Brownian limit of a tagged particle in a deterministic hard-sphere gas: removing the logarithmic density restriction and the equilibrium assumption

Bodineau, Gallagher and Saint-Raymond (Inventiones mathematicae, 2016) gave the first rigorous derivation of Brownian motion as the limit of a fully deterministic system of interacting particles: a tagged hard sphere in a gas of N identical hard spheres at Gibbs equilibrium converges, in the Boltzmann–Grad limit with the diffusive time rescaling, to a Brownian motion, and its distribution to a solution of the heat equation. The derivation, however, holds only in a linear regime (equilibrium background perturbed only through the tagged particle) and only while the inverse Knudsen number α = Nε^{d-1} grows more slowly than roughly √(log log N) — an astronomically smaller density than any physical regime — leaving open whether convergence persists for α growing polynomially or faster with N, and whether the equilibrium assumption can be removed. The companion nonlinear question stated at the time (deriving the full Boltzmann or Navier–Stokes equations from deterministic hard-sphere dynamics) was resolved in 2024–2025 by Deng, Hani and Ma. The open core is therefore the tagged-particle problem itself: establish, or refute, the Brownian/heat-equation limit at physical (polynomial or faster) densities in α, and for initial data beyond perturbations of global equilibrium.

Background

Hilbert's sixth problem asks for a rigorous passage from microscopic Newtonian dynamics to macroscopic continuum equations, with kinetic theory as the intermediate level. For a gas of N hard spheres of diameter ε on the torus $\mathbb{T}^d$ (particles move freely and undergo elastic specular collisions), the Boltzmann–Grad scaling sends $N\to\infty$, $\varepsilon\to 0$ with $\alpha=N\varepsilon^{d-1}$ fixed; $\alpha$ is the inverse Knudsen number (the mean free path is of order $1/\alpha$, so a particle undergoes about $\alpha$ collisions per unit time). Lanford's 1975 theorem — completed with explicit rates by Gallagher, Saint-Raymond and Texier — proves that in this scaling the one-particle density of the gas converges to the Boltzmann equation, but only on a short time interval, far shorter than any relaxation or macroscopic transport time; the obstacle is controlling the recollisions (re-encounters of particle pairs that have already collided and become dynamically correlated) and pathological branchings (collision histories whose number of dynamical predecessors grows super-exponentially).

The tagged-particle problem addresses the next scale, diffusion. Brownian motion of a tracer in a fluid is the classical picture, but deriving it from deterministic Newtonian dynamics of an interacting gas was out of reach even in linear regimes: van Beijeren, Lanford, Lebowitz and Spohn (1980) obtained the linear kinetic equations for equilibrium time-correlation functions only formally, without quantitative estimates on times diverging with N (the Lorentz gas, where the background scatterers are frozen, had been treated earlier). Bodineau, Gallagher and Saint-Raymond (2016) considered initial data of the form $M_{N,\beta}(Z_N)\rho_0(x_1)$: the gas is at the Gibbs equilibrium (a Maxwellian velocity distribution $M_\beta$ times the uniform spatial distribution, conditioned on non-overlap), and only the position of one distinguished ("tagged") particle carries a smooth perturbation $\rho_0$. In this linear regime the tagged particle should be governed by the linear Boltzmann equation $\partial_t\varphi_\alpha+v\cdot\nabla_x\varphi_\alpha=-\alpha\mathcal{L}\varphi_\alpha$, where $\mathcal{L}$ is the collision operator against the equilibrium bath.

Their two main theorems: (i) the first marginal of the tagged particle is close in $L^\infty$ to $M_\beta\varphi_\alpha$ with an error bounded by $C\big(t\alpha/(\log\log N)^{(A-1)/A}\big)^{A^2/(A-1)}$ for any fixed integer $A\ge 2$, which validates the linear Boltzmann approximation for time spans $t\alpha$ growing like a power of $\log\log N$ — far beyond Lanford's short times, the key being a pruning of super-exponential ("pathological") collision trees in the Duhamel/BBGKY series expansion (the hierarchy of equations for the successive particle marginals), combined with the maximum principle inherited from the invariant Gibbs measure; (ii) under the diffusive rescaling $\Xi(\tau)=x_1(\alpha\tau)$, if $\alpha\to\infty$ while $\alpha\ll(\log\log N)^{(A-1)/2A}$ (that is, essentially $\alpha\ll\sqrt{\log\log N}$), the process $\Xi$ converges in law to a Brownian motion of variance $\kappa_\beta=(1/d)\int v\mathcal{L}^{-1}v\,M_\beta(v)\,dv$, and the tagged-particle distribution converges to the solution of the heat equation $\partial_\tau\rho-\kappa_\beta\Delta\rho=0$. The diffusive limit requires error control over times of order $\alpha$ in microscopic units, i.e. over about $\alpha^2$ collisions of the tagged particle, which is what forces the logarithmic restriction on $\alpha$.

The restrictions are physically and mathematically substantive. First, $\alpha$ tunes the collision rate, and physical Brownian motion corresponds to enormous $\alpha$ (a tracer in a macroscopic gas suffers astronomically many collisions per observation time), whereas $\sqrt{\log\log N}$ grows negligibly; whether the convergence to the heat equation survives when $\alpha$ grows polynomially or faster with N was left open, and all the error estimates of the proof break down in that regime. Second, the equilibrium framework is essential to the method: the perturbation affects only the tagged particle, the background stays at global equilibrium (the paper notes that in this framework one cannot hope to retrieve the full nonlinear gas dynamics), and within this framework the authors already handle velocity-dependent and localized perturbations of the tagged particle; general non-equilibrium initial data for the interacting gas remained out of reach. A subsequent line of work by the same group derived, at exact equilibrium and for long times, the fluctuating Boltzmann equation (the stochastic equation governing typical fluctuations of the gas, with large deviations) — still inside the equilibrium setting. In parallel, the nonlinear companion question — deriving the full Boltzmann equation for long times, and from it the fluid equations, from deterministic hard-sphere dynamics — was resolved in 2024–2025 by Deng, Hani and Ma; that resolution does not touch the tagged-particle density restriction.

Problem Statement

In the setting of Bodineau, Gallagher and Saint-Raymond (2016) — N identical hard spheres of diameter ε on the torus $\mathbb{T}^d$, $d\ge 2$, with deterministic Newtonian collision dynamics, Boltzmann–Grad scaling $\alpha=N\varepsilon^{d-1}$, and diffusive observation $\Xi(\tau)=x_1(\alpha\tau)$ of one tagged particle — establish, or refute, the validity of the Brownian limit beyond the two restrictions of the known derivation, extending the rigorous microscopic derivation of diffusion for a tracer in a fully interacting deterministic gas to the regimes where it is physically needed; the two restrictions define two independent directions of the same overarching question, each addressable on its own establish-or-refute terms, so that a resolution of either of the two directions below, on its own, constitutes a resolution of this task. Specifically: (1) Density regime: the convergence in law of $\Xi$ to Brownian motion of variance $\kappa_\beta$ — where $\kappa_\beta=(1/d)\int v\,\mathcal{L}^{-1}v\,M_\beta(v)\,dv$ is the diffusion coefficient determined by the equilibrium bath, $\mathcal{L}$ being the linear Boltzmann collision operator and $M_\beta$ the Maxwellian at inverse temperature $\beta$ — and the convergence of the tagged-particle distribution to the solution of the heat equation $\partial_\tau\rho-\kappa_\beta\Delta\rho=0$, is proven only for $\alpha\ll\sqrt{\log\log N}$; determine whether it persists when $\alpha$ grows with N polynomially or faster — the gas must remain dilute, meaning the volume fraction $N\varepsilon^d=\alpha\varepsilon$ must still tend to 0, which caps $\alpha$ at growth slower than about $N^{1/d}$ — or whether convergence fails in some faster-growth regime of $\alpha$ within that cap. (2) Initial data: the derivation requires the background gas to be at (or a perturbation of) global Gibbs equilibrium, with only the tagged particle perturbed; determine whether the derivation extends to initial data for the interacting gas beyond perturbations of global equilibrium — away from equilibrium, where the background evolves nonlinearly and the tagged particle couples to that evolution — identifying the correct limiting process where such an extension holds, or showing that the derivation does not extend to general non-equilibrium initial data of the interacting gas. Admissible non-equilibrium data here are non-degenerate (carrying thermal motion — a velocity distribution of positive temperature, excluding ballistic data with no thermal motion), and a determination of direction (2) must state the class of initial data it covers: a positive answer identifies the correct limiting process for that class, while a failure for one admissible non-equilibrium datum refutes the extension to that class. The nonlinear companion question posed alongside these in 2016 (deriving the full Boltzmann equation or the Navier–Stokes equations from the deterministic dynamics) has been resolved by Deng–Hani–Ma and is not part of this problem.

References

1. Thierry Bodineau, Isabelle Gallagher, Laure Saint-Raymond, 'The Brownian motion as the limit of a deterministic system of hard-spheres', Inventiones mathematicae 203 (2016) 493–553. DOI: 10.1007/s00222-015-0593-9, https://doi.org/10.1007/s00222-015-0593-9; preprint arXiv:1305.3397, https://arxiv.org/abs/1305.3397
2. Oscar E. Lanford, 'Time evolution of large classical systems', in Dynamical Systems, Theory and Applications, Lecture Notes in Physics 38, Springer (1975), pp. 1–111. DOI: 10.1007/3-540-07171-7_1, https://doi.org/10.1007/3-540-07171-7_1
3. Isabelle Gallagher, Laure Saint-Raymond, Benjamin Texier, 'From Newton to Boltzmann: Hard Spheres and Short-range Potentials', Zürich Lectures in Advanced Mathematics 18, European Mathematical Society (2014). DOI: 10.4171/129, https://doi.org/10.4171/129
4. H. van Beijeren, O. E. Lanford, J. L. Lebowitz, H. Spohn, 'Equilibrium time correlation functions in the low-density limit', Journal of Statistical Physics 22 (1980), 237–257. DOI: 10.1007/bf01008050, https://doi.org/10.1007/bf01008050
5. Yu Deng, Zaher Hani, Xiao Ma, 'Long time derivation of the Boltzmann equation from hard sphere dynamics', arXiv:2408.07818 (2024; v3 July 2025), https://arxiv.org/abs/2408.07818
6. Yu Deng, Zaher Hani, Xiao Ma, 'Hilbert's sixth problem: derivation of fluid equations via Boltzmann's kinetic theory', arXiv:2503.01800 (2025), https://arxiv.org/abs/2503.01800
7. Thierry Bodineau, Isabelle Gallagher, Laure Saint-Raymond, Sergio Simonella, 'Statistical dynamics of a hard sphere gas: fluctuating Boltzmann equation and large deviations', Annals of Mathematics (2) 198 (2023), no. 3. DOI: 10.4007/annals.2023.198.3.3, https://doi.org/10.4007/annals.2023.198.3.3
8. Thierry Bodineau, Isabelle Gallagher, Laure Saint-Raymond, Sergio Simonella, 'Long-time derivation at equilibrium of the fluctuating Boltzmann equation', Annals of Probability 52 (2024). DOI: 10.1214/23-aop1656, https://doi.org/10.1214/23-aop1656; arXiv:2201.04514, https://arxiv.org/abs/2201.04514
9. Thierry Bodineau, Isabelle Gallagher, Laure Saint-Raymond, Sergio Simonella, 'Derivation of the Boltzmann equation from hard-sphere dynamics (after Y. Deng, Z. Hani, and X. Ma)', Séminaire Bourbaki (January 2026), arXiv:2602.04407, https://arxiv.org/abs/2602.04407
10. François Golse, 'Validité de la théorie cinétique des gaz : au-delà de l'équation de Boltzmann [d'après T. Bodineau, I. Gallagher, L. Saint-Raymond et S. Simonella]', Astérisque 446, Séminaire Bourbaki exposé 1198, pp. 29–86 (volume year 2023; published online January 2024). DOI: 10.24033/ast.1207, https://doi.org/10.24033/ast.1207
11. Corentin Le Bihan, 'Long Time Validity of the Linearized Boltzmann Equation for Hard Spheres: A Proof Without Billiard Theory', Archive for Rational Mechanics and Analysis 249 (2025). DOI: 10.1007/s00205-025-02105-z, https://doi.org/10.1007/s00205-025-02105-z
12. Corentin Le Bihan, 'Long-Time Validity of the Linearized Boltzmann Uncut-Off and the Linearized Landau Equations From the Newton Law', Communications on Pure and Applied Mathematics (2026). DOI: 10.1002/cpa.70035, https://doi.org/10.1002/cpa.70035
13. Florent Fougères, 'On the Derivation of the Linear Boltzmann Equation from the Nonideal Rayleigh Gas', Journal of Statistical Physics 191 (2024). DOI: 10.1007/s10955-024-03353-1, https://doi.org/10.1007/s10955-024-03353-1; arXiv:2404.03266, https://arxiv.org/abs/2404.03266
14. Karsten Matthies, Theodora Syntaka, 'Derivation of kinetic and diffusion equations from a hard-sphere Rayleigh gas using collision trees and semigroups', arXiv:2405.04449 (2024), https://arxiv.org/abs/2405.04449
15. Thierry Bodineau, Pierre Le Bris, 'Linear Landau equation as a limit of a tagged particle in mean field interaction with a free gas', arXiv:2602.16440 (2026), https://arxiv.org/abs/2602.16440