By OpenAISourceGitHub

Manuscripts from openai/math

Fork
Clone
git clone https://openversions.com/git/index/openai-math.git
Cite
Version link
https://openversions.com/index/openai-math/commit/dfd8c46870b48511e697cbb43a3c8c91d375fa2f
Version identifier (commit hash)
dfd8c46870b48511e697cbb43a3c8c91d375fa2f

Fork to create your own copy with credit to the source. Clone to get a Git copy on your computer.

All papers

Claims

Findings

Triangular minimality for planar Coulomb renormalized energy

For every cutoff family satisfying (1.1) and every field E∈A1E \in \mathcal{A}_1, W(E)≥W(E△)W(E) \geq W(E_\triangle), where W(E△)W(E_\triangle) is finite and inf⁡E∈A1W(E)=W(E△)\inf_{E \in \mathcal{A}_1} W(E) = W(E_\triangle).

For every integer n≥2n \geq 2 and every configuration of nn distinct points on Tn=R2/(nZ2)T_n = \mathbb{R}^2/(\sqrt{n}\mathbb{Z}^2), the mean-zero potential hh in (1.6) satisfies Wn(h)≥nW(E△)\mathcal{W}_n(h) \geq n W(E_\triangle).

For the optimal ordered-pair logarithmic energy Elog⁡(n)E_{\log}(n) on the unit sphere S2⊂R3S^2 \subset \mathbb{R}^3, as n→∞n \to \infty through the integers, Elog⁡(n)=(12−log⁡2)n2−n2log⁡n+CBHSn+o(n)E_{\log}(n) = (\frac{1}{2} - \log 2)n^2 - \frac{n}{2}\log n + C_{\mathrm{BHS}}n + o(n), where CBHS=2log⁡2+12log⁡(23)+3log⁡(πΓ(1/3))C_{\mathrm{BHS}} = 2\log 2 + \frac{1}{2}\log(\frac{2}{3}) + 3\log(\frac{\sqrt{\pi}}{\Gamma(1/3)}).

If for every integer n≥2n \geq 2 and every configuration of nn distinct points of TnT_n, the mean-zero potential hh in (1.6) satisfies Wn(h)≥nC\mathcal{W}_n(h) \geq nC, then W(E)≥CW(E) \geq C for every E∈A1E \in \mathcal{A}_1.

For every n≥2n \geq 2, a minimizing configuration of Wn\mathcal{W}_n on TnT_n has torus distance at least r∗=1/πr_* = 1/\sqrt{\pi} between distinct points, and its full lift to R2\mathbb{R}^2 has the same minimum-separation bound.

For compatible periodic Voronoi data on T=R2/Γ\mathbb{T} = \mathbb{R}^2/\Gamma with N≥1N \geq 1 distinct sites, lower edge distance d0>0d_0 > 0, and ∣T∣=N|\mathbb{T}| = N, the mean-zero potential hh and glued trial potential HH satisfy WT(h)=∑SJ(S)+12∫T∣∇(h−H)∣2 dx≥∑SJ(S)\mathcal{W}_{\mathbb{T}}(h) = \sum_S J(S) + \frac{1}{2}\int_{\mathbb{T}} |\nabla(h - H)|^2\,\mathrm{d}x \geq \sum_S J(S).

The exact constants and functions of (5.16) and (5.14) satisfy K<0K < 0, 2gd(x)≥K2g_d(x) \geq K for d≥bd \geq b and x≥0x \geq 0, and ∫0∞max⁡{0,−G(d,s)} ds≤−K\int_0^\infty \max\{0, -G(d, s)\}\,\mathrm{d}s \leq -K for d≥bd \geq b, where b=28209/100000b = 28209/100000.

Methods

Gluing a continuous H1 trial potential H across Voronoi sectors on a flat torus using common radial functions and matching traces along Voronoi edges and vertex rays, yielding an exact dual identity for the renormalized energy.

Source excerpts

“Lemma 4.2 (Gluing and singular normalization). For compatible periodic Voronoi data on T with N ≥ 1 distinct sites and lower edge distance d0, let I be finite or countable, let ki ≥ 6 be integers with repetitions permitted, and let B, Di ∈ C1((0, ∞)) satisfy (4.3)–(4.4). The sector functions (4.5) define a Γ-periodic function H that is continuous off the sites and belongs to H1(Ωη) for every admissible η < d0.” · Show in paper

“WT(h) = ∑S J(S) + 1/2 ∫T |∇(h − H)|^2 dx ≥ ∑S J(S).” · Show in paper

Screening point charges with radial disks of area one and applying the strict minimum principle to W2,sW^{2,s} supersolutions to prove uniform separation of minimizing configurations on finite square tori.

Source excerpts

“The argument uses a screened logarithmic potential and the minimum principle.” · Show in paper

Using local L^r energy estimates, mass-displacement measure bounds, and periodic averaging to reduce the full-plane infimum over A1 with fixed-width square cutoffs to a uniform lower bound on finite square tori.

Source excerpts

“We also use the following nonsharp consequences of Lemma 4.7 and Proposition 4.9 of [17].” · Show in paper

“Proposition 2.4 (Reduction to square tori).” · Show in paper

Outward-rounding interval arithmetic with denominator q = 2^{144} and degree-3 interval jets combined with analytic infinite-mode truncation bounds to certify scalar inequalities on a compact rectangle.

Source excerpts

“Put q = 2^144. A scalar interval is represented by integers u ≤ v, meaning [u/q, v/q].” · Show in paper

“To bound derivatives we use interval jets (F0, F1, F2, F3), Fj ⊇ F^(j)(t)/j!.” · Show in paper