Conic degeneration and the determinant of the Laplacian (Q2352246)

The determinant of the Laplacian is an important spectral invariant for closed Riemannian manifolds. In order to study its extension to Riemannian orbifolds one can approximate them by manifolds and study the asymptotics: the present paper carries this out in the case of a natural degeneration of Riemannian metric \(\Omega_\varepsilon\) towards an orbifold \(\Omega_0\) with an isolated conical singularity as \(\varepsilon\to 0\). The latter mean that \(\Omega_0\) is a manifold of dimension \(n\) except at a point \(P\), and that a neighbourhood of \(P\) is isometric to the cone over a manifold \(\mathcal N\) of dimension \(n-1\), that is \([0,1]\times\mathcal N\) with of \(\{0\}\times\mathcal N\) collapsed to \(P\) with the metric \(dr^2 + r^2dy^2\) where \(dy^2\) is the metric on \(\mathcal N\). The degeneration is then roughly constructed as follows: one fixes a smooth approximation \(\mathcal Z\) of the infinite cone \(C_{\mathcal N}\) over \(\mathcal N\) and one takes \(\Omega_\varepsilon\) to be the gluing of the tip of to \(\varepsilon^{-1}\mathcal Z\) with a slice of \(\mathcal C_{\mathcal N}\) that itself glues to \(\Omega_0\) minus the conic neighbourhood of \(P\) (this is better explained using pictures than words, see Figure 1 in the paper under review). The main result in the paper is Theorem 2, which gives the asymptotic expansion at \(\varepsilon\to 0\) of the logarithms of the Laplacian determinants \(\log\det \Delta_{\Omega_\varepsilon}\). The expansion has the form \[ \log\det \Delta_{\Omega_\varepsilon} = a (\log \varepsilon)^2 - b \log \varepsilon + \log\det\Omega_0 + c + o(1) \] where \(a,b,c\) depend only on the conical singularity (they are expressed in terms of spectral invariants on \(C_{\mathcal N}\) and \(\mathcal Z\)); in addition \(a = 0\) if \(n\) is odd. In particular the constant coefficient is not in general the logarithm of the determinant of the limit orbifold. As usual when dealing with analytic torsion the asymptotic analysis must deal separately with the part coming from the small-time asymptotics of the heat trace and that coming from large times. Theorem 3 in the introduction deals with the former: it expresses the heat trace on \(\Omega_\varepsilon\) as a barycenter of the heat trace on the tip of \(\mathcal Z\) and the heat trace on the nonsingular part of \(\Omega_0\) with a rest that doe not contribute to the asymptotics at \(t\to 0\). The latter is dealt with in Theorem 4, which establishes that as \(\varepsilon\to 0\) the heat trace at time \(t\) on \(\Omega_\varepsilon\) for \(t\) bounded away from 0 converges to that for \(\Omega_0\), and that its exponential decay in \(t\) is uniform.