betaFeedback

WorksPhysics and Astronomy

Interfacial Structure and Dynamics of Active-Passive Fluid Mixtures

Abstract

We investigate the interface between an active and a passive fluid using molecular dynamics simulations of active Brownian particles in contact with a passive Lennard-Jones fluid. We compute the surface tension via two independent routes, pressure anisotropy and capillary-wave analysis, and find that they disagree systematically once activity is switched on. Static density and orientation profiles reveal three distinct activity regimes, with particles preferentially oriented toward the interface. Going beyond these static properties, we characterize the interface dynamics: a threshold activity above which mixing becomes irreversible, coinciding with the onset of the mixing regime, and a transient overshoot in interfacial width before relaxation to steady state.

Introduction

Active particles are a prominent class of intrinsically out-of-equilibrium systems, with realizations ranging from synthetic microswimmers to bacterial suspensions and sperm. Because they convert stored or energy available from their surroundings into directed motion, they are promising for microrheology, drug delivery, and micromachines [1, 2, 3, 4, 5]. Most studies, however, address bulk properties, whereas many applications depend just as strongly on the behavior of active fluids near interfaces.

Consequently, the surface tension of active systems has attracted growing attention, most commonly at free interfaces between an active fluid and vacuum [6, 7, 8], in motility-induced phase separation [9, 10, 11, 12, 13, 14, 15, 16], in edge-current settings [17, 18, 19, 20, 21, 22], in wetting [23, 24, 25, 26, 27], and via capillary rise [28]. These works report anomalous values, including negative surface tensions for interfaces that nonetheless remain mechanically stable [29, 30, 31]. This paradox was recently resolved by Ref. 32 for the active–active interface between coexisting dense and dilute phases of active Brownian particles, where densities are self-selected by activity.

Active–passive interfaces, by contrast, remain largely unexplored at the particle-resolved level. Previous work has studied the mixing and structure of active–passive fluids [33, 34, 35], propagating fronts [36], and kinetic roughening [37], while Ref. 38 described giant interfacial fluctuations, active waves, and droplet emulsification in a continuum, flow-dominated regime. None of these works quantifies the surface tension of the interface, tests whether its thermodynamic and capillary-wave definitions, equivalent in equilibrium, still agree under activity, or follows the interface after a sudden change of activity.

Here, we close this gap using three-dimensional molecular dynamics simulations of active Brownian particles in contact with a passive Lennard-Jones fluid. We address two questions. Statistically, we compare the thermodynamic and capillary-wave routes to the surface tension as a function of activity. Dynamically, we treat activity as a control parameter, analogous to a quench in passive systems: we repeatedly switch it on and off, identify a threshold activity above which mixing becomes irreversible on the timescale of our simulations, and resolve the time-dependent interfacial width Δ(t)\Delta(t) and position z0(t)z_0(t). This reveals a transient overshoot in the interfacial width that is invisible in static analyses.

The active–passive setup is motivated twofold. First, fluid–fluid interfaces are the relevant geometry in many biological and soft-matter applications, unlike the fluid–vacuum interfaces considered in most of the literature above. Second, a fluid–vacuum interface fixes the pressure on one side to zero, whereas an active–passive interface lets us probe surface tension at finite, tunable ambient pressure, a regime not systematically addressed for the two-route comparison in active systems.

Beyond the stationary properties, the dynamics of active–passive interfaces is of both fundamental and practical interest. In many realizations, activity is not fixed but can be switched externally, e.g. via light or fuel availability, so that the response to a sudden change of activity determines how fast and how controllably such systems can be driven. Moreover, the (ir)reversibility of activity-induced mixing probes memory and hysteresis, which are absent for interfaces in equilibrium, and the relaxation of the interface offers an independent test of whether equilibrium concepts such as an effective surface tension remain meaningful out of equilibrium.

The paper is organized as follows. We first present the simulation method. We then analyze static density profiles, which show three distinct activity regimes, compare them quantitatively with capillary-wave theory, and examine the particle orientation profiles. Finally, we turn to the interface dynamics, covering the reversibility of activity-driven mixing and the time-resolved relaxation of Δ(t)\Delta(t) and z0(t)z_0(t) after a sudden change in activity.

Methods

We consider a binary system consisting of two particle species, AA and BB. Along the zz-direction, the system comprises two bulk regions (one containing the AA-particles, the other the BB-particles) separated by two interfaces normal to zz. Periodic boundary conditions are applied in all three directions. A schematic view of the setup is shown in Figure 1.

Figure 1. Schematic view of the simulation setup. Blue particles correspond to the A-particles in this paper, the red ones correspond to the B-particles.

All particles interact via the Lennard-Jones potential

U(rij)=4ϵαβ[(σαβrij)12−(σαβrij)6],(1)U(r_{ij}) = 4\epsilon_{\alpha\beta} \left[\left(\frac{\sigma_{\alpha\beta}}{r_{ij}}\right)^{12} - \left(\frac{\sigma_{\alpha\beta}}{r_{ij}}\right)^6\right], \tag*{(1)}

where rijr_{ij} is the distance between particles ii and jj, and ϵαβ\epsilon_{\alpha\beta} and σαβ\sigma_{\alpha\beta} (α,β∈{A,B}\alpha,\beta\in\{A,B\}) set the energy and length scales, respectively. Any quantity can thus be expressed in units of σAA\sigma_{AA}, ϵAA\epsilon_{AA}, and the mass mm; as is common practice, these units are omitted below. We use σAA=σBB=σAB=1.0\sigma_{AA}=\sigma_{BB}=\sigma_{AB}=1.0, ϵAA=ϵBB=1.0\epsilon_{AA}=\epsilon_{BB}=1.0, and ϵAB=0.5\epsilon_{AB}=0.5. The value of ϵAB\epsilon_{AB} controls whether the system mixes or demixes [39].

The dynamics of particle ii are described by the standard model of active Brownian particles (ABPs) [4, 40]:

dridt=v0in^i−μi∇U(ri)+2DTi ηtrans(t),(2)\frac{d\mathbf{r}^{i}}{dt} = v_{0}^{i}\hat{\mathbf{n}}^{i} - \mu^{i}\nabla U(\mathbf{r}^{i}) + \sqrt{2D_{T}^{i}}\,\boldsymbol{\eta}_{\mathrm{trans}}(t), \tag*{(2)}
dθidt=2DRi ηrot(t).(3)\frac{d\theta^{i}}{dt} = \sqrt{2D_{R}^{i}}\,\eta_{\mathrm{rot}}(t). \tag*{(3)}

Here, v0iv_{0}^{i} is the self-propulsion velocity, μi\mu^{i} the mobility, DTiD_{T}^{i} and DRiD_{R}^{i} the translational and rotational diffusion coefficients, and n^i\hat{\mathbf{n}}^{i} the orientation of the particle. Therefore, nzn_{z} is the zz-component of the orientation. The noise terms η\eta are Gaussian white noise,

⟨η(t)⟩=0,⟨η(t)η(t′)⟩=δ(t−t′).(4)\langle\eta(t)\rangle= 0,\qquad\langle\eta(t)\eta(t')\rangle= \delta(t-t'). \tag*{(4)}

Each particle is thus an overdamped Brownian particle that can additionally propel itself. Because inertia is negligible, the velocity of a particle is set directly by the forces acting on it, weighted by its mobility μi\mu^{i}. The first equation contains three contributions to this velocity: (i) the term v0in^iv_{0}^{i}\hat{\mathbf{n}}^{i} describes self-propulsion at constant speed along the instantaneous orientation n^i=(cos⁡θi,sin⁡θi)\hat{\mathbf{n}}^{i}=(\cos\theta^{i},\sin\theta^{i}); it represents the conversion of energy from the environment into directed motion and drives the system out of equilibrium; (ii) the term −μi∇U-\mu^{i}\nabla U is the deterministic drift due to the interaction potential, e.g. steric repulsion between particles or an external potential; (iii) the noise term 2DTiηtrans\sqrt{2D_{T}^{i}}\eta_{\mathrm{trans}} accounts for thermal fluctuations of the surrounding solvent and leads to ordinary translational diffusion with coefficient DTiD_{T}^{i}.

Rotational diffusion gradually randomizes the direction of self-propulsion, so that orientational correlations decay exponentially, ⟨n^(t)⋅n^(0)⟩=e−DRit\langle\hat{\mathbf{n}}(t)\cdot\hat{\mathbf{n}}(0)\rangle=e^{-D_{R}^{i}t} in two dimensions. This defines a persistence time τRi=1/DRi\tau_{R}^{i}=1/D_{R}^{i} and a persistence length ℓpi=v0i/DRi\ell_{p}^{i}=v_{0}^{i}/D_{R}^{i}: on short times the particle moves ballistically, while on long times its motion is diffusive with an enhanced effective diffusion coefficient

Deffi=DTi+(v0i)22DRi.D_{\mathrm{eff}}^{i}=D_{T}^{i}+\frac{(v_{0}^{i})^{2}}{2D_{R}^{i}}.

For v0i=0v_{0}^{i}=0, self-propulsion vanishes and the model reduces to a passive Brownian particle, which relaxes to thermal equilibrium. The two species can differ in activity, mobility, and diffusivities, which allows us to describe mixtures of active and passive particles, as well as particles with different levels of activity.

The simulations were performed with the open-source, GPU-accelerated molecular dynamics package gampy [41]. The system was initialized on a face-centered-cubic lattice with 15×15×3015\times15\times30 unit cells at a target number density ρ=0.8\rho=0.8. This gives a lattice constant a=(4/ρ)1/3≈1.71a=(4/\rho)^{1/3}\approx1.71, box dimensions Lx=Ly=15a≈25.6L_{x}=L_{y}=15a\approx25.6 and Lz=30a≈51.3L_{z}=30a\approx51.3, and N=27000N=27000 particles in total. Particles in the top half of the box (z>Lz/2z>L_{z}/2) were assigned to species AA, those in the bottom half to species BB.

Unless stated otherwise, we use DTA=DTB=0.005D_{T}^{A}=D_{T}^{B}=0.005, DRA=DRB=1.0D_{R}^{A}=D_{R}^{B}=1.0, and μA=μB=0.05\mu^{A}=\mu^{B}=0.05. The equations of motion were integrated with a timestep dt=0.005dt=0.005. After an equilibration of duration t=103t=10^{3}, we performed production runs of duration t=107t=10^{7}. All reported density profiles are averages over 5 independent initial configurations. For the study of the system’s dynamics in section III B 1, each phase had a duration of t=512t=512. During the production runs, we record the density profiles of both particle species as well as the height-resolved profiles of the particle orientations.

We calculate the surface tension from the pressure anisotropy of the system,

σ^=∫dz[pz(z)−12(px(z)+py(z))],(5)\hat{\sigma}=\int dz\left[p_{z}(z)-\frac{1}{2}\left(p_{x}(z)+p_{y}(z)\right)\right], \tag*{(5)}

where pxp_{x}, pyp_{y}, and pzp_{z} are the pressure components in the respective directions, determined from the virial (see [42] for details).

To compare with capillary-wave theory [43], we fit the density profiles to a tanh function,

ρfit(z)=ρ1+ρ22+ρ1−ρ22tanh⁡(z−z0Δ).(6)\rho_{\mathrm{fit}}(z)=\frac{\rho_{1}+\rho_{2}}{2}+\frac{\rho_{1}-\rho_{2}}{2}\tanh\left(\frac{z-z_{0}}{\Delta}\right). \tag*{(6)}

We call Δ\Delta the width of the density profile and z0z_{0} the position of the interface. Note that due to symmetry reasons, we continue an analysis of only one of the surfaces.

Results

Static results

Density profiles

In Figure 2 we show the resulting density profiles of the active particles for different self-propulsion velocities v0v_{0}, at a fixed translational diffusion coefficient DTA=0.005D_{T}^{A}=0.005. It is convenient to express the activity in terms of a translational Péclet number,

Density profiles of the active and passive particles

Figure 2. Density profiles of the active (colored) and passive (gray) particles for DTA=0.005D_T^A = 0.005.

Pet=v0σAADTA,(7)Pe_t = \frac{v_0 \sigma_{AA}}{D_T^A}, \tag*{(7)}

which quantifies the self-propulsion velocity relative to thermal diffusion.

Two observations stand out immediately. First, the density profiles of the passive B-particles remain essentially unchanged across the entire range of activities studied. This is expected: activity enters the particle dynamics of (3) only through the self-propulsion term v0n^iv_0\hat{n}^i, which is identically zero for the passive species; the B-particles are affected by the activity of the A-particles only indirectly, through pair interactions localized at the interface, and this local perturbation is too weak to alter the passive bulk density profile appreciably. Second, and in contrast, the activity of the A-particles has a pronounced effect on their own density profile, and we can identify three qualitatively distinct regimes.

For very small activities, v0<0.35v_0 < 0.35, corresponding to Pet<70Pe_t < 70, the density profile of the active particles remains nearly indistinguishable from the passive case: thermal diffusion dominates over self-propulsion, and the interface behaves essentially as in an equilibrium binary fluid. For intermediate activities, 0.35<v0<0.60.35 < v_0 < 0.6 (corresponding to 70<Pet<12070 < Pe_t < 120) the density profile progressively broadens as v0v_0 increases, signaling that the active particles increasingly accumulate against and press into the interfacial region. Finally, for large activities, v0>0.6v_0 > 0.6, the density profile broadens so much that active and passive particles coexist within a particular region; we refer to this as a mixing regime. Notably, even in this regime the density profile of the passive particles remains unchanged, which shows that the mixing is asymmetric: active particles invade the passive region, but the reverse process that passive particles invading the active region does not occur to a comparable degree.

The bulk plateau value of the active-particle density profile, ρbulk≈330\rho_{\mathrm{bulk}} \approx330 (in units of particle number per bin), is consistent with the nominal number density ρtotal=0.8\rho_{\mathrm{total}} = 0.8 and the box cross-section Lx×LyL_x \times L_y, confirming that the systems remain well equilibrated away from the interface irrespective of activity. As we show in the following section, the width of the interfacial region continues to grow with DTAD_T^A even at fixed v0v_0, and the activity at which the crossover into the mixing regime occurs shifts to smaller v0v_0 as DTAD_T^A increases which is consistent with the scaling suggested by PetPe_t in Eq. (7).

Figure 3 presents a systematic analysis of the density profiles and the surface tension as functions of activity. For various combinations of DTAD_T^A and v0v_0, we fitted the density profiles with Eq. (6) to extract the interfacial width Δ\Delta and position z0z_0.

Figure 3. Density width, position and surface tension against activity for different thermal diffusion coefficients.

Panel (a) shows the width of the density profile as a function of v0v_0 for different values of DTAD_T^A. The behavior is strikingly non-monotonic. For small v0v_0 and DTAD_T^A, the width remains essentially constant, consistent with the near-equilibrium regime identified in Sec. III A 1, where the density profile is barely perturbed by activity. As DTAD_T^A increases, this constant regime progressively shrinks and eventually disappears. For larger activities, the width first grows, reaches a pronounced peak, and then decreases again. Notably, the location of this peak shifts to smaller values of v0v_0 as DTAD_T^A increases, in line with the scaling suggested by the Péclet number Pet=v0σ/DTAPe_t = v_0 \sigma/D_T^A introduced above.

This non-monotonic behavior is itself a fingerprint of activity at a finite-pressure interface. At a free active–vacuum interface, the width of the density profile is expected to grow monotonically with activity until the interface eventually becomes mechanically unstable. In our simulations, by contrast, the active particles cannot spread indefinitely: once they reach the passive fluid, the finite density and pressure of the passive phase set an effective upper bound on how far the interface can broaden, so that Δ\Delta saturates and subsequently decreases again as the system enters the mixing regime. Consistent with this picture, the active–passive interface in our simulations remains mechanically stable up to substantially higher activities than what is reported for active–vacuum interfaces, where instability sets in far earlier [35, 44].

Panel (b) shows the corresponding position of the interface, z0z_0, which reveals the same three regimes identified above: for very small and for very large activities, z0z_0 is essentially independent of v0v_0, while in the intermediate regime the interface position shifts systematically toward the passive-particle side. This shift indicates that, before the onset of full mixing, the active fluid is already displacing the passive fluid and effectively gaining volume at its expense.

Finally, panel (c) shows the surface tension σ^\hat{\sigma}, obtained independently from the pressure anisotropy, Eq. (5). Here the trend differs qualitatively from that of Δ\Delta: σ^\hat{\sigma} is maximal at vanishing activity, decreases monotonically with v0v_0, and saturates at a small but non-zero plateau value of σ^≈0.05\hat{\sigma} \approx0.05–0.070.07 once the mixing regime is reached. Remarkably, this plateau value is essentially independent of both v0v_0 and DTAD_T^A, suggesting that it reflects a generic property of the fully mixed active–passive interface rather than a activity-dependent quantity. The qualitative disagreement between the non-monotonic behavior of Δ\Delta and the monotonic decay of σ^\hat{\sigma} already indicates that density width and surface tension are not simply proportional to one another once activity is present; we examine this discrepancy quantitatively in the next section.

Comparison to capillary-wave theory

The surface tension extracted from molecular dynamics simulations can, in principle, be obtained via two independent routes. The first is the thermodynamic, pressure-anisotropy definition of Eq. (5). The second follows from capillary-wave theory, which predicts that thermally excited fluctuations of the interface broaden the density profile according to [43]

Δ2=Δ02+kBT2πσ^log⁡(L/B0):=A+Bσ^−1,(8)\Delta^2 = \Delta_0^2 + \frac{k_B T}{2\pi\hat{\sigma}} \log(L/B_0) := A + B\hat{\sigma}^{-1}, \tag*{(8)}

where Δ0\Delta_0 is an intrinsic offset width, LL is the lateral size of the simulation box, and B0B_0 is a microscopic length scale. (8) predicts a linear relation between Δ2\Delta^2 and the inverse surface tension σ^−1\hat{\sigma}^{-1}, with a slope set by temperature alone. This is a relation that is well established in equilibrium fluid-fluid interfaces but has to our knowledge not been systematically tested for active-passive interfaces.

Figure 4 tests this prediction against all data presented in this paper, plotting Δ2\Delta^2 obtained from the tanh-fits of Sec. III A 1 against σ^−1\hat{\sigma}^{-1} obtained independently from the pressure anisotropy. For small activities, the data indeed collapse onto the linear relation predicted by Eq. (8), confirming that capillary-wave theory correctly captures the interfacial fluctuations of the active-passive system in the near-equilibrium regime. As activity increases, however, the data increasingly deviate from this linear relation. This is consistent with the qualitative disagreement between Δ(v0)\Delta(v_0) and σ^(v0)\hat{\sigma}(v_0) already noted in panels (a) and

(c) of Figure 2: once the density profile broadens due to genuine mixing of the two species, rather than due to thermally excited capillary fluctuations alone, Δ\Delta ceases to be a faithful proxy for σ^\hat{\sigma}, and the two routes to the surface tension no longer coincide.

This breakdown highlights a more fundamental ambiguity that is specific to active systems: whereas the mechanical (pressure-based) and fluctuation-based (capillary-wave) definitions of surface tension are equivalent in equilibrium, this equivalence relies on the assumption that pressure is a well-defined, isotropic bulk quantity away from the interface. Note that this is an assumption that is itself non-trivial for active matter, where swim pressure and interparticle stresses need not combine into a conventional thermodynamic pressure [6]. Our results suggest that this ambiguity becomes appreciable precisely when the interface is broadened by activity-driven mixing rather than by thermal capillary waves, offering a concrete diagnostic for when the two definitions of surface tension can be expected to agree.

Figure 4. Test of the data against the prediction from capillary-wave theory.

Orientations of the particles

Finally, in Figure 5 we examine one of the defining properties of active matter that has no equilibrium counterpart: the orientation of the self-propelled particles. The height-resolved orientation profile reveals two distinct features. In the bulk region, far from either interface, the mean orientation ⟨nz⟩\langle n_z \rangle fluctuates around zero, as expected: in the absence of any symmetry-breaking field, the swimming direction is uniformly distributed and no net orientation should persist on average. In the interfacial region, by contrast, the orientation develops a pronounced net component directed toward the interface, with active particles preferentially oriented so as to swim into the passive phase.

Figure 5. Height resolved orientation of the particles nzn_z (full line) and density profile ρ\rho (dashed line) for different self-propelling velocities v0v_0 and constant DTA=0.025D_T^A = 0.025. The inset shows the maximum of the orientation profile in relation to v0v_0.

This interfacial alignment has a simple mechanistic origin. Away from the interface, an active particle’s orientation evolves freely via rotational diffusion, Eq. (3), unconstrained by its neighbors. Close to the interface, however, particles oriented away from the interface are pushed back into the bulk by steric repulsion from their neighbors, while particles oriented toward the interface continue to propel themselves forward unimpeded. This asymmetry effectively biases the local population of orientations toward the interface, even though the underlying rotational dynamics, in Eq. (3), remains isotropic. The same mechanism is responsible for particle accumulation at any confining or repulsive boundary in active systems [45], and here it operates locally at the active-passive interface itself.

The inset of Figure 5 shows the maximal value of this orientation, max⁡(⟨nz⟩)\max(\langle n_z\rangle), as a function of v0v_0 for different values of DTAD_T^A. Strikingly, the data for all values of DTAD_T^A collapse onto a single curve, indicating that the degree of interfacial alignment is controlled by the self-propulsion velocity alone and is essentially insensitive to translational diffusion. The collapsed data furthermore grow approximately linearly with v0v_0, consistent with the interpretation above: since the propulsive flux of particles into the interface scales directly with v0v_0, while the counteracting steric restoring force is set by the (activity-independent) interparticle interactions, the degree of orientational bias inherits the linear dependence on v0v_0 directly.

Dynamic results

In this part, we investigate the dynamics of active-passive interfaces. This section is divided into 2 different parts: Firstly, we investigate the reversibility of active-passive interfaces. After that, we investigate how the density profile at the interface develops.

Reversibility Threshold

To probe the reversibility of the active-passive interface, we perform simulations in which the activity is repeatedly switched on and off, tracking the density profile at each stage. The protocol is illustrated in Figure 6: after equilibration, the system is first held in a passive state, denoted without (0); activity is then switched on (with (1)), subsequently switched off again (without (1)), and so on. If the underlying dynamics were fully reversible, the density profile recorded in every without stage should be identical, independent of how many activity cycles precede it.

Figure 6. Density profiles for simulations where the activity was switched on and off (DTA=0.005D_T^A = 0.005). One can see a transition between a reversible regime and a non-reversible regime: While for small activities, the passive density profile can be recovered when the activity is switched off, this is not the case for large self-propelling velocities.

This is indeed what we observe for small activities, v0<0.4v_0 < 0.4: the passive density profile is recovered essentially unchanged after each activation-deactivation cycle. For larger activities, however, the without-profiles from successive stages no longer coincide, indicating that the system fails to return to its original state once activity has been switched off – the interface has been irreversibly altered.

This loss of reversibility coincides with the onset of mixing identified from the steady-state density profiles in § 3 A 1: for low activities, where the active and passive fluids remain demixed at steady state, the process is fully reversible, whereas for larger v0v_0, where active and passive particles coexist within an extended interfacial region, the process becomes irreversible. This correspondence suggests that irreversibility is a direct consequence of mixing: Once active particles have penetrated sufficiently deep into the passive phase, purely diffusive relaxation after deactivation is no longer sufficient to re-establish the original demixed configuration on the timescale of our simulations.

Activity timescale

Having established that switching activity on and off within a single configuration does not yield reversible statistics once mixing has occurred, we instead characterize the transient dynamics of interface formation using an ensemble approach: we prepare 1000 independent initial configurations, each equilibrated in the passive state, and impose a fixed activity v0v_0 on all A-particles simultaneously at t=0t = 0. This avoids the irreversibility artifacts discussed in the previous section, at the cost of deviating from the switch-on/switch-off protocol commonly used to probe relaxation in passive systems [46, 47]. Figure 7 shows the resulting time evolution of the density profile for several representative activities.

Figure 7. Density profiles for different activities at different times for a constant DTA=0.005D_T^A = 0.005. Black dashed lines correspond to the fits with formula (6).

Figure 8 quantifies this transient by tracking the fitted width Δ(t)\Delta(t) and interface position z0(t)z_0(t), obtained from Eq. (6), as functions of time for DTA=0.005D_T^A = 0.005. Panel (a) shows a clear qualitative distinction between low- and high-activity systems. For low activities, the width remains essentially constant throughout the entire simulation, consistent with the near-equilibrium, demixed steady state discussed above. For higher activities, by contrast, Δ(t)\Delta(t) first overshoots sharply, reaching a pronounced transient peak, before relaxing back down to a lower, quasi-stationary value. Both the height and the timing of this peak depend strongly on activity: more active systems overshoot earlier and more strongly, before relaxing to their (still activity-dependent) steady-state width. This transient overshoot suggests that the active particles initially invade the passive phase faster than the interface can mechanically respond, before the system settles into the broadened, but stable, steady state characterized in § 3 A 1.

Figure 8. Time development of the width and the position of the density profile in time for a constant DTA=0.005D_T^A = 0.005

Panel (b) shows the corresponding evolution of the interface position z0(t)z_0(t).

All systems approach their steady-state position monotonically, but the time required to do so depends systematically on v0v_0: lower-activity systems reach their steady state only after a substantially longer transient than higher-activity systems. This trend is intuitive, since the rate at which particles invade the passive region and hence the rate at which the interface itself advances, is set directly by the self-propulsion velocity.

Figure 9 shows the same analysis for a larger translational diffusion coefficient, DTA=0.025D_T^A = 0.025. The qualitative features, i.e. the transient overshoot in Δ\Delta and the activity-dependent approach to steady state in z0z_0, are reproduced, but occur at systematically shifted values of v0v_0: the same qualitative behavior seen at a given v0v_0 for DTA=0.005D_T^A = 0.005 now appears at correspondingly larger v0v_0 for DTA=0.025D_T^A = 0.025. This shift is consistent with the Péclet-number scaling, Pet=v0σ/DTAPe_t = v_0\sigma/D_T^A, already identified in the static analysis of § 3 A 1, and confirms that the same underlying control parameter governs both the static and the dynamic interfacial behavior.

Figure 9. Time development of the width and the position of the density profile in time for a constant DTA=0.025D_{T}^{A}=0.025

Summary and Outlook

In this work, we investigated the interface between an active and a passive fluid using 3-dimensional molecular dynamics simulations of active Brownian particles in contact with a passive Lennard-Jones fluid. Unlike the free active–vacuum interfaces studied previously, this setup allowed us to probe the interfacial behavior of active matter at finite ambient pressure. Our results contribute in two respects: a comparison of two routes to the surface tension at finite ambient pressure, and, to our knowledge for the first time, a particle-resolved characterization of the interface dynamics following a sudden change in activity.

Both the width and the position of the interfacial density profile depend non-monotonically on activity, with three distinct regimes: a near-equilibrium regime at low PetPe_t in which the interface is essentially unperturbed, an intermediate regime in which the active fluid progressively invades the passive phase, and a high-activity mixing regime in which active and passive particles coexist over an extended region. The surface tension obtained from the pressure anisotropy, by contrast, decays monotonically with activity and saturates at a small, non-zero plateau value that is independent of both v0v_0 and DTAD_T^A. This qualitative disagreement between the two observables shows that the width of the density profile is not simply proportional to the surface tension once activity is present, in contrast to the equilibrium case. Testing this connection quantitatively against capillary-wave theory, we found good agreement in the near-equilibrium regime, but systematic deviations once the system enters the mixing regime, indicating that the broadening of the interface beyond a certain activity is driven by genuine mixing rather than by thermally excited capillary fluctuations. We further characterized the orientation of the active particles across the interface, and found that particles preferentially orient toward the interface, with the degree of alignment controlled by activity alone. Going beyond these static properties, which have so far been the focus of studies of active–passive interfaces, we investigated the interfacial dynamics. The interface exhibits a threshold in activity beyond which the mixing process becomes irreversible: activity can no longer be switched off to recover the original, demixed state within the accessible simulation time. This loss of reversibility roughly coincides with the onset of the mixing regime identified in the static analysis, suggesting that both phenomena share a common origin. Tracking the time evolution of the density profile following a sudden onset of activity, we observed a pronounced transient overshoot in the interfacial width before the system relaxes to its steady state, with the timing and magnitude of this overshoot controlled by activity in a manner consistent with the Péclet-number scaling identified in the static case. To our knowledge, neither the reversibility threshold nor the time-resolved relaxation of the interfacial width and position has previously been characterized for active–passive fluid interfaces.

Taken together, our results demonstrate that the surface tension of an active-passive interface cannot, in general, be inferred from capillary-wave theory alone, and that the two standard routes to the surface tension, which coincide in equilibrium, need not agree once the system is driven out of equilibrium by activity, even when the interface remains mechanically stable throughout. This raises a more fundamental question that our results do not resolve: how pressure, and hence surface tension, should be defined in non-equilibrium active systems in the first place, given that swim pressure and interparticle stresses need not combine into a conventional, isotropic bulk pressure far from the interface [6, 48, 49, 50].

Several directions emerge naturally from this work. First, a systematic comparison of alternative pressure definitions (e.g., mechanical versus virial routes) at the active-passive interface would help clarify when the disagreement between thermodynamic and capillary-wave surface tensions can be attributed to a genuine breakdown of capillary-wave theory, as opposed to an ambiguity in the definition of pressure itself. Second, the present dynamic analysis was restricted to the interfacial width and position; a time-resolved measurement of the surface tension itself during the transient following the onset of activity would show whether the overshoot in the width is mirrored in the thermodynamic surface tension, and would extend the comparison of the two routes to the time domain. Third, our results were obtained at a single interaction asymmetry (ϵAB=0.5\epsilon_{AB}=0.5); a systematic study of a variation of ϵAB\epsilon_{AB} to tune the equilibrium miscibility of the two species would establish how robust the observed regimes are, and allow the transition into the mixing regime to be located more precisely. Fourth, a systematic variation of the activity of the second species from passive to as active as the first would show whether the mixing threshold shifts once both species can invade each other symmetrically. Finally, given that mixed active-passive interfaces of the kind studied here are relevant to biological systems in which motile and non-motile constituents coexist, extending this analysis to more realistic interaction potentials and confinement geometries would be a natural next step toward connecting these results to biological soft matter.

Acknowledgments

The author acknowledges fruitful discussions with Jeppe Dyre, Sumeja Bureković and Søren Toxværd. This work was supported by the Independent Research Fund Denmark through grant no. 5281-00148B.

References

  1. [1]D. Patra, S. Sengupta, W. Duan, H. Zhang, R. Pavlick, and A. Sen, Intelligent, self-powered, drug delivery systems, Nanoscale 5, 1273 (2013).DOI
  2. [2]C. Bechinger, R. Di Leonardo, H. Löwen, C. Reichardt, G. Volpe, and G. Volpe, Active particles in complex and crowded environments, Reviews of Modern Physics 88, 045006 (2016).DOI
  3. [3]A. Zöttl and H. Stark, Emergent behavior in active colloids, Journal of Physics: Condensed Matter 28, 253001 (2016).DOI
  4. [4]P. Romanczuk, M. Bär, W. Ebeling, B. Lindner, and L. Schimansky-Geier, Active Brownian particles: From individual to collective stochastic dynamics, The European Physical Journal Special Topics 202, 1 (2012).DOI
  5. [5]L. Wang, X. Xu, Z. Li, and T. Qian, Active Brownian particles simulated in molecular dynamics, Chinese Physics B 29, 090501 (2020).DOI
  6. [6]A. K. Omar, Z.-G. Wang, and J. F. Brady, Microscopic origins of the swim pressure and the anomalous surface tension of active matter, Physical Review E 101, 012604 (2020).DOI
  7. [7]N. Lauersdorf, T. Kolb, M. Moradi, E. Nazockdast, and D. Klotsa, Phase behavior and surface tension of soft active Brownian particles, Soft Matter 17, 6337 (2021).DOI
  8. [8]S. Paliwal, V. Prymidis, L. Filion, and M. Dijkstra, Nonequilibrium surface tension of the vapour-liquid interface of active lennard-jones particles, The Journal of Chemical Physics 147, 084902 (2017).DOI
  9. [9]M. E. Cates and C. Nardini, Active phase separation: new phenomenology from non-equilibrium physics, Reports on Progress in Physics 88, 056601 (2025).
  10. [10]G. Fausti, E. Tjhung, M. E. Cates, and C. Nardini, Capillary interfacial tension in active phase separation, Physical Review Letters 127, 068001 (2021).DOI
  11. [11]X.-q. Shi, G. Fausti, H. Chaté, C. Nardini, and A. Solon, Self-organized critical coexistence phase in repulsive active particles, Physical Review Letters 125, 168001 (2020).DOI
  12. [12]Z. Sun, L. Li, F. Ye, and M. Yang, Interfacial stability in tensionless phase-separated quorum-sensing systems, arXiv preprint arXiv:2507.15030 (2025).arxiv.org/abs/2507.15030
  13. [13]L. Langford and A. K. Omar, Theory of capillary tension and interfacial dynamics of motility-induced phases, Physical Review E 110, 054604 (2024).DOI
  14. [14]S. Bureković, F. De Luca, M. E. Cates, and C. Nardini, Active Cahn-Hilliard theory for nonequilibrium phase separation: Quantitative macroscopic predictions and a microscopic derivation, Physical Review Research 8, 033141 (2026).DOI
  15. [15]R. Fernández-Quevedo García, E. Chacón, P. Tarazona, and C. Valeriani, Dynamics and rupture of doped motility induced phase separation, Soft Matter 21, 5413 (2025).
  16. [16]J. Díaz and I. Pagonabarraga, Activity-driven emulsification of phase-separating binary mixtures, Physical Review Letters 134, 098301 (2025).DOI
  17. [17]B. C. Van Zuiden, J. Paulose, W. T. Irvine, D. Bartolo, and V. Vitelli, Spatiotemporal order and emergent edge currents in active spinner materials, Proceedings of the National Academy of Sciences 113, 12919 (2016).
  18. [18]F. Alsallom and D. T. Limmer, Origin of edge currents in chiral active liquids, arXiv preprint arXiv:2603.18159 (2026).arxiv.org/abs/2603.18159
  19. [19]B. Adorjáni, A. Libál, C. Reichardt, and C. J. O. Reichardt, Phase separation, edge currents, and hall effect for active matter with magnus dynamics, The European Physical Journal E 47, 40 (2024).DOI
  20. [20]Y. Zhou, Q. Yin, S. Nayak, P. Bag, P. K. Ghosh, Y. Li, and F. Marchesoni, Visual quorum sensing in chiral suspensions: Hyperuniformity and edge currents, PNAS nexus 4, pgaf373 (2025).DOI
  21. [21]L. Caprini and U. Marini Bettolo Marconi, Bubble phase induced by odd interactions in chiral systems, The Journal of Chemical Physics 162 (2025).
  22. [22]X. Yang, C. Ren, K. Cheng, and H. Zhang, Robust boundary flow in chiral active fluid, Physical Review E 101, 022603 (2020).DOI
  23. [23]N. Grodtzinski, M. E. Cates, and R. L. Jack, Hydrodynamic theory of wetting by active particles, Physical Review E 113, 065423 (2026).
  24. [24]U. Thiele, What is active wetting?, The European Physical Journal E 49, 56 (2026).DOI
  25. [25]F. Turci, R. L. Jack, and N. B. Wilding, Partial and complete wetting of droplets of active Brownian particles, Soft Matter 20, 2060 (2024).DOI
  26. [26]E. Chacón, F. Alarcón, J. Ramírez, P. Tarazona, and C. Valeriani, Intrinsic structure perspective for MIPS interfaces in two-dimensional systems of active Brownian particles, Soft Matter 18, 2646 (2022).arxiv.org/abs/2202.02032
  27. [27]Y. Zhao, R. Zakine, A. Daerr, Y. Kafri, J. Tailleur, and F. Van Wijland, Wetting by active fluids, Nature Physics 22, 805 (2026).DOI
  28. [28]A. Wysocki and H. Rieger, Capillary action in scalar active matter, Physical Review Letters 124, 048001 (2020).
  29. [29]J. Bialké, J. T. Siebert, H. Löwen, and T. Speck, Negative interfacial tension in phase-separated active Brownian particles, Physical Review Letters 115, 098301 (2015).pubmed.ncbi.nlm.nih.gov/26371685
  30. [30]S. Hermann, D. de las Heras, and M. Schmidt, Non-negative interfacial tension in phase-separated active Brownian particles, Phys. Rev. Lett. 123, 268002 (2019).DOI
  31. [31]E. Tjhung, C. Nardini, and M. E. Cates, Cluster phases and bubbly phase separation in active fluids: Reversal of the Ostwald process, Physical Review X 8, 031080 (2018).
  32. [33]J. Stenhammar, R. Wittkowski, D. Marenduzzo, and M. E. Cates, Activity-induced phase separation and self-assembly in mixtures of active and passive particles, Physical Review Letters 114, 018301 (2015).pubmed.ncbi.nlm.nih.gov/25615509
  33. [34]L. F. Cugliandolo, P. Digregorio, G. Gonnella, and A. Suma, Phase coexistence in two-dimensional passive and active dumbbell systems, Physical Review Letters 119, 268002 (2017).DOI
  34. [35]A. Wysocki, R. G. Winkler, and G. Gompper, Propagating interfaces in mixtures of active and passive Brownian particles, New Journal of Physics 18, 123030 (2016).DOI
  35. [36]G. Xu, T. Huang, Y. Han, and Y. Chen, Morphologies and dynamics of the interfaces between active and passive phases, Soft Matter 17, 9607 (2021).DOI
  36. [37]C. Reichhardt and C. J. O. Reichhardt, Reversibility, pattern formation, and edge transport in active chiral and passive disk mixtures, The Journal of Chemical Physics 150, 064905 (2019).DOI
  37. [39]F. J. Martínez-Ruiz, A. I. Moreno-Ventas Bravo, and F. J. Blas, Liquid-liquid interfacial properties of a symmetrical lennard-jones binary mixture, The Journal of Chemical Physics 143, 104706 (2015).
  38. [40]M. E. Cates and J. Tailleur, Motility-induced phase separation, Annu. Rev. Condens. Matter Phys. 6, 219 (2015).DOI
  39. [41]().
  40. [42]gampdy-StressSaver, https://github.com/ThomasBechSchroeder/gampdy/blob/master/gampdy/runtime_actions/stress_saver.py (), accessed: 2026-09-22.
  41. [43]F. Buff, R. Lovett, and F. Stillinger Jr, Interfacial density profile for fluids in the critical region, Physical Review Letters 15, 621 (1965).DOI
  42. [44]J. Smrek and K. Kremer, Interfacial properties of active-passive polymer mixtures, Entropy 20, 520 (2018).DOI
  43. [45]J. Elgeti and G. Gompper, Microswimmers near surfaces, The European Physical Journal Special Topics 225, 2333 (2016).DOI
  44. [46]L. Heitmeier, T. Voigtmann, and J. S. Hansen, Spectral analysis of the velocity relaxation in nanoscale channels, Physics of Fluids 37, 122019 (2025).
  45. [47]S. Knudsen, B. D. Todd, and J. S. Hansen, Hydrodynamic modes in nano-channels, Physics of Fluids 37, 012021 (2025).DOI
  46. [48]G. Yu, R. Li, F. Li, J. Zhang, X. Li, Z. Chen, J. Mecke, and Y. Gao, Pressure in active matter, Chinese Physics B 34, 094702 (2025).arxiv.org/abs/2508.09496
  47. [49]A. P. Solon, Y. Fily, A. Baskaran, M. E. Cates, Y. Kafri, M. Kardar, and J. Tailleur, Pressure is not a state function for generic active fluids, Nature Physics 11, 673 (2015).DOI
  48. [50]S. C. Takatori, W. Yan, and J. F. Brady, Swim pressure: stress generation in active matter, Physical Review Letters 113, 028103 (2014).pubmed.ncbi.nlm.nih.gov/25062240

Paper details

vlnc:W_apueyj5y
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 2

downloads · 0

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

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents