102
FUNCTIONAL ANALYSIS ON NORMED SPACES: THE BANACH SPACE COMPARISON 103
As shown in [12], the category AC arises also in so called “Quantified Functional Anal-
ysis”. We use this term for the theory of (locally convex) approach spaces as developed
e.g. in [11, 12], where the key idea is to work with canonical numerical structures over-
lying (locally convex) vector topologies, instead of only considering the topological level.
For more information we refer to [11, 12]. In [12] we proved that AC is also the category
of Eilenberg-Moore algebras for the monad arising from a dual adjunction having the
dualization functor HomlcApVec (−, R) on lcApVecop as a right adjoint. Here lcApVec
denotes the category of locally convex approach spaces and contractive linear maps, and
R is equipped with the absolute value as norm.
If we define C : sNormop 1 → Set to be the restriction of HomlcApVec (−, R) to
sNormop 1 , then C is in fact the dualization functor HomsNorm1 (−, R) which is nothing
but the closed dual unit ball functor on sNormop 1 . Again C is the right adjoint of a (dual)
adjunction, giving rise to a monad. The question we address in this paper is finding a
description of the category of Eilenberg-Moore algebras for this monad. Speaking more
loosely as we did above this answers the question of which canonical algebraic structure
is present on the closed dual unit ball of a seninormed space.
From the previous discussion one might be tempted to guess that this category of
Eilenberg-Moore algebras is concretely isomorphic with TC. Quite surprisingly this is
not the case as we shall prove. The resulting category SC of Eilenberg-Moore algebras
has as objects sets which allow for an abstract integration with respect to certain finitely
additive measures, also called charges, of total variation at most one in the sense of [9].
The crucial ingredient to obtain this result is a Riesz-type representation theorem for
charges, to be found in [9] which we state in precise terms later on.
2. Basic definitions
Let S be a set and let F be a field of subsets of S, i.e. a collection F ⊂ 2S satisfying the
following axioms:
1. ∅, S ∈ F,
2. A ∈ F ⇒ Ac ∈ F,
3. A, B ∈ F ⇒ A ∪ B ∈ F.
A bounded charge on (S, F) is an additive map α : F → R, i.e. a map satisfying
α(A ∪ B) = α(A) + α(B) for all A, B ∈ F with A ∩ B = ∅ (note that then α(∅) = 0
follows), such that the total variation
n
X
||α|| := sup{ |α(Ai )| | {A1 , . . . , An } ⊂ F finite partition of S}
i=1
is finite. We write ba(S, F) for the space of all bounded charges on (S, F), equipped with
the total variation norm defined above. To ease the notation we put ba(S) instead of
ba(S, 2S ),
104 M. SIOEN, S. VERWULGEN
A simple function on (S, F) is a map f : S → R for which there exist a finite partition
{A1 , . . . , An } ⊂ F of S and scalars a1 , . . . , an ∈ R, such that
n
X
f= ai 1Ai .
i=1
We put B(S, F) for the closure, with respect to the supremum–norm kf k∞ := sups∈S |f (s)|,
of the set of simple functions on (S, F) in the space of all bounded functions from S to
R. The space (B(S, F)) is equipped with the norm || − ||∞ ). If F is the collection of all
subsets of S, we write BS instead of B(S, F), which in this case equals the space of all
bounded real valued functions on S. R
For α ∈ ba(S, F) and a ∈ B(S, F) the integral s∈S a(s)dα(s) can be introduced as
the limit of the integrals of a sequence of simple functions converging uniformly to a [9],
where the integral of a simple function is defined in the obvious way:
Z X n n
X
ai 1Ai (s)dα(s) := ai α(Ai ).
s∈S i=1 i=1
We will use the notational convention to write OX for the closed unit ball of a semi-
normed space X and CX for the closed dual unit ball, i.e.
In the sequel LC X denotes the topological dual of X, equipped with the dual norm
1
kϕkC := inf{k > 0 | ϕ ∈ CX}.
k
The following representation theorem identifies the dual space of (B(S, F), || − ||∞ ).
2.1. Theorem. [9, 1] The assignment
γ(S,F ) : ba(S, F) −→ LRC (B(S, F), || − ||∞ )
(1)
α 7−→ s∈S
ev(−, s)dα(s),
R R
with s∈S ev(−, s)dα(s) (a) := s∈S a(s)dα(s) is a linear isometry.
Another notational convention is to put T S for the closed unit ball of ba(S). From
Theorem 2.1 we deduce that the map
γS : T S = Oba(S, 2S ) −→ CBS
R (2)
α 7−→ s∈S ev(−, s)dα(s)
is a one-one correspondence.
It is easy to verify that we have an endofunctor
f Tf
T : Set −→ Set : (S1 → S2 ) 7−→ (T S1 → T S2 ) (3)
α 7−→ αf ,
FUNCTIONAL ANALYSIS ON NORMED SPACES: THE BANACH SPACE COMPARISON 105
exists and is finite. One moreover shows, in the same way as in any standard measure
theory course, that the assignment
Z
α(−)dβ(α) : F −→ R
α∈Oba(S,F )
Z
A 7−→ α(A)dβ(α).
α∈Oba(S,F )
with Cf (ϕ) := ϕ ◦ f (ϕ ∈ CY ). It immediately follows from the fact that the product
Q
s∈S (R, | |) in the category sNorm1 is equal to BS, that the map
i.e. σS2 ◦ CBf = T f ◦ σS1 . Since σS1 and σS2 are bijections, this follows since, for
each α ∈ T S1 and ϕ ∈ BS2 , the following string of equalities hold:
(CBf ◦ γS1 )(α)(ϕ) = CBf (σS−1 1
(α))(ϕ)
= σS−1 (α)(ϕ ◦ f )
Z 1
= ϕ ◦ f dα
S1
Z
= ϕdαf
S2
= γS2 (αf )(ϕ)
= (γS2 ◦ T f )(α)(ϕ).
is commutative.
2
σS
*
Bearing in mind the factorization T 0 2 S / T 0T S / T 2 S , we have to verify
T 0 σS σT S
so
Z
0
(µS ◦ σT S ◦ T σS )(Φ)(A) = α(A)d(σT S ◦ T 0 σS )(Φ)(α)
Zα∈T S
= ev(−, A)(α)dσT S (T 0 σS (Φ))(α)
α∈T S
= T 0 σS (Φ)(ev(−, A))
= Φ(ev(−, A) ◦ σS ).
ev(−, A) ◦ σS = ev(−, 1A ).
Fact 4. For the moment we only know that T0 = (T 0 , η 0 , µ0 ) is a monad. However, from
a lengthy yet straightforward categorical computation [13] it follows that this
information, together with Facts 1—3 suffice for the triple T = (T, η, µ) to be a
monad too. Moreover, σ : T0 → T is an isomorphism of monads.
FUNCTIONAL ANALYSIS ON NORMED SPACES: THE BANACH SPACE COMPARISON 109
Fact 5. It is another straightforward exercise in category theory to show that the assign-
ment
0 f f
Iγ : SetT −→ SetT : (X, h) → (Y, k) 7−→ (X, h ◦ γX ) → (Y, k ◦ γY ) (11)
is an isomorphism of categories.
Let T00 be the monad induced by the adjunction l1 a O. We will proceed by providing an
00
explicit description of E. By definition, an object in AlgT is a pair (N, EN ) consisting
of a set N and a structure map
EN : Ol1 N −→ N
4.1. Definition. [7] A totally convex structure is a set M together with, for each
α = (αi )i∈N0 ∈ Ol1 N0 , an operation α
b : M N0 → M , such that, with the notation
∞
X
αi xi := α
b((xi )i∈N0 ),
i=1
A totally affine map f : M → N between totally convex modules is a map between the
underlying sets such that, for all (αi )i ∈ Ol1 N0 and for all (xi )i ∈ M N0 ,
X∞ ∞
X
f( αi xi ) = αi f (xi ).
i=1 i=1
The category of totally convex modules is denoted by TC and we put U : TC −→ Set for
the forgetful functor.
FUNCTIONAL ANALYSIS ON NORMED SPACES: THE BANACH SPACE COMPARISON 111
The closed unit ball of a Banach space X is in a pointwise way a totally convex module,
which is denoted OX.
b Moreover, if f : X → Y is a linear non–expansive map between
Banach spaces, then f |OX : OXb → OY b is a totally affine map between totally convex
modules.
Let M be a totally convex module. We put
∞
X
EM (a) := a(xi )xi , (a ∈ Ol1 M ) (12)
i=1
where (xi )i∈N0 is a sequence in M such that {xi | i ∈ N0 } contains the support of a.
00
Conversely, let (N, EN ) be in AlgT . Then we define
∞
X
αi xi := EN (a), (13)
i=1
(
0 if x ∈
/ {xn | n ∈ N0 }
a(x) := P
i:xi =x αi otherwise.
S◦O
b ' id
Ban1 . (14)
with V : SC → Set the canonical forgetful functor. Suppose that we could find a concrete
functor E : SC → TC with the additional property that the square formed in diagram
(15) commutes, then from (14) we see that
LC ' S ◦ E ◦ C.
b
In other words, such E would yield a factorization of the dualisation LC via the natural
dual algebraic structure. The sequel of this section is devoted to the comparison of SC,
the algebraic theory of dual unit balls, with TC, the algebraic component of Banach
spaces.
Let (M, IM ) be a space of charges and fix a ∈ Ol1 M . Then we put
X
a := a(x)δx ∈ T M.
x∈M
As the closed unit ball of ba(M ), T M carries a natural totally convex structure, so this
assignment is well defined. Now we define
E IM (a) := IM (a),
E : SC → TC
Hence
X X
E IM ( B(a)a) = IM ( B(a)a)
a∈Ol1 M a∈Ol1 M
Z
= IM α(−)dβ(α)
α∈T M
= IM (T IM (β))
X X
= IM ( B(a)δx )
x∈M a∈Ol1 M
E IM (a)=x
X X
= IM ( ( B(a))δxf )
x∈M a∈Ol1 M
E IM (a)=x
X X
= E IM
( B(a))δxf
.
x∈M a∈Ol1 M
E IM (a)=x
Tf f
TN /N
IN
114 M. SIOEN, S. VERWULGEN
(f ◦ E IM )(a) = (f ◦ IM )(a)
= IN (T f (a))
= IN (Ol1 f (a))
= (E IN ◦ Ol1 f )(a),
where the last but one equality is true because, as one easily verifies, we have T f (a) =
Ol1 f (a).
4.4. Corollary. If (M, IM ) and (N, IN ) are isomorphic spaces of charges then (M, EIM )
and (N, EIN ) are isomorphic totally convex structures.
From the above remark, in combination with the following theorem, it is noted that
the SC theory is of strictly stronger nature than the theory of totally convex modules.
4.5. Theorem. The categories SC and TC (with their canonical forgetful functors) are
not concretely isomorphic.
Proof. It is well–known from the general theory of monads (see e.g. [3, 6]) that the
assignment F S := (T S, µS ) (S ∈ |Set|) defines a functor F : Set −→ SC that is left
adjoint to V : SC −→ Set. Now suppose SC and TC were concretely isomorphic.
Since adjunctions are determined up to natural isomorphism, this would imply that the
underlying sets of the free TC–object on R (i.e. Ol
b 1 R) and the free SC–object on R (i.e.
F R) would have the same cardinality. Now on the one hand we see that
On the other hand, we can define for every ultrafilter U on R a charge αU that is an
element of T R by (
1 if A ∈ U,
αU (A) :=
0 otherwise,
and it is easy to see that αU 6= αV if U =
6 V. If R is equipped with the discrete topology,
it therefore follows from [2], Theorem 9.2, that
R
#T R = #Oba(R, 2R ) ≥ #β(R) = #22 > #2R .
It is always nice to have the dual of a normed space represented by a concrete Banach
space. This representation puts, often in a canonical way, an SC–structure on the closed
unit ball of that Banach space. If we apply the forgetful functor E, we see from the
commutation of diagram (15) that we then recover the pointwise TC–structure on the
closed unit ball of the Banach space. As an example we reconsider the representation
theorem 2.1 in this context.
R
4.6. Theorem. The pair (Oba(S, F), Oba(S,F ) ) is an SC–object and
Z
γ(S,F ) : (Oba(S, F), ) −→ CB(S,
b F)
Oba(S,F )
is an SC–isomorphism.
Proof. We only have to show that γ(S,F ) is an SC–morphism, that is, we have to show
the commutation of the diagram
R
T Oba(S, F)
Oba(S,F )
/ Oba(S, F)
T γ(S,F ) γ(S,F )
T CB(S, F)I / CB(S, F).
CB(S,F )
Take β ∈ T Oba(S, F) and a ∈ B(S, F). Then on the one hand we have that
ICB(S,F ) (T γ(S,F ) (β)) (a) (16)
= ICB(S,F ) (βγ(S,F ) ) (a) (17)
Z
= ev(−, a)(ϕ)dβγ(S,F ) (ϕ) (18)
ϕ∈CB(S,F )
Z
= ev(−, a) ◦ γ(S,F ) (α)dβ(α) (19)
α∈T Oba(S,F )
Z Z
= a(s)dα(s) dβ(α). (20)
α∈T Oba(S,F ) s∈S
We are therefore done if we show that (20) and (23) are equal. LetP (an )n be a sequence
of F–simple functions, converging uniformly to a. We write an = m n n
i=1 ai 1Ai , with all
n
n n n n
ai ∈ R and all Ai ∈ F such that all {A1 , . . . , Amn } are a partition of S. Then
Z Z
a(s)d α(−)dβ(α) (s)
s∈S α∈Oba(S,F )
Z Z
= lim an (s)d α(−)dβ(α) (s)
n→∞ s∈S α∈Oba(S,F )
m
X n Z
= lim ani α(Ani )dβ(α)
n→∞ α∈Oba(S,F )
i=1
Z mn
!
X
= lim ani α(Ani ) dβ(α)
n→∞ α∈Oba(S,F )
Z Zi=1
= lim an (s)dα(s) dβ(α)
n→∞ α∈Oba(S,F ) s∈S
Z Z
= an (s)dα(s) dβ(α),
α∈Oba(S,F ) s∈S
R
where the last step isR valid because the sequence (α 7−→ s∈S
an (s)dα(s))n is uniformly
convergent to α 7−→ s∈S a(s)dα(s).
That Theorem 4.6 is indeed a strengthening of the Riesz–type representation theorem
2.1 is easily seen now since the latter can be obtained as a simple corollary using the
commutative diagram 15.
4.7. Corollary. (ba(S, F), k k) and LC B(S, F) are isomorphic Banach spaces.
This is of course not surprising since Corollary 4.7 has served as a starting point for
our theory.
References
[1] N. Dunford and J. T. Schwartz. Linear operators Part I, volume VII of Pure and
applied mathematics. Interscience publishers, 1971.
[2] L. Gillman and M. Jerison. Rings of continuous functions. The university series in
higher mathematics. Van Nostrand company, 1966.
[3] S. Mac Lane. Categories for the Working Mathematician. Springer, 1998.
[4] J. W. Negrepontis. Duality in analysis from the point of view of triples. Journal of
Algebra, 19:228–253, 1971.
[5] H. E. Porst and W. Tholen. Concrete dualities. Category Theory at Work, pages
111–136, 1991.
FUNCTIONAL ANALYSIS ON NORMED SPACES: THE BANACH SPACE COMPARISON 117
[7] D. Pumplün and H. Röhrl. Banach spaces and totally convex spaces i. Communica-
tions in Algebra, 12(8):953–1019, 1984.
[8] D. Pumplün and H. Röhrl. Banach spaces and totally convex spaces ii. Communi-
cations in Algebra, 13(5):1047–1113, 1985.
[9] K. B. Rao and M. B. Rao. Theory of charges. Pure and Applied Mathematics.
Academic Press, 1983.
[11] M. Sioen and S. Verwulgen. Locally convex approach spaces. Applied General Topol-
ogy, 4(2):263–279, 2003.
[12] M. Sioen and S. Verwulgen. Quantified functional analysis and seminormed spaces:
a dual adjunction. Journal of pure and applied algebra, 207:675–686, 2006.
Mark Sioen
Department of Mathematics, VUB
Pleinlaan 2
B-1050 Brussels
Stijn Verwulgen
Department of Mathematics and Computer science, University of Antwerp
Middelheimlaan 1
B-2020 Antwerp
Email: msioen@vub.ac.be
stijn.verwulgen@ua.ac.be
This article may be accessed at http://www.tac.mta.ca/tac/ or by anonymous ftp at
ftp://ftp.tac.mta.ca/pub/tac/html/volumes/18/3/18-03.{dvi,ps,pdf}
THEORY AND APPLICATIONS OF CATEGORIES (ISSN 1201-561X) will disseminate articles that
significantly advance the study of categorical algebra or methods, or that make significant new contribu-
tions to mathematical science using categorical methods. The scope of the journal includes: all areas of
pure category theory, including higher dimensional categories; applications of category theory to algebra,
geometry and topology and other areas of mathematics; applications of category theory to computer
science, physics and other mathematical sciences; contributions to scientific knowledge that make use of
categorical methods.
Articles appearing in the journal have been carefully and critically refereed under the responsibility of
members of the Editorial Board. Only papers judged to be both significant and excellent are accepted
for publication.
Full text of the journal is freely available in .dvi, Postscript and PDF from the journal’s server at
http://www.tac.mta.ca/tac/ and by ftp. It is archived electronically and in printed paper format.
Subscription information. Individual subscribers receive abstracts of articles by e-mail as they
are published. To subscribe, send e-mail to tac@mta.ca including a full name and postal address. For in-
stitutional subscription, send enquiries to the Managing Editor, Robert Rosebrugh, rrosebrugh@mta.ca.
Information for authors. The typesetting language of the journal is TEX, and LATEX2e
strongly encouraged. Articles should be submitted by e-mail directly to a Transmitting Editor. Please
obtain detailed information on submission format and style files at http://www.tac.mta.ca/tac/.
Managing editor. Robert Rosebrugh, Mount Allison University: rrosebrugh@mta.ca
TEXnical editor. Michael Barr, McGill University: mbarr@barrs.org
Transmitting editors.
Richard Blute, Université d’ Ottawa: rblute@uottawa.ca
Lawrence Breen, Université de Paris 13: breen@math.univ-paris13.fr
Ronald Brown, University of North Wales: r.brown@bangor.ac.uk
Aurelio Carboni, Università dell Insubria: aurelio.carboni@uninsubria.it
Valeria de Paiva, Xerox Palo Alto Research Center: paiva@parc.xerox.com
Ezra Getzler, Northwestern University: getzler(at)math(dot)northwestern(dot)edu
Martin Hyland, University of Cambridge: M.Hyland@dpmms.cam.ac.uk
P. T. Johnstone, University of Cambridge: ptj@dpmms.cam.ac.uk
G. Max Kelly, University of Sydney: maxk@maths.usyd.edu.au
Anders Kock, University of Aarhus: kock@imf.au.dk
Stephen Lack, University of Western Sydney: s.lack@uws.edu.au
F. William Lawvere, State University of New York at Buffalo: wlawvere@acsu.buffalo.edu
Jean-Louis Loday, Université de Strasbourg: loday@math.u-strasbg.fr
Ieke Moerdijk, University of Utrecht: moerdijk@math.uu.nl
Susan Niefield, Union College: niefiels@union.edu
Robert Paré, Dalhousie University: pare@mathstat.dal.ca
Jiri Rosicky, Masaryk University: rosicky@math.muni.cz
Brooke Shipley, University of Illinois at Chicago: bshipley@math.uic.edu
James Stasheff, University of North Carolina: jds@math.unc.edu
Ross Street, Macquarie University: street@math.mq.edu.au
Walter Tholen, York University: tholen@mathstat.yorku.ca
Myles Tierney, Rutgers University: tierney@math.rutgers.edu
Robert F. C. Walters, University of Insubria: robert.walters@uninsubria.it
R. J. Wood, Dalhousie University: rjwood@mathstat.dal.ca