Typical hyperbolic surfaces have a spectral gap greater than 2 / 9 − ϵ 2/9-\epsilon
Nalini Anantharaman 1 and Laura Monk 2 · 2026
Read Typical hyperbolic surfaces have a spectral gap greater than 2 / 9 − ϵ 2/9-\epsilon by Nalini Anantharaman 1 and Laura Monk 2 (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-2604-09792.txt.
Introduction
Anantharaman and Monk show a spectral gap of nearly two ninths for typical hyperbolic surfaces drawn from the Weil-Petersson probability measure. This intermediate result leads toward their proof of the optimal spectral gap of nearly a quarter, building on the first part of this series. Much of the proof uses explicit inclusion-exclusion counting to rule out tangles at the precision of one over the genus.
Catalog record
| CREATOR | Nalini Anantharaman 1 and Laura Monk 2 |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 82972 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.09792
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Typical hyperbolic surfaces have a spectral gap greater than 2 / 9 − ϵ 2/9-\epsilon Nalini Anantharaman 1 and Laura Monk 2 Abstract . In this article, we prove that typical hyperbolic surfaces, sampled with the Weil–Petersson probability measure, have a spectral gap at least 2 / 9 − ϵ 2/9-\epsilon . This is an intermediate result on the way to our proof of the optimal spectral gap 1 / 4 − ϵ 1/4-\epsilon , building on the results of the first part of this series, [ 2 ] . A significant part of the proof is an explicit inclusion-exclusion argument to exclude tangles at the level of precision 1 / g 1/g . 2020 Mathematics Subject Classification Primary 58J50, 32G15; Secondary 05C80, 11F72 1. Introduction This article is part of our series of work [ 2 , 3 , 4 ] , ultimately proving that the spectral gap of typical hyperbolic surfaces is near-optimal, i.e. \forall\epsilon>0,\quad\lim_{g\rightarrow\infty}\mathbb{P}_{g}^{\mathrm{\scriptsize{WP}}}\left(\lambda_{1}\geq\frac{1}{4}-\epsilon\right)=1 where \mathbb{P}_{g}^{\mathrm{WP}} is the Weil–Petersson probability on the moduli space of hyperbolic surfaces of genus g . This shows that typical hyperbolic surfaces have an optimal spectral gap, due to a result of Huber [ 7 ] . This conclusion is reached at the end of [ 4 ] , by proving a Friedman–Ramanujan type result for the volume functions associated to any fixed type of closed geodesics. This shorter paper shows how the results from [ 2 ] already allow to prove the following partial result, namely: Theorem 1.1 . For any \epsilon>0 , \lim_{g\rightarrow\infty}\mathbb{P}_{g}^{\mathrm{\scriptsize{WP}}}\left(\lambda_{1}\geq\frac{2}{9}-\epsilon\right)=1. This beats the previous record of 3/16-\epsilon by Wu–Xue [ 14 ] and Lipnowski–Wright [ 8 ] , but does not yet reach the optimal 1/4-\epsilon proven in the follow-up article [ 4 ] . The reason behind this incremental progress is that we perform asymptotic expansions in powers of 1/g , the 3/16 result corresponding to the leading order 1/g^{0} , the 2/9 to the second order 1/g , and the 1/4 to arbitrary orders 1/g^{N} . We prove 1.1 by a classic trace method, as sketched in [ 2 , Section 3] , performing all computations at the second order in the expansion in powers of 1/g . Our computations are explicit and we gain a precise understanding of the problem at this level of precision. Since this article is in direct continuation with [ 2 ] , we will not re-introduce the notations and results it contains. The main novel challenge here is that we need to make a probabilistic hypothesis in order to remove tangles , which requires to adapt the methods developed in [ 2 ] to more general averages that include an indicator function. This work is done in Sections 2 to 5 . We then conclude to the proof of 1.1 in Section 6 . In the final proof of the optimal result [ 4 ] , we use a different, more abstract approach to the removal of tangles, based on the Moebius formula [ 3 ] . We believe it is interesting to present these different approaches to remove tangles, and how this removal can be done explicitely at this level of precision. 2. The tangle-free hypothesis and its geometric implications In [ 2 , Section 3] , we present a detailed trace method, which is the method we follow in this article. We relate the spectral gap problem to the question of proving that the expansion (f_{k}^{\mathbf{T}})_{k\geq\chi(\mathbf{T})} of the volume functions V_{g}^{\mathbf{T}} satisfy the Friedman–Ramanujan property . We prove that: • for any local topological type \mathbf{T} with \chi(\mathbf{T})\leq 1 , any k\geq\chi(\mathbf{T}) , the function \ell\mapsto\ell f_{k}^{\mathbf{T}}(\ell) is a Friedman–Ramanujan function in the weak sense (see [ 2 , Theorem 1.13] ); • however, the sum \sum_{\mathbf{T}:\chi(\mathbf{T})\leq 1}f_{1}^{\mathbf{T}}(\ell) is not a Friedman–Ramanujan function in the weak sense (see [ 2 , Section 9] ). We prove the second point by exhibiting a subset of surfaces of genus g , which are tangled , and hence have a small spectral gap. We prove that the probability of this set of surfaces only decays like 1/g as g\to\infty . In particular, we observe that even though these surfaces have vanishing probability, they contribute large contributions to the averages over the whole moduli space, due to their exponential proliferation of closed geodesics. This motivates the need to make a probabilistic hypothesis, removing some “bad” surfaces containing many geodesics. Let us introduce a set of “good surfaces”. Notation 2.1 . Let 0<\kappa<1 . Let us consider the counting functions \displaystyle N_{\kappa}^{\mathrm{inj}}(X) \displaystyle=\#\{\beta\in\mathcal{G}(X):\,\ell_{X}(\beta)\leq\kappa\} \displaystyle N_{R}^{\mathrm{tang}}(X) \displaystyle=\#\{Y\text{ embedded in }X:\,\chi(Y)=1,\ell_{X}^{\mathrm{max}}(\partial Y)\leq R\}, where: • \mathcal{G}(X) is the set of primitive oriented closed geodesics on X , and for \beta\in\mathcal{G}(X) , \ell_{X}(\beta) denotes its length; • any embedded subsurface Y is assumed to have geodesic boundary; • \chi(Y)>0 denotes the absolute Euler characteristic of Y (i.e. the absolute value of its Euler characteristic); • \ell_{X}^{\mathrm{max}}(\partial Y) denotes the length of the longest boundary component of Y ; • in N_{R}^{\mathrm{tang}} , amongst pair of pants, we only count those with three boundary components forming a multi-curve, i.e. we exclude the possibility that two components are glued into a once-holed torus (because this once-holed torus is already counted once in N_{R}^{\mathrm{tang}} ). We denote as \mathrm{TF}_{g}^{\kappa,R} the set of hyperbolic surfaces of genus g such that N_{\kappa}^{\mathrm{inj}}=N_{R}^{\mathrm{tang}}=0 . We call \kappa - short loops the closed geodesics counted by N_{\kappa}^{\mathrm{inj}} , and R - tangles the embedded surfaces counted by N_{R}^{\mathrm{tang}} . Remark 2.2 . The same set of good surfaces appears in [ 8 ] , and making the probabilistic assumption that X\in\mathrm{TF}_{g}^{\kappa,R} is a crucial step of Lipnowski–Wright’s proof of the 3/16-\epsilon spectral gap result. For us, the parameter \kappa will be an arbitrary small fixed positive number, and we will take R=\kappa\log(g) . The probability of \mathrm{TF}_{g}^{\kappa,R} can then be estimated directly using [ 9 , Theorem 4.2] and [ 11 , Theorem 5] . Lemma 2.3 . For any small enough \kappa>0 , any large enough g , R=\kappa\log(g) , 1-\mathbb{P}_{g}^{\mathrm{\scriptsize{WP}}}\left(\mathrm{TF}_{g}^{\kappa,R}\right)=\mathcal{O}\left(\kappa^{2}+g^{\frac{3}{2}\kappa-1}\right). We now prove that, under the probability assumption X\in\mathrm{TF}_{g}^{\kappa,R} , the number of geodesics of length \leq L:=A\log(g) filling any pair of pants or once-holed torus is radically reduced, from being exponential in L to polynomial in L . Lemma 2.4 . Let 0<\kappa<1 , A\geq 1 . For any large enough g , if we set R=\kappa\log(g) and L=A\log(g) , then there exists a set \mathrm{Loc}_{g}^{\kappa,A} of local topological types such that: • for any X\in\mathrm{TF}_{g}^{\kappa,R} , any primitive closed geodesic of length \leq L filling a surface of absolute Euler characteristic 1 is locally equivalent to a local type in \mathrm{Loc}_{g}^{\kappa,A} ; • the cardinal of the set satisfies (2.1) \#\mathrm{Loc}_{g}^{\kappa,A}=\mathcal{O}_{\kappa,A}\left((\log g)^{c_{\kappa,A}}\right) for a constant c_{\kappa,A}>0 depending only on \kappa and A . Proof. Let X\in\mathrm{TF}_{g}^{\kappa,R} , and let \gamma be a primitive closed geodesic on X of length \leq L filling a pair of pants Y . Let b_{1},b_{2},b_{3} denote the three boundary components of Y , ordered in non-decreasing length. By definition of \mathrm{TF}_{g}^{\kappa,R} , \ell_{X}(b_{1}),\ell_{X}(b_{2})\geq\kappa\quad\mathrm{and}\quad\ell_{X}(b_{3})\geq R. Let \alpha denote the simple orthogeodesic of Y from b_{3} to itself. We fix a base point x_{0} on \alpha , and a_{1} , a_{2} two simple paths based