Distribution theories based on representations of locally
compact Abelian topological groups
Citation for published version (APA):
Elst, ter, A. F. M. (1988). Distribution theories based on representations of locally compact Abelian topological groups. (EUT-Report; Vol. 88-WSK-06). Technische Universiteit Eindhoven.
Document status and date: Published: 01/01/1988
Document Version:
Publisher’s PDF, also known as Version of Record (includes final page, issue and volume numbers)
Please check the document version of this publication:
• A submitted manuscript is the version of the article upon submission and before peer-review. There can be important differences between the submitted version and the official published version of record. People interested in the research are advised to contact the author for the final version of the publication, or visit the DOI to the publisher's website.
• The final author version and the galley proof are versions of the publication after peer review.
• The final published version features the final layout of the paper including the volume, issue and page numbers.
Link to publication
General rights
Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights. • Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain
• You may freely distribute the URL identifying the publication in the public portal.
If the publication is distributed under the terms of Article 25fa of the Dutch Copyright Act, indicated by the “Taverne” license above, please follow below link for the End User Agreement:
www.tue.nl/taverne Take down policy
If you believe that this document breaches copyright please contact us at: openaccess@tue.nl
providing details and we will investigate your claim.
Department of Mathematics and Computing Science
Distribution theories based on representations oflocally compact
Abelian topological groups by
A.F.M. ter Elst
EDT-Report 88-06 3-12-1987
AMS Subject Classifications (primary) 16A80, 22010, 46A12
(secundary) 15A69, 43A65, 46A05, 46F05, 46H25, 46M05
I. 1.1. 1.2.
1.3.
II. ll.l. 11.2. 11.3. III. III.1. III.2. III.3. IV.V.
VI. VI. 1. V1.2. VII. VIII. VIII.I. VIlI.2. Appendix A.B.
C. Introduction Some notationsThe spacesSuo
c
andTu.c
The spaceSuo
c
The spaceT
u. c
The pairing betweenSuo
c
andTu,c
Seminonns onSu,
c
and bounded sets inTu.c
The Stone-representant for a unitary representation Seminonns onSuoc
and bounded sets inTu.c
A growth condition
Nice properties of the spacesSuo
c
andTu.c
Nice properties of the spaceTu. c
Nice properties of the spaceSuo
c
Su, c andTu, c spaces which are Montel or nuclear
Examples
Continuous linear maps
Conjugate linear homeomorphims The strong dual ofSu,
c
andTu,c
The spaces
Su.c
and1U,c
Cartesian productsTensor products and kernel theorems Tensor products
Kernel theorems Topological groups . Topological vector spaces
A suitable topology for the tensor product of two locally convex topological vector spaces which have additional structure
Acknowledgement Index of symbols List of properties References Page 1 3 4 4 6 8 12 12 16 20
26
28
31 3749
62
69
69
71 75 81 81 85 102 104 107 110 111 113 114Inthe monograph [EG2] the topological spacesSH,A enTH,A' which are both inductiveand pro-jective limit of Hilbert spaces, have been studied extensively. Here the labelHdenotes a separ-able Hilbert space, the label A denotes a positive self-adjoint operator in H. IfH =L2(JR) and
A
=-
1-
d 22 it turns out thatSH A
=
C * L 2(JR) as sets, with C=
{x i--7l.
2 t 2 :t>
O} a-'I
d x ' 1t t +xsubset of LI(JR) and * the convolution product between LI(JR) and L2(JR).Ifwe put a suitable topology on C * L 2(JR) we can prove thatSH,A
=
C * L 2(JR) as locally convex topological vec-tor spaces. The map! i--7 !*. from L1(lR) into L(L2(lR» is a --representation ofthealgebraL1(JR) by operators on the Hilbert space L 2(JR) and this map corresponds to the regular representation of JR in L2(JR). (See [HRI], Theorem 22.10.) So in fact the spaceSL
2
(1R) ,"-1-
:::2
is detennined by the group JR, a subset C of LI(JR)and a representation of this group in a Hilbert space.The aim of this report is to study the results in [EG2] from the viewpoint of commutative har-monic analysis and also to explore the possibilities for generalizations which naturally arise in that new context. We start with a locally compact Abelian group G, a subset C ofLI(G), a representationU of G in a (not necessarily separable) Hilbert spaceH and we suppose that the pair (C ,U)satisfies some technical conditions (pI, P2).
Inchapter I we define the locally convex topological vector space of smoothed emelementsSu,c' the trajectory space Tu,C and a duality between them. The spaceSu,c is an inductive limit of Hilbert space and T
u.
c is a projective limit of Hilbert spaces. Assuming a weak condition(P2 ')we prove thatSuoc
=
Has sets if and onlyifSuoc=
Has topological vector spaces.Inchapter II we define a projective limit topology on the vector space Suoc and we consider a special type of bounded subsets ofTu.c. It turns out that the projective limit topology forSuoc is equal to the inductive limit topology forSu,c if and only if all bounded subsets of T
u.
c are of this special type. Further, we consider an additional condition on the set C (P3) and then we prove thatSuoc is a projective limit of Hilbert spaces.Assuming the conditions PI, P2, P3, in chapter III we show that the following conditions are equivalent: the pair (C ,U) has property P4; Tu.c is bomological; Tu.c is reflexive; Su,c is complete; Suoc is sequentially complete; every bounded sequence inSuoc has a weakly conver-gent subsequence; all bounded subsets ofSuoc look as if the inductive limit were strict; and several other conditions. Finally we present necessary and sufficient conditions for Suoc and
T
u,
c to be nuclear or Montel.Inchapter IV we include several (old and new) examples. We show that allSH,A' TH,A' t(H, A)
anda(H, A)spaces, whereHis a Hilbert space andAa positive self-adjoint operator inH,fit in the theory.
Inchapter V we consider continuous linear maps between spaces of type Suoc and Tu.c. The theorems are nearly similar to the theorems in [EG2], Section lA, but the proofs are (of course) different.
homeomorphic with the strong dual ofT
u,
c. Assuming property P3, it is shown thatTu,
Ciscon-jugate linear homeomorphic with the strong dual ofSu,c if and onlyifa symmetry property (P4) holds.
Inchapter
vn
it is investigated whether Cartesian products of twoSu.c spaces (resp. twoTu.c spaces) are again of typeSu.c (Tu,C).Here property P5 plays an essential role.Inchapter VIII we introduce explicit completions of a-topological tensor product ofSUhCI and
Tu2,C2' (Four possibilities.) It is shown that the completed a-topological tensor product of two
T
u,
c-spaces is again of typeT
u.
c. We formulate and prove four kernel theorems. The latter is a generalisation of the corresponding chapter (III) in[EG2].Appendix A contains a summary of the concepts of commutative harmonic analysis as used in this report. Appendix B summarizes the prerequisites from topological vector space theory. In Appendix C we introduce a so-called a-topology on the tensor product of two locally convex topological vector spaces which have enough seminorms which correspond to semi inner pro-ducts. This a-topology corresponds to the Hilbert space tensor product of two Hilbert spaces. Inthe final part of this introduction we sketch the relation between the results of the underlying report and the papers[EGK]and[EK].These papers also expand the results of[EG2],be it in dis-tinct directions.
The set<I>of nonnegative functions on
R
Ilin [EGK] canbecompared more or less with the sets of complex valued bounded functions
j
on G in this report. An important difference is that <I> may contain unbounded functions. On the other hand, G need not necessarilybe equal to 1R1l•
The condition A.I in[EGK]is superfluous and the (strong) condition A.III in[EGK](cf. property P3 in this report) is always assumed. By condition A.I1 in [EGK],for every
x
E JR." there exists ~E <I>such that~x),p 0, whereas in this report all functionsj,
withf
E C, maybe0 on a largesubsetofG.
In[EK]the operators
a
E Rare all bounded as is the case with the operatorsU(j)in this report.Further, the operatorsaE Rare required tobepositive Hermitian and bounded by 1. Also a
sub-semigroup property is required which is not assumed in[EGK]nor in this report. A condition like property P3 is considered at the end of the paper[EK].
In both papers [EGK] and [EK] no theorems are proved which are similar to the theorems in chapters 5-8 in this report.
Some notations
Let A be a set and let V be a subset of A. 1hen VC
is the complement of V in A, so VC
=
{aE A :a~ V}. I
v
is thecharacteristic function ofV,thusIv:A -+ JR
{
lifaEV'
lv(a)
=
(a E A)Oifa~V.
Let
I
bea complex valued bounded function onA.Then II111
00:=sup ( II
(a)I :a E A}. INis theset of positive integers, JV=={1,2,3,···}.
Let Xbea topological space and letV bea subset of X. Then cIoV==Vdenotes theclosure ofV. By Cc(X)we denote the set ofallcomplex valued continuous functions Ion X with compact sup-port, Le. there exists a compact subsetK of X such that
I
(x)=0 forallx E KC•Further, Co(X)denotes the Banach space of all complex valued continuous function
I
on X which vanish at infinity, Le. forall£>
0 there exists a compact subsetKof X such that II(x)I<
£forallxE KC•The nonn onCo(X) is 111100 •ByC(X)we denote the set ofallcomplex valued continuous
func-tions onX.
The abbreviation a.e. meansalmost everywhere.
Let Gbea locally compact Abelian topological group and let
I,
g E L1(G).Then 11/111denotesthe norm of
I
and1*
gdenotes the convolution product ofI
and g. The adjointj
off
is the function (equivalence class) withj(x)
=l(x-
1), a.e.
x
E G. For every subset C ofL l(G)letC
denote the set
{! :
lEe}.
j
is theFourier transform off.Thedualgroup of G isG.
LetHbea Hilbert space. Theinner product inHis denoted by ( , ) andthe normby II II. LetTbe
a densely defined operator fromH into a Hilbert space. ThenD
<n
denotes thedomain ofTand T* denotes theadjoint operator. IncaseTis a bounded operator, we denote by IITII thenormofChapter I. The spaces SU, c and T
u.
c 1.1. The spaceSu,cLet G be a locally compact Abelian topological group with a Haar measure~and letUbe a (con-tinuous) representation of G in a Hilbert spaceH.For definitions concerning topological groups the reader is referred to Appendix A. We suppose that every representation is continuous. For every
f
ELI(G)define a continuous operator onHby(U(f)u,v)=
J
f(x)(UJ(;u,v)d~x) (u,vEH).G
(See [HRI], Theorem 22.3.) The operators U (f),
f
ELI(G)have the following properties: II U (f)u II S II fill\IuII ,U(f)*
=
uif>,
(f,g
E LI(G), uE H)U(f*g)
=
U(f)U(g)=
U(g)U(f).Let C be a subset ofLI(G).Suppose the pair(C, U)possesses the following properties. PROPERTIES1.1.
PI. For all
f,
gEethere exists h E C such that 1 and 2 hold: 1) f=hor there existsfl E LI(G)such thatf= h*
ft,
2) g=
hor there existsgl E LI(G)such thatg=
h*
gl.n.
There exist a net (f).kJ in C such that for all x E Hholds lim U (f'A)x=
x. 'AThroughout this report we suppose that the conditions PI and
n
are satisfied by the pair(C, U).REMARK. These conditions are weak, but sufficient for the purpose of this chapter. In practice one meets sets C which satisfy much stronger properties such as:
PI'. For all
f,
gEethere existhE C andIt ,
gI ELI(G)such thatf
=
h*
It
andg=
h*
gI.n'.
There existsaLI(G)-bounded net (f'AkJ in C such that for all x E Hholds lim U(f'A)x=X.'A
n".
There exists a LI(G)-bounded net (f'A)'A.eJ in C such that for all g E LI(G) holds limA*
g=
ginLI(G).).
Of course: PI' implies PI and
n'
impliesn.
Alsonit
impliesn'.
Indeed, letE:= {xE H :lim U(f'A)x =x} andF :={U(f)x :fE LI(G), x E H} Then E is closed in Hand
'A
FeE.SinceUis continuous,F is dense inH.SoE
=
H.REMARK. For all xE Hwe have limU(f'A)*X
=
x. Indeed, U(f'A) -Iis a nonna! operator, so 'AII (U(f'A)-/)* xII =11 (U(f).)-/)xll for all I.E J. Hence, lim II U(f'A)* x-xll=
lim II(U(f>..) -/)*xII=limII(U(f~)-/)xII
=
O.~ ~
REMARK. Let
C
:=if :
I
e C}. Then the pair(C, U) satisfies propertiesPI andn.
The pair(C, U)satisfies propertiesPI',
n'
andn"
if and only if the pair(C, U)satisfies propertiesPI',n'
resp.n".
DEFINITION 1.2. Let/e L1(G).DefineNf := {x e H : U(f)x =O} , the kernel ofU(f), Rf:= U(f)(H). the range ofU(f),
Of:=U(f) I
Ny :
Ny-+ Rf .Note thatNf= NjforallI e L1(G).
LEMMA 1.3.
Let Ie L1(G). Then Of is a bijection. So there exists a unique norm II IIf on Rf such thatRf
becomes a Hilbert space and Of is a unitary map. The identity map fromRf intoHis continuous. Proof.
Of is injective. Let x e H. Since U(f) is continuous. the set Nfis closed, hence there exist v e Nf and we Ny such thatx
=
v+ w. Then U (f)x=
U (f)w=
Ofw. So Of is swjective. Let cp e Rf .ThenIIcpll~ II U(f)1I11Oi(cp)II~11/11111cpllf'so the identity map fromRf intoH iscontinu-ous.
0
Inthe next lemma the spacesRfandRgare compared ifI
=
g*
hfor someh e L I(G).LEMMA 1.4.
Let I, g. he L1(G). Suppose I=g
*
h. Then Ng cNf and RfcRg. For all cpe Rf holds O;I(cp)=U(h)(Oi(cp».
The identity map fromRf intoRgis continuous.Proof.
Letx eNg. Then U(f)x=U(h)U(g)x=0, sox e Nf . Similarly: RfcRg. Let cp e Rf and let
x:=Of\Cp). Assertion: U(h)xeN{ Let veNg. Then U(g)U(h)v=U(h)U(g)v=O, so
U (h)ve NgC Nf and 0
=
(x, U (h)v)=
(U (h)x, v). This proves the assertion. Since Og(U (h)x)=
U(g) U(h)x=
U(f)x=
cp we have proved that O;I(cp)=
U(h)Oi(cp).Hence, forall cp e Rf we obtainIIcpllg=11O;I(cp)lI~ IIU(h)II IIOi(cp)lI~ IIh11111cpllf.0
DEFINITION 1.5.
LetSu c:= u Rf ·
• feC
By Lemma 1.4 and propertyPI,Su,Cis a linear vector space. The topologyO'ind forSu,Cis the inductive limit topology generated by the Hilbert spacesRf' Le. the finest locally convex topol-ogy forSuoCfor which all natural maps fromRfintoSu,Care continuous(fE C).
For the tenninology of locally convex topological vector spaces we refer to Appendix B.
LEMMA 1.6.
The identity map fromSuoc intoH is continuous. Hence the topologyO'ind forSuoc is Hausdorff. The setSu,Cis dense inH.Let IE C. Then the mapU (f) : H ~Su,c is continuous.
Proof.
The first assertion follows from Lemma 1.3 and the definition ofO'ind'Since H is Hausdorff, also
(Su C,O'ind)is Hausdorff. By property
n,
limU(f,.)x=
xfor allx EH,soSu c is dense inH.Let, A '
IE C and letPbe the projection ofHontoNJ.ThenU (f)
=Or
P : H ~ Su,c is continuous.0
THEOREM 1.7.
The space(Su.C,O'ind)is bomological, barrelled and a Mackey space.
Proof. O'indis the inductive limit topology ofbomological and barrelled spaces, hence(Su,C,O'ind) is bomological and barrelled ([Wil], Theorem 13-1-13). Every homological space is a Mackey space. (See [Will, Theorem 8-4-9.)
1.2. The spaceT
u,
CBesides the spaceSuoc we introduce the spaceTu,Cin this section. DEFINITION 1.8.
A C-trajectoryis a map eI> from C intoH with the property that
eI>(f*g)=[U(g)]* eI>(f) for alllEe, gEL1(G)such that
I
*gEe.LetT
u,
C be the vector space of all C-trajectories. The Hausdorff topology'tproj forTu,
c is the locally convex topology for Tu,C generated by the seminonns tl' defined by tleI» := II eI>(f) II, eI>E Tu.c,/E C.Itfollows that for all
I
E C, the map eI>~ eI>(f) fromTu, cintoHis continuous.LEMMA 1.9.
o
Proof.
First, suppose there exists 1ELI(G)such that
f
=
g*
1. ThenU (g)* «'P(j)=U (g)* U(1)* <l>(g)
=
U (j)*<I>(g). Ingeneral, by property PI, there existshE C such that 1) and2)holds:
1) 1=hor there exists
h
ELI (G)such thatf
= h*
it,
2) g=hor there existsg1ELI (G)such thatg=h
*
g1.If1= horg=h,then the lemma is trivial or already proved. Suppose
I
*'
handg*'
h.Let11andg1as in 1)and2).ThenU (j)* «'P(g)= U (h)* U (jl)* U (g1)*<I>(h) =
U(h)* U(gl)* U(jt>* <I>(h)=U(g)* <I>(j).
0
The Hilben spaceHcan be embedded inT
u.
Cin a natural manner.DEFINITION 1.10. Define
emb:H~Tu.c
[emb(x)](f)
=
U(f)* x (x E H, I E C).THEOREM 1.11.
The map emb is injective and continuous. For all <I>E T
u
Cholds <I>=lim emb(<I>(f),» inTu c.
. A. .
So the rangeemb(H)of the map emb is dense inT
u. c.
Proof.
LetxE Handembx=O.By propenyP2, then alsox=limU (f).,)* x=lim [emb(x)](f),)=O. So
A. A.
emb is injective.Itis trivial that emb is continuous. Let<I>
E
Tu.
Cand letlEe.
Then by Lemma 1.9, limtlemb<l>(fA.)-«'P)=limIIU(f)* «'P(fA.)-<I>(f) II=limIIU(f),)* <I>(f)-<I>(j) II=O. SoA. A. A.
<I>=lim emb(<<'P(jA.» andemb(H)is dense inTu c.
A. .
COROLLARY 1.12.
I. Forall«'PE Tu. c,fE C andgEL I(G)such thatf
*
g=0 holdsU(g)*<I>(f)=O.II. Forall<I>E Tu. c andIE C holds«'P(j)E
Nt.
Proof.I. Let IE C, g E L1(G) and suppose1* g=0. Let E:= {«'PETu,c: U(g)* «'P(f)=0}. By definition of'tproj' the set E is closed in Tu.
c.
It is trivial that E contains embeR). So E=Tu. c .II. Letx E H andlEe. Forall vE Nf , ([emb(x)] (f), v) = (U(f)* x,v) = (x, U(f)v)= O. So [emb(x)] (f)E
Ny.
SinceNy
is closed andemb(H)is dense inTv.c. the corollary follows. I]THEOREM 1.13.
The space(T
u,
c,'tproj) is complete.Proof.
Let«(f)a)aeA be a Cauchy net inTv,c. ForalllEe, «(f)a(f»aeA is a Cauchy net inH, so there exists (f)f E H such that lim(f)a(f)=(f)f. Define (f) :C ~H by CfJ(f):=(f)f' lEe. LetlEe,
a
gE L 1(G)and suppose
I
*
gEe. Then(f)(f*g)= lim(f)a(f*g)=limU(g)*(f)a(f)=a a
U(g)* (lim(f)a(f» =U(g)*~). So (f) E Tv,c. Since limtt<(f)-(f)a) =0 for all lEe,
a a
o
The next lemma turns out to be useful. (Cf. Theorem 1.18.)
LEMMA 1.14.
Let
I,
gEe,hE L 1(G)and suppose that/=g*
h.Thentf~II h 111 tg.Hence, letBe Tv,c be a 'tprorneighbourhood ofO. Then there existI
E C andc>
0 such that {(f)E Tv.c:tt<CfJ)~c} c B.Proof.
For all (f)E Tv,c we get tt<(f)=IIU(h)*(fJ(g)II~IIU(h)III1(fJ(g)II~lIhIl1tg«fJ).From this and
property PI, the second assertion follows.
0
1.3. The pairing betweenSv,CandTv.C
In this section we define a duality
< ,
>
betweenSu,C andTu,C with the properties thatGind and 'tproj are compatible with< ,
>
and<
cp,embx>
=(ell,x) for allellESu,C andx EH. It turns out that<U (f)x,(f)>
=(x,(f)(f» foralllEe,XEHand(f) E Tv c. (See Theorem 1.16 II.)LetellESu,c and(f) E Tu,c. There existslEesuch thatellE Rf .We show that it makes sense to define<cp,(f)
>
:=(Oi(cp),(f)(f».LetgEeand suppose alsoellERg.Then we havetoprove that(Oi(eIl),(f)(f»=(Q
g
1(eIl),(f)(g». First suppose there exists 1E L1(G) such thatI=g
* I. By Lemma 1.4:In the general case there existsh E C such that 1) and 2) hold: 1)
1=
hor there existsII
ELI(G)such thatI
=
h*
II,
2) g
=
hor there existsgI ELI(G)such thatg=
h*
gI.In any case:(Qi(ell), <1>(f))
=
(Qh
l(ell),<1>(h))=
(Q;I(eIl),<1>(g )). DEFINITION 1.15.Define
< ,
> :
Su, exTu,C -7CThe pairing
< , >
is sesquilinear.THEOREM1.16.
I. LetellE Suoc andxE H.Then
<
eIl,embx>
=
(ell,x).II. Let
I
E C,x E Hand <1>E Tu, c.Then<
U (f)x,<1>>
=
(x,<1>(f)). Proof.I. Let
lEe
be such that ellE Rf . Then<
<I>,embx>= (Qi (<I>),(embx)(f)) =(Qi(<I», U(f)* x)
=
(U(f) Qi(<I», x)= (<1>,
x).II. LetPbe the projection ofH onto
Nj.
ThenQ"i (U (f)x) =Px. Hence by Corollary 1.12 II:<U (f)x,<I>
>
=
(Qjl(U (f)x),<1>(f))=
(Px, <1>(f))=
(x, P<I>(f))=(x,<fJ(f)).0
LEMMA 1.17.
The pairing
< , >
is nondegenerate. Proof.Let ellE Su,e and suppose
<
eIl,<1>>
= 0
for all <1>E Tu,c.
Then111\111
2=
(<1>,1\1)
=
<
eIl,emb1\1
>
=o.
So1\1=0.
Let <1>E TU,e and suppose
<1\1,<1»
=
0
for allellE Su, c. LetlEe.
For all xE H we obtain by Theorem 1.16 11,0=
<
U(f)x,<1>>
=
(x,<1>(f)). So <1>(f)=O.
0
THEOREM1.18.
The topology(jindforSu,c is compatible with the dual pair
<
Su,c'Tu,c>
and the topology'tprojforTu,c is compatible with the dual pair
<
Tu,c'Su,c>.
Proof.For
lEe
letif :R
f -7Su, c bethe natural map.Let <1>E T
u, c.
Define[(cjl):=<cjl,~> (cjl e Su.d.
Let / e C. For all cjl e Rf: 110 ilcjl)I= I< cjl,<1> > I= I(ni(cjl),<1>(f))IS;"<1>(f)1111 n/\cjl) 11=
II~(f)1I11cjlllf. so [0 if:Rf ~C is continuous. Hence [ : Suo e ~C is continuous. (See [Will,
Theorem 13-1-8.)
Conversely, let [ :Suo e ~C be continuous and linear. Let / e C. The map [0 U (f) : H ~ C is
continuous and linear by Lemma 1.6. So (Riesz) there exists a unique <1>(f)e H such that
[ 0 U (f)x=(X,~(f)) for allx e H. We prove that~is a C-trajectory. Letg e C,h eLl(G) and
suppose /:=g
*
h e C. Then for all xe H we have (x, U(h)* <1>(g))= (U(h)x,<1>(g))=[0 U(g)U(h)x= [0 U(f)x= (x,<l>(f)). We conclude that <1>(f)=U(h)*~(g)and ~e Tu e.
Finally, let cjl e Suo e. Let/ e C be such that cjl e Rf .Then [(cjl) = (njl(<I»,<1>(f)) = <cjl,<1». So the topology(JindforSu,eis compatible with the dual pair <Suoe,Tu.e>.
Let cjl e Suo e. Define
[: Tu.e ~C
[(~):= <<1>,<1» (<1> e T
u.
d.
The map is [ is linear. There exists /e C such that cjlE Rf . Then 11(<1»1 = I <cjl,<1>
>
I= I(ni(<I»,<1>(f))IS;II ni(cjl) IIt/~) for all ~ E Tu.e. So1is continuous.Conversely, let1 : Tu.e ~ Cbe continuous and linear. Since 1is continuous, by Lemma 1.14, there exist/ E C and c
>
0 such that Il(~)IS;ctl<1» for all<1>E Tu.e. Define ex:U(j)(H)~ C by ex := [0 emb0 nt. The map ex is linear. Let the topology forU (j)(H) be the inducedtopol-ogy ofH. Letxe U (j)(H). Then I a(x) IS; cII[emb(ni(x))](f) II = c IIxII. So ex is continuous. By Riesz' theorem there exists a unique ve U (j)(H)=Nt such that a(x)=(x, v) for all
xe U(j)(H).
.l
-We show that [(<1» = <U (f)v,~> for all ~e Tu.e. Let xEN
i'
Then [(embx) =ex(U (f)x) =(U(f)*x,v)= (v,[embx](f))= <U(f)v,embx>. Let xENj. Then Il(embx)/$
ctlembx)=O, so 1(embx) =0= (U(j)x,v)= <U(f)v,embx>. So for all xEH we obtain
1(embx) =
<
U (f)v,embx >. Because [ and ~ ~ <U (f)v,<1> > are continuous functions onTu.e andemb(H)is dense inTu•e,forall~e Tu.eholds [(<1» = <U(f)v,<1».
0
InLemma1.6 it is proved that the setSuo e is dense in the Hilbert spaceH. Itmay happen that
SUoeequalsH as a set In a particular case, it can be proved that (Su.e,(Jind),as a topological vec-tor space, is equal to the Hilbert spaceH.
THEOREM 1.19.
SupposeSuoe=H as sets and suppose property P2' holds. Then (Su,e,(Jind)=H as topological vector spaces.
Proof.
The identity map from Su, e into H is continuous by Lemma 1.6, so we have to prove that the
let> : H ---+C
let>(x):=<x, <I>> (xE H) .
Let (/;..heJbe the net as in propertyP2'and letM
>
0be such thatII1..
111<M for allAE J. LetxE H. Then lemb(x)(Y)=<y,embx>=(Y, x) for ally E H. So lemb(x) is continuous(xE H).Let <I»E T
u.
c. Assertion: the set {lemb(et>(f,» :AE J} is pointwise bounded. Indeed, letx
E H.Since H=Su,
c
as sets. there exist1
E C and y E NJ such thatx
=
U (j)y. Then for all AE J: llemb(et>(f,»(x)1 = I<x,emb<l>(j..)> I = l(y, [emb<l>(j..)](j»\ = l(y, U(j)* <I»(j')..) I= l(y, U(j..)*<I»(f»1 S lIyII IIU(f..)II II<I»(f)IIS lIyII IIhlll II<I»(j)IIS Mllyllll<l>(j)11. This provesthe assertion. By the uniform boundedness theorem ([Wil] , Theorem 3-3-6), the set {lembet>(f,) :AE J} is uniformly bounded. LetM0 be an upper bound for this set. Then for all
x E H : Ilet>(x)I = I <x,<I»
>
I=lim I<x,emb<I>(j..)>
I=lim IlembdV')(x) ISM0IIxII. Solet>is..
..""'V'
continuous for all <I»E Tu,
c.
Leta.E Su,c'. There exists <I>E Tu,c such thata(l\I)= <1\1,<I>
>
for all1\1E Suoc. Thena0 i =let>is continuous. Since a Hilbert space is bomological, the mapi :H---+Suo c is continuous.
0
We fInish this chapter with a lemma which will be used in Chapter 8, but belongs here. LEMMA 1.20.
Let
1
E C,gE L1(G) and suppose1
*
g =0. Then U(g)*x
=0 for allx
E Nj.Proof.
Let E:= (xE H: U(g)* x=O}. Then E is closed in H. Let YE H. Then
U(g)* U(j)* Y = U(f* g)* Y =0, so U(j)* YE E. Hence U(j)* (H) cEo By [Wei], Theorem
Chapter II. Seminorms onSu,cand bounded sets inTU,c II.I. TheStone-representant for a unitary representation
LetUbe a representation of a locally compact Abelian group G in a Hilbert spaceH. The group G and the representationUare taken fixed throughout this chapter.
DEFINITION 2.1.
LetAbe a locally compact topological Hausdorff space,
m
a measure onA, / an index set, for all iE/letAj be an open subset ofA with induced topology and let tj :G
~Aj be a topologicalhomeomorphism. LetWbe a unitary operator fromHontoL2(m).
The tuple(A,m, /,Aj,tj,W)willbe called aStone-representantforUif and only if:
Aj ( )Aj =0 ifi
*
j (i,j E I),A=UAj, jel
The mapY ~ m(tj(Y», Ya Borel measurable subset of
G,
is a finite regular measure onG
(i E I),o
<
m(Ai)<
00 (i E I),m
(Z)=L
m
(Z () Aj) (Zc
ABorel measurable),iel
For all
!
ELI(G)leti
be the continuous function onA such that f(ti(Y» =](y)for all iE/ andallYEG.
ThenW U(f)W-1~ =i.~forall!E L1(G) and~EL2(m).
This definition is useless without the next theorem. THEOREM 2.2.
There exists a Stone-representant forU.
Proof.
See[HRII], Remark 33.6and[HRI], Theorem C.37. REMARK. The Stone-representant is not unique.
o
REMARK.If,in addition, the Hilbert spaceH is separable, then the measuremis regular.
Proof. Fori E /letei:= (m(Ai)rt IA;'Then(ei)iel is an orthonormal set inL2(m),soI is count-able. We may suppose that the set / is infinite, and hence suppose that / = IN.
Every compact subsetKofAis contained in only finite many open subsetsAj ,hence
m
(K)<
00.Let Z be a Borel measurable set,m(Z)
<
00and let e>
O. For alln E INthere exist openU,.c
G
such that m(t,.(U,.»< m(Z()A,.)+
er"
and U,.:::>t;;l(Z()A,.), by the regularity of y ~ m(t,.(Y». Let U:= v t,.(U,.). Then V is open inA,Z c U and m(U)=
L
m(t,.(U,.»'5.00
L
m(ZnAII)+£rll=m(Z)+£.11=1
Som(Z)
=
inf{m(U) : Uopen, Zc U}.Let U be an open subset ofA. Suppose
m
(U)<
00. Let £>
O. There exists N E IN such thatN
m(U)~
L
m(U nAil)+~.
Foralln
E IN,n
~
Nthere exists compactK nc
t;;I(Un
An)suchn=1
2
N
thatm(U
nAII)~
m(tn(KII»
+~.
LetK := u tn(Kn). Since eachAll is open inA,the setK is2n
11=1 compact inA, Kc
Uand N m(U)~L
m(tll(Kn»+£=m(K)+£. 11=1 Hence,m(U)=sup{m(K): K compact,K
c
U} .Similar, if
m
(U)=
00,m(U)
=
sup(m(K) : K compact,Kc
U} .Somis a regular measure onA.
REMARK. The measure
m
is locally finite, Le. for every measurable set Zc
A withm
(Z)>
0, there exists a measurable Z1 C Z such that 0<
m
(Z1)<
00.DEFINITION 2.3.
Let BorCG,C) be the set of all complex valued Borel measurable functions on G and let Borb(G,C)be the subset of all bounded elements of Bor(O,C). Define similarly Bor(O, JR.) and
Borb
(0
,JR.).With the aid of the Stone-representant for U we can extend the set operators U(j) with
fE LtcG) to a set operatorsU[F] withF E Borb(G,C).
DEFINITION 2.4.
Let(A,
m,
I, Aj,'tj, W)bea Stone-representant forUand letFbea Borel measurable function onO.
Define the Borel measurable function F onAby F(tj(Y» :=F(y) (i EI, yEO) .For everyFE Borb(O,l')define a continuous operatorU [F, A]onHbyU [F, A]:=W-1MFW,
withMF the multiplication operator by F onL2(m).
The definition of F depends onthechoice ofA,but from the context it will follow on which setA
REMARK. It is clear that forallF, K E Borb(G,C),allAE Candall!E L1(G) U [F
+
K, A]=
U [F, A]+
U [K, A] , U [AF, A]=
AU[F, A] , U[FK, A]=
U[F, A]0 U [K, A], U[/1,A ]=
U [F, A]* , Uri,
A ]=
U (f) ,for every Stone-representant (A, M, I, Aj ,'tj,W) for U.
LEMMA 2.5.
Let(A,
m,
I,Aj,'tj,W)be a Stone-representant forU. I. LetVbean open subset ofG.
LetX:=clo{U[F,A](H):FE Co(G),O~F~ Iv}. ThenU[1
v,
A ]is the projection ofHontoX.II. Let F1,F2"" be a uniformly bounded sequence in Borb(G,C) and suppose
F('Y) := lim FII('Y)exists for every'YE
G.
ThenU [F, A] =s - lim U [FII,A].11-+"0 11-+"0
Proof.
I. LetXo :=U[Iv,A](H). ThenXo is closed inH and X cXo. Let!E X oand let~:=WI
Then Iv'~=~ in L2(m). We have to prove that ~E clo{F1'\:FE Co(G), O~F~Iv. 1'\E L2(m)}=W-1(X).We may as well assume that~is real valued and~ ~ O.Moreover, by Lebesgue's dominated convergence theorem, we may also assume that there exista,~
>
0such that~=~·Iz with Z={a E A:a~ ~(a)~~} and ZrlAj c'tj(V)for all iE I. Then
m
(Z)<
00. Let E>
O. There existn
E IV and i1 , ' •• ,illE I such thatm (Z)
<
±
m(ZrlAj})+
~.
Sincem0 'tj is a regular measure onG,
there exists compactj=1 2~
A l E
K cG such that K c'ti} (ZflAj.) and m(Z rlAj.)-m('tj,(K»
<
- - 2 for allJ J J 2n~
j E {I, ., .,n}. Then K C V. By Urysohn's lemma there exists FE Cc(G) such that
IK~F~ Iv.
II
Let K1:=U 'tj(K). Then K1 C Z and m(Z) - m(K1)
<
E~-2. Let 1'\:= IXt .~. Thenj=1 J
II~-F1'\1I2 = II ~-F IKt ~II = IIIz~-IKt ~1I2 ~ ~2I1Iz-IK, 112~ ~2. E~-2 = E. SOXo cX. II. This follows from Lebesgue's dominated convergence theorem.
o
THEOREM 2.6.
ThenU[F, AI] =U [F, A 2]forallF E BorbG,C).
Proof. Let fE LI(G). Then uli,Ad=u(f)=uli,A 2]. Since {j:fE LI(G)} is dense in
Co«h
we obtainU [F, AI]=
U [F, A 2]foraUF E Co(G)by Lemma 2.5 II. LetVc
Gbe open.Then we have seen that {U [F, Ad(H): FE Co(G),O~ F~ Iv}
={U [F,A 2](H): FE Co«h
O~ F~ Iv}. Hence by Lemma 2.5 I,U[lv,Ad
=
U[lv,A2 ].LetB :={V : V
c
G
Borel measurable and U[lv,Ad
=U[lv,A2]}' ThenB contains all open subsets of G andBis a a-algebra by Lemma 2.5 II, soBcontains all Borel measurable subsets of G. Using Lemma 2.5 II again, we obtainU [F,Ad
=
U [F, A2]for allFE Borb(G,C). [) We naturally arrive at the following definition.DEFINITION 2.7.
Let FE Borb(G,C). Define U[F]:=U[F,A] where (A,m,I,Aj,tj,W) is any Stone-representant forU.
We summarize some properties ofU [. ].
PROPOSITION 2.8.
LetF, K E Borb(G,C), AE C andfE LI(G).Then IIU [F]II~\I F1100 , U [F+K]=U [F]+U[K], U [AF]
=AU
[F] , U[FK]=U[F]o U[K], U [F]=U [F]* , UI1J
=
U(f), U[1a]=I. LEMMA 2.9.LetF E Bor(G,C) and let(A,
m,
I,Aj,tj,W) be a Stone-representant for U. For alln
E IN letVn :={YE
G:
n-l~ IF(y)1<n}.
Then the orthogonal sumL
U[F.lvJ= W-IMFW,nelN
whereMF denotes the (unbounded) multiplication operator by F onL2(m).
Proof.
In
the following definition we extend
U [.]to Bor(G,
C).This definition makes sense by Lemma
2.9.
DEFINITION 2.10.
Let
F EBor(G,
C).Define a closed, densely defined operator
U [F]on
Hby
U [F]:=
W-1MFWwhere
(A,m, /,
Aj,'tj,W)is any Stone-representant for
Uand
MFdenotes the multiplication
operator by F on
L2(m).REMARK.
U [F]is a self-adjoint operator on
Hfor every
F eBor(G,
JR).11.2.
Seminonns on
Su.cand bounded sets in
Tu.cLet C
bea fixed subset of
L1(G)which satisfies properties
PIand
n.
Corresponding to the set C
we define
threesubsets of Bor(G,C) and seminonns on
Su.c.DEFINmONS 2.11.
Let
C*
:={FEBor(G,C): for
allfECis
F·jbounded} ,
C::= {F EC*:
for
all'YEG
isF('Y)il:0} ,C:
:={FEC* : there exists
£>
0such that for
all'Y EG holds
IF('Y)I>
E} .For
allF EC* define
SF :Su.c-? JR
REMARK. Since}=! for
allfEC, it is obvious that (6)*
=
C* , (6): = C: and (6); =
C;.
We
-*
-*
-*
- *
-
1# -*
write C
,C"and C
pfor
(C) ,(C)u'resp.
(C)p.REMARK. C* is a vector space.
Borb(G,C)c
C'.For
all FE CI#and
K E Borb(G,C),also
FK
EC*
andF
E C*.LEMMA 2.12.
I.
Let
FE el#.Then
SFis O'ind-continuous on
Su,c.I. Let f E C. For all C/lE Rf holds SF(C/l)=11U [F]U(f)Oi(C/l)1I= IIU [Fh0i(C/l)IIS
IIU[Fhllil0i(C/l)IIS IIFjII00IIC/lllf.By [Will, Theorem 13-1-8,SFis <rind-continuous.
II. Trivial because16 E
e;.
0
DEFINITION 2.13.
Let <rpmj be the locally convex Hausdorff topology forSu,c generated by the seminormsSFwith
FE
e;.
COROLLARY 2.14.
The following inclusion holds for the topologies forSu,c:
Proof.
Lemma2.12I.
We introduce a specialtypeof elements ofTu.c. DEFINITION 2.15.
LetFE eNandxE H.DefineU[F].x:
e
~H(U[F].x)(f):=U[F]U(f)*x (fE e).
LEMMA2.16.
LetF E eNandxE H.Then:
I. U [F].
x
isae
-trajectory.II. IfxED (U[F]),thenU[F].x
=
emb(U[F]x).III. LetC/lE Su.c.Then
<
C/l, U [F].x>
= (U [F]C/l ,x).IV. LetKE Borb(G,C).ThenU[F]. U[K]x=U[FK].x.
V. SupposeF E e~.ThenU [F].x=0 if and only ifx=O.
VI. The mapU [F] •fromH intoTu,cis continuous.
o
Proof.
I. LetfEe, gEL1(G)and supposef*g E C.Then
(U[F].x) (f*g)= U[F] U(f*g)* x=U[F f {]x=U[{] U[Ff]x=U(g)* (U[F].x)(f).
II. Supposexe D(U[FJ).Forallfe Cholds
(U[F].x)(f)
=
U [F] U(f)* x=
U(f)* U[F]x=
(emb(U[F]x» (f).III. Letfe C and
ell
e Rf.Then<
ell,
U[F].x>
=(ni (ell),
U[F] U(f)* x)=(ni(ell)'
U[F !]x)=
(U(Fh
0:iCell) ,
x)=
CU[F]ell,
x). IV. Forallf
e C we have(U[F]. U[K]x) (f)= U[F] U(f)* U[K]x=U[FK] U(f)*x=(U[FK].x) (f).
V. Let(f',,»),eJ be the net in propertyPZ. SupposeU[F].x =0.
Then U[F](U(f~)*x)=(U[F].x)(f~)=O,so U(f~)*x=O for all AeJ. Hence
x
=
limU(f~)*x =0.VI. Let
f
e C. Forall xe Hwe obtain-
-tf(U[F].x)
=
II U[F!]xliS;II F!1100II xII.SoU[F].is continuous.DEFINITION2.17.
LetF e C~. By Lemma 2.16 V there can be defined a unique nonn II IIFonU[F].H such that
U [F]. Hbecomes a Hilbert space andU [F].is a unitary map fromHontoU [F] • H.
LEMMA 2.18.
I. The identity map fromU[F] .Hinto (Tu,c,'tproj)is continuous forall F e C~.
II. LetF, K e C~. Then IF I
+
IK I e C:, U[F].Hc
U[ IF I+
IK f ].H and theiden-tity map fromU [F] • HintoU [I F I
+
I K I ] • H is continuous.Proof.
1. This follows from Lemma 2.16 VIandDefinition 2.17.
II. It is clear that I F I
+
I K t eC~.
Define L e Borb(6, C) by L:=
I F I~
I K I . Then ( I F I+
I K I ). L=
F.Letxe H.Then by Lemma 2.16. IV,U [F]. x
=
U [ 1F I+
I K I ] • U [L ]xE U [ I F I+
I K I ]. HandIIU[F].xll'FI+IKI =11 U[L]xllS; IIxli=U U[F].xIlF.
IJ
DEFINITION 2.19.LetT:
=
u U[F].H.Fee:
By Lemma 2.18 II,T is a linear vector space. The topology'tindforT is the inductive limit topol-ogy generated by the Hilbert spacesU [F] • H , F e C~.
every 'tprorbounded subset ofTu,c is contained in a set of the formU [F] • BwithFEe; andB a bounded subset ofH.Then, in particular,T
=
Tu,c as sets and 'tind is a locally convex topology for the vector space TU,C'LEMMA 2.20.
Tis a subset ofTu,c.The identity map from (T,'tind) into(Tu,c ,'tproj) is continuous. Hence the topology 'tind forTis Hausdorff.
Proof.
From Lemma 2.16I it follows thatT
c
Tu,c and by Lemma 2.18I and[WillTheorem 13-1-8, the identity map from (T, 'tind) into (Tu,c ,'tproj) is continuous. Since'tproj is Hausdorff, also 'tindis a Hausdorff topology.
0
The main result of this section is the following theorem.
THEOREM2.21.
The following conditions are equivalent.
1. (Su,c,O'ind)=(Su,c ,O'proj) as topological vector spaces.
II. For every 'tproj -bounded subsetBofTu,cthere existFEe;and a bounded subsetB0ofH such thatB
=
U[F].B o.Proof.
I =:> II. LetBbea bounded subset ofTu,c, LetBO bethe polar ofB.By[WillTheorem 8-4-12,
Ba is an absorbing subset ofSu,c,which is also closed and absolutely convex. SoBO is a barrel inSu,c and since(Su,c, O'ind) is barrelled,BO is a O'ind-neighbourhood of 0 in Su,c, Because O'ind=O'proj, there exist FEe; and £
>
0 such that{ell
E Su,c: SF(eIl)$ e} cBO.Let <I>EB. For all
ell
ESu,c with IIU[FlcIlll$e
holds I<
ell,
<I>>
I$ I, so1
I
< ell,
<I>>
I=:; -II£ U[F]eIlllforallell
ESu'c.
Define1:U[F](Sud, ~CbyI(U[F]eIl):=
<
ell ,
<1» (eIlE SU,d. Let the topology for U[F](Su,d be the induced topology ofH.Then the map1is continuous and linear. By Riesz' theorem there exists XII>E U[F](Su,dcH such that l(x)=(x,xII» for all XE U[F](Su,d. Then\I XII>II
=
11111=:;e-
1•Forallell
ESu,cwe obtain-
-<
ell,
U[F]. XII>>
=(U[F]ell,
XII» =I(U[F]ell)
=<
ell,
<I>>,
so U[F]. XII>=
CI» by Lemma 1.17.II
=>
I. Let V be a O'ind-neighboumood of 0 inSu.c. Since O'ind is a regular locally convex topology forSu.c there exists a closed absolutely convex O'ind-neighboumoodW of0 such that Wc
V. Then W=
WO0 by [Will Theorem 8-3-8, hence (W
O
r
is absorbing. So «(Wil] Theorem 8-4-12) WO
is 'tprorbounded in T
u.c .
By assumption there exist FEC:
and a bounded subsetBo
ofHsuch thatWO=
U(F]. Bo.
Then(U[F]. Bor
=
Woo=
We V. Let M>
0 be such thatIIxliSM for allx
E B o.Let~E Su,cand supposeIIU(F]~IIS M-I•Then for allx E Bo:
-
-I<~,UfF].x> I
=
I(U[F]c1l,x)I S IIU[Fl,UUxllS 1,soc1lE: (U[F].Bor c V.Then{~E:Su.c : si(')SM-IleVandVisaO'proj neighbourhood. 11.3. A
growth
conditionLet C be a subset ofLI(G)and suppose the pair (C ,U)satisfies properties PI and
n.
Inthis sec-tion we consider an addisec-tional condisec-tion on the set C. We then prove that every 'tprorbounded subsetBofTU,cis equal toU [F] • B0for someF E C; and some bounded subsetB0ofH.Suppose the set C has the following property.
PROPERTY 2.22.
P3. There exist a sequence of Borel measurable disjoint setsQ1 ,Q2, . . . in
G
and a sequence of positive real numbers bl , b2 , •.• such thatG
=
~
Qn andi
b;/ <00 and for alln=1 11=1
f
E C there existgE Canda
>
0 suchthatfor all n E INholds: bn sup{I }(y) I :yE Qnl:S;ainf{Ig(y)I : yE Qnl.This property is inspired by (EGK] Definition 1.1, A IIIandimproved by F.J.L. Martens. REMARK. Property P3 is equivalent with the following property.
P3'. There exist a sequence of Borel measurable disjoint setsQI ,Q2, ... in
G,
a sequence of00 00
positive real numbersb1 ,b2 , •.. and v
>
0 such thatG
=
u Qn and1:
b;v<
00and for,.=1 n=1
all
f
E C there existgE Canda
>
0 such that forallnE INholds:Of course, P3 implies P3'. Suppose P3' holds. Then, inductively, for allkE INwe obtain: for all
fEe
there existg E C and 3>
0 such that for all nE N holds:b~sup{ I}(y) I :YE QnlS3inf{ 19(y) I :yE Qnl. I
Takek
>-.
vREMARK.Incase the set C is supposed to satisfy property P3 then we write
Q
1 ,Q
2, . , . and b1 ,b2 , . , . forthesequences as in property P3 without explicity defining them again, if no con-fusion can arise.REMARK. The set C satisfies property P3 if and onlyifthe set
C
satisfies property P3.We start with some useful lemmas. Let (A ,m ,I , Ai ,'ti, W) be a Stone-representant for U,
which is taken fixed throughout the remaining part of this chapter. DEFINITION 2.23.
Let nE IN. Define the projection operator Pn:= U[1QJ, define the layer
L"
of Qn in A byLn:={'ti(Y):i E I,yE Qn}.
Note thatPn
=
0 if and only ifm(Ln )=
0 (nE IN).LEMMA 2.24.
Let!E C,n E IN,£
>
0and suppose If(a) I~£for a.e.a E Ln.Then:I. Pn(H)is a subset ofRf.
II. The map
ni
PnfromHintoHis bounded andIIni
PnII$1..
£III.
ni
(<I» E Pn(H)forall<I> E Pn(H).IV. U(g)
ni
Pn=ni
Pn U(g)forallg
E L1(G).Proof.
LetV :={yE
G:
}(y)=o} and letP := U[1v]. ThenPis the projection ofHontoNf .DefineA:A~€ { (}(a)rl A(a):=
o
if If(a) I ~£andaE Ln , else. (a E A)Then A is Borel measurable and II A1100$ £-1. LetMA be the multiplication operator by A on
L2(m).
I. Let xEH. Then U(f)W-1MAWX=w-l(i.A.Wx)=w-l(1L,..WX)=Pnx, So Pn(H) cRf .
II. LetXE H. ThenW-1 M A Wx E Ntsince
tv.
(A.Wx)=Oa.e. Soni
Pnx =W-1 M A Wx andni
Pn =W-1M A W.Then IIni
Pn11$ II A 1100$ £-1.III. Let cp e P,.(H).Then
P,. ni(cp)=P,. n iP,.(cp) =P,. W-l MA Wcp= W-I(lL,..A.Wcp)=W-I(A. Wcp) =
n iP,.cp = n,i(cp). So ni(cp) e P,.(H).
IV. Letx e H.ThenU(g)ni P,.x=W-I(g.A.Wx)=ni P,. U(g)x.
o
LEMMA 2.25.
For allne INwithP,. ~0 there exists! e C and£
>
0 such that I](y) I>
£for all Ye Q".Proof.
Let ne IN and suppose m(L,.)~O. There exists ie I such that m('ti(Q,.» ~O. Let ~:=1<t;(Q.)'
Then~e L2(m) and~ ~O. Let(f,j"M:Jbethe net in property
n.
SupposeA.(Y)= 0 for all A. e J and ye Q,.. Then 0,*W-l ~ = limU(f,jW-1~ = limW-l(i..~) = 0, contradiction. So there exists..
..
A.e J and ye Q,. such that J..(y)~O. By property P3 there exist ge C and
a>
0 such that b,. sup (IJ..(oo)I :00e Q,.}::;a inf{ Ig(oo) I :00e Q,.}.Then Ig(oo) I~ a-I IJ..(y) Ib,.>
0 for all00e Q,..
0
COROLLARY2.26.
Letne IN.ThenP,. (H)
c
Su,c,Proof.
IfP,.= 0, thenP,.(H)= {OJ
c
Su,c, IfP,. ~0, the corollary is proved by Lemmas 2.25 and 2.24L
0
LEMMA 2.27.
Let<J> e Tu,candn e IN.Define'¥ :C ~H
'¥(f):=P,. <J>(f) (fe C).
Then'¥e emb(H). More explicitely, suppose there exist ge C and£
>
0 such that Ig(y)I~tfor all ye Q,..Then'¥= emb(niP,. <J>(g».
Proof.
We may assume thatP,.
'*
O. By Lemma 2.25 there existg e C and£>
0 such that Ig(y) I>
e for all ye Q,.. Then Ig(y) I>
e for all ye Q". Since the pair(C ,
U) satisfies properties PI-P3, we obtain by Lemma 2.24 I thatP,.<J>(g) e Rg.So nil P,.<J>(g)is well defined.Let! e C. Then by the Lemmas 2.24N, 1.9 and 2.24 I,
[emb n iP,. <J>(g)] (f)=U(f)* n i P,. <J>(g)= n i P,. U(f)* <J>(g)= nil P,. U (g)* <J>(f) =
n i U(g)p,. <J>(f)= n i ng P,.<J>(f)=P,. <J>(f)='¥(f).So'¥=emb(niP,. <J>(g».
0
DEFINITION 2.28.
Let c;l)E TuC and nE IN. DefineP71 • c;l)E H by the relationP71c;l)(f)=[emb(P71 •c;l»)](f) forall
fE
C.LEMMA 2.29.
N
I. Let <l>E Tu,c.Then c;l)
=
limL
emb(Pn.c;l») in(Tu,c ,'rproj).N __71=1
II. P71 • emb
x
=
P71Xforallx
E Handn
E IN.III. P71 • c;l)E Pn(H)forallc;l)E Tu,cand
n
E IN.IV. The map c;l) ~ P71 • c;l) from(Tu.c ,'tproj) intoHis continuous forallnE IN.
in N
c;l)= lim
L
emb(P11 • c;l»)N-+oo11=1
So
I. Let c;l)E Tu,c,For every
fEe
we obtain in theHilbertspaceH:00 N
<l>(f)=
L
Pn <l>(f)= lim(L
emb(Pn·c;l»))(f).71=1 N __ 71=1
(Tu,C''tproj).
II. Let n E IN and x E H. Then Pn(embx)(f)=Pn U(f)'" x
=
U(f)"'Pnx=(embPnx)(f) for allfEe.
By Definition 2.28, P71 • embx=
P71X.
LetnE IN. For the proof of III andIV we may suppose that P71
*"
O. By Lemma 2.25 there existgEe
and£>
0such that Ig(y) I~£forall'fE Qn.III. Let c;l)E Tu,c.By Lemmas 2.27 and 2.24 III,P71 • c;l)
=
ni
P11c;l)(g)E Pn(H). IV. By Lemmas 2.27 and 2.24 II we obtain forall<l>E Tu,c :IIPn.c;l)1I
=lIni
Pn<l>(g)II~IInil
Pllllllc;l)(g)II~ £-1 tg(<l».So the map c;l) ~ PII • c;l) from(Tu,c ,'tproj) intoHis continuous.
o
The main result of this section is the following theorem. THEOREM 2.30.
LetBeTu,cbe a 'tprorbounded set,B
*"
0. Then there existFEe;
and a bounded setB0c H such thatU[F]. Bo
=BandU[F]. IBo:(Bo,
II II)~(B ,'tproj IB)is a homeomorphism.Proof.
Let
n
E IN. By Lemma 2.29IV, the map c;l) ~ PII • c;l) from(Tu,c ,'tproj)intoH is continuous,sorn :=sup{1IPI I .c;l)1I : c;l)E B}
<
00.DefineF :
G
~ JRF :=
L
(1+bnr
n ) lQ.'71=1
bn sup{I}(y)1:yE Qn}S~inf{1g(y) I :yE Qn} foralln E IN.
Then 1F}IS 1I}1I00
+
~ sup{tg(W) :4> EB} <00. Indeed, for all nE IN, YE Qn such that}(y)
*'
0and forall4> E Bwe obtain by Lemma 2.24II:(l+bn"Pn.4>II)1}(y) I S 1I}1100
+
bnll0 iPnllll4>(g)1I I }(y) I S IIjlloo+
bn(~-1 bnIj(y) 1)-1 •• 11W(g)1I I](y) I S IIjlloo
+
~ sup{tg('¥):'¥EB}. So the assertion is proved, since 1F(y)j(y) I=
sup{(l+bnll Pn.WII) Ij(y)1:WEB}.00 1
For all 4>EB define x~:=1:: 1 b Pn .4>EH. Let Bo:={X~:4>EB}. Then
n=1 +n'n
00 1 00 1
IIx~1I2=
1::
2 HPn.4>1I2S1::
2
<ooforall<l>E B,soBoisboundedinH.n=1 (l+bn'n) n=1 bn
Let 4> EB.ForallfE Cwe obtain by Definition 2.28:
Pn <I>(f) =Pn U(f)* (Pn.4» =Pn U[F] U(f)*x~ =Pn«U[F].x~)(f) for all nE IN. So
<1>=U[F].x~.By Lemma 2.16 V,U[F]. 180 :Bo~Bis a bijection.
Obviously U[F]. Iso:Bo~B is continuous. (See Lemma 2.16 VI.) The inverse map is also continuous. Indeed, let(<I>a)a.eMbea net inB, let WEB and suppose lim 4>a
=
4> in'tproj lB.a
00 4 E2
Assertion: limx~. =x~ inH. Let
E>
O.
Let N E IN be so large that1::
2
ST'
For alla n=N+1 bn
nE IN,nS N,andPn
*'
0,there existfn E CandEn>
0such that Ij nIIQ.~Enby Lemma 2.25. There exist 00 E M such that for all aEM, a~00, and all nE IN, nSN, Pn*'O holds:tt.
(<I>a-4»S(l+bn'n)En E(2Nrt .Then for allaEM,a~ 00 we obtain by Lemma 2.24 II:N 00
IIXII>._x~1I2
=
1::
IIPn(x~. -XII» 112+
1::
IIPn(xll>._x~)112n=1 n=N+1
So the assertion and the theorem is proved.
COROLLARY 2.31.
(Su.c, (Jind)
=
(Su.c, (Jproj)as topological vector spaces.Proof.
Theorems 2.30 and 2.21.
o
Let Tbe the subspace ofTu.e asinDefinition 2.19. COROLLARY 2.32.
T=Tu,e as sets. We have even the stronger equality:Tu,e= U U [F] • Has sets.
Fee~
The identity map from(Tu.e, 'tind) into(Tu,e,'tproj) is continuous.
In the next chapter we give conditions such that(Tu,e ,'tind) is equal to(Tu.e ,'tproj)
as
a topolog-ical vector space.COROLLARY 2.33.
LetBbe a subset ofTu.e.Then:
I. B is bounded in(Tu,e ,'tproj) if and only if there existF E C~ and boundedB0c H such thatB
=
U[F]. Bo.
II. Bis compact in(Tu,e ,'tproj) if and only if there exist
FEe:
and a compactB0c
H suchthatB=U[F].Bo.
III. B is sequentially compactin(Tu,e ,'tproj) if and only if there existF E C~ and a
sequen-tially compactB0
c
Hsuch thatB=U [F] • Bo.
IV. Bis sequentially compact in(Tu,e ,'tproj) if and only ifBis compact in(Tu,e ,'tproj)' V. Let (<I>n)neIN be a sequence in(Tu,e ,'tproj)' Then (cI>n)neIN is a Cauchy sequence if and
only if there existFEe: and a Cauchy sequence (xn)neIN inH such thatcI>n
=
U[F] • Xnfor all
n
E IN.VI. Every'tprorbounded sequence in Tu,ehas a weakly convergent subsequence.
Proof.
Every compact and every sequentially compact set is bounded. Every Cauchy sequence is bounded. ForallF E C~,the mapU[F].fromH into(Tu.e, 'tproj) is continuous. So I, II, III, V
Chapter
m.
Nice properties of the spacesSu,c and Tu,cLetGbe a locally compact Abelian group and letU bea representation ofGin a Hilbert spaceH.
Let
e
be a subset ofL1(G) which satisfies properties PI-P3. We fix the groupG, the representa-tionUand the sete
throughout this chapter.As we have seen Tu,c carries the topology 'tproj in a natural way. In the previous chapter, a second topology,'tmd, has been introduced onT
u,
c. (Note that we used property P3 to define the topology 'trod.) It turns out that the assertion (Tu,
c ,'tind)=(Tu,
c ,'tproj) as topological vector spaces is equivalent withthe assertion thatSu,c is complete. As we shall see it is also equivalent with a lot more assertions suchas:
(Tu,
C ,'tproj) is bomological; (Tu,
c ,'tproj) is reflexive;Su,c = ("\ D(U[F]). In particular it turns out to be equivalent with the following property of
Fee;
the pair(e ,U).(Cf.
[tEl
property AIV'.) PROPERTY 3.1. P4.'VFeBor(G,C) [ ('VKeC' (F.K is bounded]) =>3feC
3
c>o(U[lheG:IF(y)1 >clj(y)l}l = 0] .REMARK. It is trivial that the following property implies property P4.
P4'. 'VFeBor(G,C) [('VKec' [F .Kisbounded]) =>3feC 3c>o [1 F I:s;c 1
j
Il] .REMARK. The pair(e , U)satisfies property P4 resp. P4' if and only if the pair
(C ,
U)satisfies property P4 resp. P4'.LEMMA 3.2.The following conditions are equivalent. I. Property P4 holds.
II. 'VFeBor(O,C) [('VKec; [F.Kisbounded])=>
3
fec3
c>o[U[lheG:IF(Y)I>Clj(Y)I}]=O~.
m.
'VFeBor(G,C) [('VKec' [F.Kis bounded])=> 3feC3c>o"t~eH
[IIU{F]xll:S;c
BU(f)X"~
. IV. "tFeBor(O,C) [('VKec' [F.Kis bounded]) =>3feC 3c>o
'V~eH
'VyeHWI U(f)xll::;; 1 A lIyII::;; 1)=> J(U[F]y,x) I::;;c~
.Proof. Note that a functionF E Bor(G ,C)is bounded ifF • K is bounded for allKEeN,since
10 E eN.So
m.
and IV. make sense.II=>I. Trivial.
I=> III. LetF E Bor(G , €) and suppose F • K is bounded for all KEeN. By property P4 there exist fEe and
c
>
0 such that U[ly]=
0 withY:= (rE
G:
IF(r) I>c Ij(r)I}. DefineL: G ~ JRo ifj(r)
=
0 OryE Y,L(r)
=
(rEG)
jg;
ifj(r)*
0 and rE Y.ThenLE Borb(G,€)andIIL1100$;c. Let
x
E H.By Proposition2.8:II U[F]x II=II U[ly.] U[F]xll=11U[ly. F]xll = II U[Lilxll$; II U[L]II IIU rhx II$;
c IIU(f)xll.
III => I. LetF E Bor(G, €) and supposeF. K is bounded for all KEeN. By assumption,
there exist fEe and c
>
0 such that II U [F ]x II $; c II U (f)x II for all x E H. Let(A ,
m ,
I , Ai ,ti ,W)be a Stone-representant forU.Let e
>
0 and define Z£:=(aE A : IF(a)I>
2c If(a) I>
e}. Suppose m(Z£)*
O.There existiE I and
n
E IVsuch that0<
m(Z£I'lti(QII»<
00.LetAI :=Z£n ti(QII)'Define~:A ~€ I
~
~(a):=o
ifa
E Al , (aE A)Then ~ is bounded, so ~EL2(m). But O*2cIlIAIII$;IIF;II=IIU[F]W-I~II$; ellU(f) W-I ;II
=
c IIIAIII. Contradiction. Hencem(Z£)=
O.Similar, define Zo :=(a EA : IF(a) I
>
0 and If(a) I=OJ. Thenm(Zo)=O. Define00
Y:=(rE G : IF(r)I
>
2c Ij(r)I}. ThenZo u U ZII-I=U ti(Y)' hence U[ly]=O. 11=1 ielSo III impliesI.
III=>IV.LetFE Borb(G. C), fEeandc >0 and suppose thatII U[F]xll$; ellU(f)xll forall
x E H. Let x, y E H and suppose that II U(f)xll$; I and lIyll$; 1. Then
I(U[F]y,x)1= l(y, U[F]* x)I$;lIyIlIlU[F]* xll=