Elliptic curves, random matrices and orbital integrals (Q2520782)

Let \(p\) be a prime and let \(\mathbb F_p\) denote the prime field of \(p\) elements. If \(E/\mathbb{F}_p\) is an elliptic curve, then its characteristic polynomial of the Frobenius has the form \(f_{E/\mathbb F_p}(T)=T^2-a_{E/\mathbb F_p}T+p\), where \(|a_{E/\mathbb F_p}|\leq 2\sqrt{p}\) which depends only on the isogeny class of \(E/\mathbb F_p\). For a given integer \(a\) with \(|a|\leq 2\sqrt{p}\), the set \[ I(a,p):=\{\, E/\mathbb F_p\,|\, a_{E/\mathbb F_p}=a\,\} \] is a single isogeny class of elliptic curves over \(\mathbb F_p\). Its weighted cardinality is \[ \#\tilde{I}(a,p):=\sum_{E\in I(a,p)} \displaystyle\frac{1}{\#\text{Aut}(E)}. \] \textit{E.-U. Gekeler} [Int. Math. Res. Not. 2003, No. 37, 1999--2018 (2003; Zbl 1104.11033)] obtained the formula for \(\#\tilde{I}(a,p)\): \[ \#\tilde{I}(a,p)=\displaystyle\frac{1}{2}\sqrt{p} \nu_{\infty}(a,p)\prod_{\ell}\nu_{\ell}(a,p) \] where, for each rational prime \(\ell\neq p\), \[ \nu_{\ell}(a,p)=\text{lim}_{n\to\infty} \displaystyle\frac{\#\{\,\gamma\in \text{GL}_2(\mathbb Z/\ell^n)\,|\, \text{tr}(\gamma)\equiv a \,\text{mod}\,\ell^n,\, \text{det}(\gamma) \equiv p\,\text{mod}\,\ell^n\,\}}{\#\text{SL}_2(\mathbb Z/\ell^n)/\ell^n}, \] for \(\ell=p\), \(\nu_p(a,p)\) is defined analogously with \(p\) in place of \(\ell\) in the above formula, and \[ \nu_{\infty}(a,p)=\displaystyle\frac{2}{\pi} \sqrt{1-\displaystyle\frac{a^2}{4p}}. \] The right hand of the formula for \(\#\tilde{I}(a,p)\) is known as the Gekeler random matrix model. The main goal of this paper is to provide a conceptual explanation of the Gekeler formula, and extend it to ordinary elliptic curves over a finite field \(\mathbb F_q\) with \(q\) a power of \(p\). The numbers \(\nu_{\ell}(a,q)\), \(\nu_p(a,q)\) and \(\tilde{I}(a,q)\) are defined similarly replacing \(p\) by \(q\). The authors method relies on the Langlands-Kottwitz formula of the points on a Shimura variety over a finite field. The cardinality of an ordinary isogeny class of elliptic curves over \(\mathbb F_q\) can be computed using orbital integrals on the finite adelic points of \(\text{GL}_2\). A main observation is to relate each local factor \(\nu_{\ell}(a,q)\) to an orbital integral \(\int_{G_{\gamma_{\ell}}(\mathbb Q_{\ell})\backslash \text{GL}_2(\mathbb Q_{\ell})} \mathbf{1}_{\text{GL}_2(\mathbb Z_{\ell})} (x^{-1}\gamma_{\ell}\,x) \,dx\), where \(\gamma_{\ell}\) is an element of \(\text{GL}_2(\mathbb Q_{\ell})\) of trace \(a\) and determinant \(q\), \(G_{\gamma_{\ell}}\) is its centralizer in \(\text{GL}_2(\mathbb Q_{\ell})\), and \(\mathbf{1}_{\text{GL}_2(\mathbb Z_{\ell})}\) is the characteristic function of the maximal compact subgroup \(\text{GL}_2(\mathbb Z_{\ell})\). The choice of the invariant measure \(dx\) on the orbit forms a technical core of the paper.