The analytic torsion and its asymptotic behaviour for sequences of hyperbolic manifolds of finite volume (Q457599)

Let \(X\) be an hyperbolic manifold of odd dimension \(d=2m+1\), which is non-compact of finite volume. Then there are defined certain spectral invariants of \(X\), the analytic torsions \(T_X(\tau)\) where \(\tau\) is a representation of \(\mathrm{SO}(d,1)\), inducing in a natural way a euclidean vector bundle \(E_\tau\) over \(X\) (their definition involves the spectra of the Laplace operators on forms on \(X\) with coefficients in \(E_\tau\)). The main result in the present paper establishes, for certain sequences of finite coverings \(X_i\to X\), the exponential growth in the degree of the analytic torsions \(T_{X_i}(\tau)\), with a rate depending only on \(d\) and the volume of \(X\). Namely, under hypotheses on \(X_i\) and \(\tau\) to be precised later, the authors prove that \[ \lim_{i\to+\infty} \frac{\log T_{X_i}(\tau)}{\deg(X_i\to X)} = t_{\widetilde X}^{(2)}(\tau)\cdot\mathrm{vol}(X) \qquad\qquad (\ast) \] where \(t_{\widetilde X}^{(2)}(\tau)\) is a nonzero constant (positive when \(m\) is even and negative if it is odd). The motivation from this result comes from work of \textit{N. Bergeron} and \textit{A. Venkatesh} [J. Inst. Math. Jussieu 12, No. 2, 391--447 (2013; Zbl 1266.22013)]. A similar result was worked out in the preprint \url{arxiv:1212.3161} of the reviewer, which was stated there only in three dimensions and is significantly weaker in some aspects. Let us explain the hypotheses under which the conclusion \((\ast)\) above holds. They are of two kinds: first one must suppose that \(\tau\circ\theta\not=\tau\) where \(\theta\) is the Cartan involution of \(\mathrm{SO}(d,1)\). This first hypothesis is of a spectral nature, as explained below. The other hypotheses concern the geometry of the \(X_i\): the authors assume that the systole of \(X_i\) (smallest length of a closed geodesic on \(X_i\)) goes to infinity and that the cusps are uniform (meaning that their Euclidean basis is a torus and that the set of all these tori when \(i\) goes to infinity stays in a compact set in the moduli space). There is a third hypothesis on the cusps which is (non-trivially) redundant with the two previous ones. We also remark that the hypothesis on the systole is unnecessary strong: indeed, the authors prove that \((\ast)\) holds also for one specific sequence \(X_i\) not satisfying it (Theorem 1.5), and their proof actually should work in the setting of a sequence of finite covers which is only Benjamini--Schramm convergent to the hyperbolic space (assuming the cusps are uniform in the sense above), a notion introduced in [\textit{M. Abert} et al., C. R., Math., Acad. Sci. Paris 349, No. 15--16, 831--835 (2011; Zbl 1223.53039)]. The proof of \((\ast)\) divides into two parts according to the two sides of Selberg's trace formula: this is natural since the definition of analytic torsions uses the spectral side to deal with the heat kernel at large times, and the geometric side to deal with the small times. The part of the proof that concerns the latter contains no real difficulties under the very strong hypotheses that the authors make on the sequence \(X_i\). The main contribution of the present paper thus lies in the study of the heat kernels of arbitrary covers of \(X\). The hypothesis on \(\tau\) ensures that the Laplace operators with values in \(E_\tau\) have a uniform spectral gap, hence that no complications can come from small eigenvalues: this already arises in the compact case studied in the work of Bergeron and Venkatesh. In the noncompact case there is an additional part of the spectral side of the trace formula which causes substantial difficulties, namely the integrals involving intertwining operators. The authors manage to control the intertwining operators well enough to overcome this problem, using techniques developped by the first author in [Ann. Math. (2) 130, No. 3, 473--529 (1989; Zbl 0701.11019)].