Effective equidistribution and the Sato-Tate law for families of elliptic curves (Q607034): Difference between revisions

Let \(E\) be an elliptic curve defined over \(\mathbb{Q}\) and let \(L(E,s)=\sum_{n\geq 1} \frac{a_E(n)}{n^s}\) be the associated \(L\)-function. It is well known that for a prime \(p\) the local factor \(a_E(p)\) is equal to \(2\sqrt{p}\cos\theta_E(p)\) for some \(\theta_E(p)\in [0,\pi]\). If \(E\) has no complex multiplication then the Sato-Tate conjecture states that the distribution of the \(\theta_E(p)\), as \(p\) varies, converges to \(\mu_{ST}:=2\sin^2\theta \frac{d\theta}{\pi}\) (the Sato-Tate measure). For a fixed prime \(p\), the authors consider two (finite) families of elliptic curves, namely \[ \mathcal{F}_1(p):\;y^2=x^3-ax-b\quad a,b\in \mathbb{Z}/p\mathbb{Z}\;p\geq 5 \] (all elliptic curves modulo \(p\)); \[ \mathcal{F}_2(p):\;y^2=x^3-A(T)x-B(T)\quad A(T),B(T)\in\mathbb{Z}[T] \] with non-constant \(j\)-invariant (where the parameter \(T\) is specialized to some \(t\in \mathbb{Z}/p\mathbb{Z}\)). For any interval \(I\subseteq[0,\pi]\) (and \(i=1,2\)) let \(V_i(p):=i\# \mathcal{F}_i(p)\) and define the \textit{discrepancy} \[ D_{I,\mathcal{F}_i(p)}(\mu_{ST}):= \mid \# \{E\in\mathcal{F}_i(p)\,:\;\theta_E(p)\in I\,\}-\mu_{ST}(I)V_i(p) \mid \] (basically this, as \(p\) varies, provides an effective estimate on the convergence of the distribution of the \(\theta_E(p)'s\) to the Sato-Tate measure). Using results from \textit{M. Ram Murty} and \textit{K. Sinha} [J. Number Theory 129, No.~3, 681--714 (2009; Zbl 1234.11055)], some combinatorial identities (proved in Appendix A) and some formulas for the moments of the \(a_E(p)\)'s (originally given, with some typos, in \textit{B. Birch} [J. Lond. Math. Soc. 43, 57--60 (1968; Zbl 0183.25503)] and correctly stated and proved in Appendix B), the authors compute explicit inequalities for \(D_{I,\mathcal{F}_i(p)}(\mu_{ST})\) (for a fixed \(p\)) and then, as \(p\rightarrow \infty\), obtain the estimates \[ D_{I,\mathcal{F}_1(p)}(\mu_{ST})\leq C_1 \frac{V_1(p)}{\log V_1(p)} \quad ,\quad D_{I,\mathcal{F}_2(p)}(\mu_{ST}) \leq C_2 V_2(p)^{\frac{3}{4}}\;, \] where the \(C_i\)'s are computable constants.