Out-of-Equilibrium Phase Transitions

• Out-of-equilibrium phase transitions are defined by non-analytic changes in dynamic observables that arise when systems are driven far from thermal equilibrium.
• They are characterized by singularities in large deviation functions and scaled cumulant generating functions, revealing critical changes in system trajectories and order parameters.
• These transitions are analyzed using field-theoretic renormalization, macroscopic fluctuation theory, and experimental probes like Loschmidt echoes and OTOCs in both classical and quantum systems.

Out-of-equilibrium phase transitions are abrupt, singular changes in the macroscopic or dynamical behavior of physical systems driven or evolving far from equilibrium, typically governed by parameters other than thermal temperature. Unlike equilibrium transitions, which are classified by changes in free energy landscapes as external control variables such as temperature or field are varied, out-of-equilibrium phase transitions emerge in steady-state, transient, or trajectory-space properties—such as currents, entropy production rates, return probabilities, or long-time averages—when the system is subject to external driving, dissipation, or global non-equilibrium protocols. This category encompasses discontinuities or singularities in large deviation functions (LDFs), dynamic order parameters, synchronization regimes, quasistationary state structures, or critical points in non-thermal steady states, and applies in both classical and quantum settings, including stochastic networks, driven lattice gases, quantum quenches, and engineered Floquet protocols.

1. Distinguishing Features and General Theory

Out-of-equilibrium phase transitions are characterized by singularities in dynamical observables and/or large deviation rate functions that describe rare fluctuations of time-integrated quantities. Key hallmarks include non-analytic behavior in scaled cumulant generating functions (SCGFs), discontinuities or cusps in derivatives of rate functions, and abrupt shifts in the structure and universality of dynamical steady states or trajectory ensembles.

Dynamical phase transitions (DPTs) can be continuous (second-order) or discontinuous (first-order), much like their equilibrium analogs, but are generically governed by fundamentally different scaling principles. For example, DPTs in Markovian or quantum jump processes are often signaled by singularities in SCGFs ψ(λ) associated with entropy production, currents, or time-integrated observables:

• ψ(λ) = lim_{τ→∞} (1/τ) ln⟨e^{-λO(τ)}⟩

The dual Legendre transform yields the large deviation rate function I(o):

• I(o) = sup_{λ}[ −λ o − ψ(λ) ]

A cusp or discontinuous derivative in ψ(λ) as system size increases indicates a dynamical phase transition, typically corresponding to coexistence of qualitatively distinct classes of trajectories or fluctuations (Vaikuntanathan et al., 2013).

The general classification and scaling framework for non-equilibrium critical phenomena relies on field-theoretic renormalization group (RG) tools, dynamic scaling hypotheses, and universal finite-size scaling forms adapted to the relevant driven or dissipative context. In many cases, macroscopic fluctuation theory (MFT) and the additivity principle provide large-deviation-based approaches for diffusive systems or models with conservation laws (Shpielberg et al., 2015, Täuber, 2016).

2. Kinetic Networks, Driven Diffusive Systems, and Entropy-Producing Trajectories

In minimal driven kinetic networks, such as ring or triangle-decorated Markov chains, phase transitions arise directly in the ensemble of dynamical trajectories. Broken detailed balance (external drive) induces nonequilibrium steady states with persistent currents and nonzero entropy production. The probability distribution P(σ) of the time-averaged entropy production rate σ over a trajectory of duration τ exhibits the large-deviation form:

• P(σ) ≈ e^{−τ I(σ)}

The scaled cumulant generating function ψ(λ), as the largest eigenvalue of a tilted rate matrix, acts as a “dynamical free energy.” In large system-size limits, ψ(λ) develops a non-analytic cusp at a critical λ*, where its first derivative jumps discontinuously. This signals a first-order dynamical phase transition between delocalized, high-entropy trajectories (typical cycling) and localized, low-entropy trajectories (trapping near a disorder or heterogeneity) (Vaikuntanathan et al., 2013).

Such phenomena generalize to boundary-driven diffusive systems governed by MFT. There, DPTs are manifested as singularities in the current large deviation function Φ(J) or its SCGF Ψ(λ), determined via stability analyses (e.g., Le Châtelier inequalities (Shpielberg et al., 2015)) or Landau-like expansions in the amplitude of symmetry-breaking density modes:

• Continuous (second-order) DPTs: breaking of particle-hole symmetry; amplitude m grows as m ~ |λ−λ_c|^{1/2}.
• First-order DPTs: symmetry absent, leading to a discontinuous jump in m at λ_d^± with phase coexistence (Baek et al., 2016).

Microscopic realizations include the KLS model, WASEP, and classical exclusion processes.

3. Quantum Dynamical Phase Transitions and Loschmidt Echoes

Dynamical quantum phase transitions (DQPTs) correspond to non-analytic temporal behavior in quantities such as the Loschmidt echo—either global or in subsystem-reduced form (RLE)—after a sudden quench. The rate function λ(t) = −(1/N) ln |G(t)|^2 may exhibit singularities at critical times t_c, where Fisher zeros of the return amplitude cross the real time axis, forming characteristic cusp-like features (Parez et al., 1 Sep 2025, Vianello et al., 23 Feb 2026).

In free fermion systems, explicit criteria for DQPTs can be derived: for a specific momentum mode k*, a DQPT occurs if the post-quench occupation satisfies 2n_{k^*}−1=0 and t^* equals an odd multiple of π/(2ε_{k^*}). In interacting bosonic Josephson junctions, DQPT critical times and the sharpness of cusps are directly influenced by pair tunneling and the system's nonlinearity (Parez et al., 1 Sep 2025, Vianello et al., 23 Feb 2026).

Experimentally, DQPTs are precisely diagnosed by the behavior of out-of-time-ordered correlators (OTOCs), which can also signal equilibrium QPTs and order parameters in integrable, non-integrable, and long-range quantum spin chains. Notably, OTOC sign-changes, zeros, or sharp crossovers as functions of control parameter or time robustly distinguish between phases in experimental quantum simulators (Nie et al., 2019, Heyl et al., 2018).

4. Out-of-Equilibrium Criticality, Universal Scaling Laws, and Spinodal Dynamics

Out-of-equilibrium dynamic criticality often requires new scaling paradigms beyond equilibrium RG. For continuous transitions, dynamic scaling and universality classes (KPZ, directed percolation, reaction-diffusion) are well established (Täuber, 2016), with scaling forms such as

• ξ ~ |Δ|^{-ν}, t_c ~ ξ^z, C(r,t) ~ r^{-(d−2+η)} F(t/r^z)

For magnetic or thermal first-order transitions, standard correlation lengths remain finite; instead, the relevant scale is the coherence length (e.g., ξ_h(h) ~ |h|^{-1/D}). Off-equilibrium driving (e.g., linear magnetic field ramp) leads to a universal scaling of the magnetization, domain sizes, and dissipated work via the freeze-out time and length:

• For first-order: $\hat t_{\rm FOT}\sim\tau_Q^{1/2}$, $\hat\ell\sim\tau_Q^{1/(2D)}$, dissipated work per cycle $W\sim\tau_Q^{-1/2}$
• For continuous: $\hat t\sim\tau_Q^{z\nu_h/(1+z\nu_h)}$, $\hat\xi\sim\tau_Q^{\nu_h/(1+z\nu_h)}$, $W\sim\tau_Q^{-(2/3)}$ (D=3 example) (Scopa et al., 2018)

Out-of-equilibrium spinodal-like transitions in short-range systems emerge upon slow (Kibble–Zurek) traversing of FOTs, where nucleation, rather than critical slowing-down, controls the scaling:

• $$E(t,t_s)\approx{\cal E}\bigl(t(\ln t)^{d/(d-1)}/t_s\bigr)$$

The critical threshold shifts as $\delta\beta_*\sim1/(\ln t_s)^{d/(d-1)}$, receding logarithmically in the adiabatic limit (Pelissetto et al., 20 Nov 2025).

Quenches across FOTs can also induce dynamic percolation transitions at finite critical times, at which the largest spin clusters percolate. Fractal dimensions remain random-percolation-like, but approach exponents $w(h)$ depend on the quench strength and vanish as $h\to0$, with the critical time scaling $t_c\sim e^{A h^{-1/2}}$ (in 2D Ising case) (Pelissetto et al., 13 Mar 2026).

5. Floquet and Mean-Field Out-of-Equilibrium Phase Transitions

Periodically driven quantum systems undergo “Floquet phase transitions” when resonances (avoided crossings) in the Floquet quasi-energy spectrum are tuned through critical lines in parameter space. These transitions are associated with sharp non-analyticities (cusps) in synchronized-state observables, (e.g., magnetization, correlators) and distinct long-time relaxation regimes (local/non-local, with exponents n^{-3/2} or n^{-1/2}) (Arze et al., 2018).

In long-range interacting systems and mean-field models (e.g., HMF, SYK+U), out-of-equilibrium QSS transitions arise as bifurcations in the Vlasov equation's one-particle effective Hamiltonian dynamics. Critical points correspond to structural changes from integrable (magnetized) to ergodic (nonmagnetized or traveling-wave resonance) phase-space topology. Rich phase diagrams containing tricritical, azeotropic, and ensemble-inequivalent regimes can be established, and QSS lifetimes diverge with system size (Firpo, 2011, Staniscia et al., 2010, Alexandrov et al., 2023).

6. Field-Theoretic and Universality Aspects: Exceptional Points and Ginzburg-Landau Descriptions

Nonequilibrium O(N) models can develop exceptional critical points (CEPs), where both the spectrum and dynamical matrix become non-diagonalizable, and gapless modes coalesce. At CEPs, order parameters undergo continuous limit-cycle rotations with the number of Goldstone modes greatly enhanced. Fluctuations diverge due to vanishing friction and finite noise, producing either fluctuation-induced first-order transitions or restoration of symmetry (no long-range order) in $d<4$. The field theory is governed by Langevin dynamics with anti-damping, generically breaking fluctuation-dissipation balance (Zelle et al., 2023).

In spatially extended systems, out-of-equilibrium phase transitions in learning dynamics of neural or diffusion models can be connected to Ginzburg-Landau theory. The critical regime is characterized by amplification of soft Fourier modes, diverging correlation lengths, and emergence of collective order—generalizing pitchfork bifurcations to spatiotemporal criticality (Ambrogioni, 20 Mar 2026).

7. Application Domains and Implications

Out-of-equilibrium phase transitions are realized in biological networks (molecular motors, cellular proofreading), nanoscale electronic circuits, spintronic devices, driven cold atomic gases, artificial neural architectures, and nonequilibrium quantum simulators. Their functional consequences range from bimodal response and enhanced fluctuations for biological fidelity to the control of generation in machine learning models.

These transitions have no true thermal analogy and therefore require new classification schemes—dynamic universality classes, trajectory-based large deviation frameworks, and physically motivated RG tools. As such, they constitute a central organizing principle in the modern theory of complex, driven, and dissipative many-body systems.