Mean-field theory of collective motion due to velocity alignment

We introduce a system of self-propelled agents (active Brownian particles) with velocity alignment in two spatial dimensions and derive a mean-field theory from the microscopic dynamics via a nonlinear Fokker-Planck equation and a moment expansion of the probability distribution function. We analyze the stationary solutions corresponding to macroscopic collective motion with finite center of mass velocity (ordered state) and the disordered solution with no collective motion in the spatially homogeneous system. In particular, we discuss the impact of two different propulsion functions governing the individual dynamics. Our results predict a strong impact of the individual dynamics on the mean field onset of collective motion (continuous vs discontinuous). In addition to the macroscopic density and velocity field we consider explicitly the dynamics of an effective temperature of the agent system, representing a measure of velocity fluctuations around the mean velocity. We show that the temperature decreases strongly with increasing level of collective motion despite constant fluctuations on individual level, which suggests that extreme caution should be taken in deducing individual behavior, such as, state-dependent individual fluctuations from mean-field measurements [Yates {\em et al.}, PNAS, 106 (14), 2009].

💡 Research Summary

The paper develops a comprehensive mean‑field description of collective motion in a system of self‑propelled agents that interact solely through velocity alignment in two dimensions. Starting from a microscopic Langevin formulation, each particle of unit mass obeys

 dr_i/dt = v_i,
 dv_i/dt = –γ(v_i) v_i + μ (u_ε,i – v_i) + √(2D) ξ_i(t),

where γ(v) v is a velocity‑dependent friction (or propulsion) term, μ = 1/τ_a is the alignment strength, u_ε,i is the average velocity of all neighbours within a metric radius ε, and ξ_i(t) is a Gaussian white noise of intensity D. Two specific friction laws are examined: (i) the Rayleigh‑Helmholtz form –γ(v)v = (α – β v²) v, which yields a preferred speed v₀ = √(α/β) and a non‑linear damping at high speeds, and (ii) a linear Schienbein‑Gruler form, which is essentially a constant‑plus‑linear term and produces almost Gaussian velocity statistics.

From the Langevin equations the authors derive a nonlinear Fokker‑Planck (FP) equation for the one‑particle probability density P(r,v,t). The alignment term makes the FP equation nonlinear because the local mean velocity u_ε(r,t) depends on moments of P itself. By defining velocity moments h vⁿ_k = (1/ρ)∫ v_kⁿ P dv (with ρ the local density) they obtain an infinite hierarchy of coupled moment equations. Closure is achieved by decomposing the velocity into a mean field u(r,t) and fluctuations δv = v – u, assuming that x‑ and y‑components of δv are statistically independent and have zero odd moments. This leads to a diagonal covariance matrix with entries T_x and T_y, which the authors interpret as directional “temperatures” measuring velocity fluctuations around the mean flow.

Using the moment expansion, the authors write explicit expressions for the first few moments (mean velocity, second‑order moments, etc.) in terms of u and the temperatures T_k, plus a higher‑order correction θ_k that accounts for non‑Gaussianity of the distribution. The resulting macroscopic equations are:

1. Continuity: ∂_t ρ + ∇·(ρ u) = 0.
2. Momentum: ∂_t(ρ u) + ∇·(ρ u⊗u) = ρ