Nodal intersections for random waves against a segment on the 3-dimensional torus (Q527389)

The main purpose of this article is to provide probabilistic estimates on the number of zeros of eigenfunctions of the Laplacian on the 3-dimensional torus \(\mathbb{T}^3=\mathbb{R}^3/\mathbb{Z}^3\), which lie on a line segment of length \(L\). The eigenvalues have the form \(4\pi^2 m\), where \(m\) is the sum of three squares of integers. The number \(N_m\) of such representations of \(m\) is the dimension of the corresponding eigenspace, as the standard orthonormal basis consists of exponential functions \(e^{2\pi i\langle \mu,x\rangle}\), where \(\mu\) takes values in \(\mathcal{E}(m)=\mathbb{Z}^3\cap\sqrt{m}\mathcal{S}^2\). The random arithmetic waves are eigenfunctions \[ F(X)=\frac{1}{\sqrt{N_m}}\sum_{\mu\in\mathcal{E}(m)}a_{\mu}e^{2\pi i\langle \mu,x\rangle} \] whose Fourier coefficients \(a_{\mu}\) with respect to this basis are independent Gaussian random variables (with the exception \(a_{-\mu}=\overline{a_{\mu}}\), as this paper deals with real valued functions). Denote by \(\mathcal{Z}=\mathcal{Z}(F)\) the number of zeros of \(F\) with eigenvalue \(4\pi^2 m\) on a smooth curve \(\mathcal{C}\) of length \(L\). The formula for the expectation was proven in [\textit{Z. Rudnick} et al., Ann. Inst. Fourier 66, No. 6, 2455--2484 (2016; Zbl 1360.60081)] and is given by \[ \mathbb{E}[\mathcal{Z}]=L\frac{2}{\sqrt{3}}\sqrt{m} \] Regarding the variance \(\mathrm{Var}(\mathcal{Z}/\sqrt{m})\), there are three different asymptotic estimates for \(m\not\equiv0,4,7\bmod 8\) given in the article, when \(\mathcal{C}\) is a line segment parallel to \(\alpha=(\alpha_1,\alpha_2,\alpha_3)\), depending on rationality conditions between the coordinates of \(\alpha\). The author obtains also the following stronger estimate: \[ \mathrm{Var}(\mathcal{Z}/\sqrt{m})\ll\frac{1}{m^{1/4-\varepsilon}} \] for every \(\varepsilon>0\) and any line segment \(\mathcal{C}\), regardless of the direction, conditional to a conjecture by \textit{J. Bourgain} and \textit{Z. Rudnick} on lattice points on spherical caps [Geom. Funct. Anal. 22, No. 4, 878--937 (2012; Zbl 1276.53047)]. This complements the aforementioned work by Rudnick, Wigman, and Yesha, who studied similar questions involving curves of nowhere vanishing curvature instead of segments.