Hermitian K-theory over topological *-algebras (Q1820371)

This large study is a sequel to the author's previous work [J. Math. Anal. Appl. 92, 452-506 (1983; Zbl 0564.18011)] and is much connected with the survey article by \textit{A. S. Mishchenko} [Usp. Mat. Nauk 31, No.2, 69-134 (1976; Zbl 0427.55001)]. Denote by A a complete locally m-convex \(C^*\)-algebra with unit element (its topology is given by a family \((p_{\alpha})\) of submultiplicative seminorms satisfying \(p_{\alpha}(x^*x)=p_{\alpha}(x)^ 2\) for all x in A and all indices \(\alpha)\). The main object of the paper are hermitian A-valued inner forms on (finitely generated projective) left A- modules. Under the additon defined through a direct sum (of modules and forms) they form so called Grothendieck group \(K_ h(A)\) \((A\to K_ h(A)\) is the hermitian K-functor). Those concepts are described in the two first sections of the paper; the author proves there, among other results that if M is a finitely generated free A-module then there exists on M a canonically defined positive definite hermitian inner form, which is uniquely determined within an equivalence (theorem 2.2), and that if \(\alpha\) is a hermitian inner product on a finitely generated projective A-module M, then the pair (M,\(\alpha)\) admits a direct sum decomposition \((M,\alpha)=(M^+,\alpha^+)\oplus (M^-,\alpha^-)\), where \(\alpha^+\) (resp. \(\alpha^-)\) is positive (resp. negative) definite (theorem 2.4). Further (sections 3-5) the author considers the category \({\mathcal E}^ h_ A(X)\) of hermitian A-bundles on a topological space X and associated Grothendieck group \(K_ h(X;A)\) and describes their connections with representations of topological *-algebras and homotopy classification of A-bundles. He proves (theorem 3.3) that if X is paracompact then every A-bundle over X can be endowed with a hermitian structure so that it becomes a hermitian A-bundle. The second part of the paper (sections 6-10) is entitled ''Applications''. In section 6 the author states (without proof) a Hirzebruch signature theorem. A large section 7 is devoted to Fredholm A-operators, i.e. A- linear continuous endomorphisms u of \(\ell_ 2(A)=\{(x_ i)_ 1^{\infty}\in A^ N:\Sigma x_ ix^*_ i\) converges in \(A\}\) (with suitable topology) such that \(\ell_ 2(A)=M_ 1\oplus N_ 1=M_ 2\oplus N_ 2\) and \(u=\left( \begin{matrix} u_ 1\\ 0\end{matrix} \begin{matrix} 0\\ u_ 2\end{matrix} \right)\) with \(u_ 1:M_ 1\to M_ 2\) being an isomorphism of A-modules and \(u_ 2\) being a continuous A-linear map of \(N_ 1\) into \(N_ 2\). For such an operator u the author defines an index. This is done under the assumption that A is also a Q-algebra, i.e. the set of its invertible elements is open. As the author remarks (Scholium 3.2) it is a rather restrictive assumption since in many cases it implies that A is a Banach algebra (a \(C^*\)-algebra), it is so if A is barrelled or commutative and it is not known whether it is not always true. The main result of section 8 entitled ''compact A-operators'' states that if A is also a Q-algebra and u,v are A-linear continuous endomorphisms of \(\ell_ 2(A)\) such that \(uv=1+\kappa_ 1\) and \(vu=1+\kappa_ 2\), with \(\kappa_ i\) being compact, then both u and v are Fredholm. In section 9 the author developes an analogon of the classical Atiyah-Singer index and in section 10 the Atiyah-Singer index formula for elliptic A-operators is given, in the latter again assuming that A is a Q-algebra.