kk-theory for Banach algebras. I: The non-equivariant case (Q2345410)

Bivariant \(K\)-theory works best for \(C^*\)-algebras. On the one hand, Kasparov's definition allows to write down interesting cycles and homotopies starting, say, from elliptic differential operators. On the other hand, the theory has nice functorial properties such as a characterisation by a universal property. The combination of both properties is very powerful. Several authors have defined analogues of bivariant \(K\)-theory for other classes of algebras. This usually loses one of the two important features of bivariant \(K\)-theory for \(C^*\)-algebras. Vincent Lafforgue defined a Banach algebra bivariant \(K\)-theory by concrete cycles like in Kasparov theory. This allowed him to prove the Baum-Connes conjecture in certain important new cases, using that certain important homotopies that cannot work in the \(C^*\)-algebra setting do exist in his Banach algebra theory. His theory lacks, however, basic features such as a composition product. Joachim Cuntz defined a bivariant \(K\)-theory for topological algebras and other categories of algebras so that it has a universal property like in the \(C^*\)-algebra case. The cycles for his theory are quite different from the ones in Kasparov theory, however, and arguments like Lafforgue's cannot be done in his framework. This article defines a bivariant \(K\)-theory for Banach algebras that combines the positive features of the existing approaches. On the one hand, it receives a map from Lafforgue's bivariant \(K\)-theory. On the other hand, it has a composition product, a universal property, and a triangulated structure. The bivariant \(K\)-theory is defined by localising a suitable triangulated category of Banach algebras at certain equivalences to make it Morita invariant and split-exact. The article starts with a discussion of Morita equivalence for possibly degenerate Banach algebras; this generalises the theory of bimodule pairs by Lafforgue, which is only intended for Banach algebras \(A\) with \(A\cdot A=A\). This leads to a homotopy category of Morita morphisms between Banach algebras. It is shown that \(K\)-theory is a functor on this category. Next the basic setup of homology theories is extended from \(C^*\)-algebras to Banach algebras, and it is shown that it has some desirable properties. Morita invariance, half-exactness and homotopy invariance imply Bott periodicity, and Morita invariance and split exactness imply homotopy invariance; the proofs of these deep results for other categories carry over to the Banach algebra setting. Then the author carries over the Spanier--Whitehead construction that formally inverts the suspension operation. This leads to a triangulated category that incorporates the homotopy theory of Banach algebras. This category is localised at suitable classes of morphisms to put in exactness and Morita invariance properties. The resulting category is the desired bivariant \(K\)-theory for Banach algebras. It is a triangulated category and has a nice universal property by construction. This theory deserves to be called a bivariant \(K\)-theory because \(\mathrm{kk}^{\mathrm{ban}}(\mathbb C,B)\) is the usual topological \(K\)-theory of \(B\). Furthermore, the quasi-homomorphism picture for Kasparov theory is carried over to Lafforgue's theory; this allows to construct a natural map from Lafforgue's bivariant \(K\)-theory to the new one.