Lecture Notes in

Mathematics
Edited by A. Dold and B. Eckmann

752
Michael Barr

*-Autonomous Categories
With an Appendix by Po-Hsiang Chu

Springer-Verlag
Berlin Heidelberg New York 1979

Author
Michael Barr
Department of Mathematics
McGill University
805 Rue Sherbrooke Ouest
Montreal, P.Q./Canada H3A 2K6

AMS Subject Classifications (1980): 18A35, 18 B15, 18 B30, 22 B99,

46A12, 46A20, 46M05, 46M10, 46M15, 46M40
ISBN 3-540-09563-2 Springer-Verlag Berlin Heidelberg NewYork
ISBN 0-387-09563-2 Springer-Verlag NewYork Heidelberg Berlin
Library of Congress Cataloging in Publication Data
Barr, Michael.
*-Autonomous categories.
(Lecture notes in mathematics ; 752)
Includes bibliographies and indexes.
1. Categories (Mathematics) I. Title. I1. Series: Lecture notes in mathematics (Berlin) ; ?52.
QA3.L28 no. 752 [QA169] 510'.8s [512'.55] ?9-21746
ISBN 0-0-38?-09563-2
This work is subject to copyright. All rights are reserved, whether the whole or
part of the material is concerned, specifically those of translation, reprinting,
re-use of illustrations, broadcasting, reproduction by photocopying machine or
similar means, and storage in data banks. Under 54 of the German Copyright
Law where copies are made for other than private use, a fee is payable to the
publisher, the amount of the fee to be determined by agreement with the publisher
by Springer-Verlag Berlin Heidelberg 1979
Printed in Germany
Printing and binding: Beltz Offsetdruck, Hemsbach/Bergstr.
2141/3140-543210

PREFACE

The category of finite dimensional

many interesting properties:
known as autonomous)
the functor

dimensional

It is a symmetric closed

category which has an object

(-,K), internal Hom

opposite category.

vector spaces over a field

into

has

monoidal (hereafter

K, with the property that

K, induces an equivalence with its

Similar remarks apply to the category of finite

(real or complex) banach spaces.

We call such a category *-autonomous.

Almost the same thing happens with finite abelian groups, except the "dualizing
object",

~ /~

or ~ / ~ , is not an object of the category.

In no case is the

category involved complete, nor is there an obvious way of extending both the
closed structure and the duality to any of the completions.
phenomena,

In studying these

I came on a fairly general construction which allows you to begin

with one of the above categories

a complete and cocomplete

(and some similar ones) to embed it fully into

category which admits an autonomous

which, using the original dualizing

In an appendix,
embeds any autonomous

object,

structure

is *-autonomous.

my student Po-Hsiang Chu describes a construction which

category into a *-autonomous

described is not, however,

full and is completely

category.

The embedding

formal.

The work described here was carried out during a sabbatical

McGill University,

I was at Universitetet

partially

leave from

academic year 1975-76 mostly at the Forschungsinstitut

Mathematik der Eidgenossische

(Louvain-la-Neuve)

and

Technische Hochschule,

Z~rich.

i Aarhus as well as l'Universite

f~r

For shorter periods

Catholique

de Louvain

and I would like to thank all these institutions.

supported during that year by a leave f e l l o w s h i p f r o m

I was

the Canada Council

and received research grants from the National Research Council and the Mlnlstere
de l'Edueation

du Quebec.

Preliminary versions of part of this material has been published

five papers by me listed in the bibliography.

in the

The current version was presented

in a series of lectures at McGill in the Winter Term, 1976.

TABLE OF CONTENTS
PREFACE
CHAPTER I:

PRELIMINARIES

i.

Symmetric

2.

Uniform Spaces

3.

Uniform Space Objects

4.

*-Autonomous

13

CHAPTER II:

Closed Monoidal

Categories

Categories

EXTENSIONS

OF STRUCTURE

i.

The Setting

19

2.

Extension of the Duality

21

3.

Extension of the Internal Hom

25

CHAPTER III:

THE CATEGORY G.

i.

Completeness

2.

Definition and Elementary Properties

3.

The Closed Monoidal Structure on

4.

Summary of the Hypotheses

CHAPTER IV:

33

of

37
41
42

EXAMPLES

i.

Vector Spaces

43

2.

Dualizing Modules

49

3.

Banach Spaces

52

4.

Modules over a Hopf Algebra

59

5.

Topological Abelian Groups

65

6.

Semilattices

96

BIBLIOGRAPHY

i01

APPENDIX
CONSTRUCTING

*-AUTONOMOUS

CATEGORIES

Po-Hs iang Chu

CHAPTER I:

CHAPTER II:

CONSTRUCTION OF

I.

The Category _ ~

2.

AX

is Enriched

CHAPTER III-

103

PRELIMINARIES

_AX

Ax

AND ITS ENRICHMENT

OVER

V
104

over

HAS A *-AUTONOMOUS

105

STRUCTURE

i.

The Hom-Functor A X (-,-)

113

2.

The Functor

122

3.

The Functor

- -

126

4.

The Dualising

CHAPTER IV:

Object and the Unit for Tensor

128

APPLICATIONS

i.

Functor Categories

129

2.

The Double Envelope

130

3.

Miscellaneous

131

Results

B IBLIOGRAPHY

138

Index of Definitions

139

Index of Notation

140

CHAPTER

i. S y m m e t r i c

(i.i)

A symmetric

i)

closed

A category
A functor

(tensor)

iii)

A functor

(internal

iv)

An object

I of

Natural

vii)

r = rV

: VI

= V

: IV

+ V

Natural

: (I,V)

tity

that

on

hom)

V(-,-):~op

x~

(V,(V',V"))

of these

equivalences
: VV'

+ V'V

: (V,(V',V"))

(V', (V,V"))

transformation
: I +

(V,V)

transformations

c = c(VrV',V")

: (V',V")(V,V')

d = d(V,V',V")

: (V',V")

e = e(V,V',V")

: (V,V')

subject

to a g r e a t

H o m ( V , V ' ) ~ H o m ( I , (V,V')
and that

important

+ V(v'ev")

are

data.

data

following

: (VV',V")

Natural

o f the

: (VV')V"

j = jV

quires

consists

a = a(V,V',V")

A natural

These

--:~x~

+ V

t = t(V,V',V")

(1.2)

Equivalences

s = s(V,V')

ix)

category

Categories.

p = p(V,V',V")
Natural

viil)

Monoidal

Equivalences

i = iV
vi)

PRELIMINARIES

ii)

v)

Closed

monoidal

I.

(V,V")

((V,V'),(V,V")-)
((V',V"),(V,V"))
many

in s u c h

axioms.
a way

The most

that

jV

corresponds

to t h e

re-

iden-

the diagram

Hom(I, (V',V")

Horn(I, ( V , V ' ) ) ~ H o m ( V ' , V " )

x Hom(V,V')

1
H o m (V, V" )

H o m (II, (V' , V " ) ( V , V ' ) )

commutes.

Here

the bottom

arrow

ternal

uses

composition
The

deal

the left hand

while

remaining

of redundancy

rI

the upper

axioms
among

arrow

(which,

is j u s t

an instance

one of the coherence

arrow

o n the r i g h t

a r e of t w o k i n d s .

t h e data.

H o m ( I , (V,V"))

For example

Some
<p>

of the
rules

functoriality

states,

is e x t e r n a l
express

the

of

= I)

a n d in-

composition.
fact

there

is an i s o m o r p h i s m

is a g r e a t

Hom(VV',V")

+ Hom(V, (V',V"))

< p((V',V"),
similarly

(V,V'),

b y means of

((vv')v",-)

and one l a w says t h a t

(V, V"))>o

(c(V,V',V"))

p, c a n d s .

(vv'(v",-))

axiom which

Another

expresses

(v,(v',(v",-)))

The o t h e r k i n d of axioms
mous pentagonal

= d(V,V',V")

v(v'(v"v"') )

Another
in terms

(v,(v'v",-))

are the c o h e r e n c e

expresses

axioms

the c o m m u t a t i v i t y

expresses
of

(v(vv"),-)

examplified

by MacLane's

fa-

of

(vv')(v"v"' ) ( (vv')v")v"'

V ( (V' V" )V'") (V (V' V") V"'

(1.3)

A category

such t h a t

is e n r i c h e d

Hom(I,v(A,B))

composition

over

is n a t u r a l l y

if there

equivalent

is a f u n c t o r

to

Hom(A,B)

V(-,-):AOPxA

+ V

Also required

is a

map

V(B,C)V_(A,B)
lying above the c o m p o s i t i o n
are t a b u l a t e d

of m o r p h i s m s

in 1.5 of [ Eilenberg,

in

+ V(A,C)

Kelly] .

as w e l l as m a n y c o h e r e n c e

--:VxA

is a functor

such t h a t

Provided

has,

by the a d j o i n t

and

-A:~

+ A

~(A,-)

functor

has a r i g h t adjoint,

is,

commutes

theorem.

denoted

for e a c h
with

Analogously

[-,A] :V Op

axioms w h i c h

A tensor

A(A

limits,

, left a d j o i n t
this

it f r e q u e n t l y

+ A

to

~(A,-) :A V

can u s u a l l y be shown to exist

happens

(note the s w i t c h

that

~(-,A):A__ p

in variance)

w h i c h deter-

mines a b i f u n c t o r

[-,A] :V p x A
c a l l e d a cotensor.

We h a v e

Hom(V,V(A,B))

(1.4)

Let

a function

S
Sn

be a set a n d
+ S .

If

a m o d e l of or a l g e b r a
n
wS:S n +
S
for e a c h
~
such that for all

+ A

~ Hom(VA,B)

be a cardinal

~ Hom(A,[V,B] ) .

number.

An

n-ary o p e r a t i o n

on

is

~ = {~ } is a class g r a d e d b y the c a r d i n a l s
n
of sets
n
for
~
is a set
S e q u i p p e d w i t h an
n-ary operation
A morphism

n , all

Sn
fn

f:S + T

of m o d e l s

of

is a f u n c t i o n

, the square
mS

~T

1
Tn

commutes.
(1.5)

a m o n g the e l e m e n t s

in
n

are a s s u m e d

to be c e r t a i n

~. , iEn
1

where

realization

in any a l g e b r a

is the p r o j e c t i o n

operation

whose

and

to the ith coordinate.

is an n - a r y operation,

value on any a l g e b r a

t h e n there

The c a t e g o r y

of a l g e b r a s

most readily be formalized

maps

include

(wi) ~

and morphisms

by building

all the f u n c t i o n s

In p a r t i c u l a r

A full

quotients
example

inverse

limits.

category

of a v a r i e t y

algebras

in the variety.

If

under quotients.

is itself
F

hence

FV

object

Thus

in

W'

in w h i c h

understand

(1.9)

is free
V ---+

Now suppose
w ,

W'

in

W'

in

W'

the free

the closure

of a

is a q u o t i e n t

VEV

is unique

Since
V

and

s h o w that

Henceforth
V

of

W'

since every

The same a r g u m e n t w o u l d

and t h a t

of

so that

is a s u b o b j e c t

variety

but
closed

to be a full sub-

of quotients

is also unique.

a containing

and

is an

contains

of a s u b o b j e c t

. Actually

groups

W'

it is a q u o t i e n t

and h e n c e

products

that it h a v e these limits

we m a y c o n s i d e r

that a product

[Linton]

of a v a r i e t y

and which

is a q u o t i e n t

is e m b e d d e d
W

merely

trans-

and p r o d u c t s

a semi-variety

and equalizers

splits

of an o b j e c t

a semi-variety

W
=

We d e f i n e

in

and

free a b e l i a n

to be a full s u b c a t e g o r y

argument

is a s e m i - v a r i e t y

of

can
whose

is a c o n t r a v a r i a n t

is a n a t u r a l

u n d e r subalgebras,

of a torsion

in a v a r i e t y

to b e a s e m i - v a r i e t y ,

the free a l g e b r a s

An a l g e b r a

of algebras

is c l o s e d

of a q u o t i e n t

FEW'
map

is a q u o t i e n t

the v a r i e t y

operation

If

This n o t i o n

are cardinals,

closed under subalgebras

under p r o d u c t s

It is a s t a n d a r d

the q u o t i e n t

of i.

t h a t w e are not r e q u i r i n g

and a subobject

a variety.

is free,

objects

m a y be found in [ L a w v e r e ]

subcategory

is a q u a s i - v a r i e t y

of the p r o d u c t

Details

in the variety.

closed

whose

and a morphism

a quasi-variety

Notice

The c a t e g o r y

full

b u t that they be the limits

(1.8)

functor

is a g a i n a variety.

We define

mi-ary

m = Zm. ,
i

of a variety which

of a n o n - v a r i e t a l

is an

and such t h a t

such functors.

subcategory

n o t quotients.
under

set v a l u e d

is c a l l e d a variety.

is the s u m of t h a t m a n y copies

formation between
(1.7)

Sn

a category

H o m (re,m) =

product preserving

(~i) , in

is the c o m p o s i t e
S m ~ ~.smi
1

(1.6)

If

is an m - a r y operation,

w e will
contains

t h a t the t h e o r y

m-ary operation

is c o m m u t a t i v e .

That means

and algebra with underlying

set

that for a n y n - a r y
S , the square

n
S n x m ~ Sm x n ~

Sn

n
6d
Sm

commutes.

The

as a c t i n g on row vectors,

and

respectively

isomorphism

in the u p p e r

a n d the i s o m o r p h i s m

of this square

is t h a t

l e f t is m o s t e a s i l y

Snm

and

as transpose.

is a h o m o m o r p h i s m

Sm x n

as

described by thinking
nxm

and

The i m p l i c a t i o n

with respect

to

mxn

of

matrices

of the c o m m u t a b i l i t y
(or v i c e versa).

If this h o l d s

for all

on the u n d e r l y i n g
(i.i0)

Suppose

and

are algebras,

operation,

~,~

the result

is t h a t each o p e r a t i o n

sets b u t is in fact a m o r p h i s m

in a d d i t i o n

that

(fi),

is a v a r i e t y

is a g a i n a m o r p h i s m .
diate

that

that

~(V,-)
I

closed

category.

FS

But since

(1.12)

Then

category.

FT

VE~

and it is imme-

(V',V)

It is easily

theorem,

seen

has an a d j o i n t

T h e n the r e q u i r e d

natural

-V

trans-

as the c a n o n i c a l

closed

structure

- but

In fact,

let

for a c o m m u t a t i v e
~0

theory

be the c a t e g o r y

of al-

as above a n d w e w i l l
W~

Then

see

has a

to

FT--~

on the sets

T h e n there

W(W,V)

w(w,v)

and

respectively,

and,

as

is an e q u a l i z e r

W(FT,V)

is a q u a s i - v a r i e t y ,

and

_W(FS,V)

',

vs

VT

and

Vs

and hence

W(W,V)

all b e -

V.
Whether

o r n o t the theory

Here

structure

b e the c a t e g o r y

~ H]-modules.

regardless
of K - l i n e a r

are the free a l g e b r a s

(1.7), b e l o n g

is c o m m u t a t i v e ,

is the e x a m p l e w h i c h

For

W
has the s t a n d a r d s t r u c t u r e
--0
ideal.
In fact, let
V~
and

closed

as

an n-ary

to s h o w that we have a s y m m e t r i c

is o n l y a s e m i - v a r i e t y

d a r d one.
V

theory.

metric monoidal

and

and

V'

If

as a c o e q u a l i z e r

and
in

long to

theory.

on Hom(V',V)

functor

on one generator.

FS

observed

a function

an a l g e b r a w h i c h we call
and, b y the a d j o i n t

it into a c l o s e d

is an e x p o n e n t i a l

presentation

where

V' --+ V

~(V',V)

This m a y be d e s c r i b e d

in case t h a t

for the theory.

m a y e a s i l y be c o n s t r u c t e d

to a c o m m u t a t i v e

Even

Vn

an o p e r a t i o n

be the free a l g e b r a

w e can still m a k e

that

limits

and e q u i v a l e n c e s

corresponding

gebras

is again

preserves

let

(i.ii)

This d e f i n e s

Hom(V',V)

formations
monoidal

for a c o m m u t a t i v e

a family of m o r p h i s m s

ion

is not m e r e l y

the c o m p o s i t e
(fi)
V' ~

Finally

in

V'V
is just

there

on the s e m i - v a r i e t y
interests

of r e p r e s e n t a t i o n s

maps w i t h

symmetric

~ action

with

each g r o u p e l e m e n t

o b j e c t on one g e n e r a t o r

which means

which
H

on K-vector

structure.

with

for

t h a t there

differs

spaces,
according

For
x~H,

x(v~v')

t h a t the f u n c t o r

as

f:V'--+ V

= xvxv'.

Hom(I,-)

a field

otherwise

~(V',V)

a c t i n g by the i d e n t i t y map.

is a sym-

f r o m the stan-

be a group,

is c o m m u t a t i v e

monoidal

over

V
Let

(xf) (v') = xf(x-lv ')

w e take the t e n s o r p r o d u c t
K

us.

of

The theory of this v a r i e t y

is a c l o s e d

it is p o s s i b l e

known

is, b u t

take the set

and

v'V ~ .

The unit o b j e c t

This

is not the free

is n o t the usual u n d e r -

lying set functor.

(1.13)

Suppose

that

is a s e m i - v a r i e t y

a n d also a c l o s e d

symmetric

monoidal

category.

The v a r i e t a l

structure

free a l g e b r a

on one g e n e r a t o r

between

the internal

gives an u n d e r l y i n g

w h i c h we denote

and external
Hom(V',V)

and that these

inclusions

between

The e x i s t e n c e

sets.

the h y p o t h e s i s
an i n j e c t i o n

that

provided

functor

hom

these o b j e c t s

c <V(V',V)>

cohere w i t h the u n i t a r y
of the r e q u i s i t e

be a c o c o m m u t a t i v e

ture a n d that b o t h the e x t e r n a l

those u n d e r l y i n g

(1.14)
in

of

< V(V',V)

<V'>)

- and

vast simplifications
like,"There
course

between

>

a map

<IAI > - - +

In these notes,

awkward.

An a d d e d a d v a n t a g e

Example.

spaces

and binary operations

u n a r y operation,
category.
f,g

we will

V'

for

unit object

K.

canonical

map,

without

functor

and equations

the t h e o r y

are maps,

required

by

f+f'

: V~+

for this

hom

is

(2.1)

Let
S

b e a set.

whose union

Let
is

u =

S .

(S 8) be a

For

xES,

are c o n t a i n e d

to define

If

for

say things

Hom(<iAI >, <IEI > ) .

The a l t e r n a t i v e

is n o t the

b u t as long as

of it.
throughout
V

denote

the c h a p t e r s
the c a t e g o r y

It has the n u l l a r y

an a b e l i a n
.

This

is to

unbearably

to be able to apply

uniformity)

field and

group as w e l l

O, u n a r y
as a

It is w e l l k n o w n to be a c l o s e d

and

is d e f i n e d b y

Uniform

>

<IBI >) - makes

vanish.

continue

V'

are v e c t o r

spaces

(f+f') (v) = f(v)+f' (v)

the o p e r a t i o n

a n d the t e n s o r p r o d u c t

2.

sets of

V'

defines

are func-

For later use, w e

t h a t at the u n d e r l y i n g

<IBI > --+

questions

lcK

struc-

W h a t this m e a n s w i l l of

f u r t h e r mention.

I , for each

l K , (If) (v) = lf(v)

I>,

it means

a variety.

is commutative.

hom

do.

- of m o s t of the p r o p o s i t i o n s

(commutative)

seem

in m i n d is that

(that is b a s i c a l l y

is a l r e a d y

is

in this case the t r a n s p o s i t i o n

and a map

the p r o o f s

be a

Then

first m a p is
hypothesis

in the future,

we m a y as w e l l take a d v a n t a g e

multiplication

In fact,

: V--+

milarly,

Let

over

The

<V(V,V')

~(B, (A,E))."

Here is the e x a m p l e w h i c h we w i l l

d e v o t e d to the theory.
of v e c t o r

c Hom(<~

(2.5) b e l o w and to m a k e sense of the idea of a c o n v e r g e n c e

(1.15)

from

Nor does the q u e s t i o n

and

~(A,B)

is t h a t all c o h e r e n c e

w e s e e m to b e stuck w i t h one,

functions

Hom(V,V')

<IEI > )

a rigid grounding

for functions

of the internal

But in all cases

- not to m e n t i o n

in

V '----+

Specifically,

H o m ( < IBI>,

maps

What additional

theway

from

~(A, (B,E))---+

statement.

isomorphism

sets that are very m u c h p a r t of their

that

this w i l l be u n d e r s t o o d

for a d o p t i n g

object

is not clear.

reference

it comes d o w n to a w e l l - k n o w n

a coherence

can be shown to follow

The m a i n p o i n t to k e e p

a pseudomap

map

b y the

and a s s o c i a t i v e

a n d the e l e m e n t s

in o u r theory.

is a c a n o n i c a l

m a k e the s t a t e m e n t s

reason

hom

for future

d e p e n d on the e x a c t

set level

, <V>)

functions

sets and c o m p a r e

The fact t h a t w e are s u p p o s i n g

Hom(<V>,

We w i s h to d e s c r i b e

coalgebra

stage in the theory.

is r e p r e s e n t e d

say that up to n a t u r a l

the counit m a p is an epimorphism.

at this

call an e l e m e n t

which

c Hom(<V'>

come p r o v i d e d w i t h u n d e r l y i n g

tions b e t w e e n

which will

n e e d e d to force the s e c o n d to be an i n j e c t i o n
worth pursuing

<->

of

on

Hom(V,V')

and

. Si.

The

is the usual one.

Spaces.

cover

let

of

u*(x)

S, t h a t
= U{UEU

is a c o l l e c t i o n

of s u b -

Ix~u }. Let u* ={u*(x) Ix~S}.

If

and

are covers,

such t h a t

xucv

that

implies

uU

the e x i s t e n c e

called a star r e f i n e m e n t
structure

on

w e say that

Then a uniform

of

of a

U.

refines

structure
v~U

A pair

on

if for

such that

(S,U)

x6v~v

there

is a c o l l e c t i o n
v*

consisting

w i l l b e c a l l e d a u n i f o r m spa~e.

refines

of a set

The covers

in

is a

u~u

of covers
.

Such a

such
v

is

and a u n i f o r m
U

are c a l l e d u n i -

form covers.
If

(SI,U_I)

and

if for all
on

($2,U_2)

~2U2

is c a l l e d

there

is

separated
and

are u n i f o r m

contains both

For a t h o r o u g h

discussion

~IUI

x~y

in

of u n i f o r m

spaces

I a n d pp.

28-29 of

X =

(S,U)

f:S 1

f-l(u_2)

there

F r o m n o w on w e w i l l

is a

suppose

including

tion g i v e n h e r e a n d t h a t g i v e n b y entourages,
chapter

a map

which refines

if for

spaces

S2

is c a l l e d u n i f o r m

A uniform

uU

str~cture

such t h a t no set in

all u n i f o r m

spaces

the r e l a t i o n b e t w e e n

see [ Isbell]

is a u n i f o r m

space,

, especially
write

S =

are separated.
the d e f i n i -

problem
IXl

8 of

for the

underlying set.
(2.2)

If

(S,~)

is a u n i f o r m

w i t h the same p o i n t

u~U .

This

tinuous

topology

uniform

spaces,

R ~

(see below)

(2.3)

If

uniformity

As u n i f o r m

is a u n i f o r m

i n d u c e d on

form t o p o l o g y
uniform

topology

on

S'

spaces

in the i n d u c e d

S'

f'

by

(S,U)~

(2.5)

in

space

dense

space

all the sets

(S,T)

u*(x)

E v e r y u n i f o r m m a p induces

where

a con-

In p a r t i c u l a r

For instance,

t h e y are i n e q u i v a l e n t

S'

is a subset

the c o l l e c t i o n
on

S'

two

as t o p o l o g i c a l

since the former

of

of all

S , there

uPS'

for

is the same as t h a t

(Sl, U_l) is c o m p l e t e

subset

uniformity,

[ Isbell]

or just b y

St

if

S'cS

there

is com-

is a n a t u r a l

u~ucU

. The u n i -

i n d u c e d b y the

if for e v e r y u n i f o r m
function

f:S --~

II h o w e v e r y u n i f o r m

as a dense

and every

is a u n i f o r m

, chapter

is u n i q u e up to a unique

Lemma.

spaces b u t not conversely.

space a n d

(i.e. h a s the i n d u c e d uniformity)

embedding

topology.

a topological

at

S .

uniform

It is shown

base

is not.

Namely

We say that the u n i f o r m

(2.4)

is a s s o c i a t e d

the same topology.

for the i n d u c e d u n i f o r m i t y

(S,U), e v e r y t o p o l o g i c a l l y

is

topological

m a y induce

b u t the latter

(S,U)

there

is c a l l e d the u n i f o r m

structures

(-i,i)

space,

Take as a n e i g h b o u r h o o d

one on the a s s o c i a t e d

distinct

plete

set.

S1

We d e n o t e

whose

space
which

is

restriction

to

space can be e m b e d d e d

s u b s e t of a c o m p l e t e

isomorphism.

f':S'--~ S 1

space a n d t h a t this

this c o m p l e t i o n

of

(S,U)

is understood.

Suppose
fl
(SI,U_I)

(TI,V._l)
is a c o m m u t a t i v e

d i a g r a m of u n i f o r m

f2

spaces

($2, ~ )

~ (T2,V2)
such that

fl

is the i n c l u s i o n

of a dense

subset, e q u i p p e d w i t h the i n d u c e d uniformity,

and

f2

is an i s o m o r p h i s m of u n d e r l y i n g sets

~ i has a b a s i s c o n s i s t i n g of sets w h o s e image u n d e r

topology of

~2

"

Then there is a unique map

f2

($2,U2)--+

is closed in the u n i f o r m

(TI,V_I) m a k i n g b o t h trian-

gles commute.
Proof.

A l t h o u g h I have t r i e d to m a k e these notes l a r g e l y self-contained,

this a r g u m e n t

requires m o r e than a p a s s i n g a c q u a i n t a n c e w i t h the theory of u n i f o r m spaces to be understood.

For this I r e f e r to [ I s b e l l ]

Otherwise,

it can b e s t b e a p p r e c i a t e d b y ima-

gining the spaces to be m e t r i c and r e p l a c i n g C a u c h y n e t s by C a u c h y sequences, n

We m a y suppose w i t h o u t loss of g e n e r a l i t y that
tity m a p w h i c h m e a n s that
w i t h r e s p e c t to
and
hs

{s }
.

~2

There is

is uniform,

~{C~l

v Iv

Vev_{

~*

f i n i t i o n of C a u c h y net).

c o v e r of

ya8

{gsy},YaB

~2

hsEV

and thus converges to

to

hs

T
hs

sS 2

a C a u c h y net

as above

{gs }

s .

(2.6)

e>0

implies

Since

gs V

Vc~*(hs).

of e l e m e n t s of

~2

"

~iE~l .

There is a

and converges,

Hence for all y>B,gs v{*(hs).

~i

b u t we may argue as follows.

S1

constructed by choosing

w h i c h converges to it.

Then

and a uniform map

Such an

f:X--~

X =
(M,~)

In fact the covers

(S,~)

and any

u_EU

It follows
there is a

such that the inverse image of

is c a l l e d a p s e u d o m e t r i c for

a c o r r e s p o n d i n g c o l l e c t i o n of

(T,VI)~.

T .

a u n i f o r m i t y c a l l e d the m e t r i c uniformity.

(M,~)

~2

is a u n i f o r m space in a natural way.

m e t r i c space

u .

N o w let

(T,V_I) and converges to a unique point, g~(s), of

determine

the 1-spheres refines

converges to
"

(that is just from the de-

g~:(S2,U2)---+ (T,VI)

, 1.14 that for any u n i f o r m space

for

~i

is c l o s e d in that t o p o l o g y and e v e r y

is u n i f o r m to

from [ Isbell]

o v e r a basis of

{hs }

s~S 2

the net is e v e n t u a l l y in one of the n e i g h b o u r h o o d s of

{s }

(M,~)

is u n i f o r m

is C a u c h y in

A m e t r i c space

e-spheres,

Suppose n o w that

Then

But we h a v e just shown t h a t all such p o i n t s a l r e a d y lie in

by

is the iden-

and b y the h y p o t h e s i s we m a y also suppose

Thus

has a u n i f o r m e x t e n s i o n to

for each

"

is a g a i n a C a u c h y net in

as well.

This does not q u i t e show that

g

~i

f2

T h e n g i v e n that
~i

is a C a u c h y net w i t h r e s p e c t to
refines

Now

gs ~V , it follows that

The map

"

w h i c h c o n v e r g e s to

such that

in the u n i f o r m t o p o l o g y of

~i

~2

and

are c l o s e d in the u n i f o r m t o p o l o g y d e t e r m i n e d b y

a n d an index

Thus in every

S1

{gs }

such t h a t

that the sets in

hs

is a r e f i n e m e n t of

w e m u s t show it is w i t h r e s p e c t to

is a C a u c h y net in

Since

set

~i

T1 = T2 = T

X .

If

ranges

is c a l l e d a b a s i s of p s e u d o m e t r i c s

X .

(2.7)

Suppose

S--+

T .

uU

(S,~)

We say that

w h i c h refines

and
F

(T,~)

are u n i f o r m spaces and

is an e q u i u n i f o r m set if for all

f-l(~)

for all

f~F .

is a set o f functions
v~V

there is a single

It is easy to see that every finite set of

u n i f o r m functions is an e q u i u n i f o r m set.
(2.8)

If, on the o t h e r hand,

o f sets of functions
in

are e q u i u n i f o r m .

p s e u d o m e t r i c s for

V .

S --~

is a set and

(T,[)

is a u n i f o r m space, a c o l l e c t i o n

T determines a uniformity

on

such that

This is m o s t r e a d i l y d e s c r i b e d in terms of a b a s i s
Let

the sets
d

of

F(s,F,d)

= {s' [d(fs,fs')
for

< 1

scS,F~,dEd

for all

f6F}

. T h e n the sets

{F (s,F,d)Is~S}
are a cover of

which

pseudometrics

for this

sup{d(fs,fs')

If~F}

(2.9)

tions

S--+

F---+

structure

suppose

case

given a c o l l e c t i o n

so d e s c r i b e d

has the u n i f o r m i t y

in

[ Isbell]

, Chapter

(2.10)

If

X---+

is an i n j e c t i v e

is u n i f o r m l y

isomorphic

We note the o b v i o u s
consisting

above

and e m b e d d i n g s

is a c o l l e c t i o n

by a pseudometric

pseudometric

and

in

where

X.

convergence

Xi

of u n i f o r m

possess

A basis

(d.F) (s,s')

and a set

of
=

of

of sets of func-

structure

A special

on all o f

on

case
S

We

is

~ = {S}

t h a t is con-

provided

is g i v e n the i n d u c e d

spaces

that

uniformity.

has a f a c t o r i z a t i o n

and monomorphic

the usual p r o p e r t i e s

(see [ Kelly]

of u n i f o r m

constructed

Pi:X---~

of

a uniform

uniformity.

as the e p i m o r p h i c

the e m b e d d i n g s

(2.11)

b e r of indices

vary.

u n i f o r m m a p we say it is e m b e d d i n g

a n d left c a n c e l l a t i o n

{X }

d-F

as a c o l l e c t i o n

to its image w h e n t h a t image

In p a r t i c u l a r ,

If

and

III.

under c o m p o s i t i o n

majorized

of subsets
Z

describes

of u n i f o r m

fact t h a t the c a t e g o r y

of s u r j e c t i o n s

spectively.

a convergence

sidered

as

is g i v e n b y the f u n c t i o n s

T h e n the d i s c u s s i o n

structure

In t h a t case we can think of

w i l l call a u n i f o r m i t y
in w h i c h

a uniform

We can instead

functions

determine

system

parts,

re-

of i n v a r i a n c e

).

spaces,

as follows.

be the c o r r e s p o n d i n g

every p s e u d o m e t r i c

on

Let

be a finite

i=l,

...

projective

, n

map.

X = HX

Let

d.
l

is
num-

be a

a n d let

d(x,y)

= s u p ( d l ( P l X , p l y ) , d 2 ( P 2 x , P 2 y) ..... dn(PnX,pny))

In the sequel we w i l l w r i t e

d = sup{di(Pi,Pi)}

m i t y d e f i n e d b y the p s e u d o m e t r i c s

so d e f i n e d

It is e a s y to show that the u n i f o r -

is the c o a r s e s t

for w h i c h the p r o j e c t i o n s

are uniform.
(2.12)

Proposition.

discrete
Proof.

uniformity.
X

Let

{X }

has a u n i f o r m

q u i r e d to b e a u n i f o r m

and

set into a single p o i n t

then
(2.13)

f((x

)) = f((x'

form.

= X

Notes o n u n i f o r m

inverse

T h e r e are really
g i v e n here.

and the t h i r d b y e n t o u r a g e s

(see [Kelley],

as the m o s t u s e f u l

if
f

sets

That

= x'

factors

except

for things

The first

every such

for ~=l,...,n,

~ X ----~

of u n i f o r m

is b y families

~ i ..... ~ n

where

takes

through

is re-

is r e f i n e d

KV

three d e f i n i t i o n s

VI).

h a v e the

n u m b e r of indices,

of all

The s e c o n d

Chapter

image of that c o v e r

a cover.

that

that

and

a finite product.

cover of a p r o d u c t

be a finite

determines

in particular,

covers,

as w e l l

The

spaces

through

A uniform

1 .... ,n

in o t h e r words,

spaces.

of u n i f o r m

factors

T h e n the c o l l e c t i o n

otherwise

implies,

)), or,

Let

space.

The first is b y the u n i f o r m

vely geometric

cover b y singletons.

in the c o r r e s p o n d i n g

V u , ~ = l , . . . , n

HX----+

cover of the product.

by a c o v e r of the f o l l o w i n g
be covers

be a c o l l e c t i o n

T h e n any m a p

XlX...xX nspaces.

of s e m i n o r m s

is the m o s t i n t u i t i -

like t o p o l o g i c a l

groups where

each

neighborhood of

over the group.

This a u t o m a t i c a l l y d e f i n e s a u n i f o r m i t y - the c o n t i n u i t y of m u l t i p l i -

gives rise to the c o v e r gotten b y t r a n s l a t i n g that n e i g h b o r h o o d all

cation is e q u i v a l e n t to the e x i s t e n c e of star r e f i n e m e n t s - c a l l e d the c a n o n i c a l uniformity. A g r o u p h o m o m o r p h i s m is c o n t i n u o u s iff it is uniform.

Thus there is no dis-

tinction in t h a t case b e t w e e n t o p o l o g i c a l and u n i f o r m notions.

The d e f i n i t i o n b y p s e u d o -

norms seems the m o s t useful in these notes.

The d e f i n i t i o n by e n t o u r a g e s is p r o b a b l y

most useful in c o n n e c t i o n w i t h c o m p a c t n e s s arguments.

canonical - and unique - ~ i f o r m i t y

A c o m p a c t H a u s d o r f f space has a

for w h i c h the e n t o u r a g e s are all n e i g h b o r h o o d s of

the diagonal. E v e r y m a p to a n o t h e r u n i f o r m space is c o n t i n u o u s

any rate there is no "best" definition.

iff it is uniform.

At

It is i m p o r t a n t that all three are equivalent.

3. U n i f o r m Space O b j e c t s

Let

(3.1)

b e a s e m i v a r i e t y w i t h u n d e r l y i n g set functor

ture on an o b j e c t

V~V

we m e a n a u n i f o r m structure on

tions are u n i f o r m w h e n the p o w e r s o f

(see [ Isbell]
If
~I

and
the

is a p r e u n i f o r m object in

let

<V>

denote the u n d e r l y i n g u n i f o r m space

has a p r e u n i f o r m s t r u c t u r e

the u n i f o r m c o m p l e t i o n of

consider

the case that

V n ~-q~

V~

Then

n ,

Vn

V .

V .

(3.3)
that

N o w suppose that

< I V I>

are

(in

V)

is dense in

V~

(V~)n

(we omit the < > )

.

For any n-ary ope-

V~ n

V~

is s a t i s f i e d w h e n r e s t r i c t e d

But an e q u a t i o n is s a t i s f i e d e x a c t l y w h e n two maps are equal

V~
V

and

is c o m p l e t a b l e

has, then, an e x t e n s i o n to a m a p

and two u n i f o r m maps w h i c h agree on a dense subset are equal.

w h a t is at issue is that

I< V > I

To e x p l a i n this more clearly,

is dense in

A n y e q u a t i o n r e q u i r e d to be s a t i s f i e d b y an a l g e b r a of
to the dense subset

Both

U , w e say that

also lies in

is a variety.

and it is r e a d i l y seen that for any

~ , the m a p

V .

V .

provided

ration

are e q u i p p e d w i t h the p r o d u c t uniformity.

denote the u n d e r l y i n g d i s c r e t e o b j e c t o f

If

By a p r e u n i f o r m struc-

such that all the opera-

, C h a p t e r II.)

same u n d e r l y i n g set of

(3.2)

<V>

<->.

<V>

b e l o n g to the s e m i - v a r i e t y
is c l o s e d

In the general case,

(by w h i c h we u n d e r s t n a d that the s i t u a t i o n is

e x p o s e d in section i, in p a r t i c u l a r that we have a symmetric m o n o i d a l structure).

We say that
object

is a d m i s s i b l e p r o v i d e d for all p r e u n i f 0 r m o b j e c t s

U n ~ ( V ' , V ) of

~(V',V)

such that

< UnV(V',V)

>---+

< V (IV' I , IVI)>--+

is a pullback.

V' , there is a sub-

Here, of course,

Un(<V'>,

<V> )

Hom(<IV'l>,l<V>l )

Un(<V'>, <V> )

is the set of u n i f o r m maps b e t w e e n

these sets.
(3.4)

If the theory o f the s e m i - v a r i e t y

has the i n d u c e d c l o s e d m o n o i d a l structure,

is a -Th-algebra.

For

~ V = ~ :Vn

is the c o m m u t a t i v e theory
it is easy to see that

Th

and

Un ~ V'> <V>)NHOm(V',V)

is an o p e r a t i o n w h i c h w e have s u p p o s e d to be

10

uniform,

t h e n for any c o l l e c t i o n

f.:V'--~
1

V,

in

(fi)

Vn

V'
is also a u n i f o r m morphism.
Thus the a d m i s s i b i l i t y
belongs

to

(3.5)

of

Th

e(f

),

the

d o w n to w h e t h e r

is not n e c e s s a r i l y

monoidal

structure,

and

are two s u b o b j e c t s

of

comes

is

the c o m p o s i t e

Th

operation

in

V(V',V)

or not t h a t p a r t i c u l a r

_Th--algebra

Even when

V1

But this

, of u n i f o r m morphisms,

<V2>

Then both

at the u n d e r l y i n g
lows that

there

is at m o s t one c a n d i d a t e
of an o b j e c t

inclusions

set.

commutative

Since

V2

VoNV 1

for

VO

and

< V0 >

For suppose

Vo~VI ---+

set functor

Thus the s u b o b j e c t

has some ad hoc closed

Un ~(V',V)

such that

the u n d e r l y i n g

V O = VoNV 1 = V 1 .

and

< Vi >

V1

reflects

Un ~(V',V)

V0

as subobjects

become

equalities

isomorphisms

is unique,

it fol-

provided

it

exists.
(3.6)

We say t h a t

missible,

completable

tegory o f u n i f o r m
follows

objects,

that a m o r p h i s m

the u n d e r l y i n g

(3.7)

is a u n i f o r m o b j e c t
and its c o m p l e t i o n

If

natural

closed monoidal
structure

structure

Thus there

is,

thereby

it is not true,

unless

uniform

iff it is a f u n c t i o n of o n l y f i n i t e l y
{U }

be a collection

uniformity

<V( IVI, I u
is a p u l l b a c k ,

for e a c h

and

let

I)>

both

<V( ~ I,I KU
which

shows

dense

in

it

in

(3.4)

spares

every pre-

structure

is

and uniform.

that every o b j e c t

uniformity.

In fact,

to a d i s c r e t e

in

it

space

is

m a n y coordinates.
Let

U = KU

the p r o j e c t i o n s
V

t h e o r y w i t h the

are uniform.

be a p r e u n i f o r m

Un(<V>

>

Hom(< IVl> ,<I U

I-I , < - >

,< U

e q u i p p e d w i t h the

object.

I claim that

T h e n since

> )

I>)

a n d the h o m functors

commute with

so is
<KV(V,U

Evidently

the ca-

and u n i f o r m on

between preuniform

is finitary,

of d i s c r e t e

for w h i c h

To see this,
< U n V ( V , U )>

products,

every p r e u n i f o r m

the d i s c r e t e

of u n i f o r m objects.

- the c o a r s e s t

is a u n i f o r m object.

(3.2)

the t h e o r y

f r o m 2.12 that a m a p f r o m a p r o d u c t

if it is addenote

F r o m the d e f i n i t i o n
in

t h e n as o b s e r v e d

in

in that case no d i s t i n c t i o n

follows

product

Un V

for a c o m m u t a t i v e

induced,

a n d as o b s e r v e d

a uniform object when equipped with

Let

as the hom.

is b o t h a m o r p h i s m

of all the a l g e b r a s

is a d m i s s i b l e

Even in that case,

(3.8)

object,

We let

sets.

uniform

becomes

< Un V ( - , - ) >

is simply one w h i c h

is the c a t e g o r y

completable.

with

if it is a p r e u n i f o r m

is adimissible.

that

HU ~
U~

is admissible.

a n d the l a t t e r
is a d m i s s i b l e

is the p r o d u c t

of the

) >

I)>

..... H o m ( < l VI>,

Moreover

is c o m p l e t e

being

Un(<V>,<HU

so t h a t

a product

> )

<IHU

is dense

in

U ~ = HU ~

of a d m i s s i b l e

1>
U

from which

and

objects.

is

is completable.
It is e v i d e n t

that

11

(3.9)

N o w let

U'

U"

be two m a p s of u n i f o r m o b j e c t s and

g i v e n the structure as a subspace of

to show that

m a t t e r entirely.

is admissible.

U'.

be t h e i r equalizer,

E x a c t l y the same a r g u m e n t as above suffices

To show that it is c o m p l e t a b l e is, however,

It is easy to see that

U~

is a subobject of

U'~

n o t seem to b e any o b v i o u s reason that it is a r e g u l a r subobject.

another

b u t there does

If w e s u p p o s e d that

was a q u a s i - v ~ r i e t y - c l o s e d u n d e r p r o d u c t s and subobjects - that w o u l d settle it.

BUt in the one e x a m p l e w e h a v e in which
(see IV. 3.) - ~

is not a v a r i e t y - that of Banaeh spaces

is not a q u a s i - v a r i e t y either. What h a p p e n s there is t h a t any c l o s e d

subobject of a u n i f o r m o b j e c t is a u n i f o r m object.

Thus the a p p r o p r i a t e h y p o t h e s i s at

this p o i n t is unclear for w a n t of e x a m p l e s and we leave it as an open question.

(3.10)

Proposition.

The u n d e r l y i n g functor

I I : UnV---+

creates products.

v i d e d the c o m p l e t i o n of an e q u a l i z e r is an a d m i s s i b l e p r e - u n i f o r m o b j e c t of
functor c r e a t e s limits.
tisfied provided

If

is the c a t e g o r y of all algebras,

is c l o s e d in

ject u n d e r l y i n g a c l o s e d

UnW

This last c o n d i t i o n m e a n s that if

of the u n i f o r m space w i t h

I U'I

~ , the

this c o n d i t i o n is sa-

under all subobjects or at least if the

s u b o b j e c t of an o b j e c t in U n V
U~UnV

, and

~ W , then

b e l o n g s to

< U '> c <U>

I U' Ic V

Pro-

ob-

V .

is a c l o s e d subspace

In the sequel, w e will simply suppose that the c o n c l u s i o n of this p r o p o s i t i o n is

a u t o m a t i c a l l y satisfied.
(3.11)

Let

V~

and

U~Un~ .

We let

[V, U]

denote

sest u n i f o r m i t y for w h i c h the map e v a l u a t i o n at

~(V, IU I)

v: [ V,U] --+ U

e q u i p p e d w i t h the coar-

is u n i f o r m for each

vV.

E q u i v a l e n t l y we require that
< [V,U] >
be a u n i f o r m embedding.

is a pullback.

Then

IV,U]

<U>

<V>

is c e r t a i n l y a p r e u n i f o r m object.

<V(V,UnV(U',U))>

U n(<U'>, < [ V , U

<V(IU'I ,I I V , U ] I ) >

Hom(< ~' I>, < H V , U ] >

N o w I c l a i m that

] >)

We b e g i n w i t h the fact that

< UnV(U' ,U)> ---+ U n ( < U ' > ,< U > )

t
< V( IU'I,IUI >
is a p u l l b a c k in

Applying

'

H o m ( < V %-)

H o m ( < V > ,<UnV(U',U)>)

H o m ( < V > , < V( IU 'I ,IUI > )

is as well.

N o w since

<V>

Horn(< IU' I>, < lUl>)

we see that
Hom(<V>, U n ( < U ' > ,<U> ))

H o m ( < V > , H o m ( < I U ' l > ,<IUJ>))

is d i s c r e t e
<V>

Hom(<V> , Un(<U'>,<U>)
and

~ Un(<U'>,<U

> )

12

H o m ( < V > , H o m ( < I U ' I>,<] U I> )) ~ H o m ( < ]U' I>,<I U < V > I>).
since b y
< V(V,UnV(U' ,U))>

>

Hom(<V>,<

< V(V,V( I U'I ,IU I > - - - +

is a pullback,

Hom( < V > , V ( IU '] ,[UI ))

it follows that so is
< ~(V,UnV(U,U))>--+

< V ( V , V ( [U'~}UI)>
If

f:V --+

(U' ,V_(V,U))>

Un V(U',U)

--+

[V, U] .

'

Un(<U'><u<V>>)

<V>

gom(<lU' I>, < ~U

c V(U',U)

is a pseudomap,

is a p s e u d o m a p
U '---+ ~(V,U) =
<V>
< V>
< U'> --+ < U
> ~ < U>

so

mines a u n i f o r m map

<[ V,U] >)

UnV(U',U)>)

then

I[V,U]

I> ) .
f < V ( V , V ( U , U ) ) >
I

<

Since it also deter-

it gives a u n i f o r m p s e u d o m a p

That is, the image of the u p p e r map in the last square lies in
In a similar w a y the image of the lower map lies in

U'

Un(<U'>,

H o m ( < ~' I>, < I[V,U] 1> )

w h i c h gives the d e s i r e d result.

(3.12)

This not only shows that

for a c o t e n s o r

Un ~ ( U ' , [ V , U ]

[ V,U ]

is a d m i s s i b l e but gives the d e s i r e d a d j u n c t i o n

) ~ V ( V , U n V(U',U)) . However, we still have to show that

[V,U] is c o m p l e t a b l e and that its c o m p l e t i o n is admissible.

that

is complete.

Then for

U'

dense in

U" , we k n o w

F i r s t w e c o n s i d e r the case
(essentially b e c a u s e we h a v e

a s s u m e d it) that
U n V(U",U) ---+
is an isomorphism.

Applying

~(V,-)

Un V(U',U)

w e see that

V(V, UnV__(U",U) ) ---+ V(V,UnV(U' ,U) )

or
U n V ( U " , [ V , U ] ----+ U n V ( U ' , [ V , U ] )
is an isomorphism.

This implies that

trary w e use the h y p o t h e s i s of

an a d m i s s i b l e p r e u n i f o r m

V--object as well.

is u n i f o r m i z e d as a s u b o b j e c t of
(3.13)

Proposition.

Proof.

In the square

Let

[V,U]

is c o m p l e t e as well.

(3.10) to infer that the c l o s u r e of

[ V,U~]

be e m b e d d e d in

is arbi[V,U~]is

To see that w e n e e d o n l y o b s e r v e that [ V , U ]

M o r e generally, we have the following
U'

<[ V,U] > ---+

<[ V , U ] > ----~

the u p p e r and r i g h t h a n d m a p are embeddings.

N o w if
[ V , U ] in

Then

<U>

<U'>

[ V , U ] is e m b e d d e d in

[V,U' ] .

<V>

< V>

Thus the c o m p o s i t e and t h e r e f o r e the left

h a n d m a p is one as well.
(3.14)

One m o r e h y p o t h e s i s w i l l have to be made.

When

~(I,-)

r e p r e s e n t s the under-

13

lying set functor,

the o p e r a t i o n s
map.

More

map

A1

pseudomaps.
(3.15)

extends

If

theory.

in

Moreover,

cations

if

AI---+

immediately

according

(i)

category

is a c a n o n i c a l

of v e c t o r

UnV

and

N o w we m u s t suppose

that there

spaces,

is simply

is s i m u l t a n e o u s l y

autonomous.

A closed

(iii)

a vector

in

is complete,

A(A,B) ---+ A ( A

it is a v a r i e t y w i t h a

(2.12)

field K

space

is a
any

the same to h o l d for

map

the c a t e g o r y

o n groups

The fact that

limit of maps

of preuniform
a uniform

. This

,B ).
com-

objects

object

is the

is the same as a

such t h a t all

~calar multipli-

are continuous.

An autonomous

(ii)

space o v e r the d i s c r e t e

*-Autonomous

F r o m h e r e on, we revive

closed

and p s e u d o m a p s .

that the u n i f o r m

is a dense e m b e d d i n g

to the remarks

vecotr

group which

b e t w e e n maps
implies

is the c a t e g o r y

4.

(4.1)

A2

A2

Thus the c a t e g o r y

same as a t o p o l o g i c a l
topological

are u n i f o r m

to a m a p

It follows

Example.

is no d i f f e r e n c e

specifically,

mutative
V

there

in the theory

By a * - a u t o n o m o u s

category

and call a s y m m e t r i c

monoidal

is m e a n t

G;

(_), : ~ op __+

An e q u i v a l e n c e

This is s u b j e c t

an o l d name due to L i n t o n

category

functor

Cateqories

d = dG

to one axiom,

~;

: G--+

G**

that the d i a g r a m
G(G,G')

G(G'*,G*)

G(G**,G'**)
whose h o r i z o n t a l

and vertical

a r r o w are the a c t i o n s

of

(-)*

on the internal

hom,

com-

mute.
(4.2)

An i m m e d i a t e

one a split epi.

mono which
(4.3)

is:

equivalence
T = I* .

G'

G p

~ G

category
which

of the former

(GI)* ~ G*

G '--~

is full a n d faithful

and such that there

Hom(G~(G2G3)*)

T
G*

have c o r r e s p o n d e n c e s

with unit

and
and
G

G
G' ~

G'**

In t h a t case,

is a i-i c o n s e q u e n c e
is syrmaetric,

G'---+

G*

G--+

G'*

---+ G'

G'*---+

lemma

to h a v e a * - a u t o n o m o u s
and tensor product

Since the tensor

G'~G--~

between

and b y the Y o n e d a

a n d thus is a l s o a split

T h e n we see that there

dence b e t w e e n m a p s
respondence

is an i n s t a n c e

far less d a t a t h a n t h a t is r e q u i r e d

Hom(GI~G2,G~)---~

Let

arrow is a split m o n o and the v e r t i c a l

that b o t h are i s o m o r p h i s m s .

Suppose w e h a v e a m o n o i d a l
w i t h a functor

the h o r i z o n t a l

But the latter

implies

In fact,

result

G*

---+ G'**

Since

maps

is s i m i l a r l y
these

equipped

is n a t u r a l

d e f i n e ~(G',G) = ( G ~ G * ) * .

between

there

Putting

category.
~

together

GAG'---+
a

T and

correspon-

gives a cor-

(-)* is full a n d f a i t h f u l w e

14

Next define

G(G',G)

(G'~G*)*

There is a 1-1 c o r r e s p o n d e n c e b e t w e e n maps

G(G 2,G 3) = ~G2~G ~) *

G1
G2~G ~
G~G 2

w h i c h d e m o n s t r a t e s that
berg-Kelly]

'

(G2~G 1 ) *

G2~G 1

G3

GI~G 2

G3

G(G2,-)

-eG
It follows from
2
f o l l o w i n g p r o p o s i t i o n 3.1, that G

comes a u t o n o m o u s w i t h this d e f i n i t i o n of hom.

is immediate.
tor.

The e q u i v a l e n c e of G(G',G)

Note that this is i n d e p e n d e n t of w h e t h e r or not

T h a t seems to be m o r e or less e q u i v a l e n t to

ples b u t one.

[Eilen-

is left a d j o i n t to

, II.3 e s p e c i a l l y the m a t e r i a l

then be-

and ~(G*,G'*)

(-)* is a m o n o i d a l func-

I* ~ I , w h i c h holds in all the exam-

Thus the data at the b e g i n n i n g of this section suffice to determine a

*-autonomous category.
(4.4)

O n the o t h e r h a n d suppose that the c a t e g o r y

~ , internal h o m functor

G(-,-),

u n i t o b j e c t I t o g e t h e r w i t h r e q u i r e d natural t r a n s f o r m a t i o n s and e q u i v a l e n c e s c o n s t i t u t e
a symmetric c l o s e d category.
G p---+

T = I* .

Suppose also that there is a full faithful functor

a n d an e q u i v a l e n c e Hom(GI,~(G2,G~))

~ Hom(G3,G(G2,G{))

As before,

(-)*:
let

T h e n there is an e q u i v a l e n c e
Hom(G',~(G,T)) ~ H o m ( I , G ( G , G ' * ) )
Hom(G,G'*)

Similarly,
Hom(G',G*) ~ Hom(G',~(I,G*))
Hom(G,G(I,G'*))
Hom(G,G'*)
C o m p a r i n g these, w e see that

G* = G(G,T)

gether w i t h Hom(G',G) ~ Hom(G*,G'*)

and b y the Y o n e d a lemma

(4.5)

Since

G* ~ ~(G,T)

Also, Hom(G',G*) ~ Hom(G,G'*) w h i c h to-

implies that

have Hom(G',G) ~ Hom(G*,G'*) ~ Hom(G',G**)

(-)* is an equivalence.

w h i l e e v i d e n t l y Hom(G*,G'*)

In fact, we
= Hom(G'**,G**)

G' ~ G'**
the internal c o m p o s i t e gives a natural t r a n s f o r m a t i o n
(G',G) ~

(G*,G'*)

w h i c h f o l l o w e d b y the i s o m o r p h i s m above gives a m a p

(G',G)---+

(G',G**)

. It follows

from c o h e r e n c e that the d i a g r a m

G(G',G)

G(G*,G'*)

" ~ G (G' ,G**)

commutes. S i n c e the v e r t i c a l and diagonal maps are isomorphisms,

so is the h o r i z o n t a l

15

one w h i c h means that

(-)* is i n t e r n a l l y full and faithful.

Now we define G ' G = G ( G ' , G * ) * .

We see that there are natural i-i c o r r e s p o n d e n c e s b e t w e e n maps

GIG 2

G3

~(GI,G~)*

---+

G3

G~ ~

~CG1,G~I

G I --'+ G(G~,G~) ~ ~(G2,G3)

Here we have freely u s e d the e q u i v a l e n c e b e t w e e n

and

G**

n o n o m o u s c a t e g o r y n o w follows from [ E i l e n b e r g - K e l l y ] , 1.3

(4.6)

That w e have a *-auto-

The m a i n p u r p o s e in these notes is to b e g i n with a good deal less than a *-auto-

n o n o m o u s c a t e g o r y and c o n s t r u c t one. This leads us to define the notion of a p r e - * - a u t o nomous situation.

This consists o f a c a t e g o r y

~ , two full s u b c a t e g o r i e s

and

D ,

an e q u i v a l e n c e of c a t e g o r i e s
(-)* : C p
a functor

(-,-)

: Cjp x D--+

(I,D)

(ii)

D ,

and an o b j e c t

of a *-autonomous c a t e g o r y insofar
(i)

>

I~.

These are subject to the axioms

as they make sense.

In particular, w e suppose that

D;

Horn(I, (C,D)) ~ Hom(C,D);

(iii)

(CI, (C2,C~)) ~

(iv)

(C3, (C2,C~)) ;

Hom(C*,C') ~ Hom(C'*,C)

These are s u b j e c t to c e r t a i n c o h e r e n c e conditions w h i c h will b e i n t r o d u c e d as needed.

(4.7)

Since

(-)* is an equivalence,

functor w h i c h w e t e m p o r a r i l y denote
(C,D) ~

(D#,C *)

which, b y

may be r e w r i t t e n
w e have, since

it has, up to a n a t u r a l isomorphism,
(-) #:D P--+ C .

(ii) implies that

Hom(D,C) ~ H o m ( C * , D #) .

(-)* is an equivalence,

Then

~iii)

an inverse

above may be r e w r i t t e n

Hom(C,D) ~ H o m ( D # , C *) .

Should it h a p p e n that

Similarly

(iv)

also belongs to

that

Hom(D,D) ~ H o m ( D * , D #)
and
Hom(D,D)

Hom(D#,D

*)

The c o h e r e n c e s a l l u d e d to above w o u l d r e q u i r e that if

~(f)

describes the m a p

+ C

and

then for
8(g)e(f) =

> 8(g)

f
(fg)#

>

H o m (C,D)

>

H o m (D#, C * )

Hom(D,C)

(C*,D #)

describes

>

C'

~(f)8(g)

(gf)*

w h i l e for

Thus c o r r e s p o n d i n g to
D

-~ D

D'

16

we have
D*
and

~(i)8(i)

identify

= i* = 1

D*

and

8(1)

while

D#

~ D#

e(1)

8(1)e(1)

(4.8)

If w e let

and

If w e let

in

(3.6.iii),

(CI,C;) ~

For

C,C'EC
C* ~

Proposition.

If

observed

A,Be~

A,BEC UD.

or

A,BeD

(C3, (I,C[))
.

Suppose

D~

(Cl,I*) ~ C[ .

~ Hom(B*,A*)

{g~ : D ~

D~}

is clear,

f* : B*-~ A*

a family

of a factorization

and

under w h i c h 4.6(iv)
C~

determines

~or all

~.
Supposing

D-representation

system.

{C--+ D~}
{C~

--

e~}

D~} u{C--+C*}w ' w h e r e

such that

m~h = g$

let a suitable
{C --+ D}

family
a

so is any larger family.

{C--+
* } wC

D~_D has a C--generation w h i c h dualizes

C

{fw:

~ D~

{C~

Of

D~}

C-generation
In particular

and

h.e =f

be called a
of

D .

If

so is the fa-

Thus we may suppose

to a C--generation.

to a D--representation.

and a C - g e n e r a t i o n

for all

is a _C-generatin"

has a D--representation which dualizes

of

D~}

have the

~ D

and the dual notion

is a D - r ~ p r e s e n t a t i o n

a D-representation

{m~:C~

Namely suppose that every pair of families

h:D -+ C

this to be the case,

C

from the remaining

such that every d i a g r a m

a unique map

of

follows

there is a family of maps

C
commutes

one of the unstated coherence hypotheses

is an involution.

{e w :C w --~ D} ' C ~ ~C
-- which collectively

C 3 = I , we get

If we let

this is just the duality while the other two follow as

for every object

and for every

properties

every

Then

of these morphisms

fHom(A,B),f

hypotheses.

mily

~ OlD

(C,C'*)

We wish to discuss a condition

(4.11)

and

Let

The n a t u r a l i t y

Cw-+ C}

CnD

in 4.7.

is that for

D~cD

Henceforth we can

we get

(C3,C~)

Hom(A,B)

(4.10)

agree on

(C,T)

(C',C*) %

Proof.

(-)#

I* = T , this may be summarized as,

Proposition.

(4.9)

D* ~ D#

and

O .

C2 = I

(3.6.i),

(-)*

a single functor

(CI, (I,C~)) ~
or, in view of

) D#

so that

Thus

(-)* : (CuD) Op
that interchanges

~(i)

= i# = 1

by this isomorphism.

w e may think of them as determining

D*

Similarly every

Now fixing a map h : D - + C ,

both a s s u m e d to dualize p r o p e r l y

17

we begin with the diagram

in which

fw

is defined as

>

D~

~ew

and similarly

g~ = m~.h

. Dualizing and using 3.9

we g e t
D*
~

then

~ + h*

has

a unique

C*

fill-in

which

we n a t u r a l l y

and the fact that it is an involution

of the

C~, f , g~

(4.12)

Theorem.

Hom(A,B)
(4.13)

C*

e*

D*
which

>

and

hypotheses

for

If we suppose that

V-functor

of

A,B~UD.
A

is a

and that the equivalence

h*

The naturality

of

4.6(i),

(ii),

(iii)

and

(4.10)

we h a v e

Thus we have a pre *-autonomous

V

category,

we can ask that

a duality

situation.

(-)*

: C

be

be that of V-enriched home,

!(c,,c) ~
In addition,

by

m~ .

Under the

Hom(B*,A*)

denote

follow readily from the s~ne properties

Z(c*,c,*)

we can suppose that the isomorphism of

(4.6.ii) be V--enriched,

V(I, (C,D)) ~ V(C,D)

In that case, we come to the notion of a V_-enriched pre-*~autonomous
case the maps used in the C--generation and D--representation
cessary,
(4.13)

In that

in the paragraphs may,if ne-

be replaced by pseudomaps.
The V-enriched versions of

to derive the analogue of

a family

situation.

{C --+~ D}

and

(4.8) and

(4.9) go through without change.

(4.12) we must modify

In order

(4.10). This is done by ~gain supposing

{C-+ D~}

We observe that there is, for all

~(D,C)

"~

~(D,D~)

~,~,

a commu-

tative square,

_v(c~,c)--+ _v(c ,D~)

We may now require that
That is, given any object

~(D,C) be the simultaneous

V

and commutative

>

V(C ,C)---+
one for each

V (q,D~)

V(C ,D~)

~,$, there should exist a unique map

Again argumenting

the representing

equalizer of all those squares.

squares

and generating

V--~ V__(D,C) inducing those squares.

families

does not change the situa-

tion so we may suppose they are invariant under dualization.

V(C ,C) ~ V(C*,C~)

and

V__(C ,D~) ~ V(D$,C~)

Then

V(D,D~) ~ V(D~,D*),

implies that V__(C*,D*) has the same uni-

18

versal mapping property

(4.14)

Example.

pre-*-autonomous
spaces w i t h

as V(D,C)

On the c a t e g o r y
situations.

This

there

is no r e a s o n n o t to e x t e n d
D

to be the c a t e g o r y

C-which

A linearly
means

space

tient modulo

open),

A linearly

of c l o s e d

linear

subvarities

where

is a p o i n t a n d

Linearly

space

there

as h a v i n g

is,

first,

form a topological

of

Lefschetz

transformation

also

finite

to a space of the f o r m

Ks

compact,

KS .

t h a t the c o n t i n u o u s

naturally
Here

equivalent

S.K

pies o f

and
K .

s i m p l y take
and,

pactness.
compactness

T.K + S.K

s t a n d for the direct

See [ L e f s c h e t z ]

pp.78-82

the set of c o n t i n u i n g
the d i s c r e t e

One observation
K

to the linear

T.K

of course,

t h a t if

As w e l l he s h o w e d

for details.

linear m a p s w i t h

That

Since the quo-

of a p r o d u c t

of

if e v e r y c o l l e c t i o n

intersection

v + V'
property

a n d p r o v e d all the e l e m e n separated

s u c h space
Here

since

to a

is isomorphic
is l i n e a r l y
KS

KT

are

o f the duality.

respectively

To get the r e q u i r e d
the usual v e c t o r

quotients

compact

linear m a p s

is the s t a t e m e n t

sum o f an S-fold,

in t h i n k i n g

t h e n linear c o m p a c t n e s s

T h e n linear c o m p a c t n e s s
f r o m finite

spaces.

a T - f o l d of co(-,-):cPxD

space

+ D

structure

topology.

that may be helpful

is finite

which

from a l i n e a r l y

s h o w e d that e v e r y
for a set

compact

linearly.

is a set of the f o r m

the n o t i o n

is to

uniformity.

an o p e n set to a point,

compact

(topologically)
so is

are c l o s e d u n d e r products,

linear

Lefschetz

in

V) w i t h the

defined

spaces

at

it is a subspace

is l i n e a r l y

subvariety

The s e c o n d

of l i n e a r l y

base

vector

category but

the d i s c r e t e

topologized

(when y o u i n d e n t i f y

space

dimensional

a *-autonomous

to a larger category.

is the same as s a y i n g

A continuous

is closed.

C = D = finite
already

considered

space w h i c h

spaces

compact

are at least two r e a s o n a b l e

D p- w e take the c a t e g o r y

a subspace

intersection.

separated

to

(a linear

V'

tary p r o p e r t i e s .

subspaces.

the s t r u c t u r e

topologized

has a non-empty

and closed

in fact,

is d i s c r e t e

this

spaces.

spaces,

, a l l spaces

(vector)

such a s u b s p a c e

that point becomes

is,

is a v e c t o r

t h a t the o p e n sub

discrete

must be equivalent
compact

of v e c t o r

they are isomorphic.

The first is to take

the usual duality.

take
For

and hence

fields

about

linearly

is e q u i v a l e n t

m a y b e t o u g h t of as the t r a n s f e r

to a r b i t r a r y

ones.

compact

to o r d i n a r y

spaces

is

topological

of the n o t i o n of

com-

C H A P T E R II.

E X T E N S I O N S OF STRUCTURE

i.

(1.1)

The m a i n goal in these notes is to c o n v e r t a p r e - , - a u t o n o m o u s

*-autonomous category.
and

The Setting.

T h a t is, g i v e n a V--category

w h i c h d e t e r m i n e an e n r i c h e d p r e - , - a u t o n o m o u s

subcategory

w h i c h contains

and can b e e q u i p p e d w i t h a *-autonomous

structure e x t e n d i n g the g i v e n s t r u c t u r e on

and

The r i g h t degree of a b s t r a c t i o n has p r o b a b l y not b e e n r e a c h e d here.

pose of e x t e n d i n g the s t r u c t u r e on

that

e q u i p p e d w i t h subcategories

situation, we w i s h to find a full

and

(1.2)

G c A

situation into a

and

is a c a t e g o r y of u n i f o r m objects

For the pur-

D , I h a v e found it e x p e d i e n t to suppose
in a closed c a t e g o r y

w h i c h is a semi-

variety.
In this chapter,

objects, ~
an o b j e c t

and

IcC

and a functor

nomous situation.
and for e v e r y

objects in
uniform

(-,-):C__
p x D - - + D

{A--+ D~}

w h i c h give a V - e n r i c h e d p r e - * - a u t o That every o b j e c t of

denote the full s u b c a t e g o r y of

UnV

of maps - or even of p s e u d o m a p s - w i t h each

The i n c l u s i o n o f

A--+ UnV

It is clear from the d e f i n i t i o n of

have a n e m b e d d i n g into a p r o d u c t in
subobjects and h e n c e u n d e r limits.
But if w e have

~ .

D~D

Given

AA ,

is called a D__-

has a left adjoint.

as the full s u b c a t e g o r y of objects w h i c h

that

is itself invariant under products and

Thus w e need o n l y verify the s o l u t i o n set condition.

UUnV ,

A~A

and a map

and m o r e o v e r lies in

A .

Thus all the algebras o f

exceed that o f

c o n s i s t i n g of all

A .

Pro~sition.

Proof.

(-)*:cP---+D,_
_

e v e r y o b j e c t of the form [V,D ] can b e e m b e d d e d in a p r o d u c t of

We n o w let

r e p r e s e n t a t i o n of

the c a t e g o r y of u n i f o r m

W e m a d e the f o l l o w i n g a d d i t i o n a l hypotheses:

V{V, D6D

D .

is such a category, U n V

V__-objects w h i c h can b e e m b e d d e d in a p r o d u c t of o b j e c t s o f

a family

(1.3)

then, V

are full s u b c a t e g o r i e s e q u i p p e d w i t h a V--enriched d u a l i t y

U---~ A , the image has c a r d i n a l i t y

and all p o s s i b l e maps of

that of

w h o s e c a r d i n a l i t y do not

to such algebras c o n s t i t u t e a solution

set.
(1.4)

For

AA

c o n s i d e r the d i a g r a m
AoP

>

(Un!) p

[-,~]

!(-,A)

+ I !(-'A)
V

with

inclusion,

v a r i a n c e exhibited,
V(U,A) ~ ~(~U,A)

the a d j o i n t above and

[-,A]

is left a d j o i n t to

and

V(-,A).

are the identity.

The a d j u n c t i o n

With the
I~

gives

w h i c h is the c o m m u t a t i v i t y r e q u i r e d to a p p l y [Barr,73], T h e o r e m 3 and

20

conclude

that the image of

is a c o t e n s o r e d
(1.5)

[-,A]

lies

in

This

It is easy to see that w e get a f a c t o r i z a t i o n

the s u r j e c t i o n s

and

have a surjection

is c l e a r l y

the cotensor.

Thus

V--category.

the e m b e d d i n g s

A'----+ A

s y s t e m on

by taking

(with the induced uniformity).

and an e m b e d d i n g

B - - + B'.

We w a n t

Suppose

to be
now we

to c l a i m that

V(A, B) -----~ V(A' ,B)

V_(A,B')---* Z(A' ,B')

is a pullback.
above s q u a r e

This

is so iff for e v e r y

is a p u l l b a c k

in

S .

V~V

Using

the functor

the e o t e n s o r

Hom(V,-)

adjointness,

applied

to the

we get a c o m m u t a -

tive square

Hom(A,[V,B])-----+ Hom(A',[V,B])

w h i c h w e need to k n o w
B'

, [V,B]

Hom(A,[V,B']) ~

Hom(A' ,[V,B'])

is a pullback.

is a s u b s p a c e

of

[V,B']

According

to

(I.3.13)

when

so that we h a v e a d i a g o n a l

is a s u b s p a c e

fill-in

of

in any c o m m u t a -

tive s q u a r e

AI - - . - - ~

Iv, B l - - . [ v ,
which
D

is e x a c t l y w h a t

a n d the r e m a i n i n g

(1.6)

Example.

is required.
structure

Let

of

vector

fulfill

for the p a i r

spaces

of v e c t o r

C, D

that is,

the k e r n e l

of

so t h a t the k e r n e l -- and w i t h
particular
to

this
A

the f o l l o w i n g

In fact,

contains
A - ~ D ix

,.the s u b c a t e g o r i e s

if

a finite
... x D~n.

the spaces

The

Not e v e r y

latter

for every o p e n s u b s p a c e

dimensional

is a linear

analogue

we say t h a t a space

BoA,

is a finite

there

Wel con-

CeC

is a power
topolo-

linearly

topologized

in

and every

product

discrete

KD

space
D 6D

of maps

A-+ D ,

is finite dimensional,
dimensional.

In

space and no such space

a D_-representation

to the p r o p e r t y

Namely,

h a v e linear

of the k e r n e l s

It is e a s y to see that a space does h a v e

b e i n g t o t a l l y bounded.

in

is e m b e d d e d

intersection

and

C = D = finite

T h e n every

it every o p e n s u b s p a c e -- is c o f i n i t e

is n o t true of any infinite

property which

o v e r the field

For the first w e take

Since

of 1 . 4 .

spaces

topology).

D- representation.

however.

then e v e r y o p e n s u b s p a c e

(with the d i s c r e t e

gies so does a n y t h i n g w i t h a

all the c o n d i t i o n s

and h e n c e h a s a D--representation.

has a D - r e p r e s e n t a t i o n

belongs

Thus the c a t e g o r y

be the c a t e g o r y

sider two p o s s i b i l i t i e s
dimensional

B' ]

of a u n i f o r m

is l i n e a r l y

number o f elements

iff it has
space

totally bounded
a I,

..., a n

iff

such

21

is in the linear

wording

that

of the h y p o t h e s i s

b y the f o l l o w i n g
Proposition.

A separated
completion

Proof.

If

dimensional.

A--+ ~ A / A

cause

A/A

I claim

and h e n c e

mensional
form

A0cA

subspace

A0

also t h a t

where

whence

B/B 0

is d i s c r e t e

the l i n e a r l y

of discrete

Now if

compact

of the l i n e a r l y

ones.

i.e.

full s u b c a t e g o r y .

temporarily
If w e w i s h

designate

we w i l l

dual of

A#

want

shows

in

that

it m a p s

is to t a k e

that

to give
= 0

be-

space

in the

compact.

to e a c h

is e m b e d d e d

in

finite diB

is the

contains

is l i n e a r l y

injectively

T h e n for any

an o p e n

topologized

to

A/A 0

and

to be the d i s c r e t e

spaces

and

iff it is l i n e a r l y

a subspace

of a p r o d u c t

in that case

consists

of the

the d u a l i t y

C~
V(C,A)

Duality.

on

CUD

to all o f

We are t r y i n g to d e f i n e

C-+ A

C-+ A

this means
we r e q u i r e

the internal

there

V(A,T)

, w h i c h we
A # - + C*

V(C,A)--* ~ ( A # , C *)

that

V(A,T)

A~

w h i c h means
C*

More-

in such a w a y that the

~#I ~ ~(A,T)

the c o a r s e s t

~(C,A),

~ V(V(A,T),V(C,T))

is to say t h a t w e e q u i p

a dual of

a pseudomap

equipped with

a n d each e l e m e n t

or at least to

s h o u l d b e one

h o m as w e l l

This r e q u i r e s

Accor-

uniformity

the c o r r e s p o n d i n g

is uniform.
with the

ele-

Another way

coarsest

unifor-

such t h a t

A#

is u n i f o r m

for all

C~C

A@

coordi-

of l i n e a r l y

and h e n c e

of

Then

Thus

to e x t e n d

under

mity

compact

a D--representation.

AA.

s u c h that c o r r e s p o n d i n g

this u n i f o r m i t y

is l i n e a r l y

If we h a v e a m a p

m e n t of

to d e s c r i b e

NA

of a p r o d u c t

to b e i n g

to b e the o b j e c t

V ( I A # I , Ic* I)

subspace

that is e q u i v a l e n t

is its i n t e r n a l h o m into

dingly we define

of

o p e n and shows

any n e i g h b o r h o o d

since

to each p s e u d o m a p

eventually

is

an o p e n sub(vector)

consisting

of

to h a v e the d u a l i t y b e V__-enriched

that corresponding
over,

BoA

Extension

Let

A #.

is c o f i n i t e

combine

spaces.

T h e first task is to e x t e n d

a large

A cA

A-+ A/A

T h e n a space h a s a D - r e p r e s e n t a t i o n

to h a v i n g

topologized

HA/A

that

This

and

2.

(2.1)

For as n o t e d e a r l i e r
spaces,

contains

and also l i n e a r l y

is finite d i m e n s i o n a l
for

iff

compact.

suppose

U DB 0 = BNA 0

totally bounded

In fact its kernel

is t h e n a c l o s e d

is a n e i g h b o r h o o d

The o t h e r c h o i c e

topologized.

first

(see [ L e f s c h e t z ] ) .

U = BAY

of

This is e v i d e n t l y

is l i n e a r l y

, A/A 0

The maps

is an embedding.

of

is l i n e a r l y

then every o p e n s u b s p a c e

dimensional.

image of the s u b s e t of

To see the c o n v e r s e
open subspace

bounded,

Any neighborhood

The c l o s u r e

space

compact.

totally

in all others.

spaces

and
al, ..., a
. Of course this is just a ren
is finite d i m e n s i o n a l .
The a n a l o g y is h i g h t e n e d

topologized

is finite

which

is the inverse

compact

linearly

A/A

is separated.

in the product.

A/B

is l i n e a r l y

is l i n e a r l y
Thus

a map

nate a n d

that

fact.

its u n i f o r m

which

s p a n of

Since

A#

, [ V(C,A) ,C* ]

can h a v e o n l y a set o f

uniform

covers,

only a

22

set of

cluded

need be

among

used

from any one

the objects

of

A#
among

these

maps.

This

map

used)

By t a k i n g

C = I

(which m a y

always

be

in-

we get

~ [Z(I,A),I*] =[ IAl, T]

is an i n j e c t i o n

since

the underlying

map

in

factors

as

injections

IA#1

Thus t h e r e

(2.2)

is a f a m i l y

This

IBI

means

) V(A,T)

= V(A,T)---+

{C

that

such

to m a p

Second

we

V(IAI,ITI) ~

that

A#

is e m b e d d e d

B---+ A # , w e

need

I[ IAI ,T]

require

in a p r o d u c t

two things.

B--+ [V(C,A),C*]

I BI

I -

K[V(C

First,

, for e a c h

C6C

we

such

,A),C~]

need

a map

that

~ V_(A,T)

l[v(c,A),c* ] I --- v(v(c,A) ,V(C,T) )

commutes,
~(B,C*)

the right
such

hand

map being

composition.

This

is e q u i v a l e n t

to m a p s

~(C,A)

that

V(C,A)

V(B,C*)

V (V (A,T) ,V(C, T) )

commutes.
(2.3)

Thus

isomorphic
second

datum

morphism.
for

to

to

map

C *-+ C #

~(C,T)
required

The

take

coherence

CeC

take

we

the

need

itself

the

< C*>

of

V(C*,C'*)

is t r i v i a l .
so t h a t

first

composite

~(C',C)~

is a m a p

required

we may

for

s o we c a n

the

~ < C#>

canonical

for w h i c h

On t h e o t h e r
identity

gives

we

hand

But each

is

isomorphisms.
take

among

The

the duality

iso-

the candidates

a map

c#--+ [ v(c,c) ,c* l

which
(2.4)

composed
Now

with

let

DeD

D*

is i s o m o r p h i c

may

follow

get

each

D ~---~
"
~C*

the
.

unit

I--+ ~(C,C)

We k n o w

that

to a s u b o b j e c t
with

the name

, and hence

of

gives

there
HC*

~ D*

C*

Thus

is a C - g e n e r a t i n g
.

Since

I---+ V, (D C) _

D#

C# - +

C* N C #

family

D#---+ I[ [ V ( C

,D),C~]

of the corresponding
To go t h e o t h e r

way,

{C --~ D}

we

maps
require

such

is a map,

that
we

C~---+ D

to

for a l l

CC

V (C, D*) ---+ V (D,C*)

for which

we

take

the duality

isomorphism.
#

We h a v e
write

A*

now established

instead

of

A#

that
We n o w

()
turn

is an e x t e n s i o n
to t h e

of

functionality

()

and will

of this

henceforth

operation.

23

(2.5)

Proposition.

Proof.

For any

CC

, c o m p o s i t i o n of maps gives a m a p

!(C,A)----+ V(A*,C*).

This is the same as a map

A*

~ [ V(C,A) ,C* ]

w h i c h w e have.
Corollary.
Proof.

For any

A , B A , c o m p o s i t i o n of maps gives a m a p

V(B,A)---~ V(A*,B*)

We r e q u i r e a map

A*

~ [ V(B,A) ,B* ]

We c e r t a i n l y have a m a p
V_(A,T)~
Moreover,

for any

C6C

V(V(B,A),V(B,T))

w e have

A* - ~

[ Z(c,A) ,c*]

--~ [ V(B,A)~V(C,B) ,C* ]

[ V(B,A) ,[ V(C,B) ,C* ] ]
and h e n c e there is the r e q u i r e d map
the fact that
(2.6)

IV,-]

A*--+ [V(B,A),B*]

Note that this a r g u m e n t uses

p r e s e r v e s embeddings.

This shows that

(-)* : A --+ A

is a V-functor.

There is nothing to guarantee

that it is an e q u i v a l e n c e and it is in fact u n l i k e l y that it is always so.

Since

I e C , c o m p o s i t i o n gives a m a p

A*---+ [V(I,A),T]
IAI = !(I,A)---+ !(A*,T)

>** I

and the b e s t y o u can u s u a l l y h o p e for is t h a t the above map

morphism.
w h i c h gives

It is always a m o n o m o r p h i s m .
{A**--+ D**~ ~ D~}

I A!--+

I A** I be an iso-

For there is a D - r e p r e s e n t a t i o n

Thus w e h a v e

IAI

I A**I --+ ~ ID I

{A---+ D }
is a m o n o m o r -

p h i s m and h e n c e the first m a p is.

(2.7)
in
AAA
Let

Let
IA** I
since

AA

A A c A ** .

{A---+ D }

so that w e have
an isomorphism,
(2.8)

denote the s u b o b j e c t of

b u t such that

AA

A**

We h a v e

I AAI

~ I AI

be a D - r e p r e s e n t a t i o n of
A A---+ A

is uniform.

q u a s i r e f l e x i v e if

The c o n d i t i o n that

such that

IAA I

has the induced uniformity.

IAI

by the inverse of the above inclusion.

A .

Then each

We say that

A A = A**

is the image of

There is no q u e s t i o n of

A--+ D

gives

is p r e r e f l e x i v e if

A**----~D**~ D
AA--~ A

is

and r e f l e x i v e if b o t h of these hold.

be q u a s i r e f l e x i v e

is e q u i v a l e n t to the a s s e r t i o n that

~(I,A) ~ !(A*,I*)
or that e v e r y p s e u d o m a p
sarily unique -- o f

A .

A*--+ T

is r e p r e s e n t e d b y e v a l u a t i o n a t an e l e m e n t -- neces-

24

(2.9)

Proposition.

Suppose

(i)

is c o s m a l l

in

(ii)

is i n j e c t i v e

(see proof);

in the V--category

with

respect

to the class of

embeddings;
(iii)

is c l o s e d

under

finite

h a v e the u n i v e r s a l
Then every p s e u d o m a p
Proof.

sums and

mapping

f : A*---+ T

under

properties

is r e p r e s e n t e d

We can f i n d a family of p s e u d o m a p s

finite p r o d u c t s

for p s e u d o m a p s

by evaluation

and these

as w e l l as maps.

at an e l e m e n t of

{C--+ A} such that the h o r i z o n t a l

arrow

in the d i a g r a m

A*

~ EC*

is an embedding.
that
say

be c o s m a l l

C[ x

d u c t in
_C p - - +

Then we have

Since

f# : ~C* -

m e a n s t h a t such a m a p

... x C*n "

~

D
_

a map

Since

_D

x(C 1 +
whose

Example.

are l i n e a r l y

tinous

l i n e a r map,

B 0 = A 0 NB

jection

choices

where

A0

is e a s i l y

spaces,

and

Since

(-)*

For if

the former

the h y p o t h e s e s

subspace
in

is e m b e d d e d

f .

till t h e n e x t section.

is e m b e d d e d

m u s t b e an o p e n

b y an

C 1 + ... + Cn---+ A

seen. to r e p r e s e n t

of

(2.10)

, in fact for any choice

is an o p e n s u b s p a c e

and b o t h are discrete,

is r e p r e s e n t e d

under the p s e u d o m a p

is p o s t p o n e d

of v e c t o r
of

topologized.
its k e r n e l

is a l s o their p r o -

so w e h a v e

t h a t the lower p s e u d o m a p

of the d u a l i t y

In t h e e x a m p l e

spaces

1.6,

it follows

The image of that e l e m e n t

(2.10)

of f i n i t e l y many,

this

s u m in

The h y p o t h e s i s

(CI+...+C)*

are the ones g i v e n o r i g i n a l l y

fied for b o t h p o s s i b l e

f .

... + C n

discussion

a product

finite p r o d u c t s

, that is their

components
Further

through

extends

> ~C*

T c
C

which

"'" + Cn)* ~ C1* x ... x C*n

i
for

under

... + Cn{ ~
(C 1 +
A*

From duality

factors

is c l o s e d

C1 + C2 +

is an equivalence,

> T

in

B0 c B
.

Since

and
.

are satis-

for w h i c h the
B---+ K

is a con-

T h e n as we showed

B/B0---+ A / A 0

in the latter.

in

is an in-

Then we have a

diagram

B ----+ B / B 0

I
A
and t h e n o r d i n a r y
A/A 0

is discrete.

vector

space

~ K

I
~ A/A 0

theory provides

the r e q u i r e d

A/A0----+ K , c o n t i n o u s

since

25

3.

(3.1)

As with the duality,

Extension of the Internal H D m

we wish to extend the functor

(-,-)

: C p D D

to a

functor denoted
~(-,-)
Although we are for convenience
is not in general.

: A_Op A A .

using notation suggesting

It is not generally

symmetric,

(3.2)

To b e g i n with we require
uniformity

for every pseudomap

C A

A(A,B)

Chapter

III is devoted to finding

on which it is well-behaved.

give it the coarsest

that

it

does not always have an adjoint and

when it does, the tensor is not always associative.

a nice subcategory

this is an internal hom,

~(A,B) I

and every pseudomap

have the weak uniformity

[ V ( C , A ) V(B,D), (C,D)].

= ~(A,B),

the V--valued hom.

such that the pseudomap

A(A,B)

B D .

Second we will

(C,D)

is uniform

More abstractly,

determined by all

CeC, D ( D and

This makes sense for the underlying map in

we require

A(A,B)---+
V

is

V(A,B) --+ V(V(C,A)~(B,D),V(C,D))

which is the transpose

under adjunction to composition

V(C,A) V(A, B) Z(B, D) --+ ~(C,D)

Since there is an epimorphic

D~)

family

{f~

: C~--+ A)

and a monomorphic

family

{g~ : B

, there is a m o n o m o r p h i s m
V(A,B)--+ ~!(C

, D~)

This map factors

V(A,B)--+ nZ(Z(c

,A)!(B,D~) ,Z(C ,D~))-+ H V ( I I , ! (Cm,D~)) ~ ]IV(Cw,D ~)

where the second map is induced by the names of the

is an injection as well.
{D~}

of objects of

Hence for some

and

f~

and

(not necessarily

respectively,

A(A,B)

g~

Thus the first map

the same)

is embedded

families

{C ) and

in

H[Z(Cm,A)~V(B,D~), (C ,D~)]
(3.3)
Proof.

Proposition.
Since

(-,-)

If

C~

and

DED

: C p x D + D

, A(C,D)

is canonically

isomorphic

is assumed to be a V_-functor,

to

(C,D)

there is for each

D~D, a natural map

which expresses

the fact that

v(c',c)

V((C,D),(C',O))

(-,D)

is a V-functor.

Similarly,

there is for each

C6C , a map
V(D,D')
Each of these maps lies alone ordinary
and replacing

by

V(C' ,C)V(D,D')

C'

+ V((C,D), (C,D'))

composition of functions.

Putting these together

in the second we get a map

V((C,D), (C',D))!((C',D), (C' ,D')) !((C,D), (C',D'))

the second map being composition.

By the cotensor

adjunction,

this gives a map,

26

(C,D) -+ [ V ( C ' , C ) V ( D , D ' ) , (C',D')]

for any

C'c~ , D'D

This implies that

(C,D)--+ A(C,D)

is uniform. To go the o t h e r

way w e o b s e r v e that
A(C,D)--+ [V(C,C)~V(D,D), (C,D)]
m u s t b e uniform.
maps of

and

We may compose this with the map induced b y names of the identity
D r e s p e c t i v e l y to get

A(C,D)

[ II,(C~)] ~

(C,D), the latter isomor-

p h i s m coming from
V(A,[ I,B]) ~ V(I,V(A,B)) ~ V(A,B)
from w h i c h
uniform.
V(C,D)
(3.4)

[ I,B] ~ B

b y the Yoneda lemma.

This implies that

A(C,D)-+

(C,D)

is

It is easy to see that both of these maps lie over the identity m a p on

.
Proposition.

Proof.

The b i f u n c t o r

A(-,-)

: A_p x A + A

is

(or lifts to) a V_-functor.

We m u s t s h o w that there are natural maps

V(A',A)--+ V(A(A,B),A__(A',B))
V(B,B')--+ V ( A ( A , B ) , A ( A , B ' ) )

for all

A,A',A,B'{A .

We do the first, the second b e i n g similar.

T h e n we require a

map

A(A,B) ~ [V_(A' ,A),A(A' ,B)]

w h i c h means first a map

V(A,B)-+ V(V(A',A),V(A',B))

and second for all

CE~, D D ,

A(A,B)--+ [V(A',A),[ V_(C,A')eV(B,D), (C,D)]]

by

(I.3.13).

The first is just c o m p o s i t i o n and the second comes from

A(A,B)--+ [V_(C,A)V(B,D), (C,D)]
-~ [V(A',A)eV(C,A')eV_(B,D), (C,D)]
[ V ( A ' , A ) , [ V ( C , A ' ) ~ V ( B , D ) , (C,D)]]

(3.5)

Proposition.

(c ,D~)]

Suppose

A(A,B)

is e m b e d d e d in the p r o d u c t

9[V(C

. Then there is a c o m m u t a t i v e d i a g r a m

_A(A,B)

+ ~[V(C
,A),_

(Cm,B)]

~[V(B,D~),(B,D~)]----+
K[V(C ,A)V(BrD~),(Cm,D~)]
e,~
The upper m a p and left h a n d m a p are embeddings.
Proof.

From

(3.4) w e h a v e a map for all

V(C

w h i c h transposes to

~,

,A)--+ V ( A ( A , B ) , A ( C

,B))

,A)V(B, D ) ,

27

A(A,B)--~ [V(C ,A) ,A(C ,B)]

and is the

component

of a map
A(A,B)--+ n[V(C

Similarly,

there are maps

which transpose

Cotensoring

,A)

component
~

the one factorization

CC

i.

If

is embedded
2.

For any

Corollary

3.

There

There

H[V(C

is a canonical

Finally

the pro-

embedded

in

in a product

,A),

(C ,D)]

~[V(A,D~),

from

~[V(B,D~),_ (C,D~)]

; if

DD,

A(I,A)

I(I,A) I .

~ A .

For some family

D~]

while

(I,A)

{D~}

of objects

is canonically

of

em-

(I,D~)] .

map

~(A,B)--~[
We have,

follows

K[V(A,D~),_

is a canonical

The last property

systems.

isomorphism

IAI = V(I,A)

in the isomorphic

Proof.

we get

(Cw,D~)]

,A)V(B,D$), (C ,D~)]

is embedded

A A , A* ~ ~(A,T)

is canonically

4.

is a V-functor,

and the other is analogous.

, A(C,B)

By hypothesis,

Corollary

_~[~(C

of factorization

in a product

Corollary

bedded

,D~)]

,A) ,A(C ,B)]--+ n [V(C ,A)V(B,D~), (C ,D~)]

Corollary

, A

[Z(B,D~),(C

,B)-~

g i v ~ a map

properties

Proof.

A(C

of a map to

from cancellation

A(A,D)

Z(A(Cw,B), (C ,D~))

,A) ,A(cm,B)] ---+ [!(Cm,A)Z(B,D~),

~[!(C
This gives

V(B,D~)-+

and using the fact that the cotensor

[Z(C

duct over all

~,~

to

~(C

This is the

for all

,A),A(C ,B)]

Im , B]

(3.4), the canonical

map

Iil = Z(I,A) -+ V(A(A,B),~(I,B))

Z V(A(A,B),B)

which has the transpose

i(A,B) ---+ [ IAi ,B]

(3.6)

Proposition.

Suppose

and

are reflexive.

The first is thus a split mono and the second a split epi.
of the first and hence

and

~(B,D)

Thus

~ ~(D*,B*)

~ V(B*,A*)

Ce~

A(B*,A*)--+

and

In particular

Thus we have

DD

[V(D*,B*)!(A*,C*)

~ A(B*,A*)

V(A**,B**)

is also a split mono from which

V(A,B)
for

V(A,B)-+~*,A*)-+

A(A,B)

lence.

are isomorphisms.

we have

Then

Since

instance

(-)* is a V--functor,

Proof.

it follows

we have

an equiva-

The second

V(C,A)

is an

that both
~ V(A*,C*)

, (D*,C*)]

[ V(C,A)e!(B,D), (C,D)]
so we have

~(B*,A*)-+

A(A,B)

and similarly

in the other direction.

Note that

(-)*

28

is not in general

an

functor.

In fact if

is q u a s i - r e f l e x i v e

and not reflexive,

the m a p
B ~ A(I,B)

.--+ A(B*,T)

= B**

is n o t uniform.
(3.7)

Let

CeC,

De~

is the p s e u d o m e t r i c

a s u b s e t of

d e f i n e d on

~ = }(C)

(d,S),

is a c o l l e c t i o n

S~,

uniformity
sets in

d
on

is separated.

in

For if

such t h a t

which

d(fx,gx)

(-,-)

#(C)

~ 0 .

there

in

sists of s o m e t h i n g
requirement

%(C)

is an

is s o m e t h i n g

is a c a n o n i c a l

is s e p a r a t e d

there

structure

uniformity

of p s e u d o m e t r i c s

If the u n i o n of the

t h e n this

union,

is g i v e n

for every p s e u d o m a p

sets or all

for each

for

CE~

of u n i f o r m

~(C)

sets or the like.

that there is a natural

of the functor
a

conver-

f : C C'

In practice,

finite

structure

hence an e l e m e n t

o n the values

has the s t r u c t u r e

f(S)~S'

separated)

is a p s e u d o m e t r i c

if there

(C,D)

such t h a t

all singletons)

is an e l e m e n t of the dense
D

~(C,C')

there

then the c o l l e c t i o n

includes

and if, m o r e o v e r ,

to s u p p o s i n g

which

in

like all c o m p a c t

(d,S)

on the sets

such that

S'~(C')

amounts

of

I x~S}

convergence
if

is a c o n v e r g e n c e
C

. Then

not n e c e s s a r i l y

Since

of

a pre-(i.e,

there

on

, defines

We say that the u n i f o r m

of subsets

gence o n the sets

(e.g,

f ~ g

fx # gx

: C__
p x D D

S(#(C)

C , t h a t of u n i f o r m

is dense

x~S~

family

on

a pseudometric

: sup {d(fx,gx)

o f subsets

a pseudometrie

and

by

(d,S) (f,g)
If

V(C,D)

and

usually

con-

The second

transformation

[--+ H o m ( ~ ( C ) , # ( C ' ) )
At a n y rate

like a V - e n r i c h m e n t .

it is e n o u g h

to g u a r a n t e e

that

~(C,C')----+ V((C',D), (C,D))

for any

DeD

The s i m i l a r
V__(D,D' ) ---+ V_((C,D) , (C,D') )

for any
(3.8)
US

CeC

We n o w let
where

~(Ci)

nerate

~(A)

Dually

because

# (A)

for each

n
set in

HD

exists

the inverse

denote

{B D

Si

{C --+ A}

in this manner,

if

the family of a l l subsets

i ~ l,...,n,

If a set

(3.9)

Proposition.
~(A)

{A(A,B)---+
Proof.

Let

If

uniform convergence

A,B~

on

represents
Then

(A,B)

A#(A,B)
~(A)

with

of the f o r m

SIU...

C.-+ A
of a
l
is s u f f i c i e n t
to ge-

C C

generates

such that

is e m b e d d e d

in

B .

A(A,B)
A

of

is a p s e u d o m e t r i c .

same p s e u d o m a p

{ C - - + A}

of p s e u d o m a p s

generates

represents

momentarily,

of p s e u d o m a p s

we say that the f a m i l y

{ C ~ - + A}

(C , D~)}

Let,

is the image under

is a c o l l e c t i o n

, t h e n w e say that the p r o d u c t

gence on

image of a p s e u d o m e t r i c

has the u n i f o r m i t y

and

{B D~}

of u n i f o r m c o n v e r -

represents

then

denote

V(A,B)

e q u i p p e d w i t h the u n i f o r m i t y

T h e n a b a s i s of p s e u d o m e t r i c s

on

A#(A,B)

of

consists

of

29

(d,S)

a pseudometric

(Ti), Ti~(Ci),

on

A , S~(A)

fiV(Ci,A)

Also let

Then it is easily seen that

m.

on

H(Ci,Dj)

A(A,B)

A_#(A,B)

To go the other way,

But

t~(fi)Ti~(A)

sets
B

di

V(C

,A)

on

and so

let

~(C ) .
sup

Similarly,

_V(B,D~)

on

(A,B)

(3.10) We say that a map

dominates

the pseudometric

of pseudometrics

A(A,B)

in

K(C ,D~)

Thus it follows

Then there is

A'--+ A

just as

Suppose

arising

A(A,B)

sup(di(gi,gi))

on

from a finite number of

is a pseudometric

on

A__#(A,B)

is uniform.
is dominating

is clear that this is dual to an embedding.

Proposition.

and

be a p s e u d o m e t r i c

Thus every p s e u d o m e t r i c o n

A__#(A,B) + A(A,B)

~C~ A

is embedded

and is thus already realized by a finite number of elements of the

and sets from

B D~

A#(A,B)

fi : C. A, gi : B Di, i = l,...,n , sets

Ti#(A i)
1
such that d ~ sup{(di(gifi,gifi),Ti)}, i = 1 ..... n .

D.i

must b e m a j o r i z e d by the

pseudomaps

Thus

function

is uniform.

a finite set o f pseudomaps,say

and pseudometrics

, i = 1 ..... n, j = i,...,

is such that any pseudomap

_V(C~,A)

(#fi)

, gjcZ(B,D j) , j = 1 ..... m .

induced bv the composite

is uniform when it is induced by elements of

A(A,B)

S = SlO...US n , S i =

d = sup{dj(gj,gj)}

A#(A,B)
But the uniformity

that the map

Let

(d,S) = sup{(dj(gjfi,gjfi),Ti)}

But this is just the pseudometric

H(C ,D~)

(see 3.7).

{B + D~}

A'~-+ A

provided

A family

represents

is dominating

and

~(A')

maps onto

#(A)

{C -+ A}

generates

iff

B B'

B ~D~

It

iff

is an embedding.

is an embedding.

Then each

map in the square

A(A,B)

> A(A',B)

A(A,B')

> A(A' ,B' )

is an embedding.
Proof.

Just repeat the previous

No special property of

argument.

or

was used

are families

such that

in that argument.
Corollary.
~A---~ A

Suppose

{A~---+ A}

is dominating

and

and

B ~B

{B --+ B~}

Then each map in the square

is an embedding.

A(A,B)---+ ~A(A

,B)

~A(A,B~)---+ ~A(A ,B~)

is an embedding.
(3.11) Proposition.
and

{ E - - + D E}

Proof.
(C~,D~)}

Let

represent

We k n o w from
represents

{C~---+ A}
E .

(3.9) that
it.

Then

and

{C~ --~ B}

A(A,A(B,E))

~(B,E)

dominate
is embedded

is embedded

A second application

We note that this implies that

of

in

A
in

H(C~,D~)

and

B , respectively,

K(C , (C~,D~))
so that

(3.9) yields the result.

A(A,A(B,E))

is embedded

I[[A(C ,A)A(C~,B)~A__(E,D E), (C , (C~,D~))]

in

{A(B,E)--~

30

By s y m m e t r y so is

A(B,A(A,E))

This does not imply that they are isomorphic but it

does imply that any inclusion b e t w e e n

as a s u b o b j e c t o f

~ ( I A ~ IBI ,IEI )

V(A,A(B,E))

(3.15).

V(B,A(A,E))

under the i s o m o r p h i s m w i t h

m a t i c a l l y lifts to an inclusion b e t w e e n
can do now is

and

A(A,A(B,E))

and

(each c o n s i d e r e d

V(IBI~IAI ,IEI))

A(B,A(A,E))

auto-

The b e s t w e

W h e n we introduce completeness h y p o t h e s e s in the next chapter,

we will have b e t t e r results

(3.12)

Proposition.

Suppose

A,B,B'6A

and

B --~ B'

is an embedding.

Then there is

a canonical p u l l b a c k
V(A,B)--

--+ V(A,B')

Proof.

V(IAI,]B[)--

+ V(IAI,IB' l)

An element of the p u l l b a c k is a p s e u d o m a p

torization

IAI ~

IBI -+

underlies a u n i f o r m

IB'I .

Since

is e m b e d d e d in

IAI ~

IBI

) A(A,B')

I
[IAI

The d i a g r a m in

,B]-

-+

[ [AI , B ' ]

w h i c h underlies this is a p u l l b a c k so the actual p u l l b a c k

has the same u n d e r l y i n g V--object as

~(A,B)

it cannot b e c o a r s e r t h a n that i n d u c e d by
uniform.

, the map

A --+ B .

A(A,B)

~(A,B)

for w h i c h there is a facB'

There is a canonical p u l l b a c k

C o r o l l a r y i.

Proof.

A - - + B'

with, possibly,
A(A,B')

a coarser uniformity.

But that means that the u n i f o r m i t y on the p u l l b a c k is the same as that of

a n d so they are isomorphic.

C o r o l l a r y 2.

If

{ B B m}

is a family such that

B + ~B~

is an embedding,

then

V(A,B) .... --+ KV(A,B )

4,

4,

Z(Im,[B!) ----+ n v ( I m , I B

I)

is a pullback.
We may put these t o g e t h e r to conclude,
Corollary

3.

U n d e r the h y p o t h e s e s of C o r o l l a r y 2, there is a canonical pullback,

A(A,B)----+ K A ( A , B m)
+
[IA,,B]

(3.13) Proposition.

Let

But

, else the induced map w o u l d not be

.... n[ I m , B j

A , B A , C 6 ~ . T h e n there is a canonical inclusion

31

A(A,A(C,B)

Proof.

As n o t e d in

c A(C,A(A,B))

(3.11) it is s u f f i c i e n t to show that there is a canonical inclusion

V(A,A(C,B))
Let

{C~ A}

and

{B + D~}

k n o w from 3.9 that the family

dominate

c V(C,A(A,B))
A

and r e p r e s e n t B , respectively.

{A(A,B)--+

(C ,D~)}

represents

~(A,B)

T h e n we
T h e n from

C o r o l l a r y 2 above there is a c a n o n i c a l p u l l b a c k
V(C,A(A,B)) ----~ nZ(C , (C ,D~))

!(Icl ,~(A,B)) ~

~!(lCl , (C ,D~))

and it is thus s u f f i c i e n t to find c a n o n i c a l maps

~,~

and

V(A,A(C,B))---+ ~( ICI , ~(A,B))

the c o m m u t a t i o n of the square.

V(A,A(C,B)) --+ V(C, (C ,D~))

for all

The fact that they are canonical implies

The first is the c o m p o s i t e

V(A,A(C,B)) ~

V(C

, (C,D0)) ~ Z(C, (Cw,D~))

and the second the composite

V ( A , A ( C , B ) ) - + Z(A,[ IcI ,B]

w h e r e the first map comes from C o r o l l a r y 4 of

(IcI ,V(A,B))

(3.5) and the second is the cotensor ad-

junction.
Corollary.

A map

(3.14) Example.
discrete f i e l d

A A(C,B)

exchanges to a m a p

C ~(A,B)

We again c o n s i d e r the c a t e g o r y of t o p o l o g i c a l vector spaces over the

K .

For both p o s s i b l e choices o f

and

is discrete w h i c h is that o f u n i f o r m c o n v e r g e n c e on all of

of all subsets o f

C .

When

C = D =

the later d e v e l o p m e n t t h a t the functor

a * - a u t o n o m o u s structure.
pact and
tor

D , the t o p o l o g y on a
C .

finite d i m e n s i o n a l spaces,
A(-,-)

~(A,-)I

commutes w i t h limits.

#(C)

(C,D)

consists

it w i l l follow f r o m

d e s c r i b e d in this section a l r e a d y gives

W h a t w e w i l l do here is to show that w h e n

D = d i s c r e t e spaces, w e do not get a c l o s e d category.

V(A,-)

Thus

Thus to show that

limits, i t is s u f f i c i e n t to show it has the r i g h t topology.

For

C = l i n e a r l y comA

A(A,-)

f i x e d the funccommutes w i t h

But this follows from

(3.9)

It is n o w a n easy a p p l i c a t i o n o f the s p e c i a l a d j o i n t functor t h e o r e m to show that

~(A,-) has an a d j o i n t w h i c h w e will designate

-~A .

Since

A(A,-)

the a d j u n c t i o n is strong w h i c h means t h e r e is a n a t u r a l e q u i v a l e n c e

(B,E))

a n d if

A(-,-)

is a V--functor,
~(AB,E)

= ~(A,

is to b e a c l o s e d c a t e g o r y structure this will h a v e to lift to

a natural equivalence

A(AB,E) ~ A(A, (B,E))

Let us now suppose that this e q u i v a l e n c e w e r e valid.

T h e n for

E = K , w e get

(A~B)*

32

A(A,B*)

. Let

~(KT,s'K) ~
reflexive,

that

topology on
tinuous

A = KT

(TxS)-K

and

B = KS

so that

I ABI ~

AB

with

~SxTI.

infinite sets.

Then

(A,B*)

This implies, since e v e r y space is quasi-

L o o k i n g ahead, w e will see in the next chapter the

is c o a r s e r than that of

(= uniform).

and

(AB)** = K SxT

(AB)**, i.e. that

(AB)** ----+ AB

is con-

Since every map from a linearly c o m p a c t space is closed, this is

an i s o m o r p h i s m and so

AB ~ K SxT

The a l g e b r a i c tensor p r o d u c t o f

c o n s i s t s of those functions

SxT--+ K

fi[K S, g f ~ K T which, w h e n

and

w h i c h have the form

are infinite,

IAI
and
IBI
n
(s,t) [ => i~Ifi(s)gi(t) ,

is not all of K ST .

the a l g e b r a i c tensor p r o d u c t e q u i p p e d w i t h the topology i n d u c e d by

N o w let

be

K SxT. The identity

map

IKSIIKTI

IE

t r a n s p o s e s to a m a p
IKS I --'+ V( IKTI
On the o t h e r hand,

,IE!

the i s o m o r p h i s m
KSK T

c o n s t r u c t e d above o n the h y p o t h e s i s that

nonical m a p

>
~

K SxT

was a c l o s e d c a t e g o r y transposes to a ca-

KS___+ A ( K T, KSxT)

w h i c h gives
~ S l _ _ ~ v ( K T ' K SxT)
From

(3.12) w e get a m a p
IKSI --+ ~(KT,E)

Since

~(KT,E)

is e m b e d d e d in

A ( K T, K ST )

the above map

KS

) A ( K T, K SxT)

factors

as a m a p
KS---~ A ( K T, E)
w h i c h t r a n s p o s e s to
K S K ~---+
Since
A(-,-)
(3.15)

K S K T ~ K SxT

this implies that

E .

E ~ K SxT

w h i c h is, as n o t e d above,

false.

Thus

does n o t give a n internal hom.

The c o u n t e r e x a m p l e suggests its own resolution.

The p r o b l e m arises out of the

n e c e s s i t y of the tensor p r o d u c t of two objects of

first that w e stick to the full s u b c a t e g o r y of

c o n s i s t i n g of complete objects. For

to lie in

C .

in that case w e w o u l d b e forced to use a c o m p l e t e d tensor product.

o r - l e s s obvious,
is

KST~

the

as required.

Since, as is m o r e -

c o n s t r u c t e d above is the actual tensor product,

However,

this leads to another diffficulty.

sure that the dual o f a complete o b j e c t is complete.

discontinuous

This suggests at

(or non-uniform)

o t h e r d i f f i c u l t i e s as well.

its c o m p l e t i o n

We c a n n o t be

It c o u l d be c o m p l e t e d

only by

maps w h i c h is o b v i o u s l y u n d e s i r a b l e and w o u l d cause

This p r o b l e m is s o l v e d by u s i n g a m o d i f i e d notion of com-

p l e t e n e s s w h i c h does not get in the way as much w i t h the duality.

C/{APTER ]

C A T ~ G O R Y G.

i. C o m p l e t e n e s s .

(i.i)

We suppose

henceforth

in their u n i f o r m i t y .
c l o s u r e of A
A~

of

belongs

to

in that e m b e d d i n g
The h y p o t h e s e s
UnV

we suppose

C *~--+ A*

in d e c i d i n g

suffice

to

whether

t h a t if

object

as w e l l as these of
A

is e m b e d d e d

and is,

in fact,

to g u a r a n t e e

that

in

~D~

are c o m p l e t e
, then the

the u n i f o r m
A~

completion

is a d m i s s i b l e

so it

A~A

is a p r o p e r

(or e q u i v a l e n t l y

Since a completeable

cient,

of

t h a t if

is also c o m p l e t e

of 1.3.10

and t h e n e v i d e n t l y

In a d d i t i o n
the i n d u c e d m a p
(1.2)

that the o b j e c t s

F r o m this it follows

closed

s u b o b j e c t of a

V(C,T)---+ V(A,T))

is d e n s e l y

embedded

C6~ then

is not injective.

in its completion,

it is suffi-

every d i a g r a m
l

-+ B 2

A
in w h i c h
map

B1

B~-- A

sion.

) B2
.

is a dense

Take the i d e n t i t y

is the i d e n t i t y
Since

embedding

TO see this take

on

is dense

in

can b e c o m p l e t e d

B1 = A

, B 2 = A~

for the v e r t i c a l

, the maps
A~

A--+

map.

, they are equal

If there

A~

to a c o m m u t a t i v e

and the m a p b e t w e e n
is a r e t r a c t i o n

a n d the i d e n t i t y

and the i n c l u s i o n

of

on

d i a g r a m by a

t h e m the incluA~----+ A

A~

which

agree o n

A--+ A~

is an isomor-

phism.
Later

in these notes w e w i l l h a v e o c c a s i o n

ness g o t t e n b y r e s t r i c t i n g
us suppose
complete
a

that

the class of

is a class

g : B2---+ A

(necessarily

(see

Proposition.
(II.3.7)).

Proof.

We h a v e

elements

of

Ai

Then
that

Suppose
AA

as these of

is

The obvious

S6~(A*),

ded

, aeA

is r e f i n e d b y the u n i f o r m

f : B~---+ A

is dense)

AA

such t h a t

is

Let
~ -

, there

is

gm = f .

has a c o n v e r g e n c e

and

in

F(a,d,S)

We m a y thus

t h i n g to do h e r e

is a b a s i s of u n i f o r m

{F(a,d,S)
where

and

is an i s o m o r p h i s m

is the set of p s e u d o m e t r i c s

and e v e r y

uniformity

~ - complete.
~ IAI

in

If

~ - complete

to s h o w that there

in

of c o m p l e t e -

is required.

We say that an o b j e c t

since the e m b e d d i n g

do that w e h a v e
A

notions

follows.

is

I AAI

weaker

for w h i c h a fill-in

embeddings.

: BI- ~ B 2

unique

The m a i n p o i n t of this g e n e r a l i t y
(1.3)

BI--+ B 2

of dense

if for e v e r y p s e u d o m a p

to c o n s i d e r

on

T , then a b a s i c
I aeA

= {b

is to use

covers of

uniform

F#(a,d,S)

= {b

[ aeA}

I d(~a,~b)

~
{b
~eS

(I.2.5).

the
To

b y sets c l o s e d
cover of

A A is

}
I d(~a,~b)

< 1

for all

cover
{F#(a,d,S)

AA

identify

, where
~ , for all

] d(~a~b)

~ %}

~eS}

~S}

Th~s

34

which
{b

is an i n t e r s e c t i o n

I d(~a,~b)

S %}

of c l o s e d

sets

is the inverse
A ~

Moreover
(1.4)

F#(a,d,S)

intersection
that a m a p

o f all

~ - complete

It does

[ 0, %]

(~a,ld)

induce

of

A~

f : A---* B

A~A

which

induces

A~

B~

that

and an o b j e c t

such a f u n c t i o n

To see that o b s e r v e

the m a p

and is thus a u n i f o r m

sabobjects

pseudomap)

closed.
under

TT

F(a,2d,S)

of dense e m b e d d i n g s

(respectively

~A---* ~B .

is r e f i n e d b y

G i v e n a class

and t h e r e f o r e

image of

cover.

, let

contain

a map

~A

A .

denote

the

We w i s h to s h o w

(respectively

pseudomap)

so that w e h a v e a c o m m u t a t i v e

square
A ~

~A

~. A ~
f

B----+ ~A-----+ B
and w h a t w e n e e d is a fill-in
that

f~ -I(~B)

hence

~A

observe
E1

So let

that

is d e n s e l y

extends

fg

El----+__E 2

possesses
embedded

lies in

We let

~A----* A

Since

denote

(1.7) We n o w suppose
mention

(~B)

that

function

for w h i c h

Ce~

satisfy

hypothesis

v) below.

on

k
~B,

and

f~k

so d o e s

In one example,
We s u p p o s e
A---+ C

~ ~A

extends

which

~A

f~g

and

which means

to a V - f u n c t o r

is the image of

for a n y

however

~ , then

iii)

If

A---* B

belongs

to

~ , then

B*

If

A--+ B

belongs

to

~ , then

extend
f~h

and

which

~ .

V~V

on

The i n c l u s i o n
~ .

If

The inclu-

A~A

, then

.
Unless

there

is

(see

(IV.4)ff.)

this class does not

the f o l l o w i n g

to

object

First

is taken to be the class of dense e m b e d d i n g s

embedding

is - t h e n

both

: E~--*~B

is a given class o f dense embeddings.

belongs

usually

[V,A]

A---+ B

We note t h a t s h o u l d

E1

and

is c o m p l e t e

is V--enriched.

If

~-complete

is a p s e u d o m a p

be a map.

A~

to also d e n o t e b y

Every dense

Every

since

find c o n v e n i e n t

ii)

v)

~ f~ -I(~B)

~ A~

the a d j o i n t n e s s

(II.l.4),

to the contrary,

A--~ C

iv)

there

subset

: E1

: E2

to s h o w

contains

Thus we h a v e proved,

and h e n c e

s~fficient

for it c e r t a i n l y

and

the full s u b c a t e g o r y

u s e d in

it is c l e a r l y

A~

takes v a l u e s

h a s a l e f t a d j o i n t which w e

by the same a r g u m e n t

i)

to

. In a d d i t i o n

The o b j e c t

~ A

of

extension

2
to t h e dense

sion is V--full a n d faithful

specific

belong

everywhere.

(1.5) P r o p o s i t i o n .
(1.6)

in

F o r this

subobject

a unique

Restricted

hence a r e equal
the image o f

the middle.

is a b-complete

with

CC

belongs

is e m b e d d e d

to

~ ;

in an o b j e c t of

is complete;
is p r e r e f l e x i v e ;

is q u a s i - r e f l e x i v e .
be stable

~-completeness

under the f o r m a t i o n o f closed

for any class

satisfying

subobjects

the above

- it

is e q u i v a l e n t

#
to

~ -completeness

A + B

completion

and
A~

tainly implies

B
.

for the class

is e m b e d d e d

in

The p o s s i b i l i t y

the p o s s i b i l i t y

~#
C6C,

of dense e m b e d d i n g s
t h e n the c l o s u r e

of e x t e n d i n g

of e x t e n d i n g

of

A----~ C
A

a map defined on

it to

B .

in
A

with
C

C~C.

F o r if

is its u n i f o r m

to all of

A t cer-

35

(1.8) We say that

(1.9) ProlDosition.

is
Let

complete and reflexive.

6A

~-,-complete p r o v i d e d
b e ~-complete.

A*

Then

is

6A =

~complete.
({A*)*

is {-complete,

M o r e o v e r there is a canonical b i j e c t i o n

~-*-

~A---+ A , m e a n i n g that

has the same u n d e r l y i n g V--object w i t h a p o s s i b l y finer uniformity.

Proof.

The inclusion

A*---+ ~A*

induces

dense i n c l u s i o n a n d

TeD

morphism.

16A I ~ -> ~**I -

That is,

is complete,

reflexive w h i c h means that

A** =

there is a c a n o n i c a l b i j e c t i 0 n
be given.

since

6A =

(~A*)*---~ A**

the induced map

Since

AA

is

and so

@A--+ A .

A*---~ ~A*

~(~A*,T)---+ ~(A*,T)

~-complete,

A**----+ A

N o w let

Since

is a

is an iso-

it is b y h y p o t h e s i s quasi-

exists and is bijective.

BI~

B2

b e l o n g to

and

Thus
BI-+~A

We get, by c o m p o s i t i o n

is

B 1 ---+

B2----+

~-complete,
A* ----+ B*
2
~A*

since

B2

B2**

~ 6A

B2

> 6A

is ~-complete.

(~A)* =

B2----+ B 2**

is a s s u m e d p r e r e f l e x i v e m e a n i n g there is a canonical m a p

easy to check that the r e s t r i c t i o n to

~A*

B2

is a s s u m e d complete,

B2

since

(~A*)** =

Since

(~A*) A

~A*

B1

is the given map

is ~-complete,

) ~A*

B1

-> 6A .

It is

This shows that

it is e v i d e n t l y quasi-reflexive.

Thus

w h i c h upon d u a l i z i n g gives the canonical map

~A----+ (6A)**

and shows that

~A

it is reflexive.
so that
(I.i0)

~A

Finally,

Let

CC

Let

, A(C,B)

{B-+ D }

r e p r e s e n t a t i o n of

~A*

As it also is ~-complete, h e n c e q u a s i - r e f l e x i v e

is {-complete, hence so is

is

b e a class of dense inclusions and

({A*)** =

(6A)*

A(C,B)

B .

Then

be

{~(C,B)--~

(see II.3.10, Corollary). F o r any

1
V(E,[ICI ,B] )
(see I I . 3 . 1 2 , C o r o l l a r y 1

it induces isomorphisms

p - complete.

p-complete.

b e a D - r e p r e s e n t a t i o n of

V(E,A(C,B))

is a p u l l b a c k

(~A*) A =

~-*-complete.

Proposition.

Then for any

Proof.

is

is p r e - r e f l e x i v e .

EeA

(C,D)}

is a D -

the d i a g r a m

~ ~V(E, ( C , D ) )

I
, ~V(E,[ ICl ,D~)
and apply V(E,-)) . If

V(E2,(C,D ~))---+ V(EI, ( C , D ) )

V(E2,[ ICl ,B]) ~ V( Icl ,V(E2,B))

---- V(ICl,V(EI,B)) ----V(EI,[Ic~B] ) ,

E1

> E2

belongs to

~,

36

!(E2,[ ]C] ,D ]) ~ ! ( [ C ] , V ( E 2 , D

))

Z(]C] ,!(EI,D ))~ ~ ( E l i [ C [ ,D ))

since

(C,D)

and

are the same w h e t h e r

Corollary.

Let

are all

E = E1

or

p - complete.
E = E2

b e reflexive and

But then three o f the four vertices

and hence so is the fourth

~-,-complete.

T h e n for any

(the pullback).

D~D

, A(A,D)

is

~-

complete.
Proof.

D u a l i z e and use

(1.11)

Theorem.

(A,B)

is

Proof.

Let

(II.3.6).
A

and

~-*-complete

b e r e f l e x i v e and

Then

~-complete.

~-complete.

P r o c e e d e x a c t l y as in the p r o o f of

(1.10)

e x c e p t replace

by

A .

The pull-

back is
V (E,A (A, B) )

~V(E,A(A,D

1
> ~V(E,[ IAI,D])

V_(E,[ IAI ,B] )

with the above c o r o l l a r y p r o v i d i n g the n e c e s s a r y

(1.12)

Example.

))

~-completeness of

We r e t u r n o n c e m o r e to the e x a m p l e of v e c t o r spaces.

are the finite d i m e n s i o n a l spaces, t o p o l o g i z e d discretely,

ly complete.

The d e f i n i t i o n o f

of a space in

Those of

crete spaces.

As well,
D

Not e v e r y space is

and

t h e n the spaces are c e r t a i n -

then,

is e v e r y space

~-,-complete.

consists of discrete, and

~-complete.

~-complete.

c o n s i s t i n g o f the elements of the

V

~-completeness requires that w e b e g i n w i t h a subspace

are e v i d e n t l y c o m p l e t e as are those of

c o m p a c t space cannot b e

that space

When

Thus the r e q u i r e d map e x t e n s i o n p r o p e r t y is t r i v i a l l y s a t i s f i e d and

~-comPlete.

tion is q u i t e d i f f e r e n t w h e n
ces.

).

~ - n e c e s s a r i l y finite d i m e n s i o n a l - and a dense subspace - n e c e s s a r i l y

the w h o l e thing.
every space is

~(A,D

The situa-

o f l i n e a r l y compact spa-

w h i c h are products o f d i s -

A n y dense p r o p e r subspace o f a linearly

For example the s u b s p a c e o f

S-fold d i r e c t sum is not

KS ,

~-complete.

infinite,
If we call

, w e have
S.K---+ V---+ K s

w i t h the first m a p b i j e c t i v e and the s e c o n d a dense embedding.

D u a l i z a t i o n gives

S . K - - ~ V*---+KS
The first m a p b e i n g the dual of a dense e m b e d d i n g is a b i j e c t i o n and the second is evidently dense.

We k n o w it is dense as soon as w e k n o w that

finite d i m e n s i o n a l linear c o m p a c t subspaces.

crete and factor through

S.K , whence

(S.K)*

and

This is p r e s u m a b l y always true b u t c e r t a i n l y is if

finite.

does not c o n t a i n any in-

F o r then all the

V*
S

C-+ V

are w i t h

S-K

is countable.

dis-

are e m b e d d e d in the sum space.

and

are b o t h c o u n t a b l y in-

F o r any infinite d i m e n s i o n a l l i n e a r l y compact space, b e i n g a p o w e r of

uncountable while

Thus in that case

V* ~ V

and so

K , is

V** ~ V

. The

former i s o m o r p h i s m is i n d u c e d b y the s t a n d a r d i n n e r p r o d u c t and h e n c e the latter one is

37

induced by the canonical map.

V

is reflexive.

If

This shows that

is closed in

and hence has continuous functionals.

vanish on

W , W/V

(2.1)

~-,-complete either, but that

Thus there are non-zero functionals on

V , confirming the hypothesis made in

2. D e f i n i t i o n

is not

is a separated linearly topologized space

and Elementary

Properties

of G

which

It will be a standing hypothesis in this chapter that

jects are quasireflexive.

(i.i).

~-complete ob-

The demonstrations of this fact in the various examples seem

quite different - save for the cases in which

have any common generalization.

(II.2.8) is satisfied - and do not seem to

It is clear that the hypotheses of

(II.2.8) are not al-

ways satisfied.
(2.2) We let

denote the full subcategory of

flexive,

~-complete and

both

and

fact

Un~

~-*-complete.

are subcategories of

It follows from hypotheses w e h a v e

~ .

Also

is complete, essentially by hypothesis

adjoint functor theorem.

consisting of objects which are remade that

is complete and cocomplete.

(see

In

(I.3.10)) and cocomplete by the

being reflexive is also complete and cocomplete as is ~A.

The last step follows from the next proposition.

(2.3)

Proposition.

Proof.
w e have

That

6A

A*---+ G*

The inclusion

belongs to

and with

G--+ ~A

has a right adjoint

follows from

G*

A*

Now if

~-complete, we have

Uniqueness follows from the fact that the maps

first because

(1.9).

is densely embedded in

~A*

GG

~A *---+ G*

~
and

whence

(~A*)*---+ A**---+ A
and

6A .
G + A

is a map,

G ~ G**--+(~A*) *.

are bijective,

the second because

is

the

~-com-

plete, hence quasi-reflexive.

(2.4)

Let

CI,C2 ~ .

The identity map

(Cl,C~)

~ (Cl,C~)

transposes , by the corollary of II.3.13 to a map

C1
Applying

[-I

' ((C1,C~),C ~) ~ (C2, (C1,C~)*)

we get a map in

~1 I---+ !(c2, (Cl,C~)*)----+ !(Ic21,1(Cl,C~)*

ICll
Let

T(CI,C2)

(2.5)
Proof.
if

~ ]c21"-"~

I(c1,c~)*

I)

I .

denote the image of that map as an embedded subobject of

Proposition.
(CI,C~)

For any

lies in

CI,C2 ~ , T(CI,C 2) is dense in

so its dual is in

~ .

(CI,C~)*

By the hypothesis made in

T(CI,C 2) is not dense, its closure is proper and the induced map
!(CI,C~)

is not injective.

> !(T(CI,C 2),T)

But the map

!(Cl,C~)

> !(T (CI,C2) , T ) ~

(CI,C~)*

i( IC1 I

Ic21,ITL)

(i.I)

88
can also be factored as

V(Cl,C~)

v(ICll, Ic~l)

....

V(IC II,V(C2,T))---+ V(]C l ),V_(IC2 ), )T)))

v(Ic ll~Ic 21, ITI) ,

each term of which is injective.

Since the first factor of an injection is an injection,

the result follows.

(2.6)

Proposition.

Let

CI, C2eC .

For any

AEA

there is a canonical map

V(CI,A(C2,A))-~ V( ~(CI,C 2) I ,IAI )

Proof.

Let

image of

{A--~ Din} be a D-representation_ of

A .

Since

IT(CI,C 2) [ is a regular

~i ~ [C2[, there is a pullback

V_( IT (CI,C 2 ) I,~I )----+ HV_( IT (CI,C2)] , ) O ))

v( ~Cl~ Ic21, IAI ) ..... nv(~cz;~Ic21, ID) )

Then map
V_(CI,A(C2,A)) ~

-~

HV(Cl, (C2,D m))

~v(c1. (D~,C{~---- ~V(D~, (Cl,C?)

--~ nv((Cl,C~)*,D ~) ~ ~V( I(Cz,C~)* (,ID

-~

~Z( I~(Cl,C2)l, I<I)

Also map

Z(Cl,#(c2 ,A) ) ---+ Z(~ l I, Z(c2,A) )

.... Z(IcII,Z(I21,1 m )) ~ ~(IcI l~ Ic2;,lal )
Since each of these maps is canonical, they give the same map to

]]V_(ICIIIC21,~3 I) and

hence the map required.

(2.7)

Pr__~gsition.

Let

C I, C2~C_, AA .

Then there is a canonical map

V(CI,A(C2,A) )---~ V(T (CI,C 2) ,A)

Proof.

Let

{A---~ D }

be a D__-representation of

A .

By (If. 3.12), there is a pullback

V(T (CI,C2),A)----~ RV(y (CI,C2),D)

V(I T (CI,C2)I ,I AI )--~[[V_(Ir (CI,C 2) [ ,[DJ )

We map

V(CI,A(C2,A)) ~

--- ~v(c l,{D~,c~)

I[V(CI, (C2,D))

~V(D~,{Cl,C~))

KV((C1,C~)*,D )---~ ZV(T(C I,C2),D )

39

while

(2.6) provides

V(CI,A(C2,A))---~ V(IT(CI,C2)I,IAI)

fromwhich

we have the required

map.
Corollary.

If, in addition,

A is ~-complete,

there is a canonical map

V(C I,A(C 2,A) )---+ V( (CI,C~)* ,A))

(2.8) Proposition.

Let

CI,C2E~A

Then the canonical map

V(CI,A(C2,6A))

~ V(CI,A(C2,A))

is an isomorphism.
Proof.

The inverse of the canonical

map is the composite

Z(CI,A(C2,A))

~ !((CI,C~)*,A)

--~ !(A*, (Cl,C~)) ~ !(~A*, (Cl,C~))

!(CI,A(~A*,C~))

Z(CI,A(C2,~A))

where the next to last map is the V - m o r p h i s m

(2.9) Proposition.

Let

Cc~,

GeG, A 6 ~

underlying

the map of II.3.13.

Then the canonical

V(C,A(G~A))

map

.7 V(C,A(G,A))

is an isomorphism.
Proof.

Let

{C

) G}

be a C--representation of
V(C,A(G,6A))

G .

~ HV(C,A(C

Then there is a pullback

,~A))

V(ICI,V(G,6A))
.
.
.

) HV(ICI,V(C

,~A))
0

We map
V(C,A(G,A))
as established

in

(2.8).

, HV(C,A(C

,A))

~ nV(C,A(C

,~A))

Also we have

V(C,A(G,A))

~ V(ICI ,V(G,A))

V(ICI,V(G,6A))

This gives the required map.

(2.10) Proposition.

Let

CC, D e D
A(G,

and

(C,D))

G~G

Then the canonical

map of

~ A(C,A(G,D))

is an isomorphism.
Proof.

As o b s e r v e d in

(II.3.11)

it is sufficient

to show that

V(G, (C,D))---+ V (C,A (G,D))

is an isomorphism.

The inverse of the canonical

map is given by

V(C,A(G,D))---+ V(C,A(D*,G*))---+V_((C,D)*,G*)
the second map coming from the corollary
(2.11) Proposition.

Let

A,B~A, G e G .

to

~ V(G, (C,D))

(2.7).

Then there is a canonical

map

(II.3.13)

40

A (A,A (G,B))
Proof.

As o b s e r v e d

in

(II.3.11)

~ A (G,A (A,B))

it is sufficient

V (A,A (G, B) )
Let

{C----~ A}

and

{B---~ D }

to have a canonical

map

~ V (G,A (A, B))

be a C - d o m i n a t i o n

and D-representation,

respectively.

Then there is a pullback

V(G,A(A,B))

> KV(G, (C ,D ))

V(IGI ,V (A, B) )---~ KV([GI,V(C

We have a canonical

map

V(A,A(G,B))
by

(2.10).

,D )) .

~ EV(C

,A(G,D~)) ~ EV(G, (C ,D~))

As well we have
V(A,A(G,B))----~ V(A,[I GI, B] ) ~ V(IGI,V(A,B))

induced by the embedding

Corollary

i.

Let

A(G,B)----+ [ IGI,B]

AeA, G, H 6 G .

Then there is a canonical

A(G,A(H,A))
Corollary

2.

Let

A~A, C~, G ~ G .

isomorphism

~ A (H,A(G,A))

Then the canonical

V (G,A (C,6A))

map

~ V (G,A (C,A))

is an isomorphism.
(2.12) Proposition.

Let

Ae~A

, G, HG .

Then the natural map

V (H,A (G, 6A) )

~ V (H,A (G,A))

is an isomorphism.
Proof.

Replace

by

everywhere

in the proof of

V(H,A(C,~A))
is e s t a b l i s h e d
Corollary.
Proof.

in corollary

A map

Apply

(2.13) Proposition.

Let

~ V(H,A(C~,A))

is equivalent

to a map

H--~ A(G,A)

to the above.
A,B,E~A

Then there is a canonical map

V (A,A (B,E))
Proof.

The necessary

2 above.

H---+ A(G,~A)

Hom(I,-)

(2.9).

V(6A,6A(~B,6E))

We have
V (A,A (B,E))

~ V (A,A (6B, E) )

V (6B,A_ (A,E))--+ V_(6B,A,(6 A, E) )

V(~B,A(~A,6E))---~ V(~A,A(6B,6E))
V(~A, ~A (~B, 6E) )

isomorphism

41

Here the first a n d t h i r d a r r o w are i n d u c e d b y b a c k adjunctions,

the second and fourth

from 2.11, the fifth f r o m 2.12 and the last i s o m o r p h i s m is the one given b y the adjunction.
Corollary.
Proof.

U n d e r the same h y p o t h e s e s a m a p

A - - + A(B,E)

gives a map

6A--~ ~A (6B,6E) .

A p p l y Hom(I,-) to the above.

3.

(3.1)

Let

= 6A(G,H)

G,HcG
.

The Closed M o n o i d a l Structure on

Since b y

Note that since

[G(G,H) I ~ V(G,H)

(i.ii), A(G,H)

is ~-complete,

is b i j e c t i v e for all

6A---+ A

In p a r t i c u l a r e v e r y e l e m e n t of

G (G,H)

6A(G,H)~G .
Ae~A,

We define G(G,H)

it follows that

is a u n i f o r m pseudomap. This

explains the p r i n c i p a l a d v a n t a g e of u s i n g ~-completeness i n s t e a d of completeness.

No

n o n - u n i f o r m maps n e e d b e a d d e d to h o m o b j e c t to make it ~-complete.

(3.2) Proposition.
Proof.

By

Let

G,H~G .

T h e n there is an e v a l u a t i o n map

G - ~ A(A(G,H),H)

(2.11) there is a canonical map

A(A(G,H),A(G,H))

~ A(G,A(A(G,H),H))

The r e q u i r e d m a p is, o f course, the image of the identity.

(3.3) Proposition.

Let

G,H,K6G .

T h e n c o m p o s i t i o n of m o r p h i s m s determines a canoni-

cal map
A(G,H)
Proof.

F r o m the d i s c u s s i o n following

into w h i c h b o t h
from

A (G,A (A (H, K), K) )

and

A(A(H,i),A(G,K))
(II.3.11) it follows that there is a single
A (A (H, K) ,A (G,K))

m a y be embedded.

AA

Moreover,

(II.3. 12 ) it follows that

is a pullback.

V (A (G,H) ,A (A (H,K) ,A(G, K) ) )

V (A (G,H) ,A)

V(V(G, H) ,V(A(H,K) ,A(G,K) ) )

V(V(G,H), ]AI)

N o w the p r e c e d i n g p r o p o s i t i o n p r o v i d e s a map

~ A_(A (H,~) ,K)

to w h i c h we may apply the functor

A(G,H)
w h i c h is an e l e m e n t o f

A(G,-)

to get

~ A(G,A(A(H,K),K))---~ A ,

V(A(G,H),A)

From

(II.3.4), w e h a v e a map

V(G, H)----+ V ( A ( H , K ) , A ( G , K ) )
w h i c h expresses the fact that

A(-,K)

is a V--functor.

This is a canonical element o f

V_(V(G,H) ,V_(A(G,H) ,A(G, K) ) )

a n d b o t h b e i n g canonical give the same element o f
element of
and

V(A(G,H),A(A(G,H),A(G,K)))

Hom(I,-)

commutes w i t h p u l l b a c k s ,

V (V (G,H), IAI)

Thus we get an

Since the elements we b e g a n w i t h were maps

this e l e m e n t too is a map

42

A (G, H)---~
(3.4) Proposition.

Proof.

Let

G,H,KG

A (A (G, H) ,A(G,K) )

Then c o m p o s i t i o n gives maps

G(G,H)

>

G(G(H,K),G(G,K))

G(H,K)

G (G (G, H) ,G(G,K) )

The first comes by a p p l y i n g the c o r o l l a r y o f

comes f r o m t r a n s p o s i n g

G(G,H)---~ G(G(H,K),G(G,K))---+

G(H,K)---+ A ( G ( G , H ) , G ( G , K ) )
(3.5)

Theorem

Proof.

Since

G---+ G(H,K*)

(2.11) to the above.

and a p p l y i n g

(G(H,K),G(G,K))

The s e c o n d

to get

once more.

is ,-autonomous.

IC, I G .

If

, t h e n we h a v e

G e G , G* ~
G~

(G,I)G

G(H,K*)~

and h e n c e

G(K,H*)

and

G* = G(G,I)

K~

If

G(G,H*) ~ G(H,G*)

T h e n b y I4.4 we h a v e all the data required.

4.

(4.1)

StuJmary of the Hypotheses.

The h y p o t h e s e s u s e d in this c o n s t r u c t i o n are rather complicated.

In a d d i t i o n

they are s c a t t e r e d all o v e r the p r e c e d i n g three chapters, b e i n g i n t r o d u c e d as needed.

Thus it seems useful to collect in one place a summary of these h y p o t h e s e s and a reference to a fuller e x p o s i t i o n o f each.
under which

(3.4) is proved.

It is u n d e r s t o o d that these are the a s s u m p t i o n s

In some cases more r e s t r i c t i v e forms of these h y p o t h e s e s

are s t a t e d that are s a t i s f i e d in some of the examples.

This will be m e n t i o n e d in the

examples.
(4.2)

The h y p o t h e s e s are
(i)

is an a u t o n o m o u s

(i.e. closed, s y m m e t r i c , monoidal)

category

(see

(I.l.l), (I.i.2)) .
(ii)

is a semi-variety;

that is a full s u b c a t e g o r y of a variety closed u n d e r

projective limits and c o n t a i n i n g all the free algebras

(I.i.8))
(iii)

a n d that the h y p o t h e s i s o f

The s u b c a t e g o r i e s

and

a pre-*-autonomous situation
representation
(C,D)
(iv)

Every
(see

(v)

(see

C~C

C*--~

is not injective;
C

(see

(I.3)) have the structure o f

(see

(II.3.7))

(III.l.4) and

and every c l o s e d p r o p e r s u b o b j e c t

has a

D--

(III.l.7)) is p r e r e f l e x i v e

equivalently

C*---+

A---~ C
Aw

and every o b j e c t of

, the map

injective iff

(see(iII.l.l)).

E v e r y o b j e c t in

C~C

(II.2.9)).

F o r every
A*

(I.i.4) -

(II.l.2)); m o r e o v e r the u n i f o r m i t y on the objects

~-complete o b j e c t
(II.2.7) also

Un V

(see 1.4.6.) and that e v e r y

is a c o n v e r g e n c e u n i f o r m i t y

dense in
(vi)

(see

of

(see

(I.3.10) is satisfied.

is complete.

is

CHAPTER

Before
Top M o d R

getting

of t o p o l o g i c a l

In fact,
A

, let

into e x a m p l e s

u(M)

= {a+M

neighborhoods
f i n e m e n t of

let

of

R-modules

is e q u i v a l e n t

be a topological

I aA}

module.

The c o l l e c t i o n

To see that,

there

is a

c~A

with

N e x t I c l a i m that s u b t r a c t i o n
traction
special

of

~(N) x u(N)

then

In fact if

f-l(~(M))

To go the other way,

logical

spaces d e s c r i b e d

1.8, p.17),

that the u n i f o r m
the o r i g i n a l
a uniform

topology

topology.

group

group and let

is a u n i f o r m

cover

VI,V 2 6 ~
0

in

T h e n for all

V~v

cated c o m p l e t e l y

for t o p o l o g i c a l
in the e a r l i e r

clear that other choices

follows we t a k e

KS

Let

and

of

K = ~

gized w i t h l o c a l l y

(1.3)
for

for

b~a+M

, the image under sub-

topological
and

or

is u n i f o r m

modules

is a

induces a

a neighborhood

of

in

from u n i f o r m

their

spaces

products

continuity.

It is clear

from a topological

of a n e i g h b o r h o o d

(Let

of

By the u n i f o r m i t y

there

is a

contain

group is

cover

with

in

So let

of a d d i t i o n

Uu

st(0,w__)

is in some set in

to topo-

([ Isbell ] ,

there

V I + V 2 c U.

for some star

In p a r t i c u l a r

for

I aA}

or

on compact

convex
D p

V e c t o r Spaces.

and

K =

However,
will result

, always
of

for special

of finite

consisting

choices

(explained below).
K

of

theories.

, it is
In w h a t

topology.

of finite

and c o u n t a b l e

is just the h o m into

field has b e e n expli-

in d i f f e r e n t

w i t h the usual

Top V

sum t o p o l o g y
> C

spaces over a d i s c r e t e

Chapters.

be the c a t e g o r y

direct

and c o u n t a b l e
sums

The d u a l i t y

, topologized

S.K

powers

topolo-

C p---~ D

by uniform

as
con-

sets.

In fact a c o m p a c t
sS

N-N c M

we m u s t s h o w that every u n i f o r m

implies

, V+M

vector

b e the s u b c a t e g o r y

well as its inverse

vergence

is a star re-

The theory

(1.2)

in

and so

1.
(i.i)

u(N)

N-N c M

group constructed

{a+M
refines

implies

b e a u n i f o r m cover.

of

is in some set in

b-aE

functor

by cover b y t r a n s l a t e s

aEV

, a+M

v.)

if

map between

of the o p e r a t i o n s

of

Then

Since t h a t f u n c t o r p r e s e r v e s

such t h a t

is a n e i g h b o r h o o d

of

ranges o v e r all

scalar m u l t i p l i c a t i o n

just use the s t a n d a r d

refinement

, I claim

is continuous,

of a u n i f o r m

be a uniform

There

That

To go the other way,

is r e f i n e d

as

in 1.2.

the u n i f o r m i t y

u(M)

so t h a t

In fact,

f : A'----+ A

= u(f-l(M))

of u n i f o r m R-modules.

is a n e i g h b o r h o o d

N-N c M

, bEc+N

case of the fact t h a t a c o n t i n u o u s

u n i f o r m one.
A

I a~c+N}

is uniform.
~(M)

Un M o d R

If

b~*(a,N)

aec+N

refines

to

of all

For if

suppose
b~c+N

w h i c h means

EXAMPLES.

we need to k n o w that for any ring R, the c a t e g o r y

0 , is a uniformity.

u(M)

ZV~

the p r o j e c t i o n

s e t in

KS

o n the

has c o m p a c t
coordinate

projection

is c o n t a i n e d

o n every coordinate.
in the c l o s e d d i s c

Thus
A(rs)

44
of radius

rs

a continuous

Hence

every c o m p a c t

linear m a p

Ks ~ ~ K

set is a subset of

K A(r
sS
b y an e l e m e n t

is r e p r e s e n t e d

It is e v i d e n t

of

(Xs)

S-K

that

b y the

formula
(x s) (Ys)

= ZXsY s

Then the n e i g h b o r h o o d

{(x s) I (xs)nA(r s) < I} = {(x s)

can b e e a s i l y d e s c r i b e d
i/r s (or all of

as follows.

where

lYsi < r s

we h a v e

is a s e q u e n c e
an e l e m e n t

ZysZ s = r < 1 .

Let

S-K

Xs6A(i/rs )

~ilsXsYsl

< Zilsl

(ys)~KA(rs)
value

Is = rsZs/r
IXsl =

= 1 .

implies
rs

denote

We h a v e

consi-

Zllsl = 1 . In other w o r d s
o
(i/r s)
in the sum.
Then

< 1 .

YsZs

S-K

On the o t h e r h a n d

Zllsl

the o p e n d i s c of radius

the subset of

and

l~YsZsl

such that

rlZsl/rslZsi

< i}

suppose

T h e n choose

(Zs)S-K
for e a c h s~S

is real and positive.

~IrsZsi/r

r/r s < i/r s .

= ZYsZs/r

This

Then

= i. If n o w

shows

that

~ KS

(KS) *

(1.3)

To go the o t h e r way,

only question
vergence.

Proof.

Every

compact

The p r o d u c t

on

on topological
contradicts

set in

is a c o m p a c t

s = 1,2,3,...,n,...
the s e m i n o r m on

(1.4)

first o b s e r v e

that a l g e b r a i c a l l y
topology

on

KS

(S.K)

The

is that of p o i n t w i s e

con-

We b e g i n with,

Suppose

seminorm

we

is the topology.

Proposition.

there

S.K
, p =

is a p o i n t

(pn)

vector

is in a finite d i m e n s i o n a l

Pn(X)

S.K

and

for

subspace.

countably

m a n y s, say

(x (n)) 6 Z w i t h
x n # 0 . Let
p = (ps)
be
s
(n)
= nix/x n
I T h e n since
Pn
is a c o n t i n u o u s

is a c o n t i n u o u s

spaces)

a n d suppose

s e m i n o r m on

p ( x (n)) a n
s

Thus

S-K
p

(this is s t a n d a r d
is u n b o u n d e d

on

result
which

its compactness.

Now a compact

set in

S'K

subspace.

is a c o m p a c t

in a space

as the set of all e l e m e n t s

mensional

S'K

subset of

s u c h that

In fact it is c o n t a i n e d
SO

A(i/rs ) c K

F{A(i/rs ) I seS}

such t h a t

of a b s o l u t e

x s = Zs/X s , we h a v e

~IXsrsl

c i r c l e d h u l l of the images of the

for w h i c h

Ys~K

Let

= 0) and

s
(IsXs)

sting of all e l e m e n t s
o
FA (i/rs)
is the c o n v e x
if

of

A compact

S0-K
S

set in

set in some

for some

necessary
S0-K

finite

to e x p r e s s

is c o n t a i n e d

finite d i m e n s i o n a l
subset

SO c S

a basis

subspaee.

Just take

for this finite di-

in a set of the f o r m

A~ = {~{~sS I s~s 0} I ~l~sl ~ ~}

where

~(s)

< ~/I

is a fixed p o s i t i v e
, ~llsl

< I

real number.

N o w if

is a linear

functional

and

implies

l~(~ss) l = IZ~s~(S) l
< Zl~sl~/Z
_<
so t h a t the basic

neighborhood

of

{~I~(A~) < ~} ~ {~l~(s) < c/~ , s~s 0} ,

which

shows

that

(S.K)

h a s the t o p o l o g y of p o i n t w i s e

convergence

and h e n c e

is

KS

45

(1.5)
and

The h y p o t h e s e s
D

of III.4.2,

and thus are still

denote

the space of c o n t i n u o u s

p a c t subsets

of

Proposition.

parts

satisfied.

(i) and

maps

C---~ D

topologized

Let

w e let,

on the c h o i c e of
for

C~C

, D{D

by uniform convergence

,(C,D)
on com-

Let

C = KS

and

D = P.K

Then

(C,D) ~
Proof.

(ii) do not d e p e n d

For the third part,

ti-li : p.K---+ ~

(SxP)-K

b e the n o r m d e f i n e d

by

ll(Xp)il = % Ixpl .
which

is d e f i n e d

be a continuous
s

coordinate

since there are o n l y

linear map.

(all other

For

finitely many non-zero

seS

, let

If

t h a t for at least one

peP

in

coordinate

in all o t h e r s

topology

also converge.
converge

and

to the e l e m e n t
But exactly

finitely many

cannot be a Cauchy

which

This shows

denote

sp
s, an

Thus the

of o r d i n a r y

of finite
infinite

converges

sums of

series

in the

f(Xs)

must

this net can

> 0 , it is the case that

(The p r o o f

is t h e same,

exceptions.

coordi-

x K
can be c h o s e n
s
the e l e m e n t w i t h
xs

sums of

net

the

and every

P(Xs)<

the finite p a r t i a l

sums

net otherwise.)
f

s
finitely many

involves

are non-zero.
peP

Thus

Now

: K---~ P . K
is d e t e r m i n e d b y
s
in all o n l y finitely m a n y
fsp ~ 0 .

that a l g e b r a i c a l l y
(KS,p.K)

tained

let

the net of finite

(x s)

as in c o n v e r g e n c e

Thus only finitely many

fs(1)

x =

o n l y if for every s e m i n o r m

w i t h at m o s t

peP,

x
Let
f : KS----+ P - K
P
the r e s t r i c t i o n of
f to the

f ~ 0
for i n f i n i t e l y m a n y
s
, fsp(Xs) = 1 . If
XseKS
denotes

so

product

denote

s
0) and for

coordinates

n a t e of that restriction.

As for the topology,

we have

in a set of the f o r m

KA(rs)

neighborhood

of

or

f : K S.

in

P-K

(SxP)-K

seen a l r e a d y
where

rs

is the set F~(tp)

that every c o m p a c t
is a n o n - n e g a t i v e
where

set in

KS

real number.

is a p o s i t i v e

is conA basic

real number

P
~

If

w i t h elements

of

> P-K
.

has components

sp

the

sp

: K---+ K

can be identified

W e h a v e that
f(HA(rs) ) c FA(tp)

iff
(*)

~ fsp r s / t p
s,p

T h e r e a s o n is t h a t
and

(~fsprs
s

I peP)

m u s t be able to b e w r i t t e n

~I~pI
which
(,)

is e a s i l y

seen to b e e q u i v a l e n t

is e q u i v a l e n t

< 1 .

to the c o n d i t i o n

FA

(lptp)

with

IpeK

< 1 ,

to
(fsp)

as

(tp/r s)

(*)

Similarly

the c o n d i t i o n

46

which
in

is an o p e n

(SxP)-K

set in

(SP)-K

To go the other way,

T h a t this is o p e n in the f u n c t i o n

space

let

FA

(Usp)

topology

be an o p e n set

is an easy c o n s e q u e n c e

of the following.
(1.6)

Lemma.

reals.

Let

be a function

T h e n there are f u n c t i o n s

from

: ~---+ R

f(i,j)
whenever
Proof.

(i,j)

is in the d o m a i n

By l e t t i n g

f : ~ x~---+ R +

f(i,j)

e.g.,

of

- or a subset t h e r e o f

, h

: ~--+ ~

- to the p o s i t i v e

such t h a t

~ g(i)h(j)

f .

wherever

was hitherto

undefined,

we can suppose

Let
g(n)

T h e n if,

= 1

~ x~

= h(n)

= inf{l,f(i,j)

I i ~ n, j ~ n}

i -< j, we h a v e
1 >- g(i)

Note:

It is the failure

the r e s t r i c t i o n
(1.7)
index

set

SP

-> h(j)

f(i,j)

-> g(i)h(j)

of this lemma to h o l d

to finite

N o w then,

f(i,j)

and c o u n t a b l e

g i v e n an o p e n s e t
where

e a c h of

w e can find two s e q u e n c e s

FA(Usp)
and

(rs)

and

for u n c o u n t a b l e

index sets t h a t

forces

index sets.

(SP)-K

is e i t h e r

(tp)

Usp

in

, Usp

is a f u n c t i o n

finite or countable.

on an

B y the lemma

such that

-> r stp

This means that

FA(Usp)
and the l a t t e r

is the set o f d o u b l y

D FA(rstp)

indexed

f(~A(i/tp))
and h e n c e
(1.8)

is a n e i g h b o r h o o d

This decribes

of

sequences

f =

(fsp)

such t h a t

~ FA(rs )

in the f u n c t i o n

space.

the b i f u n c t o r
C p D - - ~ D

which

t o g e t h e r w i t h the d u a l i t y b e t w e e n

and

describes

a pre-,-autonomous

situ-

ation.
That every object
the f u n c t i o n

space

(C,D)

(1.9)

Proposition.

(with

T = K)

Proof.

has a D--representation

is, b y d e f i n i t i o n ,

The h y p o t h e s e s

and hence

the c o n c l u s i o n

and the u n i f o r m i t y

on

uniformity.
of II.2.9

are s a t i s f i e d

factors

can b e r e p e a t e d

is a t o p o l o g i c a l
theorem.

is e v i d e n t

a convergence

The a r g u m e n t we used to s h o w t h a t any m a p

tely m a n y

Banach

in

vector

space

to s h o w t h a t

KS

so that the i n j e c t i v i t y

The third hypothesis

is evident.

~ K

is zero e x c e p t

is cosmall.
of

Evidently
follows

for fini-

any space in

f r o m the H a h n -

47

(i.i0)
means
of

If

A---~ B

is a p r o p e r

t h a t there is a n o n - z e r o

allows

an e x t e n s i o n

It is e v i d e n t
see [ S c h a e f e r ]

t h a t objects

tails

for

or

b~A

, A+Kb/A ~ K

is zero on

Thus

B*---~ A*

which

The i n j e c t i v i t y

is not injective.

For the c o m p l e t e n e s s

of

out with

sequences

above pairing.

(ri)

= 0

of finite

sums

The r e q u i r e d

Next I claim that

l i m i n+2 r. = 0

of

A sequence

de-

Pn

which means

s =

Again

lim in+2ri=0

is said t o b e slowly

large

i .
(s i)

Clearly

for e v e r y

, the d o t p r o d u c t

b e the s p a c e o f r a p i d l y d e c r e a s i n g

= ~ inlri ]

convergence

if

spaces.

is said to b e rapidly

(s i)

increasing

and

on compact

is the dual of

For

vector

it is t h e n a l s o true t h a t

We let
Pn(r)

by u n i f o r m

slowly

convergent.

in

of elements
Since

= 0

b y the s e m i n o r m s

topologized

of topological

for all s u f f i c i e n t l y

(ri) a n d e v e r y

are dense

C = D = the c a t e g o r y

of a c a t e g o r y

A sequence

is a b s o l u t e l y

are satisfied.

f r o m the above.

, Isil S i n

r =

sequences,topologized

we know that

for o u r c o n s t r u c t i o n

n , lim inr

rapidly decreasing

sequences

which
B

and

, e q u i p p e d w i t h the usual t o p o l o g y .

interesting

growing if for some

r-s = ~ ris i

are complete.

l
s e e n to imply t h a % ~ ~ inlril

is e a s i l y

nite

inferred

decreasing if for all

growing

in

can also b e w o r k e d

H e r e is a n o t h e r

stands

A+Kb---~ K

required

o f copies of

can b e r e a d i l y

(i.ii)

in

, II.6.2.

The t h e o r y
(or products)

map

embedding

of this to all of

T h u s the h y p o t h e s e s

closed

the space o f slowly

sets in

First observe

is the s e m i n o r m

t h a t for some

above and

under the

that the fir =

l.n+2 r i

, i > m

(ri)6D

< 66/ 2

or t h a t

I ~
i nr. I < Thus
r
is a p p r o x i m a t e d b y the finite s e q u e n c e w h i c h agrees
m+l
1
w i t h the first
m
terms of
r and is
0 thereafter.
T h e n a functional on
D
is det e r m i n e d b y its v a l u e
quence

(si)

can choose

on the finite

b y the f o r m u l a

a sequence

i(1) ,. ..,i(j)

i(1)

ri I
< i(2)

sequences

which means

~ Z ris i. If,
< i(3)

it is d e t e r m i n e d

in fact,s i

< ...

such t h a t

b y a se-

is n o t s l o w l y growing,
> ij .

si(j)

we

For h a v i n g

s. ~ i j+l
is not s a t i s f i e d for all s u f f i c i e n t l y
l
large
i
so there are a r b i t r a r i l y large
i
for w h i c h it fails.
Let
i(j+l)
be the
-i
first one of these larger t h a n
i(j)
T h e n let
r. = s
when
i = i(j)
and
0
.-n-~
1
otherwise.
Then
~ r.s.
diverges while
r. < 1
for all
i > i(n+l)
implies
ii
1
inri < i -I---~ 0
as
i--+ ~
This
contradiction establishes that
(si)C chosen

(i.12)

There

we know that

is a c e r t a i n

called nuclear spaces.

lowed

characterization

by

and

T.

Y. K o m u r a

of a c a r t e s i a n

class of spaces

of n u c l e a r

p o w e r of

is d e t e r m i n e d

so t h a t is a c o m p l e t e

by a sequence

w i t h the b o u n d e d

convergence

(in p a r t i c u l a r

IV.6.1;

II.7.1,

IV.9.6..

convergence

corollary

metric

II.81;

Next I claim that

on b o u n d e d

sets,

conjectured
a space

especially

of p s e u d o - n o r m s

space,

topology,

isolated

i.e.

has the t o p o l o g y

Hom(C,D)

is i s o m o r p h i c

to

is n u c l e a r
ii.i.i

and e s t a b l i s h e d

for details.

or F-space.
DF-space

of b o u n d e d
2; IV.5.6

In fact if

The topo-

s e e n to be
Its dual

and b o t h

convergence

on

and

D). See

and its first corollary;

, e q u i p p e d w i t h the t o p o l o g y
D

the fol '

iff it is a s u b s p a c e

and it is e a s i l y

nuclear

corollary

for our p u r p o s e

by Grothendieck

a Fr4chet

is a c o m p l e t e

II.7.2.,

for s t u d y b y G r o t h e n d i e c k

is r a t h e r o p a q u e b u t

Namely,

See [ P i e t s c h ]

complete

are r e f l e x i v e

spaces

is m o r e useful.

l o g y on

[Schaefer]

first

T h e usual d e f i n i t i o n

f : C --+ D

of u n i f o r m
is c o n t i n -

48

uous,

t h e n it is c o n t i n u o u s

the c o m p o s i t e

C---+ D---+ ~

f o l l o w e d by each c o o r d i n a t e
, the second m a p p r o j e c t i o n

projection.

So let

f.
3

on the jth coordinate.

be
Then

f.
is a linear functional on
C
and hence r e p r e s e n t e d b y a s e q u e n c e
(r..)
which
3
13
for each
j is r a p i d l y d e c r e a s i n g .
In o r d e r that for each
(si)EC , the values
s.r.. E D
i 13
c r e a s i n g in
sufficient

double

j .

Although

sequence

versely,

into a sequence.

logy,

it follows

hoods

of 0

Chapter

this m u s t be done
(i n ) .

for all

Thus

Clearly

(rij)

(s i)

in

IV,

and

j < k(i,j)

vector

space of

especially

in

it is s u f f i c i e n t

k(i,j)

An

I Pm(tj)

b e the r e a r r a n g i n g

t h a t the polars
sequence

by

seminorms

w h i c h are,

of b o u n d e d

< in

by
(r..)
13
~ inj TM Irij I <

iff

of fundamental

to

i---~ ~
.

a
Since

Con-

or as

j ~.

As for the topo-

sequence
.

of n e i g h b o r -

(See [ Schaefer]

of II.7.1.)

form a f u n d a m e n t a l

takes
.

function.
k---+ ~ .

sets in

with the corollary

Isil

represented

< ~}

m i n e d b y these

defined

a m a p iff

to r e s t r i c t

< i2n'2nlr3 ij I
~ 0
as
.n.n
2n
i ] r.. < k r k + 0 as
13
Hom(C,D)
is i s o m o r p h i c to
D

5.2 in c o n j u n c t i o n
C

it is c l e a r l y

represents

so

form a fundamental

r a p i d l y de-

we use the usual m e t h o d of r e a r r a n g i n g

So let

from reflexivity

is the sets

sets in

that the latter be a s e q u e n c e

-< (i+j) 2 , knlrk I < (i+j)2nlrij]

i < k(i,j)

Thus the u n d e r l y i n g

{(tj)

in p r i n c i p l e

to c o n s i d e r

k = k(i,j)

means

and s u f f i c i e n t

the test s e q u e n c e s
.n.m
n,m, w e have
lim i 3 r . = 0 .
13
To s h o w the i s o m o r p h i s m w i t h
D

for all
m = n

it is n e c e s s a r y

W h a t this

s y s t e m of b o u n d e d

the set into the n e i g h b o r h o o d

Thus the t o p o l o g y

as seen above,

equivalent

on

Hom(C,D)

is deter-

to the o n e s on D

under

the isomorphism.
(1.13)

Let

b e the full s u b c a t e g o r y

of

and

and

(=products)
--+ D

(K,K)

with
f~ = 0

f(a~)
.

= 1 .

Thus

sists,
As for

of

ces is dense.
which means
sing sequence

required

to

are satisfied.

shows

a =

, ~ ~ 0

r e s u l t of

~ IF = 0

f(a)
A ~-

example

a =
= 0

But

linear

is r e p r e s e n t e d

is n o n - z e r o

Thus

there

has
f

fac-

f~ ~ 0

III.4.2.

(see i.i0)

is satis-

H e n c e all

The category

The s u b s p a c e

is the

(a~)HA~
.

for w h i c h

one of its c o n s e q u e n c e s

coordinate

a~ = 0 w h e n e v e r

c a n n o t b e de-

(V) of

are satisfied.

sequences.

In fact
is an a~A~_-

in the r e m a i n i n g

f(a)

If now

of those

with

finite.

at such an e l e m e n t

that

if n o t there w o u l d b e a n o n - z e r o

and as s o o n as single

and

there

T. and Y. K o m u r a of e x a c t l y

w e can examine

but

(a~)

The condition

as in the p r e c e d i n g

of slowly g r o w i n g

In fact,

f~ = 0

this grows w i t h o u t bound,

limit a r g u m e n t

(-,-):cPxD

It is also cosmall.

Unless

T h e value of

for our m a i n c o n s t r u c t i o n

spaces,

consists

~D

o f t e n let

on the finite p r o d u c t

the same a r g u m e n t

~-complete

infinitely

Since

II.2.8

b y t h e above m e n t i o n e d

T h e space

is injective.

h o m functor

sums

the index sets b e i n g

in f i n i t e l y m a n y c o o r d i n a t e s

, the same

the p r o j e c t i o n

the h y p o t h e s e s

(C~,D),

f~ = flA~

sums of copies

are finite d i r e c t

the i n t e r n a l

D ) = H~,

are finite d i r e c t

objects

w i t h o n l y f i n i t e l y m a n y exceptions.
a

the c o n d i t i o n s

fied b y e x a c t l y

a~

whose

We d e f i n e

let

topology,

coordinates.

f~ = 0

that

If this h a p p e n s

for all

tors t h r o u g h
Hence

implies

in the p r o d u c t

= 0

(~C~,~

is a functional,

number o f n o n - z e r o

f~(a~)

and
and

The e l e m e n t s w i t h

converge,

fined.

= K

theorem

f : ~A~---+ K

whose objects

the full s u b c a t e g o r y

of c o p i e s of

so t h a t

The H a h n - B a n a c h
if

con-

the n u c l e a r

spaces.

fairly c o n c r e t e l y
F

of finite

functional

on

here.

sequenC/c (F)

by a rapidly decreais an

a6F

with

49

~(a) ~ 0 .

For

AA , a map

elements of

A .

Here

a.
l

nate and

elsewhere.

sequence

Pal,Pa2,...

f : F---+ A

(pa i)

say that a sequence

(Pai)
(ai)

of elements of

is a r a p i d l y d e c r e a s i n g sequence and
the sum

E c.a.l
l

on

A , the

= E ciPa i

m u s t b e a r a p i d l y d e c r e a s i n g s e q u e n c e of real numbers.

(ai)

converge.

Since

of

b y the formula

= E P(ciai)

Now

is r a p i d l y d e c r e a s i n g if the sequence

is r a p i d l y d e c r e a s i n g for all seminorms

numbers,

al,a2,..,

1 in the ith coordi-

C o n t i n u i t y r e q u i r e s that for every s e m i n o r m

b e e x t e n d a b l e to all of
pf(c) = pf((ci))

w h i c h m e a n s that

is d e t e r m i n e d b y a sequence

is the image of the s e q u e n c e w i t h a

p .

(ci)

T h e n we require that w h e n e v e r

a slowly i n c r e a s i n g sequence of real

(ciai)

is also r a p i d l y decreasing, this

can b e r e p l a c e d b y the simpler h y p o t h e s i s that every r a p i d l y d e c r e a s i n g series converge.

I m u s t emphasize that this is o n l y a c o n s e q u e n c e of ~ - c o m p l e t e n e s s .

It is

t e m p t i n g to c o n j e c t u r e that it is e q u i v a l e n t b u t there is no real evidence for that.

If the c o n j e c t u r e w e r e valid,

it w o u l d follow t h a t

is

~ - * - c o m p a c t p r o v i d e d the

sum of rapidly c o n v e r g i n g series of c o n t i n u o u s f u n c t i o n a l s w e r e always continuous.

2. D u a l i z i n g M o d u l e s .
(2.1)

This example is in r e s p o n s e to a q u e s t i o n of Robert Raphael as to w h e t h e r the

theory of v e c t o r spaces over a d i s c r e t e field had any natural g e n e r a l i z a t i o n to modules over o t h e r c o m m u t a t i v e rings.
k e l y that the d u a l i z i n g o b j e c t

If there is such a generalization,

its e n d o m o r p h i s m ring - m u s t be the g i v e n ring.

quire that

m u t a t i v e ring.

it seems li-

will h a v e to be injective and its dual - w h i c h is

Technical c o n s i d e r a t i o n s s e e m to re-

be finitely g e n e r a t e d and a cogenerator.

We say that an R - m o d u l e

A c c o r d i n g l y let

is a dualizing module p r o v i d e d

finitely g e n e r a t e d injective c o g e n e r a t o r and the canonical m a p

b e a comT

is a

R---~HomR(T,T)

is an

isomorphism.
We leave till 2.10 the q u e s k i o n of the existence of a d u a l i z i n g module.
is a field,

it is clear that

itself is the unique d u a l i z i n g module.

If

The theory in

t h a t case reduces to that o f vector spaces c o n s i d e r e d earlier.

(2.2)

N o w w e let

structure.

be the c a t e g o r y of all R - m o d u l e s with its usual m o n o i d a l closed

The c a t e g o r y

Un V

take all m o d u l e s of the form

modules of the f o r m

S.T

.
S

(2.3)

and

D*

= RS

D~D , D X S-T

w i t h the p r o d u c t t o p o l o g y and for

Hom(D,T) ~ Hom(S-T,T) ~ R S .

e q u i p p e d w i t h the p r o d u c t topology.
D

Rs

and

RF

any

we

T---+ S.T

T---~ S-T

We w o u l d like to define

Instead, define
~ R

factors through

h a v e the p r o d u c t t o p o l o g y

F-T

to b e Hom(D,T)

is c o n t i n u o u s for a l l

d u a l i z e to the p r o j e c t i o n s

t o p o l o g y is at least as fine as the p r o d u c t topology.

finitely generated,

But w e must show that this is indepen-

as a d i r e c t sum.
,
e q u i p p e d w i t h the coarsest t o p o l o g y such that D

When

For

all d i s c r e t e

such as that it be finite or countable.

d e n t of the r e p r e s e n t a t i o n of

The c o o r d i n a t e injections

V a r i a t i o n s o n the t h e o r y can b e o b t a i n e d b y p u t t i n g car-

d i n a l i t y r e s t r i c t i o n s on
If

is simply that of t o p o l o g i c a l R-modules.

RS

RS---+ R

O n the other hand,

since

for a finite subset

(the latter b e i n g discrete)

T---~ D.

so that this
T

is

F c S .

the p r o j e c -

50

tion
D*

RS----~ R F

is c o n t i n u o u s

is the p r o d u c t

topology

(2.4)

If

module

in its kernel.

kernels

C ~ RS ~ C

of m a p s

the d i r e c t

and h e n c e

RS----+ R ~

on

, w e see t h a t any c o n t i n u o u s
The p r o d u c t

RS----+ R F

limit of

topology

where

Hom(RF,T)

limit of the finite

sums

cribes

the duality between

(2.5)

Since

R .

is a finite

is just
and

is a cogenerator,

p o l o g i z e d b y o p e n submodules.

Thus the t o p o l o g y

on

m u s t have an o p e n sub-

subset of

base consisting
.

Thus

~ Hom(R,T) F ~ F.Hom(R,T)

S-T

Thus we d e f i n e

of the

Hom(RS,T)
~ F'T

= S-T

is

and the

This des-

every

Then

R S---+ T

has a n e i g h b o r h o o d

~ Hom(F-R,T)

direct

so is

RS

AEUn~

has a D_-representation

consists

of these

"linearly

iff it is to-

topologized"

mo-

dules.
Proposition.
Proof.

Let

A 0 = ker~
U D B0

be a n e m b e d d i n g

is o p e n in

where

B0

a = a0+b 0 ~ A
Thus

is i n j e c t i v e w i t h r e s p e c t

A---~ B

and

, A 0 = ANU

where

is an o p e n submodule.

where

a0eA0,

a = a0+b 0 A 0

b0~B 0

as well.

to the class of e m b e d d i n g s
q : A --+ T
U

a map.

is a n e i g h b o r h o o d

I c l a i m that

, then

Since

in

of

A 0 = AN(A0+B0)

b 0 = a-a 0 A

while

is d i s c r e t e
in
.

Then

In fact if

bB 0 c U so b ~ A N U

= A 0.

Thus w e have
A/A0-- -+ B / ( A 0 + B 0)

#
is an i n j e c t i o n
--+ T.

Then extension

injective
(2.6)

and b o t h m o d u l e s

in

to

o f II.2.9

f r o m the e a s i l y p r o v e d

For a continuous

for some finite

P-T

factors

map

subset

are n o w satisfied.

through

subsets

~ lim
and

cPD--~ D

factors,

Since

for some

The c o s m a l l n e s s

finite

(FG) .T ~

G c p

(SP)-T

P respectively.

where

and the d i s c r e t e

the h y p o t h e s e s

of III.4.2.

W e c a n now v e r i f y

In fact the r e q u i r e d

isomorphisms,

(RS,(RP,Q.T))
are i m m e d i a t e w h e n

S,P

argument

Since b o t h

as above.

in

formity o n

are p o w e r s
(C,D)

all subset of
Since
III.4.2
T

(iv)

is

and

is a c o n v e r g e n c e

follows

remark,

through

generated

a m a p to

the limits

are taken o v e r

(C,D)

to b e Hom(C,D)

topology.

The first two are clear while the

such as

and the general

sides are discrete,

in

(RQ,(RP,s.T))

are finite

of objects

of

Thus H o m ( R S , p -T) ~ lim H o m

Thus w e d e f i n e

of an R - m o d u l e

jects

that

: A/A 0

is v e r y s i m p l y the hom-

is f i n i t e l y

(2.7)

is easy.

group is u n i f o r m l y

by the p r e v i o u s

R F ~ F-R

w i t h the usual s t r u c t u r e

third

induces

from the h y p o t h e s i s

m a p on a t o p o l o g i c a l

functor

R S---+ P - T

G-T ~ T G

of

The

F c S

(F.R,T G) ~ l i m Hom(R,T) FxG

the finite

A 0 = k e r 9, ~

n o w follows

fact that a c o n t i n u o u s

t o g e t h e r w i t h 1.2.12.

functor.
RF

B/(A0+B0)----+ T

Since

The hypotheses

continuous

are discrete.

case

follows

no t o p o l o g i c a l

by a limit

question

arises.

and hence have a D--representation.

uniformity,

~(C)

consisting

of

alone

Ob-

The uni(or of

C ).
the h y p o t h e s e s
is satisfied.

is a cogenerator.

of II.2.9

are satisfied,

The fifth h y p o t h e s i s

In fact if

A ---~ B

every o b j e c t

is p r e r e f l e x i v e

is an e a s y c o n s e q u e n c e

is a p r o p e r

embedding

and

so

of the fact t h a t

x(B,

x{A

, let B 0

51

b e an open submodule of

such t h a t

x+B 0

does not m e e t

m e a n s the latter is a n o n - z e r o open submodule.

B/A+B 0

admits a n o n - z e r o map to the c o g e n e r a t o r

but

on

But then

x~A+B 0 w h i c h

Thus the non-zero d i s c r e t e module

T .

This is a m a p n o n - z e r o on

A , as required.

That the o b j e c t s in

and

are complete is evident and h e n c e all the hypo-

theses are s a t i s f i e d and the c o n s t r u c t i o n works.

(2.8)

If

is noetherian,

the full s u b c a t e g o r y of finitely g e n e r a t e d R-modules is

a s u b - , - a u t o n o m o u s category.
M

and

Hom(M,N)

In fact, T

are finitely generated,

is a submodule of

is finitely generated by h y p o t h e s i s and if

choose a s u r j e c t i o n

Hom(F.R,N) ~ F-N

F-R---+ M

with

finite. T h e n

In that case it b e c o m e s natural to ask

w h e t h e r this is a compact category in the sense of Kelly,

n e c e s s a r y that

A c t u a l l y the q u e s t i o n cannot b e p r o p e r l y formulated w i t h o u t

R ~ T .

G r a n t i n g that, there are natural maps

w h i c h compose to give

M*~M---+ (N,N)

is a field and h e n c e w h e n

is not semisimple,

M*@M--+ R

This t r a n s p o s e s to

really a s k e d is w h e t h e r this m a p is an isomorphism.

R

M = N = R

(M,N) is ca-

M*~ N .

that hypothesis.

F i r s t off, b y l e t t i n g

that is w h e t h e r

n o n i c a l l y i s o m o r p h i c to

is s e m i s i m p l e

exact w h i l e

M .
M~-

M*~N---+ (M,N)

In fact if

is zero

If

(Tor commutes w i t h
(M ,-)

is left

is a field and

R =

is a d u a l i z i n g m o d u l e and that w h e n

m a d e into an R - m o d u l e via the o b v i o u s a r g u m e n t a t i o n

K*~K---+ (K,K)

and what is

(yon N e u m a n n regular s e l f - i n j e c t i v e

T h e n S t e v e Schamuel's elegant o b s e r v a t i o n that

is not settles the question.

R

R---+ (N,N)

(i.e. a finite p r o d u c t of fields).

there is a n o n - f l a t m o d u l e

x ] / ( x 2) , then it is easily seen that

and

It is easy to see that it is w h e n

rings are fields), and h e n c e a finitely g e n e r a t e d n o n - f l a t module

li~m), call it

w e see that it is

R---+ K ,

is

t h e n the canonical map

(although they are isomorphic).

(2.10) Now w e turn to the q u e s t i o n of the existence of a d u a l i z i n g module.

The only

result I k n o w is the following.

Proposition.

Suppose

is a c o m m u t a t i v e ring w i t h a d u a l i z i n g module

a finitely g e n e r a t e d K - p r o j e c t i v e K-algebra.

and

T h e n for any c o n s t a n t l y rank

g e n e r a t e d R - p r o j e c t i v e R - m o d u l e P , T = HomK(P,Q)

is a d u a l i z i n g module for

is

finitely
R .

Before b e g i n n i n g the p r o o f let me o b s e r v e that this requires b e g i n n i n g w i t h a

ring that has a d u a l i z i n g module.

Of course

m i g h t b e a field.

The result of this

p r o p o s i t i o n a p p l i e d in t h a t case is t h a t any finite d i m e n s i o n a l c o m m u t a t i v e K - a l g e b r a

has a d u a l i z i n g m o d u l e and m a y w e l l have more than one.
(R,K) ~ R

iff

P r o o f of the p r o p o s i t i o n .
For if

and w i t h

M---+ N

It can b e shown that

Hom

is self injective.
The p r o o f t h a t the

so d e f i n e d is injective is standard.

is an i n j e c t i o n of R-modules, we have

HmR(i'HmK(P'Q))

HomR(i,HOmK(P,Q))

H o m K (P@RN ,Q)

HOmE (P~RM, Q )

P R-projective

P@RM

> P @ RN

is still an i n j e c t i o n and since

is K - i n -

52

jective it follows that the map is a surjection.

(R,Q)) ~ HomK(M,Q)
that

HomR(M,T)

and the fact that

~ 0

ted as an R-module,

so that

c o g e n e r a t e s R-modules.

there is a s u r j e c t i o n

F.R---~ p

this m a p has a left inverse.

Similarly

is a retract of

and thus as a K-module, P

tract of

QFG

Since

from

M ~ 0 , HOmR(M,HomK
it follows

Since

is finitely genera-

and since

is p r o j e c t i v e

is a retract of
(FXG)-K

G.K

whence

for some finite set

T = HomK(P, Q)

is finitely g e n e r a t e d as a K - m o d u l e so is

it is finitely generated as an R-module.

finitely additive.

Similarly,

is a c o g e n e r a t o r in K-modules,

The functor

is a reAfortiori,

HomK(HOmK(M,Q),Q))

is

The natural m a p
M ---+ H o m K ( H o ~ (M, Q) ,Q)

is an i s o m o r p h i s m w h e n

M = K , b y hypothesis,

hence is w h e n

w e l l w h e n it is f i n i t e l y g e n e r a t e d and projective.
c o n s i d e r e d as a K-module.

is finite free and as

This is, in particular,

true of

P,

Now

H O m R ( T , T ) = HomR(HOmK(P, Q) ,HOmK(P,Q)
H O m K (P@RHOmK (P, Q),Q)
H o m R ( P , H O m K (HOmK(P, Q) ,Q)
HomR(P,P)
But

is locally isomorphic to

and hence the natural m a p

e v e r y w h e r e locally an i s o m o r p h i s m and hence is an i s o m o r p h i s m

R--~ HomR(P,D )

is

(its k e r n e l and c o k e r n e l

are e v e r y w h e r e locally zero).

3. B a n a c h Spaces.
(3.1)

Let

b e the c a t e g o r y of b a n a c h spaces and c o n t i n u o u s linear maps of n o r m Nl.

It

is k n o w n that this is a semivariety.

its unit ball

uV

of all functions
lasl

The u n d e r l y i n g functor assigns to each spare

1
S the b a n a c h space (S)

The left a d j o i n t assigns to each set

a : S---+ K

(where

is the real or c o m p l e x field)

converges, w i t h n o r m d e f i n e d b y

Itai] =

~iasi

such that

A n a l g e b r a for the theory de-

sS
t e r m i n e d b y this a d j o i n t pair is a set closed under o p e r a t i o n s

(x i)

(li)

is any finite or countable sequence of scalars for w h i c h

> ~ l.x.1
1

ili] S 1 .

where

The op-

erations m u s t s a t i s f y equations w h i c h look like double s u m m a t i o n identities and are

fairly obvious.

E x a m p l e s of a l g e b r a s for this theory w h i c h are not

b a n a c h spaces are the o p e n

interval

(unit b a l l s of)

(-i,i) as w e l l as the q u o t i e n t space [ - i , i ] / (-i,i).

This latter space is more a c c u r a t e l y the c o e q u a l i z e r of the inclusion m a p and the zero
map.

It has three elements i,-i and the third w h i c h m i g h t b e d e n o h e d

sents the w h o l e i n t e r i o r of the ball.

take the value
(3.2)

Proposition.

or

-i , takes the value

Let

let

Since

uV

0 .

b e a b a n a c h space and

and invariant under the above theory.

Proof.

B c uV

b e t o p o l o g i c a l l y closed

T h e n there is b a n a c h space

is closed in V , so is

b e the linear subspace of

0 , and repre-

Any o p e r a t i o n not r e q u i r e d b y an i d e n t i t y to

B .

Since

generated by

B .

such that B ~ W.

is complete,
Since

so is

B . Now

is invariant under

53

the o p e r a t i o n s

it is c o n v e x and c i r c l e d

and hence d e t e r m i n e s

a norm

on

b y the

formula
p(w)
If

wB

, p(w)

~ 1 , while

all

( > 0 .

Now

is c o m p l e t e

Since

a IB

This is c l o s e d in

if

p(w)

is closed

finer t h a n t h a t induced b y

since

formity
(3.3)

follows,

in

b y 1.2.5,

uniform.

First

let

cause the v a l u e

be

is

is.

I---+ j

preserves

lute value ~ 1 .
(-i,i)

(3.4)

structure

and
banach

There

space w i t h n o r m

euclidean

norm determined

in

and
by

IB

of 1.3.10

choices

banach

U = max(p(al),...,P(an))

IB

on

IB
of

in
is

6B

in the

p uni-

It is not clear

object which

is not

w i t h its usual uniformity

Let

This is a p r e u n i f o r m

value

object be-

iff all the genuine

varies)

are 1 (resp.

< 1

(-i,i)

variables

-i).

A map

b y a scalar of abso-

are uniform.

Thus Hom(I,J)

Un V

for

spaces.

It is s u f f i c i e n t

consist
and

Let

to des-

of all o b j e c t s w i t h a DD

The first

is to take

b e a finite d i m e n s i o n a l

be a linear basis.

Let

li II

denote

the

, i.e.

Ullal+...+inanll

R n , w e have

1 and -i are isolated w h i l e

interval.

and let

al,...,a n

N o w let

in this norm.

is e v e n t u a l l y

.
is satisfied.

[-i,i]

in w h i c h

genuinely

al,...,a n

d u c t in

of

of a p r e u n i f o r m

to d e s c r i b e

are two n a t u r a l

of

for

and

to b e finite d i m e n s i o n a l

is closed

sequence

iff it is m u l t i p l i c a t i o n

it is n o t n e c e s s a r y

cribe the full s u b c a t e g o r i e s

representation.

to

(l-)wB

are b y t r a n s l a t e s

o n l y those of a b s o l u t e

w h i c h does n o t b e l o n g

or

covers

interval

the o p e r a t i o n

the a l g e b r a

Fortunately,

Thus

is the value of an o p e r a t i o n

on w h i c h

Of these,

Thus the c o m p l e t e n e s s

f r o m the c l o s e d

-i)

> 0

induced by

is an e x a m p l e

the closed

inherited

(i.e. all v a r i a b l e s

Here

for all

The u n i f o r m i t y

The uniform

w i t h the u n i f o r m i t y

1 (resp.

wEB

is since e v e r y C a u c h y

f r o m its c o m p l e t e n e s s

are.

be the same interval

h a s the u n i f o r m i t y

is

IB

F r o m this w e see t h a t the h y p o t h e s i s

w h a t the u n i f o r m o b j e c t s

this implies

since

and these are c l o s e d

I wclB}

~ 1 , w6(l+6)B

iff every set

V

= inf {I

(a~+...+a~) .

First observe

[(Lll[ ..... [lnl)-([l ..... i)

llll+...+il n

that f r o m the o r d i n a r y

inner pro-

~ [Illll ..... lln])[[ [[(i..... 1) I[

~ ,~

2)h
(Illl2+...+lln I

Then
I p ( l l a l + . . . + I n a n)

l l l l P ( a l ) + . . . + l l n l P ( a n)
U(llll+...+llnl)

2)%

~[n(illl2+..+llnl
= ~
so t h a t
pact

is b o u n d e d

in the n o r m

in a finite d i m e n s i o n a l

11 II

euclidean

takes on a m i n i m u m v a l u e there,

Conversely,

space

say l/M,

Lillal+...+Inanli

and

the

(n-l)

,
sphere

never v a n i s h e s

and it is clear t h a t

ilali = 1

on it.

[iall K Mp(a)

Hence

is comp

for all aA.

54

Thus each of the norms defines

(3.5)

F r o m this it follows

cessarily

the same t o p o l o g y

every

a m a p in the category)

A**

w i t h a under

and its second dual.

(3.6)

Proposition.

Proof.

F o r any

aA

If(a) I ~ p(a)

Let

p**

p** = p

p*(f)

Then

T h i s shows

spaces.

denote

identification

~ p*(f)

We may,

at a c e r t a i n l y

such t h a t

p**(a)
that

~ f(a)

A ~ A**

observations
spaces,

this means

extension

to the w h o l e

and e s t a b l i s h e s
A

nach

space

If

not a b a n a c h

is c o n t i n u o u s

is e q u i v a l e n t

"MT"

with

p*(f)

~ i,

b~A

suppose p(a) = i .

f(a)

= 1 . The Hahnto a functio-

This implies

that

which

for finite d i m e n s i o n a l

is the unit b a l l of a b a n a c h
into u n i t

uniformity

F i r s t off,

a product

does not

is a t o p o l o g i c a l
restricted

that a d d i t i o n

n o t the case

of finite d i m e n vector

The first

on the p r o d u c t

on the unit ball,

in

it makes

described

induced b y the c o a r s e s t

which

see [ S e m a d e n i ]

of d i s c r e t e

con-

As long as this is done

no d i f f e r e n c e

Thus w e take

ba-

locally

locally

is chosen.

, [Wiweger]

The

) spaces

It is clear that the first a p p r o a c h

development.

of p r o d u c t s

is to use the

of finite d i m e n s i o n a l

is continuous.

(for m i x e d topology,

In

has an

and scalar m u l t i p l i c a t i o n

for the o b j e c t

this uniformity.

space.

to the u n i t ball,

such t h a t e v e r y m a p to an a r b i t r a r y

to the c a t e g o r y

in the s p i r i t of our p r e v i o u s
are subspaces

space,

uniformity

space

in the same w a y for e v e r y o b j e c t

of

for all

is to use the u n i f o r m i t y

v e x space w h i c h

is a c a t e g o r y

t h a t we can e x t e n d

uniformity,

This is a s s u r d l y

i n d u c e d b y the p r o d u c t
The second

loss of generality,

N o t e that the p r o d u c t

space of such a nature

continuous.

spaces.

ces w h i c h

p*

induced b y a family of e m b e d d i n g s

spaces.

that the induced

o n the w h o l e

which

by

induced b y the norm.

convex topology

result

A*

of a finite d i m e n s i o n a l
induced by

the d u a l i t y

is a set

There are two w a y s of e x t e n d i n g

uniformity

We norm

a functional

guarantees

m a y be m a d e here.

although

particular

in 3.3.

(but not ne-

= 1 = p(a)

banach

w i t h the u n i f o r m i t y

Several

are u n i f o r m l y

~ i} ~ p(a)

If(b) l~ p(b)

A n o b j e c t of the c a t e g o r y

sional b a n a c h

If(a) I ~ p*(f)p(a)

without

admits

] , II.3.2.)

balls of finite d i m e n s i o n a l

(3.8)

so t h a t

space a n d has the n o r m and the u n i f o r m i t y

coincide

~ i}

the n o r m on

= sup { If(a) l lP*(f)

a ~ 0

([Schaefer

on all of

~ 1 .

(3.7)

let

generated

theorem

nal d e f i n e d

the usual

on

is c o n t i n u o u s

be the set of them.

= sup { If(a) l lp(a)

, If(a/p(a))l

To go the o t h e r way,
The s u b s p a c e

A*

A---+ R

and h e n c e
p**(a)

Banach

(or uniformity)

functional

and we let

p*(f)
We i d e n t i f y

linear

is m o r e

to b e the c a t e g o r y

of spa-

spaces w i t h the n o r m and the t o p o l o g y

induced.
It s h o u l d b e noted t h a t if
logically)

unit balls

map of

to a b a n a c h

Since

is e m b e d d e d

(3.9)

As internal

and

they are isomorphic.

space,

homfunctor

For the t o p o l o g y

is c o n t i n u o u s

in a p r o d u c t

(algebraically
on

and topo-

is such t h a t a n y

as soon as it is on the unit h a l l of

of b a n a c h

c P D ---+ D

have isomorphic

spaces,

w e take

the same is true of

(C,D)

to b e all linear m a p s b e -

55

t w e e n the spaces
spectively,

and

d e f i n e the n o r m

If

and

are the n o r m f u n c t i o n s on

(q/p) (f) = sup {qf(c)

i p(c)

~ i}

and

D re-

This c o u l d also be

described by
sup {qf(c/p(c))
except when
nomous

C = 0 .

situation.

I c ~ 0}

U s i n g this it is e a s y to see t h a t

and

form a pre * - a u t o -

The o n l y p a r t of I I I . 4 . 2 w h i c h m u s t be d e m o n s t r a t e d is that e v e r y

o b j e c t is p r e r e f l e x i v e

(the ~ - c o m p } e t e n e s s here

comes h a r d e r to v e r i f y as

and D

is a v a c u o u s h y p o t h e s i s ) .

grow larger

o b j e c t b u t its u n i f o r m i t y b e c o m e s

This be-

for the dual has the same u n d e r l y i n g

finer m e a n i n g

it m i g h t a d m i t m o r e

functionals.

Ac-

c o r d i n g l y we t u r n to the next example.

(3.10)

We let

through

d e n o t e the c a t e g o r y of b a n a c h

spaces.

u n c h a n g e d w i t h f i n i t e d i m e n s i o n a l b a n a c h s p a c e s r e p l a c e d by b a n a c h space.

we can c o n s i d e r
determined

to be the c a t e g o r y of those

b y m a p s to b a n a c h

spaces.

MT

n o r m b u t in t h a t t o p o l o g y ) .

Semademi

m a y be d e s c r i b e d as follows.
pologized by pointwise
f o r m u l a as in
is a proof.

(3.5).

D(D

(simple, weak)

([Semademi]),

, D*

D*

I uD

takes
which

which preserves

nations automatically

the

uD

of these

that

C p ~ D

in the

The d u a l i t y

is the set of l i n e a r f u n c t i o n a l s on

to

D*

KD

I = u]<

is compact.

finitary

induces a map

(not, of course,

D to-

c o n v e r g e n c e and w i t h the n o r m g i v e n by the same

It is t o p o l o g i z e d as a s u b s p a c e of

in the u n i t ball of

uD---~ I

is c o m p a c t

The fact t h a t the u n i t ball of

g i z e d as a s u b s p a c e of
map

For

shows

Thus

s p a c e s w h o s e t o p o l o g y and n o r m are

A m o n g t h e m are the full s u b c a t e g o r y

spaces w h i c h are l o c a l l y c o n v e x and w h o s e u n i t b a l l

a map

The d i s c u s s i o n of 3.8 goes

If
and

uD

is standard.

Here

is the unit ball of

D,

so that unit ball is t o p o l o -

It is in fact a c l o s e d s u b s p a c e for any

operations

D---+ ~

is c o m p a c t

essentially convex linear combi-

The p r e s e r v a t i o n of a f i n i t a r y o p e r a t i o n

i n v o l v e s o n l y a f i n i t e n u m b e r of c o o r d i n a t e s and the set of m a p s p r e s e r v i n g a g i v e n ope r a t i o n is e a s i l y seen to b e closed.

of c o n t i n u o u s l i n e a r m a p s
u n i t ball of

(3.11)

Proposition.

uD*

Proof.

E a c h such m a p

and the least u p p e r b o u n d

p o l o g y is r e q u i r e d on

and

C---+ K

On the other hand,

C*
Let

and

DeD

C(C

{9~D*
> 0 .

where

uD

i llgll ~ 1

So let

and

DeD
m

uC*

in norm.

{9D*
is a n e i g h b o r h o o d of
and

0
x6D

i [1911 ~ 1

and

be a c o m p a c t subset.

A zero

is

ig(x) I < e

N o w t h e r e is a f i n i t e s u b s e t

is the u n i t b a l l of

the l a t t e r set

No a d d i t i o n a l to-

we r e q u i r e the following.

T h e n the t o p o l o g y on the u n i t b a l l s

in the c o m p a c t c o n v e r g e n c e t o p o l o g y

where

be the set

t h a t is c l e a r since it is the t o p o l o g y of u n i f o r m c o n v e r g e n c e on the

u n i t b a l l and its scalar m u l t i p l e s .

neighborhood

C*

is b o u n d e d a b o v e on the c o m p a c t

is that of u n i f o r m c o n v e r g e n c e o n c o m p a c t sets b o u n d e d
For

, let

is the n o r m of the map.

Before continuing,

CeC

if

for all

x I .... ,x n

xcX}

such that

X c U

(xi+(e/2)uD)

The set
and

19(xi) i < {/2,

i=l ..... n}

in the p o i n t w i s e c o n v e r g e n c e topology.
, x = xi+(/2)y

for some

with

But if

ilytl ~ 1 .

9
Then

b e l o n g s to

56

1~(x) I S
(3.12)

i~(xi) I + /2

The

result

I ~(Y) t < 6/2 + E/2 = 6

of t h i s

continuous

on

uD*

tion.

the

topology

For

immediately
is a l s o

C ---~ D

a continuous

linear

norm.

with

and

which

(~elley]

, Chapter

Proposition.
A

of b a n a c h

the unit ball

neighborhood

defined

AeA

linear

functions

This means

norm,
a

of norm

(l-c)ilail

Thus

follows
C ---+ D

some scalar multiple

The previous

gives

of

us a g a i n

works

5, p.223)

implies
are

and

a6A

be

an e l e m e n t
and

space

such that

in

(a+M)

as w e l l

of

of n o r m

II~LI ~ 1

The only

> i.

are comthing

there

It f o l l o w s

whose

is a

Thus

Then there

so is t h e u n i t b a l l

for a s u b s p a c e

a{uA

N uA = ~

is a n e i g h b o r h o o d

fact that

spaces

reflexive.

is c l o s e d

and

The

the d o m a i n

the s a m e h e r e .

implies

set o f a l l b i l i n e a r

o f the u n i t b a l l s .

and

> 1

as t h e

well when

~(a)

paragraph

C--+ D

to see t h a t t h a t m a p p r e s e r v e s

identified

on the p r o d u c t

closure

in a p r o -

of such a product.

is a c o n v e x
that

does

circled

(a+M) N ( u A + M )

not contain

by

with

p(b)

an extension

the natural

~A*

c a n be

of a b a n a c h

on the one dimensional

(3.15)

norm.

It is r o u t i n e

holds

theorem,

is
func-

function

s p a c e of ~ o n t i n u o u s

is c l o s e d

, defined

Banach

a fact which

linear

that

linear

scalar

This property

of

on

a continuous

former.

D*---+ C*

by the

p(b)
is a s e m i n o r m

map

and dividing

uA

N = uA+M
of

in

such that

spaces.

so t h a t
The gauge

Let

on

duct

t h a t of t h e norm,

Thus

adjointness

7, t h e o r e m

the unit ball

Thus

induced

is t r u e of a c o n t i n u o u s

D*--+ C*

(C,(C',D))

product

is t o s h o w t h a t o b j e c t s

the

to t h e

It f o l l o w s

are continuous

hom/cartesian

Since

sup-on-the-unit-ball

(C',(C,D))

functional

than

(C,D) ---+ (D*, C*)

left

Proof.

is a m a p
same

belongs

bounded.

to be t h e

Pact

(3.14)

in

a map

(C,D)

t h e usual

CC'---~ D

the usual

The

D*---+ C * .

is a m a p

Both

maps

is c o a r s e r

, induces

We n o w d e f i n e

that there

C---+ D

D*

in n o r m a n d h e n c e

it is a m a p

equipped

on

if

on

from the representation

continuous

(3.13)

is t h a t

and hence

and thus

= inf { ~ I b 1 N }

Jlbti

for all

subspace

to a l l of

and

generated
A

p(a)

> i.

~(a)

= p(a)

by a

for w h i c h

that

no m a t t e r

how

map

A-

is n o r m p r e s e r v i n g .

~ 1

such that

+ A **

A*

~(b)

is t o p o l o g i z e d ,

l~(a/(l-E)llalL)

For

I > 1

~ p(b)

The

given

has,

by the Hahn-

K llbll .

as l o n g as

which

functional

any

means

it b e a r s

the

~ > 0 , there
that

sup
is

i~(a) I >

the
sup { i~(a) I i ii~II = 17 Z llall

while

the other

nal on

A*

(3.16)

This

topology
ous,
image
ber

on

, has

direction

norm equal

implies,
C**

it f o l l o w s
is d e n s e .
~i,...,~ n

is a u t o m a t i c .
to

C---+ C * *

Suppose
of

fC**

functionals

that

a , considered

as a f u n c t i o -

ilall.

in p a r t i c u l a r ,

that

is t h a t o f s i m p l e
that

It f o l l o w s

A --~ A**

convergence

is c o n t i n u o u s
.
on
{g

on

and

an

Let

C*

~C*

in t h a t

A neighborhood
C

is a n i n j e c t i o n .

~ > 0

I i (g-f(~i) I < }

Since

topology.

of

f
as

each

We w i l l

is d e t e r m i n e d

C~

The

is c o n t i n u -

show that the

b y a f i n i t e num-

57

Now

ker

~i

is a s u b s p a c e

codimension.
mensional

Let

subspace

t i o n of

to

of c o d i m e n s i o n

A = C/ker
of

A*

C*

(in fact for a n y

(3.7),

m a y a l s o be c h o s e n

to h a v e n o r m
C**

of

and t h a t of

uC

is

morphism

uC**

on u n i t balls

C**

< 1

of n o r m

Then

lies

< 1
C

b y an

by

If

in the n e i g h b o r h o o d
Then

Since

is compact,

C**

Since

The r e s t r i c -

is a p r e i m a g e

of

of

described

its image

of that image.

uC

is c o m p a c t

above

ILall = lJflA*Li~iifi1< 1 and so

is an i s o m o r p h i s m

C---+ C **

cEC

has finite

to a finite di-

~l,...,~n).

[Ifl] < 1 .
uC

ker ~n

is isomorphic

aA

is in the closure

is

and h e n c e

ker ~in...n
A*

they are g e n e r a t e d

N o w suppose

every e l e m e n t o f

so t h a t

represented

e l e m e n t of

> 0).

ker ~n

, (in fact,

is, b y

t h e n the c o r r e s p o n d i n g

~IN...N

is c l o s e d b u t

Thus the image

t h a t m a p is homeo-

(see the d i s c u s s i o n

in

(3.8)) .
(3.17)

Proposition.

Let

61~II < 1

h a s an e x t e n s i o n

Proof.

Since

extension

mension

nal

B/A 0

with

in

of n o r m

T h e n any functional

m a p s are u n i f o r m l y

continuous,
in w h i c h

Thus we m a y

suppose

A/A 0

is e m b e d d e d

in

ilall >
~(a)

case t h a t

1 .

> 1

B/A 0

~ = 0)

We k n o w

and

~ : A---+ K

with

< 1 .

and c o n t i n u o u s

in the trivial

= 1 , whence

on

of

In that case
(except

~(a)

is c o m p l e t e

to the c l o s u r e

A 0 = ker ~ .

that

be e m b e d d e d

to

closed
and

A/A 0

generated

The r e q u i r e d

has an

case

so is

is a space of di-

by an e l e m e n t

from 3.14 t h a t there

LI~Ii ~ 1

is a linear

functional

such

functio-

is

b ----+ ~ (b)/~ (a)

(3.18)

Proposition.

Proof.

We have

of m a p s

Every

AA

is q u a s i r e f l e x i v e .

shown t h a t e v e r y

CE~

is a functional

with

]Lfll < 1 .

whose

n o r m is still

t i o n a l on

~C *

if m o r e t h a n c o u n t a b l y

llflii+'''+lifntl > alf#11 .

119iI] = 1

and

nate and

A*---+ HC

Let

We have

Let
# 0)

{q

:C~

Then

let

A}

= f# LC~

we could

If

(~)e~C

9i

f: A * - - - + ~
f#

to a func-

~11f i] > i]f#1[ (in

find a finite

have

is a family

Suppose

has an e x t e n s i o n

~ = SlflU+'''+llfnll - llf#11 and

fi(gi ) > IIfill - /n

elsewhere.

< 1

many

Suppose

is an embedding.

T h e n we k n o w t h a t

particular
such

is reflexive.

C ~C such that the induced

9icCi

set

~=l,...,n

such t h a t

in the ith coordi-

f(9 ) = fl(~l)+...+fn(~n)
> [iflll+llf211+...+LI fnll -

= lif#11
while
C

II(~)II

= 1 , a contradiction.

and of c o u r s e

converges
= Z~g

(in norm,

(c)

(3.19)
jective

DD

, both

and p r e s e r v e s
C

and

We should
stated.

f eC

is r e p e s e n t e d

by an e l e m e n t

a~ = g(c ) , ZIIa~IL < 1

to an

a{A

T h e next to last e q u a l i t y

and

If

follows

9cA*

so t h a t

c ~

~ae

, f(9)=f#((gg

))

from the fact that

and continuous.

Now with

proof that

Then with

in the topology)

afortiori

= 9(Zg c ) = ~(a)

is a d d i t i v e

N o w each

ZIic~II converges.

norm

constitute

now v e r i f y

The c o n d i t i o n

D**

(see 3.15)

are b a n a c h

a pre-*-autonomous

III.4.2(v) .

is s a t i s f i e d

spaces,

the m a p b e t w e e n

and is thus an isomorphism.

t h e m is bi-

This f i n i s h e s

the

situation.

Unfortunately

for n o r m - p r e s e r v i n g

it is not quite true in the form

embeddings

(use the same argu-

58

ment

as in i.i0).

the topology

of

necessarily
this,

out

is to let

Then

be c o n t i n u o u s

we b e g i n

(3.19)

The way
(CI,C2)

and

T ( C I , C 2)

the map

have

constructed

can be proved

ad h o c

of

IClI@IC21

and

to t h e C o r o l l a r y

the norm

of III.27

will

to b e

norm

non-increasing.

To see

with

Proposition.

A,B6A

For any

, the

natural

embedding

A(A,B) ---+ A ( B * , A * )
is n o r m p r e s e r v i n g .
Proof.

If

f : A ---+ B

,
IIfll = s u p {ilf(a) il lllal] = i}

Suppose

llfll = 1

llf(a) ll > 1-6/2.

18f(a) i > 1-6.

Then
There

But

for a n y

6 > 0 , there

is a l i n e a r

functional

implies

this

is t r u e

for all

ilf*ll _< llfll = 1

6 > 0

and

> I-E

, it f o l l o w s

(3.20)

The natural

norm

suppose

V(B*,A*)

that

1 and
show

that
~

: (Cl,C~)*---+

has norm

-< 1
f#

still

has

norm

1 and

has

A(C2,A)

is the m a p

induced

by

the C o r o l l a r y

The map

: Ci---+ A(C2,A)---+
the

same

norm

A(A*,C~)

as

~* : A * - ~ (Cl,C ~)
NOW

if

cI6C 1 , d6A*

, c2C 2

are

such

that

llClLi = ildll = lIy211 = 1 , t h e n

IIf# (Cl) II _< 1

which

and
and

norm.

Now we

We m u s t

ilall = 1
liS]i = 1

embedding

f : Cl~-~
has

that

that

that

V(A,B)---+
preserves

such
such

lIBil = 1 ,

IIf*ll : 1
Corollary.

on

llf*Z(a) II > -6

IIf*ll >_ 1
Since

aEA

that
llf*Sil

Since

is a n

then
i]ai[ = i,

which

implies

that
llf#(Cl) (~)II <_ 1

and hence

that
If#(Cl ) (e) (C 2) I < 1

to III.2.7.

sg

But

and

f # ( C I) Ca) (C 2) = ~(a) (c I) (c 2)

1 , the

since

d,CI,C 2

are arbitrary

elements

of n o r m

inequality

I~(c~)
successively

(C I) (C2) I -< 1

implies

IIf(~) (Cl)ll

~ 1

I1~(~)11 ~ 1
I1~11 ~ i
which
nore

establishes
that

what

condition

The objects

call t h a t

the

w e need.

Since

in

are banach

"underlying

Thus t h e r e q u i s i t e s

set"

for o u r

spaces

is t h e u n i t

construction

4. M o d u l e s
(4.1)

Let

is a v e c t o r

be

a cocommutative

space

no o t h e r

use was made

of III.4.2(v),

we may

ig-

here.

over

, equipped

a Hopf

algebra

with

: K---+ H

and

are complete.

for s p a c e s

As for

in

This

, this

, re-

is c o m p a c t .

are present.

over

hopf

and hence
ball

over

K-linear

; e

: H H ---+ H

Algebra.
a field

means

that

maps

: H---+ K

; 6 : H---+ H6~H

1 : H---+ H
such

that

and

determine

a cocommutative

in the c a t e g o r y

of cocommutative

(4.2)

known

6(g)

The best
= gQg

is. T h e

and

second

best

The operations
+i

and

l(g)

are

I(~)

of groups

(4.3)

If

tive

dules

is s i m p l y

(4.4)

We a d o p t

let

for

Let

H(M,N)

the

denote
denote

algebra

Note

that

homomorphisms

Most

algebras

of the m a n y

belong

in f a c t

we u n d e r s t a n d
is g i v e n

homomorphism.

Sweedler's

is a m o d u l e ,
H

algebra

6L

algebra

a module

is t h e e n v e l o p i n g

multiplication

in p r a c t i c e ,
If

(4.5)

g(G

example

and o f lie

whose

is t h e g r o u p
for

the u n i q u e

is a h o p f

algebra

although

= g-i

known

= -

theory

example

coalgebra

and

notation

case

We

algebra

object

for w h i c h
features
to the

p .

let

H
for

let

PM

the c a t e g o r y

the abelian

: H@M---~ M
of H-modules

group

Hom(M,N)

, this

formula

o f a lie
6()

which
theory

heH

denote

~(g)

algebra

= 0 , 6()

of h o p f

= i,

iff

are common

a module

A morphism
denote

Here

is c o m m u t a t i v e

Le

= i~

to t h e

algebras.

for t h e a s s o c i a -

f : M---+ N

the category

of H-mo-

of H-modules.

6h = E h ( 1 ) h ( 2 )
h
o n t h e s u m m a t i o n is u s u a l l y

index

of a g r o u p

a group

of a g r o u p .

b y an H - m o d u l e

by

and write

~ G]

the a l g e b r a

omitted.

the a c t i o n

of

on

and morphisms

described

equipped

the H-action

with

(hf) (m) = E h ( 1 ) f ( ~ h ( 2 ) m )
In t h e

~,~,i

coalgebras.

amounts

to

above.

.
If M,NH,

defined

by

60

(xf) (m) = x f ( x - l m ) ,
which

is m o r e - o r - l e s s

standard

while

(4.6)

Proposition.

H-modules

iff

Let

M,N

for all

hH

in

, mEM

To m o t i v a t e

In t h a t case,

the

the proof
formula

is e v i d e n t l y

-i

is

that case

above becomes,

for a l l

for a l l

xG

arbitrary,

sible.

The proof

lution

-i

f(m)

is g o t t e n b y

is a g r o u p .

of t h i s

combinatorial

el = 6

into diagrammatic

argument

commutative
, 61 = @

would

diagrams.
If w e

language.

seem nearly
Since

replace

1
by

impos-

is an i n v o lh

the

(Elh) f(m)

= Elh(1) f(h(2)m)

can be written
[email protected]~M.I~H~M.6~M

where

we use

mutative

the c a n o n i c a l

isomorphism

= f. E S M

to i d e n t i f y

K~M

with

Then we have

a com-

The polygon

marked

diagram
HSM

H~H~M

H(*)

x~G

= f(xm)

juxtaposing

structure

(*)

The

where

H = ~ G]

above becomes
(6h) f(m)

This

is a h o m o m o r p h i s m

= f(x-lm)

is a t r a n s l a t i o n

that a purely

of the c o a l g e b r a

formula

f : M---+ N

this means

The proof below

mentioning

= xf(x-lm)

xf(m)

It is w o r t h

"bracket"

to
x

and since

map

the

= ~ h(1) f ( l h ( 2 ) m )

we c o n s i d e r

equivalent

- f(xm) , x ~ L

A K-linear

f(m)
which

, it b e c o m e s

(6h) f(m)
Proof.

for a lie a l g e b r a

(xf) (m) = xf(m)

of

xEG

left-lower

sequence

is e x a c t l y

that,

HSF

H~H~f

H~N

-w H ~ H ~ N

of this diagram
(*)

being

the

~N

~ HSN

is the u p p e r - r i g h t
formula

labelled

of next.

above.

6~

H@f

H~uM

H~N

7/

8~M I
H~uHef

H~H~uM

> H~N

H~C~N~
H~H~uH~M
H ~ I(9/I(9.MI

H~M

H~(*)

H~ f

H~uH~H~UM
H~H~ M
HHSf

Then we have a c o m m u t a t i v e d i a g r a m

H~M~

H~HOM

[
[

6~M

H~60M

dHM

HHSM

H@M

) H@H@M--

Hf
H~N ~ N

H~H~H~M

I
I

HH~M

HSHSz M

H@~ M

6~M

H~ISHM

+ H ~ H ~ H ~ M

H@I~M

H~H@M
H@H~f

H~Hf

+ H~H~N--

H@I@N

H~ N
HSN

H~H~N

~N

KN

>

H~N

~N

Finally w e have a c o m m u t a t i v e d i a g r a m
HZM
H(~H(gM-

/
H~M

K~N-I~M.

H@f
> H@M

K~laM

> K~M

~ H~N

K~f

~ K~N

r/~N

> H~N

> H~N

62

Putting

these

diagrams

is, o f c o u r s e ,

together,

the condition

we

see
f

that

that

(*)

implies

that

b e an H - h o m o m o r p h i s m .

f ' Z M = ~N "H@f
T o go the o t h e r

"

This

way

the

equations
f(hm)

= hf(m)

{h = E h ( l ) (ih(2))
imply

that
~ h ( l ) f (ih(2) m)

: E h ( l ) (ih(2))f(m)
=

(4.7)

If

an o b v i o u s
K ---+ M

is a h o p f

way

an H - m o d u l e

is d e t e r m i n e d

in t h e c a t e g o r y

over

structure

on

algebra

uniquely by

the element

(~h) f ( m )

an e l e m e n t
must

If
meM

satisfy,

In o r d e r
heH

module,

that

determines

in

a K-linear

map

the map be a morphism

a map
K

is d e t e r m i n e d

is a n o t h e r

M
.

for all

(eh)m = h m
In p a r t i c u l a r ,

: H ~

, the a u g m e n t a t i o n

uniquely

by

a K-linear

+ H(M,N)
map

f : M---+ N

such

that

(~h) f = h f
which

means,

in v i e w

of the H - a c t i o n

defined

(6h) (fm)
i.e.

that

Proposition.

any

M,N

in

Thus

Define

MSN

to b e t h e K - m o d u l e

the algebra

a symmetric,

is c o c o m m u t a t i v e

associative

tensor.

shown,

Hom(K,H(M,N))

M~ N
K

h(mSn)
Since

we h a v e

Hom(M,N)
(4.8)

(4.4),

: Eh(1)f(ih(2)m)

is an H - h o m o m o r p h i s m .
For

in

with

H-action

: ~h (i) m ~ h (2) n

counit

by

"

and coassociative,
The

given

it is e a s y

to see t h a t

this

gives

identity

h = ~ h ( l ) (ch(2))
gives

rise

to a n i s o m o r p h i s m
M

so t h a t
(4.9)

this

makes

Proposition.

into

For any

a symmetric

M,N,P

in

Hom(M~N,P)
Proof.

Since

isomorphism,

as a K - m o d u l e ,

M~N

= MOKN

M~K

monoidal
, there

category.
is a n a t u r a l

isomorphism,

+ Hom(M,H(N,P))
and

H(M,N)

= HomK(M,N),

there

is a c a n o n i c a l

63

HomK(M~N,P )

~ HomK(M,H(N,P))

Thus the only thing to check is that a morphism on one side corresponds
the other.
all

If

f : MN --+ P

h6H , m E M , nEN

is a K - l i n e a r

m~p it is a m o r p h i s m

to a morphism on

iff it satisfies,

for

,
(6h) f(mn)

= ~ h(l ) f(lh(2 ) (m~n))

(i)

= E h(1)f(lh(2)(1)m~lh(2 ) (2)n)

The corresponding

map

: M---+ H(N,P),

defined by

~(m) (n) : f(m@n)

must satisfy for all

hEH

, mEM
(~h)(fm) = ~ h(1)~(lh(2)m)

which means,

for all

n6N

,
({h) (fro) (n) = ~ h(1)~(lh(2)m) (n)
(Eh) f(m~n)

= ~ h(l ) (1)f(lh(2)m) (lh(l) (2)n)

(ii)

= E h(l ) (1)f(lh(2)m~lh(1) (2)n)

To see that these are equal,

of

on

M,N,P,

let

respectively

note the interchange

ZM ' ~N ' ~P

(e.g.

isomorphism.

denote the maps describing

~M : H~M---+ M)

Also let

From the cocommutativity

the action

o : HH

> H@H

and coassociativity,

deit

follows that the d i a g r a m

H

~H
H~

H~H

commutes.

This implies that for

'

H~

~ H~H~H

IIHH--

hEH

h(1)h(2) (1)h(2) (2) = ~ h(1) (1)h(2)~h(1) (2)

By applying

HI@I

, tensoring with

mn

and applying

an interchange,

we conclude that

~h(1)lh(2) (l)~m~lh(2) (2)~n = ~h(1) (1)lh(2)~mlh(1) (2)n

Now apply

~p.H~f.H~M~ N

(4.10) Theorem.

and the required

is an autonomous

Note that the set underlying

identity of

is

HomK(M,N)

distinct

have led in the current

, not

from maps.

chapter III it may have appeared that the pseudomaps

Although

(ii) results.

category.
~(M,N)

have an example which features pseudomaps

mary category.

(i) and

instance to considering

Thus we

In the development

in

should have been taken as the pri-

that point of view would have simplified

but whose maps are just K-linear.

Hom(M,N)

the language,

the category whose objects

it would

are H-modules

64

(4.11) Until n o w we h a v e not used the fact that

sults are v a l i d
field a l t h o u g h
(4.12)

for a r b i t r a r y
it w o u l d

It is e v i d e n t

h o p f algebras.

likely be s u f f i c i e n t

that

is a v a r i e t y

should be n o t e d that the theory of

G]-modules

in c o m m u t a t i v e

the c a n o n i c a l

one.

The p s e u d o m a p s

This

M--~ N

iff

that

is a field.

Thus the p r e c e d i n g

follows w i l l

require

have a dualizing

as w e l l as c l o s e d m o n o i d a l

is n o t in general

is)

is r e f l e c t e d
- i.e.

K
What

in the e x i s t e n c e
of

commutative

re-

be a

module.
It

(the theory of

of p s e u d o m a p s

H(M,N)

category.

and even if it is, the closed

the e l e m e n t s

that

structure
distinct

- are the K - l i n e a r

is not

from maps.

maps while

the m a p s are the H - h o m o m o r p h i s m s .

Since

the c a t e g o r y

gical H - m o d u l e s .

W e let

logical H - m o d u l e s

which

is

is abelian,

D
are,

as K - v e c t o r

e q u i p p e d w i t h the H s t r u c t u r e

subcategory
derlying

of o b j e c t s w h i c h h a v e

any such o b j e c t

entations.
the dual

space of the v e c t o r

spaces,

MA

K-vector

, the space

space u n d e r l y i n g

is e q u i v a l e n t
C

compact.

Evidently

denote

the v e c t o r

M(IM],K)

The same

underlying

of topo-

The d u a l i z i n g

object
the

space un-

space w i t h enough d i s c r e t e
underlying

to topolo-

b e the c a t e g o r y

As usual w e let

a D--representation.

IMI
and hence

and

linearly

by

Thus the linear t r a n s f o r m a t i o n

is an i s o m o r p h i s m

Un H

H-modules

induced

is a t o p o l o g i c a l

We also k n o w that if

M* = A(M,A)

the c a t e g o r y

b e the d i s c r e t e

repre-

is the same as

is e v i d e n t l y

the c a n o n i c a l

true of

map

+ H (M*,K)

this m a p is an isomorphism.

This

shows t h a t every o b j e c t

is q u a s i - r e f l e x i v e .
To s h o w that o b j e c t s
that every t o p o l o g i c a l
spaces,

in

K-vector

has a r e p r e s e n t a t i o n

have

a D--representation,

by discrete

spaces.

underlying

an o b j e c t of

sufficient

to find a family of pseudomaps,

maps.

For that,

structures

The p r e - * - a u t o n o m o u s

ity r e q u i r e d
ning parts
spaces

sums of copies
structure

discretely

underlying

C*

of

already belongs
linear,

to d i s c r e t e
K

is evident.

insures

that

are immediate.
and

i.e.

spaces w h i c h

The m o d u l e

A*

Thus the g e n e r a l

Recall,

applies

in s h o w i n g

space

For it is
H-linear,

can b e g i v e n H - m o d u l e
for all
m a p s of

in

(4.5).

when

the fifth,

K-linear

here.

to

of c o n t i n u o u s

described

to o b s e r v e
of o p e n sub-

is the v e c t o r

this h o l d s

is c o n t i n u o u s

are the c o n t i n u o u s
theory

but not necessarily

In p a r t i c u l a r

hf

basis at

If such a space

and g i v e n the H - s t r u c t u r e

of the H a c t i o n

of III.4.2

respectively.

, t h e n that o b j e c t

y o u can take the maps

as d i r e c t

is t o p o l o g i z e d

Un H

it is s u f f i c i e n t

space w h i c h has a n e i g h b o r h o o d

CC

C ---+ D

The c o n t i n u -

is.

The remai-

that the v e c t o r

functionals

on

and

A,

65

5.

(5.1)

In t h i s

as o b s e r v e d

section

we

Topological

l e t V be

at t h e b e g i n n i n g

of t o p o l o g i c a l

abelian

L e t A be

the c a t e g o r y

of t h i s

chapter,

Groups.

of abelian

identify

groups.

Un V w i t h

group.

By a seminorm

on A is m e a n t

p(0)

ii)

iii)
iv)
v)
vi)
(5.2)

An

p(a-a')

< p(a)

Trivial

consequences

p(a)

~< 0

p(a)

= p(-a)

Ip(a)

invariant
)

+ p(a')

= d(a+a",

iii)

d(a-a')

= d(a-a",

~(p)

on

in

) ;

(ii)

for all

a,a',a" e

such
A

that

are
;

a'-a")

= d(0,

a seminorm

) ;

a'+a"),

consequences

d(-a,-a')

a, a' E A.

on A is a p s e u d o m e t r i c

= d(-a,-a')

be

(ii)

- p(a') I ~< p ( a - a ' )

d(a,a')

in

+ p(a')

pseudometric

d(a,a')

(ii),

for all

(take a ' = 0

~< p(a)

Trivial

Let

are

(take a=a'

ii)

(To p r o v e
(5.3)

= 0

p(a+a')

a-a')
A

= d(a-a',0)

Define

: A x A

= d(a,a').

a function

by
(a,a')
for

= p(a-a')

a , a ' , a " E A,
e(p) (a,a')

= p(a-a')
~< p ( a - a " )

= e
so t h a t ~(p)

is a p s e u d o m e t r i e .

+ p(a"-a')

(p) (a,a")

+ e(p) (a",a')

Also,

e(p) (a,a')

= p(a-a')
= p (a+a"-a"-a')
= ~ a+a"-(a"+a')]

= e(p) (a+a',
and we

see t h a t ~(p)
Let

be

can and will,

the category

a function

: A--~

that
i)

Then

We

T o p V,

groups.

an a b e l i a n

p
Such

Abelian

is an i n v a r i a n t

an invariant

a'+a")

pseudometric.

seminorm
B(d)

on

and define

: A----+~

by
Bd(a)

= d(a,0)

a function

66

Then
f,d(0) = d(0,O)
~d(a-a')

= 0

: d(a-a',

0)

: d(a-a'+a',

O+a')

: d(a,a')
< d(a,0)

+ d(0,a')

: d(a,O)

+ d(a',0)

so t h a t
(5.4)

8d

8d(a)

8d(a')

is a s e m i n o r m .

Proposition.

The

correspondences
p ---+ ~ (p)
d ---+ B (d)

determine
on

correspondence

1-1

between

seminorms

on

and invariant

pseudometrics

Proof.

We have
~Bd(a,a')

- 6d(a-a')
: d(a-a',

0)

= d(a-a'+a',

a')

: d(a,a')
Also,
8~p(a)

Thus

(5.5)

and

are i n v e r s e

Proposition.

following

Let

p(a-0)

p(a)

isomorphisms.

be a t o p o l o g i c a l

is c o n t i n u o u s

ii)

is u n i f o r m l y

~(p)

is c o n t i n u o u s

~(p)

is u n i f o r m l y

The

If

is c o n t i n u o u s

the

f i r s t map,

is c o n t i n u o u s

components

ous.
of

Finally,
0

for

a seminorm.

Then

the

(resp.

(resp.

of the
suppose

uniformly

continuous)

uniformly

first map
that

are

continuous,

continuous),

a' E

a+M

the

identity

is c o n t i n u o u s .

and
For

composite

implies

p(a)

< s .

- p(a') I <_ p ( a - a ' )

< s .

composite

is.

If

is the c o m p o s i t e

and

C > 0

implies
IP(a)

is the

so the

p = ~(p)

such that

a @ A,

~(p)

A ~E+FR

s u b t r a c t i o n , is u n i f o r m l y

a E M
Then

continuous

A x A ~

~(p)

and

;
continuous

iv)

in w h i c h

group

are equivalent:
i)

iii)

Proof.

= ~p(a,O)

it is u n i f o r m l y

, choose

continu-

a neighborhood

87

(5.6)

Let

so is

-M

and

h o o d of

be

a topological

M0 = M n

contained

(-M)

in

abelian
.

The

group

set

We m a y

and

M0

be

a neighborhood

is a s y m m e t r i c

inductively

choose

of

M.

(i.e. M 0 = - M 0)

Then

neighbor-

a sequence

M I , M 2 , M 3 , ....
of s y m m e t r i c

neighborhoods

of

such
M

for a l l

+M

is a f i n i t e

set of strictly
I(~)

the c o r r e s p o n d i n g

If

is a p o s i t i v e

Proposition.

dyadic

rational

dyadic

rational,

For

If

~+v >

the c o n c l u s i o n

1 ,

define
= A
: M0

means

there

are

finite

is b y

the s m a l l e s t

a double

integer

sets

~ n

case,

v >

If

~=i

and

v=O

or v i c e

versa,

I(B)

and
=

T h u s we b e g i n

such

that

f i r s t on t h e

cardinality

of

supposing

that

by

N 8

and

second

on

we h a v e

union

: Z{2-ili
the

by

and

l(a U 9) : Z { 2 - 1 1 i

since

suppose

of integers
:

of

M(~+v)

is e v i d e n t .

M(1)

~}
D

~ n s : ~
In t h a t

i >

rationals

induction;

in

if

Thus we may

i(~)

The proof

+ M(v)

D >
This

let

~}

a subset

: ~{MiJi E

conclusion

is a l s o e v i d e n t .

integers,

number.

M(1)

any dyadic

the

positive

M(1)

M(~)
Proof.

Mn-i

= Z{2-ili

M (i(~))
(5.5)

n > 0
If

be

that

8}

e ~} + E { 2 - i l i

8}

Thus

is d i s j o i n t .

~(~ u

8)

: l(e)

(8)

Similarly
M(D+V)

= M (I(~)
=

+ I(8))

(X(~

8))

= z{Mili

~ u

: E{Mil

i e

= M (I(~))
=

Next we

suppose

that

~ N

8 # 0
:

and
has

~, n

B'

has

fewer

the s a m e n u m b e r

and

of elements

than
but

the

e ~

e} + E { M i l i E

6}

+ M (I(B))

M(v)

t h a t the

~(~')

elements

M(~)

8}

conclusion
:

is v a l i d w h e n e v e r

~(8')

and

that

least integer

it is a l s o v a l i d w h e n
of

~, n

8'

is s m a l l e r

a' n
than

8'

68

that
Now

of

e N

let

B
be

the s m a l l e s t

element

of

a N

~ .

1
2
a n d the

only

possibility

consistent

Then

if

i = 1 ,

1
2

with
~ + ~

<

is
1
2
in w h i c h

case

the

result

follows

from
M1 + M1

Now

suppose

i > 1

Then

if

they

have

which

in at l e a s t

i-i ~
the

B ,
same

is s m a l l e ~

M0

Since
i - i

it is n o t

o n e o f them.

a' n

~,

number

than

~n

Suppose,
=

~ - {i} U

B'

~-

has

fewer

of

a n

say, t h a t

~'

i-i ~

In t h a t

case,

let

{i - i}

{i}.
elements

of e l e m e n t s

that

but

the

than
least

In e i t h e r

d n

element

case,

while
of

our

if

a' A

inductive

i-i
B'

is

B ,
i-i

hypothesls

implies
M (I(~'))

+ M (ICB'))

M (ICa')

+ I(B'))

Evidently
I(~')

l(e)

- 2 -i + 2 - i + l

I(B')

I(B)

- 2 -i

so t h a t

Finally,
M (l(e))

If

Z{Mj]j

6 a} + X { M j ] j

Z{M.Ij
]

a-{i}

} + M i + X{Mj]j

X{Mj[j

a-{i}

} + M i _ 1 + ~{Mjl j e

Z{Mjlj

~'} + Z { M j [ j

S (l(a'))

c
Corollary.

+ M (I(8))

~-{i}

B}

} + M.
x

B'}

S'}

+ S (i(8'))

M (~(a')
~ < ~

+ I(~'))

are

dyadic

M (~(a)

rationals,

+ I(8))

then

M(~) C M(~)
Proof.

For

the difference

of t w o d y a d i c

M(~) M(~)
(5.6)

Proposition.

is g i v e n
Proof.

by
Let

The

a family
A

be

topology

(resp.

of continuous

a topological

rationals

+ M(~-~)

uniformity)

seminorms
group

{M(1) ] I

and

(resp.
M

a dyadic

is o n e

so t h a t

C M(~)
of a n y t o p o l o g i c a l
uniform

invariant

a neighborhood.
rational

There

abelian

group

pseudometrics).
is t h e n

a family

69

o f s!nmnetric

neighborhoods

of

such

that

M(1)
M(I)
and whenever

I ~ ~

+ M(~)

C M(I+~)

,
M(1)

Define

C M

C M(~)

a function
p

: A

by
p(a)
as

runs

(5.7)

over

The

function

(i)

p(0)

= 0

(ii)

p(a)

= p(-a)

(iv)
(v)

(ii)

(i)
Since

p(a-a')

-< 1

is e v i d e n t
each

M.
1

dyadic
p

has

the

+ p(a')

if

if a n d o n l y

if

following

properties.

implies

a E M

f r o m the

fact

is s y m m e t r i c

.
for e v e r y

0 E M(1)

so is e a c h
p(a)

if a n d o n l y

E M(1)}

rationals.

= p(a)

p is c o n t i n u o u s
p(a)

= inf{IIa

the positive

Proposition.

(iii)

Proof.

all

M(1)

positive

dyadic

rational

Thus

_< 1

a e

M(I)

-a E -M(1)
if a n d o n l y

if
- a E M(I)

if a n d o n l y

if
p (-a)

Since

the

dyadic

rationals

are

dense,

p(a)
(iii)

It is s u f f i c i e n t ,

now,

to s h o w

p(a
Now

let

e > 0

Since

the

I and

I'

dyadic
such

be

< 1

this

is o n l y

possible

= p(-a)

that

+ a')

-<

p(a)

+ p(a')

are

dense

o n the

line,

there

that
p(a)

p(a')

< I < p(a)

< I'

< p(a')

that
a E M(1)

and
a' E M(I')
so t h a t

given.

rationals

+ e/2

and

It f o l l o w s

if

+ /2

are positive

dyadic

rationals

70

a + a'

M(1)

M(Z

+ M

(i')

+ l'

so t h a t
p(a+a')

Since

s > 0

is a r b i t r a r y ,

this

Let

a C A

Then

a + M(1)

and

s > 0

I + I'

<

p(a)

is p o s s i b l e

p(a+a')
(iv)

<-

be

<

p(a)

given.

+ p(a')

only

+ c

if

+ p(a')

Let

be

a dyadic

rational

such

that

)~ < 6 .
is a n e i g h b o r h o o d

of

and

a'

a + M(1)

if
,

we h a v e ,
a'

- a

M(1)

or
p(a'-a)

<- I < a .

But
p(a)

: p(a-a'+a')

<- p ( a - a ' )

+ p(a')

gives
p(a)

- p(a')

-<

p(a-a')

Similarly,
p(a')

- p(a)

-< p ( a ' - a )

= p(a-a')

so t h a t
Ip(a')
(v)

This

(5.8)

is b y

This

borhood

essentially
of

- p(a) I

completes

, there

the

(5.9)

At

abelian

continuity

of

this point we

groups.

p(a'-a)

Let

{A

be

<

For we have

such

p-l([ 0,1])

guarantees

require
w

the a r g u m e n t .

is a s e m i n o r m
0

while

definition.

that

a digression
a family

shown

that

for a n y n e i g h -

that

p-l([ 0,i])

is a n e i g h b o r h o o d

o n the

in the

of a b e l i a n

sums

groups.

category

of 0.

of t o p o l o g i c a l

Then

A : ~A
denotes,
finite
p =

as u s u a l ,

number

(pw)

be

the subgroup

of n o n - z e r o
the s e m i n o r m

of

terms.
on

HA
If

consisting

for e a c h

defined

It is t r i v i a l

is a s e m i n o r m

on

only
A

a
let

to see t h a t
p
is a s e m i n o r m
th
w
summand, then

on

and that

if

: A

> A

is the

of the

supposing

topology
over

Pw

with

E Pwaw

PW
NOW

sequences

by

P(aw)

inclusion

of all

each

Ai

is a t o p o l o g i c a l

for the s e t of s e m i n o r m s

the s e t of s e m i n o r m s

on

PU
abelian

p =

(pw)

We

call

g r o u p we e n d o w

so d e f i n e d
this

as e a c h

the d i r e c t

A
-wP

with
runs

sum topology

the w e a k
independently
on

, a

71

term which

we w i l l

Before
seminorms

assertion

justify

later.

stating

the n e x t p r o p o s i t i o n ,

and

on

that

for a l l

> 0

refines

the

cover

{M C A l a , a '
refines

it w i l l b e

of s e m i n o r m s

(5.10)

there

is an

continuous
Proof.

Let

Supposing

p
/

on

b + M

where

and

discrete
,

p.f

let

be

if

p.f

groups.

so is
.

of

a seminorm

is a s e m i n o r m

on

p.f

on

of

if a n d o n l y

groups
f

and

the

be

such

that

f : IAI

on

seminorm

neighborhood

Let

for a n y

of

semi-

.
~IBI b e

is c o n t i n u o u s

seminorm

while

that
sq

iff

for e v e r y

.
property

is e v i -

is o f the

a neighborhood

of

form
s u c h thai

,
< i} C f - l ( s )

We have
+ M)

if
f(a)

if a n d o n l y

refines

< l} c N

a E f-l(b

b +M

if
(a) - b

is t r u e

to the

the p r o p e r t y
p

Then

Every

{ a l p (f (a))
is a n e i g h b o r h o o d

with
that

topological

{b'Ip(b')

and

is e q u i v a l e n t

< s }

is a c o n t i n u o u s

b C B

is a n e i g h b o r h o o d
let

such

be

is c o n t i n u o u s ,

To go the o t h e r w a y ,

if

This

< ~ }

--)q(a-a')

p E p

and

of t h e u n d e r l y i n g
seminorm

N + N C M

and
A

dent.

Now

of two

by

is a s e t o f s e m i n o r m s

s > 0

Proposition.

homomorphism

which

p k q

to s a y t h a t

the cover by

A basis

Then

E M =~p(a-a')

{M C Ala, a, E M

norm

if

of

convenient

p (f (a) - b)

<

if

E A

is s u c h

(f

(a)

b)

>

that
Ipf(a)

- pf(a') I

<

we have
p (f(a')

- b)

-<

p (](a')

- f(a))

<

6 + 1 - e = 1

+ p (f(a)

- b)

so t h a t
f (a')

b + M

Thus
f-l(b
is an o p e n
(5.11)
topology

set

in

Proposition.

+ M)

and hence

is c o n t i n u o u s .

Let

be

a family

{A

of topological

on
A

EA

oJ

abelian

groups.

Then

the

72

described

above

is the

finest

for w h i c h
u

each

inclusion

: A ---+A
w

is c o n t i n u o u s .
Proof.
and

It is e v i d e n t

p =

hence

(pw)

each

that

if for e a c h

, then p~ = P-u

uw

is c o n t i n u o u s

is.

Conversely,

for e a c h

Thus

'

suppose

Let

p =

is a c o n t i n u o u s

such

is a n y

(pw).

Pw

for e a c h

Then

seminorm

p-u

seminorm

on

is c o n t i n u o u s

on

such

and

that q.u

p~

for

(a w) ~ A

we have
:

Euw(a )

q (Z u

_<

Z qum(a )

so t h a t
q(a)

and consequently
defined

by

Corollary.
Proof.

The

If

for all

refines

all s u c h

defined

with

ness

direct

sum

(5.12)

in the u s u a l w a y .
on

qfu

is the

group

Then

= q[w

and henceforth

abelian

group

A, A*

ways.

Here we

proposition.

and

is a c o n t i n u o u s

: A

>B

fu

= f

If

Thus

on

is a s e m i n o r m
and hence

is a c o n t i n u o u s

Let

{A

is

in this

the g r o u p

take
}

the

be

section,

we

let

of c o n t i n u o u s

topology

a family

qf

on

, qf

is

is a s e m i n o r m

homomorphism.

For

is a n a t u r a l

of topological

each

there

A *
w
isomorphism.

dualizes

universal

The

on

unique-

abelian

~ (HA w) *

is a p r o j e c t i o n

property

of the s u m

----+ A

gives

Z A * ---+ ( ~ A ) *
is e v i d e n t l y

canonical

and natural.

Now

~/Z

For

a topological

A----+ T

topologized

convergence.

to

mapping

homomorphisms

of compact

A * ---+ ( ~ A ) *
w
W

which

homomorphism

: ~ A ----+B
w

: ~A

The

in t h e t o p o l o g y

sum.

map

and that map

which

is c o n t i n u o u s

categorical

is a s e m i n o r m

sum topology.

varying

Proof.

is c l e a r .

Here

canonical

p(a')

In p a r t i c u l a r ,

topology

is a t o p o l o g i c a l

the direct

of

E pw(a)

, let

a seminorm
A

f
be

(a))

suppose

groups.

Then

there

is a

in

73

f
is a c o n t i n u o u s

homomorphism.

:HA

> T

Let
-

be

the m a p w h i c h

is t h e

identity

)]]A

on the

coordinate

and

o n all o t h e r s .

Let

=:.u
w

I first

claim

that only

the n e i g h b o r h o o d

contains

of

no n o n - z e r o

finitely
in

[
w
represented

subgroup

of

eventually, by the archimedean

Similarly

for a

f-l(M)

Now

t C (1

, 0)

is to b e

many

Thus

= ~i'

is a n e i g h b o r h o o d
"''' m

= A

of

doubling

the i m a g e

a neighborhood

of

in

from

interval/

of the r e a l s ,

[ -i (M)
where

the

[ Repeated

property

.]

are d i f f e r e n t
by

~A

: -i (M)

and except

for

nB

where
I
= I0,
B

Since

is a s u b g r o u p

w = w I,
,

that

...,

otherwise

mn
.

and
f (B) C

it f o l l o w s

: (B) = 0

In p a r t i c u l a r

since

u (A)
w

C B

a =

(a)

for

it f o l l o w s

that
[~

Then

w I, . . . ,w n

if
~ A

the difference
--r

a'
belongs

to

SO t h a t

=
f (a')

a --

l{umawl

= 0

fa

~ : ~i'

Hence
=

[(Zua
w

Zfu

Z:

~a

L:

~T

so t h a t
:

be
M

will

(0,41-)

the i n t e r v a l / i
k4

subgroup

cannot

~3

finitely

that

many

i~

27

lie in

HM~

t E

let

Evidently

This means

In p a r t i c u l a r

of a

For

1 ~.

p u t it i n t o

in

,~ J

of a non-zero

1
k- ~

say

74

This

shows
Now

t h a t the

canonical

let

be

Mi

map

is at

the n e i g h b o r h o o d

least
of

(_9_-I,
for

i -> 2

Then

the

family

of all

an a l g e b r a i c

in

isomorphism.

represented

by

the i n t e r v a l

2 -i)

Mi

is a n e i g h b o r h o o d

base

at

In p a r t i c u -

lar
M2
(as d e f i n e d

above).

Now

If

{f
as

ranges

compact

when

over
X

compact

sets

and

over

all

to

which

contain

for all

x E X

, and

jx
for all s u c h

integer

jx +

consists

of

> 2 .

X U

Since

{0}

is

,
,

(2 i-2 - j)0

2i-2x

j ,

(X) ~ M i , t h e r e

is s a m e

j < 2 i-2

M
such

f (jx) ~
which

integers

C M2 : M

f (jx) e
If

A*

If

0 < j < 2 i-2

then

in

~ {sls(xu{0)) cMil

f (2i-2x)
since

at

is a n d

can restrict

then,

base

(X) C Mi}
i

{:Is(x) c M )1
we

a neighborhood

is a c o n t r a d i c t i o n .

In fact,

that

may

taken

as

2 raised

to the p o w e r

4 +[log 2 Irl]
where

is the a b s o l u t e l y

least

residue

of

(mod i).

Thus
{f If (X) C M i }
Since
under

2i-2x
the

is c o m p a c t

addition

map of

when

is

{f If (2i-2x)

--

that power

of

C M}

it is the

image

--

to

of t h e

2i-2nd

the sets of the

power

of

form

{flf(x) c M)
as

Now

let

least

ranges
p

over

be

residue,

the c o m p a c t

the s e m i n o r m
modulo

i.

on

sets
T

in

which

form neighborhood
assigns

to

base

, 4 times

at

in

its a b s o l u t e l y

least

Then
x E M

{--~ p(x)

< 1

Thus
f (X) C M
if a n d o n l y

if

p f

< i

on all of

~(s)
then

belongs

. If w e

if

sup {p~(x) Ixe x)

to

{gIg(x) c M}
if a n d o n l y

let

denote

the

seminorm

defined

by

75

~(f)
Thus
of

the

seminorms

seminorms
Now

determine,

<

as

ranges

X C A

is

compact,

~ X

is

sets

of

, a basis

compact

in

Evidently

that

<s
or,

equivalently,

Now

if

is

Is

<x> c M}

refined

by

is

also

sists

of

compact
~

where

A
Y

and
X

is

<s

If

<x #) c M}

X #i

Y : x u

compact

to

X D X# : ~
so

the

A*

of

returning

A = ~A
if

over

refines

{0}

(-x)

Thus

a basis

of

seminorms

of

A = HA

con-

compact
X = ~

0 E

and
X = -X
I claim

that

under

these

hypotheses

: ( (~ x) ^ )
(5.13)

At

Lemma.

Let

of

is

x.

this

p o i n t , we r e q u i r e ,

n > 1

and

x_,

...,

be

such

that

the

absolutely

least

residue

positive,
px i < 1

j = l,

...,

n-i

i=l
and
px n < 1
Then

the

absolutely

least

residue

of

g xi
i:l
is p o s i t i v e .

Moreover,

n Px i
i=l
Proof.

We will

hypotheses

are

confuse

xi

and

its

p <i!l xi 1

absolutely

least

that
0

-<

Xl

<

<-

X2

<

1
4
1

residue.

When

n =

2 ,

the

76

and,

of c o u r s e ,
1
xI + x 2 <

This

implies

residue

that

is its o w n

xI + x2

is p o s i t i v e

and moreover

absolutely

p ( x 1 + x 2)

For

larger

Then

~e

, we may

hypotheses

suppose

least

residue

and hence

that that

that
=

4 ( x 1 + x 2)

4x 1 + 4x 2

p ( x I)

inductively

are s a t i s f i e d

+ p ( x 2)

Vat

the assertion

is v a l i d

for

n-i

by
n-i
and

xn

i=l

Hence

the a b s o l u t e l y

least

residue

of

i n~
i=l

is p o s i t i v e

X x.
i=l

and

+ px n

n-1
=

~ PX i
i=l

+ px n

Px i

i=l

(5.14)

Now we

that

such

return

to

~e

proof

of 5.12.

We

H~ X

-X

suppose

t h a t is a c o m p a c t

and

Let
f
be

such

: A

~T

that
~(f)

T h e n is t h e r e

some

x E X

such

>_

p [ (x)
Now

is a f i n i t e

that
>-

sum
f (x)

Z fu~x

and
-<

p f (x)

p(Z f U

~ n)

_<

Z p f u~X

-<

Z sup

A
Z X

{p f u n
Ix e X }
c0 w
~
L0

(fu)

s~set

of

77

Thus

{siS(s)
T o go the o t h e r

way,

l}

<

{sl Z

X^

(fU a)

< I}

suppose
A

k (f u w)
Then

there

are

E X

such

that

Zp[ux
(of c o u r s e

a n d the

arguments

would

weren't.)

Now

are

complicate
since

lutely

least

residue

nitely

many

Now we may

sum
by

somewhat

is

[u

lemma,

so t h e s e

sups

but

the s a m e

principle

is s y m m e t r i c ,

finite.

induction,

this means

either

a
is p o s i t i v e .
Suppose

are

we

actually
would

work

can suppose

Since

the

attained.

[u

= 0

s e t of i n d i c e s

even
that

The
if t h e y

the abso-

for all b u t

involved

is

fi-

~l,...,~n.

that

E {p [ u
By o u r

>- 1

compact,

and hence

of each

, the

suppose,

-> 1

Iw = ~I . . . . .

~n-i }

<

that
p fu~

->

or that
p(E

In the

first

case,

there.

In t h e

where.

In e i t h e r

let

second

Z {p f u

->

x E X

case

case

{fu x
x

have

let

x E X

Iw = al . . . . .

an})

la = ~ i . . . . .

an }

in e v e r y

have

xwi

coordinate

in t h e

~.
1

but

th

th

the

coordinate

and

and x

L0
n

else-

while
f (x)

>

~(f )

>

so t h a t

This

shows

that

{~l~(s) < l}
Thus

topologies

the

on

E A

and

{slz ~ (Su a) < 1}

(~A.)

are

identical

and these

groups

are

iso-

morphic.
(5.15)

Proposition.

Proof.

Let

{A }

be

A direct

sum of complete

a family

of

complete
A

and
which

A'

be

the

same

the product

topological

group.

abelian

of any
In

group

family

fact

for

but

groups

is a n e i g h b o r h o o d

and

topologized

of open

+ M

is c o m p l e t e .

~A

sets

a neighborhood
H(a

where

groups

=
of

by

is open.

of

(a)

(a a) + H M
0

in

the b o x
First

topology
I claim

contains

a set

-- the o n e

that
of

the

A'

in

is a

form

w
Moreover

if

is a n e i g h -

78

borhood

of

such

that
N

- N

to

TIN

HN

LJ

Next

I claim

it

is

in

that

A'

sufficient,
A'

closed

is

since

consisting

of

neighborhoods

be

the

subgroup

is

a closed

is

sets

of

(of

A')

subgroup.

In

closed
the

For

~ 0

w
neighborhood

for
of

infinitely
0

such

w
elements

be

an

by

show

an o b v i o u s

additive

that

is

But

are

of

there
course

the

HM

those

of

the

of

1.2.5,

base

at

the

M
m a y b e t a k e n as
to
c l o s e d in
A . Now let

is

analogue

a neighborhood

direct

sum.

I claim

that

if

many

(a w)

For

all

to .

these

all

other

let

to

be

a symmetric

that
a

Let

TIM
tO

to

, whence

a
then

to

in
A

whose

fact,

complete,

in

LO

complete.
A

arbitrary

neighborhood

of

for

KM

(ato)

coordinates

and

let

Then

is

a neighborhood

of

(a w)

(bto)
then

have

so

that

aw

-a

to

then
0

so

(aw)

when

we

and

If

to

that
b

to

Thus
(bto)
This

is d i s j o i n t

from

In p a r t i c u l a r
The

is

topology

Hence

also
on

(a)
+
M
to
complement
of

the

for

each

to '

Pw

can be

is

described

of

be

such

on

to

in

a seminorm
M

is a n e i g h b o r h o o d
a seminorm

is o p e n

in

A'

and

is

closed.

complete.

p ( a w)
where

means

where

as

sup
A

(Note

generated

by

the

seminorms

that

the

sup

is

finite.)

For

EM

w
is a n e i g h b o r h o o d

that

Mto C

one

(Pwaw)

on

the

{atolpato < l}

of

zero

in

to

, let

pto
_

if

79

Then
(a)

to

if a n d o n l y

]]M

pwato
for all

, if a n d o n l y

<

if
sup pmaw

Thus

the

topology

determined

by

completeness
cient

on

EPw

is c o a r s e r

where

each

of t h e d i r e c t

to s h o w

that

sum

the basic

in

than

p~

<

that

means

from

another

on

A w )"

application

sum

(which,

Now we may

of

1.2.5.

recall,
infer

is

the

It is the

suffi-

_< i}

{ (a w) I~ P w a ~

Suppose
~

that
Z pwaw

Choose

of the direct

is a s e m i n o r m

(a)
w
which

neighborhood
M

is c l o s e d

to

if

e > 0

> 1

so t h a t
Z pwaw

Since

(a)
LO

Z A

number

of i n d i c e s

03

this

>

s u m is a c t u a l l y

involved.

Then

for

finite.

i = i,

Let

...,

eI

'

define

...,

be

seminorms

the
q~

finite
by

1
qw i

nP~.l/~

Let
qto

for i n d i c e s
Moreover,

w el,

.. ., Wn.

Then

q~

is a c o n t i n u o u s

if
(b)
w

is s u c h

evidently

that
sup qwb

<

1 ,

we h a v e
q~ibwi

nPwibw'l

<

or
<
Pwibwi
for

i - 1 .....

~/n

Then we have,

Z p

(a

+ b

~i a i

~,i.

i=l

>

i=l
>

tO
1

O]
1

~
P~ibwi
i=l
s/n
i=l

seminorm

on

A~.

80

If

{blsu p q b

< i}

we have that
( ( a M) + N )
and

(5.16)

is closed.
Proposition.

This completes

Let

an e l e m e n t

Let

M~

X
(a)

E A

lies in a sum of finitely many factors.

be a compact subset of
X

Let

Then every compact set in

Proof.

the proof.

A .

Let

be an index such that there is

with

be a neighborhood

of

with

a~ ~ M~;
and let

PC

be a s e m i n o r m on

A~

S~

such that
{a A~Ip~(a)

< i}

whence
p~ (a~)
Now assuming this to be possible

->

for at least countably many coordinates,

=

say

~i' ~2 . . . . .

then we can find such seminorms

P~l' P~2 . . . .
Let

be the s e m i n o r m on

w h i c h has the seminorm

qi

in the

i th

coordinate

i = 1,2,...

an e l e m e n t

and

lP~ i

elsewhere.

(a .)
el

By hypothesis

there is for each

such that
P~ia~i I

>

->

so that
qia~i I
so that
q(a i )

~qj a~j j

-> qi a~i i
>But then

is a continuous

unbounded

real valued map on the compact set

which

81

is impossible.
(5.17)

Proposition.

Let

and

b e t o p o l o g i c a l a b e l i a n groups.

T h e n the canon-

ical b i j e c t i o n
A @ B

) A x B

is an isomorphism.
Proof.

This map exists in any p o i n t e d c a t e g o r y and is, of course, a b i j e c t i o n of the

u n d e r l y i n g a b e l i a n groups.

If

and

are seminorms on

I(a,b) Ipa + qb < 1 I

I (a'b) Ipa < 1} 2

The left h a n d side is a b a s i c n e i g h b o r h o o d of

side is the i n t e r s e c t i o n of two such in
C o r o l l a r y i.

(3
in

A x B .

and

B ,respectively,

I(a,b)[qb < 12 } "

A B

w h i l e the right h a n d

Thus the m a p is also open.

The canonical map

(A B)*

, A* x B*

is an isomorphism.
Proof.

B y the c a n o n i c a l map is m e a n t the one w h o s e

A*

coordinate,

for example,

is

g o t t e n b y d u a l i z i n g the i n j e c t i o n
A

~ A ~ B

At any rate, we h a v e
(A B)

(A x B)

B*

C o r o l l a r y 2.

If

AI,

..., A

B*

are t o p o l o g i c a l a b e l i a n groups, the canonical map

(Z A i)

~ HA i

is an isomorphism.
(5.18)

Proposition.

Let

{A }
ca

be a family of t o p o l o g i c a l a b e l i a n groups.

Then

the canonical map

(~ A )

~ KA w

is an isomorphism.
Proof.

The fact that the canonical m a p is a c o n t i n u o u s b i j e c t i o n is trivial.

b a s i c n e i g h b o r h o o d of

in

(EA)

ca

is
N(X,M)

= {[ I[ (X) C M}

where
X
is c o m p a c t and

c o n t a i n e d in a finite s u m of
X

N(X,M)

ca
in

T .

It follows from 5.16 that

, say

... @ A
1

Thus

is a n e i g h b o r h o o d of

ca
n

A(cal' "''' can)

is the i n v e r s e image of the set of the same name in

A(ca 1 ,

....

can )

is

82

under

the

d u a l o f the

canonical

inclusion

A ( m l,
From

the

commutativity

...,

) Z A

w n)

(Z Am)

>

]I A m

1
...,

A(m I ,
together
it

with

follows

arrow

the p r e v i o u s l y

that

the

(5.19)

w )
n

image

follows

x...x A
mI

fact that

of t h a t
that

mn

set

in

inverse

the

lower

Z A

image

arrow

is an i s o m o r p h i s m ,

is o p e n .

m
is t h e

image

Since
of

the

N(X,M)

upper
under

map.

Before

and discrete

it

established

inverse

is a b i j e c t i o n ,

the c a n o n i c a l

of

continuing,

groups.

it is n e c e s s a r y

If

is a d i s c r e t e
D

is t o p o l o g i z e d
Proposition.

by
The

the p r o d u c t
group

to d i s c u s s

Pontijagin

group,

group

the

duality

for c o m p a c t

(D,T)

topology,

i.e.

of h o m o m o r p h i s m s

of

as a s u b s p a c e
D

~ T

of

is c l o s e d

.
in

TD

and hence

compact.
Proof.

If

x,y C D

, the m a p
T

defined

) T

by
~

~- f (x)

- f (x + y)

[ (y)

factors
TD
where

the

first map

a n d the s e c o n d

> T x T x T

is p r o j e c t e d

onto

are

continuous

and

so the

>

is c l o s e d .

The

(5.20)

inverse

of

0
=

intersection
is a c o m p a c t

of t h e s e
group,

we

over

the

discrete

of topological
Theorem.

If

to

x,y

topology.

We

all

y))

x,y E D

is t h e n

closed.

use here

(C,T)
one h i g h l y

non-trivial

fact

f r o m the t h e o r y

groups.
C

is c o m p a c t ,

then

for any

~ : C
that

x + y

,
f (x +

x E C

x ~ 0

, there

is a c o n t i n u o u s

homomorphism

such

and

let

C
with

corresponding

tI + t2 - t3

{f If (x) + f (y)

If

~ T

factors

takes
( t l , t 2 , t 3)

Both

the

$(x)

~ 0

Proof.

See [ Hewitt-Ross] , 22.17 on p.345.

(5.21)

This

implies

that

for a n y

compact

group

, the

canonical

map

83

C
is i n j e c t i v e .
known

fact

The

that

The

analogous

fact

is an i n j e c t i v e

continuity,

) C**

for d i s c r e t e
cogenerator

for a c o m p a c t

immediately

fact that every

requires,

from

map

in

of c o u r s e ,

(5.22)

Proposition.

Proof.

The

the p o i n t w i s e
C*

to a m a p
contains

~ :

0 < ~ < e .
n l E K

= 0.

rR ----+ T

In t h e

such

if

topology

The

analogous

let

and

T*

~(27 ) : 0
~
x ~

is c h o s e n

there

for the s e c o n d ,

number

case,

is d e n s e

Otherwise,

As
that

positive

latter

Then

Thus

category

from the well

of a b e l i a n

groups.

~ C**
convergence

is c o n t i n u o u s .

is o b v i o u s .

a smallest

numbers.

in t h e

readily

of t h e

fact

latter

and

for d i s c r e t e

the

groups

no p r o o f .

~*
first

follows

, of

Cfollows

groups

so

nl

The kernel

[R

s > 0

Then

and
n~

_< x <

(n+l)~

is s e p a r a t e d ,
Choose

<

( n+

n I

is e q u i v a l e n t
a map either

small positive

there

is a

~ ~- K

, Inl - x I < s

K =
m

of such

arbitrary

~ E K

-<

contains

is a s m a l l e s t

~ : T --

or e l s e

that

since

a map

such

and

@ = 0

with

and
whence

that

i)~

Then
0
Since

i ~ K

and

n X E K

this

<

is o n l y p o s s i b l e

1
if

n I = i

or

1
n
Now

choose

e > 0

so t h a t

~,~)
Choose

an i n t e g e r

k > i

such

that
2-kll

so t h a t

for

j >- k

let

residue

of

be

absolutely

~(2-k-ix)

or

a > 0 ).

the

absolutely

least

the

residue

But

if

least
of

least

can only be
a e

residue

~(2-k-Jl)

of

~,~)

residue
either

of
~a

~(2-k-ll)

or

~a +
_ i

n2ka

2-J-kl

2-k-Jl

Then

the a b s o l u t e l y

(depending

and similarly

is
=

can only be

is

implies

#(2-kl)

a - ~1 < _ 1

(0, ),

2-3a
That

,
~(2-3X)

Now

<

a
~ .

least

on w h e t h e r

in the o t h e r

Similarly,

case.

a < 0
Thus

the a b s o l u t e l y

84

(x)
This
x

continues

to be true w h e n

are c l e a r l y d e n s e in

x 6

R .

2ka

~(I)

is an integer.

~ (2k2-kl)

12kal

is a s m a l l e r p o s i t i v e

m u l t i p l i c a t i o n by

or by

2ka

If
12kal

so that

n2kax

Moreover,
0

so that

and such

Hence
~(x)

for all

n2kax

is a l i n e a r c o m b i n a t i o n of such e l e m e n t s

-n

>

Hence

e l e m e n t in the kernel.

This

is e i t h e r

shows t h a t the c a n o n i c a l map

~ (T, T)

is s u r j e c t i v e .

T h a t it is i n j e c t i v e

(5.23)

It is clear

and

are also r e f l e x i v e .

Proposition.
Proof.

Let

from

(5.12)

is c l e a r and h e n c e

and

(5.18)

of c o p i e s of

A c o m p a c t g r o u p is i s o m o r p h i c to a c l o s e d s u b g r o u p of a p o w e r of
C

be

compact.

Since

The n a t u r a l m a p

C
is i n j e c t i v e .

it is an i s o m o r p h i s m .

that sums a n d p r o d u c t s

is compact,

~ TC
it is h o m e o m o r p h i c ,

hence

isomorphic,

to its

image.
(5.24)

Proposition.

d i r e c t s u m m a n d of
Proof.

Let

be e m b e d d e d

c o m p a c t group

Then

is a

It is c l e a r l y s u f f i c i e n t to show that the map

C

i n d u c e d b y the e m b e d d i n g ,

T*

subgroup

D C

a c t e r on

T /D .

in the

> T

is s u r j e c t i v e .

is the image.
This m e a n s

Since

Suppose,
T

to the contrary,

is a c o g e n e r a t o r ,

there is a n o n - z e r o c h a r a c t e r on

there
T

t h a t the p r o p e r
is a n o n - z e r o

char.

w h i c h v a n i s h e s on

or t h a t
T

is not injective.

B u t we h a v e

~D*

a commutative

diagram

'C

T
in w h i c h

~D

~C

the top and right h a n d m a p are i n j e c t i o n s

and the left an i s o m o r p h i s m from

w h i c h it f o l l o w s t h a t the b o t t o m m a p is an i n j e c t i o n
(5.25)

Proposition.

Proof.

The p r o o f is standard.

The group

as well.

is i n j e c t i v e in the c a t e g o r y of c o m p a c t groups.

If
C1

~ C2

85

is an i n j e c t i o n

and

: C1

) T

is a map,
C1

> C2

T
This

is the

compact

group

) C

modulo

T x C2

f o r m the p u s h o u t

the

compact,

hence

closed

subgroup,

consist-

i n g o f all
{ (-#(x),
In o t h e r w o r d s ,
is a s u b g r o u p

the p u s h o u t

of

(5.26)

Proposition.

Proof.

If

nite).

Then

is the s a m e

a n d the p r e c e d i n g
Every

is c o m p a c t ,

compact
embed

0
is e x a c t
compact

with
one

to

and

Tn/c

abelian

maps

work

pact,

out

it

is

(5.27)

that

the

If

Then

is, b y

since

composite

~ C

Since

It

Every

follows

map must

discrete

on

be

from

( n

in b o t h

diagram

(5.12)
is

0,

surjective.

, it

follows

needn't

be

fi-

categories

with

exact

of

rows,

and

(5.18)

that

it is an e a s y

Since

C**

and

the m i d d l e

diagram

chase

are

com-

the v e r t i c a l

left hand

it is a l w a y s

d u a l o f the

the dual

map

true

natural

(in a n y

closed

category)

map

of t h a t m a p

is i n j e c t i o n .

Thus

~ D** /D

*
with

exact

' D *

>

is an i s o m o r p h i s m .

there

is an e x a c t

Now

sequence

70

rows,
D** / D

~ 0

+ D *
maps

canonical

~D

that

~D''~D
diagram

The

Since

+
0

reflexive.

~D

the

this map

~D

is

is c o m p a c t .

an i s o m o r p h i s m .

a commutative
0

This

Tn

~ 0

composite

group

of that map with

the

argument.

n **
n
**
) ---+(T /C) ---+ 0

the bottom

hand

is a c o g e n e r a t o r ,

Then we have

in w h i c h

is i n j e c t i v e

~ Tn/C

' Tn

---+(T

the

Thus

~ 0

) C

, say

is d i s c r e t e ,

is the i d e n t i t y
since

a commutative

left

the p r e c e d i n g ,

the

of

~ Tn/c

that

completes

groups.

is r e f l e x i v e .

groups we have

abelian

an isomorphism.

Proposition.

Proof.

proposition

in a p o w e r

injections.

is an i s o m o r p h i s m .

as of the u n d e r l y i n g

group

I X @ CI }

~ C -----+ T n

is c o m p a c t .

and discrete

the vertical

x )

are

~ D****
injections,

~ (D ** /D) **
the m i d d l e

) 0

an i s o m o r p h i s m

and hence

so is

one.

shows

that

the d u a l i t y

is v a l i d b e t w e e n

the

compact

a n d the d i s c r e t e

groups

88

(5.28)

Proposition.

topology
Proof.

The

of uniform

dual

o f the g r o u p

convergence

on

compact

of r e a l n u m b e r s ,

sets,

equipped

with

the

is

If
:

is a map,

> T

let
:

By continuity,

0(I)

= 0

inf

and

{xlx

~ = 0

> 0

&

~(x)

only when

~(i/l)

: 0}

~ : 0

.
Otherwise

let

~ T

by
@(i/~) (x) : x / A
We h a v e

an e x a c t

sequence
0

and

@(~

) = 0

~ ~

induces

> ~

~ ~/~

from 5.22

that

is an i n j e c t i o n .

Thus

If we d e f i n e ,

for all

~ 0

~ T

is m u l t i p l i c a t i o n

The o n l y m a p s

the i n v e r s e .

~ T

a map
: T

We k n o w

(mod i)

T --

by

an i n t e g e r .

which

are

Since

injections

ker
are

@ = ker @

the

identity

,
and

}(1/~)

~ E
~(I)

: ~R

~ T

by
~(~) (x)
then

the a b o v e

shows

that

every

~x

is of

(mod l)

the

form

~(i)

for

al C i~

which

is c l e a r l y

unique.
A basis
natural

for the c o m p a c t

number.

t h a t the b a s i c
(5.29)

We
We

leads

Evidently

~(I)

neighborhoods

of

are now ready

consider

in

of t h e

choice

of

are

just

the u s u a l

consist

i)

iii)
The three

IlI<

i/4n

a
so

ones.

and

finally,

a n d as m a n y
of

We

for
let

Each

consist

A power

form

of all g r o u p s

is c o m p a c t

n = 0

, n

is

finite,

in

(ii)

choices

A compact
~

iff

n ],

of the

mn

~m

form

n. [R

m.T

to

ii)

of

(_ 14 ' i )

subcategory

, of duality

D
subject

into

I-n,

theory.
for the

C
and

of the i n t e r v a l s

set

the

consists

that

to a p p l y

takes

six passibilities

to a d i f f e r e n t

of all g r o u p s

sets

group

in a p r o d u c t

of d i s c r e t e

or

and two

c a n be

can of course be

and

is f i n i t e ,
m

discrete;
or

in

(iii)

embedded,

embedded
groups.

is at m o s t

is at m o s t

the s i x p o s s i b i l i t i e s .

as m e n t i o n e d

in a p o w e r
Thus

give

of

countable;

countable.

above,
A power

for all possibile

in a p o w e r

choices

of

can be

of
of

embedded

every

group

87

in

can be e m b e d d e d

is made,

we let

of groups
(5.30)

in

in a p r o d u c t

, as usual,

of groups

consist

in

of all groups

Whatever

choice

of

t h a t can be e m b e d d e d

and

in a p r o d u c t

W e define

the functor
(-,-)

: C p

x D

) D

by letting
(C x m

be the d i r e c t

sum of nine

i)
ii)

(c, p . ~ )

= 0

(C, q.T)

= q.C*

(~m,

= 0

( [R

( m,

q.T)

(~n,

ix)
We w i s h

p.~)

maps

) D topologized

discretely;

(m x n ) .

(mxq).~

(n x q ) . T
that these

of a m a d
for same

@ q.T)

to e s t a b l i s h

qo.T

D @ p.~

as follows.

= (m x p ) . ~

(Z~ n, q.T)

the image

(77n, D) = n . D

viii)

it lies in

D)

TM, n.~)

v)
vi)
vii)

(iii)

terms

77 n

is the group of continuous

iii)

iv)

in

(C,D)

C ~

are the c o r r e c t

q.T

finite

lies

subset

Hom(C,q.T)

in a c o m p a c t

qo C q

underlying

F o r example,

groups.

subgroup

of

By

(5.16)

Hence

l i m ) H o m (C,qo.T)
lira+Horn (C, T

qo

lira)Horn (C,T)

qo

lim qo.HOm(C,T)
q. Hom(C,T)
Here

Hem

(-, -)

refers

are g e n e r a t e d b y c o m p a c t
any map
sum.
of

from either

Analogously
0

~RTM

(and,

or

in
77 n

subsets,

of these

t h a t contains

group c o n t a i n e d
power

to the a b e l i a n

if fact,

no n o n - z e r o

We c o n s i d e r

in its k e r n e l

the p r o d u c t

is

Let

the cases

C E C

shows

and

on c o m p a c t

already

When

must contain

convergence

i)

in

of

Proposition.

We have

the groups

subgroup.

0, i} n

in the k e r n e l

Hom(A,B).

Proof.

dually)

{-i,

D E D

established

rRTM

and

respectively.
factor

all h a v e

Then

(compact

Thus

case,

(C,D)

a finite

a neighborhood

a continuous

of all b u t

map

from a

abelian

Hom(C,D)

group

topologized

convergence).

t h a t the a b e l i a n

group

underlying

(C,D)

is

Hom(C,D)

separately.
C

is c o m p a c t

{ ::

and

c--+D

is o p e n in the c o m p a c t

discrete,

[ :(c)

: 0 ]

convergence

any sub-

finitely many.

the discrete

is

77n

Thus

through

is a map to such a group,

t h a t in each

subsets

Both groups

s u m m u s t also

lies

f-l(M)

underlying
(5.31)

and

to the above

by uniform

1 ]m

to a d i r e c t

If

T h e n a dual a r g u m e n t
(A,B)

[-i,

groups

group v a l u e d hom.

topology

so that the t o p o l o g y

is

88

discrete.
ii)

There is nothing

to prove.

iii)

The group
hood

q'C
of

~ q ,

is discrete.

= (-1/4,

will do.
Again,

v)

Thus

has a neighbor-

For example,

if for all

1/4),
M

iv)

On the other hand, q-T

w i t h o u t any small subgroups.

F(M )

Hom(C,q-T)

there is nothing

is discrete

in the compact open topology.

to prove.

We must show that the abstract

(n x m)- ~

isomorphism

~ (~n, m.~)

is a homeomorphism.
Here

and

one is easier.

may be finite or countable.

We first show it is open.

We consider

A neighborhood

the latter case as the other

of

on the left is of the

form
F(-rB , r~)
where

{r B}

so d e t e r m i n e d

is a doubly indexed sequence

ii)

real numbers.

consists of all doubly indexed sequences

of real numbers
i)

of positive

such that
Only

[~

finitely many

E I [~8/r~BI

Given such a sequence,

~ 0 ,

and

< 1

let
s

2 ~ sup {i, i/rsyIB,y

~ a}

and
t
Then for

2 -~ inf {i, rBy I B,y N ~}

e ~ ~ , we have
s

2~

and
t
while

~ k B

2-Br8

implies
>-

2 /r8

and
-S
t8

and in either case

ts/s ~

2-~-8r 8

On the right h a n d side the set

{/I / ( H [-s~,s~])
is a n e i g h b o r h o o d

of

O .

Every
f

(l ~8)

F(-ts,t ~) }

The n e i g h b o r h o o d

89
in that set must in p a r t i c u l a r have the property
f

8[-sa,

sa ]

that

(-tB, ts)

or

If as sa]
1[ aS l

<

ts/s a

t8

<

2-a-B r 8

-<

so that

IZ f C~B/rc~8l <- ZIf as/rasl

<-

To go the other way,

This shows that the map is open.

every compact set in

is

[An

contained in one of the form

H [-s a, s]
while sets of the form
F(-t B, t B)
are a basis

m-~

for the open sets in

s,

{fl'(~[

Thus a basic open set in

s l)

F(-t8,

( n,

m'[A)

is

ts)}

If
f
then

belongs

{f S}

to that set iff

[

I[ [ ~ 8

S~ I /

<

Its[

Now let

rc~ ~

2-~-B

tB/s a

Then supposing
{f 8}

F(-r 8, r B)

(n x m)"

we have,

I I lS~Bsl / Ihl

~ l l l s ~ BsJtSI

B
<
Thus t h e map i s

continuous

The failure of such an argument

reason for the restriction
topology on

(C,D)

in several places.
correct w i t h o u t

The remaining
arguments

for larger than countable

to finite and countable

is not a convergence

exponents.

exponents

figures

the main results of [Barr,

to at most countable

is the

Without this,

topology w h i c h hypothesis

Thus I do not know w h e t h e r

restriction

~Z2 -~-B

and h e n c e a h o m e o m o r p h i s m .

the

crucially
1977]

are

exponents.

cases are h a n d l e d b y trivial m o d i f i c a t i o n

of one of the above

and are left to the reader.

(5.32)

Proposition.

For any

C E C

Proof.

Just examine the nine cases individually.

(C, D)

D E D
~

(D , C*)
Each of them is trivial in view of

the definitions.
The error is in the proof of 7.9 in which the caveat

of 5.3 is ignored.

g0

(5.33)

Proposition.

For

any

Cl,

(C I,
In p a r t i c u l a r ,

there

C2 E C

(C2,D))~

D ~ D_ ,
( C 2,

is a i-i c o r r e s p o n d e n c e

(C1,D))

between

maps

) (C2, D)

C1
and maps
C 2 -Proof.
above
and

There
and

C2

are,

the
are

so t h a t we

in p r i n c i p a l ,

fact that some

compact

need only

27 c a s e s

are

zero,

and

discrete,

show

they have
$

is a m a p

of a c o m p a c t

a finite

number

of w h i c h

is an o p e n

has

open kernel.

subgroup

which

that map

is c o n t i n u o u s .

contains

the k e r n e l

using

the s y m m e t r y

collapsing.

If

a n d (C 2, (CI,D))

elements.

Now

C1

are d i s c r e t e

a map

(C2, D)

set and hence

The

has

a finite

image.

It thus

determines

-----+ D

intersection

Each

of

of the

element
Since

of the

induced

) ('C
I iI,

C2

kernel

(C I, (C2,D))

same

is the k e r n e l

whence

However,

is a c o n s i d e r a b l e

maps
C

each

both

the

to c o n s i d e r .

there

: C 1 --+

to d i s c r e t e

of continuous

(C 1 ' D)

D)

of

that

finitely

many open

subgroups

map

C 2 determines
is open,

each

a map

ICI]

such map

, D

whose

is c o n t i n u o u s

so

we have
(Cl, D)

C2
It is e v i d e n t l y

exactly

the

variable

and

coefficients

consider

the

cases

same

in the o t h e r

in the s e c o n d

come

C1
since

(C2,D)
what

is an

choice

in

leave

the s a m e

nothing

c a n be
(5.34)

to p r o v e .

reduced
This

to o n e

shows

it is s u f f i c i e n t
T

as the

is the c a s e

as c o e f f i c i e n t s ,

that

cosmallness

discrete

in the

first

it is s u f f i c i e n t

or

PR

or

If

is d i s c r e t e

no

to
C] p

(CI,D)

to the

C 1 = ~.

The

cases

reader

to s h o w

Cl = 77

that

using

or

C 2 : Z[

(5.32),

every

case

considered.

III.4.2(iii)

will

On the other hand,

(c I, D))

It is l e f t

to v e r i f y

exponents

, so a l s o

C 2 compact,

already

Since

~ (C2, D)

r R - v e c t o r space.

of

( c 2,
Exactly

out

C.
c o m p a c t or ~
or
~
and
l
C 2 = 59 , t h e r e a r e no n o n - z e r o m a p s

is c o m p a c t ,

matter

direction.

is s a t i s f i e d .

the h y p o t h e s e s
appear

of

as a w a y

II.2.9.

station

We n o w
We

t u r n to

turn

a n d the

(iv),

for w h i c h

to the i n j e c t i v i t y

third hypothesis

of

is e v i d e n t .

Suppose
A
is an e m b e d d i n g .
that

every

: A

Since
> T

can be
can b e

* B

embedded

extended

to

in a p r o d u c t
B

> T

of

in the

it s u f f i c e s
case

B = ~D

to s h o w
The

set

91

f-if_
is o p e n

in

and hence

contains

a set

of the
N

A
where

is o p e n

indices

Wl,

....

in

Wn

such

The

B0

is a s u b g r o u p

and that set

contains

topology

on

B0

so is

A/A 0

denote

is s u c h

K {Bwle

that

~ e I .....

A0 = A n B0 .

there

is a f i n i t e

set of

en }

Since

no s u b g r o u p s ,
.{ (A 0 )

Let

form

that

M
Since

the a b s t r a c t

quotient

group

topologized

(here only')

as a s u b g r o u p

of
B/B 0

Dwl

x...x

Den

The homomorphism

F
induced

by

the i n v e r s e

image

image

in

in

Then

is s t i l l
of

: A/A 0

continuous.

(-1/4,

of an o p e n

1/4)

set M

In

has

is c o n t i n u o u s

a unique

products,
a closed
case
A
and

subgroup

there

continuous

A0

is an

Let

finite

sums

splits

sequence

that

of

D ----+ A

topological

retraction

on

Since

is c o n t i n u o u s

to b e

iff

the i n v e r s e

neighborhoods

Since

to

D E D

~-subspaces
A

Then

of

is c o m p l e t e ,

is c l o s e d

t h a t if

We
of

A0

under

and

consider
A

finite

A C D

is

first

In fact,

is a s u b s p a c e

the

since

of

S i n c e any ~ - l i n e a r m a p o n a f i n i t e
0
v e c t o r s p a c e is c o n t i n u o u s , t h i s m a p is

subspace

of

supposing

m-~

m = ~
on

0
D

) A -----+ A / A 0

is i s o m o r p h i c

chosen

A/A 0 .

to s h o w

is c o n t i n u o u s

~ A0

A / A 0.

of

to

are b a s i c

an e x t e n s i o n

subgroup

(I am t a c i t l y

the

A/A 0

induced

sets

the s u m of all

dimensional

0
means

(A/A0)

has

retraction

finite

to p r o v e ) ,

which

be

to an a b i t r a r y

is n o t h i n g
the

A0

can be

topology

A ----+ T

such

1/4)

is t h e d i v i s i b l e

on e v e r y

But

it is s u f f i c i e n t

map

a homomorphism

since

to the c l o s u r e

Thus

R-linear

space

of the

every

D = mJR

closed,

dimensional

limit

extension

B/B 0 E D

that

is

in t h e

fact,

is o p e n .

in B / B 0

7 -l(-1/4,
a n d so

~ T

to a s u b g r o u p

Since

m-~
is

is the d i r e c t

; if

. The

restriction

finite,

there

to

~ 0
of

Let

be

a subspace

1
of

containing

A/A 0

such

that
D

algebraically.
while

A 0 and

the a b o v e

The map
A 1 being

isomorphism

A0 A 1

D ---+ A 0 A 1
subspaces

implies

is t o p o l o g i c a l

is c o n t i n u o u s
the

as w e l l .

by

continuity
Now

the s a m e

argument

in the o p p o s i t e

A/A 0 C A 1

so t h a t

as a b o v e

direction

so

92

A
For

every

hence

finite

subspace

discrete.

is t h e r e f o r e

Thus

every

relatively

The

duality

of

is

surjective

V C

and

A1

subset

open

in

A/A 0

0
V

of

A/A 0

A/A 0

A/A 0

implies

Next

the

suppose

result

+ n-T

D = m-~

Consider

the

to

a map

kernel

of each

: B ----+ T

such

the

) m ~

vertical

that

map

evidently,

Finally,

g ( n "77 ) = 0

D1

discrete.

Then

that

m.

a map,

the

D1

and
.

pullback
"

" ~

so

map

a map

A ----+ T

has,

by

the

is

the

above,

same

as

an extension

) T

induces

the

desired

map

@ m- ~

@ n.T

Let

+ n.T

is

uniquely

A N

m- ~{ @ n . T

determined,

f0

The

map

: A

to

m. ~

+ n.T

being

the

identity

component

of

D.)

vanishes

on

A0

and hence

induces

A/A 0

continuous
get

is

a subgroup

extension

to

of

D1

a map

> T

then

to

: D

gLA

a map
> T

~ T

a map

f
Since

}T

: A0

and
g

: A/A 0
, that

> T
is

: D1 ~

discrete
T

and

Composed

a map
h

: D

~ T

that
hIA

- gli

or
f
Thus

A C D

restriction

an e x t e n s i o n

such

such
have

if

has

we

and we

n-T

Such

(m + n)

f
is

in e v e r y

is

let

A0
(Note

n'~
.

, and

=
with

is

f (n~ ) = 0

g
for w h i c h ,

open

discrete

and

is

subspace

(rn + n)

!
map

relatively
A/A 0

vector

) A
for

where

no

that

follows

D = m-~

is

Thus

D
and

contains

g + h

is

the

desired

gIA

+ hIA

extension.

(g + h) IA

T
with

is

injective

the

projection

has
D

a
~ D]

93

(5. 35)

To take care of I II.4.2(v),

s u b g r o u p such t h a t
the class of

A*

C * ---+
Suppose

we

c o n s i d e r the case of

is injective.

C E D

If

g e n e r a t e d by

is a p r o p e r
x

and

and

A C C

F x~

is a f i n i t e l y g e n e r a t e d a b e l i a n g r o u p and

C E C

is done b y s u c c e s s i v e l y e n l a r g i n g

first t h a t
C

where

This

closed subgroup
Then

B/A

let

is finite.

x E C - A

and

In that case,

be the s u b g r o u p

is cyclic and has a n o n - z e r o map to

The

composite
B
can, by the p r e v i o u s
The

and

fRTM ---+ T m

where

~ T

s e c t i o n be e x t e n d e d to

Thus

---+ A

is not injective.

case
C

with

~ B/A

finite

Rewrite

F0

is e a s i l y r e d u c e d to the p r e v i o u s

one b y p u l l i n g b a c k along

t h a t as

is finite.

torus and is,

F x~R n

Then

in fact,

F 0 Tm

is a c o m p a c t s u b g r o u p of a finite d i m e n s i o n a l

the m o s t g e n e r a l such.

For the dual of a c o m p a c t s u b g r o u p of a

torus is the q u o t i e n t of a f i n i t e l y g e n e r a t e d

free group.

This

is the d i r e c t sum of a

free group and finite group and its dual is the s u m of a torus and a finite group.
N o w in the m o s t g e n e r a l case,
:

C
It follows

from the d u a l i t y

arbitrary.
topology
of

If

that

C1 x P R n x 77 k

C1

is not dense in

that there are

is n o t dense

can be e m b e d d e d in a p o w e r
C

finite sets

in the image

, Jt follows
m0 C m,

CO

of

m
T0x~0

n O C n,

CO

<

and

A0

(5.36)

k0 C k

readily

it now follows that the t h e o r y applies.

scribed here extends
here to show that.
9.6 and 9 . 8
i.

Every

in

n
x~

0 x

such that the image A 0

with

It is the

no
~k 0
x
.
fR

T h a t is,

22k
0

from 5.15.

from the p r e v i o u s

A l l the h y p o t h e s e s

case.
satisfied,

case that the d u a l i t y t h e o r y de-

a l t h o u g h not s u f f i c i e n t m a c h i n e r y

facts are these:

(See [Hewitt-Ross],

is d e v e l o p e d

89, e s p e c i a l l y

.)

locally compact group with

only at zero)
finite ;

that of P o n t r j a z i n
The m i s s i n g

C1

The r e s u l t now follows

follows

may be

k0

m)
F0 T C

is a n o n - d e n s e subgroup.
The last h y p o t h e s i s

x~Z

is the p r o d u c t of the image of

CO

where

in

m0
The group

Tm

from the d e f i n i t i o n of the p r o d u c t

is of the

form

a norm

D ~ n-rR @ m-T

(i.e. a s e m i n o r m t a k i n g on the value

where

is d i s c r e t e

and

n, m

are

zero

94

2.

Every

C x Rn

locally

zzm

For

if

it is c l e a r
compact
group

group

generated

is c o m p a c t

and

that

group

and

compact

where

are

and

is a s u b g r o u p
a seminorm,

and

the

by

a compact

are

m,

full

set

subcategories

are d u a l

under

of a p r o d u c t

is of the

form

finite.

the

of

duality.

of g r o u p s

in

and

Moreover
For

~O

if

described
every
L

above,

locally

is s u c h

let
L

{x Lip(x).

L/L

: 0}

P
h/p

P
Then
L
and

L/p E ~

that they
Let

empty

all

be

Since

lie

in

interior.

a n d is h e n c e

open.
The s e t

Let

subgroup
An open set
X C L0

to b e o p e n b y

a seminorm

compact

be

generated

shows

(5.37)
To s e e

L*

that

let

has

n ~

by

is a c o m p a c t

X + M

also has

space

be

it f o l l o w s

a compact

set with

non-empty

is l o c a l l y

set.

non-

interior

compact

so

1
1
(- ~, ~)}

topology.

the w a y ,

that every

dual has

group

of

integers

topologized

x~

x~

the

of

~ L

the c o m p a c t / o p e n

be

is a s u b g r o u p

and

X + M

compact

T[f (X) C

which

by

group

~-complete,

the m a p

It is n o t true,
this,

}
hence

Hence

L
This

compact

Then

in a l o c a l l y

on

is c o m p l e t e ,

a locally

of

L0

{f:
is f o r c e d

~{L/plp

locally

neighborhood

The

L0

a compact

every

T 0

The maps

A ----+ T

the

compact/open

as a s u b g r o u p

topology.
of

are

found among

the m a p s

}7 - - +

,
i.e.

the e l e m e n t s

of

2n~

for all s u f f i c i e n t l y

order

is a p o w e r

of

2 .

consists

large
In

A
2, 4,

converges

to

, so t h a t
X

is c o m p a c t .

the

That

s > 0

they

are

the

of

which

elements

of

annihilate
T

whose

8,

16,

...,

2 n,

...

the s e t
=

{0,

2, 4,

such

8, 16 . . . . .

2n .... }

: A

that

, there

f=
is a

~ T

that
[ (X)

If

is,

elements

sequence

I claim

the

Suppose

is a h o m o m o r p h i s m

Then

of all

For

such

if not,

1
1
(- ~, ~)
4

C
let

be

<

1
2

that
1
4

--

k
<

the a b s o l u t e l y

least

residue

of

f (2)

95

and then
Using

t h a t the
the

f (2 k) ~

the b i n a r y

group

(-1/4,

zero homomorphism
~7
2

1/4)

Thus

representation

of

is o p e n

of e l e m e n t s

of

f (2) : 0

we

in the

2 power
A*

from which

see

that

compact

order,
lim
--+

open

topology

topologized

2n

Then
A**

which

is the

2-adie

completion

=
of

*
lim ~
~---2n

f (2 k)

f (n) = 0

lira ~[
<----2n

= 0

as w e l l .
on

A*.

discretely.

for a l l

This means
Thus
In f a c t

A*

is

98

6.

(6.1)

By an i n f s e m i l a t t i c e

of e l e m e n t s
pair
y

has

an inf.

of e l e m e n t s

by

xy

in w h i c h

we

see t h a t
element

f : LI----+L 2

If

let

and

a partially

If the e m p t y

is i d e m p o t e n t .

which
denote

u : L1

preserves

obviously

a unit

is an o b j e c t
pointwise,

of

that the theory

of

Thus

parts

(6.3)

take

for

crete

d e d in a p r o d u c t

we

We

let

the

let

C*

one e a s i l y

of

III.4.2

we

take

the

of

the u n i f o r m i t y

sees,

by

and

discrete

I)

If

by

arguments,

, the

finite

and

monoid

a morphism

inf,

i.e.

a monoid

by

is a h o m o m o r p h i s m

Thus

the s e t o f m a p s

the i n t e r n a l

and hence

that

hom

and
L1

) L2

is c o m p u t e d

is an a u t o n o m o u s

satisfied.
consisting

compact

ones

(or,

we

Since

that

ultimately,
with

with

D E D

which

D*

be

of 2 e l e m e n t

is c l o s e d ,

If

ones).

C E C

uniformity

the

is c o m p a c t ,

dis-

can b e e m b e d -

0 < 1 .

the discrete

let
I

~(D,I)

of all the u n i f o r m l y

semilattices

{0,i}

C --+ I

I D.

i n f of

as a c o m m u t a t i v e

fg d e f i n e d

above.

lattice

maps
.

induced

the u s u a l
x E D

the

two e l e m e n t

L(Icl,

are

and the

set

every

semi-lattices.

x E L1

since

of U n L

all

the set of continuous

is d i s c r e t e

defined

finite

for all

is c o m m u t a t i v e

of topologically
be

and preserves

It is e v i d e n t ,

a n d ii)

For

structure

with

any

s e t as w e l l

are s e m i l a t t i c e s

of these

f(x) g(x)

= 1

the s a m e
L2

so is t h e m a p
=

u(x)

the s u b c a t e g o r y

I = T

be

lattice

equipped

i)
D

semilattices.

(6.4)

that

is d e n o t e d

and

the c a t e g o r y

f o r the m u l t i p l i c a t i o n

L(LI,L2)

variety.
We

such

s e t in w h i c h

t h a t the e m p t y

is e x a c t l y
L1

are h o m o m o r p h i s m s ,

~ L2

inf

If

fg(x)
The map

ordered

it is s u f f i c i e n t

an i n f s e m i l a t t i c e

is a f u n c t i o n
We

is m e a n t

Obviously

an inf.

every

homomorphism.
(6.2)

have

Semilattices

lattice,
so is

hence

ID

and

L(D,I),
and

compact.

If

discrete

semi-

function
: D ---+ T

defined

by
{~

~(y)

, if

=
, otherwise

is r e a d i l y
lattice

seen

to p r e s e r v e

and hence

the c a n o n i c a l

every

inf.

object

of

From
C

this,

it is i m m e d i a t e

has enough

that every

representations

into

and

that

maps
D

) D**

~ C**

are i n j e c t i o n s .
(6.5)

T o see

tice.

Then

they are

isomorphisms,

we p r o c e e d

as

follows.

Let

be

a finite

lat-

given
: L ----+ T

let
= x
tice

x = inf{y E Llg(y ) = 1 } .
.
L P

It is i m m e d i a t e
.

maps preserve

(Of c o u r s e
only

infs.

that

x -< y

a finite
rf

Since

is

implies

semilattice

finite
y -< x

and

so t h a t

is a l a t t i c e .

preserves
L*

finite

is s i m p l y

However

remember

inf,
the

lat-

that

97

f : L 0 ---+L I
is an inf p r e s e r v i n g

morphism,

the i n d u c e d
f*: L~P----+ L~ p

also p r e s e r v e s
course

infs.

The c o r r e s p o n d i n g

and is a c t u a l l y

clear.

An a r b i t r a r y

discrete

finite,

L* = lim L

of

semilattice
L

f *op : LI----+ L 0

function

the left a d j o i n t

f:)

Thus

is the d i r e c t

li T L

equipped with

preserves

sups of

in this case the d u a l i t y

is

limit of finite ones.

If

the i n v e r s e

limit topology.

If

: L ---+ 2
is u n i f o r m
means

(or e v e n continuous)

it contains

projection
inverse

a set of the form

image of i is o p e n

L* N M

where

of a s u b s e t of a finite p r o d u c t

image of

is that

the i n v e r s e

depends

a factorization

o n l y on finite m a n y

indices,

This

image u n d e r

say

el'

"''' em

Thus there is

that when

In p a r t i c u l a r

extends

~ .

L0>---+ L 1

there

Now we have

whence

In the process,

so is

of r e p r e s e n t a t i o n s

is an injection,

on finite

L1

> L0

lattices

is a

is a map

shown
EL

is a surjection,

The i d e n t i f i c a t i o n

: L * x...x
~i
which

C ~L*

of

in particular,

surjection.

x...x L
The same is true of the
~i
~n
d i f f e r e n t finite set of indices).
The r e s u l t

(with a p o s s i b l y

L 0 C L*al x . . . x L*~ m

implies,

in

is the i n v e r s e

L ---+ L 0 where

L*

----+ T

~m

that

= EL

--+ L

L ----+ L
T

we h a v e s e e n t h a t

is c o s m a l l

and that w h e n

L = +--lim L

then

L*

lim L *
+---

B u t then
L
and with each
(6.6)
(C,D)

reflexive,

This establishes
to c o n s i s t

equipped with

so is

lim L

the d u a l i t y b e t w e e n

of the s u b l a t t i c e

the d i s c r e t e

of

(ICI,D)

uniformity.

and

consisting

As

for the hom,

we take

of the u n i f o r m morphisms,

The m a p

(C,D) ---+ (D*,C*)

is the e v i d e n t

one and as for its b e i n g

uniform,

there

is n o t h i n g

to prove.

T h e com-

posite
(C,D)
is the i d e n t i t y

~ (D*,C*)

~ (C**,D**)

so the first is injective.

first it is i n j e c t i v e

as w e l l

Since

(C,D)

the s e c o n d

so t h a t e a c h is an isomorphism.
(I,D) ~ D

is an instance

of the

The equivalences

98

Horn (I, (C,D))

are clear.

In view of

(topologically)

Hom(C,D)

, 1.4.6(iii)

(e l, (C2,D))
Since b o t h sides are

(C,D) --~ (D*,C*)

(C 2, (CI,D))

discrete,

Horn (CI, (c2,m))

follows from

it is s u f f i c i e n t to show that

--~ Hom (C2, (Cl,m))

If
f : C I ----+ (C2,D)
then the image of

e q u i v a l e n c e relation
that

factors

Xl,

..., x n

is compact and discrete, hence finite.

E

on

C1

Cl---+ Cl/E ~
Let

That is, there is an

w i t h only finitely many equivalence classes such

(C2,D)

be a set of r e p r e s e n t a t i v e s mod
f(x ,
1

.
E

Each

xj

determines a function

) : C 2 ----+ D

w h i c h s i m i l a r l y factors
C2~
where

C2/Ei ~

D
Since

is an e q u i v a l e n c e relation with only finitely m a n y classes.

C2 / N E i ----+ ZC2/E i

is injective,

there are only finitely many classes mod

(N El).

Thus

determines

an e l e m e n t of
(Cl/E

, (C2/ n E i ,D))

(C2/ N E 1 ,(CI/E , D)) ---+ (C 2 , (C1,D))

the exchange p o s s i b l e there b e c a u s e all three are d i s c r e t e and
category.

is an autonomous

This determines a map

(C I, (C2,E)) ---+ (C2, (CI,E))

w h i c h is e v i d e n t l y an involution.

Thus we h a v e a pre - * - automonous situation.

The d i s c r e t e u n i f o r m i t y of

(C,D)

is evidently that of global u n i f o r m

c o n v e r g e n c e and is a c o n v e r g e n c e uniformity.

Thus the first three parts of III.4.2

are satisfied.
(6.7)

We now consider III.4.2(v).

lattice.
image

A0

Since

Let

C E C

and

A C C

is p r o f i n i t e there is a finite q u o t i e n t

remains proper,

else

w o u l d b e dense.

Since

be a p r o p e r closed subC
of C such that the
,0
C O ---+,A0 is s u r j e c t i v e

and not an i s o m o r p h i s m it cannot be i n j e c t i v e w h e n c e neither is

C ---+ A*

(6.8)

The first is already

We now show that the h y p o t h e s e s of II.29 are satisfied.

shown and the third is evident.

For the second,

it suffices to show that every dia-

gram of the form

A C

) nD

can b e completed.
such that

As in 6.5 above there is a finite set of indices

depends only on those indices.

Wl'

"''' ~n

We can thus factor the d i a g r a m as

99

~ zD

D Ix...x

A 0 C__+

= B0

n
+
T
Now

A0

and

B0

are d i s c r e t e .

If

is n o t

B 0 ----+ A 0

surjective

let

be

its

image.
We h a v e

the

composite
A0

is i n j e c t i v e ,

whence

the

proper

The

condition

last

in

C*

A0

first

and by hypothesis

the

factor

A0

is.

is

contradicts

is i m m e d i a t e

B0

compact

so

that

the p r e v i o u s

and hence

we have

is

closed

paragraph.
another

model

of

theory.

(6.9)

The

model

of a

category

category

maps

fact models
follows.

if

that

semilattiees

complete

a set,

: X ----+ Y

for

A C X

that

is n o t

is the

inf preserving
theory.

assigns

complete

homomorphisms

constructed
lattice.

category

triple

whose

TX

~ =

in the w a y

provides

described

It is u n d e r s t o o d
objects

functions.)

The

are

complete

It is a c l o s e d

(T , ~

, ~)

that

here.
the

lattices

category,

in

can be described

as

let

is a map,

T f

2X

is t h e

right

adjoint

This means

to 2f : 2 y ---+ 2x

Tf(A)
unit

and

is in f a c t a c o m p l e t e

of commutative

For

are

semilattices

category

semilattice

of c o m p l e t e

and whose

and

of complete

*-autonomous

(A c o m p l e t e

The

B0

But

and this

of III.4.2

B0

is t h e

: X ---+ T X

to a f a m i l y

of s u b s e t s

{yl/l(y)

singleton

map

C A } .

and

~X

: 2 2 x ---+ 2 x

their

intersection.

Thus

(x ,Y)
consists
by

the

of all

infs

in

inf preserving
Y

homomorphisms

map

category.
readily

X ----+ 2
Hence

seen

*-autonomous
a compact

into a complete

lattice

(X , 2)

is a l i m i t p r e s e r v i n g
X*

and

to be o r d e r
structure

category:

made

In p a r t i c u l a r ,
X* =

Every

X ---+ Y

have

inverting.

Thus

are o b v i o u s .

a *-autonomous

set-valued

isomorphic

X* ~ X p

It h a s b e e n
category

functor

underlying

such

sets

I ; T

considered

as a

a n d the i s o m o r p h i s m

from which

incorrectly
that

on

the d u a l i t y

conjectured
and such

that

that

is

a n d the
this

the

is

natural

map
X* Y - - ~
induced

by

exchanging

and

X
X*

is an i s o m o r p h i s m .

Here

(X ,Y)

in

X --~

> I

I ---+

(Y

,Y)

is e v a l u a t i o n

and

I ---+ (X,X)

t h e u n i t map.

100

Since

(X,Y)

either

of the n a t u r a l

(X Y )

, compactness

maps

(whose

is e q u i v a l e n t

constructions

an i s o m o r p h i s m .

(xP,Y p)

were

dual

isomo~hic

~derlying

In the

sets.

and,

this would

in p a r t i c u l a r ,

I have

to

(X ,Y )

c a s e of l a t t i c e s ,

lattices

category)

I = T )

---+ (X Y)

(X,Y)
being

(in a * - a u t o n o m o u s

are e a s y -- g i v e n

verified,

that

using

imply

(X,Y)

an

that

and

(X,Y)

(xOP,

HP67 progra~le

~d

yOp)

have

calculator,

that when

x:xOp

/i\

\i/

and

I\

/\1\
\i/
then there
The

are 94

computation

integers
would

inf preserving

is c a r r i e d

ordered

X ---+ Y

out by modeling

by divisibility

s e e m to be p o s s i b l e
To have

maps

such

for any

compactness,

that

finite

and only

(as w e l l

the

88 s u c h m a p s
as

yOp)

X p ~ X ---+ y O p

as a s e t o f p o s i t i v e

inf of two numbers

is t h e i r

gcd

This

lattice.

it is n e c e s s a r y

-- a n d a l m o s t

surely

sufficient

-- to h a v e

trace map
tr
such that

the

(X,I)
is the i d e n t i t y .
the

Here

first and second

lattices
triple)

of complete
has

e d to a n y

: (X,X)

---+ (X, (X,X))

the

first map

copies

of

subcategory.

---+ (X, (X,X))

is i n d u c e d
The

atomic boolean

such a trace map but

larger

its

by

the u n i t a n d the s e c o n d

(which

form suggests

Namely,

(X'tr') ~ (X,I)

full subcategory

algebras

tr
be defined

~ X* X ---+ I ,

composite

if the

: (X,X)

o f the

category

is t h e K l e i s l i
very much

lattice

interchanges
of complete

category

for the

t h a t it c a n n o t b e e x t e n d -

X = 2A

, then

let

----+ 2

by

tr(f)

In o t h e r w o r d s ,

trace

for a n y s u b s e t

A0 < A

is c o m p l e m e n t e d

and atomic

i,

if

O,

otherwise

is r e p r e s e n t e d
g(A)

= A

that omits
seems

a @ f (A - {a})

by

g(A-{a})

crucial

the m a p

at l e a s t

for a l l a @ A

= a

g : X ---+ X
,

g ( A 0)

two elements

here.

defined

by

=
of

The

fact

that

BIBLIOGRAPHY

M. Barr, Duality of vector spaces, Cahiers Topologie G~om~trie Diff~rentielle

17 (1976),3-14.
M. Barr, Duality of banach spaces, Cahiers Topologie G~om~trie Diff~rentielle
17 (1976), 15-32
M. Barr, Closed categories and topological vector spaces,
G~om~trie Diff~rentielle 17 (1976), 223-234.

Cahiers Topologie

M. Barr, Closed categories and banach spaces, Cahiers Topologie G~om~trie

Diff~rentielle 17 (1976), 335-342.
M. Barr, A closed category of reflexive topological abelian groups,
Topologie G~om~trie Diff~rentielle 18 (1977), 221-248.

Cahiers

M. Barr, The point of the empty set, Cahiers Topologie G~om~trie Diff~rentielle
13 (1973), 357-368.
8. Eilenberg, G.M. Kelly, Closed categories, Proc. Conf. Categorical Algebra
(La Jolla, 1965), Springer-Verlag, 1966, 421-562.
E. Hewitt, K.A. Ross, Abstract Hamonic Analysis, Vol. I, 1963, Springer-Verlag.
K.H. Hofmann, M. Mislove, A. Stralka, The Pontryagin Duality of Compact
O-Dimensional Semilattices and its Applications, Lecture Notes
Math. 396, (1974), Springer-Verlag.
J.R. Isbell, Uniform Spaces, Amer. Math. Soc. Surveys no. 12, 1964.
J.L. Kelley,General Topology, Van Nostrand, 1955.
G.M. Kelly, Monomorphisms, epimorphisms and pull-backs, J. Austral. Math. Soc.
9, (1969), 124-142.
F.W. Lawvere, Functional Semantics of Algebraic Theories, Dissertation,
Columbia University, 1963.
S. Lefschetz, Algebraic Topology, Amer. Math. Soc. Colloquium Publications,
Vol. XXVII, 1942.
F.E.J. Linton, Some aspects of equational categories, Proc. Conf. Categorical
Algebra (La Jolla, 1965), Springer-Verlag, 1966, 84-94.
A. Pietsch, Nuclear Locally Convex Spaces, Springer-Verlag, 1972.
H.H. Schaefer, Topological Vector Spaces,

third printing, Springer-Verlag, 1970.

Z. Samadeni, Projectivity, injectivity and duality, Rozprawy Mat. 35 (1963).

A. Wiweger, Linear spaces with mixed topology, Studia Math. 20 (1961), 47-68.

CONSTRUCTING

*-AUTONOMOUS CATEGORIES

Po-Hsiang Chu

CHAPTER I:

PRELIMINARIES

We will be dealing with closed symmetric monoidal

*-autonomous

Kelly coherence conditions

(see [MacLane,Kelly]),

following useful theorem

[to appear].

Theorem:

(autonomous)

categories as defined in the previous paper.

and

Using the MacLane-

M.F. Szabo has proved the

A diagram commutes in all closed symmetric monoidal categories

iff it commutes in the category of real vector spaces.

This theorem not only points out the notion of closed symmetric
monoidal category is a 'correct' generalization

of the category of vector

spaces, but it also provides a very easy method to check if a diagram is

commutative in any closed symmetric monoidal category.
The following is a collection of easy consequences

of this theoren

which we shall use later on:

Corollary i.

Given A,B,C objects in

and map

AB

> C, then the

following diagram commutes:

i

~ (A,A)

(B,B)
where the map~ (A,A)
(A,A)
Note.
The map

(A~B,C)

(id'fJ>(A,(B,C))
x
A

) (B,C)

(B,B)

is the composition
>(AB,C)

is the usual transpose of

AB

~ C .

(AB,C) is obtained in a similar fashion.

we simply denote either composite by

Corollary 2.

(AB,C)

From now on

f .

Given A,B,C,D,F objects in

V and map BC

> F, then the

following diagram commutes:

(A,C)(D,B)

(id,f)if

lid(id, f)
(A,C)(D,C,F))

Ip-lid
(AB,F)(D,B)

Iidp-i

I (s,id)id

(A,C)(DC,F)

(BA,F)(D,B)

lid(s,id)
(A,C)(CD, F)

Ipid
(B, (A,F))(D,B)

iid~p

iM

(A,C)(C, (D,F))

(D, (A, F))

(C,(D,F))(A,C)

(A,(D,F))

> (A,(B,F))(D,B)

p-i

, (AD,F)

'ts'id)'(DA,F)

104

PROOF.

It is easy to check that the diagram commutes in the category of

real vector spaces.

Remark. The word "coherence" is going to appear frequently throughout this

paper.

In particular,

if the commutativity of a certain diagram is said to

be implied by coherence, we understand that its commutativity follows easily

from this theorem.
Our second assumption on V is that it has pullbacks.

Since almost

all interesting examples of closed symmetric monoidal categories have this

property,

this restriction is not too drastic.

The following is a collection of examples satisfies our assumption:

(i)

The category of vector spaces over a fixed field K;

(ii)

The category of Banach spaces;

(iii)

The category of compactly generated spaces;

(iv)

The category of sets (and functions);

(v)

The category of abelian groups;

(vi)

The category of lattices.

An example of a closed symmetric monoidal category that does not have

pullbacks is the category of sets and relations.

CHAPTER II:

CONSTRUCTION OF ~ X AND ITS ENRICHMENT OVER V.

i.

The Category ~

Given an arbitrary object X in V, we shall construct a category_A X

as follows:
The objects of ~ X consist of triplets
in V and v:VeV'

> X is a morphism in V.

A morphism from (V,V',v)

and g:W'

commutes.

(V,V',v) where V,V' are objects

to (W,W',w) is a pair (f,g), where f:V

~V' are morphisms in V such that the square

VW'

idg

WW '

~ VV'

>

~W

105

If (f,g):(V,V',v)

~ (W,W',w) and (h,k):(W,W',w)

(U,U',u)

are morphisms in Ax then the following diagram commutes:

VU'

WU'

idk

id$

> VW'

idk

> VV'

+ WW'

lhid

UU'

> X

This implies the composition of (f,g) and (h,k) is (hof,gok) in AX .

Since the composition is defined explicitly in terms of morphisms in V , the
associativity of maps in AX

can now be verified:

If (f,g):(V,V',v)

(W,W',w)

(h,k):(W,W',w)

(U,U',u)

(l,m):(U,U',u)

(T,T',t)

are morphisms in AX , then

((l,m) o(h,k))o(f,g)

= (loh,kom) o(f,g)
= ((loh)of, go(kom)
= (lo(hof),(gok)om)
= (l,m) o (hof, gok)
= (l,m) o((h,k)o(f,g))

Moreover,

Id(V,V',v)

= (idv,idv,)

is the obvious identity.

Hence we have

shown that Ax is a category.

2.

__~ is ~hriched over X

Definition.

If V is a closed monoidal category,

then A is enriched over

if A is equipped with the following:

i) For each A,B in A, an object V(A,B)

in V;

ii) For each A in A, a morphism j(A):I

> V(A,A) in V ;

iii) For each A,B,C in A, a morphism

M'(A,B,C):~(B,C)V(A,B)

~ X(A,C) in ~ .

These data are required to satisfy the following axioms:

VC i.

The following diagram commutes:

M'

V(A,B)V(A,A)

V(A,B) I ~

~ V(A,B)

106

VC 2.

The following diagram commutes:

M'

V(A,A)V(B,A)

~(B,A)

j @ i d f
IV(B,A) I
VC 3.

The following diagram commutes:

(V(C,D)V(B,C))V(A,B)

) V(C,D)(V(B,C)V(A,B))

iM'eid

[idM'

V(B,D)V(A,B)

V(C,D)X(A,C)

~ X ( A , D ) / M '
Given

A = (V,V',v),B = (W,W',w)

the object in

objects in A X , define V(A,B) to be

such that the following square is a pullback.

V(A,B)

pl

> (V,W)

(W',V')

~ (V~W',X)

Here

-i
(v,w)

~ (ww',x)

(v,w)

~ (v,(w',x))

and

> (vw',x)

-i
(w',v')

~ (ww',x)

(w',v')

~ (w',(v,x))

are the right and bottom maps, respectively.

isomorphism in
Given

~(w'~v,x)

~(VW',X)

Therefore V(A,B) is defined up to

V .

A = (V,V',v)

in _~ , the following diagram commutes, by

Corollary i:
I

(v',v')

> (V,V)

, (vv',x)

Universal property of pullbacks implies that there exists a unique map j(A)
I ----~V(A,A)

in

such that the diagram

107

V(A,A)

~(V,V)

(v',v')

v + (vv',x)

commutes.
Now suppose
in ~ X "

A = (V,V',v),

In order to verify

B = (W,W',w), C = (T,T',t)

are three objects

iii) it suffices to show the outer square of the

diagram.
_V(B, C) V(A, B)

pl~pl

) (W,T)(V,W)

p2p2

IM

(T',W')(W',V')

V(A,C)

~ (V,T)

p1

It

(W',V')(T',W')

(T',V')

, (VT',X)

commutes.
Using the fact that -~_V(A,B)

pl

(W',V')

is a bifunctor and
> (V,W)

V(B,C)

> (VW',X)

(T',W')

are pullbacks (hence commute!),

tative diagram in Fig. 1.

~ (W,T)

, (WT',X)

we can get the desired result from the commu-

Note in Fig. 1 that corollary 2 of Szabo's theorem

(Chapter I) implies that (2) commutes;

commute.

pl

coherence implies that (i) and (3)

Again using the universal property of pullbacks, there exists a

unique morphism

M'(A,B,C):V(B,C)V(A,B)

V(A,C)

in

V such that the

diagram
V(B,C)V(A,B)

plpl

(T',W')(W',V')

(W',V')(T',W')
commutes.

Hence

V(A,C)

, (T',V')

i) - iii) are defined.

> (W,T)(V,W)

pl

, (V,T)

, (VT',X)

108

Now we have to show they satisfy the required axioms.

Given

A = (V,V',v),

B = (W,W',w)

in _~ , by construction we have

the pullback diagram:

V(A,B)

pl

p21

p.b.

(w',v')
But the coherence of
V(B,C)OX(A,B)

id~pl

> (V,W)
I~

~ (vaw',x)

implies that the diagrams of Fig. i commute.

V(B,C)(V,W)

plaid

~(W,T)e(V,W)

p2aid

p2id

(T' ,W')V(A,B)

id~pl

(T',W')a(V,W)

ltaid
waid

(WaT',X)a(V,W)/t

id~p2

(T' ,W')a(W' ,V')

id~

, (T',W')(VaW',X)

, (VAT'

(3)

(w' ,v')a(T' ,W')

V(A,B)I

~ V(A,B)

(V,W) I

(V,W)

(T',V')

V(A,B) I

(W' ,V')I
FIGURE i.

> V(A,B)

~ (W' ,V')

~ (V,T)

109

Hence the following diagram

plid

V(A,B)I

(V,W)I

V(A, B)
,2id

lr
1

(v,w)

pl

p2

p.b.

(VW', X)

(W' ,V' )

(W',V')l
commutes.

rI

Since the outer square commutes,

such that (1)

and (2)

property as well;

commute.

there exists a unique map V(A,B)I

But the map

V(A,B)I

V(A, B)

~ V(A,B) has this

therefore it follows from uniqueness that it is the map induced

by pulling back.
Recall that in the construction of

j(A)

we have the following commutative

diagram:

~ ~~

~ -~V ( A~, A )'

"

~ v ( v ' , v ' )
Then the defining property of M(A,B,A),
-~-

~ (v,v)

> (vv' ,x)

coherence of V_ , and the fact that

is a bifunctor imply that the diagram:

V(A, B) V(A,A)

Ip2

(W' ,V')(V' ,V')

V(A,B)

Is
IV(A,B) ip2>(V',V')(W',V')
commutes.

plpl

~ (V,W)(V,V)

pl

ip pb
M

,(U' ,V')

> (V,W)

Io
, (VW' ,X)

110

Again applying
_V(A,B)I

idj

the same argument,

M'

+ V(A,B)_V(A,A)

we conclude
+ V(A,B)

that the map

is the map induced by

pulling back.
But this is not sufficient

to conclude

that VCI. holds,

i.e.

that the

diagram:
M'

V(A,B) V(A,A)

> V(A, B)

V(A,B)I
commutes.
We are still required
V(A,B)I

plid

+ (V,W)@I
.-

/
t

(V,W)e(V,V)

to show that the following

V(A,B)I

diagrams

p2id

> (W',V')I

+ (V,W)

commute:

I
S

(W',V')(V',V')

I(W',V')

11
/ j

j/~/

I
M

(V',V')e(W',V')
That is, that the induced maps
(therefore

(V,W)I

satisfy

,(W',V')

the same commutative

square

they are same by uniqueness).

But it is trivial once we notice there exist canonical maps

idi
ieid
~ (V,W)(V,V)
in (i) and
I(W',V')
) (V',V')(W',V')

in (2) which break

(i) and (2) into two smaller

conmlutative

squares.

Hence

VCI. holds.
Applying

a similar

argument,

we conclude

VC2.

is also

true.

Next we

are going to verify VC3.

Given
in _ ~

A = (V,V',v),

, then by iii) we have

Coherence

(2) of Figure

B = (W,W',w),

of V and property

2 commute;

Now we apply

C = (T,T',t),

the commutative

diagrams

of M(A,B,C)

similarly

(i') and

imply

that subdiagrams

the same argument as in proving VCI,

idM'
+ !(C,D)V(A,C)
M'id

(V(C,D)V(B,C))V(A,B)

by pulling back.
a(V(C,D),

V(B,C),

+ V(B,D)V(A,B)
We only have

*_V(C,D)V(A,C)

M'

> V(A,D)

to show that the composition

, V(C, D) (V_(B, C) @V (A, B) )

M'

(i) and

i.e. the maps

M'
~ V(A,D)

_V(A,B))

(V(C, D) V (B, C) ) V(A, B)

idM'

objects

2.

(2') commute.

V(C,D)(V(B,C)V(A,B))

are the maps induced

D = (U,U',u)

of Figure

V(A,D)

111
pl(pl~pl)

V(C, D) (V(B, C) V(A, B) )

> (T,U)((W,T)(V,W))

(i)

(U',T')((T' ,W')(W',V'))
ids

idoM

V(C, D) _V(A, C)

plopl

> (T,U)(V,T)

(2)

(U' ,T')((W' ,V' )(r' ,W'))

(U', T')(T' ,V' )

Is
((W',V')(T',W'))(U',T')Mid~(T',V')(U',T ')

V(A,D)

pl

pl

p.b.

>(u',v')

(pl~pl)pl

(~(C,D)V(B,C))~(A,B)

, (V,U)
]~
>(VoU',X)

> ((T,U)(W,T))(V,W)
L

(p2~p2)~p2~id
((U',T')(T',W'))(W',V')

sid

(i ')
V(B, D)V(A,B)

(2')

((T',W')(U',T'))(W',V')

IMi d
pl~pl

~ (W,U)(V,W)

p2

(U' ,W')(W' ,V')

Ls
(W',V')O((T',W')(U',T')) idM~(w',v')(U';W')

V(A,D)

pl

plpb
, (U',V')

~ (V,U)

~r

I
, (VU',X)

FIGURE 2.
is also a map induced by pullback and it satisfies the same commutative
square as the map:
(V(C,D)V(B,C))V(A,B)

__

M'oid > V(B,D)V(A,B )

M T

> V(A,D)

The first part follows easily from the following commutative

diagram:

112
(T,U)((W,T)(V,W))

((T,U)(W,T))e(V,W)

pl(pl@pl)

(plpl)@pl

X(C,D)(X(B,C)@V(A,B))

(X(C,D)~(B,C))V(A,B)

p2(p2@p2)

(p2@p2)@p2

(U',T')((T',W')(W',V'))

((U' ,T')(T' ,W'))(W' ,V' )

ids

sid

(U',T')((W',V')(T',W'))

((T'.,W')e(U' ,T'))(W' ,V')

-1

(W' ,V')((T',W')e(U',T'))

((W',V')(T',W'))(U',T')

As for the second part, we observe a simple fact of

V : two

permutations of the tensor product of any three fixed objects are coherently
isomorphic.

Therefore it is enough to show the following diagrams commute:

((T,U)(W,T))(V,W)

(W,U)(V,W)

((W',V')(T',W'))(U',T')
iMid
(T' ,V')(U' ,T')

> (T,U)((W,T)(V,W))

(T,U)(V,T)

> (W',V')((T',W')(U',T'))
lidM
(W' ,V')(U' ,W')

This follows trivially from coherence and completes the proof.

113

CHAPTER III:

1.

STRUCTURE

The H o m - F u n c t o r ~ ( - , - )

Definition.
in ~

~_
--x HAS A *-AUTONOMOUS

Given any two objects

= (V,V',v)

and B = (W,W',w)

, define an object _~(A,B)

= (V(A,B), VOW', n)

First of all recall V(A,B)

is the object in

in ~ X as follows:
such that the follow-

ing diagram is a pullback.

V(A,B)

pl

~ (V,W)

(w',v')

, (vw',x)

Since we require _~(A,B)

to be an object i n _ ~

morphism in V, which sends V(A,B)(VW')

to X.

It seems there are two (canonical)

(i)

Since the above square commutes,

route) which sends V(A,B)
n:~(A,B)(VW')

(2)

alternatives

for defining

let n' be the morphism

to (VW',X)

n:

(along either

, and define

> X to be the transpose of n'.

Again since the above square commutes, we have the following

commutative

diagram:
plid

V(A,B)(VW')

> (V,W)(VW')

~id

p2id I
(W',V')(VW')
Now let

~id

ev:(VW',X)(VW')

But since

is coherent,

n', so these two definitions

) (VW',X)(VW')

> X be the evaluation map, then put

n" = ev composed with the above map

to

, n has to be a

V(A,B)(VW')

_~op

_~

to

A X(-,-)

is a bifunctor

_~ .

We have to show
i)

given any object

B = (W,W',w)

is a contravariant
ii)
iii)

G =_~(B,-)
Given A

in _ ~ , F = A_X(-,B)

functor;

is a covariant
~ B,

is identical

are same.

For the rest of this section we shall prove

which sends

~ (VW',X)(VW').

it is easy to verify that

funetor;

~ D in A X , then the diagram

114

A_x(B,C)

~ Ax(A,C)

A_x(B,D)

, Ax(A,D)

commutes.
Recall if

C = (V,V',v) and

A = (P,P',p)

in A X and (f,g):C

is a morphism in AX, then the square:

VP '

idg

PP '

~ VV'

~ X

commutes.
In order to show

is contravariant,

we must find a map (in A X)

F(f,g) = (f',g'):_Ax(A,B)
By definition

_~(A,B) = (V(A,B), PW', n I) and Ax(C,B) =

(V(C,B), VW', n 2) ;
As for

~ Ax(C,B)

so the choice for

g'

is clear:

g' = fid:VW'

f', consider the following diagram:

pl
V(A B)

~(C,B)

pl

(*) p2

p21

I
~

p.b.

(W',V,)-- ~

(P,W)

(V,W)(~

+ PW'

f,id)

(i)

> (VW',X)

d,g)

id)

(W' ,P' )

~ ( P ~ W '

,X)

We know the outer square commutes,

therefore it suffices to show (i)

and (2) are commutative.

For (i), we prove it by looking at the following commutative diagram:

(P,(W',X))

(P,W)

(f,id)

> (V,W)

~ (PW',X)

(fid,

id)

~ (VW',X) +---P

(V,(W',X))

115

As for (2) we have a similar diagramatical proof:

(w',e)

(W',(P,X))

(id,$)

(W'eP,X)

+ (W',V')

(idf,id)

But in this case the commutativity

the fact that
diagram

(*)

(f,g)

, (W'V,X) ~

(W',(V,X))

of the outer square is due to

is a morphism which sends

to

(hence the

above commutes).

This implies that there is a unique map

f':V(A,B)

+ V(C,B)

induced by pullback such that the diagram

pl

V(A,B)

+ (P,W)
id)

pl

~(C,B)

, (v,w)

p.b.
(W',V')

(VW',X)
K
(fid,id)

iN,g)
(W',P')

~'.

(PW',X)

commutes.
Therefore the following diagram commutes:
V(A,B)

p2

~(W',P) -

f'

> (PW',X)

(id,g)

V(C,B)

p2

(feid,id)

~(W',V')

(VW',X)

This implies that:

V(A,B)(VOW, )

idg'

1.

f'Qid

V(C,B)(VW')
commutes.

+ _V(A,B)(PW' )

n2
~ X

116

Therefore

(f',g')

has the property required of a morphism in

It is trivial to see that

F

F(id A) = idF(A) . Now we have to show

preserves composition.
Suppose

A = (P,P',p),

C = (V,V',v)

objects in _~ , moreover (f,g):E

~C

and
and

E = (U,U',u)
(h,k):C

~A

are three
then we

want to show that

Ax(A,B )
((hf)''(gk)')

(h',k')

> A_x(C,B)
f

(f''g')

_~(E,B)
commutes.
By definition:

_~(A,B) = (V(A,B), PW', nl)

_~(C,B) = (V(C,B), VW', n 2)
_Ax(E,B) = (V(E,B), UW', n3)

Now we consider the following commutative diagram:

pl

V(A,B)

(P,W)

pl

V(C,B)

~ (V,W)
(f, i d a /

_V(E,C)

pl ~ (U,W)
1

p2

p2

p.b.

[~

~(UW',X)

(W' ,U' )
(id,g)

(W' ,V' )

(W',P')

~(fid,id)~
v

~ (VW',X)

~(PW',X)

117

But the following diagrams also commute:

(W' ,P')

(id~k)

~ (W' ,V')

(W' ,U' )
(PW',X) (hid,id)

((hf)id'id) I

> (VW',X)

~~~(f~id,id)

(UoW',X)

(h,id)

(P,W)

(v,w)

(hof,id)

(u,w)
This implies that the map induced by pullback is identical to
and clearly

k'og' = (hid) o(fid) = ((hof)id) = (gok)'

Hence

f'oh',
F

is a

contravariant functor.
As for
B = (W,W',w),
A

> C

G,

we have a similar series of diagrammatical proofs:

A = (P,P',p),

a morphism i n _ ~

C = (V,V',v) are objects i n _ ~ with

. We need

G(f,g) = (f',g'):G(A)

Suppose

(f,g):
> G(C).

By definition
G(A) = _~(B,A) = (V(B,A) , WP', n I)
and

G(C) = _~(B,C) = (_V(B,C) , WV', n 2)

Hence the choice of
g' = idg:WV'
is clear.

> WP'

And the following commutative diagram shows the existence and

uniqueness of

f':

118

pl

V(B,A)

pl

V(B,C)
p2

> (w,e)

(W,V) ( ~ i d , f )

p.b.

]p2

) (WV',X) id)
~
> (WP',X)

(V' ,W')

(P',W')

Again the preservation of the identity is clear.

Now if
AX

and

A = (P,P',p),

(f,g):A

> C,

C = (V,V',v),
(h,k):C

E = (U,U',u) are objects in

) D are morphisms, then the

commutative diagrams of Figure 3 imply

preserves composition.

To prove (iii):
Suppose

A = (V,V'v),

objects in

~X

B = (W,W',w),

and (f,g):A

C = (P,P',p),

) B,

(h,k):C

D = (U,U',u)
~ D

are maps in

then the following diagrams commute:

pl

V(B C)

V(B,D)

p2

(w,P)

p2

, (w,u)
I~

p .b.

# .

(WU',X)

(U' ,W' )

(idk,id)~
> (WP',X)

(P',w')

pl

V(B,D)
V(A,D)

lp2

p2

(U',V')
Sd,g)
(U',W')

> (W, U)

~,idl
pl ~ (V,U)
p.b.

lfi
~ (VU',X)

(gid,id)~
> (WU',X)

are
~X '

119

(W,P)

(id,f)

(WP',X)(ideg,id) ~ (WV',X)

+ (W,V)

(id,hf)1

(id(gok),id)I

~k,id)

(wu',x)

(w,u)

(P',W')(g,id)___+ (V',W')
(gk'id)I
(u',w')

V(B

pl

,A)

v_(B,c)

> (W,P)
(id,/
(w,v)~

pl
!(B,E)
-- pl
i

~2

lp2

p.b.

, (w,u)

I~

(U',W')~ +(WU',X)
,id) (idk,id)~
+ (WV',X)

(V',w')

(idg,id)~
(WP',X)

(P',W')
FIGURE 3.

120

pl

~(A,C~

> (V,P)

(id,h)
pl

V(A,D)

lp2

~2

(V,U)

p.b.
9

(U' ,V')

(kidi ~ ~X)
> (VP',X)

(P',V')

pl

V(B,C)

V(A,C)

(w,p)

pl

(V,P)

V(A,D) pl ,(V,U)
p2 lp2 p.b. fi
~(VU' ,X)

(u' ,v')

(kid,

id)~
(VP ', X)

(P' ,V' )

(WP',X)

(P' ,W' )
FIGURE 4.

This implies that the first diagram in Figure 4 commutes which implies, in
turn, that the second one does.
Applying the same argument, the map from
by pullback is the same as

V(B,D)

V(A,D)

induced

h"of" , hence the following diagram commutes:

lh

V(B,C)

V(B,C) to

f"

f,

lh

V(A, C)

>_V(A,D)

121
Next consider Figure 5.
is a pullback,

f'oh'

Since the center square of the first diagram

is the unique map

V(B,C)

> V(A,D)

that

makes the diagram commute.

Next consider the lower diagram of Figure 5.

Using this and the

fact that the following diagram commutes:

idk

VU'

> VP '

Ifid

[fid
idk

WU'

-.%

~ WP '

pl

V(B,C)

pl

V(B,D),

--

-~I

lp2

V(A,D)
p2

(U',V')

~id,
(U',W')
(kid)

g)

pl

~ (W,U)

,(V,U)
(f'
id)k/
>(vu' ,X)

(fid,id~,
~(WU'
,X)
(idk,~
~(WP', X)

(P' ,w' )

pl

V(B,C)

V(A,C)
p2

(W,P)

[p2

pl
p.b.

(W,P)

(V,P)

(VP ',X)

(P' ,V')

( f i ~
(WP',X)

(P',W')
FIGURE 5.

122

we obtain

the desired

result

that the diagram

Ax(B,C)

~Ax(A,C)

Ax(B ,D)

~Ax(A,D )

commutes.

2.

The Functor

*.

In this section we shall define

and examine

its relationship

Definition.

Given any object

object

(V',V,vos)

where

with

a morphism

in

definition

is justified

AX

, then define
since

VoW'

in A X define

~ vev'

object

*(f,g)

+ X
in

idg

~X

*(A)

to be the

is a map in
~X

and

V .

(f,g):A

= (g,f):*(B)

the commutativity

WW '
implies

is another

,:_~op

A = (V,V',v)

s:V'V

B = (W,W',w)

Suppose

AX(-, -)

a functor

+ *(A).

> B
This

of the diagram:

> VV'

>

that the diagram

idf

W'V

+ W'W

I gOid
V'V

Iws
> X

vos

commutes.
From the above

formula on morphisms

we can easily conclude

that

a functor.
Moreover

has an inverse

The following
Proposition
V(A,B)

i. Given

~ V(*(B),*(A))

PROOF.

are some properties

A = (V,V',v)

the commutative

Notice
commute.

of

since

*o* = idAx_

*:

B = (W,W',w')

in

~X

, then

By definition

Consider

(contravariant),

*(A) = (V',V,vos)
diagram of Figure

that the coherence

It also implies

of

and

*(B) = (W',W,wos).

6.

implies

squares

that the diagram

(VW', X)

(VW',X)

(s, id)

~ (W' V, X)

(I),

(2),

(3),

(4)

is

123

commutes.
The fact that
V(* (B) ,*(A))

pl

> (V,W)

Vos

(w',v')

, (w'~v,x)

is a pullback square implies that there exists a unique

such that the diagram of Figure 6 still commutes.
square involve

V(A,B)

induces a unique map

that the diagram of Figure 6 commutes.

p:V(A,B)

~ V(*(B),*(A))

Similarly the pullback

q:V(*(B),*(A))

This implies

~ V(A,B) such

qop is the map

induced by the pullback square:

V(A, B)
Ip2

pl

(V,W)

p'b"

(w' ,v' )

If~
(vw' ,x)

p1

V(A,B)

pl

_V(* (B(,*(A))

V(A,B)
p2

~2

(v,w)

pl

, (v,w)

, (v,w)

Wos

p.b.

(W',V')

(VW',X)

~os

(W' ,V' )

2/

(W'V,X)

(VW',X)

(w' ,v' )
FIGURE 6.

124

But by the remark above

qop = idv(A,B)

V(A,B)

and

the same argument

completes

Let

For any object

Corollary.

Let

A,B

PROOF.

By definition

diagram,

and

This

that
> W'V

in

that

Ax(A,B)

and

V(A,*(B))

= V(B,*(A))

= C.

be two objects

in

*(B) = (W',W,wos)

it is sufficient
in

= (V(A,B),VW',n2)
and

idv(,(B),,(A))

S(W',V):W'V

_AX ,

which

~ VoW'

we

q:V(*(B),*(A))

= poq

~ V(A,B)

We also have s(V,W'):VW'

such that

s(V,W')os(W',V)

(p,s(W',V))

to check that the pair

~ V(*(B),*(A))

(s,id)

A,B

is indeed

the following

Ax(A,*(B))
If

is an object

Proposition

2. Let A,B,C

in

in

= Ax(B,*(A))
~X

, then

V(A,_Ax(B,*(C)))
Let

definition

*(A) = (V',V,vos)

A = (V,V',v),

Now put

Recall

(VE
(VW',X)

in

C
~X

= *(*C)
' then

V(C,Ax(B,*(A))).

B = (W,W',w),
and

C = (U,U',u).

that

Ax(B,*(A))
V(B,*(C))

Then by

*(C) = (U',U,uos).

Bc = Ax(B,*(C)) = (V(B,*(C)),WU,n I)
Ba =

pullbacks:

(s,id),

_V(A,B)

A X , then

be three objects

PROOF.

the transpose.

be two objects

PROOF.

, (W'V,X)

the proof by taking

Let

idw,v

But we see this by considering

_AX .

moreover

diagram:

(VW',X)
and complete

' then

*(*(C))

+ V(*(B),*(A))

= qop

V(A,B)

Corollary.

~X

*(A) = (V',V,vos),

p:V(A,B)

idv(A,B)

an isomorphism
commutative

'

B = (W,W',w),

s(W',V) os(V,W ') = idvw,

Hence

in

~X

AX(*(B),*(A)) = (V(~B),*(A)),W'V,n I) .

that

have isomorphism

squares

in the previous

poq = idv(,(B),,(A) ) .

(*(B),*(A))-

But recall

and

that

be two objects

A = (V,V',v),

A--x(A'B) ~ ~ X

such

Hence

the proof.

PROOF.

implies

V(*(B),*(A))

to conclude

Corollary.

then

also has this property.

Now switch
apply

idv(A,B)

and

= (V(B,*(A)),WV,n 2)
and

V(B,*(A))

make

the following

125

V(B,*(C))

pl

~ (W,U')

V(B,*(A))

pl

> (W,V')

]P2
(u,w')

(v,w')

, (weu,x)

]v~s
wos

> (w~v, x)

Now consider Figure 7. Since (U,-) and (V,-) have left

adjoints, they preserve pullbacks, hence the outer and inner squares are
still commutative.
But (i) is a pullback, hence the following diagrams commute:
(v, (u,w'))

pos

> (u,(v,w'))

(v, (wu, x))

(u, (wov, x))

~ (w(w~u),x)

(u, (w,v'))

p-1

(w (weu), x)

(w~u,v)

(u (wv) ,x)

\/

~ (ue(w~v),x)

(u, (wev,x))

pl

V(A,Bc)
pl

V(C Ba)

> (U,Ba)(id~pI~(u,(w,v'))
(id,p2)

,2

(I)

(U, (V,W')) (id,~)

(WV,U')

(V,V(B,* (C)))

w.uv)

(id,9)

(u, (w~v,x))

, (u(wv),x)

(v,(u,w'))

(S~

(id,~)

(V, (W,U'))

(id,~)

(v,(wu,x))

FIGURE 7.

-i
(w(w~u),x)

126

This implies

that there exists

that the diagram of

Figure

a unique map
still

V(C, Ba)

pl

a pullback,

there

A similar
p:V(C,Ba)

argument

If

map

8) shows

> V(C,Ba)

the existence

as in the previous

and

of a map

A, B, C are objects

PROOF.

Apply

the same argument

Corollary.

Let A, B, C be objects
_Ax(A,Ax(B,C))
_~(A,Ax(B,C))

proposition,

qop = idv(c,Ba)
in _~

Ax(A,Ax(B,*(C)))

PROOF.

V(A,Bc)

the same argument

that poq = idv(A,Bc )

Corollary.

such

(u(wv),x)
a unique

(Figure

> V(A,Bc)

Applying
conclude

exists

~ (U,Ba)

Now using the fact that

(U,Ba)

(WOV,U')
is

V(A,Bc)

commutes.

, then

~Ax(C,_~(B*(A)))

as in previous
in

we

corollary.

A X , then

~Ax(*(C),Ax(B,*(A)))

~ AX(*(*(A)),Ax(B,*(*(C))))
~ AX(*(C),Ax(B,*(A)))

Remark.

These propositions

the foundation

3.

The Functor

Note:

Henceforth

Definition.

and corollaries

of our construction,

the duality

lay

-ewe write,

Given

It is clear

concerning

as we shall see later on.

that

A,B

for an object

objects

' then define

--

in

~X

is a bifunctor,

of

~X

since

'

A* instead

of

*(A).

AB = Ax(A,B*)*

--

is the composition
AX(-,-)

~X ~X

(*'*)

+ &

AX

(id,*)

~ _~

~X

> ~X

127

p1

_V(C,Ba)

>(WoV, U'

/
pl

V(A Bc)

(V,V(B,*(C))) (id,pl) (V,(W,U'))

(id,p2) 1

~2

p2

(V,(U,W'))

(id,~)

/
(WU,V')

(U,Ba)

(id,

(id,~)

p.b.

(v,(wu,x))
p-i
(v(wu),x)

/pos

(s,id)~

, (U, (V,W'))

(u,(w,v')) ~.~-v-rr-=v+~(u,(wv,x))

tlO,v)

-i

FIGURE 8.
Proposition.

Let A,B be objects in _~ , then

AB m BA.

PROOF.
Proposition.

PROOF.

AB = Ax(A,B*)* m Ax(B,A*)* = BA.

Let A,B,C

be objects in _AX , then

(AB) C ~ A(B~C)
(AB) C = Ax(A, B*)*C
= Ax(Ax(A, B*)*, C*)*
Ax(C,Ax(A, B*) )*
Ax(C,_~(B,A*) )*
Ax(A,_Ax(B,C*) )*
Ax(Ax(B, C*)*, A*)*
Ax(A,Ax(B,C*)**)*
= AAx(B, C*)*
= A(B~C).

>(u(wv),x)

128

4.

The Dualisin$
Let

is the cannonical
Claim.

Object

and the Unit

T = (X,l,r)
isomorphism

in

is the dualising

Let

following

A = (V,V',v)

commutative

in

~X

r:Xl

' such that

V.
object,

_~(A,T)
PROOF.

for Tensor.

be the object

i.e.

for any object

in

_AX.

m A

be an object

_~

in

, then we have the

diagram

V(A,T)_

pl

~ (V,X)

pl

Ir

(1,v')

, (vI,X)

1p
(V,(I,X))

V'

(id,i)

(V,X)
But
V'

~ (V,X)

V'
is trivially
(unique)

a pullback

morphism

unique map

g:V'

Corollary.

PROOF.

Suppose

in

V,

Apply

such that

~ V(A,T)
is the identity
is

implies

V'.

f:V(A,T)

which

(V,X)

for

an object

in

AX ,

m A x ( A ,T)
~A
=A

fog = idv,

--

T A = __~(T ,A )

that we have an induced

the same argument

then

and

to get a

gof = id.(A,T)~

129

On the other hand,

Theorem.

Let

AoT

~ T A ~ A.

This completes

A,B,C be objects in

PROOF.

_~,

the proof.

then

_~(AB,C)

-~ A x ( A , ~ ( B , C ) )

Ax(AB,C)

= Ax(Ax(A,B

) ,C)

Ax(C

,Ax(A,B ))

-~ Ax(C

,Ax(B,A ))

_~(A,Ax(B,C)) .
Proposition.

Let

be an object in

_~

then

Ax(T ,A) ~- A.
PROOF.

Ax(T

,A) ~ _ ~ ( A

,T) ~ A.

Remark.
(i)

There is an obvious embedding

(V,X)

to

context
(2)

_AX
(V,X)

sending

~ X

has a *-autonomous

It is easy to verify

the MacLane-Kelly

_AX

VI

> X:

hence in this

structure.

also satisfies our first assumption,

APPLICATIONS

Functor Categories
In this chapter, we shall apply the theory developed

far to the double envelope of a symmetric monoidal

Before defining the double envelope,
elementary

We know that if

and

X is complete,

X = S , the category of sets

structure.

category

thus

C.

let us recall some

results of the functor categories.

Given categories

W = X~

i.e.

coherence conditions.

CHAPTER IV:

i.

functor from the comma category

to

we have the functor category

then so is

W,

in the case

also has a closed symmetric monoidal

The tensor is the cartesian product while the internal

is defined as the functor whose value at

transformations

F(-) IIom(D,-)

--+ G(-).

is the set of nature

GF

130

2.

The Double Envelope.

Definition.

Given a symmetric

I-I:C
objects

> S , we denote
of

E(C)

from ~o

to

morphism

from

are all triplets

S,t

is a natural

(F,G;t)

is a natural

transformation

monoidal

to

G'

with a faithful

(F,G;t)

(F',G';s)

to

in

from
G

where

from

E(~)

to

by
and

~ F(C)

and

The

are functors

to

is a pair

F'

functor

E(C).

F x G

I-~-I.

(f,g)

where

is a natural

such that the following

idxg

F(C) x G'(C')

of

transformation

transformation
from

category

the double envelope

diagram

x G(C')

fxid

F'(C)
commutes

x G'(C')

for every object

Proposition.
PROOF.

E(C)

(C,C')

in

E(C)
x CO .

of

C x C .

is a category.

Suppose

(f,g):

(F,G;t)

(f',g'):

cO

~Ic~c' I

, then the following

F(C)XG"(C')

fXid

> (F',G';s)

(F',G';s)
diagram

commutes

) F'(C)XG"(C')

idXg'

are maps

> (F",G";u)

(C,C')

for every

f'Xid

) F"(C)XG"(C')

idXg'

F(C)XG'(C')

fXid

in

>F'(C)XG'(C')

idXg
F(C)XG(C')
This implies

are maps

in

that

E(C)

) C~C'

(f,g) : (F,G;t)

) (F' ,G' ;s)

(f',g') : (F' ,G' ;s)

~ (F", G" ;u)

(f",g"):(F",G";u)

> (F'",G'";v)

, then

(f",G")o((f',g')o(f,g))

= (f",g")o(f'of,gog')
= (f" o (f' o f), (gog') og")
= ((f' of') of,go (g' og"))
= (f"of', g' og") o (f, g)
((f", g") o(f' ,g')) o (f, g).

Moreover,

given

(F,g;t)

then

(idF,id G) is the obvious

choice

for identity.

131

Before p r o v i n g the m a i n t h e o r e m of this chapter

O

v e s t i g a t e the functor categories

~o

co

two obvious embeddings of

and

into

--

where

S~
--

let us in~

T h e r e are

, namely

--

and r,
O

~(F) = F x I

and r(F) = I x F for every

F in S~
, and
C
-I is the unit in S-i.e.
I sends every object into the singleton
-C
(the terminal object) in S. H e n c e w e can regard objects in S-- as
C CO
objects in S=
-v i a either embedding.
N o w we can prove.
-O
O
C
C
Proposition.
E(C)
is e n r i c h e d over V = S--- .

PROOF.

By previous r e m a r k

is a closed symmetric m o n o i d a l

c a t e g o r y w i t h pullbacks, m o r e o v e r it is coherent.
N o w given

A = (G,F;t)

V(A,B)

an object in

and

B = (G',F';s)
O

define

--

of

object

S--

--

, then

V(A,B)

in E(C) w e have to

V(= S~C

~ ) .

Suppose

(C,C')

is an

is the functor w h o s e value at (C,C')

is defined by requiring that the d i a g r a m

V(A,B)(C,C')

pl

~ (~(G),%(G')(C,C')

(r(F'),r(F))(C,C')

~ (G x F',

I-~-I)(C,C')

be a pullback.
o
Note.

(-,-)

denotes the internal h o m - f u n c t o r of

map (~(G),~(G'))(C,C')
that in

V,

p r o p e r t y of

G x F'
V

> (G x F',

is isomorphic to

S~

cO
x

As for the

I-~-I)(C,C'),

w e simply o b s e r v e

~(G) r(F').

Then the adjoint

constructs such a map (in the same fashion as in Chapter

II, Section 2.)

A similar argument constructs

(r(F'),r(F))(C,C')

) (g x F,

map

I--I)(C,C').

N o w the enrichment follows i m m e d i a t e l y f r o m the result in Chapter

II, since this is h o w p u l l b a c k s are defined in the functor category, i.e.
b y p o i n t - w i s e evaluation.
Theorem.
A

E(~)

is e n r i c h e d over

PROOF.
3.

Put

This concludes the proof.

is a s u b c a t e g o r y of a * - a u t o n o m o u s category

A; m o r e o v e r

V.

X =

I--I

, then follow the c o n s t r u c t i o n in Chapter III.

M i s c e l l a n e o u s Results.
In this section, we are a s s u m i n g

has all the p r o p e r t i e s as given

in Chapter I and we shall prove that there is a functor

- C A T ( V - CAT

F maps

to

is the c a t e g o r y of all categories w h i c h are e n r i c h e d over

X)The functor
put

F(X) = ~ .

on objects of

is obvious:

given

in

V,

then

132

Now we have to show given a map

a

V-functor T( = F(f))
The notion of a

II, Section 6.
a function

(ii)

for each

(i)

_~

to

> S

in

V, this induces

AS.

V-functor can be found in [Eilenberg & Kelly] Chapter

In this case we have to show:

(i)

in

from

f:X

maps objects of

B,C

in

_~

to objects of

AX, a morphism

T(B,C) maps

AS.

V(B,C) to

V(T(B),T(C))

such that the following axioms are satisfied:

The following diagram commutes:
T

V(B,B)

, V(T(B) ,T(B))

(2)

The following diagram cormnutes:

M v

V(C,D)V(B,C) .

_V(B,D)

I,l t

V(T(C),T(D))V(T(B),T(C))

, _V(T(B),T(D))

Note. In both categories we denote the enriched object by

V(-,-)

it

is clear from the context which one we are referring to.

The function

on objects of

_Ax, then T(B) is the composition

_~ is obvious;

VV'

given

B =

, S

i.e.

objects in

AX,

then

~ X

(V,V',v)

T(B) = (V,V',fov).
To show (ii):
Suppose

B = (V,V',v)

C = (W,W',w)

and the following diagram commutes

T(B) = (V,V',fov), T(C) = (W,W',fow)

pl

V(B,C)

V(T(B) ,T(C))

, (v,w)

pl

p.b.

p2

(W' ,V' )

fov

, (V,W)

f~w

~(V~W',S)

( i d , f ~

(W',V')

(V~W',X)

in

133

Since the inner square is a pullback, there exists a (unique) map

T(B,C)

from

V(B,C)

to V(T(B),T(C)).

TO show (i) commutes let B = (V,V',v)

and the following diagrams commute:

in

_AX.

Then

T(B) = (V,V',fov)

i3
V(T(B), T(B))
p2

P'I"

(v',v')

(V' ,V')

~(wv',x)

--,

pl

V(T(B),T(B))

lp2
(V',V')

pl

(V,V)

(V,V)

p.b.
fov

> (v~v' ,x)

(id,f~~
(V',V')

(VV' ,X)

(v' ,v)

(v' ,v')

> (V,V)

[~

V(B,B)

p2

,(wv' ,s)

+ (wv',x)

for , (VV' ,S)

134

Hence the composition

map

> V(T(B),T(B))

> V(B,B)

> V(T(B),T(B))

are both induced by pulling back.

and

Thus by

the uniqueness property they are "equal".

To show

(2)

commutes, let

B = (V,V',v),

C = (W,W',w),

D = (U,U',u)

be three objects in _AX . Then T(B) = (V,V',fov), T(C) = (W,W',fow),

T(D) = (U,U',fou) and the following four diagrams commute:
V(C,D) V(B, C)

pl~pl

(U' ,W')(W' ,V')

V(B,D)

Is

~ (W,U)(V,W)

pl

p2[

(w',v')(u',w')

p.b.

[fi

~ (u',v')

V(T(C),T(D))V_(T(B),T(C))

(U' ,W')(W' ,V')

~ (wu',x)

plpl

> (W,U)(V,W)

V(T(B) ,T(D))

pl

, (V,U)

p.b.

(W' ,V')(U' ,W')

,(U' , V ' )

fov

~(VU',S)

pl

V(C,D)
V(T(C),T(D))
p2

,(V,U)

p21

(U',W')

(w,u)
pl

~(W,U)

f~u

fow

~(WU',S)
( i d , ~

(U',W')

> (WU',X)

135

pl

V(B,C)

(V,W)

~a

\
V(T (B) ,T(C))

~2

pl

p2

(w' ,V')

,(v,w)

f~w

p .b.

fov

~(VW',S)

(id,f)~
(VW',X)

(w' ,V')

(w,u)~(v,w)

V(C,D) V(B, C)

V(T (C) ,T (D)) V(T (B), T (C))

~2p2

pl~l

(w,u)~(v,w)

p2~2 ~'~V(T(B),T(D))

pl> (V,U)

(v,u)

p21 pb Ifou
(w' ,v')~(u' ,w')

(w' ,v')~(u' ,w')

+(u' ,v')

, (U' ,V')

fov

~(wu' ,s)

(VU',X)

136

plp1

V(C,D) V(B,C)

(w,u)(v,w)

h
~

~"~V(B,D)

plepl

~ (V,U)

V(T(B),T(D))

pl + (V,U)

p2ep2

~ U',V')
M

f~(VU'

/~d
(u', v')

(w',v')(u',w')

,S)

(id'f ) ~
, (vu',x)

This implies that the diagrams above commute, which implies

that the composition
~(C,D)(V(B,C))

T~

M'

> V(T(C),T(D))V(T(B),T(C))

--+ V(T(B),T(D))

is the map induced by pulling back.

This also implies that the composition
V(C,D) V(B,C)

M'
'

V(B,D)

+ V(T(B),T(D))

is the map induced by pulling back.

Hence by the uniqueness property, they are "equal", therefore (2)
commutes.
Now we are left to show that if
maps in

V, then

F(g) oF(f)

f:X

F(gof), i.e.

> S
F

and

g:S

>K

are

preserves composition.

All we have to cheek is that the composition is preserved in (i)

and (ii).
It is easy to show (i) is preserved.

For if

B = (V,V',v) in

_~ ,

then
(F(g) oF(f))(B) = F(g)(F(f)(B)) = F(g)(V,V',fov)
= (V,V',go(fov))
= (V,V',(gof) ov)
= F(gof)(B).
To show (ii) is preserved: Let B = (V,V',v),

C = (W,W',w)

F(f)(B) = (V,V',fov),F(f)(C) = (W,W',fow),(F(g)oF(f))(B)

in

~X' then

= F(gof)(B)

= (V,V',(gof) ov) ,F(gof)(C) = (F(g) oF(f))(C) = (W,W',(gof) ow)

and the diagrams (*), (**) and (***) commute

137

p1

V(B ,c)

+ (v,w)

~(F(f)(B),F(f)(C))

(*)

[p2

~2

(W',V')

pl

~P'b"
fov

(v,w)

f~w

~(VW'.S)
(id,f)~

(VW',X)

(W',V')

V(F(f)(B),F(f)(C))

pl

(v,w)

~
X(F(gof)(B),F(gof) (c)) pl >(v,w)

(**)

f ow

(W',V')

(gf)
f~v

(VW ',K)
(id,gof)~

(w' v')
Note.

>(VW',S)

F(gof) (-)

(F(g) oF(f)) (-).

pl

X(s,c)

+ (V,W)

V(F(gof)(B),F(gof)(C))

p2
(W',V')

p.b.
($of)ov

,(V,W)

(gof)ow ~
, (WW',K)

(VW',X)

(W',V')

But (*) and (**) imply the diagram of Figure 9 connnutes.

This implies that both F(gof) in (***) and the composition
V(B,C)

F(f)

V(F(f)(B),F(f)(C))

are induced by pulling back.

F(g)

V(F(gof)(B),F(gof)(C))

Hence it follows they are equal.

138

pl

V(B,C)

(v,w)
pl

V(F(f) (B) ,F(f) (C))

~ (v,w)

)
V(F~of)(B),F(gof)(C))

p2

pl ~ (V,W)

[p2

fow

(gof)ov ,(VW',K)

(W' ,V')

fov

(W',V')
..~

(i

(VW' ,S)
(id,f)~

(w',v')

~ (ww' ,x)

FIGURE

9.

BIBLIOGRAPHY
i.

M. BARR, Duality of vector spaces, Cahiers Topologie G~om~trie

Diff~rentielle, XVII-I (1976), 3-14.

2.

M. BARR, Duality of Banach spaces, Ibid., 15-32.

3.

M. BARR, Closed categories and topological vector spaces, Ibid.

XVII-3, 223-234.

4.

M. BARR, Closed categories and Banach spaces, Ibid. XVII-4, 335-342.

5.

M. BARR, A closed category of reflexive topological abelian groups,

Ibid. XVIII-3, 221-248.

6.

M. BARR, *-autonomous categories,

7.

S. EILENBERG, G.M. KELLY, Closed categories, Proc. Conf. Categorical

8.

M.E. SZABO, Commutativity in closed categories.

This volume.

Alg. (La Jolla, 1965), Springer (1966), 421-562.

To appear.

Index of Definitions

Admissible
Autonomous
Basis

(uniform object)
(category)
13

(for a pseudometric)

Completable
9
Convergence uniformity
Cosmall
24

Semi-norm
65
Semi-variety
3
Separated uniform
*-autonomous
13
Uniform
Uniform
Uniform
Uniform

8
8

Linearly compact
18
Linearly totally bounded
complete
Nuclear

Refine (of seminorms)

Reflexive
23
Represent
28

33

47

Pre-reflexive
23
Pre-*-autonomous
15
Pre-uniform structure
Product uniformity
i0
Pseudomap
5
Pseudometric
7

71

28

Dominating
29
Double envelope
130
Dualizing module 49
Embedding
Entourage

Quasi-reflexive
23
Quasi-variety
3

20

convergence
cover 6
object
I0
space 6

on

28

Variety
3
V-enriched pre-*- autonomous
V-enriched *-autonomous
17
~-complete
34
~-*-complete
35

17

Index of Notation

_v(-,-)

<--~-->

I-I

<-->

34

(_)~

37

UnV

i0

[-,-]

ii

33

()A

33

33

37
43
o

13

44

(-)

13

52

(-). (-)

18

h(1)h(2)
A

59

C~D

19

(-)

74

19

_Ax

104

A(-,-)

25

E(C)

130

28