Moments of the error term in the Sato-Tate law for elliptic curves (Q1788089)

Let \(E(a,b):y^2=x^3+ax+b\) be an elliptic curve (with \(a,b\in \mathbb{Z}\)) with conductor \(N_{E(a,b)}\) and let \(L(E(a,b),s)= \displaystyle{ \sum_{n\geqslant 1} a_{E(a,b)}(n)n^{-s} = \sum_{n\geqslant 1} \widetilde{a}_{E(a,b)}(n)n^{\frac{1}{2}-s}}\) be the associated Hasse-Weil \(L\)-function. For any \(x\in \mathbb{R}_{>0}\) let \(\widetilde{\pi}(x)=|\{\frac{x}{2}<p\leqslant x\,:\,p\ \mathrm{prime}\}|\) and for any interval \(I\subseteq [-2,2]\) let \(N_I(E(a,b),x)=|\{\frac{x}{2}<p\leqslant x\,:\,p\ \mathrm{prime},\ p\nmid N_{E(a,b)}\ \mathrm{and}\ \widetilde{a}_{E(a,b)}(p)\in I\}|\). The Sato-Tate law in this setting reads as \[\lim_{x\rightarrow +\infty} \frac{N_I(E(a,b),x)}{\widetilde{\pi}(x)}=\frac{1}{\pi}\int_I \sqrt{1-\frac{t^2}{4}}\,dt:=\mu(I)\ ,\] and the paper under review deals with the error term moments \[ \frac{1}{AB} \sum_{|a|\leqslant A} \sum_{|b|\leqslant B} \left( N_I(E(a,b),x) - \widetilde{\pi}(x)\mu(I)\right)^m \] for any \(m\in \mathbb{N}\), providing estimates of type \[ \frac{1+(-1)^m}{2} \frac{m!}{2^{\frac{m}{2}}(\frac{m}{2})!} (\mu(I)-\mu(I)^2)^{\frac{m}{2}} + O(x^\alpha(\log x)^\beta) \] and various formulas for the \(O\)-terms (i.e. for \(\alpha\) and \(\beta\)) depending on additional hypotheses like the MRH (modular Riemann hypothesis) or regarding the growth of the Fourier coefficients \(\widetilde{a}_{E(a,b)}(n)\) and, of course, on bounds for \(A\) and \(B\). The authors write down the \(m\)-th power \(\left( N_I(E(a,b),x) - \widetilde{\pi}(x)\mu(I)\right)^m \) in terms of sums and products of the \(\widetilde{a}_{E(a,b)}(n)\), with coefficients they are able to control. The multiplicativity of some functions (most notably the sum \[ S(n)=\frac{1}{s(n)^2}\sum_{\begin{subarray}{c} a,b=1\\ (ab\Delta(E(a,b)),n)=1\end{subarray}}^{s(n)} \widetilde{a}_{E(a,b)}(n) \] where \(s(n)\) is the largest squarefree number dividing \(n\) and \(\Delta(E(a,b))\) is the discriminant of the elliptic curve) enables them to work on prime powers: they relate \(S(p^h)\) to the Kronecker class number and use combinatorial identities by \textit{B. J. Birch} [J. Lond. Math. Soc. 43, 57--60 (1968; Zbl 0183.25503)] and \textit{Z. A. Melzak} [``Problem 4458'', Amer. Math. Monthly 58, 636, (1951)] to provide general estimates, which they are able to refine in the final sections, with the help of the additional hypotheses, to get the required formulas for the \(O\)-terms.