Analytic torsion and Reidemeister torsion of hyperbolic manifolds with cusps (Q2201988)

The authors start from the definition of analytic torsion, for which the celebrated Cheeger-Müller theorem was proved. As the authors point out, when the complex of cochains is defined over \(\mathbb{Z}\), then the Reidemeister torsion can be expressed in terms of the size of the torsion subgroup of the integer cohomology and the covolume of the lattice defined by the free part in the real cohomology. Thus to establish the exponential growth of torsion subgroups in cohomology, we need to compute the limiting behavior of analytic torsion via spectral methods. The main theorem the authors prove (page 5) is quite involved (page 913, Theorem 1.1 in the text): Theorem. If the complex irreducible representation \(\mathcal{\rho}: G\rightarrow \mathrm{GL}(V)\) satisfies \(\rho\circ \theta\not=\rho\) and if the discrete subgroup \(\Gamma\subset G\) is such that Assumption 2.2 below holds, then \[ \log(T;E, g_{X}, h_{E})=\log(\tau(\overline{X}, E, \mu_{X}))-\frac{1}{2}\log(\tau(Z,E,\mu_{Z}))-K_{\Gamma}^{\rho}c_{\rho} \] where \(c_{|\rho}\in \mathbb R\) is an explicit constant depending on \(\rho\), \(k^{\rho}_{\Gamma}\) is the number of connected components of \(Z\) on which the cohomology with values in \(E\) is non-trivial, and \(\mu_{X}\) and \(\mu_{Z}\) are cohomology bases of \(H^{*}(\overline{X},E)\) and \(H^{*}(Z,E)\) described above. Furthermore, if \(n\) is odd, then \(\tau(Z,E,\mu_{Z})=1\) and the formula simplifies to \[ \log T(X;E,g_{X}, h_{E})=\log \tau(\overline{X},E,\mu_{X})-K^{\rho}_{\Gamma} c_{\rho} \] The main technical contribution of this theorem is that they are able to control the contribution from the degeneration of the ``cusp ends''. The authors worked out explicit examples like \begin{itemize} \item \(G=\mathrm{Spin}(d,1)\). In this case the above formula simplies to \[ \log T(X;E,g_{X}, h_{E})=\log \tau(\overline{X},E)-\frac{1}{2}\log \tau(Z,E) \] \item \(G=\mathrm{Spin}(3,1)\cong \mathrm{SL}(2,\mathbb{C})\) and \(X:=\Gamma \backslash G / K\) is the complement of a hyperbolic knot. In this case the formula simplies to \[ T(X;E,g_{X},h_{E})=\tau(\overline{X},E) \] \item \(G=\mathrm{SO}_{0}(d,1)\). For full detail we refer the reader to Corollary 1.7 of the paper. In this case the formula gives \[ \tau(\overline{X}, E_{\tau(m),\mu_{X,m})}=C_{n}\mathrm{vol}(X)m^{n(n+1)/2+1}+\mathcal{O}(m^{n(n+1)/2} \log(m)) \] \end{itemize} To obtain Theorem 0.1, the authors work with singular metrics of the form \[ g_{\epsilon}=\frac{dx^2}{x^2+\epsilon^2}+(x^2+\epsilon^2)g_{\partial \overline{X}}, x\in (-\delta, \delta) \] in a tubular neighborhood \(N\cong \partial \overline{X}\times (-\delta, \delta)_{x}\). The rest of the work heavily uses b-calculus from [\textit{R. B. Melrose}, The Atiyah-Patodi-Singer index theorem. Wellesley, MA: A. K. Peters, Ltd. (1993; Zbl 0796.58050)]. It should be pointed out that the method used in the paper is not the only method available to treat space with cusp type singularity. Currently, there are at least three (independent?) alternative methods available for problem of this type: The relative hyperbolic theory invented by Sarnak, Jorgenson and Rolf Lundelius; the ``relative compact perturbation theory'' invented by Siarhei Finski and other followers of Bismut; and lastly the (recent) work by \textit{G. Freixas i Montplet} and \textit{R. A. Wentworth} [Ann. Sci. Éc. Norm. Supér. (4) 52, No. 5, 1265--1303 (2019; Zbl 1440.14043); J. Differ. Geom. 115, No. 3, 475--528 (2020; Zbl 1454.58030)] on singular spaces like modular curves. Reviewer's remark: The adoption of b-calculus to the present setting seems to be an over-kill from my naive perspective. Indeed, most of the paper (pages 923--942) are devoted to set up the b-calculus and use it to analyze the degeneration of heat kernel/resolvent in the cusp setting. While the analysis is quite involved as one needs to work through the single surgery space viewing as manifold with corners, the authors overcome a non-technical essential difficulty in Section 4 to understand the \(L^{2}\)-kernel of \(D_{b}\). To do this, the authors compute the \(L^{2}\)-cohomology of \(F_{P}(Y)\) with coefficients in \(E\) (Lemma 4.3, Proposition 4.4). The proof involves a version of the Hodge-de Rham theorem using spectral sequence of polyhomogeneous operators. From my naive perspective, this is a great achievement that they did not mention in the introduction part of their paper. This reminds me of \textit{J. I. Burgos Gil}'s work using log-singular metrics and log-log forms for Arakelov theory on Shimura varieties (for a reference, see [Lect. Notes Math. 2276, 377--401 (2021; Zbl 1458.14031)]). What is unclear to me is whether we can by-pass the technical part of the paper using alternative methods to derive the main result. In other words, is it possible for us to ``re-interpret'' the main result of the paper from an arithmetic point of view by suggesting it may be an equality of morphisms on the moduli space of (real) arithmetic varieties? The intricate explicit equalities in statement (1.6) and (1.7) suggest there may be something deeper that can provide a unified perspective on these results.