Speed-Change Exclusion Process

• The speed-change exclusion process is an interacting particle system with configuration-dependent jump rates, enabling variable particle speeds and complex phase behavior.
• It utilizes advanced techniques such as hydrodynamic scaling limits and Wiener–Itô chaos expansion to analyze nonlinear diffusion and relaxation phenomena.
• Applications span traffic flow modeling, dynamic defect analysis, and phase transitions in active matter, providing insights into transport and stochastic particle dynamics.

A speed-change exclusion process is an interacting particle system on a lattice in which the rate at which particles move (extractively, jump rates) depends on the local or global configuration, in contrast to the simple exclusion process (SEP), where jump rates are uniform. This class includes models with local environment-dependent rates, variable or adaptive particle speeds, dynamic defects, and long-range effects, providing a versatile framework for modeling transport, stochastic particle flows, traffic, and dynamical random environments. Theoretical analysis reveals rich relaxation, hydrodynamic, fluctuation, and phase transition properties shaped by the interplay between speed-heterogeneity and exclusion.

1. Fundamental Definition and Classes

A general speed-change exclusion process on a graph (e.g., $\mathbb{Z}^d$ or a finite torus) is defined by the Markov generator: $$L f(\eta) = \sum_{x, y} \eta(x) [1-\eta(y)]\, c_{x,y}(\eta) \left(f(\eta^{x,y}) - f(\eta)\right)$$ where:

• $\eta$ is the occupation configuration.
• $\eta(x), \eta(y) \in \{0,1\}$ are site occupations.
• $f$ is an observable.
• $c_{x,y}(\eta)$ is a non-negative, configuration-dependent jump rate.
• $\eta^{x,y}$ is the configuration after swapping occupations at $(x,y)$.

Classical special cases include:

• Simple exclusion process (SEP): $c_{x,y}(\eta)$ constant, nearest-neighbor symmetric/asymmetric.
• Speed-change model: $c_{x,y}(\eta) = q(x,y) c_{\eta}(x)$, with $c_{\eta}(x)$ depending locally on $\eta$, e.g., function of neighborhood (“environment” of $x$).
• Multi-speed generalizations: Each particle carries a “speed” or “class” label that may evolve, requiring rules for synchronization or renewal upon interaction (Furtlehner et al., 2011).
• Dynamic defects: Site- or bond-specific $c_{x,y}(\eta)$ modified (temporarily or permanently) in particular regions, possibly moving as in (Chatterjee et al., 2016, Das et al., 25 Mar 2025).

Stationarity is often with respect to the product Bernoulli measure $\nu_\rho$, but in presence of nonuniform rates, more intricate invariant measures or phase transitions can appear.

2. Hydrodynamic Limit and Macroscopic Equations

Speed-change exclusion models admit a variety of hydrodynamic scaling limits, often in diffusive (or hyperbolic) scaling regimes, where the empirical density field $\rho^N_t(u)$ (scaled appropriately) converges, in the limit $N \to \infty$, to the solution of a nonlinear PDE: $$\partial_t \rho = \nabla \cdot (D(\rho)\nabla \rho) \quad \text{or} \quad \partial_t \rho = \nabla \cdot J(\rho, \nabla \rho)$$ The diffusion matrix $D(\rho)$ is model-dependent and may encode slow, fast, or degenerate diffusion (Gonçalves et al., 2023, Baldasso et al., 2014). Notable features:

• Fast/slow diffusion transition: $D(\rho) = m\rho^{m-1}$ smoothly interpolates between porous medium ($m>1$) and fast diffusion ($m<1$) behavior (Gonçalves et al., 2023). The process is constructed as a linear combination of “kinetically constrained” models using generalized binomial coefficients; see also detailed construction in (Gonçalves et al., 2023) for precise generator formulas.
• Boundary phase transitions: For exclusion with slow boundaries (rate $N^{-\theta}$), the hydrodynamic limit is the heat equation with Dirichlet (if $\theta < 1$), Robin ($\theta = 1$), or Neumann ($\theta > 1$) boundary conditions, encoding how the slow/fast exchange at boundaries induces macroscopic discontinuity in profile relaxation (Baldasso et al., 2014).
• Gradient/non-gradient models: In gradient models, the microscopic current is the discrete gradient of a local function of $\eta$, simplifying hydrodynamic analysis via entropy methods. Non-gradient models require new techniques, such as two-scale expansions and homogenization theory (Gu et al., 25 Sep 2025).

3. Relaxation, Chaos Expansion, and Variance Decay

The time-dependent variance of a local function $u$ under the process semigroup $P_t$ with respect to the equilibrium measure is a signature observable: $$\operatorname{Var}\left[P_t u\right] = \mathbb{E}_{\nu_\rho}\left[(P_t u - \mathbb{E}_{\nu_\rho}[u])^2\right]$$ For the speed-change exclusion process on $\mathbb{Z}^d$, the variance decays as (Gu et al., 25 Sep 2025): $$\operatorname{Var}[P_t u] = (u'(\rho))^2 \chi(\rho)/\sqrt{(8\pi t)^d\det D(\rho)} + o(t^{-d/2})$$ Key terms:

• $u'(\rho)$: Sensitivity of $u$ to density, derivative w.r.t. $\rho$;
• $\chi(\rho) = \rho(1-\rho)$: Compressibility;
• $D(\rho)$: Homogenized diffusion matrix, determined by the (possibly degenerate, random, or configuration-dependent) jump rates.

This result extends classical relaxation results from gradient to non-gradient conservative systems, using (Gu et al., 25 Sep 2025):

• Wiener–Itô (chaos) expansion: $u = \sum_n \Pi_n u$, projecting onto chaos levels (e.g., first-order linear statistics);
• Homogenization and two-scale expansion: Approximate the non-gradient semigroup by the corresponding gradient (SEP) semigroup and correct locally via solution of a “cell problem”;
• Regularization (spatial averaging): Mixes local fluctuations, ensuring independence at large scales, facilitating application of Nash-type inequalities for variance decay.

Significance: The sharp prefactor $(u'(\rho))^2 \chi(\rho)/\sqrt{\det D(\rho)}$ describes the rate of decay of time correlations for arbitrary local observables and is a benchmark for convergence to equilibrium in such systems.

4. Coupling, Attractiveness, and Invariant Measures

Attractiveness is a strong monotonicity property implying that discrepancies between ordered initial configurations $\eta \leq \xi$ are non-increasing under the coupled dynamics. For speed-change models, necessary and sufficient conditions for attractiveness require coupled "Massey-type" inequalities balancing the configuration-dependent rates (Gobron et al., 2023). For a process with $c_{x,y}(\eta) = q(x,y) c_{\eta}(x)$, these read: $$\sum_x \eta(x) q(x,y)[c^e(x)-c^c(x)]^+ \leq \sum_x (1-\eta(x))q(x,y)c^c(x) \quad \forall y$$ (see equations (4.12), (4.13) in (Gobron et al., 2023)). The basic (SEP) coupling is not attractive for genuinely configuration-dependent rates; specialized couplings with compensatory transitions are required. Under translation-invariance and irreducibility, the only extremal invariant, translation-invariant measures are the family of Bernoulli products $\nu_\rho$ with densities $\rho \in [0,1]$.

5. Large Deviations, Fluctuations, and Random Environments

Speed-change models manifest rich large deviation phenomena. For example, in asymmetric models like TASEP, the probability of sustained macroscopic deviations from hydrodynamic behavior exhibits two “speeds” (Olla et al., 2017):

• Speed-$N$ LDs: Lower-tail deviations (e.g., blocking) achievable by local, persistent modifications;
• Speed-$N^2$ LDs: Upper-tail deviations (faster current), requiring coordinated, system-spanning manipulations.

Variational formulas using spatially inhomogeneous modifications of jump rates (Poisson clocks) characterize the rate functions for such LDs. These methods generalize to other speed-change processes, wherein the jump (mobility) function has configuration-dependence.

In dynamic random environments—when the exclusion process forms the backdrop (“environment process”) for another random walk—the speed-change exclusion process is the prototypical example of a slowly mixing, dynamically evolving background (Jara et al., 2018). The scaling limit of a random walk in such an environment decomposes as a sum of a Brownian motion and an independent Gaussian process (reflecting environmental fluctuations).

Recent advances (Conchon--Kerjan et al., 3 Sep 2024) demonstrate sharp thresholds in ballisticity for random walks in dynamic exclusion environments, with the speed $v(\rho)$ of a tagged particle being a strictly monotone function of density and transitioning sharply at a critical $\rho_c$.

6. Applications: Traffic, Jams, Condensation, and Active Matter

Speed-change exclusion processes provide mechanistic representations of phenomena such as:

• Traffic with variable speeds: Multi-speed exclusion naturally maps to tandem queueing models and zero-range processes, allowing computation of the fundamental diagram (flow-density relation) and analysis of jammed (“condensate”) phases (Furtlehner et al., 2011).
• Phase separation and clustering: Models with dynamic defects or external periodic driving forces exhibit defect-induced phase separation via velocity ordering or time-periodic “unjamming” of active matter (Das et al., 25 Mar 2025).
• Accelerated exclusion / cooperative dynamics: Distance-dependent interactions (e.g., “pushing” in RNA polymerases) lead to phase-segregated entrained states with simple current-density relations (e.g., $J = 1 - \rho$) and discontinuous transitions (Dong et al., 2012).
• Speed process (tagged particles): Asymptotic speed and velocity selection phenomena for tagged or “second-class” particles inform the joint law of individual state evolution, both in exclusion and zero-range analogues (Kumar et al., 2019, Amir et al., 2019).

7. Characteristic Model Formulas and Summary Table

Model/Class	Typical Generator Form	Key Macroscopic Behavior
SEP (SSEP/TASEP)	$$L f = \sum_{x,y} \eta(x)[1-\eta(y)](f(\eta^{x,y})-f(\eta))$$	Linear heat equation (diffusion)
Speed-change (local c(η))	$c_{x,y}(\eta) = q(x-y)c_{\eta}(x)$	Nonlinear/inhomogeneous PDE, LDs
Fast/slow diffusion	See (Gonçalves et al., 2023, Gu et al., 25 Sep 2025)	Porous medium/fast diffusion equations
Multi-speed/queueing	Sites/pattern updates, service rates variable	Condensation phase transition, jamming
Dynamic defects	Local $c_{x,y}(\eta)$ time-dependent or spatially random	Phase separation, traveling wave

Key analytical and computational tools: Wiener-Itô chaos expansion, two-scale homogenization, spatial regularization, martingale/Doob transform, entropy/hydrodynamic methods, and coupling/attractiveness constructions.

---

References:

• (Furtlehner et al., 2011): Multi-speed exclusion and queueing
• (Dong et al., 2012): Accelerated exclusion and entrainment
• (Baldasso et al., 2014): SSEP with slow boundary, hydrodynamics
• (Olla et al., 2017): Large deviations and speed-$N$, speed-$N^2$ phenomena
• (Jara et al., 2018): Random walk in dynamic speed-change environment
• (Gonçalves et al., 2023): Slow/fast diffusion transitions via speed change
• (Gobron et al., 2023): Coupling/attractiveness for speed-change exclusion
• (Conchon--Kerjan et al., 3 Sep 2024): Sharp density thresholds for ballisticity
• (Das et al., 25 Mar 2025): Persistent exclusion with time-periodic drive
• (Gu et al., 25 Sep 2025): Sharp variance decay and chaos expansion in non-gradient models

These foundational works delineate a broad and flexible toolkit for analyzing and utilizing speed-change exclusion dynamics in stochastic systems, non-equilibrium transport, large deviation theory, and mathematical physics.