---
title: Stochastically-Forced Diffusion Model
url: https://www.emergentmind.com/topics/stochastically-forced-diffusion-model
type: topic
---

# Stochastically-Forced Diffusion Model

Searching arXiv for the primary paper and closely related work on stochastic diffusion and stochastic forcing.
Searching for arXiv:2508.18882.
In contemporary arXiv usage, a stochastically-forced diffusion model can denote several classes of non-equilibrium diffusion theories; in the specific sense developed for non-motile active matter, it is a lattice-gas and continuum field theory in which reciprocal but randomly fluctuating pairwise interactions generate active suspensions and enhance the diffusion of a weakly coupled tracer, even in the absence of self-propulsion or non-reciprocity [2508.18882]. The central result is that stochastic pair forces minimally break detailed balance, create a \(q\)-dependent effective temperature that rises at short wavelength, and thereby open a distinct route to \(D_{\rm eff}>D_y\) in dense suspensions.

## 1. Microscopic construction

The model is formulated on a one-dimensional periodic lattice \(\mathcal L\) of spacing \(a\). At each site \(i\in\mathcal L\), the occupation number is \(n_i\in\{0,1,\dots,\kappa\}\), where \(\kappa\) is the site carrying capacity. Two classes of moves are allowed. The first is a single-particle hop, i.e. a partial-exclusion process, in which a particle at \(i\) hops to \(i\pm1\) with rate
\[
\alpha_{i,\pm1}(n)=D_r\,n_i\Bigl(1-\frac{n_{i\pm1}}{\kappa}\Bigr),
\]
with \(D_r^{-1}\) setting the microscopic diffusive timescale. The second is a pairwise “attractive” hop in which two particles on \(\{i-1,i+1\}\) simultaneously hop to \(i\) with rate
\[
\beta_i(n)=k_r\,n_{i-1}\,n_{i+1}\,\max\!\Bigl(0,1-\frac{n_i}{\kappa-1}\Bigr).
\]

The second rule is the source of activity. It breaks detailed balance, drives an effective attraction, and ultimately phase separation, yet preserves reciprocity and center-of-mass. This combination is the defining feature of the construction: the system is non-equilibrium not because of self-propelled particles or non-reciprocal forces, but because the pairwise dynamics inject stochasticity while retaining reciprocal interactions [2508.18882].

## 2. Coarse-grained dynamical theory

The coarse-grained description starts from the master equation, introduces the Martin–Siggia–Rose action, expands \(n_{i\pm1}=n_i\pm a\,\nabla n +\tfrac{a^2}{2}\nabla^2n+\cdots\), and keeps only leading gradient terms. In the continuum limit, fixing \(a=1\) but restricting to long wavelengths, the density field \(n(x,t)\) obeys
\[
\partial_t n
= \partial_x\Bigl[D_r\,n\bigl(1-\tfrac{n}{\kappa}\bigr)
-\,k_r\,n^2\bigl(1-\tfrac{n}{\kappa-1}\bigr)\Bigr]\partial_x n
+\partial_x\!\bigl[\sqrt{2D_r\,n(1-n/\kappa)}\,\Lambda\bigr]
+\partial_x^2\!\bigl[\sqrt{k_r\,n^2(1-n/(\kappa-1))}\,\eta\bigr].
\]

Linearization about a uniform average \(\bar n\) proceeds by setting \(\phi(x,t)=n(x,t)-\bar n\). The resulting Langevin equation is
\[
\partial_t\phi
= D_0\,\partial_x^2\phi
-\gamma\,\partial_x^4\phi
+\sqrt{2D_1}\,\partial_x\Lambda
+\sqrt{A}\,\partial_x^2\eta,
\]
with coefficients
\[
D_0=D_r-k_r\bigl(2\bar n-3\bar n^2/(\kappa-1)\bigr),
\]
\[
D_1=D_r\,\bar n\bigl(1-\bar n/\kappa\bigr),
\]
\[
A=k_r\,\bar n^2\bigl(1-\bar n/(\kappa-1)\bigr),
\]
\[
\gamma=\tfrac{1}{12}[\,D_r + k_r(14\bar n-15\bar n^2/(\kappa-1))\,].
\]
The noises \(\Lambda\) and \(\eta\) are independent, zero-mean, unit-variance white noises. The two-noise structure is essential: the \(\partial_x\Lambda\) term is equilibrium-like, whereas the \(\partial_x^2\eta\) term encodes the non-equilibrium stochastic pair forcing [2508.18882].

## 3. Effective temperature and short-wavelength amplification

In Fourier space, with \(\phi(q,t)=\int dx\,e^{-iqx}\phi(x,t)\), the two noise sources combine into an effective-temperature spectrum
\[
\mu_\phi\,T_{\rm eff}(q)=D_1+\tfrac{A}{2}\,q^2,
\]
with \(\mu_\phi=1\). Equivalently,
\[
T_{\rm eff}(q)
=
\frac{D_r\,\bar n\bigl(1-\tfrac{\bar n}{\kappa}\bigr)
+\tfrac12\,k_r\,\bar n^2\bigl(1-\tfrac{\bar n}{\kappa-1}\bigr)\,q^2}
{\mu_\phi}.
\]

Because of the \(\propto q^2\) term, the effective temperature grows at short wavelength. In the model’s interpretation, this growth reflects the nonequilibrium pairwise noise \(\sqrt A\,\partial_x^2\eta\). The effective temperature is therefore not a flat equilibrium control parameter; it is a mode-dependent spectrum. This distinction is central to the ensuing tracer dynamics, because the enhanced short-scale agitation can outweigh the usual fluctuation-induced slowdown and reverse the sign of the correction to diffusion [2508.18882].

## 4. Tracer coupling and renormalized diffusion

A tracer at position \(y(t)\), with bare mobility \(\kappa_y\), is weakly coupled to the fluctuating density field through
\[
\dot y=h\,\kappa_y\,\partial_x\phi\bigl|_{x=y}+ \sqrt{2\kappa_yT}\,\xi_y,
\qquad h\ll1.
\]
Integrating out \(\phi\) perturbatively by a Dean–Demery path-integral gives, at zero external force, the self-diffusion coefficient
\[
D_{\rm eff}
= D_y
-\frac{h^2}{d}\,\int\!\frac{d^dq}{(2\pi)^d}\;
\frac{|q|^2\,|\tilde K(q)|^2\,
\bigl[D_y\,\mu_\phi\,T_{\rm eff}(q)+(2D_y-\mu_\phi T_{\rm eff}(q))\,\Delta(q)\bigr]}
{\Delta(q)\,\bigl[D_y+\Delta(q)\bigr]^2},
\]
where \(D_y=\kappa_yT\) is the tracer’s bare diffusion, \(\Delta(q)=D_0+\gamma\,q^2\) is the field’s deterministic relaxation rate, and \(\tilde K(q)\) is the coupling form-factor, taken here as \(\tilde K=1\). In one dimension, with \(d=1\) and \(\kappa_y=\mu_\phi=1\), this reduces to Eq. (30) in the supplementary material.

The equilibrium comparison is explicit. In equilibrium Model B, where \(A=0\) and \(T_{\rm eff}\equiv D_1\), the fluctuation-dissipation theorem holds and the correction is negative, so \(D_{\rm eff}<D_y\). The same occurs when the stochastic-force amplitude vanishes, \(A\to0\): the \(q^2\) growth in \(T_{\rm eff}\) disappears, equilibrium-like noise is recovered, and fluctuations slow the tracer. In the dense-suspension limit \(\bar n\to\kappa\) (or \(\kappa-1\)), both \(D_1\to0\) and \(A\to0\); the field becomes quasi-frozen, \(T_{\rm eff}\to0\), and \(D_{\rm eff}\to D_y\). By contrast, intermediate densities with \(A>0\) generate high-\(q\) noise and can reverse the sign of the integral, yielding \(D_{\rm eff}>D_y\). This is the article’s precise sense of enhanced diffusion [2508.18882].

## 5. Relation to other stochastic diffusion frameworks

The phrase “stochastically-forced diffusion model” is not unique to non-motile active matter. In the literature represented here, it spans several distinct constructions. Donev et al. derive a stochastic advection-diffusion equation for tracer concentration from fluctuating Stokes dynamics in the large-Schmidt-number limit, with a divergence-free white-in-time random velocity \(w(r,t)\), and show that the ensemble mean obeys Fick’s law with \(\chi_{\rm eff}=\chi_0 I+\chi\) while individual realizations exhibit giant fluctuations [1306.3158]. The “Multinomial Diffusion Equation” is instead a microscopic mass-conserving discrete-particle model on a lattice whose continuum limit reproduces the classical stochastic diffusion PDE, but which remains faithful at low particle density where the classical SDE fails [1010.0719].

A different family places the stochasticity in the diffusivity itself rather than in pairwise forces or fluctuating fluxes. The generalized grey Brownian motion and diffusing-diffusivity models both yield strictly linear growth of the MSD while allowing non-Gaussian displacement statistics; in the DD model, the propagator is non-Gaussian at short times and crosses over to Gaussian at long times [1811.09531]. A bounded diffusing-diffusivity model driven by symmetric dichotomous noise likewise produces a short-time PDF with a logarithmic divergence at the origin, Gaussian tails modulated by a power law, and ordinary Gaussian diffusion at long times [2604.11800]. These comparisons clarify that the non-motile active-matter construction is not a random-diffusivity model in disguise; its stochasticity is carried by reciprocal pair interactions and their induced density-field noise.

## 6. Significance, misconceptions, and scope

A common misconception is that enhanced diffusion in active suspensions requires self-propulsion, non-reciprocity, or explicit motility. The non-motile active-matter model excludes that identification: reciprocal but randomly fluctuating interactions already suffice to generate active suspensions and enhance tracer diffusion. Another frequent misconception is to treat \(T_{\rm eff}(q)\) as a thermodynamic temperature. Here it is explicitly \(q\)-dependent, and the short-wavelength increase is the diagnostic of the nonequilibrium pairwise forcing rather than an equilibrium fluctuation-dissipation parameter [2508.18882].

The scope of the construction is equally specific. It is derived from a one-dimensional periodic lattice gas, coarse-grained in a long-wavelength limit, linearized about a uniform average density, and evaluated for a weakly coupled tracer. Within that regime, the model yields a two-noise continuum field theory, a short-scale elevated effective temperature, and a renormalized tracer diffusivity whose correction changes sign when \(A\) is large enough. The authors summarize the outcome as a generic route to enhanced diffusion via purely stochastic, center-of-mass-conserving forces. A plausible implication is that stochastic pair forcing should be regarded as an independent mechanism of transport renormalization in dense non-equilibrium suspensions, rather than as a secondary correction to more familiar motility-based activity.

Source: https://www.emergentmind.com/topics/stochastically-forced-diffusion-model