Closing the gap around the essential minimum of height functions with linear programming
José Ignacio Burgos Gil, Ricardo Menares, Binggang Qu, Martín Sombra · 2026
Read Closing the gap around the essential minimum of height functions with linear programming by José Ignacio Burgos Gil, Ricardo Menares, Binggang Qu, Martín Sombra (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2601-18978.txt.
Introduction
José Ignacio Burgos Gil and three co-authors close in on a stubborn number in number theory. For height functions, which grade how complex a point is, the essential minimum is a key floor value. For many common heights, no one has pinned that floor down in exact terms.
The team pairs two old routes. One is a lower bound method that goes back to Smyth. The other is an upper bound method based on how whole points spread out in the limit. Joining the two with linear programming narrows the gap. Each closed gap sharpens what number theorists can say about the smallest, simplest points of a given kind.
Catalog record
| CREATOR | José Ignacio Burgos Gil, Ricardo Menares, Binggang Qu, Martín Sombra |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 119 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2601-18978
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Closing the gap around the essential minimum of height functions with linear programming Thanks: J. I. Burgos was partially supported by grants PID2022-142024NB-I00 and CEX2023-001347-S funded by MICIU/AEI/10.13039/501100011033. R. Menares was partially supported by ANID FONDECYT Regular grant 1250734. B. Qu was partially supported by grant PID2022-142024NB-I00 funded by MICIU/AEI/10.13039/501100011033. M. Sombra was partially supported by the MICINN research project PID2019-104047GB-I00, the AGAUR research project 2021-SGR-01468 and the AEI project CEX2020-001084-M of the María de Maeztu program for centers and units of excellence in R&D José Ignacio Burgos Gil, Ricardo Menares, Binggang Qu, Martín Sombra Abstract . For many common height functions, the explicit determination of the essential minimum is an open problem. We consider a classical method to obtain lower bounds that goes back at least to C.J. Smyth, and a method to obtain upper bounds based on the knowledge of the limit distribution of integral points. We use an infinite dimensional linear programming scheme to show that that both methods agree in the limit, by showing that the principle of strong duality holds in our situation. As a corollary we prove that the essential minimum can be attained by sequences of algebraic integers. Recent results by A. Smith and B. Orloski–N. Sardari, furnish a characterization of compactly supported measures that can be approximated by complete sets of conjugates of algebraic integers, in terms of infinitely many nonnegativity conditions. We establish an extension of this characterization to measures with non necessarily compact support. As an application of this result and of strong duality, we show that the essential minimum is a computable real number when the Green function used to define the height is computable. We systematically use potential theory for measures that can integrate functions with logarithmic growth. 1. Introduction 1.1. The essential minimum For many common height functions, the explicit determination of the essential minimum (see ( 1.3 )) is an open problem. We consider a classical method to obtain lower bounds, that goes back at least to C.J. Smyth, and a method to obtain upper bounds based on the knowledge of the limit distribution of integral points used in [ BGMRL18 ] . In this work we use an infinite dimensional linear programming scheme to show that both methods agree in the limit, by showing that the principle of strong duality holds in our situation. This allows us to prove that the essential minimum can be attained by sequences of algebraic integers. We also establish a characterization, in terms of infinitely many nonnegativity conditions, of probability measures on \mathbb{C} with non necessarily compact support that can be approximated, in a suitable sense, by sequences of complete sets of conjugates of distinct algebraic integers. This characterization, in the case of measures with compact support, was previously established by A. Smith and B. Orloski–N. Sardari. As an application of this result and of strong duality, we show that the essential minimum is a computable real number when the Green function used to define the height is computable. These issues can be considered on any variety X defined over a number field, but in this work we will focus in the cases X=\mathbb{P}^{1} or X=\mathbb{A}^{1} . We proceed to state our results precisely. We denote by \overline{\mathbb{Q}} the set of all algebraic numbers. Given \alpha\in\overline{\mathbb{Q}} , we denote by P_{\alpha}(x)\in\mathbb{Z}[x] a primitive irreducible polynomial having \alpha as a root. The polynomial P_{\alpha} is determined by \alpha up to sign. Its set of roots O(\alpha) is called the complete set of conjugates of \alpha . We say that \mathrm{g}\colon\mathbb{C}\longrightarrow\mathbb{R} is a Green function , if it is continuous, invariant under complex conjugation and obeys the asymptotic (1.1) \mathrm{g}(z)=\log|z|+o(\log|z|),\textrm{ as }|z|\longrightarrow\infty. Such functions correspond to metrics on the line bundle \mathcal{O}(1) on \mathbb{P}^{1} with certain singularities at infinity (see Section 6.1 ). We define the height function \Ht_{\mathrm{g}}\colon\overline{\mathbb{Q}}\longrightarrow\mathbb{R} associated to such \mathrm{g} by the rule (1.2) \Ht_{\mathrm{g}}(\alpha)=\frac{1}{\#O(\alpha)}\left(\log|a_{\alpha}|+\sum_{\beta\in O(\alpha)}\mathrm{g}(\beta)\right). Here, a_{\alpha}\in\mathbb{Z} is the leading coefficient of P_{\alpha} . Height functions measure the arithmetic complexity of algebraic numbers, due to two key properties: the function \Ht_{\mathrm{g}} is bounded from below and for any A,B>0 the set \big\{\alpha\in\overline{\mathbb{Q}}\mid\Ht_{\mathrm{g}}(\alpha)<A,\;\#O(\alpha)<B\big\}\quad\textrm{ is finite.} The essential minimum of \Ht_{\mathrm{g}} is defined as (1.3) \ess(\Ht_{\mathrm{g}})\coloneqq\inf\left\{\liminf_{n\rightarrow\infty}\Ht_{\mathrm{g}}(\alpha_{n})\,\middle|\,(\alpha_{n})\subseteq\overline{\mathbb{Q}}\textrm{ is a sequence of distinct algebraic numbers}\right\}. It follows directly from the definitions that \inf(\Ht_{\mathrm{g}})\leq\ess(\Ht_{\mathrm{g}}) . Also, it can be shown that the image of \Ht_{\mathrm{g}} is dense in [\essh,\infty) (see Remark 6.8 ), so \essh coincides with the smallest limit point in \Ht_{\mathrm{g}}(\overline{\mathbb{Q}}) . When \mathrm{g}(z)=\log^{+}|z| , we obtain the Weil (or naïve) height \Ht_{W}=\Ht_{\log^{+}} . Clearly \Ht_{W}\geq 0 and for any root of unity \zeta , it holds \Ht_{W}(\zeta)=0 . This shows that \inf(\Ht_{W})=\ess(\Ht_{W})=0 and the closure of the image of \Ht_{W} is [0,\infty) . The essential minimum of a height function plays a role in the subject of equidistribution of small points. Indeed, a generic sequence of algebraic points is small if its height converges to the essential minimum. A theorem of Yuan [ Yua08 , Theorem 3.1] , subsuming previous results by many mathematicians, ensures that under an appropriate hypothesis on the essential minimum, small points are equidistributed with respect to a measure depending only on the height under consideration. There are many situations where the hypothesis in Yuan’s theorem does not hold, and hence the asymptotic distribution of small points is not known. Very few situations beyond Yuan’s theorem are well understood, see [ cBS25 ] , [ BGPRLS19 ] and [ Küh22 ] . In the case of \mathbb{P}^{1} considered here, the hypothesis in Yuan’s theorem boils down to \inf(\Ht_{\mathrm{g}})=\ess(\Ht_{\mathrm{g}}) . However, for many common heights the strict inequality \inf(\Ht_{\mathrm{g}})<\ess(\Ht_{\mathrm{g}}) holds and both the explicit determination of \essh and the asymptotic distribution of small points, are open problems. An illustrative example is the Zhang-Zagier height, defined as follows. Extend the Weil height to \mathbb{A}^{2}(\overline{\mathbb{Q}})=\overline{\mathbb{Q}}^{2} by \Ht_{W}^{2}(\alpha,\beta)=\Ht_{W}(\alpha)+\Ht_{W}(\beta) . If X\subsetneq\mathbb{A}^{2} is any subvariety, we define a height \Ht_{X} on X by \Ht_{X}\coloneqq(\Ht_{W}^{2})_{|X} . Then, a theorem of Zhang [ Zha92 ] ensures that \ess(\Ht_{X})>0 if and only if X does not contain a torsion translate of a toric subvariety. For concreteness, Zagier took X=\{x+y=1\}\subset\mathbb{A}^{2} and studied in [ Zag93 ] the resulting height, that we denote by \Ht_{\operatorname{ZZ}} . In our framework, identifying X with \mathbb{A}^{1} by projection on one coordinate, the choice \mathrm{g}(z)=\frac{1}{2}\Big(\log^{+}|z|+\log^{+}|1-z|\Big) leads to a height \Ht_{\mathrm{g}} such that \Ht_{\operatorname{ZZ}}=2\Ht_{\mathrm{g}} . In loc. cit. , it is obtained a lower bound for \ess(\Ht_{\operatorname{ZZ}}) . These results were improved by Doche ( [ Doc01a ] , [ Doc01b ] ), who also gives upper bounds for \ess(\Ht_{\operatorname{ZZ}}) , obtaining (1.4) \log(1.2817770214)\leq\ess(\Ht_{\operatorname{ZZ}})\leq\log(1.289735). An important height function is the stable Faltings height on elliptic curves. Namely, \Ht_{\operatorname{F}}\colon\overline{\mathbb{Q}}\longrightarrow\mathbb{R} is defined as \Ht_{\operatorname{F}}(\alpha)=\Ht_{\operatorname{Fal}}(E_{\alpha}), where E_{\alpha} is an elliptic curve with j -invariant j(E_{\alpha})=\alpha and \Ht_{\operatorname{Fal}}(E_{\alpha}) is the stable Faltings height of the elliptic curve E_{\alpha} . This height is also an explicit multiple of \Ht_{\mathrm{g}} for an appropriate \mathrm{g} , see section 6.5 . In [ Löb17 ] is obtained a lower bound that shows \inf(\Ht_{\operatorname{F}})<\ess(\Ht_{\operatorname{F}}) and both lower and upper bounds for \ess(\Ht_{\operatorname{F}}) are obtained in [ BGMRL18 ] . In particular, it follows from loc. cit. that (1.5) -0.748629\leq\ess(\Ht_{\operatorname{F}})\leq-0.748622. We denote by \overline{\mathbb{Z}} the set of all algebraic integers (recall