EMBEDDING IN BROWNIAN MOTION by NEIL F. FALKNER B . S c ...

205
EMBEDDING IN BROWNIAN MOTION by NEIL F. FALKNER B.Sc, University of Manitoba, 1973 M.Sc, University of Manitoba, 1974 A THESIS SUBMITTED IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE STUDIES in the Department of Mathematics We accept this thesis as conforming to the required standard THE UNIVERSITY OF BRITISH COLUMBIA September , 1978 (c) Neil F. Falkner, 1978

Transcript of EMBEDDING IN BROWNIAN MOTION by NEIL F. FALKNER B . S c ...

EMBEDDING IN BROWNIAN MOTION

by

NEIL F. FALKNER

B . S c , U n i v e r s i t y of Manitoba, 1973 M.Sc, U n i v e r s i t y of Manitoba, 1974

A THESIS SUBMITTED IN PARTIAL FULFILMENT OF

THE REQUIREMENTS FOR THE DEGREE OF

DOCTOR OF PHILOSOPHY

i n THE FACULTY OF GRADUATE STUDIES

i n the Department

of

Mathematics

We accept t h i s t h e s i s as conforming to the

req u i r e d standard

THE UNIVERSITY OF BRITISH COLUMBIA

September , 1978

(c) N e i l F. Falkner, 1978

In p r e s e n t i n g t h i s t h e s i s in p a r t i a l f u l f i l m e n t o f the requ i rement s f o r

an advanced degree at the U n i v e r s i t y o f B r i t i s h Co lumb ia , I ag ree that

the L i b r a r y s h a l l make i t f r e e l y a v a i l a b l e f o r r e f e r e n c e and s tudy .

I f u r t h e r agree t h a t p e r m i s s i o n f o r e x t e n s i v e c o p y i n g o f t h i s t h e s i s

f o r s c h o l a r l y purposes may be g r a n t e d by the Head o f my Department or

by h i s r e p r e s e n t a t i v e s . I t i s u n d e r s t o o d tha t c o p y i n g o r p u b l i c a t i o n

o f t h i s t h e s i s f o r f i n a n c i a l g a i n s h a l l not be a l l o w e d w i thout my

w r i t t e n p e r m i s s i o n .

Department o f M a t h e m a t i c s

The U n i v e r s i t y o f B r i t i s h Co lumbia

2075 Wesbrook Place Vancouver, Canada V6T 1W5

Date September 18, 1978

Supervisor: Dr. R. V. Chacon

Ab s t r a c t and H i s t o r i c a l Review: Let n be a p o s i t i v e i n t e g e r , l e t y

be a p r o b a b i l i t y measure on ]R n , and l e t (B ) be Brownian

motion w i t h i n i t i a l d i s t r i b u t i o n y .

(For those u n f a m i l i a r w i t h Brownian motion we i n s e r t a b r i e f

h e u r i s t i c e x p l a nation. Consider a drunk on a pub crawl i n ]R n .

Imagine that h i s i n i t i a l p o s i t i o n i s u n c e r t a i n , and i s described by

the p r o b a b i l i t y law y . Imagine that there i s a pub at each po i n t i n

]R n , and that the drunk wanders from pub to pub i n a t o t a l l y random

f a s h i o n , as i s to be expected from h i s i n e b r i a t e d s t a t e . Imagine that

the drunk takes only an i n f i n i t e s i m a l d r i n k at each pub, so that he

keeps moving c o n s t a n t l y . We can describe the drunk's random progress

through ]R n by c o n s i d e r i n g the set of a l l p o s s i b l e paths he can f o l l o w ,

and a s s i g n i n g each some i n f i n i t e s i m a l p r o b a b i l i t y . In more p r e c i s e

mathematical terms, we d e s c r i b e the motion of the drunk by means of a

c e r t a i n p r o b a b i l i t y measure P on the space C of continuous maps

from [0, °°) i n t o 3Rn . Then B t (to) = co(t) f o r co e C , t e [0, °°) .

The p r o b a b i l i t y space (C,P) , together w i t h the f a m i l y (B f c) of

random v a r i a b l e s defined on i t , i s c a l l e d Brownian motion, a f t e r the

b o t a n i s t Brown who, i n 1827, observed the random motion of microscopic

p a r t i c l e s suspended i n water.)

For each random time T l e t y^ be the d i s t r i b u t i o n of the random

v a r i a b l e B^ . (This random v a r i a b l e i s the one defined by

B (to) = B - .(to) .) I t i s n a t u r a l to ask which measures v on ]R n are 1 i ( to)

of the form y T where T i s a stopping time. (A stopping time i s a

- 11 -

random time which "does not depend on the f u t u r e " ; f o r example, the

f i r s t time (B^) h i t s some set i n H n i s a stopping time.)

Skorohod [1] f i r s t considered t h i s s o r t of question. He showed that

i f n = 1 and u i s the point mass at 0 , and i f v i s a p r o b a b i l i t y

easure on E. w i t h centre of mass at 0 and w i t h f i n i t e v a r i a n c e m 2

a = x dv(x) , then there i s 'a "randomized" stopping time T such

that v = and the expectation of T i s equal to the v a r i a n c e of

v . (We remark that i f u i s the point mass at 0 and T i s a

stopping time, p o s s i b l y randomized, w i t h f i n i t e e x p e c t a t i o n , and i f we

define v to be , then v has mean 0 and v a r i a n c e equal to

the expectation of T . Thus Skorohod's r e s u l t can be s t a t e d i n an

" i f and only i f " form.) Dubins [1] and Root [ 1 ] , by d i f f e r e n t methods,

showed that Skorohod's r e s u l t can be improved i n that a T which i s

not randomized can be obtained. The method of Dubins y i e l d s a " n a t u r a l "

stopping time T whenever v i s a p r o b a b i l i t y measure on ]R whose

mean i s defined and equal to 0 , even when v does not have f i n i t e

v a r i a n c e . The meaning of " n a t u r a l " here i s enlarged upon i n the paper

of Chacon [ 1 ] , where standard stopping times are defined. We give a

d i f f e r e n t but equivalent d e f i n i t i o n of standard stopping times i n 8.2,

and a l s o another c h a r a c t e r i z a t i o n of them i n 8.13. Doob (see Meyer

[3]) has pointed out that i f n = 1 and u and v are any two

p r o b a b i l i t y measures on TR then there i s a stopping time T such

that u = y . This i s the reason f o r the s p e c i a l i n t e r e s t i n standard

stopping times. But we d i g r e s s . To get back to our s t o r y , Chacon and

Walsh [ 1 ] , using p o t e n t i a l theory, gave a very transparent proof of

the r e s u l t of Dubins and Root. A l s o using p o t e n t i a l theory, Rost [1]

- i i i -

ge n e r a l i z e d Skorohod's r e s u l t to Markov processes w i t h proper p o t e n t i a l

k e rnels (eg., Brownian motion i n dimension greater than or equal to 3).

Rost's method, however, produces a randomized stopping time even when

a non-randomized one e x i s t s . Now a p h y s i c i s t has s a i d , "Give me f i v e

parameters, and I w i l l f i t an elephant; give me s i x parameters, and

w i l l make him wiggle h i s trunk!" In other words, i t i s n a t u r a l to ask

when one can get a non-randomized stopping time, f o r n >_ 2 .

Baxter and Chacon [1] have given s u f f i c i e n t c o n d i t i o n s f o r t h i s to

be so, but t h e i r hypotheses are rat h e r strong — see the d i s c u s s i o n

i n 8.21. In 7.11 and 8.20, we have succeeded i n proving r e s u l t s

along these l i n e s which appear much c l o s e r to being best p o s s i b l e .

They are not a c t u a l l y best p o s s i b l e , however, as the example 7.13

shows. I t i s much e a s i e r to get best p o s s i b l e r e s u l t s i f one allows

oneself to work w i t h randomized stopping times. In 11.12, the 2

p r o b a b i l i t y measures v on R such that v = v , where T i s a

standard randomized stopping time and u i s a p r o b a b i l i t y measure on

]R^ such that i s a p o t e n t i a l (see 1.4), are c h a r a c t e r i z e d . This

r e s u l t appears to be new. A l s o , i n 10.5, we prove a p a r t i c u l a r case of

the r e s u l t of Rost [ 1 ] , by a d i f f e r e n t method.

Now Skorohod [1] al s o showed that i f (X^) i s a square-integrable

martingale w i t h time set I = {0, 1, 2,...} , s a t i s f y i n g E(X^) = 0 ,

and i f (^t) i - s Brownian motion, i n one dimension, s t a r t i n g from 0 ,

then there i s an i n c r e a s i n g sequence (T ) of randomized stopping

times w i t h f i n i t e e x p e c t a t i o n s , such that the processes (B^ ) and i (X.) have the same j o i n t d i s t r i b u t i o n . (Note that i f X. = X,. f o r l J l 0

a l l i , t h i s reduces to the r e s u l t f o r measures which we described

- i v -

f i r s t . ) Dubins [1] remarks that h i s method can be a p p l i e d here to

y i e l d non-randomized stopping times, and that the c o n d i t i o n that 2

E(X_^) < «> can be dropped, though of course the T^'s need not have

f i n i t e e xpectation then (though one can show that they w i l l be standard

i f c onstructed as Dubins d e s c r i b e s ) . In 12.7 we prove a g e n e r a l i z a t i o n

of t h i s to Brownian motion i n n dimensions. (For n >_ 2 , we use

randomized stopping times.) We a l s o give a d e t a i l e d proof of Dubins'

a s s e r t i o n concerning the one dimensional case. Monroe [1] has shown

that a r i g h t - c o n t i n u o u s martingale ^t^0<t<°° C a n ^ e e m k e d d e d -*-n a n

"enlargement" of one dimensional Brownian motion by means of a r i g h t -

continuous i n c r e a s i n g f a m i l y (T f c) of minimal (= standard) stopping

times. In 12.16, we prove a g e n e r a l i z a t i o n of t h i s r e s u l t to n

dimensions. F i n a l l y , i n 12.18 we describe an example, discovered by

R. V. Chacon, which shows that enlargement r e a l l y i s necessary i n the

theorem of Monroe.

Now f o r a few words about the s e c t i o n s of t h i s t h e s i s from which

r e s u l t s have not yet been mentioned. Sections 1, 5, and 6 are mainly

devoted to e s t a b l i s h i n g n o t a t i o n and terminology, and to s t a t i n g

c e r t a i n known r e s u l t s , f o r ease of reference. Sections 2, 3, and 4

are devoted to e s t a b l i s h i n g those aspects of the f i n e theory of

balayage which are needed throughout the r e s t of the t h e s i s . With

the probable exception of the strong form of the domination p r i n c i p l e 2

f o r the l o g a r i t h m i c p o t e n t i a l i n ]R , the r e s u l t s of these three

s e c t i o n s are not new. I b e l i e v e , however, that some of the proofs are

novel. For example, the reader might f i n d i t amusing to compare our

proof of 2.1 w i t h the proof of 8.43 of Helms [ 1 ] , or to compare our

- v -

proof of 3.2 w i t h the d i s c u s s i o n i n s e c t i o n s 9 and 10 of the paper of

B r e l o t [ 1 ] . Secti o n 9 of t h i s t h e s i s i s concerned w i t h the development

of the m a t e r i a l on randomized random v a r i a b l e s , and enlargements of

p r o b a b i l i t y spaces, needed i n s e c t i o n s 10, 11, and 12, and i s e s s e n t i a l l y

a review of, and enlargement on, p a r t s of the papers of Baxter and

Chacon [2 and 3 ] . The theorem 9.13 however, though a simple r e s u l t ,

i s new and sheds l i g h t on the meaning of the " d i s t r i b u t i o n a l enlargements"

defined i n Baxter and Chacon [ 3 ] . ( D i s t r i b u t i o n a l enlargements are

e s s e n t i a l l y what we c a l l o p t i o n a l enlargements.)

One l a s t t h i n g : towards the end of s e c t i o n 10, we d i s c u s s a

couple of p o t e n t i a l t h e o r e t i c a p p l i c a t i o n s of the embedding theorem

10.5.

- v i -

Table of Contents Page

1. P o t e n t i a l Theoretic P r e l i m i n a r i e s 1

2. The I n t e g r a l Representation of Balayage 8

3. Thinness 23

4. The Strong Form of the Domination P r i n c i p l e 31

5. Brownian Motion P r e l i m i n a r i e s 46

6. P r e l i m i n a r i e s on Brownian Motion and P o t e n t i a l Theory 57

7. Embedding Measures i n Brownian Motion i n a Green Region, Using Non-Randomized Stopping Times 61

1 2 8. Embedding Measures i n Brownian Motion i n H or 1 ,

Using Non-Randomized Stopping Times 76 9. Randomized Stopping Times, and Enlargements of

P r o b a b i l i t y Spaces 98

10. Embedding Measures i n Brownian Motion i n a Green Region, Using Randomized Stopping Times 124

2 11. Embedding Measures i n Brownian Motion i n H , Using

Randomized Stopping Times 136

12. Embedding Processes i n Brownian Motion 147

13. Appendix of Miscellaneous N o t a t i o n and Terminology 184

References 189

Index of Selected N o t a t i o n and Terminology 192

- v i i -

Ac knowled g ement s

I wish to thank my s u p e r v i s o r , R a f a e l Chacon, whose guidance

and encouragement have b e e n i n v a l u a b l e ' t o me. He has taught me much

about how to do mathematics. I would a l s o l i k e to thank Maurice Sion,

whose l u c i d teaching made i t i n f i n i t e l y e a s i e r f o r me to l e a r n the

"heavy duty" measure theory needed i n the study of processes, and

John Walsh, who was always ready to answer my questions about

Brownian motion and P o t e n t i a l Theory. I wish to express my g r a t i t u d e

to the N a t i o n a l Research C o u n c i l of Canada, and the Izaak Walton

K i l l a m Memorial Fund of the U n i v e r s i t y of B r i t i s h Columbia, f o r

p r o v i d i n g me w i t h f i n a n c i a l support. F i n a l l y , I would l i k e to thank

Cathy Agnew and C a r o l Samson f o r t h e i r f i n e t y p i n g .

- 1 -

1. POTENTIAL THEORETIC PRELIMINARIES

This s e c t i o n i s mainly devoted to e s t a b l i s h i n g the n o t a t i o n and

terminology of p o t e n t i a l theory that we s h a l l use.

1.1. For each x e H , 6 denotes the Dirac measure at x: x

6 (A) = l . ( x ) f o r A £ B o r e l H X A.

I f x = 0 we s h a l l j u s t w r i t e 6 f o r 6

1.2. A denotes the L a p l a c i a n on E. ;

n .2

i = l 9x:

We s h a l l f r e q u e n t l y use the L a p l a c i a n i n the sense of d i s t r i b u t i o n

theory. In p a r t i c u l a r , i f u i s a superharmonic f u n c t i o n i n an

open set D i n H n then -Au i s a ( p o s i t i v e ) measure i n D which

i s c a l l e d the Riesz measure of u . On the other hand, i f T i s a

d i s t r i b u t i o n i n D such that -AT i s a ( p o s i t i v e ) measure i n D

then T a r i s e s from a unique superharmonic f u n c t i o n i n D . (We

s h a l l have no need of the l a t t e r f a c t , though.)

1.3. Let $ be the f u n c t i o n on ]Rn defined by

$(x) =

2 1_ 2TT

x i f n = 1

l o g x i f n = 2

i f n > 3 (n-2)a x n

,n-2

- 2 -

where a i s the n - 1 dimensional Lebesgue measure of the surface n of the u n i t sphere i n ]R n .

(we take $(0) = + °° i f n > 2 .)

Then <3> has the f o l l o w i n g three p r o p e r t i e s :

a) $ i s superharmonic.

b) A $ = -6 .

c) $ i s i n v a r i a n t under r o t a t i o n s about the o r i g i n .

In f a c t , these p r o p e r t i e s determine $ to w i t h i n an a d d i t i v e

constant. Of course property b) has the simple form that i t does

because of the way we normalized $ .

We s h a l l a l s o use the l e t t e r $ to denote two other f u n c t i o n s :

f i r s t , the f u n c t i o n on ]R n x ]R n defined by $(x,y) = $(x- y ) ;

second, the f u n c t i o n on [0,°°) s a t i s f y i n g 4>(x) = $(| |x| |) .

The context w i l l always make c l e a r which f u n c t i o n we mean. Note that

$ depends on n but t h i s dependence i s not made e x p l i c i t . I t i s

p a r t i c u l a r l y important to keep t h i s i n mind when $ i s regarded as

a f u n c t i o n on [O,00) .

1.4. I f y i s a measure on the B o r e l subsets of H n (not assumed

f i n i t e on compact sets) then we de f i n e U^, U^: H n —>• [0,°°] by

t r > ) =

u V ) =

$ +(x,y)dy(y)

$ (x,y)dy(y)

and we d e f i n e U^, on the set where and are not both i n f i n i t e ,

- 3 -

by

We say i s a p o t e n t i a l i f f U M i s everywhere defined and super­

harmonic on ]R n •

This happens i f f i s f i n i t e at a l l p o i n t s of E.n and

i s f i n i t e at at l e a s t one point of 3Rn .

In t h i s case i s l o c a l l y Lebesgue i n t e g r a b l e and y i s

f i n i t e on compact sets (so and y can be regarded as Schwartz

d i s t r i b u t i o n s on ]R n) and AU^ = -y; i n p a r t i c u l a r , y can be

recovered from .

The f o l l o w i n g three r e s u l t s give more e x p l i c i t c h a r a c t e r i z a t i o n s

of those measures y f o r which i s a p o t e n t i a l . We s t a t e these

r e s u l t s without proof, but the statements are so d e t a i l e d that they

almost prove themselves.

1.5. P r o p o s i t i o n : Suppose n = 1 .

a) = 0 and U y = -U y

b) I f U y i s f i n i t e at two d i s t i n c t p o i n t s then y i s f i n i t e ,

i s f i n i t e at a l l p o i n t s and U y i s continuous; indeed

U y i s L i p s c h i t z and yOR) i s a L i p s c h i t z constant f o r

x dy(x) < °° c) i s a p o t e n t i a l i f f yOR) < 0 0 and

1.6. P r o p o s i t i o n : Suppose n = 2 .

a) I f U^(x) i s f i n i t e and y({y £ K. : | |y-x| | <_ r}) i s f i n i t e 2

f o r some x i n K. and some r i n ( l , 0 0 ) then y i s

- 4 -

f i n i t e , U. i s f i n i t e at a l l p o i n t s and i s continuous; 1 2

indeed i s L i p s c h i t z and — u0R ) i s a L i p s c h i t z

constant f o r .

b) I f u i s f i n i t e on compact sets then i s l o c a l l y Lebesgue

i n t e g r a b l e .

c) U P i s a p o t e n t i a l i f f u i s f i n i t e and

log +||x||dp(x) < » .

1.7. P r o p o s i t i o n : Suppose n >_ 3 .

a) = 0 and U P = .

b) I f U P i s f i n i t e at one poin t then u i s f i n i t e on compact

sets and U P i s l o c a l l y Lebesgue i n t e g r a b l e .

c) i s a p o t e n t i a l i f f lA$du < 0 0 •

One can combine 1.5 through 1.7 i n t o a s i n g l e r e s u l t .

1.8. P r o p o s i t i o n : U P i s a p o t e n t i a l i f f u i s f i n i t e on compact •

sets and I $(x) I du (x) < 0 0 . J||x||>l

1.9. Consider an open set D i n ]R n .

I f f o r some poin t X Q i n D the f u n c t i o n $(XQ,«) has a harmonic

minorant i n D then f o r a l l x i n D, $(x,*) has a harmonic minorant

i n D, and indeed has a greatest one which we s h a l l , f o r the moment,

denote by h(x,•) .

In t h i s case we say D i s a Green re g i o n and the f u n c t i o n

G: D x D —*• [0,°°] defined by

- 5 -

G(x,y) = *(x,y) - h(x,y)

i s c a l l e d the Green f u n c t i o n of D .

For each x e D, G(x,*) i s the smallest non-negative super­

harmonic f u n c t i o n u i n D s a t i s f y i n g Au = - 6

J x

G(x,') should be thought of as the e l e c t r o s t a t i c p o t e n t i a l i n

D that would a r i s e from a u n i t p o i n t charge at x i f 3D were

made of e l e c t r i c a l l y conducting m a t e r i a l , provided 3D i s connected.

One can show that G i s j o i n t l y continuous on D x D and

G(x,y) = G(y,x) f o r a l l x,y e D .

C l e a r l y i f n >_ 3 then E n i s a Green r e g i o n and the Green

f u n c t i o n G of lR n i s given by G(x,y) = $(x,y) .

I t i s a l s o c l e a r that any open subset of a Green region i s a

Green r e g i o n . In p a r t i c u l a r , any open subset of E n , where n >_ 3,

i s a Green r e g i o n .

A moment's thought w i l l show that an open set D i n ]R i s a

Green r e g i o n i f f D ^ E . 2

One can show that i f D i s an open set i n E then D i s a 2

Green r e g i o n i f f E \D i s not a p o l a r s e t . This i s Myrberg's theorem -see 8.33 of Helms [1].

(A p o l a r set i s a set contained i n the set of " p o l e s " ( i e . ,

i n f i n i t i e s ) of a superharmonic f u n c t i o n . P o l a r sets are the

" s m a l l " sets of p o t e n t i a l theory.) 2

One can a l s o show that i f D i s an open set i n E (or f o r that

matter, i n K.n) then D i s a Green r e g i o n i f f there i s a non-constant

- 6 -

non-negative superharmonic f u n c t i o n i n D . This i s part of 8.33 of

Helms [1], but i t ' s not hard to give a proof using only the m a t e r i a l

i n the f i r s t s i x chapters of Helms; i e . , the m a t e r i a l up to and

i n c l u d i n g the chapter on Green p o t e n t i a l s .

1.10. Let D be a Green region i n ]R n w i t h Green f u n c t i o n G .

I f y i s a measure on the o - f i e l d of B o r e l sets of D, we

de f i n e Gy: D —»- [0,°°]. by

Gy(x) = G(x,y)dy(y)

I f Gy i s f i n i t e at at l e a s t one p o i n t of each component of D

then Gy i s superharmonic i n D, y i s f i n i t e on compact sets (and

so can be regarded as a Schwartz d i s t r i b u t i o n i n D) and AGy = -y .

In t h i s case we say Gy i s the p o t e n t i a l of y, and we d e s c r i b e

y as a measure i n D such that Gy i s a p o t e n t i a l .

I f y i s a measure i n D such that Gy i s a p o t e n t i a l then

the greatest harmonic minorant of Gy i n D i s zero. Gy i s thus

the smallest non-negative superharmonic f u n c t i o n i n D having y as

Riesz measure.

I f u i s a superharmonic f u n c t i o n i n D which has a subharmonic

minorant, and i f y i s the Riesz measure of u, then Gy i s a

p o t e n t i a l and u = Gy + h where h i s the grea t e s t harmonic minorant

of u . This i s known as the Riesz decomposition theorem.

1.11. Let D be an open set i n H n .

I f u i s a non-negative superharmonic f u n c t i o n i n D and E i s

- 7 -

a subset of D then the r£duite of u over E r e l a t i v e to D

(which we s h a l l denote by red(u,E,D)) i s defined to be the infimum

of the set of non-negative superharmonic f u n c t i o n s i n D which

majorize u on E . The balayage of u over E r e l a t i v e to D

(which we s h a l l denote by bal(u,E,D)) i s defined to be the lower

r e g u l a r i z a t i o n of u .

One can show that bal(u,E,D) i s superharmonic i n D and the

set where bal(u,E,D) d i f f e r s from red(u,E,D) i s a p o l a r subset

of the boundary of E .

We note that the more usual n o t a t i o n s f o r red(u,E,D) and

bal(u,E,D) are and r e s p e c t i v e l y .

Of course, i f D i s not a Green r e g i o n then re"duite and balayage

are t r i v i a l n o t i o n s , s i n c e then every non-negative superharmonic

f u n c t i o n i n D i s constant.

1.12. The main reference we recommend f o r b a s i c p o t e n t i a l theory i s

the book of Helms [ 1 ] . Other u s e f u l references are the books of

Du P l e s s i s [ 1 ] and B r e l o t [ 1 ] . In p a r t i c u l a r , Du P l e s s i s uses the

n a t u r a l and i n t u i t i v e language of d i s t r i b u t i o n theory, whereas Helms

avoids i t .

- 8 -

2. THE INTEGRAL REPRESENTATION OF BALAYAGE

Throughout t h i s s e c t i o n , n i s a p o s i t i v e i n t e g e r , D i s a

Green r e g i o n i n TRU, and G i s the Green f u n c t i o n of D .

I f E i s a subset of ]R n and y i s a measure i n D such that

Gy i s a p o t e n t i a l then bal(Gy,E,D) i s the p o t e n t i a l of a measure

i n D which we s h a l l denote by bal(y,E,D) . Thus

bal(Gy,E,D) = G bal(y,E,D) .

In t h i s s e c t i o n we s h a l l prove the f o l l o w i n g i n t e g r a l represent­

a t i o n formula f o r balayage:

I f u i s any non-negative superharmonic f u n c t i o n i n D then

f o r every x i n D,

bal(u,E,D)(x) = u ( y ) b a l ( 6 ,E,D)(dy) D X

This was proved by B r e l o t [ 1 ] .

Our method of proof d i f f e r s from B r e l o t ' s i n that i t r e q u i r e s

no appeal to the theory of the D i r i c h l e t problem; i t uses only the

c l a s s i c a l p o t e n t i a l theory contained i n the f i r s t seven chapters of

the book of Helms [ 1 ] .

We begin w i t h a weak v e r s i o n of the "domination p r i n c i p l e " .

2.1. P r o p o s i t i o n : Let E be a closed subset of D and l e t y be

a measure i n D such that Gy i s a p o t e n t i a l s y l i v e s on E, and

y does not charge polar subsets of the boundary of E.

Suppose v i s a non-negative superharmonic f u n c t i o n i n D such

- 9 -

that v >_ Gy on E\Z where Z i s some polar s et.

Then v >_ Gy throughout D .

Proof: vAGy i s a p o t e n t i a l i n D so vAGy = Gv f o r some measure

v i n D .

Now Gv >_ Gy on E\Z and we wish to show Gv >_ Gy throughout

D . Let x be i n D .

For 0 < r < distance(x^R n\D) l e t a^ be the uniform u n i t

d i s t r i b u t i o n on the closed b a l l of rad i u s r centred at x, and

l e t g = b a l ( a ,E,D) . r r

Now B l i v e s on E and Ga = Gg on E\P where P i s some r r r

polar subset of the boundary of E . Al s o B r does not charge p o l a r

s e t s , by 2.2 below. Hence

Gvda = Ga dv r r

GB dv r

f Gv dB

Gy dB

GB r dy

Ga dy r

Gy da

L e t t i n g r go to 0, we obt a i n

Gv(x) > Gy(x) •

- 10 -

2.2. Lemma: Let u be a superharmonic f u n c t i o n i n an open set V

i n H n , and l e t y be the Riesz measure of u . Then y does not

charge p o l a r subsets of the set where u i s f i n i t e .

Proof: By a standard argument we can reduce to the case where V i s

a Green region (even an open b a l l ) and u i s a p o t e n t i a l i n V .

Then u = G^u . I t s u f f i c e s to show that i f K i s any compact sub­

set of {u < 0 0 } then y does not charge polar s e t s . Now G y K V K

i s f i n i t e so by 6.21 of Helms [ 1 ] , given e > 0 there i s a compact

se t C £ K such that y(K\C) < e and G y i s continuous on V

I f P i s a polar set i n V then there i s a measure y w i t h

compact support i n V such that G^y = 0 0 o n P n K .

Now V d y c = G y y c dy < °°, so y c ( { G v y = «>})= 0

Hence the y -outer measure of P i s l e s s than e . K

As e > 0 was a r b i t r a r y , y does not charge p o l a r s e t s . K •

2.3. Lemma: Let E be a closed subset of D and l e t u be a

p o t e n t i a l i n D which i s f i n i t e on the boundary of E . Let h be

the greatest harmonic minorant of u i n the open set D\E .

Let v = h i n D\E

u on E

Then v = reM(u,E,D) .

Proof: Let w = r£d(u,E,D) . Then w £ u and w i s harmonic i n

D\E so w <_ v . Let be a sequence of open b a l l s such that

- 11 -

D\E = uB. i

and V i , { j: B. = B.} i s i n f i n i t e . Let u = u and l e t

PI(u . ; B .) i n B . 1 1 i

u i+1 u. l i n D\B.

l

Then each u. i s superharmonic i n D, and u. 4- v . l l

Hence v, the lower r e g u l a r i z a t i o n of v, i s superharmonic i n

D and i s , i n f a c t , the p o t e n t i a l of a measure v which l i v e s on the

closed set E and (by 2.2) does not charge polar subsets of the

boundary of E .

Now w >_ v on E and {w < w} i s a polar set. Thus by the

domination p r i n c i p l e 2.1, w >_ v throughout D . Now i n D\E,

w = w and v = v . A l s o , on E, w = u = v . Thus v = w .

Remark: I f u i s not f i n i t e on the boundary of E the co n c l u s i o n

of the above lemma may f a i l . For instance suppose E = {x} and

u = G(x,*) f o r some x e E, and n (the dimension of lR n) i s at

l e a s t 2 . Then v = u but red(u,E,D) = ° ° l r v ; .

be p o t e n t i a l s i n D which are f i n i t e on the boundary of E . Then

Proof: The greatest harmonic minorant of a sum of two superharmonic

2.4. C o r o l l a r y : Let E be a closed subset of D and l e t u ,u

b a l ( u + u

= baKu^E.D) + bal(u 2,E,D) .

- 12 -

fun c t i o n s (each of which has a subharmonic minorant) i s the sum of

t h e i r greatest harmonic minorants - see 5.22 of Helms [ 1 ] . Hence

red(u^ + U £ , E, D)

= red(u 1 }E,D) + r£d(u2,E,D) .

I t f o l l o w s that bal(u 1+u 2,E,D) and b a l ( u ,E,D) + bal(u 2,E,D)

d i f f e r at most on a po l a r s e t . But two superharmonic f u n c t i o n s

which are equal except on a po l a r set are equal everywhere. •

The f o l l o w i n g r e s u l t i s proved i n chapter 8 of Helms [1] using

the D i r i c h l e t problem theory developed i n that chapter. To sub­

s t a n t i a t e our c l a i m that the r e s u l t s of t h i s s e c t i o n can be e s t a b l i s h e d

without t h i s part of the theory, we i n c l u d e a proof.

2.5. Lemma: Let Z be a po l a r subset of D and l e t x e D\Z .

Then there i s a f i n i t e measure u i n D such that Gu = 0 0 on Z

but Gy(x) < °° .

Proof: By the d e f i n i t i o n of a polar set (Helms [ 1 ] , p. 126) there i s

an open set V <=_ R n and a superharmonic f u n c t i o n v i n V such that

Z _c V and v = 0 0 on Z . Let v be the Riesz measure of v . Let

(B_^) be a sequence of open b a l l s such that

(D\{x}) n V = uB. . I

For each i , l e t v. be the measure on the B o r e l s e t s of D defined I

by v^(A) = v(AnB_^) . Each i s a f i n i t e measure i n D s a t i s f y i n g

Gv. (x) < 0 0 . A l s o , i n B. , AGv. = -v = Av . Hence i n B., Gv. and i l i l i

- 13 -

v d i f f e r by a harmonic f u n c t i o n . In p a r t i c u l a r ,

{Gv = °°} n B. = {v = •»} n B. . 1 1 1 Now choose a sequence (a.) of

p o s i t i v e r e a l numbers such that

Y a.v. (D) < . 1 1 00

1 and

y a.Gv.(x) < . 1 1 i 00

The d e s i r e d measure y can be taken to be £ i •

2.6. C o r o l l a r y : Let u be a non-negative superharmonic f u n c t i o n

i n D and l e t E be any subset of D . Then bal(u,E,D) i s the

smallest non-negative superharmonic f u n c t i o n v i n D such that

v >_ u on E except f o r a polar set.

Proof: Let w = bal(u,E,D) . Then w i s a non-negative super­

harmonic f u n c t i o n i n D and E n {w<u} i s a p o l a r set. Let v

be a non-negative superharmonic f u n c t i o n i n D such that Z = E n {v<u}

i s a p o l a r set. Let x e D\Z . By 2.5 there i s a f i n i t e measure u

i n D such that Gy = 0 0 on Z but Gy(x) < 0 0 . For any e > 0,

v + eGy ^ u on E, so v + eGy >_ w . L e t t i n g e go to 0, we f i n d

that v (x) >_ w(x) . Hence v >_ w on D\Z As Z i s p o l a r , v >_ w

throughout D . •

2.7. C o r o l l a r y : Let ( u^) ^ e a n i n c r e a s i n g sequence of non-negative

superharmonic f u n c t i o n s i n D whose supremum u i s not i d e n t i c a l l y

i n f i n i t e on any component of D, and hence i s superharmonic. Let

- 14 -

(E_^) be an i n c r e a s i n g sequence of subsets of D, w i t h union E .

Then b a l ( u ,E ,D) + bal(u,E,D) .

Proof,: Let w = bal(u,E,D) . C l e a r l y bal(u_^,E^,D) increases to

a superharmonic f u n c t i o n v <_ w . I f x e E and v(x) < u(x) then

f o r some i , x e E^ and v(x) < u_^(x) . Thus

E n {v<u} £ u (E. n {bal(u.,E.,D) < u.}) . i

Hence E n {v<u} i s a p o l a r s e t , so v >_ w by 2.6. •

2.8. Lemma: Let u be any non-negative superharmonic f u n c t i o n

i n D . Then u i s the l i m i t of an i n c r e a s i n g sequence of bounded

p o t e n t i a l s i n D whose Riesz measures are f i n i t e .

Proof: Let be a sequence of open r e l a t i v e l y compact subsets

of D which increases to D . For each i , l e t u. = bal(iAu,D. ,D) l l

Then each u_ i s a p o t e n t i a l i n D, bounded by i , and s a t i s f y i n g

u. = uAi i n D. . Hence u. i u . A l s o , i f u. i s the Riesz l l l l measure of u. then u.(D) = u.(D.) < 0 0 .

i x i i n

2.9. Theorem: Let u be a non-negative superharmonic f u n c t i o n i n

D and l e t E be any subset of D . Then f o r each x e D,

bal(u,E,D)(x) = u ( y ) b a l ( 6 ,E,D)(dy) .

Proof: I . Assume u i s a f i n i t e p o t e n t i a l w i t h Riesz measure u . a) Assume E i s closed i n D . Let

a = y E B = yD\E

- 15 -

Then u = Ga + GB and Ga and G3 are f i n i t e p o t e n t i a l s so by 2.4,

bal(u,E,D) = bal(Ga,E,D) + bal(GB,E,D) . F i r s t consider Ga . By

the domination p r i n c i p l e 2.1, bal(Ga,E,D) = Ga . Al s o

bal(G6 x,E,D) = G8^ on E\Z where Z i s p o l a r ; hence

bal(G6 ,E,D) = G6 almost everywhere with respect to a, by 2.2. X X

Thus

Ga(y)bal(6 ,E,D)(dy) X

G bal(<$x,E,D) (z)a(dz)

G6 x(z)a(dz) = Ga(y)6 x(dy)

= Ga(x)

Now consider GB . i ) F i r s t suppose x I E or n = 1 . Let V be

the open set D\E and l e t G' be the Green f u n c t i o n of V. Let h

(resp. k) be the greatest harmonic minorant of GB (resp. GS^) i n

V . As GB and G6 are f i n i t e on the boundary ( i n D) of E, X

and

bal(G8,E,D) i n V

k = bal(G6 x,E,D) i n V,

by 2.3. ( I f n = 1, every superharmonic f u n c t i o n i s f i n i t e . ) Now

G'B = G8 - h i n V . But

G'B(x) = G' (x,z)g(dz)

G(x,z) - k(z)3(dz) V

= G8(x) - bal(G6 ,E,D)(z)B(dz) .

- 16 -

Thus

bal(GB,E,D)(x) = bal(G6 ,E,D)(z)B(dz) X

G bal(6 ,E,D)(z)B(dz)

GB(y)bal(6 x,E,D)(dy)

i i ) Now suppose x e E and n _> 2 . Then {x} i s p o l a r so

bal(v,E\{x),D) = bal(v,E,D) f o r each non-negative superharmonic

f u n c t i o n v i n D, by 2.6. For each i e U l e t B^ be the open

b a l l of ra d i u s 2 centred at x and l e t E = E\B^ . Then each

E. i s closed i n D and E. + E\{x} . Hence, by 2.7, l i

ba l ( v , E ,D) + bal(v,E\{x},D)

f o r each non-negative superharmonic f u n c t i o n v i n D . Now by i ) ,

bal(GB,E ,D)(x) = ba l ( G 6 x , E i , D ) ( z ) 8 ( d z )

f o r each i . L e t t i n g i go to i n f i n i t y , we o b t a i n

bal(GB,E,D)(x) = bal(G6 x,E,D)(z)g(dz)

G8(y)bal(6 x,E,D)(dy)

Combining our r e s u l t s so f a r we have

bal(Gy,E,D)(x) = Gy(y)bal(6 x,E,D)(dy)

f o r a l l x e D, where Gy i s f i n i t e and E i s closed i n D .

b) Now assume that E i s merely a countable union of closed subsets

of D . Let (E^) be an i n c r e a s i n g sequence of closed subsets of D

- 17 -

whose union i s E . Then f o r each i ,

bal(Gy,E ,D)(x) = G y ( y ) b a l ( 6 x , E i , D ) ( d y ) ,

by a ) . L e t t i n g i go to i n f i n i t y , we o b t a i n

bal(Gy,E,D)(x) = Gy(y)bal(6 x,E,D)(dy) .

c) Now l e t E be an a r b i t r a r y subset of D . Let us impose the

a d d i t i o n a l assumption that y(D) < °° . F i x e > 0 . Let

v = e + bal(Gy,E,D)

w = e + b a l ( G 6 x , E , D )

S = {v > Gy}

T = {w > G6 } x

Then E i s contained i n S to w i t h i n a p o l a r set so f o r any non-

negative superharmonic f u n c t i o n f i n D,

b a l ( f ,E,D) <_ b a l ( f ,S,D) .

The same statement holds w i t h S replaced by T . A l s o ,

S = u {v>r} n {Gy <_ r} reQ

Now f o r each r , {v>r} i s open and {Gy <_ r} i s closed i n D .

Thus S i s a countable union of closed subsets of D . S i m i l a r l y ,

so i s T . Hence

- 18 -

bal(Gy,E,D)(x)

< bal(Gy,T,D)(x)

Gy(y)bal(6 x,T,D)(dy)

bal(G6 x,T,D)(z)y(dz)

w dy

bal(G6 ,E,D)dy = ey(D) +

<_ sy (D) +

= ey(D) +

= ey(D) + bal(Gy,S,D)(x)

(by b)

bal(G6 x,S,D)dy

Gy(y)bal(6 x,S,D)(dy)

(by b)

< ey(D) + v

= ey(D) + e + bal(Gy,E,D)(x)

As y(D) < «° and e > 0 was a r b i t r a r y ,

bal(Gy,E,D)(x)

Gy (y)bal(5 x,E,D)(dy)

< bal(Gy,E,D)(x)

I I . Now l e t us consider the general case. By 2.8, there i s an

i n c r e a s i n g sequence (u-^) °^ f i n i t e p o t e n t i a l s i n D such that

u. + u and each u. has f i n i t e Riesz measure. By I . c ) , f o r each 1 1 i we have

b a l ( u ,E,D)(x) = u (y ) b a l ( 6 ,E,D)(dy) 1 X

- 19 -

L e t t i n g i go to i n f i n i t y and applying 2.7 on the l e f t and the

monotone convergence theorem on the r i g h t we ob t a i n at l a s t

bal(u,E,D)(x) = u(y)bal(S x,E,D)(dy)

2.10. C o r o l l a r y : Let E be any subset of D and l e t y be a

measure i n D such that Gy i s a p o t e n t i a l . Then f o r each- x e D,

bal(Gy,E,D)(x) = bal(G6 x,E,D)dy

Proof:

bal(Gy,E,D)(x) = Gy(y)bal(6 x,E,D)(dy)

G b a l ( 6 x , E , D ) ( z ) y ( d z )

bal(G6 x,E,D)dy

2.11. C o r o l l a r y : Let E be any subset of D . Then f o r a l l

x,y e D,

bal(G6 ,E,D)(x) = bal(G<5 ,E,D)(y) y x

Proof: Take y = & i n 2.10. y •

2.12. C o r o l l a r y : Let E be any subset of D and l e t u,v be

non-negative superharmonic f u n c t i o n s i n D . Then

bal(u+v,E,D) = bal(u,E,D) + bal(v,E,D) .

Proof: This f o l l o w s immediately from the l i n e a r i t y of the i n t e g r a l . •

- 20 -

2.13. C o r o l l a r y : Let E be any subset of D . Let x e D and

l e t y be a measure i n D such that

bal(u,E,D)(x) = u(y)y(dy)

f o r every non-negative superharmonic f u n c t i o n u i n D of the form

u = G(z,«) f o r z e D . Then

y = bal(6 x,E,D) .

Proof: For any z e D,

G Y(z) = G ( z , y ) Y ( d y )

= bal(G(z,-),E,D)(x)

G(z,y)bal(6 x,E,D)(dy)

= G ba l ( 6 ,E,D)(z) .

That i s , Y a n d bal(6 x,E,D) have the same Green p o t e n t i a l r e l a t i v e

to D . •

2.14. Lemma: Let f be a twice c o n t i n u o u s l y d i f f e r e n t i a b l e

f u n c t i o n w i t h compact support i n D . Then f i s e x p r e s s i b l e as

the d i f f e r e n c e of two non-negative bounded continuous superharmonic

f u n c t i o n s i n D .

Proof: Let g = -Af and l e t u = Gg +, v = Gg . Then u and v

are f i n i t e continuous superharmonic f u n c t i o n s i n D, by 6.22 of

Helms [ 1 ] . Now A(u-v) = -g = Af i n D . Hence there i s a harmonic

f u n c t i o n h i n D such that u - v = f + h . Now u and v are

- 21 -

bounded on the s u p p o r t of g and hence a r e bounded on D by the

d o m i n a t i o n p r i n c i p l e 2.1. Thus h i s a l s o bounded on D . L e t

M = sup h . Then f = (u+M) - (v+M-h), and u + M , v + M - h

a r e bounded n o n - n e g a t i v e c o n t i n u o u s superharmonic f u n c t i o n s i n D .

Remark: U s i n g D i r i c h l e t problem t h e o r y , one can show t h a t the h

i n the above p r o o f must be 0, but we don't need t h i s h e r e .

2.15. Theorem: L e t E be any subset of D . L e t y be a measure

i n D such t h a t Gy i s a p o t e n t i a l . I f A i s any B o r e l subset of

D, then:

a) x>—>• b a l ( 6 x > E , D ) (a) i s a B o r e l f u n c t i o n i n D

b) b a l ( y , E , D ) ( A ) = y ( d x ) b a l ( 6 x , E , D ) ( A )

P r o o f :

a) b a l ( 6 ,E,D)(.D) = 1 bal(6 ,E,D)(dy)

= b a l ( l , E , D ) (x) <_ 1

f o r each x e D; i n p a r t i c u l a r , b a l ( 6 x > E , D ) i s a f i n i t e measure

i n D . Suppose $ i s a t w i c e c o n t i n u o u s l y d i f f e r e n t i a b l e f u n c t i o n

w i t h compact support i n D . Then

<Ky)bal(6x,E,D)(dy)

i s a B o r e l f u n c t i o n i n D by 2.14 combined w i t h 2.9. By a monotone

c l a s s argument, i t f o l l o w s t h a t x b a l ( 6 x , E , D ) (A) i s a B o r e l

f u n c t i o n i n D f o r any B o r e l subset A of D .

- 22 -

b) By a) we can define a measure v on the B o r e l subsets of

D by

v(A) = u(dx)bal(6 x,E,D)(A) .

Then f o r any z e D,

Gv(z) = v(dy)G(y,z)

U(dx) bal(6 x,E,D)(dy)G(y,z)

bal(G6 z,E,D)du

= bal(Gy,E,D)(z),

where we have a p p l i e d 2.9 and 2.10. Thus Gv = G bal(u,E,D), so

v = bal(y,E,D) . •

- 23 -

3. THINNESS

3.1. D e f i n i t i o n . Let D be a Green region i n ]Rn . Let E £ D

and l e t x e D . We s h a l l say E i s t h i n at x r e l a t i v e to D

i f f there i s a non-negative superharmonic f u n c t i o n u i n D such

that bal(u,E,D)(x) < u(x) .

A l s o , we s h a l l use the f o l l o w i n g n o t a t i o n s :

fringe(E,D) = {x e E: E i s t h i n at x r e l a t i v e to D}

base(E,D) = {x e D: E i s not t h i n at x r e l a t i v e to D} .

3.2. Theorem: Let D be a Green region i n ]R n . Then there i s

a bounded p o t e n t i a l v i n D such that f o r every E £ D,

base(E,D) = (bal(v,E,D) = v} .

Proof: Let be a sequence of open r e l a t i v e l y compact subsets

of D such that the range of 1 S a n open base f o r D . For v — i each i , l e t v. = bal(l,V.,D) . Let v = )2 v. . Then v i s a l l l

bounded p o t e n t i a l i n D . Suppose E £ D and x e D\base(E,D) .

We'll show that bal(v,E,D)(x) < v(x) . W e l l , there i s a non-negative

superharmonic f u n c t i o n u i n D such that bal(u,E,D)(x) < u(x) .

Let c be a number s t r i c t l y between bal(u,E,D)(x) and u(x) .

Now {u > c} i s open i n D and contains x . Thus f o r some j ,

x e V £ {u > c} . Then u >_ c bal(l,V^,D) i n D . Hence

bal(u,E,D) > c bal(v.,E,D) . Now cv.(x) = c > b a l ( u , E , D ) ( x ) , so - 1 3

v.(x) > b a l ( v . ,E,D) (x) . Since 2 _ : l b a l ( v . ,E,D) + £ 2~\. i s a 3 3 3 1

non-negative superharmonic f u n c t i o n i n D which i s greater than or

- 24 -

equal to v on E except f o r a p o l a r s e t ,

< v ( x ) , by 2.6. •

3.3. C o r o l l a r y : Let D be a Green r e g i o n i n ]R n Let E c D

Then fringe(E,D) i s a p o l a r set and base(E,D) i s a countable

i n t e r s e c t i o n of open s e t s .

Proof: Let v be as i n 3.2. Then fringe(E,D) = E n {bal(v,E,D) < v},

and so i s a p o l a r set by 7.40 of Helms [1]. Also

D\base(E,D) = {bal(v,E,D) < v } , and so i s a countable union of closed

subsets of D as v and bal(v,E,D) are lower semicontinuous.

f u n c t i o n G . Let y be a measure i n D such that Gy i s a

p o t e n t i a l . Let E be any subset of D . Then bal(y,E,D) l i v e s

on base(E,D) .

Proof: Let v be as i n 3.2. Consider any x e D . Let w = bal(v,E,D).

Then bal(w,E,D) = w by 2.6. Thus

3.4. C o r o l l a r y : Let D be a Green r e g i o n i n H n , w i t h Green

w(y)bal(S ,E,D)(dy)

= w(x) = v ( y ) b a l ( 6 ,E,D)(dy),

by 2.9. A l s o , w(x) i s f i n i t e and w <_ v . Thus

b a l ( 6 ,E,D)({w < v}) = 0 .

- 25 -

As t h i s i s true f o r a l l x e D, we have

bal(y,E,D)({w < v}) = 0,

by 2.15. But

{w < v} = D\base(E,D> . •

3.5. C o r o l l a r y : Let D be a Green region i n ]R n, l e t E £ D,

and l e t u,v be non-negative superharmonic f u n c t i o n s i n D . Then

the f o l l o w i n g are e q u i v a l e n t :

a) u <_ v on E\Z f o r some polar set Z

b) u <_ v on base(E,D) .

Proof: a) => b) . Let f = bal(u,E,D) and l e t g = bal(v,E,D) .

Then f <_ g by 2.6. But by the d e f i n i t i o n of base(E,D), f = u

on base(E,D) and g = v on base(E,D) . Hence u <_ v on base(E,D) .

b) => a) . Let Z = fringe(E,D) . Then Z i s a p o l a r set

by 3.3, and u <_ v on E\Z . •

Now we are going to show that i f D and D' are Green regions

i n ]R n, E £ H n , and x e D n D' then E n D i s t h i n at x r e l a t i v e

to D i f f E n D' i s t h i n at x r e l a t i v e to D' .

3.6. P r o p o s i t i o n : Let D be a Green r e g i o n i n JRn and l e t E £ D .

a) I f x 6 D\E then E i s t h i n a t x .

b) I f n = 1, x e D, and E i s t h i n at x then x i E .

- 26 -

Proof: a) Let G be the Green f u n c t i o n of D, l e t u = G(x,*)»

and l e t v = bal(u,E,D) . Let W be a connected open subset of D

such that W n E = 0 and x e W . Then v i s harmonic i n W but

u i s not. Thus u - v i s a non-negative superharmonic f u n c t i o n

i n W which i s s t r i c t l y p o s i t i v e at some point of W and hence at

a l l p o i n t s of W . In p a r t i c u l a r , v(x) < u(x) . Thus E i s t h i n

at x .

b) Let u be a non-negative superharmonic f u n c t i o n i n D . Then

red(u,E,D) = bal(u,E,D) since only the empty set i s polar i n dimension

one. Thus u = bal(u,E,D) on E . But i n one dimension, a l l super­

harmonic f u n c t i o n s are continuous. Hence u and bal(u,E,D) agree

on the c l o s u r e of E i n D . •

3.7. Lemma: Let D be a Green r e g i o n i n ]R n, where n >_ 2 . Let

W be an open subset of D, and l e t x e W . Then there i s a f i n i t e

continuous non-negative superharmonic f u n c t i o n w i n D such that

w <_ w(x) i n D

and

w(x) > sup w . D\W

(Remark: This r e s u l t i s f a l s e i f n = 1 and x belongs to an

unbounded component of D .)

Proof: Let m = sup G(x,*) where G i s the Green f u n c t i o n of D . D\W

( I f D\W i s empty, l e t m = 0 .) Then m i s f i n i t e , by 5.8 of

Helms [ 1 ] . Let w = G(x,«)A(m+l) . Then w <_m + 1 = w(x), and

sup w < m + 1 . D\W •

- 27 -

3.8. Theorem. Let D be a Green r e g i o n i n ]R n, where n >_ 2 .

Let E £ D and l e t x e D . Then the f o l l o w i n g are e q u i v a l e n t :

a) E i s t h i n at x r e l a t i v e to D .

b) There i s an open subset W of D, a non-negative super­

harmonic f u n c t i o n g i n D and a constant a such that

x e W, g(x) < a, and g >_ a on W n (E\{x}) .

Proof: Let F = E\{x} .

a) => b) Let u be a non-negative superharmonic f u n c t i o n i n D

such that bal(u,E,D)(x) < u(x) . Now Z = F n (bal(u,E,D) < u} i s

a p o l a r set so by 2.5, there i s a non-negative superharmonic f u n c t i o n

v i n D such that v = °° on Z but g(x) < u(x) where

g = v + bal(u,E,D) . Let a be a number s t r i c t l y between g(x)

and u(x) and l e t W = {u > a} . Then x e W and g(x) < a .

A l s o , since g^_u on F, g > a , on W n F .

b) => a) By 3.7, there i s a f i n i t e non-negative superharmonic

f u n c t i o n w i n D such that

w <_ w(x) i n D

and

w(x) > sup w . D\W

We'll show that bal(w,E,D)(x) < w(x) . Let

m = sup w . D\W

( I f D\W i s empty, l e t m = 0 .) Since w(x) > m, there i s a

p o s i t i v e r e a l number b such that w(x) - ba >_ m . Let

- 28 -

v = w(x) + b(g-a) . Then v i s superharmonic i n D . Now

v >_ w(x) - ba >_ m, so v >_ 0 i n D and v >_ w on D\W . Also

v >_ w(x) on W n F as g - a >_ 0 there. Hence v >_ w on W n F

as w(x) >_ w i n D . Thus v >_ w on F . Now {x} i s p o l a r , as

n >_ 2 . Thus v >_bal(w,E,D) i n D, by 2.6. But

v(x) = w(x) + b(g(x) - a)

< w(x) . •

3.9. Theorem: Let E be any subset of IR n and l e t D,D' be

Green regions i n E.n . Suppose x £ D n D' . Then E n D i s t h i n

at x r e l a t i v e to D i f f E n D' i s t h i n at x r e l a t i v e to D' .

Proof: C l e a r l y we need only prove the forward i m p l i c a t i o n . I f

n = 1, the r e s u l t f o l l o w s from 3.6. Hence assume n >_ 2 and

E n D i s t h i n at x r e l a t i v e to D . Then by 3.8, there i s an

open subset W of D, a non-negative superharmonic f u n c t i o n g

i n D, and a number a such that x e W, g(x) < a, and g >_ a

on W n (E\{x}) . Let G (resp. G 1) be the Green f u n c t i o n of D

(resp. D') . Let U be an open r e l a t i v e l y compact neighbourhood

of x i n W n D' . Let u be the Riesz measure of g . Then

u(U) < 0 0 so we can de f i n e a f i n i t e measure v i n D' by

v(A) = y(AnU) f o r A e B o r e l D' . Let g' = G'v . Now i n U,

Ag' = -v = Ag . Thus there i s a harmonic f u n c t i o n h i n U such

that g' = g + h i n U . Now g'(x) = g(x) + h(x) < a + h(x) .

Choose e > 0 such that g'(x) < a + h(x) - e . Let W be an

open neighbourhood of x i n U such that h >_ h(x) - e i n W .

- 29 -

Let a' = a + h(x) - e . Then x e W, g' (x) < a' , and g' >_ a'

on W n (E\{x}) . Thus E n D' is thin at x relative to D',

by 3.8. •

Note that the above theorem implies that thinness is a local property.

3.10. Definition. Let E £ ]Rn and let x e TRU . We shall say

E is thin at x i f f there is a Green region D in TRn such that

x e D and E is thin at x relative to D . Also, we shall use

the following notations:

fringe(E) = {x e E: E is thin at x}

base(E) = {x e ]Rn: E is not thin at x} .

3.11. Corollary: Let E £ ]Rn .

a) If x e ]Rn and E is thin at x then for any Green region

D in H n with x e D, E n D is thin at x relative to

D •

b) For any Green region D in H n , fringe(EnD,D) = D n fringe(E)

and base(EnD,D) = D n base(E) .

c) fringe(E) i s a polar set and base(E) is a countable

intersection of open subsets of E.n .

d) If n = 1, fringe(E) is empty and base(E) = E .

Proof: a) and b) follow immediately from 3.9.

c) If n >_ 3, 3Rn i t s e l f i s a Green region. If n = 1 or

2, ]Rn can be written as the union of two Green regions.

Now apply 3.3.

- 30 -

d) apply 3.6. •

3.12. Let us round out t h i s s e c t i o n w i t h some remarks on the f i n e

topology. We don't a c t u a l l y need the f i n e topology i n what f o l l o w s ,

but i t provides a h e l p f u l p e r s p e c t i v e .

Let D be an open subset of ]R n . Then the f i n e topology on

D i s the weakest topology i n D which makes a l l superharmonic

f u n c t i o n s i n D continuous. I t i s stronger than the usual topology

of D, w i t h e q u a l i t y i f f n = 1.. One can show that the f i n e topology

on D i s a l s o generated by the f u n c t i o n s of the form Uy|D, where

y ranges over measures on 3Rn w i t h compact support contained i n D .

I t f o l l o w s that the f i n e topology on D i s equal to the r e s t r i c t i o n

to D of the f i n e topology on H n .

I f D i s a Green region w i t h Green f u n c t i o n G then the f i n e

topology on D i s a l s o generated by the f u n c t i o n s of the form Gy,

where y ranges over measures w i t h compact support i n D .

I f D i s any open subset of ]R n and S i s any c o l l e c t i o n of

superharmonic f u n c t i o n s i n D which i s l a r g e enough to generate

the f i n e topology on D, and i f u + v e S f o r a l l u,v e S, then

the sets of the form W n {u < c} (W open £ D, u e S , c e H )

c o n s t i t u t e a base f o r the f i n e topology on D . Using t h i s i n

co n j u n c t i o n w i t h 3.8, one f i n d s that f o r E £ H n and x e ]R n

(where n >_ 2 ) , E i s t h i n a t x i f f x i s not a f i n e l i m i t p o i n t

of E . Also i f V £ H n (where n i s a r b i t r a r y ) , V i s f i n e l y

open i f f ]Rn\V i s t h i n at each point of V .

- 31 -

4. THE STRONG FORM OF THE DOMINATION PRINCIPLE

In t h i s s e c t i o n we are going to prove the s o - c a l l e d strong form

of the domination p r i n c i p l e , f i r s t f o r a Green r e g i o n , and then f o r 2

E. (where we use the l o g a r i t h m i c p o t e n t i a l ) . For the case of a Green 2

re g i o n , t h i s r e s u l t i s due to B r e l o t [ 2 ] . For E , i t may be new.

We a l s o give a proof of the domination p r i n c i p l e i n E 1 . Of course

t h i s i s q u i t e easy, and the a d j e c t i v e " s t r o n g " i s superfluous i n t h i s

case, owing to the f a c t that i n dimension one a l l superharmonic

f u n c t i o n s are f i n i t e and continuous.

4.1. Theorem ( B r e l o t ' s strong domination p r i n c i p l e )

Let D be a Green r e g i o n i n E n w i t h Green f u n c t i o n G . Let

y be a measure i n D such that Gy i s a p o t e n t i a l . Let E be any

subset of D . Then the f o l l o w i n g are eq u i v a l e n t :

a) y(D\base(E,D)) = 0 .

b) Whenever v i s a non-negative superharmonic f u n c t i o n i n

D such that v >_ Gy on E\Z where Z i s a p o l a r s e t ,

then v >_ Gy throughout D .

c) Whenever v i s a p o t e n t i a l i n D such that v >_ Gy on

E, then v >_ Gy throughout D .

Proof:

a) —> b) I f x e base(E,D) then f o r any non-negative superharmonic

f u n c t i o n u i n D,

- 32 -

u ( y ) b a l ( 6 ,E,D)(dy) x

bal(u,E,D)(x) (by 2.9)

= u(x) u(y)6 (dy), x

so b a l ( 6 ,E,D) = 6 x by 2.13. Hence bal(y,E,D) = y by 2.15. Thus

bal(Gy.E,D) = Gy . Hence v >_ Gy throughout D by 2.6.

b) => c) c) i s j u s t a s p e c i a l case of b) .

c) => a) Let f = bal(Gy,E,D) . Then f = Gv where v = bal(y,E,D)

Let Z = { f < G y } n E . Then Z i s a p o l a r s e t . Now

v(D\base(E,D)) = 0 by 3.4. Suppose y(D\base(E,D)) 4 0 . Then

y 4 v, so Gy 4 Gv . As Gv <_ Gy and Gv, Gy are superharmonic,

t h i s i m p l i e s that {Gv < Gy} i s not a p o l a r s e t . Hence f o r some

x e D\Z, we have Gv(x) < Gy(x) . By 2.5, there i s a p o t e n t i a l g

i n D such that g = °° on Z but Gv(x) + g(x) < Gy(x) . Let

v = Gv + g . Then v i s a p o t e n t i a l i n D and v >_ u on E,

but v(x) < u ( x ) . •

4.2. Theorem. Let D be a Green region i n R n w i t h Green f u n c t i o n

G . Let y be a measure i n D such that Gy i s a p o t e n t i a l , and

l e t v be a non-negative superharmonic f u n c t i o n i n D w i t h Riesz

measure v . Then the f o l l o w i n g are e q u i v a l e n t :

a) v >_ Gy almost everywhere w i t h respect to y, and

y(Z) £ v(Z) f o r every B o r e l p o l a r set Z £ D .

b) v > Gy everywhere i n D .

- 33 -

Proof:

a) => b) Let E = {v ^Gu} and l e t P = f r i n g e ( E ) . Then

y(D\E) = 0 . Also P i s a p o l a r set (and i s B o r e l by 3.3, s i n c e

f r i n g e ( E ) = E\base(E,D)) so y(Z) <_ v(Z) f o r every B o r e l set

Z £ P . Let a = Mn\p > 6 = y p . Then 6 <_ v, so there i s a

(unique) measure y i n D such that 8 + y = v . Now Gy = Ga + G8

and v = GB + Gy + h where h i s the g r e a t e s t harmonic minorant of

v i n D . Let w = Gy + h . Then w >_ Ga on E, except p o s s i b l y

on the p o l a r set {GB = 0 0) . A l s o , a(D\(E\P)) = 0 so

a(D\base(E,D)) = 0 . Thus w > Ga throughout D, by 4.1. Hence

v >_ Gy throughout D . •

b) => a) Obviously v >_ Gy y - a.e. For the proof that y

charges p o l a r sets l e s s than v, look ahead to 10.8. •

2 We now take up the proof of the strong domination p r i n c i p l e i n 3R

4.3. Lemma: Let y be a measure on such that i s a

p o t e n t i a l . Then:

a) l i m (U^(x) - yCR 2)1>"(x)) = 0 . | | x | |-*»

b) I f y has compact support, or i f y(dx) = f ( x ) d x where 2

f i s non-negative and l o c a l l y Lebesgue i n t e g r a b l e on H

and tends to zero at i n f i n i t y , then l i m (U M(x) - y ( R 2 ) $ ( x ) ) = 0 .

| |x| |-*»

- 34 -

Proof:

a) U y(x) - yOR 2)$ (x)

(x-y) - $ (x)du(y)

Now i f x > 1 and | |x-y| | >_ 1 then

(x-y) - $ (x) = l o § 1 x-y H

Thus $ (x-y) - $ (x) —*• 0 as | |x| | —>• «>, f o r each f i x e d y e TR.^ .

The d e s i r e d r e s u l t now f o l l o w s from the Lebesgue dominated convergence

theorem, once we have e s t a b l i s h e d the f o l l o w i n g c l a i m ( f o r the

assumption that U y i s a p o t e n t i a l i s equivalent to the f i n i t e n e s s

of l o g + | | x | | d y ( x ) , by 1.6).

2 Claim: For a l l x,y £ H ,

I $ (x-y) - (x) I ± Y n ( l o g 2 + l o g + l |y|

F i r s t note that f o r any r >_ 0,

l o g ( l + r ) <_ l o g 2 + l o g + r .

(Consider the two cases 0 <_ r <_ 1 and r > 1 .) F i r s t suppose

| | x | | •> 1 and || x-y | | >_ 1 . Then

$~(x-y) - $ (x) = j- l o g

Thus i f | |x-y| | >_ | |x| | then

1 x-y 1 |x|

- 35 -

(x-y) - $ (x) 2TT l o g

1 x-y 1 |x |

1 27 l o g 1 _l_ y 1 T X

< ^ l o g d + ||y||)

< |^ (log 2 + l o g + | | y | | )

w h i l e i f I I x-y I I <_ ||x|| then

|$~(x-y) - $ ( x ) | = l o g 1*1 1 x-y | 1 (x-y) + y

= o - l°g —V\ —F\ 2TT ° x-y

i 2 7 l o § 1 + |y| 1 x-y

- 27 l o g ( 1 + I' yl

< ( l o g 2 + log +||y||)

Now suppose ||x-y|| £ 1 . Then

|$"(x-y) - <2>~(x)| = $~(x)

l o g + | |x| | <_j^ l o g + ( | |x-y| | + | |y|

- 17 l o g ( 1 + I'yl

x < 1 . Then

<_ ( l o g 2 + l o g + | |y

F i n a l l y suppose

|$ (x-y) - $ ( x ) | = <3> (x-y)

= ^ l o g + | | x - y | | l ^ l o g + ( | | x | | + ||y||)

< i ^ l o g d + ||y||) < ( l o g 2 + l o g + | | y |

- 36 -

The c l a i m i s now e s t a b l i s h e d .

b) I f u has compact support K then U y ( x ) = 0 f o r a l l x e 1^

whose d i s t a n c e from K i s a t l e a s t 1 . On the o t h e r hand, i f

y(dx) = f ( x ) dx where f s a t i s f i e s the s t a t e d c o n d i t i o n s then

c l e a r l y U y ( x ) —>- 0 as | |x| | —*- » . Now U y = U y - U y and

$ = $ + - $ , and $ + ( x ) = 0 f o r | |x| | >_ 1 . Hence

U y ( x ) - yQR )$(x) —• 0 as

4.4. C o r o l l a r y : L e t y be a measure on such t h a t U y i s

a p o t e n t i a l . Then

l i m i n f ( U y ( x ) - yQR )*(x)) > 0 .

The p r o o f i s c l e a r .

4.5. Lemma: L e t y be a measure on ]R 2 such t h a t U y i s a

p o t e n t i a l . F o r each p o s i t i v e r e a l number r l e t B denote the 2

open b a l l of r a d i u s r c e n t r e d a t 0 i n ~R and l e t u = » r

y' = U 7 • Then: K\B

r 2

a) F o r a l l x e H and a l l p o s i t i v e r e a l numbers r ,

u ^ O O > - ^ y O R 2 \ B r ) i o g ( l + -LiiLL ) + $ ( y ) d y ( y )

| y| |>r

b) F o r a l l p o s i t i v e r e a l numbers e and a l l p o s i t i v e i n t e g e r s

k, t h e r e e x i s t s a p o s i t i v e r e a l number r ^ such t h a t f o r

y y r r0 — r < °° w e n a v e U + e >_ U on B^ r .

- 3 7 -

P r o o f :

a) C o n s i d e r any r i n (0,°°) . Then f o r any x i n R and any y

i n E 2 \ B , r

x-y j | < — r y + M 1 +

so

$(x-y) = - 27 l o g j |x-yI

i " 27 log 1 + x

Thus f o r a l l x i n TR ,

y r U r ( x ) = $ ( x - y ) d y ( y )

E 2 \ B

> - ^ y O R 2 \ B r ) l o g 1 +

+ $ ( y ) d y ( y ) y !>r

b) F o r any x i n TR and r i n (O, 0 0),

y y' U y ( x ) = U r ( x ) + U r ( x )

> U r ( x ) - y O R 2 \ B r ) l o j 1 +

1_ 2IT

y >r l o g | | y | | d y ( y ) ,

by a ) . Now choose r ^ i n ( l , 0 0 ) so t h a t

^yOR 2\B )iog(i+k) + ^ log||yI|dy(y) y | | > r r

The c o n c l u s i o n of b) then f o l l o w s . •

- 38 -

2 4.6. Lemma: Let K be a compact non-polar subset of ]R . Then 2

there i s a non-zero measure A on R which i s supported by K, A 2 such that U i s bounded above on TR

2

Proof: Let D be a Green region i n R c o n t a i n i n g K and l e t G be

the Green f u n c t i o n of D . (We can take D to be an open b a l l contain­

ing K, f o r instance.) Let u = bal(l,K,D) . Then u i s a p o t e n t i a l

i n D . Let A be the Riesz measure of u . Then A i s supported by 2

K . We s h a l l denote the obvious extension of A to a measure on R

by the same l e t t e r A . Now there i s a harmonic f u n c t i o n h i n D

such that = u + h i n D . Since u i s bounded i n D and h, by

c o n t i n u i t y , i s bounded i n a neighbourhood of K, i s bounded i n a A 2 neighbourhood of K . A l s o , U i s continuous on ]R \K . I t now A 2 f o l l o w s from 4.3(b) that U i s bounded above on R

2 Now here i s the strong domination p r i n c i p l e f o r R

4.7. Theorem: Let u be a non-zero measure on R 2 such that U y

2 i s a p o t e n t i a l . Let E be a subset of ]R and l e t v be a super-

2 harmonic f u n c t i o n on R such t h a t :

2 a) l i m i n f (v(x) - pQR )*(x)) > -«

b) v >_ U y on E\P where P i s a po l a r set

c) uQR 2\base(E)) = 0 .

Then v > U P on a l l of R 2 .

Proof: The proof proceeds by reducing to the case of a Green r e g i o n .

For each r > 0 l e t denote the open b a l l of rad i u s r centred 2 at 0 i n ]R and l e t u denote u_ . Choose e > 0 . By r B r

4.5(b), there e x i s t s r ^ i n (0,°°) such that f o r a l l r i n [ r ^ , 0 0 )

- 39 -

u _ U y + -| > U r on B . 2 — r

2

Next, as u ^ 0 but yQR \base(E)) = 0, base(E) ± 0 . Hence E i s

not a p o l a r s e t . (Combine 2.6 and 3.11). Now base(E) contains

E\fringe(E) and, by 3.11, f r i n g e ( E ) i s p o l a r . Thus base(E) i s

not p o l a r . Hence there e x i s t s r ^ i n [ r ^ , 0 0 ) such that base(E) n B i s not p o l a r . Now base(E) n B i s a n a l y t i c ; i n f a c t

r l r i i t i s a countable i n t e r s e c t i o n of open s e t s , by 3.11. Thus there i s a compact non-polar set K contained i n base(E) n B , by 6.23,

r l 7.32, and 7.33 of Helms [ 1 ] . By 4.6, there i s a non-zero measure X

2 X e 2 on TR. supported by K such that U <_ y on TR . Choose a i n

(0,1] . Then U a X = aU A <_ -| on TR2 . Thus f o r a l l r i n [r 0,«0,

y _ U y + e >• U r ' a on B ,

— r

where y = y + aA . As a,A. 0, there e x i s t s r„ i n [ r , ,°°) r,a r 2 1' such that f o r a l l r i n [ r ^ , 0 0 ) ,

y r a0R 2) > yOR 2) •

Then f o r any r i n t ^ , 0 0 ) ,

2 l i m i n f (v(x) - y QR ) * ( x ) ) = +°° . i i i i r ,a I l x l

Choose r , i n [r9,°°) . Let y = M • Then there e x i s t s r , i n 2

[r„,°°) such that on 1R \B we have J r4

2 v > y(]R )$

and

yOR2)* + e > U Y

where the second estimate f o l l o w s from 4.3(b), since y has compact

- 40 -

support. Now choose r,. i n (r ,°°) and consider the Green region

B , which we s h a l l denote by B . Let r 5

h = U Y on 8B

PI(U Y;B) i n B

(Note that the support of y i s a compact subset of B .) Then h

i s continuous on B, and h|B i s the greatest harmonic minorant of

U i n B . Now v + e >_ h on 8B and v + E - h i s lower semi-

continuous on B and superharmonic i n B . Hence v + e - h i s

a non-negative superharmonic f u n c t i o n i n B . Now

v + £ - h ^ U y + e - h on (EnB)\P and U y + e - h >_ U Y - h on

B . Thus v + e - h ^ U T - h on (EnB ) \P . Now base(EnB ,B) r 3 r 3 r 3

contains base(E) n B by 3.11. Hence y(B\base(EnB ,B)) = 0 . r 3 r 3

Y Y Therefore v + e - h > _ U - h throughout B by 4.1, since U - h Y

i s the Green p o t e n t i a l of y r e l a t i v e to B . Thus v + e >_U

i n B . As t h i s i s true f o r a l l r,. i n [r^,°°), v + e >_ U Y on

]R2. Now l e t t i n g r ^ go to <*>, we o b t a i n v + e >_ u y + a ^ on H 2 .

Next, l e t t i n g a decrease to zero we f i n d that v + e >_ U y on ]R2 .

As e > 0 was a r b i t r a r y , i t f o l l o w s that v >_ U y on H 2 , and we

are done. •

4.8. Theorem: Let u,v be measures on H 2 such that U y, U V

are p o t e n t i a l s and u 4 v . Let c be a r e a l number. Then a) and

b) below are e q u i v a l e n t .

- 41 -

a) U V + c >_ U y on K

b) i ) U V + c >_ U y almost everywhere with respect to y 2

i i ) y(Z) <_ v(Z) f o r a l l B o r e l p o l a r subsets of H 2 2 i i i ) V0R ) £ y(R ) .

Proof:

a) => b) i ) i s obvious. The proof of i i ) w i l l be def e r r e d . 2

See 10.8. i i i ) For each x i n H l e t y be the uniform u n i t x

2 i i

d i s t r i b u t i o n on (y e E : j |y-x|| = 1} . Then U V + c dy

x Y

U X dv + c

= U V(x) + c

and

Thus

V1 d y x = U y(x) + c

U V + c > U y

2 2 Hence vQR ) £ yOR ) • This f o l l o w s from 4.3(a) ..

b) = > a) Let E = {U V + c >_ U y} and l e t P = f r i n g e ( E ) . Then

P i s a B o r e l p o l a r set by 3.11(c). Thus y(Z) <_ v(Z) f o r a l l B o r e l

sets contained i n P . Let 8 = y p , a = y - B , y = v - & . Then 2 ct 6 y

a,B,Y are ( p o s i t i v e ) measures on H , U , U , U are p o t e n t i a l s , 2 Y a aQR \base(E)) = 0, and U + c >_ U on E except p o s s i b l y on the

B 2 2 p o l a r set {U = °°} . Al s o yOR) <_aCJR ) so

l i m i n f ( ( U Y ( x ) + c) - a ( R 2 ) $ ( x ) ) >_ c, I l x | |-^°°

- 42 -

by 4.4. Thus, i f a ^ 0, we may conclude that U + c >_U on 2 6 R by applying 4.7. Then by adding U to both sides of t h i s

v y 2 i n e q u a l i t y we o b t a i n U + c >_ U on R

On the other hand, suppose a i s zero. Then y = B <_ v but 2 2

v 0 R ) < y Q R ) so y = v . But we are assuming y f v . •

4.9. C o r o l l a r y . Let y be a measure on R 2 such that U y i s a

p o t e n t i a l and y ^ 0 . Let c be a r e a l number. Suppose U y <_ c

y - a.e. Then U y < c everywhere.

Proof: Take v = 0 i n 4.8. •

F i n a l l y we g i v e a proof of the domination p r i n c i p l e f o r R

F i r s t of a l l , r e c a l l (1.5) that i f y i s a measure on R then

U y i s a p o t e n t i a l i f f x dy(x) i s f i n i t e . Now we prove a lemma.

4.10. Lemma. Let y be a measure on R such that U y i s a

p o t e n t i a l . Let 5 be the centre of mass of y . ( I f y = 0 j u s t

l e t £ be any r e a l number.) Consider the f u n c t i o n

f = y(R)$(£>*) - U y . Then f >_ 0, f i s i n c r e a s i n g on

f i s decreasing on [£,°°), and f (x) —• 0 as [x [ — - «> .

Proof: By the choice of £, y(IR)5

(-°°,C] . Then

y dy(y) . Suppose x i s i n

f (x) = - j yOR)(?-x) + \

1 2

x-y|dy(y)

x - y - |x-y|dy(y)

( x - y ) l ( _ T O ) X ] ( y ) d y ( y )

- 43 -

Hence f i s non-negative and i n c r e a s i n g on . Also f (x) —>- 0

as x —• by the Lebesgue dominated convergence theorem. The

i n t e r v a l i s t r e a t e d s i m i l a r l y . •

4.11. C o r o l l a r y . Let y,v be measures on TR such that U y, U V

are p o t e n t i a l s , and l e t c be a r e a l number. Suppose

U V + c > U y .

Then:

a) vCJR) <_ uQR)

b) i f vQR) = uQR) then c >_ 0 and y and v have the same

centre of mass.

Proof: By 4.10,

l i m (U y(x) - y(IR)$(C,x)) = 0 | x | -**>

and

l i m (U V(x) - vOR)$(n,x)) = 0 |x|-**>

where E, (resp. n) i s the centre of mass of y (resp. v) . The

c o r o l l a r y f o l l o w s immediately from t h i s . •

Now here i s the domination p r i n c i p l e f o r ]R .

4.12. Theorem. Let y be a non-zero measure on E. such that U y

i s a p o t e n t i a l and l e t v be a superharmonic f u n c t i o n on R such

that

- 44 -

a) l i m i n f (v(x) - y ( E ) $ ( x ) ) > -» | x | -*»

b) v > U y almost everywhere w i t h r e s p e c t to y

Then v > U y everywhere on TR .

P r o o f : A f u n c t i o n on TR i s superharmonic i f f i t i s f i n i t e and

concave. Note t h a t such a f u n c t i o n i s a u t o m a t i c a l l y c o n t i n u o u s .

L e t E = {v ^ U y } and l e t W = H\E . Then W i s open. A l s o

y(W) = 0 so U y i s harmonic i n W . S i n c e we a r e i n d i m e n s i o n

one, t h i s j u s t amounts to s a y i n g t h a t on each component of W,

the graph of U y i s a s t r a i g h t l i n e . By the c o n t i n u i t y of U y ,

t h i s a c t u a l l y h o l d s on the c l o s u r e of each component of W . L e t

p e W and l e t C be the component of W c o n t a i n i n g p . Then

C = (a,b) where a e {-°°} u E, b e E u {<*>}, and a < p < b .

A l s o E 4 0 s i n c e y 4 0, so a t l e a s t one of a and b i s f i n i t e

Case 1. a and b b o t h f i n i t e . Then v ( a ) >_U y(a) and

v ( b ) >^U y(b) . A l s o v i s concave, and U y i s a s t r a i g h t - l i n e

f u n c t i o n on [a,b] . Hence v >_ U y on [a,b] . In p a r t i c u l a r ,

v ( p ) >. U P ( p ) .

Case 2. a = b f i n i t e . By a) and 4.10,

l i m i n f ( v ( x ) - U y ( x ) > -°° . x->-°°

Hence v - U y i s bounded below on (-ro,b] . For each x e (-°°,p] ,

t h e r e i s a unique number c ( x ) e [0,1] such t h a t

p = (1 - c ( x ) ) x + c ( x ) b

- 45 -

As x —>• -<*>, c ( x ) —>- 1 . Now

v(p) 1 (1 - c ( x ) ) v ( x ) + c ( x ) v ( b )

1 (1 - c ( x ) ) ( v ( x ) - U U ( x ) ) + U y ( p ) ,

s i n c e U y i s a s t r a i g h t - l i n e f u n c t i o n on (-°°,b] and v ( b ) _> U y ( b ) .

L e t t i n g x go to we o b t a i n

v ( p ) > D y ( p ) .

Case 3. a f i n i t e , b = 0 0 . T h i s i s s i m i l a r to case 2 . •

4.13. C o r o l l a r y . L e t u,v be measures on TR such t h a t U y , U*V

a r e p o t e n t i a l s , and l e t c be a r e a l number. Assume a l s o t h a t

P ^ 0 . Then a) and b) below a r e e q u i v a l e n t .

a) U V + c >_ U P on a l l of TR

b) vQR) <_ uQR), and U V + c >_ U y almost everywhere w i t h

r e s p e c t t o u .

P r o o f :

a) ==> b) A p p l y 4.11(a).

b) = > a) By 4.10, we can take v t o be U V + c i n 4.12.

- 46 -

5. BROWNIAN MOTION PRELIMINARIES

In t h i s s e c t i o n we s h a l l l a y down the n o t a t i o n and terminology

that we s h a l l need f o r Brownian motion. Due to the l a r g e number of

l e t t e r s used up by t h i s n o t a t i o n , we reserve the r i g h t to use these

l e t t e r s f o r other purposes l a t e r on. However, whenever they are used

without e x p l a n a t i o n t h e i r meanings w i l l be as defined i n t h i s s e c t i o n .

5.1. For 0 < t < °°, p t w i l l denote the f u n c t i o n on ]R n defined

by

, s f 1 *n/2 -|| x | | 2 / 4 t

Then:

a) p s a t i s f i e s the heat equation:

8 7 P t = A p t

,n b) p i s a p r o b a b i l i t y d e n s i t y on ]R w i t h mean 0 and

varian c e 2nt:

p t > 0

P t ( x ) d x = 1

xp t ( x ) d x = 0

x| p t ( x ) d x 2nt

c) ( p t) i s a con v o l u t i o n semigroup:

p g ( x - y ) p t ( y ) d y

- 47 -

5.2. The p o i n t a t i n f i n i t y f o r H n w i l l be denoted by 9, and ]R n

o

w i l l denote the space ]R n u {3} where 8 i s c o n s i d e r e d as an

i s o l a t e d p o i n t .

5.3. F o r 0 < t < °°, P w i l l denote t h e f u n c t i o n on H n x B o r e l t 0 0

d e f i n e d by

P t ( x , A ) = \

P t ( y - x ) d y i f x 4 3

AfHRn

1 A ( 3 ) i f x = 3

Then (P ) i s a t e m p o r a l l y homogeneous Markov t r a n s i t i o n f u n c t i o n

on . P (x,A) s h o u l d be thought of as the p r o b a b i l i t y t h a t i f o t

a p a r t i c l e s t a r t s a t x a t a c e r t a i n time and moves a c c o r d i n g to

Brownian motion, i t w i l l be found i n the s e t A t u n i t s of time

l a t e r . Note t h a t a p a r t i c l e s t a r t i n g a t 3 does not go anywhere.

Now suppose 0 <_ s < t ^ < ... < t ^ and A^,. . . ,A^ e B o r e l

A l s o l e t x e ]R n . We wish t o d e f i n e P (x,A ,...,A, ),

O S y t ^ , • • . , t ^ 1 K

which i s to be thought of as the p r o b a b i l i t y t h a t i f a p a r t i c l e

s t a r t s a t x a t time s, i t w i l l be i n A. a t time t . f o r 1 x

i = l , . . . , k . W e l l , c l e a r l y we sho u l d have

^ t / ^ V • pV s ( x ' V

and

P s

. j_ (x,A ,. . . ,A. ) , t ^ , . . . , J-

P (x,da )P (a ,A ,...,A.) ' 1 1' 2' k

A l

- 48 -

so by r e c u r s i o n we are lead to de f i n e

V V

P (x,da ) A ' 1

P (a da ) A 2

C1' C2 1 P t - t ( a k - l ' d

5.4. We d e f i n e ft to be the set of a l l f u n c t i o n s u>: [0,°°] —*• E„

such t h a t :

i s continuous on [0,°°)

a) (« ) = 3

(Then i f co(t) = 3 f o r some t e [0,°°), io i s i d e n t i c a l l y equal to

3 .) We w r i t e CJ f o r the element of P. which i s i d e n t i c a l l y equal 3

to 3 . For each t e [0,°°] we de f i n e

B : fi-*lR3 by Bfc(u)) = u ( t ) ,

and we l e t be the a - f i e l d on Q generated by {B g: 0 <_ s <_ t}

We a l s o w r i t e 8 f o r 8 . CO

I f y i s any a-f i n i t e measure on B o r e l R^ then there i s a o

unique B^-outer measure P y on ft such that i f Ap,A^,...,A^ are

i n B o r e l E ^ and 0 < t n < ... < t. then 3 l k

P y ( B Q e A Q, B e A^-.-.B , V

w ( d x ) P 0 ; t 1 , . . . , t . ( s t » A r - " » A k ) • •v 1 k

(In p a r t i c u l a r , t a k i n g A^ = ... = A^ = R^ we have P M ( B Q e A Q) = y

- 49 -

P P i s c a l l e d the Wiener measure on Q w i t h i n i t i a l measure y .

E y w i l l denote the i n t e g r a l w i t h r e s p e c t t o P y . C u s t o m a r i l y ,

one t a k e s y t o be a p r o b a b i l i t y measure, i n which case P y i s

a l s o a p r o b a b i l i t y measure. However, we s h a l l have o c c a s i o n t o

c o n s i d e r y's which a r e n o t p r o b a b i l i t y measures, and which i n

some c a s e s a r e even i n f i n i t e . P y can a l s o be d e f i n e d as the

unique 8^-outer measure on Q, such t h a t f o r A e B o r e l ]R n and

0 < s < t < ° ° we have

P y ( B Q e A) = y(A)

P y ( B t e A | B ° ) = P s ; t ( B s , A )

The second e q u a l i t y h e r e i s known as the Markov p r o p e r t y . I f

y = S then P y i s a l s o denoted by P X . I f H e B° then x

x \—y P X ( H ) i s a B o r e l f u n c t i o n on and

P y ( H ) = y ( d x ) P X ( H )

f o r any y . F o r any t e [0,°°] we d e f i n e 8 t o be the a - f i e l d

on Q g e n e r a t e d by ^ Bs

: t £. s and we d e f i n e 0 t: —>• by

( 0 t c j ) ( s ) = u)(s+t) . The 9{.'s a r e c a l l e d t r a n s l a t i o n o p e r a t o r s .

S i n c e B • B = B , , 6„ i s ( ^ B , B^)-measurable. Now here i s s t s+t t t

another v e r s i o n o f the Markov p r o p e r t y : F o r any n o n - n e g a t i v e B -

measurable f u n c t i o n <j> on (and any o - f i n i t e measure y on

B o r e l ]Rg) and f o r 0 <_ t < °°, we have

B. Ey(4> • e t |B°). = E

- 50 -

Now we i n t r o d u c e the a - f i e l d s By, By, B, and . We take By

to be the c - f i e l d of a l l P y - m e a s u r a b l e s u b s e t s o f ft . S i n c e P y

i s o - f i n i t e on B ( i n d e e d , on Bg) and o u t e r r e g u l a r w i t h r e s p e c t

to B°,

By = { H e f t ; L c H c M and P U(M\L) = 0 f o r some L,M e B°},

Next, we l e t

By = { H e By: P y(HAF) = 0 f o r some F i n B°},

F i n a l l y we l e t

1 . . 8 "

y

where the i n t e r s e c t i o n s r u n over a l l a - f i n i t e measures y on

B o r e l . (We would get the same t h i n g i f we j u s t l e t t h e i n t e r -

s e c t i o n s r u n over p r o b a b i l i t y measures.) These a - f i e l d s a r e a l l

c l o s e d under the S o u s l i n o p e r a t i o n , and B y+ = By, B = B .

For any A c we l e t

o

D A = i n f { t >_ 0: B t e A}

T A = i n f { t > 0: B t e A} .

i s c a l l e d the debut o f A and T^ i s c a l l e d the h i t t i n g time

of A . I f A i s a n a l y t i c (or more p a r t i c u l a r l y , B o r e l ) then

and T • a r e (B ) - s t o p p i n g t i m e s . T h i s i s the r e a s o n f o r i n t r o d u c i n g

the completed a - f i e l d s . I f we d i d not do t h i s we would have D , T A A

- 51 -

b e i n g (8^)-stopping times f o r A c l o s e d , and ( B ^ + ) - s t o p p i n g times

f o r A open, but f o r A m e r e l y B o r e l , we would be a t a l o s s .

F o r each t , and each f i n i t e y, the map 8 i s ( y B,8 V)-

measurable, where

y8 = {H £ Q: L c H c M f o r some L,M e °8 w i t h P P(M\L) = 0}

and v = yP^ . I t f o l l o w s t h a t 6fc i s (8,8)-measurable. I f $ i s

u x a n o n - n e g a t i v e 8 -measurable f u n c t i o n on Q, then (j) i s P -

measurable f o r y - a.a. x and x t — E (<j>) i s measurable w i t h

r e s p e c t to the c o m p l e t i o n of y, and

E % ) = y(dx)E X((j.) .

I f $ i s B-measurable, then x i — * • E (tj>) i s u n i v e r s a l l y measurable.

In terms of B and B t > the Markov p r o p e r t y can be s t a t e d as

f o l l o w s : F o r any n o n - n e g a t i v e B-measurable f u n c t i o n on fi,

any a - f i n i t e measure y on B o r e l E.^, and any t e [0,°°) a

Ey(<|) • 8 t |B t ) = E t(<f>)

The system (ti,B,B^,B^,B) w i l l be c a l l e d s t a n d a r d Brownian

motion ( i n n d i m e n s i o n s ) .

5.5. Consider a system (A,M,Mt,Xt,P) where:

A i s a set M i s a o-field of subsets of A (M )_ i s an increasing family of sub-o-fields of M t 0<t<°°

- 52 -

f o r each t e [0,°°], X t i s an M t~measurable map of A

i n t o

f o r each co e A, t 1—>• X t(to) i s c o n t i n u o u s on [0,°°)

X^ i s i d e n t i c a l l y e q u a l t o 9

P i s a measure on M which i s a - f i n i t e on MQ

i f A e B o r e l and 0 < s < t < 0 0 then 0

P ( X t e A|M8) = P s ; t ( X s , A ) .

Such a system w i l l be c a l l e d a g e n e r a l i z e d Brownian motion p r o c e s s

(GBMP f o r s h o r t ) . I f i n a d d i t i o n P ( A ) = 1, we s h a l l s i m p l y say

t h a t the system i s a Brownian motion p r o c e s s (BMP f o r s h o r t ) . I f

(A,M,M t,X t,P) i s a GBMP and T i s an ( M F C + ) - s t o p p i n g time then

( A , M , M T + t + , X T + t , P ) i s a l s o a GBMP. T h i s i s known as the s t r o n g

Markov p r o p e r t y . For s t a n d a r d Brownian motion, a v e r y u s e f u l

v e r s i o n of the s t r o n g Markov p r o p e r t y i s : i f <f> i s a n o n - n e g a t i v e

B-measurable f u n c t i o n on Q, T i s a (8 t)-stopping time and u i s

n 9

a a - f i n i t e measure on B o r e l JR™ t h e n

B„ Ey(<j> . eT|BT) = E T(<f>) .

(Note t h a t 6 T i s (B,B)-measurable and B^ i s measurable from B

to the u n i v e r s a l c o m p l e t i o n of B o r e l .) o

Suppose (A,M,M t,X t,P) i s a GBMP . L e t u be the " d i s t r i b u t i o n "

of X Q :

u(A) = P ( X n e A) f o r A e B o r e l .

- 53 -

Suppose y i s o - f i n i t e . For each w e A, the map t *• X t(uj)

i s an element of ft . Thus we have a n a t u r a l map of A i n t o ft;

l e t us c a l l i t 4> . For each H e 8°, i^ - 1[H] e M and

P ( * _ 1 [ H ] ) = P y(H) . Al s o B • IJJ = X . Thus we may regard

(A,M,M ,X t,P) as an "enlargement" of (ft,B°,8°,Bt,Py) . In t h i s

manner we can prove things about GBMP's by f i r s t proving them

about standard Brownian motion, where we can use t r a n s l a t i o n

operators and v a r i o u s other aspects of the s t r u c t u r e of standard

Brownian motion which give us more "handles" f o r computations.

5.6. Theorem. Let u e C 2 ( R n ) and suppose u and a l l i t s f i r s t

two p a r t i a l s are bounded on E. N . Let y be a p r o b a b i l i t y measure t

on R n Au(B )ds . Then s For 0 <_ t < «>, l e t M = u(B t) -

(M t) i s a martingale over (ft,8,B t,P y) .

Sketch of Proof: Using 5.1(a) and i n t e g r a t i n g by p a r t s , one f i n d s

that E y(M t) = udy f o r a l l t . Now apply the Markov property. •

5.7. C o r o l l a r y : Let u e C OR ) and l e t y be a p r o b a b i l i t y rt

.n

,,2^n ry: L e t u e (.

n

measure on TR

Suppose that f o r a l l t e (0,°°), we have

For 0 <_ t < °°, l e t M = u(B t) - Au(B )ds s

TR E

u(x) p t ( x ) d y ( a ) d x < 0 0

and rt

0 n J v(x)p (x)dy(a)dx < °°,

E E

- 54 -

where v = |u|, | |grad u||, and |Au | , r e s p e c t i v e l y . Then (Mfc)

i s a m a r t i n g a l e over ( f i , B , 8 t , P y ) .

2

S k e t c h o f P r o o f : Cut u by C f u n c t i o n s w i t h compact s u p p o r t s ,

a p p l y 5.6 to the c u t f u n c t i o n s , and take a l i m i t . •

5.8. C o r o l l a r y : L e t u e C 20R n) . F o r 0 <_ t < °°, l e t

ft M t = u ( B t ) - A u ( B g ) d s . Suppose t h e r e a r e numbers a,B e [0,°°)

and y e [1,2) such t h a t

II I I Y

v ( x ) <_ a e 3 ' ' X ' I f o r a l l x e ]R n,

where v = |u|, ||grad u||, and |Au | , r e s p e c t i v e l y . Then (M t)

i s a m a r t i n g a l e over (f2,B,8t ,P a) , f o r any a e E. n .

5.9. Theorem: L e t a e R n and l e t T be a ( 8 t ) - s t o p p i n g time

s a t i s f y i n g P 3 ( T < °°) = 1 . Then:

a) E 3 ( | | B T - a | | 2 ) < 2 n E a ( T ) .

b) E a ( T ) i s f i n i t e i f f tf = { | | B _ - a | | 2 : 0 < t < »}

i s P - u n i f o r m l y i n t e g r a b l e . I n t h i s c a s e , we have

e q u a l i t y i n a ) .

P r o o f : A l l m a r t i n g a l e s w i l l be over (Q,B,E^,Pa) . L e t

u(x) = ||x - a | | 2 , f o r x e ]R n . Then Au = 2n, so ( u ( B t ) - 2nt)

i s a m a r t i n g a l e , by 5.8. Hence E a ( | | B g - a|| 2) = 2 n E a ( S ) f o r any

bounded ( B t ) - s t o p p i n g time S, by the o p t i o n a l sampling theorem.

P a r t a) now f o l l o w s from F a t o u ' s lemma. Moreover, i f H i s u n i f o r m l y

- 55 -

i n t e g r a b l e then t a k i n g S = T A t and l e t t i n g t go to i n f i n i t y ,

a i i i 12 a we f i n d t h a t E ( | | B^ - a | | ) = 2nE (T) < °° . C o n v e r s e l y , suppose

E a ( T ) < « . Then E a ( | | B r p t - a l l 2 ) < E 3 ( T ) f o r a l l t , so 2

{ | | B ^ A - a||: 0 <_ t < °°} i s L -bounded and hence u n i f o r m l y

i n t e g r a b l e . Now each component of ( B t - a) i s a m a r t i n g a l e , by

5.8. Thus B T A t - a = E a ( B T - a|5 T A t) f o r a l l t , by the o p t i o n a l

sampling theorem. But then f o r a l l t ,

l|B T, t-a|| 2<EM|B T-a|| 2|B T A t),

by Jensen's i n e q u a l i t y . Thus H i s u n i f o r m l y i n t e g r a b l e . •

5.10. C o r o l l a r y : L e t y be a p r o b a b i l i t y measure on H n and

l e t T be a (B t)-stopping time s a t i s f y i n g P y ( T < °°) = 1 . Then:

a) E y ( | | B T - B Q | | 2 ) < 2 n E y ( T ) .

b) E V ( T ) i s f i n i t e i f f H = { | | B - B | | : 0 <_ t < »}

i s P y - u n i f o r m l y i n t e g r a b l e . In t h i s c a s e, we have

e q u a l i t y i n a ) .

(Note t h a t 5.10 h o l d s even i f y does not have f i n i t e v a r i a n c e . )

5.11. C o r o l l a r y : L e t y be a p r o b a b i l i t y measure on ]R n and

l e t A be a B o r e l s u b s e t of ]R n . L e t T = i n f { t ^ 0 : B t i A} .

Then E y ( T ) < ^ - ( d i a m e t e r ( A ) ) 2 . — zn

5.12. Dynkin's f o r m u l a : L e t D be an open s u b s e t o f E. N and l e t

2

u be bounded and c o n t i n u o u s on D, and C i n D, w i t h Au

bounded i n D . L e t R = i n f { t >_ 0: B f c I D}, l e t a e D, and

- 56 -

l e t T be a ( B t ) - s t o p p i n g time such t h a t T <_ R and E (T) <

Then

(Note t h a t i f D i s bounded then E d ( T ) < » f o l l o w s from T <_R .)

T h i s i s e s s e n t i a l l y Theorem 2, paragraph 1, c h a p t e r 4 of Rao [ 1 ] .

We p o i n t out t h a t the p r o o f Rao g i v e s i s c o r r e c t f o r a e D, but

s e r i o u s l y f l a w e d f o r a e 9D . However, t h i s d i f f i c u l t y can be

overcome.

- 57 -

6. PRELIMINARIES ON BROWNIAN MOTION AND POTENTIAL THEORY

In t h i s s e c t i o n we s t a t e the r e s u l t s that we s h a l l need on the

connections between Brownian motion and p o t e n t i a l theory. These

r e s u l t s are a l l well-known, but i t i s a l i t t l e d i f f i c u l t to give

convenient references f o r them. For the general theory of Markov

processes and p o t e n t i a l theory, the reader may consult the fundamental

papers of Hunt [ 1 ] , [ 2 ] , [ 3 ] , or the books of Meyer [2] or Blumenthal

and Getoor [ 1 ] . In these works, the p o t e n t i a l theory i s defined i n

terms of the process, and the connection between Brownian motion

and c l a s s i c a l p o t e n t i a l theory i s mentioned, but not proved. For

the theory of Brownian motion and c l a s s i c a l p o t e n t i a l theory, the

l e c t u r e notes of Rao [1] are e x c e l l e n t , but some r e s u l t s which we

need f o r B o r e l sets are proved only f o r compact s e t s .

Throughout t h i s s e c t i o n , D i s an open subset of ]R n and

R = i n f { t >_ 0: B I D} .

6.1. Theorem: Let u be a non-negative superharmonic f u n c t i o n i n

D . Let S,T be (B t ) - s t o p p i n g times, w i t h S <_ T . Then f o r any

x e D,

u(x) l E X ( u ( B s ) l { s < R } ) > E X ( u ( B T ) l { T < R } ) .

Thus i f u(x) i s f i n i t e then (u(B ) 1 , ,) i s a supermartingale t t t < K . r

over ( f t , B , ( 8 t ) , P x ) .

6.2. Theorem: Let u be superharmonic i n D . Let

ft^ = { CJ e ft: t *—> u(B t(w)) i s continuous on [0,R(<JJ))

and f i n i t e on (0,R(co))} .

- 58 -

Then:

a) e S o u s l i n B ^ ; hence e B .

b) For any x e D, P x ( f i \ n ) = 0 . (This i s due to Doob [1].)

( I f x e H^\D then {B = x} c {R = 0} so b) i s a c t u a l l y t r u e f o r

a l l x e . )

6.3. Theorem: Let A be an a n a l y t i c subset of lR n and l e t u be

a non-negative superharmonic f u n c t i o n i n D . Then f o r a l l x e D,

b a l ( u , A n D, D)(x) = E X ( u ( B T ) l { T < R } ) ,

where T = T . A

6.4. Theorem: Let A be an a n a l y t i c subset of H n , and l e t

T = T A . Then f o r each x e ]R n, A i s t h i n at x i f f P X(T > 0) = 1.

Thus base(A) = {x e ]R n: P X(T > 0) = 0} and

f r i n g e ( A ) = {x e A: P X ( T > 0) = 1} . A l s o , we have

P X ( B T i base(A), T < °°) = 0 f o r a l l x e ]R n .

6.5. Theorem: Let A be an a n a l y t i c subset of ]R n and l e t T = T.. A Then the f o l l o w i n g are e q u i v a l e n t :

a) A i s p o l a r .

b) P X(T < °°) = 0 f o r at l e a s t one x e ]R n .

c) P X(T < ») = 0 f o r a l l x e ]R n .

Moreover, i f n = 1 or 2 then e i t h e r P (T < <») = 0 f o r a l l

x e H n , or P X(T < ») = 1 f o r a l l x e IR n .

- 59 -

6.6. C o r o l l a r y : Suppose n = 1 or 2 . Then the f o l l o w i n g a r e

e q u i v a l e n t :

a) D i s a Green r e g i o n .

b) P X ( R < °°) = 1 f o r some x e D, or D i s empty.

c) P X ( R < °°) f o r a l l x e D .

6.7. Theorem: L e t E be any su b s e t o f TRn, and l e t x e H n .

Then the f o l l o w i n g a r e e q u i v a l e n t :

a) E i s t h i n a t x .

b) There i s a G s e t H c ]R n w i t h E £ H and P X ( T U > 0) = o H

c) There i s a B o r e l s e t Y £ ]R n w i t h E £ Y and P X ( T y > 0)

d) There i s an a n a l y t i c s e t A £ R n w i t h E £ A and

P X ( T A > 0) = 1 . A

6.8. P r o p o s i t i o n : Suppose n >_ 3 . L e t

ft = {_ e ft: l i m ] |B (to) | | = °°} .

Then ft e B ° , and P X ( f t 1 ) = 1 f o r any x e ]R n . ( T h i s may be

deduced from 6.3 w i t h D = ]R n, u s i n g the s t r o n g Markov p r o p e r t y

and the f a c t t h a t the p o t e n t i a l of a measure w i t h compact support

goes to z e r o a t i n f i n i t y . )

6.9. Theorem: Suppose D i s a Green r e g i o n , and G i s the Green

f unct i o n o f D . Then f o r any x e D and any A & B o r e l D,

f G(x,y)dy =

A P X ( B e A, t < R)dt 0

- 60 -

Sketch of Proof: I t ' s enough to prove the theorem w i t h 1 replaced ________________________ CO

by <f>, where <j> i s a non-negative C f u n c t i o n w i t h compact

support i n D . Let u = G<f> . Then u(x) —>• 0 as x —>• z e 3D,

except f o r a p o l a r set of z's, by 8.31 of Helms [ 1 ] . A l s o , i f

n >_ 3 then u(x) —»- 0 as | |x| | —>- » . Thus by 6.5, 6.6, and

6.8, u(B t) —> 0 as t + R, P X - a.s.

f o r any x e D . Now approximate D by an i n c r e a s i n g sequence of

r e l a t i v e l y compact open subsets and apply Dynkin's formula 5.12. •

- 61 -

7. EMBEDDING MEASURES IN BROWNIAN MOTION IN A GREEN REGION, USING NON-RANDOMIZED STOPPING TIMES.

7.1. Throughout t h i s s e c t i o n D i s a Green region i n ]R n, G i s

the Green f u n c t i o n of D, and R = i n f { t >_ 0: Bfc I D} .

7.2. I f y i s a measure i n D and T i s a (8^)-stopping time

then l e t y^ be the measure on B o r e l D defined by

y T ( A ) = P y ( B T e A, T < R) . The measure y T i s thus obtained by

l e t t i n g y d i f f u s e under Brownian motion up to time T, where only

Brownian paths which stay i n D f o r the whole i n t e r v a l of time [0,T]

c o n t r i b u t e to y^ . Given y, i t i s n a t u r a l to ask what measures

v can be w r i t t e n i n the form v = y^ . I f T i s allowed to be

"randomized" (see 9.2) then Rost [1] has shown that one gets

p r e c i s e l y the measures v s a t i s f y i n g Gv <_Gy . However, i f we

r e q u i r e T to be a genuine ( B t ) - s t o p p i n g time then there i s at

l e a s t one a d d i t i o n a l c o n s t r a i n t on v . For example, i f we take

n = 3, D = H n , y=<5, v = y 6 + -j y where y i s the uniform

u n i t d i s t r i b u t i o n on the u n i t sphere centred at 0 i n ]R n, then

v can be reached from y using a randomized stopping time (the

obvious one!) but not w i t h a genuine stopping time as b) of the

f o l l o w i n g r e s u l t shows.

7.3. P r o p o s i t i o n . Let y be a measure i n D such that Gy i s

a p o t e n t i a l , and l e t v = y^ where T i s some (B )-stopping time.

Then:

a) Gv <_ Gy (so v i s f i n i t e on compact subsets of D)

- 62 -

b) For every B o r e l p o l a r subset Z of D,

v(Z) = y(Z n E ) ,

where E = {x e D: P X(T = 0) = 1}

(Note that E i s u n i v e r s a l l y measurable.)

Proof:

a) Let x e D . Then

Gv(x) = v(dy)G(y,x)

= E y ( G ( B _ , x ) l { T < R } )

y ( d z ) E ^ ( G ( B _ , x ) l { _ < R } )

y(dz)G(z,x) (by 6.1)

= Gy(x)

b) v(Z) - P y ( B T e Z, T < R)

= P y(B_ £ Z, T = 0, T < R) (by 6.2)

y ( d x ) P X ( B Q e Z, T = 0)

y ( d x ) l z ( x ) P * ( T = 0)

= y(Z n E ) ,

x, s i n c e f o r every x, P (T = 0) = 0 or 1 •

S i m i l a r l y , we can show:

- 63 -

7.4. P r o p o s i t i o n . Let u be a measure i n D such that Gy i s

a p o t e n t i a l . Let S,T be (B t)-stopping times w i t h S <_ T .

Then Gy g >_ Gy^

7.5. Lemma. Let y be a measure i n D such that Gy i s a

p o t e n t i a l . Let A be an a n a l y t i c subset of 3Rn and l e t T = T

Let v = y T . Then f o r a l l x i n D,

Gv(x) = E X ( G y ( B T ) l { T < R } )

Proof:

E X ( G y ( B T ) l { T < R } )

y ( d y ) E X ( G ( y , B T ) l { T < R } )

y(dy)bal(G(y,-),A n D,D)(x)

y(dy)bal(G(x,«),A n D,D)(y)

y ( d y ) E y ( G ( x , B T ) l { T < R } )

= E y ( G ( x , B T ) l { T < R } )

v(dz)G(x,z) = Gv(x) .

(by 6.3)

(by 2.11)

To put i t another way, y T = bal(y,A n D,D) .

7.6. C o r o l l a r y . Let y be a measure i n D such that Gy i s a

p o t e n t i a l . Let U be a B o r e l (or j u s t c o a n a l y t i c ) subset of D,

and l e t T = . Suppose D\U i s t h i n at each p o i n t of U .

- 64 -

(That i s , suppose U i s f i n e l y open.) Then y^,(U) = 0 .

P r o o f : As we have j u s t o b s e r v e d , y T = bal(y,D\U,D) . As D\U

i s t h i n a t each p o i n t o f U, base(D\U,D) £ D\U . Now a p p l y 3.4. •

7.7. C o r o l l a r y : L e t y be a measure i n D such t h a t Gy i s a

p o t e n t i a l , and l e t v be a superharmonic f u n c t i o n i n D, and

suppose Gy >_ v i n D . L e t U be a B o r e l ( o r j u s t c o a n a l y t i c )

r e l a t i v e l y compact s u b s e t of D, and l e t T = . Suppose

t h e r e i s a f u n c t i o n h which i s harmonic i n some open s e t c o n t a i n i n g

the c l o s u r e of U, such t h a t Gy >_ h >_ v on U . Then Gy^ >_ v

i n D .

P r o o f : By 7.5, G y T ( x ) = Gy(x) f o r any x e D such t h a t

P X ( T > 0) = 0 . That i s , G y T = Gy on base(D\U,D) (see 6.4).

Suppose x e U . Then P (T < R) = 1 s i n c e U i s r e l a t i v e l y

compact i n D . A l s o , f o r e v e r y • co e {BQ = x} and e v e r y

t £ [ 0 , T ( c o ) ) , Gy(B t(a))) >_h(B t(u>)) . Hence, by 6.2, Gy(B^) >_ h(B T>

P X - a.s. Thus G y T ( x ) >_ E X(h(B^,)) . But by Dynkin's f o r m u l a

5.12, and by the h a r m o n i c i t y of h,

E X ( h ( B T ) ) = h(x) .

Thus

G y T ( x ) >_ v ( x ) .

We have now shown t h a t Gy^, >_ v on D \ f r i n g e ( D \ U ) . But by 3.11,

f r i n g e ( D \ U ) i s p o l a r . Hence Gy^ > v throughout D .

- 65 -

7.8. Lemma: L e t y be a measure i n D such t h a t Gy i s a

p o t e n t i a l . L e t N = {v: v i s a measure i n D and Gv <_ Gy}

Then f o r any compact s e t K c_ D,

sup{v(K): v e N} < 0 0 .

P r o o f : L e t K be a compact subset of D . L e t V,W be open

r e l a t i v e l y compact s u b s e t s of D such t h a t K £ V c w . L e t

y' = bal(y,W,D) . Then y' l i v e s on W, so y'(D) < » . L e t

v = b a l ( l , V , D ) . Then v i s a p o t e n t i a l i n D . L e t X be the

R i e s z measure of v . Then X l i v e s on V . Now f o r any v e N,

v(K) = l d v =

GXdv =

GydX =

GXdv K

GvdX

Gy'dX

GXdy' < y'(D)

7.9. Lemma: L e t y be a measure i n D such t h a t Gy i s a

p o t e n t i a l . L e t (T_^) be a sequence of (B^)-stopping times and

suppose T_ —y T p o i n t w i s e on Q . Then:

a) T i s a (B^.)-stopping time

b) F o r any B o r e l f u n c t i o n t(>: D

j<j>|Gy < 0 0, we have N>Gy_

•' j

c) F o r any c o n t i n u o u s f u n c t i o n

D, we have

H such t h a t

<f>Gy_ .

<(> w i t h compact support i n

- 66 -

d>dv T. 1

P r o o f :

a) i s a s t a n d a r d r e s u l t , and f o l l o w s from the r i g h t - c o n t i n u i t y

of (Bt) . b) I t s u f f i c e s t o c o n s i d e r <J> >_ 0 . F o r t e [0,°°] and

CA) e ft, l e t

f R(o») Z t(o») =

t A R ( _ )

<j>(B_(_))ds

I t f o l l o w s from 6.9 t h a t f o r any (B ) - s t o p p i n g time S,

( * ) J * G y s = E y ( Z s ) .

Now f o r w e { Z Q < » } , the map t —* Z (to) i s f i n i t e

and c o n t i n u o u s on [0,°°] . A l s o Z f c <_ Z^ and

E " ( Z 0 ) - <j)Gp < 0 0 . Thus b) f o l l o w s from the Lebesgue

dominated convergence theorem.

c) From b) i t f o l l o w s t h a t i f <j> i s C w i t h compact s u p p o r t

i n D then

'S'dPn 4>dyn

By 7 . 7 , sup y_ (K) < 0 0 f o r any compact s e t K _£ D . The i i

a s s e r t i o n c) then f o l l o w s by an a p p r o x i m a t i o n argument.

The above p r o o f seems a l i t t l e roundabout. I f y i s f i n i t e

then a more d i r e c t p r o o f can be g i v e n , u s i n g 6.6 and 6.8.

- 67 -

7.10. Lemma: L e t u be a measure i n D . C o n s i d e r a s e t U o f

the form U = V n { v < c } , where V i s open i n D, v i s s u p e r ­

harmonic i n D, and c i s a r e a l number . L e t (T ) be a

sequence of (B^)-stopping times and suppose 1\ —*• T p o i n t w i s e

on ft . Suppose a l s o t h a t u_ (U) = 0 f o r a l l i . Then i

P - ( U ) = 0 .

P r o o f : Suppose u_(U) 4 0 . Then t h e r e i s an open s e t V' which

i s r e l a t i v e l y compact i n V and a r e a l number c' < c such t h a t

u-(U') 4 0, where U ' = V ' n { v < c ' } . L e t f be a n o n - n e g a t i v e

f i n i t e c o n t i n u o u s f u n c t i o n i n D such t h a t f = 1 on V and

f = 0 o u t s i d e some compact s u b s e t of V . L e t g be a non-

n e g a t i v e f i n i t e c o n t i n u o u s f u n c t i o n on (-00 , 0 0] such t h a t g = 1

on (-°°.c'] and g = 0 on [c,°°] . L e t <f>(x) = f ( x ) g ( v ( x ) ) f o r

x e D . Now cf> = 0 o u t s i d e U, so f o r a l l i ,

E y ( * ( B _ H ) - 0 . l l

L e t A = {x e D: f o r a l l i , EX(<t>(B_ ) l r - „ i ) = 0} . Then (A i s T. {T.<R>

u n i v e r s a l l y measurable and) u(D\A) = 0 . L e t x e A . By 6.6

and 6.8 (**t) i s P - a.s. u l t i m a t e l y o u t s i d e any compact s u b s e t

of D . Thus on {T >_ R}, P X - a.s.

•<V 0 = * ( BT ) : L{T<R} •

On t h e o t h e r hand on {T < R}, P X - a.s.

^Tn{T.<R} - * ( BT ) : L{T<R}

- 68 -

s i n c e t *- (B^) i s P A - a.s. c o n t i n u o u s on [0,R) by 6.2.

Thus

• ( BT. ) : L{T.<R} ^ V ^ R } 1 l

P - a.s. on . By the Lebesgue dominated convergence theorem,

i t f o l l o w s t h a t

E X ( < K B T ) 1 { T < R } ) - 0 .

As t h i s i s t r u e f o r a l l x e A, and as u l i v e s on A,

E y ( * ( B T H { T < R } ) = 0 .

Now <j> = 1 on U' . Hence u (U') = 0 . As t h i s i s a c o n t r a d i c t i o n ,

i t must be t r u e t h a t u (U) = 0 . •

Remark. 'The s e t s of the form V n {v < c} (V open £ D, v s u p e r ­

harmonic i n D, c e H) c o n s t i t u t e a base f o r the f i n e t o p o l o g y of

D . A l s o , any c o l l e c t i o n of f i n e l y open s e t s has a c o u n t a b l e sub-

c o l l e c t i o n whose u n i o n d i f f e r s from the u n i o n of the whole c o l l e c t i o n

by o n l y a p o l a r s e t . (See remark f o l l o w i n g V.1.17 o f Blumenthal

and Getoor [1].) Thus, u s i n g the " f i n i t e n e s s p a r t " of 6.2, we see

t h a t U i n the statement of the above lemma can be r e p l a c e d by any B o r e l

f i n e l y open s e t . We s h a l l not need t h i s , though.

At l a s t we can prove the main r e s u l t o f t h i s s e c t i o n , which i s

a l s o one of the main r e s u l t s of t h i s t h e s i s .

- 69 -

7.11. Theorem: Let u be a measure i n D such that Gy i s a

p o t e n t i a l . Let v be a measure i n D such that Gv <_ Gy and

y(Z) <_ v(Z) for every Borel set Z £ {Gv = °°} . Then there i s

a ( B t )-stopping time T such that v = y_ . (Note well that

T i s not randomized.)

Proof: Let S be the set of a l l ( B t )-stopping times S such

that Gy >_ Gv . One can then show that a ( B )-stopping time

T s a t i s f i e s y_ = v i f f T e S and whenever T <_ S e S then

T = S P y - a.s. on {T < R} . (For the "only i f " part, use the f a

that for any non-negative Borel function f i n D and any (8 )-

stopping time S, we have

fR (Gy s)f = E P ( f(B )dt)

SAR

This follows from 6.9 and the strong Markov property.)

One can also show that S does contain such "maximal"

elements. What i s more, t h i s approach i s reasonably constructive

since we don't need Zorn's lemma to produce a "maximal" T; the

p r i n c i p l e of dependent choice s u f f i c e s .

Nevertheless we s h a l l describe another way of producing a

suit a b l e T . We prefer t h i s second approach f o r i t s e x p l i c i t n e s s ,

i n s p i t e of the f a c t that i t i s not quite as s l i c k as the method

outlined above.

Let 1/ be a countable open base f o r D co n s i s t i n g of

r e l a t i v e l y compact subsets of D . Let G be the weakest topology

on D which i s stronger than the usual topology of D and which

- 70 -

makes Gv c o n t i n u o u s . L e t U be the c o l l e c t i o n o f s u b s e t s of

D o f the form V n {Gv < c} where V e V and c i s a p o s i t i v e

r a t i o n a l . Then U i s c o u n t a b l e . A l s o , U i s an open base f o r

the t o p o l o g y G . ( T h i s a s s e r t i o n i s not used below; indeed we

do n o t e x p l i c i t l y make use of the t o p o l o g y G, but i t g i v e s a

p e r s p e c t i v e on the p r o o f . )

L e t ^ - ^ i > i ^ e a sequence i n U i n which each element of

U o c c u r s i n f i n i t e l y many t i m e s . F o r each i , l e t

H = i n f { t > 0: B t I V±} . L e t T Q = 0, and f o r i >. 1 l e t

T._^ + IL • 6 T i f t h i s s t o p p i n g time i - 1 . . c i s i n i l e t T. = I

T. , o t h e r w i s e , i - i

(Note t h a t T. , + H. • 6„, = i n f { t > T. 1 : B„ I U.} .) Then i - i l T. 1 i - i t l i - i

(T^) i s an i n c r e a s i n g sequence of s t o p p i n g times i n S . L e t

T = l i m T. . Then T e S by 7.9. That i s , GX >_ Gv, where i-x»

X = v T . We s h a l l show t h a t X = v . By 7.3(b), X charges p o l a r

s e t s l e s s than u . Hence X(Z) <_ v(Z) f o r ev e r y B o r e l s e t

Z c {Gv = <*>} .

L e t A = {GX > Gv} . We c l a i m X(A) = 0 . Suppose X(A) > 0 .

Then t h e r e i s a p o s i t i v e r a t i o n a l c such t h a t

X({GX > c} n {c > Gv}) > 0 . But {GX > c} i s open i n D, and

hence i s a c o u n t a b l e u n i o n of elements of 1/ . Thus t h e r e e x i s t s

V e V such t h a t

GX > c on V and X(U) > 0,

where U = V n {c > Gv} . L e t I - { i _> 1: U = U} . Then I i s

i n f i n i t e .

B e f o r e p r o c e e d i n g f u r t h e r , l e t us n o t e t h a t i f X,Y a r e

(B ) - s t o p p i n g times then the s t r o n g Markov p r o p e r t y i m p l i e s t h a t

( u x ) Y = UX + Y . Q • Now i f i e I then Gy_ > c > Gv on

X i - 1

U. = U so by 7.7, G ( y T ) > Gv i n D ; hence 1-1 1

T. = T. . + H. • 6_ . Hence by 7.6, y_ (U) = 0 f o r i e I . l l - l l T. n T. l - l l

But then by 7.10, u (U) = 0 . That i s , A(U) = 0 . As t h i s i s

a c o n t r a d i c t i o n , we must have A (A) = 0 . Thus GA <_ Gv A - a.e.,

and, by 2.2, A(Z) <_ v(Z) f o r e v e r y B o r e l p o l a r subset Z of D .

Hence GA <_ Gv, by the d o m i n a t i o n p r i n c i p l e 4.2. T h e r e f o r e

GA = Gv . From t h i s i t f o l l o w s t h a t A = v . •

As we s h a l l see s h o r t l y , the above theorem i s not the b e s t

p o s s i b l e r e s u l t o f i t s t y p e . I b e l i e v e the f o l l o w i n g i s the b e s t

p o s s i b l e r e s u l t , but I am u n a b l e to prove i t a t the p r e s e n t time.

7.12. C o n j e c t u r e : L e t y,v be measures i n D, and suppose Gy

i s a p o t e n t i a l . Then the f o l l o w i n g a r e e q u i v a l e n t :

a) There i s a ( B f c ) - s t o p p i n g time T such t h a t y_ = v .

b) Gy >_ Gv, and t h e r e i s a B o r e l s e t A £ D such t h a t

v(Z) = y(ZnA) f o r a l l B o r e l p o l a r s e t s Z £ D .

(Note a) = > b) h o l d s by 7.3.)

Now l e t us c o n s i d e r an example t h a t shows t h a t 7.11 i s not

b e s t p o s s i b l e .

- 72 -

7.13. Example: Suppose D = ]Rn where n = 3 . (We make these

assumptions to keep the computations from getting too messy.)

Let u = 6 . Let ( r j ) ^ e a s e c l u e n c e of p o s i t i v e r e a l numbers

decreasing to 0, such that there e x i s t m,M e ( l , 0 0 ) with

r. m < — ^ — < M for a l l j . — r —

3 + 1

Let

S. = {x e D: I I x l I = r.} 3 1 1 1 1 3

H. = {x e D: I I x l I < r.} J 3

A. = { x e D : r.,_ < l l x l l <r.} . 3 3 + 1

1 1 1 1 - j

Let (s.) be a sequence i n [O,00) with £s. £ 1, and l e t v be 3 j 3

the s p h e r i c a l l y symmetric measure on D which l i v e s on u S_. and

assigns mass s . to S. f o r each j . Suppose v does not p i l e

up too f a s t around 0, i n the sense that £js. < °° . Then there 3 J

i s a (Bj.)-stopping time T such that u = v .

Proof: F i r s t l e t us show that Gv grows very slowly at 0 compared

to Gy . For each i , l e t v. = v„ . Then 3 S.

3

Gv(x) = <

s. 4 i T r .

3

^— for l l x l l < r,

— r r for x > r . x 1 1 1 - 3

(It follows that Gv <_ Gy .) Thus for x e Afc,

- 73

Gv(x) = 4lT

k s.

"3=0 r j l*lT j=k+l ^

so

Gv(x) = v . I | x |

j=0 J ' j j=k+l Gy(x) j r. .Jr.. j

0 0 r, A r . < I s. 1

L e t us c a l l the l a t t e r q u a n t i t y a ^ . Then

oo oo c o r A r

k=0 j=0 2 k=0 j

CO /- CO \

j=0 J 1 k=j r j J

j=0 J ^ £=0

F o r the remainder of the p r o o f , l e t us not make such s p e c i a l

assumptions on v . L e t us assume, r a t h e r , t h a t v i s such t h a t

Gv s a t i s f i e s the p r o p e r t i e s we have j u s t v e r i f i e d . E x p l i c i t l y ,

we s h a l l assume t h a t v i s a measure i n D such t h a t Gv <_ Gy

and t h e r e a r e p o s i t i v e r e a l s ( a i . ) such t h a t £ a, < 0 0 and f o r k

each k,

Gv < a. G i n - k y

Now f o r any E _c D, l e t C(E) denote the o u t e r c a p a c i t y of E

r e l a t i v e t o D . (See Helms [ 1 ] , c h a p t e r 7.) F o r each j , l e t

- 74 -

be a c e r t a i n compact s u b s e t " o f S.. which w i l l be more e x p l i c i t l y

s p e c i f i e d s h o r t l y , and l e t X^ = bal(y,E^.,D) . Now A l i v e s on

E. . A l s o Gy = -,—— on E., so 1 4irr. i 3

VV = 4-r. = C(S.) '

J J

Thus f o r x 6 H., J

A (E ) C(E ) r \ ( v ) > J J 3 ___ VJA . v,A7 / / . _ \ <")_ -\ /-i/-c \ o_ j v ' - (ATr)(2r ) C(S ) 8TT_

C(E ) = C ( ^T ^ r T ^ N x l |Gy(x)

C(E )||x|| = 2C(S ) r G y ( x ) '

Thus f o r x e A., 3

C(E.) G A j ( x ) ^ 2 C ( S * ) S G y ( x ) •

Choose k so t h a t a. < ^ — f o r a l l i > k . Now the E.'s can J - 2M - j

be chosen so C(E^)/C(S_.) i s any number between 0 and 1 .

Thus f o r i > k we can choose E. so t h a t - 3

C(E.)

2C(S.)M " " j = a.

L e t E = (D\KL ) u ( u E.) . L e t A = b a l ( y , E , D ) . Then GA = Gy j__k 2

on DXH^., and f o r j k, GA •> GA_. >_ a^.Gy >_ Gv on A . Hence

GA > Gv . Now A = y T T, where U = T_, by 7.5. A l s o — U L

C(E.)

j _ k L ( S j ; j>k 2

- 75 -

so E i s t h i n a t 0 . (See p r o o f o f theorem 10.21 of Helms [1] ~

t h i s i s the W i e n e r - B r e l o t t e s t f o r t h i n n e s s . ) Thus P°(U > 0) = 1,

so Ujj({0}) = 0 . (Once (B^) l e a v e s 0 i t cannot come back to

i t , because {0} i s p o l a r . ) Thus n e i t h e r p^ nor v charges

p o l a r s e t s . Thus, by 7.11, t h e r e i s a (8 t)-stopping time V

such t h a t

( V v = v '

But then v = p^, by the s t r o n g Markov p r o p e r t y , where T i s the

(8 ) - s t o p p i n g time U + V • 9 .

t u Q

The p r o o f of the above example i s a r a t h e r d e l i c a t e " b a l a n c i n g

a c t " . We have to take the E j ' s b i g enough so t h a t

bal(Gp,E,D) >_ Gv . On the o t h e r hand, i f we took them too b i g

then E would not be t h i n a t 0, and bal(p,E,D) would j u s t be

p, so t h a t 7.11 would not a p p l y .

- 76 -

8. EMBEDDING MEASURES IN BROWNIAN MOTION IN I T OR ]R_, USING

NON-RANDOMIZED STOPPING TIMES.

8.1. I f y i s a measure on H n and T i s a (B^)-stopping time

t h e n , f o r the purposes of t h i s s e c t i o n , y^ w i l l denote the measure

on B o r e l ]R n d e f i n e d by y^,(A) = P y ( B T e A, T < °°) . (Of c o u r s e

the c o n d i t i o n "T < °°" i s s u p e r f l u o u s ; we i n c l u d e i t j u s t f o r

emphasis.)

As i n the p r e v i o u s s e c t i o n , we c o n s i d e r the q u e s t i o n : What

measures v can be w r i t t e n i n the form v = y^ ?"

Of c o u r s e i f n >_ 3, t h i s i s a s p e c i a l case of the q u e s t i o n

c o n s i d e r e d i n the p r e v i o u s s e c t i o n . However, i f n = 1 or 2 then n

-K i s not a Green r e g i o n , so the above q u e s t i o n i s then not subsumed

under the one of the p r e v i o u s s e c t i o n .

I f T i s a l l o w e d to be randomized (and y,v a r e f i n i t e ) then

Rost [2] has g i v e n a n e c e s s a r y and s u f f i c i e n t c o n d i t i o n f o r v to

be e x p r e s s i b l e i n the form v = y T ; R o s t ' s c r i t e r i o n i s i n terms

o f a - e x c e s s i v e f u n c t i o n s .

Skorohod [1] showed t h a t ( i n the case n = 1) i f y = & f o r

some x e 1R and i f v i s a p r o b a b i l i t y measure on H w i t h f i n i t e

v a r i a n c e and mean e q u a l to x, then t h e r e i s a randomized (8 )-

s t o p p i n g time T such t h a t E (T) < °° and v = u . (Then of x 1 1 c o u r s e E (T) = y v a r i a n c e ( v ) ; the f a c t o r o f - j a r i s e s because we

have n o r m a l i z e d Brownian motion to get the c l e a n e s t form f o r Dynkin's

f o r m u l a 5.12.) Note t h a t by 4.10, a p r o b a b i l i t y measure v on H

has c e n t r e of mass e q u a l t o x i f f U V i s a p o t e n t i a l and U V <_ U y

- 77 -

where y = . Of c o u r s e , t h i s r e s u l t o f Skorohod's i s the g r a n d ­

f a t h e r o f a l l t h e s e embedding r e s u l t s .

Dubins [1] and Root [ l j gave p r o o f s of Skorohod's r e s u l t which

produce non-randomized T's .

Doob (see Meyer [3]) has noted t h a t i f y,v a r e any p r o b a b i l i t y

measures on H t h e n t h e r e i s a ( B ^ ) - s t o p p i n g time T such t h a t

y_ = v . T h i s somewhat d i s q u i e t i n g r e s u l t l e a d s us to r e s t r i c t the

c l a s s of s t o p p i n g times t h a t w i l l be " a l l o w e d " , as f o l l o w s .

8.2. D e f i n i t i o n . L e t T be a ( B f c ) - s t o p p i n g time, and l e t y be

a measure on H n such t h a t U y i s a p o t e n t i a l . We s h a l l say t h a t

T i s y - s t a n d a r d i f f whenever R,S a r e ( B t ) - s t o p p i n g times and y R V S y R y S R < S < T then U and U a r e p o t e n t i a l s and U > U

8.3. D i s c u s s i o n of 8.2. I f n >_ 3 then any T i s y - s t a n d a r d , by y T

7.4. Suppose n <_ 2 and T i s y - s t a n d a r d . Then U i s a p o t e n t i a l

and i s l e s s than or e q u a l to U y . Hence y_ has t o t a l mass g r e a t e r

than or e q u a l to t h a t o f y, by 4.8 ( i f n = 2) or 4.11 ( i f

n = 1) . Hence i n d i m e n s i o n one or two, i f T i s y - s t a n d a r d then

T i s P y - almost s u r e l y f i n i t e . Suppose n = 1, y = 6^,

T = i n f { t > 0: B t = 1}, and v = y_ . Then v = &1 . Thus T i s 2

not y - s t a n d a r d . Now suppose n = 2 , S = {x e H : | | x | | = r } ,

y i s the u n i f o r m u n i t d i s t r i b u t i o n on S-, T = i n f { t > 0: B t e S^},

and v = y_ . Then T i s P y - a.s. f i n i t e by 6.5, s i n c e S^ i s

not a p o l a r s e t . C l e a r l y t h e n , v i s the u n i f o r m u n i t d i s t r i b u t i o n

on S^ . Now

- 78 -

irV) =

w h i l e

U V ( x ) =

2TT

2TT

log 2 for ]Ix < 2

l o g I|x I f o r x > 2

0 f o r | | x | | <_ 1

1 2TT log I|x|| for | |x j | > 1

Thus T i s not u-s t a n d a r d .

8.4. Lemma. L e t y,v be measures on H n such t h a t U y , U V a r e

p o t e n t i a l s . Then

$ ( x , y ) d y ( x ) d v ( y ) < °°

Hence

U y d v and U Vdy

b o t h make sense, and they a r e e q u a l . A l s o , they a r e not e q u a l t o -°°

P r o o f : I f n ^ 3, we have n o t h i n g t o p r o v e . I f n = 1 or 2 the n

y and v a r e f i n i t e . I f n = 1 then by 4.10,

l i m (yOR)$a,y) -I y I _ H x j

$ ( x , y ) d y ( x ) ) = 0,

and by 1.5,

then by 4.3,

$(£,y)dv(y) i s f i n i t e ; a l s o $ = -$ . i f n = 2

2 -l i m (yOR )$ (y) -

Ixl

$ ( x , y ) d y ( x ) ) = 0,

and by 1.6, ( y ) d v ( y ) i s f i n i t e . A l s o , f o r any n,

y

s e t s .

- 79 -

$ ( x , y ) d y ( x ) i s c o n t i n u o u s , and so i s bounded on compact

8.5. Theorem: L e t y be a measure on E n (where n = 1 or 2)

such t h a t U y i s a p o t e n t i a l . L e t T be a (8^)-stopping time

such t h a t E y ( T ) < °° . Then T i s y - s t a n d a r d , and

E y ( f ( B f c ) d t ) = ( U y - U T ) f

f o r any n o n - n e g a t i v e B o r e l f u n c t i o n f on E

P r o o f : F i r s t l e t f be a non-negative compactly supported C

f u n c t i o n on E n , and l e t u = U f . Then u i s C°°, grad u = U g r a d f ,

and Au = - f . L e t M = u(B t> + f (B ) d s , f o r 0 <_ t < » . By 0

5.8, (M t) i s a m a r t i n g a l e over (ft,8,8 t,P ) f o r any a e I T

But u + c* i s bounded, where c = (See 4.3 f o r n = 2 and

4.10 f o r n = 1 .) Thus E y(|M f c|) < » f o r a l l t . Hence (M )

i s a m a r t i n g a l e over (ft,8,8 t,P y) . Of c o u r s e , (M ) has r i g h t -

c o n t i n u o u s ( i n f a c t , c o n t i n u o u s ) p a t h s . Now u(x) = g(x) - c$ ( x ) ,

where g i s bounded on E ; a l s o $ (x) <_ | |x - y| | + $ (y) f o r

a l l x,y i n

E y ( $ ( B Q ) ) =

From 5.10, t o g e t h e r w i t h the f a c t t h a t

dy < «>, i t then f o l l o w s t h a t {u(B_ ) : 0 < t < °°} TAt —

i s P y - u n i f o r m l y i n t e g r a b l e . (Here we use the f a c t t h a t E y ( T ) < °° .)

A l s o TAt fT

f ( B )ds < s — f ( B )ds and E M ( f ( B )ds) < b E y ( T ) < o ° ,

where b = sup f . Thus {M_ A t: 0 <_ t < °°} i s P y - u n i f o r m l y

- 80 -

i n t e g r a b l e . Hence E^CHj.) = E y ( M Q ) , by the o p t i o n a l sampling

theorem. That i s ,

U f d u T - E y ( f ( B )ds) = U fdy .

In p a r t i c u l a r , Urdy^, i s f i n i t e . Hence (as f can be non-zero)

yT

U i s a p o t e n t i a l . Now interchanging orders of i n t e g r a t i o n (which

i s j u s t i f i e d by 8.4) we o b t a i n rT E y ( f ( B )ds) = 0

y yT (U y - U ) f .

As t h i s holds f o r a l l compactly supported non-negative C f u n c t i o n s

f on R n, we f i n d that U y >_ U ^, and then a monotone c l a s s

argument shows that the formula holds f o r a l l non-negative B o r e l

f u n c t i o n s f on H n .

Now c l e a r l y we can apply the above argument to any (B )-

stopping time which i s <_ T, s i n c e such a stopping time w i l l a l s o

have f i n i t e P y- e x p e c t a t i o n . I t then f o l l o w s that i f R,S are

(8 t)-stopping times and R <_ S <_ T then

yR yS U , U are p o t e n t i a l s ,

and

U R > u s

That i s , T i s y-standard. •

8.6. Terminology. A measure y on ]R w i l l be c a l l e d good i f f

y has compact support and U Y i s f i n i t e and continuous. Note that

- 81 -

n i f <j) i s a compactly s u p p o r t e d bounded B o r e l f u n c t i o n on H and

y(dx) = <j>(x)dx, then y i s good. A l s o , by 4.6 of t h i s work t o g e t h e r

w i t h 6.21 o f Helms [ 1 ] , any compact n o n - p o l a r s u b s e t o f ]R n c a r r i e s

a non-zero good measure.

8.7. Theorem: L e t y be a measure on ]R n, where n = 1 or 2,

n y ' and l e t ( y . ) . T be a n e t of measures on H such t h a t U i s a l l e i

p o t e n t i a l f o r a l l i e I , and cfidy. <j>dy f o r a l l compactly

s u p p o r t e d c o n t i n u o u s f u n c t i o n s <f> on H Then

a) U ^ 1 dy U y dy f o r a l l good measures y on ]R n

Now suppose a l s o t h a t the n e t ( U dv) converges t o a f i n i t e l i m i t

f o r some non-zero measure v on 3Rn such t h a t U V i s a p o t e n t i a l .

Then

b) U y i s a p o t e n t i a l .

u i c) The net (U ) converges u n i f o r m l y on compact s e t s t o

U + C, where C i s some f i n i t e n o n - n e g a t i v e c o n s t a n t ,

d) U 1 dy U y - C dy f o r a l l good measures y on ]R n .

(Remark: C need not be z e r o ; see 8.15 f o r a n a t u r a l example of t h i s . )

n y

P r o o f : I f y i s a good measure on H then U + i s c o n t i n u o u s and,

as n <_ 2, i s a l s o compactly s u p p o r t e d . P a r t a) f o l l o w s i mmediately

from t h i s , upon i n t e r c h a n g i n g o r d e r s of i n t e g r a t i o n . Now suppose v

i s as i n t h e statement of the theorem. Then f o r some 1^,

sup

^ 0

U_ dy < Hence l i m ( s u p y.({x e H n : ||x|| >^r})) = 0

- 82 -

But t h e n , s i n c e we a l s o know t h a t the net (j<|>dy^) converges t o a

f i n i t e l i m i t f o r each c o n t i n u o u s compactly s u p p o r t e d f u n c t i o n $

on ]R n, t h e r e e x i s t s i ^ >_ i ^ such t h a t

(*) sup y . (R n) < 0 0

^ 1

Thus u0R n) < °° and | f dy,^ fd y f o r a l l bounded c o n t i n u o u s

f u n c t i o n s f on H Next, i t i s easy t o show t h a t

(**)

Thus by Fat o u ' s lemma,

u. U y < l i m i n f U 1

U_dv <_ l i m i

U X d v

Hence U ydv < 0 0 . T h e r e f o r e , as v 4 0, U y i s not i d e n t i c a l l y

i n f i n i t e . Hence, by 1.5 ( i f n = 1) or 1.6 ( i f n = 2 ) , U_ i s

f i n i t e everywhere, and U y i s a p o t e n t i a l . A l s o , f o r any x i n

Rn, U V d i f f e r s from vOR n)$ (x,') by a bounded c o n t i n u o u s f u n c t i o n .

I t f o l l o w s t h a t the n e t (U_ (x)) converges t o a f i n i t e l i m i t f o r

each x i n ]R n . But f o r any x,y i n ]R n,

| u \ x ) - U ^ ( y ) | < y.0R n)||x - y| I .

y i Thus by (*), {U_ : i >_ i^} i s e q u i c o n t i n u o u s ( i n d e e d , u n i f o r m l y

u.

L i p s c h i t z ) so i n f a c t (U ) converges u n i f o r m l y on compact s u b s e t s

of R n . L e t u be i t s l i m i t , which i s f i n i t e and c o n t i n u o u s on H n . L e t u = U y - u_

(-°°,00]-valued on H n . A l s o

Then u i s l o w e r - s e i n i c o n t i n u o u s and

U 1 dy udy f o r a l l good measures

- 8 3 -

Y on 1 R In p a r t i c u l a r , t h i s convergence h o l d s f o r y of the CO

form y(dx) = <j>(x)dx, where <j> i s C and compactly s u p p o r t e d

on ]R n . Thus y = -Au . ( C l e a r l y u i s l o c a l l y Lebesgue

i n t e g r a b l e , and so d e f i n e s a Schwartz d i s t r i b u t i o n on H n .) Hence

t h e r e i s a harmonic f u n c t i o n h on H n such t h a t u = v almost

everywhere w i t h r e s p e c t t o Lebesgue measure on ]R n, where

v = U Y - h . By a v e r a g i n g over a b a l l o f r a d i u s 6, c e n t r e d a t

x, and l e t t i n g 6-1-0, one e a s i l y checks t h a t a c t u a l l y

u(x) = v ( x ) f o r a l l x e E. n .

Now by (**) , U Y <_ u_ . Thus h >_ 0 . But a n o n - n e g a t i v e harmonic

f u n c t i o n on H n i s c o n s t a n t . ( T h i s i s P i c a r d ' s theorem - 1 . 1 1 of

Helms [ 1 ] ; t r u e f o r any n .) Thus C i s j u s t the c o n s t a n t v a l u e

of h .

8.8. Lemma. L e t y,a be measures on H n , where n = 1 o r 2,

such t h a t U Y , U 0 a r e p o t e n t i a l s . Suppose U Y - C >_ U A , where

C e [0,°°) . Suppose a l s o t h a t yQR n) = aOR n) . Then C = 0 .

P r o o f : I f n = 1, t h i s f o l l o w s e a s i l y from 4.10. Suppose n = 2 .

L e t y be the u n i f o r m u n i t d i s t r i b u t i o n on the sphere o f r a d i u s 1

2 c e n t r e d a t 0 i n TR. . Then

U Y * Y - C = ( U Y - C)*y

> U ° * Y = U A * Y ,

2 where * denotes c o n v o l u t i o n . L e t m = y(R ) . Then

m = ( u * y ) 0 R 2 ) = ( a * y ) 0 R 2 ) . A l s o U Y * Y = U Y * y = - $ ~ * u = - U Y and

- 84 -

U ° * Y = -u° . Thus by 4.3, as ||x|

|U P Y ( x ) - m*(x) | 0, and

|u°* Y(x) - m*(x) | 0

From t h i s i t f o l l o w s t h a t C must be 0 . •

8.9. Lemma: Suppose M i s a f a m i l y of measures on E.n, where

n = 1 or 2, such t h a t

a) For each u e M, U y i s a p o t e n t i a l .

b) sup{pOR n): u e M} < °° .

c) i n f { U y ( x ) : y e M , x e K} > -» f o r some compact n o n - p o l a r

s e t K c n n .

Then l i m sup y({x eE. n: | |x| | >_r}) = 0 . r-x» yeM

P r o o f : By 4.6, t h e r e i s a non-zero measure y c a r r i e d by K such

t h a t U Y i s bounded above on ]Rn . L e t L = i n f { U y ( x ) : y e M , x e K}

Then f o r any y e M ,

U Ydy = U Mdy > Ly(K)

Y

Then from b ) , the upper boundedness of U , and the f a c t t h a t

U Y ( x ) — • - c o as | j x | | —> °°, i t f o l l o w s t h a t

l i m sup y ( { x eE. n: ||x|| >_r}) = 0 . r-x» yeM Q

- 85 -

8.10. Lemma: L e t u be a f i n i t e measure on R n . L e t T be a

c o l l e c t i o n of (B^)-stopping times such t h a t i f T e T and S i s

a ( B t ) - s t o p p i n g time s a t i s f y i n g S <_ T, then S e T . Suppose

t h a t

l i m sup u_({x e ] R n : ||x|| >_r}) = 0 . r-x° TeT

Then

l i m sup P y ( T > t ) = 0 . t-*» TeT

P r o o f : F o r each n a t u r a l number i , l e t = i n f { t _L 0: I l Bt l I _. i ^ '

Then each R± i s P y - a.s. f i n i t e , by 5.11. I f T e T then

T A R. e T and P y ( T _> R.) = y _ A R ({X e ]R n: | |x| | >_ i } ) . Thus i

l i m sup P y ( T >_ R.) = 0 . TeT 1

F i x e > 0 . Then f o r some i , sup P y ( T >_R.) __ f" • Next, f o r TeT 1 1

some t e [ 0, °°) ,

Then

P y ( R ± > t ) < |

sup P y ( T >_ t ) <_ £ TeT •

8.11. Theorem: Let u be a measure on ]R n (where n = 1 or 2)

such t h a t U y i s a p o t e n t i a l . L e t T be a s e t of good measures on H n

such t h a t whenever a, B a r e measures on ]R n such t h a t U a , a r e

p o t e n t i a l s and U dy >_ U dy f o r a l l y e f , then u " > U £

Suppose (T_^) i s a sequence o f y - s t a n d a r d ( B t ) - s t o p p i n g times

c o n v e r g i n g p o i n t w i s e on ft to a f u n c t i o n T : ft -*- [0, °>] . (Note t h a t

then T i s a (B^)-stopping time because (B^) is right-continuous.)

- 86 -

C o n s i d e r t h e f o l l o w i n g s t a t e m e n t s :

a) There i s a measure a on ]R n , such t h a t U ° i s a p o t e n t i a l yT

and aQR n) = P QR n) , w i t h U 1 >_ U a f o r a l l i .

y T b) U i s a p o t e n t i a l and

T i f y T U dy + D dy

f o r a l l measures y e V

c) T i s y - s t a n d a r d .

Then a) => b) => c)

P r o o f : a) => b ) .

L e t H = {H : H i s a (B ) - s t o p p i n g time and H <_ T. f o r some i} . y H a Z 1

Then U > U f o r a l l H e H s i n c e each T. i s y - s t a n d a r d . T h e r e f o r e — x

l i m sup y„({x e !Rn : | |x| | >_ r}) = 0 , by 8.9. Hence r+~ Heh1 H

l i m sup P y(T^. >_ t ) = 0 , by 8.10. T h e r e f o r e T i s P y - a.s. t-*» i l —

f i n i t e , so f o r each bounded c o n t i n u o u s f u n c t i o n <j> on H ,

<f> dy_ •+ i

4> dy_ ,

s i n c e ( Bt ) i s c o n t i n u o u s on [0, °°) .

L e t v be the u n i f o r m u n i t d i s t r i b u t i o n on t h e s u r f a c e o f t h e

u n i t sphere c e n t r e d a t 0 . Then P

r o > T. -U 1 dv = -U V dy T.

x

U V dy T. x

U dv

U a dv .

- 86a -

Thus t h e sequence ( T.

U 1 dv) i s bounded. By a sub-subsequence

argument, u s i n g 8.7 and 8.8, we o b t a i n t h e statement b) .

b) => c ) . C o n s i d e r any bounded ( 8 t ) - s t o p p i n g time Q . We have

<f>dp T.AQ l TAQ

n H T ± A Q y

f o r a l l bounded c o n t i n u o u s f u n c t i o n s <j> on ]R , and U > U

f o r a l l i . Thus, by t h e same method we used i n c o n c l u d i n g t h e p r o o f

of (a => b) , we f i n d t h a t f o r a l l measures y e f ,

T.AQ U 1 dy

TAQ , U v dy .

PT.AQ Y T . Now f o r each i , U 1 >_ U 1 as T_ i s y - s t a n d a r d . I t f o l l o w s

t h a t

U ^ ^ dy > f dy

f o r a l l y e T , whence U ^ >_ U

In p a r t i c u l a r , i f R, S a r e ( B t ) - s t o p p i n g times s a t i s f y i n g

R < S then

y y y T T TARAt „ TASAt T T T u > u > u

y y T

f o r a l l t e [0, °°) . ( A l s o U >_ U , so t h e t o t a l mass of y T i s

>_ t h e t o t a l mass of y , whence t h e s e two t o t a l masses must i n f a c t be

e q u a l and thus T must be f i n i t e P y - a . s . ) . I f i n a d d i t i o n S <_T ,

we have / R A t > / s A t >

- 87 -

f o r a l l t e [0, °°) . L e t t i n g t -*• 0 0 and a p p l y i n g 8.7 and 8.8 we

o b t a i n

U R > U S . •

8.12. C o r o l l a r y : L e t y be a measure i n H n , where n = 1 o r

2, such t h a t U y i s a p o t e n t i a l . L e t T be a ( 8 t ) - s t o p p i n g y T

time. Then T i s y - s t a n d a r d i f f U i s a p o t e n t i a l and

y y U T A t > U T f o r a l l t e [0,°°) .

P r o o f : (==>) f o l l o w s immediately from 8.2, the d e f i n i t i o n o f

" y - s t a n d a r d " .

(<==) T a k i n g t = 0 and a p p l y i n g 4.8 ( i f n = 2) or 4.11

( i f n = 1) we f i n d t h a t y T ( R n ) = yOR n) . For each t ,

T A t i s y - s t a n d a r d , by 8.5. Now a p p l y 8.11, w i t h a = y

and T. = T A i .

T

8.13. Theorem: L e t y be a measure on H n , where n = 1 or 2,

such t h a t U y i s a p o t e n t i a l . L e t T be a (B^)-stopping time.

L e t m be Lebesgue measure on H n . Then:

a) F o r any measure v on ]R n, U y - U V i s d e f i n e d m - a.e.,

and i t s m - i n t e g r a l over any compact subset of H n makes

sense, though i t may be -H>° .

b) T i s y - s t a n d a r d i f f f o r each compact s u b s e t K of ]R n,

u T y U - U dm i s f i n i t e and e q u a l t o E ( K

1 K ( B )ds) 0

- 88 -

P r o o f : a) U M i s everywhere d e f i n e d , does not assume the v a l u e

and i s m - i n t e g r a b l e over any compact su b s e t of 3Rn . Now

v v

e i t h e r the same i s t r u e of U , or e l s e U i s i d e n t i c a l l y -°°

on ]R n except p o s s i b l y f o r a s e t of m-measure 0 where i t i s

u n d e f i n e d . T h i s p r o v e s a ) . M T A i y

T

b) (=>) As T i s y - s t a n d a r d , U + U except

p o s s i b l y on a p o l a r s e t , by 8.7. A l s o U y - U T A ± dm K

= E y ( T A i

l K ( B s ) d s )

f o r any i , and any compact su b s e t K of R. L e t t i n g i go

to °°, we o b t a i n the d e s i r e d e q u a l i t y by the monotone convergence

y y T theorem. The f i n i t e n e s s i s c l e a r , s i n c e U and U a r e

p o t e n t i a l s .

(<=) F o r any t e [0,°°) and any compact s e t K £]R",

r M - i 'K

U y - U T A t dm

= E y (

< E y (

(•TAt l K ( B s ) d s )

l K ( B s ) d s )

(by 8.5)

u K T I T - U dm ; K

l^TAt P I J » P<j» hence U >_ U m - a.e. A l s o U i s a p o t e n t i a l by the

y T A t y T f i n i t e n e s s assumption. T h e r e f o r e U >_ U everywhere. Now

a p p l y 8.12.

- 89 -

Remark: Note t h a t E y ( T 1 (B )ds) i s t h e P y - e x p e c t e d amount of

0 K s

time t h a t (B f c) spends i n K up t o time T . Observe the a n a l o g y

between (*) o f the p r o o f of 7.9, and the f o r m u l a e s t a b l i s h e d i n the

above theorem.

8.14. C o r o l l a r y : L e t y be a measure on E. n, where n = 1 o r

2, such t h a t U y i s a p o t e n t i a l . L e t R,S be (8 t)-stopping

t i m e s , and c o n s i d e r the (8 ) - s t o p p i n g time T = R + S • 8 . t R

Suppose R i s y - s t a n d a r d and S i s y_,-standard. Then T i s K

y - s t a n d a r d .

P r o o f : A p p l y 8.13 i n c o n j u n c t i o n w i t h the s t r o n g Markov p r o p e r t y .

8.15. In o r d e r t h a t a (8 t)-stopping time T be y - s t a n d a r d , i t

i s not s u f f i c i e n t t h a t

fT E y ( 1„(B ) d s )

0 K s

be f i n i t e f o r each compact s e t K _c ]R n . T h i s i s shown by the

f o l l o w i n g example. L e t n = 2 . L e t y be the u n i f o r m u n i t

d i s t r i b u t i o n on {x e H : ||x|| = 2} . L e t A = {x e TR : ||x|| = 1},

and l e t T = i n f { t > 0: B^ e A} . As A i s not p o l a r , T i s P P -

a.s. f i n i t e (because n < 3 ) , by 6.5. Then c l e a r l y v = y_ i s

the u n i f o r m u n i t d i s t r i b u t i o n on A . Thus

U V ( x ) =

0 f o r | | x | | <_ 1

- ^ l o g | | x | | f o r ||x|| > 1

- 90 -

w h i l e

U M ( x ) =

- ^ l o g 2 f or

_ 1_

x|| < 2

^ l o g | | x | | f o r ||x|| > 2

Thus U M ^_U V, so T cannot be y - s t a n d a r d . Now f o r i = 3,4,5,...

l e t A± = {x e H 2 : ||x|| = 1} and l e t = i n f { t > 0: B t e k± u A}

Then each T^ has f i n i t e P y - e x p e c t a t i o n , by 5.11, and so i s y-

s t a n d a r d , by 8.5. Now v. E y^, c o n s i s t s o f a u n i f o r m d i s t r i b u t i o n i

o f mass e. on A. p l u s a u n i f o r m d i s t r i b u t i o n of mass 1 - E. on X I 1

A, where e^ e (0,1) i s determined by the e q u a t i o n

l o g i + (1 - s ^ l o g 1 = l o g 2

T h i s f o l l o w s from Dynkin's f o r m u l a 5.12, and the f a c t t h a t l o g | | • | |

2> i s harmonic i n H^\{0} . Thus E. = 2

l l o g 1 Now one e a s i l y computes t h a t

v. U 1 4- v _ l o g 2

2TT '

A l s o , i t i s c l e a r t h a t T. + T l

(The f a c t t h a t e. x

0 can then

be used t o g i v e a n o t h e r p r o o f t h a t T i s P y - a.s. f i n i t e . ) I t

2 f o l l o w s t h a t f o r any compact s e t K £ H ,

E y ( l K ( B s ) d s ) = K

( Uy - ( D T _ l o g _ 2 ) ) f

which of c o u r s e i s f i n i t e .

We a l s o remark t h a t the sequence o f measures (v^) y i e l d s a

n a t u r a l example i n which the C of theorem 8.7 i s non-zero;

C = _ l o g 2 2TT

h e r e . •

- 91 -

8.16. Lemma: L e t H be a bounded c o - a n a l y t i c subset of E. n . L e t

T = i n f { t > 0: _ t I H} . Then f o r any measure y on TRn such t h a t

U y i s a p o t e n t i a l , T i s y - s t a n d a r d , and

U T ( x ) = E X ( U y ( B _ ) ) f o r a l l x e R n .

P r o o f : I f v i s any measure on E. n such t h a t U V i s a p o t e n t i a l

then E V ( T ) < » by 5.11, so T i s v - s t a n d a r d by 8.5 ( i f n £ 2) .

In p a r t i c u l a r , t h i s i s t r u e i f v = y, and a l s o i f v = 6^ f o r any

x e H n . Hence

E X ( U P ( B _ ) ) = E X ( 4 > ( y , B _ ) ) y ( d y ) ,

where the r e q u i r e d i n t e r c h a n g e i n o r d e r of i n t e g r a t i o n i s j u s t i f i e d

by 8.4 ( i f n <_ 2) . We a l s o have

U T ( x ) = E y ( $ ( x , B _ ) ) y ( d y )

Now we c l a i m t h a t f o r a l l x,y e ]R n,

E ( $ ( y , B _ ) ) = E y ( $ ( x , B _ ) ) .

L e t V be a bounded open s u b s e t of ]R n such t h a t H c V

Case 1. x e IR n\V . Then P X ( T = 0) = 1, so

E X ( $ ( y , B _ ) ) = *(y,x) = *(x,y) .

subcase a. y e E. n\V . Then E y ( $ ( x , B _ ) ) = $(x,y) .

subcase b. y e V . L e t W be an open s u b s e t of V such t h a t

H £ W and W £ V . Then $(x,*) i s bounded and c o n t i n u o u s on W,

- 92 -

and harmonic i n W . A l s o T <_ i n f { t >_ 0: B t £ W} . Thus by

Dynkin's f o r m u l a 5.12,

E y ( $ ( x , B _ ) ) = $(x,y) .

Case 2. y e ]R n\V . T h i s i s s i m i l a r t o case 1.

Case 3. x,y e V . For each z i n V, l e t h ( z , ' ) be the g r e a t e s t

harmonic minorant of $(z,«) i n V . L e t G be the Green f u n c t i o n

of V . Then G = $ - h on V x V . A l s o T < i n f { t > 0: B t I V} P Z -

a.s. f o r each z e V . Thus

E x ( G ( y , B _ ) )

A l s o

= b a l ( G ( y , - ) , V \ H , V ) ( x ) (by 6.3)

= b a l ( G ( x , - ) , V \ H , V ) ( y ) (by 2.11)

= E y ( G ( x , B _ ) ) (by 6.3) .

E X(h(y,B_).)

= h(y,x) (by 5.12)

= h ( x , y )

= E y ( h ( x , B _ ) ) (by 5.12)

E X ( $ ( y , B _ ) ) = E y ( $ ( x , B _ ) )

i n t h i s c a s e t o o , and the c l a i m i s e s t a b l i s h e d .

Hence

- 93 -

8.17. C o r o l l a r y : L e t H be a bounded c o - a n a l y t i c subset of H ,

and l e t y be a measure on H n such t h a t U y i s a p o t e n t i a l .

L e t

T = i n f { t > 0: B I H} .

Suppose v i s superharmonic on R.n, v <_ U y , and t h e r e i s a

f u n c t i o n h which i s harmonic i n some open s e t c o n t a i n i n g the

y y T c l o s u r e of H, such t h a t v <_ h <_ U i n H . Then v <_ U

P r o o f : I f n _ 3, t h i s i s j u s t a s p e c i a l case of 7.7. I f n = 1

or 2, we need o n l y emulate the p r o o f of 7.7, u s i n g 8.16 i n p l a c e

of 7.5. •

8.18. Lemma: L e t y be a measure on H n . L e t H be a c o -

a n a l y t i c subset of H n such t h a t ]R n\H i s t h i n a t each p o i n t of

H . L e t

T = i n f { t > 0: B I H} .

Then y_(H) = 0 .

P r o o f : L e t ( v^) be a n i n c r e a s i n g sequence of bounded open s u b s e t s

of E. n whose u n i o n i s ]R n . For each i , l e t

R. = i n f { t > 0: B_ i V.} and l e t y. be the r e s t r i c t i o n of y l — t I I y.

to B o r e l V , . Then by 7.6, P ^"(B- e H, T < R ) = 0 f o r a l l i .

L e t t i n g i — • ~, we o b t a i n y_(H) = 0 . ^

8.19. Lemma: L e t (A,F,Q) be a a - f i n i t e measure space. L e t

T be a c o l l e c t i o n of [0,°°]-valued F-measurable f u n c t i o n s on A .

- 94 -

Suppose the l i m i t of each i n c r e a s i n g sequence i n T b e l o n g s to T,

and T 4 0 • Then T c o n t a i n s Q - e s s e n t i a l l y maximal elements.

P r o o f : By the H a u s d o r f f m a x i m a l i t y p r i n c i p l e , T c o n t a i n s a maximal

c h a i n C . As Q i s a - f i n i t e , some c o u n t a b l e subset CQ of C

has the same Q - e s s e n t i a l supremum as C . Then T = sup CQ belongs

to T and T i s a Q - e s s e n t i a l l y maximal element of T . However,

the f u l l axiom of c h o i c e i s not r e a l l y needed to prove t h i s lemma.

Here i s a p r o o f which uses o n l y the p r i n c i p l e of dependent c h o i c e .

L e t f be a s t r i c t l y p o s i t i v e F-measurable f u n c t i o n on A such

t h a t fdQ < 0 0 . D e f i n e g: T —> [0,°°) by g(T) = (1 - e T ) f d Q

L e t TQ e T and l e t M Q = s u p { g ( T ) : T Q <_ T e T} . I f i _> 1 and

T Q,...,T ,MQ,...,M have been d e f i n e d , l e t

M = s u p { g ( T ) : T <_ T e T} and choose ^ e T such t h a t

T <_ T± and g ( T ± ) _> M ± - 2 _ 1 . Then ( T ^ i s an i n c r e a s i n g

sequence i n T . L e t . T = l i m T. . Then T e T . I f S e T and i - y o o

T <_ S then f o r any i , g(S) <_ <_ g(T ) + 2 _ 1 <_ g(T) + 2" 1; hence

g(S) <_ g(T) . Thus S = T Q - a.e. Hence T i s Q - e s s e n t i a l l y

maximal i n T . •

8.20. Theorem: L e t u,v be measures on ]R n such t h a t U y, U V

a r e p o t e n t i a l s . Suppose

a) U y > U V .

b) u(Z) <_ v(Z) f o r every B o r e l s e t Z £ {U V = »} .

c) yOR n) = v O R n ) , on n >_ 3 .

Then t h e r e i s a y - s t a n d a r d ( 8 t ) - s t o p p i n g time T such t h a t y T = v

- 95 -

P r o o f : We c o u l d prove t h i s i n much the same way as we proved 7.11.

For the sake of v a r i e t y though, l e t us g i v e a somewhat d i f f e r e n t

p r o o f . L e t T be the s e t of a l l y - s t a n d a r d ( 8 ^ ) - s t o p p i n g times

T such t h a t U >_U . Then 0 e T, so T 4 0 . A l s o the l i m i t

of any i n c r e a s i n g sequence i n T b e l o n g s to T, by 8.11 ( i f

n <_ 2) or 7.9 ( i f n >_ 3 ) . Thus, by 8.19, T c o n t a i n s a P M -

e s s e n t i a l l y maximal element T . L e t A = y T . We s h a l l show t h a t

A = v . Of c o u r s e , we have o n l y to show t h a t U A <_ U V . L e t

A = {U A > U V} . We c l a i m A(A) = 0 . Suppose n o t . Then t h e r e

e x i s t s a number c and a bounded open s e t W such t h a t

A(H) > 0,

where

H = W n {U* > c > U V} .

L e t

S = i n f { t > 0: B I H} .

AS v

Then S i s A-standard by 8.16, and U >_ U V by 8.17. A l s o , by

the s t r o n g Markov p r o p e r t y , A = y , where T' = T + S • 0 . In

the case n <_ 2, T' i s y - s t a n d a r d by 8.14. Thus T' e T . Now

T <_T' . Hence P P ( T 4 T') = 0 . T h e r e f o r e

u T = u T , . But u T ( H ) > 0 w h i l e , by 8.18, y^,, (H) = 0 . T h i s

c o n t r a d i c t i o n shows t h a t we must have A(A) = 0 . Thus

A v U -<_U A - a . e . I f Z i s a B o r e l p o l a r s e t then

- 96 -

A(Z) = A(Z n {U X <_ U V})

= A(Z n {U X = 00} n { n X <_ u V}) (by 2.2)

= A(Z n {U V = 0 0 } )

<_ y(Z n {U V = »}) (by 6.5)

1 y(z) .

Thus by the d o m i n a t i o n p r i n c i p l e , <_UV . (See 4.2 i f n > 3,

4.8 i f n = 2, or 4.13 i f n = 1 ; note t h a t i n the case n <_ 2

we are assuming t h a t uQR n) = vOR n) .)

8.21. Remark: Baxter and Chacon [1] have proved the f o l l o w i n g

theorem: " L e t y,v be p r o b a b i l i t y measures on ]R n such t h a t

y v U , U are p o t e n t i a l s . Suppose

a) U y > U V .

b) l i m ( U P ( x ) - U V ( x ) ) = 0 . I I x I I "><J0

c) U V i s f i n i t e and c o n t i n u o u s on TRU . Then t h e r e i s a

( B t ) - s t o p p i n g time T such t h a t y_ = v . I f v has a f i n i t e

second moment so does y, and T can be chosen so t h a t

I xI I 2dv(x) = |x|| 2dy(x) + 2 n E M ( T )

I f n >_ 3, c o n d i t i o n b) can be dropped."

Now by Lemma 5 of Bax t e r and Chacon [1], i f y,v a r e f i n i t e

measures on ]R n such t h a t U y , U V are p o t e n t i a l s and U P >_ U V ,

then

- 97 -

IxI I^dv (x ) = | x| |^du(x) + 2n U y ( x ) - U V ( x ) d x

A l s o , by 8.13, i f T i s y - s t a n d a r d then

U y ( x ) - U P T ( x ) d x = E y ( T ) .

( I f n >_ 3, use 6.9 i n p l a c e of 8.13). Thus 8.20 i s a g e n e r a l i z a t i o n of the

c i t e d theorem of Baxter and Chacon. In p a r t i c u l a r : I f n >_ 3, we no l o n g e r

r e q u i r e y and v to have the same t o t a l mass; moreover, they need

not even be f i n i t e measures. C o n d i t i o n b) of the theorem of Bax t e r

and Chacon has been c o m p l e t e l y e l i m i n a t e d . C o n d i t i o n c) has been

p a r t l y e l i m i n a t e d . No c o n t i n u i t y assumption on U V i s made. A l s o ,

the f i n i t e n e s s assumption on U V has been r e l a x e d ; f o r example, i t

s u f f i c e s t h a t U V be f i n i t e y - a.e. Of cou r s e 8.20 i s not b e s t

p o s s i b l e ( u n l e s s n = 1 ) . T h i s i s shown by 7.13, i n the case n = 3.

I b e l i e v e the f o l l o w i n g r e s u l t i s "best p o s s i b l e " .

8.22. C o n j e c t u r e : L e t y,v be measures on ]R n such t h a t U y ,

U V a r e p o t e n t i a l s . Then t h e r e i s a y - s t a n d a r d ( 8 t ) - s t o p p i n g

time T such t h a t y^ = v i f and o n l y i f the f o l l o w i n g 3 c o n d i t i o n s

a r e s a t i s f i e d :

a) U y > U V .

b) There i s a B o r e l s e t A £ ]R n such t h a t v(Z) = y(Z n A)

f o r every B o r e l p o l a r s e t Z £ R n .

c) n > 3, or y ( R n ) = vCR n) .

- 9 8 -

9. RANDOMIZED STOPPING TIMES, AND ENLARGEMENTS OF PROBABILITY SPACES

Suppose y i s the u n i t p o i n t mass a t 0 i n H n , and v i s

the u n i t measure on 3Rn which has h a l f o f i t s mass a t 0 and the

o t h e r h a l f u n i f o r m l y d i s t r i b u t e d on S = {x e H n : ||x|| = 1 } .

I f v = u T f o r some (8^) - s t o p p i n g time T then

P°(T>0) = 1 so {0} cannot be a p o l a r s e t , f o r i f i t were we would

have P ° ( B T e {0}) = 0 , c o n t r a d i c t i n g v({0}) 4 0 . Thus we can

o n l y have v = y^ i f n = 1 . But f o r any n , t h e r e i s an o b v i ous

"randomized s t o p p i n g time" T such t h a t v = "y " . Namely, l e t

0 w i t h p r o b a b i l i t y 1 / 2

T w i t h p r o b a b i l i t y 1 / 2

T h i s m o t i v a t e s the f o l l o w i n g d e f i n i t i o n s .

9 . 1 . D e f i n i t i o n : L e t (A,F) and ( X , A ) be measurable spaces.

A randomized random v a r i a b l e ( r r v , f o r s h o r t ) i n ( X , A ) (over

(A,F)) i s a map x : A p r o b a b i l i t y measures on A , such t h a t f o r

e v e r y A i n A , x(')(A) i s F-measurable. I f X i s a

t o p o l o g i c a l space, we s h a l l speak s i m p l y o f r r v ' s i n X ; i t w i l l

be u n d e r s t o o d t h a t A = B o r e l X , the a - f i e l d g e nerated by the

open s e t s of X .

9 . 2 . D e f i n i t i o n : I f (A,F) i s a measurable space and ^t^Q<t«=°

i s an i n c r e a s i n g f a m i l y of sub - a - f i e l d s of F then a randomized

- 99 -

( F t ) - s t o p p i n g time ( ( F ^ J - r s t , f o r s h o r t ) i s a r r v x i n [ O , 0 0 ]

o v e r ( A , F ) such t h a t f o r 0 _ < t < ° ° , x ( ' ) ( [ 0 , t ] ) i s F -

measurable. (That i s , i n t u i t i v e l y , the c o n d i t i o n a l p r o b a b i l i t y t h a t

we s t o p by time t depends o n l y on the i n f o r m a t i o n t h a t we have by

time t .)

9.3. Examples: L e t ( A , F ) be a measurable space.

L e t ( X , A ) be another measurable space and l e t f : A •+ X .

D e f i n e x : A -*• p r o b a b i l i t y measures on A by

= 6 f ( _ ) ;

t h a t i s , x(u)(A) = 1 ( f ( _ ) ) .

F o r any A £ X , xA ( f ( ' ) ) 1 S F - m e a s u r a b l e i f f f X [ A ] e F .

Thus x i s a r r v i n ( X , A ) i f f f i s a r v i n ( X , A ) ; t h a t i s ,

i f f f i s measurable. I n t h i s c ase we c a l l x the r r v a r i s i n g

from the r v f .

Now l e t ( F ) be an i n c r e a s i n g f a m i l y of sub - a - f i e l d s t 0_t<°°

of F , l e t T be a r v i n [ O , 0 0 ] , and l e t x be the r r v i n

[0,°°] a r i s i n g from T . For any t e [0,°=) , x ( - ) ( [ 0 , t ] ) i s

F t - m e a s u r a b l e i f f {T <_ t} e F • Thus x i s a r s t i f f T i s a

s t o p p i n g time.

9.4. L e t ( A , F ) and ( X , A ) be measurable s p a c e s .

Then r e l a t i v e t o a measure Q on F , r r v ' s i n ( X , A ) may

be more or l e s s i d e n t i f i e d w i t h measures on the p r o d u c t o - f i e l d

F ® A , whose p r o j e c t i o n on A i s e q u a l to Q . T h i s i s made

- 100 -

p r e c i s e by the f o l l o w i n g r e s u l t ,

Theorem:

a) I f x i s a n r r v i n (X ,A) then to x(^) (H(w)) i s

F-measurable f o r any H i n F ® A . (Where H(„)

= the s e c t i o n of H over _ .)

b) I f x i s an r r v i n (X ,A) and Q i s a measure on F ,

and i f we d e f i n e y on F ® A by

y(H) = Q(d_)x(o))(H(_))

then y i s a measure on F ® A and y(F><X) = Q(F) f o r

F i n F .

c) I f Q i s a s e m i f i n i t e measure on F and A i s c o u n t a b l y

generated, and i f x a n d x ' a r e r r v ' s i n (X ,A) which

g i v e r i s e to the same measure y on F ® A ( a c c o r d i n g

to the fo r m u l a i n p a r t b ) ) then the s e t

E = (X r" X M

bel o n g s t o F and Q(E) = 0 .

d) Suppose Q i s a o - f i n i t e measure on F , y i s a measure

on F ® A such t h a t y(F*X) = Q(F) f o r a l l F i n F ,

A i s c o u n t a b l y g e n e r a t e d , and f o r each F i n F w i t h

Q(F) f i n i t e , the measure A •-»• y(A*A) i s i n n e r r e g u l a r w i t h

r e s p e c t to a semicompact c l a s s c o n t a i n e d i n A . Then t h e r e

e x i s t s a r r v x i n (X ,A) such t h a t y(H) = Q(d_)x(u)(H(_))

- 101 -

f o r a l l H e F ® A . Note t h a t x i s Q - e s s e n t i a l l y unique by c ) .

(A semicompact c l a s s i s a c o l l e c t i o n C of s e t s , such t h a t

c o u n t a b l e £ C , n = 0 i m p l i e s n = 0 f o r some f i n i t e

= c0 .)

We s h a l l not prove t h i s theorem h e r e . L e t us remark t h a t

P a c h l [1] has r e c e n t l y shown t h a t i n d ) , the c o n d i t i o n t h a t A be

c o u n t a b l y generated can be dropped, p r o v i d e d we a r e w i l l i n g to have

each x ( w ) d e f i n e d o n l y on a sub - a - f i e l d A^ of A , where

f o r each H e F ® A , H(co) i s i n A f o r Q -almost a l l co and co * *

co >->• x(w) (H(co)) i s Q -measurable, where Q i s the c o m p l e t i o n of

Q . I n the same paper, P a c h l showed t h a t the requirement of i n n e r

r e g u l a r i t y w i t h r e s p e c t to a semicompact c l a s s i s , i n a sense,

n e c e s s a r y .

9.5. L e t us a l s o remark t h a t f o r a s e p a r a t e d c o u n t a b l y generated

measurable space (X , A ) , the f o l l o w i n g a r e e q u i v a l e n t :

a) Every f i n i t e measure on A i s i n n e r r e g u l a r w i t h r e s p e c t

to a semicompact c l a s s (depending on the measure).

b) (X , A ) i s B o r e l i s o m o r p h i c to a u n i v e r s a l l y measurable

subspace of TR. .

c) (X , A ) i s B o r e l i s o m o r p h i c to a u n i v e r s a l l y measurable

subspace of a P o l i s h space.

d) (X , A ) i s u n i v e r s a l l y measurable i n every c o u n t a b l y

s e p a r a t e d measurable space i n which i t i s B o r e l embedded.

(A P o l i s h space i s a s e p a r a b l e c o m p l e t e l y m e t r i z a b l e t o p o l o g i c a l

space.)

- 102 -

L e t us say t h a t (X,A) i s a u n i v e r s a l l y measurable space i f f

(X,A) i s a measurable space which i s B o r e l i s o m o r p h i c t o a

u n i v e r s a l l y measurable subset of 3R .

9.6. P r o p o s i t i o n : L e t (A,F,Q) be a measure space and l e t

(F ) . be a r i g h t - c o n t i n u o u s i n c r e a s i n g f a m i l y of sub - a - f i e l d s

t U t "^00

of F • Assume Q i s a - f i n i t e on F^ .

L e t x be a r r v i n [0,°°] over (A,F) . Then the f o l l o w i n g

a r e e q u i v a l e n t :

a) There i s an ( F t ) - r s t T such t h a t x = X Q " a.e. f dQ = 0 b) Whenever t e [0,°°), f e L 1 ( Q ) . such t h a t

F

f o r a l l F e F , and g e C[0,°°] such t h a t g = 0 on

[t,°°] , then Q(du>) X(w)(ds)f(_)g(&) = 0

c) For each t e [0,°°) , x ( # ) ( [ 0 , t ] ) i s Q - a.e. e q u a l to

an F^-measurable f u n c t i o n .

P r o o f : a) => b) I f A e B o r e l [ 0 , t ] , then x ( . ) ( A ) i s F -

measurable.

b) => c) L e t t e [0,~) and l e t ^ = x(»)([0,t)) . Then

\\) h dQ = ^ E(h|F t)dQ

E(ip|F t)E(h|F t)dQ

E(i|>|F )h dQ for a l l h e L 1(Q) . Hence i|> = E(i^ | Ffc)

- 103 -

(Note t h a t we can t a l k of c o n d i t i o n a l e x p e c t a t i o n s w i t h r e s p e c t t o

F because Q i s a - f i n i t e on ^ t •)

Now x ( ' ) ( [ 0 , t ] ) = l i m x ( * ) ( [ 0 , t + 2 ~ k ) ) . U s i n g the r i g h t -k-*>°

c o n t i n u i t y o f (F^_) we o b t a i n the d e s i r e d c o n c l u s i o n .

c) => a) For r r a t i o n a l , 0 < r < 0 0 we can choose f u n c t i o n s

A r on A such t h a t :

0 < A < 1 — r —

A i s F -measurable r r

r < r ' => A < A , — r — r

A r = x ( O ( [ 0 , r ] ) Q - a.s.

Now f o r each to e A , t h e r e i s a unique p r o b a b i l i t y measure x(to)

on [0,°°] such t h a t t e [0,°°) , x ( t o ) ( [ 0 , t ] ) = i n f A (to) . r>t r

Then x i s the sought r s t . (The r i g h t c o n t i n u i t y o f (F^) i s

used h e r e too.) •

9.7. D e f i n i t i o n : An enlargement o f a measure space (A,F,Q) i s

a p a i r c o n s i s t i n g o f a measure space (M, G, R) and a (G,F)-

measurable map \p : M + A such t h a t ip(R) = Q ( i . e . R ( ^ _ 1 [ F ] ) = Q(F)

f o r a l l F e F) .

9.8. L e t (M,G,R,IJ0 be an enlargement of a a - f i n i t e measure

space (A,F,Q) . L e t (X,A) be a u n i v e r s a l l y measurable space,

and l e t f be a r v i n (X,A) over (M,G) .

- 104 -

D e f i n e g : M ->• A x X by g(p) = (Mp) . f (p)) • Then g i s

(G, F ® A)-measurable. L e t p = g(R) on F ® A . Then the p r o j e c t i o n

of u on A i s ijJ(R) = Q . Thus, by 9.4, t h e r e i s a Q - e s s e n t i a l l y

unique r r v x i n (X,A) over (A,F) such t h a t

V(H) = Q(d_)x(u)(H(_)) f o r a l l H e F ® A

In t h i s way, a r v over an enlargement g i v e s r i s e to a r r v over

the o r i g i n a l space. One e a s i l y checks t h a t i f (R ) . i s a _ _eA

d i s i n t e g r a t i o n o f R w i t h r e s p e c t to , F , then x(^) = ^R^)

f o r Q-a.a. _ e A .

9.9. Any r r v over a measure space a r i s e s , i n the f a s h i o n d e s c r i b e d

i n 9.8, from a r v i n an enlargement. L e t us d e s c r i b e e x p l i c i t l y why

t h i s i s s o.

L e t (A,F) , (X,A) be measurable spaces and l e t x be a r r v i n

(X,A) over (A,F) . L e t Q be a measure on F > and l e t us show

how x a r i s e s from a r v over an enlargement of (A.F,Q) .

W e l l , l e t M = A x x

G = F ® A

R(G) = Q(d_))xCw) (G(_)) f o r G e G

= p r o j e c t i o n o f M on

f = p r o j e c t i o n of M on X .

Then (M,6,R,^) i s an enlargement of (A,F,Q) , f i s an r v i n

(X,A) over (M,6) , and x a r i s e s from f i n the manner d e s c r i b e d

i n 9.8.

- 105 -

9.10. Given two r r v ' s , they may be r e a l i z e d as r v ' s i n a common

enlargement, by a p r o c e s s s i m i l a r to t h a t d e s c r i b e d i n 9.9. However,

the j o i n t d i s t r i b u t i o n o f the two r v ' s i s not determined by the two

r r v ' s a l o n e ; i t depends on the enlargement. T h i s i s why we cannot

j u s t work w i t h r r v ' s i n g e n e r a l , b u t must a l s o c o n s i d e r enlargements.

9.11. D e f i n i t i o n . A f i l t e r e d measurable space i s a system

( A , F , F T ) » where ( A , F ) i s a measurable space and ^t^0<t<°° ^ S

an i n c r e a s i n g f a m i l y o f sub - a - f i e l d s of F .

A f i l t e r e d measure space i s a system ( A , F , F ,Q) , where

( A J F J F ^ ) i s a f i l t e r e d measurable space and Q i s a measure on

F which i s a - f i n i t e on FQ . (The l a s t assumption i s made to

ensure t h a t we can c o n s i d e r c o n d i t i o n a l e x p e c t a t i o n s w i t h r e s p e c t to

any F .)

9.12. D e f i n i t i o n . An enlargement o f a f i l t e r e d measure space

( A , F , F T , Q ) i s a f i l t e r e d measure space (M,G,G^,R) t o g e t h e r w i t h

a map ip such t h a t (M,G,R,ijj) i s an enlargement of the measure

space (A »F,Q) and such t h a t i n a d d i t i o n , \JJ i s (G^, F^)-measurable

f o r 0 £ t < °° .

An o p t i o n a l enlargement of ( A , F , F T , Q ) i s an enlargement

(M,G,Gt,R,iJ;) such t h a t f o r each t e [0,«) and each g e l^CM.G ,R)

we have E(g|*,F) = E(g|i|;,F ) Q-a.s.

We remark t h a t o p t i o n a l enlargements a r e e s s e n t i a l l y the

" d i s t r i b u t i o n a l e n largements" of Bax t e r and Chacon [ 3 ] . These a u t h o r s

i n t r o d u c e d t h i s n o t i o n to f a c i l i t a t e the d i s c u s s i o n o f randomized

time changes.

- 106 -

Next we prove a r e s u l t which c l a r i f i e s the meaning o f the

n o t i o n o f o p t i o n a l enlargement.

9.13. Theorem: L e t ( A , F , , Q ) be a f i l t e r e d measure space such

t h a t (F ) i s r i g h t - c o n t i n u o u s and F = a(uF ) . L e t (M , G , G ,R,i|0 t t t t

be an enlargement of (A,F,F._,Q) . Then the f o l l o w i n g a r e

e q u i v a l e n t :

a) (M , G , G t,R , i | j ) i s an o p t i o n a l enlargement.

b) E v e r y ( G ^ ) - s t o p p i n g time T g i v e s r i s e , as i n 9.8, to

a randomized ( F t ) - s t o p p i n g time T .

c) Whenever (X i s a m a r t i n g a l e over (A , F F Q) t 0<t<°° t

(Y^) E ( X t ° i | i ) i s a m a r t i n g a l e over (M , G , G t,R) .

d) Whenever (E,E) i s a measurable space and ( X _ ) „ ' ' r x t 0<t < o°

i s a Markov p r o c e s s i n (E,E) over ( A , F , F t > Q ) then

(Y^) = (X^oijj) i s a Markov p r o c e s s over (K,C,G^,R) .

Moreover, w i t h r e g a r d to a) => d ) , i f (^t) n a s a t r a n s i t i o n

f u n c t i o n , then (Y^) has the same t r a n s i t i o n f u n c t i o n .

P r o o f : a) => b) . L e t x be the r r v i n [0,°°] a r i s i n g from T as

d e s c r i b e d i n 9.8. L e t t e [O, 0 0) . Then f o r any F e F ,

Q ( d _ ) x ( u ) ( [ 0 , t ] ) - = R ( ^ _ 1 [ F ] n { T < t}) .

' F

Thus x ( * ) ( [ 0 , t ] ) = E ( l { _ < t ^ U , F ) , which i s Q - a.e. e q u a l to an

F^-measurable f u n c t i o n by o p t i o n a l i t y . Hence, by 9.6, x i s Q - a.e.

e q u a l to an ( F ^ ) - r s t T .

b) => a) . L e t t e [0,°°) and l e t G e G . By c o n s i d e r i n g the

( G t ) - s t o p p i n g time T = t 1 £ + 0 0 L ^ . » one f i n d s t h a t E(l_|i | > ,F ) i s

- 107 -

F^-measurable mod Q .

B e f o r e p r o c e e d i n g to the p r o o f of the r e m a i n i n g i m p l i c a t i o n s , we

make the f o l l o w i n g o b s e r v a t i o n :

L e t 0 <_ s < t < 0 0 . Then (*.) and (**) below a r e e q u i v a l e n t .

(*) F o r any g i n ^(M.G^R) ,

E ( g | * , F ) = E(g|i |»,F ) Q - a.e. L S

(**) F o r any f i n i-^A . F ,Q) ,

E ( f ° i H G ) = E ( f | F W • s s

T h i s can be prov e d by an " o r t h o g o n a l i t y argument" such as was used i n

the p r o o f o f (b => c) of 9.6.

The e q u i v a l e n c e (a <=> c) f o l l o w s immediately from t h i s t o g e t h e r

w i t h F = a(uF ) . t t

a) => d ) . L e t f be a n o n - n e g a t i v e E-measurable f u n c t i o n on E

L e t 0 <_ s < t < » . Then

E ( f ( Y f c ) | 6 g ) = E ( f ( X t W | G , )

= E ( f ( X t ) | F s ) o ^ (by (*) => (**))

= E ( f ( X J | X - 1 ( E ) W ((X,) i s Markov) t s t

= E ( f ( Y . ) | Y _ 1 ( E ) ) . t s

Thus ) i s Markov. A s i m i l a r c a l c u l a t i o n shows t h a t any

t r a n s i t i o n f u n c t i o n f o r ( X t ) i s a l s o a t r a n s i t i o n f u n c t i o n f o r

( Y t ) .

- 107a -

d) => a ) . L e t (E,E) be a measurable space and (X^) a

p r o c e s s i n (E,E) ov e r (A,F,Ft,Q) such t h a t

F = X _ 1(E) f o r 0 < t < » . t t -

Then ( X t ) i s Markov.

Hence, by h y p o t h e s i s , (Y ) = (X t°4>) i s a l s o Markov. L e t

0 <_ s < t < °° and l e t f be a no n - n e g a t i v e F^-measurable f u n c t i o n

on A . S i n c e = X ^(E) , f = h ( X t ) f o r some no n - n e g a t i v e

E-measurable f u n c t i o n h on E . Then

E(f°^|G ) = E(h(Y )|G ) s t 1 s

= E(h(Y )|Y _ 1(E)) ((Y,) i s Markov) t s t

= E ( h ( X t ) | X 81 ( E ) ) o ^

= E ( h ( X t ) | F g ) o ^ .

Thus, u s i n g (**) => (*) , and the f a c t t h a t F = o(uF ) , we t t

see t h a t the enlargement i s o p t i o n a l . Of c o u r s e , we must show t h a t

t h e r e i s a measurable space (E,E) and a p r o c e s s (X ) i n (E,E)

- 108 -

such t h a t F = X~ X ( E ) f o r a l l t .

L e t E = Q x [0,~)

E = the a - f i e l d o f ( F t ) - p r o g r e s s i v e l y measurable s e t s

and l e t X t ( _ ) = (_,t) f o r _ e A , t e [0,°°) . T h i s does the

t r i c k . •

Remark: The r i g h t - c o n t i n u i t y o f ('F ) i s not needed f o r the

e q u i v a l e n c e of a ) , c ) , and d) i n the above theorem; a l s o , the

assumption t h a t F = o ( u F ) i s not needed f o r the e q u i v a l e n c e of t t

a) and b ) .

9.14. Observe t h a t i f (M,G,G^,R,u)) i s an o p t i o n a l enlargement

of (A, F , F Q) then ( M , G , G ,R,iJ.) i s an o p t i o n a l enlargement of

( A . F . F . + . Q ) •

9.15. P r o p o s i t i o n . L e t ( A , F , F , Q ) be a f i l t e r e d measure space,

where (^t) 1 S r i g h t - c o n t i n u o u s , and l e t X be a r r v i n [O, 0 0]

over ( A , F ) . L e t M = A x [0,»]

G = F ® A , where A = B o r e l [0,°=]

G F ® A f o r 0 < t < 0 0 , where t t t -

A = {A e A : A n (t,»] = 0 or (t,»]}

R ( G ) = Q(dw ) x(u)(G(u)) f o r G e G

and l e t \p = p r o j e c t i o n o f M on A . Then the f o l l o w i n g a r e e q u i v a l e n t .

a) There i s an ( F ) - r s t x such t h a t X = T Q-a.e.

b) ( M , G , G ,R,IJJ) i s an o p t i o n a l enlargement of (A , , ^ t » Q ) •

c) ( M , G , G T + , R , i J ; ) i s an o p t i o n a l enlargement of ( A , F , F t , Q ) .

- 109 -

P r o o f : C l e a r l y (M , G , G t,R,iJj) i s an enlargement of ( A , F > F t » Q ) •

As (F ) i s r i g h t - c o n t i n u o u s , the e q u i v a l e n c e o f b) and c) f o l l o w s

from 9.14.

b) => a) . L e t T = p r o j e c t i o n o f M on [0,°°] . Then

{T <_ t} = A x [ 0 , t ] f o r 0 _< t < °° , so T i s a ( G ^ - s t o p p i n g

time. Hence, by (a => b) of 9.13, t h e r e i s an (F ) - r s t T such

t h a t R ( ^ _ 1 [ F ] n T _ 1 [ A ] ) = Q (dto) x (to) (A) f o r F e F and A e A .

But then x = T Q - a.e.

a) => b) . One e a s i l y checks t h a t i f t e [0,°°) and G e G

then T(»)(G) i s F -measurable and E ( l J i J > . F ) = T(»)(G) Q - a.e. • t G

Note t h a t i n the above p r o o f , T i s r e a l i z e d as the ( G ^ - s t o p p i n g

time T i n the o p t i o n a l enlargement (M , G , G t,R,i(j) . In 9.18 below,

we d e s c r i b e another way of r e a l i z i n g x i n an o p t i o n a l enlargement.

9.16. Lemma: L e t y be a p r o b a b i l i t y measure on [0,°°] , and

l e t m be Lebesgue measure on the B o r e l s e t s of (0,1) .

Then t h e r e a r e unique maps f , g : (0,1) ->• [0,°°] such t h a t :

a) f and g a r e i n c r e a s i n g

b) f i s l e f t - c o n t i n u o u s

c) g i s r i g h t - c o n t i n u o u s

d) y = f(m) = g(m)

Moreover, i f h : (0,1) •+ [O, 0 0] i s i n c r e a s i n g and h(m) = y then

f <_ h _< g and f o r a l l u i n (0,1) we have f (u) = h(u-) and

g(u) = h(u+) .

- n o -Sketch o f P r o o f : F o r 0 < u < 1 , l e t

f ( u ) = sup{t e [0,»] : y ( [ 0 , t ] ) < u}

(where sup 0 = 0 )

and l e t g(u) = i n f { t e [0,°°] : y ( [ 0 , t ) ) > u} . •

9.17. D e f i n i t i o n : L e t ( A , F , F T , Q ) be a f i l t e r e d measure space,

where (F ) i s r i g h t - c o n t i n u o u s . L e t (X,A,m) be a measure space

w i t h m(X) = 1 . L e t M = A * X

G = F ® A

G = (F F C ® A) (= n F 9 A) e>0 t + e

f o r 0 _< t < °°

R = Q <_. m

i> = p r o j e c t i o n o f M on A .

Then c l e a r l y (M,G,Gt,R,^) i s an o p t i o n a l enlargement o f

(A,F,F^.Q ) ; w e c a l l i t the p r o d u c t enlargement o f ( A , F , F T , Q ) by

(X,A,m) .

9.18. P r o p o s i t i o n : L e t ( A . F , F T , Q ) be a f i l t e r e d measure space,

w i t h (F ) r i g h t - c o n t i n u o u s . L e t (X,A,m) be a measure space w i t h

m(X) = 1 , and l e t ( H , G , G ^ , R , M ) ) be the p r o d u c t enlargement o f

( A , F , F T , Q ) by (X,A,m) .

Then:

a) I f T i s a ( G t ) - s t o p p i n g time then T(',x) i s an

( F t ) - s t o p p i n g time f o r each x e X , and T = T(',m)

i s a randomized (F ) - s t o p p i n g time.

- I l l -

b) Suppose X = (0,1) , A = B o r e l X , and m = Lebesgue i

measure on A . I f T i s a randomized ( F ^ - s t o p p i n g

time, and i f f o r each to i n A we l e t T(to,») be

the unique i n c r e a s i n g l e f t - c o n t i n u o u s map f of (0,1)

i n t o [0,°°] such t h a t f(m) = x(to) (see 9.16) then

x i s a ( G ^ - s t o p p i n g time.

P r o o f : a) { T < t } € ' F ® A .

Hence {T(«,x) < t} = {to e A: (to,x) e {T < t}} e F , and

T ( * ) ( [ 0 , t ) ) = m({x e X : T(»,x) < t}) i s F t - m e a s u r a b l e . The

d e s i r e d c o n c l u s i o n s now f o l l o w from the r i g h t - c o n t i n u i t y o f (F' ) .

b) T ( t o,u) = sup{t e [0,~] : x ( t o ) ( [ 0 , t ] ) < u} . (Here sup 0 = 0.)

As x ( t o ) ( [ 0 , t ] ) i s r i g h t - c o n t i n u o u s i n t , T(to,u) = sup{r e [0,°°) :

r i s r a t i o n a l and i ( t o ) ( [ 0 , r ] ) < u} . Thus T(to,u) < t i f f t h e r e

i s a r a t i o n a l r e [0, t) such t h a t x ( t o ) ( [ 0 , r ] ) _> u . That i s ,

{T < t} = r £ $ t ) ^ , u ) e M : u < T (to) ( [ 0 , r ])} .

r r a t i o n a l

As x(»)([0,r]) i s F^-measurable, {T < t} e F ^ ® A . Thus T

i s a (G ) - s t o p p i n g time.

9.19. L e t (A,F,Q) be a measurable space and l e t E be a s e p a r a b l e

normed sp a c e .

I f Z i s a weak*-dense su b s e t of the u n i t b a l l of the *

t o p o l o g i c a l d u a l E of E then the s m a l l e s t o - f i e l d on E

making a l l the elements of Z measurable c o i n c i d e s w i t h the

- 112 -

a - f i e l d generated by the norm-open s u b s e t s o f E .

Thus t h e r e i s no ambiguity about what we mean by an F-measurable

f u n c t i o n from A i n t o E . We w r i t e L X(A , F,Q;E) f o r the space

of a l l f u n c t i o n s f : A E such t h a t f i s F-measurable and

f(w)||Q(du.) < 0 0 , equipped w i t h the seminorm

f •* | |f | | = | | |f (oi) | |Q(d_) .

We wish to d e s c r i b e the d u a l o f t h i s space.

Now E need not be s e p a r a b l e , and E may not be the d u a l *

o f E , so we have to be c a r e f u l about what we mean by a measurable *

E - v a l u e d f u n c t i o n .

We s h a l l say t h a t a f u n c t i o n g : A -»• E i s weak * - F-measurable

i f f _ H- <x,g((~)> i s F-measurable f o r each x e E . In t h i s c a s e,

u s i n g the s e p a r a b i l i t y o f E , we can show t h a t the n u m e r i c a l

f u n c t i o n u> •->• | |g(aj)| | i s F-measurable. *

A l s o , i f f : A -> E i s F-measurable and g : A E i s

F-measurable, then _ •->• <f (_), g(t_)> i s F-measurable. T h i s can be

p r o v e d by a p p r o x i m a t i n g f by c o u n t a b l y - v a l u e d F-measurable

f u n c t i o n s . A l s o , i f g, g' : A ->• E a r e weak*-F-measurable then

{g 4 g'} e F , s i n c e i t i s e q u a l to {||g - g'|| = 0} . Now , g

d e f i n e s a c o n t i n u o u s l i n e a r f u n c t i o n a l T on L X(A , F,Q;E) by the

f o r m u l a <f,r> = < f ( _ ) , g(_)>Q(d_)

S i m i l a r l y g' d e f i n e s T' . I f Q i s s e m i f i n i t e , then

- 113 -

||r|| = the Q - e s s e n t i a l supremum o f ||g(»)|| , and r = I" i f f

Q({g t g')) = 0 . On the o t h e r hand, i f Q i s c - f i n i t e and i f

r i s any c o n t i n u o u s l i n e a r f u n c t i o n a l on l?~(1\,T,Q;E) then t h e r e

i s a weak*-F-measurable f u n c t i o n g :A E such t h a t

||g(w)|| 1 I M l f o r a 1 1 w i n A

f and <f,T> = < f ( „ ) , g(_)>Q(d_) f o r a l l f i n _ 1(A,F,Q;E)

We s h a l l not prove t h i s r e s u l t h e r e . See Ionescu T u l c e a and

I o n e s c u T u l c e a [ 1 ] , V I I . 4 , f o r a p r o o f i n the more g e n e r a l case i n

which E i s not assumed s e p a r a b l e . In t h i s case the l i f t i n g

theorem i s used, so one assumes Q to be complete. I n our case,

i n which E i s s e p a r a b l e , t h i s i s u n n e c e s s a r y . ( L e t us a l s o

remark t h a t i f E i s not assumed s e p a r a b l e , the F-measurable

f u n c t i o n s i n t o E a r e d e f i n e d to have s e p a r a b l e range.)

9.20. N o t a t i o n : I f X and Y are s e t s and f , g a r e r e a l - v a l u e d

f u n c t i o n s on X, Y r e s p e c t i v e l y , then f ® g w i l l denote the

f u n c t i o n on X * Y d e f i n e d by

(f®g)(x,y) = f ( x ) g ( y ) .

9.21. L e t (A,^,Q) be a o - f i n i t e measure space, and l e t K be

a compact m e t r i z a b l e s p a c e .

Then C ( K ) , the space of c o n t i n u o u s r e a l - v a l u e d f u n c t i o n s on

K , equipped w i t h the supremum norm, i s a s e p a r a b l e normed sp a c e .

Thus C(K) can p l a y the r o l e o f E i n 9.19.

- 114 -

I f X i s a r e a l - v a l u e d f u n c t i o n on A x K such t h a t

X(co,») i s c o n t i n u o u s on K f o r each to i n A and X(»,p) i s

F-measurable f o r each p i n K (such an X i s c a l l e d a C a r a t h e o -

dory f u n c t i o n ) then X i s F ® ( B o r e l K)-measurable and hence

to •-»• X(to,») i s an F-measurable C ( K ) - v a l u e d f u n c t i o n .

On the o t h e r hand, i f f i s an F-measurable C ( K ) - v a l u e d

f u n c t i o n then f(»)(p) i s an F-measurable r e a l - v a l u e d f u n c t i o n

f o r each p i n K , s i n c e the p o i n t - e v a l u a t i o n s a r e c o n t i n u o u s

l i n e a r f u n c t i o n a l s on C(K) .

Thus F-measurable C ( K ) - v a l u e d f u n c t i o n s on A can be

i d e n t i f i e d w i t h Caratheodory f u n c t i o n s on A x K . Under t h i s

i d e n t i f i c a t i o n , l?~(A,F,Q;C(K)) becomes i d e n t i f i e d w i t h the s e t of

Car a t h e o d o r y f u n c t i o n s X on A x K s a t i s f y i n g

sup |x(»,p) |dQ < °° . ' p e K

I f f e L 1 ( A } F J Q ) and g e C(K) then f o g be l o n g s to t h i s

s e t .

L e t RRV(A,F,Q;K) be the space o f randomized random v a r i a b l e s

i n K over (A,F) , equipped w i t h the weak t o p o l o g y i n d u c e d by the

maps

Q(dto) X ( t o ) ( d p ) h ( t o ) ( p )

where h ranges over L 1(A,F,Q;C(K)) . To save w r i t i n g , l e t us

denote t h i s space by j u s t RRV, f o r the remainder o f 9.21.

I f x e R R V > then h >->- <x>h> i s a c o n t i n u o u s p o s i t i v e l i n e a r

- 115 -

f u n c t i o n a l on L. 1(A , F,Q;C(K)) s a t i s f y i n g <x,f«l> = fdQ f o r a l l

f e L^"(A,F,Q) . On the o t h e r hand, any f u n c t i o n a l w i t h t h e s e

p r o p e r t i e s a r i s e s from a ( Q - e s s e n t i a l l y unique) x b e l o n g i n g to RRV.

Thus { < X » # > : X e RRV} i s a weak*-closed s u b s e t of the u n i t

1 *

b a l l of L (A , F,Q;C(K)) , and so i s weak*-compact by the Banach-

A l a o g l u theorem.

I t f o l l o w s t h a t RRV i s a compact space. A l s o , i f H has

dense l i n e a r span i n L*~(A, F,Q;C(K)) then the maps o f the form

X •->• <X>h> (heH) i n d u c e the o r i g i n a l t o p o l o g y of RRV. Hence i f F

i s c o u n t a b l y generated mod Q, RRV i s p s e u d o - m e t r i z a b l e (and two

elements of RRV a r e z e r o d i s t a n c e a p a r t i f f they a r e e q u a l Q - a . e . ) .

Now the elements of any c o u n t a b l e s u b s e t of RRV are a l l randomized

random v a r i a b l e s over ( A , F ' ) where F ' i s some c o u n t a b l y generated

sub - o - f i e l d of F . I t f o l l o w s t h a t RRV i s s e q u e n t i a l l y compact,

even when i t i s not p s e u d o - m e t r i z a b l e .

9.22. L e t us g i v e an example o f a sequence of random v a r i a b l e s which

converges to a randomized random v a r i a b l e which i s not a random

v a r i a b l e .

L e t A = [0,1)

F = B o r e l A

Q = Lebesgue measure on F

K = [0,1] .

F o r i = 1,2,..., l e t X^ be the c h a r a c t e r i s t i c f u n c t i o n o f

u [ ^ r - , ^j) . Then each X. i s a r v i n K over (A , F ) ; j - 1 2 1 2 1 1

j odd

- 116 -

l e t be the c o r r e s p o n d i n g r r v .

L e t x be the r r v i n K over (A,F) d e f i n e d by

Then u s i n g the f a c t t h a t the bounded c o n t i n u o u s f u n c t i o n s are dense i n

9.23. Now here i s a v e r s i o n o f a r e s u l t due to Bax t e r and Chacon [ 2 ] .

The r e a d e r may a l s o f i n d the review o f t h i s a r t i c l e , by Meyer [ 4 ] ,

to be e n l i g h t e n i n g .

Theorem. L e t (A,F,F ,Q) be a f i l t e r e d measure space, w i t h (Ffc)

r i g h t - c o n t i n u o u s . L e t RST = RST(A,F,F t >Q) be the space o f

randomized ( F ^ ) - s t o p p i n g t i m e s , endowed w i t h the topo l o g y i t i n h e r i t s

as a subspace o f RRV = RRV(A,F,Q; [0,°°]) .

Then:

a) RST i s compact and s e q u e n t i a l l y compact.

b) I f (x.) i s a n e t c o n v e r g i n g to x i n RST, and i f

(Z ) t'0<t< CO i s a r e a l - v a l u e d c a d l a g q u a s i - l e f t - c o n t i n u o u s

p r o c e s s s a t i s f y i n g

sup |z | dQ < 0<t<°°

00

t h e n <x. ,Z> -»• <x,Z> ; t h a t i s , Q(du>) x i ( _ ) ( d t ) Z t ( _ )

- 117 -

P a r t i a l P r o o f :

a) S i n c e (F^) i s r i g h t - c o n t i n u o u s , we can a p p l y (b => a) of

9.6 to c o n c l u d e t h a t i f x b e l o n g s to the c l o s u r e of RST

i n RRV then t h e r e i s a T i n RST such t h a t x = t Q - a.e.

Thus RST i s compact and s e q u e n t i a l l y compact because RRV

i s . (See 9.21).

b) We remark t h a t " c a d l a g " stands f o r " c o n t i n u a d r o i t avec

des l i m i t e s a gauche" ( r i g h t c o n t i n u o u s w i t h l e f t l i m i t s ) .

I f ( Z t ) i s c o n t i n u o u s then the d e s i r e d c o n c l u s i o n f o l l o w s

immediately from the d e f i n i t i o n o f the t o p o l o g y of RRV, and does not

depend on the f a c t t h a t the T ^ ' s a r e randomized s t o p p i n g t i m e s .

F o r the g e n e r a l case, i n which (Z^~) i s merely q u a s i - l e f t

c o n t i n u o u s , the f a c t t h a t the T ^ ' s a r e randomized s t o p p i n g times,

r a t h e r than a r b i t r a r y randomized random v a r i a b l e s i n [O, 0 0] , i s

c r u c i a l . (To see t h i s , j u s t c o n s i d e r the non-randomized case.)

S i n c e we s h a l l be concerned o n l y w i t h c o n t i n u o u s p r o c e s s e s , we

r e f e r the r e a d e r to B a x t e r and Chacon [2] or Meyer [ 4 ] .

9.24. F o r the moment, l e t M denote the space o f measures y on

[0,°°) s a t i s f y i n g y({0,°°)) £ 1 , endowed w i t h i t s u s u a l vague

t o p o l o g y . M i s a compact m e t r i z a b l e space. F o r each y i n M ,

l e t F be the d i s t r i b u t i o n f u n c t i o n o f y : u

F ( t ) = y ( [ 0 , t ] ) (0<t<») .

The map y — * F t | i s a 1-1 map of M onto the s e t of i n c r e a s i n g

- 118 -

r i g h t - c o n t i n u o u s maps o f [O, 0 0) i n t o [0,1] . I f ( P ^ ) i s a

sequence i n M and y e M then y. ->• y i n M i f f F ( t ) -> F ( t ) i u± y

f o r each t i n [0,°°) a t which F^ i s c o n t i n u o u s , as i s shown i n

most any s t a n d a r d t e x t on p r o b a b i l i t y t h e o r y . One can show t h a t

t h i s r e s u l t h o l d s f o r n e t s as w e l l as sequences. Indeed, i f f o r

each y e M we l e t U be the c o l l e c t i o n o f s e t s of the form y

{ v e M : |F ( t i ) - F v ( t i ) | < e f o r i = 0,...,k}

where e > 0 , k e IN , and t - , . . . , t , a r e c o n t i n u i t y p o i n t s of F , O k y

then (U ) w c o n s t i t u t e s a neighbourhood base f o r a H a u s d o r f f y yeM

t o p o l o g y on M which i s weaker than the g i v e n t o p o l o g y o f M and

which t h u s , by compactness, must be i d e n t i c a l to the g i v e n t o p o l o g y o f

M . (Note t h a t we do not c l a i m t h a t the elements o f U a r e open — y

o n l y t h a t they are neighbourhoods of y .)

Now l e t H be the s e t of a l l i n c r e a s i n g r i g h t - c o n t i n u o u s maps

o f [0,°°) i n t o [0,°°] . S i n c e [0,°°] may be i d e n t i f i e d w i t h

[0,1] by means o f an o r d e r - p r e s e r v i n g homeomorphism, H may be made

i n t o a compact m e t r i z a b l e space i n a n a t u r a l way. I f (IK) i s a

n e t i n H and h e H , then h -»• h i n H i f f h ^ t ) h ( t )

f o r each t i n [0,°°) at which h i s c o n t i n u o u s . Now any h e H

i s c o n t i n u o u s except a t c o u n t a b l y many t ' s i n [0,°°) . Thus i f

(h^) i s a sequence i n H c o n v e r g i n g to h e H then h^ h

p o i n t w i s e almost everywhere w i t h r e s p e c t to Lebesgue measure on

[0,°°) . I t f o l l o w s t h a t i f <j> i s any bounded c o n t i n u o u s f u n c t i o n on

[0,°°] and f any Lebesgue i n t e g r a b l e f u n c t i o n on [O, 0 0) then

- 119 -

f ( t ) 4 > ( h ( t ) ) d t i s a co n t i n u o u s r e a l - v a l u e d f u n c t i o n on H

U s i n g the r i g h t - c o n t i n u i t y o f the h's we can deduce from t h i s t h a t

f o r any t e [O, 0 0) , the e v a l u a t i o n map h i — > h ( t ) i s a B o r e l

f u n c t i o n on H ; i n d e e d i t i s the p o i n t w i s e l i m i t of a sequence o f

c o n t i n u o u s f u n c t i o n s on H . Now c o u n t a b l y many o f these e v a l u a t i o n

maps s u f f i c e to s e p a r a t e the p o i n t s o f H .

I t f o l l o w s from t h i s t h a t H = B o r e l H i s the s m a l l e s t o - f i e l d

on H making a l l the e v a l u a t i o n maps measurable. F o r each t e [0,°°]

l e t H be the a - f i e l d o f s u b s e t s of H generated by the s e t s of

the form {h e H : h(a) e E} where a e [0,°°) and E e B o r e l [ 0 , t ] .

C l e a r l y (H^) i s an i n c r e a s i n g f a m i l y of c o u n t a b l y generated sub-o-

f i e l d s o f H and H = H .

9.25. D e f i n i t i o n . L e t (A,F,F ) be a f i l t e r e d measurable space

and l e t H be as i n 9.24. A r i g h t - c o n t i n u o u s (F )-time change

i s a map T : A H such t h a t f o r each a e [0,») , T(»)(a) i s

an" ( F ^ ) - s t o p p i n g time.

Observe t h a t T : A ->• H i s a r i g h t - c o n t i n u o u s (F )-time change

i f f f o r a l l t e [0,») and a l l A e H , T _ 1 [ A ] e F f c . (Here Hfc

i s as i n 9.24.)

9.26. P r o p o s i t i o n . L e t (A,F,F ,Q) be a f i l t e r e d measure space, '

where (F ) i s r i g h t - c o n t i n u o u s . L e t H, H, be as i n 9.24. L e t

X be a r r v i n H ov e r (A,F) . A l s o , l e t

M = A x H

G = F ® H

- 120 -

( G t ) = ( ( F t ® H t ) + )

R(G) = Q(d_ ) x(u)(G(u))) f o r G e G

ij. = p r o j e c t i o n o f M on A .

Then the f o l l o w i n g a r e e q u i v a l e n t :

a) (M ,G,G ,R,iJ.) i s an o p t i o n a l enlargement of (A ,F,F._,Q)

b) F o r a l l t e [0,») and a l l A e H , x ( * ) ( A ) i s F f c -

measurable mod Q .

c) Whenever t e [0,°°) , k e ]N , a Q , . . . , a k e [0,°°) , e £ (0,°°) ,

(f>i e C([0,°°)) such t h a t $± = 0 on [t,°°) f o r

i = 0,...,k , g i s the f u n c t i o n on H d e f i n e d by

k g(h) = n F

a.+e l <j>. ( h ( s ) ) d s ] ,

i=0 * U. I

and f £ L ( A , F , Q ) such t h a t

then

fdQ = 0 f o r a l l F £ F ,

Q(dui) x(to)(dh ) f (u)g(h) = 0

P r o o f : F i r s t note t h a t ij. i s (G, F)-measurable, (G^, F^)-measurable

f o r a l l t , and ^(R) = Q .

a) => b) I f A eH then A x A e G f c . A l s o , E (_ A > < A | i j ; , F)

= X(«)(A) Q - a.e.

b) => c) F o r each i , = 0 on [t,°°) so h <—*• i(;^(h(s))

i s H^-measurable (i n d e e d H -measurable) f o r each s . I t f o l l o w s

t h a t g i s H -measurable. Thus X(»)(dh)g(h) i s e q u a l Q - a.e.

- 121 -

to an F^-measurable f u n c t i o n . Hence Q(dto)f (to) x ( c o ) ( d h ) g ( h ) = 0

c) => b) F i r s t , f o r any g o f the s o r t d e s c r i b e d i n c) we

can c o n c l u d e t h a t x ( * ) ( d h ) g ( h ) i s e q u a l Q - a.e. to an F -

measurable f u n c t i o n . Then ( l e t t i n g e -+ 0) we f i n d t h a t whenever

t € tO, 0 0) , k e U , a ^ . . . , ^ e [0,°°) and \Ji e C([0,°°)) such t h a t

\\>. = 0 on [ t , 0 0 ) f o r i = 0,...,k , then the map

x(co)(dh) II i(j.(h(a.)) i s F -measurable mod Q . From t h i s i=0 1 1 *

we can con c l u d e t h a t i f t e [0,°°) , A e H , and E > 0 then

x(«)(A) i s F^ +^-measurable mod Q . Then u s i n g the r i g h t - c o n t i n u i t y

o f (F ) we o b t a i n the d e s i r e d c o n c l u s i o n .

b) => a) By a monotone c l a s s argument, one f i n d s t h a t i f

B e F^ ® H then co I—• x (w) (B (co)) i s F^-measurable mod Q .

U s i n g the r i g h t - c o n t i n u i t y o f (^ ) one f i n d s t h a t t h i s c o n c l u s i o n

i s s t i l l t r u e i f B i s o n l y i n G .

But f o r any B e G ,

X ( 0 ( B ( . ) ) = E ( 1 B U , F ) Q - a.e.

9.27. Now he r e i s a v e r s i o n o f a r e s u l t due to Bax t e r and Chacon [3]

Theorem. L e t ( A »F , F t ,Q ) be a f i l t e r e d measure space, where )

i s r i g h t - c o n t i n u o u s . L e t H, H, H be as i n 9.24, and l e t M, G, G^

be as i n 9.26. L e t R be the s e t of measures R on G such t h a t

(M ,G,G t»R , i|0 i s an o p t i o n a l enlargement o f (A ,F,F ,Q) . L e t R be

g i v e n the weak t o p o l o g y i n d u c e d by the maps of the form

- 122 -

f ( u ) ( h ) R ( d _ , d h )

where f e L X(A,F,Q;C(H)) . Then R i s compact and s e q u e n t i a l l y

compact.

P r o o f : L e t RRV denote the compact, s e q u e n t i a l l y compact, t o p o l o g i c a l

space RRV(A,F,Q;H) , as d i s c u s s e d i n 9.21. We can do t h i s , s i n c e H

i s compact and m e t r i z a b l e .

L e t S be the s e t of measures S on G whose p r o j e c t i o n on

A i s Q , and l e t S be t o p o l o g i z e d a n a l o g o u s l y to R . L e t

d> : RRV -»- S be d e f i n e d by

*(X)(G) = Q(d_)x(o>) (G(oj))

Then d> i s a c o n t i n u o u s map o f RRV onto S . I f x> x ' e RRV

then <Kx) = M x ' ) i f f X = x ' Q - a.e. C l e a r l y S i s H a u s d o r f f .

Thus d> maps c l o s e d s u b s e t s o f RRV onto c l o s e d s u b s e t s o f S . Now

R £ S and, s i n c e <|> i s onto , R = <f>[d> X [ R ] ] . Thus t o complete

the p r o o f , i t s u f f i c e s to show t h a t d) X [ R ] i s c l o s e d i n RRV. But

by (a<=>b) of 9.26,

4'~ 1[R] = " X e RRV : f o r a l l t e [0,°°) and a l l A" e H .

x ( * ) ( A ) i s e q u a l Q - a.e. to an

F -measurable f u n c t i o n , t

and t h i s s e t i s c l o s e d i n RRV, by (b<=>c) o f 9.26. (The f u n c t i o n s

g of the form c o n s i d e r e d i n 9.26(c) a r e c o n t i n u o u s on H , as p o i n t e d

out i n 9.24.)

- 123 -

Remark: In view of (a<=>b) of 9.25, and the observation following

the definition in 9.24, R may be regarded as the set of "randomized

right-continuous (F^)-time changes" (where we identify pairs of time

changes which are equal Q - a.e.). Thus from any sequence of

right-continuous (F^)-time changes, we may extract a subsequence

converging (in the sense explained above) to a randomized right-

continuous (F^)-time change.

- 124 -

10. EMBEDDING MEASURES IN BROWNIAN MOTION IN A GREEN REGION, USING RANDOMIZED STOPPING TIMES.

10.1. Throughout t h i s s e c t i o n , D w i l l be a Green r e g i o n i n R n w i t h

Green f u n c t i o n G , and R w i l l be i n f { t >_ 0 : I D} .

I f y i s a measure i n D and T i s a randomized ( 8 t ) - s t o p p i n g

time, then y_ w i l l denote the measure on B o r e l D d e f i n e d by

V_(A) = P P(du.)x(„) ({t e [0,R(o))) : B (ID) e A})

Note t h a t i f T a r i s e s from a genuine ( B ^ - s t o p p i n g time T then

P T = V- where

y_(A) = P M ( B _ e A, T < R) .

More g e n e r a l l y , i f (X,A,m) i s a measure space w i t h m(X) = 1 ,

and i f T a r i s e s as i n 9.18(a) from a s t o p p i n g time T over the

p r o d u c t enlargement of ( f t , B , B t , P y ) by (X,A,m) , then f o r

A £ B o r e l D ,

V T(A) = m(dx)P y(B , s £ A , T(',x) < R)

m ( d x ) y _ ( . ) X ) ( A ) X

10.2. N o t a t i o n . I f ( A , F ) i s a measurable space and a, x a r e r r v ' s

i n [0, 0 0] over ( A , F ) , we w r i t e a <_ x to mean

a(„) ( ( t , 0 0 ] ) <_ x(_)((t,»]) f o r a l l u> £ A and a l l

t e [0,«) •

10.3. Lemma. L e t p be a measure i n D such t h a t Gy i s a p o t e n t i a l .

- 125 -

Let a, T be randomized (B t)-stopping times such that a <_ x . Then

Gy > Gy > Gy . (It follows that y_ and y are fi n i t e on — 0 — T O T

compact subsets of D .)

Proof: Let A = Borel (0,1) and let S, T be the ((8 ® A ) + ) -

stopping times associated with a, T by 9.18(b). '1

Then S < T . Also y = — o 0 du Vg^. uy a n c* similarly for y^

Moreover, S(-,u) and T(-,u) are (8 t)-stopping times for

0 < u < 1 . Now app ly 7.4. •

10.4. Lemma. Let y be a measure in D such that Gy is a

potential. Suppose (T ) is a net converging to x in

RST(fi,B,Bt,Py) . Let v = y T , y ± = yx_ Then:

a) For any compactly supported continuous function <f> in D ,

4>Gv

b) For any compactly supported continuous function <j> in D ,

<(>dy , 4>dv

Proof:

a) Let <J> be a non-negative compactly supported continuous

function in D , and let (Zj)n be the non-negative decreasing t 0<t<°°

process defined by

Z t = <t>(Bs)ds t A R

- 126 -

Then E (ZQ) = <()Gy by 6.9. S i n c e Gy i s a p o t e n t i a l , i t f o l l o w s

t h a t E Y ( Z Q ) i s f i n i t e . Now t Z^(IXI) i s co n t i n u o u s on [0,°°]

f o r any co f o r which ZQ(CO) i s f i n i t e .

Thus <T , Z> + <T, Z>. (See 9.23.) But i f c i s any

randomized ( B ^ ) - s t o p p i n g time then

<o, Z> P y(dco) a(co)(dt)Z t(co)

P y(dco) du Z . , v (co) S(co,u)

du E V [ S ( - , u ) A R

* ( B g ) d s ]

du * G y S ( - u) ^ b y 6 ' 9 3 n d t h e s t r o n g M a r k o v

p r o p e r t y )

4>Gyo ,

where S i s the ( ( 8 ^ ® B o r e l ( 0 , 1 ) ) + ) - s t o p p i n g time a s s o c i a t e d to

a by 9.18(b) .

b) From a ) , i t f o l l o w s t h a t the c o n c l u s i o n o f b) h o l d s i f <f)

2 i s a compactly s u p p o r t e d C f u n c t i o n i n D . But Gy^ <_ Gy f o r

a l l i , by 10.3. Thus, by 7.8, sup y. (K) < 0 0 f o r any compact

i 1

s u b s e t K o f D . We can t h e r e f o r e complete the p r o o f o f b) by

an a p p r o x i m a t i o n argument.

The n e x t r e s u l t i s a p a r t i c u l a r case of a r e s u l t due to Rost [1]

Our method o f p r o o f , which was sugg e s t e d to me by R. V. Chacon, i s

- 127 -

q u i t e d i f f e r e n t from R o s t ' s .

10.5. Theorem. L e t y, v be measures i n D , and suppose Gy i s

a p o t e n t i a l . Then the f o l l o w i n g a r e e q u i v a l e n t :

a) Gy _> Gv .

b) There e x i s t s a randomized ( 8 t ) - s t o p p i n g time T such t h a t

y = v . x

P r o o f :

b) => a ) . See 10.3.

a) => b ) . F o r each n a t u r a l number i , l e t v. be the measure — l

i n D such t h a t Gv. = iAGv . l

Then, by 7.11, t h e r e are (8^)-stopping times T_ such t h a t

v. = y T f o r each i . By 9.23, th e r e i s a randomized ( 8 t ) - s t o p p i n g 1 i

time T such t h a t some subsequence o f (T^) converges to T i n

RST(fi,8,8 t,P V) . Now Gv ± + Gv . Hence, by 10.4, y^ = v . ^

10.6. C o r o l l a r y : L e t y,v be measures i n D such t h a t Gy i s a

p o t e n t i a l and Gv <_ Gy .

Then v(Z) <_ y(Z) f o r ev e r y B o r e l p o l a r s u b s e t Z of D .

P r o o f : By 10.5, v = y^ f o r some randomized ( 8 t ) - s t o p p i n g time T .

L e t T be the ( (B f c ® B o r e l ( 0 , 1 ) ) + ) - s t o p p i n g time a s s o c i a t e d to

T by 9.18(b). (-1

Then v = du y , x . Now i f Z i s any B o r e l p o l a r s u b s e t J Q T(-,u)

of D then y , , (Z) <_y(Z) f o r each u i n (0,1) , by 7.3(b);

- 128 -

hence v(Z) <_ y(Z) .

10.7. The p r o o f o f 10.6 t h a t we have j u s t g i v e n depends on a r a t h e r

l a r g e f r a c t i o n of what has gone b e f o r e . T h e r e f o r e , l e t us i n d i c a t e

how t h i s r e s u l t can be proved more d i r e c t l y , u s i n g o n l y c l a s s i c a l

p o t e n t i a l t h e o r y .

F i r s t of a l l t h e r e i s a B o r e l s e t A £ D such t h a t v _> u on

B o r e l s u b s e t s o f A and v u on B o r e l s u b s e t s o f D\A . Now by

2.2, we have o n l y t o prove t h a t v(Z) <_ y(Z) f o r every B o r e l s e t

Z £ {Gv = °°} . L e t H = A n {Gv = °°} . Then t h e r e a r e unique

measures a, B i n D such t h a t a + u = u , 3 + u = v . (Note n rl

t h a t u < v .) rl

Now v < u on B o r e l s u b s e t s o f {Gv = <*>}\H , and u <_ u , so — n

i t c e r t a i n l y s u f f i c e s to show t h a t 8(H) = 0 . Observe t h a t

GB <. Ga and a(H) = 0 .

Now a ( f r i n g e ( H , D ) ) = 0 . (H i s a p o l a r s e t , so f r i n g e ( H , D )

Hence, by a theorem of c l a s s i c a l p o t e n t i a l t h e o r y , bal(Ga,H,D) i s

e q u a l to the lower r e g u l a r i z a t i o n of

in f { b a l ( G a , U , D ) : H £ U open £ D} .

But H i s p o l a r . Thus t h i s infimum i s e q u a l to zero except on a

p o l a r s e t .

Now i f H £ U open £ D then U £ base(U,D) so 3 l i v e s on n

base(U.D) . Hence bal(Ga,U,D) >_ G3 U by the d o m i n a t i o n p r i n c i p l e H

4.1. T a k i n g the infimum over a l l such U , we f i n d t h a t { 8 > ^

i s a p o l a r s e t , whence i t must be empty. Thus 3(H) = 0 .

- 129 -

10.8. Now we e s t a b l i s h the f o l l o w i n g improvement of 10.6. Note t h a t

i t completes the p r o o f s o f (b => a) of 4.2 and (a => b) of 4.8.

P r o p o s i t i o n : L e t W be an open subset o f H n and l e t u, v be

superharmonic f u n c t i o n s i n W w i t h R i e s z measures y, v r e s p e c t i v e l y ,

L e t E = b a s e ( { u J> v}) n W . Then v(Z) <_ y(Z) f o r every B o r e l

p o l a r s e t Z £ E .

P r o o f :

Case 1. Assume W i s a Green r e g i o n and u, v are p o t e n t i a l s

i n W . L e t F = {u >_ v} , and l e t a = bal(y,F,W) , g = bal(v,F,W) .

y(dx)bal(<5_,F,W) , and s i m i l a r l y f o r g , by 2.15. Now a = x

I f x e E then bal(<!>x,F,W) = 6^ by the d o m i n a t i o n p r i n c i p l e

4.1. On the o t h e r hand, i f x e W\E then G__6 i s f i n i t e on E w x

so b a l ( 6 x > F , W ) does not charge p o l a r s u b s e t s of E by 2.2.

Thus ot(Z) = y(Z) and g(Z) = v(Z) f o r e v e r y B o r e l p o l a r s e t

Z £ E . But u •> v on F so bal(u,F,W) >_ bal(v,F,W) . Hence

a(Z) _> g(Z) f o r e v e r y B o r e l p o l a r s e t Z c W .

Case 2. L e t W, u, v be as i n the statement of the p r o p o s i t i o n .

L e t Z be a B o r e l p o l a r s u b s e t of E . Then Z i s a c o u n t a b l e u n i o n

of r e l a t i v e l y compact B o r e l s u b s e t s o f W , so i t s u f f i c e s t o c o n s i d e r

the case i n which Z i t s e l f i s r e l a t i v e l y compact i n W .

Then t h e r e a r e open s e t s U and V such t h a t Z £ u , U i s

r e l a t i v e l y compact i n V , and V i s r e l a t i v e l y compact i n W .

Then u, v a r e bounded below i n V so t h e r e i s a r e a l number c

such t h a t u+c and v+c a r e n o n - n e g a t i v e i n V .

- 130 -

L e t f = b a l ( u + c , U, V)

g = b a l ( v + c , U, V) .

Then f and g a r e p o t e n t i a l s i n V whose R i e s z measures

c o i n c i d e w i t h y and v r e s p e c t i v e l y on B o r e l s u b s e t s o f U ;

i n d e e d f = u+c i n U and g = v+c i n U .

Now {f _> g} n U = {u _> v} n U . Thus, by 3.9 and 3.10, i f

x e U then {f >_ g} i s t h i n a t x r e l a t i v e t o V i f f

{f >_ g} n U i s t h i n a t x r e l a t i v e to U i f f {u >_ v} n U i s t h i n

a t x r e l a t i v e to U i f f {u >_ v} i s t h i n a t x . Hence

Z c b a s e ( { f _> g}, V) . T h e r e f o r e y(Z) _> v(Z) by case 1.

10.9. Remark. L e t T be the s e t of r r v ' s T i n [0,°°] over

(ft,B) of the form

T ( _ ) = f ( _ ) 6 Q + ( 1 - f ( w ) ) « T ( a ) ) .

where f ranges over B^-measurable [ 0 , 1 ] - v a l u e d f u n c t i o n s on „

and T ranges over non-randomized ( B ^ ) - s t o p p i n g t i m e s . The element

of T a r e randomized ( B t ) - s t o p p i n g times which we might say a r e

"randomized o n l y a t time 0" . We c l a i m t h a t the f o l l o w i n g a r e

e q u i v a l e n t :

a) The c o n j e c t u r e 7.12 i s v a l i d f o r the Green r e g i o n D .

b) Whenever u, v a r e measures i n D such t h a t Gy i s a

p o t e n t i a l and Gy >_ Gv , t h e r e e x i s t s T e T such t h a t

- 131 -

P r o o f :

a) => b) : By 10.6, v <_ y on B o r e l s u b s e t s of E = {Gv = °°}

L e t 1)1 be a [ 0 , l ] - v a l u e d B o r e l f u n c t i o n on E such t h a t

v(Z) = <f> dy f o r every B o r e l s e t Z £ E . Extend <|> to be 0 on

H^\E , and l e t f = <}>(B ) . Then f i s B -measurable; i n d e e d , f o 0 u

i s B^-measurable.

Now t h e r e a r e unique measures a, 8 i n D such t h a t a + v = y ,

6 + v = v . We have Ga _> Gg , and g({Gg = « = } ) = 0 . Hence, by E

2.2, g does not charge p o l a r s e t s . But then a c c o r d i n g to the

c o n j e c t u r e 7.12, t h e r e i s a ( 8 t ) - s t o p p i n g time T such t h a t

T

L e t T(U>) = f(u>)6 n + ( 1 - f ( u ) ) 6 _ / , , (to e Q) . Then T e T 0 1 (.to;

I f A i s any B o r e l s u b s e t o f D then

y x ( A ) = P y(dto)x(to)({t e [0,R(to)) : B t ( u ) e A})

P y(dto) [f ( t o ) l A ( B 0 ( t o ) ) l { ( ) < R } ( t o )

+ ( l - f ( t o ) ) l A ( B T ( t o ) ) l { T < R } ( t o ) ]

y(dx)<j>(x)l A(x) + P a ( d a J ) l A ( B T ( t o ) ) l { T < R } ( t o )

= v £ ( A ) + a T ( A ) = v(A)

Thus y = v T

- 132 -

b) => a ) : L e t y, v be measures i n D such t h a t Gy i s a

p o t e n t i a l , Gy _> Gv , and t h e r e e x i s t s a B o r e l s e t A £ D such t h a t

v(Z) = y(ZnA) f o r a l l B o r e l p o l a r s e t s Z £ D .

L e t T e T such t h a t y_ = v . ( A c t u a l l y , any randomized

( B ^ - s t o p p i n g time T such t h a t y_ = v w i l l do here.)

L e t E = { G y = ° ° } , y ' = y_^_ , y" = y_ , v' = y^ , and l e t

v" = y" . Then Gv' _< Gy' and Gv" <_ Gy" . A l s o y' does not

charge p o l a r s e t s . Thus, by 7.11, t h e r e i s a ( B ^ - s t o p p i n g time T'

such t h a t y_j,, = v' .

Now f o r any B o r e l p o l a r s e t Z £ D ,

v"(Z) = v(Z) = y(ZnA) = y(ZnAnE)

= y"(ZnA) .

Thus, a c c o r d i n g to the h y p o t h e s i s o f b ) , t h e r e e x i s t s T" e T such t h a t

y",, = v" . By the d e f i n i t i o n o f T , t h e r e i s a 8 ..-measurable T U

[ 0 , l ] - v a l u e d f u n c t i o n f and a ( B t ) - s t o p p i n g time T" such t h a t

T ( _ ) = f ( u)6Q + ( 1 - f ( _ ) ) 6 _ „ ^ ^ f o r a l l cu i n . C l e a r l y we may

assume t h a t f = 1 on {T = 0} .

L e t g = _ A ° B Q . Then f o r any B o r e l s e t H £ D ,

g d P y " = y"(HnA)

{BQ e H}

= y(HnAnE) = v(H E)

= v"(HnE) ( s i n c e v' does not charge p o l a r s e t s )

= y"(HnE) T P V " ( d „ ) [ f ( _ ) l { B o £ H n E j Q < R } ( _ ) + < l - f ( o > » l { B _ e HnE, T < R}

(03) ]

- 133 -

f dP u {B Q e H}

I t f o l l o w s t h a t g dP1" f dP M f o r a l l F i n Br S i n c e

g and f a r e b o t h B 0-measurable, g = f P - a.e.

L e t T = <

T' on { B Q I E}

0 on {BQ e E n A}

T" on {BQ £ E \ A}

Then T i s a ( B t ) - s t o p p i n g time, and y T = v .

10.10. Here i s a c o r o l l a r y o f 10.5 which s h o u l d make i t p o s s i b l e t o

reduce to the case where u i s a p o i n t mass i n a t t e m p t i n g to prove

the c o n j e c t u r e 7.12.

P r o p o s i t i o n : L e t y, v be measures i n D such t h a t Gy i s a

p o t e n t i a l and Gy _> Gv .

Then t h e r e i s a f a m i l y (v ) „ o f measures i n D such t h a t : x xeD

a) F o r a l l x e D , Gv < G6 x — x

b) F o r a l l A e B o r e l D , x H- ^ x ( A ) i s a B o r e l f u n c t i o n i n

y ( d x ) v x ( A ) D and v(A) =

P r o o f : By 10.5, t h e r e i s a randomized (B^)-stopping time x such

- 134 -

t h a t v = v . By 9.15, we can modify T on a s e t of P y-measure T

0 so t h a t i t becomes a randomized ( B t + ) - s t o p p i n g time.

Now l e t v = (5 ) f o r each x i n D . X X T g

10.11. Here i s a n o t h e r a p p l i c a t i o n o f 10.5.

P r o p o s i t i o n : L e t D^, be Green r e g i o n s i n ]R n w i t h Green

f u n c t i o n s G^, r e s p e c t i v e l y . Suppose - 2 *

L e t u, v be measures i n such t h a t u ( D 2 \ D ^ ) = 0 = vCD^XD^) ,

G 2 U i s a p o t e n t i a l i n T)^ , and G^u > G^v . Then G^V 2. 2V '

P r o o f : L e t R ± = i n f { t >_ 0 : B t I D ±} ( i = 1, 2) . Now G^ <_ G 2

on x ; hence G^u £ G^P i n .' In p a r t i c u l a r , G^u i s

a p o t e n t i a l i n . Thus by 10.5, t h e r e i s a randomized (8 t)-

s t o p p i n g time x such t h a t f o r a l l A i n B o r e l ,

v(A) = P U(dco)x(co)({t e [0, R-^to)) : Bt(u>) £ A})

L e t a ( u ) ( E ) = x (co) (En [ 0 ^ (co) ) ) + x (co) ( (R^co) , »])6oo(E) f o r co e n ,

E e B o r e l [0,°°] . Then f o r any t e [0,°°] and any co £ ft ,

a(co)([0,t]) = x(co)(H(co))

where H = (ft * [0,t]) n [0,R) £ 8 ® B o r e l [0,t] . Thus a i s a

randomized (8^)-stopping time . I f A £ B o r e l , then

v(A) = v(AnD x)

P y(dco)a(co) ({t £ [0,R 1(w)) : Bt(u>) e A n D ^ )

- 135 -

P y(d_ ) a ( _ )({t e [0,R (_)) : _ t(_) e A})

P y(d_)a(_)({t e [0,R 2(u)) : B (_) e A})

Thus G 2v £ G 2u , by 10.3.

- 136 -

2 11. EMBEDDING MEASURES IN BROWNIAN MOTION IN R , USING RANDOMIZED

STOPPING TIMES

11.1. I n 10.5, we saw t h a t i f y, v a r e measures i n a Green r e g i o n

D i n n n such t h a t Gy i s a p o t e n t i a l , t h e n Gy > Gv i f f t h e r e

e x i s t s a randomized (B._)-stopping time T such t h a t y_ = v ,

where y i s as d e f i n e d i n 10.1. A l s o , i n 8.20, we saw t h a t i f T

y v

y, v a r e measures on R such t h a t U and U a r e p o t e n t i a l s ,

then t h e r e e x i s t s a y - s t a n d a r d ( 8 t ) - s t o p p i n g time T such t h a t

y_ = v i f f U y _> U V and y(R.) = v(lR) .

I n t h i s s e c t i o n , we a r e g o i n g to prove the analogue o f the 2

l a t t e r r e s u l t f o r measures on E . In t h i s c a s e , the example g i v e n

a t the b e g i n n i n g o f s e c t i o n 9 shows t h a t we must use randomized

s t o p p i n g t i m e s .

11.2. Throughout t h i s s e c t i o n , i f T i s a randomized ( B t ) - s t o p p i n g

time and y i s a measure on E n , then y_ w i l l denote the measure

on B o r e l E n d e f i n e d by

P T(A) = P y(dco)x(u.) ( { t e [0,°°) : B^w) e A})

Of c o u r s e i f T a r i s e s from a genuine ( B t ) - s t o p p i n g time T then

y_. = y_ where y_ i s as d e f i n e d i n 8.1.

More g e n e r a l l y , i f (X,A,m) i s a measurable space and T a r i s e s

as i n 9.18(a) from a s t o p p i n g time T over the p r o d u c t enlargement

o f (n,B,8t,Py) by (X,A,m) then

xm ( d x ) y T ( . , x ) '

- 137 -

11.3. D e f i n i t i o n : L e t y be a measure on ]R n such t h a t U y i s

a p o t e n t i a l . L e t T be a randomized ( 8 ^ )-stopping time.

We s h a l l say T i s y - s t a n d a r d i f f whenever p, 0 a r e

randomized (B ..-stopping times such t h a t p <_ a <_ T (see 10.2 f o r

the meaning o f <_ her e ) then U P and U 0 a r e p o t e n t i a l s and

U p > U ° .

11.4. D i s c u s s i o n o f 11.3: I f n >_ 3 then any x i s y - s t a n d a r d , by

10.3.

Suppose n <_ 2 and x i s y - s t a n d a r d . Then y^0R n) 2. u0R n)

by 4.8 ( i f n=2) o r 4.11 ( i f n=l) . Thus "x i s P y - a.s. f i n i t e " ;

more e x p l i c i t l y P y ( { _ : X(CJ) ({«=}) 4 0}) = 0 .

11.5.

a) L e t m be Lebesgue measure on B o r e l (0,1) . By 9.16, the map

f f(m) i s a 1-1 map o f the s e t o f i n c r e a s i n g l e f t - c o n t i n u o u s

maps o f (0,1) t o [0,°°] onto the s e t of p r o b a b i l i t y measures

on [0,°°] . C l e a r l y t h i s map i s an o r d e r isomorphism, where

f o r p r o b a b i l i t y measures a, 6 on [0,°°] , we say a <_ 8 i f f

a((t,»]) £ 8((t,°°]) f o r a l l t e [0,°°) . I t f o l l o w s t h a t t h i s

o r d e r i n g o f p r o b a b i l i t y measures on [0,°°] i s a l a t t i c e o r d e r i n g .

G i v e n any two p r o b a b i l i t y measures a, 6 on [0,°°] , we denote

t h e i r l e a s t upper bound i n t h i s o r d e r i n g by a v B , and t h e i r

g r e a t e s t lower bound by ot A 8 .

b) I f A i s a s e t and a, T a r e maps from A to p r o b a b i l i t y

measures on [0,°°] we d e f i n e a A X and a v x i n the o b v i o u s

- 138 -

p o i n t w i s e f a s h i o n .

c) I f (A , F ) i s a measurable space and a , T a r e r r v ' s i n [0,°°]

over (A , F ) then so a r e O A T and a v T , by 9.18 (where

we take F = F f o r 0 <_ t < 0 0) .

d) I f ( A , F , F t ) i s a f i l t e r e d measurable space, where (F ) i s

r i g h t - c o n t i n u o u s , and i f a, x a r e randomized ( F t ) - s t o p p i n g

times then so a r e a A T and a v x , by 9.18.

e) I f a i s a p r o b a b i l i t y measure on [O, 0 0] and t e [O, 0 0] then

a A 6 can be d e s c r i b e d q u i t e e x p l i c i t l y ; a A 6 = a r n , t t [ 0 , t ]

+ a((t,°°])6 t I f f i s any non - n e g a t i v e B o r e l f u n c t i o n on

[0,°°] t h e n

( a A 6 t ) ( d s ) f ( s ) = < x ( d s ) f ( s A t ) .

11.6. Lemma: L e t (A ,F,Q) be a a - f i n i t e measure space.

L e t (x ) be a net c o n v e r g i n g to x i n RRV = RRV(A , F,Q;[0,«])

L e t T : A ->- [0,°°] be F-measurable. L e t a.(a)) = x.(co) A 5 ,

a(co) = x(co) A 6r T(u)

Then a . ->• a i n RRV l

P r o o f : L e t (Z )„ be any c o n t i n u o u s r e a l - v a l u e d p r o c e s s over t 0<_t<°° J

(A , F ) s a t i s f y i n g

Q(dco) sup IZ (to) I < »

0<t<°°

Then (Z m ) i s a l s o such a p r o c e s s , t AT

- 139 -

Hence Q(doj) T . ( _ ) ( d t ) Z (_) 1 tAT

Q(du>) x ( _ ) ( d t ) Z t A _ ( _ )

That i s , by 1 1 . 5 ( e ) ,

Q(dco) a 1 ( _ ) ( d t ) Z t ( _ )

Q(da») o ( _ ) ( d t ) Z t ( _ ) •

11.7. Lemma: L e t (A,F,Q) be a a - f i n i t e measure space,

L e t ( x ± ) be a n e t c o n v e r g i n g to x i n RRV = RRV(A,F,Q;[0,»])

Suppose l i m sup t-*» i

Q(d_)x (_)((t,»]) = 0 . Then:

a) Q(d_)x.(_)({»}) = 0 .

b) I f (Z ) i s any c o n t i n u o u s r e a l - v a l u e d p r o c e s s w i t h time

s e t [0,°°) (not [0,°°]) such t h a t

Q(dui) sup |Z ( u ) | < 0<t<~

then Q(du) x. (uO(dt)Z (ui) l t

Q(dco) T ( _ ) ( d t ) Z t ( _ )

P r o o f :

a) L e t f be a s t r i c t l y p o s i t i v e F-measurable f u n c t i o n on A

- 140 -

such that f dQ

For each t in [0,°°) let gfc be a continuous [0,1]-

valued function on [0,°°] such that g = 0 on [0,t]

and gfc = 1 on [t+1, 0 0 ] . Then

Q(du) T i ( u ) ) ( d s ) f ( o ) ) g t ( s ) + Q(do)) T ( u ) ( d s ) f ( a > ) g t ( s )

for each t . But

lim sup t-*» i '

Q(dco) T_ ((JJ) (ds)f (co)gt(s) = 0 . Hence

lim Q(doj) x(to)(ds)f(to)g (s) = 0 . That i s ,

Q(dto)f (to) T (to) ({<»}) = 0 . As f is s t r i c t l y positive, a)

is proved.

For each s e [0,°°) let h be a continuous [0,l]-valued s

function on [0,°°] such that h = 1 on [0,s] and

h = 0 on [s+1, <*>] . Then s

Q(dto) x. ( to)(dt)h (t)Z. (to) X S t

T(t0)(dt)h g(t)Z t(t0)

for each s . Letting s go to <*> , and using the

"tightness condition" on (x ) , we obtain the desired

result.

- 141 -

11.8. Lemma: L e t T be a s e t of randomized ( 8 ^ ) - s t o p p i n g times

such t h a t i f a, x a r e randomized ( B t ) - s t o p p i n g times w i t h a <_ x

and xeT then a e T .

L e t y be a f i n i t e measure on ]R n . Suppose t h a t

l i m sup y ({x e ]R n : | |x| | r } ) = 0 T T

r-x» Tel

Then l i m sup t-**> xeT

P y ( d u ) ) x ( _ ) ( ( t , °°]) = 0

P r o o f : Emulate the p r o o f o f 8.10,

n 11.9. Lemma: L e t y be a measure on IR (where n = 1 o r 2)

such t h a t U y i s a p o t e n t i a l .

L e t (ft' , B ' ,8J_,Q,if>) be the p r o d u c t enlargement o f ( f t , 8 , B t , P y )

by (X,A,m) , where X = (0,1) , A = B o r e l X , and m = Lebesgue

measure on A .

L e t x be a randomized ( B ^ - s t o p p i n g time, and l e t T be

the ( B p - s t o p p i n g time a s s o c i a t e d to x by 9.18(b). Suppose

T dQ < 0 0 . Then x i s y - s t a n d a r d .

P r o o f : F o r each t , l e t B^ = B t ° v) . Then (ft' ,8' ,8|.,Bj.,Q) i s

a g e n e r a l i z e d Brownian motion p r o c e s s w i t h i n i t i a l d i s t r i b u t i o n y

Note t h a t i f a i s any randomized (B ) - s t o p p i n g time and i f

S i s the ( B ^ ) - s t o p p i n g time a s s o c i a t e d t o a by 9.18(b), then

y (E) = Q(B e E , S < °°) f o r each B o r e l s u b s e t E o f ]R n .

A l s o , i f p, a a r e randomized ( B ^ - s t o p p i n g times such t h a t

p <_ a , and i f R, S a r e the ( B | _ )-stopping times a s s o c i a t e d to

- 142 -

p, a by 9.18(b), then R _< S . Now emulate the p r o o f of 8.5,

wo r k i n g w i t h the p r o c e s s (ft' ,8' ,BJ_,Bj_,Q) i n s t e a d of ( f t , 8 , B t , B t , P U )

11.10. C o r o l l a r y : L e t u be a measure on ]R n such t h a t U P i s

a p o t e n t i a l .

L e t T be a randomized ( B ^ J - s t o p p i n g time such t h a t

P y({u) : x ( t o ) ( ( t , «>]) 4 0}) = 0 f o r some t i n [0, °°) . Then x

i s y - s t a n d a r d .

P r o o f :

C l e a r . •

11.11. Theorem: L e t y be a measure on 1R (where n = 1 o r 2)

such t h a t U P i s a p o t e n t i a l .

L e t T be a s e t of good measures on ]R n such t h a t whenever

a, B a r e measures on ]R n such t h a t U a, U B a r e p o t e n t i a l s and

U dy > U Bdy f o r a l l y e T , then U a >_ U 6

Suppose (x^) i s a net c o n v e r g i n g to x i n RST(ft,B,B t,P ) ,

and each x_ i s y - s t a n d a r d . C o n s i d e r the f o l l o w i n g s t a t e m e n t s :

a) There i s a measure a on ]R n such t h a t U a i s a p o t e n t i a l ,

a C R n ) = yQR n) , and U 1 >_ U 0 1 f o r a l l i .

b) l i m sup t-x» i

P M ( d a i ) x . ( c o ) ( ( t , - ] ) = 0 .

c) f dy + x. I f dy^ f o r every bounded c o n t i n u o u s f u n c t i o n on

T d) U i s a p o t e n t i a l and

y < T i U dy -> U dy f o r a l l

- 143 -

e) x i s y - s t a n d a r d .

Then a) => b) => c ) , and a) => d) => e ) .

P r o o f : a) => b ) . L e t S be the s e t of randomized ( B t ) - s t o p p i n g

times a such t h a t a <_ x_^ f o r some i . Then a e S i m p l i e s

y v t h a t U i s a p o t e n t i a l and U _. U , s i n c e the T

i ' s a r e

y - s t a n d a r d . Hence l i m sup y ({x e l R n : | |x| | >_ r } ) = 0 , by r-x» aeS

8.9. The statement b) then f o l l o w s by 11.8.

b) => c ) . Apply 11.7, w i t h (Z ) = (f(B..)) . (Note t h a t

P y ( f t ) = y ( R n ) < 0 0 as U y i s a p o t e n t i a l and n <_ 2 .)

a) => d ) . S i n c e a) => c ) , we have t h a t y -»• y weakly. — x . x

l

Then, r e a s o n i n g as i n the p r o o f of (a => b) of 8.11, we can deduce

the statement d ) .

d) => e ) . For each randomized ( B ^ ) - s t o p p i n g time a and

each s i n [0, °°) , l e t oAS denote the randomized ( B t ) - s t o p p i n g

time d e f i n e d by ( O A S ) ( _ J ) = O ( O ) ) A 6 . Then f o r each s i n [0, °°) , s

x ^ s -> X A S i n RST , by 11.6, and y x . A S y u 1 > u s

by 11.10. Thus, a p p l y i n g (a => d) w i t h ("^^ r e p l a c e d by ( x ^ A s ) ^ ,

x r e p l a c e d by x A s , and a e q u a l to y g , we f i n d t h a t

y. X . A s

U 1 dy -»-y

T T X A s , U dy

f o r a l l measures y e Y . Now f o r each i ,

X . A S X . u 1 > u 1

- 144 -

s i n c e T i s y - s t a n d a r d . Thus we f i n d t h a t 1

U A U

u T A s > u T

f o r a l l s i n [0, °°) . From t h i s , w i t h s = 0 , i t f o l l o w s t h a t t h e

t o t a l mass of y i s > t h a t of y , whence t h e s e two measures have T —

the same t o t a l mass, and so P y ( { u e Q : x(w)({°°}) 4 0}) = 0 .

Now suppose p and a are randomized ( 8 t ) - s t o p p i n g times w i t h

p <_ a <_ T . Then f o r any s i n [0, 0 0) ,

U V V u

u p A s > u a A s > u T A s > u T . L e t t i n g s °° and a p p l y i n g 8.7 and 8.8 (we can t a k e a = y^ i n 8.8

y y here) we c o n c l u d e t h a t U D and U a r e p o t e n t i a l s and

y y u 0 > u 0 .

(As s ->• 0 0 , y -*• u weakly and y y weakly, by p A s p O A s a

(b => c ) . T h i s i m p l i c a t i o n a p p l i e s because p , a <_ x and

» p y ( T = ») » = o .)

11.12. C o r o l l a r y : L e t y be a measure on 3R (where n = 1 or 2)

PC

time. Then the f o l l o w i n g a r e e q u i v a l e n t :

such t h a t U P i s a p o t e n t i a l . L e t x be a randomized ( B ) - s t o p p i n g

a) x i s y - s t a n d a r d .

b) U T i s a p o t e n t i a l and U T A t > U T f o r a l l t i n [0,™)

( x A t i s as d e f i n e d i n the p r o o f o f 11.11(d).)

P r o o f : a) => b ) . C l e a r .

- 145 -

u. b) => a) . We always have y TOR n) £ yOR n) • Now s i n c e U P > U T

and n <_ 2 , we have y T Q R n ) >_ yQR n) , by 4.8 ( i f n = 2) o r 4.11

( i f n = 1) . Now T A t i s y - s t a n d a r d f o r each t i n [0,°°) ,

by 11.10, and T A t T i n RST(ft,B,8 t,P y) as t -> °° . To con c l u d e

the p r o o f , a p p l y 11.11 w i t h a = y^ and ^ x ^ ) ^ = ( T A t ) t •

11.13. C o r o l l a r y : L e t y be a measure on ]R n (where n = 1 or 2)

such t h a t U P i s a p o t e n t i a l .

L e t (ft* ,8' ,8j.,Q,<Jj) be an o p t i o n a l enlargement o f (ft,B,B t,P y)

and l e t B' = B t ° I|J f o r each t . For each (8J_)-stopping time S

l e t y g be the measure on H n d e f i n e d by y g ( E ) = Q ( B g e E, S < °°)

L e t T be a (8 j_)-stopping time, and l e t T be (a v e r s i o n o f ) the

randomized ( 8 t ) - s t o p p i n g a r i s i n g from T as d e s c r i b e d i n 9.8 and

a) => b) o f 9.13. Then the f o l l o w i n g a r e e q u i v a l e n t :

a) Whenever R and S a r e (8J . )-stopping time and R <_ S <_ T

y R y S P R y S then U and U a r e p o t e n t i a l s , and U >_ U

b) T i s y - s t a n d a r d .

P r o o f : a) => b) . F o r each t i n [0,°°) l e t ( T A t ) (co) = T(to )A<5 t

(co e Q) . T A t a r i s e s from T A t i n the same way t h a t T a r i s e s y

from T . A l s o , y = y m and y = y m . Thus U i s a

T T T A t T A t

P T A t PT

p o t e n t i a l , and U > U f o r a l l t i n [0,°°) . Thus T i s

y - s t a n d a r d , by 11.12.

b) => a ) . L e t p, o be ( v e r s i o n s o f the) randomized (B ) -

s t o p p i n g times a r i s i n g from R, S r e s p e c t i v e l y , as d e s c r i b e d i n 9.8

- 146 -

and a) => b) o f 9.13. Then p <_ a <_ T Q - a.e. L e t a' = O A X , y a y p

p ' = p A a ' . Then y . = y and y , = y . Thus U and U a r e a a p p

y y 0

p o t e n t i a l s , and U >_ U . But y = y and y = y . p R o o |-j

11.14. Theorem: L e t y, v be measures on H 2 such t h a t U y, U V

a r e p o t e n t i a l s . Then the f o l l o w i n g a r e e q u i v a l e n t : 0 0

a) U P > U V and p ( l ) = v ( l ) .

b) There i s a y - s t a n d a r d randomized (B^_)-stopping time x

such t h a t y = v . x

P r o o f : b) => a ) . C l e a r .

a) => b) . For 0 < r < 0 0 , l e t A be the u n i f o r m u n i t 2

d i s t r i b u t i o n on the open b a l l o f r a d i u s r c e n t r e d a t 0 i n 1R ,

and l e t v = v*A , where * denotes c o n v o l u t i o n . Then r r

v v r v r U = U *\ . Thus U i s f i n i t e (and c o n t i n u o u s ) . A l s o , by 4.18 v r v of Helms [ 1 ] , U tU as r+0 . F o r each r , t h e r e e x i s t s a

y - s t a n d a r d (B ) - s t o p p i n g time T_ such t h a t y_ = v_ , by 8.20.

r

By 9.23, t h e r e i s a d e c r e a s i n g sequence ( r ( i ) ) i n (0,°°) and a

randomized ( 8 ^ - s t o p p i n g time x such t h a t r ( i ) -»- 0 and Tr ( i ) T

i n RST ( n , B , B_,P y) . Then y = v and x i s y - s t a n d a r d , by 11.11 t x

(we can take a = v i n o r d e r t o a p p l y 11.11 . ) r ( 0 ) ^ g

- 147 -

12. EMBEDDING PROCESSES IN BROWNIAN MOTION

In t h i s s e c t i o n we d e f i n e a c l a s s o f p r o c e s s e s which, f o l l o w i n g

Chacon [ 1 ] , we c a l l p o t e n t i a l p r o c e s s e s . E s s e n t i a l l y , t h e se t u r n out

to be the p r o c e s s e s which can be embedded i n an o p t i o n a l enlargement

o f Brownian motion by means of an i n c r e a s i n g f a m i l y o f s t a n d a r d

s t o p p i n g times - see 12.7, 12.16, and 12.17.

12.1. N o t a t i o n : I f y i s a measure on B o r e l ]R^ then we s h a l l o

w r i t e U P f o r U V where v i s the r e s t r i c t i o n of y to B o r e l H n .

12.2. D e f i n i t i o n : L e t I £ [-«=, °°] , and l e t n be a p o s i t i v e

i n t e g e r . An n - d i m e n s i o n a l p o t e n t i a l p r o c e s s w i t h time s e t I

i s a system (A,F,F^,X^,P) where:

a) (A,F,P) i s a p r o b a b i l i t y space.

b) ^ j ^ i e i "*"s a n i n c r e a s i n g f a m i l y of sub - a - f i e l d s of F .

c) ( X . ) . T i s a f a m i l y o f ]R n-valued random v a r i a b l e s over l l e i J 3

(A,F,P) .

d) For each i e I ,

i ) X. i s F.-measurable, l l

. .. T Ilaw(X-j) . . n ) U -1- xs a p o t e n t i a l .

i i i ) I f n <_ 2 , P C X ^ S ) = 0 .

e) Whenever S and T a r e ( F ^ ) - s t o p p i n g times t a k i n g on o n l y

f i n i t e l y many v a l u e s , and S <_ T , we have

law(X ) law(X ) U b > U 1 .

- 148 -

law(X-) (We remark t h a t i t f o l l o w s from d) t h a t U and

law(X_)

U a r e p o t e n t i a l s . A l s o , an ( F ^ ) - s t o p p i n g time i s

assumed to take v a l u e s i n I .)

12.3. P r o p o s i t i o n : L e t I _ [-°°, 0 0] , l e t n be a p o s i t i v e i n t e g e r ,

and l e t ( A ,F , F^.X^P) be a system s a t i s f y i n g a) through d) of 12.2.

Then a) and b) below are e q u i v a l e n t :

a) ( A , F , F ,X ,P) i s an n - d i m e n s i o n a l p o t e n t i a l p r o c e s s .

b) F o r each y e ] R n , ( $ ( X i 9 y ) l ^ x ^ 9}^ i s a s u P e r m a r t i n g a l e i

i n the extended sense over (A , F , F^,P) .

Moreover, i f n = 1 then a t h i r d e q u i v a l e n t c o n d i t i o n i s :

c) (X_^) i s a m a r t i n g a l e over (A , F , F^,P) .

P r o o f : I f S i s an ( F _ ^ )-stopping time t a k i n g on o n l y f i n i t e l y

many v a l u e s and y e ] R n then

law(X )

U (y) = E ( * ( X - , y ) l ^ 3 } ) .

Thus (b => a) i s a consequence of the o p t i o n a l s a m p l i n g theorem. a) => b) . L e t i , j e I w i t h i <_ j and l e t F e F_ . We

w i s h to show t h a t Z Y dP > F 1

Z. dP f o r each y e ]R , where F 3

( X ^ j y ) ! ^ . ^ ^ 3 j f o r each k e I and each y e l R n .

L e t S = i on F

j on A\F

- 149 -

and l e t T = j on a l l of A . Then S and T a r e (F ) - s t o p p i n g

times t a k i n g on o n l y f i n i t e l y many v a l u e s and S £ T so

law(X_) law(X ) U > U

That i s , zl dP > Z y dP f o r a l l y e R n From t h i s we can

c o n c l u d e t h a t

law(X_)

Z Y . dP > ZY dP f o r each y e TR such t h a t

U " (y) < 0 0 • Thus i f we l e t y, v be the measures on TR

d e f i n e d by

y(A) = P ( { X ± e A} n F)

v(A) = e A} n F)

law(X_) then U y > U V except p o s s i b l y on the p o l a r s e t {U = 00}

Hence U y > U V on a l l o f TRn . That i s , Z y dP > Z Y . dP f o r

a l l y e H n , as d e s i r e d .

Now suppose n = 1 .

Then P ( X ± = 9) = 0 f o r each i e I

c) => b) . S i n c e n = 1, 3>(x,y) = - -|-|x-y| . Thus we need o n l y

a p p l y Jensen's i n e q u a l i t y f o r c o n d i t i o n a l e x p e c t a t i o n s .

a) => c ) . L e t S, T be (F ) - s t o p p i n g times assuming o n l y

f i n i t e l y many v a l u e s , such t h a t S <_ T .

law(X ) law(X ) Then U _. U , so law(X.) and lawCX.,) have the

same c e n t r e o f mass, by 4.11. Hence E ( X g ) = E(X_) . I t f o l l o w s t h a t

(X^) i s a m a r t i n g a l e over (A,F,F_^,P) .

- 150

12.4. D e f i n i t i o n : L e t y be a measure on H n such t h a t U y i s

a p o t e n t i a l , and l e t (ft' , B ' , B ',Q,ij;) be an o p t i o n a l enlargement o f

( f t , B , 8 t , P y ) . L e t B' = B f c ° i|; f o r 0 <_ t < » .

A ( B p - s t o p p i n g time T w i l l be c a l l e d s t a n d a r d ( r e l a t i v e to

Q) i f f whenever R, S a r e ( B p - s t o p p i n g times and R <_ S <_ T then

l a w ( B R ) l a w ( B g ) l a w ( B R ) law (Bp

U and U a r e p o t e n t i a l s and U ^ U

12.5. D i s c u s s i o n o f 12.4:

a) Note t h a t the d e f i n i t i o n 12.4 i s a g e n e r a l i z a t i o n o f the

d e f i n i t i o n 8.2.

b) By 11.13, T i s s t a n d a r d r e l a t i v e to Q i f f " t h e " randomized

( B p - s t o p p i n g time a r i s i n g from T i s y - s t a n d a r d .

c) One can show t h a t T i s s t a n d a r d r e l a t i v e to Q i f f f o r

eve r y compact s e t K £ ]R n ,

T(w) Q(du>) l K ( B ' ( o 3 ) ) d t

law (Bp i s f i n i t e and e q u a l to I (U - U ) . T h i s i s a si m p l e

'K g e n e r a l i z a t i o n o f 8.13, and we omit the p r o o f .

d) I f n >_ 3 , e v e r y ( B p - s t o p p i n g time i s s t a n d a r d r e l a t i v e to

Q •

e) I f n <_ 2 and y ( { 8 } ) = 0 and T i s s t a n d a r d r e l a t i v e to

Q then T i s f i n i t e Q - a . s .

12.6. P r o p o s i t i o n : L e t y be a p r o b a b i l i t y measure on ]R n such t h a t

U y i s a p o t e n t i a l , and l e t (ft' , B ' ,8pQ,iJ>) be an o p t i o n a l enlargement

- 151

of ( f t , B , B t , P y ) . L e t = B ° 1J1 f o r 0 <_ t < 00

L e t I c [ - o o o o ] and l e t ( T . ) . _ be an i n c r e a s i n g f a m i l y o f - l i e l

( B p - s t o p p i n g times which a r e s t a n d a r d r e l a t i v e to Q . Then

(ft',8',8.j, jB^-Q) i s an n - d i m e n s i o n a l p o t e n t i a l p r o c e s s w i t h time i i

s e t I .

P r o o f : C l e a r l y the i n d i c a t e d system s a t i s f i e s a) through d) of the

d e f i n i t i o n 12.2. To f i n i s h the p r o o f we s h a l l show t h a t i t s a t i s f i e s

b) of p r o p o s i t i o n 12.3.

L e t i , j e I w i t h i <_ j and l e t F e B_ . i

L e t S = \

T. on F l

T. on S7'\F J

and l e t T = T. . Note t h a t S i s a J

law(B^) law(B^)

( B p - s t o p p i n g time. As T i s s t a n d a r d , U > B . For

law (Bp y e {U < °°} , we can conclude from t h i s t h a t

* ( Y > y ) 1 { B T ^ } d Q -r 1 1 . X _ $ ( B T . ' y ) 1 { B _ * _} d Q ' ? 1 •

Then, by the method used i n the p r o o f of (a => b) of 12.3, we can

co n c l u d e t h a t t h i s i n e q u a l i t y a c t u a l l y h o l d s f o r a l l y e ]R n . •

12.7. Theorem: L e t I = {0,...,k} where k e U , o r l e t I = K .

L e t n be a p o s i t i v e i n t e g e r , and l e t (A,F,F^,X^,P) be an

n - d i m e n s i o n a l p o t e n t i a l p r o c e s s . L e t u = law(X^) . Then:

a) There i s a p r o d u c t enlargement (ft' ,8' ,B|. ,Q,^) of ( f t , B , B t , P y )

- 152 -

and an i n c r e a s i n g f a m i l y ^ ± ^ ± e j _ °^ ( B j . )-stopping times

which a r e s t a n d a r d r e l a t i v e to Q , such t h a t (B*, ) . and i

( X . ) . _ have the same j o i n t d i s t r i b u t i o n , where o f cou r s e 1 1 e l

B£ = B ° IJJ f o r 0 <_ t <_ <=° . '

b) I f n = 1 , no enlargement i s n e c e s s a r y i n a ) . That i s ,

t h e r e i s an i n c r e a s i n g f a m i l y ^ ± ^ ± e j °^ u - s t a n d a r d

(8 f c)-stopping times such t h a t ^ B x . ^ i e l 3 n d ^ X i ^ i e l n a v e

the same j o i n t d i s t r i b u t i o n .

P r o o f : a) L e t (ft 1,8' ,B^,Q,IJJ) be the p r o d u c t enlargement of

(f t , B , 8 t , P M ) by ( L , L , A ) , where L = ( 0 , 1 ) , L = B o r e l L, and

A = Lebesgue measure on L .

L e t B^ = B ° f o r 0 <_ t <_ °° . E v i d e n t l y we can l e t T Q = 0 .

Suppose j e I such t h a t j+1 e I , and suppose s t a n d a r d (8*.)-stopping

times T n < ... < T. have been chosen to t h a t (Y.,...,Y.) 0 — — j 0 j

= (B' ,...,B' ) and X ,...,X. have the same j o i n t d i s t r i b u t i o n . We 0 j 2

s h a l l show t h a t one can choose a s t a n d a r d ( B j.)-stopping time

T.,, > T. so t h a t Y_,...,Y., B' and X.,...,X., X... have the J+1 ~ J 0' j T j + 1 0' j ' j+1

same j o i n t d i s t r i b u t i o n .

L e t D = ]R n and l e t E = E." . L e t a be the j o i n t d i s t r i b u t i o n

o f X_,...,X. and l e t 3 be the j o i n t d i s t r i b u t i o n o f X_.,...,X. , X.,., 0 3 0 3 j+1

Then a i s a p r o b a b i l i t y measure on E ^ + ^ , t

p r o b a b i l i t y measure on E2+^ , and the p r o j e c t i o n map

( XQ 5 • • • > X j ' X j +1^ ' ^ ^ TJ ' * * ' ' j

i s a

- 153 -

c a r r i e s onto a

Hence by 9.4, t h e r e i s an a - e s s e n t i a l l y unique B o r e l measurable

f a m i l y (8 ) . . of p r o b a b i l i t y measures on E such t h a t f o r a l l X x e E 3 l

A e B o r e l ( E j + 2 ) ,

8(A) a(dx)B ( A ( x ) ) .

S i n c e X.,...,X., X.,. i s a p o t e n t i a l p r o c e s s , 0' J j+1

F x E

$ ( x j , y ) l D ( x j ) d 8 ( x 0 , . . . , x j + 1 ) > FxE

l ( x . + 1 , y ) l D ( x . + 1 ) d 8 ( x 0 , . . . , x . + 1 )

f o r a l l F e B o r e l ( E ^ + ± ) and a l l y e D . T h i s may be deduced from

(a => b) of 12.3, by change o f v a r i a b l e s . Now the f i r s t i n t e g r a l i n

t h i s i n e q u a l i t y may a l s o be w r i t t e n as

•KXj , y ) l _ ( x ) d a ( x Q , . .. ,x..)

w h i l e the second may be w r i t t e n as

U J ( y ) d a ( x Q , . . . , X j ) .

From t h i s one may deduce, by a v e r a g i n g w i t h r e s p e c t to y over each

element of a c o u n t a b l e s e t o f open b a l l s which forms a base f o r the

t o p o l o g y o f D , t h a t f o r a-almost a l l ( X Q , . . . , X _ . ) e E ^ + ± we have

X Q , • • • , x. J 1 • ) l - ( X j ) on D

Of c o u r s e we s h o u l d v e r i f y t h a t U i s a p o t e n t i a l f o r a - a.a. x e

- 154 -

e As the $ 's a r e f i n i t e measures, the U 's a r e OK. We need o n l y

X B check t h a t f o r a - a.a. x , U i s f i n i t e - v a l u e d . F i x any y e D

Then U x ( y ) d a ( x )

. + 2 ^ " ( x j + 1 , y ) 1 D ( x j + 1 ) d B ( x Q , . . . , x . , x . + 1 )

$ _ (Vl' y ) 1{X. + 1 * o)

law(X ) = U_ J (y) < » .

B x i+1 Thus, i n p a r t i c u l a r , U_ (y) < °° f o r a - a.a. x e E J , f o r t h i s v

p a r t i c u l a r y . But i f y i s a f i n i t e measure on E and U_ i s

f i n i t e a t one p o i n t t h e n i t i s f i n i t e a t a l l p o i n t s ' — see 1.5 f o r 8

the case n = 1 and 1.6 f o r the case n = 2 ; i f n > _ 3 , U_ = 0

f o r a l l x , as § >_ 0 i n t h i s c a s e .

Thus f o r a - a.a. x = (x.,...,x.) e E^ x D , we can choose a 0 j

6 - s t a n d a r d randomized (B ) - s t o p p i n g time T such t h a t X. t X J

B x ( A ) -x.

P J(dco)x (u))({t e [0, » ] : B (CJ) e A})

f o r a l l A e B o r e l E (more b r i e f l y , B = (6 ) ) . x x. T

J x

(For the p r o o f of t h i s see 10.5 f o r the case n > 3 and 11.4

n = 2 . The' method used to prove 11.14 a l s o works f o r n = 1 , though

i n t h i s c ase we can g e t non-randomized s t o p p i n g times; see 8.20.) Now

i f n < 2 then t h i s takes c a r e of a - a.a. x e E^"*" s i n c e then

- 155 -

B P(X = 3) = 0 . On the o t h e r hand i f n >_ 3 then we have U X <_ 0

f o r a - a . a . x e E ^ x { 3 } ; f o r any such x the p r o b a b i l i t y

measure B must be e q u a l to 6„ so we can take T = "0" . x 3 x

Thus we o b t a i n a f a m i l y (x ) . ,.. o f randomized (8 )-x -J+1 t x e E J

s t o p p i n g times such t h a t f o r a - a.a. x = (x-.,. . . ,x^) e E"^ + ± ,

x i s 6 - s t a n d a r d and 8 = (6 ) x x . x x. x 3 3 x

Now i t w i l l become ap p a r e n t , from ja s e r i e s o f lemmas which f o l l o w ,

t h a t the x 's can be chosen so t h a t f o r each t e (0,°°) the map x

( X , U > ) «• T x ( _ ) ( [ 0 , t ) _

i s ( B o r e l E ^ + X ) ® B^-measurable.

F o r x e E ^ + 1 l e t T be the ((8 ® A ) ) - s t o p p i n g time

a s s o c i a t e d to x^ by 9.18(b), and f o r _' = (_,u) e ft1 l e t

e ' c o ' = ( 9 _) ( f o r 0 < t < °°) t t - —

F i n a l l y l e t T... = T. + S . J + l 3

F i r s t of a l l we c l a i m t h a t S i s a (8^ + t ) - s t o p p i n g time. 3

W e l l , f o r any t e (O, 0 0) ,

{S < t} = u {_'-(_,u) e ft' : u < x f n M ( 6 - , , ^ ) ( [ 0 , r ] ) } r e [ 0 , t ) Y 0 ( U ) ' - - " Y j ( w > T j ( w }

r r a t i o n a l

(See p r o o f of 9.18(b).)

A l s o , f o r any r e [0,t) , the map

156 -

= ( _ , _ ) • * ((Y 0(_'),...,Y j(a»')), 6-

i s (B^ + r > ( B o r e l E"^ + ±) ® B ^ )-measurable and the map j

(( X f.,...,x.),w) •> x ( w ) ( [ 0 , r ] ) u j XQ , .. . 5 x j

i s ( B o r e l E ) ® B^-measurable.

I t f o l l o w s t h a t {S < t} e B_j, + t f o r 0 < t < 0 0 . Thus, as j

i s r i g h t - c o n t i n u o u s , S i s a (B , + t _ ) - s t o p p i n g time, as j j

c l a i m e d . But then, u s i n g the r i g h t - c o n t i n u i t y o f ( B p , one e a s i l y

checks t h a t T_ + S i s a ( B p - s t o p p i n g time. That i s , T j + i i s

a (8|_)-stopping time.

Now l e t Y = B' and suppose f f a r e non-

n e g a t i v e B o r e l f u n c t i o n s on E . Then

f 0 ( x 0 ) . . . f . ( x . ) f . + 1 ( x . + 1 ) d P

f 0 ( x Q ) . . . f . ( x . ) f j + 1 ( x j + 1 ) d B ( x 0 , . . . , x . , x . + 1 )

fo (V"- fj ( x. ) [j f - _ i ( x - . i ) d 3 v (x ) ] d a ( x , j+1 j+1 x Q,...,x j+1 0

(*)

Vi ( u , ) [ n f . ( Y . ( u . ' ) ) ] E i=0 1 1

[ n f . ( Y , (_'))]_,_,__ ( B l

[ fj + l ( B T .

)]dQ(

i=0 j + l x T

Y 0 ( u ' ) , . . . , Y (a ) ' )

(6' „'))dQ (a> ' ) Y. ( a ) ' ) , . . . , Y ( i _ ' ) j

- 157 -

f 0 ( Y 0 ) . . . f . ( Y . ) f . + 1 ( Y . + 1 ) d Q

Thus Y„,...,Y. , Y.,., and X„,...,X. , X.,. have the same j o i n t 0' j j+1 0' j j+1

d i s t r i b u t i o n .

(The s t e p (*) above, and a l s o the s t e p (**) below, f o l l o w from

a "souped up" v e r s i o n o f the s t r o n g Markov p r o p e r t y d e s c r i b e d i n

Meyer [ 2 ] . F o r the r e a d e r ' s convenience we s t a t e and prove t h i s

r e s u l t i m mediately f o l l o w i n g the p r e s e n t p r o o f . )

Now l e t us show t h a t T.,, i s s t a n d a r d . We s h a l l prove t h i s J+1

by a p p l y i n g 1 2 . 5 ( c ) . L e t K be any compact s u b s e t o f D . Then

T. r J

l K ( B ' ) d t dQ

law(B^ )

[ U y - U j ] = K

law(X.) [UM - U ]

K

w h i l e Lj+1

T'. J

l K ( B ' ) d t dQ

(**){

T v ( I N T , , * (6 ' io') Y Q(oo ) , . . . , Y (co ) 1\ J V Bt ( 9T a ) ' ) ) d t A W )

j

Y.(co') Y 0 ( c o ' ) , . . . , Y . ( W ' ) 1 (B')dt]dQ(co*)

IN. L

6 law(Bj, X . x_» • • • > x.

U J - u 0 2

;P J ® A)

K ] d a ( x Q , . . . , X j )

K

X X j • • • j x

U j ( y ) - U °' j ( y ) d a ( x 0 , . . . , x j d y

- 158 -

law(X.) law(X ) U 2 (y) - U J (y)dy

K

Thus 1 ( B ' ) d t dQ IN. L

law(B^ )

[u y - U 2 + 1 ] K

l a w ( X . ^ ) [UP - U 2 + 1 ] < - .

K

I t f o l l o w s t h a t T.,, i s s t a n d a r d . J+1

(b) The p r o o f of t h i s p a r t i s q u i t e s i m i l a r to t h a t of p a r t a ) ,

except t h a t i n p l a c e of the f a m i l y (T ) . of randomized (8 ) -X xeE-5

s t o p p i n g times, we a p p e a l to 8.20 to get a f a m i l y (T ) . - of X x e E 3 + i

genuine ( B p - s t o p p i n g t i m e s . I t w i l l become a p p a r e n t , from the

s e r i e s of lemmas which f o l l o w , t h a t the T 's can be chosen so t h a t x

f o r each t e (O, 0 0) the map

(x,a>) + 1 { T < t j ( u )

i s ( B o r e l E2+^~) ® B ^ m e a s u r a b l e , •

Now he r e i s the v e r s i o n o f the s t r o n g Markov p r o p e r t y mentioned

i n the p r o o f of 12.7.

12.8. P r o p o s i t i o n . L e t ( W , M , M t > Z t , 9 t > P ) be a s t r o n g Markov p r o c e s s

- 159 -

w i t h t r a n s l a t i o n o p e r a t o r s , w i t h s t a t e space (E,E)

L e t T be an ( M ^ - s t o p p i n g time, and l e t f : W x W [O, 0 0]

be M T ® M-measurable. L e t u be any p r o b a b i l i t y measure on E .

Z T ( w ) u , u Then E (f(w,«)) = E M ( f ( w , » ) ° 6 |M )(w) f o r P M - a.a. w £ W

More p r e c i s e l y : Z T(w)

a) The map w E (f(w,«)) i s M - m e a s u r a b l e .

b) F o r any M-measurable f u n c t i o n g : W -> [0, 0 0 ] ,

we have

Z (w) g(w)E ( f ( w , . ) ) d P P ( w )

g(w)f(w,6 Tw)dP^(w)

P r o o f : I t s u f f i c e s to c o n s i d e r f of the form a ® b , where a and

b a r e n o n - n e g a t i v e f u n c t i o n s on W which a r e M-measurable and Z T(w) T

z T(w) M-measurable r e s p e c t i v e l y . Then E (f(w,«)) = a(w)E (b) ,

which i s M^,-measurable, and the f i r s t i n t e g r a l i n b) i s

T y g a E x ( b ) d P y

g a E y ( b o 0x|MT) d P y

g(w)a(w)b(e Tw)dP H(w) ,

where the second s t e p f o l l o w s from the f a c t t h a t g a i s M - m e a s u r a b l e .

- 160 -

Before concluding, let us emphasize that, in a slight departure

from the usual formulation of the translation operator version of the

strong Markov property, we are not assuming here that

M = a(Z t : 0 _< t <_ oo) mod P y . We must work in a more general setting

than this, since we want to apply this proposition to the process

(ft* ,B' ,8^,Bj.,eppX » A) considered i n the proof of 12.7(a).

Now we take up the proof of the measurability results used in

the proof of 12.7.

12.9. Lemma. Let X and Y be sets, let R and S be a-rings

of subsets of X and Y respectively, and l e t y be a measure on

5 such that each set belonging to S is of o-finite y-measure. Also.

let T = R ® 5 , and for each 5-measurable function f : Y H , let

[f] = {g : Y -*-]R| g is S-measurable and y({f 4 g}) = 0} .

Then the following are equivalent, for F : X ->• L 1 = L±(Y,S,y) .

a) range(F) is a separable subset of L X , and

and x F(x)dy is R-measurable for each S e S S

b) range(F) is a separable subset of L X , and

F X[U] n {F 4 0} e R for every norm-open set U £ L X .

c) There is a T-measurable function f : X x Y TR such

that F(x) = [f(x,«)l for a l l x e X .

Proof: a) => b). One f i r s t verifies that x ^ f(x) g dy is

R-measurable for each bounded S-measurable function g on Y .

- 161 -

Next, as range (F) i s s e p a r a b l e , t h e r e i s a s e t S e S o f a - f i n i t e

y-measure such t h a t

r a n g e ( F ) c {f e L 1 : f = 0 y - a.e. on Y\S} .

Thus f o r any bounded l i n e a r f u n c t i o n a l A on L"^ , t h e r e i s a bounded

S-measurable f u n c t i o n g on Y such t h a t

A ( F ( x ) ) F ( x ) g dy

x

f o r a l l x e X ; thus A ° F i s R-measurable. The c o n c l u s i o n b)

now f o l l o w s from the g e n e r a l t h e o r y of m e a s u r a b i l i t y f o r Banach

space v a l u e d f u n c t i o n s ; see P e t t i s [ 1 ] .

b) => c ) . U s i n g the s e p a r a b i l i t y o f range(F) , and the f a c t

t h a t F 1 [ B ] n { F ^ 0 } e R f o r each open b a l l B i n L 1 , we can

c o n s t r u c t a sequence (F_^) of c o u n t a b l e - v a l u e d R-measurable

f u n c t i o n s from X i n t o such t h a t f o r each i , and f o r each

i n X ,

||F(x) - F.(x) || ± < 2"1 . L

Now f o r each i , t h e r e i s a T-measurable r e a l - v a l u e d f u n c t i o n f .

on X x Y such t h a t

F.(x) = [ f . ( x , • ) ]

f o r a l l x e X . Moreover, f o r a l l x e X ,

f j [(x,») F(x) y - a.e.,

because of the r a t e o f L^-convergence o f (F^.(x)) to F(x) .

- 162 -

L e t E = {(x,y) e X x Y : ( f \ ( x , y ) ) does n o t converge i n E }

and d e f i n e f : X x Y -> E by

f i l m f . ( x , y ) i f (x,y) I E

f ( x , y ) = \

0 i f (x,y) e E

Then E e T and f has the d e s i r e d p r o p e r t i e s .

c) => a ) . The m e a s u r a b i l i t y p a r t f o l l o w s from F u b i n i ' s theorem.

L e t us e s t a b l i s h the s e p a r a b i l i t y p a r t . W e l l , t h e r e a r e c o u n t a b l e

s e t s RQ £ R and S^ £ S such t h a t f i s measurable w i t h r e s p e c t

to the o - r i n g g e n e r a t e d by the s e t s o f the form RxS where

R e R„ and S e Sn . L e t S., be the o - r i n g g e n e r a t e d by , 0 0 and l e t Z = u S . Then Z e S , so Z i s of a - f i n i t e u-measure,

and f o r each x e X , f(x,«) i s S^-measurable and v a n i s h e s o u t s i d e

Z . Thus range(F) £ " L ^ Z , S ^ y ) " , which i s s e p a r a b l e . •

Remark: The assumption t h a t each s e t i n S i s of a - f i n i t e y-measure

r e a l l y i s needed f o r the p r o o f o f (c => a) of the above lemma. T h i s

i s shown by the f o l l o w i n g example. L e t X = Y = [0,1] ,

R = S = B o r e l X , y = c o u n t i n g measure on S , and l e t

f ( x , y ) = {

1 i f x = y

0 i f x 4 y

Then f i s R ® S-measurable, but {[f(x,»)] : x e X} i s n o t a

- 163 -

s e p a r a b l e s u b s e t o f L"*" .

12.10. Lemma: L e t ( A , F , F t > P ) be a f i l t e r e d measure space, where

P i s f i n i t e . L e t RST = RST(A,F , F ,P) . Suppose F i s c o u n t a b l y

g e n e r a t e d mod P . Then t h e r e i s a map T •+ x o f RST i n t o i t s e l f

s uch t h a t

a) F o r e v e r y x , x = x P - a . e .

b) The map (x,to) H- x ( t o ) ( [ 0 , t ) ) i s ( B o r e l RST) ® F - m e a s u r a b l e

f o r each t e (0,°°) .

P r o o f : F o r each p o s i t i v e r a t i o n a l number r , l e t f : RST x A [0,1]

be d e f i n e d by f (x,co) = x ( w ) ( [ 0 , r ) ) .

Then f o r each x , f (x,«) i s F^_-measurable; a l s o , i f F e F ,

then

f (x,-)dP = F r

P(dco) x ( a 3 ) ( d t ) l [ 0 ) r ) ( t ) l F ( t o ) .

Thus f o r F e F , the map x f (x,«)dP i s B o r e l ( R S T ) - m e a s u r a b l e . F r

Indeed i t i s the p o i n t w i s e l i m i t o f a sequence of c o n t i n u o u s f u n c t i o n s

on RST. A l s o note t h a t L ^ ( A , F ,P | F r ) i s s e p a r a b l e s i n c e F i s

c o u n t a b l y g e n e r a t e d mod P .

Thus, by 12.9, f o r each r we can s e l e c t a map : RST x A [0,1]

such t h a t

i ) g i s ( B o r e l RST) ® F -measurable r r

i i ) F o r every x , g (x,«) = f r ( x , « ) P - a.e.

Fo r each t e (0,») , l e t h = sup g r . Then each h t i s a 0<r<t

- 164 -

( B o r e l RST) ® F - m e a s u r a b l e map of RST x A i n t o [0, 1] , and f o r

each (x,co) e RST * A , the map t H- h t(x , c o ) i s i n c r e a s i n g and

l e f t - c o n t i n u o u s . A l s o , f o r each T e (0,°°) , h (x,») = x(»)([0,t))

P - a.e.

Now f o r each x , l e t T be d e f i n e d by

x(co) = the unique p r o b a b i l i t y measure u on [0, 0 0] such t h a t

y ( [ 0 , t ) ) = h (x,co) f o r 0 < t < » .

Then c l e a r l y the map x •-»- x has the r e q u i r e d p r o p e r t i e s .

12.11. Lemma: L e t (A , F,P) be a f i n i t e measure space, and l e t

RRV = RRV(A ,F,P;[0, «]) . L e t (E , E ) be a measurable space, and l e t

(X V be a measurable p r o c e s s over (A , F ) w i t h s t a t e space t 0<t<°°

(E , E ) . F o r each x i n RRV , l e t y(x) be the measure on E

d e f i n e d by

y(x ) (A) = P(dco)x(co) ( { t e [0, » ] : X (u>) e A})

Then the map x y(x) i s measurable, i n the sense t h a t f o r each

A e E , u(»)(A) i s a Borel(RRV)-measurable f u n c t i o n on RRV .

P r o o f : L e t A e E , and l e t

H = {(co,t) e A x [0, o o ] : x (u>) e A} .

Then H e F ® ( B o r e l [ 0 , «=]) } and H(co) = {t e [0, o o ] : X (w) e A}

f o r each co e A . Note t h a t f o r each x e RRV , the map

co x(co)(H(co)) i s F-measurable, so the d e f i n i t i o n of y(x) i s

a l r i g h t .

- 165 -

L e t y be the s e t o f bounded measurable r e a l - v a l u e d p r o c e s s e s

(Y )_ over (A,F) such t h a t the map t 0<t«=°

T P(dco) x(o))(dt)Y t((_)

i s B o r e l ( R R V ) - m e a s u r a b l e . I f (Y ) has c o n t i n u o u s sample paths

then t h i s map i s a c t u a l l y c o n t i n u o u s by the d e f i n i t i o n o f the

t o p o l o g y o f RRV (see 9.21), so i n t h i s case 0^) e V . By a

monotone c l a s s argument, i t f o l l o w s t h a t V c o n s i s t s o f a l l

bounded measurable r e a l - v a l u e d p r o c e s s e s over (A,F) . In p a r t i c u l a r

1 e V . That i s , the map H

X H - P(dw)x(a.) (H(u.))

i s B o r e l ( R R V ) - m e a s u r a b l e .

12.12. Lemma: L e t (A,F,F ,P) be a f i l t e r e d measure space, where

P i s f i n i t e , ) i s r i g h t - c o n t i n u o u s , and F i s c o u n t a b l y

generated mod P . L e t RST = RST(A,F,F ,P) , and l e t T be a B o r e l

s u b s e t of RST . L e t (E,E) be any c o u n t a b l y generated measurable

space and l e t (X__)„ be a measurable p r o c e s s over (A,F) w i t h t 0<t<°°

s t a t e space (E , E ) . F o r each x e RST , l e t u(x) be as i n 12.11.

L e t (Z,A , a ) be a a - f i n i t e measure space and suppose z H - V ( Z ) i s

a measurable map from Z to f i n i t e measures on E . L e t

Z ^ = { z e Z : v ( z ) = u ( x ) f o r some x e T } . Then:

a) Z^ i s measurable w i t h r e s p e c t to the c o m p l e t i o n of a .

b) I f a(Z\Z.) = 0 , then t h e r e i s a f a m i l y (x ) „ of I z ze_

randomized ( F ^ - s t o p p i n g times such t h a t :

- 166 -

i ) x e T and u(x ) = v ( z ) f o r a - a.a. z e Z . z z

i i ) F o r each t e (0,°°) , the map

(z,w) H- T z ( a ) ) ( [ 0 , t ) )

i s A ® F^-measurable.

P r o o f : L e t M be the s e t of f i n i t e measures on E , and l e t M be

the s m a l l e s t a - f i e l d o f s u b s e t s o f M which makes the maps of the

form m H- m(A) (A e E) measurable. Then (M,M) i s a s e p a r a t e d

c o u n t a b l y g e n e r a t e d measurable space. The maps x >-> u(x) and

z H- v ( z ) a r e ( B o r e l RST, M)-measurable and (A,M)-measurable

r e s p e c t i v e l y . Now RST i s a compact p s e u d o m e t r i z a b l e space, by

9.21 and 9.23(a). Hence (RST, B o r e l RST) i s a B l a c k w e l l space i n

the sense of Meyer [ 1 ] . Hence M^ E {u(x) : x e T} b e l o n g s to

S o u s l i n M . T h e r e f o r e Z^ = {z e Z : v ( z ) e M^} b e l o n g s to

S o u s l i n A ; p a r t a) f o l l o w s from t h i s . Now assume a(Z\Z^) = 0 .

Now by the theorem on p. 251 o f D e l l a c h e r i e and Meyer [ 1 ] ,

t h e r e i s a f u n c t i o n ¥ : M^ -> T such t h a t :

¥ i s ( B o r e l f i e l d (Souslin(M|M )) , B o r e l RST)-measurable and

y(H'(m)) = m f o r a l l m e M^ .

L e t p z

=

¥(v(z)) i f z e Z

"0" i f z e Z\Z,

Then p e T and u(p ) = v ( z ) f o r a - a.a. z e Z , and the map z z

z H- p i s ( B o r e l f i e l d ( S o u s l i n A) , B o r e l RST)-measurable. In z

p a r t i c u l a r , z H- p i s (A, B o r e l RST)-measurable, where X i s

- 167

the c o m p l e t i o n o f A w i t h r e s p e c t t o a . Now B o r e l RST i s

c o u n t a b l y g e n e r a t e d , so f o r some e A w i t h a ( Z \ Z ^ ) = 0 ,

we have t h a t z »-»• p i s A-measurable on Z .

L e t a = { z

p i f z e Z A z U

'0" i f z e Z\Z 0

Then a = p f o r a - a.a. z € Z , and z a i s (A, B o r e l RST)-z z z

measurable. F i n a l l y , l e t x ^ x be as i n 12.10 and l e t x^ = 5 z

f o r a l l z e Z . Then c l e a r l y the f a m i l y (x ) „ has the d e s i r e d z zeZ

p r o p e r t i e s , •

12.13. Lemma: L e t y be a measure on H n such t h a t U y i s a

p o t e n t i a l . L e t RST = R S T ^ B . B ^ P ^ ) and l e t T = {x e RST : x i s

y-standard} .

Then T e B o r e l RST .

P r o o f : I f n >_ 3 then T = RST , and we a r e done. Suppose n = 1

or 2 . Then y i s f i n i t e . F o r each x e RST , l e t y_ be as i n

11.2. u y A . y X X A i T

Now x i s y - s t a n d a r d i f f U i s a p o t e n t i a l and U >_ U

f o r each n a t u r a l number i . T h i s i s e s s e n t i a l l y 11.12. (See 11.5 f o r the meaning o f x A i .)

Now U i s a p o t e n t i a l i f f |$(x)|dy (x) i s f i n i t e ,

x >1

y . y A l s o , U >_ U T i f f u T A 1 > U whenever V i s an open b a l l

- 168 -

whose r a d i u s i s r a t i o n a l and whose c e n t r e has r a t i o n a l c o o r d i n a t e s .

Thus the c o n d i t i o n t h a t x be u - s t a n d a r d can be e x p r e s s e d i n terms

o f c o u n t a b l y many measurable c o n d i t i o n s .

12.14. Lemma: L e t (A,F,P) be a a - f i n i t e measure space, where F

i s c o u n t a b l y generated mod P . L e t H be a compact m e t r i z a b l e

sp ace, and l e t RRV = RRV(A,F,P;H) . L e t

RV = {Y e RRV X = P - a.e. f o r some F-measurable f u n c t i o n

f : A -»• H}

Then:

a) F o r x e RRV > we have x e RV i f f whenever x'» x" e RRV

and x = \ X 1 + \ x" p " a- e- t h e n X = x' = x" P - a.e.

b). RV i s a G - s e t i n RRV o

P r o o f : a) (=>) i s c l e a r .

(<=) Suppose x e RRV\RV .

L e t U be a c o u n t a b l e open base f o r H . L e t A = {oo e A : x(w)

i s n ot a p o i n t mass} . Then A = {to e A : f o r some U e U ,

0 < x(w)(U) < 1} . Thus A e F . As x I RV , P(A) > 0 . But then

f o r some U e U , P(B) > 0 , where B = {u> e A : 0 < x(^) (U) < 1} .

D e f i n e a : A [0, 1] by a(w) = x(^)(U) • D e f i n e a, x e RRV by

a (GO) (E) = y(u>) (E\U) l-a(cj)

x(oo)(E) = y(oj)(EnU) a(oo)

a(u>) = x(u) = x(w)

} (UJ e B , E e B o r e l H)

(_ e A\B)

- 169 -

Then {a 4 x} = B , so a and x d i f f e r on a s e t of p o s i t i v e

P-measure. A l s o x ( w ) = [1 _ a(co)]a(co) + a(co)x(co) f o r a l l co e A .

Now d e f i n e x '» x" E R R V as f o l l o w s :

i f a (to) L \ > l e t

X' (co) = o(co)

X"(ai) = [1 - 2a(co)]o(co) + 2a(co)x(co) ;

i f a(co) >_ , l e t

X'(co) = [1 - 2(l-a(co))]x(co) + 2(1 - a(oi))a(ai)

x"(co) = x(co) .

Then (x* 4 x " l = B , so x ' and x " d i f f e r on a s e t o f p o s i t i v e

P-measure. A l s o , x = ^ x ' + "J x" •

b) By 9.21, RRV i s a compact p s e u d o - m e t r i z a b l e space; l e t K

be i t s m e t r i c i d e n t i f i c a t i o n , and l e t ij; be the c a n o n i c a l map o f

RRV onto K . F o r a , x e RRV , \p(o) = \J/(x) i f a = x P - a.e.

L e t D be the d i a g o n a l i n K x K . Then (K><K)\D i s

a-compact. D e f i n e iji : K x K -»• K by <()(x,y) =" y x + -| y " . Then

()> i s c o n t i n u o u s , and by b) , <|>[(KxK)\D] = K \ ^ [ R V ] . Thus K\\JJ[RV]

i s a-compact i n K . Hence RV = \p "*"[^[RV]] i s a G^-set i n RRV

12.15. At l a s t we can prove the m e a s u r a b i l i t y a s s e r t i o n s made i n

the p r o o f of 12.7.

Lemma: L e t (Z,A,a) be a a - f i n i t e measure space.

- 170 -

L e t D = 3Rn , E = . Suppose IT i s a measurable map o f Z o

i n t o D and g i s measurable map from Z to p r o b a b i l i t y measures

on E such t h a t f o r a - a.a. z e Z ,

U ^ Z ^ i s a p o t e n t i a l

and U 6 T r ( z ) > U 6 ( Z ) .

Then:

a) There i s a f a m i l y (x ) _ o f randomized (8 ) - s t o p p i n g z zeZ t

times such t h a t :

i ) f o r a - a.a. z e Z , x i s 6 . . - s t a n d a r d and z T T ( Z )

(« /_0_ = S(z) . TT (. Z ) X

Z

i i ) f o r each t e (0,°°) , the map (z,u>) ^ x (w)([0, t ) ) z

i s A ® B -measurable, t

b) I f n = 1 , t h e r e i s a f a m i l y (T ) of genuine (B ) -2 Z £ _ J t

s t o p p i n g times such t h a t :

i ) f o r a - a.a. z e Z , T i s 6 , ^ - s t a n d a r d and z T T ( Z )

TT (.z; T z

i i ) f o r each t e (0,°°) , the map (z,o.) 1^_ t ^

i s A ® 8^-measurable.

P r o o f : F i r s t l e t us e x p l a i n some o f the n o t a t i o n used i n the statement

o f the lemma. I f u i s a measure on B o r e l E , x i s a randomized

( B p - s t o p p i n g time, and T i s a genuine ( B p - s t o p p i n g time, then

u_ denotes the measure on B o r e l E d e f i n e d by

- 171 -

V T ( A ) = P y (dto)x(co)({t e [0, »] : B (a>) £ A})

w h i l e y^, denotes the measure on B o r e l E d e f i n e d by

y T ( A ) = P y ( B T £ A) .

Now f o r each z e Z , l e t v ( z ) be B(z) t r a n s l a t e d by

- T T ( Z ) , so t h a t

u v ( z ) < u 6

f o r a - a.a. z .

I f n >_ 2 , l e t T be the s e t o f 6 - s t a n d a r d randomized ( B p -

s t o p p i n g t i m e s . I f n = 1 , l e t T be the s e t o f 6 - s t a n d a r d

randomized (8 ) - s t o p p i n g times x such t h a t x(co) = 6 , . f o r t t (co;

P - a.a. co e ft , f o r some B-measurable f u n c t i o n f : ft •+ [0, °°] .

In e i t h e r c a s e , T i s a B o r e l s e t i n RST(ft,8,8 t,P ) . (Apply 12.13

i f n >_ 2 ; a p p l y 12.13 and 12.14 i f n = 1 .)

Now f o r a - a.a. z e Z , t h e r e e x i s t s x e T such t h a t

<5T = v ( z ) . (Apply 10.5 i f n >_ 3 , 11.14 i f n = 2 , o r 8.20 i f

n = 1 .) Thus by 12.12, t h e r e i s a f a m i l y (a ) such t h a t : z Z £ Z

i ) f o r a - a.a. z £ Z , a e T and 6 = v(z) . z a z

i i ) f o r each t £ (O, 0 0) , the map (z,co) o (co)([0,c°)) i s z

A ® B -measurable, t

F o r each z , l e t x be "the t r a n s l a t i o n o f a by + T T ( Z ) " . The z z

f a m i l y ( x ) then has the p r o p e r t i e s a s s e r t e d i n a ) . Now suppose z

n = 1 . F o r each z , l e t T be the (B ) - s t o p p i n g time d e f i n e d by z t

- 172 -

T (GJ) = i n f { t : T ( u ) ) ( [ 0 , t ) ) > 0} . 2 2

Then x (u ) = 6_ , f o r P Z - a.a. co e ft , by the c h o i c e o f T . z T (_)

z

A l s o {T < t} = {x (»)([0,t)) > 0} . C l e a r l y then, the f a m i l y z z

(T_) has the p r o p e r t i e s a s s e r t e d i n b) . ~ l

12.16. Theorem: L e t n be a p o s i t i v e i n t e g e r , and l e t ( A , F , F t > X t , P )

be an n - d i m e n s i o n a l p o t e n t i a l p r o c e s s , w i t h time s e t [0,°°) ,

and w i t h r i g h t - c o n t i n u o u s sample p a t h s . L e t u = law(X^) . Then

t h e r e i s an o p t i o n a l enlargement (ft' ,8' ,8|.,Q,i>) of ( f t , 8 , 8 t , P y ) and

an i n c r e a s i n g f a m i l y ( T t ) p < t < _ o f ( B ^ ) - s t o p p i n g times which a r e

st a n d a r d r e l a t i v e to Q , such t h a t :

a) F o r each w e f t ' , t H- T t (w) i s r i g h t - c o n t i n u o u s on

[0,-) .

b) The p r o c e s s e s ( x p and (B_j_ ) have the same f i n i t e

d i m e n s i o n a l j o i n t d i s t r i b u t i o n s , where B' = B ° ii f o r J s s

0 <_ s <_ »• .

P r o o f : L e t (ft",8",8J],R,d)) be the p r o d u c t enlargement o f

(f t , 8 , 8 t , P P ) by (L, L ,A) , where L = (0,1) , L = B o r e l L , and

A = Lebesgue measure on L . A l s o , l e t B' = B ° <j) f o r 0 <_ t <_ 00 .

Then by 12 . 7 ( a ) , f o r each n a t u r a l number k , t h e r e i s an i n c r e a s i n g

k f a m i l y (S ) ^ o f ( B ' p - s t o p p i n g times which a r e s t a n d a r d r e l a t i v e

k to R such t h a t the p r o c e s s e s (X ) and (B", ) have the

' t e l . c K k S t t e l k

-k same j o i n t d i s t r i b u t i o n , where 1^ = {j2 : j = 0 , 1, 2,...} . Now

173 -

d e f i n e X k , T k f o r a l l t e [0 , ») by

X? = X j 2

-k

T = T j 2

-k

} f o r ( j - l ) 2 ~ k < t < j 2 k , j = 1, 2, 3,

Then (X ) and (B ' \ )„ have the same f i n i t e d i m e n s i o n a l t 0<t<°° 0<t<°°

j o i n t d i s t r i b u t i o n s , and X -»• X^ as k -»• °° , f o r each t . t t

L e t H be the space of r i g h t - c o n t i n u o u s i n c r e a s i n g maps o f

[O, 0 0) i n t o [0, 0 0] , t o p o l o g i z e d as i n 9.24. A l s o , l e t H and

(H ) „ be as i n 9.24.

t 0<t<?°

L e t ft' = ft x H

B' = 8 9 H

( B p = ( ( B t 9 H f c ) + )

ijj = p r o j e c t i o n of ft' on ft

B ' = B o ib (0<t<°°) . t t Y

For each k , l e t T be the map of ft" i n t o ft' d e f i n e d by

r, (co) = ((f) (co) , T . ( c o ) ) .

Then r i s ( B " , B ' ) - m e a s u r a b l e , and a l s o ( B " , B ' ) - m e a s u r a b l e f o r k t t

each t . (co T^(co) i s (B'p H p -measurable f o r each t , and

(B") i s r i g h t - c o n t i n u o u s . ) Next, f o r each k , l e t Q, be the t K

- 174 -

measure on 8' d e f i n e d by Q F C(A) = R O ^ C A ) ) . Suppose f i s a

no n - n e g a t i v e 8'-measurable f u n c t i o n on ft' . Then f ° r i s

B'^-measurable, and for any A e 8 ,

f dQ ,-1

f ° r , dR - 1 - 1 k

ip [A] r k " [ A ] ]

f ° r , dR . -1 k

<j> X [ A ]

Thus E, ( f | , 8) = E ( f o r |<f>, 8) , which i s 8 -measurable mod P V .

Thus f o r each k , (ft' ,8' , B^,Q k , i j j ) i s an o p t i o n a l enlargement o f

(f t , B , 8 t , P y ) . F o r each t e [0, ») , d e f i n e T f c on ft' by

T t ( u , h ) = h ( t ) . Then each T f c i s a ( B p - s t o p p i n g time s a t i s f y i n g

T t o = T f c . F o r each k , the p r o c e s s (B^ ) has the same f i n i t e

d i m e n s i o n a l j o i n t d i s t r i b u t i o n s , r e l a t i v e t o Q K , as (X^) has

( r e l a t i v e t o P ) .

Each T i s s t a n d a r d r e l a t i v e to Q . A l s o , f o r each (co,h) e ft' ,

t H- T t(_),h) i s i n c r e a s i n g and r i g h t - c o n t i n u o u s on [0, °°) . By

9.27, t h e r e i s a subsequence ( \ ( ^ ) ) o f (\) a n d a p r o b a b i l i t y

measure Q on 8' such t h a t (ft',8',8^,Q,IJJ) i s an o p t i o n a l

f ( u ) ) ( h ) Q k ( £ ) ( d o j , d h ) enlargement o f (ft,B,B t,P K) and

f o r a l l f e L 1(ft,B,P y;C(H)) . We c l a i m t h a t the s t o p p i n g times

T f c a r e s t a n d a r d r e l a t i v e to Q and t h a t the p r o c e s s (B_ ) Q < t < a o

has the same f i n i t e d i m e n s i o n a l j o i n t d i s t r i b u t i o n s r e l a t i v e t o Q

as (X ) . has ( r e l a t i v e to P ) . T h i s i s somewhat e a s i e r to t 0<t<°°

f (u) (h)Q(doo,dh)

- 175 -

prove when n 21 3 , so l e t us c o n s i d e r t h i s case f i r s t . I n t h i s c a s e,

e v e r y s t o p p i n g time i s s t a n d a r d , and we need o n l y check t h a t the

j o i n t d i s t r i b u t i o n s a r e r i g h t .

L e t D = ]R n , E = , l e t 0 <_ t < ... < t . <°° , and l e t 0 1 J

g be a c o n t i n u o u s f u n c t i o n on E 3 which v a n i s h e s o u t s i d e a compact

s u b s e t o f D 3 . Then

g(B' ,...,B' )dQ fcl fcj

(*)

l i m e+0

1 e

l i m l i m e4-0 I-**

re

g ( B l 0 t±+s

1 e

re

,B; )ds dQ t.+s 3

g ( B l ,...,B'^ Jds d Q k W

j 0 t±+s

l i m l i m e+0 £->~

1 e g(B" k(£) '

t 1 + s

. . B « ( i ) ) d . dR

t.+s 3

l i m l i m e+0 ; 0 £ i + s

l i m e+0

g(X ,...,X )ds dP 0 V S 3

g(X ,...,X )dP 1 3

T h i s s u f f i c e s t o show t h a t (B^, ) Q < t < O 0 h a s t n e r i g h t f i n i t e

- 176 -

d i m e n s i o n a l j o i n t d i s t r i b u t i o n s . However, the s t e p (*) r e q u i r e s

f u r t h e r e x p l a n a t i o n . (We remark t h a t a l l the o t h e r s t e p s i n the

above c a l c u l a t i o n h o l d f o r any n , but the method o f v e r i f i c a t i o n

of the s t e p (*) depends on n , and i s s i m p l e s t when n _> 3 .)

W e l l , g(B' ( o ) , h ) , ...,B' (_,h))ds

0 l i + s l 2 + S

re

Q g ( B h ( t 1 + s ) ( w ) ' - - " B h ( t . + s ) ( a , ) ) d s

Now i f h -»• h i n H then h (t.+s) •+ h(t.+s) f o r i = l , . . . , j , f o r m m i l

a l l b u t c o u n t a b l y many s (see 9.24). Now ( B ^ depends c o n t i n u o u s l y

on t , except a t t = 0 0 ; but g i s s u p p o r t e d by a compact s u b s e t

o f D J and, as n _ _ 3 , | | B t | | ->- °° P P - a .s. as t -»- 0 0 .

Thus f o r P y - a.a. w e f t , the map

re

Q 8 ( B h ( t 1 + s ) ( w ) ' - - " B h ( t j + s ) ( w ) ) d s

i s c o n t i n u o u s on H . T h i s j u s t i f i e s the s t e p (*) i n the case n >_ 3

Now suppose n = 1 o r 2 .

L e t D, E, t ^ , . . . , t j , and g be as b e f o r e . F i x e e [0, °°) .

D e f i n e f on ft x H by

re f ( o i , h ) =

Q g ( B h ( t 1 + s ) ( t ° ) ' " - ' B h ( t . + s ) ( w ) ) d s

and f o r each a e [0, °°) d e f i n e f on ft x H by cL

re f ( 0),h)

0

8 ( B h ( t 1 + s ) A a ( w ) ' - - - ' B h ( t j + s ) A a ( w ) ) d S

- 177 -

Then f , f (0<a<°°) a r e u n i f o r m l y bounded, each f i s c o n t i n u o u s

i n i t s second v a r i a b l e , and f = f on ft x H , where a a

H = {h e H : h < a on [0, t . + e)} . Now u s i n g 8.9 and a s l i g h t a — J

g e n e r a l i z a t i o n o f 8.10, one f i n d s t h a t

l i m sup Q (f2'\(fixH )) = 0 . a-*>° k

But i f h h i n H and each h i s bounded by a on [0, t . + e)

then h ( t ) <_ a f o r each c o n t i n u i t y p o i n t t o f h i n [0, t ^ + e) ,

whence h < a on [0, t . + e) . Thus each H i s c l o s e d i n H . - 3 a

Hence Q(ftxH ) _> l i m sup Q . (fixH ) . I t f o l l o w s t h a t a i + oo k W a

l i m Q(S7*\ftxH ) = 0 . a

But f dQ - f dQ k(£)

f - f dQ a

+ f dQ -a f a d\(i)

+ f a " f d C W )

Now the f i r s t term on the r i g h t hand s i d e h e r e can be made

s m a l l by t a k i n g a l a r g e , the t h i r d term can be made s m a l l u n i f o r m l y

i n I by t a k i n g a l a r g e , w h i l e the second term goes to 0 as

I •+ 0 0 , f o r each a . Thus f dQ . T h i s s u p p l i e s the

j u s t i f i c a t i o n o f the st e p (*) above i n the case n <_ 2 , so we now

- 178 -

know t h a t (B^ ) has the r i g h t f i n i t e d i m e n s i o n a l j o i n t d i s t r i b u t i o n s t

r e l a t i v e to Q i n a l l c a s e s . I t remains to show t h a t the s t o p p i n g

times T a r e s t a n d a r d r e l a t i v e to Q when n = 1 o r 2 . L e t t

t e [0, 0 0) . L e t o = law(X . Then a i s a p r o b a b i l i t y t+1

measure on E and a(D) = 1 . Now

re law(B' ;Q) = l i m -

Lt e4-0 law(B' ;Q)ds

0 t+s

= l i m l i m — e+0 £-*=° £ o l a " ( B i t + 6

; Q M ^ ) ) d s •

where the second s t e p f o l l o w s from the f a c t t h a t

law(B' ;Q , %) law(B' ;Q) f o r each I , as was shown above, T t + s k W T t + s

(The l i m i t s a r e w i t h r e s p e c t to the vague t o p o l o g y . ) Now c o n s i d e r

any a e [0, °°) .

Then l a w ( B l ;Q) = l i m -T t a e+0 e

re law(B' A a ; Q ) d s

0 t+s

l i m l i m e+0 l-**>

l a w ( B T t + s A a ; Q k ( £ ) ) d S '

where i n t h i s c a s e , the second s t e p f o l l o w s from the f a c t t h a t i f

f i s a bounded c o n t i n u o u s f u n c t i o n on E then the map

(o),h) f(B, , , s ( u ) ) d s h(t+s)Aa

i s c o n t i n u o u s i n h

Now f o r 0 <_ s <_ — and k(£) >_ 1 ,

- 179 -

U

law(B t+s

;Q. ) > u

a

A p p l y i n g 8.7 and 8.8, one f i n d s t h a t

law(Bj, A a;Q) law(B^, ;Q)

U > U

Then, by 11.12, T i s s t a n d a r d r e l a t i v e t o Q •

I n the case n = 1 , the above theorem j u s t says t h a t any r i g h t -

c o n t i n u o u s m a r t i n g a l e can be embedded i n an o p t i o n a l enlargement of

Brownian motion by means of a r i g h t - c o n t i n u o u s i n c r e a s i n g f a m i l y o f

s t a n d a r d s t o p p i n g times — a r e s u l t due to Monroe [ 1 ] . (We remark

t h a t Monroe works w i t h minimal s t o p p i n g times, r a t h e r than s t a n d a r d

ones. I t i s easy to show t h a t a s t a n d a r d s t o p p i n g time i s m i n i m a l .

Monroe i n e f f e c t p r o v e s t h a t the converse i s t r u e when n = 1 ,

though he does not e x p l i c i t l y d e f i n e s t a n d a r d s t o p p i n g t i m e s . T h i s

i s an i n t e r e s t i n g r e s u l t i n i t s own r i g h t , b u t f o r the purpose of

embedding, i t i s perhaps e a s i e r t o work w i t h the d e f i n i t i o n o f

s t a n d a r d s t o p p i n g times than w i t h the d e f i n i t i o n of minimal s t o p p i n g

times.)

12.17. C o r o l l a r y : L e t I = {0, -1, -2,...} and l e t (A,F,F ±,X ,P)

be an n - d i m e n s i o n a l p o t e n t i a l p r o c e s s w i t h time s e t I . Then

t h e r e i s a p r o b a b i l i t y measure u on ] R n such t h a t U y i s a 0

p o t e n t i a l (and p({9}) = 0 i f n = 1 o r 2 ) , and an o p t i o n a l enlargement

- 180 -

( f t ' , B ' , B ' , Q , i f O of ( f t , B , B t , P U ) , such t h a t t h e r e a r e BJ.-stopping

txmes

T 0 ^ - T - l - T - 2 ^

which a r e s t a n d a r d r e l a t i v e to Q , such t h a t the p r o c e s s e s (X^)

and (B^ ) have the same j o i n t d i s t r i b u t i o n (where B| = • f i

f o r 0 <_ t <_ °° ) .

P r o o f : There a r e i n d i c e s i ^ > i ^ > i ^ > .•• i n I such t h a t

f ( x ) l a w ( X . ) ( d x ) 1 j

f ( x ) y ( d x )

f o r e v e r y compactly s u p p o r t e d c o n t i n u o u s f u n c t i o n f on E = 3Rg ,

where y i s some measure on E s a t i s f y i n g y(E) <_ 1 .

law(X.) law(X ) Now U 1 i u f o r e v e r y i . Thus, by 8.7 and

8.8 ( i f n = 1 o r 2) , U y i s a p o t e n t i a l and

U

law(X ) 2 dy -> U y dy

f o r e v e r y good measure y on D = ]R n . ( I f n _> 3 i t i s t r i v i a l

to check t h a t t h i s convergence h o l d s . )

law(X.) law(X ) But U 1 i n c r e a s e s as i d e c r e a s e s . Thus U + U

as i -> -°° . (Hence we a l s o have t h a t law(Xj ->• y v a g u e l y as

i -»• -°° .)

I f n = 1 o r 2 then, by 8.9,

l i m sup law(X.) ({x e D : | |x| | >. r}) = 0 sup r-x» i

- 181 -

In t h i s case we a l s o have law(X^)({9}) = 0 f o r a l l i .

Thus i f n = 1 o r 2 then y i s a p r o b a b i l i t y measure and law(X.)

y({9}) = 0 . On the o t h e r hand, i f n j> 3 then from U y >_ U 1

we can deduce t h a t u(D) _> law(X i>(D) (see p r o o f of 7.8); a l s o ,

as 9 i s an i s o l a t e d p o i n t o f E , l a w ( X i ) ( { 9 } ) •*• y({3}) as

!->•-«>. Thus i n t h i s case too, y i s a p r o b a b i l i t y measure.

Now f o r each y e D , ($(X^,y)) i s a s u p e r m a r t i n g a l e o v e r

(A , F , F.,P) , and sup E ( $ ( X . , y ) ) = U P ( y ) . Thus by V, T21 o f Meyer [1] 1 i 1

f o r each y e D such t h a t U y ( y ) < » t $(X^,y) converges almost s u r e l y and i n L X , as !->•--, to an i n t e g r a b l e f u n c t i o n .

But from t h i s i t f o l l o w s t h a t t h e r e e x i s t s X : A ->• E such t h a t —GO

X i s F -measurable, where F = nF. , and X. -*• X P - a.s. —oo —oo —oo ^ 2. i — 0 0

as 1 -> -« . C l e a r l y law(X ) = y and ( A , F , ( F . ) (x ) P)

i s a p o t e n t i a l p r o c e s s , where J = Iu{-°°} • Now l e t

G

0

= F - o o . Y

0

= X - »

and l e t

G = F Y = Xn f o r 1 < t < 0 0

t 0 t 0 -

G t - 1=_v Y t = X_ x f o r | < t < 1 ,

and so on.

Then ( A , F , G t , Y t » P ) i s a p o t e n t i a l p r o c e s s , w i t h time s e t

[0, 0 0) , w i t h r i g h t - c o n t i n u o u s sample p a t h s , and so can be embedded

i n an o p t i o n a l enlargement o f Brownian motion w i t h i n i t i a l d i s t r i ­

b u t i o n law(Yg) = y , by means of a r i g h t - c o n t i n u o u s i n c r e a s i n g

f a m i l y o f s t a n d a r d s t o p p i n g times, by 12.16. I t i s now c l e a r how to

- 182 -

complete the p r o o f . < g

I n the above c o r o l l a r y , i t i s n e c e s s a r y t o c o n s i d e r an e n l a r g e ­

ment o f Brownian motion, even i n the one d i m e n s i o n a l c a s e . T h i s i s

shown by the f o l l o w i n g example, which was i n v e n t e d by R. V. Chacon.

12.18. Example: L e t y be the u n i t p o i n t mass a t 0 i n ]R n ,

and l e t (ft' , B ',8j.,Q,40 be the p r o d u c t enlargement of ( f t , B , B T , P M )

by ( L , L , A ) , where L = {0, 1} , L = power s e t o f L , and A i s o

t h e measure on L which a s s i g n s mass to {0} and to {1} .

L e t B' = B ° ij; f o r 0 < t < » . For i = 0, 1, 2,... d e f i n e t t — —

T! on ft' by

T:(OJ,0) = i n f { t > 0 : ||B.(a>)|| = 2 1" 1} x — t

T! (W,1) = i n f { t > 0 : | |B. (u>) | | = 2 1 } x — t

Then each T^ i s a ( B ^ ) - s t o p p i n g time which i s s t a n d a r d r e l a t i v e t o

and _> T 1 _> • • • .

Thus ( f t ' , B ' , B ^ , , B ^ , , Q ) i s a p o t e n t i a l p r o c e s s w i t h time s e t i i

{0, -1, -2, ...} , by 12.6.

Suppose TQ T ^ >_ ... a r e y - s t a n d a r d ( B ^ - s t o p p i n g times

such t h a t ( B T ) and (B^,) have the same j o i n t d i s t r i b u t i o n , i i

We s h a l l deduce a c o n t r a d i c t i o n . W e l l by 8.13, as T± i s y-y.

( U y - U 1 ) , where y.. = l a w ( B T ) = law(B^,) s t a n d a r d , E^(T.) x x

- 183 -

But law(B^,,) = j a±-i + \ °± w n e r e f o r e a c h J > °j i s t h e

u n i f o r m u n i t d i s t r i b u t i o n on the sphere of r a d i u s 2~* c e n t r e d a t 0

I t f o l l o w s t h a t E M ( T . ) -> 0 as i -»•-<» . Hence T. 4- 0 P P - a.s. ^ l i

. . As (8y) i t

as i -> -°° . L e t T = l i m T_. . As (8y) i s r i g h t - c o n t i n u o u s ,

8y = n8y . Now B y = {E e 8y : E n {T £ t} e B y f o r 0 £ t < »} 1 i

s i n c e P y(T=0) = 1 and s i n c e e v e r y s e t of^ P y-measure zero b e l o n g s

to 8y . But s i n c e u i s a p o i n t mass, 8y = {0, ft} mod P y .

Now f o r each i , l e t E = {||B || = 2 1} . Then P y ( E ± ) = j . i

But we a l s o have P y ( E . n E. ,) I x-1

= Q ( | | B l , | | = 2 1 , ||B', || = 2 1 " 1 ) i i - 1

= Q(ftx{l}) = | .

Thus, mod P y , E = E = E = ••• . L e t E = n u E. . 0 -1 -2 i = 0 j = i l

Then E e n B y = 8y , b u t P y ( E ) = ~ , which c o n t r a d i c t s the i i

t r i v i a l i t y of B y .

- 184 -

13. APPENDIX OF MISCELLANEOUS NOTATION AND TERMINOLOGY

13.1. We take p o s i t i v e to mean s t r i c t l y p o s i t i v e (though by a

p o s i t i v e measure we r e a l l y j u s t mean a n o n - n e g a t i v e one; a l s o , by

i n c r e a s i n g we j u s t mean n o n - d e c r e a s i n g ) . A , rea d meet, denotes

g r e a t e s t lower bound, w h i l e v , rea d j o i n , denotes l e a s t upper

bound. I f f i s any [-00,°°]-valued f u n c t i o n then f + denotes

fvO and f ~ denotes (-f)vO .

13.2. ]R denotes the s e t of r e a l numbers.

]N denotes the s e t {0,1,2,...} o f n a t u r a l numbers.

Q denotes the s e t of r a t i o n a l numbers.

13.3. I f x = ( x l 5 . . . , x ) e ]R n then l l x l l denotes the u s u a l 1 n

e u c l i d e a n norm of x :

2 2 1 / 2

||x|| = ( x j + ...+ x 2 ) .

13.4. We use the symbol A i n two ways; t o denote the L a p l a c i a n :

n g2 A = £ — j , and to denote symmetric d i f f e r e n c e o f s e t s :

1=1 3x. l

AAB = (A\B)u(B\A) .

13.5. L e t X be a t o p o l o g i c a l space. F o r E c X , i n t ( E ) denotes

the i n t e r i o r of E , E denotes the c l o s u r e o f E , and 8E denotes

the f r o n t i e r o f E (3E = E \ i n t (E)) . I f f i s a [-°°, °°]-valued

f u n c t i o n on X then we s h a l l say f i s c o n t i n u o u s i f f f ''"[U] i s

open i n X f o r each U open i n [-°°, 0 0] ; the p o i n t o f t h i s

remark i s to emphasize t h a t when we say " f i s c o n t i n u o u s " , we

- 185 -

are not s u g g e s t i n g t h a t i t assumes o n l y f i n i t e v a l u e s . We s h a l l

n o t use the e x p r e s s i o n " c o n t i n u o u s i n the extended s e n s e " . I f Y

i s a n o t h e r t o p o l o g i c a l s p a c e , then C(X,Y) denotes the s e t of

c o n t i n u o u s f u n c t i o n s from X i n t o Y ; C(X) denotes C(X^R) .

B o r e l X denotes the o - f i e l d g e n e r a t e d by the c l o s e d s u b s e t s of

X . I f u i s a [-°°, °°]-valued f u n c t i o n on X then the lower

r e g u l a r i z a t i o n of u , denoted by u , i s the l a r g e s t lower-

s e m i c o n t i n u o u s f u n c t i o n v on X s a t i s f y i n g v <_ u ; a l a r g e s t

such f u n c t i o n always e x i s t s s i n c e the supremum of any c o l l e c t i o n of

l o w e r - s e m i c o n t i n u o u s f u n c t i o n s i s a g a i n l o w e r - s e m i c o n t i n u o u s . F o r

each x e X , we have

u(x) = u(x) i f x i s an i s o l a t e d p o i n t o f X

l i m i n f u(y) i f x i s a l i m i t p o i n t of X y x

We remark t h a t l i m i n f u(y) denotes sup i n f u(y) , where y -y x Vet/_ yeV\{x}

x

f i s the c o l l e c t i o n of neighbourhoods of x x

13.6. When we speak o f a measure, w i t h o u t a q u a l i f i e r such as

" s i g n e d " or " r e a l - v a l u e d " , we s h a l l always mean a p o s i t i v e measure.

An o u t e r measure on a s e t X i s a [0, <=°]-valued f u n c t i o n y on

the power s e t of X such t h a t whenever S i s a c o u n t a b l e c o l l e c t i o n

o f s u b s e t s of X and R c u S we have y (R) < \ y (S) ; (In SeS

p a r t i c u l a r , t a k i n g R = 0 and S = 0 , we f i n d t h a t y(0) = 0) .

I f y i s an o u t e r measure on X , then M denotes the a - f i e l d y

of y-measurable s u b s e t s of X :

- 186 -

M = {R c X : f o r any S c X , y(S) = y (SnR) + y(S\R} . y

I f H i s a c o l l e c t i o n of s u b s e t s o f X then an tf-outer

measure on X i s an o u t e r measure y on X such t h a t H c M and ~~ y

f o r e v e r y R c X , y(R) = i n f { y ( H ) : R c H e H} . I f X i s a

t o p o l o g i c a l space then by a measure on (or i n ) X , we mean a

measure y on B o r e l X such t h a t y(K) i s f i n i t e f o r each c l o s e d

compact s e t K c x . I f (X ,A) i s a measurable space and y i s

a measure on A and A e A then y. denotes the measure on A A

d e f i n e d by y.(B) = y(BnA) .

13.7. We have made some use of the th e o r y of S o u s l i n s e t s and of

a n a l y t i c s e t s . We r e f e r the r e a d e r to Meyer [ 1 ] , D e l l a c h e r i e and

Meyer [ 1 ] , and B r e s s l e r and S i o n [1] f o r e x p o s i t i o n s o f t h i s t h e o r y .

13.8. I f D i s a Green r e g i o n then the symbol G^ , used w i t h o u t

e x p l a n a t i o n , denotes the Green f u n c t i o n of D .

13.9. Suppose V i s an open b a l l i n l R n and f i s a [-00 , 0 0 ] - v a l u e d

f u n c t i o n on 3V . Then P I ( f ; V ) denotes the f u n c t i o n d e f i n e d by

P I ( f ; V ) ( x ) = 1

a r n

f 2 2 r ~ I i x " p | 1 f (z) d a ( z )

3V l | x - z l | n

f o r a l l x e V f o r which the i n t e g r a l makes sense, where p i s

the c e n t r e o f V , r i s the r a d i u s of V , a i s the " s u r f a c e a r e a "

measure on 3V , and a = the " s u r f a c e a r e a " o f the u n i t sphere i n n

]R n . Note PI sta n d s f o r " P o i s s o n i n t e g r a l " . I f f i s i n t e g r a b l e

w i t h r e s p e c t t o a over 3V , then P I ( f ; V ) i s harmonic i n V and

- 187 -

and l i m P I ( f ; V ) ( x ) = f ( z ) f o r each z € 8V a t which f i s x->z

c o n t i n u o u s — see 2.4 and 2.7 of Helms [ 1 ] .

13.10. We f o l l o w the c o n v e n t i o n t h a t 0 times 0 0 e q u a l s 0 . In

a d d i t i o n , we adopt the c o n v e n t i o n t h a t 0 times u n d e f i n e d e q u a l s 0 ;

see 6.1 f o r an example o f t h i s .

13.11. Suppose (A ,F) and (E ,E) a r e measurable s p a c e s , P i s a

measure on F , and : A -»• E i s (F ,E)-measurable. Then ^(P) ,

law(i(j) , and law(^;P) a l l denote the measure u on E d e f i n e d

by y (A) = P (ip X [ A ] ) ; we use whic h e v e r n o t a t i o n i s most s u g g e s t i v e

and l e a s t ambiguous i n a g i v e n s i t u a t i o n . ' I f g i s a [-°°, °°]-

v a l u e d F-measurable f u n c t i o n on A , then E(g|i | ; ,E) denotes any

E-measurable f u n c t i o n h on E such t h a t

h(x)lawGjj) (dx) = A

g dP

f o r a l l A e E ; i f law Op) i s a - f i n i t e and g i s e i t h e r n o n - n e g a t i v e

or i n t e g r a b l e , such a f u n c t i o n h e x i s t s , and i s unique mod lawGjO .

F i n a l l y , a d i s i n t e g r a t i o n of P w i t h r e s p e c t t o , E i s a f a m i l y

(P ) of p r o b a b i l i t y measures on F such t h a t f o r each A e E x xeE

and each F e F ,

x B- P (F) is E-measurable, and x

P ( F n ^ 1 [ A ] ) = P (F)law(i | 0(dx) . J A X

I f lawdlO is a-finite, F is countably generated, and P is

- 188 -

i n n e r r e g u l a r w i t h r e s p e c t t o a semicompact c l a s s , then a d i s i n t e ­

g r a t i o n o f P w i t h r e s p e c t t o \\i, E e x i s t s . (See a l s o 9.4.)

- 189 -

REFERENCES

R. B a x t e r and R. V. Chacon

1. P o t e n t i a l s of Stopped D i s t r i b u t i o n s . 111. J . Math. 18 (1974) 649-656.

2. Compactness o f S t o p p i n g Times, Z. W a h r s c h e i n l i c h k e i t s t h e o r i e verw. G e b i e t e , 40 (1977) 169-181.

3. Enlargement of o - a l g e b r a s and Compactness of Time Changes, Can. J . Math., 29_ (1977) 1055-1065.

M. Blumenthal and R. K. Getoor

1. Markov P r o c e s s e s and P o t e n t i a l Theory, Academic P r e s s , 1968.

B r e l o t

1. M i n o r a n t e s Sous-harmoniques, Extremales e t C a p a c i t e s , J . de Math. Pure e t App., IX, 24. (1945) 1-32.

2. Elements de l a T h e o r i e C l a s s i q u e du P o t e n t i e l , 3 e d i t i o n (1965), C e n t r e du Documentation U n i v e r s i t a i r e , 5 P l a c e de l a Sorbonne, P a r i s .

W. B r e s s l e r and M. S i o n

1. C u r r e n t Theory of A n a l y t i c S e t s , Can. J . Math., 16. (1964) 207-230.

V. Chacon

1. P o t e n t i a l P r o c e s s e s , T r a n s . Amer. Math. S o c , 226 (1977) 39-58.

V. Chacon and J . B. Walsh

1. One D i m e n s i o n a l P o t e n t i a l Embedding, U n i v e r s i t e de S t r a s b o u r g , S e m i n a i r e de P r o b a b i l i t e s X, 19 74/75, S p r i n g e r - V e r l a g , L e c t u r e Notes i n Mathematics, no. 511.

- 190 -

C. D e l l a c h e r i e and P. A. Meyer

1. P r o b a b i l i t i e s e t p o t e n t i e l , e d i t i o n e n t i e r e m e n t r e f o n d u e , Hermann, 19 75.

J . L. Doob

1. Se m i m a r t i n g a l e s and Subharmonic F u n c t i o n s , T r a n s . Amer. Math. S o c , 77 (1954) 86-121.

L. E. Dub i n s

1. On a Theorem of Skorohod, Ann. Math. S t a t . 39. (1968) 2094-2097.

N. Du P l e s s i s

1. An I n t r o d u c t i o n to P o t e n t i a l Theory, O l i v e r and Boyd, Edi n b u r g h , 1970.

L. L. Helms

1. I n t r o d u c t i o n to P o t e n t i a l Theory, John W i l e y and Sons, Inc . , 1969.

G. A. Hunt

1. M a r k o f f P r o c e s s e s and P o t e n t i a l s I , ' 44-93.

2. M a r k o f f P r o c e s s e s and P o t e n t i a l s I I , 316-369 •

3. M a r k o f f P r o c e s s e s and P o t e n t i a l s I I I 151-213.

A. I o n e s c u F u l c e a and C. Ionescu F u l c e a

1. T o p i c s i n the Theory of L i f t i n g , S p r i n g e r - V e r l a g , 1969.

P. A. Meyer

1. P r o b a b i l i t y and P o t e n t i a l s , B l a i s d e l l , 1966.

2. P r o c e s s u s de Markov, S p r i n g e r - V e r l a g , L e c t u r e Notes i n Mathematics, no. 26, 1967.

- 191 -

3. Sur un A r t i c l e de Dubins, U n i v e r s i t e de S t r a s b o u r g , Seminaire de P r o b a b i l i t e s V, 1969/70, S p r i n g e r - V e r l a g , L e c t u r e Notes i n Mathematics, no. 191.

4. Convergence F a i b l e e t Compacite des Temps d ' A r r e t , d'apres B a x t e r e t Chacon, U n i v e r s i t e de S t r a s b o u r g , S e m i n a i r e de P r o b a b i l i t e s X I I , 1976/77, S p r i n g e r - V e r l a g , L e c t u r e Notes i n Mathematics, no. 649.

I. Monroe

1. On Embedding R i g h t Continuous M a r t i n g a l e s i n Brownian M o t i o n , Ann. Math. S t a t . , 43 (1972) 1293-1311.

J . K. P a c h l

1. D i s i n t e g r a t i o n and Compact Measures, to appear i n Math. Scand.

B. J . P e t t i s

1. On I n t e g r a t i o n i n V e c t o r Spaces, T r a n s . Amer. Math. S o c , 44 (1938) 277-304.

M. Rao

1. Brownian M o t i o n and C l a s s i c a l P o t e n t i a l Theory, Feb., 1977, Aarhus U n i v e r s i t y L e c t u r e Notes S e r i e s , no. 47.

D. H. Root

1. The E x i s t e n c e o f C e r t a i n S t o p p i n g Times of Brownian M o t i o n , Ann. Math. S t a t . 40 (1969) 715-718.

H. R o s t

1. D i e S t o p p v e r t e i l u n g e n e i n e s M a r k o f f - P r o z e s s e s m i t L o k a l e n d l i c h e m P o t e n t i a l , M a n u s c r i p t a Math., 3 (1970) 321-330.

2. The S t o p p i n g D i s t r i b u t i o n s of a Markov P r o c e s s , I n v e n t . Math., 14 (1971) 1-16.

A. B. Skorohod

1. S t u d i e s i n the Theory o f Random P r o c e s s e s , Addison-Wesley, 1965.

- 192 -

INDEX OF SELECTED NOTATION AND TERMINOLOGY

a n a l y t i c s e t s 13.7

B 5.4 t

8 ° , B°, B\ B j , 8, B t , °B , J B 5.4

b a l a y a g e , b a l ( , , ) 1.11, 2.

base( , ) 3.1

bas e ( ) 3.10

D 5.4

A

d i s i n t e g r a t i o n 13.11

E( | , ) 13.11

enlargement 9.7, 9.12

f i l t e r e d measurable space 9.11

f i l t e r e d measure space 9.11

f i n e t o p o l o g y 3.12

f r i n g e ( , ) 3.1

f r i n g e ( ) 3.10

G D 13.8

Gy 1.10

g e n e r a l i z e d Brownian motion p r o c e s s 5.5

good measure 8.6

Green f u n c t i o n 1.9

Green p o t e n t i a l 1.10

Green r e g i o n 1.9

law( ) , law( , ) 13.11

lower r e g u l a r i z a t i o n 13.5

- 193 -

measure 13.6

o p t i o n a l enlargement 9.12

o u t e r measure 13.6

p 5.1

P 5.3 t

P y , P X 5.4

P I ( ; ) 13.9

p o l a r s e t 1.9

p o t e n t i a l 1.4, 1.10

p o t e n t i a l p r o c e s s 12.2

p r o d u c t enlargement 9.17

randomized random v a r i a b l e , r r v 9.1

randomized s t o p p i n g time, r s t 9.2

r e d u i t e , r e d ( , , ) 1.11

R i e s z measure 1.2

RRV( , , ; ) 9.21

RST( , , , ) 9.23

S o u s l i n s e t s 13.7

s t a n d a r d randomized s t o p p i n g time 11.3

s t a n d a r d s t o p p i n g time 8.2, 12.4

T 5.4 A

t h i n n e s s 3.1, 3.10

U y, uj, U_ 1.4

U y 12.1

- 194 -

u n i v e r s a l l y measurable space 9.5

6, 6 1.1

t 5.4

u A 13.6

$ 1.3

Q 5.4

( f t , B , B t , B t , e t , P x ) 5.4

5.2