Effective equidistribution of eigenvalues of Hecke operators (Q1011686)

Let \(S(N,k)\) be the space of cusp forms of even weight \(k\) on \(\Gamma_0(N)\), and let \(s(N,k)\) be its dimension. For primes \(p\) and \(1\leq i\leq s(N, k)\), let \(a_{p,i}\) be the eigenvalues of the \(p\)th Hecke operator acting on \(S(N,k)\). By Deligne's theorem, \(a_{p,i}p^{-(k-1)/2}\) lies in the interval \([-2, 2]\). The Sato-Tate conjecture predicts that in the generic case these numbers are equidistributed in \([-2, 2]\) with respect to the measure \(\mu_\infty= {1\over\pi}(1- (x/2)^2)^{1/2}\) when \(p\) varies and \(N\), \(k\) are fixed. In [J. Am. Math. Soc. 10, No. 1, 75--102 (1997; Zbl 0871.11032)] \textit{J.-P. Serre} proved that these quantities are equidistributed in \([-2,2]\) with respect to a certain measure \(\mu_p\) when \(p\) is fixed and \(N\), \(k\) run through a sequence for which \(N+ k\to\infty\). Serre's theorem and its proof do not yield effective bounds for error terms. In the paper under review the authors derive an effective version of Serre's theorem. Their main result states that for primes \(p\nmid N\) and every subinterval \([\alpha,\beta]\) of \([-2,2]\) one has \[ {1\over s(N,k)}\#\left\{1\leq i\leq s(N, k)\mid a_{p,i} p^{-(k-1)/2}\in [\alpha, \beta]\right\}= \int^\beta_\alpha \mu_p+ O\Biggl({\log p\over\log kN}\Biggr), \] where the implied constant is effectively computable. There are several applications: For a given \(d\) and prime \(p\nmid N\) one can estimate the number of eigenvalues \(a_{p,i}\) which are algebraic integers of degree \(\leq d\). There is an effectively computable constant \(B_d\) such that \(J_0(N)\), the Jacobian of the modular curve \(X_0(N)\), is isogenous to a product of \(\mathbb Q\)-simple abelian varieties of degree \(\leq d\). The authors also study the effective equidistribution of eigenvalues of the Frobenius acting on a family of curves over a fixed finite field as well as the eigenvalue distribution of adjacency matrices of families of regular graphs. The results are derived from a general distribution theorem.