• 検索結果がありません。

AND WORDS

N/A
N/A
Protected

Academic year: 2022

シェア "AND WORDS"

Copied!
6
0
0

読み込み中.... (全文を見る)

全文

(1)

Internat. J. Math. & Math. Scl.

VOL. 16 NO. 2 (1993) 277-282

277

A METRIC SPACE ASSOCIATED WITH A PROBABILITY SPACE

KEITHF. TAYLORandXIKUI WANG

Department

ofMathematics Universityof Saskatchewan Saskatoon,Sask. CanadaSTN 0W0

(Received January 7, 1992)

ABSTRACT.

For

acomplete probabilityspace

(9t, I,P),

theset of allcomplete sub-(z-algebras of )2,

S(Y:.),

is given a natural metric and studied. The questions of when

5’(Z)

is compact or connected are awswered and the important subset consisting of all continuous sub-a-algebras is shown tobe closed. Con,aections with Christensen’s metricon theyon

Neumann

subalgebrasofa

Type

II]-factorarebriefly discussed.

KEY

WORDS

AND PHRASES.

a-algebras,conditional expectations,metricspace,yon

Neu-

mannalgebra.

1992 AMS SUBJECT CLASSIFICATION

CODES.

Primary"

28A05,

28A20. Secondary:

46L10.

1. INTRODUCTION.

Let

(f/, E, P)

beacomplete probabilityspaceandlet

S(I)

denotethe set ofallcompletesub-(z-

algebrasofI]. Wewill defineametric don

5’(Z)

and investigatesometopological propertiesofthe resultingmetricspace

(S(Z),d).

In

[1],

Boylanintroduceda metric

d’

on

3’(Z)

for thepurpose of

studying convergence propertiesoftheconditional expectationoperatorsdefinedby varyingsub-(z- algebrasofZ. Ourmetricturns out to beequivalent toBoylan’sand seemstobemore convenient forstudyusing fitnctionalanalytic methods.

Weprove, in section3,that

(S(Z),d)

iscompact if and only if

Z

ispurelyatomic. Wealso show that

(S(Y_,),d)

isconnected if andonlyif

Z

has at mostoneatom.

In

section4,weconsider the continuous sub-(z-algebrasof

Z

and show that they formaclosed nowheredense subsetof

(S(Z), d).

Thereis acloseanalogy between probabilityspaces and von

Neumann

algebras witha faithful finitenormal trace.

In

fact,ourdefinitionof dismodeledon ametricdefinedbyChristensen

[2]

on

the set of all yon Neumann subalgebrasofa

Type II-factor.

Christensen’s metrichasbeen useful in the studyofthe index in

II-factors (see [3]

and

[5]).

Wegiveashort discussionofacommon generalizationofourmetricand Christensen’s in section5.

2.

THE

METRICS.

In

thissection,wedefinethemetric dandBoylan’smetric

d’

and show thattheyareequivalent.

If

Zo S(Z),

then let

L(Zo) L(f, Zo, P)

be consideredas aclosed subspaceof

L(f, Z, P)

in thenatural way. Since bounded functionsaresquareintergrableonaprobabilityspace,wemay consider

L(Z0)

as a

(non-closed)

subspace ofthe Hilbert space

L(f, Z,P).

Itis easy to check that the unit ball in

L(Z0)

is closed in

L(f, Z, P).

Themetric dis essentially the Hausdorff metriconthe unit balls in theL2-norm. Let

(2)

L

(o),

{f

G

L’’(0) ilflloo _< 1}.

ForEl,

E

ES(E),let

d(Vl, 2)

7.J-{

sll|) inf

Ilf gll2,

sup inf

II.f gll2}.

Itiseasytocheckthat, disa metricon

S().

Forthe metric

d’,

we use a definitiondueto

Rogge [6]

which isaslight variationon Boylan

[1].

For

EI,E

ES(Z), let

d’(Ea,_E)

sup inf

P(AAB)V

sup inf

P(AAB).

AZBEE BEEA

(Boylan’s metric d" has Vreplacedby

+

andclearly

1/2d* < d’ _<

d’,sod and d* areessentially the

same).

Thearguments in

[1]

showthat

(S(E),d’)is

acompletemetricspace.

Forany

Eo S(E),

let

0

denote theorthogonal projectionof

L(,Z,P)

onto

L(fl,

E0,

P),

considered as asubspace of

Lz(E).

Let

E

denote the rtriction of e to

L(E).

As is well k.own, e and

E

are restrictions of the conditional expectation mapping of

L(fl, E,P)

onto

L(,

E0, P). We will use any of the well known special properties ofconditional expectation withoutgiving references.

Wenownsider the relationship between themetricdand

d’.

THEOREM 1. Forany

E, E S(E),

d’(E,, E)

d(E,,

) 22d’(,, Z)(1 d’(E,, E))

Thus

d’

and dareequivalent metrics.

PROOF. To prove the left. hand inequality, fix

A E.

It is shown in 2.1 of

[4]

that

if P(AAB)

isachieved at theset

C {EE(XA) > },

where XA isthe indicator function of theset A.

Some

elementary manipulations show that

P(AC)

- I- Er(XA)

dP

Now

II- XAII, ,

o

P(AC) I1 x.ll, -I E(XA)II, IIX* E(*)ll, IIX* E(XA)II*

Thus

inf P(AAB) d(, ),

for all

A .

Symmetric arguments apply for2,which gives

d’(,, ) 5 d(,, ,).

E ,

For

the right inequality, note that, for any

f

E

L(), (f) f

and

E*(f)

is the

(L -

norm)

closest element of

L(E)

to

f. Thus,

sup inf

Ilf-gll

sup

I[f-E(f)ll, g

sup

IIW’(f)-E(f)ll,

lL(hai() IL() fL(h

Similarly for theother term in d(E,),sowehave the inequality

(,,2:)

p

IIE’(f)-S(f)ll

However,

for any

.f

L’() the function

f’ (1 + f)

takes valuesin

[0, 1]

almosteverywhere

(3)

METRIC SPACE ASSOCIATED WITH A PROBABILITY SPACE 279 and theorem 3in

[6]

shows that

Thus,

lIE -’ (f’) E2(.f’)[[2 <_ v/2d’(E,, E2)(1 he(E,, Z;))

lIE"’ (.f) E(/)II _<

Thiscompletes theproofofthetheorem.

COROLLARY. Theidentity map isahomeomorphismof

(S(E),d’)

with

(S(E),d).

REMARK

1.

Not

onlyarcd and

d’

equivalentinthesensethat theydefine thesameopenscts inS(E), but thesamesequences areCauchy ineachof

(._q(E),d)

and

(S(E),d’).

Since

(S(),d’)

is complete,wehave that(._q(S).d) isacomplete metricspace.

REMARK

2. The right hand inequality intheorem cannot be improved tooneoftheform d

_< kd’

forsomepositiveconstant I,’,asisshownbythefollowing example.

EXAMPLE.

Let12

[0, 1],

with

P

denotingLebesguemeasureonZ,the Lebesgue measurable stbsets of

[0, 1].

For each a

>

0, let

E

denote the a-algebra generated by an atom

[0, a]

and

El"l(a, 1] {B -

S C (a,

1]}.

Forany

A ,

let

B Af’l(a, 1]

and

B [O, a] U( A l"l(

a,

1]).

Then

B,B

fi

E,

and either

P(AAB)

or

P(AAB)

islessthan orequalto

.

If

A [0, ],

then

P(AAB) > ,

forall

B E,

Thus

d’(E,, E)=

To

compute

d(E,Z),

note that, for any

f L(E),

thefunction

f..(,,,]

is in

L(Y:.o)

and

]]f f.Y(.]]l -< .

On theotherhand,if

f

.([0,]- .X(,,],then

Eo(])

0almosteverywhere.

Thisimplies that

]].f- gll -> ll.f[]2 v/.

forany g(

L(E,).

Thus

d(Zo, E) yrs.

With

E E,

and

E ,

the inequalities in theorem become

_< _< 2-() (_<

’).

Thusnok existswith

d(Z,, E) _< kd’(,E)

for alla

(0,1).

3.

THE MAIN

RESULTS.

Weconsiderthemetric space

(S(), d)

andestablishthefollowing two theorems. Since

(S(), d’)

ishomeomorphicto

(S(F_,),d),

these characterizationsholdfor eithermetric.

THEOREM

2.

(S(F_,),d)

iscompact ifand only if

:E

ispurely atomic.

THEOREM 3.

(S(F.,),d)

isconnectedifand only if has atmost oneatom.

Before proving these theorems,weneed thefollowinglemma whichisafolkloreresult ofmeasure theory andcan beproven withaneasy applicationof

Zorn’s

lemma.

LEMMA

If

(X,.T’,u)

is a finite measurespace with no atoms, then there existsa map

F"

[0,1] .T"

such that

F(s)

C_

F(t)

and

,(F(t)) tI(X)

forall s,t

. [0,1]

withs

_<

t.

Wewill usethe notation, forA

E

and

Ao {AB" B c= 0}

PROOF

OF

THEOREM

2.

Assume

that

(12, E,P)

is not completely atomic.

Let fie

and

12

be the continuous and atomic parts of fl respectively.

We

are assuming

P(fl) >

0.

Let

F

[0,1] E

be such that

F(I)

C_

12,P(F(t)) tP(fl),

and

F(s)

C_

F(t)

for s,t

. [0,1],s _<

t.

Define

F[s,t] F(t)\F(s)

and let

(4)

280 K.F. TAYLOR AND X. NANG

A,,

U{ F[(k )/2"-’, {2k I)/2"]

k 1, :2 ,2

"-

foreachn 1,2,... Then P(A,,)

P(f/c)/2,

for eachn and

P(A,AA,,) P(f/c)/2,

forn

-

m.

Let

E,

debotethe a-algebra

generated

by

A,,

f]

E

and theatom

f\A,,

for eachn.

Let f

denote the

indicatorfunction of

F/\A,.

Then

f

E L(E,,) and forany9 5

L(Em),

ifm n,then9differs from

f

byatleast 7oneither

(F/\A,,) CI

A,or

(f\A,)f’l(fl\A,,).

Eachofthesesetshas probability

P(flc)/4.

Thus

Ill 91122 _> p(flc)/16,

whichimplies

d(E, E,,,) _> /P(fc)/4,

ifn

-fl

m.

Hence,

the

sequence

(E,,),,__

has nocluster points and

S(E)

isnotcompact.

Conversely, suppose

(f,E,P)

is completelyatomic. Since

(S(.,),d)

iscomplete, compactness will follow ifwe show that

S(E)

istotally bounded. If

E

hasonly afinitenumber of atoms, then

S(E)

is afinite set and obviously compact. So suppose

{A, A2

aredisjointatomsgenerating

E. For >

0, chooseNsuch that

:__N+ P(A,,) < . There

areonlyafinitenumberof

algebras

of

setscontained in the

algebra generated

by

{A A2v }. Let .A,...,

,4 denote these

algebras. For

<

j

<

k,let denotethea-algebra generated by

,4j

and

{AN+I, AN+,. ..}.

Let

A For

any

S(E) A

Iq

.Ai

forsomej,

<

j

_<

k. Then

C_

E

andit is easy tocheck that

d(E;, Zj) <

c. Therefore,

S(Z)

iscoveredbythe finite setof-balls centered at

, <

j

<

k. Since

c

>

0wasarbitrary,

S(E)

istotally bounded.

PROOF OF THEOREM 3. Suppose Aand

B

aretwo distinct atoms in

.

Let

S’

denote

the set of all

E0

in

S(E)

which haveanatomcontainingboth

A

and

B,

let

S

denotethe set ofall

0

in

S(Z)

whichhave two disjoint atoms containing

A

and

B,

respectively. Itiseasy toseethat

S(E) S

[J

S.

Let

S

and

E S.

Let

A’

and

B’

be disjoint atoms of

E

containing

A

and

B,

respectively. Let Cbe an atom of

El

which contains

AI.JB. Let f XA’ X’ - L()

For

anyg

L(Z),g

a,almost everywhereon

C,

forsomeconstant a,

lal <

1. Thus

IIf gll > fA If gl

de

+/ If gl

de

/( -a)P(a) + ( + a)P(B)

> /2min{P(a),P(B)}

Therefore,

d(,E) > /min{P(A),P(B)},

for all

r S

and

Z

E

S.

Noticealso that the trivialo-algebra

E {f, 1} S

and

E S.

So

S

and

S

arenonempty disjointopensubsets of

S(E)

with union

S(E). Hence, S(E)

is notconnected if therearetwoor moreatomsin ).?,.

Conversely,suppose

(f/, E, P)

has at mostoneatom. If thereis anatom, callit

A

and if there isnoatom, let

A .

If

P(A)

1, then

S(E)

isaonepoint space, whichisconnected.

So

assume

P(Ft\A) >

0. Wewillshow that any

E0 S(E)

canbe connected toE; byanarcin

S(E).

Firstassume

E0 S(E)

and

E0

iscompletelyatomic,generated bythe disjoint atomsA0,

A,...

(a

finiteor infinite collection). We may assume

A

C_

Ao.

Foreach n

_>

1, by the lemma, there exists

F,, [0,1]

A,,f’IE such that

P(F,(t)) tP(A)

and

F(s)

C_

F,(t)if

s,t

[0,1]

with

s

<

t. Similarly, let

Fo [0,1] A01DE

be such that

F0

is increasing and

P(Fo(t)) P(A) +

tP(Ao\a),

for all

[0, 1].

Foreach

[0,1],

define

Z(t) S’()

as ther-algebra generated by

{F,(t)CIE, A,,\F,(t)’n

0,1,2

}. It

isclear that

E(0) Eo, E(1) E

and

E(s)

C_

Z(t)if

s_<t. Fixs,

t[0,1]withs<t.

(5)

METRIC SPACE ASSOCIATED WITH A PROBABILITY SPACE 281

For

any

.[

6_.

L((t))l

and for each

,

0,1,2,...,

f

isconstant,saya,,,almost everywhereon the atom

.4,,\F,,(t)i,,

E(/). Define g E

L’’(E()),,

by making 9 a,, on

A,,\F,,(s)

and9

f

on

Fn(s),

forn 0, 1,2 Then

f

andy differonlyon

U(F,,(t)\F,,(s))

and therebyatmost2.

Now

P{(F,,(t)\F,(s)))

(t

s)[P(Ao\A) + P(A,) + P(A2) +...]

Th

I1- 11 <

This implies that

d(V(s),V(t)) <

2

//-’Z-

s. Therefore,any atomic element

:E0

of

S()

is con- nected byan arcto v

If

:E0

is not completely atomic, let f

tic

(J fl, where

flc["l 0

is continuous and

fl :E0

is

atomic.

Let :E

be theatomic a-algebra generated by an atomequal to

fie

and

fl

]0.

By

methodsimilar tothe above, but justworking with fie,an arcfrom

:El

to

:E0

canbefound. Also,

beingatomicis connected byan arc to

.

Thusanyelement

:E0

of

S(:E)

can be connectedby

an arcto :E. Therefore

S()

isconnected, in fact,arcwise connected if

(fl, :E, P)

has atmostone atom.

The two theorems ofthis sectionshow that thereis somemeaningflfl connection between the topologyof

S(:E)

and thestructure of theoriginal probabilityspace. Since this

topology

isreallyan embodimentof the equiconvergence property for conditional expectations,wefeel that

S(:E)

with

this metricd

(or

ifoneprefers

d’)

will provideagoodlocale forstudyingconditionalexpectations.

4.

CONTIIIUOUS SUB-a-ALGEBRAS.

Let S(:E)= {0

6

S(:E)’ (fl:E0, P)has

no

atoms},

the space ofcontinuous sub-a-algebras of E. Ofcourse, if is not acontinuous a-algebra, then

S() . On

the other

hand,

if

:E

is

continuous, then

S(:E)

isavery rich set tolookat. For example,if

G

isagroup of :E-measurable transformations of12, let denote thea-algebraof G-invariant sets in

.

IfGis afinitegroup,

then

e S(:E).

This providesaway in whichtoconstructmanyinteresting examplesofelements of

S(:E). In

spite ofthis,

Sc(:E)

isa"small" subset of

S(:E).

THEOREM

4.

S(E)

isaclosed nowhere dense subset of

S(]).

PROOF. We

firstobserve that if

:E0

E

S(:E)

and

A E,

then

{P(AB)" B e :Eo} [0, P(A)].

Nowfix

e S()\S(:E)

and letAbeanatomof

. By

the aboveobservation,forany

:Eo e

there exists

B e :Eo

such that

P(A["IB) 1/2P(A).

Let

f

X -Xu\

L(0).

Forany g

L(F,),g

is constanton

A

andoneeasily checks that

Ill- gll2 >-(P(A)) /:.

Thus

d(:E0, :E) _> (P(A))

/forany

0 e S(:E).

Hence

S()\S(:E)

isopen and

S(:E)is

closed.

To

see that

S(])

is nowhere dense, let

:E0

E

Sc()

and e

>

0. Choose

A

" :Eo

such that

0<

P(A)

<e

.

Define

:E

tobethea-algebra generatedby

A

and

{B :E0" B

C

fl\A}.

Then

has anatomA. Forany

f L(:E0),

let g

Xu\Af e L(:E).

Since

Ill- gll] f.a Ill <- P(A)

and

:E

C:E0,we have that

d(0, E) _< (P(A))

/

<

e. Combined with thegeneralestimateabove,

(6)

forthis particular

E0

and1,

d(

x"

,-,o,) (P(A))

/ Thusthe open ballof radiusecenteredat

E0

is not contained in

S.(E)

forany

>

0.Since

5’c(E)

is

closed,

it isnowhere densein

S().

Thus

(So(E), d)

is itselfaconplete metric space whose topological propertiesshould be inter- esting to study.

We

donot consider suchastudyinthis paper.

5. CONNECTIONS WITH VON

NEUMANN ALGEBRAS.

There is aone-.to-onecorrespondence between the sub-a-algebrasof

E

and the yonNeumann subalgebrasof

L(2, E,

P).

In

this section,weassumethe readeris familar with the basictheory ofvon

Neumann

algebrasasfound ina reference such asSa"kai

[7].

Let

.M

beafixedfiniteyon

Neumann

algebrawithadistinguished faithful, finite, normaltrace r. For example,

.M could

be

L(,E,P)

or a

Type lI-factor. Let S(.JM)

denote thesetofallyon

Neumann subalgebrasof21 whichhave thesameidentityas.A4.

In

the case that

.M

is a

Type

ll-factor, Christensen defined a metric on

S(.M)

in

[2].

He

showed that

S(AJ)

then becameacompletemetricspace.

He

used hismetrictostudy perturbation properties of subfactors and his metrichas also been used in thestudyof the index ofasubfactor in a

Type II-factor (see [3]

and

[5]).

Christensen’s definition works well in thesituation we are

considering. Forx .A4,let

IIllz 9r(’)

For

A/, S(.A4),

let

A/’

and denote theunit

balls inA/’and

,

respectively.

Then define

d(A/’, ) max{

sup inf x

YlI,

sup inf

II

x

YlI}

It

is routinetocheck thatd isametricon

S(.A4)

and Christensen’sproof that

S(.A4)

iscomplete carriesovertothis moregeneralsituation.

Someoftheproofsofsection3canbeadaptedtotheyon

Neumann

algebra situationbut not all.

We

list below what we can prove and leave the details of the adaption of the proofs tothe interestedreader.

THEOREM 5. Let .A4 be a finite yon Neumann algebra with a fixed faithful, finite trace r. With the Christensen’smetric,

S(.M)

iscompact if and onlyif.A4

i.s

generated byitsminimal projections.

PROPOSITION6. If

S(./t4)

isconnected,then.A4 hasat mostoneminimalprojection.

OPEN PROBLEM Istheconverseof proposition 6

true?

ACKNOWLEDGEMENT.Thisresearchwassupported byan

NSERC

Canada operating grant.

!EFERENCES

1. Boylan, E. S. EquiconvergenceofMartingales,

Ann_.

Math._.__.

Stst.

42

(1971),

552-559.

2. Christensen, E. SubalgebrasofaFiniteAlgebra, Math____:.

Ann___.. 243 (1979),

17-29.

3.

Jones, V. F.

1.Index for

Subfactors, Invent.

Math.

(1983),

1-25.

4. Kudo, H.

A Note

onthe

Strong Convergence

ofr-algebras, Ann____. _2

(1974),

"/6-83.

5.

Mashhood, B.

and Taylor,K. F. On Continuityof theIndexofSubfactorsofa Finite

Factor,

J.

Func.

Anal. 12

(195S),

56-66.

6.

Rogge, L.

Uniform Inequalities forConditional Expectations,

Ann;

Prob.___..2

(1974),

486-489.

7. Sakai,S.

C’-algebras and W’-algebr,

Ergebnisse derMath. 60, Berlin, Heidelley, NewYork:

Springer-Verlag

1971.

参照

関連したドキュメント

So in this paper generalized groups, completely simple semigroups, and Ree’s matrix semigroups are the same.. One must pay attention to this point that each generalized group is

An example of a database state in the lextensive category of finite sets, for the EA sketch of our school data specification is provided by any database which models the

We also note that taking A = id in Theorem 2 gives the known existence theorem (cf. 141] and the references given therein) for the classical sweeping process in infinite

The authors there gave an example of a function which is contraction in the category of cone metric spaces but not contraction if considered over metric spaces and hence, by proving

Debnath’s in- equality is given, and a generalization of Alzer’s inequality is established.. We call the left-hand side of this inequality Alzer’s inequality [2], and the

In [FHR18], the authors characterize the multipliers from one model space to another in terms of kernels of Toeplitz operators and Carleson measures for model spaces.. However, it

Definition 1 Given two piles, A and B, where #A ≤ #B and the number of to- kens in the respective pile is counted before the previous player’s move, then, if the previous player

In order to keep the formulas short, we use the following multi-index notation.. We will show now that this condition characterizes polyfunctions. Since there exists an