Interacting Diffusion Systems at Stationarity

• Interacting diffusion systems are models where multiple species diffuse and interact, leading to complex stationary states and sharp interfaces defined by reaction kinetics and spatial constraints.
• Rigorous methods including barrier functions, spectral analysis, and variational principles establish the existence, uniqueness, and structure of stationary profiles in reaction-diffusion, cross-diffusion, and lattice models.
• These models reveal practical insights into stationary currents, phase segregation, and finite-size transport corrections with applications across biological, chemical, and physical systems.

Interacting diffusion systems at stationarity comprise a broad spectrum of models in which multiple particle species, fields, or chemical components undergo both spatial diffusion and inter-species interactions, yielding rich stationary behaviors and interface phenomena. The stationary configurations are typically determined by the interplay of diffusion, reaction kinetics, coupling mechanisms, confinement, and system geometry. Rigorous analyses for key classes—reaction-diffusion systems with singular limits, spatial segregation, cross-diffusion, lattice models, and infinite-volume interacting diffusions—unveil the universal mechanisms governing phase separation, equilibrium profiles, and stationary currents.

1. Reaction-Diffusion Systems: Fast-Reaction Segregation and Stationary Interfaces

In singularly reactive, two-component systems, stationarity is achieved via sharp segregation enforced by fast reaction limits. The archetypal system

$$\partial_t u_1 = D_1 \partial_{xx} u_1 - k R(u_1, u_2), \qquad \partial_t u_2 = D_2 \partial_{xx} u_2 + k R(u_1, u_2)$$

with $R(u_1, u_2) = u_1^{m_1} u_2^{m_2}$, $m_1, m_2 \geq 1$ and segregated initial data $u_1(0,x)\cdot u_2(0,x)=0$, exhibits a rigorous decoupling as $k \to \infty$: solutions converge to heat-equation profiles on disjoint supports separated by a stationary phase interface determined by the initial condition. The construction leverages explicitly designed barrier functions (supersolutions based on cosh/polynomial profiles), a comparison principle in weak formulation, and patching of local neighborhoods to globally control solutions near the interface. In the fast-reaction limit, the system pins the interface at the support boundary of the initial data:

• For $x$ where $u_2(0,x)=0$, $u_1$ solves the heat equation with Dirichlet condition on the interface;
• For $x$ where $u_1(0,x)=0$, $u_2$ solves its own heat equation analogously. No Stefan-type free-boundary motion occurs: the interface remains stationary, with each species instantaneously excluded from the other's domain. Mass is conserved and splits between the two sides in the stationary regime. Rigorous estimates using Hölder and $W^{2,1}_p$ bounds guarantee uniform convergence and strict segregation $u_1 \cdot u_2 \to 0$ (Tsukamoto, 11 Jun 2025).

2. Stationary States in Competitive and Cross-Diffusion Systems

Spatially extended interacting diffusions governed by competitive kinetics (e.g., Lotka–Volterra competition) and cross-diffusion exhibit stationary solutions dictated by the domain, boundary conditions, and diffusion coefficients. For multi-component systems on bounded domains,

$$\partial_t u_i = d_i \Delta u_i + f_i(u), \quad u = (u_1, ..., u_n)^\top \geq 0,$$

where $f_i(u)$ encodes competitive and possibly nonlinear reaction terms, stationarity reduces to solving the elliptic system $d_i \Delta U_i + f_i(U) = 0$ under specified boundary conditions. The existence and uniqueness of stationary solutions depend on:

• For large diffusion or convex domains with Neumann BC, only spatially constant equilibria arise (no nontrivial patterns).
• Under Dirichlet BC, spatial heterogeneity emerges if the domain exceeds explicit critical sizes given by, e.g., $L_{KISS} = \pi \sqrt{D}$ for scalar diffusion.
• For systems supporting ODE-level periodic orbits, the extended system converges to time-periodic spatially homogeneous stationary states. Stability analysis employs spectral decomposition, linearization about constant equilibria, and Floquet theory. Nontrivial interface and pattern formation requires small diffusion, nonconvex geometry, or boundary-driven bifurcations (Przeradzki, 2011).

In nonlinear cross-diffusion systems with weak (small parameter) coupling,

$u_{i,t} - \nabla \cdot [D_i (I + \delta \Phi(u)) \nabla u + (\diag(\nabla V) + \delta \Psi(u)) u] = 0,$

stationary states exist uniquely for sufficiently small $\delta$, constructed via the Implicit Function Theorem. The steady-state profiles are perturbed from the uncoupled solution $u_i^*(x) = Z_i^{-1}\exp(-V_i(x))$, and convergence to stationarity is exponentially fast in time as long as the Poincaré/fixed-point inequalities and ellipticity conditions are satisfied (Alasio et al., 2019).

3. Stationarity in Interacting Lattice and Discrete Geometries

Lattice-based models of interacting diffusions provide microscopic mechanisms underpinning macroscopic equilibrium and stationary transport laws. In systems of cavities/sites with occupation $n_i \in \{0, 1, ..., n_{\max}\}$,

• The equilibrium probability $p_n^{eq}$ is derived from the grand canonical partition function constructed from both ideal and interaction free energies.
• Transport coefficients (self-diffusion $D_s$, transport diffusion $D_t$, Maxwell–Stefan $D_{\text{ms}}$) are calculated analytically in the mean-field regime and via kinetic Monte Carlo simulations.
• Stationarity is defined via constant average occupation and fluctuation statistics at equilibrium, and the stationary diffusion coefficients trace to the equilibrium variance and thermodynamic factor $\Gamma = \langle n \rangle / \mathrm{Var}[n]$.
• The role of correlations (both single-particle memory and interparticle effects) significantly affects the stationary values; $D_t > D_s$ is typical but negative friction and $D_{\text{ms}} < D_s$ may emerge in specially engineered interaction landscapes.
• Dimensionality and system size enter explicitly via $D_s(L) = D_s(\infty) + A/L$, indicating finite-size corrections to stationarity (Becker et al., 2014).

Deterministic interacting lattice systems (e.g., particle-vacancy exchange with hard-core constraints) further admit closed-form stationary measures and transport regimes. The stationary current autocorrelation, diffusion constant $\mathcal{D}$, and Drude weight are computed exactly. The post-quench charge profile relaxes to a stationary error-function shaped profile reflecting strictly diffusive relaxation with no persistent ballistic component in the equilibrium regime (Medenjak et al., 2017).

4. Infinite-Volume Interacting Diffusions: Existence and Uniqueness of Stationary Laws

Infinite systems of interacting diffusions (on, e.g., $\mathbb Z^d$ or stratified Lie groups) require careful construction of invariant stationary measures. In models with degenerate-elliptic generators (such as the Heisenberg group operators), the stationary law is obtained as the weak limit of finite-volume invariant measures after demonstrating exponential ergodicity for finite subsystems, Lyapunov drift conditions, and tightness in appropriately weighted configuration spaces. The martingale problem approach ensures that the constructed measure $\nu$ satisfies invariance under the generator $L$:

$$\int_S L f(a) \, \nu(da) = 0, \quad \forall f \in C^2_{\text{cyl}}(S) \implies L^* \nu = 0,$$

where $S$ is the weighted configuration space. Existence (but not necessarily uniqueness) of a stationary law is rigorously established under fairly minimal assumptions; stronger uniqueness and convergence rate estimates require further hypotheses on interaction strength or system growth (Zak, 2015).

5. Stationary Distributions in Irreversible Reaction-Diffusion Processes

Stationarity in irreversible reaction-diffusion systems (coalescence, annihilation, aggregation) is characterized by exact distributions for extreme-value observables and infiltration statistics. For the initial condition of a filled half-line and empty complement, diffusion-driven infiltration yields:

• A stationary distribution $P(N)$ for the number $N$ of infiltrated particles due to the competition between reaction and diffusion;
• Explicit universal mean values, e.g., $\langle N \rangle = \frac{3}{8} + \frac{1}{2\pi}$ for coalescence, with corresponding results for annihilation via a duality mapping.
• The empty-interval and even-interval techniques solve for the stationary density profiles and joint distribution of extremal particles.
• Cluster aggregation leads to a stationary mass spectrum subject to specified sum rules and scaling, though detailed forms may remain open problems. These exact stationary results highlight nontrivial, finite infiltration and non-Gaussian statistics emerging at stationarity in one-dimensional irreversible systems (Krapivsky et al., 2015).

6. Stationary States, Interfaces, and Currents in Coupled and Multi-Species Systems

Advanced models involving cross-diffusion, phase coupling via moving interfaces, and multi-species reactions demonstrate structured stationary states determined by explicit algebraic and variational conditions:

• Coupled cross-diffusion systems with moving interfaces yield stationary piecewise-constant profiles: each phase is described by a constant vector $(c^s, c^g)$, interface position $X^*$ fixed by Butler–Volmer-type jump laws and mass conservation constraints. The stationary free-energy minimizers coincide with physical stationary states, and explicit formulas apply under distinguishability conditions for species mobilities (Cancès et al., 2024).
• Multi-species reaction-diffusion models with boundary driving can exhibit partial uphill diffusion in stationary current profiles, but in the hydrodynamic/macroscopic scaling, this uphill phenomenon vanishes—macroscopic PDEs revert to conventional diagonal diffusion operators, illustrating the inherent scale dependence of stationarity and transport behaviors (Casini et al., 2022).
• Switching interacting particle systems produce stationary density profiles violating Fick’s law: the coupled fast/slow particle layers support uphill diffusion regimes where stationary current can point up the total density gradient. The exact stationary solution is computed, including boundary-layer analysis and the precise parameter regimes producing non-Fick stationary transport (Floreani et al., 2021).

7. Stationary Laws in Interacting Particle Systems and Yaglom Limits

A class of interacting particle systems approximates quasi-stationary distributions (Yaglom limits) for diffusions with unbounded drift via boundary-resurrection rules (e.g., Fleming–Viot mechanisms). For $N$ particles each driven by a possibly noncompact SDE with absorption at the domain boundary, stationarity is realized by:

• Exponential ergodicity of the $N$-particle system for appropriate boundary-jump measures and Lyapunov conditions;
• Large $N$ (propagation-of-chaos), followed by large domain approximations $m \to \infty$, yield empirical stationary distributions converging in probability to the true quasi-stationary law of the single-particle diffusion under conditioning to avoid absorption.
• The method provides practical Monte Carlo approximation algorithms for stationary distributions relevant to population biology and stochastic processes beyond the reach of spectral theory (Villemonais, 2010).

---

Overall, the study of interacting diffusion systems at stationarity displays a wealth of emergent phenomena: sharp interfaces, strict segregation, spectral and algebraic characterization of stationary laws, scale-dependent transport, and rigorous multi-scale constructions of invariant measures. The mathematical techniques range from barrier and comparison arguments, spectral stability analyses, entropy and variational formulations, to infinite-dimensional martingale-problem approaches. Explicit parameter thresholds, functional inequalities, and algebraic minimization criteria universally govern the existence, uniqueness, and structure of stationary states throughout these diverse models.