Counting locally supercuspidal newforms (Q6937283)

The Eichler-Selberg trace formula provides an explicit formula for the trace of a Hecke operator \(T_n\) acting on a space of cuspidal modular forms \(S_k(N,\psi)\). Later this was reproved with adelically by interpreting the trace of \(T_n\) as the trace of an operator \(R(\phi)\) on a space of automorphic forms for a suitable test function \(\phi = \otimes \phi_v\). When the nebentypus \(\psi\) is trivial and \(p\) is a finite place, one can take \(\phi_p\) to be the characteristic function of a suitable compact open subgroup \(K_p(N)\) of \(\mathrm{GL}_2(\mathbb Q_p)\) to pick out the \(p\)-part of the global level \(N\). Computing the trace of \(R(\phi)\) reduces to determining the finite collection of global orbital integrals that contribute and computing the corresponding local orbital integral.\N\NThe present paper aims to compute traces of Hecke operators on subspaces of \(S_k(N,\psi)\) with given local ramified components of sufficiently nice type. Namely, suppose \(N = S^2 T^3\) where \(S, T\) are squarefree coprime integers, and for each \(p \mid N\) fix a supercuspidal representation \(\sigma_p\) of conductor exponent \(v_p(N)\). In other words, \(\sigma_p\) is a depth 0 supercuspidal if \(p \mid S\) and a simple supercuspidal if \(p \mid T\). Let \(S_k(\hat \sigma)\) be the subspace of \(S_k(N,\psi)\) with local component \(\sigma_p\) for each \(p \mid N\). We remark that when \(\psi\) is trivial, every minimal newform \(f \in S_k(N,\psi)\) lies in a unique \(S_k(\hat \sigma)\).\N\NTheorem 1.1 gives a (rather involved) formula for the trace of \(T_n\) on \(S_k(\hat \sigma)\) assuming \((n,N) = 1\) and \(k > 2\). This is achieved by computing the adelic trace formula using test functions \(\phi\) such that \(\phi_p\) is the matrix coefficient of \(\sigma_p\) for \(p \mid N\). The main technical hurdle is to compute the local elliptic orbital integrals for such \(\phi_p\). The reason to exclude \(k=2\) is to avoid convergence issues.\N\NTheorems 1.2 and 1.3 specialize to this formula to cases where \(n=1\), \(\psi = 1\), and either \(S=1\) or \(T=1\). In this case, the trace formula computes \(\dim S_k(\hat \sigma)\). These formulas turn out to be rather simple, and refine various existing dimension formulas. Some connections are made with biases of root numbers. We remark that asymptotic dimension formulas (i.e., the main term in the trace formula) were already known, so the novel aspect is an explicit computation of the secondary terms, which is considerably harder. One can view the secondary terms as a bias toward certain local components in \(S_k(N,\psi)\).\N\NThe author emphasizes that the main value of the paper is in the methodology, which in principle should apply to more general situations. The paper is well written and should be a useful reference to researchers interested in explicit trace formulas.\N\N\textit{Remark:} A subsequent joint paper with the reviewer [\textit{A. Knightly} and \textit{K. Martin}, ``Counting newforms with prescribed ramified supercuspidal components'', Preprint, \url{arXiv:2511.14587}] extends Theorem 1.2 to allow for supercuspidal components of arbitrary odd conductor.