Boundaries, spectral triples and \(K\)-homology (Q2327767)

In this technical paper, the authors extend the concept of a spectral triple of non-commutative geometry to a relative spectral triple. Efforts to find a both conceptually satisfactory and technically feasible definition of a relative spectral triple fits into the known open problem of how to grasp the classical concept of a manifold-with-boundary in a non-commutative framework. Recall that a classical (i.e. ``commutative'') closed Riemannian spin manifold always gives rise to a triple consisting of a Hilbert space \(H\) (the \(L^2\) sections of the spinor bundle) together with an algebra \(A\) of bounded operators (the smooth functions acting on spinors by multiplication) and a densly defined unbounded self-adjoint operator \(D\) (the Dirac operator acting on spinors); from this spectral triple \((H,A,D)\) the original space can be recovered at least as a metric space. There are roughly two ways to approach the corresponding classical relative problem: Either impose APS-like self-adjoint boundary conditions (having the disadvantage of being non-local) for the Dirac operator or to make the Dirac operator itself self-adjoint by pushing the boundary into infinity (again quite a non-geometric procedure). Meanwhile the concept of a spectral triple admits a natural generalization to the non-classical (i.e. ``non-commutative'') situation, it is not straightforward how to obtain a satisfactory relative version. In this paper, the authors define a relative spectral triple by relaxing the self-adjointness of the Dirac operator to a symmetric condition together with additional analytic data (for a precise formulation see Definition 2.1 in the article). The article contains two main results: The first says that the relative spectral triple as defined is compatible with the boundary map in the associated \(K\)-homology theory (see Theorem on p. 408 or equivalently Theorem 2.13 in the article); the second says that the doubling of a relative triple as defined gives rise to an (absolute) spectral triple (see Theorem on p. 409 or Theorem 4.22 and Proposition 4.23 in the article). Note that the latter result is the non-commutative analogue of doubling i.e. a way of constructing a closed manifold out of a manifold-with-boundary by glueing two copies of the latter along their boundaries.