Non-Equilibrium Statistical Mechanics of Anharmonic Chains Coupled to Two Heat Baths at Different Temperatures

Jean‐Pierre Eckmann, Claude‐Alain Pillet, Luc Rey-BelletCommunications 1999 · DOI · arXiv

Introduction

In classical statistical mechanics, the behavior of systems in thermal equilibrium is well-defined. When a closed system is placed in a reservoir at a constant temperature, it eventually settles into the Boltzmann distribution. However, real-world physical systems are frequently "open" and driven by external gradients, such as temperature differences. Understanding these non-equilibrium steady states (NESS) is significantly more complex, particularly for non-linear systems.

This paper provides a rigorous mathematical foundation for the non-equilibrium statistical mechanics of a classic model: a finite chain of anharmonic (non-linear) oscillators coupled at its boundaries to two infinite heat baths maintained at different temperatures, $T_L$ and $T_R$. The authors address the fundamental question of whether such a system will converge to a unique, stable steady state or exhibit chaotic, non-convergent behavior.

The Mathematical Challenge

While linear systems—where forces are strictly proportional to displacement—were well-understood prior to this work, non-linear systems presented a formidable challenge. In a linear chain, the equations of motion can be solved explicitly. In an anharmonic chain, the interaction between particles is non-linear, making it difficult to prove that the system's energy will not accumulate indefinitely or that the thermal noise from the baths will effectively "mix" throughout the entire chain. The authors sought to prove the existence and uniqueness of an invariant measure—a mathematical description of the steady state—under these non-equilibrium conditions.

Methodology

1. Microscopic Derivation

The researchers began with a full Hamiltonian description of the system, including the oscillator chain and the infinite heat baths (modeled as wave equations). To make the problem tractable, they "integrated out" the infinite degrees of freedom associated with the baths. This process yielded generalized Langevin equations, which describe the motion of the chain as a result of deterministic internal forces, random thermal noise from the baths, and friction (dissipation).

2. Quasi-Markovian Approximation

Generalized Langevin equations often involve "memory effects," where the system's future state depends on its past. By selecting specific coupling functions between the chain and the baths, the authors introduced auxiliary variables to represent the state of the baths. This transformation allowed them to describe the system using a set of Markovian stochastic differential equations, where the future state depends only on the current state.

3. Spectral Analysis and Hypoellipticity

The core innovation of the paper lies in its treatment of the "generator" of the diffusion process. The authors extended Hormander’s commutator method—a technique originally developed for partial differential equations—to a non-compact phase space (where particle positions and momenta can be infinitely large). They used the concept of hypoellipticity to demonstrate that the internal coupling between oscillators is strong enough to distribute the thermal noise from the boundaries to every degree of freedom in the chain. This ensured that the system's evolution operator had a "compact resolvent," a technical requirement for proving the existence of a smooth ($C^\infty$) invariant measure.

Key Theoretical Results

Existence of the Steady State (Theorem 2.1)

The authors proved that for any temperature difference between the two baths, there exists at least one invariant measure with a smooth density. This confirms that the system will always reach some form of steady state, regardless of how steep the thermal gradient is.

Uniqueness and Mixing (Theorem 2.2)

While a steady state always exists, proving it is the only possible state is more difficult. The authors demonstrated that if the temperature difference ($T_L - T_R$) is sufficiently small, the invariant measure is unique. Furthermore, they proved the system is "mixing," meaning it eventually "forgets" its initial conditions and converges to this unique steady state over time.

Energy Bounds and Confinement

A critical physical insight provided by the proof is that the chain's temperature is bounded by the temperatures of the reservoirs; the system cannot spontaneously become hotter than the hottest bath or colder than the coldest one. This is mathematically represented by showing that the steady-state density decays rapidly in Schwartz space, provided the potential is "quadratic at infinity," ensuring the particles remain confined and do not fly off to infinity.

Limitations and Future Directions

The primary limitation of this work is its perturbative nature regarding uniqueness. The proof for a unique, mixing steady state strictly applies only when the temperature gradient is small. Whether uniqueness holds for very large temperature differences remained an open question at the time of publication. Additionally, the results rely on specific constraints on the potential (quadratic plus bounded). The authors suggested that for potentials that grow faster than quadratically, the coupling requirements might be relaxed, though this was not formally proven in this text.

Conclusion

By generalizing the commutator method to handle operators on non-compact spaces, this paper established a new standard for rigor in non-equilibrium statistical mechanics. It successfully bridged the gap between microscopic Hamiltonian dynamics and the macroscopic reality of steady-state heat conduction, providing a robust framework for studying random processes in complex physical systems.

Summary updated

BibTeX
@inproceedings{eckmann1999non,
  author = {Jean‐Pierre Eckmann and Claude‐Alain Pillet and Luc Rey-Bellet},
  title = {{Non-Equilibrium Statistical Mechanics of Anharmonic Chains Coupled to Two Heat Baths at Different Temperatures}},
  booktitle = {Communications},
  year = {1999},
  doi = {10.1007/s002200050572},
}
Figure