Non-Equilibrium Deterministic Dynamics

Updated 12 March 2026

Non-equilibrium deterministic dynamics is the study of systems evolving via reversible microscopic rules that yield irreversible, autonomous macroscopic behavior upon coarse-graining.
It employs methods such as rate equations, large-deviation theory, and deterministic PDE closures to predict relaxation, fluctuations, and phase transitions.
Applications range from quantum spin systems and deterministic billiard models to data-driven hydrodynamic closures, demonstrating feedback-induced stationary currents and negative mobility.

Non-equilibrium deterministic dynamics refers to the study of dynamical systems in which the time evolution is governed by deterministic, often reversible, equations of motion, but the macroscopic observables of interest are far from thermodynamic equilibrium and exhibit irreversible relaxation, sustained currents, or complex pattern formation. This area encompasses models ranging from driven Hamiltonian and quantum systems through interacting classical billiards with feedback rules, to data-driven continuum closures learned from stochastic non-equilibrium trajectories. Central themes include the emergence of autonomous, Markovian or macroscopic deterministic behavior from microscopically reversible dynamics, the role of coarse-graining and large system size, and the mathematical structure underlying relaxation, fluctuations, and phase transitions.

1. Deterministic Relaxation and Macroscopic Autonomy

Deterministic non-equilibrium dynamics often exhibits macroscopic observables undergoing Markovian-type evolution toward stationary or steady states, independently of many details of the initial conditions. A paradigmatic example is provided by finite, closed quantum spin-½ ladders with two coupled legs. Considering the magnetization difference $\Delta M$ between subsystems as a coarse observable, the exact unitary Schrödinger evolution of the closed system gives rise to deterministic, exponential relaxation of $\langle\Delta M(t)\rangle$ and its variance, following an autonomous rate equation $\frac{d}{dt}\delta X = -\Gamma \delta X$ , with the same relaxation timescale governing both mean and fluctuations. This deterministic relaxation persists even though the underlying evolution remains reversible, and the fluctuations become relatively negligible in the thermodynamic limit ( $\sigma/\langle\Delta M\rangle \to 0$ as $N\to\infty$ ), yielding macroscopic determinism. The full time-evolving probability $P_X(t)$ is also well reproduced by effective Markov chain models with detailed-balance rates, and the system's stationary state is the infinite-temperature distribution—demonstrating a quantum realization of the "autonomy" and "memory-loss" that typify non-equilibrium thermodynamics (Niemeyer et al., 2014).

2. Structural Properties: Markovianity, Typicality, and Emergence

The autonomy and deterministic trend in such dynamics can be understood in light of large-deviation theory and the concentration of measure for extensive systems. In open Markov-jump systems, the ensemble-averaged entropy production $\Sigma$ and the change in system self-information $\mathcal I(x) = -\ln P_{\rm ss}(x)$ along a deterministic macroscopic trajectory $x_t$ are related by a tighter emergent "second law" $\Sigma + k_b[\mathcal I(x_t) - \mathcal I(x_0)] \geq 0$ . This bound is saturated near equilibrium and generalizes the usual entropy production inequality to non-equilibrium steady states, linking deterministic relaxation to the steady-state fluctuation spectrum (Freitas et al., 2021).

In the thermodynamic limit, the dynamics concentrates on a deterministic path in state space, governed by a rate equation whose Lyapunov function is given by the steady-state rate function $I_{\rm ss}(x)$ . Stochastic simulation (e.g., Gillespie) becomes unnecessary, as non-equilibrium distributions can be approached by integrating deterministic ODEs for the macroscopic drift, with entropy production rates providing deterministic estimates on the probability of rare events (Freitas et al., 2021).

3. Phase Transitions, Irreversibility, and Uphill Currents

Purely deterministic, reversible, and energy-conserving models can nonetheless exhibit genuine non-equilibrium phase transitions and sustained currents when equipped with non-local, feedback-type rules. In deterministic billiard models consisting of two cavities connected by channels with a threshold-activated bounce-back rule (inverting velocity if the number of forward movers exceeds $T$ ), the system displays sharp symmetry-breaking transitions as $T/N$ is varied. For intermediate thresholds and appropriate geometric parameters, the stationary state develops a persistent inhomogeneity (order parameter $\chi > 0$ ) and a loop current, with one channel carrying "uphill" current—directed from the region of lower to higher particle density. This scenario models a deterministic "battery" with negative absolute mobility, despite strict time-reversal invariance and phase-space volume preservation at the microscopic level (Cirillo et al., 2020, Cirillo et al., 2021). The transitions and steady states are quantitatively predicted by combining kinetic theory with large- $N$ stochastic (urn-model) analogs, confirming that feedback-mediated deterministic dynamics can sustain macroscopic irreversibility and fluxes.

4. Detailed Balance, Fluctuation Relations, and Transport

Irreversible deterministic maps, even when non-invertible in parts of phase space (as in certain generalized baker maps), may still produce equilibrium-like properties and standard fluctuation theorems when observables are projected to relevant subspaces. Detailed balance, linear-response (Green–Kubo) relations, and equilibrium fluctuation relations can survive microscopic irreversibility as long as the irreversibility acts on "irrelevant" degrees of freedom, and the projected dynamics for the observables of interest satisfies the necessary stochastic symmetry (Colangeli et al., 2011). Transport coefficients, entropy production, and macroscopic currents therefore emerge as robust properties of the effective dynamics, insensitive to certain violations of micro-reversibility, provided the coarse variables obey detailed-balance-type relations.

5. Fractal Steady States, Lyapunov Spectra, and Irreversibility

Time-reversible deterministic thermostats—Nosé–Hoover, isokinetic, or cubic—when subject to external driving or constraints, generate irreversible macroscopic behavior, including relaxation to fractal invariant measures with zero Lebesgue volume. The phase-space contraction rate $\nabla\cdot\dot{\bf x}$ becomes negative away from equilibrium, concentrating the invariant measure on a fractal (strange attractor) whose dimension and Lyapunov exponents can be computed and related to entropy production rates. In models such as the field-driven Galton board or thermostated oscillators, positive Lyapunov exponents quantify chaotic mixing, while the sum of exponents (Kolmogorov-Sinai entropy) reflects the dynamical entropy. At steady state, these systems display non-trivial relations between transport, Lyapunov spectra, and the macroscopic irreversibility imposed by deterministic constraints (Hoover et al., 2015).

6. Macroscopic Hydrodynamics and Data-driven Deterministic Closures

Deterministic macroscopic equations can be systematically inferred from high-dimensional, stochastic, non-equilibrium data by leveraging physical priors and sparse regression techniques. Starting from particle trajectories in active matter or collective systems, coarse-graining and adaptive kernel-density estimation yield smoothed fields (density, polarization). Candidate libraries of nonlinear terms, reflecting symmetry and conservation properties, are used to fit (via group-sparse regression and stability selection) closed deterministic PDEs for the macroscopic fields. Learned equations accurately recover key transport mechanisms—advective, diffusive, pressure, Landau-type non-linearities—and encode the typical deterministic hydrodynamics underlying the non-equilibrium system. The validity of these deterministic closures is confirmed by recovering band formation, stripe propagation, and stability boundaries directly from the data, supporting the emergence of macroscopic deterministic flows from stochastic microscopics (Maddu et al., 2022).

7. Non-normality, Curl Flux, and Non-equilibrium Potential Structure

Many non-equilibrium deterministic systems, especially in neural dynamics and active matter, exhibit non-normal linearizations at fixed points, resulting in transient amplification (reactivity) even for linearly stable systems. The drift can be Helmholtz-decomposed into a potential (gradient) part and a non-conservative curl flux. When the curl flux is nonzero, detailed balance is broken: the steady-state has nonzero circulating probability currents and the system cannot be described by a scalar potential alone. This structural insight, rigorously formulated in Wilson–Cowan-type models and linear–nonlinear reactivity frameworks, underpins deterministic non-equilibrium features such as avalanche statistics and noise-induced transitions (Santo et al., 2018).

Non-equilibrium deterministic dynamics therefore provides a unifying mathematical and conceptual framework for understanding how irreversible macroscopic relaxation, stationary currents, and complex phase behavior arise from underlying deterministic, and often reversible, microscopic laws. The field synthesizes spectral, probabilistic, analytical, and computational techniques to explain how large-system typicality, Markovian closure, and feedback constraints enable the emergence of statistically stable yet intrinsically non-equilibrium phenomena.