Rigorous Mesoscopic Limits of Many-Particle Systems

Rigorous Mesoscopic Limits of Many-Particle Systems

OpenAI's claim

Hard-sphere fluctuation limit: The paper claims finite-dimensional Gaussian fluctuations for a nonequilibrium hard-sphere gas throughout any finite interval of the regular Boltzmann lifespan; centering subtracts the exact expectation in the particle model, rather than its kinetic approximation.

Boltzmann--Grad limit for stable potentials: The manuscript claims convergence of all fixed-order rescaled factorial marginals for stable finite-range radial interaction potentials throughout regular kinetic intervals; attractive wells and dynamically created clusters are permitted under the stated regularity and Gaussian data bounds.

Mathematical statement of the claimed result

Mathematical claim 1.

Let $\|g\|_{\mathrm{Bol},\beta}=\sum_{a\in\mathbb Z^3}\sup_{|x-a|\le1,v}e^{\beta|v|^2}|g(x,v)|$. Suppose the smooth probability density $f_0\ge0$ has $\|f_0\|_{\mathrm{Bol},2\beta}+\|\nabla_xf_0\|_{\mathrm{Bol},2\beta}<\infty$ and a nonnegative classical hard-sphere Boltzmann solution $f$ on $[0,T]$ satisfies $\sup_{t,x,v}e^{2\beta|v|^2}f<\infty$. Start diameter-$\epsilon$ Newtonian hard spheres from the grand-canonical law proportional to $\mu^N/N!\prod_if_0(z_i)\mathbf1_{|x_i-x_j|>\epsilon}\,dZ_N$, $\mu=\epsilon^{-2}$. Define \[\zeta_t^\epsilon(\phi)=\sqrt\mu\left[\mu^{-1}\sum_i\phi(z_i(t))-\mathbb E\left(\mu^{-1}\sum_i\phi(z_i(t))\right)\right].\] For every finite collection of times and compactly supported smooth real tests, their joint law is claimed to converge to a centered Gaussian field $\zeta$ with $d\zeta_t(\phi)=\zeta_t(A_t\phi)\,dt+dW_t(\phi)$. Put $B=[(v-v_*)\cdot\omega]_+$, $v^\prime=v-((v-v_*)\cdot\omega)\omega$, $v_*^\prime=v_*+((v-v_*)\cdot\omega)\omega$, and $\Delta\phi=\phi(x,v^\prime)+\phi(x,v_*^\prime)-\phi(x,v)-\phi(x,v_*)$. Then \[A_t\phi=v\cdot\nabla_x\phi+\int Bf_t(x,v_*)\Delta\phi\,dv_*\,d\omega,\qquad C_t(\phi,\psi)=\tfrac12\int Bf_tf_{t,*}\Delta\phi\Delta\psi\,dx\,dv\,dv_*\,d\omega.\] The initial covariance is $\int f_0\phi\psi$, and $\mathbb E[W_t(\phi)W_s(\psi)]=\int_0^{t\wedge s}C_r(\phi,\psi)\,dr$, with independent increments and initial field independent of $W$. The assertion is finite-dimensional convergence with exact microscopic centering, conditional on the regular kinetic lifespan.

Mathematical claim 2.

Fix a radial potential $\Phi(q)=\phi(|q|)$ vanishing for $|q|\ge1$, either $C^2$ on $\mathbb R^3$ or $C^2$ off zero and tending to $+\infty$ at zero. Require thermodynamic stability $\sum_{i<j}\Phi(q_i-q_j)\ge-BN$ for every finite distinct-point configuration. Let $f_0\ge0$ be a $C^1$ probability density with finite Gaussian Boltzmann norms $\|f_0\|_{\mathrm{Bol},2\beta}+\|\nabla_xf_0\|_{\mathrm{Bol},2\beta}$, where $\|g\|_{\mathrm{Bol},\gamma}=\sum_{a\in\mathbb Z^3}\sup_{|x-a|\le1,v}e^{\gamma|v|^2}|g|$. Assume a classical probability-density solution of $(\partial_t+v\cdot\nabla_x)f=Q_\Phi(f,f)$ on fixed $[0,T]$ has $\sup_{t,x,v}e^{2\beta|v|^2}f<\infty$, with $Q_\Phi$ defined by the actual two-body scattering of $\Phi$. Evolve Newtonian particles with interaction $\Phi((x_i-x_j)/\epsilon)$ from the exactly normalized grand-canonical law proportional to $\mu^N/N!\prod_if_0(z_i)\mathbf1_{|x_i-x_j|>\epsilon}$, $\mu=\epsilon^{-2}$. Its scaled ordered factorial densities are characterized by $\int aF_s^\epsilon=\mu^{-s}\mathbb E\sum_{i_1,\ldots,i_s\text{ distinct}}a(z_{i_1},\ldots,z_{i_s})$. The claim is \[\sup_{0\le t\le T}\|F_s^\epsilon(t)-f(t)^{\otimes s}\|_{L^1((\mathbb R^6)^s)}\longrightarrow0\quad\text{for every fixed }s\ge1.\] Attractive wells and noninjective scattering are permitted; $\Phi$ and $T$ are fixed independently of $\epsilon$.

Outline of the argument

Hard-sphere fluctuation limit: Layered collision histories are expanded by connected components while retaining finite observation records and exact factorial cancellation. Relative translations give a merger gain, and excluded-volume links and redundant contacts are controlled at the sharp cumulant scale. Only the appropriate signed trees survive in the limit, producing the linear fluctuating Boltzmann covariance and Gaussian finite-dimensional laws.

Boltzmann--Grad limit for stable potentials: Fine time cells give an exact inclusion--exclusion expansion of entire interaction components, while coarser layers are centered against the known kinetic solution. Energy is retained across complete components, including clusters with many internal encounters. Polynomial cutoffs and separated detached trajectories control the signed terms and yield uniform $L^1$ convergence after summing the finite layers.

Formal verification

Hard-sphere fluctuation limit: The pinned catalogue lists no Lean formalization for this result.

Boltzmann--Grad limit for stable potentials: The pinned catalogue lists no Lean formalization for this result.

This site has not run the Lean development. These links describe OpenAI’s selected formal statements and should be read with their stated scope.

Status caveat

OpenAI reports that an internal model produced the manuscripts. Their arguments have not been independently verified as proofs by this site. Any portion outside the documented formal scope may contain mistakes, and the catalogue label reports the announced mathematical claim.

Sources

• OpenAI, Hard-sphere fluctuations on the regular Boltzmann lifespan, 2026-09-23; citation key OAI:Hard-sphere-fluctuations-on-the-regular-Boltzmann-lifespan-September-23-2026. PDF (GitHub); PDF (direct); Manuscript page and citation; Latest version; Statement source: thm:main.
• OpenAI, The Boltzmann–Grad limit for stable radial potentials on regular kinetic intervals, 2026-09-23; citation key OAI:The-Boltzmann-Grad-limit-for-stable-radial-potentials-on-regular-kinetic-intervals-September-23-2026. PDF (GitHub); PDF (direct); Manuscript page and citation; Latest version; Statement source: thm:main.
• OpenAI catalogue entry.