Exponents for the equidistribution of shears and applications (Q2009147)

In an earlier work [\textit{D. Kelmer} and \textit{A. Kontorovich}, Math. Ann. 370, No. 1--2, 381--421 (2018; Zbl 1400.11105)], the authors of the paper under review introduced ergodic methods to prove the effective (i.e. with power saving error terms) equidistribution of ``shears'' in cusped hyperbolic surfaces. In this paper, they study the same problem but now allow full use of the spectral theory of automorphic forms to produce explicit exponents, and uniformity in parameters, also giving applications to counting square values of quadratic forms. They also take the opportunity to correct an error in the analysis in loc. cit., which has no effect on the qualitative power gain, but does affect the exponents as explicitly quantified there. To state their results, let \(\Gamma < G:=\mathrm{PSL}_2(\mathbb{R})\) be a non-uniform lattice, and let \(\mathbf{x}_0\in T^1(\Gamma\backslash \mathbb{H})\cong \Gamma\backslash G\) be a base point in the unit tangent bundle of the punctured surface \(\Gamma\backslash \mathbb{H}\), so that the visual limit point \(\mathfrak{a}=\lim_{t\to\infty} \mathbf{x}_0\cdot a_t\) is a cusp of \(\Gamma\), where \(a_t=\mathrm{diag}(e^{t/2},e^{-t/2})\) is the geodesic flow. With \(A^+=\lbrace a_t:t>0\rbrace\), the \textbf{shear} of the cuspidal geodesic ray \(\mathbf{x}_0\cdot A^+\) is defined to be its left translate by \[ \mathfrak{s}_T:=\begin{pmatrix} (T^2+1)^{-1/4} & T(T^2+1)^{-1/4} \\ 0 & (T^2+1)^{1/4} \end{pmatrix}. \] A special and interesting case of one of the main theorems they prove is: Theorem 1.1. Assume the Ramanujan conjecture for the exponent bounding the Fourier coefficients of Maaß forms on \(\Gamma\). Let \(K=\mathrm{SO}(2)\) denote the standard maximal compact subgroup of \(G\). Then for any left-\( K \)-invariant test function \(\Psi\in C^\infty_c(\Gamma\backslash G)^K\), any \(T\geq 2\) and any \(\epsilon >0\), there are constants \(C_j=C_j(\Psi), j=1,2\) so that \[ \int_{a\in A^+} \Psi(\mathbf{x}_0\cdot a\cdot \mathfrak{s}_T) \mathrm{d}a=C_1\log T+C_2+O_{\Psi,\varepsilon}\left(T^{-1/4+\varepsilon}\right). \] A standard application of such equidistribution results is to counting problems in discrete orbits, of which they give the following example of counting integer points on the inhomogeneous Pythagorean quadric \(W_d:x^2+y^2-z^2=d\). For a perfect square \(d=n^2\), the number \(\mathcal{N}_d(T)\) of integer points on \(W_d\) within the ball \(\lbrace x^2+y^2+z^2<T^2\rbrace\) satisfies \[ \mathcal{N}_d(T)=\mathcal{M}_d(T)+O\left(T^{1-\eta}d^{\beta\eta}\right) \] for any \(\eta<\frac{3}{40+72\theta}, \beta>\frac{3}{2}+2\theta\) and \(T\geq d^\beta\), where \( \theta=\frac{7}{64} \) is the bound towards the Ramanujan conjecture (that \( \theta=0 \)). Here, the main term \(\mathcal{M}_d(T)\) is given by \[ \mathcal{M}_d(T)=\frac{\sqrt{128}T}{\pi}\left(\log T+C-D(n)+(\frac{1}{3}-\frac{1}{2^{\nu+2}})\log 2\right) \] where \(\nu=\nu_2(n)\) is the \( 2 \)-adic valuation of \(n\), \(C\) is a constant roughly equal to \(0.616174...\), and \(D(n)=\frac{1}{n}\sum_{a|n}\phi(\frac{n}{a})\log a\).