66
37? N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate Council of the University of North Texas in Partial Fulfillment of the Requirements For the Degree of DOCTOR OF PHILOSOPHY By Catherine Ann Abbott, B.S., M.A, Denton, Texas August, 1988

N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

  • Upload
    others

  • View
    8

  • Download
    0

Embed Size (px)

Citation preview

Page 1: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

3 7 ?

N81J /Vc? ia&SS'S

OPERATORS ON CONTINUOUS FUNCTION SPACES

AND WEAK PRECOMPACTNESS

DISSERTATION

Presented to the Graduate Council of the

University of North Texas in Partial

Fulfillment of the Requirements

For the Degree of

DOCTOR OF PHILOSOPHY

By

Catherine Ann Abbott, B.S., M.A,

Denton, Texas

August, 1988

Page 2: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

Q-i.

Abbott, Catherine Ann, Operators on Continuous Function

Spaces and Weak Precompactness Doctor of Philosophy

(Mathematics), August, 1988, 62 pp., bibliography, 36 titles.

If T:C(H,X)-+Y is a bounded linear operator then there

exists a unique weakly regular finitely additive set function

m:5fr(H)-«(X,Y**) so that

T(f) - J H f d m .

In this paper, bounded linear operators on C(H,X) are studied

in terms the measure given by this representation theorem.

The first chapter provides a brief history of representation

theorems of these classes of operators. In the second

chapter the represenation theorem used in the remainder of

the paper is presented.

If T is a weakly compact operator on C(H,X) with

representing measure m, then m(A) is a weakly compact

operator for every Borel set A. Furthermore, m is strongly

bounded. Analogous statements may be made for many

interesting classes of operators. In chapter III, two

classes of operators, weakly precompact and QSP, are studied.

Examples are provided to show that if T is weakly precompact

(QSP) then m(A) need not be weakly precompact (QSP), for

every Borel set A. In addition, it will be shown that weakly

precompact and GSP operators need not have strongly bounded

representing measures. Sufficient conditions are provided

Page 3: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

which guarantee that a weakly precompact (QSP) operator has

weakly precompact (QSP) values. A sufficient condition for a

weakly precomact operator to be strongly bounded is given.

In chapter IV, weakly precompact subsets of L 1(fi ,X) are

examined. For a Banach space X whose dual has the

Radon-Nikodym property, it is shown that the weakly

precompact subsets of L1(p,X) are exactly the uniformly

integrable subsets of ,X). Furthermore, it is shown that

this characterization does not hold in Banach spaces X for

* which X does not have the weak Radon-Nikodym property.

Page 4: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

TABLE OF CONTENTS

Ohapter P a g e

I. INTRODUCTION 1

II. REPRESENTATION OF OPERATORS ON C(H,X) 7

III. WEAKLY PRECOMPACT AND GSP OPERATORS

ON CONTINUOUS FUNCTION SPACES 24

IV. WEAKLY PRECOMACT SUBSETS OF L 47

BIBLIOGRAPHY 6 0

111

Page 5: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

CHAPTER I

INTRODUCTION

This paper will concentrate on the study of bounded

linear operators on continuous function spaces. The vehicle

for this study is the Riesz Representation Theorem; that is,

if T: C (H,X)-4Y is an operator and m is a vector measure so

that T<f) = /jjfdm for each f € C(H,X), then we study the

relationship of T to m. It seems appropriate at this time to

give a brief overview of the Riesz Representation Theorem.

We will mention only selected highlights from its abundant

history and refer the reader to [DU, p. 180-183], [Di, p.

416], and [DS, p. 373,380-381] for additional references.

In a note in Gomptes Sendus in 1909 [R], Frederic Riesz

established the following theorem

Etant donnee I'operation lineaire A[f(x)], on peut

determiner la fonciion a variation bornee «(&), telle que

quel Ie que so i t la fonetion continue f(x), on ait

W ( * ) ] = ll0 f (x)da(x) .

The following expanded version of the theorem appears in [RN,

p. 110].

The Stieltjes integral

J b f(t)i,(x) 3.

formed with a fixed function of bounded variation ot(x),

Page 6: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

defines a linear functional in the space C of continuous

functions f(x), and conversely, every linear functional can

be written in the integral form.

In 1913, Radon [Ra, p. 1333] extended the representation

theorem to Include functionals on C(Rn), and this time the

representation theorem was expressed in terms of measures.

In a note at the end of Saks" Theory of the Integral [Sa,

p. 322], Banach demonstrates an extension of the Riesz

Representation Theorem to C(H) for H compact and metric. An

extension for functionals on C(H) for H a compact Hausdorff

space was accomplished by Kakutani in 1941 [K, Theorem 9, p.

1009]. The theorem appeared as follows.

Let 0 be a compact iausdorff space, and let C(U) be

defined as usual. Then every bounded linear functional f(x)

defined on C(0) which satisfies the conditions

11/ II - '

x > 0 implies f(x) > 0

can be expressed in the form

f(x) = \^x( t)fi(dt)

where p>(E) is a completely additive non-negative set function

defined for all Borel sets B of ft such that p(U) - 1.

At this point the study branched in several directions.

Various authors, e.g. Halmos [Ha], Hewitt [He], Edwards [Ed],

and Bourbaki [Bour] have examined linear functionals on C(H)

for H locally compact. Gelfand [G] and Bartle, Dunford and

Schwartz [BDS] examined operators on C(H) which take their

Page 7: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

ranges in a Banach space. The following theorem appears in

the paper of Bartle, Dunford, and Schwartz.

^ ® ® oper&tor on C (S) to X, there exists a unique

set function p:Yhl** such that

(a) fi(')x* 6 &(S) for each x* € /*

(b) the mapping x ^(*)x of I into &($) is continuous

with ike X and C(S) topologies in these spaces, respectivelyj

(c) x*7f = $ s f ( s ) p ( i s ) x * , f € C(S), x* € I*;

(*) |^| - semi-variation of )i over S.

Conversely, if (i is a set function on £ to X** satisfying (a)

and (b), then equation (c) defines an operator T:C(S)->X with

norm given by ( i ) , and such that t*x* = p(*)x*.

The studies of Singer took another direction. He chose

instead to examine the linear functionals on C(H,X). In a

1959 paper [Si, Theorem 2, p. 35] of his, we find the

following theorem.

Le dual de I'espace C^ est equivalent a I'espace de tous

les champs de formes f completement a d d i t i f s , a variation

bornee et regvliers, la norme etant

|/| * y*r f . e£Q e

I 'equivalence $ <—> / entre ces deux espaces est donnee par la

relation

<x, - ig<*(9),ifq> (z € CA).

The paths merge in the study of operators T:C(H,X)->Y, where X

and Y are Banach spaces. In Dinculeanu's book [Di, p.

398-399] we find the following theorem.

Page 8: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

L e t Z C /' be a n o r m i n g s p a c e . I f U : % g ( T ) -» / i s a

l i n e a r n a p p i n g w i t h | | ^ | | < 03 f o r A € 56, t h e n t k e r e e x i s t s an

a d d i t i v e s e t f u n c t i o n m:% -» x (E,Z*) w i t h f i n i t e

s e m i - v a r i a t o n s u c h t h a t

9 ( f ) - - i f d m , f o r f € * g ( T ) .

M o r e o v e r , f o r e v e r y o p e n s e t G we h a v e

IM - ' ( ' ) •

Additional contributions to the representation theorem

(some quite technical in nature) were made by Brooks and

Lewis [BrL] and Batt and Berg [BB]. The theorem in [BrL] is

stated as follows.

I f L : B -4 F i s an o p e r a t o r , t h e n t h e r e i s a u n i q u e w e a k l y

r e g u l a r s e t f u n c t i o n -> B ( E , F * * ) s o t h a t

L ( f ) = f j [ f t * , f ^ 3 .

A representation theorem for operators more in keeping

the nature of the original representation theorem of

Riesz may be found in [Tu].

We conclude this chapter with a discussion of some of the

terminology and notation used in the remainder of this paper.

In the chapters which follow X , Yf and Z will represent real

Banach spaces. The dual of the Banach space X will be •

denoted by X . We identify X with its cannonical image in * *

X . The symbol H will denote a compact Hausdorff space;

$(H) will denote the <r-algebra of Borel subsets of H and S an

arbitrary <r-algebra. The set of all X! simple functions

taking their range in the Banach space X will be denoted by

Page 9: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

SJJ(S) . We will use U^(£) to denote the uniform closure of

the X valued simple functions; that is, UX(E) is the closure

of SX(E) where the topology is given by the norm defined by

II'HOD = sup<||f(h)||x:h€H>.

The Banach space of continuous X valued functions on H will

be represented by C(H,X), where the norm on C(H,X) is again

given by the formula

ll fL - »uP<||«<h)||x:h€8>. In the case where X = R, we will use S(£), U(£), and C(H) in

place of SR(S), UR(S), and C(H,R).

will use ca(£,X), or ca(£) if X = IR, to denote the set

of all countably additive set functions x . We say that

an element /t of ca(S,X) is regular if for each E € E and e >

0 there exists a set F G S whose closure is contained in E

and a set G whose interior contains E so that J/t(C)j < e for

all C € S with C C G\F. The set of all such elements of

ca(E,X) will be denoted by rca(£,X).

We say that a function T:X -» Y is a bounded linear

operator if T is both continuous and linear. A function

x :X-*IR is said to be a continuous linear functional if it is

both continuous and linear. The term operator (functional)

will be understood to mean bounded linear operator

(continuous linear functional). We shall use £(X,Y**) to

denote the Banach space of bounded linear operators having X * *

for a domain and subsets of Y for a range. For T €

«(X,Y**)f ||T|| will be given by

Page 10: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

11*11= ""pfllToolHMi*1*-$ j(c

Let m:E-*2(X,Y ) be an additive set function on a

*-algebra E. The semivariation m of m is defined by

m(A) = sup {II S m(B. ) (x.) It: II (A) , llx.lKl}.

"B IKA) 1 1 11 11 1,1

(We use 11(A) to denote a partition of A.) If m(A) < oo for

all A6£, then m is said to have finite semivariation. A

predecessor of this property, known as the Gowurin a-property, appears in [Go]. The reader should note that finite semivariation does not imply finite total variation.

rsj

If m(A^)-K) whenever (A^) is a pairwise disjoint sequence from

S, then m is said to be strongly bounded. Equivalent

formulations of this property may be found in [BrL, Lemma

3.1]. Examples in [Le] and [Dol] show that countable

additivity does not in general imply strong boundedness. The

measure m is said to be weakly regular provided m * (•) = (x,y ) '

* <m(•)x,y > is a finite regular Borel measure for each x € X

He *

and y €Y . For terms used but not defined in this paper, the

reader may consult [DU] and [DS].

Page 11: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

CHAPTER II

REPRESENTATION OF OPERATORS ON C(H,X)

For the convenience of the reader, the representation

theorem for bounded linear operators on C(H,X) used in this

paper will be presented in this chapter. We will begin with

the Riesz Representation Theorem for linear functionals on a

compact Hausdorff space. The proof which is given here comes

from [DS]. Next, a representation theorem for bounded linear

operators on UX(E) due to Dinculeanu [Di] will be given.

Finally, the representation for bounded linear operators

T:C(H,X)-4Y found in [BrL] will be demonstrated. The proof of

the representation theorem for operators on C(H,X) will

employ the technique of embedding UX(»(H)) in C(H,X)**. Once

this objective is accomplished, the representation theorem

for bounded linear operators on UX(#(H)) may be put to work.

The uniqueness and weak regularity of the representing

measure will follow from properties of the representing

measure given in the Riesz Representation Theorem.

1. Theorem.[DS, p.258-265] If M is a compact Mausdorff

space, there is an isometric isomorphism between C(i}* and

rca(H), where E is the stgma algebra of Bore I subsets of H,

so thai corresponding elements of x* and p satisfy the

Page 12: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

identity

x*f s tgfil*

£E2fif: First it will be shown that each /t € rca(E)

defines an element x* of C(H)* by the above formula and in *

this case ||x || = ||/t||. Let ft € rca(S) and f € C(H). It is

easily shown that f is p integrable. Define X*:C(H)-HR by

x (f) = JHfd/t

* Clearly x is linear. Furthermore,

5 ® o } l f | t , l H < H ) - tl'lUkll-

Thus x* defines an element of C(H)* for which ||x*|| < ||/*||.

1c

Let e > 0. Let b® a pairwise disjoint collection

of elements of £ such that

^ IkII - f -

Choose s i € {1, -1} so that s ^ f E j ) • | /t (E±) | . Since p is

regular, we may choose closed sets Cj C Ej so that |/»|(E1\C1)

< | for l<i<k and open sets G ± D C i so that |^|<Gi\Ci) K |

for l<i<k. Since H is a compact Hausdorff space we may

insist that the collection (G^) be pairwise disjoint. For

l<i<k, choose continuous functions so that c K f ^ h j O for h€H,

ffCj) - 0, and f(G*) = 0. Set f - E s ^ . Hence

* k k k II* II 2 / f d* ^ ) " S |f|(E \0 ) - £ U|(G.\C.)

i=l i=i a -1 i=i1 1 1 1

> - 2f > || |j - s,.

Thus ||x || > ||/»||. Therefore ||x*|| =

Page 13: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

Next, it will be shown that every element x* of C(H)* may

be represented by an element ft of rca(E). Let x* € C(H)*.

Since C(H) is a subspace of U(£), x has a norm preserving

V* * * extension to U(L) . Let y denote one such. Define X by . * M E ) = y (*E) for each E € £. Clearly X is additive, p(E)|

$ •

< ||y ||, and y f = JHfdJ for f € S(£) . Since S(£) is dense in

U(E), it follows that x*f • JHfdJI for all f 6 U(E) .

Now X « X1 - X2, where X and X2 are non-negative. Thus

to complete the proof, it suffices to consider the case where

4 is non-negative and obtain /( € rca(E) so that /Hfd/i =

for f € C(H).

For a closed set F, define

/»1(F) - inf{X (G) :GDF,G open), and for ECH, define

P 2 ( E ) = sup{/t1(F) :FCE, F closed}.

Observe that /»1 and are non-negative and non-decreasing.

We note that /^(F) = /»2(F) for F a closed set.

Let Fj be closed and G1 be open. If G is an open set so

that GDFFJXGJ), then (GJUJGDF^ Hence j (G^UG) < JL (G ) +

-1(G). Thus

= inf (C) rCDFj ,c open)

< X (G1UG) < X (G1) + X (G).

Since the above statement is true for all open sets

G^fFjXGj), the following holds:

^(Gj) + inf{J(G):G3(F1\G)1, G open)

= X (G1)+/r1(F1\G1) .

Page 14: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

10

Let P be an a r b i t r a r y c l o s e d s e t . The p r e v i o u s s t a t e m e n t

h o l d s f o r a l l open s e t s . In p a r t i c u l a r t h e s t a t e m e n t h o l d s

f o r a l l open s e t s G^DFflFj. Thus

/ ' i < F i ) ^ i n f { i (Gj )+ /<j (F^\Gj) : G j 3 ( F j f l F ) , Q± open}

< i n f U (G 1) : G 1 3 ( F 1 f l F ) , G1 open}

+ s u p { / t 1 ( F 1 \ G 1 ) :G1D(F1f lF) , Gj open}

< ^ 1 ( F i r i F ) + sup{/» 1 (C) :CC(F 1 \F) , C c l o s e d }

- ^ ( F j H F ) + / » 2 ( F 1 \ F ) .

I f ECH, t h e n

P 2 ( E ) » s u p { / t 1 ( F 1 ) : F 1 C E , FX c l o s e d }

< sup{/J 1 (FriF 1 ) iFJCe, F x c l o s e d }

+ s u p ^ f F ^ F ) ^ C e , F 1 c l o s e d }

< s u p { ^ 1 ( C ) :CC(EflF) , C c l o s e d }

+ s u p { / t 2 ( C ) :CC(E\F) , C c l o s e d }

- ^2(EflF) + / t 2 ( E \ F ) .

I t w i l l now be shown t h a t

^ / » 2( E n p ) + / , 2 ( E X F ) *

Let Ka and K2 be d i s j o i n t c l o s e d s e t s . S i n c e H i s a compact

H a u s d o r f f s p a c e , we may c h o o s e d i s j o i n t open n e i g h b o r h o o d s

and W2 o f K1 and K2 r e s p e c t i v e l y . I f G i s an open

n e i g h b o r h o o d o f KjUKg, t h e n } (G) > MGflWj) + ^(GflWg)

Thus

/ j 1 (K 1 UK 2 ) - i n f {J (G) iGDfKjflKg) , G open}

> i n f U (Gf1W1) jGDd^flKg) , G open}

+ i n f {J (GflW2) : GO (KjflKg } , G open}

> ^ ( K j ) + / ( ! (K 2) .

Page 15: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

11

Hence

/»2(E) > 8up{/»1(K1)+/ll(K2):K1C(EnP), K2C(E\F), ^ ^ closed)

> fi 2 (EfiF) + /»2(E\F).

Thus

/ t2

( E ) 85 / t2

( E n F ^ + * 2 ( E \ p )

for ECH and F a closed set.

Thus every closed set is measurable. Define ft to be

the restriction of n 2 to the algebra generated by the closed

subsets of H. By [DS, III. 5.2, p. 133], it follows that is

additive. Observe that for a closed set F, /i (F) = /« (F) = 1 2

/MF). Hence by the definition of and it follows that

/» is regular. Thus by [DS, III.5.14,p. 138], /t has a regular

countably additive extension to the Borel subsets of H.

To complete the proof, it will be shown that

lBtu -

for f € C(H). Let f 6 C(H). Without loss of generality

assume that f is non-negative and that 0 < f(h) < l for all h

€ H. Let e > 0. Partition H into disjoint sets,

from the ^-algebra of Borel subsets of H so that

t-

£ a ji(E ) + e > J fd/t, i==l 1 1 H

where a^ = inf {f (h) :h€E^}. Since /J is regular, we may choose

closed sets F^CE^ SO that

n ^S^a1/t(F1) + 2e >

Since H is a compact Hausdorff space and f is continuous,

there exist disjoint open sets so that

Page 16: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

12

b =inf<f(h):h€G.} > a, - -A 1 i - i n

for i - l,. . . n. Thus

n S + 3e >

Recall once again that if F is a closed set, then p^(F) =

= M * ) ' Now if G is an open set so that FCG, clearly

M F ) < /1(G). Since p( G) = sup {/j (F) :FCG} , /{(G) < /1(G). Thus

n n S b »(G ) < E b.i(G ) < JHfdJ. 1=1 i=i 1 H

Hence

<*> j H(a^>j Hfd,.

Now /»< H} = -1(H); thus

/(l-f)dJ < JH(l-f)d/».

Note that the function g defined by g = 1—f is a function for

which 0 < g(h) < 1 for all h € H. Thus the inequality (*)

must hold for g. Hence JHgd/» = JHgd/i and consequently JHfd/»

= /HfdJ.

Before proceeding with the proof of the representation

theorem on UX(E), we must make the following definition.

2- fiatinltlorj. Let m:E -+ 2(X,Y) be an additive set

function on a sigma algebra E. For f = E* x. € S (E), i X

define

J fdm = Em(A A) ( x ± ) .

For f € Uy(E), let (f ) be a sequence from S„(E) so that f A n X n

Page 17: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

13

converges to f uniformly. Define

Jfdm = lim Jf dm. n-tao

The reader may consult [Di, p.106-142] for properties of this

integral.

3- lhfi2ZSm.[Di, Theorem 1, p. 145] for each bounded

linear operator T: V f there exists a unique finitely

additive set function m with finite seaivariat ion so

that

f ( f ) - tfd*

for f e flys;.

Furthermore if 7 corresponds to n, then a(S) *

PVQQf; I»et T:Ujj(S)-fY be a bounded linear operator.

Define m:S-»a»(X,Y) by

m(A)(x) = T(xAx) for A € £, x € X.

For each A € E, the map m(A) :X-4Y is clearly linear. Now

||m(A) (x)j| - ||TCxjfA)|j < ||X||||T||

for all x € X. Thus m takes its range in X(K,Y). Let A and

B be in S so that AOb = ^. Then

m(AUB) (x) - T U A U Bx ) * T(XaX+XbX) = m(A) (x)+m(B) (x) .

Thus m is finitely additive. If f € SX(E), then it is clear

that T(f) = Jfdm. Let A € S; then

m(H) » sup{ ||E m(A1)(x1)||:Aieir(H) ,||xi||<l>

-SUp<||/*Hfd»||:(6Sx(£), IKIL-D

" IITH-

Page 18: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

14

Thus m has finite semivariation. Let f € Uv(£). Choose a A

sequence {fn) from SX(E) so that converges uniformly to f.

Then

T(f) - lim T(f ) = lim Jf dm = ffdm. n-too

n n-ta>

The uniqueness of m is clear, as it must be the case that

t U a x ) » J*Axdm - m(A) (x) .

We shall now proceed with the task of embedding U (£) in ** x

C(H,X) . First we make the following definition.

4. D e f i n i t e . For f* € C(H,X)* and x € X define

<x,f*>:C(H) 4 R by

* * <x,f >(p) > f (p«x) for f € C(H).

* Now <x,f > defines a bounded linear functional on C{H). Let

^(x, f*) b e t h e u n i <J u e regular Borel measure in the Riesz

Representation Theorem so that

<x,f*>(F) - J H F ^ ( X f f * j .

For f € C(H,X) and f* € C(H,X)*, define <f,f*>:C(H) -4 R by

<f,f*>(?) = f*(pf),

where pf(h) » p(h)f(h). As before, note that <f,f*> defines

a bounded linear functional on C(H). Let p.. * be the (r / r )

unique regular Borel measure which represents it.

For A € :»(H) and x € X, define £ Ax on C(H,X)* by

XAx(f*) = A) for f* € C(H,X) .

Thus Sjj(se(H)) may be viewed as a subspace of C(H,X)**. For A

€ &(H) and f € C(H,X), may be defined analogously.

Page 19: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

15

5. UiSfiCSm.[BrL, Lemma 2.1] If I and Y are Banack

sfaces . . i , X j ^ <= S j M D ) , Hen • ||-||M.

n Pfgpf: Let s = II y x. be a simple function In

1 * 1 A. 1 1

canonical form, I.e. A.fl A. • <f> and x t x for 1 ± j. i i

Without loss of generality assume J|x1 jj = JJsjJ . Choose x* €

Bx* so that ||xjJ| = x (Xj). Let t € Aj. Define ?:$(H) -» X* by f 0 if t (f A

?(A) = | x* if t € A

Now Jfdtf = <f(t),x > for f € C(H,X); thus y may be viewed as

an element of C(H,X)*. It is easily shown that y is

countably additive and Jyj is regular. In addition

9(H) - sup{||S?(B)xi|| : B € 11(A) , ||xi|| < 1}

- SU P < | x * ( x ) | : ||x|| < 1} = 1 « X*(X )

ll^llc(H,X) Now,

8(11) = )(7) " S r . (?) » #,(A.). 1 - 1 Ki l-l 1-1 (Xj/?) i i-r(xi,9)lrti

Let e > 0. For each i > l, choose a compact set C A^ SO

k i ? i ( x ^ , if) I (^i^i) < e. Next, choose a collection

pairwise disjoint open sets so that K. C v,, t ^ i i A A

k k

i =2 Vi ' a n d i52l/t(x4 ,») I ( Vi X Ki ) < e' Choose >k C C (H) so JL 1 B1

that p1(Ki)=l, ^i(V^)=0 and 0 < ^ < l. Then for i > l the

support of yi is contained in V1. Set G1 = {x€H:(x)=l).

oo Note that G1 is a Gg set as Gi = (x€H: p i( X)

>l~}. Now

Page 20: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

16

But ,

Thus,

. S . I ' U j . f j ' V - 'H ' i * ( V f ) l

- , 5 , l ' « < e .

< r i a ' ( K 1 , f ) * < * 1 - T > < r i ) = K f j X j )

= / j j P i x i ^ ' = P ^ * ) = 0 .

i l 1 l ' ' ( x 1 . , ) ( G i , l * f -

Now

( A T A ^ K ? ) = / . ( X i , ¥ ) (*!> - < * ! . ? > U A i ) - ? ( * ! > * » )

= l a - » < V X 1 - **<*!> " | K | | .

Hence,

l i l / A / i ' " ! " I I N I + <Ai> I

- IKII " i 5 2 I'1 < X ; i , » ) ( A i ' I

* I M + j U ^ ^ ' V l ]

* IKII - , j l « W - £

> HxJI - 2 c .

T h e r e f o r e | H | c ( h . x ) - > | K | | > H ^ .

Now, f o r t h e r e v e r s e i n e q u a l i t y , l e t s be a s b e f o r e and (

^ B C ( H , X ) * ' c h o o s e ( K ^)» {G^}» { V l } , and {p^} a s b e f o r e ,

e x c e p t l e t i range from l t o k . Then

l < S - H " l i l / ( x . , ( ) < A i ' l

Page 21: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

17

* li5iI'l(x1,o<Ai' " '(Xj.o ( 0i' ]I

+ 3 f

k 5 iii5,'ixiiic(H.x)+ 3 c

< IjxJI + 3c .

Therefore |H| c ( h > x,*- < Hs^.

6. IbfifljEfiffi.CBrL, Theorem 2.1] Let f € C(M,I). There

exists a sequence simultaneously

approximating f in the uniform and C(M,I)** topologies.

SXSSL: Let A„ = <x: ||f(x)||x > 1, = n <x: ||f(x)||x > i-|>.

Note that A n € ®(H) as A n is a compact . Let f € C(H,X).

Define to be f* . n

Note that

IIfn - f L - ll*An

f - f L < i t 0 -

Thus f n -• f uniformly.

By Definition 4, f^ may be viewed as an element of * *

C(H,X) . We will next show that f converges to f In

* * 2 C(H#X) . Let € > 0. Choose n so that - < e for n > N. Let *

f € BC(H,X)*' s l n c e X A is the pointwise limit of elements n

of C(H), by applying the Lebesgue Dominated Convergence

Theorem, we may choose f € C(H) so that

I ' u . f V V - '(f,f*j<"l < e-

Page 22: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

18

Hence

- f*(f)| - k ( f, f*,(A n) " f*(f)|

- I ' u . f V V +

< € + |f*(pf - f)|

< € + ii' ii ii" - 'ii

< e + - < 2e . n

oo

Thus f n -t f in the C(H,X)** topology.

We will now proceed with the construction of the desired

simple functions. Partition An into sets <A .} n so that n ni i a-1

Ani € f o r e a c h n a n d 1 a n d ||f(x) " f(Y)|| < ^ for x and y

k ft

Now in A n l. Fix x n l € ((Ani) and set s n > .S x A x n l

n i

HSn ~ f L = Hi?, *a . xni " fll loo n i

k

S Hi?, *A xni " *A fllao + ll*A f " f L < n'

n i n n

Thus s n -» f uniformly.

Let e > 0. Choose N so that ^ < e for n > N, and fix n >

N. For convenience set Bi - a r 1, xa - X r 1 , and kn= k. Let

'* € BC(H,X) * ' H o t e t h a t

k s "n " «n)(f »l " I i?i *BjXi "

= l i l . ^ X j . f W " ' ( f . f W l l -

Choose compact sets Ki C b, so that

i ^ l ' i x ^ f ' i h w <<• an<s i l 1 i " ( f . f M < vc i > -

Next, choose a pairwise disjoint sequence {V.}k of open 1 i - 1

Page 23: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

19

sets so that K± C ,

k

For each i choose pi € C(H) so that p± (K±) » l, ^(Vj) = 0,

and 0 < < l. Thus

lv f*> - V*>l - - Nf,,*)'8!)! 5 I ^ I X j . f V W l + R f f . f V W l

+ l^< X l,e*)< Ki' " ^ (f,f*)

( Ki'l

S 2 f + l ^ l x ^ f V i ' -

S 2 f + Sl'*(x1,f*)l<Vi\Ki'

+ |Sf*(x1p1 - f P l)|

+ s k((,f * ) l'V Ki» < 4f + Hf IIIIVl - f'illoo

< 4e + i < 5e. * *

Thus s n 4 f in C(H,X)

7. lhfi2Eeffl. [BrL, Theorem 2.2] If T:C(M,X) -» Y is a

bounded linear operator, then there is a unique weakly

regular set function n:m(H) -» z(X, Y**) so that

Uf) • for f € C(t,l).

furthermore, ||7|j * %(g)m

a|c g|c & ale

EEfiof: Now SX(»(H)) C C(H,X) . Hence T :Sx(® (H) ) -t

Page 24: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

20

** * * Y . By Theorem 5, it follows that T has a continuous

extension T to all of U^(»(H)). Hence by Theorem 6, T(f) » * *

T (f) - T(f) for f € C(H,X). By Theorem 4, there exists

m:«8(H) -+ 2(X,Y ) representing T. Thus,

T < V > = T <*AX) = ^ H V d m = m< AM x)-

Let x 6 X and y* € Y*. Then

<m(A)(x),y*> = <T**(^ax),y*>

- A r A x ( T * ( y * ) ) = > ( x > T« ( y' ) ) (A)

where ^(x,T*(y*)) ***© regular Borel measure

guaranteed by the Riesz Representation Theorem. Hence the

choice of m is unique.

In addition, m(H) =||T|| = ||T|| .

If T:C(H,X)->Y is a bounded linear operator with

representing measure m, we may actually show that |m *J is

regular. To this end, consider the following.

8. Pefjpitjon. Let H be a compact Hausdorff space. An

operator T:C(H,X)~*Y is said to be dominated if there exists a

positive linear functional p € C(H)* so that ||T(f)|| <

<P.||f< ->||>.

9. Theorem. [BrL, Theorem 2.8] If T:C(H,X)-+Y is a

bounded linear operator with representing measure m, then T

is dominated if and only if J m I € rca(s&(H)).

Page 25: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

21

£caa£: Suppose that |m| € rca(»(H)). Thus ||/Hfdm|| <

si**ce /H( )d|m| defines a positive functional, we

conclude that T is dominated.

Conversely suppose that T is dominated by a positive

functional p. Using the Riesz Representation Theorem we

obtain a positive regular Borel measure (i for which p(f) =

JHfd/i for all f € C(H) .

Let e > 0. Let A € E, x 6 X, and y* € Y*. Since m, * (x,y )

is a regular Borel measure, there exists a sequence (K ) of n

compact sets and a sequence (G ) of open set so that K C A C n n

Gn a n d lm(x,y*) H Qn^ Kn* < n* L e t {^n^ b e a se<3uence of

continuous functions so that Pn(Kn) = 1 and Pn(Gnc) • 0. By

the Lebesgue Dominated Donvergence Theorem, lim f » dm * jHB, J" P n <X'Y )

= m(x,y*)(A) a n d l l m J V n ^ = ^ ( A )* N o t e t h a t

n-to

IVn d m( X,y*)l = l<**'/H'nxd">l

< II/hVHI * ^Hlknxlld^ *

Thus m(Xfy*)(A) < /»(A) for all x € X, y* € Y*, and A € £.

Hence ||m(A)|| < /t (A) for all A € S. Thus |m| (A) < ^ (A) for

all A € S, and it follows that |mj € rca(E).

10. Defjnitipn. Let m:E-tt(X,Y**) be a finitely additive

set function. For y* € Y*, define my*:E-«* by

®y*(A)(x) = <m(A)(x),y*>.

11. Corollary. If T:C(H,X)-+Y is a bounded linear

Page 26: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

22

operator with representing measure m, then | * | € rca($(H))

Pffppf: If m is the representing measure for a bounded

linear operator T, then for y* € Y* we have (y*oT)(f) •

Y (T(f)) = /Hfdmy*. Thus ||y*oT|| - m *(H) = |» *1* Hence $ - y

y OT is dominated. Therefore m^* 6 rca(36(H)).

We may now use Theorem 9 to establish a representation

theorem for continuous linear functionals on C(H,X).

12. Corollary. If x* € C(H,X)\ then there exists m €

rca(SB(H),X ) so that

x*(f) = /Hfdm.

* •

PrQQt• Let x € C(H,X) . By Theorem 11 there exists a

measure m:s®(H)->(ie(X,R)=X*) so that |mr| € rca(S) for all r €

R. Since nij • l«m, m € rca(»(H) ,X*) .

In this paper we will be primarily concerned with weakly

precompact and QSP operators on continuous function spaces

and how these operators influence their representing

measures. In addition we will find it convenient to use

several properties of unconditionally converging and weakly

compact operators. We close this chapter with the formal

definitions of these classes of maps.

Page 27: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

23

13• A subset S of a Banach space X is said

to relatively weakly compact if the weak closure of S is

weakly compact.

An operator T:X-*Y is said to be weakly compact if T(B )

is relatively weakly compact.

A subset S of a Banach space X is said to be a GSP set if

for all finite measure spaces (fi,S,p), bounded linear

operators T:X-H^f/O, and e > o there exists B € £ such that

/»(fl\B) < e and T(S) | B • {T(x)*B:x€s> is totally bounded.

A bounded linear operator T:X-+Y is said to be GSP if

T(BJJ) is a G S P s e t .

A subset of S of a Banach space X is said to be weakly

precompact if every bounded sequence of elements of S has a

weak Cauchy subsequence.

A bounded linear operator T:X->Y is said to be weakly

precompact if T(B^) is weakly precompact.

A bounded linear operator T:X-+Y is said to be

unconditionally converging if ST(xn) is unconditionally

convergent whenever £xnis weakly unconditionally Cauchy.

Page 28: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

CHAPTER III

WEAKLY PRECOMPACT AND GSP OPERATORS

ON CONTINUOUS FUNCTION SPACES

In this chapter GSP operators and weakly precompact

operators on C(H,X) will be examined. The focus of this

discussion will be on the relationship between these classes

of operators and their representing measures. For many

important classes of operators on C(H,X), properties of the

operator are reflected in the values of the representing

measure. For example, if T:C(H,X)-+Y is an operator with

representing measure m, then

(i) m(A) is weakly compact for each Borel set A,

whenever T is weakly compact,

(ii) m(A) is compact for each Borel set A whenever T is

compact,

(iii) m(A) is absolutely summing (p-summing) for each

Borel set A whenever T is absolutely summing (p-summing),

(iv) m(A) is unconditionally converging for each Borel

set A whenever T is unconditionally converging,

(v) m(A) is completely continuous for each Borel set A

whenever T is completely continuous,

(vi) m(A) is weakly completely continuous for each

Borel set A whenever T is weakly completely continuous.

24

Page 29: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

25

The implications in (i) and (ii) are in Batt and Berg

[BB] and Brooks and Lewis [BrL]. The conclusions in (iii)

are from Swartz [Sw2] and Bilyeu and Lewis [BiL]. The result

in (iv) may be found in Swartz [Swl], Dobrakov [Do2] and

Bilyeu and Lewis [BiL], The assertion in (v) is obtained by

Bilyeu and Lewis[BiL], and (vi) is established by Bombal and

Cembranos in [BC]. (The reader may refer to these references

for the definitions of classes of operators not explicitly

discussed in this paper.)

In [Swl] Swartz showed that if T is unconditionally

converging, then m is strongly bounded. As each member of

the six classes of operators listed above is unconditionally

converging, it follows easily that each must have strongly

bounded representing measure.

In this chapter examples will be given to show that

neither of the above conclusions necessarily hold for the

representing measures of weakly precompact operators or QSP

operators. That is, if T is weakly precompact (GSP), then m

need not be strongly bounded and m(A) need not be weakly

precompact (GSP). However, sufficient conditions will be

given for a weakly precompact (GSP) operator to have a

representing measure with weakly precompact (GSP) values.

The following fundamental theorem of H. Rosenthal [Ro]

characterizes weakly precompact Banach spaces. His theorem

has proved important in the study of Banach spaces.

Page 30: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

26

1« Theorem• A Banach space contains a subspace

isomorphic to if and only if it has a bounded sequence

with no weakly Cauchy subsequence. Consequently, a Banach

space is weakly precompact if and only if it contains no copy

of l t .

It is well known that If an operator T:X-»Y Is weakly

compact, then Its adjoint, T , Is weakly compact. However,

an analogous statement does not hold for weakly precompact

operators. In fact, T weakly precompact need not even Imply * *

that T is weakly precompact.

2. • If a bounded linear operator T:X-+Y is

weakly precompact then neither T* nor T** need be weakly

precompact.

Let I:cQ-+co be the identity operator. The operator I is

weakly precompact as cQcontains no copy of lt . However, I* 9fe *

and I are the identity operators on lt and / respectively .

Now, l^ contains an isomorphic copy of lt. Thus neither I* * *

or I can be weakly precompact.

It is the case, however, that the following holds.

3. Theorem. If T:i-*Y is a bounded linear operator so

that 7 is weakly precompact, then I is weakly precompact.

Page 31: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

27

ECflfif: First, we observe that if a Banach space Z

contains a copy of lt then so must Z*. To see this, let

J:ij-*Z be an isomorphism. Then by [DS, VI.2.7] we see that * *

J : Z is onto. Since contains a copy of lt, J

contains a bounded sequence with no weak Cauchy subsequence.

This sequence must be the image of a bounded sequence in Z*

which has no weak Cauchy subsequence. Thus, by Theorem 1, Z*

contains a copy of lt .

Now, suppose that T:X-+Y is a bounded linear operator for

which T is not weakly precompact. Then X contains a copy, W,

of lt on which T acts as an isomorphism [Lo, Theorem 3], Let

J: Z i-+X be an isomorphism taking lt to W. Note that TJ is an

isomorphism. Hence (TJ) — J T is onto I . Since I 00 oo

contains a copy of It , I^ contains a bounded sequence with no

weak Cauchy subsequence. This sequence must be the image of

some bounded sequence in Y*. Thus J*T* cannot be weakly

precompact. Therefore T* is not weakly precompact.

A fundamental theorem of C. Stegall [St], proves to be

very useful in the discussion of GSP operators. Stegall's

Theorem is a consequence of the proof of the factorization

theorem of Davis, Figiel, Johnson and Pelczynski in [DFJP].

We state these two theorems and requisite definitions for

reference purposes.

4. Definition. Let (QX.it) be a finite measure space.

Page 32: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

28

A function f:fl->X is said to be p-measurable if there exists a

sequence (fn) of /»-simple functions so that lim llf - fll = o n-x» n 11

H-almost everywhere.

A /t-measurable function f:fi-»X is said to be Bochner

integrable if there exists a sequence of simple functions

(fn) so that

/Oll'n " '11^ " °" n-to

In this case we define / fdft for E € £ by E

= lim / f c n-»oo

The collection of all ft -Bochner integrable functions f:fl-+X is

denoted by L^/^X). For a study of properties of the Bochner

integral the reader should consult [DU, II.2].

A Banach space X is said to have the Radon-Nikodya

property with respect to <0,E,/e) if for each ^-continuous

vector measure G:S-»X of bounded variation, there exists a g €

L1(/t,X) so that

G(E) = JEgd/»

for all E € S.

If X has the Radon-Nikodym property with respect to every

finite measure space then X is said to have the

Radon-Nikodym property RNP.

5. Theorem• A bounded linear operator 7;1HF is weakly

compact if and only if 7 factors through a refelexive space.

That is, T:X-*Y is weakly compact if and only if there exists

Page 33: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

29

*

a reflexive space 2, and bounded linear operators A:X-+Z and

B:Z->Y so that I = 8A.

6. ThQQCQifl. A bounded linear operator T:I-*¥ is QSP if

and only if T factors through a space 2 so that 2* has the

Hadon-Nikodnt property.

Theorem 6 was proved by demonstrating that if T:X-+Y is a

QSP operator, then the same construction in [DFJP] of a

space through which T factors must yield a space Z so that Z

has the Radon Nikodym property.

It should be observed that if I:X-« is the identity

operator on X, then the construction in [DFJP] of the space

through which T factors yields a space isomorphic to X.

Stegall's proof together with this observation makes possible

the following easy example.

Esssmls* If T:X-4Y is a bounded linear operator which

is QSP then neither T nor T** need be QSP.

Let I:cQ->co be the identity operator. As CQ* « L{ has

the Radon-Nikodym property, it follows from Theorem 6 that I

is a GSP operator. Now, I i s the identity on lt. In

this case the space Z obtained via the construction in

[DFJP1] is isomorphic to lt. However, It* « J does not have

the Radon-Nikodym property. Hence, I* is not a QSP operator. * *

Similarly, I is the identity operator on I and the space

Page 34: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

30

* *

constructed in [DFJP] is isomorphic to I . Now, I * does 00 00

not have the Radon-Nikodym property. As a consequence I

cannot be a GSP operator.

The following example shows that the analog of Theorem 3

does not hold for GSP operators.

8. Example• If T:X-+Y is a bounded linear operator so *

that T is GSP, then T need not be GSP.

Let I:JT-*JT be the identity operator on the James Tree

space [Jam]. Now, all even duals of JT have the

Radon-Nikodym property and all odd duals do not have the

Radon-Nikodym property.[LSt] In view of Theorem 6, all odd

adjoints of I are GSP operators. Now, as every adjoint of

the identity operator is an identity operator, in light of

our earlier observation, it is clear that I and all its even

adjoints fail to be GSP operators.

The following theorem of Riddle, Saab, and Uhl [RSU] is

an additional consequence of the factorization theorem in

[DFJP]. Their theorem provides a nice characterization of

weakly precompact operators.

9. Theorem. The operator T:I-*Y is weakly precompact if

and only if T factors through a space which does not contain

ii-

Page 35: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

31

The next theorem, due to S. Musial [Mu] in the case where

each separable subspace of X is contained in a separable and

complemented subspace of X and L. Janika [Jan] in the general

case, aids in the further understanding of weakly precompact

operators. We first make an essential definition.

PsfAnitAon- Let (fl,E,/i) be a finite measure space.

Suppose that f:fl-*X is a function so that x*f € L (/t) for all * * ** *

x € X . For E € E, there exists x_ € X for which £

) 38 fEx*fd/t

for all x* € X* [DU, II.3.1]. If x £ € X for all E € E, then

we say that f is Pettis integrable. We say that x is the E

Pettis integral of f over E and we write

* E = P .

A Banach space X is said to have the weak Radon-Nikodym

property with respect to (Q,E,/») if for every /i-continuous

vector measure G:E-« of bounded variation, there exists a

Pettis integrable function g so that

G(E) - P - JEfd/i

for all E € E.

If X has the weak Radon-Nikodym property with respect to

every finite measure space (ft,E,/i), then X is said to have

the weak Radon-Nikodym property.

Theorem *[Jan, Mu] A Banack space 2 does not contain

a copy of if and only if Z* has the weak iaiou-Nikoiya

Page 36: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

32

property.

We may now use the conclusions of [DFJP], [RSU], [St],

and [Mu], [J] to establish a relationship between weakly

compact, GSP, and weakly precompact operators.

12. Uisacfil. Suppose that T:I->Y is a bounded linear

operator. If T is weakly compact, then T is GSP. I f T:X-*¥

is GSP, then f is weakly pre compact .

Ecfifil: Suppose T:X->Y is weakly compact. By Theorem 5, T

must factor through a reflexive space. However, every

reflexive space has the Radon-Nikodym property. Hence, from

Theorem 6, one must conclude that T is a GSP operator.

Suppose T:X-*Y is GSP. By Theorem 6, T factors through a

space Z whose dual has the Radon-Nikodym property. Since any

space having the Radon-Nikodym property must have the weak

Radon-Nikodym Property, by Theorem 10, Z cannot contain a

copy of ij. Hence, by Theorem 9, T is weakly precompact.

At this point examples may be easily produced to show

that weakly compact, GSP, and weakly precompact are distinct

classes of operators.

13. Example• GSP operators need not be weakly compact.

Earlier, we observed that the identity operator on c is o

Page 37: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

33

a GSP operator. However, this operator is riot weakly

compact.

14. Example. Weakly precompact operators need not be

GSP operators.

Consider again the identity operator I:JTMJT on the James

Tree space. Recall (Example 8) that all odd adjoints are

GSP, but the operator I and all its even adjoints fail to be

GSP. Neither JT nor any of its duals contain lt [LSt,

Corollary 1], Consequently, the operator I and all its

adjoints are weakly precompact.

We now direct our attention to operators on C(H). Here

we find a much closer relationship among the three classes of

operators under scrutiny. To assist us in our investigation

we shall use the following theorem [DU, VI.1.3, p. 149].

15. ThQQrem. // £ a „ - f i e \ i 0f subsets of S, and

T:0(E)-*Y is a bounded linear operator that is not weakly

compact, then there is a linear subspace of V(Yt) that is an

isomorphic copy of I^ on which T acts as an isomorphism.

16. Theorem. If T:C(B)-*¥ is a bounded linear operator,

then the following statements are equivalent.

( i ) I is weakly compact.

( i i ) T** is GSP.

Page 38: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

34

(iii) 7** is weakly precompact.

( * v ) 7 \fJ(Yi) 1 3 w e a k l 9 precompact

Prpof: (i)-f(ii) if T is weakly compact, then T** is

weakly compact. Thus, by Theorem 12, T** is GSP. 9|C 3§6

If T is GSP, by Theorem 10, T is weakly

precompact.

(iii)-»(iv) This implication is clear.

(iv)-»(i) Suppose T is not weakly compact. Then • *

T |U(£)ls n o t w e a k lY compact. By Theorem 15, there exists a

linear subspace of U(£) that contains an isometric copy of t GO

* * on which T a c t s a s a n isomorphism. Now I^ contains a

copy of I a n d thus T j u(S) a c* s a s a n isomorphism on a * *

copy of I j. Hence T |u(£) canno"t be weakly precompact.

Of course, the representing measure of any operator on

C(H) will have weakly precompact values. In fact, it should

be observed that if X contains no copy of lt, then every

bounded linear operator T on C(H,X) will have a weakly

precompact valued representing measure. However, the next

example will show that even if an operator on an arbitrary

C(H,X) space is weakly precompact, then the representing

measure need not have weakly precompact values.

17. Example. If T:C(H,X)-*Y is a weakly precompact

operator, then m(A) need not be weakly precompact.

Page 39: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

35

Let H = {0} U n = 1,2,...}. Let £ denote the Borel

subsets of H. Define an operator T on C(H,J ) as follows-00

T(f) = (f(0)n - f(i) n)™ = 1

for f € C(H»^00) • Here f(0)n and f(^)n denote the nth

coordinates of f(0) and f(^) respectively. Let f € C(H,i ) " 00

Then

l , < 0 ,n " '<5'nl S " "n'lli ? «• 00 11

Thus T(f) € cq. Hence TtCfH,!^) *4 cq, and clearly T is

linear. To show that T is bounded, note that

||T(f)||c - sup|f(0) -f(i)| o n

< suplf(O) I + supIf(~) I n n

< ||f(0)||, + ||,(i)||( < 2||,L. 00 00

Thus T is bounded.

consider the operator T on Uj(£)=U defined in the Now _ _ . — -

00

same manner; that is

*<«> - (f(0)n -

for f € U. The same arguments used for T show that T is

bounded and linear. Furthermore, T takes its range in I as 00

sup|5(f)n| - sup|f(0)n - f(i,n| < ,||i;||ro.

Note that T restricted to C(H,1 ®) is T. Define m:£-*e(Z J ) 00 00

by

m(A) (x) - T(^Ax) = ((ATAx(0))n - <*Ax(i) )®al.

Clearly, m is the measure guaranteed by Theorem II.3 for

Page 40: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

36

which

T(f) - JHfdm

for all f € U. To show that m is the representing measure

for T, it suffices to show that m is weakly regular. Let x =

( xn ) € a n d y = € c0* = Without loss of

generality assume ||x|| = 1. Let e > 0. Choose N so that

GO £ |Yn| < e. Note that for A € E, n»N

m<x,y)<A> " <•<*>*.*> = J [<***«»>„ " ( V n " n l y n -

Furthermore note that any subset of H is either open or

closed. In fact, any set containing 0 must be closed and any

set not containing 0 must be open. It suffices to consider

two cases.

First, suppose A C h and 0 A. Then A is open. Set F =

A fl 1 < n < N-l}, and notice that F is certainly compact.

In addition

lm(*.y)l(AVP)

"" S U^{ Bg| im(x,y) I partitions A\p|

" ne?\F | U A^ X ,° , ,'> " ( W R " n l | v „ |

< £ I x I I y I < e . n€A\F nl

Now suppose A C H and 0 € A. Then A is closed and in fact,

compact. Let Q = [A D <i:i < n < N-l>] U > N> U {0>.

The set Q is open and

lB(x,y)l<°U> * ^ A k a \ * " l » l n " W ' R ' n l

Page 41: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

37

ao

- n?„lynl < f"

n=N

Thus m is weakly regular and consequently m must be the

representing measure for T.

Now T must be weakly precompact as Cq contains no copy of

I!. However,

m(0)(x) - T(*{0)x) = (xn - 0 ^ - x

for all x = (XR) € l^. Thus m(0) Is the Identity operator on

• Hence m(0) is not weakly precompact.

Example 17 also provides us with a GSP operator whose

representing measure has some values which are not GSP

operators. To see this, observe that T factors through c , a o

space whose dual has the Radon Nikodym property. In view of

Theorem 6, T must be GSP. Since m(0) is not weakly

precompact, by Theorem 12, m(0) is not GSP. Furthermore,

since the representing measure takes its range in 2(/ I ) 00 00

rather than in by [Do2] the representing measure is

not countably additive and hence not strongly bounded. This

example leads to the following theorem.

18. Theo^m. The Banach space I contains a copy of lt

if and only if there exists a Banack space Yf a compact

Mausdorff space M, a Borel subset A of M, and an operator

L:G(B,I)-*Y so that L is weakly precompact, but m(A) is not.

Page 42: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

38

PfQflf * Suppose X contains an isomorphic copy of 2 j. Let

Y be the subspace of X isomorphic to Let J:Y -> J be an oo

embedding. As is injective, J may be extended

isometrically to all of X. Continue to call this extension

J. Let H = {0} U {~in € N>. Let £ denote the ^-algebra of

Borel subsets of H. Define J:on (5j) by J(f)(h) * J(f(h))

for h € H. Note that if f = then J(f) = S^A f(Xl).

Thus J maps Ux(£) to Uj (E ) . Furtermore, for f € C(H,X), J(f) 00

€ C(H, l^). Define L : C ( H , X H C q by L - ToJ. The reader may

observe that

t(f) = (J(f(0))n - J(«(i))n).

Note that L = TOJ defines an extension of L to U (S) having A

representing measure n:£-te(X, defined by n(A)(x) =

m(A)(J(x)). As n is weakly regular, n must be the

representing measure for L. Hence, n(0)(x) = m(0)(J(x)) «

J(x). Thus n(0) = J, which is not weakly precompact.

We now turn our attention to establishing some sufficient

conditions for m(A) to be weakly precompact for every Borel

set A. These conditions were motivated somewhat by the proof

in [BiL] of statement (v) in the introductory paragraph of

this chapter. We will begin by making the following

definition.

19- Definition. Let T:C(H ,XHY be a bounded l i n e a r

Page 43: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

39

operator. For x € X, define T :C(H)-»Y by

Tx(

f) • T(fx) for f € C(H).

Observe that T x has representing measure m (IR,Y** )=Y**)

defined by n»x(A) = m(A)(x). (In Chapter II, m * was defined * * ^

for y € Y . Although the same notation has been used for

two different meanings, the intent will always be clear from

context.)

20. Theorem. If T :C(H ,X)-*Y is weakly precompact, then

the following statements are equivalent.

(*) I j , s precompac t .

( H ) is weakly precompact for all x in X.

( H i ) weakly compact for all x in I .

(*v) I j . ( B f i f f l j ) 15 dense in T^ ( ^ f f ( t y ) f o r x € / .

'Sr(Z)' (*) f (1C(l)!)! " '«»»« <» (%))•

P£oq£: (iH(ii) Suppose that T (£)is weakly * X

# * precompact. Hence T x .y ^ is weakly precompact for all x €

X. Thus by Theorem 16 T x is weakly precompact for all x€ X.

(ii )-*( iii) This implication follows from Theorem 16.

(iiiH(iv) Suppose that T is weakly compact for all x €

X. Then m is countably additive for all x. Hence W = * x

{m(x,y*) :llY ||-1} l s uniformly countably additive for all x.

Thus W x is uniformly regular. Let A € E. Choose an open set

Gi and a compact set so that K1CaCg1 and m ( x y* )(G 1\K 1) <

Page 44: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

40

1 i' ^ o r Y ^ ®y*' i € II. Let be a continuous function

so that = 1 and = 0. Note that

<V*. V l » = * /„*»"»•(*. Y*) « < y \ T " < * a »

uniformly for y* G By*. Therefore T**(B C ( H )) is dense in

Tx (BU(S))*

(iv)-4( v) Suppose that T**(B C ( H )) is dense in T**(B U ( S ))

for all x. Hence n»x takes its values in Y. Thus by [Do2] m

is countably additive for all x. Consequently the set

{m(x,y*) :IIY II-1* i s uniformly countably additive and hence

uniformly regular. Let A 6 S and x € X. We may now proceed

as in (iii)-»(iv) to obtain a sequence (f.) in B_,„. so that * $ $ 4c

<y <Y »TX (JtA)> uniformly for y*€By*. Thus

<Y fT(xf>1)> -4 <y*,T**(xjtA)> uniformly for y* € By*. Since bSx(£)

1 s d e n s e l n bUx(£)' t (bC(H,X)) 1 s d e n s e l n

T (BUx(E))*

* * (v)->(i) Suppose that T (BC(H X ) ) is dense in • * '

T *BU (E)* a n d t h a t T i s w e a kly precompact. Let <f ) be a X ^

bounded sequence from "X(E) • Choose a bounded sequence (p )

from C(H,X) so that ||T**(p1) - T**f 1|| 0. Since T is weakly

precompact we may choose a subsequence f of f. so that * * * *

T ) is weakly Cauchy. Then T (f. ) must be weakly ii

* * Cauchy also. Thus T ^ is weakly precompact.

1 X

It should be noted that a stronger implication than

Page 45: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

41

(iii)-+(iv) actually holds. It is actually the case that the

statement In (ill) implies that T x (B C ( H )) is dense in

Tx ' This implication holds for the following very

fundamental reason.

21. Theorem- A bounded linear operator T:I-*Y is weakly

compact if and only if is dense in

Proof: Suppose T is weakly compact. Let (xff) be a net

from Bx which converges to a point x in the w* topology. By

* * * * [VI.4.7,DS,p. 484] T (xtf) converges to T (x) in the weak

s|c 4c $ topology. Thus T (w -closure(Bx)) C w-closure(T**(B )).

* X

Recall that w -closure(Bx) = B **. Thus T**(B **) = «V X

• * *

T (w -closure(Bx) C w-closure(T**(Bx)) =* T**(BX). Thus

T (Bx) is dense in T**(BX**).

For the converse, suppose that T**(BV) is dense in A

T**(BX**). Thus T**(BX**) C T**(Bx) C Y. Hence T is weakly

compact by [DS, VI.4.2, p. 482].

22. Corollary. If T:C(M,X)-*Y is weakly precompact and

any of conditions (i)-(v) in Theorem 26 hold, then m(A) is

weakly precompact for every Borel set A.

Propf• Let T:C(H,X) -» Y be weakly precompact and suppose

that any of conditions (i)-(v) in Theorem 20 hold. Thus

Page 46: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

42

* * T |U (£) 1 8 w e a k l Y precompact. Let A 6 S and (x ) be a

I X 1

bounded sequence from X. Note that (tf.x.) is a bounded A JL

sequence from Ux<£). Thus ( T * * ( V l ) ) . (.(AKxp) has a weak

Cauchy subsequence.

A theorem similar to 20 may be demonstrated for GSP

operators. First consider the following lemma.

23. Lemma. If ACJCI and B is a GSP set, then A is a GSP

set.

Pr99t: Let (fi,Etft) be a finite measure space and

b e a b o u n d e d linear operator. Let € > 0. Since B

is a GSP set, we may choose K € E so that /t(ft\K) < e and

{L(X)*K:X€B} is totally bounded. Hence {L(X)*„:X€B} is

compact. But {L(X)*K:X€A> C {L(X)ATK:X€B> . Thus,

<L(X)*k:X€A} is totally bounded. Therefore A is a GSP set.

24. Theorem. If f : 0(H, X)-*Y is a GSP operator, then the

following are equivalent:

(*) f I/? /vi *3 SSP. f j ( s;

( i i ) 7** is GSP for all x € I.

(***) Tx is weakly compact for all x € J.

(iv) fs**(*C(M)) is dense in

'e.(E) (v) i» de.se in t"(Mf - . ) .

Page 47: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

43

( v * ) T \y (Y,) *s weakly precompact. I /

£ESSl: (iH(ii) Suppose that T**.^0 j> is QSP. Then • X

|U (S) l s w e a k l y precompact. Thus from Theorem 20 we see < X

* •

that T x is weakly precompact for all x € X. Hence by

Theorem 16, T** is QSP for all x € X.

This implication follows from Theorem 16.

The implications (iii)-f(iv) and (iv)-4(v) follow from

Theorem 20.

(v)-4(i) Suppose that T**(B C ( H X ) ) is dense in

T (BUx(E))* s i n c e T (BC(H,X)) i s a Q S P s e t a n d T**<Bu ( S ) )

^ 3lc sic

^ Lemma 23, T (B^ (£)) *s a set. Thus X

T lu x (S) 1 3 GSP-

Since every GSP operator is weakly precompact, it follows

from Theorem 20 that (vi) is equivalent to (i)-(v) under the

hypothesis that T is GSP.

** T

25. Coronary. If T :C(i ,X)-*Y is a GSP operator, n

represents T, and any of conditions ( i ) - ( v i ) of Theorem 24

hold, then n(A) is a GSP operator for every Bore I set A.

PrffOf: Let T: C (H, X) -4Y be a GSP operator and suppose that

any of conditions (i)-(vi) in Theorem 24 hold. Thus * * * *

T |ux(£) l s a Q S P °P e r a t o r• Let A € S. Now T (By s ) is

X

Page 48: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

44

a GSP set as is any of its subsets. Since m(A)(B ) C * * ^

T ^BU ( S ) ^ * s a 6 S P s e t- Thus m(A) is a GSP X

operator.

26, Lemma. If I C cabv(%,X) and £ is weakly precompact,

then {|/t|.-/i € /} is uniformly covntably additive.

Efaof: Suppose K is weakly precompact and {J/»|:/e € K} is

not uniformly countably additive. Choose e > 0, a sequence

(/*j) from K, and a pairwise disjoint sequence (A^) from £ so

that > e for i = l, 2, . . .. Pick partitions

/n >n(i) _ m ^ _ { ij'j=l o f Ai a n d xij € B

x* s o that

n(i) .

for i = l, 2, . . . Define F:cabv(£,X) -4 Zj by F { f t ) =

^Xij^ (Bi j)) j=l, i=l = xi2^^Bi2^' * * ''

xln( 1 )P Bln( 1) ' x2l',(B21) X2n(2)^t(B2n(2))' * '

Note that ||F(/i)|| < ||i||. Thus F is bounded and clearly F is

linear. Thus F(K) must be weakly precompact. Let (x )°° = ij j=l

oo

Kk,ii < *• k2

F(/t1). Let kj = l. There exists k so that 2 |x..| < -j=k^ V 2

Thus ||F(/i1) - F(/ik )|| > |. Pick k so that S |x. I < §. 2 j=kg 2 2

Then ||F(/,1) ~ P(/»k )|| > | and ||F(*k ) - F(/»k )|| > |. This 2 3

process may be continued to obtain a sequence (/j. ) so that i

l l P ( ' k . ) " F ('k } l l - I f o r T h u s >)* has no norm 1 3 Ki 1

Page 49: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

45

convergent subsequence. Since weak convergence and norm

convergence in l^ coincide, this sequence has no weak

convergent subsequence. Furthermore, I is weakly

sequentially complete. Hence (F(/» )) has no weak Cauchy i

subsequence. Consequently F(K) is not weakly precompact.

27. Theorem. If T:C(Jf,X)-tY is a bounded linear operator

so that J is weakly precompact, then a is strongly bounded.

E£S2£2l: Suppose T is weakly precompact. Then T ( B * ) = * *

{my*:y €By*} is weakly precompact. Thus {|m *|:y €By*> is

uniformly countably additive by Lemma 26. Hence m is

strongly bounded.

28. Corollary. If T:C(M,I)-*Y is a bounded linear

operator so that T* is weakly precompact, then T** 'X

weakly precompact and-, consequently, n(A) is weakly

precompact for every Borel set A.

h & ) t s

£ £ 0 0 f : Suppose T:C(H,X) -» Y is a bounded linear operator •

so that T is weakly precompact. Thus by the previous

theorem, m is strongly bounded. Thus m is countably

additive. This implies m is countably additive for all x €

X. Hence T x is weakly compact for all x € X. Since T* is

weakly precompact, from Theorem 3 we know that T is weakly

Page 50: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

46

precompact. Thus by Theorem 20, T j0 j, is weakly • X

precompact. By Corollary 22, m(A) is weakly precompact for

every A € S.

It should be observed that for T to be weakly precompact

it is not sufficient for T to be weakly precompact. * X

The following example may be given.

29. Example. If T:C(H,X)-»Y is a bounded linear operator # • #

so that T |u (J]) 1 8 weakly precompact, then T need not be I X

weakly precompact.

Let {p} denote a set containing one point. Note that

C({p>,co) = co. Then U j £ ) - C q. Let I:C<{p},co) -4 C q be

the identity map. Observe that I and I**|u are weakly

co precompact, but I is not.

Page 51: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

CHAPTER IV

WEAKLY PRECOMPACT SUBSETS OP L^/^X)

In [GS] N. Ghoussoub and P. Saab characterize Banach

spaces X for which weakly compact subsets K of 1^(0,X) are

defined by the following three properties:

(i) K is bounded.

(ii) K is uniformly integrable.

(Hi) for each E € S, the set {/Efd/t:f€K> is relatively

weakly compact.

This characterization may be stated as follows.

1- Tijeofeig. [GS, Corollary 3] A Banach space I and its

dual X have the ladon-Nikodym property if and only if for

every finite measure space any hounded and uniformly

integrable subset I of Lj(n,I) is relatively weakly compact

whenever for every B G E, the set is relatively

weakly compact in /.

In this chapter we will consider weakly precompact

subsets K C ,X) in terms of conditions analogous to (i),

(ii), and (iii). We begin by considering the following

lemma. The arguments for Lemma 2 and Theorem 3 are similar

to those for Lemma 3.2 and Proposition 3.1 in [BrL]. Let

47

Page 52: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

48

f:0-4X and g:fl-+x . in the discussion which follows, <f,g>

will be the function mapping 0 to R defined by <f,g>(W) =

<g(w),f(w)> for w € 0.

2. IiSfflia. Suppose that I is a Ban&ch space so that X*

has ike iadon-jfikoiym property with respect to ( U I f

( f j is a sequence from L j ( n , I ) so that (*) < °°>

19

(**) {fB:»€ll} is uniformly integrable,

and

is weakly Cauchy for all A € E,

then ( f n ) is weakly Cauchy in L j f p , ! ) .

P r o o f ; Suppose that X has the Radon-Nikodym property,

(Q,£,/t) is a finite measure space, and (f ) is a sequence

from L1(/j,X) satisfying (i), (ii), and (iii). Let (fn) be a

sequence from K. Since X* has the RNP, L^p.X)* « L (p.X*)

[DS, IV. l.l]. Let g € L^ (/i, X ). Let sn be a sequence of

simple functions so that sn->g a. e. p and ||sn(t)|| < 2||g(t)||

for each n [DS, III.3.8].

Let e > 0. Since {f^} is uniformly integrable, there

exists a S > 0 so that if p (A) <6 for A € E, then JJfJd/j <

e. By Egoroff's Theorem, there exists A € £ so that ft (A) < 6

and sn->g uniformly on 0\A. Let A be one such element of £.

Pick N 6 0 so that for all n > N,

llsn(t) " 9(t)|| < <r

Page 53: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

49

for all t€fi\A. Since (fR) satisfies ill, pick I G IN so that

if i and j > I,

< e*

Let i and j > I; then

|/0<f1,g>d/i - S a < t y g > f y \

- Un < fi , s N" g > d''l + + Uo < fj ' s N " 9 > d' (l

* I W ^ r V 3 ^ ! + |/*<fi'8H-B><V| + (

+ IW<frV0>d',l + < cl!filli+ »«IWL + £ + 'H'jB,+ "IWL-

Thus (fn) is weakly Cauchy in L1(^,X).

We may now characterize weakly precompact subsets of

L1(p,X) for Banach spaces X whose duals have the

Radon-Nikodym property.

3. Theoyqm. Let I be a Banach space so that I* has the

laion-N ikoiym property. If I C Lj(n,X) then I is uniformly

iategr&ble if and only if [ is weakly precompact .

Proof: Let (f^) be a bounded sequence from K. Define

</*n) bY

VA> - /Afn*'

Since K is uniformly integrable, it follows that (a ) is r n

uniformly countably additive. Construct a countable

subalgebra x - {^j of E so that ^JfO) - J ,0) for ail i j w4

n € N. Let £ be the *-algebra generated by A. Define

Page 54: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

50

I I: Span{/t n:n€NHca ( E \ x ) by D(y) - Clearly II Is an

isometry. Since X has the Radon-Nikodym property, and hence

the weak Radon-Nikodym property, by Theorem III.l and Theorem

III.11 we see that for A € £, {/*n(A) :n€N} is weakly

precompact. Thus (/t ) has a subsequence (/t ) so that nk k

^Hfc^i^k l s w e a k l Y Cauchy for all i € N. Let x* € X*.

Hence lim x*(ft (A)) exists for all A € A. Hence by k-ta> nk *

[DS,IV.8.8], lim x (a (A)) exists for all A G E1 . Therefore k-to k

*^n = ^ A f n l s w e a k l Y Cauchy for all A € S* . Thus by ic k

Lemma 2, (f^) is weakly Cauchy in L 1(p^ l,X). Since II is an

isometry, (fR ) is weakly Cauchy in L (p,X). k 1

For the converse, let K be a weakly precompact subset of

Lj (/*, X). For f € L 1(/t,X), define vf € ca(£,X) by v (A) =

^ ' S e* K' = Since K is weakly precompact, the

same may be said of K1. Applying Lemma III.26, we see that

K1 is uniformly countably additive. The uniform countable

additivity of K1 implies the uniform integrability of K.

The conclusion in Theorem 3 may not be obtained for

Banach spaces X for which X lacks weak Radon-Nikodym

property, for if every weakly precompact subset of L1((i ,X) is

uniformly integrable, then X* must have the weak

Radon-Nikodym property. To see this, suppose that X is a

Banach space so that uniformly integrable subsets of L1(/t,X)

Page 55: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

51

are weakly precompact. Note that if (xn) is a bounded

sequence from X, then (*Q*xn) is uniformly integrable in

L1(/t,X). Thus (XQ'X^) has a weak Cauchy subsequence

(*ft'xn }* However, this produces the subsequence (x ) which nk

is weakly Cauchy in X. Hence we conclude that X* possesses

the weak Radon-Nikodym property. Thus if X* does not have

the weak Radon-Nikodym property, then L± <t ,X) must have a

subset which is uniformly integrable and not weakly

precompact. In fact, by extending Example IV.2.2 in [DU] we

may make a stronger statement about those spaces X for which *

X does not have the weak Radon-Nikodym property.

4. Theorem. Suppose that X* does not have the weak

iadon-Kikodym Property. There exists a finite measure space

and a set I C £j(p,X) so that

( t ) i is bounded,

( i i ) I is uniformly integrable,

( H i ) { f £ f d p : f € [ } is relatively compact for all B € E,

and

(iv) I is not weakly precompact.

P r Q o f : Suppose X is a Banach space so that X* does not

have the weak Radon-Nikodym Property. Thus by Theorem

III.11, X contains a copy of I v Let (xn) be a copy of the

l± basis in X. Without loss of generality, assume ||xn|| < l

for all n € N. Since (xn) is a copy of the l% basis, we may

Page 56: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

52

choose ^ € R so that

n n

* '£1-11-for every n € II and choice of scalars (a } n

i i=l*

Let 0 = [0,1], E = #([0,1]), and p, = Lebesgue measure.

Use r n to denote the nth Rademacher function on [0,1]; that

i s' r n ^ = s9n(sin(2n*t)). (Take sgn(0) = l.) For n € W,

define fn:[0,l]4X by f R - r ^ . Let K « <fn:n€N).

Observe that ||fj = J|fn|<^ < 1. In addition, lim llZ-fd^l

n-ta» '

0 for all E € S. Thus the set K = {fn:n€N) satisfies

conditions (i) and (Hi). Furthermore, as |j^n(^)J| ^ 1 for

all n € N and t € [0,1], K is uniformly integrable. Let

C IR. Observe that

> f/l £ |a1r1|d/» • ^ S |aA i=l i=l

Thus (fR) is a copy of the basis. Therefore K is not

weakly precompact.

Theorem 3 and Theorem 4 leave several questions

unanswered. First, may Banach spaces whose duals have the

Radon-Nikodym property be characterized as those Banach

spaces X for which weakly precompact subsets of L 1(/<,X) are

exactly those subsets of L^(^i,X) which are uniformly

integrable? If not, how may one characterize such spaces?

In addition, the question as of whether or not Banach spaces

Page 57: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

53

having the weak Radon-Nikodym property may be characterized

as those spaces X for which subsets K of LjL(/(,X) satisfying

<i), (ii) and (iii) of Theorem 4 are weakly precompact

remains.

In [GS], N. Ghoussoub and P. Saab showed by construction *

that if X fails to have the Radon-Nikodym Property, then

there exists a finite measure space (fl,E,ft) and a subset K of

L1(/t,X) so that K is bounded, uniformly integrable, and

{ J E f d / t : f € K } is weakly compact for all E € E, but K is not

weakly compact. The question as of whether or not their set

K is weakly precompact arises. Suppose that X* has the

Radon-Nikodym Property. The following is a sketch of the

construction in [GS]. Let A denote the Cantor group with

Haar measure /, and let {A^ . s l<i<2n> denote the standard nth

partition of A. Set A = A. Thus A . = A U °'1 n,i n+1,2i-l

An+l,2i 1 8 c l°P e n' and p(An ..) = 2 n. We use x n d to denote

*a . The identity operator i:x-« factors the Haar operator

H. I ) which takes the basis of I into the usual Haar

basis of C(A) as a subspace of LJp). That is, there exist

bounded linear operators and V:X-»Lm(ft) so that H =

VOIOU. If {enl:n>0,l<i<2n} is an enumeration of the usual I

1 basis H may be defined as follows:

H ( e . ) = y — y — K ni' *n+l,2i-l *n+l,2i ~ hni

Set

f n ( t ) a n hji ( t ) eji' t € j=l i=l

Page 58: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

54

In [QS] the sequence (fn) was shown to be bounded and

uniformly integrable. Furthermore, it was shown that

{/Afnd/»:n € IN) is weakly compact for all A € E.

For an operator T:X->Y, define T:!^ (/»,X)-+L1 (ft, Y) by

<T(f)(t)) = T(f(t)). Then

H(f )(t) - i S S L J t l h , , , t € A. j=l i=l 3 1 3 1

Note that for t € C{A), H(fn)(t) C c(A). Ghoussoub and Saab

show that {I(fn):n€ll} is not weakly compact by demonstrating

that {H(f n):n€ll} is not weakly compact. Unfortunately, the

same technique may not be used to determine whether or not

{fn:n € N} is weakly precompact, as <H(fn):n € N} is weakly

precompact. To see this recall that a bounded sequence (xn)

in a Banach space is equivalent to the basis if and only

if there exists a real number $ > 0 so that for any fc c 1' 2'

. . .,cn> C R

m m

'n=l|Cnl " l^Vnll Let {c1, . . c } C R. Let t € A. Note that t 6 A D

0,1

l.Pj"5 A2,p 3 * ' * f o r s o m e sequence of positive integers A, ^ 3 . 2

(Pk). Thus

Im f n 2^

s c [i S E h (t)h,,(«)] I wta n=l j=i i=i J1 31 1

Page 59: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

55

I m j- n 2

2 ^ 1 C " > 1 i»i i X3 + 1. 21-1' *' -*J+1.21'l) 1

I m r n = sup S c I - S w€A n=l n S = l

(X. _ i (t) -AT. j = 1 j+l,2p.-1*~' *j+l,2p. (t))

Let q be the smallest non-negative integer so that w € A

and w £ A q + , then

q + 1

q»p ' q

(X j+l , 2 pi-i ( t )" J f j + 1' 2 pi ( t ) )' Uj+i,2p. ^ " ^ j + i ^ p . ( w ) )

1 if j+l < q+i

-1 if j+l = q+l

0 if j+l > q+i

Thus

S c S(f )(t) n=l " n

k m S c

n + (k-1) E c3>? n=l i=k+l l l

for some k G {0, 1, . . ., m). Let (fn ) be a subsequence of

(fR)- Set c - (-l)s and for all n £ n for s s some s set c =

n

0. Thus

£ c i?(f ) s=l ns V * / I L V » 8 ( f n 'H* - 2-

S=1 s s

Page 60: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

56

S. However 2LJ|g | = r. Hence (H(fn )) is not equivalent to the

s-i s s Jj basis. Thus (H(fn)) is weakly precompact.

An essential part of the proof of Theorem 1 is the

following characterization of Banach spaces X whose duals

have the Radon-Nikodym property.

5. Theorem. TDU, IV.1.1] let be a finite measure

space, 1 < p < oo, and I a Banach space. Then I (pfX) -

£ (H,I ), where p * q = 1, if and only if / has the

iadon-ff ikodym property with respect to p.

A similar characterization may be made for Banach spaces *

X for which X has the weak Radon-Nikodym property.

6. Theorem, let I be a Banach space and a

finite measure space. The space I* has the weak

iadon-Jf ikodym property with respect to p if and only if for

each I € l^(p,X) , 1 < p < oo, there exists a Pettis

integrable function so that

L(f) * for f € l^(/t,I),

where <f,g>(m) = (9 (<*))( f (<»)) for w G ft .

Proof: Let X be a Banach space and (ft ,£,/<) a finite

* measure space. Suppose that X has the weak Radon-Nikodym

Page 61: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

57

property with respect to p. Let L € L Define G on £ P

by

G(E)(x) = L(*eX) for E € S.

Thus

|G(E)(X)| - |L U EX,| < ||L||.||V||p - ||L||-||,e||p-||x||,

• and consequently G:E-»X .

To see that G is ft continuous and of bounded variation,

let E€S, {Ej, . . E n) a partition of E, and {xJf . .

x } C B . Thus n j\

n n I S G(E.){x.)| < IL( £ G(E,)(x.)J i-1 1 1 1 ' i=l 1 1 1

* IIHI• IIj/E^illp - W'/nlF^/if'5

*

Since X has the weak Radon-Nikodym property with respect to

there exist a Pettis integrable function g-.TrtX* so

that G(E) =* P - J gd/» for all E € E. If f is a simple

function, then L(f) - Jft<f,g>d/i. Let f € L (/i,X) and let tr (f ) be a sequence of simple functions so that f -+ f a. e. 11 n

on 0 and so that |fnU)| < 2|f(«)| for all u € 0. Note that

fft<fn'g> = L ( fn ) L ( f ) * n

Define fR(A) = JA<fn,g>d/t. Observe that

*n(A) = L(^Afn) -f L(f) .

Page 62: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

58

In addition fn«/» for all n. Thus by [DS,p.l58],

lim ^„(E) = 0 /J( E)->0

uniformly in n. Let c > 0, and choose S > 0 so that if p(A)

< S, then fn(A) < e for all n € N. Choose N so that

M<t:||f(t)||>N) < s . Let B = < t: ||f (t) ||<N>. Hence

/fl|<fn.9>|^ » /B'=l<fn-9>ld' + /Bl

<fn'9>ld''

< £ + ' BI< S*A <

xni' 9 >l d' ni

< € + sup{JB £ <^Exi,g>:||x1||<2N, 11(B) partitions B} EGII ( B )

< t + suP^ ||/g<^^'(B^Afgxjd/'||: H ^ i l l — H ( B > >

• e + 2N|6|(B).

Since f n -» f a. e. , <fn>9> <f,g> a.e. . Certainly <fn»g>

is measurable for all n as f is measurable for all n.

Furthermore | | < 2|<f,g>|. Hence by Fatou's lemma,

/ | <f»g> | d/t < JJji J| <fn,g>|d/t.

Now applying the Lebesgue convergence theorem we obtain

lim/<f ,g>d/t = J<f,g>d/t. n-to

Therefore L(f) = J < f , g > d / t .

For the converse, suppose that for every L € L (p,X) P

there exists a Pettis integrable function g:£->X* so that L(f)

Let G:£-« be a /» continuous vector measure of bounded

variation. Let Eq € E so that ^(E ) > 0. It will be shown

that G has a Pettis integrable weak Radon-Nikodym derivative

on some B€S, bCEq, with /t(B)>0. A version of the Exhaustion

lemma will complete the proof [DU,III.2.5]. Choose k € IN so

Page 63: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

59

that JG| (EQ) < K/T(EQ). Apply the Hahn decomposition theorem

to Kp - |G| to obtain BGE so that BCEq, /t(B) > 0, and |G|(E)

< k/j (E) for E € E, E C B. For f = x,, where x. € X, E. G i

E, and E^^Ej = t for i£j, define

L(f) - EG(EinB)(x1) .

Thus

XG(E FLB)

(i (E.flB)x. /MEJlB)

1

< Ek||^{(EinB)(xi)|| < ^jf^ < k/»(fl)q||f||p.

Thus since L is linear on simple functions, L defines a

continuous linear functional on the simple functions of

Lp(/t,X). Hence L has an extension to all of L (£,X). By

hypothesis there exists a Pettis integrable function G so

that L(f) » J<f,g>dm for all f € L (/t,X). Now

G(EflB) (x) = " JEnB<x,g-d/»>

for all x € X. Therefore G(EOB) = P - /EpB9<3/».

Page 64: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

BIBLIOGRAPHY

[BDS] R. G. Bartie, N. Dunford, and J. Schwartz, Veah Compactness and Vector Measures, Canad. J. Math. 7 (1955), 289-305.

[BB] J. Batt and E. J. Berg, linear Bounded Transformations on the Space of Continuous Functions, J. Func. Anal. 4 (1969), 215-239.

[BiL] R. G. Bllyeu and P. Lewis, Some Mapping Propert ies of iepresent ing Measures, Ann. Mat. Pura Appl. 109 (1976), 273-287.

[BC] F. Bombal and P. Cembranos, Characterization of Some Classes of Operators on Spaces of Vector-Valued Continuous functions. Math. Proc. Cambridge Philos. Soc. 97 (1985), 137-146.

[Bour] N. Bourbaki, Integration, chap. I-VI, Hermann, Paris, 1952-1959.

[BrL] J. K. Brooks and P. W. Lewis, Linear Operators and Vector M e a s u r e s , Trans. Amer. Math. Soc. 192 (1974), 139-162

[DFJP] W. J. Davis, T. Figiel, W. B. Johnson, and A. Pelczynski, Factoring Veakly Compact Operators , J. Funet. Anal. 17 (1974), 311-327.

[Di] N. Dinculeanu, Vector Measures, Pergamon Press, Oxford, 1967.

[DU] J. Diestel and J. J. Uhl, Jr., Vector Measures, Amer. Math. Soc. Providence, 1977.

[Dol] I. Dobrakov, On Integration in Banach Spaces, I, Czech. Math. J. 20 (1970), 511-536.

[Do2] I. Dobrakov, On iepresentat ion of Linear Operators on C o ( T , I ) , Czechoslovak Math. J. 21 (1971), 13-30.

[DS] N. Dunford and J. T. Schwartz, Linear Operators Part I: General theory, Interscience, New York, 1957.

[Ed] R. E. Edwards, A Theory of Madon Measures on Locally Compact Spaces, Acta Math. 89 (1953), 133-164.

[G] I. Gelfand, Abs trakte Funkt ionen und lineare

60

Page 65: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

61

Operatoren, Mat. Sb. 4 (1938), 235-384.

CGS1 N. Ghoussoub and P. Saab, Jleak Compactness in Spaces of Bochner Integrable Functions and the iadon-Nikodym Property, Pacific J. Math. 110 (1984), 65-70.

• i [Go] M. Gowurin, liber die St iel tjessche Integrat ion abstrakter Funkt ionen, Fund. Math. 27 (1936), 255-268.

[Ha] P. R. Halmos, Measure Theory, van Nostrand, Inc, Princeton, 1950.

[He] E. Hewitt, Integrat ion on Locally Coup act Spaces I , University of Washington Publ. in Math. 3 (1952), 71-75.

[Jam] R. c . James, A Separable Somewhat Reflexive Banach Space with Ifonseparable Dual, Bull. Amer. Math. Soc. 80 (1974), 738-743.

[Jan] L. Janicka, Some Measure-Theoretical Characterization of Banach Spaces Hot Containing l J t Bull.

Acad. Polon. Sci. 27 (1979), 561-565.

[K] S. Kakutani, Concrete Representation of Abstract (M)-Spaces (A Characteri zat ion of the Space of Continuous Functions) , Ann. of Math. (2), 42 (1941), 994-1024.

£Lel P« Lewis, Some Regulari ty Conditions on lector Measures with Finite Semi-lariat ion, Rev. Roumain Math. Pures Appl. 15 (1970), 375-384.

[LSt] J. Lindenstrauss and C. Stegall, Examples of Separable Spaces which do not Contain l j and Vhose Duals are

Non-Separable , Studia Math. 54 (1975), 81-105

[Lo] R. H. Lohman, A Vote on Banach Spaces Containing Ij, Canad. Math. Bull. 19 (1976), 365-367.

[Mu] K. Musial, The Veak iadon-Nikodym Property in Banach Spaces, Studia Math. 64 (1979), 151-173.

[Ra] J. Radon, Theorie und Anwendungern der absolut additiven Mengenfunktionen , S. B. Akad. Wiss. Wien, 122 (1913), 1295-1438.

[RSU] L. H. Riddle, E. Saab, and J. J. Uhl, Jr., Sets with the Veak iadon-Nikodym Property in Dual Banach Spaces, Indiana Univ. Math. J. 32 (1983), 527-540.

[R] . F- Riesz, Sur les Operations Fonct ionnelles Lineaires, C. R. Acad. Sci. Paris, 149 (1909), 974-977.

Page 66: N81J - Digital Library/67531/metadc331171/m2/1/high_res... · N81J /Vc? ia&SS'S OPERATORS ON CONTINUOUS FUNCTION SPACES AND WEAK PRECOMPACTNESS DISSERTATION Presented to the Graduate

62

C R N 1 r.«o'pR1BSZ a n dB ,

B * JS 2 " N a g y ' factional Analysis,

YorkT 1955? B o r o n ' Frederick Ungar Publishing Co., New Trans.

^ Rosenthal, A Characterization of Banack Spaces Containing l J f proc. Nat. Acad. Sci. USA, 71 (1974) 2411-2413.

"riiz'luir:/ l " i s e i

Y o r k ' D o v e r

f S 1 J I* s l n Srer , Les Duals de Certains Espaces de Banack

\ U i i ] ? U t \ 0 [e c U n r a ' B u l 1- d e s Sciences math. 2° serie, 82

c St®?a11' T h e t*don-#ikod9m Property in Coniuaate Banack Spaces. II, Trans. Amer. Soc. 264 (llei)? 507-Zi 9

[Swl] C Swartz, Unconditionally Converqinq Operator*

g a t o r s °on T p T e ^ o f V e \ u l \ \ ^

Trans. Amer. Math. Soc. 179 (1973), 123-131 F * n c t t o n s > £ J b L « m „ « « D / H* T uf k e r'/ Representation Theorem for a Continuous Linear Transformation on a Space of Continuous F u n c t i o n s , Proc. Amer. Math. Soc. 16 (1965), 946-953