Active Field Theory Overview

• Active Field Theory is a continuum framework that models active matter via stochastic PDEs for coarse-grained fields, capturing key effects like motility-induced phase separation.
• It employs both bottom-up microscopic coarse-graining and top-down symmetry-based modifications to introduce nonvariational terms and colored noise.
• The approach links microscopic dynamics to macroscopic observables, explaining interfacial phenomena, entropy production, and critical behavior in active systems.

Searching arXiv for recent and foundational papers on active field theory to ground the article. Active field theory is the continuum statistical description of active matter in terms of coarse-grained fields whose dynamics is not restricted by equilibrium time-reversal symmetry. In the scalar setting most relevant to motility-induced phase separation (MIPS), the basic variable is a conserved density-like field, and activity appears either through nonvariational terms in the chemical potential or current, through colored stochastic forcing, or through coefficients derived from microscopic active-particle dynamics. More broadly, the subject also includes microscopic Doi–Peliti formulations for active particles, liquid-state field theories for active fluids, and Schwinger–Keldysh effective theories for active liquid crystals. Across these approaches, the unifying idea is that active systems can retain equilibrium-like structure in some sectors or scales while violating detailed balance in others, especially in interfacial physics, finite-frequency response, entropy production, and coupling to external fields (Cates, 2019).

1. Foundations and scope

Active field theory, in the sense emphasized in the lecture notes “Active Field Theories” (Cates, 2019), treats active matter as a set of stochastic partial differential equations for coarse-grained order parameters such as density $\rho$, composition $\phi$, fluid velocity $\mathbf v$, polarization $\mathbf p$, or nematic tensor $Q_{ij}$. For passive equilibrium systems, such equations are constrained by microscopic time-reversal symmetry, with consequences including a free-energy functional $F[\Psi]$, a steady-state distribution $P_0[\Psi]\propto e^{-\beta F[\Psi]}$, detailed balance, and fluctuation-dissipation relations. Active systems lift these constraints because particles are continuously driven by converting supplied energy into motion, so the allowed continuum terms are not restricted by equilibrium time-reversal symmetry (Cates, 2019).

The subject contains two complementary construction principles. A bottom-up route begins from microscopic active-particle models, coarse-grains them, and derives hydrodynamic or mesoscopic field equations. A top-down route starts from familiar passive stochastic field theories and adds the lowest-order symmetry-allowed terms that break time-reversibility. The scalar active-matter problem provides the clearest illustration: density may remain the only hydrodynamic field in isotropic systems without alignment interactions, but the current or chemical potential acquires explicitly nonequilibrium contributions (Cates, 2019).

The modern literature also broadens the meaning of active field theory. Some works derive predictive continuum equations directly from interacting active Brownian particles in three dimensions (Bickmann et al., 2020) or from active Brownian particles in external fields (Kafri et al., 23 Mar 2026). Others formulate microscopic Doi–Peliti field theories that preserve particle identity and enable exact statements about entropy production (Pruessner et al., 2022). Still others construct effective field theories for active nematics using the Schwinger–Keldysh contour and modified dynamical KMS structure (Landry, 2023). This suggests that active field theory is best understood not as a single model class but as a hierarchy of nonequilibrium continuum and field-theoretic descriptions linked by coarse-graining, symmetry, and stochastic thermodynamics.

2. Scalar conserved theories and the Model B lineage

The canonical passive reference point is Model B, the conserved $\phi^4$ theory of diffusive phase separation. Its dynamics is

$$\dot\phi=-\nabla\cdot\mathbf J,\qquad \mathbf J=-M\nabla\mu+\mathbf J^N,$$

with

$$\mu=\frac{\delta F}{\delta\phi}=a\phi+b\phi^3-\kappa\nabla^2\phi,$$

and

$\phi$0

This is the passive Cahn–Hilliard theory, and its coexistence is governed by equal chemical potential and equal pressure, equivalently the common-tangent construction (Cates, 2019).

A bottom-up derivation for active scalar matter arises from self-propelled spheres or discs without alignment interactions. After harmonic expansion of the orientational distribution and adiabatic elimination of polarization, one obtains a fluctuating density equation

$\phi$1

with effective chemical potential

$\phi$2

Here $\phi$3 is a density-dependent propulsion speed. In uniform states this gives an effective local free-energy density

$\phi$4

and the spinodal criterion

$\phi$5

This is the field-theoretic expression of MIPS: particles accumulate where they move slowly, and sufficiently rapid decrease of $\phi$6 produces a bulk instability (Cates, 2019).

Once gradient corrections are retained, the scalar active theory ceases to be purely equilibrium-like. The general nonequilibrium square-gradient form written in the lecture notes is

$\phi$7

In equilibrium the coefficients would satisfy $\phi$8, because $\phi$9 would derive from a free energy. In active matter this relation fails. The resulting phase coexistence is therefore “anomalous”: equality of bulk chemical potential remains, but equality of usual thermodynamic pressure is replaced by equality of a pseudo-pressure constructed from transformed variables (Cates, 2019). This scalar framework is the conceptual origin of Active Model B and Active Model B$\mathbf v$0.

3. Active Model B and the interfacial structure of nonequilibrium scalar theory

The paper “Scalar $\mathbf v$1 field theory for active-particle phase separation” (Wittkowski et al., 2013) introduced Active Model B as a minimal nonequilibrium scalar field theory for phase separation in active matter. Its conserved order parameter $\mathbf v$2, linearly related to particle density, obeys

$\mathbf v$3

with

$\mathbf v$4

and chemical potential

$\mathbf v$5

The passive baseline corresponds to the $\mathbf v$6 free energy

$\mathbf v$7

while the active correction

$\mathbf v$8

is nonintegrable: there is no free-energy functional $\mathbf v$9 such that $\mathbf p$0 (Wittkowski et al., 2013).

A central conceptual point is that detailed balance is restored in uniform bulk states but broken at interfaces. If $\mathbf p$1 is spatially uniform, $\mathbf p$2 and the active term vanishes, so the bulk chemical potential reduces to $\mathbf p$3. The nonequilibrium structure therefore enters only through gradients, making interfacial physics the locus of time-reversal-symmetry breaking (Wittkowski et al., 2013). This mechanism directly reflects the earlier coarse-graining logic in which activity first appears at square-gradient order (Cates, 2019).

The static consequence is the “uncommon tangent” construction. For passive Model B, coexistence between $\mathbf p$4 and $\mathbf p$5 requires

$\mathbf p$6

which yields the common tangent on $\mathbf p$7. In Active Model B, a static planar interface exists only at a shifted nonequilibrium chemical potential $\mathbf p$8, determined implicitly from the interface profile equation

$\mathbf p$9

For small $Q_{ij}$0,

$Q_{ij}$1

The bulk phases still have equal $Q_{ij}$2, but they have unequal thermodynamic pressures. The paper identifies a jump in thermodynamic pressure across a planar interface,

$Q_{ij}$3

and interprets active coexistence as tangents to $Q_{ij}$4 that are parallel but vertically displaced (Wittkowski et al., 2013).

This result is significant because interfacial phenomena are usually implicated in coarsening but irrelevant for passive phase equilibria. In Active Model B, the opposite hierarchy appears: coarsening remains close to passive Model B, with domain scale approximately obeying $Q_{ij}$5, while the static phase diagram is qualitatively altered by the interfacial active term (Wittkowski et al., 2013). The lecture notes later distilled this as the defining scalar signature of broken time-reversal symmetry in active phase separation: coexistence is no longer governed by equal pressure, even though the dynamics remains diffusive and conserved (Cates, 2019).

4. Beyond minimal scalar models: AMB$Q_{ij}$6, predictive derivations, and external fields

Active Model B is not the most general scalar theory at the same order. The lecture notes emphasize that the symmetry-allowed current in Active Model B$Q_{ij}$7 is

$Q_{ij}$8

The $Q_{ij}$9-term makes the deterministic current non-gradient, allowing $F[\Psi]$0, circulating currents, and microphase separation via reverse Ostwald dynamics (Cates, 2019). In one dimension, the two active terms combine into an effective $F[\Psi]$1, but in higher dimensions the distinction is essential.

A detailed microscopic route to AMB$F[\Psi]$2 is provided by the tutorial “How to derive a predictive active field theory: a step-by-step tutorial” (Vrugt et al., 2022). There the interaction-expansion method (IEM) starts from overdamped interacting active Brownian particles in two dimensions, derives the one-body equation, expresses two-body correlations through a pair-distribution function $F[\Psi]$3, performs a gradient expansion of the nonlocal interaction term, expands angular dependence in Fourier harmonics, and projects onto density $F[\Psi]$4 and polarization $F[\Psi]$5. Under a quasi-stationary approximation for $F[\Psi]$6, the density dynamics becomes

$F[\Psi]$7

which can be rewritten as an AMB$F[\Psi]$8-type equation

$F[\Psi]$9

The importance of this derivation is that the continuum coefficients become explicit radial integrals of the microscopic interaction potential and Fourier modes of the pair-distribution function (Vrugt et al., 2022).

A related but more general predictive route appears in the three-dimensional ABP theory of (Bickmann et al., 2020). There the orientationally resolved one-body density is expanded in tensorial moments, the interaction kernel is subjected to an untruncated gradient expansion, and the resulting local hierarchy contains infinitely many order parameters and spatial derivatives. Reduced models then recover scalar density theories. The simplest second-order model yields

$P_0[\Psi]\propto e^{-\beta F[\Psi]}$0

with an explicit density-dependent diffusivity

$P_0[\Psi]\propto e^{-\beta F[\Psi]}$1

When certain coefficients are neglected, this reduces to the Cates–Tailleur active diffusion equation with

$P_0[\Psi]\propto e^{-\beta F[\Psi]}$2

thereby giving a microscopic interpretation of the density-dependent swim speed often inserted phenomenologically (Bickmann et al., 2020).

The coupling of active scalar field theories to external potentials was addressed directly in (Kafri et al., 23 Mar 2026). Starting from active Brownian particles and expanding in persistence time, that work derived the time-dependent equation

$P_0[\Psi]\propto e^{-\beta F[\Psi]}$3

with $P_0[\Psi]\propto e^{-\beta F[\Psi]}$4. The distinctive nonequilibrium feature is the mixed tensorial coupling

$P_0[\Psi]\propto e^{-\beta F[\Psi]}$5

which cannot be captured by simply adding $P_0[\Psi]\propto e^{-\beta F[\Psi]}$6 to an effective free energy. This explains boundary accumulation and far-field density modulation around localized obstacles already at leading order beyond effective equilibrium (Kafri et al., 23 Mar 2026). A plausible implication is that external-field coupling is a structurally independent ingredient of active scalar theories, not a secondary correction.

5. Critical dynamics, noise structure, and universality

Active field theory near criticality need not be governed primarily by deterministic nonequilibrium currents. The paper “Critical active dynamics is captured by a colored-noise driven field theory” (Maggi et al., 2021) studied active Ornstein–Uhlenbeck particles near the MIPS critical point and found that fluctuation-dissipation-theorem violation is strong at short times and wavelengths but progressively restored at long scales. The proposed coarse-grained theory is a conserved scalar field $P_0[\Psi]\propto e^{-\beta F[\Psi]}$7 obeying Model-B-like dynamics with colored noise: $P_0[\Psi]\propto e^{-\beta F[\Psi]}$8 with

$P_0[\Psi]\propto e^{-\beta F[\Psi]}$9

and noise covariance

$\phi^4$0

whose Fourier-space form is

$\phi^4$1

Here the nonequilibrium character is encoded in the noise covariance rather than a nonintegrable drift (Maggi et al., 2021).

At Gaussian level this yields

$\phi^4$2

$\phi^4$3

$\phi^4$4

and an effective FDT-violation factor

$\phi^4$5

Thus $\phi^4$6 as $\phi^4$7, but $\phi^4$8 at high frequency. The theory reproduces the observed suppression of fluctuations relative to response in the ultraviolet sector (Maggi et al., 2021).

Renormalization-group analysis then shows

$\phi^4$9

so the colored-noise correlation length and time are RG irrelevant, the upper critical dimension remains $$\dot\phi=-\nabla\cdot\mathbf J,\qquad \mathbf J=-M\nabla\mu+\mathbf J^N,$$0, and the fixed point reduces to the Wilson–Fisher one. The conclusion is that the asymptotic universality class remains conserved Ising / Model B, even though finite-scale dynamics remains measurably nonequilibrium (Maggi et al., 2021). This sharpens a common issue in active field theory: equilibrium universality in the infrared does not imply equilibrium dynamics at finite wavevector and frequency.

A related mean-field effective-equilibrium approach was developed for scalar active matter under the Unified Colored Noise Approximation (Paoluzzi et al., 2019). There the steady-state AOUP distribution is approximated by

$$\dot\phi=-\nabla\cdot\mathbf J,\qquad \mathbf J=-M\nabla\mu+\mathbf J^N,$$1

with activity-induced terms $$\dot\phi=-\nabla\cdot\mathbf J,\qquad \mathbf J=-M\nabla\mu+\mathbf J^N,$$2 and $$\dot\phi=-\nabla\cdot\mathbf J,\qquad \mathbf J=-M\nabla\mu+\mathbf J^N,$$3 generating renormalized two-body and three-body couplings. Near criticality the effective action reduces to an active Landau form

$$\dot\phi=-\nabla\cdot\mathbf J,\qquad \mathbf J=-M\nabla\mu+\mathbf J^N,$$4

For isotropic interactions $$\dot\phi=-\nabla\cdot\mathbf J,\qquad \mathbf J=-M\nabla\mu+\mathbf J^N,$$5, yielding Ising-symmetric $$\dot\phi=-\nabla\cdot\mathbf J,\qquad \mathbf J=-M\nabla\mu+\mathbf J^N,$$6 criticality, whereas anisotropic interactions generate a genuine cubic term and may replace an Ising critical point by first-order behavior (Paoluzzi et al., 2019). This suggests that the universality question in scalar active matter depends strongly on whether one studies dynamic criticality, steady-state effective-equilibrium mappings, or explicitly nonequilibrium interfacial theories.

6. Microscopic field theories, entropy production, and orientational systems

Microscopic field theories preserve information lost in additive-noise density theories. The paper “Field theories of active particle systems and their entropy production” (Pruessner et al., 2022) argues that coarse-grained additive-noise density theories can reproduce densities and correlations yet give spurious entropy production because smoothing changes the effective state space. In contrast, the Doi–Peliti framework starts directly from a one-particle Fokker–Planck equation

$$\dot\phi=-\nabla\cdot\mathbf J,\qquad \mathbf J=-M\nabla\mu+\mathbf J^N,$$7

and maps it to the action

$$\dot\phi=-\nabla\cdot\mathbf J,\qquad \mathbf J=-M\nabla\mu+\mathbf J^N,$$8

This preserves particle ontology while allowing path-integral and diagrammatic methods (Pruessner et al., 2022).

The same work derives exact entropy-production expressions. At stationarity,

$$\dot\phi=-\nabla\cdot\mathbf J,\qquad \mathbf J=-M\nabla\mu+\mathbf J^N,$$9

and for pair interactions the exact stationary entropy production becomes

$$\mu=\frac{\delta F}{\delta\phi}=a\phi+b\phi^3-\kappa\nabla^2\phi,$$0

A notable result is that pair interactions require at most the equal-time three-point density, and more generally an $$\mu=\frac{\delta F}{\delta\phi}=a\phi+b\phi^3-\kappa\nabla^2\phi,$$1-body interaction requires at most the $$\mu=\frac{\delta F}{\delta\phi}=a\phi+b\phi^3-\kappa\nabla^2\phi,$$2-point density (Pruessner et al., 2022). This sharply distinguishes microscopic field theory from mesoscopic active scalar models, where entropy production is often inferred from coarse-grained path weights.

Microscopic Doi–Peliti methods have also been applied directly to free active Ornstein–Uhlenbeck particles (Bothe et al., 2021), active Brownian particles in external potentials (Zhang et al., 2022), transiently chiral active particles (Britton et al., 2 Jul 2025), and active Brownian particles with dry friction (Zhang et al., 2024). These works use phase-space field theories rather than density-only hydrodynamics. For example, (Bothe et al., 2021) derived the Gaussian Doi–Peliti action

$$\mu=\frac{\delta F}{\delta\phi}=a\phi+b\phi^3-\kappa\nabla^2\phi,$$3

diagonalized it in a Fourier–Hermite basis, and obtained the propagator

$$\mu=\frac{\delta F}{\delta\phi}=a\phi+b\phi^3-\kappa\nabla^2\phi,$$4

This preserves particle identity and provides a microscopic starting point for interactions and reactions (Bothe et al., 2021).

At the orientational hydrodynamic level, active field theory extends beyond scalars. The lecture notes discuss dry flocks, dry active nematics, and wet active liquid crystals (Cates, 2019). For a dry active nematic, the active current takes the form

$$\mu=\frac{\delta F}{\delta\phi}=a\phi+b\phi^3-\kappa\nabla^2\phi,$$5

while wet active systems have active stresses such as

$$\mu=\frac{\delta F}{\delta\phi}=a\phi+b\phi^3-\kappa\nabla^2\phi,$$6

or, for nematics, $$\mu=\frac{\delta F}{\delta\phi}=a\phi+b\phi^3-\kappa\nabla^2\phi,$$7 (Cates, 2019). The paper “Active actions: effective field theory for active nematics” (Landry, 2023) systematized such theories using the Schwinger–Keldysh formalism. There the passive orientational probe action is

$$\mu=\frac{\delta F}{\delta\phi}=a\phi+b\phi^3-\kappa\nabla^2\phi,$$8

while activity is introduced either via a slow fuel sector or via sources $$\mu=\frac{\delta F}{\delta\phi}=a\phi+b\phi^3-\kappa\nabla^2\phi,$$9 with modified dynamical KMS transformation. The resulting constitutive law is

$\phi$00

which includes both the standard isotropic active flow and an anisotropic-friction correction at the same derivative order (Landry, 2023). This suggests a broader active-field-theory principle: nonequilibrium drive can be encoded as a controlled deformation of local-equilibrium effective field theory rather than as an arbitrary abandonment of symmetry.

A plausible implication is that the field has bifurcated into two complementary traditions. One focuses on predictive microscopic or mesoscopic derivations tied to active-particle models, especially for scalar MIPS. The other imports nonequilibrium EFT machinery from quantum and classical statistical field theory to classify active hydrodynamics by symmetry and fluctuation structure. The two approaches are not contradictory; together they define the contemporary research program of active field theory.

7. Related directions and conceptual synthesis

Several adjacent developments enlarge the domain of active field theory without fitting the scalar hydrodynamic template. A nonequilibrium mean-field liquid-state theory for active Ornstein–Uhlenbeck particles derives two-point structure from a Dean-type density equation and a tracer-bath perturbation expansion, yielding

$\phi$01

with

$\phi$02

This provides a microscopic route to nonequilibrium direct correlation functions rather than a mesoscopic phase-field equation (Tociu et al., 2022). The resulting picture decomposes active structure into equilibrium local packing encoded in $\phi$03 and active temporal persistence encoded in $\phi$04.

Discrete-symmetry flocking models have likewise been treated field-theoretically. In the Doi–Peliti formulation of Active Ising Models, the coarse fields are conserved density $\phi$05 and nonconserved magnetization $\phi$06, with hydrodynamic structure

$\phi$07

$\phi$08

For one model class,

$\phi$09

giving a first-order flocking scenario at nonzero self-propulsion and a non-Model-C critical structure at zero self-propulsion because the density dynamics is spin-independent (Scandolo et al., 2023). This extends active field theory beyond scalar phase separation to coupled conserved and nonconserved fields with broken detailed balance.

The surveyed literature also includes “active field theory” in driven photonic systems, specifically non-adiabatic mean-field theory for active cavities (Burghoff, 2024). There the microscopic gain medium is integrated out into response operators

$\phi$10

and the cavity field obeys an operator mean-field equation rather than a derivative-expanded adiabatic one. Although this belongs to active photonics rather than active matter in the soft-matter sense, it reinforces a recurring theme: activity is often most naturally represented by memory kernels, operator-valued constitutive laws, or auxiliary sectors rather than by simple local drift terms.

Taken together, the literature supports a layered definition of active field theory. At the most phenomenological level it is the symmetry-based modification of passive continuum theories by leading time-reversal-symmetry-breaking terms (Cates, 2019). At the mesoscopic predictive level it is the coarse-grained continuum equation derived from active-particle dynamics, with coefficients expressed through microscopic interactions and pair correlations (Bickmann et al., 2020, Vrugt et al., 2022, Kafri et al., 23 Mar 2026). At the microscopic statistical-field-theory level it is a Doi–Peliti or related path-integral representation that preserves particle degrees of freedom and permits exact statements about observables such as entropy production (Pruessner et al., 2022, Bothe et al., 2021). At the modern EFT level it is a symmetry-organized Schwinger–Keldysh description of open, driven media and their fluctuation structure (Landry, 2023).

This suggests that the enduring conceptual lesson of active field theory is not merely that detailed balance is broken, but that the location of this breaking matters. In Active Model B it is localized in interfacial gradient terms (Wittkowski et al., 2013). Near MIPS criticality it may reside predominantly in colored noise at finite scales (Maggi et al., 2021). In entropy production it can be hidden or distorted by inappropriate coarse-graining (Pruessner et al., 2022). In external-field coupling it reappears as mixed $\phi$11-$\phi$12 terms absent from equilibrium theories (Kafri et al., 23 Mar 2026). In that sense, active field theory is the study of how nonequilibrium survives coarse-graining—sometimes in currents, sometimes in noise, sometimes in memory kernels, and sometimes only in the precise observables used to probe it.