Gilbert and Lefevere [Phys. Rev. Lett. 101, 200601 (2008)] studied chains of locally confined hard-core particles - each particle trapped forever in its own billiard cell and exchanging energy with nearest neighbors through rare elastic collisions - and showed that under a molecular chaos (statistical independence) assumption on the collision variables, the one-particle distributions obey a Boltzmann-like equation from which the Green-Kubo heat conductivity equals the equilibrium collision frequency, κ(T)=ν(T)=T ν\kappa(T)=\nu(T)=\sqrt{T}\,\nu, a prediction confirmed numerically on the square-strings model as collisions become rare. The proof of molecular chaos itself was explicitly left open. Later literature resolved the neighboring questions only for stochastic surrogate models: for the associated Markov energy-exchange processes the conductivity is now rigorously computable, and Sasada proved that the exact identity κ=ν\kappa=\nu holds if and only if a structural gradient condition is satisfied, which these models fail; Gaspard and Gilbert (2017) computed the resulting small negative dynamical correction, κ/ν=1−0.000373\kappa/\nu=1-0.000373 at the three-dimensional conduction threshold. For the deterministic billiard dynamics the central questions remain open: (a) prove or refute asymptotic independence of the variables of successive interparticle collisions in the rare-collision, large-NN limit, with correlation-decay bounds and finite-size corrections; (b) identify the minimal dynamical hypotheses on the single-cell motion (ergodicity versus stronger mixing, and the role of the cell geometry); (c) determine the rare-collision limit of κ/ν\kappa/\nu, in particular whether it approaches unity. A 2020 expert review confirms that the required correlation control in these high-dimensional semi-dispersing billiards 'defies a rigorous approach' at present.

Background

Fourier's law of heat conduction states that the heat current in a material at rest is proportional to the temperature gradient, j=−κ(T)∇Tj=-\kappa(T)\nabla T, with a material-dependent conductivity κ(T)\kappa(T). Deriving Fourier's law and computing κ\kappa from the deterministic Hamiltonian dynamics of interacting particles is a classical open problem of nonequilibrium statistical mechanics; most established approaches concern weakly anharmonic lattices, where a phonon kinetic equation (the Peierls-Boltzmann equation) applies. Gilbert and Lefevere (2008) studied the opposite, extremely anharmonic regime of locally confined systems.

Locally confined hard-core systems. NN unit-mass particles sit on a one-dimensional lattice with positions and momenta (qi,pi)1≤i≤N(q_i,p_i)_{1\le i\le N}, qi,pi∈Rdq_i,p_i\in\mathbb{R}^d, evolving under the Hamiltonian H(p,q)=∑i=1Npi2/2+V(qi)+U(qi−qi+1)H(p,q)=\sum_{i=1}^{N}p_i^2/2+V(q_i)+U(q_i-q_{i+1}). The pinning potential VV vanishes inside a bounded cell ΩV\Omega_V and is infinite outside, so each particle remains trapped in its own cell forever: mass transport is blocked and only energy is transported. The nearest-neighbor potential UU vanishes inside a region ΩU\Omega_U and is infinite outside, so that when the separation qi−qi+1q_i-q_{i+1} reaches the boundary Λ=∂ΩU\Lambda=\partial\Omega_U the two neighbors undergo an elastic hard-core collision, exchanging the components of their velocities normal to Λ\Lambda. Such a system is equivalent to a billiard - free flight punctuated by elastic reflections - in a high-dimensional configuration space. Physically these arrays model aerogels, materials in which gas particles are trapped in nano-size pores and rarely meet. The original ergodicity result for a class of such systems is due to Bunimovich, Liverani, Pellegrinotti and Suhov (1992).

A canonical example: the square-strings model. Point particles move freely in unit square cells [−1/2,1/2]2[-1/2,1/2]^2 and interact when their mutual distance equals a parameter aa with 0<a<50<a<\sqrt{5}; at interaction they exchange the velocity components along their relative motion. The motion of a single particle in its cell is pseudo-integrable: it is ergodic (time averages equal phase-space averages for almost every initial condition) for almost every velocity direction, but not mixing (correlations of observables need not decay). Dispersing cell geometries (convex scatterers that defocus incoming trajectories) instead produce strongly chaotic, mixing single-cell dynamics; the billiard arrays of Gaspard and Gilbert use such cells.

Molecular chaos and the Boltzmann-like equation. In the regime of rare interactions the mean time τb\tau_b between collisions of neighboring particles is much larger than the mean time τw\tau_w between collisions of a particle with the walls of its own cell (in the square-strings model τb/τw→∞\tau_b/\tau_w\to\infty as a→5a\to\sqrt{5}; in dispersing arrays the same separation is produced by letting the ball radius approach a critical geometric value, the conduction threshold, at which the array turns from conductor to insulator). Gilbert and Lefevere assumed molecular chaos - the Stosszahlansatz that the dynamical variables (collision times, exchanged energies) of successive collisions of a neighboring pair are statistically independent, so that the two-particle distribution entering a collision factorizes into one-particle marginals - and derived from the Liouville equation a Boltzmann-like equation for the one-particle distributions fi(p,q,t)f_i(p,q,t). A Chapman-Enskog expansion (a systematic expansion of the Boltzmann-like equation in the small ratio of the microscopic collision timescale to the macroscopic timescale) around local equilibrium together with a Green-Kubo computation (the Green-Kubo formula expresses transport coefficients through equilibrium autocorrelations of the conserved current) then yields the central prediction κ(T)=ν(T)=T ν\kappa(T)=\nu(T)=\sqrt{T}\,\nu, where ν(T)\nu(T) is the equilibrium mean frequency of nearest-neighbor collisions - a static average computable to arbitrary precision. Numerics on the square-strings model coupled to stochastic thermal baths confirm the prediction as interactions become rare: κ/ν=1.0037\kappa/\nu=1.0037 at 5−a=0.155\sqrt{5}-a=0.155, growing to 6.0006.000 at 5−a=1.029\sqrt{5}-a=1.029.

What is rigorous today. The source paper proves none of the chaos assumption: it states that how the required conditions are realized in these models 'is not yet fully elucidated' and that the only dynamical property 'a priori necessary' in the derivation is ergodicity of the single-cell dynamics. The follow-up literature developed a two-step program (Gaspard-Gilbert): first reduce the rare-interaction deterministic dynamics to a stochastic energy-exchange model - a Markov jump process on the lattice of local energies - then derive Fourier's law for that process. Step two is essentially rigorous today: exponential mixing and spectral gaps for the stochastic models (Grigo-Khanin-Szasz; Sasada), hydrodynamic limits and Fourier's law (Yao Li and collaborators), macroscopic fluctuation theory (Lefevere-Mariani-Zambotti). Step one, the deterministic molecular chaos justification, is heuristic in every treatment. Moreover, within the stochastic models Sasada proved that the exact equality between κ\kappa and the static collision-frequency contribution holds if and only if the model satisfies the gradient condition (a structural condition on stochastic lattice gases guaranteeing that the conserved current decomposes as a gradient plus a dissipative part), which these models fail; Gaspard-Gilbert computed the resulting small negative dynamical correction, κ/ν=1−3.7×10−4\kappa/\nu=1-3.7\times10^{-4} at the three-dimensional conduction threshold. The identity κ(T)=ν(T)\kappa(T)=\nu(T) is therefore an asymptotic, leading-order statement in the rare-collision limit, and its deterministic status is unknown.

Problem Statement

For the class of deterministic locally confined hard-core Hamiltonian systems H(p,q)=∑i=1Npi2/2+V(qi)+U(qi−qi+1)H(p,q)=\sum_{i=1}^{N}p_i^2/2+V(q_i)+U(q_i-q_{i+1}) with hard-wall pinning and hard-core nearest-neighbor coupling - including the square-strings model of Gilbert and Lefevere and the dispersing billiard arrays of Gaspard and Gilbert - establish, or refute, the molecular chaos assumption underpinning the Boltzmann-like equation and the conductivity identity of Gilbert and Lefevere (2008), and delineate the regime in which they hold. Concretely:

(a) Molecular chaos / propagation of chaos. Prove, or disprove, that in the rare-collision regime (τw/τb→0\tau_w/\tau_b\to0, e.g. a→5a\to\sqrt{5} in the square-strings model, or ball radius approaching the critical radius at the conduction threshold in dispersing arrays) together with the thermodynamic limit N→∞N\to\infty, the dynamical variables governing successive nearest-neighbor collisions become asymptotically independent, so that the one-particle distributions satisfy the Boltzmann-like equation of the source to controlled error. The required quantitative content includes bounds on the decay of correlations between successive interparticle collisions as functions of the timescale ratio τb/τw\tau_b/\tau_w, and estimates of the finite-size corrections (Gilbert and Lefevere note that the rest of the chain must act as an energy reservoir for a colliding pair).

(b) Minimal dynamical hypotheses. Determine the weakest dynamical assumptions on the single-cell billiard under which (a) holds: is ergodicity of the cell dynamics sufficient in all geometries, as Gilbert and Lefevere claim is the only property 'a priori necessary', or are stronger mixing properties required, and how does the answer depend on the cell geometry (for instance, configurations in which a slow particle can 'hide' from its neighbors, where mixing of the chain is provably no faster than t−2t^{-2}, versus configurations in which adjacent fast particles can always reach it)?

(c) Conductivity identity. Determine, for the deterministic dynamics, the rare-collision limit of the Green-Kubo heat conductivity ratio κ(T)/ν(T)\kappa(T)/\nu(T) - in particular whether the leading-order identity κ(T)≈ν(T)\kappa(T)\approx\nu(T) of the source holds asymptotically, i.e. whether κ(T)/ν(T)→1\kappa(T)/\nu(T)\to1 - and identify what fixes the limiting value. At the level of the stochastic energy-exchange surrogates the exact identity is known to fail: it holds if and only if the gradient condition holds, which these models fail, with a computed correction of −3.7×10−4-3.7\times10^{-4}; the corresponding deterministic statements are open.

These components jointly support one objective: a first-principles, molecular-chaos-based derivation of Fourier's law with the conductivity formula κ≈ν\kappa\approx\nu for deterministic locally confined systems. Any answer must state explicitly whether it concerns the deterministic dynamics or a stochastic surrogate, since a result about the surrogate does not solve the deterministic problem. Completion is recognizable by an expert: for at least one canonical geometry of the stated class - the square-strings model or a dispersing billiard array - a full solution resolves (a) by proof or by rigorous refutation, resolves (c), and, when (a) is answered affirmatively, resolves (b); when (a) is refuted instead, (b) becomes the identification of the obstruction to asymptotic independence and of the dynamical hypotheses under which it does or does not occur. In every case the dynamical hypotheses, the scaling limit and the error bounds must be stated explicitly, and extending the result from one canonical geometry to the whole stated class is desirable but not required for completion.

References

  1. Thomas Gilbert and Raphaël Lefevere, 'Heat Conductivity from Molecular Chaos Hypothesis in Locally Confined Billiard Systems', Physical Review Letters 101, 200601 (2008). DOI: 10.1103/PhysRevLett.101.200601; arXiv:0808.1179; https://arxiv.org/abs/0808.1179
  2. Pierre Gaspard and Thomas Gilbert, 'Heat Conduction and Fourier's Law by Consecutive Local Mixing and Thermalization', Physical Review Letters 101, 020601 (2008). DOI: 10.1103/PhysRevLett.101.020601; https://doi.org/10.1103/PhysRevLett.101.020601
  3. Pierre Gaspard and Thomas Gilbert, 'Heat conduction and Fourier's law in a class of many particle dispersing billiards', New Journal of Physics 10, 103004 (2008). DOI: 10.1088/1367-2630/10/10/103004; https://doi.org/10.1088/1367-2630/10/10/103004
  4. Pierre Gaspard and Thomas Gilbert, 'On the derivation of Fourier's law in stochastic energy exchange systems', Journal of Statistical Mechanics: Theory and Experiment (2008) P11021. DOI: 10.1088/1742-5468/2008/11/P11021; arXiv:0809.3967; https://arxiv.org/abs/0809.3967
  5. Pierre Gaspard and Thomas Gilbert, 'Heat transport in stochastic energy exchange models of locally confined hard spheres', Journal of Statistical Mechanics: Theory and Experiment (2009) P08020. DOI: 10.1088/1742-5468/2009/08/P08020; arXiv:0905.1051; https://arxiv.org/abs/0905.1051
  6. Thomas Gilbert, 'On the waiting time distributions of systems of locally confined particles with rare interactions', Journal of Statistical Mechanics: Theory and Experiment (2011) P06015. DOI: 10.1088/1742-5468/2011/06/P06015; https://doi.org/10.1088/1742-5468/2011/06/P06015
  7. Pierre Gaspard and Thomas Gilbert, 'A two-stage approach to relaxation in billiard systems of locally confined hard spheres', Chaos 22, 026117 (2012). DOI: 10.1063/1.3697689; arXiv:1111.6272; https://arxiv.org/abs/1111.6272
  8. Pierre Gaspard and Thomas Gilbert, 'Dynamical contribution to the heat conductivity in stochastic energy exchanges of locally confined gases', Journal of Statistical Mechanics: Theory and Experiment (2017) 043210. DOI: 10.1088/1742-5468/aa60ce; arXiv:1611.05809; https://arxiv.org/abs/1611.05809
  9. Makiko Sasada, 'Spectral gap for stochastic energy exchange model with nonuniformly positive rate function', The Annals of Probability 43, 1663-1711 (2015). DOI: 10.1214/14-AOP916; https://doi.org/10.1214/14-AOP916
  10. Makiko Sasada, 'Thermal conductivity for stochastic energy exchange models', arXiv:1611.08866 (2016). https://arxiv.org/abs/1611.08866
  11. Makiko Sasada, 'On the Green-Kubo formula and the gradient condition on currents', The Annals of Applied Probability 28, no. 5 (2018). DOI: 10.1214/17-AAP1369; https://doi.org/10.1214/17-AAP1369
  12. Péter Bálint, Thomas Gilbert, Péter Nándori, Domokos Szász and Imre Péter Tóth, 'On the limiting Markov process of energy exchanges in a rarely interacting ball-piston gas', Journal of Statistical Physics 166, 903-925 (2017). DOI: 10.1007/s10955-016-1598-5; arXiv:1510.06408; https://arxiv.org/abs/1510.06408
  13. Péter Bálint, Thomas Gilbert, Domokos Szász and Imre Péter Tóth, 'What mathematical billiards teach us about statistical physics?', arXiv:2009.06284 (2020). https://arxiv.org/abs/2009.06284
  14. Péter Bálint, Péter Nándori, Domokos Szász and Imre Péter Tóth, 'Equidistribution for Standard Pairs in Planar Dispersing Billiard Flows', Annales Henri Poincaré 19, 979-1042 (2018). DOI: 10.1007/s00023-018-0648-8; https://doi.org/10.1007/s00023-018-0648-8
  15. Zsolt Pajor-Gyulai and Domokos Szász, 'Energy Transfer and Joint Diffusion', Journal of Statistical Physics 146, 1001-1025 (2012). DOI: 10.1007/s10955-012-0426-9; https://doi.org/10.1007/s10955-012-0426-9
  16. Yao Li, 'On the stochastic behaviors of locally confined particle systems', Chaos 25, 073121 (2015). DOI: 10.1063/1.4927300; https://doi.org/10.1063/1.4927300
  17. Yao Li and Lai-Sang Young, 'Polynomial convergence to equilibrium for a system of interacting particles', The Annals of Applied Probability 27, 65-90 (2017). DOI: 10.1214/16-AAP1197; arXiv:1601.00717; https://arxiv.org/abs/1601.00717
  18. Yao Li, 'On the polynomial convergence rate to nonequilibrium steady states', The Annals of Applied Probability 28, no. 6 (2018). DOI: 10.1214/18-AAP1406; arXiv:1607.08492; https://arxiv.org/abs/1607.08492
  19. Yao Li and Lingchen Bu, 'From billiards to thermodynamic laws: Stochastic energy exchange model', Chaos 28, 093105 (2018). DOI: 10.1063/1.5040296; arXiv:1805.02133; https://arxiv.org/abs/1805.02133
  20. Yao Li and Wenbo Xie, 'Thermal conductivity and local thermodynamic equilibrium of stochastic energy exchange models', Journal of Statistical Mechanics: Theory and Experiment 2019, 043205 (2019). DOI: 10.1088/1742-5468/ab0c16; arXiv:1808.07953; https://arxiv.org/abs/1808.07953
  21. Yao Li, 'From deterministic dynamics to thermodynamic laws II: Fourier's law and mesoscopic limit equation', arXiv:1908.06219 (2019). https://arxiv.org/abs/1908.06219
  22. François Huveneers, 'Energy transport through rare collisions', Journal of Statistical Physics 146, 73-97 (2012). DOI: 10.1007/s10955-011-0374-9; arXiv:1106.5991; https://arxiv.org/abs/1106.5991
  23. Alexander Grigo, Konstantin Khanin and Domokos Szász, 'Mixing rates of particle systems with energy exchange', Nonlinearity 25, 2349-2376 (2012). DOI: 10.1088/0951-7715/25/8/2349; https://doi.org/10.1088/0951-7715/25/8/2349
  24. Leonid Bunimovich, Carlangelo Liverani, Alessandro Pellegrinotti and Yurii Suhov, 'Ergodic systems of n balls in a billiard table', Communications in Mathematical Physics 146, 357-396 (1992). DOI: 10.1007/BF02102633; https://doi.org/10.1007/BF02102633
  25. Wael Bahsoun and Fanni M. Sélley, 'Map Lattices Coupled by Collisions: Hitting Time Statistics and Collisions Per Lattice Unit', Annales Henri Poincaré 23, 2919-2947 (2022). DOI: 10.1007/s00023-022-01164-2; https://doi.org/10.1007/s00023-022-01164-2
  26. Wael Bahsoun and Maxence Phalempin, 'Rare Events Statistics for Zd\mathbb{Z}^d Map Lattices Coupled by Collision', Communications in Mathematical Physics 407, no. 3, article 57 (2026). DOI: 10.1007/s00220-026-05557-w; arXiv:2412.12803; https://arxiv.org/abs/2412.12803
  27. Krzysztof Burdzy, Jeremy G. Hoskins and Stefan Steinerberger, 'From Pinned Billiard Balls to Partial Differential Equations', Journal of Theoretical Probability 39, no. 1, article 16 (2026). DOI: 10.1007/s10959-025-01472-3; arXiv:2209.01503; https://arxiv.org/abs/2209.01503
  28. Raphaël Lefevere and Lorenzo Zambotti, 'Hot Scatterers and Tracers for the Transfer of Heat in Collisional Dynamics', Journal of Statistical Physics 139, 686-713 (2010). DOI: 10.1007/s10955-010-9962-3; https://doi.org/10.1007/s10955-010-9962-3
  29. Raphaël Lefevere, Mauro Mariani and Lorenzo Zambotti, 'Macroscopic fluctuation theory of aerogel dynamics', Journal of Statistical Mechanics: Theory and Experiment (2010) L12004. DOI: 10.1088/1742-5468/2010/12/L12004; https://doi.org/10.1088/1742-5468/2010/12/L12004
  30. Nikolai Chernov and Dmitry Dolgopyat, 'Brownian Brownian Motion I', Memoirs of the American Mathematical Society 198, no. 927 (2009). DOI: 10.1090/memo/0927; https://doi.org/10.1090/memo/0927
  31. Viviane Baladi, Mark F. Demers and Carlangelo Liverani, 'Exponential decay of correlations for finite horizon Sinai billiard flows', Inventiones mathematicae 211, 39-177 (2018). DOI: 10.1007/s00222-017-0745-1; https://doi.org/10.1007/s00222-017-0745-1
  32. Gerhard Keller and Carlangelo Liverani, 'Map Lattices Coupled by Collisions', Communications in Mathematical Physics 291, 591-597 (2009). DOI: 10.1007/s00220-009-0835-z; https://doi.org/10.1007/s00220-009-0835-z