Analytic torsion and \(L^2\)-torsion of compact locally symmetric manifolds (Q366951)

The aim of the reviewed paper is to study the asymptotic expansion of the analytic torsion of a compact quotient of a negatively curved symmetric space as one varies the system of coefficients, following the work of the first named author on hyperbolic three-manifolds [Prog. Math. 297, 317--352 (2012; Zbl 1264.58026)]; the same problem was also studied in more general setting in the work of \textit{J. M. Bismut} et al. announced in [C. R., Math., Acad. Sci. Paris 349, No. 17--18, 977--981 (2011; Zbl 1227.58010)]. Let us describe the setting more precisely: fix \(G\) a real semisimple non-compact Lie group and \(\widetilde X\) the associated Riemannian symmetric space, and fix a uniform, torsion free lattice \(\Gamma\) in \(G\). Then the quotient \(X=\Gamma\backslash\widetilde X\) is a compact Riemannian manifold, and for any Hermitian bundle \(E\) on \(X\) there is a well-defined Ray-Singer analytic torsion \(T_X\). By a classical construction of Matsushima-Murakami, to each complex, finite-dimensional representation \(\tau\) of \(G\) there is associated such a bundle on \(X\), and the authors denote by \(T_X(\tau)\) its analytic torsion. The main result of the paper is then the following: suppose that \(G\) has a maximal compact Cartan subgroup of dimension \(\mathrm{rank}_{\mathbb{C}}(G)-1\) (otherwise they prove that the analytic torsion is always trivial); fix a positive weight \(\lambda\) of the complexified Lie algebra of \(G\) and for \(m\geq 1\) let \(\tau_\lambda(m)\) be the irreducible representation of highest weight \(m\lambda\); then, under the hypothesis that \(\lambda\) is not fixed by the Cartan involution they prove the asymptotic development \[ \log T_X(\tau_\lambda(m)) = C_{\widetilde X}\mathrm{vol}(X) P_\lambda(m) + O(e^{-cm}) \] where \(P_\lambda\) is a polynomial. The highest-degree monomial of \(P_\lambda\) is proportional to that of \(m\dim(\tau_\lambda(m))\), and they compute the constant \(C_{\widetilde X}\) explicitely for all groups satisfying the first hypothesis, namely \(\mathrm{SL}_3(\mathbb{R})\) and the connected components of the orthogonal groups \(\mathrm{SO}(p,q)\) for \(pq\) odd. The proof has two steps: first comparing the analytic torsion \(T_X(\tau_\lambda(m))\) to the \(L^2\)-analytic torsion associated to \(\tau_\lambda(m)\), which yields that the logarithmic difference of the two is \(O(e^{-cm})\): this is where the hypothesis that \(\lambda\) be not fixed by the Cartan involution is used. The second step consists in computing the \(L^2\)-torsion which yields the polynomial term of the expansion. Let us finally remark that the hypothesis that \(\Gamma\) be uniform is quite important: similar results can be proven in the non-uniform case (at least for real hyperbolic manifolds, see the work [J. Funct. Anal. 263, No. 9, 2615--2675 (2012; Zbl 1277.58018)] of the authors) but the proofs become much more involved.