Professional Documents
Culture Documents
Editors:
A. Dold, Heidelberg
B. Eckmann, Ztirich
E Takens, Groningen
Wojciech Banaszczyk
Additive Subgroups
of Topological Vector
Spaces
Springer-Verlag
Berlin Heidelberg NewYork
London Paris Tokyo
Hong Kong Barcelona
Budapest
Author
Wojciech Banaszczyk
Institute of Mathematics
L6d~ University
Banacha 22
90-238 L6d~, Poland
This work is subject to copyright. All rights are reserved, whether the whole or part of
the material is concerned, specifically the rights of translation, reprinting, re-use of
illustrations, recitation, broadcasting, reproduction on microfilms or in other ways, and
storage in data banks. Duplication of this publication or parts thereof is only permitted
under the provisions of the German Copyright Law of September 9, 1965, in its current
version, and a copyright fee must always be paid. Violations fall under the prosecution
act of the German Copyright Law.
© Springer-Verlag Berlin Heidelberg 199 I
Printed in Germany
Printing and binding: Druckhaus Beltz, Hemsbach/Bergstr.
2146/3140-543210 - Printed on acid-free paper
I~REFACE
Preface
Chapter 1, PRELIMINARIES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . i
1. Topo]ogical groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . i
2. Vector spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3. Geometry of numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
Index of symbo]s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
Subject index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
Chapter 1
~ E S
i. T o p o l o g i c a l groups
The set
{X • G : Ix(A) I ~ }
is c a l l e d the p o l a r of A in G - ; we d e n o t e it by o
A G. If the m e a n -
A° = {X • G^ : XIA ~ 0};
1
N {g • G : l×(g) l ~ ~}
X~A °
A m = {a I + ... + am : a l , . . . , a m E A}.
W': = (K + A °) n U ° c W.
(i) A c { g i } ~ = 1 + V.
The set
W" = {X E U O : Ix(gi) i
- <(gi )I ~ i-~ for i = i ' . "" ,n}
commutative diagram:
G- ~ >G-/A
The formula
p(a,f) = a + a(f) (a ~ A, f ~ F)
(I) p(A x F) = G .
rectangular one.
Thus ~ V n c ~* U . •
n= 1 n= 1 n
U1 + U2 + ... := U (U 1 + ... + U n)
n=l
where Un ~ N o ( G n) for e v e r y n. It f o l l o w s directly from (i.15) that
B is a b a s e at zero in G.
Fix an i n d e x m and c h o o s e an a r b i t r a r y V ~ No(Gm). We are to
find some U ~ No(G) with U N G m c V. Naturally, we m a y assume that
m = I. There is some U 1 E No(G1) with U 1 + U 1 c V, and a simple
inductive argument allows us to find, for e a c h n ~ 2, some Un,W n E
cally q u a s i - c o n v e x group.
Let G,H be abelian t o p o l o g i c a l groups and let # : G ~ H be a
continuous homomorphism. Then the formula
<~(x),g> = <X,#(g)> (X E H ; g ~ G)
~KL : II G i x II H i ~ II G i × II H i
ieK i~K ieL i~L
((ui),(vi)) = ~ (ui,vi).
ieI
¢(g).(ui)iE I = (¢i(g).ui)i~ I
el(g) .u i .¢(g).u. I . (g ~ G; ul ~ Hi )
n n
5q. 5q li~ ~(g~igj) > 0
i=l j =i J
(a) ¢(0) ~ 0;
For the proof, see [38], (32.8) (a), (b) and (f).
2. V e c t o r spaces
1
sup {If(u) : u e U} = Ix(U)[ < ~.
,U+W, U (~
X~--m--~ = x(~) + x ),
1 1 1
i.e. ±f(Um + w) = m±f(u) + ~(w), which proves that f is an a d d i t i v e
mapping. Now, take any scalar I ~ R. Let [x] and {x} denote the
integer and the fractional parts of x, respectively. Since f is
additive, for e a c h positive integer p, we have
1 1
f(lu) = f(~plu) = ~f(plu)
al i s o m o r p h i s m of Ec onto Ec.
The set
is c o m p a c t , so t h a t H° ~ No(E~)'-v It r e m a i n s to p r o v e that P E ( K 0) D
HO .
SO, take any X ~ H°. There is s o m e f ~ E* with pE(f) = X.
Due to (2.2), we have
f
Set h = 4-~" Since U is s y m m e t r i c , we have
Then ph E E and
1
sup {Iph(u) l : u ~ U} < ~ < Iph(v) l,
AO = {u E E : f(u) E 1 1
[-~,~] + Z for all f ~ A},
1 1
AO = u ~ E : f(u) ~ [-~,~] for all f e A}.
dk(X,Y;E) = inf d ( X , Y ; L )
L
where the infimum is t a k e n over all linear subspaces L of E with
d i m L < k. If E is c o n t a i n e d in s o m e other vector space F, then
dk(X,Y;F) = dk(X,Y;E). Therefore we may simply write dk(X,Y) in-
stead of dk(X,Y;E). The numbers dk(X,Y) are called Kolmo~orov
diameters of X with respect to Y. We take the following natural
convention: for t > 0, the expression "dk(X,Y) < t" should be read
as "X - < Y and dk(X,Y) < t".
dk+l_l(X,Z) ~ dk(X,Y)dl(Y,Z).
d l + l ( X n , Y n) ~ 2 - m n ( l + i) -m < 2 - m n ( k 2 - n ) -m = k -m.
dk(~ : E ~ F) = d k ( ~ ( B E ) , B F ) -
for all u,w E E. This holds if and only if the norm on satisfies E
P
the p a r a l l e l o g r a m identity, i.e. if and only if ~D is a u n i t a r y space.
The following lemma is a direct c o n s e q u e n c e of our definitions.
id
E > E
Ep Pq > Eq.
= dk(Bp,Bq). •
Bq
0 = {f ~ E # : If(u)l & 1 for all u ~ Bq}.
Then dk(B0,B0)q
p = dk(B p,Bq) for every k.
P = dk(F-l( B0%F-I,B0,.
dk(B0,B0)q q,, ~ pJ)
3. Geometry of numbers
d(L*) = 2 -n v o l n
> (rB n) > r n n-n/2
: 2-nrn~ n =
O n the o t h e r hand, from the Minkowski theorem and the assumption that
L N D : {0} we get
27
diam
n 2 1/2 n 2)1/2.
!2 d i a m P = (k=l
~ sk) = (k:iZ H¢*ektl
(i) M n B # 0.
n
Let us s u p p o s e that
28
-2 -2 < 1
(3) N1 + "'" + D n - i "
(4) ~i = dk(Bn,D) (k = l , . . . , n ) ,
-i
(5) nk = dk(B n N Mo,(D n M) - u) (k = l , . . . , n - l ) .
2<1.
(6) ~ + "'" + ~n "
(7) ~(p) c (B n n N) - w.
n-i
X d ~n ( ~ (~E ) , ( B A N) - w) < i.
k=l
[0,i] and let E be the set of all those u ~ M for w h i c h the inter-
section of D and the straight line passing through u and perpen-
dicular to M is a s e g m e n t with length not less than 2r. Then E is
R n-I with Ln_ 1 n Dn_ 1 = {0}. Let D1 ~ "'" & nn-i be the p r i n c i p a l
n k2 -2/k]i/2
(1) llfll S 1 + [ Z (El ... ~k )
k=l
k -I ~k )I/k
d(Uk, span {ui}i< k) (~I "'"
la n - P n C n n I ~ {Cnn.
1
lan_l - PnCn,n_l - Pn_iCn_l,n_l i --< ~ C n _ l , n _ I.
n
lak_ 1 - PiCi,k_l I <= l2 c k - l , k - l '
i=k-i
a n W n. Then, clearly,
1 3
(6) a{f I + ... + anf n ~ [~,~].
n 1
a{f I + 7 akh k g [~,-~].
k=2
n
1 3
qfl + "'" + a{fk + Z aih i E
[~,~] -
i=k+l
1 3
From (6) we get f(a') E [~,~], so t h a t
1 3
f(a) = f(a" + a O) = f(a') + f(a O) E [~,~] + Z.
n 211/2
llfll ~ {lhll + llf - hll ~ llhll + [ ~ If k - hkl
k=l
n
~ 2 1/2 n k2 -2/k~I/2
1 + [ X Ckk~ 1 + [ ~ (~I " .. ~k )
k=l k=l
principal semiaxes ql' .... ~n-i such that q~2 + " " + q n --2
i =< 1 " There-
n
(1) Z k ~ .I- < I.
2%
k=l
Proof. We m a y a s s u m e that
n
D = { ( X l , . . . , x n) ~ R n : k=l~k~
"-2X2k "
< I}.
n
E = {(Xl,...,Xn) g Rn : ~ nk-2x2k ~ i}
k=l
n
From (i) w e get ~ S I. Thus, by (3.9), there exists an ortho-
k=l
gonal projection ~ : Rn ~ Rn with m : = d i m ~ ( R n) = l , . . . , n , such
that ~(K) 0 2E = {0} and (n(a) + E) n ~(K) = ~. Hence, by (i),
d(~(a),~(K)) > nl
: i~ /2 > 1 > 4"
i
m
llhll < 1 + [ ~ k2(~l ... ~ k ) - 2 / k ] I/2
k=l
1 m
ilfll = llh~ll < iihll.iIwit = llhli < 1 + ~ [ )q (n I ... q k ) - 2 / k ] I/2
k=l
1 e 3 / 2 [ m k21]k2]i/2 1 m kl]i/2
<1+ 5 ~ = 1 + [ }= k~
k=l k=l
< 1 + 1 e3/2 L
r ~=
n k~ k l ] i / 2 < 1 + 1 e3/2 < 4. •
k=l
34
(1) q12 + . . . 1
+ q2 < 4"
1
(2) Ix(K n Bn) I < ~.
1
(3) IX'(K" n ~Bn) I < I×(K n Bn) I.
1 1 1 1
IX(K N ~Bn) n < ~ I x ( K N Bn) I < ~ < ~.
1
(4) I~(K A B°)In --< I×(K N Bn) I <
A 1 = {u E B °n : u + W l , U - w I E B~}
In v i r t u e of (4) a n d (i.I), we h a v e
J~(K A BO)J.
n
A 2 = {u E A 1 : u + w2, u - w 2 ~ AI}.
Hence, by (i),
ImmDaf. We m a y a s s u m e K to be a l a t t i c e in R n. According to
(3.7), w e c a n f i n d a b a s i s Ul, .... u n of K such that
k-i 1 ~k)i/k
d(Uk, span {ui)i=l) Z k- (~I "'"
u k = a k l e I + ... + a k k e k (k = l , . . . , n )
37
n n n
)ifl12 = ~ f~ ~ ~ (2akk)-2 < l k2(~l -2/k
k=l k=l " ~ k=iX ... ~k )
= ! ~n k2(12~2.22~2 . . . . k2~2)i/k(k!)_2/k
4 k=l
n 1
(1) 7. k ~k : i.
k=l
Next, we may assume that
n 2 -2 ~ i}.
D = {(Xl,...,x n) E R n : ~ Xk~ k
k=l
Define
n
E = {(Xl,...,x n) g R n : Z k-lx2~k I < 2}
k=l
7 d (E,D) = %2
n21
3q k~ < ~.
k=l k=l
1/2 1/2
(4) ~k ~ %k ~k (k = i .... ,r).
m m
g p ( { U k } k = I) = L N s p a n { U k } k = I.
m
L e t us w r i t e M = span { U k } k = I. If L N (U + M) c M, then
L = gp (L N U) c gp (L N (U + M)) c gp (M) = M,
Set
It is n o t h a r d to see t h a t {u + P } u E L is a d i s j o i n t covering of R n.
From (i) we get p c ½ B n. •
(2) K = gp (K N Bp).
Then s p a n K c K + iBq.
n
~q d 2 ( B p N M,Bq N M) < i.
k=l
(i) K + Bn = Rn .
41
and, consequently,
!
4 < liwll - l~fll-l'f(v) <z llw - vll<" !4 '
which is i m p o s s i b l e . •
n ~(Bp,Bq) 1
(i) X d ~ 4"
k=l
(2) M N ½ Bq c c o n v (K n Bq).
The quantity
[~(L,U)] n ~ d(L)/VOln(U).
< 4 -2/n 2n
I I ( L * , B n ) I I ( L , B n)
1 2 ( L n , B n ) > (5 n l ) 2 / n - 1 = n
2 2~e
for each lattice L in Rn; here c and cI are some universal con-
stants. On the other hand, from Siegel's mean value theorem one can de-
duce that there exists a universal constant c2 such that to each sym-
metric convex body U in Rn there c o r r e s p o n d s a lattice L with
kI(L,U)XI(L*,U0) ~ c2n;
½ 3/2 [ ~n k 2 ~ 2 ] 1/2
llfll ~ -e
k=l
Then, by (I),
n
~IT*,Bn)klCL,D) S ~3/2 [ X k2~ 211/2.
k=l
n 1/2
~ ( L * , B n ) X I ( L , B n) ~ ½e 3/2 [ Z k 2] < 2.25n 3/2
k=l
~ ( L * , B n ) X I ( L , B n) & ½n 3/2
44
n
X i ( L , D ) X n _ i + I ( L * , B n) & 6 ~ k ~i (i = l,...,n).
k=l
X i ( L , B n ) X n _ i + I ( L * , B n) ~ 3n(n + I) (i = l,...,n).
Xi(L,Bn)Xn_i+I(L*,Bn) ~ ~n 2 (i = l,...,n)
EKOTICGRO~PS
2~i@
defines an Ls(X,~)-representation ¢ of E; we write ¢ = e
1 3 2
X({t ~ (0,I) : tf(x) E [~,~] + Z}) _>- ~.
1
1 3
f ~({x ~ X : tf(x) ~ [~,~] + Z})dt
0
This implies that the function under the first integral must assume a
value not less than ~(X) for a certain t E 0,i). •
f If(x)12d~(x) < e.
Y
47
Then
l]¢wf - ¢vfll
= f + f ]exp [2~i(0w)(x)3
Y X\Y
- exp [2~i(Sv)(x)]]2.]f(x)]2d~(x)
< 4E + 4~2E2]Ifll 2.
Proof. Let ~ be a n o n - t r i v i a l c o n t i n u o u s u n i t a r y r e p r e s e n t a t i o n
of E/K in a H i l b e r t space H. In v i e w of (1.21), we m a y assume # to
be cyclic. This implies that H is separable. Let A be the algebra
of o p e r a t o r s in H g e n e r a t e d by o p e r a t o r s ~g, g ~ E/K. Then A is
an a b e l i a n s e l f - a d j o i n t algebra in H c o n t a i n i n g the identity operator.
The c l o s u r e A of A in the strong o p e r a t o r t o p o l o g y is an abelian
yon N e u m a n n algebra in H. T h e r e f o r e we can d e c o m p o s e H into an at
most c o u n t a b l e H i l b e r t sum of A-invariant subspaces Hn such that,
for each n, either dim H n = 1 or the r e s t r i c t i o n of A to Hn is
u n i t a r i l y i s o m o r p h i c to the a l g e b r a LC(0,1) in the Hilbert space
49
9
L~(0,1). The last s e n t e n c e follows from the s t a n d a r d results on the
s t r u c t u r e of a b e l i a n von N e u m a n n algebras; perhaps the best r e f e r e n c e
here will be [47], S e c t i o n 9.4.
The s u b s p a c e s Hn, b e i n g i n v a r i a n t for A, are i n v a r i a n t for ~.
So, ~ can be d e c o m p o s e d into a H i l b e r t sum of some r e p r e s e n t a t i o n s
~n w h i c h are either o n e - d i m e n s i o n a l or unitarilyequivalent to LS(0,1)-
-representations. One of these r e p r e s e n t a t i o n s is non-trivial, there-
fore we m a y assume that ~ itself is o n e - d i m e n s i o n a l or an LS(0,1)-
-representation. In the first case, E/K admits a non-trivial con-
tinuous character; let us d e n o t e it by ×. Let ~ : E ~ E/K be the
natural projection. Then X~ e E . By (2.3), we have X~ = Pf for
some f E E . Since X~ # 0 and x~(K) = {0}, it follows that f # 0
and f(K) c Z. Now, we m a y d e f i n e 8 by the formula (0u)(x) = f(u)
for u ~ E and x ~ (0,i).
It remains to c o n s i d e r the case w h e n ¢ is an LS(0,1) - repre-
sentation. Then ~ is an LS(0,1) - r e p r e s e n t a t i o n of E and, by
(4.4), there exists a c o n t i n u o u s linear o p e r a t o r 8 : E + L~(0,1) with
e 2~i8 = ~. It is clear that 8 # 0 and 0
8(K) c LZ(0,1 ). •
A n o n - t r i v i a l H a u s d o r f f a b e l i a n g r o u p is c a l l e d exotic if it does
not admit any n o n - t r i v i a l c o n t i n u o u s u n i t a r y r e p r e s e n t a t i o n s . We say
that G is s t r o n g l y exotic if it does not admit any w e a k l y c o n t i n u o u s
r e p r e s e n t a t i o n in H i l b e r t spaces.
In c o n n e c t i o n w i t h the above definition, let us notice that e v e r y
topological group (abelian or not) admits a f a i t h f u l s t r o n g l y c o n t i n u -
ous r e p r e s e n t a t i o n by b o u n d e d o p e r a t o r s in a suitably chosen Banach
space. It suffices, for instance, to c o n s i d e r the r e p r e s e n t a t i o n by
shift o p e r a t o r s in the space of b o u n d e d and u n i f o r m l y c o n t i n u o u s func-
tions on the group.
An a b e l i a n t o p o l o g i c a l g r o u p G is said to be b o u n d e d if to each
U E No(G ) there c o r r e s p o n d a p o s i t i v e integer n and a finite subset
A of G, such that A + Un = G (this d e f i n i t i o n m a k e s sense also for
n o n - a b e l i a n groups). For instance, all c o m p a c t or c o n n e c t e d groups are
bounded. If K is a s u b g r o u p of a n o r m e d space E, the quotient g r o u p
E/K is b o u n d e d if and o n l y if there exists a n u m b e r r > 0 such that
E = K + rB (cf. the proof of (5.1) (b) below).
f
A > R
X
exp A - - > T
normal basis ( e n ) n = I. L e t us s e t
2
e n = e 2 + en for n = 3,4,...
5. Q u o t i e n t s of n o r m e d spaces
co
(see (2.1)). Suppose that {an}n= 1 is a dense subset of 11 and let
K = gp {a n + an}n= I. Then
(b) the groups Co/K and IP/K, 1 < p < =, are bounded;
u = kl(a I + e I ) + ... + k m ( a m + e m)
(2) K + 2B E = E.
u ~ (a n + e n) - e n + B E c K + B E + B E = K + 2BE,
= ~(E) = E/K.
23
Q = {t ~ (0,I) : (su)(t) ~ z + [~,~])
(Oakn)(t) ~ Z + [~,~]
54
for t ~ Q and n sufficiently large (say, for n > no). All func-
tions 0(a k + ek ) assume integer values only, therefore
n n
1 3
(0ekn)(t) E Z + [7,7]~
~ for t E Q and n > no •
1
(3) l@ek )(t) I ~ for t ~ Q and n > n .
O
n
f~(t) < ~ for almost all t E (0,i) (see [73], Lemma on p. 29).
n=l ~
On the other hand, (3) implies that Z f2(t) = ~ for t ~ Q.
n=l n
The contradiction obtained completes the proof of (d).
m
~. X n e n ~ u as m ~ ~.
n=l
m
m
W : ----- 7. X n e n ~ L : : gp {a n + e n } n = l ,
n=l
!
(4) Z x~ < 9"
n=m+l
m
Let M = span {a n + e n } n = I. Then L is a l a t t i c e in M and w ~ L,
(5) f(w) e Z + 12
[~,~].
n-i
(6) an = k=l
Z ~knek (n = 2 , 3 , . . . )
55
hl,h2,.., on (0,i) such that lhm+ll, lhm+21 .... < I, the functions
n-i
(7) hn + ~q ~ k n h k E LZ2(0,1)
k=l
c : = b + (b - a ) ( [ p ] - p) ~ (a,b)
and we define
l
ip] - p for t ~ (a,c)
hn(t) =
[p] - p + 1 for t ~ (c,b).
Here [p] denotes the integer part of p. Then lhn(t)l < 1 and
n-I
hn(t) + ~.. a k n h k ( t ) = hn(t ) + p e Z
k=l
= f(e n + a n ) e Z.
56
m m
@w = 8 n=l~ Xnen ~ n=l~ Xn f(en) = f(w) ~ Z + [~,~]i
2
It follows from (5.1) (c) and (f) that the group K o c c u r r i n g there
is w e a k l y dense in ip for p > i, but w e a k l y c l o s e d in 1 I. This
57
(i) f(ei) E ½ + Z.
It follows from (5.4) that the group K from (5.3) is not dually
embedded. Similarly, the group K from (5.1) is neither dually e m b e d d e d
in co nor in I p, p > i; it is clear, however, that K is d u a l l y
e m b e d d e d in 11 .
(5.6) B ~ K S . The fact that the groups K in (5. i) and (5.3) are
free is not accidental; Sidney [82], p. 983, p r o v e d that a countable and
58
quence in L0 If 0
0 : LR(0,1) ~ L 10,i) is a continuous linear operator
with 8{f n} = 0 for every n, then 8 = 0.
~ ~k(t)f(~k(t)) if t ~ X
=i
fF(t) =
if t ¢ X.
m
(2) ~k(t)f(~k(t)) = ~ #k(t)f(~k(t))
k=l k=l
m
(3) 5q %k(t)fn(~k(t) ) = 0 (n = 1,2 .... ).
k=l
such that ~kl(t) < "'" < ~k (t). T h e n we can find some coefficients
r
El' .... ~r such that
m r
(4) k=iX ~k(t)f(~k(t)) = j=IZ ~jf(~kj(t))
r
(5) j=IX ~jfn(~kj(t)) = 0 (n = 1,2 .... ).
nk
the s i m p l e s t one is the sequence of functions t ~ t where (nk)k= 1
nk
tions t are not l i n e a r l y dense in L (0,i) is a c o n s e q u e n c e of a
result of S. M a z u r [64].
6. Q u o t i e n t s of n o n - n u c l e a r spaces
~ = vol ~(U)
vol ~(W)
I@(B E A N) I i/k
Vk(¢ : E ~ F) = sup [
N IB F N ~(N)] ~
V k ( ¢ 1 ... Cs ) ~ V k ( ~ I) ... V k ( ~ s) (k = 1 , 2 , . . . ) .
J(~I~2(B(E3 ) A N) I i/k
]
IB(E 1 N (~i~2)(N) I
62
& Vk(¢2)'Vk(~l).
Pz~Qf. Take any operators X,~ with lixli,ll~ll ~ 1 such that the
composition X¢~ is defined and acts between unitary spaces. Fix an ar-
bitrary k = 1,2, . . . . We have to show that Vk(¢) ~ dk(X¢~). Denote
e = X~. Since e acts between unitary spaces, we have Vk(e) =
[dl(8) ... dk(e)]i/k Hence, applying (6.3) with s = 3, we derive
Hence, by (6.3),
63
k k
II dj(A) < ekk! II hj(A)
j :i j :i
In view of (i) and the inequality cs > 5, this implies that dk(A) =
o(k-4). Consequently, A is a nuclear operator ([79], Proposition
7.2.2). We have thus shown that, for each continuous seminorm on E,
there exists another seminorm such that the corresponding operator is
nuclear. This means that E is a nuclear space ([79], Theorem 7.2.7). •
Iu n MI ~ m_i . Iul
Iw n HI n! Iwl
It is clear that
Hence
64
lu n MI z m!
tUN Mm+iI
lw n M i (m + 1 ) ! t w n Mm+ll
n! Iw n Mnl
= m! IU n NI ,
n~ Iw n NI
The proof is s t a n d a r d .
(2) l i m s u p k l / 2 v k ( ¢ i N) =
k~
T h e n we m a y w r i t e
i/(k-s)
A ~ (k - s) I/2 kl/2 [(k - s)!] Ql/(k-s)
k k!
= (k - s, I/2 (k - s ) ! ] i / ( k - S ) k - S / 2 ( k - S ) [ k l / 2 Q l / k ] k / ( k - s )
T ' [ k!
vol
n ({u ~ u : tf(u) I ~ i}) ~ ½ vol n (u),
then
then
llvll2 = 1 f (v,u)2du, v ~ R n,
vol n (U) U
0c ( v , u ) 2 d u = oc ( v , u ) 2 d u for all v ~ R n.
U C
(n + 2) n/2 v o l n (C 0)
v o l n (U) 1
= (n + 2) n/2 vol n (C) vol n (C 0) V ° i n (C) v o l n (U)
(n + 2) n/2 2 1
n vol (U)
n
Hence, by (i),
limk~sup kl/2vk(ApnPolNn) = ~.
I/k(n)
(4) k ( n ) I / 2 [ I M n A B(Pn)I] > 2~e,
IMn N BlPo) I
(6) L n N U ° = {0},
1
(8) yOlk(n) ({u a U n : If(u)l > }) ->_ ~ V O l k ( n ) (Un).
1
(i0) Wni = - ~Cni(Wmj + amj) (i = 1 ..... k(n)).
k(n)
u = v + ~ ri(Wni + ani)
i=l
k(n) 1
From (6) we get ~ riani ~ B(Po) , which implies that Pu (u) > 7"
i=l
Finally, denote F = span K and s u p p o s e that F/K admits a non-
-trivial continuous unitary representation. Then, d u e to (4.5), there
exists a continuous non-zero linear operator @ : F + L~(0,1) with
0
0(K) c L Z ( 0 , 1 ). For e a c h n = 1,2 ..... we h a v e
k(n)
(13) u = Z Z Xni(Wni + ani)
n=l i=l
w i t h all b u t f i n i t e l y m a n y c o e f f i c i e n t s x. e q u a l to zero. N a t u r a l l y ,
1
such a representation is u n i q u e . For t ~ (0,i), let us w r i t e
k(n)
ft(u) = Z Z Xni~ni(t).
n=l i=l
k(n)
e(u) =
Xni0(Wni + ani).
n=l i=l
f t ( W n i ) = - l-cnift(w
p m3. + amj ) (i -- I, . . . .,k)
Thus, by (16),
Consequently, we h a v e
NUCLEAR GROUPS
7. N u c l e a r groups
-m
d k ( X , y ) < Co k o (k = 1,2 .... )
where co and m° are some universal constants. One may take, for in-
d k ( B p , B q) ~ c k -m (k = 1,2 .... ),
73
Pmaof. The "if" part is t r i v i a l . The "only if " part follows di-
rectly from (2.14). •
dk(W,V) ~ ck -m (k = 1 , 2 , . . . ) .
d k ( X n , Y n) < c 2 - m n k -m (k = 1 , 2 , . . . ) ,
P P P
E = II E x II Ei, X = II X n × II Ei, Y = II Y n x If E..
n= 1 n i~J n=l i~J n=l i~J i
Then X,Y are symmetric and convex subsets of the vector space E.
From (2.7) we get
Next, define
P
K = II K n × 1[ Ki and ( P x (~i) : K ~ G.
= ~n)n=l i~J
n=l i~J
P
IT ~ n ( K n A Xn) × l'I G i c ~(K A X),
n=l i~J
d k ( X n , Y n) < c 2 - m n k -m (k = 1 , 2 , . . . ) ,
II ( II Di) -- II D
~N i~I (v,i)6NxI ~,i
wj : H + Dj : = II D
(~,i)ENxI ~,i
and y - Z u i E Bq.
igI
n n
¢( Z u i) = Z ~(u I) (Ul,...,u n ~ K N Bp)
i=l i=l
(2) Z ui -
1
-- i ~ u i E Bq.
iEI 2 1 •
m
4( X u i) = %( + Z u i) = ~( ~ u i) + 4( X u i)
i=l iel i~I i~I i~I
m
= Z ~(ui) + X ~ ( u i) = X $(ui).
i~I i¢I i=l
n m n m
5q. %(ui) - 5q. ~(wj) = >q ~(ui) - 51 ~(-wj)
i=l j =i i=l j =I
n m
= ~( ~ u i + m )) = %(0) = 0.
i=l j=l (-wj
1 . -m-i
d k ( B p , B r) <. ~ c K (k = 1 , 2 , . . . ) ,
= ~'(u) + $'(w).
8. C h a r a c t e r s of n u c l e a r groups
i~m~f. We m a y a s s u m e that
(1) 7 k d k ( B p , B q) = i.
k=l
id
E > E
E > E
P q
1 1
]X(U - w)] < ~Ix(K N Bq) I < 4--m
1
(2) I<(¢q(K} N B{Eq}) I = ]x(K n Bq) I < 4"
A u = {h e Ep : ph~p(U) = X(u)}.
81
n
Now, choose any Ul,...,u n ~ K. Let M = span {~q(Uk)}k= 1 c Eq
and m = dim M. It follows from (2.11) that ~q(Bp) is an absorbing
subset of Eq and its Minkowski functional r is a pre-Hilbert semi-
norm on Eq. Therefore M n ~q(Bp) and M n B(Eq) are n-dimensional
ellipsoids in M. So, by (2) and (3.15), there exists a linear func-
tional g on M such that pg = < on M A ~q(K) and
m
(3) sup {Ig(~ I : u E M N ~q(Bp)} ~ 5 ~ kdk(M N ~q(Bp),M N B(Eq)).
k=l
Applying (2.13) and (2.8) (a), we see that
= dk(~q(Bp),~q(Bq)) ~ dk(Bp,B q)
= X(Uk).
, n
Thus h ~ 5B(Ep) N k=iN Auk.
Since the sets Au are all weakly closed and 5B(Ep) is weakly
compact, it follows that there exists some h ~ 5B(Ep) n N A u. It
u~K
remains now to take f = h~p. •
- 1
Y = {K ~ G : IK($(K n Bp)) I ~ 4"
It is c l e a r that Y is an e q u i c o n t i n u o u s subset of G . It r e m a i n s to
show that each × ~ X is a r e s t r i c t i o n of s o m e X 6 Y.
So, take any X 6 X. let K" = $-I(H). Then X# is a character
1
JK(~(K n Bp)) I < s u p {If(u) I : u 6 Bp} < 4"
Next, the f o r m u l a
~c~ Define
Q = {h ~ E p : h~p(a) E [ , ] + Z}
and, for e a c h u ~ K,
83
A u = {h 6 E p : h~p(U) ~ Z}.
1
(i) X kdk(Bp,Bq) 5'
k=l
n n
Z tkU k = Z t k C s # ( U k) (t k ~ R, u k E K)
k=l k=l
n n n
5q t k ¢ s ¢ ( U k ) = Z t k O ( U k) = a( Z tkU k) = 0(0) = 0,
k=l k=l k=l
¢o
,1- o0
2
k=l k=l = ~-~ < 4"
Hence, in v i r t u e of (3.20),
hand, from the proof of Lemma 6 in [4] it follows that if E,F are
a r b i t r a r y n o r m e d spaces and ¢ E G(E,F), then kVk(~) remains b o u n d e d
as k ~ ~.
9. N u c l e a r v e c t o r groups
d k ( U j , U i) < k -I (k = 1,2,...).
Therefore
88
g~U EU o ~ ( X ) = x(g)}
1
J×(g)l = lo(g)(×)J = IPG(~)(×)} = IP(¢(X))I =< } & ( × ) l --<-~-
This implies that g ~ U because we h a v e assumed U to be q u a s i - c o n -
vex. Hence pG(~) = a(g) ~ a(U), which yields PG(H N YU) c a(U) since
~ H N YU was arbitrary. From this and (2) we o b t a i n (i).
90
I
so Ix,(K N Bp) I < 9" In virtue of (8.1), there is some f ~E # with
X
Pfx = X~ on K and
) if X ~ V°,
(¢u)(x)
=Ifx(u if X ~ V°.
91
X E VO ,
(P~)(X) =I~0 (X) if
if X C V °-
We shall p r o v e that
= p(~u(x)) = pfx(U).
1 (X ~ V°).
Ix(g) l = IPfx(U)I --< Ifx(U) < ~"
hence, by (5),
1 Zk-2<~.
X d (¢(Bp),¢(Br)) & ~ k=l
k=l
~9.9) ~m~m~. It follows from (9.6), (9.4) and (7.5) that we might
define nuclear groups as H a u s d o r f f q u o t i e n t s of subgroups of nuclear
vector groups. Such a d e f i n i t i o n is m u c h simpler and allows us to omit
the long and c o m p l i c a t e d proof of (9.6). However, it has also some dis-
advantages. Firstly, it is not "intrinsic". Secondly, a p p l y i n g this de-
finition, we w o u l d e n c o u n t e r d i f f i c u l t i e s in s e c t i o n 7, with proving
that groups l o c a l l y i s o m o r p h i c to nuclear groups are nuclear. Also,
which is m o r e important, we w o u l d have t r o u b l e s in section 16 with show-
ing that the dual group of a m e t r i z a b l e nuclear group is nuclear. In
fact, we w o u l d have to repeat there the a r g u m e n t from the proof of
(9.6).
S( ~ (-l)i ;Q)
= Q and S ( X (-l)i ;R) = R.
i=l i i=l i
Jn
s = lim ~ g~(i)"
n~ i=l
n
~q d (D,E) < 1 and 71 d (E,F) < ~.
k=l k=l
m
Let Ul,...,u m ~ D and a e E be such that a + Z u~ E E. Then
i=l
there exists a permutation o of {i, .... m} such that
J
a + 7 uo(i) ~ F (j = 1 ..... m).
i=l
This is L e m m a 6 of [i0].
Jn
s = lim Z u (i)-
n+= i=l
co
3n+l
(6) s - Z u(i)
i=l Un+2
Jn
over the v e c t o r s - ~ u (i) and the v e c t o r s ui for i = in+l,...,
i=l
Jn+l" Set 1 = d i m L. From (3) and (4 we get L c M n + 2. Therefore,
1 1
7 d ~ ( U n + 2 N L , U n + 1 N L) < I, 5q. d 2 1
k=l k=l k ( U n + l q L'Un n L) < 4"
Thus, by (4), (3) and (6), from (10.5) we i n f e r that there exists a
permutation ~n of the set {~(Jn+l),...,~(Jn+l)}, such that
Jn j
(7) s - Z u (i) Z u E U (Jn + 1 < j < Jn+l )
i=l i = J n + l On~(i) n = = "
J
S - Z g U (J > Jn; n = 1 2,...)
i= 1 U p ( i ) n ' "
s g S( Z ui). •
ui E U for = m;
i > hence A m c s p a n U = s p a n U. This implies that
have to p r o v e that A is r a d i a l .
Choose arbitrary s g A and % g (0,i). It is to be shown that
%s ~ A. So, take any m = 1,2 .... and any U ~ No(F). Due to (9.3)
and (2.14), we can find some W ~ No(F) and two pre-Hilbert seminorms
p,q on s p a n W, such that W N B n B Q U and
P q
a0
(3) s ~ 7 u i + U.
i~l
such that
% Z u i E B q + Z u..
iEI i~J 1
%s E %U + % ~ u i c (i + %) + Z u i.
iEI iEJ
space of F contained in F .
o
S(~(ui);TX) = ~(S(Xui;RX)).
J
(i) ~ ~(U~(i)) ~ g as j ~ ~.
i=l
J
Z ~(u )(x) ~ g(x) as j ~ ~.
i~ 1 (i)
100
which c a n be w r i t t e n as
J
(2) p( ~q U (X)) ~ g(x) as j ~ co.
i= 1 (i)
J
(3) ~q. U (i)(X) ~ g(x) + k(x) as j ~ ~.
i=l
~(Zgi;G) = G n S(Zgi;G),
(I) Z d i ( U n + l , U n N M n + I) & 1
k=l
oo
(4) 57 2
k=l d k ( B s ' B r ) --< i.
id
E > E
Asr t
E > E
s r
~ = {h ~ E : h~r(K N Un+l) c Z}
r
onto r .
n
hl~ r # h 2 ~ r because ~r is s u r j e c t i v e .
Un, f = U n + {u e L m + 1 : f(u) = 0}
d k ( U I , U n) + d I ( N , N ) < k -I
(6) U n + K E No(F).
Ln+ 1 : = span (K D U n + I) = s p a n (K n B s) c ½B r + gp (K n B s)
c ½ U n + K.
U n + Ln+ 1 ~ No(FT).
The c o m p l e t i o n F
of F is a n u c l e a r v e c t o r group dueGo (9.5).
T T
Let ~ : F ~ G be the c a n o n i c a l e x t e n s i o n of ~.
Y
Without loss of g e n e r a l i t y we m a y a s s u m e that Z gi = 0; then
(mT)o-
It r e m a i n s to s h o w that ~(Q) = p. The inclusion ~(Q) c p is
trivial. To p r o v e the o p p o s i t e one, choose any a ~ P. There is a p e r -
mutation ~ of p o s i t i v e integers such that the series ~g~(i) con-
verges to a. Choose some w • F with #(w) = a. We have
9 J
~(W - i=l~ U (i)) = a - i=iZ g~(i) j~= > 0.
J
zj + Z . > w in F .
i=l u~(i) 3~ o
for e v e r y i, and
J
(8) ~ [U (i) + Vi] j+~ > W in F
i=l
is a c l o s e d linear subspace of E.
(i) S( Z u i) = Zu i + F0(Zui).
S( Z u i) c C( ~ u i) = ~ u i + A( Z u i) c Z u i + T0(~ui).
106
£. = ±i.
1
J
II ~ Ei¢uill ~ r (j = l,...,n).
i=l
THE ~ - ~ u ~ ( l ~ l
Ii. P r e l i m i n a r i e s
(i) the W e i l - R a i k o v theorem asserts that (*) holds for any LCA
group G;
(ii) the fact that (*) is true for every nuclear locally convex
space G is k n o w n as the M i n l o s theorem (see [67], Theorem i, p. 508
or [63], Ch. IV, T h e o r e m 4.3, p. 318); it follows easily from the M i n l o s
theorem that (*) is true w h e n G is a H a u s d o r f f quotient group of a
nuclear locally convex space (see Y a n g [104]);
f f~d~ = f fd~ .
X Y
112
we c a n find some U E No(G ) such that Ix(U) I < ~ for all X E Q. The~
for e a c h g e U, we h a v e
12. The B o c h n e r t h e o r e m
59
dn i/2(Bp,Bq) < ~_~.i
k=l
m ~/2 1
(3) X n < ~.
k=l
m
D = { ( X l , . . . , x m) ~ Rm : Z (Xk/~k)2 < i}.
k=l
m
2 ~<i}.
E = { ( X l , . . . , x m) E Rm : ~ (Xk/~k)
k=l
We have L = gp (L D D) and, by (3),
m 2 m )2 m 2 ~2
Z d (D,E) = 5q ( n k / ~ k = ~ qk < < 4.
k=l k=l k=l
f A
f(K)d~(K) ~ 1 - E,
L
(5) l<(w) l s ¼.
Since P c E, and the sets u + P, u E L, are pairwise disjoint, we
have
It is clear that
for k = l,...,m},
116
whence
m
(6) IAI v o l m (P) ~ 2m Z (a + ~ k ).
k=l
V = (A \ A w) U (A w \ A) = (A \ B) U (A w \ B).
Let us write
1 1
s -- - - exp [2~i~(u)]," t = Z exp [2~i<(u)].
lal u~Lna lat u~Lna
w
Is - t[ < ~1
= Z I exp [2~iK(u)][ ~ }V I .
IAI u~nnv " IAI
m
~q (2a k - 2 ~ k - lWkl)
k=l
I s - tl s 211 - ~ j < 211 .
m
2 m II (a k + ~k )
k=l
m m
1 - x - y ~ 1 - 2x - y and II (i - x k ) h 1 - Z Xk,
l+x k=l k=l
we obtain
m m
Is - t I ~ 4 Z (~k/ak) + 2 Z (lWk}/ak)-
k=l k=l
m
Since w ~ D, we have Z (Wk/Dk)2 _-< i, and thus, lWk I ~ q k for
k=l
k = l,...,m. Therefore, by (i),
117
> i - 2~.
n
dl/2 1
Z k (Bpm'Bqm) < i--~ (m = 1,2,...),
k=l
118
Hence
(2) ~ K L ~ L M = ~KM
-i
(3) ~K(A) = ~L(~KL(A))
lar B o r e l m e a s u r e ~ on H with ~ = ~.
P
119
(2) 5 ~ kdk(Bp,B q) ~ ¼,
k=l
1/2 1
(3) 5-:. d k (Bq'Br) < I--2
k=l
d k ( B s , B t) ~ d k ( B q , B r) (k : 1 , 2 , . . . ) .
Hence, by (3),
co
1
7. d l / 2 ( B s , B t ) < ~.
k=l
~L((L N B s)Lo ) ~ 1 - E,
w h i c h m a y be w r i t t e n as
-i o
(5) ~K(~LK((L n Bq)L)) ~ 1 - £.
Since K N Bq L N Bq, we h a v e
O
(6) ~ L-I
K((L N B q )~) c (K N Bq) K.
a regular measure.
As we h a v e just seen, to e a c h n there co,rresponds some Wn~ No(H)
- - 1
P(Y) < ~ ( ~ K I(K \ ~ K ( Q n ))) = PK (K \ n K ( Q n )) < n"
- 1
This implies that ~(H \ Qn ) ~ ~ since p is a R a d o n measure. Thus p
is r e g u l a r .
It r e m a i n s to c o n s i d e r the case ~(0) # i. If ~(0) # 0, then it
suffices to c o n s i d e r the p.d. function ~/~(0) instead of ~. On the
other hand, if ~(0) = 0, then, by (1.22) (b), w e h a v e ~ £ 0 and
is the t r a n s f o r m of the zero m e a s u r e . •
= X(Gp),
X(Y \ U (Y N Qn )) ~ ~( n (G \ Qn )) = 0.
n=l n=l P
< E + ~. •
B
(H D p)Ho = n ~LI((L A Bp)L),
Lg~
whence
O
Pi(X) < pi[~Ll(L - \ Bp)L] : PL(L- \ (L N Bp) L) < E.
of span ~(G). So, take any compact subset X of G and any ~ > 0.
T
G- \ X with
T
(3) X c U U~.
%
(4) X • V c U.
II llsn denote the norms in t h e spaces L (G- ,~) and L c2( S n ,~$), re-
we can find some complex numbers ll,...,l p and some continuous homo-
morphisms (characters) ~k : Sn + S, k = l,...,p, such that
P
IIX¢(x) k Z= l lkDkJlsn < £.
Then
P P 2d~]i/2
]bXV - Z lkDk~[l _ = [f _ [X V - Z Xk~k$ [
k=l G G k=l
P 2d~]i/2
= [• [X~(x) ~ lkq k] < e.
Sn k=l
P P
IIXx - ~ >k~k¢IL _ < fIXx ×Vl] _ + IIXv - Z kk~k¢Ib _
k=l G G k=l G
sI sn
Dk(Zl,...,z n) = zI ... zn
n sk n
(Dk$)(<) : qk(<(gl),...,<(gn)) : II < ( g k ) : <( Z S k g k ) ,
k=l k=l
n
which means that Dk$ : ~( Z Skgk) e e(G).
k:l
A ; 2G
for g ~ G, X 6 G and f 6 L (G ,~i ). Let H =~ LC( ~,~i ).
i~I
For e a c h A ~ B(G~), we d e f i n e a projection P(A) : H ~ H by
P(A)(fi)i~ I = (×Afi)i~i
We s h a l l n o w d e a l w i t h the p r o b l e m of the e x t e n d i n g of c o n t i n u o u s
p.d. functions and unitary representations.
2
r e p r e s e n t a t i o n in the space Lc(H ,p), given by the formula
(i) lim d k ( B p , B q) = 0,
k~
(2) 5 5: k d k ( B q , B r) < i,
k=l
~"(h + u) = ¢(h)~'(u) (h ~ H, u ~ N)
C o n s i d e r the c a n o n i c a l d i a g r a m
129
id id
F >F >F
Fp Fq Fr
[~(X) if X • Q
o(X)
if X ~ Q.
130
2
Hence, for each f • L c(H-,p), we have
(Ff)(<) = f(KIH) (K • F )
f If(o(X))12d~(x) = f If(<)12dpa(<),
H F
14. P r e l i m i n a r i e s
= <~G(h),x> = <~¢,X>.
~G
G > G
~G/H
G/H > (G/H)
is commutative. •
Let H be a c l o s e d s u b g r o u p of an a b e l i a n t o p o l o g i c a l group G.
The c a n o n i c a l h o m o m o r p h i s m s G - / H ° ~ H- and (G/H)- + H °, d e f i n e d in
the obvious way, will be d e n o t e d by %H and CH, respectively. Ob-
serve that ¢H is a c o n t i n u o u s injection; it is a s u r j e c t i o n if and
only if H is d u a l l y e m b e d d e d in G. The m a p p i n g ¢H is a c o n t i n u o u s
i s o m o r p h i s m of (G/H)- onto H° .
~Q
Q > Q
H
136
: (IIiG i) ~ }qiGi
1
~.ilx(Ki)] = Ix(K) I _-< ~.
Ir 1 7 -1
xi/U i ~ ~[4--~iJ
(Xi)ig I e Z i U i = U.
I~ o 1
Ix(K) 1 <-- >qilXi(Ki )I < 4 i(xi/ni ) < ~"
( ~ iGi
* ) -- ( IIiG ~) -- ( Z i G i ) -- Z iGi . •
id
E > E
- PE*
PE * - c * *
E > (Ec) < (Ec)
~ A
(i) 8(E) = (E c) .
one has
Let us d e n o t e
(4) v E M ± + cony A.
morphism of the space E endowed with its weak topology onto the space
tinuity of ~E"
Next, suppose that SL(U) • ~L(M). Then ~L(U) = ~L(W) for some
w g M. Let N = ~E(M n (E x {0})). Since u ~ M, it follows that
~E(U - w) ~ N. So, there is a c o n t i n u o u s linear functional f on E
K(~L(V)) = pf~E(V) (v • M)
M ~L : F ~ E × Ln
¢n = S n ~ n × n n "
co
~E"
144
on E* and E**.
(i) d k X,Y) ~ c k -m (k = 1 , 2 , . . . ) ,
I
Let I be a set of indices. By Ro we d e n o t e the corresponding
In
may identify Ln with a quotient group of Z° for some set In; let
I
~n : Zo n + Ln be the canonical projection.
(3) p c U n + 1 + B n + @n(Dn).
In
Similarly, there is a f i n i t e and symmetric subset Dn of Rn such
that
Let us d e f i n e
Now, take any k = 1,2, . . . . Applying (2.6) (a) and t h e n (2) and (4),
we get
Hence
(7) d 2 k _ l ( X 2 n _ l , Y 2 n _ I) < c 2 - m ( 2 n - l ) ( 2 k _ i) -m
(8) d 2 k ( X 2 n _ l , Y 2 n _ I) ~ d 2 k _ l ( X 2 n _ l , Y 2 n _ I)
< c2 - 2 m n k - m = c 2 - m ( 2 n - l ) ( 2 k ) -m.
L e t us n o w c o n s i d e r the v e c t o r space
E = II (M n × Ron).
n=l
I
It c o n s i s t s of s e q u e n c e s (Un, f n ) n~= 1 where un g Mn and fn g R on
we shall prove that the E,X,Y,K and ¢ thus defined have the de-
sired properties.
Inequality (i) follows from (5), (9) and (2.7). To p r o v e the in-
clusion p c ~(K N X), take any g ~ P. By (3), to each n there
correspond some u n E Un+l, bn e Bn and d n E Dn, s u c h that g = un +
co
~ ( ( u n + b n , d n ) n = I) = g.
In
It is c l e a r that the set D~ = Z ° N conv Dn is finite.
I
g Un + ~ n ( f n ) e Y 2 n - i + ~n(Y2n N Zo n) c 2U n + B~ + ~n (D "n ) .
(7) 1
5 Z kdk(Br,Bq) ~ 5.10 -2 ~ k-2 < 4"
k=l k=l
be the r e s t r i c t i o n of PG to K'.
if g E ¢(K N Bp)
if g ~ %(K N Bp)
1
B °r = {f • F # : If(u) I < ~ for all u • B r}
1
B p° = {f E F # : If(u)l < ~ for all u e Bp}
dk(X',Y') ~ d k ( ¢ ( B ~ ) , ¢ ( B ~) + L) + d I ( L , ¢ ( B ~) + L)
and that ~'(K" N X') e No(Fpc ) because [~(K N y)]O c ¢'(K" D X') and
~(K D Y) is precompact. •
17. Stron9 r e f l e x i v i t y
(a) H is d u a l l y c l o s e d in G;
(b) H is d u a l l y e m b e d d e d in G;
(~G ~
G > G
1
G/H 13
;
> G--/H °°
152
Ho H° Ho - ^ .
> (H°) -- > (G -/H°°) - ~ > (G/H)
C o n s i d e r the c a n o n i c a l d i a g r a m
H > G
¢ : G - ' / H OO ~ (H°) -
Ho
¢
G--/HOO H° > (H°) -
H° is r e f l e x i v e b e c a u s e cH is a t o p o l o g i c a l isomorphism.
G × II A n is a s t r o n g l y r e f l e x i v e group.
n=l
P = (G × II An) -- G ~ Z A-.
n=l n=l n
Pmmmf. It is e n o u g h to s h o w t h a t id : M- + E- is continuous.
c pc
Let X b e an a r b i t r a r y precompact subset of E. We have to find a
the set
E onto (ET)T.
is a r e f l e x i v e binuclear group.
For each n ~ Z, let e E E be the sequence with 1 in the
n
n-th place and 0 elsewhere. Let en, n E Z, be the c o n t i n u o u s lin-
U m = e - m + em' Um = (i - %)-l(e -m - % e m ) ,
co
Let K be the subgroup of E generated b y the set H = {Co} U { U m } m = 1
~em = o when m = 0
o
(i) [¢(H')]E/K = {0}.
* -i
f(u n) = p[(l - %) (X_n %Xn)],
which c a n be w r i t t e n as
W
x n = f(Un) + f(w n) + k n + i n,
X_n= f(u n) + % f ( w n) + k n + %1 n,
(3) f(u n
) + % f ( W n*) ~ Z + %Z.
{HE}E< ~ where
subgroup of G.
such that ge = 0 when ~ & ~. It is not hard to see that every pre-
compact subset of G is finite; therefore G- m a y be i d e n t i f i e d w i t h
Q topologically. For each ~ < ~, let g8 = (gi~)~<~ be the e l e m e n t
Let (A n)n=l
~ be a s e q u e n c e of LCA groups Set G = II A n and H =
n=l
Z A . Let P and Q be a r b i t r a r y c l o s e d s u b g r o u p s of G and H,
n=l n
respectively.
Kaplan [49] p r o v e d that G and H are reflexive, and that there
18. Groups w i t h b o u n d e d n e s s
Let G be an a b e l i a n group. By a b o u n d e d n e s s on G we m e a n a
family of subsets of G, called bounded sets, satisfying the follow-
ing conditions:
i) if X is bounded, so is -X;
Let G = Z (G i : H i ) be a r e d u c e d p r o d u c t of g r o u p s w i t h b o u n d -
ieI
edness. We i n t r o d u c e b o u n d e d n e s s into G in the f o l l o w i n g way. For
each i ~ I, let ~i : G + Gi be the c a n o n i c a l projection. A subset
X of G is b o u n d e d if ~i(X) is a b o u n d e d subset of Gi for all
i ~ I and if ~i(X) c Hi for all but finitely many i.
(18.4) R r n [ s . Let G =
Z (G i : H i ) be a reduced product of
iEI
topological groups with boundedness. Suppose that Hi is dually c l o s e d
in Gi for a l m o s t all i. Then Gd is c a n o n i c a l l y b-isomorphic to
where Ui E No(Gi) for all i and H I• c U.l for almost all i. Let
If Xi ~ Hi , then H~1 ~ X o
i because Hi is d u a l l y closed in G i. T h i s
means that Xi c Hi for almost all i. Then the set
o o it
If H i c Ui, then U i c Hi; since this holds for almost all i,
follows that %(U °) is a b o u n d e d subset of G'.
(a) H is d u a l l y c l o s e d and d u a l l y e m b e d d e d in G;
[G] eG > [G __ ] •
j-i ~ l.d
> [G ]d > [G] dd
Section i: Z, R, C, S, T
P
<x,g>
I×(A) I
G , G c, Gp, Gpc, G
¢-
o Ao Ao Ao
A°' AG' p' c' pc
Am
gp A
N (G)
0
G -- H
II Gi, Z G.
iEI igI i
g/U
Ui, Z * U.
i~I ieI 1
(G i : H i)
iEI
g
GL(E)
Hi' ~ ~i
ieI iEI
Section 2: E# E* E~, E* E*
' ' c' pc
OE
span A
conv A
XWY
dk(X,Y)
175
(utv)
d(u,A)
BE, B(E)
dk(¢ : E + F), dk(¢)
Bp, B(p)
E
P
~p
A
Pq
section 3: B
n
vol n
n
d(L)
L*
p(L,U)
Xk(L,U)
D
~i ~ "'" ~ ~n
Lc(X,P), L~(0,1)
L~(X,~), L~(0,1)
L~(X,~), L~(0,1)
e 2~i@
Section 5: c
o
1p
e
n
{f}
Lo
Section 6: lul
lwl
Vk(¢ : E ~ F), Vk(¢)
hk(¢ : E ~ F), hk(¢)
176
Section 7: Co(N,G)
Section 8: R~
6(E,F)
Section 9: Wo(G)
C(Zgi;G), C( Z q i )
A(Zgi;G), A( Z g i )
F( Z u i )
F0( ~ u i )
~C~), ~(~), o(~)
Section 14: aG
CH
~H'
Section 17: ~R, R~
Section 18: Gd
E
[G]
#d
~G
SUBJECT I_~DEX