Simplified proof of the Theorem of Varopoulos in the commutative case
Marco Thill

TL;DR
This paper provides a simplified proof of the Theorem of Varopoulos for commutative Banach *-algebras by analyzing the continuity of bitraces using the Closed Graph Theorem.
Contribution
It introduces a streamlined proof of the Varopoulos theorem in the commutative setting through continuity properties of bitraces.
Findings
Simplified proof of the Varopoulos theorem in the commutative case.
Establishment of continuity properties of bitraces on Banach *-algebras.
Application of the Closed Graph Theorem to the proof.
Abstract
We give continuity properties of bitraces on (possibly non-commutative) Banach *-algebras based on the Closed Graph Theorem, leading to a simplified proof of the Theorem of Varopoulos in the commutative case.
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsAdvanced Operator Algebra Research · Advanced Topics in Algebra · Holomorphic and Operator Theory
Simplified Proof of the Theorem of Varopoulos in the Commutative Case
Marco Thill
bd G. - D. Charlotte 53, L - 1331 Luxembourg City, Europe
Abstract.
We give continuity properties of bitraces on (possibly non-commutative) Banach -algebras based on the Closed Graph Theorem, leading to a simplified proof of the Theorem of Varopoulos in the commutative case.
Key words and phrases:
Varopoulos, commutative, positive linear functional, bitrace, continuity
2000 Mathematics Subject Classification:
Primary: 46K05; Secondary: 46K10
Many thanks to Torben Maack Bisgaard for his excellent comments
1. Introduction
The Theorem of Varopoulos [8] says that a positive linear functional on a Banach -algebra with continuous involution and with a bounded one-sided approximate unit is continuous. Its proof depends on two closely related Factorisation Theorems, one by Paul J. Cohen, and the other by Nicholas Th. Varopoulos himself. We quote them here:
Theorem 1** (Paul J. Cohen [2]).**
Let be a Banach algebra with a bounded one-sided approximate unit. Then every element of is a product of two elements of .
Theorem 2** (Nicholas Th. Varopoulos [8]).**
Let be a Banach algebra with a bounded left approximate unit. Then for a sequence in converging to [math], there exists a sequence in converging to [math] as well as with for all .
Proofs of the above two results can also be found in the following books: [9, § 6], [1, § 11], and [4, Section 5.2].
Satish Shirali [6] extended the Theorem of Varopoulos to the case of a Banach -algebra with a bounded one-sided approximate unit (and with a possibly discontinuous involution). (This was non-trivial also because Varopoulos used the fact that a bounded left approximate unit in a Banach -algebra with continuous involution gives rise to a bounded right approximate unit.) Proofs of the Theorem of Varopoulos and Shirali can also be found in the books [5, Theorem 11.3.7 pp. 1189–1190] and [7, § 27].
We shall show how to prove the Theorem of Varopoulos in the commutative case without using the more technical Factorisation Theorem of Varopoulos. This will depend on continuity properties of bitraces on (possibly non-commutative) Banach -algebras. These continuity properties are based in a novel way on the Closed Graph Theorem.
The Factorisation Theorem of Varopoulos heavily depends on additional bounds on the factors in Cohen’s Factorisation Theorem. These additional bounds are hard to establish, and are not needed for our proof. Furthermore, a number of simpler proofs for Cohen’s Factorisation Theorem without these sharp bounds have been given. See the paragraph preceding the statement of [4, Theorem 5.2.2] and the references given there.
2. Terminology and Notation
We use the terminology of [3, Chapter XV, Sections 6 & 7]. In the following, let be a -algebra, and let be a positive Hilbert form on , that is, a positive semi-definite sesquilinear form on such that
[TABLE]
holds for all . For example, any positive linear functional on induces a positive Hilbert form on by putting
[TABLE]
See [3, pp. 348–349].
The isotropic subspace
[TABLE]
is a left ideal. The quotient space
[TABLE]
is a pre-Hilbert space with respect to the inner product
[TABLE]
where shall denote the projection of . Since is a left ideal, it follows that for an operator mapping to is well defined. We denote the completion of by , which is a Hilbert space.
If is a Banach -algebra, then the operators are bounded, and thus have a continuation to , which we denote by the same symbol. It follows that the GNS representation
[TABLE]
is bounded. For proofs, see for example [7, Theorem 22.14 p. 83].
One says that is a bitrace, if furthermore
[TABLE]
holds for all . One then also has
[TABLE]
for all , and it is this property that we shall need in lemma 3. For example, if is induced by a positive linear functional , then is a bitrace if and only if
[TABLE]
holds for all . One then says that is a trace. See [3, pp. 356–358].
As usual, let .
The pair is called a Hilbert algebra, if it is a pre-Hilbert space with dense, if is a bitrace, and if the operators all are bounded, cf. [3, p. 358].
In the following, we shall let be a Banach -algebra, and let denote the projection on the closed subspace of given by
[TABLE]
3. The Proofs
Lemma 3**.**
Let be a bitrace on a Banach -algebra . The mapping
[TABLE]
(with notation as above) is bounded.
Proof.
We shall show that the mapping in question has closed graph. So let in , and in . We have to show that . The vector lies in . It is thus enough to prove that for all . One calculates
[TABLE]
by continuity of . ∎
A positive Hilbert form on a normed -algebra is called bounded if there exists such that
[TABLE]
holds for all .
Theorem 4**.**
A bitrace on a Banach -algebra is bounded if the corresponding GNS representation is non-degenerate. The last condition is satisfied, if is a Hilbert algebra.
From the lemma also follows:
Theorem 5**.**
A bitrace on a Banach -algebra is bounded on .
Theorem 6**.**
A bitrace on a Banach -algebra with a bounded one-sided approximate unit is bounded.
Proof.
This follows now from the Factorisation Theorem of Cohen (theorem 1 above). ∎
Theorem 7**.**
Let be a Banach -algebra with a bounded right approximate unit . Let be a non-zero bitrace on . The corresponding GNS representation then is cyclic. Furthermore, there exists a cyclic vector for , such that the positive linear functional induces the bitrace . The vector can be chosen to be any adherence point of the net in the weak topology.
(We give the proof here for the sake of staying self-contained. A more general result is given in [7, Theorem 27.4]. See also [7, Theorem 27.5].)
Proof.
By boundedness of (theorem 6), the net is bounded in , and therefore has an adherence point in the weak topology on . Let be any such adherence point. By going over to a subnet, we can assume that converges weakly to . It shall be shown that for all , which implies that is cyclic for . It is enough to show that for , one has . So one calculates
[TABLE]
where we have used the continuity of . Consider now the positive linear functional on defined by
[TABLE]
We have
[TABLE]
whence, after polarisation,
[TABLE]
that is, induces the bitrace . ∎
A positive linear functional on a normed -algebra is called representable, if there exists a non-degenerate continuous representation of in a Hilbert space, and a vector such that
[TABLE]
holds for all . The functional then is continuous by the continuity assumption on .
It is known that a positive linear functional on a Banach -algebra is representable, if and only if it has finite variation, in the sense that there exists such that
[TABLE]
holds for all . See for example [7, §§ 24–25].
Theorem 8**.**
A trace on a Banach -algebra with a bounded one-sided approximate unit has finite variation. Hence it is representable, and thus also continuous.
Proof.
Let be a trace on a Banach -algebra with a bounded one-sided approximate unit. In case of a bounded right approximate unit, it follows from theorem 7 and the Factorisation Theorem of Cohen (theorem 1 above), that is representable. Then has finite variation and is Hermitian in the sense that for all . In case of a bounded left approximate unit, one changes the multiplication in to , and one applies the preceding case of a bounded right approximate unit. Using the fact that turns out to be Hermitian, it follows again that has finite variation. ∎
Corollary 9**.**
A positive linear functional on a commutative Banach -algebra with a bounded one-sided approximate unit has finite variation. Hence it is representable, and thus also continuous.
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1[1] F. F. Bonsall, J. Duncan. Complete Normed Algebras, Ergebnisse der Mathematik und ihrer Grenzgebiete, Band 80, Springer-Verlag, Berlin Heidelberg New York, 1973.
- 2[2] P. Cohen, Factorization in group algebras, Duke Math. J., 26, pp. 199–205, 1959.
- 3[3] J. Dieudonné. Treatise on Analysis, vol. 2, enlarged and corrected printing, Academic Press, 1976.
- 4[4] Th. W. Palmer. Encyclopedia of Mathematics and its Applications, vol. 49: Algebras and Banach Algebras, Cambridge University Press, 1994.
- 5[5] Th. W. Palmer. Encyclopedia of Mathematics and its Applications, vol. 79: Banach algebras and the General Theory of ∗ * -Algebras, Cambridge University Press, 2001.
- 6[6] S. Shirali. Representability of positive functionals, J. London Math. Soc., 2 (3), pp. 145–155, 1971.
- 7[7] M. Thill. Introduction to Normed ∗ * -Algebras and their Representations, 5th ed., ar Xiv:math/0701306 v 2 [math.RT], 2007.
- 8[8] N. Th. Varopoulos. Sur les formes positives d’une algèbre de Banach, Comptes Rendus Acad. Sc. Paris, tome 258, Groupe 1, pp. 2465–2467, 1964.
