On the exceptional set of the generalized Ramanujan conjecture for \(\mathrm{GL}(3)\) (Q2040192)

Let \(\phi\) be a Hecke-Maass cusp form for \(\mathrm{SL}(3, \mathbb{Z})\), \(\{\alpha_{\phi, 1}(p), \alpha_{\phi, 2}(p), \alpha_{\phi, 3}(p)\}\) the Satake parameters of \(\phi\) at prime \(p\). The generalized Ramanujan conjecture (GRC) preditcs that \(|\alpha_{\phi, i}(p)| = 1\) for \(1 \leq i \leq 3\). The paper under review shows that for nonself-dual \(\phi\) the natural density of primes for which GRC fails does not exceed \(12/25\) under certain hypothesis. The authors also remark that for self-dual \(\phi\) such density is at most \(1/35\), which follows from a similar density result for \(\mathrm{GL}(2)\) forms [\textit{W. Luo} and \textit{F. Zhou}, Am. J. Math. 141, No. 2, 485--501 (2019; Zbl 1446.11082)] and the fact that \(\phi\) is self-dual if and only if \(\phi\) is a symmetric square lift of a \(\mathrm{GL}(2)\) form. One of the main inputs in the proof is the following analogue of prime number theory \[ \sum_{p \leq x} |\alpha_{\phi, 1}(p) + \alpha_{\phi, 2}(p) + \alpha_{\phi, 3}(p)|^2 \sim \pi(x) \] where \(\pi(x)\) is the number of primes not exceeding \(x\). The hypothesis in the result concerns the Rankin-Selberg \(L\)-function \[ L(s, \phi \times \tilde{\phi} ) = \sum_{n \geq 1} \lambda_{\phi \times \tilde{\phi}}(n)n^{-s} \] and the symmetric square \(L\)-function \[ L(s, \mathrm{sym}^{2}\phi) = \sum_{n \geq 1} \lambda_{\mathrm{sym}^{2}\phi}(n)n^{-s}. \] It asserts that \begin{align*} \sum_{p \leq x} \lambda_{\phi \times \tilde{\phi}}(p)^{2} \sim 2\pi(x), \quad \sum_{p \leq x} |\lambda_{\mathrm{sym}^{2}\phi}(p)|^{2} \sim \pi(x). \end{align*} The authors show the constants in the hypothesis can be predicted by the Sato-Tate conjecture on the equidistribution of the Satake parameters of \(\phi\). In addition to the above result, the authors also show that for any real number \(\alpha \geq 1\), \[ \liminf_{x \rightarrow \infty} \frac{1}{\pi(x)} |\{p \leq x : \underset{1 \leq i \leq 3}{\max} \, |\alpha_{\phi, i}(p)| \leq \alpha \}| \geq 1 - \frac{1}{(\alpha + \alpha^{-1} -1)^2} \] refining Walji's result [\textit{N. Walji}, J. Number Theory 133, No. 10, 3470--3484 (2013; Zbl 1297.11040)].