Nonhomogeneous Relaxation Techniques

• Nonhomogeneous relaxation techniques are methods where relaxation rates vary with state, time, or space, enabling more precise modeling of complex systems.
• They employ advanced methods such as exponential integrators, self-scattering algorithms, and fractional derivatives to address stiffness and variability in simulations.
• Applications in multi-phase flows, charge transport, and heterogeneous media validate these techniques by capturing nonlocal effects, memory dynamics, and anomalous behaviors.

Nonhomogeneous relaxation techniques constitute a class of methods for modeling and numerically solving systems where relaxation rates—temporal or spatial—depend on variables such as position, phase, energy, or momentum. These techniques arise in domains including multi-phase fluid dynamics, charge transport, stochastic processes with memory, and the mechanics of heterogeneous media. They share the principle of relaxing away the assumption of constant (homogeneous) relaxation times, adopting instead operators, kernels, or time-distribution functions that are nonuniform in space, time, or underlying state variables. Analytical frameworks, exponential integrators, mixture-distribution constructions, and fractional differential operators are typical components for capturing nonhomogeneous relaxation phenomena.

1. Mathematical Foundations of Nonhomogeneous Relaxation

Nonhomogeneous relaxation generalizes classical relaxation, where the return to equilibrium is typically governed by linear, constant-coefficient operators (e.g., $\dot{x} = -\lambda x$). In nonhomogeneous settings, relaxation rates may be functions of the local state, time, spatial position, or other structural variables. Consider $dV/dt = S(V)$ for a vector of primitive variables $V$; nonlinearity and inhomogeneity appear naturally in the source term $S(V)$. Linearization at a reference state $V^*$ yields an affine ODE,

$$\frac{dV}{dt} = J V + R,\quad J = \left.\frac{\partial S}{\partial V}\right|_{V^*},\quad R = S(V^*) - J V^*,$$

leading to solution methods that must treat the nonconstant structure of $J$ and the nontrivial kernel $R$ (Chiocchetti et al., 2020).

In stochastic models, nonhomogeneity enters via position- or energy-dependent waiting times or transition kernels. For instance, the probability $P(t)$ that a particle undergoes a transition at time $t$ in a process with variable relaxation time $\tau(\epsilon,k)$ is

$$P(t) = \frac{1}{\tau(\epsilon(t),k(t))} \exp\left(-\int_0^t \frac{dt'}{\tau(\epsilon(t'),k(t'))}\right),$$

a nonhomogeneous Poisson process (McDonough et al., 21 Aug 2025).

Fractional-order derivatives, spatially-varying memory kernels, and subordination techniques further extend the reach of nonhomogeneous relaxation, allowing for power-law, nonlocal, and history-dependent behaviors (Srokowski, 2015).

2. Exponential Integrators and Semi-Exact Solvers

Nonhomogeneous relaxation ODEs, particularly in engineering models such as the Baer-Nunziato two-phase flow systems, are often stiff and high-dimensional. Exponential integrators provide a robust strategy for integrating such ODEs:

$$V(t_{n+1}) = e^{\Delta t J} V(t_n) + \int_0^{\Delta t} e^{(\Delta t - \tau) J} R\,d\tau,$$

where $J$ is block-structured and admits closed-form exponentials in subsystem variables (e.g., velocity and pressure). Rather than inverting a full Jacobian, block-diagonalization and scalar-exponential formulas are exploited, ensuring both efficiency and numerical stability for arbitrary or stiff relaxation rates (Chiocchetti et al., 2020, Pelanti, 2021). For systems reducible to a single relaxation variable, analytic solutions for arbitrary rates yield updates such as

$\Delta p^* = \Delta p^0\,e^{-K_p \Delta t},$

where $K_p$ is a composite of phase impedances or coupling coefficients (Pelanti, 2021).

3. Nonhomogeneous Relaxation in Charge Transport: Self-Scattering Algorithms

In semi-classical charge-transport models governed by the Boltzmann transport equation (BTE), carrier scattering rates $\tau(\epsilon,k)$ depend on energy and momentum. The nonhomogeneous relaxation time method, implemented via the self-scattering (SS) technique, reconstructs the correct free-flight time distribution through mixture sampling:

• A constant overestimate $\Gamma$ of the maximal rate is set.
• Flight times $t$ are drawn from an exponential PDF with rate $\Gamma$.
• Real scattering events are chosen probabilistically from the true energy/momentum-dependent rates; all others are self-scattering events, which advance time but do not alter the state.

It is analytically proven that summing over all possible self-scattering histories reproduces the Poisson kernel corresponding to the full variable-dependent rate. Numerical, stochastic, and mixture-theoretical frameworks are supplied for validation and implementation (McDonough et al., 21 Aug 2025).

System	Relaxation Kernel	Solution Framework
Baer-Nunziato flows	Affine in variables	Exponential integrators, blockwise closed-form
Charge transport (BTE)	$\tau(\epsilon,k)$	Self-scattering, mixture distributions
Lévy flights w/ memory	$g(x)$ position-dependent	Fractional FP, subordination

4. Role of Spatial and Structural Heterogeneity

Nonhomogeneous relaxation is intimately connected to spatial or structural heterogeneity. In in-series chains of viscoelastic (Kelvin-Voigt) blocks modeling soft materials, even simple two-block chains ($n=2$) with differing elastic or viscosity coefficients exhibit nonmonotonic relaxation: transient overshoots in strain or force arise only because the blocks examine disparate time scales and coupling (Bedulina et al., 2013). Excessive or insufficient relaxation in one domain propagates and perturbs state evolution in distant blocks, a phenomenon inaccessible to homogeneous relaxation models. This effect generalizes to mechanical, thermal, and chemical relaxation in multi-phase flows, where phase-specific rates and mixture constraints must be dynamically re-imposed after each substep (Pelanti, 2021).

5. Fractional and Anomalous Relaxation: Memory Effects

Nonhomogeneous memory is most naturally captured via fractional Fokker–Planck equations. Subordination techniques—where operational time $\tau$ is linked to physical time $t$ via a random process with position-dependent intensity $g(x)$—give rise to time-fractional equations:

$$\frac{\partial p(x,t)}{\partial t} = {}_0D_t^{1-\beta} \left[ \lambda\,\frac{\partial}{\partial x}(x\,p(x,t)) + \frac{\partial^\alpha}{\partial|x|^\alpha}(|x|^{-\theta}\,p(x,t)) \right],$$

where ${}_0D_t^{1-\beta}$ is the Riemann–Liouville fractional derivative. For quadratic potentials, relaxation to the stationary state is governed by a series of Mittag-Leffler functions, inducing power-law decay; nonlinear potentials induce exponential (Debye-type) relaxation, with rates and amplitudes strongly affected by the heterogeneity kernel $g(x)$ (Srokowski, 2015).

6. Practical Implementation and Numerical Stability

Nonhomogeneous relaxation techniques demand tailored numerical algorithms:

• Blockwise exponential integration and avoidance of matrix inversion for stiff source solvers.
• Adaptive timestep control using defect-based estimators and physical admissibility checks (positivity, volume fraction bounds).
• In charge transport, adaptive estimation of the maximal rate $\Gamma$, domain segmentation, and post-processing of self-scatter events to ensure statistical fidelity to the physical model (Chiocchetti et al., 2020, McDonough et al., 21 Aug 2025).
• Physical consistency constraints (energy, mass) must be enforced at each relaxation substep and often require root-finding in a mixture-pressure or energy equation (Pelanti, 2021).

Pathologies such as unbounded rates, narrow peaks in kernel distributions, or excessive self-scattering must be detected and remedied through regularization, energetic windowing, and algorithmic monitoring.

7. Applications and Representative Results

Nonhomogeneous relaxation methods are integral to:

• Robust simulation of multi-phase compressible flows with arbitrary-rate heat and mass transfer and complex equations of state, enabling physically accurate reproduction of metastability, shock contact invariance, and phase-transition fronts (Pelanti, 2021).
• Monte Carlo simulation of charge transport with ab initio scattering rates, enabling correct mobility, conductivity, and diffusion statistics, even for singular kernel forms (McDonough et al., 21 Aug 2025).
• Stochastic modeling of systems with position-dependent trap densities, anomalous diffusion, or nonlocal effects in biological and polymeric materials; capturing nonmonotonic relaxation, overshoots, and memory-induced slow decay (Bedulina et al., 2013, Srokowski, 2015).

These methods underpin modern approaches to finite-rate and stiff relaxation in scientific computing, allowing accurate, efficient, and physically consistent modeling of complex heterogeneous systems.