Doi-Peliti Field Theory

• Doi-Peliti field theory is a second-quantized framework for classical stochastic processes, converting master equations into coherent-state path integrals.
• It employs operator mappings, coherent states, and the Doi shift to capture discrete particle dynamics, enabling systematic diagrammatics and renormalization.
• Widely applied to reaction-diffusion, active matter, and age-structured branching, the formalism bridges microscopic stochastic rules and macroscopic nonequilibrium phenomena.

Searching arXiv for recent and foundationally relevant papers on Doi-Peliti field theory and related applications. arxiv_search(query="Doi-Peliti field theory active matter entropy production reaction-diffusion review", max_results=10) Doi-Peliti field theory is a second-quantized framework for classical stochastic processes in which a master equation or Fokker-Planck equation is recast in Fock space and then in a coherent-state functional integral. In this representation, configuration probabilities become state vectors, stochastic generators become Liouvillians or Hamiltonians written in creation and annihilation operators, and observables become normally ordered operators or functionals of conjugate fields. The formalism is effective for investigating classical stochastic processes, has wide applications, and is applicable not only to master equations but also to stochastic differential equations, for which one can derive a kind of discrete process from stochastic differential equations (Takahashi et al., 2023). Modern applications span reaction-diffusion models, active matter, nonequilibrium work relations, age-structured branching, partial exclusion, extreme-event theory, and non-Hermitian fragmentation models (Pruessner et al., 2022).

1. Operator formulation and state-space structure

The operator core of the formalism is the Doi mapping from stochastic kinetics to a Fock-space evolution problem. For discrete occupations one introduces bosonic ladder operators satisfying

$$[a_i,a_j^\dagger]=\delta_{ij}, \qquad [a_i,a_j]=[a_i^\dagger,a_j^\dagger]=0,$$

with number operator $a_i^\dagger a_i$ counting particles on site $i$ (Bothe et al., 2022). A probability distribution is encoded as a state vector, for example

$$|P(t)\rangle = \sum_{n=0}^\infty P(n,t)|n\rangle, \qquad \frac{d}{dt}|P(t)\rangle = \mathcal{L}(a^\dagger,a)|P(t)\rangle,$$

or, equivalently, in Schrödinger-like form $d|\psi(t)\rangle/dt=-\hat H|\psi(t)\rangle$ for a suitable stochastic Hamiltonian (Takahashi et al., 2023, Itakura et al., 2009).

This operatorization is closely tied to generating functions. In one standard correspondence,

$$a^\dagger \leftrightarrow x,\qquad a \leftrightarrow \frac{d}{dx},\qquad |n\rangle \leftrightarrow x^n,$$

so Doi-Peliti calculations can often be translated into generating-function manipulations (Takahashi et al., 2023). In the functional-integral language of chemical reaction networks, Doi’s quantization is expressed as

$$z_p \to a_p^\dagger,\qquad \frac{\partial}{\partial z_p}\to a_p,$$

and the generator acts on a state $|\Phi)$ through

$$\frac{\partial}{\partial \tau}|\Phi)=-\mathcal{L}(a^\dagger,a)|\Phi)$$

(Smith et al., 2018).

The formalism is not restricted to simple occupancy models. In age-structured branching, the creation and annihilation operators carry a continuous age label,

$[\psi_q,\psi_p^\dagger]=\delta(q-p),$

and aging, birth, and death are encoded by operator terms such as

$a_i^\dagger a_i$0

so that age dependence enters directly at the operator level (Greenman, 2015). For systems with finite maximum occupation $a_i^\dagger a_i$1, the bosonic Fock space is replaced by para-Fock spaces, the occupation number ranges from $a_i^\dagger a_i$2 to $a_i^\dagger a_i$3, and partial exclusion is handled by parafermi statistics and paragrassmannian coherent states (Greenman, 2018).

2. Coherent states, Doi shift, and functional integrals

The passage from operator algebra to field theory is achieved with coherent states and a resolution of the identity. In one standard representation,

$a_i^\dagger a_i$4

with path integral

$a_i^\dagger a_i$5

and action

$a_i^\dagger a_i$6

(Smith et al., 2018). In a formulation starting from a Fokker-Planck kernel $a_i^\dagger a_i$7, the noninteracting Doi-Peliti action is

$a_i^\dagger a_i$8

which is exact for non-interacting particles and valid for continuous state variables such as coordinates and angles (Pruessner et al., 2022).

A recurring step is the Doi shift, written in different conventions as $a_i^\dagger a_i$9 or $i$0. In reaction-diffusion systems this shift isolates the response field and exposes the interaction structure. For the annihilation process $i$1, for example, the shifted action becomes

$i$2

so the quartic term appears as a noise vertex and can then be decoupled by a Hubbard-Stratonovich transformation in an auxiliary-field treatment (Cooper, 2014).

The same logic extends beyond standard reaction terms. In age-dependent branching, death appears as a quadratic interaction and birth as a cubic one in the path-integral action; in the notation of the paper, the bulk weight contains a linear propagation term $i$3, a quadratic death term $i$4, and a cubic birth term proportional to $i$5 (Greenman, 2015). This interaction-order bookkeeping is one of the main reasons the formalism supports systematic diagrammatics.

3. Particle entity, density variables, and common conceptual pitfalls

A central structural property of the Doi-Peliti formalism is that it enforces what has been called “particle entity”: the discrete nature of the constituents is preserved at the operator level. A diagnostic criterion is

$i$6

and in Doi-Peliti this follows directly from the integer spectrum of the number operator, for instance through

$i$7

(Bothe et al., 2022). The same paper argues that this discreteness is “hard-coded” in Doi-Peliti through the Hilbert-space structure and ladder-operator algebra.

This feature distinguishes Doi-Peliti from response-field formulations built from Dean’s equation. Dean’s equation contains multiplicative conservative noise and a nonlinear “Dean vertex,”

$i$8

and in that framework particle entity is not fundamental at the operator level but emerges through perturbative cancellations among diagrams involving that vertex (Bothe et al., 2022). This difference is not merely formal. In active matter, approximate field theories obtained by coarse-graining or smoothing that draw on additive noise can capture densities and correlations well, but they generally ignore the microscopic particle nature of the constituents, thereby producing spurious results for the entropy production; Doi-Peliti field theories are presented as an exact alternative that captures the microscopic dynamics, including reactions and interactions with external and pair potentials (Pruessner et al., 2022).

A second recurrent misconception concerns the interpretation of the path-integral fields. In the coherent-state representation, the stochastic differential equation obtained directly from the action need not describe the physical density. For the reaction $i$9, one Langevin equation derived from the coherent-state path integral is

$$|P(t)\rangle = \sum_{n=0}^\infty P(n,t)|n\rangle, \qquad \frac{d}{dt}|P(t)\rangle = \mathcal{L}(a^\dagger,a)|P(t)\rangle,$$0

but the paper stresses that $$|P(t)\rangle = \sum_{n=0}^\infty P(n,t)|n\rangle, \qquad \frac{d}{dt}|P(t)\rangle = \mathcal{L}(a^\dagger,a)|P(t)\rangle,$$1 is an auxiliary variable and not the physical density itself. After a Cole-Hopf transformation,

$$|P(t)\rangle = \sum_{n=0}^\infty P(n,t)|n\rangle, \qquad \frac{d}{dt}|P(t)\rangle = \mathcal{L}(a^\dagger,a)|P(t)\rangle,$$2

and a system-size expansion, one instead obtains the density-fluctuation equation

$$|P(t)\rangle = \sum_{n=0}^\infty P(n,t)|n\rangle, \qquad \frac{d}{dt}|P(t)\rangle = \mathcal{L}(a^\dagger,a)|P(t)\rangle,$$3

which agrees with the van Kampen expansion and directly describes density fluctuations (Itakura et al., 2009). The distinction between auxiliary coherent-state variables and physical density fields is therefore structural, not notational.

4. Solvable Gaussian sectors and exact propagators

One of the most developed uses of Doi-Peliti field theory is the exact treatment of free or Gaussian sectors. For free active Ornstein-Uhlenbeck particles, equivalently free inertial Brownian particles without a potential, the phase-space Fokker-Planck equation

$$|P(t)\rangle = \sum_{n=0}^\infty P(n,t)|n\rangle, \qquad \frac{d}{dt}|P(t)\rangle = \mathcal{L}(a^\dagger,a)|P(t)\rangle,$$4

maps to the Doi-Peliti action

$$|P(t)\rangle = \sum_{n=0}^\infty P(n,t)|n\rangle, \qquad \frac{d}{dt}|P(t)\rangle = \mathcal{L}(a^\dagger,a)|P(t)\rangle,$$5

After Fourier transformation in space and time, a Gaussian rescaling, a shift $$|P(t)\rangle = \sum_{n=0}^\infty P(n,t)|n\rangle, \qquad \frac{d}{dt}|P(t)\rangle = \mathcal{L}(a^\dagger,a)|P(t)\rangle,$$6, and expansion in Hermite polynomials, the action becomes diagonal in mode index $$|P(t)\rangle = \sum_{n=0}^\infty P(n,t)|n\rangle, \qquad \frac{d}{dt}|P(t)\rangle = \mathcal{L}(a^\dagger,a)|P(t)\rangle,$$7, with propagator

$$|P(t)\rangle = \sum_{n=0}^\infty P(n,t)|n\rangle, \qquad \frac{d}{dt}|P(t)\rangle = \mathcal{L}(a^\dagger,a)|P(t)\rangle,$$8

(Bothe et al., 2021). The paper emphasizes that this construction respects the underlying particle nature and can be expanded to include potentials and arbitrary reactions.

Free run-and-tumble motion admits an equally explicit treatment. In $$|P(t)\rangle = \sum_{n=0}^\infty P(n,t)|n\rangle, \qquad \frac{d}{dt}|P(t)\rangle = \mathcal{L}(a^\dagger,a)|P(t)\rangle,$$9 dimensions, with speed $d|\psi(t)\rangle/dt=-\hat H|\psi(t)\rangle$0, tumble rate $d|\psi(t)\rangle/dt=-\hat H|\psi(t)\rangle$1, and translational diffusion $d|\psi(t)\rangle/dt=-\hat H|\psi(t)\rangle$2, the bare propagator takes the form

$d|\psi(t)\rangle/dt=-\hat H|\psi(t)\rangle$3

and the full propagator is obtained by Dyson resummation over tumble vertices (Zhang et al., 2021). The mean-square displacement is

$d|\psi(t)\rangle/dt=-\hat H|\psi(t)\rangle$4

with short-time ballistic behavior and long-time diffusion characterized by $d|\psi(t)\rangle/dt=-\hat H|\psi(t)\rangle$5. The same paper derives the field theory of free active Brownian particles in two dimensions and finds the same MSD after the formal identification $d|\psi(t)\rangle/dt=-\hat H|\psi(t)\rangle$6 (Zhang et al., 2021).

Run-and-tumble motion in a harmonic potential can also be solved exactly in Doi-Peliti form by introducing two species, right-moving and left-moving, rotating to density and polarity fields, and diagonalizing in a Hermite basis. In that setting the stationary entropy production is

$d|\psi(t)\rangle/dt=-\hat H|\psi(t)\rangle$7

(Garcia-Millan et al., 2020). A related extension is the field theory of transiently chiral active particles in two dimensions, where the reorientation angle $d|\psi(t)\rangle/dt=-\hat H|\psi(t)\rangle$8 diffuses between Poissonian tumbles. The exact action contains a nonlocal tumble term,

$d|\psi(t)\rangle/dt=-\hat H|\psi(t)\rangle$9

and yields exact expressions not only for the MSD but also for orientation-resolved observables such as

$$a^\dagger \leftrightarrow x,\qquad a \leftrightarrow \frac{d}{dx},\qquad |n\rangle \leftrightarrow x^n,$$0

(Britton et al., 2 Jul 2025).

5. Interactions, loop expansions, and critical phenomena

Beyond Gaussian sectors, Doi-Peliti theory provides a systematic perturbative framework for interacting nonequilibrium systems. In stochastic predator-prey dynamics, the lattice master equation for Lotka-Volterra reactions is mapped to a continuum action

$$a^\dagger \leftrightarrow x,\qquad a \leftrightarrow \frac{d}{dx},\qquad |n\rangle \leftrightarrow x^n,$$1

from which a perturbative analysis in the coexistence phase yields fluctuation corrections to the oscillation frequency and diffusion coefficient (Tauber, 2012). The associated Langevin representation has multiplicative demographic noise with explicitly derived correlators, and the paper argues that spatial degrees of freedom and stochastic noise induce instabilities toward structure formation (Tauber, 2012).

At absorbing-state transitions, the same machinery supports renormalization-group calculations. For the conserved directed percolation class with active particles $$a^\dagger \leftrightarrow x,\qquad a \leftrightarrow \frac{d}{dx},\qquad |n\rangle \leftrightarrow x^n,$$2 and passive particles $$a^\dagger \leftrightarrow x,\qquad a \leftrightarrow \frac{d}{dx},\qquad |n\rangle \leftrightarrow x^n,$$3, the action density is

$$a^\dagger \leftrightarrow x,\qquad a \leftrightarrow \frac{d}{dx},\qquad |n\rangle \leftrightarrow x^n,$$4

At Gaussian level the total density structure factor is

$$a^\dagger \leftrightarrow x,\qquad a \leftrightarrow \frac{d}{dx},\qquad |n\rangle \leftrightarrow x^n,$$5

so $$a^\dagger \leftrightarrow x,\qquad a \leftrightarrow \frac{d}{dx},\qquad |n\rangle \leftrightarrow x^n,$$6 at criticality. With one-loop RG below the upper critical dimension $$a^\dagger \leftrightarrow x,\qquad a \leftrightarrow \frac{d}{dx},\qquad |n\rangle \leftrightarrow x^n,$$7, the fixed point gives

$$a^\dagger \leftrightarrow x,\qquad a \leftrightarrow \frac{d}{dx},\qquad |n\rangle \leftrightarrow x^n,$$8

and the hyperuniformity exponent becomes

$$a^\dagger \leftrightarrow x,\qquad a \leftrightarrow \frac{d}{dx},\qquad |n\rangle \leftrightarrow x^n,$$9

while

$$z_p \to a_p^\dagger,\qquad \frac{\partial}{\partial z_p}\to a_p,$$0

The paper states that the result for $$z_p \to a_p^\dagger,\qquad \frac{\partial}{\partial z_p}\to a_p,$$1 disproves a previously conjectured scaling relation and attributes hyperuniformity to anticorrelation of strongly fluctuating active and passive densities (Ma et al., 2023).

A complementary nonperturbative route is the auxiliary-field loop expansion. For $$z_p \to a_p^\dagger,\qquad \frac{\partial}{\partial z_p}\to a_p,$$2, the quartic interaction is decoupled by a Hubbard-Stratonovich transformation introducing composite fields $$z_p \to a_p^\dagger,\qquad \frac{\partial}{\partial z_p}\to a_p,$$3, and the effective action becomes

$$z_p \to a_p^\dagger,\qquad \frac{\partial}{\partial z_p}\to a_p,$$4

The leading-order auxiliary-field approximation then yields an effective potential, gap equations, and a beta function

$$z_p \to a_p^\dagger,\qquad \frac{\partial}{\partial z_p}\to a_p,$$5

for arbitrary spatial dimension $$z_p \to a_p^\dagger,\qquad \frac{\partial}{\partial z_p}\to a_p,$$6 (Cooper, 2014). Taken together, these examples show that Doi-Peliti theory is suitable for expansions and renormalization/group analysis, while retaining a direct link to microscopic stochastic rules (Bothe et al., 2021).

6. Nonequilibrium thermodynamics, dualities, and major extensions

The formalism has become a tool for nonequilibrium thermodynamics because it encodes short-time transition structure at the microscopic level. For active particle systems, the entropy production is written as

$$z_p \to a_p^\dagger,\qquad \frac{\partial}{\partial z_p}\to a_p,$$7

with a stationary decomposition

$$z_p \to a_p^\dagger,\qquad \frac{\partial}{\partial z_p}\to a_p,$$8

For pair interactions, the exact entropy production depends only on the one-, two-, and three-point equal-time densities, and more generally an $$z_p \to a_p^\dagger,\qquad \frac{\partial}{\partial z_p}\to a_p,$$9-point interaction requires at most the $|\Phi)$0-point density (Pruessner et al., 2022). In free run-and-tumble motion the steady-state entropy production rate is $|\Phi)$1 in any dimension (Zhang et al., 2021), while in the harmonically trapped one-dimensional model it is $|\Phi)$2 (Garcia-Millan et al., 2020).

A distinct line of work uses Doi-Peliti theory to formulate exact nonequilibrium work relations. For interacting particles hopping on a lattice under a time-dependent potential, the coherent-state field theory has a time-reversal symmetry that takes the form of a gauge-like transformation. Under this transformation the action is invariant up to a generated work term, and one obtains

$|\Phi)$3

together with the Crooks relation and a far-from-equilibrium generalization of the fluctuation-dissipation relation (Baish et al., 2024). Closely related work on adjoint fluctuation theorems shows that dualization in the Doi-Peliti functional integral is realized as a change of integration variables and exchanges the roles of retarded and advanced Green’s functions, thereby making explicit a duality between dynamics and inference (Smith et al., 2018).

The range of current extensions is broad. Partial exclusion requires paragrassmannian coherent states, and because non-commutativity is generic, a Magnus expansion may be required so that actions containing a finite number of terms are not always feasible (Greenman, 2018). For monomolecular chemical master equations, the Doi-Peliti path integral reproduces Jahnke and Huisinga’s exact time-dependent solution and extends beyond it to autocatalytic reactions and arbitrary zero- and first-order systems (Vastola, 2019). For extreme events of non-Markovian processes, a Doi-Peliti field theory combined with the Martin-Siggia-Rose formalism maps survival and first-passage problems to a two-species reaction-diffusion theory and yields perturbative corrections in self-correlated noise (Walter et al., 2021). In fragmentation and comminution, homogeneous population-balance kernels lead to an exact Markov jump generator in log-size, a non-self-adjoint Lindblad embedding, and second-quantized hopping and branching actions whose one-body sector reproduces the deterministic population balance equation while higher correlators encode finite-population fluctuations (Segura, 10 Jan 2026). A different extension replaces the conventional monomial basis by a redundant non-orthogonal basis,

$|\Phi)$4

yielding long-range hopping in the effective discrete process and improved finite-state approximations for stochastic differential equations such as the noisy van der Pol system (Takahashi et al., 2023).

These developments collectively indicate that Doi-Peliti field theory is less a single model than a general operator and path-integral architecture for stochastic many-body dynamics: it encodes discreteness exactly, admits systematic diagrammatics and RG, and can be adapted to continuous internal variables, exclusion constraints, age structure, memory, work relations, and non-Hermitian generators without leaving the basic Fock-space framework.