Arithmetic, zeros, and nodal domains on the sphere (Q2353997)

In the paper under review, the author studies the growth of the nodal domains for Hecke-Laplace eigenfunctions on the \(2\)-dimensional unit sphere \(S^2\). Let \(f\) be an eigenfunction of the spherical Laplacian \(\Delta_{S^2}\) of eigenvalue \(\ell(\ell+1)\). Let also \(Z(f) = f^{-1}(0)\) be the nodal set of \(f\). A connected component of \(S^2-Z(f)\) is called a nodal domain of \(f\), and the number of nodal domains of \(f\) is denoted by \(\mathcal{N}(f)\). For \(f\) a spherical eigenfunction of eigenvalue \(\ell (\ell+1)\), \textit{R. Courant} and \textit{D. Hilbert} [Methods of mathematical physics. Vol. I. Translated and revised from the German original. First English ed. New York: Interscience Publishers (1953; Zbl 0053.02805)] proved that \[ \mathcal{N}(f) \leq (\ell+1)^2. \] Let \(\mathcal{H}_{\ell}\) denote the space of spherical eigenfunctions of eigenvalue \(\ell (\ell+1)\). \textit{H. Lewy} [Commun. Partial Differ. Equations 2, 1233--1244 (1977; Zbl 0377.31008)] proved that \(\mathcal{N}(f) \leq 3\) for infinitely many \(\ell\) and \(f \in \mathcal{H}_{\ell}\), hence one cannot obtain non-trivial lower bounds for the growth of \(\mathcal{N}(f)\) for every \(f \in \mathcal{H}_{\ell}\). The author proves non-trivial lower bounds for \(N(f)\) when \(f\) is taken from special subsets of \(\mathcal{H}_{\ell}\), namely for eigenfunctions of certain Hecke operators arising from a certain maximal order in the Hamilton quaternions. The author considers the order \[ \mathcal{O} = \left\{ \frac{ \alpha+\beta \mathrm{i} + \gamma \mathrm{j} + \delta \mathrm{k}}{2} : \alpha, \beta, \gamma, \delta \in \mathbb{Z}, \alpha \equiv \beta \equiv \gamma \equiv \delta \mod 2 \right\}. \] The units \(\mathcal{O}^{\times}\) of \(\mathcal{O}\) form a group of order \(24\). For \(m\) odd, let \(T_m\) denote the \(m\)-th Hecke operator arising from \(\mathcal{O}\), and \(\mathcal{H}_{\ell}^{\mathcal{O}^{\times}}\) the space of real valued on \(S^2\) which are \(\Delta_{S^2}\) and \(T_m\) invariant. As a first result, the author proves the following theorem. Theorem 1. Let \(\ell\) be even and \(\mathcal{B}_{\ell}\) denote an orthonormal basis for \(\mathcal{H}_{\ell}^{\mathcal{O}^{\times}}\) consisting of Hecke eigenfunctions. Then for any \(\varepsilon >0\) we have the lower bound for the total number of nodal domains amongst members of \(\mathcal{B}_{\ell}\) \[ \sum_{\phi \in \mathcal{B}_{\ell}} \mathcal{N}(\phi) \gg_{\varepsilon} \ell^{5/4-\varepsilon}. \] As a corollary, the number of nodal domains grow on average. With everything as in Theorem 1 we have for the expected number of nodal domains amongst members of \(\mathcal{B}_{\ell}\), for any \(\varepsilon >0\) \[ \sum_{\phi \in \mathcal{B}_{\ell}} \mathcal{N}(\phi) \gg_{\varepsilon} \ell^{1/4-\varepsilon} |\mathcal{B}_{\ell}|. \] In the negative curvature case, \textit{A. Ghosh} et al. [Geom. Funct. Anal. 23, No. 5, 1515--1568 (2013; Zbl 1328.11044)] proved, under a Lindelöf hypothesis assumption, uniform lower bounds for the nodal domains on the modular surface. If \(\phi\) is a Maass form for the \({\mathrm{SL}_2( {\mathbb Z})} \backslash \mathbb{H}\) of eigenvalue \(\lambda>0\), they proved that, assuming the Lindelöf hypothesis for \(L(1/2+it,\phi)\), for every \(\varepsilon >0\) we have \[ \mathcal{N}(\phi) \gg_{\varepsilon} \lambda^{1/24 -\varepsilon}. \] The second theorem of the paper, is the analogue of the Ghosh-Reznikov-Sarnak result in the case of the sphere \(S^2\). Theorem 2. Assuming the generalized Lindelöf hypothesis, for any \(\varepsilon>0\) we have \[ \mathcal{N}(\phi) \gg_{\varepsilon} \ell^{1/12-\varepsilon} \] when \(\ell\) is even and \(\phi \in \mathcal{H}_{\ell}^{\mathcal{O}^{\times}}\) is a Hecke eigenfunction. The main tools in the proof of Theorem 1 are explicit kernel estimates in a twisted pre-trace formula for the sphere applied to a basis of Hecke eigenfunctions, combined with lattice counting estimates. The proof of Theorem 2 is analogous to the Ghosh-Reznikov-Sarnak proof. We would like to emphasize few key results in the proof of this theorem. The first is a subconvex result of \textit{J. M. VanderKam} [Int. Math. Res. Not. 1997, No. 7, 329--347 (1997; Zbl 0877.58056); correction ibid. 1998, No. 1, 65--66 (1998)] for the \(L^{\infty}\)-norm of the spherical eigenfunctions: \[ \| \phi\|_{L^{\infty}(S^2)} \ll_{\varepsilon} \ell^{5/12+\varepsilon} \| \phi \|_{L^2(S^2)} \] for a Hecke eigenform \(\phi \in \mathcal{H}_{\ell}^{\mathcal{O}^{\times}}\) (this is analogous to the Iwaniec-Sarnak bound from [\textit{H. Iwaniec} and \textit{P. Sarnak}, Ann. Math. (2) 141, No. 2, 301--320 (1995; Zbl 0833.11019)] for the \(L^{\infty}\)-norm of a Maass form on the modular surface). The second result is a lower bound for the \(L^2\)-norm of the restriction of a spherical eigenfunction on a fixed geodesic \(\gamma\) on \(S^2\): \[ \| \phi|_{\gamma}\|_{L^2(\gamma)} \gg \| \phi\|_{L^2(S^2)}. \] The \(L^2\)-norm of the geodesic restriction of the spherical eigenfunction \(\phi\) is then related via a Waldspurger-type formula to the central value \(L(1/2, \Pi \otimes \Omega_n)\) of an \(L\)-function arising from an automorphic representation of \(\mathrm{GL}_2(\mathbb{A})\), and the proof of Theorem 2 follows from a geometric argument and the generalized Lindelöf hypothesis. The paper is well written, and the reviewer enjoyed the exposition of the arguments and the fruitful ideas involved in the proofs of the main results. The author has also included in the Introduction a short discussion on the random wave model and the quantum chaos, and their relation with the theory of automorphic forms.