Sato-Tate equidistribution for families of automorphic representations through the stable trace formula (Q2114117): Difference between revisions

Sato-Tate equidistribution questions are a local version of Plancherel equidistribution theorems: they ask what can be said about the distribution of the local components of global (or adelic) objects, more precisely which limiting statistics of these local components can be established. These studies originated with Sato and Tate on the distribution of the Frobenius ``eigenangles'' of an elliptic curve. In the realm of classical Hecke eigenforms, \textit{P. Sarnak} [Clay Math. Proc. 4, 659--685 (2005; Zbl 1146.11031)] proved such equidistribution for Maass forms and \textit{J.-P. Serre} [J. Am. Math. Soc. 10, No. 1, 75--102 (1997; Zbl 0871.11032)] for modular forms. More recently, such results have been obtained in the realm of automorphic representations by \textit{S. W. Shin} and \textit{N. Templier} [Invent. Math. 203, No. 1, 1--177 (2016; Zbl 1408.11042)] for families of automorphic forms having local restrictions, precisely being in a fixed discrete series L-packet at the Archimedean place and having bounded non-Archimedean ramification. The author establishes analogous results when restricting the Archimedean component to a fixed discrete series representation instead of a whole L-packet as in Shin-Templier, and also allows for a nontrivial center. These features make possible to single out certain specific aspects of automorphic representations in applications of statistical results, for instance distinguishing between holomorphic and nonholomorphic discrete series. While the main tool in Shin-Templier work was the use of the invariant trace formula with a Euler-Poincaré function at the Archimedean place in order to write the Plancherel distribution as the spectral side of a simple trace formula, the author appeals to a hyperendoscopic version of the stable trace formula with a pseudocoefficient at the Archimedean place. For a more precise statement of the main result of the paper, let \(G\) be a reductive group over \(F\). Let \(U\) be an open compact subgroup of \(G(\mathbb{A}^{\infty, S})\) for some finite set of places \(S\), \(\xi\) a regular weight of \(G_{\mathbb{C}}\), \(\Pi_\mathrm{disc}(\xi)\) the discrete series L-packet corresponding to \(\xi\) and \(\pi_0 \in \Pi_\mathrm{disc}(\xi)\). The family of automorphic forms under consideration is \[ \mathcal{F}_{U, \infty} = \{\pi \in \mathcal{A}_{\mathrm{disc}}(G) \ : \ \pi_\infty \simeq \pi_0, \ \dim(\pi^\infty)^U \geq 1\} \] where \(\mathcal{A}_{\mathrm{disc}}(G)\) denotes the space of all the discrete automorphic representations of \(G\). Equidistribution questions can be phrased as the study of the distribution measure \[ \mu_{\mathcal{F}_{U, \infty}} := \sum_{\pi \in \mathcal{F}_{U, \infty}} a_\pi \delta_{\pi_S}\tag{1} \] where the \(a_\pi\) are the natural spectral weights \[ a_\pi = m_{\mathrm{disc}}(\pi) \dim(\pi^{S, \infty})^U. \] Shin and Templier studied analogous measures (but with the prescribed condition \(\pi_\infty \in \Pi_{\mathrm{disc}}(\xi)\)) in the weight aspect (\(\xi \to \infty\)) and in the level aspect (\(U \to 1\)) separately, and proved that the limiting measure is the Plancherel measure on the dual group \(\hat{G}_S\). In fact they even obtained a quantitative version of this statement, including an explicit error term. The author of the present article gives a variation of this result, essentially by allowing for a nontrivial center and requiring that \(\pi_\infty \simeq \pi_0\). His result is the following statement, concerning the weight aspect (\(\xi \to \infty\)). Theorem. Let \((\pi_k)_k\) be a sequence of discrete series representations of \(G_\infty\) such that the Archimedean components have regular weights going to infinity. Denote \(S\) a certain finite sets of finite places. There are constants \(A\) and \(B\) such that for any \(\phi_{S}\) in the Hecke algebra \(\mathcal{H}(G_{S}, K_{S}, \chi_{S})^{\leq \kappa}\), we have \[ \frac{\bar{\mu}^{\mathrm{can}}(U_{\mathfrak{X}}^{S, \infty}) |\Pi_{\mathrm{disc}(\xi_k)}|}{\tau'(G) \dim(\xi_k)} \sum_{\pi \in \mathcal{A}_{\mathrm{disc}}(G, \chi)} a_{\mathcal{F}_k}(\pi) \widehat{\phi}_S(\pi) = E(\widehat{\phi}_S | \omega_\xi, L, \chi_S) + O(q_{S_1}^{A+B\kappa} m(\xi_k)^{-1}) \] and the implied constants only depend on \(G\), \(\mathfrak{X}\), \(\chi\), \(\phi_{S}\) and \(U^{S, \infty}\). Here, the main term \(E(\widehat{\phi}_S | \omega_\xi, L, \chi_S)\) is a conditional Plancherel expectation, i.e., an average of measures of the form \(\mathrm{d}\mu_{\hat{f}}/\mathrm{d}\mu^{\mathrm{pl}}_{Z_G}\) where \(\mathrm{d}\mu_{\hat{f}}\) is the pushed forward of \(\hat{f}\mathrm{d}\mu^{\mathrm{pl}}\) to \(\hat{Z_G}\), and \(\mathrm{d}\mu^{\mathrm{pl}}\) denotes the Plancherel measure. Also, \(S_1\) denotes a subset of finite places in \(S\) such that \(U_{S_1}\) is a maximal compact subgroup, therefore ensuring the selected representations to be unramified at \(S_1\). The error term in particular has explicit dependence upon both the Archimedean weight \(m(\xi_k)\) and the prescribed unramified finite places \(S_1\) of the test-function. This result leads to obtain various consequences, namely Plancherel and Sato-Tate equidistribution results under various restrictions. The approach to prove these results is via trace formula methods. A rough outline of the strategy is as follows. The trace formula is a distributional equality between two distributions \[ I_{\mathrm{spec}}(\phi) = I_{\mathrm{geom}}(\phi) \] where \(\phi\) is in a certain space of functions \(\mathcal{H}(G)\), \(I_{\mathrm{spec}}\) is a sum over the automorphic spectrum of modified traces of \(\phi\) against components of \(L^2(G(\mathbb{Q})\backslash G(\mathbb{A}))\), and \(I_{\mathrm{geom}}\) is a sum over conjugacy classes of rational points of certain modified orbital integrals. The distributing measure (1) is then studied in two steps: \begin{itemize} \item identifying or approaching the measure (1) by a spectral side \(I_{\mathrm{spec}}(\phi)\) for a suitable function \(\phi \in \mathcal{H}(G)\) ; \item studying the geometric terms in \(I_{\mathrm{geom}}(\phi)\), that are expected to behave mostly like the sole contribution of the central elements, and where the other terms contribute as an error term. \end{itemize} Both steps are generally difficult and involve deep results about harmonic analysis on reductive groups, representation theoretic and automorphic results about the involved automorphic representations, and a fine geometric, arithmetic and/or analytic study of the geometric side. This strategy is indeed the heart of the results of both Shin-Templier and the author. The invariant trace formula of Arthur is difficult to state involving both a continuous part of the spectrum (splitting into a continuous integral along Eisenstein series) and weighted unipotent orbital integrals (multiplied by global coefficients). With some extra conditions (called \textit{stable cuspidality}) on the test function \(\phi\) (hence, inducing conditions on the automorphic family effectively selected by \(I_{\mathrm{spec}}(\phi)\)), these two more difficult parts disappear: this is the so-called \textit{simple trace formula} of \textit{J. Arthur} [J. Am. Math. Soc. 1, No. 3, 501--554 (1988; Zbl 0667.10019)]. Such restrictions are chosen in the work of Shin-Templier as well as in the author's work and is the reason of the two strong assumptions in the family under consideration: \begin{itemize} \item being a discrete series at the Archimedean place; \item having prescribed ramification at some finite places. \end{itemize} We emphasize some differences, difficulties and points of interest (mainly compared to Shin-Templier). An easier entrance door towards these methods may be the article of \textit{J. Binder} [Isr. J. Math. 222, No. 2, 973--1028 (2017; Zbl 1431.11063)] that we suggest to those unfamiliar with the methods and willing a guide to dive into the details. An important difference between this article and the work of Shin and Templier arises in the fact that the author selects a single representation at the Archimedean place instead of an L-packet. This requires to take a pseudocoefficient for the Archimedean test-function in the trace formula, while Shin and Templier take the Euler-Poincaré test-function of \textit{L. Clozel} and \textit{P. Delorme} [Ann. Sci. Éc. Norm. Supér. (4) 23, No. 2, 193--228 (1990; Zbl 0724.22012)]. A key feature in Shin-Templier is that the Euler-Poincaré function satisfies the stable cuspidality conditions, reducing the Arthur trace formula to its simple form, which is not the case with the pseudocoefficient in the author's work. To address this difficulty, the idea is to use the stable trace formula to write the invariant distribution \(I^G(f)\) attached to such a function as a linear combination of \textit{stable} distributions of (transfers of) this function on smaller endoscopic subgroups, in the form \[ I^G(f) = \sum_{H \in \mathcal{E}_{\mathrm{ell}}} S^H(f^H), \] where \(\mathcal{E}_{\mathrm{ell}}\) is a set of certain endoscopic subgroups. A crucial fact is to notice that the pseudocoefficient and the Euler-Poincaré function have the same stable orbital integral, so that in the above equation the transfer \(f^H\) can be modified to have Euler-Poincaré component at the Archimedean place. The next step is to appeal to (a generalization of) \textit{A. Ferrari}'s hyperendoscopic formula [Manuscr. Math. 124, No. 3, 363--390 (2007; Zbl 1146.58012)]. The hyperendoscopic formula is a rearrangement of the stabilized trace formula featuring usual \textit{invariant} distributions. It takes the form \[ I^G(f) = I^G(f^\star) + \sum_{\mathcal{H} \in \mathcal{HE}_{\mathrm{ell}}} \iota(G, \mathcal{H}) I^{\mathcal{H}}((f-f^\star)^{\mathcal{H}}), \] where \(\mathcal{HE}_{\mathrm{ell}}\) is a set of certain endoscopic subgroups, \(f^\star\) is a function having the same stable orbital integrals as \(f\), and the \(\iota(G, \mathcal{H}) \) are harmless constants. The interest of appealing to the hyperendoscopic trace formula is that it features invariant orbital integrals on the geometric side, so that the machinery developed by Shin and Templier remains available. Moreover, the transfers \((f-f^\star)^{\mathcal{H}}\) can in fact be taken to be explicit linear combinations of Euler-Poincaré functions, for which bounds established by Shin and Templier can be used. It has to be noted that fine work is still required in order to obtain estimates that are uniform in the endoscopic subgroup \(\mathcal{H}\) and to deal with the existence of central characters.