Active Brownian Particles: Theory & Applications

Updated 12 April 2026

Active Brownian particles are self-propelled entities that combine deterministic propulsion and stochastic noise to model non-equilibrium behaviors in synthetic, bacterial, and cytoskeletal systems.
Their microscopic dynamics are governed by Langevin equations featuring both translational and rotational noise, leading to a transition from ballistic motion to effective diffusive behavior in 2D and 3D.
Collective phenomena such as motility-induced phase separation, dynamic clustering, and glassy dynamics emerge from interparticle interactions, anisotropy, and confinement effects.

Active Brownian particles (ABPs) are prototypical self-propelled agents that obey non-equilibrium stochastic dynamics combining deterministic propulsion, stochastic orientational changes, and interparticle interactions. ABPs function as minimal models for active matter in synthetic colloidal systems, bacterial suspensions, and cytoskeletal agents. Their emergent phenomena—anomalous transport, motility-induced phase separation (MIPS), collective coherence, and novel glassy and interfacial states—arise from the interplay of persistence, noise, steric effects, and geometry.

1. Microscopic Dynamics and Generalized Langevin Equations

The canonical ABP in $d$ spatial dimensions is modeled by coupled overdamped Langevin equations: $\dot{\mathbf{r}}(t) = v_0 \,\mathbf{n}(t) + \sqrt{2D}\,\boldsymbol{\xi}(t)$

$\frac{d\mathbf{n}}{dt} = \sqrt{2D_r}\,\boldsymbol{\zeta}(t) \times \mathbf{n}(t)$

where $\mathbf{r}(t)$ is the particle position, $\mathbf{n}(t)$ a body-fixed director (uniformly distributed on the unit sphere or circle), $v_0$ the propulsion speed, $D$ the translational diffusion, $D_r$ the rotational diffusion coefficient, and $\boldsymbol{\xi}(t)$ , $\boldsymbol{\zeta}(t)$ are uncorrelated Gaussian white noises.

For interacting systems, interparticle forces enter additively: $\dot{\mathbf{r}}(t) = v_0 \,\mathbf{n}(t) + \sqrt{2D}\,\boldsymbol{\xi}(t)$ 0 with $\dot{\mathbf{r}}(t) = v_0 \,\mathbf{n}(t) + \sqrt{2D}\,\boldsymbol{\xi}(t)$ 1 and $\dot{\mathbf{r}}(t) = v_0 \,\mathbf{n}(t) + \sqrt{2D}\,\boldsymbol{\xi}(t)$ 2 a typically short-range repulsive (e.g., Weeks–Chandler–Andersen) potential (Zhang et al., 2022, Moore et al., 2023).

In anisotropic ABPs, the propulsion force acts off-center and varies in both magnitude and orientation, leading to translation-rotation coupling and noise-induced drifts. For a 2D particle with mass $\dot{\mathbf{r}}(t) = v_0 \,\mathbf{n}(t) + \sqrt{2D}\,\boldsymbol{\xi}(t)$ 3, moment of inertia $\dot{\mathbf{r}}(t) = v_0 \,\mathbf{n}(t) + \sqrt{2D}\,\boldsymbol{\xi}(t)$ 4, anisotropic drags $\dot{\mathbf{r}}(t) = v_0 \,\mathbf{n}(t) + \sqrt{2D}\,\boldsymbol{\xi}(t)$ 5, $\dot{\mathbf{r}}(t) = v_0 \,\mathbf{n}(t) + \sqrt{2D}\,\boldsymbol{\xi}(t)$ 6, $\dot{\mathbf{r}}(t) = v_0 \,\mathbf{n}(t) + \sqrt{2D}\,\boldsymbol{\xi}(t)$ 7, and a point-force at offset $\dot{\mathbf{r}}(t) = v_0 \,\mathbf{n}(t) + \sqrt{2D}\,\boldsymbol{\xi}(t)$ 8, the overdamped Stratonovich SDEs read (Thiffeault et al., 2021): $\dot{\mathbf{r}}(t) = v_0 \,\mathbf{n}(t) + \sqrt{2D}\,\boldsymbol{\xi}(t)$ 9 This formulation supports explicit calculation of the full translational–rotational diffusion tensor and noise-induced (“spurious”) drifts resulting from anisotropy and fluctuating propulsion.

2. Single Particle Transport and Effective Diffusion

For an isolated ABP, the mean-squared displacement (MSD) admits an exact form in 2D (Zhang et al., 2022): $\frac{d\mathbf{n}}{dt} = \sqrt{2D_r}\,\boldsymbol{\zeta}(t) \times \mathbf{n}(t)$ 0 so the long-time effective diffusion is

$\frac{d\mathbf{n}}{dt} = \sqrt{2D_r}\,\boldsymbol{\zeta}(t) \times \mathbf{n}(t)$ 1

Ballistic motion is observed at short times ( $\frac{d\mathbf{n}}{dt} = \sqrt{2D_r}\,\boldsymbol{\zeta}(t) \times \mathbf{n}(t)$ 2), crossing over to persistent Brownian diffusion governed by $\frac{d\mathbf{n}}{dt} = \sqrt{2D_r}\,\boldsymbol{\zeta}(t) \times \mathbf{n}(t)$ 3 for $\frac{d\mathbf{n}}{dt} = \sqrt{2D_r}\,\boldsymbol{\zeta}(t) \times \mathbf{n}(t)$ 4. In 3D, $\frac{d\mathbf{n}}{dt} = \sqrt{2D_r}\,\boldsymbol{\zeta}(t) \times \mathbf{n}(t)$ 5.

When a confining external potential $\frac{d\mathbf{n}}{dt} = \sqrt{2D_r}\,\boldsymbol{\zeta}(t) \times \mathbf{n}(t)$ 6 is added, the stationary density is barometric,

$\frac{d\mathbf{n}}{dt} = \sqrt{2D_r}\,\boldsymbol{\zeta}(t) \times \mathbf{n}(t)$ 7

with higher-order non-equilibrium corrections (Zhang et al., 2022).

In polymeric or viscoelastic environments, ABP motion is generically anomalous. For an ABP crosslinked to a star polymer, the MSD at moderate Pe exhibits subdiffusion, $\frac{d\mathbf{n}}{dt} = \sqrt{2D_r}\,\boldsymbol{\zeta}(t) \times \mathbf{n}(t)$ 8 with $\frac{d\mathbf{n}}{dt} = \sqrt{2D_r}\,\boldsymbol{\zeta}(t) \times \mathbf{n}(t)$ 9, which decreases as activity increases—contrasting with intuition from passive Brownian micro-rheology. At large Pe, the MSD grows logarithmically, and the velocity autocorrelation decays as $\mathbf{r}(t)$ 0, described exactly by a fractional Langevin equation with viscoelastic memory kernel and two noises (Joo et al., 2020).

For ABP diffusion in polymer solutions, the long-time $\mathbf{r}(t)$ 1 can exhibit a nonmonotonic dependence on particle size $\mathbf{r}(t)$ 2: for large activity, $\mathbf{r}(t)$ 3 shows a maximum due to competition between persistence-induced superdiffusion and cage-induced subdiffusion, with the optimal $\mathbf{r}(t)$ 4 shifting with Pe and polymer volume fraction (Du et al., 2018).

3. Collective Phenomena: Motility-Induced Phase Separation and Pattern Formation

Motility-induced phase separation (MIPS) is a generic instability of repulsive ABPs. At high Pe and above a critical density $\mathbf{r}(t)$ 5, ABPs separate into coexisting dilute and dense phases, despite purely repulsive interactions. This is captured at the continuum level by coarse-graining the microscopic dynamics: $\mathbf{r}(t)$ 6 with $\mathbf{r}(t)$ 7 the polarization density and $\mathbf{r}(t)$ 8 a density-suppressed propulsion speed. Adiabatic elimination of $\mathbf{r}(t)$ 9 gives a modified Cahn–Hilliard equation for $\mathbf{n}(t)$ 0: $\mathbf{n}(t)$ 1 where $\mathbf{n}(t)$ 2 and $\mathbf{n}(t)$ 3 is a nonequilibrium chemical potential (Stenhammar et al., 2013). Unlike equilibrium, the gradient terms violate detailed balance, but phase coexistence and domain coarsening (with exponent $\mathbf{n}(t)$ 4) closely mirror equilibrium Cahn–Hilliard behavior. Binodals can be found via a common-tangent construction, and spinodal stability by $\mathbf{n}(t)$ 5.

The coupled density–polarization field theory provides a framework for wetting and interfacial phenomena. Near a wall, persistent ABPs accumulate, forming a steady wetting layer whose thickness and polarization structure depend on the persistence and relaxation timescales. Transitions from homogeneous (bulk-aligned) to interfacial (wetting) polarization regimes are controlled by the ratio of system sweep time $\mathbf{n}(t)$ 6 to polarization relaxation $\mathbf{n}(t)$ 7 (Perez-Bastías et al., 17 Apr 2025).

In random or porous media, obstacles spatially template MIPS: obstacles stabilize phase coexistence against re-entrant mixing at large Pe, and can pin or localize domains, giving rise to new classes of absorbed, trapped, and interface particles. The phase diagram, spinodal, and binodal boundaries shift in response to disorder, but the qualitative structure remains robust (Moore et al., 2023).

ABPs on substrates with spatially inhomogeneous $\mathbf{n}(t)$ 8 display density and flux patterns that mirror the substrate negativity: particles are depleted from high-activity zones, accumulate at activity interfaces, and reproduce the “inverse image” of $\mathbf{n}(t)$ 9 in $v_0$ 0 (Mishra et al., 2022).

4. Non-Equilibrium Glass Transition and Mode-Coupling Theory

At high density, ABPs exhibit glassy dynamics: caging, slow relaxation, and non-ergodicity. Mode-coupling theory (MCT) has been extended to the active context by explicitly including orientational degrees of freedom: $v_0$ 1 The high-dimensional equation for the transient correlation matrix $v_0$ 2 includes novel driving terms $v_0$ 3 and memory kernels that depend on the persistence length $v_0$ 4 and Pe (Liluashvili et al., 2017, Reichert et al., 2020).

Key MCT predictions for ABPs:

The glass transition surface in $v_0$ 5 space interpolates between passive and athermal limits.
High-density dynamics at large Pe or $v_0$ 6 show persistent deviations from equilibrium, such as off-diagonal non-ergodicity parameters ( $v_0$ 7), manifesting orientation–position memory.
Collective and self-intermediate scattering functions show broken time-reversal symmetry on transient timescales, confirmed in event-driven simulations.
Activity melts the glass (increases $v_0$ 8), but also gives rise to new nonergodic states with persistent orientational memory.

5. Environmental Effects, Confinement, and Interfaces

Interactions with confining boundaries or heterogeneous environments profoundly impact ABP dynamics and steady-state distributions. In harmonic or more general potentials, ABPs exhibit a smooth crossover from Boltzmann-like distributions to pronounced nonequilibrium steady states as the persistence time increases relative to the potential's relaxation time (Wang et al., 2019). In channels, ABPs accumulate at boundaries, reflecting their persistent orientation until randomization occurs (Bressloff, 2023). For partially absorbing boundaries, the mean first passage time (MFPT) for absorption is governed by an exact renewal equation incorporating both angular randomization at the wall and bulk excursions.

Colloidal membranes enclosing ABPs (active vesicles) display a rich phase diagram of regimes—adhered, tethered, fluctuating, and fissioning—depending on activity, ABP volume fraction, and membrane adhesion strength. Membrane tension, shape fluctuations, clustering, and asphericity all depend sensitively on activity–adhesion interplay (Iyer et al., 2023).

6. Multiscale Control, Prediction, and Machine-Learned Surrogates

Given their collective complexity, ABPs require multiscale modeling. Modern approaches couple particle-level Brownian dynamics (with, e.g., WCA interactions), continuum advection–diffusion PDEs for density evolution, and optimal control in spectral coordinates. Model-predictive control (MPC) schemes are implemented by formulating control updates in Fourier space, with constraints on actuator fields and mass conservation rigorously enforced (Saremi et al., 7 Sep 2025).

Deep hybrid architectures, combining 1D CNNs, LSTMs, and multi-head attention, learn to forecast future density fields with subpercent errors, serving as surrogates in fast feedback and control pipelines, and efficiently tracking phase separation, vortex, and density wave patterns.

7. Applications and Nonequilibrium Statistical Mechanics

ABPs provide a rigorous platform for inferring or deriving non-equilibrium thermodynamic analogs, such as chemical potentials, free energy densities, and fluctuation–response relations for particle-number fluctuations. Subsystem additivity, large-deviation scaling, and Maxwell-type constructions for binodals and spinodals can be formulated and matched to simulation (Chakraborti et al., 2016).

ABP-based epidemiological models generalize SIR and other compartmental models by embedding agent-based contact and mobility processes. The “microscopic” reproduction number, infection rate, and outbreak dynamics can be derived from collisions of persistent ABPs, giving parameter-free predictive models for contagion spread in active populations (Norambuena et al., 2020).

ABP pseudo-chemotaxis arises in transiently fuel-depleted (active) or food-injected environments: ABPs can outperform passive Brownian particles in food acquisition, even without explicit sensory couplings, due solely to the dynamic feedback between activity zones and propulsion (Merlitz et al., 2019).

References: