Heat kernel asymptotics of the Zaremba boundary value problem (Q1876635)

From MaRDI portal
scientific article
Language Label Description Also known as
English
Heat kernel asymptotics of the Zaremba boundary value problem
scientific article

    Statements

    Heat kernel asymptotics of the Zaremba boundary value problem (English)
    0 references
    0 references
    20 August 2004
    0 references
    Let \(M\) be a compact Riemannian manifold with metric \(g\) and smooth boundary, \(V\) a vector bundle on \(M\) and \(F_B\) a second order partial differential operator acting on the smooth sections of \(V\). The Zaremba boundary conditions, discussed in this paper, are discontinuous boundary conditions given by Dirichlet conditions on one part of the boundary and Neumann conditions on another part of the boundary. The main scope of the paper is to discuss the heat kernel asymptotics of this boundary problem. More precisely, it is assumed that the boundary is decomposed as a disjoint union \[ \partial M=\Sigma_1\cup\Sigma_2\cup\Sigma_0, \] where \(\overline\Sigma_1\) and \(\overline\Sigma_2\) are smooth codimension \(1\) submanifolds with common boundary \(\Sigma_0\), which is a smooth compact codimension \(2\) submanifold without boundary. Consider the Laplace type operator \[ F_B=-g^{\mu\nu}\nabla_{\mu}\nabla_{\nu}+Q \] where \(\nabla\) is the natural extension of the Levi-Civita connection and \(Q\) is a smooth Hermitian endomorphism of \(V\). The Zaremba boundary conditions are then: \[ \phi| _{\Sigma_1}=0, \quad (\partial_r+\Lambda)\phi| _{\Sigma_2}=0, \] where \(\Lambda\) is a smooth Hermitian endomorphism of \(V\) restricted to the boundary and \(r\) is the normal geodesic distance to \(\partial M\). However, the author stresses the fact that, for the correct formulation of the problem, one needs an additional boundary condition on the singular set \(\Sigma_0\). For that purpose, he considers one of the following conditions: \[ (\sqrt{\rho}\phi)| _{\Sigma_0}=0, \quad \text{resp.} \quad (\partial_{\rho}-s)(\sqrt{\rho}\phi)| _{\Sigma_0}=0,\tag{1} \] where \(\rho\) is the normal geodesic distance to \(\Sigma_0\) and \(s\) is a real parameter. In the above notation, one has the following heat trace asymptotic expansion as \(t\to 0^+\): \[ \begin{aligned} \text{Tr}_{L^2}\exp(-tF_B) =&(4\pi t)^{-m/2}\biggl\{\dim V \text{vol}(M) +t^{1/2}\frac{\sqrt\pi}{2}\dim V[\text{vol}(\Sigma_2)- \text{vol}(\Sigma_1)]+\\ &+t\biggl[\frac16\dim V\int_MR-\int_M \text{tr}_VQ+ \frac13\dim V\int_{\partial M}K+\\ &+ 2\int_{\Sigma_2} \text{tr}_V\Lambda +\frac{\pi\alpha}{4}\dim V \text{vol}(\Sigma_0)\biggr]+O(t^{3/2})\biggr\} \end{aligned} \] where \(R\) is the scalar curvature of the metric \(g\), \(K\) is the trace of the extrinsic curvature of the boundary and \(\alpha\) is a numerical constant depending on the choice made in (1): \(\alpha=-1\) (resp. \(\alpha=7\)). It should also be observed that, thanks to a result of \textit{R. Seeley}'s [Commun. Partial Differ. Equations 27, 2403--2421 (2002; Zbl 1055.58010)] the above expansion does not contain logarithmic terms. The computation does not go through a rigorous construction of the parametrix, but rather depends on the derivation of an asymptotic solution of the heat equation which allows explicit formulas for the coefficients.
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    boundary value problem
    0 references
    heat kernel
    0 references
    spectral asymptotics
    0 references
    spectral geometry
    0 references
    0 references
    0 references