Tambara-Yamagami, loop groups, bundles and KK-theory (Q6046526)

This article interprets some constructions related to conformal field theory using K-theory and KK-theory of C*-algebras, continuing previous work by the authors in the same direction. The starting point is to interpret the fusion ring of a conformal field theory as a K-theory group. Modular invariants are endomorphisms of the fusion ring, so that they may be represented by KK-elements. Some examples of modular invariants have rather nice descriptions as KK-elements associated to ``correspondences''. The main examples covered in this article are the Tambara-Yamagami categories. These are approached in different ways, including a Potts model that interprets their nontrivial pieces of data through certain endomorphisms of the Cuntz algebra, and proving a reconstruction theorem for them, which shows that they may be obtained from a vertex operator algebra or a conformal net. The article also classifies the modular invariants in certain situations, using simpler data such as a subgroup with a 2-cocycle on it. A few points in the article are not completely precise. First, the Universal Coefficient Theorem in~(2.2) only applies if~\(A\) belong to the bootstrap class; fortunately, this is the case for all the examples considered in the article. There currently is no geometric description of the KK-theory \(KK(C^*(G),C^*(H))\) for two groupoid C*-algebras that would include a description of the Kasparov product. One may describe this KK-group more geometrically using Poincaré duality to replace it by an ordinary K-theory group of the tensor product of \(C^*(H)\) with the Poincaré dual of \(C^*(G)\). For many groupoids, there are nice models for this dual, so that we get a rather concrete description of the KK-group. This may, nevertheless, be of limited use because it does not consider the Kasparov product. In the examples discussed in Section~2, all the groupoid C*-algebras are finite-dimensional. In this case, it is known and rather elementary to prove that KK-elements may be replaced by \(\mathbb Z/2\)-graded finite-dimensional bimodules: any finite-dimensional bimodule carries an inner product making it a Hilbert bimodule, and we may take the zero Fredholm operator. Up to homotopy, this is the only choice we have. It is rather easy to translate the correspondences described in the article, such as in~(2.6) into bimodules, so that they give KK-elements. An advantage of using bimodules is that they form the arrows of a bicategory in a natural way, with bimodule maps as 2-arrows. This extra structure is, in fact, used in Section~5.5 to describe the associator in the fusion category. An important theme in the article is the fusion ring structure, which involves a product on a K-group. Unless the underlying C*-algebra is commutative, there is no canonical choice of such a product. The authors show how the product may be induced by correspondences or bimodules. They do not discuss, however, why a product defined in this way is associative. This may, of course, be checked by hand in examples, but producing more examples where this works may depend on a better understanding of when products defined in this way are associative.