Local Algebraic K-Theory

TL;DR

This paper develops a local algebraic K-theory framework for pseudo-differential operators and singular integral operators, enabling refined index formulas and applications to Lipschitz and quasi-conformal manifolds.

Contribution

It introduces a general construction of local algebraic K-theory for filtered algebras, avoiding projective modules and focusing on idempotent matrices, with potential extensions to higher degrees.

Findings

Constructed local K-theory for filtered algebras.

Applied framework to pseudo-differential and singular integral operators.

Facilitated index formulas on Lipschitz and quasi-conformal manifolds.

Abstract

In this article we address the first part of the programme presented in \cite{Teleman_arXiv_III}, \S 2; we construct the local $K$- theory level of the index formula. Our construction is sufficiently general to encompass the algebra of pseudo-differential operators of order zero on smooth manifolds, elliptic pseudo-differential operators of order zero, their abstract symbol (see Introduction \S 2.) and their local $K$- theory analytical and topological index classes, see \cite{Teleman_arXiv_III}, \S 5, Definition 5 and 6. Our definitions are sufficiently general to apply to exact sequences of singular integral operators, which are of interest in the case of the index theorem on Lipschitz and quasi-conformal manifolds, see \cite{Teleman_IHES}, \cite{Teleman_Acta}, \cite{Donaldson_Sullivan}, \cite{Connes_Sullivan_Teleman}. In this article we introduce localised algebras (Definition 3) $\mathit{A}$ and in \S 6 we define their local algebraic $K$-theory. A localised algebra $\mathit{A}$ is an algebra in which a decreasing filtration by vector sub-spaces $\mathit{A}_{\mu}$ is introduced. The filtration $\mathit{A}_{\mu}$ induces a filtration on the space of matrices $\mathbb{M}(\mathit{A}_{\mu})$. Although we define solely $K^{loc}_{\ast}(\mathit{A})$ for $\ast= 0, \; 1$, we expect our construction could be extended in higher degrees. We stress that our construction of $K^{loc}_{0}(\mathit{A})$ uses exclusively idempotent matrices and that the use of finite projective modules is totally avoided. (Idempotent matrices, rather than projective modules, contain less arbitrariness in the description of the $K_{0}$ classes and allow a better filtration control). The group $K^{loc}_{0}(\mathit{A})$ is by definition the quotient space of the space of the Grothendieck completion of the space of idempotent matrices through three equivalence relations: -i) stabilisation $\sim_{s}$, -2) local conjugation $\sim_{l}$, {\em and} -3) projective limit with respect to the filtration. By definition, the $K_{1}^{loc} (\mathit{A})$ is the projective limit of the local $K_{1}(\mathit{A}_{\mu})$ groups. The group $K_{1}(\mathit{A}_{\mu})$ is by definition the quotient of $\mathbb{GL}(\mathit{A}_{\mu})$ modulo the equivalence relation generated by: -1) stabilisation $\sim_{s}$, --2) local conjugation $\sim_{l}$ and -3) $\sim_{\mathbb{O}(\mathit{A}_{\mu})}$, where $\mathbb{O}(\mathit{A}_{\mu})$ is the sub-module generated by elements of the form $ u \oplus u^{-1} $, for any $u \in \mathbb{GL}(\mathit{A}_{\mu})$. The class of any invertible element $u$ modulo conjugation (inner auto-morphisms) we call the Jordan canonical form of $u$. The local conjugation preserves the local Jordan canonical form of invertible elements. The equivalence relation $\sim_{\mathbb{O}(\mathit{A}_{\mu})}$ insures existence of opposite elements in $K_{1}(\mathit{A}_{\mu})$ and $K_{1}^{loc}(\mathit{A})$. Our definition of $K^{loc}_{1}(\mathit{A})$ does not use the commutator sub-group $[\mathbb{GL}(\mathit{A}), \mathbb{GL}(\mathit{A})]$ nor elementary matrices in its construction. We define short exact sequences of localised algebras. To get the corresponding (open) six terms exact sequence (Theorem 51) one has to take the tensor product of the expected six terms exact sequence by $\mathbb{Z}[\frac{1}{2}]$. We expect the factor $\otimes_{\mathbb{Z}[\frac{1}{2}}]$ to have important consequences. Our work shows that the basic structure of $K_{1}$ resides in the {\em additive} sub-group generated by elements of the form $u \oplus u^{-1}$, $u \in \mathbb{GL}(\mathit{A})$, rather than in the {\em multiplicativ} commutator sub-group $[\mathbb{GL}(\mathit{A}), \mathbb{GL}(\mathit{A})]$. Even into the case of trivially filtered algebras, $\mathit{A}_{\mu} = \mathit{A}$, for all $\mu \in \mathbb{N}$, the introduced group $K^{loc}_{1}(\mathit{A})$ should provide more information than the classical group $K_{1}(\mathit{A})$.