Derivation of the Boltzmann equation from hard-sphere dynamics (after Y. Deng, Z. Hani, and X. Ma)

TL;DR

This work presents a rigorous derivation of the Boltzmann equation from hard-sphere dynamics for times extending beyond Lanford’s original short-time result, under the assumption that the Boltzmann solution remains smooth. The authors develop a dynamical cluster expansion and a cumulant-based iterative scheme that isolates the maximally factorised (chaotic) part and treats correlation errors via a detailed graph-based analysis of collision histories, including a sophisticated handling of recollisions with a UP algorithm and Burago dispersion bounds. The result shows convergence of the one-particle density $f_1^{\varepsilon}(t)$ to the Boltzmann solution $f(t)$ in $L^1$ (and stability in $L^{\infty,1}_β$) on the lifespan of $f$, thereby connecting microscopic reversible dynamics with macroscopic irreversibility and providing a framework for studying fluctuations and hydrodynamic limits. The techniques yield insights into how chaos propagates over long times and open pathways to extensions to torus settings, noncompact interactions, and stochastic or fluctuating regimes, with several challenging open problems remaining.

Abstract

Consider a microscopic system of $N$ hard spheres that are initially independent (modulo the exclusion condition on particle positions) and identically distributed in $\mathbb{R}^3$. When the number $N$ of particles goes to infinity and the diameter $\varepsilon$ of the particles goes to zero, and under the weak density assumption $N\varepsilon^2=1$, it has been known since the work of Lanford that the empirical measure for the particles converges to the solution of the Boltzmann equation in a short time interval. In particular, the particles remain dynamically independent, in this limit and in the short time interval where the correlations induced by their collisions remain under control. In a recent work, Y. Deng, Z. Hani and X. Ma successfully obtained the same convergence result in arbitrary large time; more precisely, the convergence result holds for any time interval on which the Boltzmann equation has a regular solution. In this note, we explain a few elements of their proof.

Figures (18)

• Figure 1: The hard-sphere scattering law.
• Figure 2: The hard-sphere dynamics and the deflection condition.
• Figure 3: This figure encodes the hard-sphere dynamics of 11 particle trajectories during the time interval $[0,\tau]$. The particles (which are "independent") at initial time are represented by red bullets and their paths by lines. A collision between 2 particles is represented by a circle and leads to a scattering of the particles : as an example, a particle trajectory is represented by the green broken line. This evolution leads to the collision graphs $\{ \lambda_1, \dots, \lambda_5 \}$ which contain different numbers of particles (in particular $\lambda_3$ and $\lambda_5$ contain only one particle).
• Figure 4: Compared to Figure \ref{['figure: trajectoires [0,tau] no cycle']}, the collision graphs $\lambda_1$ and $\lambda_4$ display cycles. The top (bottom) collisions in those graphs are recollisions if we follow the cluster trajectory forward (backward) in time.
• Figure 5: Above, the collisions graphs in Figure \ref{['figure: trajectoires [0,tau]']} have been associated to form two clusters : on the left $\lambda_1,\lambda_2, \lambda_3$ are now linked by three overlaps, represented by black dots and on the right $\lambda_4, \lambda_5$ are linked by one overlap. Note that the overlaps on the left generate one cycle.