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

BKT universality for height and planar spin fields

For every admissible finite-range square-symmetric interaction J⊂Z2∖{0}J \subset \mathbb{Z}^2 \setminus \{0\} containing nearest neighbors, the discrete Gaussian height model has a finite positive roughening threshold 0<βc(J)<∞0 < \beta_{\mathrm{c}}(J) < \infty, and for β≥βc(J)\beta \ge \beta_{\mathrm{c}}(J) its mean-subtracted torus field HNH_N converges in H−3(T2)H^{-3}(\mathbb{T}^2) to βeff(J,β)/vJ2ΦT\sqrt{\beta_{\mathrm{eff}}(J,\beta)/v_J^2}\Phi_{\mathbb{T}} with βeff(J,βc(J))=8πvJ2\beta_{\mathrm{eff}}(J,\beta_{\mathrm{c}}(J)) = 8\pi v_J^2, infinite right slope at βc(J)\beta_{\mathrm{c}}(J), differentiability on (βc(J),∞)(\beta_{\mathrm{c}}(J),\infty), high-temperature asymptotics βeff(J,β)=β+OJ(e−cJβ)\beta_{\mathrm{eff}}(J,\beta) = \beta + O_J(e^{-c_J\beta}), and spread-out onset βc(Jρ)/(8πvJρ2)→1\beta_{\mathrm{c}}(J_\rho)/(8\pi v_{J_\rho}^2) \to 1 as ρ→∞\rho \to \infty.

For each M∈{Villain,XY}M \in \{\mathrm{Villain}, \mathrm{XY}\} there exist a finite b0(M)>0b_0(M) > 0 and functions KM(b)>1/(4π)K_M(b) > 1/(4\pi) and AM(b)>0A_M(b) > 0 for b≥b0(M)b \ge b_0(M) with KM(b)/b→1K_M(b)/b \to 1 as b→∞b \to \infty such that the Green-function-normalized spin field SM,b,nS_{M,b,n} on D=(−1,1)2D = (-1,1)^2 with zero boundary angles converges in Hloc−3(D)H^{-3}_{\mathrm{loc}}(D) along the full sequence of integers n→∞n \to \infty to the imaginary Gaussian multiplicative chaos VKM(b),DV_{K_M(b),D}, with joint convergence of smeared fields and convergence of all mixed moments.

The choices in Theorem 1.2 for the Villain model can be made so that π2b≥βc(Jnn)\pi^2 b \ge \beta_{\mathrm{c}}(J_{\mathrm{nn}}) throughout the asserted Villain range and KVillain(b)=βeff(Jnn,π2b)/π2K_{\mathrm{Villain}}(b) = \beta_{\mathrm{eff}}(J_{\mathrm{nn}}, \pi^2 b)/\pi^2.

At every positive structural coefficient ah=a(kh)>0a_h = a(k_h) > 0, noisy block averages Y=Bsh+ηY = B_s h + \eta of the discrete Gaussian height model admit an exact local activity representation Km(X;k,a′)K_m(X; k, a') on a neighborhood of (kh,ah)(k_h, a_h) that is arbitrarily small in the analytic regulator norm, analytic in kk at fixed a′a', and exact on fitting dyadic tori.

Methods

Finite lattice Gaussian comparisons, free-rectangle cutting, and FKG-based folded tests used to construct the structural height coefficient and prove its forbidden range.

Source excerpts

“Lemma 2.1 (Finite Gaussian comparisons).” · Show in paper

“Theorem 2.2 (All free-rectangle limits).” · Show in paper

Sigma-finite cable representation, annular exploration, and path anchoring used to transfer free-rectangle Gaussian limits to Dirichlet domains, tori, and mixed-boundary domains.

Source excerpts

“For the cable representation, give every retained unoriented bond of conductance ce a unit-time independent Brownian bridge of variance parameter k/ce, conditional on its endpoint heights.” · Show in paper

“Proposition 3.7 (Vanishing reach).” · Show in paper

Finite-range polynomial covariance decomposition and analytic polymer/local partition algebra for renormalization group flows.

Source excerpts

“Lemma 4.1 (Finite-range decomposition).” · Show in paper

“Proposition 4.8 (Exact local partition map).” · Show in paper

Transfer from separated signed spin correlations and a uniform two-point bound to full complex spin-field convergence in H−3H^{-3}_{loc}(D) and mixed-moment convergence using Lee–Yang moment bounds.

Source excerpts

“Lemma 5.10 (Lee–Yang absolute moment bound).” · Show in paper

“Theorem 5.12 (Transfer from signed correlations to the full spin field).” · Show in paper

Exact branch representation, mixed Gaussian histories, and three-dimensional topological shooting for the ordinary XY model.

Source excerpts

“Proposition 6.2 (Unwrapping and sources).” · Show in paper

“Proposition 6.15 (The exact XY activity calculus).” · Show in paper

Noisy block-average observation identity, covariant isolation expansion, and physical terminal diagnostics for identifying the critical height trajectory and its infinite endpoint slope.

Source excerpts

“Theorem 9.1 (Local entry at every positive coefficient).” · Show in paper

“10.1 Physical terminal diagnostics” · Show in paper