The obstruction to excision in \(K\)-theory and in cyclic homology (Q2491155)

Even before Quillen's definition of higher algebraic \(K\)-theory, it was a central problem whether or not algebraic K-theory satisfied excision. Both [\textit{H. Bass}, ``Algebraic \(K\)-theory''. W. A. Benjamin, Inc., New York-Amsterdam (1968; Zbl 0174.30302) and \textit{J. Milnor}, ``Introduction to algebraic \(K\)-theory'', Ann Math. Stud. 72 (1971; Zbl 0237.18005)] books on the subject offer the problem considerable attention, and Karoubi-Villamajor's definition of higher algebraic \(K\)-theory enjoys variants of excision at the price of ``wrong'' \(K_1\)-groups. Given algebraic \(K\)-theory's close ties with topological \(K\)-theory this question is natural, and it seems that it was with some reluctance the community accepted that algebraic \(K\)-theory could not possibly enjoy this property. See \textit{R. G. Swan}'s paper [J. Pure Appl. Algebra 1, 221--252 (1971; Zbl 0262.16025)] for a well written exposition killing all hope of excision. It was only in the middle of the 80s that the beginning of a fuller understanding of the phenomenon seems to have appeared. Low-dimensional calculations of the first obstruction groups led Loday and Weibel independently to conjecture that rationally the obstruction groups for excision in algebraic K-theory coincided with the same groups in Connes' cyclic homology. With \textit{T. G. Goodwillie}'s isomorphism [Ann. Math. (2) 124, 347--402 (1986; Zbl 0627.18004)] between rational relative algebraic \(K\)-theory and cyclic homology in the nilpotent situation, the conjecture gained credibility, and in a series of papers, especially by \textit{S. Geller, L. Reid} and \textit{C. Weibel} [J. Reine Angew. Math. 393, 39--90 (1989; Zbl 0649.14006)], interesting calculations of algebraic \(K\)-theory were obtained under the hypothesis that the conjecture would prove to be true. In the paper under review, the author proves this conjecture, and hence also verifies that the above mentioned results give valid calculations of algebraic \(K\)-theory. To be precise, the excision result appearing in this paper can be formulated as follows. Given a square of (unital and associative) rings \[ \mathcal X\quad=\quad \begin{tikzcd} A\rar\dar & B\dar["f"]\\C\rar & D \end{tikzcd} \] and integer \(i\), let \(K_i(\mathcal X)\) be the ``multirelative'' \(K\)-groups gotten by taking the \(i\)th homotopy group of the iterated fiber of the cube obtained by applying the \(K\)-theory functor to \(\mathcal X\). These groups vanish for all \(i\) exactly when the Mayer-Vietoris sequence \[ \dots\to K_n(A)\to K_n(B)\oplus K_n(C)\to K_n(D)\to K_{n-1}(A)\to\dots \] is exact. Performing the same construction on cyclic homology we get groups \(HC_i(\mathcal X)\). The excision result proved in this paper gives that when the square is cartesian and \(f\) surjective, then the trace induces a natural isomorphism \[ K_i(\mathcal X)\otimes\mathbb{Q}\cong HC_{i-1}(\mathcal X)\otimes\mathbb{Q}. \] In the paper an equivalent statement involving negative cyclic homology is stated as the main result. The crucial ingredient to the proof of this isomorphism is \textit{A. A. Suslin} and \textit{M. Wodzicki}'s theorem [Ann. Math. (2) 136, 51--122 (1992; Zbl 0756.18008)] on the algebraic \(K\)-theory of non-unital rings, and the present paper involves proving a version of Suslin and Wodzicki's theorem for a certain type of pro-rings. Very roughly, one could say that Goodwillie's theorem gives the result ``infinitesimally'', and Suslin and Wodzicki's theorem shows why this actually is enough. In the first section the author considers the infinitesimal situation, giving criteria for rational excision extending those given by Cuntz and Quillen for periodic cyclic homology (where excision is always true). This is similar in spirit to previous work by the author, and in the second section it is shown how these considerations give criteria for when the trace from \(K\)-theory to cyclic homology will give the desired isomorphism. The adaption of Suslin and Wodzicki's procedure is performed in section three, and in section four it is shown why this together with the results of the second section actually give the desired result. The corresponding \(p\)-adic result (involving topological cyclic homology instead of Connes' version) has since been obtained by Geisser and Hesselholt, and for rings up to homotopy by Kittang and the reviewer.