Markovian Relaxations: Dynamics & Equilibrium

• Markovian relaxations are defined as memoryless dynamical processes that drive systems to equilibrium using CPTP maps and GKLS equations.
• They play a crucial role in nonequilibrium statistical mechanics, quantum thermodynamics, and open system dynamics.
• These relaxations are applicable in classical stochastic models and numerical iterative methods for Markov chains.

Markovian relaxations describe the approach of a system—classical or quantum, stochastic or dissipative—to an asymptotic state or trajectory under evolution governed by Markovian (memoryless or time-local) dynamics. In quantum systems, this is typically formalized through completely positive, trace-preserving (CPTP) dynamical maps generated by time-local Gorini–Kossakowski–Lindblad–Sudarshan (GKLS) master equations. Such dynamics are called Markovian when the evolution is CP-divisible, meaning that the state at any later time can be obtained from any earlier state via a CPTP map. Markovian relaxation is central to nonequilibrium statistical mechanics, quantum thermodynamics, and the theory of open systems, providing the mathematical underpinning for the irreversible approach to steady state, limit cycles, or equilibrium in systems interacting with an environment.

1. Markovian Quantum Dynamics and Relaxation Criteria

For a finite-dimensional quantum system with density matrix $\rho(t)$, generic time-dependent Markovian evolution is governed by a time-local GKLS master equation: $$\frac{d\rho(t)}{dt} = \mathcal{L}_t[\rho(t)] = -i[H(t),\rho] + \sum_\alpha \gamma_\alpha(t)\left(L_\alpha(t)\rho L_\alpha^\dagger(t) - \frac{1}{2}\{L_\alpha^\dagger(t)L_\alpha(t),\rho\}\right),$$ where $\gamma_\alpha(t)\ge0$, $L_\alpha(t)$ are (possibly time-dependent) jump operators, and $H(t)=H(t)^\dagger$ is the system Hamiltonian (Meglio et al., 2024).

Relaxation describes the property that, irrespective of the initial state, the system evolves towards a unique asymptotic object. Two precise notions:

• Strong relaxation: There exists a unique stationary state $\rho_{\mathrm{st}}$ such that $$\lim_{t\to\infty} \|\Lambda_{t,0}(\rho)-\rho_{\mathrm{st}}\|_1 = 0$$ for all states $\rho$.
• Weak relaxation: There exists a unique trajectory $t\mapsto\rho^*(t)$ such that $$\lim_{t\to\infty} \|\Lambda_{t,0}(\rho)-\rho^*(t)\|_1 = 0$$ for all $$\frac{d\rho(t)}{dt} = \mathcal{L}_t[\rho(t)] = -i[H(t),\rho] + \sum_\alpha \gamma_\alpha(t)\left(L_\alpha(t)\rho L_\alpha^\dagger(t) - \frac{1}{2}\{L_\alpha^\dagger(t)L_\alpha(t),\rho\}\right),$$0 (Meglio et al., 2024).

Markovianity, in this context, is equivalent to complete positivity (CP)-divisibility: for all $$\frac{d\rho(t)}{dt} = \mathcal{L}_t[\rho(t)] = -i[H(t),\rho] + \sum_\alpha \gamma_\alpha(t)\left(L_\alpha(t)\rho L_\alpha^\dagger(t) - \frac{1}{2}\{L_\alpha^\dagger(t)L_\alpha(t),\rho\}\right),$$1, there exists a CPTP map $$\frac{d\rho(t)}{dt} = \mathcal{L}_t[\rho(t)] = -i[H(t),\rho] + \sum_\alpha \gamma_\alpha(t)\left(L_\alpha(t)\rho L_\alpha^\dagger(t) - \frac{1}{2}\{L_\alpha^\dagger(t)L_\alpha(t),\rho\}\right),$$2 such that $$\frac{d\rho(t)}{dt} = \mathcal{L}_t[\rho(t)] = -i[H(t),\rho] + \sum_\alpha \gamma_\alpha(t)\left(L_\alpha(t)\rho L_\alpha^\dagger(t) - \frac{1}{2}\{L_\alpha^\dagger(t)L_\alpha(t),\rho\}\right),$$3.

2. Algebraic Relaxation Theorems and Time-Dependent Spohn Criteria

A central result is the extension of Spohn's irreducibility theorem to time-dependent Lindblad generators. For a family of bounded GKLS generators $$\frac{d\rho(t)}{dt} = \mathcal{L}_t[\rho(t)] = -i[H(t),\rho] + \sum_\alpha \gamma_\alpha(t)\left(L_\alpha(t)\rho L_\alpha^\dagger(t) - \frac{1}{2}\{L_\alpha^\dagger(t)L_\alpha(t),\rho\}\right),$$4, let $$\frac{d\rho(t)}{dt} = \mathcal{L}_t[\rho(t)] = -i[H(t),\rho] + \sum_\alpha \gamma_\alpha(t)\left(L_\alpha(t)\rho L_\alpha^\dagger(t) - \frac{1}{2}\{L_\alpha^\dagger(t)L_\alpha(t),\rho\}\right),$$5 denote the set of (time-dependent) jump operators. Define the set to be:

• Self-adjoint: $$\frac{d\rho(t)}{dt} = \mathcal{L}_t[\rho(t)] = -i[H(t),\rho] + \sum_\alpha \gamma_\alpha(t)\left(L_\alpha(t)\rho L_\alpha^\dagger(t) - \frac{1}{2}\{L_\alpha^\dagger(t)L_\alpha(t),\rho\}\right),$$6 for all $$\frac{d\rho(t)}{dt} = \mathcal{L}_t[\rho(t)] = -i[H(t),\rho] + \sum_\alpha \gamma_\alpha(t)\left(L_\alpha(t)\rho L_\alpha^\dagger(t) - \frac{1}{2}\{L_\alpha^\dagger(t)L_\alpha(t),\rho\}\right),$$7.
• Irreducible: The commutant $$\frac{d\rho(t)}{dt} = \mathcal{L}_t[\rho(t)] = -i[H(t),\rho] + \sum_\alpha \gamma_\alpha(t)\left(L_\alpha(t)\rho L_\alpha^\dagger(t) - \frac{1}{2}\{L_\alpha^\dagger(t)L_\alpha(t),\rho\}\right),$$8 consists only of scalar multiples of the identity, i.e., $$\frac{d\rho(t)}{dt} = \mathcal{L}_t[\rho(t)] = -i[H(t),\rho] + \sum_\alpha \gamma_\alpha(t)\left(L_\alpha(t)\rho L_\alpha^\dagger(t) - \frac{1}{2}\{L_\alpha^\dagger(t)L_\alpha(t),\rho\}\right),$$9.

Theorem (Generalized Spohn, (Meglio et al., 2024)):

Suppose that for a measurable set $\gamma_\alpha(t)\ge0$0 of infinite Lebesgue measure, the set $\gamma_\alpha(t)\ge0$1 is self-adjoint and irreducible. Then $\gamma_\alpha(t)\ge0$2 is weakly relaxing.

If additionally the generator preserves the identity, $\gamma_\alpha(t)\ge0$3 for all $\gamma_\alpha(t)\ge0$4, then relaxation is strong and the unique stationary state is the microcanonical state $\gamma_\alpha(t)\ge0$5 (Meglio et al., 2024).

The proof leverages Hilbert–Schmidt decomposition and analysis of the contraction properties induced by the generator, using irreducibility to preclude nontrivial invariant subspaces and Grönwall-type arguments for norm decay.

3. Beyond Instantaneous Markovianity: Transients and Mixing Phenomena

Markovian relaxation properties persist under a broad class of situations:

• Transient Non-Markovianity: If the evolution becomes CP-divisible (i.e., Lindbladian with nonnegative rates and irreducible, self-adjoint jumps) after some finite time $\gamma_\alpha(t)\ge0$6, then asymptotic relaxation still holds, independent of initial non-Markovian memory effects (Meglio et al., 2024).
• Mixtures of Channels and Non-Convexity: The set of CP-divisible (Markovian) channels is non-convex. Convex mixing of two Markovian channels can yield a non-Markovian channel, and conversely, suitable mixtures of non-Markovian channels can generate a Markovian semigroup (Uriri et al., 2019, Wudarski et al., 2016). This non-convex geometry is both theoretically and experimentally relevant, with experimental demonstrations using photonic channels (Uriri et al., 2019).

Scenario	Mixture Result	Source
Markovian + Markovian	Can be non-Markovian	(Uriri et al., 2019)
non-Markovian + non-Markovian	Can be Markovian	(Wudarski et al., 2016)
Markovian + non-Markovian	Can be either	(Uriri et al., 2019)

4. Markovian Relaxations in Classical and Quantum Stochastic Systems

Markovian relaxation underpins the irreversible approach to equilibrium or steady state in stochastic models, both quantum and classical. In classical contexts, master equations and Fokker–Planck operators with detailed balance support relaxation to unique stationary distributions.

In quantum irreversible thermodynamics, the quantum Fokker–Planck equation provides a Markovian (memoryless) description: $\gamma_\alpha(t)\ge0$7 where $\gamma_\alpha(t)\ge0$8 is a positive-definite kinetic coefficient matrix, $\gamma_\alpha(t)\ge0$9 is a thermodynamic potential, and the equation remains linear and time-local. This framework reproduces the Gibbs–Boltzmann equilibrium and is consistent with fluctuation–dissipation relations (Tsekov, 2015). The Markovian assumption here excludes persistent memory (nonlocal) effects and system–bath entanglement.

5. Markovian Relaxation in Markov Chains and Iterative Solvers

In numerical analysis of Markov chain models, "Markovian relaxation" can also refer to relaxations in iterative fixed-point schemes for finding stationary distributions or minimal solutions to matrix equations associated with stochastic processes. Accelerated or "relaxed" two-step fixed-point iterations, such as the staircase splitting for M/G/1-type Markov chains, introduce relaxation parameters $L_\alpha(t)$0 to blend updates from separate steps, with convergence to the minimal nonnegative solution under mild monotonicity (M-matrix) assumptions (Gemignani et al., 2022).

Explicit matrix-inequality eligibility criteria for over-relaxation ($L_\alpha(t)$1) allow adaptive acceleration, and the theoretical framework provides spectral estimates for the asymptotic rate of convergence. Such techniques yield significant speed-ups, especially in nearly singular problems and large-scale queueing models.

6. Spectral Duality and Interlacing in Reversible Markov Processes

In ergodic, reversible Markovian dynamics, the relaxation spectrum of the generator (eigenvalues $L_\alpha(t)$2) is intimately linked to the first-passage spectrum ($L_\alpha(t)$3) via a spectral interlacing theorem: $L_\alpha(t)$4 with strict inequalities $L_\alpha(t)$5 for one-dimensional or nearest-neighbor systems endowed with detailed balance. This duality allows the computation of the entire first-passage time distribution from the relaxation spectrum alone, providing a universal bridge between relaxation kinetics and first-passage processes (Hartich et al., 2018, Hartich et al., 2018).

This interlacing is constructive: the Laplace transform of the propagator at the absorbing site encodes both sets of eigenvalues, and Newton-type root-finding procedures use the relaxation spectrum to explicitly extract all first-passage rates.

7. Markovian Embedding of Non-Markovian Dynamics

Thermodynamics of non-Markovian generalized Langevin equations with linear memory kernels can be formulated by embedding the system into a higher-dimensional Markovian system with auxiliary variables. The resulting Markovian representation satisfies detailed balance, and all thermodynamic quantities—work, heat, entropy production—become uniquely defined and monotonically increasing, in contrast to the original non-Markovian formulation where entropy production may transiently decrease. The apparent reduction in entropy in the projected non-Markovian system is attributed to information and heat exchange with the auxiliary degrees of freedom. The embedding approach is constructive for arbitrary linear memory kernels compatible with the fluctuation–dissipation relation (Dechant et al., 28 Apr 2026).

---

Markovian relaxations thus encompass the algebraic, dynamical, spectral, and algorithmic aspects of approach to equilibrium or steady-state in both quantum and classical stochastic systems. Advances in generalized Spohn-type theorems, the geometry of quantum channels, constructive dualities in kinetic theory, and Markovian embeddings continue to refine the theoretical and practical understanding of relaxation phenomena under memoryless dynamics.