The spectrum of Fermat curves (Q1190143)

Let \(\pi:\Gamma(2)\to H_ 1(\Gamma(2))\cong\mathbb{Z}^ 2\) be the canonical map and let \(R_ N:\mathbb{Z}^ 2\to(\mathbb{Z}/N\mathbb{Z})^ 2\) denote the reduction \(\bmod N\) where \(N\geq 1\) is an integer. The group \(\Phi(N):=\text{ker}(R_ N\circ\pi)\) defines an abelian Galois covering \(\Phi(N)\backslash\mathbb{H}\) of \(\Gamma(2)\backslash\mathbb{H}\) such that \(\overline{\Phi(N)\backslash\mathbb{H}}=F_ N\) where \(F_ N:x^ N+y^ N=1\) is the Fermat curve. The paper under review studies the eigenvalue theory of the hyperbolic Laplacian on \(L^ 2(\Phi(N)\backslash\mathbb{H})\). Since \(\Phi(N)\) is a congruence group if and only if \(N=1,2,4,8\), these groups constitute an interesting class of examples for investigations on the distribution of eigenvalues of \(\Delta\). Let \(m(\lambda,N)\) be the dimension of the eigenspace for the eigenvalue \(\lambda\) of \(-\Delta\) on \(\Phi(N)\) and denote by \(M(\lambda,N)\) the sum of the \(m(\mu,N)\) with \(0<\mu<\lambda\). The superscripts ``cusp'' and ``Eis'' indicate that only the corresponding cusp forms or residues of Eisenstein series are counted. In particular, \(M(1/4,N)\) is the number of exceptional eigenvalues of \(-\Delta\) on \(\Phi(N)\). Theorem 1. (i) There are continuous strictly increasing functions \(F,G:[0,1/4]\to\mathbb{R}\) with \(F(0)=G(0)=0\) such that as \(N\to\infty\) \[ M^{cusp}(\lambda,N)=F(\lambda)N^ 2-G(\lambda)N+o(N)\quad\text{ for } \lambda\in]0,1/4[, \] \[ M^{Eis}(\lambda,N)=G(\lambda)N^ 2+O(1)\quad\text{ for } \lambda\in]0,1/4]. \] (ii) \(\lambda_ 1(N)\sim 2/N\), \(\lambda_ 1^{cusp}(N)\sim 4/N\), where \(\lambda_ 1\) is the first non-trivial eigenvalue. (iii) For all but finitely many \(\lambda\in[0,1/4]\) we have \(m(\lambda,N)=O(N^{2/3})\). A more refined information based on numerical calculations (without detailed error analysis) is given in Theorem \(1'\). The authors' investigation of \(-\Delta\) on \(L^ 2(\Phi(N)\backslash\mathbb{H})\) rests upon an analysis of \(-\Delta\) on \(L^ 2(\Gamma(2)\backslash\mathbb{H},\chi)\) where \(\chi:\Gamma(2)\to U(1)\) is a character. Note that \(m(\lambda,N)=\sum_{\chi^ N=1}m(\lambda,\chi)\). A basic result is that \(\lambda_ 1(\chi)\geq 1/4\) for all \(\chi\). This reduces the study of the exceptional spectrum of \(\Phi(N)\) to the investigation of \(\lambda_ 0(\chi)\) and its level curves \(C_ \lambda=\{\theta:\lambda_ 0(\theta)=\lambda\}\) where \(\theta\in\mathbb{R}^ 2/\mathbb{Z}^ 2\) parametrizes the characters on \(\Gamma(2)\). (Part of the analysis of the \(C_ \lambda\) is postponed to a future paper of the authors.) The proof of Theorem 1 then follows from standard results about lattice points in convex regions and on curves. The authors also examine the spectrum of \(\Phi(N)\) on \([0,\infty[:\) Theorem 4. There is a strictly increasing function \(F(\lambda)\) on \([0,\infty[\) such that \[ M^{cusp}(\lambda,N)\sim F(\lambda)N^ 2\quad\text{ as } N\to\infty. \] The theoretical basis for the authors' computer calculations stems from a result due to Selberg the proof of which is reproduced in an appendix.