Twisted \(K\)-theory, \(K\)-homology, and bivariant Chern-Connes type character of some infinite dimensional spaces (Q2258612)

Twisted \(K\)-theory for pointed compact spaces may be defined through \(C^*\)-algebras with continuous trace. Similarly, twisted \(K\)-homology is the \(K\)-homology of the same \(C^*\)-algebra. Twisted cohomology can also be described in this fashion, using either the local cyclic homology of the same continuous trace \(C^*\)-algebra or the periodic cyclic homology of a suitable dense subalgebra of smooth, trace-class-valued functions. This paper extends this setup to certain non-compact spaces. The continuous, possibly unbounded functions on a non-compact space form only a pro-\(C^*\)-algebra, that is, a projective limit of \(C^*\)-algebras. Since uncountable projective systems cause technical difficulties, most of the article deals only with countable projective limits of \(C^*\)-algebras, which are also called \(\sigma\)-\(C^*\)-algebras. The extension of \(K\)-theory and \(K\)-homology to this class is rather well-understood, and the smooth subalgebras to which one may want to apply periodic cyclic homology are Fréchet, so that everything is already defined. This allows to treat twisted \(K\)-theory, \(K\)-homology, and cohomology for spaces that are countable inductive limits of compact spaces, briefly called \(\sigma\)-compact spaces. Given a principal PU-bundle over a \(\sigma\)-compact space \(X\), we may form the \(\sigma\)-\(C^*\)-algebra \(A\) of continuous sections of the associated bundle with fibre the \(C^*\)-algebra of compact operators. The (representable) \(K\)-theory of this \(\sigma\)-\(C^*\)-algebra is a good definition of the twisted \(K\)-theory associated to the given PU-bundle. Let \(X=\bigcup X_n\) be a cofinal sequence of compact subsets \(X_n\) in \(X\), and let \(A_n\) be the restriction of \(A\) to \(X_n\). The \(K\)-theory of \(A_n\) is the twisted \(K\)-theory of \(X_n\), and \(A\) is the limit of the projective system \((A_n)\). Hence the representable \(K\)-theory of \(A\) is computed by a Milnor \(\varprojlim^1\)-sequence. The twisted \(K\)-homology of the given PU-bundle should be the \(K\)-homology of \(A\). \(K\)-homology for \(C^*\)-algebras is a special case of \(KK\)-theory, and there are several closely related ways to extend \(KK\)-theory to \(\sigma\)-\(C^*\)-algebras. Here an approach of Cuntz based on universal properties is explored. This work is largely unnecessary, however, because any cohomology theory for \(\sigma\)-\(C^*\)-algebras that is compatible with products maps projective limits into inductive limits; the proof is the same as for the Milnor \(\varprojlim^1\)-sequence. That is, \(K^*(A) \cong \varinjlim K^*(A_n)\), and the right hand side only uses \(K\)-homology for \(C^*\)-algebras. Twisted cohomology is approached here first via the periodic cyclic homology of the algebra of smooth sections of the bundle of trace-class operators on \(X\) associated to the PU-bundle; here a section is smooth if its restriction to each \(X_n\) is smooth, and this definition makes sense if \((X_n)\) is an inductive system of smooth manifolds. A second approach is via local cyclic homology. This version of the cyclic theories also gives good results on \(C^*\)-algebras, and it is defined on \(\sigma\)-\(C^*\)-algebras as well. While it is plausible that it still behaves well on that larger category, this is not proved yet. In particular, the proof of homotopy invariance in [the author, ``Continuous homotopy invariance of bivariant local cyclic homology for \(\sigma-C^\ast\)-algebras'', Preprint , \url{arXiv:1202.1333v2}] has a gap. The issue is that many results about local cyclic homology require all bounded subsets to have finite spectral radius. This is true for Banach algebras and their smooth subalgebras, but fails for \(\sigma\)-\(C^*\)-algebras. Thus it is unclear whether the local cyclic homology of the \(\sigma\)-\(C^*\)-algebras described above gives twisted cohomology.