betaFeedback

WorksPhysics and Astronomy

Curvature-Driven Density and Fluctuations in Interacting AOUPs under Confinement

Abstract

We investigate interacting active Ornstein-Uhlenbeck particles (AOUPs) confined in harmonic and double-well potentials, combining mean-field theory, analytical approximations, and numerical simulations. We derive the stationary density, position and velocity correlations, correlation length, dynamical structure factor, and leading-order entropy production. Unlike established harmonic normal-mode descriptions, our framework incorporates steric interactions, persistent forcing, and spatially varying confinement curvature through the eigenvalues and eigenvectors of the many-body Hessian. Within a locally homogeneous approximation, the correlation length is controlled by persistence and local curvature, increasing as the effective restoring stiffness decreases and formally diverging at the instability of the linearized description. Entropy production, at leading nontrivial order in persistence, is governed by gradients of the effective curvature. Harmonic confinement provides a constant-curvature reference, whereas double-well confinement creates equilibrium-like regions near the minima and strongly nonequilibrium regions near the barrier. The barrier exhibits enhanced velocity and density fluctuations, soft collective modes, long-lived correlations, and increased entropy production. Persistence amplifies these effects, while steric crowding suppresses collective fluctuations. The dynamical structure factor reveals enhanced low-frequency contributions from soft modes. Our results establish how confinement geometry, active persistence, and interactions jointly regulate collective fluctuations and irreversibility in active matter.

Introduction

Active matter systems are intrinsically out of equilibrium because their microscopic constituents continuously consume energy and convert it into self-propelled, temporally persistent motion. We emphasise at the outset that “active” here refers to this persistence and not to a net drift: the self-propulsion force has zero mean, and it is the finite correlation time of the propulsion — equivalently the coloured, non-thermal character of the

driving — that breaks detailed balance and violates the fluctuation-dissipation relation. This persistent energy injection breaks detailed balance and generates a wide range of collective phenomena, including flocking, clustering, giant number fluctuations, pattern formation, defect dynamics, anomalous transport, and collective flows in biological and synthetic systems [3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14]. A central goal of nonequilibrium statistical physics is to understand how microscopic activity modifies macroscopic observables such as density profiles, transport coefficients, fluctuations, correlations, collective modes, and entropy production.

Interacting active particles are of particular interest because interactions and persistence together produce collective behavior with no direct equilibrium analogue [4, 7, 15, 16, 17, 18, 19, 20]. A paradigmatic example is motility-induced phase separation, where purely repulsive active particles undergo phase separation in the absence of attractive forces [21, 22, 23, 24, 25]. Considerable progress has been made using hydrodynamic theories, coarse-grained descriptions, and perturbative approaches [26, 27, 28, 29, 30, 31]. Nevertheless, an exact analytical description of interacting active matter remains difficult because activity introduces memory, nonlocal correlations, and irreversible probability currents. Exact or controlled results are therefore mostly available in special limits, such as two-particle systems [32, 33, 34, 35, 36, 37, 38, 39, 40, 41], harmonic chains [42, 43, 44, 45], or specially constructed lattice models [46, 47, 48, 49, 50, 51, 52].

A particularly important and less understood regime is confined active matter, where external potentials break translational invariance and impose spatially varying curvature [53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70]. In such systems, transport, escape, fluctuations, and steady-state distributions are controlled not only by activity and interactions, but also by the local geometry of the trap. The interplay between persistence, crowding, and curvature can produce strongly heterogeneous nonequilibrium behavior that is absent in translationally invariant systems. Although confined active particles have been studied extensively, analytical theories for interacting confined systems remain limited because the dynamics depend sensitively on both many-body correlations and position-dependent curvature. Recent theoretical developments have begun to address this problem systematically [71, 72, 73].

The double-well potential provides a natural setting to study these effects because it contains regions with qualitatively different local geometry within the same system. Near the well minima, the curvature is large and positive, and the dynamics is locally close to that in a harmonic trap. Near the barrier, the curvature becomes small or negative, restoring forces weaken, and persistent active motion can dominate relaxation. This spatial variation allows one to ask how activity reorganizes density, transport, fluctuations, correlations, and dissipation across different regions of a confined system. Caprini et al. [74] and more recently Parashar et. al. [75] demonstrated that the dynamics of an active particle in a double-well potential are strongly controlled by the confining landscape. Near the minima of the wells, termed Equilibrium Regions (ERs), the particle behaves approximately as an equilibrium system: the velocity distribution is nearly Gaussian, fluctuations are relatively small, mobility remains positive, and entropy production is weak. In contrast, near the barrier, where the curvature becomes small or negative, persistence dominates over the restoring force, leading to a qualitatively different dynamical regime. In this Nonequilibrium Region (NeR), the velocity distribution becomes bimodal, the mobility can become negative, entropy production is strongly enhanced, and barrier-crossing events are facilitated by persistent active motion. The coexistence of ER and NeR within the same potential illustrates how confinement geometry can spatially organize nonequilibrium behavior.

Motivated by the emergence of equilibrium-like and nonequilibrium regions in confined active systems, we develop a theoretical framework for interacting active Ornstein-Uhlenbeck particles (AOUPs) under confinement [26, 76, 77, 78, 79]. Using the perturbative steady-state approach of Martin et al. [26], we derive a mean-field description for the single-particle distribution starting from the microscopic Langevin dynamics. Under the zero-current approximation, the active dynamics can be mapped onto an effective passive system characterized by a position-dependent diffusion coefficient, effective force, and effective potential. This framework allows us to obtain steady-state densities, transport coefficients, fluctuations, correlations, structure factors, and entropy production for interacting AOUPs in harmonic and double-well confinements.

We show that confinement geometry along with particle number and activity plays a central role in organizing nonequilibrium behavior. While harmonic traps produce smooth, nearly equilibrium-like density and fluctuation profiles, double-well confinement generates pronounced spatial heterogeneity. Persistence and particle number through excluded-volume interactions reshape the density distribution and renormalize the effective transport properties. By linearizing the dynamics around the steady-state configuration, we derive analytical expressions for velocity and position correlations in terms of the Hessian spectrum of the effective many-body potential. The local curvature controls the crossover between equilibrium-like regions (ERs), where stiff modes suppress fluctuations and correlations, and nonequilibrium regions (NeRs), where soft barrier modes amplify fluctuations, increase correlation lengths, and promote collective barrier-crossing events. These soft modes also dominate the dynamical structure factor, producing enhanced low-frequency spectral weight and persistence-induced memory effects. Increasing the particle number increases the stiffness of the Hessian eigenmodes, leading to reduced fluctuations. Finally, we show that entropy production is directly governed by spatial variations of the effective curvature, leading to weak dissipation in harmonic confinement but strong localized irreversibility near the barrier and inflection regions of the double-well potential. We investigate how confinement geometry spatially organizes nonequilibrium behavior: stiff, positively curved regions behave approximately as equilibrium-like regions, whereas soft barrier regions act as nonequilibrium regions with enhanced fluctuations, long-lived correlations, collective crossings, and localized entropy production. This establishes curvature and particle number as a central organizing principle for interacting active matter under confinement.

We now state explicitly the novelty of the present work and the gap that it addresses. Correlation functions for a harmonically confined degree of freedom driven by exponentially correlated active noise are well established [1, 2]. In particular, Samanta and Chakrabarti extended the harmonic active-noise calculation to the normal modes of Rouse, Zimm, and semiflexible polymer chains [2], where each mode obeys an effective linear harmonic equation. The spatial organization of equilibrium-like and nonequilibrium regions in a double-well potential has also been studied for a single active particle [74]. Analytical treatments of interacting active systems have therefore covered several important limiting cases, including single-particle confinement, linear or harmonic normal-mode systems, translationally invariant interacting systems, and few-particle models. Much less explored is the regime in which excluded-volume interactions, persistent active forcing, and a spatially varying external curvature act simultaneously. The present work addresses this regime and develops the following analytical results within the mean-field, small-persistence, and linearized approximations specified below.

  • Mean-field theory for interacting AOUPs in a general external confinement. Starting from the many-particle dynamics, we derive a closed single-particle density equation from the BBGKY hierarchy using a two-body factorisation closure, Eq. (4). This leads to an effective description in terms of the position-dependent diffusivity Deff(x)D_{\mathrm{eff}}(x), the effective force F(x)F(x), and the effective potential Ueff(x)U_{\mathrm{eff}}(x), Eqs. (5)–(6), (8), and (12). In these expressions, interparticle interactions and the spatial structure of the external confinement enter simultaneously. The corresponding non-interacting UCNA expression is recovered as the limiting case V→0V \to0, Eq. (10).

  • Collective velocity and position correlations from the interacting many-body Hessian. The velocity and position correlations of the confined interacting system, Eqs. (20)–(22), are expressed in terms of the eigenvalues λα\lambda_\alpha and eigenvectors ψα\psi_\alpha of the many-body Hessian, Eq. (19). This should be distinguished from the earlier linear harmonic-mode calculations [1, 2]. In those problems, the relevant degrees of freedom can be represented by independent harmonic modes with prescribed mode stiffnesses. In the present system, by contrast, the fluctuation operator is configuration dependent and contains both the local curvature of the external potential and the stiffness generated by interparticle interactions, Hij=Φ′′(xi∗)δij+∂xi∂xjV(xi∗−xj∗)H_{ij} = \Phi^{\prime\prime}(x_i^*)\delta_{ij} + \partial_{x_i}\partial_{x_j}V(x_i^* - x_j^*). The resulting spectrum therefore depends simultaneously on confinement geometry and steric coupling. The corresponding harmonic linear-mode expressions are recovered in the appropriate limiting case and are used here as a consistency check rather than as a new result. The eigenvectors ψα\psi_\alpha additionally provide the spatial structure of the collective modes and allow regions of enhanced fluctuations and soft

collective response to be identified inside a spatially nonuniform confinement.

  • A position-dependent correlation length controlled by local confinement curvature. Within a locally homogeneous approximation, λ(q;x∗)≃Φ′′(x∗)+Kq2\lambda(q;x^*) \simeq\Phi^{\prime\prime}(x^*) + Kq^2, the correlation function defines the local correlation length ξ(x∗)=τK/(γ+τΦ′′(x∗))\xi(x^*) = \sqrt{\tau K/(\gamma+ \tau\Phi^{\prime\prime}(x^*))}, showing explicitly that the spatial range of collective active fluctuations is controlled jointly by persistence and the local curvature of the confinement. As the local curvature decreases, ξ(x∗)\xi(x^*) increases. A formal divergence occurs when γ+τΦ′′(x∗)→0+\gamma+ \tau\Phi^{\prime\prime}(x^*) \to0^+, corresponding to the loss of stability of the locally linearized branch. For a harmonic external potential, Φ′′(x)=k\Phi^{\prime\prime}(x) = k is spatially constant, and therefore no position dependence of ξ\xi originates from the external confinement itself.

  • Entropy production associated with spatial variation of the effective curvature. Within the stated small-persistence approximation, the leading nontrivial contribution to the entropy production is S˙tot∝∫dx ρ(x)[∂xφeff(x)]2\dot{S}_{\mathrm{tot}} \propto\int dx\,\rho(x)\left[\partial_x\varphi_{\mathrm{eff}}(x)\right]^2 which relates irreversibly to the spatial variation of the effective curvature. A non-interacting harmonic potential has constant curvature and therefore gives no contribution at the order retained here. Interactions can nevertheless generate a finite contribution through the interaction-dependent part of φeff\varphi_{\mathrm{eff}}. For a nonlinear and spatially nonuniform confinement such as the double well, the external potential itself also contributes because its curvature varies with position.

  • Competition between persistence and crowding. Within the parameter regime investigated, increasing the persistence time τ\tau enhances the contribution of soft collective modes, particularly those associated with the barrier region, and increases the spatial heterogeneity of the nonequilibrium response. Increasing particle number or steric interaction strength, on the other hand, generally stiffens the relevant collective modes and suppresses the corresponding fluctuations. Persistence, confinement curvature, and crowding therefore act as competing controls of the collective nonequilibrium response. The simulations are used to test and illustrate these analytical predictions rather than serving as the sole basis for the conclusions.

We also clarify the role of the harmonic trap in the present study. The harmonic problem and related linear normal-mode systems are well studied [1, 2], and we do not present the harmonic trap itself as a new active-matter problem. Instead, it is used as a constant-external-curvature reference system, for which Φ′′(x)=k\Phi''(x)=k, and Φ′′′(x)=0\Phi'''(x)=0 everywhere. Consequently, the external potential introduces no spatial variation of curvature. For an interacting system, however, the many-body Hessian can still possess a nontrivial spectrum because of the interaction contribution. The harmonic confinement therefore provides a useful reference for separating interaction-induced collective effects from effects generated specifically by a spatially varying external potential.

The double-well potential provides the complementary case in which the external curvature changes strongly with position. It is positive and relatively large around the well minima, decreases on approaching the barrier, and becomes negative in the barrier region. Consequently, different parts of the same confined system sample locally stiff, soft, and locally unstable curvatures. Comparing harmonic and double-well confinements at fixed persistence, interaction strength, and particle number therefore provides a controlled way of identifying the additional effects produced by spatially nonuniform confinement and soft barrier modes. This comparison motivates our central analysis of how local confinement geometry, persistent active forcing, and crowding jointly control density profiles, fluctuations, collective correlations, soft modes, and irreversibility.

The paper is organized as follows. In Section II, we introduce the Langevin dynamics of interacting AOUPs and derive the corresponding mean-field Fokker-Planck equation. In Section III, we obtain the steady-state density, effective diffusion constant, and effective force for harmonic and double-well confinements. In Section IV, we analyze velocity and position fluctuations by linearizing the dynamics around the steady-state particle positions. In Section V, we derive the static and dynamical structure factors and compare passive and active responses. In Section VI, we compute the entropy production rate and relate it to curvature variations, particle distributions, and nonequilibrium fluctuations.

Dynamics of AOUP and Passive Particles in a Confining Potential

Consider a one-dimensional system of NN AOUPs interacting through a potential V(xi−xj)V(x_i-x_j) where {xi}i=1,2,…,N\{x_i\}_{i=1,2,\ldots,N} are the positions of the particles. They diffuse in a confining potential Φ(xi)\Phi(x_i). Throughout this section Φ(x)\Phi(x) is an arbitrary confining potential; the harmonic and double-well forms, Eqs. (13) and (14), are specified only from Sec. III onwards. Their dynamics is governed by

γx˙i=−Φ′(xi)+Fi+ηi.(1)\gamma\dot{x}_i=-\Phi'(x_i)+F_i+\eta_i. \tag*{(1)}

ηi\eta_i is an Ornstein-Uhlenbeck process: ⟨ηi(t)ηj(t′)⟩=v022exp⁡(−∣t−t′∣/τ)\langle\eta_i(t)\eta_j(t')\rangle=\frac{v_0^2}{2}\exp(-|t-t'|/\tau) with persistence time τ\tau and variance v0v_0. The active diffusion constant is given by D=v02τD=v_0^2\tau. γ\gamma is the friction coefficient. The time evolution of the Ornstein-Uhlenbeck noise can also be written as

η˙i=−1τηi+2Dτξi,(2)\dot{\eta}_i=-\frac{1}{\tau}\eta_i+\frac{\sqrt{2D}}{\tau}\xi_i, \tag*{(2)}

where ⟨ξi(t)ξj(t′)⟩=δijδ(t−t′)\langle\xi_i(t)\xi_j(t')\rangle=\delta_{ij}\delta(t-t').

Here ηi\eta_i has zero mean by construction, ⟨ηi⟩=0\langle\eta_i\rangle=0, while its finite correlation time τ\tau introduces persistent nonequilibrium forcing. A vanishing mean particle velocity is not, by itself, a signature of passivity: in a confined stationary state, standard active-particle models such as active Brownian particles and run-and-tumble particles also satisfy ⟨x˙i⟩=0\langle\dot{x}_i\rangle=0, since there is no sustained translational current in the confining direction. The term “active” therefore refers here not to a nonzero mean drift, but to the presence of a persistent self-propulsion whose statistics are not fixed by the equilibrium fluctuation–dissipation relation [1, 20]. With our convention, ⟨ηi(t)ηj(t′)⟩=(v02/2)δijexp⁡[−∣t−t′∣/τ]\langle\eta_i(t)\eta_j(t')\rangle=(v_0^2/2)\delta_{ij}\exp[-|t-t'|/\tau], the parameter v0v_0 sets the root-mean-square amplitude of the active propulsion rather than a mean drift velocity. For generic nonlinear confinement or interacting systems, this persistent driving produces a nonequilibrium steady state. The harmonic potential is an important special case: because its curvature is constant, the AOUP dynamics can remain equilibrium-like and the leading small-τ\tau entropy production vanishes. Nonlinear confinement, such as the double-well potential considered here, removes this special property and allows persistent driving to generate spatially heterogeneous nonequilibrium behaviour. In the white-noise limit τ→0\tau\to0 at fixed active diffusivity DD, the Ornstein–Uhlenbeck propulsion reduces to an effective thermal-like white noise and the passive Brownian limit is recovered.

Thermal fluctuations can be included by adding an independent Gaussian white-noise term γ2DTξi(t)\gamma\sqrt{2D_T}\xi_i(t) to (1), with ⟨ξi(t)ξj(t′)⟩=δijδ(t−t′)\langle\xi_i(t)\xi_j(t')\rangle=\delta_{ij}\delta(t-t') with DT=kBTγD_T=\frac{k_BT}{\gamma}. At the level of the coarse-grained density equation, this produces the usual translational-diffusion term DT∂x2ρD_T\partial_x^2\rho and therefore adds a thermal contribution to the effective diffusivity. Thermal noise also contributes the standard equilibrium fluctuations; for example, in the harmonic approximation the equal-time positional variance of a normal mode α\alpha acquires the equilibrium contribution kBT/λαk_BT/\lambda_\alpha. We set DT=0D_T=0 here in order to isolate the effects of persistent active forcing. Since thermal noise satisfies the equilibrium fluctuation–dissipation relation, it is not itself a source of nonequilibrium driving. Its inclusion would primarily smooth the steady-state density profiles and reduce the contrast of activity-induced spatial heterogeneities. The results presented below should therefore be understood primarily in the activity-dominated regime.

Fi=−∂xi(∑j<iV(xi−xj))F_i=-\partial_{x_i}\left(\sum_{j<i}V(x_i-x_j)\right) is a conservative force due to the interactions between the particles. Following [80], we rewrite the entire velocity term on the right-hand side of Eq. (1) as vi=1γ[−Φ′(xi)+Fi+ηi]v_i=\frac{1}{\gamma}[-\Phi'(x_i)+F_i+\eta_i], which has a conservative part originating from the external or interaction potential and a dissipative part due to selfpropulsion. Eq. (1) and Eq. (2) become

x˙i=vi\dot{x}_i = v_i
v˙i=−∑j[1τδij+1γΦ′′(xi)δij+1γ∂xixjV(xi−xj)]vj−1γτΦ′(xi)+1γτFi+2Dτξi.(3)\dot{v}_i = -\sum_j\left[\frac{1}{\tau}\delta_{ij}+\frac{1}{\gamma}\Phi^{\prime\prime}(x_i)\delta_{ij}+\frac{1}{\gamma}\partial_{x_i x_j}V(x_i-x_j)\right]v_j-\frac{1}{\gamma\tau}\Phi^\prime(x_i)+\frac{1}{\gamma\tau}F_i+\frac{\sqrt{2D}}{\tau}\xi_i. \tag*{(3)}

Integrating the multiparticle Fokker-Planck equation over the positions and velocities of all particles except one in Appendix § 9, we obtain the single particle probability distribution in the mean-field approximation

P˙1(x,v,t)=−∂x[vP1(x,v,t)]+∂v[1τvP1(x,v,t)]+∂v[∫dx′dv′1γ∂xx′V(x−x′)v′P1(x,v,t)P1(x′,v′,t)]+∂v[1γ∂x2Φ(x)P1(x,v,t)]+1γτ∂xΦ(x)∂vP1(x,v,t)+1γτ∂vP1(x,v,t)×∫dx′dv′∂xV(x−x′)P1(x′,v′,t)+Dτ2∂v2P1(x,v,t).(4)\dot{P}_1(x,v,t)=-\partial_x[vP_1(x,v,t)]+\partial_v\left[\frac{1}{\tau}vP_1(x,v,t)\right]+\partial_v\left[\int dx^\prime dv^\prime\frac{1}{\gamma}\partial_{xx^\prime}V(x-x^\prime)v^\prime P_1(x,v,t)P_1(x^\prime,v^\prime,t)\right]+\partial_v\left[\frac{1}{\gamma}\partial_x^2\Phi(x)P_1(x,v,t)\right]+\frac{1}{\gamma\tau}\partial_x\Phi(x)\partial_vP_1(x,v,t)+\frac{1}{\gamma\tau}\partial_vP_1(x,v,t)\times\int dx^\prime dv^\prime\partial_xV(x-x^\prime)P_1(x^\prime,v^\prime,t)+\frac{D}{\tau^2}\partial_v^2P_1(x,v,t). \tag*{(4)}

The evolution equation for the density of particles ρ(x,t)=∫dvP1(x,v,t)\rho(x,t)=\int dvP_1(x,v,t) is

∂tρ(x,t)=∂xj(x)=∂x[Deff(x)∂xρ(x,t)+F(x)ρ(x,t)].(5)\partial_t\rho(x,t)=\partial_xj(x)=\partial_x\left[D_{\mathit{eff}}(x)\partial_x\rho(x,t)+\mathcal{F}(x)\rho(x,t)\right]. \tag*{(5)}

The detailed derivation is provided in the Appendix § 9. We have defined

Deff(x)=D(1+τγΦ′′(x))Γ(x),(6)D_{\mathit{eff}}(x)=\frac{D}{\left(1+\frac{\tau}{\gamma}\Phi^{\prime\prime}(x)\right)\Gamma(x)}, \tag*{(6)}

and

F(x)=1Γ(x)[1γΦ′(x)+1γ∫dx′ρ(x′)∂xV(x−x′)+D∂x[11+τγΦ′′(x)]].(7)\mathcal{F}(x)=\frac{1}{\Gamma(x)}\left[\frac{1}{\gamma}\Phi^\prime(x)+\frac{1}{\gamma}\int dx^\prime\rho(x^\prime)\partial_xV(x-x^\prime)+D\partial_x\left[\frac{1}{1+\frac{\tau}{\gamma}\Phi^{\prime\prime}(x)}\right]\right]. \tag*{(7)}

It is important to state explicitly which approximations enter Eqs. (4)–(6), and in what sense the description is a mean-field one. (a) Closure of the BBGKY hierarchy. The exact first hierarchy equation, Eq. (35), couples P1P_1 to the two-body distribution P2P_2. We close it by the factorisation

P2(x,v,x′,v′,t)=P1(x,v,t)P1(x′,v′,t),P_2(x,v,x^\prime,v^\prime,t)=P_1(x,v,t)P_1(x^\prime,v^\prime,t),

i.e. we neglect two-body correlations in the interaction term and replace the force on a tagged particle by the force generated by the average one-body density of the remaining particles. This is the only sense in which the term “mean field” is used in the present work. The closure is expected to be more reliable when correlations are weak, when each particle samples many neighbours, or when the density varies slowly over the interaction range. It is not a controlled approximation for strongly correlated, dilute, or strongly ordered configurations. Since the present interactions are short-ranged and steric, we therefore regard this factorisation as an approximation whose accuracy must ultimately be assessed against the microscopic dynamics. (b) Truncation of the velocity-moment hierarchy. Equations (A9)–(A10) for jj and VV involve the third moment χ\chi. We assume that V(x,t)V(x,t) varies slowly on time scales long compared with τ\tau and that ∂xχ≃0\partial_x\chi\simeq0. This is the standard adiabatic (local-equilibrium) truncation, valid when the velocity distribution relaxes rapidly compared with the density [26]. (c) Neglect of the current gradient. In deriving Eq. (41) we neglect ∂xj\partial_xj, which is justified in and near the steady state when the current is small and slowly varying. In Sec. III we impose j=0j=0 exactly for the steady-state solutions. (d) Small-persistence expansion. Equations (6) and (27) are obtained to leading non-trivial order in τ\tau, with corrections of O(τ3)O(\tau^3). The corresponding coarse-grained expressions are therefore controlled only when the dimensionless persistence correction, for example τΦ′′(x)/γ\tau\Phi^{\prime\prime}(x)/\gamma, remains sufficiently small over the region in which the approximation is applied. (e) Smooth-density approximation. The approximation ρ′′′′(x)≈0\rho^{\prime\prime\prime\prime}(x)\approx0 is used only in the simplified estimate of Deff(x)D_{\mathrm{eff}}(x) quoted below Eq. (16); it is not used in generating the figures.

A further condition must be satisfied for the effective passive description to remain physically meaningful. From Eq. (6), Deff(x)=D(1+τγΦ′′(x))Γ(x)D_{\mathrm{eff}}(x)=\frac{D}{\left(1+\frac{\tau}{\gamma}\Phi^{\prime\prime}(x)\right)\Gamma(x)}, and we require 1+τγΦ′′(x)>01+\frac{\tau}{\gamma}\Phi^{\prime\prime}(x)>0, and Γ(x)>0\Gamma(x)>0, throughout the spatial region sampled by the particles.

The steady-state solution in the zero current approximation j=0j=0 in Eq. (41) has the following solution

ρ(x)=Nexp⁡(−1Dγ[Φ(x)+τ2γ(Φ′(x))2−Dγln⁡(1+τγΦ′′(x))+∫dx′V(x−x′)ρ(x′)+τγ∫Φ′′(x′)∫xV′(x′−x′′)ρ(x′′)]).(8)\rho(x)=\mathcal{N}\exp\left(-\frac{1}{D\gamma}\left[\Phi(x)+\frac{\tau}{2\gamma}\left(\Phi^\prime(x)\right)^2-D\gamma\ln\left(1+\frac{\tau}{\gamma}\Phi^{\prime\prime}(x)\right)+\int dx^\prime V(x-x^\prime)\rho(x^\prime)+\frac{\tau}{\gamma}\int\Phi^{\prime\prime}(x^\prime)\int^x V^\prime(x^\prime-x^{\prime\prime})\rho(x^{\prime\prime})\right]\right). \tag*{(8)}

N\mathcal{N} is the normalization factor determined from ∫dxρ(x)=N\int dx\rho(x)=\mathcal{N}.

The right-hand side of the equation is the effective potential obtained within the UCNA approximation in [72, 81].

Eqn (5) can be mapped to a passive system [27] with dynamics

x˙i=Feff(xi)+2Deff(xi)ϵi(t).(9)\dot{x}_i=F_{\mathrm{eff}}(x_i)+\sqrt{2D_{\mathrm{eff}}(x_i)}\epsilon_i(t). \tag*{(9)}

In the non-interacting case setting V=0V=0 in Eq. (8) immediately leads to

ln⁡ρ(x)=−1Dγ[Φ+τ2γ(Φ′)2−Dγln⁡[1+τγΦ′′]]+const.(10)\ln\rho(x)=-\frac{1}{D\gamma}\left[\Phi+\frac{\tau}{2\gamma}(\Phi^\prime)^2-D\gamma\ln\left[1+\frac{\tau}{\gamma}\Phi^{\prime\prime}\right]\right]+\mathrm{const}. \tag*{(10)}

{ϵi}\{\epsilon_i\} are Gaussian white noises with unit variances. The effective force acting on each particle is

Feff(x)=−F(x)+∂xDeff(x)(11)F_{\mathrm{eff}}(x)=-\mathcal{F}(x)+\partial_xD_{\mathrm{eff}}(x) \tag*{(11)}

The effective potential experienced by individual particles can be defined by the term in the exponent of the density in Eq. (8)

Ueff(x)=Φ(x)+τ2γ(Φ′(x))2−Dγln⁡(1+τγΦ′′(x))+∫dx′V(x−x′)ρ(x′)+τγ∫Φ′′(x′)∫xV′(x′−x′′)ρ(x′′)(12)U_{\mathrm{eff}}(x) = \Phi(x) + \frac{\tau}{2\gamma}(\Phi'(x))^2 - D\gamma\ln\left(1+\frac{\tau}{\gamma}\Phi''(x)\right) + \int dx' V(x-x')\rho(x') \\ \qquad+ \frac{\tau}{\gamma}\int\Phi''(x')\int^{x} V'(x'-x'')\rho(x'') \tag*{(12)}

We consider NN AOUPs interacting via short-ranged potential in two kinds of confinements: a harmonic trap

Φ(x)=k2x2(13)\Phi(x)=\frac{k}{2}x^2 \tag*{(13)}

centered at the origin, and a double-well potential

Φ(x)=k4(x2−22)2(14)\Phi(x)=\frac{k}{4}(x^2-2^2)^2 \tag*{(14)}

with wells located at x=±2x=\pm2. These two confinements are studied together for a specific reason. The harmonic trap is not presented here as a new problem—it is very well studied—but is used as a constant-external-curvature reference system, for which Φ′′(x)=k\Phi''(x)=k, and Φ′′′(x)=0\Phi'''(x)=0 everywhere. Thus, the external confinement itself introduces no spatial variation of curvature. In the interacting harmonic system, however, the many-body Hessian can still possess a nontrivial spectrum because of the interaction contribution. The harmonic trap therefore provides a useful reference in which interaction-induced spatial heterogeneity can be distinguished from that generated by spatial variations of the external confinement. The double-well potential of Eq. (14), by contrast, has a strongly position-dependent curvature. For this, one has Φ′′(x)=k(3x2−4)\Phi''(x)=k(3x^2-4), so that Φ′′(±2)=8k\Phi''(\pm2)=8k, and Φ′′(0)=−4k\Phi''(0)=-4k. The well minima therefore correspond to regions of large positive curvature, whereas the barrier is a region of negative curvature. Comparing the harmonic and double-well confinements at fixed persistence time τ\tau, interaction strength V0V_0, and particle number NN provides a controlled way of identifying the additional effects associated with spatially nonuniform external confinement and the presence of soft barrier regions. This comparison is particularly useful for interpreting the density, local diffusivity, fluctuation spectrum, correlation length, dynamical structure factor, and entropy production. In particular, because Φ′′′(x)=0\Phi'''(x)=0 in the harmonic trap, any contribution to the leading-order entropy production arising there must originate from the interaction-dependent part of the effective potential, whereas in the double-well confinement the spatial variation of the external curvature provides an additional contribution to irreversibility.

First we focus on harmonic confinement. For short ranged potential VV, the integral ∫dx′V(x−x′)ρ(x′)=V0ρ(x)+V1ρ′′(x)\int dx'V(x-x')\rho(x')=V_0\rho(x)+V_1\rho''(x) [82] in Eq. (8), the density simplifies to

ln⁡ρ(x)∝−k+k2τ/γ2Dγx2−(1+kτγ)(V0ρ(x)+V1ρ′′(x)).(15)\ln\rho(x) \propto-\frac{k+k^2\tau/\gamma}{2D\gamma}x^2-\left(1+\frac{k\tau}{\gamma}\right)(V_0\rho(x)+V_1\rho''(x)). \tag*{(15)}

The steady-state density profile of an AOUP reflects the interplay between confinement, persistence time τ\tau, and interparticle repulsion controlled by the WCA strength V0V_0. In a single-well harmonic potential (Figure 1-(A)), the density remains unimodal and centered at the trap minimum, but it deviates from a simple Gaussian as V0V_0 increases. For small τ\tau (low activity), the AOUP behaves close to an effective equilibrium particle, yielding a narrow, nearly Boltzmann-like profile. As τ\tau increases, persistent self-propulsion pushes particles outward, broadening the distribution and enhancing the density near the trap edges; strong repulsive interactions (V0V_0 large) further flatten the central peak by suppressing local crowding, leading to a wider and more uniform density around the center. The double-well trap (Figure 1-(B)) produces a bimodal density with peaks near the minima of each well. At low activity (τ\tau), the density is strongly localized in wells with weak interwell population due to rare crossings. Increasing τ\tau improves persistence-driven barrier crossing, redistributing probability towards the barrier region, and reducing depth contrast between wells. Stronger repulsive interactions amplify this effect by penalizing accumulation at the minima, promoting broader peaks and, at large τ\tau, a partially filled barrier region, signaling a strongly nonequilibrium steady-state where activity effectively lowers the barrier and reshapes the density far from equilibrium expectations.

Steady-state density profiles of interacting AOUPs under confinement

Figure 1. Steady-state density profiles of interacting AOUPs under confinement. (A) Density distribution in a harmonic trap, with enhanced accumulation and broadening controlled by the interaction strength V0V_0, particle number NN, and persistence time τ\tau. The solid (dashed) curves represent N=10N=10 (N=15N=15) particles. (B) Density distribution in a double-well potential, illustrating activity-driven localization in the wells and modified barrier occupancy, where increasing V0V_0, NN, and τ\tau amplifies crowding effects and sharpens nonequilibrium density contrasts between the wells and the barrier region. The solid (dashed) curves represent N=10N=10 (N=15N=15) particles.

The effective force and effective diffusion constant of the effective system in a harmonic trap (Eq. (13)) reads

Deff(x)=D1+kτγ+k2τ2γ2−τγ(V0ρ′′(x)+V1ρ′′′(x)),D_{\mathrm{eff}}(x)=\frac{D}{1+\frac{k\tau}{\gamma}+\frac{k^2\tau^2}{\gamma^2}-\frac{\tau}{\gamma}(V_0\rho''(x)+V_1\rho'''(x))},
F(x)=(1+kτγ)(kγx+V0γρ′(x))1+kτγ+k2τ2γ2−τγ(V0ρ′′(x)+V1ρ′′′(x)).(16)\mathcal{F}(x)=\frac{\left(1+\frac{k\tau}{\gamma}\right)\left(\frac{k}{\gamma}x+\frac{V_0}{\gamma}\rho'(x)\right)}{1+\frac{k\tau}{\gamma}+\frac{k^2\tau^2}{\gamma^2}-\frac{\tau}{\gamma}(V_0\rho''(x)+V_1\rho'''(x))}. \tag*{(16)}

Assuming sufficiently smooth densities such that ρ′′′′(x)≈0\rho^{\prime\prime\prime\prime}(x)\approx0 and using the non-interacting density (implying ρ′′(x)≈−V0ρ(x)/V1\rho^{\prime\prime}(x)\approx-V_0\rho(x)/V_1 from Eq. (15)), we get Deff(x)≈v02τ1+kτγ+k2τ2γ2+τγV02ρ(x)D_{\rm eff}(x)\approx\frac{v_0^2\tau}{1+\frac{k\tau}{\gamma}+\frac{k^2\tau^2}{\gamma^2}+\frac{\tau}{\gamma}V_0^2\rho(x)}. The local diffusion constant depends non-monotonically on τ\tau reaching maximum when it equals the relaxation time in the trap γ/k\gamma/k. The excluded-volume interactions strongly suppress diffusion in a density-dependent manner: so regions of high local density exhibit a markedly reduced DeffD_{\rm eff}. Overall, larger V0V_0 and NN promote strong local suppression of diffusion via crowding, while increasing τ\tau enhances nonequilibrium correlations that compete with and magnify confinement and interaction effects, resulting in pronounced spatial variation of the local diffusion constant. Figure 2 summarizes how activity and interactions renormalize transport and forces in confined AOUP systems. In panels (A) and (B), the local diffusion constant reflects the combined effect of persistence time τ\tau, interaction strength V0V_0, and crowding NN: increasing τ\tau enhances nonequilibrium persistence, leading to spatially heterogeneous diffusion that deviates from the constant passive value, with stronger suppression near regions of large restoring force. In the harmonic trap (A), this manifests as a reduction of the local diffusion near the trap center for large V0V_0 and NN, due to caging and interaction-induced hindrance, while for larger τ\tau the diffusion becomes more position dependent because active correlations couple to the trap curvature. In the double-well potential (B), the effect is more pronounced: diffusion is strongly reduced near the barrier and well minima as V0V_0 and NN increase, reflecting crowding-induced trapping and reduced barrier-crossing efficiency, while larger τ\tau enhances accumulation and sharpens spatial contrasts. Panels (C) and (D) show the corresponding effective force, which deviates from the bare conservative force due to activity–interaction coupling: in the harmonic trap (C), increasing τ\tau and V0V_0 renormalizes the force slope, typically softening the effective confinement at moderate activity but leading to nonlinear distortions at high density. In the double-well case (D), the effective force becomes highly nontrivial, with interaction- and activity-induced modifications that flatten the barrier and skew the force near the wells, signaling an effective nonequilibrium potential shaped jointly by persistence, excluded-volume interactions, and particle number.

Four-panel plots of local diffusion constants and effective forces for interacting AOUPs under harmonic and double-well confinement

Figure 2. Local transport and effective force in interacting AOUPs under confinement. (A) Spatially resolved local diffusion constant in a harmonic trap for two persistence times, τ=0.1\tau= 0.1 and τ=1\tau= 1, showing interaction- and density-induced suppression and enhanced heterogeneity with increasing WCA strength V0V_0 and particle number NN. (B) Local diffusion constant in a double-well potential, highlighting strong spatial variation near the well minima and barrier, with reduced diffusion at larger V0V_0, NN, and τ\tau. (C) Effective force experienced by a particle in a harmonic trap, illustrating activity- and interaction-driven renormalization of the trap stiffness. (D) Effective force in a double-well potential, exhibiting pronounced nonlinearity and barrier reshaping due to the combined effects of persistence time τ\tau, excluded-volume interactions V0V_0, and crowding NN.

Steady-State Velocities

The interplay between activity, confinement, and steric interactions produces a spatially heterogeneous nonequilibrium steady-state. For a single AOUP in a double-well potential, Caprini et al. [74] showed that the barrier region is characterized by strong violations of equilibrium behavior. In this region the mobility becomes negative, entropy production is large, and the local velocity distribution develops a bimodal structure. These regions were termed Nonequilibrium Regions (NeR). In contrast, the vicinity of the potential minima behaves approximately as an Equilibrium Region (ER), where the velocity distribution remains nearly Gaussian and entropy produc-

tion is comparatively small. Here we extend this picture to interacting many-particle systems and investigate how fluctuations and correlations are organized by the local curvature of the confining landscape, thereby deriving analytically the behaviour we previously obtained from simulations [75].

The particles interact through a Weeks-Chandler-Andersen (WCA) potential, which prevents particle crossings and preserves the ordering of particles throughout the dynamics. As a consequence, each particle index acts as a proxy for its spatial location within the trap. To understand the origin of the observed fluctuations, we linearize Eq. (3) around the energy-minimized configuration {xi∗}\{x_i^*\} and obtain

x˙=v,(17)\dot{x} = v, \tag*{(17)}
v˙=−[1τI+1γH]v−1τγHδx+2Dτξ,(18)\dot{v} = -\left[\frac{1}{\tau}\mathbf{I} + \frac{1}{\gamma}\mathbf{H}\right]v - \frac{1}{\tau\gamma}\mathbf{H}\delta x + \frac{\sqrt{2D}}{\tau}\xi, \tag*{(18)}

where δxi=xi−xi∗\delta x_i = x_i - x_i^* and

Hij=ϕ′′(xi∗)δij+∂xi∂xjV(xi∗−xj∗)(19)H_{ij} = \phi^{\prime\prime}(x_i^*)\delta_{ij} + \partial_{x_i}\partial_{x_j}V(x_i^* - x_j^*) \tag*{(19)}

is the Hessian matrix of the many-body potential energy landscape. The eigenvalues λα\lambda_\alpha and eigenvectors ψα\psi_\alpha of H\mathbf{H} characterize the local stiffness and collective normal modes of the system. In the absence of interactions, λα=ϕ′′(xα)\lambda_\alpha= \phi^{\prime\prime}(x_\alpha) reduces to the local curvature of the external potential. Large positive eigenvalues correspond to stiff, strongly confining regions, whereas small eigenvalues identify soft regions where particles can fluctuate over larger distances. In a double-well potential, the smallest eigenvalues are localized near the barrier, while the largest eigenvalues occur near the well bottoms.

The velocity and position correlation functions are given by

⟨vi(t)vj(0)⟩=∑αv02γγ+τλαλαγe−λαt/γ−1τe−t/τλαγ−1τψiαψjα,(20)\left\langle v_i(t)v_j(0)\right\rangle= \sum_\alpha\frac{v_0^2\gamma}{\gamma+\tau\lambda_\alpha}\frac{\frac{\lambda_\alpha}{\gamma}e^{-\lambda_\alpha t/\gamma}-\frac{1}{\tau}e^{-t/\tau}}{\frac{\lambda_\alpha}{\gamma}-\frac{1}{\tau}}\psi_i^\alpha\psi_j^\alpha, \tag*{(20)}

and

⟨δxi(t)δxj(0)⟩=∑αv02τγ2λα(γ+τλα)λαγe−t/τ−1τe−λαt/γλαγ−1τψαiψαj(21)\left\langle\delta x_i(t)\delta x_j(0)\right\rangle= \sum_{\alpha}\frac{v_0^2\tau\gamma^2}{\lambda_{\alpha}(\gamma+\tau\lambda_{\alpha})}\frac{\frac{\lambda_{\alpha}}{\gamma}e^{-t/\tau}-\frac{1}{\tau}e^{-\lambda_{\alpha}t/\gamma}}{\frac{\lambda_{\alpha}}{\gamma}-\frac{1}{\tau}}\psi_{\alpha}^{i}\psi_{\alpha}^{j} \tag*{(21)}

The functional forms of Eqs. (20) and (21) are closely related to the established correlation functions for harmonically confined degrees of freedom driven by exponentially correlated active noise [1, 2]. In particular, Samanta and Chakrabarti extended the single-particle harmonic calculation to the normal modes of Rouse, Zimm, and semiflexible polymer chains, for which each normal mode obeys an effective linear harmonic equation [2]. We therefore regard the corresponding harmonic-mode limit of Eqs. (20)–(22) as a consistency check of the present formalism rather than as a new result. The distinction of the present calculation is not simply the passage from a single degree of freedom to a many-body system. Rather, the correlations are expressed in terms of the complete spectrum {λα,ψα}\{\lambda_{\alpha},\psi_{\alpha}\} of the configuration-dependent many-body Hessian, Eq. (19), which contains simultaneously the local curvature of the external confinement and the stiffness generated by interparticle interactions. Thus, unlike a prescribed set of harmonic normal modes, the collective mode spectrum in the present system depends on both the particle configuration and the local geometry of the confining potential. The eigenvalues λα\lambda_{\alpha} determine the stiffness and relaxation of the collective modes, while the corresponding eigenvectors ψαi\psi_{\alpha}^{i} describe their spatial structure across the confined system. Equations (20)–(22) therefore allow the fluctuations and correlations to be resolved in space and make it possible to identify where soft and stiff collective modes are concentrated within the trap. For a harmonic external potential, Φ′′(x)=k\Phi''(x)=k is spatially uniform, so the confinement itself does not generate any spatial variation of the local curvature. An interacting harmonic system may nevertheless possess a nontrivial many-body spectrum because of the interaction contribution to the Hessian. The double-well confinement introduces an additional and qualitatively distinct source of spatial heterogeneity: its external curvature varies strongly with position, from positive curvature near the minima to soft

or negative curvature near the barrier. Comparison of the harmonic and double-well cases therefore allows us to separate collective effects generated by interactions from those associated specifically with spatially nonuniform confinement. The corresponding equal-time fluctuations are

⟨vi2⟩=∑αv02γγ+τλα(ψαi)2,⟨δxi2⟩=∑αv02τγ2λα(γ+τλα)(ψαi)2.(22)\langle v_i^2\rangle= \sum_{\alpha} \frac{v_0^2\gamma}{\gamma+\tau\lambda_\alpha}(\psi_\alpha^i)^2,\qquad\langle\delta x_i^2\rangle= \sum_{\alpha} \frac{v_0^2\tau\gamma^2}{\lambda_\alpha(\gamma+\tau\lambda_\alpha)}(\psi_\alpha^i)^2. \tag*{(22)}

(22) immediately reveals the role of curvature. The contribution of each mode is inversely proportional to γ+τλα\gamma+\tau\lambda_\alpha, implying that stiff modes with large positive curvature suppress fluctuations while soft modes enhance them. Therefore, the spatial variation of the fluctuations directly reflects the underlying curvature landscape experienced by the particles.

Figure 3 illustrates this behavior for interacting AOUPs confined in harmonic and double-well potentials. In the harmonic trap (Figure 3A), the curvature is positive and approximately uniform throughout the system. Consequently, the velocity fluctuations vary smoothly across space and exhibit a minimum at the trap center. This minimum arises primarily from crowding: particles accumulate near the center of the trap, increasing the effective local stiffness through steric interactions and suppressing fluctuations. Since the curvature does not vary strongly with position, no sharply localized nonequilibrium region emerges.

Spatial variation of the steady-state velocity fluctuations in a harmonic trap and a double-well potential

Figure 3. Spatial variation of the steady-state velocity fluctuations in (A) a harmonic trap and (B) a double-well potential. Red curves correspond to a system of N=10N = 10 particles, while blue curves correspond to N=50N = 50 particles. Fluctuations are suppressed near the minima of the potentials due to crowding.

The double-well potential exhibits qualitatively different behavior (Fig. 3B). Near the well bottoms, where the curvature is strongly positive, particles are trapped by large restoring forces and the velocity fluctuations are strongly suppressed. These regions correspond to the ER identified in Ref. [74]. In contrast, the barrier region is characterized by small or negative curvature. The restoring force becomes weak, allowing the persistent active force to dominate the dynamics. Consequently, velocity fluctuations are significantly enhanced and the system enters a NeR. The contrast between ER and NeR becomes particularly evident for small particle numbers (N=10N = 10). The fluctuation minima coincide with the well bottoms, while a pronounced maximum develops at the barrier. This enhancement reflects the fact that particles approaching the barrier can maintain their active velocities over longer times before being redirected by the external force. For larger systems (N=50N = 50), increasing particle density stiffens the collective modes, with larger λα\lambda_\alpha suppressing fluctuations throughout the trap. The enhanced effective stiffness generated by neighboring particles reduces the amplitude of active motion and partially masks the curvature-induced heterogeneity. Thus increasing particle number drives the system toward a more equilibrium-like state despite the presence of activity.

Increasing τ\tau strengthens the memory of the active force and reduces the efficiency of local relaxation. From (22), modes with small λα\lambda_\alpha (NeR regions) receive the largest enhancement because the denominator γ+τλα\gamma+\tau\lambda_\alpha grows slowly. As a result, increasing τ\tau preferentially amplifies fluctuations in the NeR while producing only modest changes in the ER. This selective enhancement of barrier fluctuations leads to increased barrier-crossing events at high persistence.

The differences between the two regions also emerge from the velocity and position correlation functions in (20) and (21). Two timescales govern the dynamics, τ\tau and γλα\frac{\gamma}{\lambda_\alpha}. The persistence time τ\tau characterizes the memory of the active force, while γλα\frac{\gamma}{\lambda_\alpha} is the relaxation time associated with a normal mode of stiffness λα\lambda_\alpha. In the ER γλα\frac{\gamma}{\lambda_\alpha} is short and correlations decay rapidly. The dynamics therefore resembles that of an equilibrium system with weak temporal memory. In the NeR, however, λα\lambda_\alpha becomes small and γλα\frac{\gamma}{\lambda_\alpha} grows substantially. Correlations persist for much longer times, producing coherent motion and enhanced collective fluctuations. The spatial extent of these correlations can be estimated by expanding the Hessian spectrum around a locally homogeneous region, λ(k)=ϕ′′(x∗)+Kk2\lambda(k)=\phi''(x^*)+Kk^2, where KK is determined by the interaction stiffness. Substituting this spectrum into (22) gives

⟨∣v(k)∣2⟩=v02γγ+τ[ϕ′′(x∗)+Kk2].(23)\langle|v(k)|^2\rangle= \frac{v_0^2\gamma}{\gamma+\tau[\phi''(x^*)+Kk^2]}. \tag*{(23)}

Fourier transformation yields an exponentially decaying spatial correlation function with correlation length

ξ=τKγ+τϕ′′(x∗).(24)\xi= \sqrt{\frac{\tau K}{\gamma+\tau\phi''(x^*)}}. \tag*{(24)}

(24) demonstrates explicitly that the correlation length is controlled by both persistence and curvature. Near the barrier (NeR) as shown in Figure 4-D, where ϕ′′(x∗)\phi''(x^*) is small or negative, ξ\xi becomes large and particles remain correlated over longer distances. Near the well bottoms (ER), where the curvature is strongly positive, the correlation length decreases and collective motion is rapidly suppressed. Increasing τ\tau enlarges ξ\xi, promoting synchronized motion and cooperative barrier-crossing events. ξ∼0\xi\sim0 for passive systems.

Steady-state single-particle position distributions and schematic illustration of equilibrium and nonequilibrium regions

Figure 4. (A-C) Steady-state single-particle position distributions Pi(x)P_i(x) for five interacting AOUPs out of N=10N=10 particles in the left well of a double-well potential. (Adapted from [75]) The distributions are symmetric for the particles in the right well and are not shown. These distributions were obtained from simulations of Eq. (1) with hard potential interactions using the Euler–Maruyama method with a time step dt=10−3dt=10^{-3} over 10710^7 iterations. Increasing persistence broadens the distributions and induces bimodality for particles located closer to the barrier, reflecting enhanced barrier exploration and nonequilibrium fluctuations. (D) Schematic illustration of the equilibrium regions (ERs) near the well minima and the nonequilibrium region (NeR) near the barrier. ERs are characterized by small fluctuations, short correlation lengths, and low entropy production, whereas the NeR exhibits enhanced fluctuations, long-ranged correlations, and large entropy production due to persistent active motion.

Structure Factor

Now, we analyze how the curvature of the trap and the activity of the particles dictate their density fluctuations in the different regions of the confining potential. The dynamical structure factor quantifies the spectrum of density-density correlations in both space and frequency. In steady-state, the dynamical structure factor of NN interacting AOUP particles is derived (see Appendix C for details) as

S(k,ω)=2Dk2Nτ2∑α∣Aα(k)∣2(λατγ−ω2)2+ω2(1τ+λαγ)2(25)S(k,\omega) = \frac{2Dk^2}{N\tau^2} \sum_\alpha\frac{\lvert A_\alpha(k)\rvert^2}{\left(\frac{\lambda_\alpha}{\tau\gamma}-\omega^2\right)^2+\omega^2\left(\frac{1}{\tau}+\frac{\lambda_\alpha}{\gamma}\right)^2} \tag*{(25)}

where

Aα(k)=∑ie−ikxi∗ψα(i),(26)A_\alpha(k) = \sum_i e^{-ikx_i^*}\psi_\alpha(i), \tag*{(26)}

and λα\lambda_\alpha and ψα\psi_\alpha are the eigenvalues and eigenvectors of the Hessian matrix H\mathbf{H} discussed in Sec. IV. The eigenvalues λα\lambda_\alpha determine the relaxation frequencies of the collective modes, while the eigenvectors enter through the form factor Aα(k)A_\alpha(k), which controls how strongly each mode contributes at a given wavevector kk. Soft modes with small λα\lambda_\alpha, typically associated with particles near the barrier (NeR), generate enhanced low-frequency spectral weight and slow relaxation, reflecting density transitions between the wells and long-lived collective rearrangements. Stiff modes with large λα\lambda_\alpha, associated with strongly confined particles near the minima, lead to faster relaxation, similar to equilibrium systems. Increasing NN increases the particle density and stiffens the Hessian eigenvalues λα\lambda_\alpha, density fluctuations become increasingly distributed among many modes and the low-frequency response is suppressed leading to more equilibrium-like behavior. In contrast, for small NN, a few low-lying modes can dominate the sum, leading to stronger fluctuations and more pronounced spectral features. The difference between passive and active systems arises through the persistence time τ\tau.

Entropy Production

AOUPs are intrinsically nonequilibrium systems because their persistent self-propulsion continuously injects energy into the dynamics and breaks detailed balance. As a consequence, the system exhibits irreversible probability currents and violates time-reversal symmetry. The steady-state entropy production provides a quantitative measure of this irreversibility and therefore characterizes the degree to which the system departs from equilibrium. In this section, we investigate how confinement generates spatially heterogeneous entropy production and how this heterogeneity is controlled by the local curvature of the confining potential, particle interactions, and persistence time. We used the Spinney-Ford [83] formula to obtain the steady-state entropy production rate in the mean field approximation

S˙tot=Dτ24⟨(∑i,j∂ijφ)2⟩+O(τ3)=Dτ24∫dx ρ(x)[∂x3φeff(x)]2+O(τ3),\dot{S}_{\mathrm{tot}} = \frac{D\tau^2}{4}\left\langle\left(\sum_{i,j}\partial_{ij}\varphi\right)^2\right\rangle+\mathcal{O}(\tau^3)=\frac{D\tau^2}{4}\int dx\,\rho(x)\left[\partial_x^3\varphi_{\mathrm{eff}}(x)\right]^2+\mathcal{O}(\tau^3),

where the effective potential is

φeff(x)=Φ(x)+2∫dx′ρ(x′)V(x−x′).(27)\varphi_{\mathrm{eff}}(x)=\Phi(x)+2\int dx'\rho(x')V(x-x'). \tag*{(27)}

time τ\tau. In the passive limit (τ→0\tau\to0), each mode has only the curvature-controlled relaxation rate λα/γ\lambda_\alpha/\gamma, whereas the active system contains two timescales: the structural relaxation time γ/λα\gamma/\lambda_\alpha and the active persistence time τ\tau. In the limit τ→0\tau\to0, the active memory relaxes rapidly and the passive Lorentzian form is recovered. For finite τ\tau, however, soft modes with small λα\lambda_\alpha are strongly affected by persistence and produce enhanced low-frequency spectral weight. These modes are especially important near the barrier of the double-well potential, whereas near the well minima large λα\lambda_\alpha suppresses persistent fluctuations and the response becomes closer to the passive one. In harmonic confinement, the curvature is uniform, the Hessian spectrum is relatively regular, and there are no strongly localized barrier modes; therefore active persistence mainly broadens or shifts the relaxation spectrum compared with the passive case, but the response remains smooth and equilibrium-like. The harmonic case therefore acts here as the spectral baseline: it shows what activity alone does to S(k,ω)S(k,\omega) when the curvature does not vary. The additional low-frequency weight seen in the double well, which is absent from this baseline at the same τ\tau, V0V_0 and NN, is consequently a curvature-induced effect and constitutes a direct spectral signature of the soft barrier modes.

(27) shows that entropy production is governed by the square of the third derivative of the effective potential. Since ∂x3φeff\partial_x^3\varphi_{\mathrm{eff}} measures the spatial variation of the local curvature, entropy production is generated only in regions where the curvature changes with position. Thus, entropy production provides a direct link between the nonequilibrium dynamics and the geometry of the effective energy landscape.

The leading-order dependence on persistence follows directly from Eq. (27),

S˙tot∝Dτ2.(28)\dot{S}_{\mathrm{tot}} \propto D\tau^2. \tag*{(28)}

Consequently, the passive limit τ→0\tau\to0 restores equilibrium behavior and causes the entropy production to vanish. In this limit, the active force loses its temporal memory and reduces to thermal white noise, recovering detailed balance and time-reversal symmetry.

The role of confinement becomes particularly evident when comparing harmonic and double-well potentials. This comparison is the sharpest illustration of why the harmonic trap is retained as a reference system: Eq. (27) depends on the confinement only through ∂x3ϕeff\partial_x^3\phi_{\mathrm{eff}}, so a constant-curvature trap is, at this order, the exactly solvable null case in which activity is present but produces no irreversibility whatsoever. Any entropy production observed in the double well at the same activity is therefore unambiguously attributable to curvature variation, and any entropy production observed in the interacting harmonic trap is unambiguously attributable to the interactions. For a harmonic trap, the curvature is uniform, Φ′′(x)=k\Phi^{\prime\prime}(x)=k and therefore Φ′′′(x)=0\Phi^{\prime\prime\prime}(x)=0. In the absence of interactions, the entropy production vanishes to order τ2\tau^2, despite the presence of active forces. The dynamics therefore remain effectively equilibrium-like. Interparticle interactions introduce a weak spatial dependence through the density-dependent contribution to ϕeff(x)\phi_{\mathrm{eff}}(x), producing a small but finite entropy production, primarily near the edges of the particle cloud where the density profile varies most strongly.

In a double-well potential, Eq. (29) demonstrates that the entropy production of a given particle depends on how strongly its position distribution overlaps with the nonequilibrium regions of the trap. We simulate the single-particle distribution ρi(x)=⟨δ(xi(t)−x)⟩\rho_i(x)=\langle\delta(x_i(t)-x)\rangle in Figure 4. For the iith particle the entropy production rate is

S˙i=Dτ24∫dxρi(x)[∂x3ϕeff(x)]2+O(τ3).(29)\dot{S}_i=\frac{D\tau^2}{4}\int dx\rho_i(x)\left[\partial_x^3\phi_{\mathrm{eff}}(x)\right]^2+O(\tau^3). \tag*{(29)}

The density in Eq. (27) is related to the single-particle distribution by ρ(x)=∑iρi(x)\rho(x)=\sum_i\rho_i(x). Since ∂x3ϕeff(x)=6kx\partial_x^3\phi_{\mathrm{eff}}(x)=6kx, from Eq. (29) we obtain S˙i=Dτ24⟨x2⟩i\dot{S}_i=\frac{D\tau^2}{4}\langle x^2\rangle_i, i.e. depends on the width of the single-particle distribution ρi\rho_i or the variance of the particle position. The entropy production is therefore directly related to the spatial extent of the particle distribution. Particles localized near the minima of the wells have narrow, unimodal distributions with small positional variance and consequently generate little entropy. These particles predominantly occupy the equilibrium-like regions (ERs), where the local curvature varies slowly and fluctuations remain weak. In contrast, particles near the barrier exhibit broader and often bimodal distributions, reflecting higher entropy production.

In Figure 4, for small persistence times, the distributions remain narrow and localized near the minima, leading to weak entropy production. As τ\tau increases, the distributions broaden and develop pronounced bimodality, indicating that particles increasingly explore both wells and the barrier region separating them. This broadening increases the positional variance ⟨x2⟩i\langle x^2\rangle_i and consequently enhances the entropy production. Therefore, the increase in entropy production with persistence arises from two complementary mechanisms: the explicit Dτ2D\tau^2 dependence in Eq. (27) and the increased sampling of the nonequilibrium barrier region caused by broader, more strongly bimodal particle distributions.

Conclusion

In this work, we investigated the steady-state dynamics, density fluctuations, velocity correlations, structure factor, and entropy production of interacting active Ornstein-Uhlenbeck particles confined in harmonic and double-well potentials. The central novelty of the study is the identification of how activity, interactions, particle number, and the local curvature of the confining potential together organize the system into equilibrium-like and nonequilibrium-like spatial regions. The novelty of the present work does not lie in the correlation functions of harmonically confined degrees of freedom driven by exponentially correlated active noise, which are well established [1, 2]. In particular, the latter work also extends the harmonic active-noise description to the normal modes of polymer chains. Rather, the present contribution is the analytical description, within the stated mean-field, small-persistence, and linearized approximations, of an interacting confined AOUP system in which steric interactions, persistent active forcing, and spatially varying external curvature act simultaneously. For the interacting system, the velocity and position correlations, Eqs. (20)–(22), are resolved in terms of the eigenvalues and eigenvectors of the configuration-dependent many-body Hessian. Its spectrum contains contributions from both the local curvature of the external confinement and the interparticle interactions, rather than from prescribed harmonic mode stiffnesses alone. Within a locally homogeneous description, this framework further leads to the curvature-dependent correlation length, Eq. (24), and to the dynamical structure factor, Eq. (25). In addition, the small-persistence expansion yields the leading-order curvature-gradient contribution to the entropy production, Eq. (27), thereby relating irreversibility to spatial variations of the effective curvature. The harmonic trap is therefore used as a constant-curvature reference system rather than as a new active-matter problem. Although interactions can still generate a nontrivial many-body mode spectrum in this case, the external curvature remains spatially uniform. The double-well potential provides the complementary situation in which the external curvature changes strongly between the well minima and the barrier region. Comparing these two confinements consequently allows us to separate interaction-induced collective effects from those associated specifically with spatially nonuniform confinement and soft barrier modes. In this way, the present analysis identifies how persistence, crowding, and local confinement geometry jointly control density fluctuations, collective correlations, mode softening, and irreversibility in interacting confined active systems. While the harmonic trap remains largely equilibrium-like due to its uniform positive curvature, the double-well potential produces strong spatial heterogeneity, with the well minima behaving as equilibrium regions and the barrier acting as a nonequilibrium region.

Starting from the many-particle AOUP dynamics, we derived a mean-field single-particle Fokker-Planck equation and obtained an effective density equation with position-dependent diffusion and effective force. This formulation shows that active persistence and excluded-volume interactions renormalize both the confining potential and the local transport properties. In the harmonic trap, activity broadens the density profile and interactions suppress diffusion in crowded regions. In the double-well potential, persistence reshapes the steady-state density by enhancing barrier exploration and modifying the population contrast between the wells and the barrier.

A major finding is that the nonequilibrium character of the system is controlled by the local curvature landscape. Linearizing the dynamics around the minimum-energy configuration shows that the eigenvalues of the Hessian matrix determine the amplitude and lifetime of velocity and position fluctuations. Stiff modes localized near the well minima suppress fluctuations and produce equilibrium-like behavior, whereas soft modes near the barrier enhance fluctuations, correlations, and collective motion. This provides a direct microscopic explanation for the emergence of nonequilibrium regions in the double-well trap. Increasing the persistence time strengthens these soft-mode fluctuations and increases the spatial correlation length, thereby promoting cooperative and synchronized barrier-crossing events. In contrast, increasing the particle number enhances crowding and collective stiffness, suppressing fluctuations and driving the system toward a more equilibrium-like state.

The dynamical structure factor further reveals how active persistence modifies density fluctuations. Soft barrier modes generate enhanced low-frequency spectral weight and slow relaxation, while stiff modes near the well bottoms relax rapidly and resemble passive equilibrium modes. Thus, the structure factor provides a spectral signature of the spatial separation between equilibrium-like and nonequilibrium-like regions.

Finally, we showed that entropy production is governed by the spatial variation of the effective curvature, through the third derivative of the effective potential. Consequently, entropy production vanishes to leading order in a noninteracting harmonic trap but becomes finite in the double-well potential and in interacting systems where the effective landscape is spatially inhomogeneous. Particles localized near the barrier region exhibit higher entropy production, whereas those near the well minima contribute weakly.

Overall, our results demonstrate that confined active matter cannot be characterized only by global activity parameters. Instead, nonequilibrium behavior is highly localized and is selected by the geometry of the effective energy landscape. The interplay of persistence, curvature, and crowding interactions produces spatially resolved transport heterogeneity, enhanced barrier fluctuations, long-ranged correlations, and localized entropy production. These findings provide a unified framework for understanding how confinement controls collective nonequilibrium dynamics in interacting active systems.

Acknowledgements

Author SD acknowledges financial support by Science and Engineering Research Board (SERB), Government of India (File SRG/2022/000598) and (File MTR/2022/000281). The author also acknowledges financial support from the Birla Institute of Technology and Science, Pilani (File PLN/AD/2022-23/3). The author also thanks DST-FIST for the computational resources provided to the Department of Physics, BITS Pilani.

Declaration

During the preparation of this manuscript, the authors used ChatGPT and Grammarly for the sole purpose of English language polishing and grammar correction. After using this tool, the authors reviewed and edited the content as needed and take full responsibility for the content of the publication.

Derivation of (4)

For a general stochastic equation of motion of the form

y˙=Ai(y,t)+∑kBik(y,t)ϵk,(30)\dot{y}=A_i(y,t)+\sum_k B_{ik}(y,t)\epsilon_k, \tag*{(30)}

where ϵ\epsilon’s satisfy

⟨ϵk⟩=0,⟨ϵkϵl⟩=δklδ(t1−t2),(31)\langle\epsilon_k\rangle=0,\quad\langle\epsilon_k\epsilon_l\rangle=\delta_{kl}\delta(t_1-t_2), \tag*{(31)}

the Fokker Planck equation is given by

P˙=−∑i∂yi{[Ai+12∑jkBjk∂yiBik]P}+12∑ij∂yi∂yj{[∑kBikBjk]P}.(32)\dot{P}=-\sum_i\partial_{y_i}\left\{\left[A_i+\frac{1}{2}\sum_{jk}B_{jk}\partial_{y_i}B_{ik}\right]P\right\}+\frac{1}{2}\sum_{ij}\partial_{y_i}\partial_{y_j}\left\{\left[\sum_k B_{ik}B_{jk}\right]P\right\}. \tag*{(32)}

For our system in Eq. (3) we get

∂tP=−∑i∂xi(νiP)−∑i∂νi[AiP]+Dτ2∑i∂νi2P,(33)\partial_tP=-\sum_i\partial_{x_i}(\nu_iP)-\sum_i\partial_{\nu_i}[A_iP]+\frac{D}{\tau^2}\sum_i\partial_{\nu_i}^2P, \tag*{(33)}

where

Ai(x,ν)=−∑j[1τδij+1γΦ′′(xi)δij+1γ∂xixjV(xi−xj)]νj−1γτΦ′(xi)+1γτFi.(34)A_i(\mathbf{x},\boldsymbol{\nu})=-\sum_j\left[\frac{1}{\tau}\delta_{ij}+\frac{1}{\gamma}\Phi^{\prime\prime}(x_i)\delta_{ij}+\frac{1}{\gamma}\partial_{x_i x_j}V(x_i-x_j)\right]\nu_j-\frac{1}{\gamma\tau}\Phi'(x_i)+\frac{1}{\gamma\tau}F_i. \tag*{(34)}

The first equation of the BBGKY hierarchy for the reduced distribution functions is

P˙1(x,ν)=−∂x[νP1(x,ν)]+∂ν[1τνP1(x,ν)]+∂ν[∫dx′dν′1γ∂xx′V(x−x′)P2(x,ν,x′,ν′)]+∂ν[1γ∂x2Φ(x)P1(x,ν)]+1γτ∂xΦ(x)∂νP1(x,ν)+1γτ∫dx′dν′∂xV(x−x′)∂νP2(x,ν,x′,ν′)+Dτ2∂ν2P1(x,ν).(35)\begin{aligned} \dot{P}_1(x,\nu)={}&-\partial_x[\nu P_1(x,\nu)]+\partial_\nu\left[\frac{1}{\tau}\nu P_1(x,\nu)\right]+\partial_\nu\left[\int dx'd\nu'\frac{1}{\gamma}\partial_{xx'}V(x-x')P_2(x,\nu,x',\nu')\right]\\ &+\partial_\nu\left[\frac{1}{\gamma}\partial_x^2\Phi(x)P_1(x,\nu)\right]+\frac{1}{\gamma\tau}\partial_x\Phi(x)\partial_\nu P_1(x,\nu)+\frac{1}{\gamma\tau}\int dx'd\nu'\partial_xV(x-x')\\ &\qquad\partial_\nu P_2(x,\nu,x',\nu')+\frac{D}{\tau^2}\partial_\nu^2P_1(x,\nu). \tag*{(35)} \end{aligned}

P1(x,ν)P_1(x,\nu) is the joint distribution of the position xx and the ”velocity” ν\nu. P2(x,ν,x′,ν′)P_2(x,\nu,x',\nu') is the two-body correlation function.

Equation (A6) is obtained from Eq. (33) by integrating over the positions and velocities of the N−1N-1 particles i=2,…,Ni=2,\ldots,N, using the reduced distributions

P1(x,ν)=∫∏i=2Ndxidνi P(x,x2,…;ν,ν2,…),P_1(x,\nu)=\int\prod_{i=2}^{N}dx_i d\nu_i\,P(x,x_2,\ldots;\nu,\nu_2,\ldots),
P2(x,ν,x′,ν′)=∫∏i=3Ndxidνi P(x,x′,x3,…;ν,ν′,ν3,…).(36)P_2(x,\nu,x',\nu')=\int\prod_{i=3}^{N}dx_i d\nu_i\,P(x,x',x_3,\ldots;\nu,\nu',\nu_3,\ldots). \tag*{(36)}

Every term of Eq. (33) that is a total derivative with respect to an integrated variable contributes only a boundary term and are neglected. (35) is not closed as because P˙1\dot{P}_1 contains P2P_2. In order to solve for P2P_2 we need to consider the next equation of the hierarchy. However here we use the mean field approximation P2(x,ν,x′,ν′)=P1(x,ν)P1(x′,ν′)P_2(x,\nu,x',\nu')=P_1(x,\nu)P_1(x',\nu') which closes (35) and we get (4). Substituting this factorisation into the two P2P_2 terms of (35) replaces them by the mean interaction field generated by the one-body density, ∫dx′dν′∂xx′V(x−x′)P1(x,ν)P1(x′,ν′)\int dx'd\nu'\partial_{xx'}V(x-x')P_1(x,\nu)P_1(x',\nu') and ∫dx′dν′∂xV(x−x′)∂νP1(x,ν)P1(x′,ν′)\int dx'd\nu'\partial_xV(x-x')\partial_\nu P_1(x,\nu)P_1(x',\nu') respectively, which is precisely the mean-field (two-body factorisation) closure discussed in Sec. II; its conditions of validity are stated there.

To extract macroscopic observables, we compute moments of (4) over ν\nu. We obtain the continuity equation for the particle density ρ(x,t)=∫dνP1(x,ν,t)\rho(x,t)=\int d\nu P_1(x,\nu,t)

∂tρ(x,t)=∂xj(x,t)(37)\partial_t\rho(x,t)=\partial_xj(x,t) \tag*{(37)}

where j(x,t)=∫dννP1(x,ν,t)j(x,t)=\int d\nu\nu P_1(x,\nu,t) is the current. Similarly, multiplying Eq. (4) by ν\nu and ν2\nu^2 and integrating over ν\nu we get the following equations for j(x,t)j(x,t) and V(x,t)=∫dνν2P1(x,ν,t)\mathcal{V}(x,t)=\int d\nu\nu^2P_1(x,\nu,t)

∂tj(x,t)=−∂xV(x,t)−1τj(x,t)−1γΦ′′(x)j(x,t)−ρ(x,t)γτ∫dx′j(x′,t)∂xx′V(x−x′)−1γτΦ′(x)ρ(x,t)−ρ(x,t)γτ∫dx′∂xV(x−x′)ρ(x′,t)(38)\partial_tj(x,t)=-\partial_x\mathcal{V}(x,t)-\frac{1}{\tau}j(x,t)-\frac{1}{\gamma}\Phi''(x)j(x,t)-\frac{\rho(x,t)}{\gamma\tau}\int dx'j(x',t)\partial_{xx'}\mathcal{V}(x-x')-\frac{1}{\gamma\tau}\Phi'(x)\rho(x,t)-\frac{\rho(x,t)}{\gamma\tau}\int dx'\partial_x\mathcal{V}(x-x')\rho(x',t) \tag*{(38)}
∂tV(x,t)=−∂xχ(x,t)−2τV(x,t)−2γΦ′′(x)V(x,t)−2j(x,t)γ∫dx′j(x′,t)∂xx′V(x−x′)−2γτ∂xΦ(x)j(x,t)−2j(x,t)γτ∫dx′∂xV(x−x′)ρ(x′,t)+2Dτ2ρ(x,t)(39)\partial_t\mathcal{V}(x,t)=-\partial_x\chi(x,t)-\frac{2}{\tau}\mathcal{V}(x,t)-\frac{2}{\gamma}\Phi''(x)\mathcal{V}(x,t)-\frac{2j(x,t)}{\gamma}\int dx'j(x',t)\partial_{xx'}\mathcal{V}(x-x')-\frac{2}{\gamma\tau}\partial_x\Phi(x)j(x,t)-\frac{2j(x,t)}{\gamma\tau}\int dx'\partial_x\mathcal{V}(x-x')\rho(x',t)+\frac{2D}{\tau^2}\rho(x,t) \tag*{(39)}

where χ(x,t)=∫dνν3P1(x,ν,t)\chi(x,t)=\int d\nu\nu^3P_1(x,\nu,t).

Equations (A8)-(A10) are the first three members of a velocity-moment hierarchy: the equation for ρ\rho involves jj, the equation for jj involves V\mathcal{V}, and the equation for V\mathcal{V} involves χ\chi. The hierarchy is truncated at this order by two assumptions, each of which we now state together with its regime of validity. (i) On time scales long compared with the velocity relaxation time, the second moment is quasi-stationary, ∂tV(x,t)≃0\partial_t\mathcal{V}(x,t)\simeq0; this is the adiabatic (local-equilibrium) elimination of the velocity degrees of freedom and requires τ\tau and γ/λ\gamma/\lambda to be short compared with the time scale over which the density evolves. (ii) The third moment is assumed to be slowly varying in space, ∂xχ(x)≃0\partial_x\chi(x)\simeq0, which holds when the velocity distribution is locally close to Gaussian, so that χ\chi is determined by V\mathcal{V} and inherits its smoothness. Under (i) and (ii), Eq. (39) becomes algebraic in V\mathcal{V} and can be solved for it.

On long time scales, V(x,t)\mathcal{V}(x,t) has a weak variation with time. In addition, we assume a small gradient of χ(x)\chi(x) and neglect the ∂xχ(x)\partial_x\chi(x) term. Eq. (39) leads to

V(x)=Dτρ(x)1+τγΦ′′(x)−j˙(x)Ψ(x),where(40)\mathcal{V}(x)=\frac{D}{\tau}\frac{\rho(x)}{1+\frac{\tau}{\gamma}\Phi''(x)}-\dot{j}(x)\Psi(x), \quad\text{where} \tag*{(40)}
Ψ(x)=11+τγΦ′′(x)[1γ∫dx′j(x′)∂xx′V(x−x′)+1γΦ′(x)+1γ∫dx′ρ(x′)∂xV(x−x′)].\Psi(x)=\frac{1}{1+\frac{\tau}{\gamma}\Phi''(x)} \left[ \frac{1}{\gamma}\int dx'j(x')\partial_{xx'}\mathcal{V}(x-x') +\frac{1}{\gamma}\Phi'(x) +\frac{1}{\gamma}\int dx'\rho(x')\partial_x\mathcal{V}(x-x') \right].

Similarly the local current j(x)j(x)

j(x)=1Γ(x)[D∂x(ρ(x)1+τγΦ′′(x))+ρ(x)γ[Φ′(x)+∫V′(x−x′)ρ(x′)]](41)j(x)=\frac{1}{\Gamma(x)} \left[ D\partial_x\left(\frac{\rho(x)}{1+\frac{\tau}{\gamma}\Phi''(x)}\right) +\frac{\rho(x)}{\gamma}\left[\Phi'(x)+\int\mathcal{V}'(x-x')\rho(x')\right] \right] \tag*{(41)}

where Γ(x)=1+τγΦ′′(x)−τΨ′(x)\Gamma(x)=1+\frac{\tau}{\gamma}\Phi''(x)-\tau\Psi'(x). In the derivation we have neglected the gradient of the local current ∂xj(x)\partial_xj(x). This last step is the third and final approximation of the derivation: substituting V(x)\mathcal{V}(x) from Eq. (39) into Eq. (38) and setting ∂tj≃0\partial_tj\simeq0 produces a term ∂xj\partial_xj, which is negligible in and near the steady state where the current is small and slowly varying, and which vanishes identically in the zero-current solution j=0j=0 used in Sec. III. Comparing Eq. (41) with the constitutive form j=−[Deff(x)∂xρ+F(x)ρ]j=-[D_{\mathrm{eff}}(x)\partial_x\rho+\mathcal{F}(x)\rho] and inserting the result into the continuity equation (A8) then gives Eq. (5), and identifies Deff(x)D_{\mathrm{eff}}(x) and F(x)\mathcal{F}(x) as quoted in Eqs. (6). The positivity condition that Γ(x)\Gamma(x) and 1+τγΦ′′(x)1+\frac{\tau}{\gamma}\Phi^{\prime\prime}(x) must satisfy for this identification to be meaningful is discussed in § 2.

Correlation functions

Let the Hessian H\mathbf{H} in Eq. (18) be diagonalized as

Hψα=λαψα.(42)\mathbf{H}\psi^\alpha=\lambda_\alpha\psi^\alpha. \tag*{(42)}

so that the two relaxation rates of a normal mode are exactly the active persistence rate 1/τ1/\tau and the structural rate λα/γ\lambda_\alpha/\gamma. This factorisation of the characteristic polynomial is the origin of the two exponentials, e−t/τe^{-t/\tau} and e−λαt/γe^{-\lambda_\alpha t/\gamma}, and of the denominator λαγ−1τ\frac{\lambda_\alpha}{\gamma}-\frac{1}{\tau} appearing in Eqs. (47) and (B7) below; the degenerate case λα=γ/τ\lambda_\alpha=\gamma/\tau is regular and gives the usual critically damped form ∝te−t/τ\propto t e^{-t/\tau}.

The equal-time moments follow from stationarity. Since ddt⟨qα2⟩=2⟨qαuα⟩\frac{d}{dt}\langle q_\alpha^2\rangle=2\langle q_\alpha u_\alpha\rangle and ⟨qα2⟩\langle q_\alpha^2\rangle is time independent in the steady state, one has ⟨qαuα⟩=0\langle q_\alpha u_\alpha\rangle=0. Multiplying Eq. (45) by qαq_\alpha and averaging, the noise term drops out by causality and ⟨qαq¨α⟩=−⟨uα2⟩\langle q_\alpha\ddot{q}_\alpha\rangle=-\langle u_\alpha^2\rangle, which gives λα⟨qα2⟩=γ⟨uα2⟩\lambda_\alpha\langle q_\alpha^2\rangle=\gamma\langle u_\alpha^2\rangle; multiplying instead by uαu_\alpha and averaging fixes the velocity variance through the noise strength, ⟨uα2⟩=Dγτ(γ+τλα)=v02γγ+τλα\langle u_\alpha^2\rangle=\frac{D\gamma}{\tau(\gamma+\tau\lambda_\alpha)}=\frac{v_0^2\gamma}{\gamma+\tau\lambda_\alpha} using D=v02τD=v_0^2\tau. Together these give Eq. (49) below without solving the dynamics.

Normal-mode correlation-function derivation

Figure B.

Defining the normal modes

δxi(t)=∑αqα(t)ψiα,vi(t)=∑αuα(t)ψiα,\delta x_i(t)=\sum_\alpha q_\alpha(t)\psi_i^\alpha,\qquad v_i(t)=\sum_\alpha u_\alpha(t)\psi_i^\alpha,

the equations of motion in Eq. (18) become

q˙α=uα,(43)\dot{q}_\alpha=u_\alpha, \tag*{(43)}
u˙α=−(1τ+λαγ)uα−λατγqα+2Dτξα.(44)\dot{u}_\alpha=-\left(\frac{1}{\tau}+\frac{\lambda_\alpha}{\gamma}\right)u_\alpha-\frac{\lambda_\alpha}{\tau\gamma}q_\alpha+\frac{\sqrt{2D}}{\tau}\xi_\alpha. \tag*{(44)}

Then

q¨α+(1τ+λαγ)q˙α+λατγqα=2Dτξα.(45)\ddot{q}_\alpha+\left(\frac{1}{\tau}+\frac{\lambda_\alpha}{\gamma}\right)\dot{q}_\alpha+\frac{\lambda_\alpha}{\tau\gamma}q_\alpha=\frac{\sqrt{2D}}{\tau}\xi_\alpha. \tag*{(45)}

Equation (B4) is a damped harmonic oscillator driven by white noise. Its homogeneous solutions are ezte^{zt} with

z2+(1τ+λαγ)z+λατγ=0⟹z1=−1τ,z2=−λαγ,(46)z^2+\left(\frac{1}{\tau}+\frac{\lambda_\alpha}{\gamma}\right)z+\frac{\lambda_\alpha}{\tau\gamma}=0\quad\Longrightarrow\quad z_1=-\frac{1}{\tau},\qquad z_2=-\frac{\lambda_\alpha}{\gamma}, \tag*{(46)}

give Eq. (49) below without solving the dynamics. The two-time correlators are then obtained from the Onsager regression theorem: the correlation functions decay with the same two rates (B5) as the deterministic relaxation, with amplitudes fixed by the equal-time moments just computed, i.e. Cα(v)(0)=⟨uα2⟩C_\alpha^{(v)}(0)=\langle u_\alpha^2\rangle, C˙α(x)(0)=⟨qαuα⟩=0\dot{C}_\alpha^{(x)}(0)=\langle q_\alpha u_\alpha\rangle=0. Imposing these two conditions on the general combination c1e−t/τ+c2e−λαt/γc_1e^{-t/\tau}+c_2e^{-\lambda_\alpha t/\gamma} determines c1c_1 and c2c_2 uniquely and yields Eqs. (47) and (B7).

Using ⟨qαuα⟩=0\langle q_\alpha u_\alpha\rangle=0 and D=v02τD=v_0^2\tau, we obtain the velocity-velocity correlations as

Cα(v)(t)=⟨uα(t)uα(0)⟩=v02γγ+τλαλαγe−λαt/γ−1τe−t/τλαγ−1τ.(47)C_\alpha^{(v)}(t)=\langle u_\alpha(t)u_\alpha(0)\rangle=\frac{v_0^2\gamma}{\gamma+\tau\lambda_\alpha}\frac{\frac{\lambda_\alpha}{\gamma}e^{-\lambda_\alpha t/\gamma}-\frac{1}{\tau}e^{-t/\tau}}{\frac{\lambda_\alpha}{\gamma}-\frac{1}{\tau}}. \tag*{(47)}

Similarly, the position-position correlation is

Cα(x)(t)=⟨qα(t)qα(0)⟩=v02τγ2λα(γ+τλα)λαγe−t/τ−1τe−λαt/γλαγ−1τ.(48)C_\alpha^{(x)}(t)=\langle q_\alpha(t)q_\alpha(0)\rangle=\frac{v_0^2\tau\gamma^2}{\lambda_\alpha(\gamma+\tau\lambda_\alpha)}\frac{\frac{\lambda_\alpha}{\gamma}e^{-t/\tau}-\frac{1}{\tau}e^{-\lambda_\alpha t/\gamma}}{\frac{\lambda_\alpha}{\gamma}-\frac{1}{\tau}}. \tag*{(48)}

From these equations we obtain the equal time variance

Cα(v)(0)=⟨uα2⟩=v02γγ+τλα,Cα(x)(0)=⟨qα2⟩=Dγ2λα(γ+τλα)=v02τγ2λα(γ+τλα).(49)C_\alpha^{(v)}(0)=\langle u_\alpha^2\rangle=\frac{v_0^2\gamma}{\gamma+\tau\lambda_\alpha},\qquad C_\alpha^{(x)}(0)=\langle q_\alpha^2\rangle=\frac{D\gamma^2}{\lambda_\alpha(\gamma+\tau\lambda_\alpha)}=\frac{v_0^2\tau\gamma^2}{\lambda_\alpha(\gamma+\tau\lambda_\alpha)}. \tag*{(49)}

Therefore,

⟨vi(t)vj(0)⟩=∑αCα(v)(t)ψiαψjα,(50)\langle v_i(t)v_j(0)\rangle=\sum_\alpha C_\alpha^{(v)}(t)\psi_i^\alpha\psi_j^\alpha, \tag*{(50)}

and

⟨δxi(t)δxj(0)⟩=∑αCα(x)(t)ψiαψjα.(51)\langle\delta x_i(t)\delta x_j(0)\rangle=\sum_\alpha C_\alpha^{(x)}(t)\psi_i^\alpha\psi_j^\alpha. \tag*{(51)}

Derivation of Structure factor

The microscopic density field is

ρ(x,t)=∑iδ(x−xi(t)),(52)\rho(x,t)=\sum_i\delta(x-x_i(t)), \tag*{(52)}

with Fourier probe

ρk(t)=∑ie−ikxi(t).(53)\rho_k(t)=\sum_i e^{-ikx_i(t)}. \tag*{(53)}

Expanding around the steady-state positions {xi∗}\{x_i^*\}

xi(t)=xi∗+δxi(t),(54)x_i(t)=x_i^*+\delta x_i(t), \tag*{(54)}

and expanding the exponential to first order in the displacement,

ρk(t)≃∑ie−ikxi∗[1−ik δxi(t)].(55)\rho_k(t)\simeq\sum_i e^{-ikx_i^*}\left[1-ik\,\delta x_i(t)\right]. \tag*{(55)}

The linearisation in Eq. (55) is controlled provided the displacements are small on the scale of the probed wavelength,

k⟨δxi2⟩≪1,(56)k\sqrt{\left\langle\delta x_i^2\right\rangle}\ll1, \tag*{(56)}

with ⟨δxi2⟩\left\langle\delta x_i^2\right\rangle given by Eq. (22); the results below therefore describe the small-kk (collective) part of the spectrum, and are least accurate for particles in the soft barrier region, where ⟨δxi2⟩\left\langle\delta x_i^2\right\rangle is largest. The kk-independent first term of Eq. (55) is the static lattice factor: being time independent, it contributes only an elastic δ(ω)\delta(\omega) contribution to the structure factor and is excluded from the fluctuating part below, which is what S(k,ω)S(k,\omega) at ω≠0\omega\ne0 measures.

The fluctuating part is therefore

δρk(t)=−ik∑ie−ikxi∗δxi(t).(57)\delta\rho_k(t)=-ik\sum_i e^{-ikx_i^*}\delta x_i(t). \tag*{(57)}

Writing the displacement field in terms of the eigenvalues λα\lambda_\alpha and eigenvectors ψα\psi_\alpha of the Hessian matrix H\mathbf{H} (Appendix B) as

δxi(t)=∑αqα(t)ψα(i).(58)\delta x_i(t)=\sum_\alpha q_\alpha(t)\psi_\alpha(i). \tag*{(58)}

Each normal mode obeys

q¨α+(1τ+λαγ)q˙α+λατγqα=2Dτηα(t).(59)\ddot q_\alpha+\left(\frac{1}{\tau}+\frac{\lambda_\alpha}{\gamma}\right)\dot q_\alpha+\frac{\lambda_\alpha}{\tau\gamma}q_\alpha=\frac{\sqrt{2D}}{\tau}\eta_\alpha(t). \tag*{(59)}

The dynamical structure factor is defined as

S(k,ω)=1N∫−∞∞dt eiωt⟨δρk(t)δρ−k(0)⟩.(60)S(k,\omega)=\frac{1}{N}\int_{-\infty}^{\infty}dt\,e^{i\omega t}\left\langle\delta\rho_k(t)\delta\rho_{-k}(0)\right\rangle. \tag*{(60)}

Define the projection of density fluctuations onto mode α\alpha:

Aα(k)=∑ie−ikxi∗ψα(i).(61)A_\alpha(k)=\sum_i e^{-ikx_i^*}\psi_\alpha(i). \tag*{(61)}

Substituting Eq. (57) into the definition of S(k,ω)S(k,\omega) gives a double sum over modes, ∑αβAα(k)Aβ∗(k)⟨qα(ω)qβ∗(ω)⟩\sum_{\alpha\beta}A_\alpha(k)A_\beta^*(k)\langle q_\alpha(\omega)q_\beta^*(\omega)\rangle. The normal modes are driven by the independent noises ηα=∑iψαiξi\eta_\alpha=\sum_i\psi_\alpha^i\xi_i, which satisfy ⟨ηαηβ⟩∝δαβ\langle\eta_\alpha\eta_\beta\rangle\propto\delta_{\alpha\beta} because the eigenvectors of the symmetric Hessian are orthonormal; the modes are therefore statistically independent, ⟨qα(ω)qβ∗(ω)⟩=δαβ⟨∣qα(ω)∣2⟩\langle q_\alpha(\omega)q_\beta^*(\omega)\rangle=\delta_{\alpha\beta}\langle|q_\alpha(\omega)|^2\rangle, and the double sum collapses to the single mode-diagonal sum

S(k,ω)=k2N∑α∣Aα(k)∣2⟨∣qα(ω)∣2⟩.(62)S(k,\omega)=\frac{k^2}{N}\sum_\alpha|A_\alpha(k)|^2\left\langle|q_\alpha(\omega)|^2\right\rangle. \tag*{(62)}

The spectral density of a single mode follows by Fourier transforming Eq. (59). With qα(t)=∫dω2πe−iωtqα(ω)q_\alpha(t)=\int\frac{d\omega}{2\pi}e^{-i\omega t}q_\alpha(\omega) the equation becomes algebraic,

[−ω2−iω(1τ+λαγ)+λατγ]qα(ω)=2Dτηα(ω),(63)\left[-\omega^2-i\omega\left(\frac{1}{\tau}+\frac{\lambda_\alpha}{\gamma}\right)+\frac{\lambda_\alpha}{\tau\gamma}\right]q_\alpha(\omega)=\frac{\sqrt{2D}}{\tau}\eta_\alpha(\omega), \tag*{(63)}

so that, using ⟨∣ηα(ω)∣2⟩=1\langle|\eta_\alpha(\omega)|^2\rangle=1 for unit-intensity white noise and taking the squared modulus of the bracket, whose real and imaginary parts are λατγ−ω2\frac{\lambda_\alpha}{\tau\gamma}-\omega^2 and −ω(1τ+λαγ)-\omega\left(\frac{1}{\tau}+\frac{\lambda_\alpha}{\gamma}\right) respectively, one reads off Eq. (64) below. Using the following expression for each normal mode

⟨∣qα(ω)∣2⟩=2D/τ2(λατγ−ω2)2+ω2(1τ+λαγ)2,(64)\left\langle|q_\alpha(\omega)|^2\right\rangle= \frac{2D/\tau^2} {\left(\frac{\lambda_\alpha}{\tau\gamma}-\omega^2\right)^2+\omega^2\left(\frac{1}{\tau}+\frac{\lambda_\alpha}{\gamma}\right)^2}, \tag*{(64)}

we obtain Eq. (25) in the main text.

Derivation of Entropy production

The probability currents of Eq. (33) are

Jxi=νiP,(65)J_{x_i}=\nu_iP, \tag*{(65)}
Jνi=AiP−Dτ2∂νiP.(66)J_{\nu_i}=A_iP-\frac{D}{\tau^2}\partial_{\nu_i}P. \tag*{(66)}

We rewrite

Ai(x,ν)=−∑jΛij(x)νj−1γτΦ′(xi)+1γτFi,(67)A_i(\mathbf{x},\boldsymbol{\nu})=-\sum_j\Lambda_{ij}(\mathbf{x})\nu_j-\frac{1}{\gamma\tau}\Phi'(x_i)+\frac{1}{\gamma\tau}F_i, \tag*{(67)}

with

Λij(x)=1τδij+1γΦ′′(xi)δij+1γ∂xixjV(xi−xj).\Lambda_{ij}(\mathbf{x})=\frac{1}{\tau}\delta_{ij}+\frac{1}{\gamma}\Phi''(x_i)\delta_{ij}+\frac{1}{\gamma}\partial_{x_i x_j}V(x_i-x_j).

Following [83], under time reversal, we split AiA_i into reversible and irreversible components

Aνiirr=−∑jΛij(x)νj,(68)A_{\nu_i}^{\mathrm{irr}}=-\sum_j\Lambda_{ij}(\mathbf{x})\nu_j, \tag*{(68)}
Aνirev=−1γτΦ′(xi)+1γτFi.(69)A_{\nu_i}^{\mathrm{rev}}=-\frac{1}{\gamma\tau}\Phi'(x_i)+\frac{1}{\gamma\tau}F_i. \tag*{(69)}

The irreversible probability current is therefore

Jνiirr=AνiirrP−Dτ2∂νiP=−∑jΛij(x)νjP−Dτ2∂νiP.(70)J_{\nu_i}^{\mathrm{irr}}=A_{\nu_i}^{\mathrm{irr}}P-\frac{D}{\tau^2}\partial_{\nu_i}P=-\sum_j\Lambda_{ij}(\mathbf{x})\nu_jP-\frac{D}{\tau^2}\partial_{\nu_i}P. \tag*{(70)}

We use the probability function calculated in [26] in the equation above

P∼e−ϕ+∑i=1Nτνi22D{1−τ2D∑i=1N[(∂iϕ)2+τ∑j=1Nνiνj∂i∂jϕ−3D∂i2ϕ]+τ26D∑i=1Nνi∂i[τ∑j=1N[∑k=1Nνjνk∂j∂k−3D∂j2]ϕ+O(τ2)]}.(71)P \sim e^{-\frac{\phi+\sum_{i=1}^{N}\frac{\tau\nu_i^2}{2}}{D}}\left\{1-\frac{\tau}{2D}\sum_{i=1}^{N}\left[(\partial_i\phi)^2+\tau\sum_{j=1}^{N}\nu_i\nu_j\partial_i\partial_j\phi-3D\partial_i^2\phi\right]+\frac{\tau^2}{6D}\sum_{i=1}^{N}\nu_i\partial_i\left[\tau\sum_{j=1}^{N}\left[\sum_{k=1}^{N}\nu_j\nu_k\partial_j\partial_k-3D\partial_j^2\right]\phi+O(\tau^2)\right]\right\}. \tag*{(71)}

The cancellation can be made explicit. Writing P=P0(1+τP1+…)P=P_0(1+\tau P_1+\ldots) with P0∝e−(ϕ+∑iτνi2/2)/DP_0\propto e^{-(\phi+\sum_i\tau\nu_i^2/2)/D}, the diffusive part of the irreversible current is −Dτ2∂νiP-\frac{D}{\tau^2}\partial_{\nu_i}P. Differentiating the Gaussian velocity factor of P0P_0 gives ∂νiP0=−τνiDP0\partial_{\nu_i}P_0=-\frac{\tau\nu_i}{D}P_0, hence

−Dτ2∂νiP0=νiτP0,(72)-\frac{D}{\tau^2}\partial_{\nu_i}P_0=\frac{\nu_i}{\tau}P_0, \tag*{(72)}

which cancels exactly the 1τδij\frac{1}{\tau}\delta_{ij} part of −∑jΛijνjP-\sum_j\Lambda_{ij}\nu_jP. Differentiating the O(τ)O(\tau) bracket of PP, whose velocity-dependent piece is −τ22D∑jkνjνk∂j∂kϕ-\frac{\tau^2}{2D}\sum_{jk}\nu_j\nu_k\partial_j\partial_k\phi, gives a contribution 1τ∑j∂i∂jϕ νjP\frac{1}{\tau}\sum_j\partial_i\partial_j\phi\,\nu_jP at the same order in τ\tau, which cancels the remaining curvature part 1γ[Φ′′(xi)δij+∂xixjV(xi−xj)]νj\frac{1}{\gamma}[\Phi^{\prime\prime}(x_i)\delta_{ij}+\partial_{x_i x_j}V(x_i-x_j)]\nu_j of Λij\Lambda_{ij}, since ∂i∂jϕ\partial_i\partial_j\phi is exactly that Hessian. All terms linear in ν\nu therefore cancel identically, and the first surviving contribution comes from the O(τ2)O(\tau^2) term of PP, which upon differentiation produces the velocity-independent third-derivative term quoted below. −∑jΛij(x)νj-\sum_j\Lambda_{ij}(x)\nu_j in JiirrJ_i^{\mathrm{irr}} is canceled by the exponential term in PP and the second term on the right hand side of the equation in PP, leaving only the term

Jνiirr=D2∂ijjϕP+O(τ).(73)J_{\nu_i}^{\mathrm{irr}}=\frac{D}{2}\partial_{ijj}\phi P+O(\tau). \tag*{(73)}

The steady-state entropy production rate can be calculated from Spinney–Ford formula [83] with the diffusion matrix (D(ν))ij−1=τ2Dδij(D^{(\nu)})_{ij}^{-1}=\frac{\tau^2}{D}\delta_{ij}

S˙tot=∫dx dν∑i,jJνiirr(D(ν))ij−1JνjirrP=Dτ24⟨(∑ij∂ijjϕ)2⟩+O(τ3).(74)\dot{S}_{\mathrm{tot}}=\int dx\,d\nu\sum_{i,j}\frac{J_{\nu_i}^{\mathrm{irr}}(D^{(\nu)})_{ij}^{-1}J_{\nu_j}^{\mathrm{irr}}}{P}=\frac{D\tau^2}{4}\left\langle\left(\sum_{ij}\partial_{ijj}\phi\right)^2\right\rangle+O(\tau^3). \tag*{(74)}
S˙tot=Dτ24⟨[∑i(Φ′′(xi)+2∑j≠iV′′(xi−xj))]2⟩+O(τ3).(75)\dot{S}_{\mathrm{tot}}=\frac{D\tau^2}{4}\left\langle\left[\sum_i\left(\Phi^{\prime\prime}(x_i)+2\sum_{j\ne i}V^{\prime\prime}(x_i-x_j)\right)\right]^2\right\rangle+O(\tau^3). \tag*{(75)}

The final entropy production under the mean field approximation reduces to

S˙tot=Dτ24∫dx ρ(x)[∂x3ϕeff(x)]2,(76)\dot{S}_{\mathrm{tot}}=\frac{D\tau^2}{4}\int dx\,\rho(x)\left[\partial_x^3\phi_{\mathrm{eff}}(x)\right]^2, \tag*{(76)}

where the effective potential is given by

ϕeff(x)=Φ(x)+2∫dx′ρ(x′)V(x−x′).(77)\phi_{\mathrm{eff}}(x)=\Phi(x)+2\int dx'\rho(x')V(x-x'). \tag*{(77)}

Similarly the medium entropy production is

dSmed=∑i[τ2D(−∑jMijνj)∘dνi−τ2D(AνirevAνiirr)dt+1p∂(Aip)∂νidt−Dτ21p∂2p∂νi2dt].(78)dS_{\mathrm{med}}=\sum_i\left[\frac{\tau^2}{D}\left(-\sum_jM_{ij}\nu_j\right)\circ d\nu_i-\frac{\tau^2}{D}\left(A_{\nu_i}^{\mathrm{rev}}A_{\nu_i}^{\mathrm{irr}}\right)dt+\frac{1}{p}\frac{\partial(A_ip)}{\partial\nu_i}dt-\frac{D}{\tau^2}\frac{1}{p}\frac{\partial^2p}{\partial\nu_i^2}dt\right]. \tag*{(78)}

Averaging this we obtain using ⟨νTν⟩=Dτ2M−1\langle\nu^{T}\nu\rangle= \frac{D}{\tau^{2}}\mathbf{M}^{-1}

⟨S˙med⟩=τ2D∑i⟨(∑jMijνj)2⟩=τ2D⟨νTMTMν⟩=M.(79)\left\langle\dot{S}_{\mathrm{med}}\right\rangle= \frac{\tau^{2}}{D}\sum_{i}\left\langle\left(\sum_{j}M_{ij}\nu_{j}\right)^{2}\right\rangle= \frac{\tau^{2}}{D}\left\langle\nu^{T}\mathbf{M}^{T}\mathbf{M}\nu\right\rangle= M. \tag*{(79)}

References

  1. [1]G. Szamel, Self-propelled particle in an external potential: Existence of an effective temperature, Physical Review E 90, 021111 (2014).pubmed.ncbi.nlm.nih.gov/25122255
  2. [2]N. Samanta and R. Chakrabarti, Chain reconfiguration in active noise, Journal of Physics A: Mathematical and Theoretical 49, 195601 (2016).DOI
  3. [3]M. C. Marchetti, J.-F. Joanny, S. Ramaswamy, T. B. Liverpool, J. Prost, M. Rao, and R. A. Simha, Hydrodynamics of soft active matter, Reviews of modern physics 85, 1143 (2013).DOI
  4. [4]C. Bechinger, R. Di Leonardo, H. Löwen, C. Reichhardt, G. Volpe, and G. Volpe, Active particles in complex and crowded environments, Reviews of Modern Physics 88, 045006 (2016).DOI
  5. [5]V. A. Shaik, Z. Peng, J. F. Brady, and G. J. Elfring, Confined active matter in external fields, Soft Matter 19, 1384 (2023).DOI
  6. [6]S. Ramaswamy, R. A. Simha, and J. Toner, Active nematics on a substrate: Giant number fluctuations and long-time tails, Europhysics Letters 62, 196 (2003).DOI
  7. [7]S. Ramaswamy, Active matter, Journal of Statistical Mechanics: Theory and Experiment 2017, 054002 (2017).
  8. [8]J. Prost, F. Jülicher, and J.-F. Joanny, Active gel physics, Nature physics 11, 111 (2015).DOI
  9. [9]A. Doostmohammadi, J. Igńes-Mullol, J. M. Yeomans, and F. Sagués, Active nematics, Nature communications 9, 3246 (2018).DOI
  10. [10]L. Giomi, M. J. Bowick, P. Mishra, R. Sknepnek, and M. Cristina Marchetti, Defect dynamics in active nematics, Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences 372, 20130365 (2014).DOI
  11. [11]L. Balasubramaniam, R.-M. Mège, and B. Ladoux, Active nematics across scales from cytoskeleton organization to tissue morphogenesis, Current Opinion in Genetics & Development 73, 101897 (2022).
  12. [12]K. Kruse, J.-F. Joanny, F. Jülicher, J. Prost, and K. Sekimoto, Masters, vortices, and rotating spirals in active gels of polar filaments, Physical review letters 92, 078101 (2004).pubmed.ncbi.nlm.nih.gov/14995891
  13. [13]K. Kruse, J.-F. Joanny, F. Jülicher, J. Prost, and K. Sekimoto, Generic theory of active polar gels: a paradigm for cytoskeletal dynamics, The European Physical Journal E 16, 5 (2005).pubmed.ncbi.nlm.nih.gov/15688136
  14. [14]G. Wang, T. V. Phan, S. Li, M. Wombacher, J. Qu, Y. Peng, G. Chen, D. I. Goldman, S. A. Levin, R. H. Austin, et al., Emergent field-driven robot swarm states, Physical review letters 126, 108002 (2021).DOI
  15. [15][15*] M. E. Cates, Diffusive transport without detailed balance: does microbiology need statistical physics?, Reports on Progress in Physics 75, 042601 (2012).DOI
  16. [16]É. Fodor and M. C. Marchetti, The statistical physics of active matter: From self-catalytic colloids to living cells, Physica A: Statistical Mechanics and its Applications 504, 106 (2018).DOI
  17. [17]L. Touzo and P. Le Doussal, Run-and-tumble particles with 1D Coulomb interaction: The active jellium model and the non-reciprocal self-gravitating gas, Journal of Statistical Mechanics: Theory and Experiment 2025, 073204 (2025).arxiv.org/abs/2502.09466
  18. [18]L. Touzo, P. Le Doussal, and G. Schehr, Interacting, running and tumbling: The active Dyson Brownian motion, Europhysics Letters 142, 61004 (2023).DOI
  19. [19]L. Touzo, P. Le Doussal, and G. Schehr, Spatio-temporal fluctuations in the passive and active Riesz gas on the circle, Journal of Statistical Physics 192, 79 (2025).
  20. [20]É. Fodor, C. Nardini, M. E. Cates, J. Tailleur, P. Visco, and F. Van Wijland, How far from equilibrium is active matter?, Physical review letters 117, 038103 (2016).DOI
  21. [21]Y. Fily and M. C. Marchetti, Athermal phase separation of self-propelled particles with no alignment, Physical review letters 108, 235702 (2012).DOI
  22. [22]M. E. Cates and J. Tailleur, Motility-induced phase separation, Annu. Rev. Condens. Matter Phys. 6, 219 (2015).DOI
  23. [23]I. Buttinoni, J. Bialké, F. Kümmel, H. Löwen, C. Bechinger, and T. Speck, Dynamical clustering and phase separation in suspensions of self-propelled colloidal particles, Physical Review Letters 110, 238301 (2013).DOI
  24. [24]Y. Fily, S. Henkes, and M. C. Marchetti, Freezing and phase separation of self-propelled disks, Soft Matter 10, 2132 (2014).DOI
  25. [25]J. O’Byrne, A. P. Solon, J. Tailleur, and Y. Zhao, An introduction to motility-induced phase separation, in Out-of-Equilibrium Soft Matter: Active Fluids (The Royal Society of Chemistry, London, UK, 2023) pp. 1–45.
  26. [26]D. Martin, J. O’Byrne, M. E. Cates, É. Fodor, C. Nardini, J. Tailleur, and F. Van Wijland, Statistical mechanics of active ornstein-uhlenbeck particles, Physical Review E 103, 032607 (2021).DOI
  27. [27]J. Tailleur and M. E. Cates, Statistical mechanics of interacting run-and-tumble bacteria, Physical review letters 100, 218103 (2008).DOI
  28. [28]J. Bialké, H. Löwen, and T. Speck, Microscopic theory for the phase separation of self-propelled repulsive disks, EPL (Europhysics Letters) 103, 30008 (2013).DOI
  29. [29]R. Wittkowski, A. Tiribocchi, J. Stenhammar, R. J. Allen, D. Marenduzzo, and M. E. Cates, Scalar φ4 field theory for active-particle phase separation, Nature Communications 5, 4351 (2014).DOI
  30. [30][30*] A. P. Solon, J. Stenhammar, M. E. Cates, Y. Kafri, and J. Tailleur, Generalized thermodynamics of motility-induced phase separation: phase equilibria, Laplace pressure, and change of ensembles, New Journal of Physics 20, 075001 (2018).DOI
  31. [31]A. P. Solon, J. Stenhammar, M. E. Cates, Y. Kafri, and J. Tailleur, Generalized thermodynamics of phase equilibria in scalar active matter, Physical Review E 97, 020602 (2018).DOI
  32. [32]A. B. Slowman, M. R. Evans, and R. A. Blythe, Jamming and attraction of interacting run-and-tumble random walkers, Physical Review Letters 116, 218101 (2016).DOI
  33. [33]A. B. Slowman, M. R. Evans, and R. A. Blythe, Exact solution of two interacting run-and-tumble random walkers with finite tumble duration, Journal of Physics A: Mathematical and Theoretical 50, 375601 (2017).DOI
  34. [34]E. Mallmin, R. A. Blythe, and M. R. Evans, Exact spectral solution of two interacting run-and-tumble particles on a ring lattice, Journal of Statistical Mechanics: Theory and Experiment 2019, 013204 (2019).DOI
  35. [35]A. Das, A. Dhar, and A. Kundu, Gap statistics of two interacting run and tumble particles in one dimension, Journal of Physics A: Mathematical and Theoretical 53, 345003 (2020).DOI
  36. [36]M. J. Metson, M. R. Evans, and R. A. Blythe, From a microscopic solution to a continuum description of interacting active particles, Physical Review E 107, 044134 (2023).
  37. [37]M. J. Metson, M. R. Evans, and R. A. Blythe, Tuning attraction and repulsion between active particles through persistence, EPL (Europhysics Letters) 141, 41001 (2023).DOI
  38. [38]P. Le Doussal, S. N. Majumdar, and G. Schehr, Stationary nonequilibrium bound state of a pair of run and tumble particles, Physical Review E 104, 044103 (2021).
  39. [39]P. Dolai, S. Krekels, and C. Maes, Inducing a bound state between active particles, Physical Review E 105, 044605 (2022).DOI
  40. [40]L. Hahn, A. Guillin, and M. Michel, Jamming pair of general run-and-tumble particles: Exact results and universality classes, arXiv preprint arXiv:2306.00831 (2023).arxiv.org/abs/2306.00831
  41. [41]L. Hahn, Steady state and mixing of two run-and-tumble particles interacting through jamming and attractive forces, arXiv preprint arXiv:2501.11379 (2025).arxiv.org/abs/2501.11379
  42. [42]S. Put, J. Berx, and C. Vanderzande, Non-gaussian anomalous dynamics in systems of interacting run-and-tumble particles, Journal of Statistical Mechanics: Theory and Experiment 2019, 123205 (2019).
  43. [43]P. Singh and A. Kundu, Crossover behaviours exhibited by fluctuations and correlations in a chain of active particles, Journal of Physics A: Mathematical and Theoretical 54, 305001 (2021).arxiv.org/abs/2012.13910
  44. [44]S. Prakash, U. Basu, and S. Sabbahpandit, Tagged particle behavior in a harmonic chain of direction reversing active brownian particles, Journal of Statistical Mechanics: Theory and Experiment 2024, 083211 (2024).DOI
  45. [45]S. Paul, A. Dhar, and D. Chaudhuri, Dynamical crossovers and correlations in a harmonic chain of active particles, Soft Matter 20, 8638 (2024).arxiv.org/abs/2402.11358
  46. [46]M. Kourbane-Houssine, C. Erignoux, T. Bodineau, and J. Tailleur, Exact hydrodynamic description of active lattice gases, Physical review letters 120, 268003 (2018).
  47. [47]T. Agranov, S. Ro, Y. Kafri, and V. Lecomte, Exact fluctuating hydrodynamics of active lattice gases—typical fluctuations, Journal of Statistical Mechanics: Theory and Experiment 2021, 083208 (2021).DOI
  48. [48]T. Agranov, S. Ro, Y. Kafri, and V. Lecomte, Macroscopic fluctuation theory and current fluctuations in active lattice gases, SciPost Physics 14, 045 (2023).arxiv.org/abs/2208.02124
  49. [49]I. Mukherjee, A. Raghu, and P. K. Mohanty, Nonexistence of motility induced phase separation transition in one dimension, SciPost Physics 14, 165 (2023).DOI
  50. [50]A. G. Thompson, J. Tailleur, M. E. Cates, and R. A. Blythe, Lattice models of nonequilibrium bacterial dynamics, Journal of Statistical Mechanics: Theory and Experiment 2011, P02029 (2011).DOI
  51. [51]R. Dandekar, S. Chakraborti, and R. Rajesh, Hard core run and tumble particles on a one-dimensional lattice, Physical Review E 102, 062111 (2020).DOI
  52. [52]R. Mukherjee, S. Saha, T. Sadhu, A. Dhar, and S. Sabhapandit, Hydrodynamics of a hard-core active lattice gas, Physical Review E 111, 024128 (2025).DOI
  53. [53]E. Woillez, Y. Kafri, and N. S. Gov, Active trap model, Physical Review Letters 124, 118002 (2020).DOI
  54. [54]E. Woillez, Y. Kafri, and V. Lecomte, Nonlocal stationary probability distributions and escape rates for an active ornstein–uhlenbeck particle, Journal of Statistical Mechanics: Theory and Experiment 2020, 063204 (2020).
  55. [55]T. Debnath and P. K. Ghosh, Activated barrier crossing dynamics of a janus particle carrying cargo, Physical Chemistry Chemical Physics 20, 25069 (2018).DOI
  56. [56]Y. Fily, The journal of chemical physics, The Journal of Chemical Physics 150, 174906 (2019).
  57. [57]L. Caprini, F. Cecconi, and U. Marini Bettolo Marconi, Correlated escape of active particles across a potential barrier, The Journal of Chemical Physics 155 (2021).DOI
  58. [58]A. Scacchi, J. M. Brader, and A. Sharma, Escape rate of transiently active brownian particle in one dimension, Physical Review E 100, 012601 (2019).DOI
  59. [59]A. Dhar, A. Kundu, S. N. Majumdar, S. Sabhapandit, and G. Schehr, Run-and-tumble particle in one-dimensional confining potentials: Steady-state, relaxation, and first-passage properties, Physical Review E 99, 032132 (2019).DOI
  60. [60]D. Wexler, N. Gov, K. Ø. Rasmussen, and G. Bel, Dynamics and escape of active particles in a harmonic trap, Physical review research 2, 013003 (2020).
  61. [61]S. Gu, T. Qian, H. Zhang, and X. Zhou, Stochastic dynamics of an active particle escaping from a potential well, Chaos: An Interdisciplinary Journal of Nonlinear Science 30 (2020).DOI
  62. [62]S. Chaki and R. Chakrabarti, Escape of a passive particle from an activity-induced energy landscape: emergence of slow and fast effective diffusion, Soft Matter 16, 7103 (2020).DOI
  63. [63]B. Walter, G. Pruessner, and G. Salbreux, First passage time distribution of active thermal particles in potentials, Physical Review Research 3, 013075 (2021).
  64. [64]T. GrandPre and D. T. Limmer, Current fluctuations of interacting active brownian particles, Physical Review E 98, 060601 (2018).DOI
  65. [65]K. S. Olsen, L. Angheluta, and E. G. Flekkøy, Escape problem for active particles confined to a disk, Physical Review Research 2, 043314 (2020).DOI
  66. [66]M. Paoluzzi, L. Angelani, and A. Puglisi, Narrow-escape time and sorting of active particles in circular domains, Physical Review E 102, 042617 (2020).DOI
  67. [67]A. Biswas, J. Cruz, P. Parmananda, and D. Das, First passage of an active particle in the presence of passive crowders, Soft Matter 16, 6138 (2020).
  68. [68]Z. Schuss, A. Singer, and D. Holcman, The narrow escape problem for diffusion in cellular microdomains, Proceedings of the National Academy of Sciences 104, 16098 (2007).pubmed.ncbi.nlm.nih.gov/17901203
  69. [69]E. Locatelli, F. Baldovin, E. Orlandini, and M. Pierno, Active brownian particles escaping a channel in single file, Physical Review E 91, 022109 (2015).pubmed.ncbi.nlm.nih.gov/25768460
  70. [70]P. K. Ghosh, Communication: Escape kinetics of self-propelled janus particles from a cavity: Numerical simulations, The Journal of chemical physics 141 (2014).DOI
  71. [71]U. M. B. Marconi and C. Maggi, Towards a statistical mechanical theory of active fluids, Soft matter 11, 8768 (2015).DOI
  72. [72]C. Maggi, U. M. B. Marconi, N. Gnan, and R. Di Leonardo, Multidimensional stationary probability distribution for interacting active particles, Scientific reports 5, 10742 (2015).DOI
  73. [73]L. Touzo and P. Le Doussal, Nonuniqueness of the steady state for run-and-tumble particles with a double-well interaction potential, Physical Review E 113, 024106 (2026).DOI
  74. [74]L. Caprini, U. Marini Bettolo Marconi, A. Puglisi, and A. Vulpiani, Active escape dynamics: The effect of persistence on barrier crossing, The Journal of chemical physics 150 (2019).DOI
  75. [75]R. Parashar and S. Dutta, Confinement geometry regulates fluctuations, densities, and collective escape in interacting active matter, Physica Scripta (2026), in press.DOI
  76. [76]L. Dabelow, S. Bo, and R. Eichhorn, Irreversibility in active matter systems: Fluctuation theorem and mutual information, Physical Review X 9, 021009 (2019).DOI
  77. [77]L. Berthier, E. Flenner, and G. Szamel, New journal of physics, New Journal of Physics 19, 125006 (2017).
  78. [78]D. Mandal, K. Klymko, and M. R. DeWeese, D. mandal, k. klymko and m. r. deweese, physical review letters, 2017, 119, 258001., Physical Review Letters 119, 258001 (2017).
  79. [79]G. P. Nguyen, R. Wittmann, and H. Löwen, Active ornstein–uhlenbeck model for self-propelled particles with inertia, Journal of Physics: Condensed Matter 34, 035101 (2021).DOI
  80. [80]L. Caprini and U. M. B. Marconi, Active particles under confinement and effective force generation among surfaces, Soft matter 14, 9044 (2018).arxiv.org/abs/1812.03694
  81. [81]P. Jung and P. Hänggi, Dynamical systems: a unified colored-noise approximation, Physical review A 35, 4464 (1987).pubmed.ncbi.nlm.nih.gov/9898048
  82. [82]P.-H. Chavanis, Hamiltonian and brownian systems with long-range interactions: V. stochastic kinetic equations and theory of fluctuations, Physica A: Statistical Mechanics and its Applications 387, 5716 (2008).DOI
  83. [83]R. E. Spinney and I. J. Ford, Entropy production in full phase space for continuous stochastic dynamics, Physical Review E 85, 051113 (2012).DOI

Paper details

vlnc:W_yapexmdb
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 3

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

ChatGPT, Grammarly

View statement
Looking for related papers…
Contents