Rigorous global-in-time derivation of the Boltzmann equation from the finite-$N$ BBGKY hierarchy in polynomially weighted $L^\infty$ spaces

The rigorous derivation of the Boltzmann equation from a finite system of $N$ hard spheres — Lanford's theorem — is valid only for a short time, and its extension to arbitrarily long times has been a central open problem of kinetic theory. Ampatzoglou, Miller, Pavlović and Tasković (2024) proved global-in-time well-posedness of the limiting Boltzmann hierarchy in polynomially weighted $L^\infty$ spaces, combining $L^\infty$ a priori bounds on the collision term with a Klainerman–Machedon board-game reorganization of the Dyson series, and posed the finite-$N$ question recorded here: can this program be lifted to the BBGKY hierarchy of the actual particle system, so that the Boltzmann equation is rigorously derived globally in time in polynomially weighted $L^\infty$ spaces, despite the singular collision geometry (the non-overlap exclusion domain, recollisions, and finite-$N$ collision trees) that the limiting hierarchy does not see? A major later development — Deng, Hani and Ma's long-time derivation of the Boltzmann equation from hard-sphere dynamics, valid for arbitrarily long times conditional on a regular Boltzmann solution — resolves the headline question for data with one spatial derivative and Gaussian velocity weights, by a different method. The problem as stated remains open: no result by any method derives the Boltzmann equation globally in time for data merely bounded in polynomially weighted $L^\infty$ norms without spatial regularity, and the $L^\infty$-based BBGKY route is untouched.

Background

The Boltzmann equation $\partial_t f + v\cdot\nabla_x f = Q(f,f)$ describes the evolution of the one-particle phase-space density $f(t,x,v)$ of a dilute gas; the bilinear collision operator $Q(f,f)$ accounts for binary elastic collisions that redistribute the velocities of colliding pairs according to the interaction law. Whether this effective equation can be rigorously derived from the deterministic mechanics of finitely many particles is a classical question going back to Hilbert's sixth problem.

For the standard model of $N$ hard spheres of diameter $\varepsilon$ in $\mathbb{R}^d$, $d\geq 2$, the dynamics is free transport on the non-overlap domain — the set of configurations in which no two sphere centers are closer than $\varepsilon$ — with specular reflection at collisions. The $k$-particle correlation functions, i.e. the marginals of the $N$-particle probability measure, satisfy a coupled linear system of transport equations with collision boundary conditions, called the BBGKY hierarchy (after Bogoliubov, Born, Green, Kirkwood, and Yvon). In the Boltzmann-Grad scaling $N\varepsilon^{d-1}\simeq 1$, in which the total collision cross-section per particle stays of order one as $N\to\infty$ and $\varepsilon\to 0$, the BBGKY hierarchy formally converges to the Boltzmann hierarchy: the same coupled system, but with the exclusion constraint between distinct particles dropped. If the correlation functions remain approximately factorized, $f^{(k)}\approx f^{\otimes k}$ — a property called propagation of chaos — the Boltzmann hierarchy collapses to the single nonlinear Boltzmann equation for $f$.

Lanford (1975) proved this convergence rigorously for hard spheres, thereby validating the Boltzmann equation, but only up to a time that is a small fraction of the mean free time; his argument was later rewritten and completed in full rigor for hard spheres and short-range potentials by Gallagher, Saint-Raymond and Texier (2014), still on short times. The barrier is combinatorial: mild solutions, meaning solutions represented by the iterated Duhamel expansion (the Dyson series), involve sums over ordered chains of binary collisions, and the number of such collision histories grows factorially with the number of collisions, so term-by-term estimates diverge beyond short times. The only early escape from the short-time regime is due to Illner and Pulvirenti (1989), who proved global-in-time validity but only for initial data exponentially close to vacuum, an assumption that tames the combinatorics. At the level of the Boltzmann equation itself (not its derivation), long-time behavior is much better understood: Illner and Shinbrot (1984) proved global existence of mild solutions for small data on $\mathbb{R}^d$ in weighted $L^\infty$ spaces, with polynomial velocity weights available in later refinements.

The source paper (Ampatzoglou, Miller, Pavlović, Tasković, 2024; published 2025) addresses the limiting Boltzmann hierarchy directly. It proves global-in-time existence and uniqueness of mild solutions in polynomially weighted $L^\infty_{x,v}$ spaces, in which each marginal $f^{(k)}$ is bounded in a sup-norm with polynomial weights in position and velocity, while the growth of the bounds in the marginal index $k$ is controlled by a chemical potential — a parameter representing the energy carried by the addition of one particle to the system. Uniqueness combines an $L^\infty$-based a priori bound on the hierarchy's collision term with a reorganization of the Dyson series via the Klainerman–Machedon board game, a combinatorial algorithm introduced for proving uniqueness of the Gross–Pitaevskii hierarchy in dispersive PDE, which groups collision histories so that only exponentially, rather than factorially, many grouped terms must be estimated. Existence is constructive for admissible data — data that are marginals of a probability measure, decomposed by a de Finetti-type representation (a theorem expressing exchangeable families of marginals as mixtures of tensorized states) into components evolved by the small-data global $L^\infty$ theory of the Boltzmann equation. In its closing discussion the paper states the open direction recorded here: whether this program can be lifted to the BBGKY hierarchy at finite $N$, i.e. whether the Boltzmann equation can be rigorously derived globally in time in polynomially weighted $L^\infty$ spaces; the authors note that the presence of actual particle collisions makes this a subtler problem requiring further investigation. The subtlety is geometric: at finite $N$ the dynamics lives on the non-overlap domain, particles can recollide, and collision trees are genuinely finite, producing singular error terms absent from the Boltzmann hierarchy.

After the source paper appeared, Deng, Hani and Ma (2024) proved a long-time derivation of the Boltzmann equation from hard-sphere dynamics by a different route: they propagate a long-time cumulant ansatz — correlation corrections beyond the tensorized state that keep memory of the full collision history — and prove its smallness in $L^1$ through a cutting algorithm applied to diagrammatic representations of collision histories. Their convergence holds for arbitrarily long times, conditional on the existence of a regular Boltzmann solution, but for initial data possessing one spatial derivative and Gaussian velocity decay. For data merely bounded in polynomially weighted $L^\infty$ norms — the source's admissible class, matching the small-data global well-posedness theory of the equation — no long-time derivation from particle dynamics is known by any method.

Problem Statement

Determine whether the Boltzmann equation can be rigorously derived globally in time from the finite-$N$ BBGKY hierarchy in polynomially weighted $L^\infty$ spaces, as posed in the source paper's closing discussion.

Concretely: consider the $N$-particle hard-sphere system (or a cutoff interaction with cross-section $B(\sigma,u)=|u|^{\gamma}b(\hat u\cdot\sigma)$, $\gamma\in(1-d,1]$, $b\in L^{\infty}$ — the source's range from moderately soft potentials up to hard spheres, with hard spheres as the model case) in $\mathbb{R}^d$, $d\geq 3$ as in the source's hierarchy theory, under the Boltzmann-Grad scaling $N\varepsilon^{d-1}\simeq 1$, evolving on the non-overlap domain with specular reflections. The initial data are initially chaotic (factorized) at the particle level — generated by a single one-particle profile $f_0$, as in the grand-canonical ensemble of the classical derivation setting — with $f_0$ bounded in the source's polynomially weighted $L^\infty_{x,v}$ class: norms $\|g^{(k)}\|_{k,p,q,\alpha,\beta}=\sup \langle\langle\alpha X_k\rangle\rangle^{p}\langle\langle\beta V_k\rangle\rangle^{q}\,|g^{(k)}|$ (which impose polynomial decay in both position and velocity), with $p>1$, $q>\max\{d-1+\gamma,d-1\}$, $\alpha,\beta>0$, and growth in the marginal index $k$ controlled by a chemical potential $\mu$ with $e^{\mu}>4C_{p,q,\alpha,\beta}$ (the source's uniqueness threshold, its Theorem 2.7); no spatial derivatives are assumed. Establish or refute the following: on every time interval $[0,T]$ on which the limiting Boltzmann equation possesses a mild solution bounded in the same polynomially weighted $L^\infty$ class — uniqueness in that class holding for $e^{\mu}>4C_{p,q,\alpha,\beta}$ by the source's Theorem 2.7, so that the tensorized limit below is well-defined — the BBGKY correlation functions converge, as $N\to\infty$ with $N\varepsilon^{d-1}\simeq 1$, to the tensorized Boltzmann solution $f(t)^{\otimes k}$ for each fixed $k$ (propagation of chaos), uniformly on $[0,T]$ with $T$ arbitrarily large — so that for small data, where global-in-time $L^\infty$ Boltzmann solutions exist by the Illner–Shinbrot-type theory, the validity of the Boltzmann equation from particle dynamics holds for all time.

In particular, the question is whether the Boltzmann equation can be rigorously derived globally in time from the finite-$N$ BBGKY hierarchy in polynomially weighted $L^\infty$ spaces, where genuine particle collisions produce the non-overlap exclusion domain, recollisions, and finite-$N$ collision trees — singular error terms absent at the Boltzmann-hierarchy level. A complete answer must establish convergence for initial data bounded only in the source's polynomially weighted $L^\infty$ class, with no spatial derivatives; a proof may internally regularize or approximate the data, but conclusions established only for data with extra spatial regularity or Gaussian rather than polynomial velocity weights address the already-settled regular-data regime (Deng–Hani–Ma) and do not resolve this problem.

References

1. Ioakeim Ampatzoglou, Joseph K. Miller, Nataša Pavlović, Maja Tasković, On the global in time existence and uniqueness of solutions to the Boltzmann hierarchy, Journal of Functional Analysis 289 (2025) 111079, DOI: 10.1016/j.jfa.2025.111079 (preprint arXiv:2402.00765), https://doi.org/10.1016/j.jfa.2025.111079
2. Yu Deng, Zaher Hani, Xiao Ma, Long time derivation of the Boltzmann equation from hard sphere dynamics (2024), arXiv:2408.07818, https://arxiv.org/abs/2408.07818
3. Yu Deng, Zaher Hani, Xiao Ma, Hilbert's sixth problem: derivation of fluid equations via Boltzmann's kinetic theory (2025), arXiv:2503.01800, https://arxiv.org/abs/2503.01800
4. 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) (2026), arXiv:2602.04407, https://arxiv.org/abs/2602.04407
5. Oscar E. Lanford, Time evolution of large classical systems, 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
6. Isabelle Gallagher, Laure Saint-Raymond, Benjamin Texier, From Newton to Boltzmann: Hard Spheres and Short-range Potentials, Zurich Lectures in Advanced Mathematics, European Mathematical Society (2014), DOI: 10.4171/129, https://doi.org/10.4171/129
7. R. Illner, M. Pulvirenti, Global validity of the Boltzmann equation for two- and three-dimensional rare gas in vacuum: Erratum and improved result, Communications in Mathematical Physics 121 (1989) 143–146, DOI: 10.1007/BF01218628, https://doi.org/10.1007/BF01218628
8. Reinhard Illner, Marvin Shinbrot, The Boltzmann equation: Global existence for a rare gas in an infinite vacuum, Communications in Mathematical Physics 95 (1984) 217–226, DOI: 10.1007/BF01468142, https://doi.org/10.1007/BF01468142
9. Sergiu Klainerman, Matei Machedon, On the Uniqueness of Solutions to the Gross-Pitaevskii Hierarchy, Communications in Mathematical Physics 279 (2008) 169–185, DOI: 10.1007/s00220-008-0426-4, https://doi.org/10.1007/s00220-008-0426-4
10. Thierry Bodineau, Isabelle Gallagher, Laure Saint-Raymond, Sergio Simonella, Long-time derivation at equilibrium of the fluctuating Boltzmann equation (2022), arXiv:2201.04514, https://arxiv.org/abs/2201.04514
11. Ioakeim Ampatzoglou, Joseph K. Miller, Nataša Pavlović, Maja Tasković, Inhomogeneous wave kinetic equation and its hierarchy in polynomially weighted L∞ spaces (2024), arXiv:2405.03984, https://arxiv.org/abs/2405.03984
12. Shuchen Guo, From Kac particles to the Landau equation with hard potentials: BBGKY hierarchy method (2025), arXiv:2508.10697, https://arxiv.org/abs/2508.10697
13. Xuwen Chen, Shunlin Shen, Zhifei Zhang, l²-decoupling and the unconditional uniqueness for the Boltzmann equation (2025), arXiv:2501.14697, https://arxiv.org/abs/2501.14697
14. Chenguang Liu, Liping Xu, An Zhang, Quantitative propagation of chaos for the Boltzmann equation with moderately soft potentials (2026), arXiv:2607.13825, https://arxiv.org/abs/2607.13825