On the frequency of height values (Q2028703)

The paper under review deals with the question of estimating the cardinality of the set \(A(k,d,\mathcal{H})\) of algebraic numbers \(\alpha \in \overline{\mathbb{Q}} \subseteq \mathbb{C}\) with fixed degree \([\mathbb{Q}(\alpha) \colon \mathbb{Q}] = d\) and Weil height \(H(\alpha) = \mathcal{H} \in \overline{\mathbb{Q}} \cap [1,+\infty)\) which have exactly \(k \in \{0,\dots,d\}\) conjugates lying inside the unit circle. This set \(A(k,d,\mathcal{H})\) is finite by Northcott's theorem, and (as the title suggests) the estimates provided by the paper under review compare the size of the image of \(H \colon \overline{\mathbb{Q}} \to \overline{\mathbb{Q}} \cap [1,+\infty)\) with the size of its fibers. Let us be more precise. First of all, the author shows in Lemma 3.1 that \(\sum_{\mathcal{H}' \leq \mathcal{H}} \lvert A(k,d,\mathcal{H}') \rvert\) grows as \(\mathcal{H}^{d (d+1)}\) as \(\mathcal{H} \to +\infty\), which is the same order of growht obtained by \textit{D. Masser} and \textit{J. D. Vaaler} [Dev. Math. 16, 237--243 (2008; Zbl 1211.11115)] without specifying the number of conjguates \(k\) inside the unit circle. This implies in particular that \(A(k,d) = \bigcup_{\mathcal{H} \geq 1} A(k,d,\mathcal{H})\) is always infinite, hence \(B(k,d) := H(A(k,d))\) is unbounded. Using this fact, the author defines \(a(k,d)\) as the limit of \(\log\lvert A(k,d,\mathcal{H}) \rvert/\log(\mathcal{H})\) as \(\mathcal{H} \to +\infty\) within the set \(B(k,d)\). Analogously, the author defines \(b(k,d)\) to be the limit of \(\log\lvert B(k,d) \cap [1,\mathcal{H}] \rvert/\log(\mathcal{H})\) as \(\mathcal{H} \to +\infty\). Then, the author shows in Theorem 2.1 that \(a(0,d) = a(d,d) = d^2\) and \(b(0,d) = b(d,d) = d\). While the proof of this second fact uses only Eisenstein's criterion, obtaining the first one requires already the usage of a the language of o-minimal structures, in order to employ a theorem of \textit{F. Barroero} and \textit{M. Widmer} [Int. Math. Res. Not. 2014, No. 18, 4932--4957 (2014; Zbl 1315.11056)], which is in fact also used in the proof of Lemma 3.1. The author of the paper under review devotes then Theorem 4.1 to prove that \(b(k,d) = d (d+1)\) if \(d \geq 2\) and \(0 < k < d\), whereas \(a(1,d) = a(d-1,d) = 0\), and \(a(k,d) = 0\) for each \(0 < k < d\) if \(d\) is prime. The author of the paper under review uses then these results in Theorem 4.3, to show that the function \(\alpha \mapsto ([\mathbb{Q}(\alpha) \colon \mathbb{Q}],H(\alpha))\) is ``almost injective'', in a precise sense, whenever \([\mathbb{Q}(\alpha) \colon \mathbb{Q}] \geq 2\). Moving on, an explicit upper bound for \(\lvert A(k,d,\mathcal{H}) \rvert\) is provided in Theorem 5.2, which depends on \(\gcd(k,d)\), and in particular shows that \(a(k,d) = 0\) whenever \(\gcd(k,d) = 1\). Moreover, Section 6 contains a very explicit study of the case \((k,d) = (2,4)\), for which it is shown that the sequence \(\log\lvert A(2,4,\mathcal{H}) \rvert/\log(\mathcal{H})\) does not have a limit as \(\mathcal{H} \to +\infty\) within \(B(2,4)\). This explicit computation is generalized in Section 7, where the author of the paper under review proves that the sequence \(\log\lvert A(k,d,\mathcal{H}) \rvert/\log(\mathcal{H})\) never has a limit if \(0 < k < d\) and \(\gcd(k,d) > 1\). However, he shows in Theorem 7.5 that the limit superior of this sequence equals \(d (\gcd(k,d) - 1)\), whereas he remarks in Lemma 7.6 that its limit inferior always equals \(0\). Now, Section 8 and 9 of the paper under review are devoted to problems of a different nature. First of all, Theorem 8.1 provides analogous results for sets of non-zero polynomials \(P \in \mathbb{Z}[t]\) with fixed degree \(\deg(P) = d\) and Mahler measure \(M(P) := \exp(\int_0^1 \log\lvert P(e^{2 \pi i t}) \rvert dt) = \mathcal{M}\), having exactly \(k \in \{0,\dots,d\}\) complex zeros inside the unit circle. Since \(M(P) \deg(P) = H(\alpha)\) for each irreducible polynomial \(P\) such that \(P(\alpha) = 0\), the novelty of this result lies in the consideration of reducible polynomials as well, which leads in fact to simpler formulas. Lastly, Theorem 9.1 characterizes those algebraic numbers \(\alpha \in \overline{\mathbb{Q}}\) for which there exists \(n \geq 1\) such that \(H^n(\alpha) = \alpha\) as being the roots of natural numbers. Moreover, Theorem 9.3 shows more precisely that either \(\lim_{n \to +\infty} H^n(\alpha) = 1\) or \(H^n(\alpha) = H^{n+1}(\alpha)\) for some \(n \geq 1\). To conclude, the paper under review provides new interesting insights and perspectives to the classical problem of making Northcott's theorem effective. Given its witty style, this paper is certainly recommended to someone who would like to learn about Weil heights and o-minimal geometry through the lenses of a concrete research problem.