Algebraic integers with conjugates in a prescribed distribution (Q7019339)

Let \(\alpha\) be an algebraic integer of degree \(n=\deg (\alpha)\) over the field of rational numbers \({\mathbb Q}\) with conjugates \(\alpha_1=\alpha, \alpha_2, \dots, \alpha_n\), and let \(\mathrm{tr} (\alpha)=\alpha_1+\alpha_2+\dots+\alpha_n\) be the trace of \(\alpha\). Define \(\lambda_{SSS}\) to be the least real number such that, for any \(\varepsilon>0\), there are only finitely many totally positive algebraic integers \(\alpha\) for which\N\[\N\mathrm{tr}(\alpha)<(\lambda_{SSS}-\varepsilon) \deg (\alpha).\N\]\NIn 1918, Schur proved that \(\sqrt{e} \leq \lambda_{SSS} \leq 2\). Note that \(\sqrt{e}=1.648721\dots\). In 1945, Siegel improved the lower bound to \(1.7336\). Later, in [Ann. Inst. Fourier 34, No. 3, 1--28 (1984; Zbl 0534.12002)], \textit{C. J. Smyth} further improved the lower bound on \(\lambda_{SSS}\) to \(1.7719\). The approach of Smyth is based on the method of auxiliary functions. The aim is to find several auxiliary polynomials \(Q_1, \dots, Q_N \in {\mathbb Z}[x]\) and appropriate positive numbers \(a_1, \dots, a_N\) for which the inequality\N\[\Nx \geq \lambda +\sum_{j=1}^N a_j \log |Q_N(x)|\N\]\Nholds for all \(x>0\). Each such construction immediately gives the lower bound\N\[\N\lambda_{SSS} \geq \lambda.\N\]\NSince then Smyth's original constant \(1.7719\) has been improved many times, and the problem of finding the value of \(\lambda_{SSS}\) is known as the \textit{Schur-Siegel-Smyth trace problem.} The best known constant attained in this way so far seems to be \(\lambda=1.793145\); see [\textit{C. Wang} et al., Math. Comput. 90, No. 331, 2317--2332 (2021; Zbl 1469.11483)]. In their paper, the authors used \(N=130\) auxiliary polynomials, which means that a lot of computation is involved.\N\NInitially, it was believed that \(\lambda_{SSS}=2\), namely, that there are only finitely many totally positive algebraic integers \(\alpha\) for which \(\mathrm{tr}(\alpha)<(2-\varepsilon) \deg (\alpha)\). In [in: Number theory. Fifth conference of the Canadian Number Theory Association, Ottawa, Ontario, Canada, August 17--22, 1996. Providence, RI: American Mathematical Society. 315--321 (1999; Zbl 0935.11041)], \textit{C. J. Smyth} proved that the constant \(\lambda=2\) cannot be achieved by just improving the set of auxiliary polynomials \(Q_1,\dots,Q_N\), and gave an explicit limit of this method which the resulting \(\lambda\) that cannot be attained just smaller than \(2\). Then Serre showed that \(\lambda\) obtained in this way cannot be greater than \(1.8983\). Note that this only shows that one cannot get anything better than \(\lambda_{SSS} \geq 1.8983\) with the method of auxiliary functions. However, Smyth already then raised some doubts whether the initial conjecture \(\lambda_{SSS}=2\) is true at all.\N\NIn this breaktrough paper, the author finally proves that the constant \(\lambda_{SSS}\) is strictly smaller than \(2\), namely, \(\lambda_{SSS} < 1.89831\) (Theorem 1.1). Hence, unlike initially expected, there are infinitely many totally positive algebraic integers \(\alpha\) satisfying\N\[\N\mathrm{tr}(\alpha)<1.89831 \deg (\alpha).\N\]\NAs an application, the author also shows how this work can be applied to find simple abelian varieties over finite fields with extreme point counts (Corollary 3.1).