On the relative \(K\)-group in the ETNC (Q2335703)

As shown by the title, the main object of this article is to give a (far reaching) reinterpretation of the relative \(K\)-group \(K_0(\mathfrak{A}, \mathbf{R})\) which appears in the Burns-Flach formulation of the Equivariant Tamagawa Number Conjecture (ETNC). Let us recall that \(\mathfrak{A}\) denotes an order in a (not necessarily commutative) finite-dimensional semi-simple \(\mathbf{Q}\)-algebra \(A\), and the relative \(K\)-group takes place in a long exact sequence \(\dots\to K_1(\mathfrak{A})\to K_1(A_{\mathbb{R}})\to K_0(\mathfrak{A}, \mathbf{R})\to K_0(\mathfrak{A})\dots\) Tamagawa numbers classically stem from evaluations of integrals on adelic points of algebraic groups, and by the canonical isomorphism \(K_0(\mathfrak{A}, \mathbf{Z})\cong\mathbf{R}^\times_{>0}\), the interpretation as a volume is still visible. In this paper the author proposes a different viewpoint, where the relative \(K\)-groups are replaced by ``absolute'' \(K\)-groups of a suitable category, and the volume interpretation is even strengthened. He introduces the category LCA\(_{\mathfrak{A}}\) of locally compact topologized right \(\mathfrak{A}\)-modules, which he shows to be an exact category, thus endowed with its own \(K\)-theory groups, determinant functors, etc. Then, if \(\mathfrak{A}\) is a \textit{regular} order, for all \(n\) there exist canonical long exact sequences \(\dots\to K_n(\mathfrak{A})\to K_n(A_{\mathbf{R}})\to K_n(\text{LCA}_{\mathfrak{A}})\to K_{n-1}(\mathfrak{A})\dots\) and canonical isomorphisms \(K_n(\text{LCA}_{\mathfrak{A}})\cong K_{n-1}(\mathfrak{A}, \mathbf{R})\) (thm. 11.3 of the text). A quite analogous result is available for \(G\)-theory in all degrees (thm. 11.4), so that, roughly speaking, \(\text{LCA}_{\mathfrak{A}}\) corresponds to the \(G\)-theory groups \(G_n(\mathfrak{A})\) in general. The author stresses that his results are by no means a formal construction transforming relative \(K\)-groups into absolute ones. Actually his methods work specifically for relative \(K\)-groups with respect to the reals, exploiting the fact \(\mathbf{R}\) is a locally compact topological field. The main result also yields a presentation ``à la Nenashev'' for \(K_0(\mathfrak{A},\mathbf{R})\), where generators are double short exact sequences of locally compact right \(\mathfrak{A}\)-modules, modulo suitable relations coming from \((3\times 3)\)-diagrams (thm. 2.5). The underlying idea for the above is inspired from [\textit{P. Deligne}, Contemp. Math. 67, 93--117 (1987; Zbl 0629.14008)]: every exact category has a universal determinant functor to its virtual objects. The author deduces from \textit{D. Clausen} [``A K-theoretic approach to Artin maps'', Preprint, \url{arXiv:1703.07842}], that the universal determinant functor of the category LCA of locally compact abelian groups is precisely the Haar measure (thm. 12.8). Since classical Tamagawa numbers stem from the Haar measure, and \(\text{LCA}_{\mathfrak{A}}\) is just LCA endowed with an action of \(\mathfrak{A}\), it is natural to expect that an \textit{equivariant} Haar measure should be the universal determinant functor of the category \(\text{LCA}_{\mathfrak{A}}\). The paper concludes with a reciprocity law: Let \(\mathfrak{A}\) be an arbitrary order in a finite-dimensional semi-simple \(\mathbf{Q}\)-algebra \(A\), and \(\mathbb{A}\) the adèle ring of the rationals. Then the composition of natural maps \(K(A)\to K(A\otimes\mathbb{A})\to K(\text{LCA}_{\mathfrak{A}})\) is zero.