Time-Dependent Generalized Gibbs Ensemble

• Time-dependent generalized Gibbs ensemble (t-GGE) is a variational ansatz that models the evolving density matrix using time-dependent Lagrange multipliers on conserved charges.
• It accurately captures transient, prethermal, and relaxation regimes in open and nearly integrable systems like quantum spin chains and bosonic lattice gases.
• Extensive numerical and analytical validations confirm that the t-GGE framework reliably predicts observable dynamics, including algebraic decays and effective power-law behaviors.

A time-dependent generalized Gibbs ensemble (t-GGE) is a variational ansatz for the instantaneous state of a quantum (or classical) many-body system governed by a large set of (quasi-)local conserved charges, in the presence of weak integrability-breaking perturbations, dissipative processes, or slow inhomogeneous transport. The t-GGE describes the evolving density matrix as an exponentiated sum over conserved charges with time-dependent Lagrange multipliers. This approach enables analytically tractable, quantitatively accurate predictions for the full nonequilibrium dynamics in interacting and open integrable systems, capturing transient, prethermal, and relaxation regimes inaccessible to traditional thermal or stationary GGE descriptions. The t-GGE has found central application in open quantum spin chains with non-reciprocal dissipation, lossy one-dimensional gases, weakly driven Heisenberg and XX models, and generalized hydrodynamics of classical chains (Marché et al., 13 Jan 2026, Lumia et al., 2024, Rossini et al., 2020, Lange et al., 2018, Pandey et al., 2024).

1. The t-GGE Ansatz and Foundations

The essential structure of the t-GGE is

$$\rho_{\mathrm{GGE}}(t) = \frac{1}{Z(t)} \exp\left[-\sum_i \lambda_i(t) Q_i\right],$$

where $\{Q_i\}$ are a (quasi-)complete set of local or quasi-local conserved charges of the unperturbed ("integrable") Hamiltonian, and $\{\lambda_i(t)\}$ are their time-dependent Lagrange multipliers. This Gaussian or generalized exponential form remains practical even for infinite families of charges, as in the XX or Heisenberg chains (Marché et al., 13 Jan 2026, Lange et al., 2018).

The choice of charges reflects the physical symmetries and integrable structure of the system (e.g., fermionic occupation numbers $n(k)=c_k^\dagger c_k$ in the XX/HCB mappings, local and higher-spin conserved quantities in XXX/XXZ chains, or normal modes in harmonic chains). In dissipative or non-integrable settings, the approximate conservation of these charges enables the t-GGE to capture "prethermal" plateaus and slow relaxation.

The normalization

$$Z(t) = \operatorname{Tr}\exp\left[-\sum_i \lambda_i(t) Q_i\right]$$

ensures a proper density matrix.

2. Closed Evolution Equations for Lagrange Multipliers and Occupations

To guarantee that the t-GGE reproduces the true time-evolution of the relevant conserved quantities, the Lagrange multipliers are dynamically fixed by imposing that for each $Q_i$,

$$\frac{d}{dt} \langle Q_i \rangle_{\text{t-GGE}} = \operatorname{Tr}[Q_i\, \mathcal{L}(\rho_{\text{t-GGE}})],$$

where $\mathcal{L}$ is the Liouvillian or full generator (unitary plus dissipative/Lindblad parts).

For quadratic or free-fermion models, this leads to closed nonlinear differential equations for the rapidity distributions $n_k(t)$:

• For the XX chain with non-reciprocal loss (Marché et al., 13 Jan 2026):

$$\partial_t \rho(k) = -2\kappa\{\rho(k)[1-\rho(k)][1+\cos(k+\varphi)]+\text{Hilbert/transformed\ kernels}\}$$

involving circular Hilbert transforms and quadratic terms reflecting dissipation-induced mixing among modes.

• For hard-core boson or free-fermion chains under balanced gain/loss (Lumia et al., 2024, Rossini et al., 2020):

$$\partial_t n_k = -\gamma_\text{L} \mathcal{F}_\text{L}[n](k) - \gamma_\text{G} \mathcal{F}_\text{G}[n](k),$$

where the functionals $\mathcal{F}$ involve nontrivial Hilbert transforms and nonlinear mixing.

For interacting models, the equations generalize to a finite set of rate equations for the most relevant approximately conserved observables (Hamiltonian, currents, higher charges), whose explicit evolution is set by projected Lindbladian or dissipative terms (Lange et al., 2018). In all regimes, the equations are self-consistently closed on the GGE parameters.

3. Physical Observables and Their Dynamical Relations

All local observables accessible by the $Q_i$ are calculable as explicit functionals of the time-dependent Lagrange multipliers or rapidity/occupation distributions in the t-GGE. In the open XX chain with non-reciprocal dissipation (Marché et al., 13 Jan 2026):

• Magnetization density $m(t)$ is recovered from $m(t) = \int_0^{2\pi} \rho(k,t)\,dk/2\pi$,
• Hamiltonian (coherent) magnetization current $$J_H(t) = J \int_0^{2\pi} \sin k\, \rho(k,t) dk/2\pi$$,
• Energy density $$\varepsilon(t) = -J \int_0^{2\pi} \cos k\,\rho(k,t) dk/2\pi$$.

Late-time and asymptotic relations emerge between these observables, determined by the structure of the emerging peak in the occupation distribution. Specifically, ratios such as $J_H/m \to J\sin\varphi$ and $\varepsilon/m\to -J\cos\varphi$ appear, with subleading corrections set by the time-dependent peak width $\sigma(t)$ of $\rho(k,t)$.

4. Benchmarking and Accuracy: Numerical and Analytical Validation

Extensive comparisons with exact tensor-network simulations, exact diagonalization, and block-diagonal (diagonal ensemble) approaches confirm that t-GGE provides quantitatively accurate predictions for both transient and late-time dynamics under weak integrability-breaking perturbations and dissipation (Marché et al., 13 Jan 2026, Lumia et al., 2024, Rossini et al., 2020, Lange et al., 2018). Notably:

• The t-GGE reproduces the algebraic decay of magnetization and density, with exponents and finite-time corrections in quantitative agreement with exact data.
• Non-Gaussianity measures (e.g., Rényi-2 relative entropy between exact and GGE-projected states) remain small, showing proximity of the real state to the t-GGE form.
• The evolution of local observables, inhomogeneous profiles (under GHD scaling), and approach to non-equilibrium steady states are matched.
• Truncating the set of charges to the dominant ones is sufficient for virtually all practical observables, as verified by systematic convergence checks.

An important result is that in the presence of two-body or non-reciprocal Lindblad loss, algebraic—rather than simple exponential—decay emerges, characterized by nontrivial exponents and time-dependent corrections, all accounted for by the full t-GGE equations.

5. Anomalous Power-Law Exponents and Their Origin

In certain realizations, such as the XX chain with non-reciprocal Lindblad losses, the decay of observables (e.g., magnetization) at intermediate times follows effective power laws $m(t)\sim t^{-\chi}$ with exponents $\chi$ numerically found in the range $0.515$–$0.58$ (Marché et al., 13 Jan 2026). However, analysis of the instantaneous exponent $\chi_{\text{inst}}(t) = -d \ln m / d \ln t$ shows slow, non-monotonic drift, indicating the absence of a true universal power-law. The source of this behavior is:

• The slow, time-dependent narrowing of the occupation distribution's peak, encoded in the nonlinear t-GGE equations,
• The presence of nontrivial string operators in Lindblad terms arising from fermion mappings.

In the ultimate scaling limit, the exponent reverts to the universal value (e.g., $\chi=1/2$ for decaying hard-core boson density), but practical timescales accessible in experiments and numerics exhibit effective exponents modified by logarithmic corrections $m(t) \sim t^{-1/2} [1 + \mathcal{O}(1/\ln t)]$.

6. Relation to Stationary and Optimized GGEs

The t-GGE generalizes the stationary GGE of closed integrable systems. While the stationary GGE is constructed by matching the late-time expectation values of the conserved charges, the t-GGE dynamically evolves these values, tracking the instantaneous expectation of each conserved quantity (Sels et al., 2014). The optimal GGE is found by minimizing the time-averaged relative entropy distance (Kullback–Leibler divergence) between the actual time-dependent state and the exponential ansatz constrained by the observed expectation values.

In closed systems, inclusion of higher-body conserved quantities (e.g., quadratic in occupations) produces correlated GGEs that match higher-order correlations. In open/dissipative or slowly driven systems, the t-GGE evolves all (or the most relevant) Lagrange multipliers via the set of derived rate equations, providing a low-dimensional, controlled representation for the full dynamics far from equilibrium.

7. Applications and Regimes of Validity

The t-GGE formalism applies in a broad range of contexts:

• Open quantum spin chains with non-reciprocal or local Lindblad dissipation, capturing correlation decay, current generation, and the approach to NESS beyond free-fermion solvable limits (Marché et al., 13 Jan 2026, Lumia et al., 2024).
• Bosonic lattice gases under strong two-body loss (quantum Zeno regime) and their relaxation dynamics, revealing algebraic decay and momentum distribution structures inaccessible to mean-field (Rossini et al., 2020).
• Weakly driven and nearly integrable Hamiltonians, where prethermal and nonequilibrium steady states can be tracked by evolving only a few Lagrange multipliers (Lange et al., 2018).
• Classical chains (e.g., harmonic chains), coupled with generalized hydrodynamics, where the local occupations (Wigner function) evolve as per GHD and relax toward GGE equilibrium in a controlled fashion (Pandey et al., 2024).

The regime of highest accuracy is weak dissipation (perturbative in the Lindblad/Kolmogorov rates compared to the energy scale of $H$), and/or long-wavelength Euler and adiabatic limits. The physical mechanism is the emergence of time-scale separation: rapid local prethermalization to a GGE, followed by slow motion along the GGE manifold set by the weak non-integrable or dissipative terms.

Table: t-GGE Framework Across Different Systems

System Type	Conserved Charges	Governing t-GGE Equation
XX spin chain with Lindblad loss (Marché et al., 13 Jan 2026)	$c_k^\dagger c_k$, non-local $Q_m$	Nonlinear ODE for $\rho(k,t)$
HCB under loss/gain (Lumia et al., 2024)	Fermionic occupations $n(k)$	Hilbert-transformed kinetic equation
HCB with two-body loss (Rossini et al., 2020)	$c_k^\dagger c_k$ (free-fermion)	Quadratic nonlinear rate equation
Heisenberg chain, weak drive (Lange et al., 2018)	Local $C_i$, truncated set	ODEs for $\lambda_i(t)$
Classical harmonic chain (Pandey et al., 2024)	Mode occupations, $Q_m$	GHD for $n(x,k,t)$

In all cases, the key t-GGE feature is a closed, deterministic evolution for the relevant set of variables, determined by the physical symmetry and perturbation structure.

• "Open quantum spin chains with non-reciprocity: a theoretical approach based on the time-dependent generalized Gibbs ensemble" (Marché et al., 13 Jan 2026)
• "Accuracy of time-dependent GGE under weak dissipation" (Lumia et al., 2024)
• "Strong correlations in lossy one-dimensional quantum gases: from the quantum Zeno effect to the generalized Gibbs ensemble" (Rossini et al., 2020)
• "Time-dependent generalized Gibbs ensembles in open quantum systems" (Lange et al., 2018)
• "Generalized hydrodynamics and approach to Generalized Gibbs equilibrium for a classical harmonic chain" (Pandey et al., 2024)
• "Stationary ensemble approximations of dynamic quantum states: Optimizing the Generalized Gibbs Ensemble" (Sels et al., 2014)