• Aucun résultat trouvé

IIl-factor. If

N/A
N/A
Protected

Academic year: 2022

Partager "IIl-factor. If "

Copied!
6
0
0

Texte intégral

(1)

Hyperfinitc product factors

EI~LING STORMEtr University of Oslo, Norway

1. Introduction

I t is an open question whether all hyperfinite factors are *-isomorphic to factors obtained as the infinite tensor product of finite t y p e I factors. I n order to s t u d y this problem it is neeessary to have criteria which tell us when a hyperfinite factor is *-isomorphic to such a product faetor. The present paper is devoted to a result of this kind, the criterion being t h a t all, or equivalently, just one normal state is in a sense asymptotically a product state. This result is an intrinsic characterization of product factors in t h a t it is independent of a n y weakly dense U g F - a l g e b r a and also of a n y tensor product factorization of the underlying g i l b e r t space.

We first recall some terminology. A U g F - a l g e b r a is a C*-algebra 9~ with i d e n t i t y I in which there is an increasing sequence of Ins-factors Mn~ containing I such t h a t nl--> ~ and L l ~ l Mn~ is uniformly dense in 9~, see [2]. A factor

~R is said to be hyperfinite if there is a UHF-algebra which is weakly dense in ~ . More specially ~ is said to be an I T P F I - f a e t o r (infinite tensor product of finite t y p e I factors) if there exists an infinite sequence of In~-factors Mnl with n~ ~_ 2 for an infinite number of i's, and a product state co = | of the C*-algebraic tensor product 9~ --~ | Mni, such t h a t ~ equals the weak closure of z~(9~), where s~ is the representation of ?I induced by co. I t was shown by Murray a n d yon Neumann, see [1, Th6or~me 3, p. 280], t h a t all hyperfinite IIl-factors are *-isomor- phic, and hence *-isomorphic to ITPFI-faetors. I t is not known whether all hyper- finite factors of types IIoo or I I I are *-isomorphic to ITPFI-factors. We refer the reader to the book of Dixmier [1] for the theory of yon N e u m a n n algebras and to the paper of Guichardet [3] for t h a t of infinite tensor products.

The author is indebted to J. Tomiyama for pointing out a gap in an early version of the paper. I n this version there was also a rather long proof of the implication

(2)

166 AI~KIV FOR ]YLA_TEMATIK. Vol. 9 NO. 1

(iii) -+ (i) in the theorem. I am grateful to G. Elliott, O. A. Nielsen, and J. Woods for communicating to me the short proof of this implication, which t h e y have kindly let me publish here. G. Elliott has also pointed ou~ t h a t the theorem can be generalized from hyperfinite factors to factors generated by a commuting (not necessarily countable) family of finite t y p e I subfactors.

2. Product factors

Definition. L e t !}t be a factor and w a normal state of !}1. We say co is asymptotically a product state if given e > 0 and a t y p e /n-factor M C ~ then we can find a t y p e /m-factor 5[ ---- N ( s , M ) such t h a t M c iV C !}t a n d such t h a t

11o-, - N | N"II <

where 5[c ~ iV, rl ~ , and we i d e n t i f y ~R a n d N | N ~ . ~ is said t o be a pro- duct factor if every normal state of ~ is asymptotically a product state.

We have here implicitly assumed t h a t both M and N contain the identity I of ~ . This will always be done when we write 9~ C ~ for two C*-algebras and ~5 with identities.

l~emarlc. The property of being a product factor is a *-isomorphic invariant.

Indeed, if 9t is a product factor and cr is a *-isomorphism of ~ onto a factor

~J~ t h e n the dual map of cr carries the normal states of ~ onto those of ~ , hence t h e y are all asymptotically product states.

THEOREM. Let ~ be a hyperfinite factor. Then the following three conditions are equivalent:

(i) ~ is a product factor.

(ii) There is a normal state on ~ which is asymptotically a product state.

(iii) ~ is *-isomorphic to an 1TPFl-factor.

We first prove two lemmas both of which are probably well known.

L]~MM)~ 1. Let ~ be a countably decomposable infinite factor or the hyperfinite

IIl-factor. If

5[ is a type In-factor contained in ~ then ~ _~ ~ | 5[.

Proof. I f ~ is the hyperfinite IIl-factor then ~ @ 5[ is also hyperfinite and of t y p e II1, hence ~ ~ ~ @ 5[ by the isomorphism theorem for such factors [1, Th6or~me 3, p. 280]. Assume ~ is an infinite factor. L e t {e~ i : i , j ~ 1 , . . . , n}

be a complete set of matrix units in 5[. Then the projection e n is infinite, hence equivalent to the identity, since ~ is countably decomposable. Thus ~ ~--- e n ~ e l r Clearly ~ e n ~ e n @ N . l~ence ~ - ~ @ 5 [ .

L]~MA 2..Let ~ be a hyperfinite factor and N a type In-factor contained in 9l. Let N ~ - - 5['fl ~ . Then N ~ is a hyperfinite factor isomorphic to Ot, and { N u =

Proof. B y L e m m a I t h e r e is a *-isomorphism ~ of ~ onto ~R | N. L e t M --- cr | N), where we write I | N for t h e s e t of:operators I | x , x E N.

(3)

H Y P E R F I N I T E P R O D U C T XT'ACTORS 167 T h e n M is a t y p e I~-factor c o n t a i n e d in 2 , so t h e r e is a u n i t a r y o p e r a t o r u in such t h a t u N u - l = M [4, L e m m a 3.3]. L e t f i ( x ) = o~(uxu -1) for x E 2 . T h e n fl is a *-isomorphism o f ~ o n t o ~ | such t h a t f i ( N ) = o ~ ( M ) = I | T h u s fl(N c) = (I | N) ~ = (I | N)' fl (~R | N) = (~3(9(~) | I) [3 (!R | N) =

| I ~ 91, where c)~ is t h e u n d e r l y i n g H i l b e r t space, a n d ~(9g) all b o u n d e d operators on 9g. Therefore No__ ~ ~R. F i n a l l y we h a v e

{N O ~Ve} u =/~-l({fi(~V) U/~(NC)} ") = fl-l({(i Q _~) U (~}~ @ i)}0) =/~-1(~}{ @ / ~ ) = ~}~, Proof of theorem. W e show (i) - + (ii) --> (iii) - + (i). Clearly (i) --> (ii). W e first show (ii) --> (iii).

L e t ~ be a U H F - a l g e b r a which is w e a k l y dense in 2 , s a y 2 is t h e n o r m closure o f [ J ~ I M , , of a n increasing sequence o f t y p e I,,-factors M ~ . L e t ~o be a n o r m a l s t a t e o f ~ which is a s y m p t o t i c a l l y a p r o d u c t state. :Let {e~}~>_l be a u n i f o r m l y dense sequence o f operators in t h e u n i t ball o f [J~~ M~,. W e shall b y i n d u c t i o n c o n s t r u c t a sequence {N~} o f t y p e /=z-factors c o n t a i n e d in ~ such t h a t

N ~ c 2v~c . . . . O)

IIo~ - ~IN~ | o~lN~lt < 2 - ~ , k = ] , 2 , . . . . ( 2 )

( J i ~ l N i is strongly dense in ~R. (3) (Here a n d later we i d e n t i f y N | N ~ a n d {N U N~}" in order to m a k e n o t a t i o n more explicit). Since o~ is a n o r m a l s t a t e t h e r e p r e s e n t a t i o n it induces i s normal, hence a *-isomorphism o f 9t onto a factor. W e m a y t h u s assume (o is a v e c t o r s t a t e % , where ~ is a cyclic u n i t v e c t o r in t h e u n d e r l y i n g t t i l b e r t space 9~. Since 9~ is separable a n d ~ cyclic, 9d is separable. L e t {~}~>_1 be a dense sequence of vectors in 9~ w i t h ~1 = ~. Since e 1 belongs to some M.~ t h e r e is f r o m t h e a s s u m p , tion t h a t co is a s y m p t o t i c a l l y a p r o d u c t s t a t e a I ~ - f a c t o r N~ c ~R such t h a t e 1 E N~ a n d

1

Suppose we h a v e chosen N k D N k _ I D . . . D N ~ a n d f o u n d a [ , . . . , a ~ in N~

w i t h lia~[I < 1 such t h a t

I](a~ - e~)$~ll < ~ - ~ , j , r = 1 , . . . , k , (~)

a n d such t h a t (2) holds. W e shall construct N~+~ D N~ a n d a2 +1 , . . . , ~,+~-~+~ in N~+~

such t h a t (2) a n d (4) hold w i t h k replaced b y /c + 1.

B y L e m m a 2 N~ is a h y p e r f i n i t e factor a n d {N~ U ~,~7" v, = 2 . L e t N~ be t h e w e a k closure of [3.~=~Pf, where {P~} is a n increasing sequence of I~-factors.

T h e n we can, using t h e K a p l a n s k y d e n s i t y t h e o r e m [l, Th6or~me 3, p. 43], f i n d a n integer n, operators a~ +~ , " " 9 , 0~+~ in t h e u n i t ball of {N~ U ' ~ k + l P,,}" such t h a t

(4)

168 ARKIV ~F61~ ~IATEI~IATIKo Vol. 9 No. 1 ll(a] +x - e;)~,ll < 2 - ~ - 1 , j , r = 1 ~ * 9 9 ~ k § 1 9

N o w {N~ U P.}" is a finite t y p e / - f a c t o r (which is *-isomorphic to N~ | P=).

Since ~o is a s y m p t o t i c a l l y a p r o d u c t s t a t e there is a t y p e I ~ + l - f a c t o r N~+I con- t a i n i n g (N~ U P.}" such t h a t

I1~' - ~ I N ~ , + I @ 0-,INi+,ll < 2 - ' ~ - ' 9

Therefore, b y i n d u c t i o n (2) a n d (4) hold for all k. Since t h e a r g u m e n t also defines t h e sequence {N~} we h a v e also shown (1).

W e n e x t show t h a t (3) holds. L e t a e i}t. S a y l[a[I _< 1. L e t I0~ . . . W~ e 9d, a n d let fl > 0. T h e n we can f i n d ~ , , . . . . ~it in t h e sequence {&} such t h a t I1~, - ~dl < ~/4, ~ = 1 . . . . ,1. L e t k be a positive integer such t h a t 2 -~ < ~//4 a n d such t h a t k > m a x { j , : r = 1 , . . . , 1}. Since {e~} is a dense sequence in t h e u n i t ball in 2 , it is strongly dense in t h e u n i t ball in gt b y t h e K a p l a n s k y d e n s i t y t h e o r e m [1, Th6orbme 3, p. 43]. Therefore t h e r e exists e, ~{e~} such t h a t II(a - e,)~,ll < V/4, r • ~ , . . . , 1. B y (~) w e c a n f i n d N~,, k 1 > k, a n d b e N<

w i t h llbll <_ 1 such t h a t II(b -- e,)&II < V / 4 , s = 1 . . . . , k r I n p a r t i c u l a r this holds for s = j , , r = 1 , . . . , l . Therefore we have

II(b - a)~,li _< ll(b - - e,)w, ll § II(a - - e,)!v,Jl < lib - - e,II II~, - - ~,11 § [](b - - e,)~)l § ~//4

< 2V/~ § ~/~ + ~/~ = ~ .

Thus U

i~l

~ f is s t r o n g l y dense in ~ , a n d we h a v e shown t h a t (1), (2), (3) all hold as asserted.

L e t ~ denote t h e u n i f o r m closure of U ~ I N ~ . T h e n ~ is a U t t F - a l g e b r a which is s t r o n g l y dense in ~ . L e t N O = {21} a n d p u t M k = N kl3 N ; _ I , k = 1 , 2 . . . T h e n we can i d e n t i f y !3 w i t h @k~=lMk. L e t ok =eo[Mk. B y (2) we h a v e t h a t if 1 < k < l t h e n

= 11~1 | 9 - . | ~.,, | (~]N~, - - E0kA_ 1 @ o o o ~ ) (D 1 (~) ~[~i')11

< II~I.N~ - - ~.,,+, | 9 . . | ~'l | ~IN~'II

_< I I ~ l X ~ - - ~.,,+1 | olN~,+lll + I1~.,,+1 | ( ~ I N ~ + I - ~,,+,, | 9 9 9 | ~z | ~ l N ' i ) l l

< 2 - ' - 1 + I I ~ I N ~ + I - - ~k+~, | . . . | ~ t @ ~IN~I!

<~ 2 --k-1 _}_ 2 - k - 2 § 9 . . § 2 - l : 2 - k ( 1 _ _ 2 - ( z - ~ ) ) < 2 - k .

Therefore t h e n o r m a l states ~1 | 9 9 9 | c~ | colN~, f o r m a C a u c h y sequence, and accordingly converge u n i f o r m l y to a n o r m a l s t a t e @~1 w~, which is clearly a produe~

c o O )

s t a t e on !8. I f ~ denotes t h e r e p r e s e n t a t i o n of !8 i n d u c e d b y @~=1 ~ t h e n r~

e x t e n d s n a t u r a l l y t o a *-isomorphism o f ~ o n t o t h e I T P F I - f a c t o r ~(!8)". H e n c e we h a v e shown t h a t (ii)--> (iii).

(5)

I I Y F E R F I N I T E P R O D U C T F A C T O R S 169 Finally we show (iii)--> (i). Assume there is a UHF-algebra 9A which is the infinite tensor product of t y p e / , c f a c t o r s M,~, so 9~ = | Suppose there is a product state | i of 9~, where r is a state of M.~. L e t w~ %~ o ~ be the Gelfand-Segal decomposition of w~ and % o z t h a t of | eo~. Say ~ and are cyclic vectors in the Hilbert spaces 9 ~ and ~ respectively. Then $ = | ~ ,

= | ~ , and ~ = | ~ , see [3, Prop. 2.9]. Since the property of being a pro- duct factor is an isomorphism invariant we m a y assume ~ = ~(9~)".

Suppose t h a t a t y p e /,-factor M contained in ~R, t h a t a normal state o~

on ~ , and t h a t e > 0 are given. Then M | ~R acts on 9~ | 9r and b y L c m m a 1 there is a *-isomorphism cr of ~ onto M | ~ . As in the proof of L e m m a 2 it follows from [4, L e m m a 3.3] t h a t there is an inner automorphism of M | carrying c~(M) onto M | I . Composing this automorphism with ~ we m a y assume a(M) = M | I. Now since M is a / , - f a c t o r and 9d is infinite dimensional there is a u n i t vector ~ in 9 { @ c ~ such t h a t c o o ~ - 1 = o ~ on M | s e e [ l , Corollaire 10, p. 302]. From [3, Ch. 1] or [5, L e m m a 3.1] there are an integer N >__ 2 and a vector ~ in 9 d | 1 | 1 7 4 1 such t h a t

co

i = _ ~ r

moreover, ~ can be taken to be a unit vector. L e t

_Y = ~-~(M | 2 / , , | . . . | M,N_I | I) ,

where we identify M,i with ~(M,~), and I stands for the identity on |176 N 9/~.

Then N is a finite t y p e I factor, and M ~ N ~ . I t remains to show t h a t [[co -- co[N | eoIN~]] < s, or equivalently, t h a t []c% -- o} [~(N) | %,]~(N)~I] < s.

L e t ~ = ~ @ ( | Then % I M | ~, and so

I1% - % I~,(N) | % I~,(N)ll

-< II% - ~r + II%[~(N) | %I~(N) ~ - - % I ~ ( N ) | %[~,(N)11 + ]I%Io~(N) | %lo~(N) * -- %l~x(N) | %[~x(N)~

< e/3 4-l[%[~x(N) -- %lo~(N)[I 4--I[%l~x(N) ~ -- %I~(N)*[[

< e/3 -1- el3 + e/3 = ~.

Thus w is asymptotically a product state, hence ~R is a product factor. The proof is complete.

(6)

170 ARKIV F O R M A T E M A T I K . Vol. 9 No. 1

References

1. DIXmER, J., Les alg~bres d'opdrateurs duns l'espace hilbertien, Gauthier-Villars, P a r i s (1969).

2. GLIM~, J . , On a certain class of o p e r a t o r algebras, Trans. Amer. Math. Soe. 95 (1960), 318--340.

3. GUICttARDET, A., P r o d u i t s tensoriels infinis et reprdsentations des relations d ' a n t i c o m - m u t a t i o n s , Ann. Scl. Ecole Norm. Super. 83 (1966), 1--52.

4. POWERS, 1~. T., R e p r e s e n t a t i o n s of u n i f o r m l y h y p e r f i n i t e algebras a n d t h e i r associated rings, Ann. Math. 86 (1967), 138--171.

5. A_RAKI, t I . , a n d WOODS, E. J., Complete Boolean algebras of t y p e I factors. Publ. ~ I M S , Kyoto Univ. Set. A , 2 (1966), 157--242.

Received October 20, 70. E r l i n g St~rmer

Universi~etot i Oslo M a t e m a t i s k I n s t i t u t t Blindern

Oslo 3, N o r w a y

Références

Documents relatifs

We describe a CHSH-Bell test where one photon of an entangled pair is amplified by a Measure and Prepare cloner and detected by using threshold detectors and a postselection

safekeeping. The pynabyte utility DYNASTAT displays your current system configuration. 'The steps described below will change the drive assignrnentsso that you will

In this Science and Society Essay, Alex Mauron examines the traditional notions of the &#34;soul&#34; expounded by Aristotle and medieval scholastics such as Thomas Aquinas and asks

[r]

We use three of these no- tions, unitary, dense and planar subcategories in Section 2 to describe the kernels of certain morphisms of categories (functors), called quotient maps, from

FINANCE: How the moneymen are moving in on Hollywood 40 , SCIENCE: DNAevidenceis upending our knowledge of early humans 42 ARTS:Met simulcasts are bringing live opera to Everyman

This assumption is however not used in the definition of the ttelson constant, and we shall therefore consider also another type of Kelson- constant, which is more

all the q~k correspond to finite energy states, i.e.. Our theorem says that states with non-zero initial decay rate could be found among infinite energy states1). What