View source for Long-time derivation of the Boltzmann equation

'''Long-time derivation of the Boltzmann equation''' concerns the Boltzmann–Grad limit of a classical gas: whether the microscopic Newtonian dynamics of ''N'' particles, as the interaction range tends to zero with the collision frequency kept of order one, is described by the nonlinear Boltzmann equation on the whole time interval on which that equation has a regular solution, rather than only for short times. Two manuscripts dated 23 September 2026 in OpenAI's public mathematics collection ([https://github.com/openai/math github.com/openai/math]) claim such a derivation for stable radial potentials and a Gaussian fluctuation theorem for hard spheres on the same interval. This page summarizes the published claims and lists publicly available follow-up work. Statements are the respective authors' claims; this page does not certify them.

== Background ==
* As recounted in the introductions of the two manuscripts:
** The scaling goes back to Grad. Lanford established the hard-sphere limit on a short time interval, and King's thesis treated positive finite-range potentials. Gallagher, Saint-Raymond and Texier analysed recollisions in detail, with a convergence rate for hard spheres, and Pulvirenti, Saffirio and Simonella proved a short-time theorem for stable radial potentials that may have attractive parts.
** Global-time results were obtained by Illner and Pulvirenti for sufficiently rare gases dispersing in vacuum. For general data, Pulvirenti and Simonella quantified correlation errors, and Bodineau, Gallagher, Saint-Raymond and Simonella developed a cluster expansion on physical trajectories.
** Deng, Hani and Ma proved the hard-sphere limit throughout every finite interval on which the Boltzmann equation has a regular solution, and discussed smooth potentials and long-time nonequilibrium fluctuations as further problems.
** For fluctuations, van Beijeren, Lanford, Lebowitz and Spohn derived short-time equilibrium covariances and Spohn the nonequilibrium second-moment limit. Bodineau, Gallagher, Saint-Raymond and Simonella developed the short-time nonequilibrium fluctuation theory, including convergence of the fluctuation process, and long-time results at equilibrium, for a spatially periodic gas.

== The original results ==
* '''Claim (stable radial potentials).''' The main theorem of ''The Boltzmann–Grad limit for stable radial potentials on regular kinetic intervals'' concerns a grand-canonical Newtonian gas in three dimensions with initial pair exclusion and a stable, finite-range radial potential that is ''C''<sup>2</sup> except for a possible repulsive singularity at the origin. For ''C''<sup>1</sup> initial densities with spatially summable Gaussian bounds on the density and its spatial gradient, on every finite interval on which the Boltzmann equation has a classical solution with a uniform Gaussian bound, every fixed-order rescaled factorial marginal converges in ''L''<sup>1</sup> to the corresponding tensor power of the solution, uniformly in time. Attractive wells, long scattering delays and noninjective scattering maps are allowed. A corollary gives convergence in probability of empirical observables.
* '''Claim (hard-sphere fluctuations).''' ''Hard-sphere fluctuations on the regular Boltzmann lifespan'' states a finite-dimensional central limit theorem away from equilibrium for the deterministic grand-canonical hard-sphere gas in the whole space: on every finite interval on which the Boltzmann solution keeps uniform Gaussian velocity decay, the empirical fluctuation field, centered by its exact microscopic expectation, converges in finite-dimensional laws to the Gaussian solution of the linear fluctuating Boltzmann equation. The manuscript states that it does not prove convergence of the fluctuation process in path space, and that it uses the layered expansion of Deng, Hani and Ma.
* Neither manuscript asserts global regularity of the Boltzmann equation: the regular solution on the given interval is a hypothesis.
* '''Sources.''' At commit <code>1581d4c</code>: [https://github.com/openai/math/tree/1581d4cbdd69a0faf34b2089614fe2a814316a6a/preprints/The-Boltzmann-Grad-limit-for-stable-radial-potentials-on-regular-kinetic-intervals-September-23-2026 The Boltzmann–Grad limit for stable radial potentials on regular kinetic intervals] and [https://github.com/openai/math/tree/1581d4cbdd69a0faf34b2089614fe2a814316a6a/preprints/Hard-sphere-fluctuations-on-the-regular-Boltzmann-lifespan-September-23-2026 Hard-sphere fluctuations on the regular Boltzmann lifespan].
* '''Formalization.''' The repository's Lean documentation has no scope note for this family at that commit.

== Public follow-up work ==
{| class="wikitable"
! Date !! Work !! Relation !! Public claim !! Public status (checked 23 October 2026)
|-
| 2026-10-07 || mxym, ''Functional hard-sphere fluctuations on regular kinetic intervals'', research manuscript 009, version 1 ([https://github.com/mxym/math/tree/c86f853d067e2105c508d6c25e2739593c166199/preprints/009-functional-hard-sphere-fluctuations mxym/math, preprints/009-functional-hard-sphere-fluctuations]) || Extension (convergence in path space) || For hard spheres in the Boltzmann–Grad scaling, the exactly centered fluctuation field converges weakly as a process on a prescribed regular Boltzmann interval, in the Skorokhod space of paths in tempered distributions on '''R'''<sup>6</sup> with the strong dual topology; the limit law is supported on continuous paths and its finite-dimensional evaluations are the Gaussian target of the fluctuation manuscript. The analytic estimates and the finite-dimensional limit of that manuscript are stated, pinned inputs; the new parts include recollision bounds, increment estimates and tightness. The manuscript does not claim a rate of convergence or uniqueness for a broader stochastic equation. || Repository manuscript prepared with AI assistance; its README describes model-assisted reviews, not external peer review or proof-assistant verification, and makes no priority claim.
|-
| 2026-10-07 || mxym, ''Full-density virial and collisional stress'', research supplement ([https://github.com/mxym/math/tree/c86f853d067e2105c508d6c25e2739593c166199/notes/full-density-virial-stress mxym/math, notes/full-density-virial-stress]) || Uses (pinned analytic package) || For hard spheres with compactly supported initial positions and velocities, first-order mean corrections for the virial ''x''·''v'' and the spatial second moment at full amplitude, and a local momentum-balance defect identified as the classical collisional-stress tensor. It is conditional on an explicit history package taken from the fluctuation manuscript and does not import its fluctuation conclusions; the README states that the whole one-particle mean correction remains open. || Repository supplement prepared with AI assistance; not peer reviewed.
|-
| 2026-10-16 || J. Jardine, Lattice24, referee reviews [https://github.com/JJardine919/openai-math-audit/blob/a06a8c6e77974d4c6021338a55a8e91b634f67f7/reviews/rank_072.json rank_072] (stable radial potentials) and [https://github.com/JJardine919/openai-math-audit/blob/a06a8c6e77974d4c6021338a55a8e91b634f67f7/reviews/rank_118.json rank_118] (hard-sphere fluctuations) in [https://github.com/JJardine919/openai-math-audit JJardine919/openai-math-audit] || Review (AI-assisted audit) || Both report no gap. The first records one minor bookkeeping slip in the closure section, a bound on a stopped selected union that should read ''K''<sub>''j''</sub>''D''<sub>''j''</sub> rather than ''U''<sub>''j''</sub>''D''<sub>''j''</sub>, and states that the larger quantity is still dominated by the weight used afterwards; it checked the exposure and replay argument for logical structure rather than every edge case. The second states that it did not read Section 4, the marked component expansion, or the later parts of Section 6, and that it took the imported Deng–Hani–Ma propositions on trust. || Part of an audit of the collection's manuscripts that the audit describes as complete.
|}

== Notes ==
* The earlier results are due to the authors named in the Background section, as recounted by the manuscripts.
* The list covers follow-ups found in public sources up to the date above; it is not a priority or completeness claim.

[[Category:OpenAI results]]