Asymptotic behaviour of analytic torsion and cohomological torsion for ℚ \mathbb{Q} -rank 1 1 arithmetic groups
Tim Berland · 2026
Read Asymptotic behaviour of analytic torsion and cohomological torsion for ℚ \mathbb{Q} -rank 1 1 arithmetic groups by Tim Berland (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-22711.txt.
Introduction
Berland extends refined asymptotics for analytic torsion from congruence subgroups of special linear groups to a large family of reductive groups. The extension yields new asymptotics and bounds on torsion growth in cohomology. Cases covered include special linear groups over rings of integers of number fields and odd-dimensional orthogonal groups. Analytic torsion here measures spectral asymmetry of a manifold.
Catalog record
| CREATOR | Tim Berland |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 96040 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.22711
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Asymptotic behaviour of analytic torsion and cohomological torsion for ℚ \mathbb{Q} -rank 1 1 arithmetic groups Tim Berland Abstract Abstract. We extend the refined asymptotics of analytic torsion associated to congruence subgroups of \operatorname{SL}(n) in [ 12 ] , to congruence subgroups in a large family of reductive groups. This is applied to give new asymptotics and bounds on the growth of torsion in the cohomology of congruence subgroups of \operatorname{SL}(2,\mathcal{O}_{F}) for F a number field, and of congruence subgroups in \operatorname{SO}(n,1) with n odd. 1. Introduction For more than a decade, analytic torsion has been a valuable tool in studying growth of torsion in the cohomology of arithmetic groups. In the seminal paper of Bergeron and Venkatesh [ 11 ] , they use analytic torsion to establish exponential growth for an alternating sum of torsion cohomology with certain non-trivial coefficient systems of families of cocompact arithmetic groups. Below we introduce their conjectures and results before moving on to more recent generalizations. 1.1. Conjectures and the cocompact case Let G be a connected semisimple algebraic group defined over \mathbb{Q} , and let \Gamma\subseteq G(\mathbb{Q}) be an arithmetic subgroup. A good example to keep in mind is G=\operatorname{SL}(n) and \Gamma=\operatorname{SL}(n,\mathbb{Z}) . Let M be a \Gamma -module arising from a rational finite-dimensional representation of G(\mathbb{R}) . Then the group cohomology H^{*}(\Gamma,M) has long been understood to have number theoretic significance, such as its link to automorphic forms by the work of Franke [ 20 ] . If one now considers a lattice L\subseteq M stabilized by \Gamma , one can also consider the cohomology groups H^{*}(\Gamma,L) . These are finitely generated abelian groups, and these, too, carry number theoretic information. In particular, it was conjectured by Ash [ 9 ] , and later partially proven by Scholze [ 42 ] , that there is a correspondence between Hecke eigensystems in H^{*}(\Gamma,\mathbb{F}_{p}) and certain Galois representations \rho:\operatorname{Gal}(\overline{\mathbb{Q}}/\mathbb{Q})\to\operatorname{GL}(n,\overline{\mathbb{F}}_{p}) . Investigating this cohomology, Bergeron–Venkatesh [ 11 ] gave the following heuristic. Let K\subseteq G(\mathbb{R}) be a maximal compact subgroup, and define the deficiency of G as \delta(G)\coloneqq\operatorname{rank}_{\mathbb{C}}G(\mathbb{R})-\operatorname{rank}_{\mathbb{C}}K . Then there is "a lot" of torsion in H^{i}(\Gamma,L) when \delta(G)=1 and i is the middle dimension, and "little" torsion otherwise. To make this precise, they make the following conjecture, generalized in [ 8 ] . Conjecture 1.1 (Bergeron–Venkatesh) . Let \Gamma\eqqcolon\Gamma_{0}\supset\Gamma_{1}\supset\dots\supset\Gamma_{N}\supset\dots be a sequence of congruence subgroups satisfying \bigcap_{N}\Gamma_{N}=\{0\} . Then \displaystyle\lim_{N\to\infty}\frac{\log|H^{i}(\Gamma_{N},L)_{\text{tors}}|}{[\Gamma:\Gamma_{N}]} exists for any i , and is equal to a computable constant c_{\Gamma,L} . Furthermore, c_{\Gamma,L} is non-zero exactly if \delta(G)=1 and i is the top of the cuspidal range. In the paper they prove an Euler-characteristic variant of this ( [ 11 ] , (1.4.2) ). Let X=G(\mathbb{R})/K . Then, under further assumptions, they show that \displaystyle\lim_{N\to\infty}\sum_{i}(-1)^{i+\frac{\operatorname{dim}X-1}{2}}\frac{\log|H^{i}(\Gamma_{N},L)_{\text{tors}}|}{[\Gamma:\Gamma_{N}]}=c_{\Gamma,L}. (1.1) Their assumptions include that \Gamma is cocompact in G(\mathbb{R}) , equivalently that G has \mathbb{Q} -rank 0 , and that L must be a strongly acyclic \Gamma -module. We will return to these assumptions in a moment. It is striking how significant the assumption that \delta(G)=1 seems to be, and how precise the conjecture and result are when it is satisfied. On the other hand, they both say very little about what happens when \delta(G)\neq 1 , except that there is not full exponential growth of torsion. In fact, as phrased in ( [ 8 ] , § 9 ), it is still an open question whether one should expect polynomial or subexponential growth in these cases. Inspired by this question, this paper establishes subexponential upper bounds on the size of torsion in the cohomology of certain families of subgroups. On the way we also establish results on second order terms when \delta(G)=1 . The results are obtained for non-cocompact subgroups, and we introduce this generalization now. 1.2. Generalizations and results on analytic torsion Since many important arithmetic subgroups are not cocompact, a lot of consequent work has focused on removing this assumption from ( 1.1 ). This was first successful for arithmetic groups associated to hyperbolic manifolds of finite volume, see [ 38 ] , [ 39 ] , [ 37 ] , [ 33 ] , and it was recently extended to other \mathbb{Q} -rank 1 arithmetic groups in [ 34 ] . All of the results quoted above are consequences of approximation theorems on analytic torsion, combined with Cheeger-Müller theorems , identities relating analytic torsion to Reidemeister torsion (see [ 35 ] , [ 14 ] , also [ 36 ] , [ 13 ] ). Analytic torsion and Reidemeister torsion are both invariants of Riemannian manifolds, the former built from the spectra of p -form Laplacians (see Section 4 ), and the latter from topological data of the manifold ( [ 41 ] , also [ 14 ] ). The consequence of Cheeger-Müller theorems is that one may investigate cohomology by analytic methods. At the time of the Bergeron–Venkatesh paper, there was no general definition of analytic torsion outside of compact Riemannian manifolds. To remedy this, Matz–Müller gave a definition and the necessary convergence results in [ 25 ] , [ 26 ] , [ 27 ] to show that a non-cocompact generalization of analytic torsion also satisfies the desired approximation theorem, namely ( [ 27 ] , Theorem 1.5 ). Our method of getting improved bounds and second order terms on torsion cohomology is sharpening the asymptotics of their approximation theorem. This is the main technical result of this paper, and its preparation and proof will occupy most of the paper. Let us present it here. We adopt the setting from the beginning of the previous section. For simplicity, we keep the assumption that G is semisimple, and present a more general result later (Theorem 9.2 ). We will further assume that \Gamma_{N}=\Gamma(N) , the principal congruence subgroup of level N in \Gamma . Assume for convenience that G has strong approximation. Set \tilde{X}=G(\mathbb{R})/K and define \displaystyle X(N)=\Gamma(N)\backslash\tilde{X}. For N large enough, this is then a locally symmetric manifold. Let d=\operatorname{dim}\tilde{X} . For \tau a finite-dimensional complex representation of G(\mathbb{R}) we then set T_{X(N)}(\tau) to be the analytic torsion of X(N) associated to \tau , defined in Section 4 . We also denote by T_{X(1)}^{(2)}(\tau) the L^{2} -torsion of X(1)=\Gamma\backslash\tilde{X} associated to \tau ( [ 22 ] ). We will need the notion of a \lambda -strongly acyclic representation of G(\mathbb{R}) given in Definition 3.1 , first introduced in [ 12 ] . The main theorem is the following. Theorem 1.2 . Let \tau be an irreducible and \lambda -strongly acyclic representation of G(\mathbb{R}) , for some \lambda>0 only depending on G . Assume that G satisfies the properties (TWN) and (BD) (as defined in [ 16 ] ). Assume either Conjecture 9.1 or that N varies over a set with fixed prime divisors (such as N=p^{k} , k\in\mathbb{N} ). Then there exists some a>0 such that \displaystyle\frac{\log T_{X(N)}(\tau)}{[\Gamma:\Gamma(N)]}=T^{(2)}_{X(1)}(\tau)+O\left(\frac{(\log N)^{a}}{N^{k(G)}}\right)\qquad\text{as}\;\;\;N\to\infty. Here k(G) is a positive computable integer depending on G defined in ( 8.5 ). We refer to Theorem 9.2 for the more general result. Conjecture 9.1 concerns the size of certain coefficients appearing in the Arthur trace formula. It is believed to hold for all reductive groups, and has been established for \operatorname{GL}(n) over number fields in [