Non-Equilibrium Thermodynamics

• Non-equilibrium thermodynamics is a comprehensive framework for describing irreversible, driven processes in systems far from thermodynamic equilibrium.
• It integrates classical, stochastic, and geometric methodologies to quantify entropy production, transport phenomena, and relaxation processes across scales.
• Applications extend from modeling self-replicating biological systems to exploring quantum statistical mechanics and covariant general relativistic formulations.

Non-equilibrium thermodynamics is the systematic theoretical framework for describing irreversible processes in systems away from thermodynamic equilibrium, encompassing macroscopic, mesoscopic, and stochastic behaviors, and generalizing the equilibrium canonical and microcanonical ensembles to evolving, driven states. It incorporates entropy production, transport, and relaxation phenomena, and bridges classical, statistical, dynamical-systems, and information-theoretic perspectives across scales. Modern research integrates geometric, variational, and operator methods, unifying linear near-equilibrium and strongly nonlinear far-from-equilibrium regimes, and linking to fundamental structures in statistical mechanics, field theory, and quantum statistical physics.

1. Foundations: Classical and Statistical Structures

The traditional foundation of non-equilibrium thermodynamics is Onsager’s theory of near-equilibrium processes based on the local-equilibrium hypothesis. This posits that even in macroscopic non-equilibrium, each infinitesimal element is instantaneously in equilibrium, so that local thermodynamic variables (internal energy $u(\mathbf{r},t)$, specific volume $v(\mathbf{r},t)$, number density $n(\mathbf{r},t)$) and entropy density $s(u,v,n)$ are defined as in the equilibrium theory. Macroscopically, the entropy balance is written as $dS = d_eS + d_iS$ with non-negative entropy production rate density $$\sigma(\mathbf{r},t) = \mathbf{J}\cdot\mathbf{X} \ge 0$$, where $\mathbf{J}$ are fluxes and $\mathbf{X}$ the thermodynamic forces (gradients of temperature, chemical potential, etc) (Ben-Naim, 2018).

Linear phenomenological laws relate fluxes and forces, encoding reciprocal relations (Onsager/Casimir): $J_i = \sum_j L_{ij} X_j$, with $L_{ij}$ the kinetic coefficients. The entropy production integrates as $v(\mathbf{r},t)$0. However, Ben-Naim has argued that this entire construction—and the very notion of local equilibrium entropy density—is generally ill-founded for strongly interacting, condensed, or few-body systems, where mutual information and correlations are essential and cannot be partitioned as $v(\mathbf{r},t)$1 (Ben-Naim, 2018).

Statistically, nonequilibrium extensions of the microcanonical ensemble lead to ensembles over evolving probability densities $v(\mathbf{r},t)$2 for a set of microsopic variables $v(\mathbf{r},t)$3 (Leadbetter et al., 2023). For classical stochastic systems, the Fokker–Planck or master equation replaces Liouville's theorem; steady nonequilibrium states are typically characterized by the Sinai–Ruelle–Bowen (SRB) measure, which is singular and encodes entropy production as phase-space contraction (Gallavotti, 2019).

2. Modern Dynamical and Geometric Approaches

Contemporary developments include the dynamical-systems, variational, and geometric perspectives:

• The dynamical-systems formalism (Gallavotti–Cohen) interprets stationary nonequilibrium states (NESS) as statistical attractors supporting unique SRB measures. Entropy production is rigorously equated with mean phase-space contraction $v(\mathbf{r},t)$4, and the Gallavotti–Cohen fluctuation theorem quantifies the symmetry of entropy production fluctuations in time-reversible chaotic systems (Gallavotti, 2019).
• Geometric representations map thermodynamic variables to coordinates on a manifold, kinetic coefficients to a metric tensor, and driving forces to vector fields. The Onsager–Machlup Lagrangian, $v(\mathbf{r},t)$5, underpins both fluctuation theory and extended (gravity-analog) models (Aibara et al., 2018). Entropy production along cycles splits into topological (time-odd), curvature-induced (time-even, rapid operation), and dissipative (quasi-equilibrium) parts.
• Gauge-theoretic formulations interpret thermodynamic forces as gauge fields. Linear Onsager relations arise from a pure-gauge (potential) structure, while Maxwell-like kinetic terms in the entropy lead to nonlinear constitutive equations, enabling nonlinear irreversible dynamics (Katagiri, 2018).
• Symplecto-contact reduction establishes a canonical geometric backbone: statistical phase-space is reduced via moment maps to mesoscopic manifolds, with the relative information entropy (Kullback–Leibler) generating Legendrian submanifolds in contact phase-space. The classical Maxwell construction for phase transitions appears as a graph-selection problem in contact geometry (Lim et al., 2022).

3. Far-from-Equilibrium Extensions and Internal Variables

Classical theory is fundamentally limited to near-equilibrium, locally additive, and correlation-free systems. Far-from-equilibrium frameworks address these limits:

• Stochastic Thermodynamics and Internal Variables (STIV): A modern formalism postulates an approximate parametric density $v(\mathbf{r},t)$6 over relevant collective variables $v(\mathbf{r},t)$7 (e.g., reaction coordinates, order parameters), with $v(\mathbf{r},t)$8 internal variables. The macroscopic (nonequilibrium) free energy $v(\mathbf{r},t)$9 constructed from $n(\mathbf{r},t)$0 and the underlying potential $n(\mathbf{r},t)$1 serves as a Lyapunov functional; its gradient flow with respect to a suitable metric yields kinetic equations for $n(\mathbf{r},t)$2 and identifies entropy production (Leadbetter et al., 2023).
• Nambu Non-equilibrium Thermodynamics (NNET): Axiomatizes the full decomposition of dynamics into reversible (multi-Hamiltonian, Nambu-bracket) and irreversible (gradient/entropy) components. For $n(\mathbf{r},t)$3 variables, reversible evolution is generated by distinct $n(\mathbf{r},t)$4 Hamiltonians, enabling simultaneous conservation of multiple invariants even in the presence of cyclic/oscillatory flows, while irreversible drift is governed by symmetric tensors ensuring non-negative entropy production. The formalism supports transient entropy decreases and high-order nonlinearities, unifying and extending Onsager, GENERIC, and Prigoginian approaches (Katagiri et al., 31 Jul 2025, Katagiri et al., 26 Aug 2025).
• Non-equilibrium Evolution Thermodynamics (NEET): Asserts the internal energy as the fundamental thermodynamic potential, incorporating nonequilibrium entropy and internal variables (e.g., defect densities in solid-state systems). Kinetic equations are constructed as relaxation flows towards extremal values of internal energy or its Legendre transforms, with generalized forces empirically or statistically determined (Metlov, 2011).

4. Covariant, Quantum, and Relativistic Formulations

Advancements in fluctuating, quantum, and relativistic systems include:

• Stochastic Thermodynamics with Covariance: For small systems under multiplicative noise, the covariant Ito–Langevin formalism produces coordinate-invariant SDEs with unique entropy-production definitions. Path-probability ratios yield exact fluctuation theorems, Jarzynski/Crooks relations, and ensure the ensemble- and trajectory-level second law is satisfied, even in the presence of nontrivial manifold geometry (Ding et al., 2021).
• Non-equilibrium Quantum Thermodynamics and Open Systems: Thermodynamic consistency for quantum Brownian motion shows that commonly used completely positive master equations (Gaussian Lindblad form) generically break detailed balance and induce persistent phase-space currents with non-zero steady-state entropy production, in contrast to the Caldeira–Leggett model which upholds equilibrium microreversibility but sacrifices complete positivity. This highlights fundamental tensions between quantum consistency and thermodynamic equilibration (Artini et al., 31 Jul 2025).
• Covariant General Relativistic Thermodynamics: Entropic forces are introduced into the Einstein–Hilbert action via gradients of entropy density, leading to modified Einstein equations with explicit non-conservation of the energy–momentum tensor attributed to irreversibility, and extra source terms in the Raychaudhuri and Friedmann equations. Bulk entropy gradients drive acceleration (cosmic or otherwise), and in gravitational collapse, sufficiently large entropic forces can avert singularities despite the strong energy condition (Espinosa-Portales et al., 2021).

5. Limitations of the Local-Equilibrium and Additivity Assumptions

Critical analyses reveal foundational limitations of the local-equilibrium and additivity assumptions:

• The existence of a local entropy density $n(\mathbf{r},t)$5 functionally equivalent to equilibrium $n(\mathbf{r},t)$6 is unjustified in systems with few particles per cell, strong interactions, or long-range correlations. Standard forms (e.g., Sackur–Tetrode) fail when particle number per cell is small or Stirling’s approximation is invalid. Intermolecular correlations contribute global (mutual information) entropy not recoverable by integration over local densities, thereby undermining the definition and calculation of entropy production in such systems (Ben-Naim, 2018).
• Inhomogeneous systems are rigorously treated by exact Gibbsian entropies, but only when subsystems are quasi-independent—i.e., the correlation length is much less than subsystem size—so that the additivity $n(\mathbf{r},t)$7 holds. Breakdown of quasi-independence leads to nonadditive correction terms. Thermodynamic potentials for subsystems are determined by the (constant) field parameters of the medium, even if the subsystem is far from equilibrium (Gujrati, 2011).

6. State Functions, Thermodynamic Potentials, and Work Relations Far from Equilibrium

Recent frameworks have clarified the construction of generalized thermodynamic potentials and state functions:

• Non-equilibrium Massieu–Planck and Helmholtz–Gibbs potentials, defined along arbitrary non-equilibrium, time-resolved processes, incorporate explicit entropy production rates and are constructed via minimum-work principles extending the equilibrium second law. These potentials serve as true state functions and enable non-equilibrium work diagrams that are quantum-consistent and do not require initial factorizations, enabling treatment of strong system-bath entanglement in quantum thermodynamic cycles (Koyanagi et al., 2024).
• In stationary non-equilibrium states supporting persistent macroscopic flows (e.g., Rayleigh–Bénard convection), global energy balances—rather than local field equations—are emphasized. The first and second laws are written in terms of global potentials (internal, kinetic, potential energies, and a new inertial function $n(\mathbf{r},t)$8), and spontaneous transitions between stationary states are governed by a monotonic decrease of $n(\mathbf{r},t)$9, supplanting entropy maximization as the Lyapunov function for pattern selection and stability (Hołyst et al., 6 Oct 2025).

7. Applications and Extensions: Biological, Self-Replicating, and Field Theories

Non-equilibrium thermodynamics underpins the analysis of complex systems well outside equilibrium:

• Minimal physical models of self-replicating protocells are framed as non-equilibrium thermodynamic cycles: system state variables are coupled to metabolic chemistry and droplet formation, and explicit entropy production is computed along the replication–division cycle. This elucidates fundamental constraints on origin-of-life scenarios, biological growth, and reproduction as fundamentally dissipative processes (Fellermann et al., 2015).
• Information-theoretic and geometric approaches using symplectic/contact/Legendrian structures unify equilibrium and nonequilibrium sectors, and clarify the geometric origin of phase transitions, graph selectors, and Maxwell constructions (Lim et al., 2022).
• The field-theoretic and gauge-theoretic perspectives suggest deep connections between non-equilibrium thermodynamics, quantum field theory, and gravity analog models, both at the macroscopic (thermodynamic geometry, effective actions) and microscopic (path integrals, action–entropy correspondences) levels (Munkhammar, 2015, Aibara et al., 2018).

---

Non-equilibrium thermodynamics is thus a multi-scale, multi-framework discipline employing methods from classical irreversible thermodynamics, modern stochastic processes, symplectic and gauge geometry, quantum statistical mechanics, and information theory. Its evolution continues to be shaped by foundational debates on locality, additivity, and the physical meaning of entropy and state functions in far-from-equilibrium regimes (Ben-Naim, 2018, Leadbetter et al., 2023, Katagiri et al., 31 Jul 2025, Aibara et al., 2018, Hołyst et al., 6 Oct 2025, Koyanagi et al., 2024).