P. MURAVERA, Maximal abelian subrings in an approximately finite factor, Sibirsk. Mat. i. 9 (1%8) 614-622 (in Russian). [1088) G. P. MURAVERA, Finite transformations of X-generators in a maximal abelian subalgebra of an approximately finite factor, Izv. Vyss. Ucebn. Zaved. Mat. 10 (1970) 61-68 (in Russian). [1089) G. P. MURAVERA, On regular abelian subalgebras in an approximately finite factor, Sibirsk. Mat. z. 13 (1972) 805-816 (in Russian).

REFERENCES

453

[1090] I. S. MURPHY, A note on B*-algebras, Glasgow Math. J. 14 (1973) 185-186. [1091] F. J. MURRAY, Bilinear transformations in Hilbert space, Trans. Amer. Math. Soc. 4S (1939) 474-507. [1092] F. J. MURRAY and J. VON NEUMANN, On rings of operators, Ann. Math. 37 (1936) 116-229. [1093] F. J. MURRAY and J. VON NEUMANN, On rings of operators, II, Trans. Amer. Math. Soc. 41 (1937) 208-248. [1094] F. J. MURRAY and J. VON NEUMANN, On rings of operators, IV, Ann. Math. 44 (1943) 716-808. [1095] L. N ACHBIN, On the finite dimensionality of every irreducible unitary represen tation of a compact group, Proc. Amer. Math. Soc. 12 (1961) 11-12. [1096] B. NAGEL, Some results in non-commutative ergodic theory, Commun. Math. Phys. 26 (1972) 247-258. [1097] M. A. NAIMARK, Rings with involution, Uspehi Mat. Nauk 3 (1948) 52-142 (in- Russian). [1098] M. A. NAIMARK, Rings of operators in Hilbert spaces, Uspehi Mat. Nauk 4 (1949) 83-147 (in Russian). [1099] M. A. NAIMARK, On a problem in the theory of rings with involution, Uspehi Mat. Nauk 6 (1951) 160-164 (in Russian). [1100] M. A. NAIMARK, On continuous analogue of Schurs lemma. Dokl. Akad. Nauk SSSR 98 (1954) 185-188 (in Russian). [1101] M. A. NAIMARK, Normed Rings (Moscow, 1956) (in Russian). [1102] M. A. NAIMARK, Factorial representations of a locally compact group, Dokl. Akad. Nauk SSSR 134 (1960) 275-277 (in Russian). [1103] M. A. NAiMARK, The decomposition in factorial representations of unitary re presentations of locally compact groups, Sibirsk: Mat. t. 2 (1961) 89-99 (in Russian). [1104] M. A. NAiMARK, The structure of factorial representations of a locally compact group, Dokl. Akad. Nauk SSSR 148 (1963) 775-778 (in Russian). [1105] M. A. NAiMARK and S. V. FOMIN, Continuous direct sums of Hilbert spaces and some applications, Uspehi Mat. Nauk 10 (1955) 111-142 (in Russian). [1106] Y. NAKAGAMI, Infinite tensor products of Von Neumann algebras, I, Kodai Math. Sem. Rept. 22 (1970) 341-354. [1107] Y. NAKAGAMI, Infinite tensor products of Von Neumann algebras, II, Publ. Res. Inst. Math. Sci. Kyoto Univ. 6 (1970) 257-292. [1108] Y. NAKAGAMI, Infinite tensor products of operators, Publ. Res. Inst. Math. Sci. Kyoto Univ. 10 (1974) 111-145. [1109] Y. NAKAGAMI, Duality for crossed products of Von Neumann algebras by locally compact groups, Bull. Amer. Math. Soc. 81 (1975) 1106-1108. [1110] M. NAKAI, Some expectations in AW*-algebras, Proc. Japan Acad. 34 (1958) 411-416. [1111] R. NAKAMOTO and M. NAKAMURA, On theorems of Korovkin, II, Prof. Japan Acad. 41 (1965) 433-435. [1112] M. NAKAMURA, The two-sided representations of an operator algebra, Proc. Japan Acad. 27 (1951) 172-176. [1113] M. NAKAMURA, On the direct product of finite factors, Tohoku Math. J. 6 (1954) 205-207.

454

REFERENCES

[1114] M. NAKAMURA, On operators of Schaefer class in the theory of singular integral equations, Proc. Japan Acad. 33 (1957) 455-456. [1115] M. NAKAMURA, A proof of a theorem of Takesaki, Kodai Math. Sem. Rept. 10 (1958) 189-190. [1116] M. NAKAMURA, Y. OTOMI and Y. URATA, A remark on the carrierprojections in abelian operator algebras, Mem Osaka Gakugei Univ. 13 (1964) 25-26. [1117] M. NAKAMURA and H. TAKAI, Non-degeneracy and discrete models, Proc. Japan Acad. 48 (1972) 566--508. [1118] M. NAKAMURA and H. TAKAI, A note on the dilation theorems, II, Proc. Japan Acad. 48 (1972) 302-307. [1119] M. NAKAMURA and Z. TAKEDA, The Radon-Nikodym theorem of traces for a certain operator algebra, Tohoku Math. J. 4 (1952) 275-283. [1120] M. NAKAMURA and Z. TAKEDA, Normal states of commutative operator algebras, Tohoku Math. J. 5 (1953) 109-121. [1121] M. NAKAMURA and Z. TAKEDA, On some elementary properties of the crossed products of Von Neumann algebras, Proc. Japan Acad. 34 (1958) 489-494. [1122] M. NAKAMURA and Z. TAKEDA, On certain examples of the crossed product of finite factors, r, II, Proc. Japan Acad. 34 (1958) 495-499, 500-502. [1123] M. NAKAMURA and Z. TAKEDA, On the extensions of finite factors, I, Proc. Japan Acad. 3S (1959) 149-154. [1124] M. NAKAMURA and Z. TAKEDA, A Galois theory for finite factors, Proc. Japan Acad. 36 (1960) 258--260. [1125] M. NAKAMURA and Z. TAKEDA, On the fundamental theorem of the Galois theory for finite factors, Proc. Japan Acad. 36 (1960) 313-318. [1126] M. NAKAMURA and Z. TAKEDA, On inner automorphisms of certain finite factors, Proc. Japan Acad. 37 (1961) 31-32. [1127] M. NAKAMURA and Z. TAKEDA, On outer automorphisms of certain finite factors, Proc. Japan Acad. 37 (1961) 215-216. [1128] M. NAKAMURA, M. TAKESAKI and H. UMEGAKI, A remark on the expectation of operator algebras, Kodai Math. Sem. Rept, 12 (1960) 82-90. [1129] M. NAKAMURA and T. TURUMARU, Simple algebras of completely continuous operators, Tohoku Math. J. 4 (1952) 303-308. [1130] M. NAKAMURA and T. TURUMARU, On extensions of pure states of an abelian operator algebra, Tohoku Math. J. 6 (1954) 253--257. [1131] M. NAKAMURA and T. TURUMARU, Expectations in an operator algebra, Tohoku Math. J. 6 (1954) 182-188. [1132] M. NAKAMURA and H. UMEGAKI, On a proposition of Von Neumann, Kodai Math. Sem. Rept. 8 (1956) 142-144. [1133] M. NAKAMURA and H. UMEGAKI, A note on the entropy for operator algebras, Proc. Japan Acad. 37 (1961) 149-154. [1134] M. NAKAMURA and H. UMEGAKI, On the Blackwell theorem in operator algebras, Proc. Japan Acad. 37 (1961) 312-315. [1135] M. NAKAMURA and H. UMEGAKI, On Von Neumanns theory of measurements in quantum statistics, Math. Japon. 7 (1962) 151-157. [1136] H. NAKANO, Hilbert algebras, Tohoku Math. J. 2 (1950) 4-23. [1137] K. NAPIORKOWSKI, On the independence of local algebras, Rept. Math. Phys. 3 (1972) 33-35.

REFERENCES

455

[1138) H. NARNHOFER, Self-adjoint operators, derivations and automorphisms on C algebras, J. Math. Phys. 16 (1975) 2192-2194. [1139) J. NAUDTS, Observables at infinity, Ann. Soc. Sci. Bruxelles (I) 84 (1970) 257-266. [1140) J. NAUDTs, A generalized entropy function, Commun. Math. Phys. 37 (1974) 175-182. [1141] J. NAVDTS and A. VERBEURRE, Number operator states on the CAR algebra, Ann. Soc. Sci. Bruxelles (I) 86 (1972) 101-108. [1142) J. NAuDTs, A. VERBEURRE and R. WEDER, Linear response theory and the KMS condition, Commun. Math. Phys. 44 (1975) 87-99. [1143) E. NELSON, Notes on non-eommutative integration, J. Functional Anal. 15 (1974) 103-116. [1144) E. NELSON and W. F. STINESPRING, Representation of elliptic operators in the enveloping algebra, Amer. J. Math. 81 (1959) 547-560. [1145) R. NEST, Invariant weights of operator algebras satisfying the KMS condition, Kumi (1973, no. 10). [1146) S. M. NEWBERGER, R-algebras and locally convex spaces, Math. Ann. 176 (1968) 145-156. [1147) O. A. NIELSEN, Maximal abelian subalgebras in hyperfinite factors, Bull. Amer. Math. Soc. 75 (1969) 579-581. [1148) O. A. NIELSEN, Maximal abelian subalgebras in hyperfinite factors, Trans. Amer. Math. Soc. 146 (1969) 259-272. [1149) O. A. NIELSEN, Maximal abelian subalgebras of hyperfinite factors, II, J. Func tional Anal. 6 (1970) 192-202. [1150) O. A. NIELSEN, The asymptotic ratio set and direct integral decompositions of a Von Neumann algebra, Can. J. Math. 23 (1971) 598-607. [1151] O. A. NIELSEN, An example of a Von Neumann algebra of global type II, J. Functional Anal. 11 (1972) 207-210. [1152] O. A. NIELSEN, Borel sets of Von Neumann algebras, Amer. J. Math. 95 (1973) 145-164. [1153) O. A. NIELSEN, New proof of Sakais theorems on global Von Neumann algebras,

to appear.
[1154] O. A. NIELSEN, Reduced crossed products of Von Neumann algebras, I, to appear. [1155) F. NIIRO, Sur lunicite de la decomposition dune trace, Sci. Papers College Gen. Ed. Univ. Tokyo 13 (1963) 159-162. [1156) V. N. Nosov, Strongly closed ideals in operator -a1gebras, Vestnik Moskov. Univ. 4 (1970) 34-37 (in Russian). [1157) V. N. Nosov, Extensions of C-algebras with continuous trace, Funkcion. Anal. Priloi. 6 (1972) 91-92. [1158) A. E. NUSSBAUM, On the reduction of C-algebras, Bull. Amer. Math. Soc. 15 (1964) 567-573. [1159] T. OGASAWARA, Finite dimensionality of certain Banach algebras, J. Sci. Hiroshima Univ. 17 (1954) 359-364. [1160) T. OoASAWARA, A theorem on operator algebras, J. Sci. Hiroshima Univ. 18 (1955) 307-309. [1161) T. OoASAWARA, A structure theorem for complete quasi-unitary algebras, J. Sci. Hiroshima Univ. 19 (1955) 79-85.

456

REFERENCES

[1162] T. OOASAWARA, Topologies on rings of operators, J. Sci. Hiroshima Univ. 19 (1955) 255-272. [1163] T. OOASAWARA and S. MAEDA, A generalization of a theorem of Dy, J. Sci. Hiroshima Univ. 20 (1956) 1-4. [1164] T. OOASAWARA and K. YOSHINAOA, Weakly completely continuous Banach *-algebras, J. Sci. Hiroshima Univ. 18 (1954) 15-36. [1165] T. OOASAWARA and K. YOSHINAOA, A characterization of dual B*-algebras, J. Sci. Hiroshima Univ. 18 (1954) 179-182. [1166] T. OOASAWARA and K. YOSHINAOA, A non-commutative theory of integration for operators, J. Sci. Hiroshima Univ. 18 (1955) 311-347. [1167] T. OOASAWARA and K. YOSHINAOA, Extension of application to unbounded operators, J. Sci. Hiroshima Univ. 19 (1955) 273-299. [1168] Y. OKA, Uniformly hyperfinite algebras and locally compact transformation groups, J. Math. Soc. Japan 25 (1973) 357-362. [1169] T. OKAYASU, On the tensor product of C*-algebras, Tohoku Math. J. 18 (1966) 325-331. [1170] T. OKAYASU, A structure theorem of automorphisms of Von Neumann algebras, Tohoku Math. J. 20 (1%8) 199-206. [1171] T. OKAYASU, On GCR operators, Tohoku Math. J. 21 (1%9) 573-579. [1172] T. OKAY ASU, On representations of tensor products in involutive Banach algebras, Proc. Japan Acad. 46 (1970) 404-408. [1l73] T. OKAYASU, Some cross norms which are not uniformly cross, Proc. Japan Acad. 46 (1970) 54-57. [1l74] T. OKAYASU, Polar decomposition for isomorphisms of C*-algebras, Tohoku. Math. J. 26 (1974) 541-554. [1l75] T. OKAYASU and M. TAKESAKI, Dual spaces of tensor products of C*-algebras, Tohoku Math. J. 18 (1%6) 332-337. [1l76] D. OLESEN, On norm-continuity and compactness of spectrum, Math. Scand. 35 (1974) 223-236. [1177] D. OLESEN, Derivations of A W*-algebras are inner, Pacific J. Math. 53 (1974) 555-561. [1178] D. OLESEN, Inner *-automorphisms of simple C*-algebras, Commun. Math. Phys. 44 (1975) 175-190. [1179] D. OLESEN, On spectral subspaces and their applications to automorphism groups, in: Conv. su C*-algebre e Appl. Fis. Teor., to appear. [1180) D. OLESEN, A generalization of a theorem by Bochner. Proc. Amer. Math. Soc. 57 (1976) 115-118. [1181] D. OLESEN and G. K. PEDERSEN. Derivations of C*-algebras have semi-continuous generators, Pacific J. Math. 53 (1974) 563-572. [1182] D. OLESEN and G. K. PEDERSEN, Groups of automorphisms with spectrum condition and the lifting problem, Commun. Math. Phys. 51 (1976) 85-95. [1183) C. L. OLSEN, Thin operators in a Von Neumann algebra, Acta Sci. Math. 35 (1974) 211-216; Correction: Ibid. 36 (1974) 377-378. [1l84] C. L. OLSEN, A characterization of thin operators in a Von Neumann algebra, Proc. Amer. Math. Soc. 39 (1973) 571-578. [1185] M. P. OLSON, The self-adjoint operators of a Von Neumann algebra form a conditionally complete lattice, Proc. Amer. Math. Soc. 28 (1971) 537-544. [1186] T. ONO, Local theory of rings of operators, I, II, J. Math. Soc. Japan 10 (1958) 184-216, 438-458.

REFERENCES

457

[1187) T. DND, Note on a B*-algebra, J. Math. Soc. Japan 11 (1959) 146--158. [1188] T. DND, Local theory in function analysis, J. Math. Soc. Japan 15 (1963) 9-30. [1189] T. DND, Deformation of the set of projections, Bull. Nagoya Inst. Technol. 20 (1968) 131-135 (in Japanese). [1190] T. DND, A real analogue of the Gelfand-Naimark theorem, Proc. Amer. Math. Soc. 2S (1970) 159-160. [1191) T. DND, In a finite AW* algebra, x*x .,;xx* implies x*x = xx*, Bull. Nagoya Inst. Technol. 24 (1972) 61-66 (in Japanese). [1192] T. DND, Hypernormalcy in finite AW*-algebras, to appear. [1193] T. DNa, Trace on finite AW*-algebra, to appear. [1194] M. DRIHARA, Sur les anneaux des operateurs, I, Proc. Imper. Acad. Tokyo 20 (1944) 399-405. [1195] M. DRIHARA, Sur les anneaux des operateurs, II, Proc. Imper. Acad. Tokyo 20 (1944) 545-553. [1196) M. DRIHARA, Rings of operatorsand their traces, Mem. Fac. Sci. Kyushu Univ. (A) 5 (1950) 107-138; Correction: Ibid. 8 (1953) 89-91. [1197] K. DSTERWALDER, Duality for free Bose fields, Commun. Math. Phys. 29 (1973) 1-14. [1198] V. 1. DVCINNIKOV, Symmetric spaces of measurable operators, Dokl. Akad. Nauk SSSR 191 (1970) 769-771 (in Russian; English trans!.: Soviet Math. 11 (1970) 448-452. [1199] V. 1. DVCINNIKDV, s-numbers of measurable operators, Funkcion. Anal. Priloi. 4 (1970) 78-85 (in Russian). [1200] V. 1. DVCINNIKOV, Compact operators relative to a Von Neumann algebra, Funkcion. Anal. Priloi. 6 (1972) 37-40 (in Russian). [1201] A. R. PADMANABHAN, Some dominated convergence theorems in a Von Neumann algebra, Proc. Japan Acad. 42 (1966) 347-350. [1202] A. R. PADMANABHAN, Stability and mixing in Von Neumann algebras, Kodai Math. Sem. Rep/. 18 (1966) 335-342. [1203] A. R. PADMANABHAN, Convergence in measure and related results in finite rings of operators, Trans. Amer. Math. Soc. 128 (1967) 359-378, 132 (1967) 563. [1204) W. PAGE, Characterizations of B*- and A*-algebras, Rev. Roumaine Math. Pures Appl. 18 (1973) 1241-1244. [1205) PAK-KEN WONG, Continuous complementors on B*-algebras, Pacific J. Math. 33 (1970) 255-260. [1206] PAK-KEN WONG, The Arens product and duality in B*-algebras. II, Proc. A mer. Math. Soc. 27 (1971) 535-538. [1207] R. PALLU DE LA BARRIERE, Algebres auto-adjointes faiblement ferrnees et algebres hilbertiennes de classe finie, C.R. Acad. Sci. Paris. 232 (1951) 1994-1995. [1208] R. PALLU DE LA BARRIERE, Decomposition des operateurs non bornes dans les sommes continues despaces de Hilbert, CiR. Acad. Sci. Paris 232 (1951) 2071 2073. [1209) R. PALLU DE LA BARRIERE, Algebres unitaires et espaces de Ambrose, c.R. Acad. Sci. Paris 233 (1951) 997-999. [1210] R. PALLU DE LA BARRIERE, Isomorphismes des *-algebres faiblement ferrnees doperateurs, C.R. Acad. Sci. Paris 232 (1951) 2071-2073. [1211] R. PALLU DE LA BARRIERE, Algebres unitaires et espaces dAmbrose, Ann. Ecole Norm. Sup. 70 (1953) 381-401.

458

REFERENCES

[1212] R. PALLU DE LA BARRIERE, Sur les algebres doperateurs dans les espaces hilbertiens, Bull. Soc. Math. France 82 (1954) 1-517. [1213] T. W. PALMER, Characterizationsof C*-algebras, Bull. Amer. Math. Soc. 74 (1968) 538-540. [1214] T. W. PALMER, Characterizations of C*-algebras, II, Trans. Amer Math. Soc. 148 (1970) ~77-588. [1215] T. W. PALMER, Real C*-algebras, Pacific J. Math. 35 (1970) 195-204. [1216] T. W. PALMER, The Gelfand-Naimark pseudo-norm on Banach *-algebras, J. London Math. Soc. 3 (l97I) 59-66. [1217] T. W. PALMER, Arens multiplication and a characterization of W*algebras, Proc. Amer. Math. Soc. 44 (1974) 81-87. [1218] T. W. PALMER, Jordan *-homomorphisms between reduced Banach *-algebras, Pacific J. Math. 58 (1975) 169-178. [1219] W. L. PASCHKE, Inner product modules over B*-algebras, Trans. Amer. Math. Soc. 182 (1973) 443-468. [1220] W. L. PASCHKE, The double B-dual of an inner product module over a C*-algebra, Can. J. Math. 26 (1974) 1272-1280. 0221] A. L. T. PATERSON, Isometries between B*-algebras, Proc. Amer. Math. Soc. 22 (1969) 570-572. [1222] A. L. T. PATERSON and A. M. SINCLAIR, Characterization of isometries between C*-algebras, J. London Math. Soc. 5 (1972) 755-761. [1223] C. PEARCY, A complete set of unitary invariants for operators generating finite W*-algebras of type I, Pacific J. Math. 12 (1962) 1405-1416. [1224] C. PEARCY, W*-algebras with a single generator, Proc. Amer. Math. Soc. 13 (1962) 831-832. [1225] C. PEARCY, On unitary equivalence of matrices over the ring of continuous complex-valued functions on a Stonian spaces, Can. J. Math. 15 (1963) 323-331. [1226] C. PEARCY, On certain Von Neumann algebras which are generated by partial isometries, Proc. Amer. Math. Soc. 15 (1964) 393-395. [1227] C. PEARCY, Entire functions on infinite Von Neumann algebras of type I, Michigan Math. J. 11 (1964) 1-7. [1228] C. PEARCY, On commutators of operators on Hilbert space, Proc. Amer. Math. Soc. 16 (1965) 53-59. [1229] C. PEARCY and J. R. RrNGROSE, Trace-preserving isomorphisms in finite operator algebras, Amer. J. Math. 90 (1968) 444-455. [1230] C. PEARCY and N. SALINAS, Finite-dimensional representations of separable C*-algebras, Bull. Amer. Math. Soc. 80 (1974) 970-972. [1231] C. PEARCY and N. SALINAS, Finite-dimensional representations of C*-algebras and the reducing matricial spectra of an operator, Rev. Roumaine Math. Pures Appl. 20 (1975) 567-598. [1232] C. PEARCY and N. SALINAS, The reducing essential spectra of an operator, Duke Math. J. 42 (1975) 423-434. [1233] C. PEARCY and D. M. TOPPING, Sums of small numbers of idempotents, Michigan Math. J. 14 (1967) 45~5. [1234] C. PEARCY and D. TOPPING, Commutators and certain III-factors, J. Functional Anal. 3 (1969) 69-78.

REFERENCES

459

[1235] C. PEARCY and D. TOPPING, On commutators in ideals of compact operators, Michigan Math. J. 18 (1971) 247-252. [1236] E. A. PEDERSEN, A decomposition theory for rings of operators,Thesis, Louisiana State Univ., Baton Rouge, La. (1966). [1237] G. K. PEDERSEN, Measure theory for C*-algebras, Math. Scand. 19 (1966) 131-145. [1238] G. K. PEDERSEN, Measure theory for C*-algebras,II, Math. Scand. 22 (1968) 6~74. [1239] G. K. PEDERSEN, A decomposition theorem for C*-algebras, Math. Scand. 22 (1968) 266-268. [1240] G. K. PEDERSEN, On weak and monotone e-closures of C*-algebras, Commun. Math. Phys. 11 (1969) 221-226. [1241] G. K. PEDERSEN, Measure theory for C*-algebras, III, IV, Math. Scand. 25 (1969) 71-93, 121-127. [1242] G. K. PEDERSEN, The "up-down" problem for operator algebras, Proc. Natl. Acad. Sci. U.S.A. 68 (1971) 18%-1897. [1243] G. K. PEDERSEN, Atomic and diffuse functionals on a C*-algebra, Pacific J. Math. 37 (1971) 795-800. [1244] G. K. PEDERSEN, C*-integrals, Thesis, Univ. of Copenhagen, (1971). [1245] G. K. PEDERSEN, Monotone closures in operator algebras, Amer. J. Math. 94 (1972) 955-962. [1246] G. K. PEDERSEN, Operator algebras with weakly closed abelian subalgebras, Bull. London Math. Soc. 4 (1972) 171-175. [1247] G. K. PEDERSEN, Some operator monotone functions, Proc. Amer. Math. Soc. 36 (1972) 309-310. [1248] G. K. PEDERSEN, Applications of weak* semi-continuity in C*-algebra theory, Duke Math. J. 39 (1972) 431-450. [1249] G. K. PEDERSEN, Borel structure in operator algebras, Mat.-Fys. Medd. Danske Vid. Selsk. 39 (5) (1974); abbreviated translation of: Math.-Phys. Commun. Roy.

Danish Acad. Sci. Letters.

[1250] G. K. PEDERSEN, A non-commutative version of Souslins theorem, Bull. London Math. Soc. 8 (1976) 87-90. [1251] G. K. PEDERSEN, The trace in semi-finite von Neumann algebras, Math. Scand.37 (1975) 142-192. [1252] G. K. PEDERSEN, Lifting groups of automorphisms, to appear. [1253] G. K. PEDERSEN, Lifting derivations from quotients of separable C*-algebras, to

appear.
[1254] G. K. PEDERSEN, Derivations of operator algebras, to appear. [1255] G. K. PEDERSEN, Similarities between UHF-algebras, to appear. [1256] G. K. PEDERSEN and N. H. PETERSEN, Ideals in a C*-algebra, Math. Scand. 27 (1970) 19~204. [1257] G. K. PEDERSEN and E. ST0RMER, Automorphisms and equivalence in Von Neumann algebras, II, Indiana Univ. Math. J. 23 (1973) 121-130. [1258] G. K. PEDERSEN and M. T AKESAKI, The Radon-Nikodym theorem for a Von Neumann algebra, Acta Math. 130 (1973) 5~. [1259] J. PEETRE and G. SPARR, Interpolation and non-commutative integration, Ann. di Mat. 104 (1975) 187-207. [1260] C. PELlGRAD and L. ZSIDO, A Riesz decomposition theorem in W*-algebras, Acta Sci. Math. Szeged 34 (1973) 317-322.

460

REFERENCES

[1261] F. PERDRIZET, Sur les sous-algebres involutives abeliennes dun facteur fini, CR. Acad. Sci. Paris 268 (1969) 872-875. [1262] F. PERDRIZET, Espaces de Banach ordonnees et ideaux, CR. Acad. Sci. Paris 269 (1969) 393-396. [1263] F. PERDRIZET, Elements positifs relativement a une algebre hilbertienne a gauche, C.R. Acad. Sci. Paris 270 (1970) 322-325. [1264] F. PERDRIZET, Traces et topologies dans les C*algebres, CR. Acad. Sci. Paris 271 (1970) 427-430. [1265] F. PERDRIZET, Espaces de Banach ordonnes et ideaux, J. Math. Pures Appl. 49 (1970) 61-98. [1266] F. PERDRIZET, Elements positifs relativement une algebre hilbertienne gauche, Compositio Math. 23 (197!) 25-47. [1267] F. PERDRIZET, Topologie et traces sur les C*-algebres, Bull. Soc. Math. France 99 (197!) 193-239. [1268] N. H. PETERSON, Invariant weights on semi-finite Von Neumann algebras, Math. Scand. 32 (1973) 133-144. [1269] J. PHILLIPS, Positive integrable elements relative to a left Hilbert algebra, J. Functional Anal. 43 (1973) 390-409. [1270] J. PHILLIPS, Perturbations of type I Von Neumann algebras, Pacific J. Math. 52 (1974) 505-511. [1271] J. PHILLIPS, Complementation for right ideals in generalized Hilbert algebras, Trans. Amer. Math. Soc. 197 (1974) 409-417. [1272] J. PHILLIPS, Perturbations of C*-algebras, Indiana Univ. Math. J. 23 (1974) 1167-1176. [1273] J. PHILLIPS, A note on square-integrable representations, J. Functional Anal. 20 (1975) 83-92. [1274] J. PHILLIPS. Autorncrphisms of fulllI, factors. with applications to factors of type III, Duke Math. J. 43 (1976) 375-386. [1275] R. J. PLYMEN, A modification of Pironsaxioms, Helv. Phys. Acta 41 (1968) 69-74. [1276] R. J. PLYMEN, C*-algebras and Mackeys axioms, Commun. Math. Phys. 8 (1968) 132-146. [1277] P. PORCELLI and E. A PEDERSEN, On rings of operators, Bull. Amer. Math. Soc. 73 (1967) 142-144. [1278] H. PORTA, A note on homomorphisms of operator algebras, Colloq. Math. 20 (1969) 117-119. [1279] H. PORTA, Ideals and universal representations of certain C*-algebras, Rev. Un. Mat. Argentina 25 (1970) 27-36. [1280] H. PORTA and J. T. SCHWARTZ, Representations of the algebra of all operators in Hilbert space, and related analytic function algebras, Commun. Pure Appl. Math. 20 (1967) 457-492. [1281] R. T. POWERS, Representations of uniformly hyperfinite algebras and their as sociated Von Neumann rings, Bull. Amer. Math. Soc. 73 (1967) 572-575. [1282] R. T. POWERS, Representations of uniformly hyperfinite algebras and their as sociated Von Neumann rings, Ann. Math. 86 (1967) 138-171. [1283] R. T. POWERS, UHF algebras and their applications to representations of the anticommutation relations, in: Cargese Lectures in Phys. 4 (1970) 137-168. [1284] R. T. POWERS, Fermi field agebra, in: Cargese Lectures in Phys. 4 (1970) 363-367.

REFERENCES

461

[1285] R. T. POWERS. Self-adjoint algebras of unbounded operators. Commun. Math. Phys. 21 (1971) 85-124. [1286] R. T. POWERS, Simplicity of the C*-algebra associated with the free group on two generators, Duke Math. J. 42 (1975) 151-156. [1287] R. T. POWERS, A remark on the domain of an unbounded derivation of a C*-a1gebra, J. Functional Anal. 18 (1975) 85-95. [1288] R. T. POWERS. Representations of the canonical anticommutation relations, to

appear.
[1289] R. T. POWERS and S. SAKAl, Existence of ground states and KMS states for approximately inner dynamics, Commun. Math. Phys. 39 (1975) 273-288. [1290] R. T. POWERS and S. SAKAI, Unbounded derivations in operator algebras, J. Functional Anal. 19 (1975) 81-95. [1291] R. T. POWERS and E. ST0RMER, Free states of the canonical anticommutation relations, Commun. Math. Phvs. 16 (1970) 1-33. [1292] E. PRESUTTI and E. SCACCIATELLI, Local bounded perturbations of KMS states, J. Math. Phys. 15 (1974) 1620-1625. [1293] E. PRESUTTI. E. SCACCIATELLI, G. L. SEWELL and F. VANDERLINGH. Studies

in the C*-algebraic theory of non equilibrium statistical mechanics: Dynamics of open and of mechanically driven systems, J. Math. Phys. 13 (1972) 1085-1098. [1294] D. PROMISLOW, The Kakutani theorem for tensor product of W*-algebras. Pacific J. Math. 36 (1971) 507-514. [1295] D. PROMISLOW, Serni-metrics on the normal states of a W*-algebra. Can. J. Math. 25 (1973) 1238-1253. [1296] R. T. PROSSER. On the ideal structure of operator algebras. Mem. Amer. Math. Soc. 45 (1963). [1297] V. PTAK, Banach algebras with involution, Manuscripta Math. 6 (1972) 245-290. [1298] V. PTAK. A theorem of the closed graph type, Manuscripta Math. 13 (1974) 109-130, [1299] V. PTAK, Banach algebras with involution, in: Proc. 3d Prague Topological Symp. (1971) 363-369. [1300] L. PUKANSZKY. On a theorem of Mautner, Acta. Sci. Math. Szeged 15 (1954) 145-148. [1301] L. PUKANSZKY, The theorem of Radon-Nikodym in operator rings, Acta. Sci. Math. Szeged 15 (1954) 149-156. [1302] L. PUKANSZKY. On the theory of quasi-unitary algebras. Acta Sci. Math. Szeged 16 (1955) 103-121. [1303] L. PUKANSZKY, Some examples of factors, Publ. Math. 4 (1956) 135-156. [1304] L. PUKANSZKY, On maximal abelian subrings of factors of type II" Can. J. Math. 12 (1960) 289-296. [1305] L. PUKANSZKY, On maximal abelian subrings in factors, R.I.A.S. Techn. Rept. 60-11 (1960) 25 pp. [1306] M. PULVIRENTI and B. TrROZZI, Time evolution of a quantum lattice system, Commun. Math.Phys. 30 (1973) 83-98. [1307] C. RADIN, Automorphisms of Von Neumann algebras as point transformations, Proc. Amer. Math. Soc. 39 (1973) 343-346. [1308] H. RADJAVI. Nonselfadjoint representations of C*-algebras. Proc. Amer. Math. Soc. 47 (1975) 133-136.

462

REFERENCES

[1309] D. RAIKOV, On some kinds of convergence of functions of positive type, Dok/. Akad. Nauk SSSR 58 (1947) 1279-1282 (in Russian). [1310] D. RAINJONNEAU, Existence des sommes dans certaines categories dalgebres, C.R. Acad. Sci. Paris 262 (1966) 283-285. [1311] M. C. REED, Torus invariance for the Clifford algebra. I, Trans. Amer. Math. Soc. 154 (1971) 177-183. [1312) M. C. REED, Torus invariance for the Clifford algebra, II, J. Functional Anal. 8 (1971) 450-468. [1313] H. REEH and S. SCHLIEDER, Bemerkungen zur Unitarequivalenz von Lorentzinvarianten Feldern, Nuovo Cimento 22 (1%1) 1051-1068. [1314] G. A. REID, A generalisation of W*-algebras, Pacific J. Math. IS (1%5) 1019-1026. [1315] G. A. REID, Epimorphisms and surjectivity, Invent. Math. 9 (1970) 295-307. [1316] G. A. REID, On a question of Dixmier, J. London Math. Soc. 3 (1971) 544-548. [1317) G. A. REID, On the Calkin representations, Proc. London Math. Soc. 23 (1971) 547-564. [1318] B. B. RENSHAW, On the tensor product of W*-algebras, Trans. Amer. Math. Soc. 94 (1974) 337-347. 11319] C. RICKART, Banach algebras with an adjoint operation. Ann. Math. 47 (1946) 528-550. [1320] C. RICKART, The uniqueness of norm problem in Banach algebras, Ann. Math. 51 (1950) 615"{)28. [1321] C. RICKART, Representation of certain Banach algebras on Hilbert space, Duke Math. J. 18 (1951) 27-39. [1322] C. RlcKART, Spectral permanence for certain Banach algebras, Proc. Amer. Math. Soc. 4 (1953) 191-1%. [1323] C. RlcKART, General Theory of Banach Algebras (Van Nostrand, New York, 1%0). [1324) G. RIDEAU, On some representations of the anticommutation relations, Cammun. Math. Phys. 9 (1%8) 229-241. [1325] M. A. RIEFFEL, Square integrable representations of Hilbert algebras, J. Func tional Anal. 3 (1%9) 265-300. [1326] M. A. RIEFFEL, Induced representations of C*-algebras, Bull. Amer. Math. Soc. 78 (1972) 605-609. [1327] M. A. RIEFFEL, Morita equivalence for C*-algebras and W*-algebras, J. Pure Appl. Algebra 5 (1974) 51-%. [1328] M. A. RIEFFEL, Induced representations of C*-algebras, Advan. Math. 13 (1974) 176-257. [1329) M. A. RIEFFEL, A commutation theorem and duality for free Bose fields, Commun. Math. Phys. 39 (1974) 153-164. [1330] M. A. RIEFFEL, Unitary representations of group extensions; an algebraic ap proach to the theory of Mackey and Blattner, to appear. [1331] M. A. RIEFFEL and A. VAN DAELE, The commutation theorem for tensor products of Von Neumann algebras, Bull. London Math. Soc. 7 (1975) 257-260. [1332] M. A. RIEFFEL and A. VAN DAELE, A bounded operator approach to Tomita Takesaki theory, to appear. [1333] J. R. RINGROSE, On some algebras of operators, Proc. London Math. Soc. IS (1%5) 61-83.

REFERENCES

463

[1334] J. R. RlNGROSE, On subalgebras .of a C*-algebra, Pacific J. Math. 15 (1965) 1377-1382. [1335] J. R. RINGROSE, Lectures on the trace in a finite Von Neumann algebra, in: Lecture Notes in Math. 247 (Springer, Berlin, 1972) 309-354. [1336] J. R. RlNGROSE, Cohomology of operator algebras, in: Lecture Notes in Math. 247 (Springer, Berlin, 1972) 355-434. [1337] J. R. RlNGROSE, Automatic continuity of derivations of operatoralgebras, J. London Math. Soc. 5 (1972) 432-438. [1338] J. R. RlNGROSE, Linear functionals on operator algebras and their abelian subalge bras, J. London Math. Soc. 7 (1974) 553-560. [1339] J. R. RINGROSE, Operator algebras and their abelian subalgebras, Intern. Congr.

Vancouver (1974).
[1340] J. R. RlNGROSE, Linear mappings between operator algebras, to appear. [1341] J. E. ROBERTS, Some applications of dilatation invariance to structural question in the theory of local observables, Commun. Math. Phys. 37 (1974) 273-286. [1342] J. E. ROBERTS and G. ROEPSTROFF, Some Basic Concepts of Algebraic Quantum Theory (Desy, Hamburg, 1968). [1343] A. G. ROBERTSON, A note on the unit ball in C*-a1gebras, Bull. London Math. Soc. 6 (1974) 333-335. [1344] A. G. ROBERTSON, Standard states of C*a1gebras, J. London Math. Soc. 11 (1975) 423-433. [1345] D. W. ROBINSON. Statistical mechanics of quantum spin systems, I, Commun. Math. Phys. 6 (1967) 151-160. [1346] D. W. ROBINSON, Statistical mechanics of quantum spin systems, II, Commun. Math. Phys. 7 (1968) 337-348. [1347] D. W. ROBINSON, Normal and locally normal states, Commun. Math. Phys. 19 (1970) 219--234. [1348] D. W. ROBINSON, Return to equilibrium, Commun. Math. Phys. 31 (1973) 171-189. [1349) D. W. ROBINSON and D. RUELLE, Mean entropy of states in classical statistical mechanics, Commun. Math. Phys. 5 (1967) 288-300. [1350) D. W. ROBINSON and D. RUELLE, Extremal invariant states, Ann. Inst. H. Poincare. 6 (1967) 299--310. [1351) F. ROCCA, M. SIRUGUE and D. TESTARD, Les etats quasi-Iibres comme etats dequilibre so us les conditions de Kubo-Martin-Schwinger, c.R. Acad. Sci. Paris 267 (1968) 672-673. [1352] F. ROCCA, M. SIRUGUE and D. TESTARD. Etats quasi-libres invariants de translation et transformations de Bogoliubov, C.R. Acad. Sci. Paris 267 (1968) 722-723. [1353] F. ROCCA, M. SIRUGUE and D. TESTARD, On a class of equilibrium states under the Kubo-Martin-Schwinger boundary condition, I: Fermions, Commun. Math. Phys. 13 (1969) 317-334. (1354] F. ROCCA, M. SIRUGUE and D. TESTARD, Translation invariant quasi-free states and Bogoliubov transformations, Ann. Inst. H. Poincare (A) 10 (1969) 247-258. [1355] V. ROBLIN, Unitary rings, Dokl. Akad. Nauk SSSR 59 (1948) 643-646 (in Russian). [1356] J. E. Roos, Sur lanneau maximal de fractions des AW*-algebres et des anneaux de Baer, c.R. Acad. Sci. Paris 266 (1968) 120-123.

464

REFERENCES

[1357] N. Roos, Independence of local algebras in quantum field theory, Commun. Math. Phys. 16 (1970) 238-246. [1358] A. ROSENBERG, The number of irreducible representations of simple rings with no minimal ideals, Amer. J. Math. 75 (1953) 523-530. [1359] P. ROSENTHAL, Weakly closed maximal triangular algebras are hyperreducible, Proc. Amer. Math. Soc. 24 (1970) 220. [1360] R. ROUSSEAU, A. V AN DAELE and L. V ANHEESWIJCK, A note on the com mutation theorem for tensor products of Von Neumann algebras, to appear. [1361] D. RUELLE, States of physical systems, Commun. Math. Phys. 3 (1%6) 133-150. [1362] D. RUELLE, A variational formulation of equilibrium statistical mechanics and the Gibbs phase rule, Commun. Math. Phys. 5 (1%7) 324-329. [1363] D. RUELLE, Some remarks on the ground state of infinite systems in statistical mechanics, Commun. Math. Phys. 11 (1%9) 339-345. [1364] D. RUELLE, Statistical Mechanics (Benjamin, New York, 1969). [1365] D. RUELLE, Integral representation of states on a C*-algebra, J. Functional Anal. 6 (1970) 116-151. [1366] D. RUELLE, Symmetry breakdown in statistical mechanics, to appear. [1367] M. B. RUSKAI, Inequalities for traces in Von Neumann algebras, Commun. Math. Phys. 26 (1972) 280-289. [1368] B. Russo, Linear mappings of operator algebras, Proc. Amer. Math. Soc. 17 (1966) 1019-1022. [1369] B. Russo, Unimodular contractions in Hilbert space, Pacific J. Math. 26 (1968) (163-169. [1370] B. Russo, Isometries of U-spaces associated with finite Von Neumann algebras, Bull. Amer. Math. Soc. 74 (1968) 228-232. [1371] B. Russo, Isometries of the trace class, Proc. Amer. Math. Soc. 23 (1969) 213. [1372] B. Russo, Trace preserving mappings of matrix algebras, Duke Math. J. 36 (1969) 297-300. [1373] B. Russo and H. A. DYE. A note on unitary operators in C*-algebras, Duke Math. J. 33 (1966) 413-416. [1374] K. SAITO, Non-commutative extension of Lusins theorem, Tohoku Math. J. 19 (1967) 332-340. [1375] K. SAITO, On the algebra of measurable operators for a general A W*-algebra, T6hoku Math. J. 21 (1969) 249-270. [1376] K. SAITO, A non commutative integration theory for a semi-finite AW*-algebra and a problem of Feldmann, Proc. Japan Acad. 46 (1970) 463-467. [1377] K. SAITO, A non-commutative theory of integration for a semi-finite AW*-algebra and a problem of Feldman, T6hoku Math. J. 22 (1970) 420-461. [1378] K. SAITO, On the preduals of W*-algebras, Tohoku Math. J. 19 (1967) 324-331. [1379] K. SAITO, On the algebra of measurable operators for a general A W*-algebra, II, Tohoku Math. J. 23 (1971) 525-534. [1380] K. SAITO, On the embedding as a double commutator in type I AW*-algebra, Tiihoku Math. J. 23 (1971) 541-557. [1381] K. SAITO, On the embedding as a double commutator in a type I AW*-algebra, II, Tohoku Math. J. 26 (1974) 333-340. [1382] K. SAITO, Automorphism groups of Von Neumann algebras and ergodic type theorems, Acta Sci. Math. Szeged 36 (1974) 119-130.

REFERENCES

465

[1383] K. SAITO. On the algebra of measurable operators for a general AW*-algebra, III, to appear. [1384] K. SAITO, Groups of *-automorphisms and invariant maps of Von Neumann algebras, Pacific J. Math. 57 (1975) 553-558. [1385] T. SAITO, On incomplete infinite direct product of W*-algebras, Tohoku Math. J. 10 (1958) 165-171. [1386] T. SAITO, The direct and crossed product of rings of operators, Tohoku Math. J. 11 (1959) 299-304. [1387] T. SAITO, Some remarks on a representation of a group, Tohoku Math. J. 12 (1960) 383-388. [1388] T. SAITO, On representation of a countably infinite group, Tohoku Math. J. 13 (1961) 268-273. [1389] T. SAITO, On groups of automorphisms of finite factors, Tohoku Math. J. 13 (l96J) 427-433. [1390] T. SAITO, Some remarks on a representation of a group, II, Tohoku Math. J. 17 (1965) 206-209. [1391] T. SAITO, On generators of Von Neumann algebras, Michigan Math. J. 15 (1968) 373-376. [1392] T. SAITO, Generators of certain Von Neumann algebras, Tohoku Math. J. 20 (1968) 101-105. [1393] T. SAITO, Generators of Von Neumann algebras, in: Lecture Notes in Math. 247 (Springer, Berlin, 1972) 435-531. [1394] T. SAITO, Hyponormal operators and related topics. in: Lecture Notes in Math. 247 (Springer, Berlin, 1972) 533-664. [1395] T. SAITO and J. TOMIYAMA, Some results on the product of W*-algebras, Tohoku Math. J. 12 (1960) 455-458. [1396] S. SAKAI, On the group isomorphism of unitary groups in A W*-algebras, Tohoku Math. J. 7 (1955) 87-95. [1397] S. SAKAI, The absolute value of W*-algebras of finite type, Tohoku Math. J. 8 (1956) 70-85. [1398] S. SAKAI, On the e-weak topology of W*-algebras, Proc. Japan Acad. 32 (1956) 329-332. [1399] S. SAKAI. A characterization of W*-algebras. Pacific J. Math. 6 (1956) 763-773. [1400] S. SAKAI. On topological properties of W*-algebras, Proc. Japan Acad. 33 (1957) 439-444. [1401] S. SAKAI, On linear functionals of W*-algebras, Proc. Japan Acad. 34 (1958) 571-574. [1402] S. SAKAI, On some problems of C*-algebras. Tohoku Math. J. 11 (1959) 453-455. [1403] S. SAKAI, On a conjecture of Kaplansky, Tohoku Math. J. 12 (1960) 31-33. [1404] S. SAKAI, The theory of W*-algebras, Mimeographed Notes, Yale Univ., New Haven, Conn. (1962). [1405] S. SAKAI. Weakly compact operators on operator algebras, Pacific J. Math. 14 (1964) 659-664. [1406] S. SAKAI, On the reduction theory of Von Neumann, Bull. Amer. Math. Soc. 70 (1964) 393-398. [1407] S. SAKAI, A Radon-Nikodym theorem in W*-algebras, Bull. Amer. Math. Soc. 71 (1965) 149-151.

466

REFERENCES

(1408) S. SAKAI, On topologies of finite W*-algebras, Illinois J. Math. 9 (1%5) 236-241. [1409] S. SAKAI. On the central decomposition for positive functionals on C*-algebras, Trans. Amer. Math. Soc. 118 (1%5) 406-419. [1410] S. SAKAI, Derivations of W*-algebras, Ann. Math. 83 (1%6) 273-279. (1411) S. SAKAI, On a problem of Calkin. Amer. J. Math. 88 (1%6) 935-941. (1412) S. SAKAI, On a characterization of type I C*-algebras, Bull. Amer. Math. Soc. 72 (I %6) 508-512. (1413) S. SAKAI. On pure states of C*-algebras, Proc. Amer. Math. Soc. 17 (1%6) 86-87. (1414) S. SAKAI. On the preduals of W*-algebras, T6hoku Math. J. 19 (1%7) 324-331. (1415) S. SAKAI. On type 1 C*-algebras, Proc. Amer. Math. Soc. 18 (1967) 861-863. [1416] S. SAKAI, Derivations of uniformly hyperfinite C*-algebras. Publ. Res. Inst. Math. Sci. Kyoto Univ. 3 (1967) 167-175. (1417) S. SAKAI, Derivations of simple C*-algebras, J. Functional Anal. 2 (1%8) 202-206. (1418) S. SAKAI. On the tensor product of W*-algebras, Amer. J. Math. 90 (1%8) 335-341. (1419) S. SAKAI. On the hyperfinite Ilv-factor. Proc, Amer. Math. Soc. 19 (1%8) 589-591. (1420) S. SAKAI, Asymptotically abelian Il--factors, Publ. Res. Inst. Math. Sci. Kyoto Univ. 4 (1%8) 299-307. [1421] S. SAKAI, On global W*-algebras. J. Functional Anal. 3 (1%9) 79-84. (1422) S. SAKAI, An uncountable number of II,. I1~ factors, J. Functional Anal. 5 (1970) 236-246. [1423] S. SAKAI. An uncountable family of non-hyperfinite type III factors, in: Functional Analysis (Academic Press. New York, 1970) 65-70. (1424) S. SAKAI. On the Stone-Weierstrass theorem of C*-algebras, T6hoku Math. J. 22 (1970) 191-199. (1425) S. SAKAI, On global type II, W*-algebras, in: Hilbert Space Operators. Colloq. Math. Soc. Janos Bolyai, Vol. 5,(North-Holland. Amsterdam. 1970) 493-497. (1426) S. SAKAI, C*-algebras and W*-algebras. Ergebn. Math. Grenzgeb. Vol. 60 (Sprin ger, Berlin, 1971). (1427) S. SAKAI. On global type II, W*-algebras. J. Functional Anal. 8 (l971l 95-100. (1428) S. SAKAI, Derivations of simple C*-algebras, II. Bull. Soc. Math. France 99 (1971) 259-263. [1429) S. SAKAI. Derivations of simple C*-algebras. III, T6hoku Math. J. 23 (1971) 559-564. (1430) S. SAKAI. Derived C*-algebras of primitive C*-algebras, Tohoku Math. J. 25 (1973) 307-316. [1431] S. SAKAI, On automorphism groups of II ,-factors, Tohoku Math. J. 26 (1974) 423-430. (1432) S. SAKAI. On commutative normal *-derivations. Commun. Math. Phys. 43 (1975) 39-40. [1433] S. SAKAI, Derivations of simple C*-algebras, IV, to appear. (1434) S. SAKAI, Automorphisms and tensor products of operator algebras, Amer. J. Math. f)7 (1975) 889-896. (1435) S. SAKAI, On one-parameter subgroups of *-automorphisms on operator algebras and the corresponding unbounded derivations. Amer. J. Math. 98 (1976) 417-440. (1436) N. SALINAS, Extensions of C*-algebras and essentially n-normal operators, Bull. Amer. Math. Soc. 82 (1976) 143-146. [1437) M. SAMUELIDES and J. L. SAUVAGEOT. Algebre de Krieger dun systerne dynamique, c.R. Acad. Sci. Paris 280 (1975) 709-712.

REFERENCES

467

[1438) S. SANKARAN, The *-algebra of unbounded operators, J. London Math. Soc. 34 (1959) 337-344. [1439) S. SANKARAN. Ordered decompositions of Hilbert spaces, J. London Math. Soc. 36 (1961) 97-107. [144OJ S. SANKARAN, Decomposition of Von Neumann algebras of type I, Math. Ann. 142 (1961) 399-406. [1441) S. SAN KARAN and S. A. SELESNICK. Some remarks on C*-algebras and duality, Semigroup Forum 3 (1971) 108-129. [1442] U. SASAKI, Lattices of projections in AW*-algebras, J. Sci. Hiroshima Univ. 19 (1955) 1-30. [1443] J. L. SAUVAGEOT, Sur Ie type du produit croise dune algebre de Von Neumann par un groupe localement compact dautomorphismes, c.R. Acad. Sci. Paris 278 (1974) 941-944. [1444] D. G. SCHAEFFER, An application of Von Neumann algebras to finite difference equations, Ann. Math. 95 (1972) 117-129. [1445] R. SCHATTEN. Norm Ideals of Completely Continuous Operators, Ergebn. Math. Grenzgeb. (Springer, Berlin, 1960). [1446] J. M. SCHWARTZ, Sur la structure des algebres de Kac, to appear. [1447) J. T. SCHWARTZ, Two finite, non-hyperfinite. non-isomorphic factors, Commun. Pure Appl. Math. 16 (1963) 19-26. [1448] J. T. SCHWARTZ, Non-isomorphism of a pair of factors of type 111. Commun. Pure Appl. Math. 16 (1963) 111-120. [1449] J. T. SCHWARTZ, Type II factors in a central decomposition. Commun. Pure Appl. Math. 16 (1963) 247-252. [1450] J. T. SCHWARTZ, W*-algebras (Gordon and Breach, New York, 1967). [1451] J. T. SCHWARTZ, Recent progress in the structure theory of factors, in: Functional Analysis (Academic Press, New York, 1970) 37-54. [1452) Z. SEBESTYEN, A weakening of the definition of C*-algebras, Acta Sci. Math. Szeged 35 (1973) 17-20. [1453) Z. SEBESTYEN, On the definition of C*-algebras. Publ. Math, 21 (1974) 207 217. [1454) I. E. SEGAL, The group algebra of a locally compact group, Trans. Amer. Math. Soc. 61 (1947) 69-105. [1455] I. E. SEGAL, Irreducible representations of operator algebras, Bull. Amer. Math. Soc. 53 (1947) 73-88. [1456] I. E. SEGAL, Two-sided ideals in operator algebras, Ann. Math. 50 (1949) 856-865. [1457] I. E. SEGAL, The two-sided regular representation of a unimodular locally compact group, Ann. Math. 51 (1950) 293-298. [1458) I. E. SEGAL, An extension of Plancherelsformula to separable unimodular groups, Ann. Math. 52 (1950) 272-292. [1459) I. E. SEGAL, Decomposition of operator algebras, I, II, Mem. Amer. Math. Soc. 9 (1951) 1-67, 1-66. [1460) I. E. SEGAL, A class of operator algebras which are determined by groups, Duke Math. J. 18 (1951) 221-265. [1461) I. E. SEGAL, A non-commutative extension of abstract integration, AnI!. Math, 57 (1953) 401-457; correction: Ibid. 57 (1953) 595-596. [1462) I. E. SEGAL, Abstract probability spaces and a theorem of Kolmogoroff, Amer. J. Math. 76 (1954) 721-732.

468

REFERENCES

[1463] I. E. SEGAL, Tensor algebras over Hilbert spaces, Trans. Amer. Math. Soc. 81 (1956) 106-134. [1464] I. E. SEGAL, Tensor algebras over Hilbert spaces, II. Ann. Math. 63 (1956) 160-175. [1465] I. E. SEGAL, Distributions in Hilbert space and canonical systems of operators. Trans. Amer. Math. Soc. 88 (1958) 12-41. [1466] I. E. SEGAL, Foundations of the theory of dynamical systems of infinitely many degrees of freedom, I. Math. Fys. Medd. Danske Vid. Selsk. 31 (1959) 1-38. [1467] I. E. SEGAL, A note on the concept of entropy, J. Math. Mech. 9 (1960) 623-630. [1468] I. E. SEGAL. Foundations of the theory of dynamical systems of infinitely many degrees of freedom, II, Can. J. Math. I3 (1%1) 1-18. [1469] I. E. SEGAL. Mathematical characterization of the physical vacuum for a linear Bose-Einstein field. Illinois J. Math. 6 (1%2) 500-523. [1470] I. E. SEGAL, Algebraic integration theory. Bull. Amer. Math. Soc. 71 (1965) 419--4S9. [1471] I. E. SEGAL, Notes toward the construction of nonlinear relativistic quantum fields. I. The Hamiltonian in two space-time dimensions as the generator of a C* automorphism group. Proc. Natl. Acad. Sci. U.S.A. 57 (1967) 1178-1183; II. The basic nonlinear functions in general space-times, Bull. Amer. Math. Soc. 75 (1%9) 1383-1389; III. Properties of the C*-dynamics for a certain class of interactions, Bull. Amer. Math. Soc. 75 (1%9) 1391-1395. [1472] I. E. SEGAL, Non-linear quantum processes and automorphism groups of C* algebras, in: Actes Congo Intern. Math., Nice, 1970, Vol. 2, pp. 419-426. [1473] I. E. SEGAL and J. VON NEUMANN. A theorem on unitary representations of semi-simple Lie groups, Ann. Math. 52 (1950) 509-517. [1474] G. L. SEWELL, States and dynamics of infinitely extended physical systems. Commun, Math. Phys. 33 (1973) 43-51. [1475] D. SHALE and W. F. STINESPRING, States of the Clifford algebra, Ann. Math. 80 (1964) 365-381. [1476] S. SHERMAN, The second adjoint of a C*-algebra, in: Proc. Intem. Congr. Math. Cambridge. 1950. Vol. 1, pp. 470. [1477] S. SHERMAN, Order in operator algebras, Amer. J. Math. 73 (1951) 227-232. [1478] A. N. SHERSTNEV. States on Von Neumann algebras, Funkcion. Anal. Priloi. 8 (3) (1974) 89-90 (in Russian). [1479] P. C. SHIELDS, A new topology for Von Neumann algebras, Bull. Amer. Math. Soc. 65 (1959) 267-269. [1480] V. S. SHULMAN, Multiplication operators in C*-algebras and the problem of reflexivity of algebras which contain a m.a.s.a.. Funkcion. Anal. Priloi, 8 (1974) 92-93 (in Russian). [1481] A. M. SINCLAIR. Jordan homomorphisms and derivations of semi-simple Banach algebras, Proc. Amer. Math. Soc. 24 (1970) 209-214. [1482] A. M. SINCLAIR, Annihilator ideals in the cohomology of Banach algebras, Proc. Amer. Math. Soc. 33 (1972) 361-366. [1483] A. M. SINCLAIR, Homomorphisms from C*-algebras, Proc. London Math. Soc. 29 (1974) 435-452 and 32 (1976) 322. [1484] I. M. SINGER, Uniformly continuous representations of Lie groups, Ann. Math. 56 (1952) 242-247.

REFERENCES

469

[1485J I. M. SINGER, Automorphisms of finite factors, Amer. J. Math. 57 (1955) 117-133. [1486J M. SIRUGUE, Les etats quasi-libres de lalgebre de Clifford comme solution des conditions de K.M.S., in: Cargese Lectures in Phys. 4 (1970) 335-348. [1487] M. SIRUGUE and D. TESTARD, Some connections between ground states and temperature states of thermodynamical systems, Commun. Math. Phys. 22 (1971) 223-237. [1488J M. SIRUGUE and W. WINNINK, Constraints imposed upon a state of a system that satisfies the KMS boundary condition, Commun. Math. Phys. 19 (1970) 161-168. [1489] M. SIRUGUE and M. WINNINK, Une generalisation du theorerne dunicite de lautomorphisme modulaire dune algebre de Von Neumann de genre denombrable, c:s. Acad. Sci. Paris 272 (1971) 1185-1187. [1490] M. SIRUGUE and M. WINNINK, Translations dans Ie temps comme groupe d*-automorphisms, to appear. [1491] C. F. SKAU, Commutative projection and abelian subalgebras, Dissertation, Univ. of Pennsylvania, Philadelphia, Pa., (1973). [1492J J. SLAWNY, Representations of canonical anticommutation relations and imp le mentability of canonical transformations, Commun. Math. Phys. 22 (1971) 104 114. [1493J J. SLAWNY, On factor representations and C*-algebra of canonical commutation relations, Commun. Math. Phys. 24 (1972) 151-170. [1494] M. S. B. SMITH, On automorphism groups of C*-algebras, Trans. Amer. Math. Soc. 152 (1970) 623-648. [1495] M. R. F. SMYTH, Riesz theory in Banach algebras, Math. Z. 145 (1975) 145-155. [14%J M. G. SONIS, On a class of operators in Von Neumann algebras with Segal measure on the projectors, Mat. Sb. 84 (1971) 353-368 (in Russian). [1497J J. S0RENG, Pure states of simple C*-algebras,J. Functional. Anal. 22 (1976) 390-404. [1498] P. G. SPAIN, On commutative V*-algebras, Proc. Edinburgh Math. Soc. 17 (1970) 173-180. [1499J P. G. SPAIN, V*-algebras with weakly compact unit spheres, J. London Math. Soc. 4 (1971) 62-64. [1500] P. G. SPAIN, The W*-closure of a V*-algebra, J. London Math. Soc. 7 (1974) 385-386. [1501J J. G. STAMPFLI, The norm of a derivation, Pacific J. Math. 33 (1970) 737-747. the range, Illinois J. Math. 17 (1973) [1502J J. G. STAMPFLI, Derivations on ell(~: 518-524. [1503J J. R. STEFANSSON, On a problem of J. Dixmier concerning ideals in a Von Neumann algebra, Math. Scand. 24 (1969) 111-112. [1504] J. D. STEIN, Continuity of homomorphisms of Von Neumann algebras, Amer. J. Math. 91 (1%9) 153-159. [1505] J. D. STEIN, Homomorphisms of B*-algebras, Pacific J. Math. 28 (1969) 431-439. [1506] W. F. STINESPRING, Positive functions on C*-algebras, Proc. Amer. Math. Soc. 6 (1955) 211-216. [1507J W. F. STINESPRING, A semi-simple matrix group is of type I, Proc. Amer. Math. Soc. 9 (1958) 965-967. [1508] W. F. STINESPRING, Integration theorems for gages and duality for unimodular groups, Trans. Amer. Math. Soc. 90 (1959) 15-56. [1509J W. F. STINESPRING, Integrability of Fourier transforms of unimodular Lie groups, Duke Math. J. 26 (1959) 123-131.

470

REFERENCES

[1510] E. ST0RMER, Positive linear maps of operator algebras, Acta Math. 1I0 (1963) 233-278. [1511] E. ST0RMER, On the Jordan structure of C*-algebras, Trans. Amer. Math. Soc. 120 (1965) 438-447. [1512] E. ST0RMER, Jordan algebras of type I, Acta Math. lIS (1966) 165-184. [1513] E. ST0RMER, On anti-automorphisms of Von Neumann algebras, Pacific J. Math. 21 (1967) 349-370. [1514] E. ST0RMER, Types of Von Neumann algebras associated with extremal invariant states, Commun. Math. Phys. 6 (1967) 194-204. [1515] E. ST0RMER, Two-sided ideals in C*-algebras, Bull. Amer. Math. Soc. 73 (1967) 254-257. [1516] E. ST0RMER, Positive linear maps and Jordan homomorphisms of C*-algebras, in: Symp. on C*-algebras, Baton Rouge (1967). [1517] E. ST0RMER, Large groups of automorphisms of C*-algebras, Commun. Math. Phys. S (1967) 1-22. [1518] E. ST0RMER. On partially ordered vector spaces and their duals. with applications to simplexes and C*-algebras, Proc. London Math. Soc. 18 (1968) 245-265. [1519] E. ST0RMER, Irreducible Jordan algebras of self-adjoint operators. Trans. Amer. Math. Soc. 130 (1968) 153-166. [1520] E. ST0RMER, A characterization of pure states of C*-algebras, Proc. Amer. Math. Soc. 19 (1968) 1100-1102. [1521] E. ST0RMER. Symmetric states of infinite tensor products of C*-algebras. J. Functional Anal. 3 (1969) 48-68. [1522] E. ST0RMER, States and invariant maps of operator algebras, J. Functional Anal. S (1970) 44-65. [1523] E. ST0RMER. The even CAR algebra. Commun. Math. Phys. 16 (1970) 136-137. [1524] E. ST0RMER, Asymptotically abelian systems, in: Cargese lectures in Phys. 4 (1970) 195-213. [1525] E. ST0RMER. On infinite tensors of Von Neumann algebras. Amer. J. Math. 93 (1971) 810-818. [1526] E. ST0RMER, Automorphisms and invariant states of operator algebras, Acta Math. 127 (1971) 1-9. [1527] E. ST0RMER. Hyperfinite product factors. Ark. Mat. 9 (1971) 165-170. [1528] E. ST0RMER, Invariant states of Von Neumann algebras. Math. Scand. 90 (1972) 253-256. [1529] E. ST0RMER. Spectra of states and asymptotically abelian C*-algebras. Commun. Math. Phys. 28 (1972) 279-294. [1530] E. ST0RMER, Hyperfinite product factors, II, J. Functional Anal. 10 (1972) 471-480. [1531] E. ST0RMER, On projection maps of Von Neumann algebras, Math. Scand. 30 (1972) 46-50. [1532] E. ST0RMER, Automorphisms and equivalence in Von Neumann algebras, Pacific J. Math. 44 (1973) 371-383. [1533] E. ST0RMER. Spectra of ergodic transformations. J. Functional Anal. IS (1974) 202-215. [1534] E. ST0RMER, Inner automorphisms of Von Neumann algebras. Commun. Math. Phys. 36 (1974) 115-122. [1535] E. ST0RMER. Hyperfinite product factors. III. Amer. J. Math. 97 (1975) 589-595.

REFERENCES

471

[1536] S. STRATI LA and L. ZSIDO, Une theorie algebrique de la reduction pour les Acad. Sci. Paris 272 (1971) 1453-1456. W*-a1gebres, [1537] S. STRATI LA and L. ZSIDO, Sur la theorie algebrique de la reduction pour les W*-algebres, C.R. Acad. Sci. Paris 275 (1972) 451-454. [1538] S. STRATI LA and L. ZSIDO, An algebraic reduction theory for W*-algebras, J. Functional Anal. II (1972) 295-313. (1539) S. STRATI LA and L. ZSID6, An algebraic reduction theory for W*-algebras, II, Rev. Roumaine Math. Pures Appl. 18 (1973) 407-460. (1540) S. STRATI LA and L. ZSID6. On the range of a derivation in a W*-algebra. Rev. Roumaine Math. Pures Appl. 18 (1973) 101-103. [1541) S. STRATI LA and L. ZSIDO. A spectral characterization of the maximal ideal in factors, Acta Sci. Math. Szeged 36 (1974) 155-159. [1542) R. F. STREATER, Intensive observables in quantum theory, J. Math. Phys. 5 (1964) 581-590. [1543] R. F. STREATER, Current commutation relations and continuous tensor products, Nuovo Cimento 53 (1968) 487-495. [1544) R. F. STREATER, On certain non-relativistic quantized fields, Commun. Math. Phys. 7 (1968) 93-98. (1545) R. F. STREATER, A continuum analogue of the lattice gas. Commun. Math. Phys. 12 (1969) 226-232. [1546) R. F. STREATER and I. F. WILDE. The time evolution of quantized fields with bounded quasi-local interaction density, Commun. Math. Phys. 17 (1970) 21-32. [1547) R. F. STREATER and A. WULFSOHN. Continuous tensor product of Hilbert spaces, Nuovo Cimento (B) 57 (1968) 330-339. [1548] C. SUNOUCHI. A generalization of Schatten-Von Neurnann-Dixrnier theorem for type I AW*-algebras. Tohoku Math. J. 23 (1971) 727-734. [1549] H. SUNOUCHI, On rings of operators of infinite classes. Proc. Japan Acad. 28 (1952) 9-13. [1550) H. SUNOUCHI, On rings of operators of infinite classes. II, Proc. Japan Acad. 28 (1952) 330-335. [1551] H. SUNOUCHI, The irreducible decomposition of the maximal hilbert algebras of the finite class. Tohoku Math. J. 4 (1952) 207-215. [1552) H. SUNOUCHI, An extension of the Plancherel formula to unimodular groups, Tohoku Math. J. 4 (1952) 216-230. [1553) H. SUNOUCHI. A characterizationof the maximal ideal in a factor of the case 1100, I. II. Kodai Math. Sem. Rept. 6 (1954) 7. 7 (1955) 65-66. [1554] H. SUNOUCHI. Infinite Lie rings. Tohoku Math. J. 8 (1956) 291-307. [1555] C. E. SUTHERLAND. Direct integral theory for weights. and the Plancherel formula. Bull. Amer. Math. Soc. 80 (1974) 456-461. [1556] C. E. SUTHERLAND. Crossed products. direct integrals and Connes classification of type III factors. to appear. [1557) N. SUZUKI. On the invariants of Wt-algebras. T6hoku Math. J. 7 (1955) 177-185.

c:s.

(1558) N. SUZUKI. On automorphisms of W*-algebras leaving the center elementwise invariant. T6hoku Math. J. 7 (1955) 186-191. [1559] N. SUZUKI. A linear representation of a countably infinite group. Proc. Japan Acad. 34 (1958) 575-579.

472

REFERENCES

[1560] N. SUZUKI, Crossed products of rings of operators, Tohoku Math. J. 11 (1959) 113-124. [1561] N. SUZUKI, Certain types of groups of automorphisms of a factor, Tohoku Math. J. 11 (1959) 314-320, 12 (1959) 477-478. [1562] N. SUZUKI, Extensions of rings of operators in Hilbert spaces, Tohoku Math. J. 14 (1962) 217-232. [1563] N. SUZUKI, Isometries on Hilbert spaces, Proc. Japan Acad. 39 (1963) 435-438. [1564] N. SUZUKI, On the type of completely continuous operators. Proc. Japan Acad. 40 (1964) 683-685. [1565] N. SUZUKI, Algebraic aspects of non self-adjoint operators. Proc. Japan Acad. 41 (1965) 706-710. [1566] N. SUZUKI, On the spectral decomposition of dissipative operators, Proc, Japan Acad. 42 (1%6) 577-582. [1567] N. SUZUKI, The algebraic structure of non self-adjoint operators, Acta Sci. Math. Szeged 28 (1%6) 173-184. [1568] N. SUZUKI, On a weakly convergent sequence of normal functionals on a Von Neumann algebra. Proc. Amer. Math. Soc. 22 (1969) 697-701. [1569] N. SUZUKI, Representation of certain Banach *-aIgebras.Proc. Japan Acad. 4S (1969) 696-699. [1570] N. SUZUKI and T. SAITO, On the operatorswhich generate continuous Von Neumann algebras. Tohoku Math. J. 15 (1963) 277-280. [1571] W. SZYMANSKI. Characters of finitely generated C*-algebras. Ann. Polon. Math. 27 (1973) 317-322. [1572] I. B. T ASOR, On the theory of categories of Hilbert Spaces. v estnik Moskov. Univ. (I) Mat. Meh. 26 (l97\) 65-68 (in Russian). [1573] R. TAKAHASHI. Sur les representations unitaires des groupes de Lorentz general ises, Bull. SOl;. Math. France 91 (1963) 289-433. [1574] A. TAKAHASHI, The Cauchy-Schwarz inequality. Rev. Colombiana Mat. 7 (1973) 101-107 (in Spanish). [1575] S. TAKAHASI. A note on the theorem of Stone-Weierstrass. Bull. Fac. Sci. Ibaraki Univ. 2 (1969) 5-7. [1576] S. TAKAHASI. On Urysohns lemma for spectrums of C*-algebras, Bull. Fac, Sci. Ibaraki Univ. 3 (1971) 5-11. [1577) S. TAKAHASI. Duality and Von Neumann algebras. Bull. Fac. Sci. Ibaraki Univ. 4 (1972) 33-35. [1578] S. T AKAHASI, On locally inner automorphisms of certain C*-algebras, Bull. Fac. Sci. Ibaraki Univ. 5 (1973) 41-46. [1579) S. T AKAHASI, A note on the homomorphic images of the center of C*-algebras, Bull. Fac. Sci. Ibaraki Univ. 6 (1974) 29-32. [1580] H. TAKAI, On an algebraic model for Von Neumann algebras. Proc. Japan Acad. 46 (1970) 941-943. [1581] H. TAKAI. On an algebraic model for Von Neumann algebras. II. Proc. Japan Acad. 46 (1970) 1091-1095. [1582] H. TAKAI, On an algebraic model for Von Neumann algebras, III, Mem. Osaka Kyoiku Univ. 20 (l97\) 121-123. [1583] H. T AKAI, Dualite dans les produits croises de C*-algebres. CiR, Acad. Sci. Paris 278 (1974) 1041-1043.

REFERENCES

473

[1584] H. TAKAI, On a duality for crossed products of C-algebras, J. Functional Anal. 19 (1975) 25-39. [1585] H. TAKAI, The quasi-orbit space of continuous Colo-dynamical systems, Trans. Amer. Math. Soc. 216 (1976) 105-113. [1586] H. TAKAI and H. YAMADA, A note on the dilation theorem, Proc. Japan Acad. 48 (1972) 216-220. [1587] Z. TAKEDA, On a theorem of R. Pallu de la Barriere, Proc. Japan Acad. 28 (1952) 558-563. [1588] Z. TAKEDA, Perfection of measure spaces and W*-algebras, Kodai Math. Sem. Rept. 5 (1953) 23-26. [1589] Z. TAKEDA, Conjugate spaces of operator algebras, Proc. Japan Acad. 30 (1954) 90-95. [1590] Z. TAKEDA, On the representations of operator algebras, Proc. Japan Acad. 30 (1954) 299-304. [1591] Z. TAKEDA, On the representations of operator algebras, II, Tohoku Math. J. 6 (1954) 212-219. [1592] Z. TAKEDA, Inductive limit and infinite direct product of operatoralgebras, Tohoku Math. J. 7 (1955) 67-86. [1593] Z. TAKEDA, On the extensions of finite factors, II, Proc. Japan Acad. 35 (1959) 215-220. [1594] Z. TAKEDA, On the extension theorem of the Galois theory for finite factors, Proc. Japan Acad. 37 (l96I) 78-82. [1595] Z. TAKEDA, On the normal basis theorem of the Galois theory for finite factors, Proc. Japan Acad. 37 (1961) 144-148. [1596] Z. TAKEDA and T. TURUMARU, On the property "Position p", Math. Japon. 2 (1952) 195-197. [1597] H. TAKEMOTO, W*-algebra with a non-separable cyclic representation, Tohoku Math. J. 20 (1968) 567-576. [1598] H. TAKEMOTO, On the homomorphism of Von Neumann algebra, Tohoku Math. J. 21 (1969) 152-157. [1599] H. TAKEMOTO, On the integral representation of some functional on a Von Neumann algebra, Tohoku Math. J. 21 (1969) 237-248. [1600] H. TAKEMOTO, A complement to "On the homomorphism of Von Neumann algebra", Tohoku Math. J. 22 (1970) 210-211. [1601] H. TAKEMOTO, On the C.-classes in the maximal CCR ideal of a Von Neumann algebra, Tohoku Math. J. 23 (1971) 301-311. [1602] H. TAKEMOTO, On a characterization of AW*-modules and a representation of Gelfand type of non-commutative operator algebras, Michigan Math. J. 20 (1973) 115-127. [1603] H. TAKEMOTO, Decomposable operators in continuous fields of Hilbert spaces, Tohoku Math. J. 27 (1975) 41~35. [1604] H. TAKEMOTO and 1. TOMIYAMA, On the reduction of finite Von Neumann algebra, Tohoku Math. J. 25 (1973) 273-289. [1605] O. TAKENOUCHI, On the maximal Hilbert algebras, Math. J. Okayama Univ. 1 (1952) 1-31. [1606] O. TAKENOUCHI, On the structure of maximal Hilbert algebras, Math. J. Okayama Univ. 1 (1952) 1-31.

474

REFERENCES

[1607] O. T AKENOUCHI, Sur une c1asse de fonctions continues de type positif sur un groupe localement compact, Math. J. Okayama Univ. 4 (1955) 153-173. [1608] O. T AKENOUCHI, Sur la facteur representation dun groupe de Lie resoluble de type (E), Math. J. Okayama Univ. 7 (1957) 151-161. [1609] O. T AKENOUCHI, Sur les sous-algebres dune algebre de Hilbert, Ann. Sci. Ecole Norm. Sup. 78 (1961) 211-240. [1610] O. TAKENOUCHI, On type classification of factors constructed as infinite tensor products, Publ. Res. Inst. Math. Sci. Kyoto Univ. 4 (1968) 467-482. [1611] M. TAKESAKI, On the direct product of W*-factors, Tohoku Math. J. 10 (1958) 116-119. [1612] M. TAKESAKI, On the conjugate space of an operator algebra, Tohoku Math. J. 10 (1958) 194-203. [1613] M. TAKESAKI, A note on the cross norm of the direct product of operatoralgebras, Kodai Math. Sem. Rept. 10 (1958) 137-140. [1614] M. TAKESAKI, A note on the direct product of operator algebras, Kodai Math. Sem. Rept. II (1959) 178-181. [1615] M. TAKESAKI, On the singularity of a positive linear functional on operator algebras, Proc. Japan Acad. 35 (1959) 365-366. [1616] M. TAKESAKI. On the Hahn-Banach type theorem and the Jordan decomposition of module linear mapping over some operator algebra, Kodai Math. Sem. Rept. 12 (1960) 1-10. [1617] M. TAKESAKI. On the non-separability of singular representations of operator algebra, Kodai Math. Sem. Rept 12 (1960) 102-1081 [1618] M. TAKESAKI. On some representations of C*-algebras. Tohoku Math. J. 15 (1963) 79-95. [1619] M. TAKESAKI. On the unitary equivalence among the components of decom positions of representations of involutive Banach algebras and the associated diagonal algebras, To/wku Math. J. 15 (1963) 365-393. [1620] M. T AKESAKl, On the cross-norm of the direct product of C*-algebras, Tohoku Math. J. 16 (1964) 111-122. [1621] M. TAKESAKI, A complement to "On the unitary equivalence among the com ponents of decompositions of representations of involutive Banach algebras and the associated diagonal algebras", Tohoku Math. J. 16 (1964) 226-227. [1622] M. TAKESAKI. Covariant representations of C*-algebras and their locally compact automorphism groups. Acta Math. 119 (1967) 273-303. [1623] M. T AKESAKI. A duality in the representation theory of C*-algebras, Ann. Math. 85 (1967) 370-382. [1624] M. TAKESAKt, Remarks on the reduction theory of Von Neumann algebras. Proc. Amer. Math. Soc. 20 (1969) 434-438. [1625] M. T AKESAKI, A characterization of group algebras as a converse of Tannaka Stinespring-Tatsuuma duality theorem, Amer. J. Math. 91 (1969) 529-564. [1626] M. TAKESAKI. A generalized commutation relation for the regular representation, Bull. Soc. Math. France 97 (1969) 289-297. [1627] M. TAKESAKI, The theory of operator algebras, Lecture Notes, Univ. of Califor nia, Los Angeles (1969-70). [1628] M. T AKESAKI. Disjointness of the KMS-states of different temperature, Commun. Math. Phys. 17 (1970) 21-32.

REFERENCES

475

(1629) M. TAKESAKI. Tomitas theory of modular Hilbert algebras and its applications, Lecture Notes in Math. 128 (Springer, Berlin, 1970). (1630) M. T AKESAKI, Algebraic equivalence of locally normal representations, Pacific J. Math. 34 (1970) 807-816. (1631) M. TAKESAKI. The quotient algebra of a finite Von Neumann algebra, Pacific J. Math. 36 (197 I) 827-831. (1632) M. T AKESAKI, A liminal crossed product of a uniformly hyperfinite C*-algebra by a compact abelian automorphism group, J. Functional Anal. 7 (1971) 140--146. [1633] M. TAKESAKI, Duality and Von Neumann algebras. Bull. Amer. Math. Soc. 77 (1971) 553-557. (1634) M. TAKESAKI, Conditional expectations in Von Neumann algebras, J. Functional Anal. 9 (1972) 306-321. (1635) M. TAKESAKI, Duality and Von Neumann algebras, in: Lecture Notes in Math. 247 (Springer, Berlin, 1972) 663-786. [1636] M. TAKESAKI, A short proof for the commutation theorem (M, 0 M 2 ) = M; 0 M;, in: Lecture Notes in Math. 247 (Springer. Berlin. 1972) 780-786. [1637] M. TAKESAKI. Dualite dans les produits croises dalgebres de Von Neumann, C.R. Acad. Sci. Paris 276 (1973) 41-43. [1638] M. TAKESAKI. Algebres de Von Neumann proprement infinies et produits croises, c.s. Acad. Sci. Paris 276 (1973) 125-127. [1639] M. TAKESAKI, Periodic and homogeneous states on a Von Neumann algebra, I, II. III. Bull. Amer. Math. Soc. 79 (1973) 202-206, 416-420, 559-563. [1640] M. TAKESAKI, The structure of a Von Neumann algebra with a homogeneous periodic state. Acta Math. 131 (1973) 79-121. [1641] M. T AKESAKI. Duality in crossed products and Von Neumann algebras of type III. Bull. Amer. Math. Soc. 79 (1973) 1004--1005. [1642] M. TAKESAKI, Duality for crossed products and the structure of Von Neumann algebras of type III. Acta Math. 131 (1973) 249-310. [1643] M. TAKESAKI, Faithful states on a C*-algebra, Pacific J. Math. 52 (1973) 605-610. [1644] M. T AKESAKI. Disintegration of locally normal states and algebraic equivalence of locally normal representations. to appear. [1645] M. TAKESAKI and N. TATSUVMA, Duality and subgroups, Ann. Math. 93 (1971) 344-364. [1646] M. TAKESAKI and N. TATSUUMA, Duality and subgroups. II. to appear. [1647] M. TAKESAKI and M. WINNINK, Local normality in quantum statistical me chanics, Commun. Math. Phys. 30 (1973) 129-152. [1648] P. K. TAM, On an ergodic abelian M-group, Proc. Japan Acad. 47 (1971) 456-457. [1649] P. K. TAM, On the unitary equivalence of certain classes of non-normal operators, I, Can. J. Math. 23 (1971) 849-856. [1650] P. K. TAM, On the commutant of certain automorphism groups, Can. J. Math. 25 (1973) 1165-1169. [1651] R. J. TAUER. Maximal abelian subalgebras in finite factors of type II. Trans. Amer. Math. Soc. 114 (1965) 281-308. [1652] R. J. T AVER, Semi-regular maximal abelian subalgebras in hyperfinite factors. Bull. Amer. Math. Soc. 71 (1965) 606-608. [1653] R. J. TAUER. M-Semiregular subalgebras in hyperfinite factors. Trans. Amer. Math. Soc. 129 (1967) 530-541.

476

REFERENCES

[1654] D. C. TAYLOR. The strict topology for double centralizer algebras, Trans. Amer. Math. Soc. 150 (1970) 633-643. [1655] D. C. TAYLOR, Interpolation in algebras of operator fields, J. Functional Anal. 10 (1972) 159-190. [1656] D. C. TAYLOR, A general Phillips theorem for C*-algebras and some applications, Pacific J. Math. 40 (1972) 477-488. [1657] D. C. TAYLOR, A general Hoffman-Wermer theorem for algebras of operator fields, Proc. Amer. Math. Soc. 52 (1975) 212-216. [1658] J. L. TAYLOR, The Tomita decomposition of rings of operators, Trans. Amer. Math. Soc. 113 (1%4) 30-39. [1659] S. TELEMAN, Sur les algebres de J. Von Neumann, Bull. Sci. Math. 82 (1958) 117-126. [1660] S. TELEMAN, La representation des algebres de Von Neumann finies par fais ceaux, Rev. Roumaine Math. Pures Appl. 15 (1970) 143-151. [1661] S. TELEMAN, Representation of finite Von Neumann algebras by sheaves, Studia Math. 38 (1970) 465-466. [1662] S. TELEMAN, Representations of Von Neumann algebras by sheaves, in: Hilbert Space Operators, Colloq. Math. Soc. Janos Bolyai, Vol. 5 (North-Holland, Am sterdam) 519-538. [1663] S. TELEMAN, Theory of harmonic algebras with applications to Von Neumann algebras and cohomology of locally compact spaces (de Rhams theorem), in: Lecture Notes in Math. 248 (Springer, Berlin, 1971) 100-315. [1664] S. TELEMAN, On algebraic reduction theory, Rev. Roumaine Math. Pures Appl. 17 (1972) 287-307. [1665] S. TELEMAN, Algebraic reduction and dimension theory, Rev. Roumaine Math. Pures Appl. 17 (1972) 1265-1281. [1666] S. TELEMAN, The commutation theorem in the algebraic reduction theory, Rev. Roumaine Math. Pures Appl. 17 (1972) 1451-1472. [1667] S. TELEMAN, Extended Von Neumann algebras and their representation by sheaves, Rev. Roumaine Math. Pures. Appl. 18 (1973) 727-760. [1668] S. TELEMAN, Sur la reduction des algebres de Von Neumann, CiR, Acad. Sci. Paris 280 (1975) 1685-1688. [1669] S. TELEMAN, Sur I~ reduction des algebres de Von Neumann, Rev. Roumaine Math. Pures Appl. 20 (1975) 693-742. [1670] D. TESTARD, Algebre de covariance, Ann. Inst. H. Poincare 6 (1%7) 267-297. [1671] M. R. TETRANASVILI, Measures of Haar type which are not invariant with respect to the symmetry of a group, Sakharth. SSR Mecn. Akad. Moambe 47 (1%7) 519-524. [1672] E. THOMA, Zur Reduktionstheorie in separabelen Hilbert-Railmen, Math. Z. 67 (1957) 1-9. [1673] E. THOMA, Zur Reduktionstheorie in allgemeinen Hilbert-Raiimen, Math. Z. 68 (1957) 153-188. [1674] E. THOMA, Uber unitiire Darstellungen abzahlbarer, diskreter Gruppen, Math. Ann. IS3 (1964) 111-138. [1675] TI YEN, Trace on finite AW*-algebras, Duke Math. J. 22 (1955) 207-222. [1676] TI YEN, Isomorphisms of unitary groups in A W*-algebras, Ttmoku Math. J. 8 (1956) 275-280.

REFERENCES

477

[1677] Tt YEN, Quotient algebra of a finite A W*-algebra, Pacific J. Math. 6 (1956) 389-395. [1678] Tt YEN. Isomorphisms of AW*-algebras, Proc. Amer. Math. Soc. 8 (1957) 345-349. [1679] M. TOMITA, On rings of operators in non-separable Hilbert spaces, Mem. Fac. Sci. Kyushu Univ. 7 (1953) 129-168. [1680] M. TOMITA. Representations of operator algebras. Math. J. Okayama Univ. 3 (1954) 147-173. [1681] M. TOMITA. Harmonic analysis on locally compact groups. Math. J. Okamaya Univ. 5 (1956) 133-193. [1682] M. TOMITA. Spectral theory of operator algebras, I, Math. J. Okayama Univ. 9 (1959) 63-98. [1683] M. TOMITA, Spectral theory of operator algebras, II. Math. J. Okayama Univ. 10 (1960) 19--60. [1684] M. TOMITA. The second dual of a C*algebra, Mem, Fac. Sci. Kyushu Univ. 21 (1%7) 185-193. [1685] M. TOMITA, Quasi-standard Von Neumann algebras, Mimeographed Notes (1967). [1686] B. J. TOMIUK, Duality and the existence of weakly completely continuous elements in a B*-algebra, Glasgow Math. J. 13 (1972) 5fHiO. [1687] B. J. TOMIUK and K. WONG, The Arens product and duality in B*-algebras. Proc. Amer. Math. Soc. 25 (1970) 529-535. [1688] J. TOMIYAMA. On the projection of norm one in W*-a1gebras, Proc. Japan Acad. 33 (1957) 608--612. [1689] J. TOMIY AMA, On the projection of norm one in W*-algebras, II, Tohoku Math. J. 10 (1958) 204-209. [1690] J. TOMIY AMA, A remark on the invariants of W*-algebras, Tohoku Math. J. 10 (1958) 37-41. [1691] J. TOMIY AMA. Generalized dimension function for W*-algebras of finite type. Tohoku Math. J. 10 (1958) 121-129. [1692] J. TOMIY AMA, On the projection of norm one in W*-algebras. III. Tohoku Math. J. 11 (1959) 125-129. [1693] J. TOMIY AMA. On the product projection of norm one in the direct product of operator algebras, Tohoku Math. J. 11 (1959) 305-313. [1694] J. TOMIY AMA. Topological representationsof C*-algebras, Tohoku Math. J. 14 (1962) 187-204. [1695] J. TOMfY AMA, A characterization of C*-algebras whose conjugate spaces are separable, Tohoku Math. J. IS (1%3) %-102. [16%] J. TOMIYAMA. Applications of Fubini type theorem to the tensor product of C*-algebras, Tohoku Math. J. 19 (1%7) 213-226. [1697] J. TOMIY AMA, A remark on representations of CCR algebras, Proc. Amer. Math. Soc. 19 (1968) 1506. [1698] J. TOMIY AMA. On the tensor products of Von Neumann algebras, Pacific J. Math. 30 (1%9) 263-270. [1699] J. TOMIYAMA. Tensor products and projections of norm one in Von Neumann algebras, Lecture Notes, Univ. of Copenhagen (1970). [1700] J. TOMIYAMA, Applications de Fubini dans les produits tensoriels infinis dalgebres de Von Neumann, Bull. Yamagata Univ. Natur. Sci. 8 (1973) 361-368. [1701] J. TOMIYAMA, On some types of maximal abelian subalgebras, J. Functional Anal. 10 (1972) 373-386.

478

REFERENCES

[1702] J. TOMIYAMA, Derivations of C*-algebras which are not determined by multipliers in any quotient algebra, Proc. Amer. Math. Soc. 47 (1975) 265-267. [1703] J. TOMIY AMA, Functional fields and module duals of operator algebras, to appear. [1704] J. TOMIYAMA, The extension property of Von Neumann algebras and a class of C*-algebras associated to them, to appear. [1705] J. TOMIYAMA, Primitive ideals in tensor products of Banach algebras, to appear. [1706] J. TOMIYAMA, Tensor products and approximation problems of C*-algebras, Pub/. Res. Inst. Math. Sci. Kyoto Univ. 11 (1975) 163-183. [1707] J. TOMIY AMA, Derived algebras of C*-algebras, to appear. [1708] J. TOMIY AMA and M. T AKESAKI, Applications of fibre bundles to the certain class of C*-algebras, Tohoku Math. J. 13 (1%1) 498-523. [1709] D. M. TOPPING, Vector lattices of self-adjoint operators, Bull. Amer. Math. Soc. 69 (1%3) 251-255. [1710] D. M. TOPPING, Vector lattices of self-adjoint operators, Trans. Amer. Math. Soc. 115 (1%5) 14-30. [1711] D. M. TOPPING, Jordan algebras of self-adjoint operators. Bull. Amer. Math. Soc. 71 (1%5) 160-164, [1712] D. M. TOPPING, Jordan algebras of self-adjoint operators, Mem. Amer. Math. Soc. 53 (1%5). [1713] D. M. TOPPING, An isomorphism invariant for spin factors, J. Math. Mech. 15 (1%6) 1055-1064. [1714] D. M. TOPPING, Asymptoticity and semi-modularity in projection lattices, Pacific J. Math. 20 (1%7) 317-325. [1715] D. M. TOPPING, UHF algebras are singly generated, Math. Scand. 22 (1%8) 224-226. [1716] D. M. TOPPING, Transcendental quasi-nilpotents in operator algebras, J. Func tional Anal. 2 (1%8) 342-351. [1717) D. M. TOPPING, On linear combinations of special operators, J. A/gebra 10 (1%8) 516-521. [1718) D. M. TOPPING, Lectures on Von Neumann Algebras (Van Nostrand, New York, 1971). [1719) E. TORRANCE, Maximal C*-subalgebrasof a Banach algebra, Proc. Amer. Math. Soc. 25 (1970) 622-{)24. [1720] C. S. TSAu, Non-isomorphic factors of type III without property L, to appear. [1721] K. TSUJI, Representation theorems of operator algebras and their applications, Proc. Japan Acad. 31 (1955) 272-277. [1722] K. TSUJI. Harmonic analysis on locally compact groups, Bull. Kyushu Inst. Technol. 2 (1956) 16-32. [1723) K. TSUJI, W*-algebras and abstract L-spaces, Bull. Kyushu Inst. Techno/. 3 (1957) 11-13. [1724] K. TSUJI, Annihilators of Von Neumann algebras (annihilating spaces), Bull. Kyushu Inst. Technol. 10 (1%3) 25-39, 13 (1%3) 19. [1725] K. TSUJI, The J. Von Neumann commutation theorem. Bull. Kyushu Inst. Techno/. 13 (1%6) 1-3. [1726] K. TSUJI, Tensor product of annihilating spaces, Bull. Kyushu Inst. Technol. 13 (1%6) 15-18. [1727] K. TSUJI. On nest algebras of operators. Proc. Japan Acad. 46 (1970) 337-340.

REFERENCES

479

(1728] T. TURUMARU, On the direct product of operator algebras, I, Tohok Math. J. 4 (1952) 242-251. [1729] T. TURUMARU, On the direct product of operator algebras, II, Tohoku Math. J. 5 (1953) 1-7. [1730] T. TURUMARU, On the direct product of operator algebras, III, Tohoku Math. J. 6 (1954) 208-211. [1731] T. TURUMARU, On the direct product of operator algebras, IV, Tohoku Math. J. 8 (1956) 281-285. [1732] T. TURUMARU, Crossed product of operator algebra, Tohoku Math. J. 10 (1958) 335-365. [1733] H. UMEGAKI, On some representation theorems in an operator algebra. I. II. Proc. Japan Acad. 27 (1951) 328-333, 501-505. (1734] H. UMEGAKI. On some representation theorems in an operator algebra. III, Proc. Japan Acad. 28 (1952) 29-31. [1735] H. UMEGAKI, Operator algebra of finite class, Kodai Math. Sem. Rept. 4 (1952) 123-129. [1736] H. UMEGAKI, Decomposition theorems of operator algebra and their applications. Japan J. Math. 22 (1952) 27-50. [1737] H. UMEGAKI, Operator algebra of finite class. II, Kodai Math. Sem. Rept, 5 (1953) 61-63. [1738] H. UMEGAKI. Note on irreducible decompositions of a positive linear functional. Kodai Math. Sem. Rept. 6 (1954) 25-32. [1739] H. UMEGAKI, Conditional expectation in an operator algebra, I. Tohoku Math. J. 6 (1954) 177-181. [1740] H. UMEGAKI. Weak compactness in an operator space. Kodai Math. Sem. Rept. 8 (1956) 145-151. [1741] H. UMEGAKI, Conditional expectation in an operator algebra. II. Ttihoku Math. J. 8 (1956) 86-100. [1742] H. UMEGAKI, Conditional expectation in an operator algebra. III, Kodai Math. Sem. Rept. 11 (1959) 51-64. [1743] H. UMEGAKI. On information in operator algebras. Proc. Japan Acad. 37 (\961) 459--461. [1744] H. UMEGAKI, Entropy functional in stationary channels. Proc. Japan Acad. 38 (1962) 668-672. [1745] H. UMEGAKI, Conditional expectation in an operator algebra. IV: Entropy and information. Kodai Math. Sem. Rept. 14 (1962) 59-85. [1746] L. I. V AINERMAN, Characterization of objects dual to locally compact groups, Funkcion. Anal. Priloi. 8 (1974) 75-76 (in Russian). [1747] L. I. VAINERMAN and G. I. KAC, Non-unimodular ring groups and Hopf-Von Neumann algebras, Math. Sb. 136 (1974) 194-225 (in Russian). [1748] A. VAN DAELE, Quasi-equivalence of quasi-free states in the Weyl algebra, Commun. Math. Phys. 21 (1971) 171-191. [1749] A. VAN DAELE, The upper envelope of invariant functionals majorized by an invariant weight, Pacific J. Math. 46 (1973) 283-302. [1750) A. VAN DAELE, A new approach to the Tornita-Takesaki theory of generalized Hilbert algebras, J. Functional Anal. 15 (1974) 378-393. [1751] A. VAN DAELE, A Radon-Nikodym theorem for weights on Von Neumann algebras, to appear.

480

REFERENCES

[1752] A. VAN DAELE and A. VERBEURE, Unitary equivalence of Fock representations on the Weyl algebra, Commun. Math. Phys. 20 (1971) 268-278. [1753] J. VARELA, Duality of C*-algebras, Mem. Amer. Math. Soc. 148 (1974) 97-108. [1754] J. VARELA, Sectional representations of Banach modules, Math. Z. 139 (1974) 55-61. [1755] N. T. V AROPOULOS, Sur les formes positives dune algebre de Banach, C.R. Acad. Sci. Paris 258 (1964) 2465-2467. [1756] F. H. V ASILESCU and L. Zsroo, Uniformly bounded groups in finite W*-a1gebras, Acta. Sci. Math. Szeged 36 (1974) 189-192. [1757] N. B. VASILEV, C*-algebras with finite-dimensional irreducible representations, Uspehi Mat. Nauk 21 (1966) 135-154 (in Russian; English trans!: Russian Math. Surv. 21, 137-155). [1758] A. VERBEURE and R. A. WEDER, Stability in linear response and clustering properties. Commun. Math. Phys. 44 (1975) 101-105. [1759) A. M. VERSIK, Non-measurable decompositions, orbit theory, algebras of opera tors, Dokl. Akad. Nauk SSSR 199 (1971) 1004-1007 (in Russian; English trans!.: Soviet Math. 12 (1971) 1218-1222. [1760] A. M. VERSIK, A geometric theory of states, Von Neumann boundary, duality of C*-algebras, Zap. Naucn. Sem. Leningr. Otdel. Mat. Inst. Steklov. (L.O.M.I.) 29 (1972) 147-154 (in Russian). [1761] J. VESTERSTR0M, Quotients of finite W*-algebras, Bull. Amer. Math. Soc. 77 (1971) 235-238. [1762] J. VESTERSTR0M, On the homomorphic image of the center of a C*-algebra, Math. Scand. 29 (1971) 134-136. [1763] J. VESTERSTR0M, Quotients of finite W*-a1gebras, J. Functional Anal. 9 (1972) 322-335. [1764) J. VESTERSTR0M, On open maps, compact convex sets, and operator algebras, J. London Math. Soc. 6 (1973) 289-297. [1765) J. VESTERSTR0M. Positive linear extension operators for spaces of affine func tions, to appear. [1766] J. VESTERSTR0M and W. WILS. Direct integrals of Hilbert spaces, II. Math. Scand. 26 (1970) 89-102. [1767) I. VIDAV, Sur un systerne daxiomes caracterisant les algebres C*, Glasnik Mat. Fiz. Astron. Drustco 16 (1961) 189-192. [1768] G. F. VINCENT-SMITH, Positive maps between partially ordered bimodules, Proc. London Math. Soc. 19 (1965) 661-674. (1769) D. V. VOICULESCU, Sur les sous-espaces paraferrnes invariants dune algebre de Von Neumann, Bull. Sci. Math. 96 (1972) 161-168. [1770] D. V. VOICULESCU, Sur un theorerne de Misonou, Rev. Roumaine Math. Pures Appl. 17 (1972) 300-310. [1771] D. V. VOICULESCU, Un theoreme du type Weyl-Von Neumann non commutatif, c.s. Acad. Sci. Paris 281 (1975) 735-736. [1772) J. VON NEUMANN, Zur Algebra der Funktionaloperationen und Theorie der normalen Operatoren, Math. Ann. 182 (1929) 370-427. [1773] J. VON NEUMANN, On a certain topology for rings of operators, Ann. Math. 37 (1936) 111-115. (1774) J. VON NEUMANN, On an algebraic generalization of the quantum-mechanical

REFERENCES

481

formalism, I, Mat. Sb. 1 (936) 415-484. [1775] J. VON NEUMANN, Some matrix inequalities and metrization of matrix spaces, Tomsk Univ. Rev. 1 (1937) 286-300. [1776] J. VON NEUMANN. On infinite direct products, Compositio Math. 6 (1938) 1-77. [1777] J. VON NEUMANN. On rings of operators, III. Ann. Math. 41 (1940) 94-161. [1778] J. VON NEUMANN, On some algebraical properties of operator rings. Ann. Math. 44 (1943) 709-715. [1779] J. VON NEUMANN, On rings of operators. Reduction theory, Ann. Math. 50 (1949) 401-485. [1780] B. J. VOWDEN, On the Gelfand-Naimark theorem, J. London Math. Soc. 42 0%7) 725-731. [1781] B. J. VOWDEN. A new proof in the spatial theory of Von Neumann algebras, J. London Math. Soc. 44 (1%9) 429-432. (1782) B. J. VOWDEN. Normalcy in Von Neumann algebras. Proc. London Math. Soc. 27 (1973) 88-100. [1783) B. J. VOWDEN, C*-norms and tensor products of C*-algebras, J. London Math. Soc. 7 (1974) 595-596. [1784] L. J. WALLEN, Semi-groups of partial isometries, Bull. Amer. Math. Soc. 75 (1969) 763-764. [1785) M. E. WALTER, Group duality and isomorphisms of Fourier and Fourier-Stieltjes algebras from a W*-algebra point of view. Bull. Amer. Math. Soc. 76 (1970) 1321-1325. (1786) S. P. WANG. On integrable representations, Math. Z. 147 (1976) 201-203. (1787) S. WASSERMANN, Tensor products of *-automorphisms of C*-algebras, Bull. London Math. Soc. 7 (1975) 65-70. [1788] S. WASSERMANN, Extension of normal functional on W*-tensor products, Math. Proc. Cambridge Phi/os. Soc. 78 (1975) 301-307. [1789] S. WATANABE, Positive linear maps of Banach algebras with an involution, Sci. Rept. Niigata Univ. 8 (1971) 1-6. [1790] A. WElL, Lintegration dans les groupes topologiques et ses applications. 2nd ed . Actualites Sci. Indust. 1145 (Hermann, Paris. 1953). (1791) M. WEINLESS, Existence and uniqueness of the vacuum for linear quantized fields, J. Functional Anal. 4 (1969) 350-379. (1792) H. WIDOM, Embedding in algebras of type I, Duke Math. J. 23 (1956) 309-324. (1793) H. WIDOM, Approximately finite algebras, Trans. Amer. Math. Soc. 83 (1956) 170-178. [1794] H. Wroox, Nonisomorphic approximately finite factors, Proc. Amer. Math. Soc. 8 (1957) 537-540. (1795) A. S. WIGHTMAN. La theorie quantique locale et la theorie quantique des champs, Ann. Inst, H. Poincare 1 (1964) 403-420. (1796) A. WIGHTMAN and S. SCHWEBER, Configuration space methods, Phys. Rev. 98 (1955) 812-837. [1797] I. F. WILDE, Hypercontractivity for fermions, J. Math. Phys. 14 (973) 791-792. (1798) J. J. WILLIAMS, Non-isomorphic tensor products of Von Neumann algebras, Can. J. Math. 26 (1974) 492-512. [1799) J. J. WILLIAMS, Spatial product isomorphisms of tensor products of finite Von Neumann algebras. to appear.

482

REFERENCES

[1800] J. J. WILLIAMS, A representation for infinite tensor products of semi-finite factors, to appear. [1801] J. P. WILLIAMS, Finite operators, Proc. Amer. Math. Soc. 26 (1970) 129-136. [1802] J. P. WILLIAMS, On the range of a derivation, Pacific J. Math. 38 (1971) 273-279. [1803] J. P. WILLIAMS, On the spectrum in the Calkin algebra, to appear. [1804] P. WILLIG, On trace-norms in factors of type II, Commun. Pure App/. Math. 21 (1968) 585-590. [1805] P. WILLIG. Trace norms, global properties and direct integral decompositions of W*-algebras, Commun. Pure Appl. Math. 22 (1969) 839-862. [1806] P. WILLIG, Properties rand L for type II, factors, Proc. Amer. Math. Soc. 2S (1970) 83~37. [1807] P. W.ILLIG, B(H) is very non-commutative. Proc. Amer. Math. Soc. 24 (1970) 204-205. [1808] P. WILLIG, Property P and direct integral decomposition of W*-algebras, Proc. Amer. Math. Soc. 29 (1971) 494-498. [1809] P. WILLIG, Property L and direct integral decompositions of W*-algebras, Proc. Amer. Math. Soc. 30 (1971) 87-91. [1810] P. WILLIG, Type II W*-algebras are not normal, Proc. Amer. Math. Soc. 40 (1973) 115-119. [1811] P. WILLIG, On hyperfinite W*-algebras. Proc. Amer. Math. Soc. 40 (1973) 120-122. [1812] P. WILLIG, Generators and direct integral decompositions of W*-algebras, Tohoku Math. J. 26 (1974) 35-37. [1813] P. WILLIG, Continuous W*-algebras are non-normal, Tohoku Math. J. 26 (1974) 483-486. [1814] P. WILLIG, Central sequences, global properties and direct integral decompositions of W*-algebras, to appear. [1815] W. WILS, Desintegration centrale des formes positives sur les C*-algebres, C.R. Acad. Sci. Paris 267 (1968) 810-812. [1816] W. WILS, Stone-tech compactification and representations of operator algebras, Doctoral Dissertation, Catholic Univ., Nijmegen (1968). [1817] W. WILS. Central decomposition of C*-algebras, Proc. of the Functional Analysis Week (1969) 55-63. [1818] W. WILS, Desintegration centrale dans une partie convexe compacte dun espace localement convexe, CiR. Acad. Sci. Paris 269 (1969) 702-704. [1819] W. WILS, Two-sided ideals in W*-algebras, J. Reine Angew. Math. 24 (1970) 55-68. [1820] W. WILS, Direct integrals of Hilbert spaces, I, Math. Scand. 26 (1970) 73-88. [1821] W. WILS, The ideal center of partially ordered vector spaces, Acta Math. 127 (1971) 41-77. [1822] W. WILS, A remark on the preceding paper by L. Gross. J. Functional Anal. 10 (1972) IIO-IB. [1823] M. WINNINK, An application of C*-algebras to quantum statistical mechanics of systems in equilibrium, Thesis, Univ. of Groningen (1968). [1824] M. WINNINK, Algebraic aspects of the Kubo-Martin-Schwinger condition, Car gese Lecture Notes (1969). [1825] M. WINNINK, Some general properties of thermodynamic states in an algebraic approach, to appear.

REFERENCES

483

[1826] W. R. WOGEN, On generators for Von Neumann algebras, Bull. Amer. Math. Soc. 75 (1969) 95-99. [1827] W. R. WOGEN, On special generators for properly infinite Von Neumann algebras, Proc. Amer. Math. Soc. 28 (1971) 107-113. [1828] W. R. WOGEN, Von Neumann algebras generated by operators similar to normal operators, Pacific J. Math. 37 (1971) 539-543. [1829] P. WOJTASZCZYK, On linear properties of separable conjugate spaces of C* algebras, Studia Math. 52 (1974) 143-147. [1830] J. C. WOLFE, Invariant states and conditional expectations of the anticommutation relations, Commun. Math. Phys. 44 (1975) 53-72. [1831] J. C. WOLFE, Free states and automorphisms of the Clifford algebra, Commun. Math. Phys. 45 (1975) 53-58. [1832] J. C. WOLFE and G. G. EMCH, C*-algebraic formalism for coarse graining. I,

General theory: II. Momentum coarse graining for Fermi systems in infinite volume; III, Momentum coarse graining for Fermi systems in infinite free space, J. Math. Phys. 15 (1974) 1343-1347. 1348-1350, 1351-1365. [1833] G. WOLFSON, The algebra of bounded operators on Hilbert space, Duke Math. J.
20 (1953) 533-538. [1834] P. K. WONG. Continuous complementors on B*-algebras. Pacific J. Math. 33 (1970) 255-260. [1835] P. K. WONG. The Arens product and duality in B*-algebras. II, Proc. Amer. Math. Soc. 27 (1971) 535-538. [1836] P. K. WONG. *-actions in A*-algebras, Pacific J. Math. 44 (1973) 775-779. [1837] P. K. WONG. The p-class in a dual B*-algebra, Trans. Amer. Math. Soc. 200 (1974) 355-368. [1838] E. J. WOODS, The classification of factors is not smooth. Can. J. Math. 25 (1973) 96-102. [1839] S. L. WORONOWICZ. On the purification of factor states, Commun. Math. Phys. 28 (1972) 221-235. [1840] S. L. WORONOWICZ, On the purification maps, Commun. Math. Phys. 30 (1973) 55--f,7. [1841] F. B. WRIGHT, A reduction for algebras of finite type, Ann. Math. 66 (1954) 560-570. [1842] F.. B. WRIGHT, The ideals in a factor, Ann. Math. 68 (1958) 475-483. [1843] J. D. M. WRIGHT, An extension theorem and a dual proof of a theorem of Gleason, J. London Math. Soc. 43 (1968) 699-702. [1844] J. D. M. WRIGHT, Order completeness and topological completeness, J. London Math. Soc. 1 (1969) 444-446, [1844] J. D. M. WRIGHT, Every monotone e-complete C*-algebrais the quotient of its Baire *-envelope by a two-sided IT-ideal, J. London Math. Soc. 6 (1973) 210-214. [1846] J. D. M. WRIGHT, On minimal o-completions of C*-algebras, Bull. London Math. Soc. 6 (1974) 168-174. [1847] A. WULFSOHN, Produit tensoriel de C*-algebres, Bull. Sci. Math. 87 (1963) 13-21. [1848] A. WULFSOHN, Le produit tensoriel de certaines C*-algebres, CR. Acad. Sci. Paris 258 (1964) 6052-6054. [1849] A. WULFSOHN, The primitive spectrum of a tensor product of C*-algebras, Proc. Amer. Math. Soc. 19 (1968) 1094-1096.

484

REFERENCES

[1850] A WULFSOHN, Von Neumann factors are perfect, Proc. Amer. Math. Soc. 41 (1973) 189-192. [1851] F. J. YEADON, A new proof of the existence of a trace in a finite Von Neumann algebra, Bull. Amer. Math. Soc. 77 (1971) 257-260. [1852] F. J. YEADON, Convergence of measurable operators, Proc. Cambridge Philos, Soc. 74. (1973) 257-268. [1853] F. J. YEADON, A note on the Mackey topology of a Von Neumann algebra, J. Math. Anal. Appl. 45 (1974) 721-722. [1854] F. J. YEADON, On a result of P. G. Dixon, J. London Math. Soc. 9 (1975) 610-612. [1855] F. J. YEADON, Non-commutative U-spaces, Math. Proc. Cambridge Philos, Soc. 77 (1975) 91-102. [1856] K. YLINEN, Compact and finite-dimensional elements of normed algebras, Ann. Acad. Sci. Fenn. (AI) 428 (1968) 37 pp, [1857] K. YLINEN, A note on the compact elements of C*-a1gebras, Proc. Amer. Math. Soc. 35 (1972) 305-308. [1858] K. YLINEN, Vector space isomorphisms of C*-algebras, Studia Math. 46 (1973)31-34. [1859] K. YLINEN, Dual C*-algebras,weakly semi-completely continuous elements, and the extreme rays of the positive cone, Ann. Acad. Sci. Fenn. (A) 599 (1975) 1-9. [1860] K. YLINEN, Weakly completely continuous elements of C*-algebras, Proc. Amer. Math. Soc. 52 (1975) 323-326. [1861] B. YOOD, Faithful *-representations of normed algebras, Pacific J. Math. 10 (1960) 345-363. [1862] B. YOOD, On Kadisons condition for extreme points of the unit ball in a B*-algebra, Proc. Edinburgh Math. Soc. 16 (1968) 245-250. [1863] B. YOOD, On algebras which are pre-Hilbert algebras, Duke Math. J. 36 (1969) 261-271. [1864] B. YOOD, On axioms for B*-algebras, Bull. Amer. Math. Soc. 76 (1970) 80-82. [1865] B. YOOD, Hilbert algebras as topological algebras, Ark. Mat. 12 (1974) 131-151. [1866] T. YOSHIMO, Nearly normal operators, T6hoku Math. J. 20 (1968) 1-4. [1867] H. YOSHIZAWA, On some types of convergence of positive definite functions, Osaka Math. J. I (1949) 90-94. [1868] H. YOSHIZAWA, Some remarks on unitary representations of the free group, Osaka Math. J. 3 (1951) 55--63. [1869] H. YOSHIZAWA, A proof of the Plancherel theorem, Proc. Japan Acad. 30 (1954) 276-281. [1870] K. YOSIDA, On the unitary equivalence in general Euclid space, Proc. Japan Acad. 22 (1946) 242-245. [1871] Y. YUEN, Groups of invertible elements of Banach algebras, Bull. Amer. Math. Soc. 79 (1973) 82-84. [1872] W. W. ZACHARY, Non-relativistic time-dependent scattering theory and Von Neumann algebras, I: Single channel scattering, J. Math. Phys. 13 (1972) 609--615. [1873] G. ZELLER-MEIER, Produits croises dune C*-algebre par un groupe dautomor- . phismes, c.u. Acad. Sci. Paris 263 (1966) 20-23. [1874] G. ZELLER-MEIER, Representations fideles des produits croises, C.R. Acad. Sci. Paris 264 (1967) 679--682. [1875] G. ZELLER-MEIER, Sur les automorphismes des algebres de Banach, C.R. Acad. Sci. Paris 264 (1967) 1131-1132.

REFERENCES

485

(1876) G. ZELLER-MEIER, Produits croises dune C*-algebre par un groupe dautomor phismes, J. Math. Pures Appl. 47 (1%8) 101-239. (1877) G. ZELLER-MEIER, Deux autres facteurs de type II .. Invent. Math. 7 (1%9) 235-242. (1878) J. ZEMANEK, A remark on transitivity of operator algebras, Casopis Pest. Mat. 100 (1975) 176-178. [1879) L. ZSIDO, I. Cuculescus proof for the commutant theorem, in: Hilbert Space Operators, Col/oq. Math. Soc. Janos Bolyai, Vol. 5 (North-Holland, Amsterdam, 1970) 539-542. (1880) L. ZSIDO, The norm of a derivation in a W*-algebra, Proc. Amer. Math. Soc. 38 (1973) 147-150. [1881) L. ZSIDO. Note on Dixrniers trace type sets in properly infinite W*-algebras, Rev. Roumaine Math. Pures Appl. 19 (1974) 269-274. [1882) L. ZSIDO, The Weyl-Von Neumann theorem in semi-finite factors, J. Functional Anal. 18 (1975) 60-72. [1883) L. ZSIDO. A proof of Tomitas fundamental theorem in the theory of standard Von Neumann algebras, Rev. Roumaine Math. Pures Appl. 20 (1975) 609-619.

SUPPLEMENT ARY BIBLIOGRAPHY

[1884] M. B. ABRAHAMSE and R. G. DoUGLAS. A class of subnormal operators related to multiply connected domains. Advances in Math. 19 (1976) 106-148. [1885] C. A. AKEMANN .and P. A. OSTRAND, Computing norms in group C*-algebras. Amer. J. Math. 98 (1976) 1015-1047. [1886] C. A. AKEMANN and G. K. PEDERSEN, Ideal perturbations of elements in C*-algebras, to appear. [1887] L. AKKARDI. The non commutative Markov property (in Russian), Funkcion, Anal. Priloien 9 (1975) 1-8. [1888] V. V. ANSELEVIC. The central limit theorem for a "non commutative" stationary random process (in Russian) Uspehi Mat. Nauk 28 (1973) 227-228. [1889] C. ApOSTOL, Inner derivations with closed range. Rev. Roumaille Math. PI/res Appl. 21 (1976) 249-265. [1890] H. ARAKI. R. HAAG. D. KASTLER and M. TAKESAKI, Extension of KMS states and chemical potentials. to appear, [1891] H. ARAKI and A. KISHIMOTO. Symmetry and equilibrium states, to appear, [1892] H. ARAKI and A. KISHIMOTO. On clustering property, to appear. [1893] H. ARAKI and G. L. SEWELL. KMS conditions and local thermodynamical stability of quantum lattice systems. to appear. [1894] R. J. ARCHBOLD, On commuting C*-algebras of operators. Math. Scand. 38 (1976) 106-114. [1895] W. ARVESON. Interpolation problems in nest algebras. J. Functional Anal. 20 (1975) 208-233. [1896] V. A. ARZUMANJAN. Factor-states on crossed products that are constructed in terms of a dynamical system (in Russian) lzv. Akad. Nauk Armjall. SSR Sa. Mat. 10 (1975) 255-263 and 290. [1897] M. F. ATiYAH. Elliptic operators, discrete groups and Von Neumann algebras. Asterisque 32-33 (1976) 43-72. [1898] P. L. AUBERT, Theorie de Galois pour une W*-algebre. Comment. Math. Helv. 51 (1976) 411-433. [1899] B. A. BARNES. The similitary problem for representatiors of a B*-a1gebra. Mi chigan Math. J. 22 (1975) 25-32. [1900] B. A. BARNES. On this operators relative to an ideal in a Von Neumann algebra, Acta Sci. Math. 38 (1976) 265-274. [1901] C. J. K. BATTY, On relative commutants in tensor products of C*-algebras, Math. Z. 151 (1976) 215-218. [1902] S. K. BERBERIAN, The character space of the algebra of regulated functions, to appear. [1903] A. BERNARD. Algebres non autoadjointes de champs continus doperateurs, in: Lecture Notes in Math. 336 (Springer. Berlin, 1973). [1904] J. J. BISOGNANO and E. H. WICHMANN. On the duality condition for quantum fields. J. Math. Phys. 17 (1976) 303-321.

SUPPLEMENTARY BIBLIOGRAPHY

487

[1905] O. BJ0RNESTAD, On factors of type II and quantum mechanics, He/v. Phys. Acta 46 (1973) 371-387. [1906] B. E. BLACKADAR, Infinite tensor products of C*-algebras, to appear. [1907] W. Bos, Direct integrals of selfdual cones and standard forms of Von Neumann algebras. Invent. Math. 37 (1976) 241-251. [1908] O. BRATTELI, A non-simple crossed product of a simple C*-algebra by a properly outer automorphic action. to appear. [1909] O. BRATTELI. Structure spaces of approximately finite-dimensional C*-algebras, II, to appear. [1910] O. BRATTELI, R. H. HERMAN and D. W. ROBINSON, Perturbations of flows on Banach spaces and operator algebras, to appear. [1911] O. BRATTELI and D. W. ROBINSON, Greens functions, Hamiltonians and modular automorphisms, Commun. Math. Phys. 50 (1976) 133-156. [1912] M. BREUER and R. S. BUTCHER, Fredholm theories of mixed type with analytic index functions, Math. Ann. 209 (1974) 31-42. [1913] G. T. BRINKE and M. WINNINK, Spectra of Liouville operators, to appear. [1914] L. G. BROWN, Operator algebras and algebraic K-theory, Bull. Amer. Math. Soc. 81 (1975) 1119-1121. [1915] L. G. BROWN, Stable isomorphisms of hereditary subalgebras of C*-algebras, to appear. [1916] J. W. BUNCE, Finite operators and amenable C*-algebras, Proc. Amer. Math. Soc. 56 (1976) 145-151. [1917] W. M. CHING and C. S. TSAU, A family of type III factors without property L, Math. Ann. 222 (1976) 195-204. [1918] M. D. CHOI and E. G. EFFROS, The completely positive lifting problem, to appear. [1919] M. D. CHOI and E. G. EFFROS. Lifting problem and the cohomology of C* algebras, to appear. [1920] E. CHRISTENSEN, G. A. ELLIOTT and G. K. PEDERSEN, Separability and hereditary subalgebras, to appear. [1921] A. CONNES, On the cohomology of operator algebras, to appear. [1922] J. B. CONWAY, On the Calkin algebra and the covering homotopy property, II, to appear. [1923] J. C. CORTET, Etats G-factoriels, c.s. Acad. Sci. Paris 275 (1972) 563-566. [1924] M. J. CRABB, J. DUNCAN and C. M. MCGREGOR, Characterizations of com mutativity for C*-algebras, Glasgow Math. J. 15 (1974) 172-175. [1925] J. CUNTZ, Eine Klasse von postliminalen gewichteten Schiftoperatoren, Arch. Math. 27 (1976) 188-198. [1926] J. CUNTZ, The structure of multiplication and addition in simple C*-algebras, to appear. [1927] P. DE LA HARPE and M. KAROUBI, Perturbations compactes des representations dun groupe dans un espace de Hilbert, Bull. Soc. Math. France 46 (1976) 41-65. [1928] D. PYO CHI, Derivations in C*-algebras, to appear. [1929] R. G. DOUGLAS, Banach algebra techniques in operator theory, in: Pure and Applied Mathematics 49 (Academic Press, New York, 1972). [1930] R. G. DOUGLAS, Banach algebra techniques in the theory of Toeplitz operators, Expository lectures from the CBMS Regional Conference held at the University of Georgia. AMS (1973).

488

SUPPLEMENTARY BIBLIOGRAPHY

[1931] K. E. EKMAN. Unitaries and partial isometries in a real W*-algebra, Proc. Amer. Math. Soc. 54 (1976) 138-140. [1932] G. A. ELLIOTT. Some C*-algebras with outer derivations. III, to appear. [1933] G. A. ELLIOTT. On approximately finite dimensional Von Neumann algebras, Math. Scand. 39 (1976) 91-101. [1934] G. A. ELLIOTT and E. J. WOODS. The equivalence of various definitions for a properly infinite Von Neumann algebra to be approximately finite dimensional, Proc. Amer. Math. Soc. 60 (1976) 175-178. [1935] M. ENOCKH and J. M. SCHWARTZ, Algebres de Kac et produits croises, C. R. Acad. Sci. Paris 283 (1976) 321-323. [1936] D. E. EVANS, Positive linear maps on operator algebras. Commun. Math. Phys. 48 (1976) 15-22. [1937] D.E. EVANS. Scattering in the CAR agebra, Cornmllll. Math. Phys. 48 (1976) 23-30. [1938] D. E. EVANS and J. T. LEWIS. Dilations of dynamical semi-groups. Commun. Math. Phvs. 50 (1976) 219-227. [1939] D. E. EVANS and T. SUNG. Spectral subspaces for compact actions, to appear. [1940] M. FANNES. Quasi-free states and automorphisms of the CCR-algebra, COl1lmun. Math. Phys. 51 (1976) 55---66. [1941] J. FELDMAN and C. C. MOORE. Ergodic equivalence relations. cohomology and Von Neumann algebras, II, to appear. [1942] M. FUJI and M. NAKAMURA. A comment on the inequalities of Arveson-Parrott and Mlak, Math. Japon. 20 (1975) 65---67. [1943] G. GALLAVOTTI and M. PULVIRENTI, Classical KMS condition and Tomita Takassaki theory. Commun. Math. Phys. 46 (1976) 1-10. [1944] R. M. GILLETTE and D. C. TAYLOR, A characterization of the Pedersen ideal of Co(T. Bo(H)) and counterexample. to appear. [1945] V. JA GOLODEC, Reduction of type III algebras, Funkcion. Anal. Priloien 9 (1975) 1-7. [1946] V. JA GOLODEC. Structure of Von Neumann algebras dual to those constructed according to dynamic systems, Funkcion. Anal. Priloien 9 (1975) 87-88. [1947] V. JA GOLODEC, Spectral properties of modular operators. and the asymptotic ratio set (in Russian) Izr. Akad. Nauk SSSR Ser. Mat. 39 (1975) 635-656 and 704. [1948] E. C. GOOTMAN and D. KANNAN. Zero one laws in finite W*-algebras. J. Math. Anal. Appl. 55 (1976) 743-756. [1949] P. GREEN, C*-algebras of transformation groups with smooth orbit space. to appear. [195:)] P. GREEN. The local structures of twisted covariance algebras. to appear. [1951] W. L. GREEN. Compact groups of automorphisms of Von Neumann algebras, Math. Scand. 37 (1975) 284-296. [1952] W. A. GREENE. Ambrose modules. Mem. Amer. Math. Soc. 148 (1974) 109-133. [1953] H. HALPERN. Normal expectations and integral decomposition of type III Von Neumann algebras, to appear. [1954] F. HANSEN. The cohomology of R is trivial, to appear. [1955] F. HANSEN and D. OLESEN, Perturbations of centre-fixing dynamical systems. to appear. [1956] J. F. HAVET, Esperances conditionnelles permutables a un groupe dautomor-

SUPPLEMENTARY

BIBLIOGRAPHY

489

(1957) (1958) (1959) (1960) [1961] (1962) [1%3] [1964] [1965] [1966] (1967) [1968] [1969] (1970) (1971) [1972] [1973] [1974] (1975) [1976] [1977] [1978] [1979] [1980]

phismes sur une algebre de Von Neumann, CiR. Acad. Sci. Paris 282 (1976) 1095-1098. J. F. HAVET, Quelques proprietes des systernes dynamiques non commutatifs, to appear. 1. N. HERSTEIN, On the multiplicative group of a Banach algebra, in: Symposia Math. VIII, (Academic Press. London, 1972) 227-232. A. IGARASHI, Ideals in a C*-algebra, Bull. Electrotech. Lab. 39 (1975) 399-414. V. ISTRATESCU, 1. ISTRATESCU and GH. CONSTANTIN, On Weyl spectrum of an operator, II, Math. Balkanica 4 (1974) 277-281. V. P. JASNIKOV, Harmonic functions connected with finite Von Neumann algebras (in Russian) Uspehi Mat. Nauk 26 (1971) 229-230. R. E. KALLMAN, Certain quotient spaces are countably separated, J. Functional Anal. 21 (1976) 52-62. D. KASTLER. Stability and equilibrium in quantum statistical mechanics, in: Lec ture Notes in Math. 466 (Springer, Berlin, 1975) 86-100. S. V. KEROV, Duality of finite-dimensional -algebras (in Russian) Vestnik Lenin grad Univ. 7 (1974) 23-29. A. KISHIMOTO, Dissipations and derivations, Commun. Math. Phys. 47 (1976) 25-32. A. KISHIMOTO, On uniqueness of KMS states of one-dimensional quantum lattice systems, Commun. Math. Phys. 47 (1976) 167-170. 1. Kov ACS and GH. MOCANU, Unitary dilations and C*-algebras, Acta. Sci. Math. 38 (1976) 79-82. F. KUBO, On theorems of Phelps, Russo and Dye, Math. Japan. 20 (1975) 69-71. K. KURIJAMA, On left Hilbert algebras with respect to Minkowsky forms, Mem. Fac. Sci. Kyushu Univ. SeT. A, 30 (1976) 103-111. E. C. LANCE, Tensor products of non-unital C*-algebras, J. London Math. Soc. 12 (1976) 160-168. E. C. LANCE, Refinement of direct integral decompositions, Bull. London Math. Soc. 8 (1976) 49-56. E. C. LANCE, A strong non commutative ergodic theorem, Bull. Amer. Math. Soc. 82 (1976) 925-926. E. C. LANCE, Ergodic theorems for convex sets and operator algebras, Invent. Math. 37 (1976) 201-214. M. B. LANDSTAD, Duality for dual covariance algebras, to appear. A. LAU, W*-algebras and invariant functionals. to appear. A. J. LAZAR and D. C. TAYLOR, Multipliers of Pedersens ideal, Mem. Amer. Math. Soc. 169 (1976). P. LEGISA, Well embedded Hilbert subspaces in C*-algebras, Studia Math. 69 (1976) 37-40. A. LIEBERMAN, Spectral distribution of the sum of self-adjoint operators, Pacific J. Math. S3 (1974) 211-216. G. LINDBLAD, On the generatorsof quantum dynamical sernigroups, Commun. Math. Phys.48 (1976) 119-130. A. A. LODKIN, A lemma on the approximation of finite diemensional *-algebras (in Russian) Zap. Naucn. Sem. Leningrad Otdel. Math. Inst. Steklov 47 (1974) 175-178. 190 and 194-195.

490

SUPPLEMENTARY BIBLIOGRAPHY

[1981] R. I. LOEBL, Contractive linear maps on C*-algebras, Michigan Math. J. 22 (1976) 361-366. [1982] A. I. LOGINOV and V. S. SULMAN, Hereditary and intermediate reflexivity of W*-algebras, Izv. Akad. Nauk SSSR 39 (1975) 1260-1273. [1983] H. LOHOFER, Verallgemeinerte Doppelverhiiltnisse relativ Von Neumannscher Algebren, Dissertation, Philipps-Universitat Marburg (1975). [1984] R. LONGO, On perturbed derivations of C*-algebras, to appear. [1985] L. E. LUNDBERG, Quasi-free "second quantization". Commun. Math. Phys. 50 (1976) 103-112. [1986] J. MAURER. Positive lineare Abbildungen auf der n-dimensionalen Matrizenalge bra, Arch. Math. 28 (1977) 193-199. [1987] R. J. McGovERN. Quasi-free derivations on the canonical anticommutation relation algebra, to appear. [1988] C. R. MIERS, Lie * triple homomorphisms into Von Neumann algebras, Proc. Amer. Math. Soc. 58 (1976) 169-172. [1989] C. C. MOORE. Group extension and cohomology for locally compact groups IV, Trails. Amer. Math. Soc. 221 (1976) 35-58. [1990] R. D. MOSAK, Banach algebras, in: Chicago Lectures in Mathematics (The University of Chicago Press, Chicago, 1975). [1991] K. NAPIORKOWSKI and W. Pusz, Particle representations of canonical com mutation relations, Rep. Math. Phys. 3 (1972) 221-225. [1992] J. NAUDTS and A. VERBEURE, Bounds on the admittance for KMS states, J. Math. Phys. 17 (1976) 419--423. [1993] V. N. Nosov, Invariant S, weights, and states on W*-algebras (in Russian) Funk cion, Anal. Priloien 9 (1975) 85-86. [1994] D. P. ODONOVAN, Weighted shifts and covariance algebras, Trans. Amer. Math. Soc. 208 (1975) 1-25. [1995] D. P. ODONDY AN, Bounded matrices and some algebras of operators, Proc. Amer. Math. Soc. 58 (1976) 134-138. [1996] D. OLESEN and G. K. PEDERSEN, Applications of the Connes spectrum to C*-dynamical systems. to appear. [1997] D. OLESEN, G. K. PEDERSEN and E. STORMER, Compact abelian groups of automorphisms of simple C*-algebras, to appear. [1998] C. L. OLSEN and W. R. ZAME, Some C*-algebras with a single generator, Trans. Amer. Math. Soc. 215 (1976) 205-217. [1999] S. OTA, A certain operator algebra in an indefinite inner product space, Mem, Fac. Sci. Kyushu. Univ. Ser. A 29 (1975) 203-210. [2000] W. L. PASCHKE, Inner product modules arising from compact automorphism groups of Von Neumann algebras, Trans. Amer. Math. Soc. 224 (1976) 87-102. [2001] C. PEARCY and N. SALINAS, Extensions of C*-algebras and the reducing essential matricial spectra of an operator, to appear. [2002] G. K. PEDERSEN, Spectral formulas in quotient C*-algebras, Math. Z. 148 (1976) 299-300. [2003] G. K. PEDERSEN, Groupes localement compacts dautornorphisrnes dune C* algebre et conditions spectrales, to appear. [2004] G. K. PEDERSEN, An introduction to C*-algebra theory, to appear.

SUPPLEMENTARY

BIBLIOGRAPHY

491

[2005] C. PELIGRAD, Invariant subspaces of Von Neumann algebras, Acta. Sci. Math. 37 (1975) 273-277. [2006] C. PELIGRAD, Remarks on a theorem by R. G. Douglas and C. Pearcy, Rev. Roumaine Math. Pures Appl. 20 (1975) 359-361. [2007] V. J. PELLEGRINI, Numerical range preserving operators on a Banach algebra, Studia Math. 54 (1975) 143-147. [2008] G. PISIER, Grothendiecks theorem for non-commutative C*-algebras with an appendix on Grothendiecks constants, to appear. [2009] S. C. POWER, C-modules and an odd-even decomposition for C-algebras, Bull. London Math. Soc. 8 (1976) 268-272. [2010] R.T. POWERS, Resistance inequalities for KMS states of the isotropic Heisenberg model, Commun. Math. Phys. 51 (1976) 151-156. [2011] W. M. PRIESTLEY, A noncom mutative Korovkin theorem, J. Approximation Theory 16 (1976) 251-260. [2012] C. RADIN, Ergodicity in Von Neumann algebras, Pacific J. Math. 48 (1973) 235-239. [2013] M. REYNOLDS, Perturbations of groups of automorphisms of Von Neumann algebras, Proc. Amer. Math. Soc. 55 (1976) 326-328. [2014] M. A. RIEFFEL, Strong Morita equivalence of certain transformation group C* algebras, Math. Ann. 222 (1976) 7-22. [2015] J. E. ROBERTS. Local cohomology and superselection structure. Commun. Math. Phys. 51 (1976) 107-119. [2016] D. W. ROBINSON, The approximation of 1I0ws, J. Functional Anal. 24 (1977) 280-290. [2017] S. SAKAI, Commutative normal *-derivations, II, J. Functional Anal. 21 (1976) 203-208. [2018] S. SAKAI, Commutative normal .-derivations, III, Tohoku Math. J. 28 (1976) 583-590. [2019] J. L. SAUVAGEOT, Image dun homomorphisme et 1I0t des poids dune relation dequivalence mesuree, C.R. Acad. Sci. Paris 282 (1976) 619-622. [2020] J. A. SCHOEN, Factor representations and factor states on a C*-algebra, Can. J. Math. 28 (1976) 130-134. [2021] J. M. SCHWARTZ, Relations entre "ring-groups" et algebres de Kac, Bull. Sci. Math. 100 (1976) 289-300. [2022] Z. SEBESTYEN, Remarks on the paper of H. Araki and G. A. Elliott "On the definition of C*-algebras", Ann. Univ. Sci. Budapest Eotvos Sect. Math. 17 (1974) 35-39. [2023] A. N. SERSTNEV, On the general theory of states on Von Neumann algebras (in Russian) Funkcion. Anal. Priloien. 8 (1974) 89-90. [2024] A. M. SINCLAIR, Automatic continuity of linear operators, in: London Math. Soc., Lecture Note Series 21 (Cambridge Univ, Press, London, 1976). [2025] J. P. SPROSTON, Derivations and automorphisms of homogeneous C-algebras, Proc. London Math. Soc. 32 (1976) 521-536. [2026] P. J. STACEY, Admissible split faces in the state space of a separable C*-algebra, Quarterly J. Math. 26 (1975) 485-490. [2027] S. STRATI LA, D. vorcur.sscu and L. ZSIDO, Sur les produits croises, c.R. Acad. Sci. Paris. 282 (1976) 779-782.

492

SUPPLEMENTARY BIBLIOGRAPHY

[2028] S. STRATI LA, D. VOICULESCU and L. ZSIDO, On crossed products. to appear. [2029] S. T AKAHASt, On Urysohns lemma and division of the set of pure states, Bull. Fac. Sci. lbaraki Univ. 7 (1975) 61-66. [2030] S. T AKAHASI, A simple proof of the Stone-Weierstrass theorem for CCR-algebras with Hausdorff spectrum, Math. Scand. 38 (1976) 304-306. [2031] H. TAKEHANA. An automorphism on tensor products of operator algebras. Math. Japan 20 (1975) 99-100. [2032] H. TAKEMOTO, On the weakly continuous constant field of Hilbert space and its application to the reduction theory of Von Neumann algebra. Tohoku Math. J. 28 (1976) 479-496. [2033] D. C. TAYLOR, The strong bidual of F(K), to appear. [2034] S. TELEMAN. On reduction theory, Rev. Roumaine Math. Pures Appl. 21 (1976) 465-486. [2035] F. J. THAYER. Obstructions to lifting -morphlsms into the Calkin algebra. Illinois J. Math. 20 (1976) 322-328. [2036] J. TOMIY AMA and K. YABUTA. Toeplitz operators for uniform algebras, to appear. [2037] J. VARELA, Fields of automorphisms and derivations of C*-algebras, Pacific J. Math. 59 (1975) 277-294. [2038] D. VOICULESCU, A non-commutative Weyl-Von Neumann theorem, Rev. Rou maine Math. Pures Appl. 21 (1976) 97-117. [2039] S. P. WANG, On algebras of operators of finite rank, to appear. [2040] S. WASSERMANN, The slice map problem for C*-algebras. Proc. London Math. Soc. 32 (1976) 537-559. [2041] S. WASSERMANN, On tensor products of certain group C*-algebras, J. Functional Anal. 23 (1976) 239-254. [2042] J. P. WILLIAMS, Diagonalizable normal operators, Proc. Amer. Math. Soc. 54 (1976) 106-108. [2043] P. K. WONG. A note on annihilator and complemented Banach algebras, J. Austral. Math. Soc. 18 (1974) 474-481. [2044] S. L. WORONOWICZ, Nonextendible positive maps, Commun. Math. Phvs, 51 (1976) 243-282. [2045] J. D. M. WRIGHT, On semifinite AW*-algebras, Math. Proc. Camb. Phil. Soc. 79 (1975) 443-445. [2046] J. D. M. WRIGHT, Wild AW*-factors and Kaplansky-Richart algebras, J. London Math. Soc. 13 (1976) 83-89. [2047] J. D. M. WRIGHT. On Von Neumann algebras whose pure states are separable, J. London Math. Soc. 12 (1976) 385-388. [2048] J. D. M. WRIGHT. Regular u-completions of C*-algebras, J. London Math. Soc. 12 (1976) 299-309. [2049] J. D. M. WRIGHT, On AW*-algebras of finite type, J. London Math. Soc. 12 (1976) 431-439. [2050] T. YOSHINO. Hyponormal operators in Von Neumann algebras, Tohoku Math. J. 28 (1976) 129-134. [2051] L. ZSIDO, Topological decompositions of W*-algebras, I-II (in Romanian) Stud. Cere. Math. 25 (1973) 859-945 and 1037-1112. [2052] V. A. ZUK and G. I. KAC, Algebras with canonical generators and the structure of C*-algebras of metabelian Lie groups (in Russian) Dokl. Acad. Nauk SSSR 229 (1976) 1313-1316.