betaFeedback

WorksMathematics

Triangular minimality for planar Coulomb renormalized energy

Abstract

We prove the Sandier–Serfaty conjecture: the triangular lattice of covolume one minimizes planar Coulomb renormalized energy over all admissible curl-free fields with a unit uniform background. Combined with Bétermin and Sandier's asymptotic formula, this also proves the Brauchart–Hardin–Saff conjecture for the linear term of optimal ordered-pair logarithmic energy on the unit two-sphere. The proof uses a direct Voronoi-cell comparison with rigorous interval arithmetic.

Introduction

The Coulomb interaction in two dimensions is logarithmic. An infinite configuration of equal point charges against a uniform neutralizing background has an infinite electrostatic energy, even in a bounded region containing a charge. Subtracting the logarithmic self-energy at each charge and then taking energy per unit area gives the renormalized energy studied here. Its proposed minimizer is the triangular lattice.

The energy and the main results

For R>1R>1, let KR=[−R,R]2K_R=[-R,R]^2. Fix smooth cutoffs χR∈Cc∞(R2)\chi_R\in C^\infty_c(\mathbb{R}^2) satisfying

0≤χR≤1,supp⁡χR⊂KR,χR=1 on KR−1,sup⁡R>1∥∇χR∥∞<∞.(1)0\leq\chi_R\leq1,\qquad\operatorname{supp}\chi_R\subset K_R,\qquad\chi_R=1\text{ on }K_{R-1},\qquad\sup_{R>1}\lVert\nabla\chi_R\rVert_\infty<\infty. \tag*{(1)}

Let Λ⊂R2\Lambda\subset\mathbb{R}^2 be locally finite and simple, so every point has multiplicity one, and put νΛ=∑p∈Λδp\nu_\Lambda=\sum_{p\in\Lambda}\delta_p. The admissible class A1A_1 consists of locally integrable vector fields EE such that

div⁡E=2π(νΛ−1),curl⁡E=0,sup⁡R>1νΛ(BR)∣BR∣<∞,(2)\operatorname{div}E=2\pi(\nu_\Lambda-1),\qquad\operatorname{curl}E=0,\qquad\sup_{R>1}\frac{\nu_\Lambda(B_R)}{\lvert B_R\rvert}<\infty, \tag*{(2)}

where the equations hold in distributions and BRB_R is the disk of radius RR centered at the origin. These conditions imply E∈Llocq(R2;R2)E\in L^q_{\mathrm{loc}}(\mathbb{R}^2;\mathbb{R}^2) for every 1≤q<21\leq q<2: locally subtract the finitely many Coulomb singularities and a smooth field with the prescribed constant divergence; the remaining field has harmonic components. Section 2 supplies the details. Consequently the same class results if local LqL^q integrability is required for any one, or for all, of these exponents.

For E∈A1E\in A_1 define

W(E,χR)=lim⁡η↓0[12∫R2∖⋃p∈ΛB(p,η)χR∣E∣2 dx+πlog⁡η∑p∈ΛχR(p)].(3)W(E,\chi_R)=\lim_{\eta\downarrow0}\left[\frac{1}{2}\int_{\mathbb{R}^2\setminus\bigcup_{p\in\Lambda}B(p,\eta)}\chi_R\lvert E\rvert^2\,dx+\pi\log\eta\sum_{p\in\Lambda}\chi_R(p)\right]. \tag*{(3)}
W(E)=lim sup⁡R→∞W(E,χR)∣KR∣.(4)W(E)=\limsup_{R\to\infty}\frac{W(E,\chi_R)}{\lvert K_R\rvert}. \tag*{(4)}

The local puncture limit exists and is finite. The outer limsup is initially allowed to take either infinite value. The puncture limit is taken first, and the cutoff layer has the fixed width in (1.1) throughout.

Identify R2\mathbb{R}^2 with C\mathbb{C} when convenient. The triangular lattice of covolume one is

Λ△=23(Z(1,0)+Z(12,32)).(5)\Lambda_\triangle=\sqrt{\frac{2}{\sqrt{3}}}\left(\mathbb{Z}(1,0)+\mathbb{Z}\left(\frac{1}{2},\frac{\sqrt{3}}{2}\right)\right). \tag*{(5)}

Let h△h_\triangle be the mean-zero periodic distributional solution of −Δh△=2π(δ0−1)-\Delta h_\triangle=2\pi(\delta_0-1) on R2/Λ△\mathbb{R}^2/\Lambda_\triangle, and set E△=−∇h△E_\triangle=-\nabla h_\triangle.

Theorem 1.1. For every cutoff family satisfying (1.1) and every field E∈A1E\in A_1,

W(E)≥W(E△).W(E)\geq W(E_\triangle).

The triangular value is finite, and inf⁡E∈A1W(E)=W(E△)\inf_{E\in A_1}W(E)=W(E_\triangle).

In particular W(E)=−∞W(E)=-\infty is impossible in this class. The inequality includes fields of energy +∞+\infty. No separation, periodicity, or Voronoi-cell bound is imposed on an arbitrary competitor.

The finite-volume statement used to prove Theorem 1.1 has its own exact normalization. For an integer n≥2n \ge2, write Tn=R2/(nZ2)T_n = \mathbb{R}^2/(\sqrt{n}\mathbb{Z}^2). Given distinct points a1,…,an∈Tna_1,\ldots,a_n \in T_n, let hh have mean zero and satisfy −Δh=2π(∑i=1nδai−1)-\Delta h = 2\pi(\sum_{i=1}^n \delta_{a_i} - 1) on TnT_n. Define

Wn(h)=lim⁡η↓0[12∫Tn∖⋃i=1nB(ai,η)∣∇h∣2 dx+πnlog⁡η].(6)W_n(h) = \lim_{\eta\downarrow0} \left[\frac{1}{2}\int_{T_n \setminus\bigcup_{i=1}^n B(a_i,\eta)} |\nabla h|^2\,dx + \pi n \log\eta\right]. \tag*{(6)}

Theorem 1.2. For every integer n≥2n \ge2 and every configuration of nn distinct points on TnT_n, the potential in (6) satisfies

Wn(h)≥nW(E△).W_n(h) \ge nW(E_\triangle).

This is a lower bound on square tori. It does not assert equality on any particular square torus. Constant harmonic fields increase the energy by n∣c∣2/2n|c|^2/2; the required orthogonality follows by integrating the periodic gradient, as proved in (2.12).

The planar minimum also determines the linear term in optimal logarithmic energy for finitely many points on the sphere.

Corollary 1.3 (Spherical logarithmic energy). Let S2={y∈R3:∥y∥=1}S^2 = \{y \in\mathbb{R}^3 : \lVert y\rVert= 1\} be the unit sphere. For each integer n≥2n \ge2, define

Elog⁡(n)=min⁡(y1,…,yn)∈(S2)n(−∑1≤i,j≤ni≠jlog⁡∥yi−yj∥),E_{\log}(n) = \min_{(y_1,\ldots,y_n)\in(S^2)^n}\left(-\sum_{\substack{1\le i,j\le n\\i\ne j}}\log\lVert y_i-y_j\rVert\right),

where the sum is over ordered pairs, ∥⋅∥\lVert\cdot\rVert is the Euclidean norm in R3\mathbb{R}^3, log⁡\log is the natural logarithm, and a collision has energy +∞+\infty. Then, as n→∞n \to\infty through the integers,

Elog⁡(n)=(12−log⁡2)n2−n2log⁡n+CBHSn+o(n),(7)E_{\log}(n) = \left(\frac{1}{2}-\log2\right)n^2-\frac{n}{2}\log n+C_{\mathrm{BHS}}n+o(n), \tag*{(7)}

where

CBHS=2log⁡2+12log⁡(23)+3log⁡(πΓ(1/3)),(8)C_{\mathrm{BHS}} = 2\log2+\frac{1}{2}\log\left(\frac{2}{3}\right)+3\log\left(\frac{\sqrt{\pi}}{\Gamma(1/3)}\right), \tag*{(8)}

and Γ\Gamma is Euler’s gamma function.

This proves the d=2d=2 case of the logarithmic-energy conjecture of Brauchart, Hardin, and Saff [4]. Bétermin and Sandier had already proved the existence of the linear coefficient and shown that it equals (8) precisely when the density-one triangular lattice minimizes their planar renormalized energy [1]. Section 8 checks the normalization and applies that equivalence to Theorem 1.1. The conclusion is an asymptotic value, not a construction of near-optimal point sets or an algorithm for Smale’s seventh problem.

Context and prior work

The finite-vortex renormalized energy of Bethuel, Brezis, and Hélein [2] is an antecedent of the infinite-system energy considered here. Sandier and Serfaty introduced the planar functional in their analysis of microscopic vortex patterns in the Ginzburg–Landau model. They proved existence of a minimum, approximation of its value by square-periodic fields, and triangular optimality among Bravais lattices [17]. Their Conjecture 1 asks whether the triangular value is also the minimum over the full admissible class. Theorem 1.1 resolves this conjecture positively. The current normalization in that paper has one point per area 2π/m2\pi/m; our density-one convention corresponds to its background m=2πm=2\pi.

The lattice comparison belongs to a longer history. The minimization of convergent Epstein zeta sums at fixed covolume was developed by Rankin [14] and Cassels [5], with an error noted in [6] and a repair supplied by Ennola [8] and simplified by Diananda [7]. Montgomery proved triangular minimality of the lattice theta function for every positive Gaussian parameter [11], [Theorem 1]. These are comparisons within the lattice class; the issue here is the lower bound for arbitrary admissible configurations.

For the confining-potential class in their Theorem 2, Sandier and Serfaty show that microscopic limits of finite Coulomb minimizers concentrate on minimizers of the planar renormalized energy [18], [Theorem 2]. Its minimum also determines a next-order coefficient for optimal logarithmic energy on the sphere [1], [Theorem 1.5]. Petrache and Serfaty prove a conditional implication from universal optimality to minimality for Coulomb and Riesz renormalized energies in their configuration-energy normalization [13], [Theorem 2 and Corollary 4]. Their periodic comparison uses tori compatible with the reference lattice, not arbitrary square tori, and their planar conclusion is conditional on the Cohn–Kumar conjecture. The companion universal-optimality theorem supplies the Gaussian comparison used in that implication [12], [Theorem 1.1 and Corollary 1.3]. The present proof treats the logarithmic functional directly and uses no universal-optimality result as an input.

Two geometric precedents help explain the finite-torus argument. Screening a point charge by a uniform disk yields a separation argument for planar confined logarithmic minimizers; see Lieb, Rougerie, and Yngvason [10], [Lemma 3 and Appendix A]. Voronoi moment inequalities combine sharp cell comparisons with Euler’s bound on the average number of sides, as in Gruber’s proof of a theorem of Fejes Tóth [9]. Affine bounds on cell energies, sharp at the regular hexagon, also appear in the Wasserstein model of Bourne, Peletier, and Theil for cells with at least three sides and mass v≥1.5⋅10−4v \ge1.5\cdot10^{-4} [3], [Lemma 12]. Here the cell comparison must arise from a trial potential that is continuous off the sites and has the prescribed logarithmic singularities. Its nonradial correction and the two scalar estimates of Section 5 address that requirement, including sectors whose perpendicular foot lies outside the corresponding edge.

The whole-plane reduction below uses Sandier and Serfaty’s local energy estimates and periodic approximation. The finite-torus argument provides the comparison with the triangular value. Throughout, the conclusion concerns energy per unit area: compact changes to a configuration can preserve this value, so the minimum statement alone does not classify all minimizing configurations.

How the comparison works

The proof separates a finite-torus comparison from the passage to arbitrary fields. The finite comparison identifies the triangular value. Sandier and Serfaty’s approximation theorem supplies periodic fields approaching the unknown full-plane minimum; it does not identify that minimum.

Section 2 checks the conversion from that approximation theorem to our normalization and fixed-width square cutoffs. The main point is to control the boundary error for an arbitrary field whose energy limsup is bounded above, without assuming separation or density one in advance. Local energy estimates give the needed charge density and boundary-strip bound. Proposition 2.4 then transfers any common lower bound on square tori to every field in A1\mathcal{A}_1, including the possible infinite-energy cases.

We next prove the finite comparison. On each square torus the point energy attains a minimum. A screened logarithmic potential and the minimum principle separate the minimizing points, as shown in Section 3. Separation is used only for that minimum; once it is bounded below, every other configuration is bounded below by minimality. This is why the main theorems impose no separation condition on their competitors.

The lifted minimizer has convex Voronoi cells, the regions closer to one site than to any other. Each edge and its incident cell determine a sector joining that edge to the site. Counting incidences on the torus gives at most six sectors per point on average. The comparison is not an estimate of the electrostatic energy of an isolated cell. Instead, we construct a single trial potential HH on the torus. Its sector formulas have matching traces and satisfy H=−log⁡r+O(r2)H=-\log r+O(r^{2}) near each site. Thus HH and the Green potential hh have matching logarithmic singularities, while the constant term is normalized to zero in HH. Proposition 4.3 gives the exact identity

Wn(h)=∑SJ(S)+12∫Tn∣∇(h−H)∣2.W_n(h)=\sum_S J(S)+\frac{1}{2}\int_{T_n} \lvert\nabla(h-H)\rvert^2.

Here J(S)J(S) is the trial potential’s sector contribution: its linear background-interaction term minus one half of its renormalized Dirichlet integral, as defined in (4.10). The identity cancels the matching logarithmic singularities and retains the nonnegative square exactly.

The construction of HH uses radial functions common to every sector. At a shared Voronoi edge its trace depends only on the equal distances to the two neighboring sites; along a ray to a vertex it depends only on the vertex radius. These observations glue the local formulas continuously off the sites and give an H1H^1 function on every punctured torus. The identity is proved for any compatible flat torus whose area equals its number of sites, not just for square tori. A separate one-site triangular torus supplies six regular sectors; there the chosen potential has exactly the triangular Green gradient, so the square term vanishes.

Section 5 constructs the radial functions from the triangular Green expansion and subtracts fixed multiples of sector area and angle. The desired bound has the form

J(S)≥K+λA(S)+μθ(S),K<0.J(S)\ge K+\lambda A(S)+\mu\theta(S),\qquad K<0.

Areas and angles sum exactly to nn and 2πn2\pi n on a square torus. The incidence count then gives n(6K+λ+2πμ)n(6K+\lambda+2\pi\mu), and exact calibration on the triangular torus identifies this with nW(E△)nW(E_{\triangle}).

References

  1. [1]Laurent Bétermin and Etienne Sandier. Renormalized energy and asymptotic expansion of optimal logarithmic energy on the sphere. Constructive Approximation, 47(1):39–74, 2018. Numbered references in this paper use arXiv:1404.4485v5.
  2. [2]Fabrice Bethuel, Haïm Brezis, and Frédéric Hélein. Ginzburg–Landau Vortices, volume 13 of Progress in Nonlinear Differential Equations and Their Applications. Birkhäuser, Boston, MA, 1994.
  3. [3]David P. Bourne, Mark A. Peletier, and Florian Theil. Optimality of the triangular lattice for a particle system with Wasserstein interaction. Communications in Mathematical Physics, 329:117–140, 2014. Numbered references in this paper use arXiv:1212.6973v3.arxiv.org/abs/1212.6973
  4. [4]J. S. Brauchart, D. P. Hardin, and E. B. Saff. The next-order term for optimal Riesz and logarithmic energy asymptotics on the sphere. In Recent Advances in Orthogonal Polynomials, Special Functions, and Their Applications, volume 578 of Contemporary Mathematics, pages 31–61. American Mathematical Society, Providence, RI, 2012. Numbered references in this paper use arXiv:1202.4037v1.DOI
  5. [5]J. W. S. Cassels. On a problem of Rankin about the Epstein zeta-function. Proceedings of the Glasgow Mathematical Association, 4(2):73–80, 1959. See corrigendum, volume 6 (1963), page 116, and the 1964 corrections by Ennola and Diananda.DOI
  6. [6]J. W. S. Cassels. Corrigendum. Proceedings of the Glasgow Mathematical Association, 6(2):116, 1963. Corrigendum to “On a problem of Rankin about the Epstein zeta-function”.
  7. [7]P. H. Diandada. Notes on two lemmas concerning the Epstein zeta-function. Proceedings of the Glasgow Mathematical Association, 6(4):202–204, 1964.
  8. [8]Veikko Ennola. A lemma about the Epstein zeta-function. Proceedings of the Glasgow Mathematical Association, 6(4):198–201, 1964.DOI
  9. [9]Peter M. Gruber. A short analytic proof of Fejes Tóth’s theorem on sums of moments. Aequationes Mathematicae, 58:291–295, 1999.DOI
  10. [10]Elliott H. Lieb, Nicolas Rougerie, and Jakob Yngvason. Rigidity of the Laughlin liquid. Journal of Statistical Physics, 172:544–554, 2018.DOI
  11. [11]Hugh L. Montgomery. Minimal theta functions. Glasgow Mathematical Journal, 30(1):75–85, 1988.DOI
  12. [12]OpenAI. Universal optimality of the triangular lattice. OpenAI Math Release preprint OAI:Universal-optimality-of-the-triangular-lattice-September-23-2026, 2026.
  13. [13]Mircea Petrache and Sylvia Serfaty. Crystallization for Coulomb and Riesz interactions as a consequence of the Cohn–Kumar conjecture. Proceedings of the American Mathematical Society, 148:3047–3057, 2020. Numbered references in this paper use arXiv:1908.09714v3.arxiv.org/abs/1908.09714
  14. [14]R. A. Rankin. A minimum problem for the Epstein zeta-function. Proceedings of the Glasgow Mathematical Association, 1(4):149–158, 1953.DOI
  15. [15]Nathalie Revol and Fabrice Rouillier. Motivations for an arbitrary precision interval arithmetic and the MPFI library. Reliable Computing, 11:275–290, 2005.DOI
  16. [16]Etienne Sandier and Sylvia Serfaty. Improved lower bounds for Ginzburg–Landau energies via mass displacement. Analysis & PDE, 4(5):757–795, 2011.arxiv.org/abs/1011.4616
  17. [17]Etienne Sandier and Sylvia Serfaty. From the Ginzburg–Landau model to vortex lattice problems. Communications in Mathematical Physics, 313(3):635–743, 2012. Numbered references in this paper use arXiv:1011.4617v2.arxiv.org/abs/1011.4617
  18. [18]Etienne Sandier and Sylvia Serfaty. 2D Coulomb gases and the renormalized energy. The Annals of Probability, 43(4):2026–2083, 2015. Numbered references in this paper use arXiv:1201.3503v2.arxiv.org/abs/1201.3503
  19. [19]Michael Struwe. On the asymptotic behavior of minimizers of the Ginzburg–Landau model in 2 dimensions. Differential and Integral Equations, 7:1613–1624, 1994.DOI

Paper details

vlnc:W_e5nd7m4v
Version
v1 · latest
Published
License
LicenseRef-OpenAI-public
Usage

views · 19

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

unreleased internal OpenAI model

View statement
Looking for related papers…
Contents