betaFeedback

WorksPhysics and Astronomy

The active chiral random walk

Abstract

We introduce a minimal and intuitive lattice model for an active chiral particle which exhibits persistence, reversal, and chirality in its dynamics. It is fully defined by a small set of microscopic parameters that provide transparent control over the resulting dynamics. By coarse-graining, we extract continuum equations and closed-form expressions for the velocity autocorrelation, off-diagonal directional correlation, mean-squared displacement, and long-time diffusion and odd diffusion coefficients. The complex parameter λ=qp+iqcλ=q_p+\mathrm{i}q_c thereby encodes the microscopic information on persistence (qpq_p) and chirality (qcq_c) of the walker. In a weak-turning limit, microscopically realized by infrequent reorientation events, the continuum description can be mapped onto well-established models of active matter. Outside this limit the discrete model exhibits finite-step dynamics, such as anti-persistent motion, that are not captured by a smooth angular-diffusion description. The model therefore provides a compact, exactly solvable, and readily tunable framework in which a small number of transparent microscopic parameters control chiral active transport.

Introduction

Random walks are among the simplest models of stochastic transport. The term was introduced by Karl Pearson in 1905 in the context of biological dispersal [1, 2], building on Ronald Ross’s 1905 studies regarding the spatial spreading of mosquitoes [3, 4]; Lord Rayleigh then pointed out that his earlier work already contained the many-step solution [5]. The discrete picture was connected to continuum diffusion by Einstein and Smoluchowski [6, 7, 8, 9]. Langevin’s complementary approach, in which unresolved microscopic degrees of freedom are represented by fluctuating forces, became another central framework for non-equilibrium statistical physics. Since then, lattice-based stochastic models have become a versatile framework for diffusion, transport, and collective phenomena, with applications ranging from random walks and percolation [10, 11, 12, 13, 14, 15, 16, 17] to reaction–diffusion processes [18, 19, 20, 21], financial markets [22], and biological search strategies [23, 24, 25, 26, 27].

The study of systems far from thermal equilibrium has substantially expanded the scope of non-equilibrium statistical physics [28, 29]. One such class of systems is active matter, in which an agent converts energy from an internal source or its local environment into motion through non-conservative forces coupled to its propulsion and orientation [30, 31]. Examples of such systems include Brownian motors [32, 33], motile cells [34, 35], and synthetic self-propelled particles [36, 37, 38]. Unlike passive Brownian particles, active agents typically maintain their direction over a finite persistence time because their propulsion orientation relaxes only through rotational noise or discrete reorientation events such as randomization, reversal, or turning. This directional persistence is a key characteristic of active matter [39]. At larger biological scales, trajectories can additionally exhibit long-range temporal correlations and anomalous transport, as observed in systems extending from molecular motion to migrating birds [40]. In many systems this directional memory is also chiral, meaning the particle has a preferred handedness, producing circular trajectories, biased turns, and edge-following motion [41, 42]. Such chiral active dynamics appear in a broad range of physical and biological systems, including swimming microorganisms [43, 44, 45, 46, 47], asymmetric self-propelled particles [48], chiral fluids [49, 50], and driven active particles [48, 51]. Understanding how persistence and chirality jointly shape transport is therefore a fundamental problem in active matter.

Continuum models such as active Brownian particles (ABPs) [30, 52] and active Ornstein–Uhlenbeck particles [53] provide descriptions of persistent active motion. In chiral variants of these models, the orientation or active velocity both decorrelates and rotates with a specific handedness, producing exponentially damped oscillatory velocity correlations [42, 54, 55]. However, continuum descriptions can obscure the discrete structure of microscopic reorientation events. In particular, the distinction between persistence, reversal, left turns, right turns, and random reorientation is often hidden inside coarse-grained rotational diffusion and drift parameters. Simple lattice models are therefore valuable because they retain microscopic interpretability, are efficient to simulate at large scales, permit exact calculations, and provide a natural setting for studying transport in structured environments, obstacles, networks, and percolating media.

Persistent motion has a long history in statistical physics, both in the continuum [56, 57] and on lattices [58, 59, 60, 61, 62, 63, 64, 65, 66, 67]. A way to introduce persistence into lattice models is via equipping the walker with an orientational diffusion, which has been studied in one and two dimensions [68]. A conceptually distinct way to encode persistence is via an internal state that stores the previous direction of motion. This internal degree of freedom makes the walk Markovian in an enlarged position–orientation space, while producing non-trivial correlations in the projected spatial trajectory. By comparison, lattice models with an explicitly chiral internal degree of freedom remain relatively sparse [69]. Recent work introduced the chiral random walk (CRW) on a lattice [70] using the language of a “coin” operator, inspired by discrete-time quantum walks [71]. In that formulation, the coin acts on the orientational degree of freedom and determines the stochastic rule by which the walker re-orients itself. This provides an intuitive and compact way to encode chirality and memory at the microscopic level.

Here we introduce a minimal active chiral random walk in which persistence, reversal, clockwise and counter-clockwise rotation, and random reorientation are controlled by explicit microscopic probabilities. These microscopic rules reduce, at the level of the first angular harmonic, to the complex parameter λ=qp+iqc\lambda= q_p + \mathrm{i}q_c. The real part qpq_p controls persistence against reversal, while the imaginary part qcq_c controls chirality. This provides a direct connection between microscopic reorientation rules and macroscopic transport.

Relative to the chiral random walk of [70], the present construction introduces persistence as a separately tunable property. We use the resulting active chiral random walk (aCRW) to derive the continuum density and polarization equations directly from the stochastic dynamics. We obtain exact directional correlations, the finite-time mean-squared displacement (MSD), and the long-time even and odd transport coefficients. We then establish the asymptotic mapping to the chiral active Brownian particle model and quantify the finite-step corrections away from this limit. The resulting comparison identifies which transport properties are preserved under coarse-graining and which depend on the discrete dynamics. In particular, the leading long-time transport remains exact even when the transient polar dynamics are only approximately represented. Strong chirality and anti-persistent motion define intrinsically discrete regimes beyond the standard smooth angular-diffusion description. Complementary to Larralde’s analysis of chiral persistent transport [66], our focus here is the active-matter interpretation and continuum relations made explicit by the intuitive coefficients qpq_p and qcq_c. We use “active” here in the kinematic sense of a particle that moves at a fixed self-propulsion speed while its orientation evolves stochastically; the model does not resolve the energetic mechanism generating propulsion. Further, our model also provides a simple, exactly tractable test of what a continuum approximation preserves. Its correlations and long-time transport follow explicitly from the same microscopic probabilities, allowing us to identify the temporal corrections omitted by the approximation and show that these need not change the leading bulk transport.

This paper is organized as follows. § 2 defines the coin-based lattice model and its stochastic interpretation. § 3 derives the continuum density and polarization equations, including the even and odd currents. § 4 obtains the exact directional correlations, finite-time MSD, and Green–Kubo transport coefficients. § 5 compares the aCRW with the continuum chiral active Brownian particle (cABP) model, identifies the weak-turning limit, and distinguishes finite-step correlation matching from transport matching. The appendices collect the directional-correlation derivation (Appendix § 7), the polar reduction and hydrodynamic expansion (Appendix § 8), the boundary-scattering rules (Appendix § 9), the derivation of the cABP field equation (Appendix § 10), and the finite-step rate and transport corrections (Appendix § 11).

Model

Active motile agents often combine stochasticity, persistence, and chirality in their dynamics. Introducing these properties into a lattice model in an intuitive and controllable manner enables the study of a range of static and dynamical properties, including the evolution and steady states of the displacement probability distribution [72, 73, 74], percolation dynamics [75, 76, 77], first-passage properties, and search efficiency [78, 79, 80], as recently demonstrated for the non-persistent CRW in Ref. [81]. We introduce a lattice model that unifies these ingredients within a single coin-based framework while allowing each of these properties to be controlled independently.

Our starting point is the CRW introduced in Ref. [70], whose formalism we adopt and extend. There, the walker’s displacement is coupled to an internal degree of freedom that stores information about its orientation, and the orientation is propagated by a stochastic operator, the coin CC. The structure of the coin dictates how consecutive steps are correlated and is therefore the object from which the internal dynamical properties of a walker arise in the model.

Microscopic state and update

The walker occupies a site (x,y)=r∈ℓZ2(x,y)=\mathbf{r}\in\ell\mathbb{Z}^{2} on a two-dimensional square lattice with lattice constant ℓ\ell, and carries an internal orientation

d∈D={→,↑,←,↓},(1)d\in D=\{\rightarrow,\uparrow,\leftarrow,\downarrow\}, \tag*{(1)}

which records the direction of its most recent step1. Each internal state is associated with a lattice displacement

e→=(10),e↑=(01),e←=(−10),e↓=(0−1),(2)\mathbf{e}_{\rightarrow}=\begin{pmatrix}1\\0\end{pmatrix},\qquad \mathbf{e}_{\uparrow}=\begin{pmatrix}0\\1\end{pmatrix},\qquad \mathbf{e}_{\leftarrow}=\begin{pmatrix}-1\\0\end{pmatrix},\qquad \mathbf{e}_{\downarrow}=\begin{pmatrix}0\\-1\end{pmatrix}, \tag*{(2)}

and we collect these as the columns of the lattice-geometry matrix

E=(10−10010−1),(3)E=\begin{pmatrix} 1 & 0 & -1 & 0\\ 0 & 1 & 0 & -1 \end{pmatrix}, \tag*{(3)}

With this notation, ed=E∣d)\mathbf{e}_{d}=E\lvert d), where ∣→)=(1,0,0,0)T\lvert\rightarrow)=(1,0,0,0)^{\mathsf{T}}, ∣↑)=(0,1,0,0)T\lvert\uparrow)=(0,1,0,0)^{\mathsf{T}}, ∣←)=(0,0,1,0)T\lvert\leftarrow)=(0,0,1,0)^{\mathsf{T}} and ∣↓)=(0,0,0,1)T\lvert\downarrow)=(0,0,0,1)^{\mathsf{T}} are the basis vectors of the internal space. The joint distribution over configurations is the probability vector

∣ρ(t))=∑r,dPd(r,t)∣r)⊗∣d),(4)\lvert\rho(t))=\sum_{\mathbf{r},d}P_{d}(\mathbf{r},t)\lvert\mathbf{r})\otimes\lvert d), \tag*{(4)}

where Pd(r,t)=(r,d∣ρ(t))P_{d}(\mathbf{r},t)=(\mathbf{r},d\lvert\rho(t)) is the joint position–orientation probability for the walker to be at r\mathbf{r} with internal state dd, and ∣r,d)=∣r)⊗∣d)\lvert\mathbf{r},d)=\lvert\mathbf{r})\otimes\lvert d). Marginalizing over the internal degrees of freedom returns the spatial probability density function (PDF), ρ(r,t)=∑d∈DPd(r,t)\rho(\mathbf{r},t)=\sum_{d\in D}P_{d}(\mathbf{r},t).

In each time step, the walker performs two sequential operations: the coin CC updates the internal state, and the conditional-step operator SS then displaces the walker along the updated direction,

∣ρ(t+Δt))=S(Ir⊗C)∣ρ(t)),(5)\lvert\rho(t+\Delta t))=S(I_{\mathbf{r}}\otimes C)\lvert\rho(t)), \tag*{(5)}

with IrI_{\mathbf{r}} the identity on the spatial degrees of freedom. The step operator is the conditional shift S=∑r,d∣r+ℓed,d)(r,d∣S=\sum_{\mathbf{r},d}\lvert\mathbf{r}+\ell\mathbf{e}_{d},d)(\mathbf{r},d\lvert, so that, for example, S∣(x,y),→)=∣(x+ℓ,y),→)S\lvert(x,y),\rightarrow)=\lvert(x+\ell,y),\rightarrow). In other words, the walker orients according to the coin and moves in its new orientation. The corresponding component master equation is introduced at the start of § 3; explicit operator matrices and the simulation rule are given below.

Coin and angular modes

The coin is built from five elementary operations: keeping the orientation, reversing it, rotating clockwise (right-handed chirality), rotating counter-clockwise (left-handed chirality), and mixing (randomizing) the orientation. Each operator is assigned a fixed probability denoted by the corresponding prefactor, so that

C=qII+qRR+qLL+qBB+qMM,(6)C=q_I I+q_R R+q_L L+q_B B+q_M M, \tag*{(6)}

where II, RR, LL, BB, MM are the identity, right-chiral, left-chiral, reversal and mixing operators, given explicitly below. Since the prefactors are probabilities, they are bounded in [0,1][0,1] and obey the normalization

qI+qR+qL+qB+qM=1.(7)q_I+q_R+q_L+q_B+q_M=1. \tag*{(7)}

We note that the four non-randomizing operators are mutually orthogonal and obey a symmetric cyclic relation. Within the cyclic ordering of DD, we have R∣dm)=∣dm−1)R|d_m)=|d_{m-1}), where m=1,2,3,4m=1,2,3,4 indexes DD cyclically (modulo four), and B=R2B=R^2, L=R3=R−1L=R^3=R^{-1} and R4=IR^4=I. These four matrices therefore form the regular representation of the cyclic group C4C_4, and their real span is the algebra of 4×44\times4 circulant matrices. The randomization operator is not a fifth independent element, but rather a combination of the four other operators, M=(I+R+B+L)/4M=(I+R+B+L)/4, so that Eq. (6) is an over-parametrization and can be reduced. Yet we keep this form, as having MM explicit helps in making the model easier to control and interpret. For analytical simplification, however, it is useful to present the coin in its minimal form by absorbing MM as C=∑k=14q^kRkC=\sum_{k=1}^{4}\hat{q}_kR^k, where q^i=qi+qM/4\hat{q}_i=q_i+q_M/4 for i∈{I,R,B,L}i\in\{I,R,B,L\} are the reduced weights that inherit the normalization from Eq. (7). We note that by construction, they are exactly the matrix elements of the coin, i.e., q^I=(d∣C∣d)\hat{q}_I=(d|C|d), q^B=(Bd∣C∣d)\hat{q}_B=(Bd|C|d), q^R=(Rd∣C∣d)\hat{q}_R=(Rd|C|d), and q^L=(Ld∣C∣d)\hat{q}_L=(Ld|C|d).

Because the coin matrix is circulant, it is diagonalized by the angular Fourier modes of C4C_4, which are

∣v0)=(1111),∣v1)=(1−i−1i),∣v2)=(1−11−1),∣v3)=(1i−1−i),(8)|v_0)=\begin{pmatrix}1\\1\\1\\1\end{pmatrix},\qquad |v_1)=\begin{pmatrix}1\\-i\\-1\\i\end{pmatrix},\qquad |v_2)=\begin{pmatrix}1\\-1\\1\\-1\end{pmatrix},\qquad |v_3)=\begin{pmatrix}1\\i\\-1\\-i\end{pmatrix}, \tag*{(8)}

with eigenvalues that follow from C∣vn)=λn∣vn)C|v_n)=\lambda_n|v_n). Explicitly,

λ0=1,(9)\lambda_0=1, \tag*{(9)}

density mode,

λ1=(q^I−q^B)+i(q^L−q^R)=qp+iqc,(10)\lambda_1=(\hat{q}_I-\hat{q}_B)+i(\hat{q}_L-\hat{q}_R)=q_p+iq_c, \tag*{(10)}

polar mode,

λ2=(q^I+q^B)−(q^R+q^L)≡qn,(11)\lambda_2=(\hat{q}_I+\hat{q}_B)-(\hat{q}_R+\hat{q}_L)\equiv q_n, \tag*{(11)}

nematic mode,

λ3=(q^I−q^B)−i(q^L−q^R)=λ1‾,(12)\lambda_3=(\hat{q}_I-\hat{q}_B)-i(\hat{q}_L-\hat{q}_R)=\overline{\lambda_1}, \tag*{(12)}

conjugate polar mode.

Each eigenvalue has a direct microscopic interpretation. The unit density eigenvalue reflects that the coin redistributes probability without creating or destroying it, making the associated PDF ρ(r,t)\rho(\mathbf{r},t) conserved, as shown in § 3. The polar eigenvalue splits into the persistence, qp=qI−qBq_p=q_I-q_B, i.e., the excess of keeping the walker orientation (qp>0q_p>0) over flipping it (qp<0q_p<0), and the chirality, qc=qL−qRq_c=q_L-q_R, i.e., the excess of left (qc>0q_c>0) over right (qc<0q_c<0) turns (counter-clockwise rotation is defined positive). The nematic eigenvalue qnq_n measures the preference for collinear (qn>0q_n>0) over orthogonal (qn<0q_n<0) steps and is the third free parameter of the coin. The three walk parameters qpq_p, qcq_c, and qnq_n now lie in [−1,1][-1,1] and inherit shared constraints from the non-negativity of the transition probabilities Eq. (7), ∣qp∣≤(1+qn)/2|q_p|\leq(1+q_n)/2, and ∣qc∣≤(1−qn)/2|q_c|\leq(1-q_n)/2, which projected is ∣qp∣+∣qc∣≤1|q_p|+|q_c|\leq1. Note that qpq_p, qcq_c, and qnq_n can equally be defined with respect to the original weights qiq_i instead of the rescaled q^i\hat{q}_i, as qMq_M cancels in the eigenvalue decomposition.

Microscopic operators and stochastic update

The orientation basis and its ordering are defined above. Here we give the explicit operators needed to implement the walk. The step operator acts as

S∣(x,y),→)=∣(x+ℓ,y),→),S∣(x,y),←)=∣(x−ℓ,y),←),S∣(x,y),↑)=∣(x,y+ℓ),↑),S∣(x,y),↓)=∣(x,y−ℓ),↓).(13)\begin{aligned} S|(x,y),\rightarrow)&=|(x+\ell,y),\rightarrow),\qquad&S|(x,y),\leftarrow)&=|(x-\ell,y),\leftarrow),\\ S|(x,y),\uparrow)&=|(x,y+\ell),\uparrow),\qquad&S|(x,y),\downarrow)&=|(x,y-\ell),\downarrow). \tag*{(13)} \end{aligned}

The five elementary 4×44 \times4 operators used in the propagation of the direction are

I=(1000010000100001),R=(0100001000011000),L=(0001100001000010),I = \begin{pmatrix} 1 & 0 & 0 & 0 \\ 0 & 1 & 0 & 0 \\ 0 & 0 & 1 & 0 \\ 0 & 0 & 0 & 1 \end{pmatrix},\quad R = \begin{pmatrix} 0 & 1 & 0 & 0 \\ 0 & 0 & 1 & 0 \\ 0 & 0 & 0 & 1 \\ 1 & 0 & 0 & 0 \end{pmatrix},\quad L = \begin{pmatrix} 0 & 0 & 0 & 1 \\ 1 & 0 & 0 & 0 \\ 0 & 1 & 0 & 0 \\ 0 & 0 & 1 & 0 \end{pmatrix},
B=(0010000110000100),M=14(1111111111111111).(14)B = \begin{pmatrix} 0 & 0 & 1 & 0 \\ 0 & 0 & 0 & 1 \\ 1 & 0 & 0 & 0 \\ 0 & 1 & 0 & 0 \end{pmatrix},\quad M = \frac{1}{4}\begin{pmatrix} 1 & 1 & 1 & 1 \\ 1 & 1 & 1 & 1 \\ 1 & 1 & 1 & 1 \\ 1 & 1 & 1 & 1 \end{pmatrix}. \tag*{(14)}

Here II, RR, LL, BB, MM are, respectively, the identity, right-chiral, left-chiral, reversal and mixing (randomization) operators. Their weighted combination is the coin defined in Eq. (6).

The coin parameters transfer immediately to the stochastic update rule for the orientation. Given a current direction ∣dt⟩\lvert d_t \rangle, the next direction is sampled according to

∣dt+1⟩={∣dt⟩,prob. qI,B∣dt⟩,prob. qB,R∣dt⟩,prob. qR,L∣dt⟩,prob. qL,∣dunif⟩,dunif∼Uniform⁡(D),prob. qM.(15)\lvert d_{t+1} \rangle= \begin{cases} \lvert d_t \rangle, & \text{prob. } q_I, \\ B\lvert d_t \rangle, & \text{prob. } q_B, \\ R\lvert d_t \rangle, & \text{prob. } q_R, \\ L\lvert d_t \rangle, & \text{prob. } q_L, \\ \lvert d_{\mathrm{unif}} \rangle,\quad d_{\mathrm{unif}} \sim\operatorname{Uniform}(D), & \text{prob. } q_M. \end{cases} \tag*{(15)}

Thus a mixing event samples one of the four orientation states uniformly; M∣dt⟩M\lvert d_t \rangle is the corresponding probability vector and not itself a realized orientation. The associated transition probabilities are the matrix elements of the stochastic coin operator, (d′∣C∣d)=P(d′∣d)(d' \lvert C \rvert d) = P(d' \lvert d).

Representative dynamics

Being a discrete random walk, the aCRW is considerably cheaper to simulate than continuum models of active matter, which is essential when dynamical properties have to be sampled. The coin in Eq. (6) also makes the aCRW directly tunable, since every microscopic parameter has an immediate interpretation. With these two notions in mind, we begin the analysis by observing four different types of walkers: random (qM=1q_M = 1), anti-persistent (qB≠0q_B \ne0), chiral (qRq_R or qL≠0q_L \ne0), and active chiral (qI≠0q_I \ne0 and qRq_R or qL≠0q_L \ne0). Figure 1 shows the representative trajectories where different characteristics are directly visible; the compact loops of the chiral walker, the extended runs of the persistent-chiral walker, and the frequent backtracking of the anti-persistent walker.

Figure 1. Representative trajectories for four walker types. The starting position is denoted by a circle and the endpoint by a square, while the colour gradient represents the passing of time (from light to dark). Each trajectory consists of 300 steps for a single walker. Microscopic parameters: random (dark grey), qM=1q_M = 1; anti-persistent (muted orange), qB=0.5q_B = 0.5, qL=0.2q_L = 0.2, and qM=0.3q_M = 0.3; chiral (purple), qL=0.5q_L = 0.5 and qM=0.5q_M = 0.5; active chiral (green), qI=0.5q_I = 0.5, qL=0.2q_L = 0.2, and qM=0.3q_M = 0.3. The trajectories display the characteristic geometry of each turning rule.

This phenomenology raises the question of how these discrete microscopic parameters relate to and control continuum properties. We address this independently by both hydrodynamic coarse-graining in Sec. 3 and Green–Kubo calculation in § 4.3, before comparing the result with established continuum models in § 5.

From lattice to continuum: density and polarization fields

The detailed description of the model allows us to coarse-grain it to the continuum limit, resulting in a set of equations that describe the non-interacting active chiral gas. To do so, we begin by projecting Eq. (5) onto (r,d∣(\mathbf{r},d|, which yields the orientation-resolved master equation in component form,

Pd(r,t+Δt)=∑d′Cdd′Pd′(r−ℓed,t),(16)P_d(\mathbf{r},t+\Delta t)=\sum_{d'}C_{dd'}P_{d'}(\mathbf{r}-\ell\mathbf{e}_d,t), \tag*{(16)}

where Pd(r,t)P_d(\mathbf{r},t) is the joint position–orientation probability for the walker to be at r=(x,y)\mathbf{r}=(x,y) with orientation d∈Dd\in D, and Cdd′=(d∣C∣d′)C_{dd'}=(d|C|d') is the conditional probability of updating from orientation d′d' to dd. Note that the displacement is set by the outgoing state dd and not by d′d', since the coin acts before the step. Summing over all possible orientations gives the spatial PDF,

ρ(r,t)=∑dPd(r,t).(17)\rho(\mathbf{r},t)=\sum_d P_d(\mathbf{r},t). \tag*{(17)}

The spatial probability is normalized as ∑rρ(r,t)=1\sum_{\mathbf{r}}\rho(\mathbf{r},t)=1. Because there is no rest state, all probability at a site is redistributed among the four outgoing orientations at every update. Writing the change in density over one time step as ρ(r,t+Δt)−ρ(r,t)\rho(\mathbf{r},t+\Delta t)-\rho(\mathbf{r},t) and replacing this finite difference by a time derivative gives the time evolution for the PDF, where ν=1/Δt\nu=1/\Delta t is the update frequency,

∂ρ(r,t)∂t≈ν[∑d,d′Cdd′Pd′(r−ℓed,t)−ρ(r,t)].(18)\frac{\partial\rho(\mathbf{r},t)}{\partial t}\approx\nu\left[\sum_{d,d'}C_{dd'}P_{d'}(\mathbf{r}-\ell\mathbf{e}_d,t)-\rho(\mathbf{r},t)\right]. \tag*{(18)}

From this dynamical equation, the spatial and orientational moments of the process can be obtained directly.

Slow fields and polar reduction

The resulting lattice-cell PDF in Eq. (17) is the zeroth orientational moment of PdP_d. In addition, one can retain the leading orientational information, the first orientational moment, which is given by the local polarization

m(r,t)=∑d∈DedPd(r,t)=(P→−P←P↑−P↓)(r,t).(19)\mathbf{m}(\mathbf{r},t)=\sum_{d\in D}\mathbf{e}_dP_d(\mathbf{r},t)=\begin{pmatrix}P_{\rightarrow}-P_{\leftarrow}\\ P_{\uparrow}-P_{\downarrow}\end{pmatrix}(\mathbf{r},t). \tag*{(19)}

Equivalently, in terms of the lattice geometry matrix E\mathbf{E} introduced in Eq. (3), m=E(P→,P↑,P←,P↓)T\mathbf{m}=\mathbf{E}(P_{\rightarrow},P_{\uparrow},P_{\leftarrow},P_{\downarrow})^{\mathsf{T}}. This marginalization is followed by a hydrodynamic coarse-graining in which ρ\rho and m\mathbf{m} are retained as the slow fields. The coarse-grained density is not closed under the dynamics since transport between neighbouring cells is controlled by the local polarization which results in an active-matter hierarchy [82, 83]. At the polar level, however, the four microscopic reorientation processes enter through only two independent combinations; persistence and chirality. As derived in Appendix § 8, Eq. (67), the projection of the coin operator onto the real lattice space satisfies

EC=C(2)E,C(2)=qp1+qcϵ,(20)\mathbf{E}\mathbf{C}=\mathbf{C}_{(2)}\mathbf{E},\qquad\mathbf{C}_{(2)}=q_p\mathbf{1}+q_c\boldsymbol{\epsilon}, \tag*{(20)}

where C(2)\mathbf{C}_{(2)} is the real two-dimensional polar representation of the coin and ϵ\boldsymbol{\epsilon} generates a counter-clockwise rotation by π/2\pi/2. The eigenvalues of C(2)\mathbf{C}_{(2)} are the polar eigenvalue pair qp±iqcq_p\pm iq_c, which are exactly λ1\lambda_1 and λ3\lambda_3, defined in Eq. (2.9b). We note that only these polar eigenvalues enter the projected dynamics. Therefore, to simplify, we denote λ≡λ1=qp+iqc=∣λ∣exp⁡(iΩ)\lambda\equiv\lambda_1=q_p+iq_c=|\lambda|\exp(i\Omega), so that ∣λ∣=qp2+qc2|\lambda|=\sqrt{q_p^2+q_c^2} encodes the decay of the polar memory and Ω=arg⁡(qp+iqc)\Omega=\arg(q_p+iq_c) its rotation.

We retain the density and polarization to describe leading bulk diffusion. The nematic moment describes alignment along an axis without distinguishing its two opposite directions, and affects the density only beyond the second-order gradient approximation used here [84]. Consequently, the transport coefficients depend on qpq_p and qcq_c, while qnq_n can still affect short-scale and boundary behaviour. The derivation and the lattice-dependent higher-order corrections [85, 86] are discussed in Appendix § 8.

Hydrodynamic equations and closure

The ordering in gradients is made explicit by expanding Eq. (16) in the lattice constant ℓ\ell. Assuming that the shift r−ℓed\mathbf{r}-\ell\mathbf{e}_d is small compared with the scale on which ρ\rho and m\mathbf{m} vary, the detailed Taylor expansion and moment projection in Appendix § 8.§ 2 yield the density equation. To second order in gradients one obtains

∂ρ(r,t)∂t=D0∇2ρ(r,t)−νℓ∇⋅[(qp1+qcϵ)m(r,t)]+O(∇4),(21)\frac{\partial\rho(\mathbf{r},t)}{\partial t}=D_0\nabla^2\rho(\mathbf{r},t)-\nu\ell\nabla\cdot\left[(q_p\mathbf{1}+q_c\boldsymbol{\epsilon})\mathbf{m}(\mathbf{r},t)\right]+O(\nabla^4), \tag*{(21)}

where D0=ℓ2/(4Δt)D_0=\ell^2/(4\Delta t) is the diffusion coefficient for a memoryless lattice random walk in two dimensions. To arrive at a closed expression for the density we turn to the dynamics of the polarization field, which obeys, to leading order in gradients,

∂m(r,t)∂t=−ν[(1−qp)1−qcϵ]m(r,t)−νℓ2∇ρ(r,t)+O(∇2).(22)\frac{\partial\mathbf{m}(\mathbf{r},t)}{\partial t}=-\nu\left[(1-q_p)\mathbf{1}-q_c\boldsymbol{\epsilon}\right]\mathbf{m}(\mathbf{r},t)-\frac{\nu\ell}{2}\nabla\rho(\mathbf{r},t)+O(\nabla^2). \tag*{(22)}

This equation has a direct physical interpretation. In this leading continuous-time description, the polarization relaxes with rate ν(1−qp)\nu(1-q_p), while the chiral part rotates the polarization with angular frequency νqc\nu q_c. Thus positive qpq_p increases the orientational persistence time, negative qpq_p produces anti-persistent reversal, and nonzero qcq_c produces chiral rotation of the local polarization. Their finite-step corrections are discussed in § 5.3.

On diffusive timescales, the density is the slow conserved mode, whereas the polarization is a non-conserved orientational mode with a finite relaxation gap. Neglecting spatial gradients in Eq. (22) leads to exponential relaxation with characteristic time τm=[ν(1−qp)]−1\tau_m=[\nu(1-q_p)]^{-1} and rotation frequency νqc\nu q_c, whose handedness is dictated by the sign of qcq_c. The exceptional cases ∣λ∣=1|\lambda|=1 retain orientational memory indefinitely and thus fall out of the diffusion interpretation. These limiting cases have a simple dynamical interpretation: λ=1\lambda=1 yields ballistic motion, λ=−1\lambda=-1 yields deterministic reversal, and λ=±i\lambda=\pm i corresponds to counter-clockwise or clockwise four-step cycles, respectively, the latter having been analyzed in detail in Ref. [70].

For a density field varying over scales much larger than ℓ\ell, the density evolves diffusively and therefore more slowly than the polarization it induces. The polarization is first order in the density gradient, while its time derivative enters only at the next diffusive order. It can consequently be eliminated locally when its relaxation time is short compared with the evolution time of the density field. This standard multiple-scale argument is the basis of the following quasi-stationary closure [87]. The orientational relaxation matrix must be invertible, (1−qp)2+qc2≠0(1-q_p)^2+q_c^2\ne0, and the long-wavelength density rate must remain small compared with its relaxation scale. The approximation therefore fails near the ballistic limit qp→1q_p\to1, qc→0q_c\to0 unless correspondingly longer length and time scales are considered.

These requirements distinguish three steps in the reduction. First, the gradient expansion assumes ℓ∣∇∣≪1\ell|\nabla|\ll1. Second, the quasi-stationary closure assumes a time-scale separation between the slow density dynamics and the non-conserved polar mode. Third, interpreting the result as the smooth continuous-time limit of an active particle requires the orientational generator [(1−qp)1−qcϵ]/Δt[(1-q_p)\mathbf{1}-q_c\boldsymbol{\epsilon}]/\Delta t to remain finite as Δt→0\Delta t\to0, so 1−qp=O(Δt)1-q_p=O(\Delta t) and qc=O(Δt)q_c=O(\Delta t). We call this the effective weak-turning regime of the coarse-grained polar mode. It is microscopically realized by infrequent finite-angle reorientations and is distinct from a genuine small-angular-step limit; § 5.2 returns to this distinction in detail. Outside this scaling, the density equation can still describe the long-time, long-wavelength limit of a discrete walk, but it should not be interpreted as a smooth cABP limit, as we detail in § 5. The exact discrete Green–Kubo calculation in § 4.3 independently establishes the long-time transport coefficients for ∣λ∣<1|\lambda|<1, including regimes in which the continuum-time equation does not resolve the step-scale transient dynamics.

Setting ∂m/∂t≈0\partial\mathbf{m}/\partial t\approx0 at leading hydrodynamic order gives the local quasi-stationary polarization

m(r,t)=−ℓ2(1−qp)1+qcϵ(1−qp)2+qc2∇ρ(r,t).(23)\mathbf{m}(\mathbf{r},t)=-\frac{\ell}{2}\frac{(1-q_p)\mathbf{1}+q_c\boldsymbol{\epsilon}}{(1-q_p)^2+q_c^2}\nabla\rho(\mathbf{r},t). \tag*{(23)}

Substituting Eq. (23) into the density equation, Eq. (21), produces the closed current–density relation,

∂ρ(r,t)∂t=−∇⋅j(r,t),(24)\frac{\partial\rho(\mathbf{r},t)}{\partial t}=-\nabla\cdot\mathbf{j}(\mathbf{r},t), \tag*{(24)}

with

j(r,t)=−Deff∇ρ(r,t)−Doddϵ∇ρ(r,t).(25)\mathbf{j}(\mathbf{r},t)=-D_{\mathrm{eff}}\nabla\rho(\mathbf{r},t)-D_{\mathrm{odd}}\boldsymbol{\epsilon}\nabla\rho(\mathbf{r},t). \tag*{(25)}

The effective and odd diffusion coefficients can then be expressed in terms of the microscopic parameters as

Deff=D01−qp2−qc2(1−qp)2+qc2,Dodd=2D0qc(1−qp)2+qc2,(26)D_{\mathrm{eff}}=D_0\frac{1-q_p^2-q_c^2}{(1-q_p)^2+q_c^2},\qquad D_{\mathrm{odd}}=2D_0\frac{q_c}{(1-q_p)^2+q_c^2}, \tag*{(26)}

where D0=νℓ2/4D_0=\nu\ell^2/4 as before. Equation (25) separates the regular diffusive current from a chiral current perpendicular to the density gradient. For spatially constant transport coefficients, the latter is divergence-free because ∇⋅(ϵ∇ρ)=0\nabla\cdot(\boldsymbol{\epsilon}\nabla\rho)=0. Therefore, the bulk scalar density obeys the effective diffusion equation

∂ρ(r,t)∂t=Deff∇2ρ(r,t).(27)\frac{\partial\rho(\mathbf{r},t)}{\partial t}=D_{\mathrm{eff}}\nabla^2\rho(\mathbf{r},t). \tag*{(27)}

Odd current and boundary signature

The odd diffusion coefficient DoddD_{\mathrm{odd}} in Eq. (25) describes a probability current perpendicular to the density gradient. This contribution is divergence-free for homogeneous coefficients and therefore does not appear explicitly in the scalar density equation, Eq. (27), although chirality still affects spreading through DeffD_{\mathrm{eff}}. Figure 2 shows the density and local polarization m\mathbf{m}, rather than the probability current. Because the coin acts on the orientation before displacement, the outgoing vector current is

Figure 2. Figure 2. Probability density and orientational polarization of an active chiral random walker with qp=0.1q_p = 0.1 and qc=−0.8q_c = -0.8. Panels (a,c) show the evolution from a central source, while (b,d) show the evolution from a near-wall source. (a,b) Normalized probability density (black–red–yellow–white colour scale). (c,d) Direction and normalized magnitude of the local polarization m=∑dedPd\mathbf{m} = \sum_d \mathbf{e}_d P_d (blue–white–red colour scale), showing clockwise orientational circulation and boundary-induced redistribution. The near-wall density exhibits a transient wall enhancement, rather than stationary accumulation. Both initial conditions occupy a single lattice site with uniform orientations. Results are obtained after 2000 updates by exact iteration of Eq. (16) on a 150 × 150 lattice, with the specular boundary rule defined in Appendix § 9.

jlat(r,t)=νℓ(qp1+qcϵ)m(r,t).(28)\mathbf{j}_{\mathrm{lat}}(\mathbf{r},t)=\nu\ell\left(q_p\mathbf{1}+q_c\boldsymbol{\epsilon}\right)\mathbf{m}(\mathbf{r},t). \tag*{(28)}

The coin can therefore rotate the current relative to the polarization. The microscopic flux definition and its connection to the continuum current are given in Appendix § 8.§ 3. The plotted polarization illustrates clockwise orientational circulation, not an independent measurement of transverse current response.

Near a wall, the density and polarization also depend on the microscopic scattering rule and on orientational modes omitted from the bulk closure [88, 89]. The redistribution in Figure 2 must therefore be interpreted together with the specified boundary rule, rather than as a consequence of DoddD_{\mathrm{odd}} alone; the wall dynamics are defined in Appendix § 9.

Velocity correlations

A characteristic signature of chiral dynamics is an antisymmetric off-diagonal correlation between the components of the velocity vector [90, 91]. We therefore consider the full directional correlation tensor of a freely moving walker.

Directional correlations

As described in § 2.3, the conditional-step operator in Eq. (13) moves the walker along the lattice direction associated with its updated internal state. Let dnd_n denote the orientation after the nnth update. For update frequency ν\nu and lattice spacing ℓ\ell, the discrete velocity is vn=νℓedn\mathbf{v}_n = \nu\ell\mathbf{e}_{d_n}, where edn=E∣dn⟩\mathbf{e}_{d_n} = \mathbb{E}|d_n\rangle. Apart from the constant factor νℓ\nu\ell, velocity correlations are therefore directional correlations.

The normalized directional correlation tensor is

Cv(n)=2⟨edm+n⊗edm⟩=Cvdiag(n)1+Cvoff(n)ϵ.(29)\mathbf{C}_v(n) = 2\langle\mathbf{e}_{d_{m+n}} \otimes\mathbf{e}_{d_m}\rangle= C_v^{\mathrm{diag}}(n)\mathbf{1} + C_v^{\mathrm{off}}(n)\boldsymbol{\epsilon}. \tag*{(29)}

Here Cvdiag(n)=⟨edm+n⋅edm⟩C_v^{\mathrm{diag}}(n) = \langle\mathbf{e}_{d_{m+n}} \cdot\mathbf{e}_{d_m}\rangle is the standard diagonal velocity autocorrelation function (VACF), while Cvoff(n)=⟨edm+n⋅ϵedm⟩C_v^{\mathrm{off}}(n) = \langle\mathbf{e}_{d_{m+n}} \cdot\boldsymbol{\epsilon}\mathbf{e}_{d_m}\rangle is the off-diagonal correlation. For an isotropic initial distribution the correlation tensor equals the identity, Cv(0)=1\mathbf{C}_v(0) = \mathbf{1}. The conditional-average derivation in Appendix § 7 shows that successive powers of the projected coin determine both components. We find

Cvdiag(n)=∣λ∣ncos⁡(nΩ)=Re⁡λn,(30)C_v^{\mathrm{diag}}(n) = |\lambda|^n \cos(n\Omega) = \operatorname{Re}\lambda^n, \tag*{(30)}

and

Cvoff(n)=∣λ∣nsin⁡(nΩ)=Im⁡λn,(31)C_v^{\mathrm{off}}(n) = |\lambda|^n \sin(n\Omega) = \operatorname{Im}\lambda^n, \tag*{(31)}

where λ=∣λ∣exp⁡(iΩ)\lambda= |\lambda|\exp(\mathrm{i}\Omega) is the persistence-chirality parameter defined in § 3.1. Essentially, the coin correlates consecutive steps through a constant complex one-step factor, producing an exponentially decaying memory for ∣λ∣<1|\lambda| < 1 and oscillations when Ω≠0\Omega\ne0. These characteristics appear in the diagonal and off-diagonal correlations in Figure 3(c,Figure 3d). This makes explicit the connection between chirality and oscillatory directional correlations. In the model they are directly governed by intuitive microscopic constants qpq_p and qcq_c.

Figure 3. Simulation tests of the predicted correlations and transport coefficients. (a) Measured MSDs (markers) and long-time predictions MSD = 4Defft4D_{\mathrm{eff}}t (solid lines), using Eq. (26). (b) Diffusion-coefficient estimates (circles) and theoretical values (stars). DeffD_{\mathrm{eff}} is extracted from the MSD slope over 100≤t≤500100 \le t \le500, and DoddD_{\mathrm{odd}} from measured off-diagonal correlations using the discrete Green–Kubo as described in § 4.3. Markers show the mean across batches; error bars indicate ±1\pm1 standard deviation across 50 independent batches of 2000 trajectories. (c,d) Diagonal and off-diagonal velocity correlations: simulations (markers) and predictions from Eqs. (30) and (31) (solid lines). In panels (a,c,d), dark-grey circles represent the random walker with (qp,qc)=(0,0)(q_p,q_c)=(0,0), purple squares the chiral walker with (0,0.5)(0,0.5), green triangles the active chiral walker with (0.5,0.2)(0.5,0.2), and muted-orange diamonds the anti-persistent walker with (−0.5,0.2)(-0.5,0.2). Each ensemble contains 10510^5 trajectories evolved for 500 timesteps, with ℓ=Δt=1\ell=\Delta t=1.

We note that exponentially damped and oscillatory velocity correlations are also characteristic of active-chiral systems such as cABPs [54].

Exact finite-time mean-squared displacement

Now that we have identified the velocity correlations, we can calculate the exact finite-time MSD. After NN steps, the displacement is

rN−r0=ℓ∑j=1Nedj.(32)\mathbf{r}_N - \mathbf{r}_0 = \ell\sum_{j=1}^{N} \mathbf{e}_{d_j}. \tag*{(32)}

We denote the mean-squared displacement by MSDN=⟨∣rN−r0∣2⟩\mathrm{MSD}_N = \langle|\mathbf{r}_N - \mathbf{r}_0|^2\rangle. Using stationarity of the step increments and Cvdiag(n)=Re⁡(λn)C_v^{\mathrm{diag}}(n) = \operatorname{Re}(\lambda^n), we obtain

⟨∣rN−r0∣2⟩=ℓ2[N+2∑n=1N−1(N−n)Cvdiag(n)]=ℓ2[N+2Re⁡∑n=1N−1(N−n)λn].(33)\langle|\mathbf{r}_N - \mathbf{r}_0|^2\rangle= \ell^2\left[N + 2\sum_{n=1}^{N-1}(N-n)C_v^{\mathrm{diag}}(n)\right] = \ell^2\left[N + 2\operatorname{Re}\sum_{n=1}^{N-1}(N-n)\lambda^n\right]. \tag*{(33)}

For λ≠1\lambda\ne1, the geometric sum can be evaluated explicitly, leading to

⟨∣rN−r0∣2⟩=ℓ2[N+2Re⁡(λ[(N−1)−Nλ+λN](1−λ)2)].(34)\langle|\mathbf{r}_N - \mathbf{r}_0|^2\rangle= \ell^2\left[N + 2\operatorname{Re}\left(\frac{\lambda\left[(N-1)-N\lambda+\lambda^N\right]}{(1-\lambda)^2}\right)\right]. \tag*{(34)}

Equation (34) describes both the transient and long-time behaviour of the walk. It remains valid for λ=−1\lambda=-1 and λ=±i\lambda=\pm i; only λ=1\lambda=1 must be evaluated directly from Eq. (33). The corresponding special cases are

MSD⁡N={ℓ2N2,λ=1,0 for even N,ℓ2 for odd N,λ=−1,0, ℓ2, 2ℓ2, ℓ2 for (N mod 4)=0,1,2,3,λ=±i.(35)\operatorname{MSD}_{N} = \begin{cases} \ell^{2}N^{2}, & \lambda=1,\\ 0\ \text{for even }N,\quad\ell^{2}\ \text{for odd }N, & \lambda=-1,\\ 0,\ \ell^{2},\ 2\ell^{2},\ \ell^{2}\ \text{for }(N\bmod4)=0,1,2,3, & \lambda=\pm i. \end{cases} \tag*{(35)}

Thus, the ballistic case grows quadratically, while deterministic reversal and deterministic chiral rotation remain spatially bounded. For ∣λ∣<1|\lambda|<1, the leading large-NN contribution is

⟨∣rN−r0∣2⟩∼ℓ2N[1+2Re⁡(λ1−λ)]=4Defft,(36)\langle|\mathbf{r}_{N}-\mathbf{r}_{0}|^{2}\rangle\sim\ell^{2}N\left[1+2\operatorname{Re}\left(\frac{\lambda}{1-\lambda}\right)\right]=4D_{\mathrm{eff}}t, \tag*{(36)}

where t=NΔtt=N\Delta t. The long-time diffusion form, unlike the exact finite-time expression, requires ∣λ∣<1|\lambda|<1. At ∣λ∣=1|\lambda|=1, λ=1\lambda=1 gives ballistic motion, λ=−1\lambda=-1 deterministic reversal, and λ=±i\lambda=\pm i counter-clockwise or clockwise four-step cycles respectively.

Green–Kubo transport coefficients

The full even and odd transport tensor follows directly from a one-sided discrete Green–Kubo relation [59, 70], without the quasi-stationary closure or an assumption of slowly varying fields,

DGK=vℓ24[Cv(0)+2∑n=1∞Cv(n)]=DeffGK1+DoddGKϵ.(37)\mathbf{D}^{\mathrm{GK}}=\frac{v\ell^{2}}{4}\left[\mathbf{C}_{v}(0)+2\sum_{n=1}^{\infty}\mathbf{C}_{v}(n)\right]=D_{\mathrm{eff}}^{\mathrm{GK}}\mathbf{1}+D_{\mathrm{odd}}^{\mathrm{GK}}\boldsymbol{\epsilon}. \tag*{(37)}

The effective diffusion coefficient determines the long-time scalar spreading, whereas the odd-diffusion coefficient cancels from the MSD and the homogeneous scalar diffusion equation but remains in the current–density relation, Eq. (25). For ∣λ∣<1|\lambda|<1, evaluating the geometric sum ∑n=1∞λn=λ/(1−λ)\sum_{n=1}^{\infty}\lambda^{n}=\lambda/(1-\lambda), together with Cv(0)=1C_{v}(0)=1, leads to

DeffGK=vℓ24[1+2Re⁡(λ1−λ)]=vℓ241−qp2−qc2(1−qp)2+qc2,(38)D_{\mathrm{eff}}^{\mathrm{GK}}=\frac{v\ell^{2}}{4}\left[1+2\operatorname{Re}\left(\frac{\lambda}{1-\lambda}\right)\right]=\frac{v\ell^{2}}{4}\frac{1-q_{p}^{2}-q_{c}^{2}}{(1-q_{p})^{2}+q_{c}^{2}}, \tag*{(38)}
DoddGK=vℓ22Im⁡(λ1−λ)=vℓ22qc(1−qp)2+qc2.(39)D_{\mathrm{odd}}^{\mathrm{GK}}=\frac{v\ell^{2}}{2}\operatorname{Im}\left(\frac{\lambda}{1-\lambda}\right)=\frac{v\ell^{2}}{2}\frac{q_{c}}{(1-q_{p})^{2}+q_{c}^{2}}. \tag*{(39)}

The diagonal and off-diagonal correlations therefore determine the effective and odd current coefficients, respectively. In particular, DoddGKD_{\mathrm{odd}}^{\mathrm{GK}} changes sign under qc→−qcq_{c}\to-q_{c} and vanishes in the absence of chirality, qc=0q_{c}=0. Equations (38) and (39) are identical to the hydrodynamic coefficients in Eq. (26): DeffGK=DeffD_{\mathrm{eff}}^{\mathrm{GK}}=D_{\mathrm{eff}} and DoddGK=DoddD_{\mathrm{odd}}^{\mathrm{GK}}=D_{\mathrm{odd}}. Thus the two calculations retain the same leading long-time transport information, even though the hydrodynamic equation applies only after its additional scale-separation assumptions. This preservation of lattice transport, however, does not imply exact agreement with cABP correlations or transport at finite Δt\Delta t, as discussed in § 5.

Figure 3 tests these predictions: panel (a) compares the measured MSD with Eq. (36); panel (b) compares coefficient estimates with Eq. (26); and panels (c,d) compare the diagonal and off-diagonal correlations with Eqs. (30) and (31). Note that the odd-diffusion-coefficient comparison is a trajectory-based Green–Kubo check, not a separate measurement of transverse response to an imposed density gradient. The simulations agree closely with the predicted long-time MSD slopes and exact correlation functions. For each batch, Cvoff(n)C_{v}^{\mathrm{off}}(n) is estimated by averaging edm+n⋅edm\mathbf{e}_{d_{m+n}}\cdot\mathbf{e}_{d_m} over trajectories and all available time origins. The DoddD_{\mathrm{odd}} estimate is vℓ2/2v\ell^{2}/2 times the sum of these measured correlations over lags 1–20, which, as one can note in panels (c,d), is sufficient. The numerical implementation of the model and figure generation is openly available in Ref. [92].

Enhanced, reduced, and intrinsically discrete regimes

Normalizing Eq. (38) by the free lattice diffusion coefficient D0=ℓ2/(4Δt)D_0=\ell^2/(4\Delta t) identifies enhanced (Deff>D0D_{\mathrm{eff}}>D_0) and reduced (Deff<D0D_{\mathrm{eff}}<D_0) long-time diffusion. The boundary between enhanced and reduced diffusion is the circle

(qp−12)2+qc2=14.(40)\left(q_p-\frac{1}{2}\right)^2+q_c^2=\frac{1}{4}. \tag*{(40)}

restricted by normalization to the physically accessible diamond ∣qp∣+∣qc∣≤1|q_p|+|q_c|\le1, as shown in Figure 4. Hence the model provides a direct microscopic route for tuning the walker between persistence-enhanced and chirality- or reversal-suppressed diffusion. Further analysis of Eq. (38) separates the system into distinct dynamical regimes. For qc=0q_c=0,

Ratio of the effective diffusion coefficient as a function of persistence and chirality

Figure 4. Ratio of the effective diffusion coefficient in Eq. (38) to the free diffusion coefficient D0=ℓ2/(4Δt)D_0 = \ell^2/(4\Delta t) as a function of persistence qpq_p and chirality qcq_c. The dashed black circle is the boundary Deff/D0=1D_{\mathrm{eff}}/D_0 = 1 in Eq. (40); it separates enhanced from reduced long-time diffusion, which is also specifically accounted for in the two-scale colour-coding of the figure. The solid black diamond is the boundary imposed by the normalization condition, ∣qp∣+∣qc∣≤1|q_p| + |q_c| \le1; the region outside it is inaccessible.

DeffD0=1+qp1−qp.(41)\frac{D_{\mathrm{eff}}}{D_0}=\frac{1+q_p}{1-q_p}. \tag*{(41)}

Thus qp>0q_p>0 enhances diffusion by persistence, while qp<0q_p<0 suppresses diffusion by anti-persistence. In the limit qp→−1q_p\to-1, the walker repeatedly reverses direction and the long-time diffusion vanishes. This regime is intrinsically discrete and does not correspond to the usual smooth ABP limit. Indeed, for $q_p<0时时彩计划 and qc=0q_c = 0, one has λ=−∣λ∣\lambda= -|\lambda|, or equivalently Ω=π\Omega= \pi, far outside the weak-turning continuum condition imposed by the quasi-stationary closure [70].

For qp=0q_p = 0, chirality alone gives

DeffD0=1−qc21+qc2.(42)\frac{D_{\mathrm{eff}}}{D_0} = \frac{1-q_c^2}{1+q_c^2}. \tag*{(42)}

Thus, pure chirality reduces the scalar diffusion coefficient by bending trajectories into rotating motion. At the deterministic chiral point ∣qc∣=1|q_c| = 1, the walker circulates around elementary lattice four-step loops and the long-time scalar diffusion vanishes, even though the dynamics remains strongly chiral [70]. We note that the scalar diffusion coefficient measures spatial spreading and not the full degree of activity or irreversibility. In particular, the handedness of the walker does not affect it, as qcq_c enters only quadratically in Eq. (38), contrary to the odd-diffusion coefficient in Eq. (39) where the sign matters.

Connection to the chiral active Brownian particle model

Motivated by the connections identified throughout the manuscript, this section compares, maps, and distinguishes the aCRW and cABP models. Specifically, the polar mode of the aCRW displays the relaxation and rotation characteristic of persistent chiral particles, as prominently studied in the cABP model [42, 54, 90]. Here we compare the long-time even and odd transport coefficients and correlations with those of a cABP. The comparison is made first in the weak-turning limit, where the models best match. We then identify the temporal truncation behind this leading mapping and explain why exact matching of sampled velocity correlations does not imply exact transport matching.

Continuum cABP model

We consider a standard two-dimensional cABP with position r(t)\mathbf{r}(t) and orientation u(ϕ)=(cos⁡ϕ,sin⁡ϕ)\mathbf{u}(\phi) = (\cos\phi,\sin\phi) [54, 93]. Its overdamped Langevin dynamics are

dr=v~0u dt+2D~ dB,dϕ=ω~ dt+2D~r dW.(43)d\mathbf{r} = \widetilde{v}_0\mathbf{u}\,dt + \sqrt{2\widetilde{D}}\,d\mathbf{B}, \qquad d\phi= \widetilde{\omega}\,dt + \sqrt{2\widetilde{D}_r}\,dW. \tag*{(43)}

Here v~0\tilde{v}_0 is the self-propulsion speed, D~\tilde{D} is an independent translational diffusion coefficient, D~r\tilde{D}_r is the rotational diffusion coefficient, and ω~\tilde{\omega} is the intrinsic angular velocity. We denote B\mathbf{B} and W\mathbf{W} are standard Wiener processes. We use tildes to distinguish the cABP coefficients from those of the discrete model. Throughout, positive chirality denotes counter-clockwise rotation: ω~>0\tilde{\omega} > 0 produces counter-clockwise angular drift, and ϵ\boldsymbol{\epsilon} represents a counter-clockwise rotation by π/2\pi/2.

The corresponding Fokker–Planck equation for the angular PDF p~(r,ϕ,t)\tilde{p}(\mathbf{r},\phi,t) is

∂p~(r,ϕ,t)∂t=D~∇2p~−v~0u⋅∇p~+D~r∂2p~∂ϕ2−ω~∂p~∂ϕ.(44)\frac{\partial\tilde{p}(\mathbf{r},\phi,t)}{\partial t} = \tilde{D}\nabla^2\tilde{p} - \tilde{v}_0\mathbf{u}\cdot\nabla\tilde{p} + \tilde{D}_r\frac{\partial^2\tilde{p}}{\partial\phi^2} - \tilde{\omega}\frac{\partial\tilde{p}}{\partial\phi}. \tag*{(44)}

The cABP PDF and polarization field are

ρ~(r,t)=∫02πp~(r,ϕ,t) dϕ,(45)\tilde{\rho}(\mathbf{r},t) = \int_0^{2\pi} \tilde{p}(\mathbf{r},\phi,t)\,d\phi, \tag*{(45)}

and

m~(r,t)=∫02πu(ϕ)p~(r,ϕ,t) dϕ.(46)\tilde{\mathbf{m}}(\mathbf{r},t) = \int_0^{2\pi} \mathbf{u}(\phi)\tilde{p}(\mathbf{r},\phi,t)\,d\phi. \tag*{(46)}

Following standard projection schemes [82, 83], we apply a first-harmonic, long-wavelength closure in which the nematic moment is neglected. This yields

∂ρ~(r,t)∂t=D~∇2ρ~(r,t)−v~0∇⋅m~(r,t),(47)\frac{\partial\tilde{\rho}(\mathbf{r},t)}{\partial t} = \tilde{D}\nabla^2\tilde{\rho}(\mathbf{r},t) - \tilde{v}_0\nabla\cdot\tilde{\mathbf{m}}(\mathbf{r},t), \tag*{(47)}

and

∂m~(r,t)∂t=D~∇2m~(r,t)−D~rm~(r,t)+ω~ϵm~(r,t)−v~02∇ρ~(r,t).(48)\frac{\partial\tilde{\mathbf{m}}(\mathbf{r},t)}{\partial t} = \tilde{D}\nabla^2\tilde{\mathbf{m}}(\mathbf{r},t) - \tilde{D}_r\tilde{\mathbf{m}}(\mathbf{r},t) + \tilde{\omega}\boldsymbol{\epsilon}\tilde{\mathbf{m}}(\mathbf{r},t) - \frac{\tilde{v}_0}{2}\nabla\tilde{\rho}(\mathbf{r},t). \tag*{(48)}

The factor 1/21/2 in the last term follows from the isotropic part of the second angular moment. The complete marginalization is shown in Appendix § 10. In the slow-density, fast-polarization regime, quasi-stationary elimination of m~\tilde{\mathbf{m}} gives j~=−D~eff∇ρ~−D~oddϵ∇ρ~\tilde{\mathbf{j}} = -\tilde{D}_{\mathrm{eff}}\nabla\tilde{\rho} - \tilde{D}_{\mathrm{odd}}\boldsymbol{\epsilon}\nabla\tilde{\rho}, where

D~eff=D~+v~02D~r2(D~r2+ω~2),D~odd=v~02ω~2(D~r2+ω~2).(49)\tilde{D}_{\mathrm{eff}} = \tilde{D} + \frac{\tilde{v}_0^2\tilde{D}_r}{2(\tilde{D}_r^2+\tilde{\omega}^2)}, \qquad\tilde{D}_{\mathrm{odd}} = \frac{\tilde{v}_0^2\tilde{\omega}}{2(\tilde{D}_r^2+\tilde{\omega}^2)}. \tag*{(49)}

Thus, the spatial diffusion coefficient D~\tilde{D} contributes only to the even coefficient, while D~odd\tilde{D}_{\mathrm{odd}} is generated entirely by the coupling between propulsion and persistent angular rotation. Positive ω~\tilde{\omega} gives D~odd>0\tilde{D}_{\mathrm{odd}} > 0, consistently with the counter-clockwise convention.

Weak-turning transport correspondence

We now compare these coefficients with the exact aCRW coefficients in Eq. (26). The relevant continuum limit keeps the cABP rates and speed fixed while the discrete time and length scales vanish together

qp=1−D~rΔt+O(Δt2),qc=ω~Δt+O(Δt2),ℓ=v~0Δt.(50)q_p = 1 - \tilde{D}_r\Delta t + O(\Delta t^2), \qquad q_c = \tilde{\omega}\Delta t + O(\Delta t^2), \qquad\ell= \tilde{v}_0\Delta t. \tag*{(50)}

At finite Δt\Delta t, the corresponding leading-order identification used below is

D~r=ν(1−qp),ω~=νqc,v~0=νℓ.(51)\tilde{D}_r = \nu(1-q_p), \qquad\tilde{\omega} = \nu q_c, \qquad\tilde{v}_0 = \nu\ell. \tag*{(51)}

In this scaling, the lattice reference coefficient vanishes,

D0=ℓ24Δt=v~02Δt4→Δt→00.(52)D_0 = \frac{\ell^2}{4\Delta t} = \frac{\tilde{v}_0^2\Delta t}{4} \xrightarrow{\Delta t\to0} 0. \tag*{(52)}

The vanishing of the aCRW coefficient D0D_0 in the continuum limit is expected, since it arises from the finite lattice step size. To compare the two models on equal footing, we consider a cABP without independent translational Brownian diffusion, D~=0\widetilde{D}=0, and compare only the contributions arising from the active dynamics.

Substitution of Eq. (50) into the aCRW coefficients gives

Deff→Δt→0v~02D~r2(D~r2+ω~2)=D~eff∣D~=0,D_{\mathrm{eff}} \xrightarrow{\Delta t \to0} \frac{\widetilde{v}_0^2 \widetilde{D}_r}{2\left(\widetilde{D}_r^2+\widetilde{\omega}^2\right)}=\left.\widetilde{D}_{\mathrm{eff}}\right|_{\widetilde{D}=0},
Dodd→Δt→0v~02ω~2(D~r2+ω~2)=D~odd.(53)D_{\mathrm{odd}} \xrightarrow{\Delta t \to0} \frac{\widetilde{v}_0^2\widetilde{\omega}}{2\left(\widetilde{D}_r^2+\widetilde{\omega}^2\right)}=\widetilde{D}_{\mathrm{odd}}. \tag*{(53)}

The even and odd long-time transport coefficients of the aCRW therefore approach those of a cABP without independent translational diffusion. This establishes an asymptotic transport correspondence between the models in the weak-turning limit.

The stochastic coin restricts which cABP rates can be reached by this four-direction construction. Near qp=1q_p=1, the physical domain ∣qp∣+∣qc∣≤1\lvert q_p\rvert+\lvert q_c\rvert\le1 becomes ∣qc∣≤1−qp\lvert q_c\rvert\le1-q_p. With Eq. (50), this yields ∣ω~∣≤D~r\lvert\widetilde{\omega}\rvert\le\widetilde{D}_r. In this limit, the aCRW reorientation events become infrequent, although the walker still changes direction through a finite angle of 90∘90^\circ. This differs from the cABP, for which the change in orientation becomes infinitesimally small. We therefore use weak turning for the effective coarse-grained polar dynamics and infrequent reorientation for its microscopic realization. A genuine small-angular-step limit would instead require an orientational lattice with zz directions and z→∞z\to\infty [66].

Finite-step correlations and the continuum approximation

This comparison connects the intuitive active-matter coefficients qpq_p and qcq_c to cABP relaxation and rotation rates, while making the finite-step approximation explicit. The underlying relation is general: for any mode multiplied by a nonzero factor λ\lambda at each update, its exponential continuous-time rate is log⁡λ/Δt\log\lambda/\Delta t. In the aCRW, the factor is simply λ=qp+iqc\lambda=q_p+iq_c, so the exact rate and its continuum approximation follow directly from the microscopic probabilities. The aCRW correlations contain the exact factor λn\lambda^n, whereas the cABP first angular harmonic evolves as exp⁡[(−D~r+iω~)t]\exp[(-\widetilde{D}_r+i\widetilde{\omega})t]. For 0<∣λ∣<10<\lvert\lambda\rvert<1, matching them at t=nΔtt=n\Delta t gives

D~rcorr=−νln⁡∣λ∣=−νln⁡qp2+qc2,ω~corr=νarg⁡λ.(54)\widetilde{D}_r^{\mathrm{corr}}=-\nu\ln\lvert\lambda\rvert=-\nu\ln\sqrt{q_p^2+q_c^2},\qquad\widetilde{\omega}^{\mathrm{corr}}=\nu\arg\lambda. \tag*{(54)}

The phase is chosen continuously near λ=1\lambda=1.

The leading continuum map in Eq. (51) instead uses (λ−1)/Δt=−D~r+iω~(\lambda-1)/\Delta t=-\widetilde{D}_r+i\widetilde{\omega}. Their relationship follows directly from

log⁡λ=(λ−1)−(λ−1)22+O(∣λ−1∣3).(55)\log\lambda=(\lambda-1)-\frac{(\lambda-1)^2}{2}+O\left(\lvert\lambda-1\rvert^3\right). \tag*{(55)}

The leading map therefore retains only the first term of the exact correlation-rate expansion. The omitted corrections vanish in the weak-turning limit and are calculated in Appendix E.1.

The simplicity of the model makes the omitted temporal corrections and preserved long-time transport explicit. The homogeneous polar projection in Eq. (20) preserves λ\lambda; replacing a finite time difference by a derivative approximates its evolution, rather than discarding the polar mode. This differs from neglecting higher angular moments. Nevertheless, the hydrodynamic coefficients coincide with the exact long-time lattice coefficients obtained in § 4.3. The aCRW therefore gives an explicit microscopic example of a general distinction: preserving long-time transport does not require reproducing the exact transient dynamics of the retained mode.

Consequences for transport

Matching correlations at update times does not guarantee matching transport coefficients: the lattice Green–Kubo relation sums correlations, whereas the cABP relation integrates them over continuous time. Even identical sampled values need not give the same sum and integral. Appendix E.2 shows that the difference under correlation matching is of order Δt2\Delta t^2 at fixed correlation-matched rates and speed.

For comparison, evaluating the cABP coefficients with the leading map, equal speeds v~0=ℓ/Δt\tilde{v}_0 = \ell/\Delta t, and D~=0\tilde{D} = 0 gives

D~eff−Deff=D0,D~odd=Dodd.(56)\tilde{D}_{\mathrm{eff}} - D_{\mathrm{eff}} = D_0,\qquad\tilde{D}_{\mathrm{odd}} = D_{\mathrm{odd}}. \tag*{(56)}

Thus the leading map has an exact offset in the even coefficient, but reproduces the odd coefficient. At fixed speed, D0=v~02Δt/4D_0 = \tilde{v}_0^2\Delta t/4 vanishes in the continuum limit.

The microscopic distinction is also visible in the current. The cABP active current is v~0m~\tilde{v}_0\tilde{\mathbf{m}}, whereas the aCRW current contains (q1p1+qcϵ)m(q_1\mathbf{p}_1 + q_c\boldsymbol{\epsilon})\mathbf{m}: the coin acts directly on the outgoing flux during each update. This finite-step correction disappears in the weak-turning limit. Strong chirality and anti-persistence remain natural aCRW regimes outside the standard cABP correspondence, although related dynamics can be represented by velocity-jump, run-and-tumble, and telegraph models [94, 95, 96].

Figure 5 illustrates these distinctions. The leading map describes the asymptotic transport correspondence, while the correlation map reproduces the sampled bulk first angular harmonic exactly. For both parameter sets shown, the correlation-matched cABP also agrees well with the measured MSD, although correlation matching alone does not guarantee this agreement. The relative long-time diffusion errors are 0.2% and 3.7% for correlation matching, versus 11% and 67% for the leading map (weak and strong turning, respectively). Neither map identifies the higher angular harmonics or the full spatial distributions; the confined contours additionally depend on propagation and wall scattering.

Figure 5. Comparison of the aCRW and cABP in the weak- and strong-turning regimes, with D~=0\widetilde{D}=0 in all cABP calculations. Green denotes weak turning, with qp=0.9q_p=0.9 and qc=0.1q_c=0.1, whereas purple denotes strong turning, with qp=0.6q_p=0.6 and qc=0.4q_c=0.4. We use lattice units ℓ=Δt=1\ell=\Delta t=1. (a) MSD as a function of time. (b) Diagonal velocity autocorrelation CvdiagC_{\mathrm{v}}^{\mathrm{diag}}. Markers show aCRW ensemble averages over N=104N=10^4 trajectories. (c,d) Contours enclosing 90% of the probability in the highest-density regions for weak and strong turning, respectively. Both models start with uniform orientations and the same Gaussian spatial distribution, centred at (x0,y0)=(75.5,10.5)(x_0,y_0)=(75.5,10.5) in a square domain of side length 151. The Gaussian has standard deviation σ≈2.043\sigma\approx2.043 per coordinate, corresponding to a nominal radius R95=5R_{95}=5 containing 95% of its unbounded probability, and is conditioned to lie inside the domain. Crosses mark its centre. All four walls are specularly reflecting, see Appendix § 9 for details. Solid contours show the aCRW probability density obtained by iterating the lattice master equation. The cABP densities are obtained numerically using a finite-volume spatial grid with spacing of 0.5, 64 angular bins, and integration time step 0.2. Both densities are binned onto a common 76 ×\times 76 spatial grid before contour extraction. Lighter to darker shades indicate t=50t=50, 100, 200, 500, 1000. Throughout the figure, dashed curves and contours use the continuum mapping in Eq. (51), whereas dotted curves and contours use the correlation map in Eq. (54). The continuum mapping captures the asymptotic transport correspondence in the weak-turning limit, whereas the correlation map reproduces the bulk first angular harmonic exactly at integer update times. For the two parameter sets shown, the correlation-matched cABP also agrees well with the measured bulk MSD, even though correlation matching alone does not guarantee agreement of the MSD or the confined spatial distributions.

Conclusion

We introduced a minimal, intuitive, and computationally efficient active chiral random walk in which persistence, reversal, chirality, and randomization are controlled directly by a small set of microscopic probabilities with explicit dynamical meanings. The complex polar eigenvalue of the coin operator, λ=qp+iqc\lambda= q_p + \mathrm{i}q_c, governs the exact directional correlations, finite-time MSD, and leading second-order bulk transport coefficients, while qnq_n controls nematic and higher-gradient information outside the regime analysed here. The hydrodynamic and discrete Green–Kubo calculations yield identical long-time even and odd coefficients, showing which transport information survives the coarse-graining.

In the weak-turning limit generated by infrequent finite-angle reorientations, the aCRW reproduces the first-angular-harmonic dynamics and leading transport of the cABP. Beyond this correspondence, the minimal model provides a transparent benchmark for a general issue in coarse-graining: a retained mode can have approximate transient dynamics while the leading transport remains exact. Here both the omitted temporal corrections and the preserved lattice coefficients can be traced directly to the microscopic probabilities. Matching sampled correlations also leaves a distinction between their discrete sum and the continuous-time integral governing cABP transport. The four-state model does not become a continuous angular process: strong turning, deterministic chiral cycles, and anti-persistent motion remain intrinsically discrete regimes.

A distinctive result is the closed density-current relation

j=−Deff∇ρ−Doddϵ∇ρ,\mathbf{j} = -D_{\mathrm{eff}}\nabla\rho- D_{\mathrm{odd}}\boldsymbol{\epsilon}\nabla\rho,

in which chirality generates an odd current perpendicular to the density gradient. This term is divergence-free in a homogeneous bulk. At a boundary, however, rejected bonds and the microscopic scattering kernel alter the outgoing orientational populations, so wall accumulation must be interpreted jointly with the boundary rule rather than from DoddD_{\mathrm{odd}} alone.

The model is especially suited to confined, disordered, networked, and interacting systems because its microscopic rules remain transparent and inexpensive to simulate. It also suggests a direct experimental realization with programmable self-propelled robots or “Hexbugs”, for which persistence, reversal, and handed turning can be imposed as discrete control instructions and both trajectories and boundary currents can be measured [97, 98]. Further extensions could test how interactions, external forcing, or spatially varying parameters convert the transverse probability current into measurable odd mobility or Hall-like response. Beyond the MSD, the model also provides a tractable setting for studying non-Gaussian displacement statistics in active motion [72, 99]. Its finite microscopic propagation speed may

likewise be useful in search problems for which telegrapher-like dynamics alter extreme first-passage statistics [100, 101]. The exact directional correlations also make the velocity power spectrum directly accessible, providing an experimentally convenient frequency-domain characterization of persistence and chirality.

Data accessibility

The Python source code and figure-generation scripts supporting this article are archived as aCRW codes, version v1.2.0 [92]: https://doi.org/10.5281/zenodo.24060170. No external experimental data were used in this work.

Acknowledgements

N. S. and E. K. acknowledge support from the Tel Aviv University–University of Potsdam "Research Cooperation Program" financing mutual 3-month stays that established their collaboration and eventually led to this work. R. M. acknowledges support by the Deutsche Forschungsgemeinschaft (grants No. ME 1535/22-1, and No. 318763901 within the CRC 1294 “Data Assimilation”, project B10). Y.R. and N.S. acknowledge support from the Israel Science Foundation (grants No. 385/21 and 2013/25).

Derivation of the directional correlations

Let dt∈Dd_t \in D denote the internal orientation at time tt, and let edt=E∣dt⟩\mathbf{e}_{d_t} = \mathbb{E}|d_t\rangle be the corresponding direction in the two-dimensional physical space, with E∈R2×4\mathbf{E} \in\mathbb{R}^{2 \times4} defined in Eq. (3). The following derivation applies to the spatially homogeneous walk, for which the same coin C\mathbf{C} is used at every update. Although C\mathbf{C} does not determine the direction of an individual stochastic realization, it determines its conditional mean. At a lag of nn updates each occurring after Δt\Delta t,

E[edt+nΔt∣dt=d]=ECn∣d⟩=C(2)nE∣d⟩=C(2)ned,(57)\mathbb{E}[\mathbf{e}_{d_{t+n\Delta t}} \mid d_t=d] = \mathbf{E}\mathbf{C}^{n}|d\rangle= \mathbf{C}_{(2)}^{n}\mathbf{E}|d\rangle= \mathbf{C}_{(2)}^{n}\mathbf{e}_{d}, \tag*{(57)}

where EC=C(2)E\mathbf{E}\mathbf{C} = \mathbf{C}_{(2)}\mathbf{E}, with C(2)=qp1+qcϵ\mathbf{C}_{(2)} = q_p\mathbf{1} + q_c\boldsymbol{\epsilon} as introduced in Eq. (20). Here 1∈R2×2\mathbf{1} \in\mathbb{R}^{2 \times2} is the identity matrix and ϵ∈R2×2\boldsymbol{\epsilon} \in\mathbb{R}^{2 \times2} generates a counter-clockwise rotation by π/2\pi/2.

For an isotropic stationary distribution, each lattice direction occurs with probability 1/41/4, and hence its expectation is

⟨edt⊗edt⟩=14∑d∈DededT=121.(58)\langle\mathbf{e}_{d_t} \otimes\mathbf{e}_{d_t} \rangle= \frac{1}{4}\sum_{d\in D}\mathbf{e}_d\mathbf{e}_d^{\mathsf{T}} = \frac{1}{2}\mathbf{1}. \tag*{(58)}

Conditional averaging of Eq. (57) then gives the normalised velocity-autocorrelation tensor (VACT) used in the main text,

Cv(n)≡2⟨edt+nΔt⊗edt⟩=2C(2)n⟨edt⊗edt⟩=C(2)n.(59)\mathbf{C}_v(n) \equiv2\langle\mathbf{e}_{d_{t+n\Delta t}} \otimes\mathbf{e}_{d_t}\rangle= 2\mathbf{C}_{(2)}^n\langle\mathbf{e}_{d_t} \otimes\mathbf{e}_{d_t}\rangle= \mathbf{C}_{(2)}^n. \tag*{(59)}

In particular, Cv(0)=1\mathbf{C}_v(0)=\mathbf{1}. Its decomposition into diagonal and antisymmetric parts is [90, 91]

Cv(n)=Cvdiag(n)1+Cvoff(n)ϵ,(60)\mathbf{C}_v(n) = C_v^{\mathrm{diag}}(n)\mathbf{1} + C_v^{\mathrm{off}}(n)\boldsymbol{\epsilon}, \tag*{(60)}

where

Cvdiag(n)=⟨edt+nΔt⋅edt⟩,(61)C_v^{\mathrm{diag}}(n) = \langle\mathbf{e}_{d_{t+n\Delta t}}\cdot\mathbf{e}_{d_t}\rangle, \tag*{(61)}
Cvoff(n)=⟨edt+nΔt⋅ϵedt⟩.(62)C_v^{\mathrm{off}}(n) = \langle\mathbf{e}_{d_{t+n\Delta t}}\cdot\boldsymbol{\epsilon}\mathbf{e}_{d_t}\rangle. \tag*{(62)}

The eigenvalues of C(2)\mathbf{C}_{(2)} are the conjugate polar pair λ=qp+iqc\lambda=q_p+\mathrm{i}q_c and λˉ\bar{\lambda}, already obtained from the full coin in Eqs. (2.9b) and (2.9d). Writing λ=∣λ∣exp⁡(iΩ)\lambda=|\lambda|\exp(\mathrm{i}\Omega), with ∣λ∣=qp2+qc2|\lambda|=\sqrt{q_p^2+q_c^2} and Ω=arg⁡λ\Omega=\arg\lambda, gives

C(2)n=∣λ∣n[cos⁡(nΩ)1+sin⁡(nΩ)ϵ]=∣λ∣n(cos⁡(nΩ)−sin⁡(nΩ)sin⁡(nΩ)cos⁡(nΩ)),(63)\mathbf{C}_{(2)}^n = |\lambda|^n[\cos(n\Omega)\mathbf{1}+\sin(n\Omega)\boldsymbol{\epsilon}] = |\lambda|^n \begin{pmatrix} \cos(n\Omega) & -\sin(n\Omega) \\ \sin(n\Omega) & \cos(n\Omega) \end{pmatrix}, \tag*{(63)}

The sign of the off-diagonal entries follows from the counter-clockwise-positive convention for ϵ\epsilon. Comparing Eqs. (60) and (63) yields

Cνdiag(n)=∣λ∣ncos⁡(nΩ)=Re⁡(λn),Cνoff(n)=∣λ∣nsin⁡(nΩ)=Im⁡(λn),(64)C_{\nu}^{\mathrm{diag}}(n)=|\lambda|^{n}\cos(n\Omega)=\operatorname{Re}(\lambda^{n}),\qquad C_{\nu}^{\mathrm{off}}(n)=|\lambda|^{n}\sin(n\Omega)=\operatorname{Im}(\lambda^{n}), \tag*{(64)}

which are Eqs. (30) and (31). Thus ∣λ∣|\lambda| sets the decay of polar memory and Ω\Omega sets its rotation and handedness. The nematic eigenvalue qnq_n does not enter either correlation.

Polar reduction and hydrodynamic expansion

Projection onto the orientation basis

In the ordered orientation basis D={→,↑,←,↓}D=\{\rightarrow,\uparrow,\leftarrow,\downarrow\}, the operator ϵ\epsilon introduced in Eq. (20) rotates physical vectors counter-clockwise by π/2\pi/2. Thus a clockwise turn, reversal, and counter-clockwise turn are represented in the physical plane by ϵ3\epsilon^3, ϵ2\epsilon^2, and ϵ\epsilon, respectively. Projecting the reduced coin with the geometry matrix gives

EC=(q^I1+q^Rϵ3+q^Bϵ2+q^Lϵ)E,(65)E C=\left(\hat q_I\mathbf{1}+\hat q_R\epsilon^3+\hat q_B\epsilon^2+\hat q_L\epsilon\right)E, \tag*{(65)}

where, in two-dimensional space,

ϵ2=−1,ϵ3=−ϵ,ϵ4=1.(66)\epsilon^2=-1,\qquad\epsilon^3=-\epsilon,\qquad\epsilon^4=1. \tag*{(66)}

It follows that

EC=[(q^I−q^B)1+(q^L−q^R)ϵ]E=(qp1+qcϵ)E≡C(2)E.(67)E C=\left[(\hat q_I-\hat q_B)\mathbf{1}+(\hat q_L-\hat q_R)\epsilon\right]E=(q_p\mathbf{1}+q_c\epsilon)E\equiv C_{(2)}E. \tag*{(67)}

The polar projection therefore retains only the two contrasts qp=q^I−q^Bq_p=\hat q_I-\hat q_B and qc=q^L−q^Rq_c=\hat q_L-\hat q_R. Since the eigenvalues of ϵ\epsilon are ±i\pm i, the eigenvalues of C(2)C_{(2)} are qp±iqc=λ1,3q_p\pm iq_c=\lambda_{1,3}, as in Eqs. (2.9b) and (2.9d).

Gradient expansion and moment equations

We now show explicitly where the omitted nematic mode enters the gradient hierarchy. Let P(r,t)=(P→,P↑,P←,P↓)T(r,t)\mathbf{P}(\mathbf{r},t)=(P_{\rightarrow},P_{\uparrow},P_{\leftarrow},P_{\downarrow})^{\mathrm{T}}(\mathbf{r},t). Its scalar, polar, and nematic moments are

ρ(r,t)=14TP(r,t),m(r,t)=EP(r,t),a(r,t)=P→(r,t)+P←(r,t)−P↑(r,t)−P↓(r,t),(68)\rho(\mathbf{r},t)=\mathbf{1}_4^{\mathrm{T}}\mathbf{P}(\mathbf{r},t),\qquad\mathbf{m}(\mathbf{r},t)=E\mathbf{P}(\mathbf{r},t),\qquad a(\mathbf{r},t)=P_{\rightarrow}(\mathbf{r},t)+P_{\leftarrow}(\mathbf{r},t)-P_{\uparrow}(\mathbf{r},t)-P_{\downarrow}(\mathbf{r},t), \tag*{(68)}

where 14=(1,1,1,1)T\mathbf{1}_4=(1,1,1,1)^{\mathrm{T}}, and the last quantity is the amplitude of the symmetric traceless nematic tensor

N(r,t)=a(r,t)2(100−1).(69)N(\mathbf{r},t)=\frac{a(\mathbf{r},t)}{2}\begin{pmatrix}1&0\\0&-1\end{pmatrix}. \tag*{(69)}

Greek indices below denote Cartesian coordinates and repeated Greek indices are summed. Expanding the master equation, Eq. (16), for fields that vary slowly over one lattice spacing gives

∂Pd(r,t)∂t=ν{∑d′Cdd′[1−ℓed,α∇α+ℓ22ed,αed,β∇α∇β]Pd′(r,t)−Pd(r,t)}+O(ℓ3∇3).(70)\frac{\partial P_d(\mathbf{r},t)}{\partial t}=\nu\left\{\sum_{d'}C_{dd'}\left[1-\ell e_{d,\alpha}\nabla_\alpha+\frac{\ell^2}{2}e_{d,\alpha}e_{d,\beta}\nabla_\alpha\nabla_\beta\right]P_{d'}(\mathbf{r},t)-P_d(\mathbf{r},t)\right\}+O(\ell^3\nabla^3). \tag*{(70)}

The first outgoing moment is C(2)mC_{(2)}\mathbf{m} by Eq. (67). The second outgoing moment is

Mαβout(r,t)≡∑d,d′ed,αed,βCdd′Pd′(r,t),Mout=12(ρ(r,t)+qna(r,t)00ρ(r,t)−qna(r,t)),(71)M_{\alpha\beta}^{\mathrm{out}}(\mathbf{r},t)\equiv\sum_{d,d'}e_{d,\alpha}e_{d,\beta}C_{dd'}P_{d'}(\mathbf{r},t),\qquad\mathbf{M}^{\mathrm{out}}=\frac{1}{2}\begin{pmatrix}\rho(\mathbf{r},t)+q_na(\mathbf{r},t)&0\\0&\rho(\mathbf{r},t)-q_na(\mathbf{r},t)\end{pmatrix}, \tag*{(71)}

where qnq_n is the nematic eigenvalue in Eq. (2.9c). Summing Eq. (70) over dd now produces

∂ρ(r,t)∂t=−νℓ∇⋅[C(2)m(r,t)]+νℓ24[∇2ρ(r,t)+qn(∇x2−∇y2)a(r,t)]+O(ℓ3∇3m).(72)\frac{\partial\rho(\mathbf{r},t)}{\partial t}=-\nu\ell\nabla\cdot\left[C_{(2)}\mathbf{m}(\mathbf{r},t)\right]+\frac{\nu\ell^2}{4}\left[\nabla^2\rho(\mathbf{r},t)+q_n(\nabla_x^2-\nabla_y^2)a(\mathbf{r},t)\right]+O(\ell^3\nabla^3\mathbf{m}). \tag*{(72)}

The operator ∇x2−∇y2\nabla_x^2-\nabla_y^2 here is the first explicit imprint of the Cartesian lattice symmetry.

Projecting the same expanded master equation onto the polarization yields

∂m(r,t)∂t=−ν[(1−qp)1−qcϵ]m(r,t)−νℓ2∇ρ(r,t)−νℓqn2(∇xa(r,t)−∇ya(r,t))+O(ℓ2∇2m).(73)\frac{\partial\mathbf{m}(\mathbf{r},t)}{\partial t}=-\nu\left[(1-q_p)\mathbf{1}-q_c\boldsymbol{\epsilon}\right]\mathbf{m}(\mathbf{r},t)-\frac{\nu\ell}{2}\nabla\rho(\mathbf{r},t)-\frac{\nu\ell q_n}{2}\begin{pmatrix}\nabla_x a(\mathbf{r},t)\\-\nabla_y a(\mathbf{r},t)\end{pmatrix}+O(\ell^2\nabla^2\mathbf{m}). \tag*{(73)}

Finally, projecting onto aa gives, to the order needed here,

∂a(r,t)∂t=−ν(1−qn)a(r,t)−νℓ[∇x(C(2)m(r,t))x−∇y(C(2)m(r,t))y]+νℓ24[(∇x2−∇y2)ρ(r,t)+qn∇2a(r,t)]+O(ℓ3∇3m).(74)\frac{\partial a(\mathbf{r},t)}{\partial t}=-\nu(1-q_n)a(\mathbf{r},t)-\nu\ell\left[\nabla_x\left(C_{(2)}\mathbf{m}(\mathbf{r},t)\right)_x-\nabla_y\left(C_{(2)}\mathbf{m}(\mathbf{r},t)\right)_y\right]+\frac{\nu\ell^2}{4}\left[(\nabla_x^2-\nabla_y^2)\rho(\mathbf{r},t)+q_n\nabla^2a(\mathbf{r},t)\right]+O(\ell^3\nabla^3\mathbf{m}). \tag*{(74)}

Equations (B9) and (B10) make the closure transparent. Provided the nematic mode is non-conserved, 1−qn≠01-q_n\ne0, the density first generates m=O(∇ρ)\mathbf{m}=O(\nabla\rho), while aa is generated only at O(∇2ρ)O(\nabla^2\rho). Its feedback into the polarization is therefore third order in gradients, and its explicit contribution to Eq. (72) is fourth order. The second-order bulk theory consequently retains only ρ\rho and m\mathbf{m}; keeping qnq_n consistently would also require the square-lattice, fourth-gradient corrections. The quasi-stationary elimination and the resulting even and odd currents are carried out in the main text, from Eq. (22) through Eq. (26).

Microscopic probability current

The microscopic current is fixed by the probability transported along lattice bonds. Since the coin acts before the displacement, the outgoing populations are

Pd+(r,t)=∑d′Cdd′Pd′−(r,t).(75)P_d^+(\mathbf{r},t)=\sum_{d'}C_{dd'}P_{d'}^-(\mathbf{r},t). \tag*{(75)}

The one-way probability flux from r\mathbf{r} to r+ℓed\mathbf{r}+\ell\mathbf{e}_d is

jd(r+ℓ2ed,t)=νPd+(r,t).(76)j_d\left(\mathbf{r}+\frac{\ell}{2}\mathbf{e}_d,t\right)=\nu P_d^+(\mathbf{r},t). \tag*{(76)}

The argument locates this flux at the bond midpoint; the displacement remains the full ℓed\ell\mathbf{e}_d. Weighting the outgoing fluxes at a site by their displacements gives

jlat(r,t)=νℓ∑dedPd+(r,t)=νℓ ECP(r,t)=νℓC(2)m(r,t),(77)\mathbf{j}_{\mathrm{lat}}(\mathbf{r},t)=\nu\ell\sum_d\mathbf{e}_dP_d^+(\mathbf{r},t)=\nu\ell\,\mathrm{ECP}(\mathbf{r},t)=\nu\ell C_{(2)}\mathbf{m}(\mathbf{r},t), \tag*{(77)}

where Eq. (67) gives the final equality. This establishes Eq. (28) and distinguishes the current from the pre-coin polarization.

To express the bond fluxes at a common spatial point, their source populations are evaluated at r−ℓed/2\mathbf{r}-\ell\mathbf{e}_d/2. Expanding these populations to first order in spatial gradients leads to

jα(r,t)≈(jlat)α(r,t)−νℓ22∇βMαβout(r,t),(78)j_\alpha(\mathbf{r},t)\approx(j_{\mathrm{lat}})_\alpha(\mathbf{r},t)-\frac{\nu\ell^2}{2}\nabla_\beta M_{\alpha\beta}^{\mathrm{out}}(\mathbf{r},t), \tag*{(78)}

with the outgoing second moment defined in Eq. (71). In the second-order bulk closure, its leading part is Mαβout≈ρ δαβ/2M_{\alpha\beta}^{\mathrm{out}}\approx\rho\,\delta_{\alpha\beta}/2, so

j(r,t)≈jlat(r,t)−D0∇ρ(r,t),D0=νℓ24.(79)\mathbf{j}(\mathbf{r},t)\approx\mathbf{j}_{\mathrm{lat}}(\mathbf{r},t)-D_0\nabla\rho(\mathbf{r},t),\qquad D_0=\frac{\nu\ell^2}{4}. \tag*{(79)}

The omitted nematic and higher-gradient contributions are beyond the retained bulk order. Substituting the quasi-stationary polarization, Eq. (23), recovers Eq. (25) with the coefficients in Eq. (26). Thus the odd coefficient belongs to a specified microscopic current, even though its homogeneous contribution disappears from the scalar density equation.

Boundary scattering rules

In the homogeneous bulk, one update is generated by

Ubulk=S(Ir⊗C),(80)U_{\mathrm{bulk}} = S(I_{\mathrm{r}} \otimes C), \tag*{(80)}

where CC updates the orientation and SS then shifts the walker in that direction, as in Eq. (5). Confinement requires an additional prescription for attempted displacements that cross a wall. We adopt specular reflection as the reference boundary rule as it is a standard idealized reflecting condition used in active-particle models and can be applied consistently to both the aCRW and cABP [88, 89]. Randomized scattering is considered in Figure 6 to illustrate the dependence on this microscopic choice.

Figure 6. Boundary-dependent density evolution for qp=0.6q_p = 0.6 and qc=0.2q_c = 0.2 on a 41×4141 \times41 lattice, starting from a Gaussian distribution restricted to the central 5×55 \times5 sites with uniform orientations. Each square combines the upper half-domain at t=150t = 150 and the lower half-domain at t=5000t = 5000; the dashed line separates the two times. Colours show the density excess ρ/ρˉ−1\rho/\bar{\rho} - 1, with separate scales for each time. Specular reflection produces a uniform stationary density, whereas randomized scattering produces wall accumulation. The lower panel shows the final projected density excess E(x)=⟨ρ⟩y/ρˉ−1E(x) = \langle\rho\rangle_y/\bar{\rho} - 1. Densities are obtained by deterministic iteration of the lattice master equation, with ℓ=Δt=1\ell= \Delta t = 1.

At each update, the bulk coin first selects a proposed outgoing orientation dd. If the target site lies inside the domain, the displacement is accepted. Otherwise, the displacement is rejected, the position remains unchanged, and the wall produces a new orientation d′d'. Time advances by one step in either case. The scattering is specified by the conditional probability

Kwall(d′∣d),d′∈D.(81)K_{\mathrm{wall}}(d' \mid d), \qquad d' \in D. \tag*{(81)}

For specular reflection at a wall with outward unit normal n\mathbf{n}, the normal component of the heading is reversed while its tangential component is preserved. In continuous space this is

ed′=ed−2(ed⋅n)n.(82)\mathbf{e}_{d'} = \mathbf{e}_{d} - 2(\mathbf{e}_{d} \cdot\mathbf{n})\mathbf{n}. \tag*{(82)}

With four cardinal headings and axis-aligned walls, a rejected displacement is normal to the wall. Specular reflection therefore reduces to reversal of the orientation, Kwallspec(d′∣d)=δd′,BdK_{\mathrm{wall}}^{\mathrm{spec}}(d' \mid d) = \delta_{d',Bd}. This is the boundary rule used for the confined calculations in the main text.

An alternative is randomized scattering,

Kwallrand(d′∣d)=14(83)K_{\mathrm{wall}}^{\mathrm{rand}}(d' \mid d) = \frac{1}{4} \tag*{(83)}

which redistributes the rejected probability equally among all four headings, including those pointing outside. Subsequent attempted displacements remain subject to the same rejection rule.

Figure 6 compares the two prescriptions at identical bulk parameters. For the parameters shown, specular reflection yields a uniform stationary density, whereas randomized scattering produces enhanced wall occupation. The boundary dynamics therefore provide an additional microscopic ingredient: wall accumulation depends on the scattering prescription and is not determined by the bulk odd-diffusion coefficient alone [88, 89].

Derivation of the cABP field equation

For completeness, we derive the cABP density and polarization equations used in § 5. In continuous two-dimensional space, the position r(t)\mathbf{r}(t) and orientation angle ϕ(t)\phi(t) obey

dr=v~0u(ϕ) dt+2D~ dB,dϕ=ω~ dt+2D~r dW,(84)\mathrm{d}\mathbf{r} = \tilde{v}_0\mathbf{u}(\phi)\,\mathrm{d}t + \sqrt{2\tilde{D}}\,\mathrm{d}\mathbf{B}, \qquad\mathrm{d}\phi= \tilde{\omega}\,\mathrm{d}t + \sqrt{2\tilde{D}_{r}}\,\mathrm{d}W, \tag*{(84)}

where u(ϕ)=(cos⁡ϕ,sin⁡ϕ)\mathbf{u}(\phi) = (\cos\phi,\sin\phi), dB\mathrm{d}\mathbf{B} and dW\mathrm{d}W are independent Wiener increments with unit variance. The corresponding probability density p~(r,ϕ,t)\tilde{p}(\mathbf{r},\phi,t) satisfies

∂p~(r,ϕ,t)∂t=D~∇2p~−v~0u⋅∇p~+D~r∂2p~∂ϕ2−ω~∂p~∂ϕ.(85)\frac{\partial\tilde{p}(\mathbf{r},\phi,t)}{\partial t} = \tilde{D}\nabla^{2}\tilde{p} - \tilde{v}_0\mathbf{u}\cdot\nabla\tilde{p} + \tilde{D}_{r}\frac{\partial^{2}\tilde{p}}{\partial\phi^{2}} - \tilde{\omega}\frac{\partial\tilde{p}}{\partial\phi}. \tag*{(85)}

This is Eq. (44), written here to make the marginalization self-contained. We use the counter-clockwise-positive convention, so ω~>0\tilde{\omega} > 0 increases ϕ\phi. The angular identities required below are

∂u∂ϕ=ϵu,∂2u∂ϕ2=−u.(86)\frac{\partial\mathbf{u}}{\partial\phi} = \boldsymbol{\epsilon}\mathbf{u}, \qquad\frac{\partial^{2}\mathbf{u}}{\partial\phi^{2}} = -\mathbf{u}. \tag*{(86)}

Angular marginalization

The first two angular moments are

ρ~(r,t)=∫02πp~(r,ϕ,t) dϕ,m~(r,t)=∫02πu(ϕ)p~(r,ϕ,t) dϕ.(87)\tilde{\rho}(\mathbf{r},t) = \int_{0}^{2\pi} \tilde{p}(\mathbf{r},\phi,t)\,\mathrm{d}\phi,\qquad\tilde{\mathbf{m}}(\mathbf{r},t) = \int_{0}^{2\pi} \mathbf{u}(\phi)\tilde{p}(\mathbf{r},\phi,t)\,\mathrm{d}\phi. \tag*{(87)}

Integrating Eq. (85) over ϕ\phi, periodicity eliminates both angular-derivative terms and ∫02πu⋅∇p~ dϕ=∇⋅m~\int_{0}^{2\pi} \mathbf{u}\cdot\nabla\tilde{p}\,\mathrm{d}\phi= \nabla\cdot\tilde{\mathbf{m}}. Hence

∂ρ~(r,t)∂t=D~∇2ρ~(r,t)−v~0∇⋅m~(r,t),(88)\frac{\partial\tilde{\rho}(\mathbf{r},t)}{\partial t} = \tilde{D}\nabla^{2}\tilde{\rho}(\mathbf{r},t) - \tilde{v}_0\nabla\cdot\tilde{\mathbf{m}}(\mathbf{r},t), \tag*{(88)}

which is Eq. (47).

To obtain the first-moment equation, we multiply Eq. (85) by u(ϕ)\mathbf{u}(\phi) and integrate over ϕ\phi. Integration by parts and Eq. (86) give

∫02πu∂p~∂ϕ dϕ=−ϵm~,∫02πu∂2p~∂ϕ2 dϕ=−m~.(89)\int_{0}^{2\pi}\mathbf{u}\frac{\partial\tilde{p}}{\partial\phi}\,\mathrm{d}\phi= -\boldsymbol{\epsilon}\tilde{\mathbf{m}},\qquad\int_{0}^{2\pi}\mathbf{u}\frac{\partial^{2}\tilde{p}}{\partial\phi^{2}}\,\mathrm{d}\phi= -\tilde{\mathbf{m}}. \tag*{(89)}

The propulsion term contains the second angular moment

Mαβ(r,t)=∫02πuαuβp~(r,ϕ,t) dϕ=ρ~(r,t)2δαβ+N~αβ(r,t),(90)M_{\alpha\beta}(\mathbf{r},t) = \int_{0}^{2\pi}u_{\alpha}u_{\beta}\tilde{p}(\mathbf{r},\phi,t)\,\mathrm{d}\phi= \frac{\tilde{\rho}(\mathbf{r},t)}{2}\delta_{\alpha\beta} + \tilde{N}_{\alpha\beta}(\mathbf{r},t), \tag*{(90)}

where Greek indices denote Cartesian components and N~\tilde{\mathbf{N}} is the symmetric traceless nematic tensor. The exact first-moment equation is therefore

∂m~(r,t)∂t=D~∇2m~(r,t)−D~rm~(r,t)+ω~ϵm~(r,t)−v~02∇ρ~(r,t)−v~0∇⋅N~(r,t).(91)\frac{\partial\tilde{\mathbf{m}}(\mathbf{r},t)}{\partial t} = \tilde{D}\nabla^{2}\tilde{\mathbf{m}}(\mathbf{r},t) - \tilde{D}_r\tilde{\mathbf{m}}(\mathbf{r},t) + \tilde{\omega}\boldsymbol{\epsilon}\tilde{\mathbf{m}}(\mathbf{r},t) - \frac{\tilde{v}_0}{2}\nabla\tilde{\rho}(\mathbf{r},t) - \tilde{v}_0\nabla\cdot\tilde{\mathbf{N}}(\mathbf{r},t). \tag*{(91)}

Here (∇⋅N~)α=∇βN~αβ(\nabla\cdot\widetilde{\mathbf{N}})_\alpha= \nabla_\beta\widetilde{N}_{\alpha\beta}. The first-harmonic, long-wavelength closure sets N~≈0\widetilde{\mathbf{N}} \approx0, reducing Eq. (91) to Eq. (48). The sign of the angular-drift term is now explicit: positive ω~\widetilde{\omega} rotates the polarization counter-clockwise through +ω~ϵm~+\widetilde{\omega}\boldsymbol{\epsilon}\widetilde{\mathbf{m}}.

Quasi-stationary closure

In the slow-density, fast-polarization regime, neglecting ∂tm~\partial_t\widetilde{\mathbf{m}} and D~∇2m~\widetilde{D}\nabla^2\widetilde{\mathbf{m}} in the closed first-harmonic equation gives

(D~r1−ω~ϵ)m~(r,t)≈−v~02∇ρ~(r,t),m~(r,t)≈−v~02D~r1+ω~ϵD~r2+ω~2∇ρ~(r,t).(92)\begin{aligned} (\widetilde{D}_r\mathbf{1}-\widetilde{\omega}\boldsymbol{\epsilon})\widetilde{\mathbf{m}}(\mathbf{r},t) &\approx-\frac{\widetilde{v}_0}{2}\nabla\widetilde{\rho}(\mathbf{r},t),\\ \widetilde{\mathbf{m}}(\mathbf{r},t) &\approx-\frac{\widetilde{v}_0}{2}\frac{\widetilde{D}_r\mathbf{1}+\widetilde{\omega}\boldsymbol{\epsilon}}{\widetilde{D}_r^2+\widetilde{\omega}^2}\nabla\widetilde{\rho}(\mathbf{r},t). \tag*{(92)} \end{aligned}

The inverse follows from ϵ2=−1\boldsymbol{\epsilon}^2=-1:

(D~r1−ω~ϵ)−1=D~r1+ω~ϵD~r2+ω~2.(93)(\widetilde{D}_r\mathbf{1}-\widetilde{\omega}\boldsymbol{\epsilon})^{-1}=\frac{\widetilde{D}_r\mathbf{1}+\widetilde{\omega}\boldsymbol{\epsilon}}{\widetilde{D}_r^2+\widetilde{\omega}^2}. \tag*{(93)}

Substitution into j~(r,t)=−D~∇ρ~(r,t)+v~0m~(r,t)\widetilde{\mathbf{j}}(\mathbf{r},t)=-\widetilde{D}\nabla\widetilde{\rho}(\mathbf{r},t)+\widetilde{v}_0\widetilde{\mathbf{m}}(\mathbf{r},t) gives the even and odd coefficients in Eq. (49). The term proportional to D~r\widetilde{D}_r is antiparallel to the density gradient, whereas the term proportional to ω~\widetilde{\omega} is rotated counter-clockwise by π/2\pi/2.

Finite-step corrections to the cABP comparison

We detail the rate and transport corrections discussed in § 5. Throughout, D~=0\widetilde{D}=0, v~0=ℓ/Δt\widetilde{v}_0=\ell/\Delta t, and the correlations refer to an isotropic, homogeneous bulk.

Expansion of the correlation map

Without spatial gradients, the exact polarization update is m(t+Δt)=C(2)m(t)\mathbf{m}(t+\Delta t)=C_{(2)}\mathbf{m}(t). Replacing the left-hand side by m+Δt ∂tm\mathbf{m}+\Delta t\,\partial_t\mathbf{m} gives the leading rates in Eq. (51). The exact correlation rates instead follow from λ=exp⁡[(−D~rcorr+iω~corr)Δt]\lambda=\exp[(-\widetilde{D}_r^{\mathrm{corr}}+i\widetilde{\omega}^{\mathrm{corr}})\Delta t].

Using λ=1+(−D~r+iω~)Δt\lambda=1+(-\widetilde{D}_r+i\widetilde{\omega})\Delta t and expanding its logarithm gives

log⁡λΔt=−D~r+iω~−Δt2(−D~r+iω~)2+O(Δt2).(94)\frac{\log\lambda}{\Delta t}=-\widetilde{D}_r+i\widetilde{\omega}-\frac{\Delta t}{2}(-\widetilde{D}_r+i\widetilde{\omega})^2+O(\Delta t^2). \tag*{(94)}

Separating real and imaginary parts yields

D~rcorr=D~r+Δt2(D~r2−ω~2)+O(Δt2),(95)\widetilde{D}_r^{\mathrm{corr}}=\widetilde{D}_r+\frac{\Delta t}{2}\left(\widetilde{D}_r^2-\widetilde{\omega}^2\right)+O(\Delta t^2), \tag*{(95)}

and

ω~corr=ω~+Δt D~rω~+O(Δt2).(96)\widetilde{\omega}^{\mathrm{corr}}=\widetilde{\omega}+\Delta t\,\widetilde{D}_r\widetilde{\omega}+O(\Delta t^2). \tag*{(96)}

These expansions hold as Δt→0\Delta t\to0 with bounded leading rates. They quantify the temporal corrections omitted by the continuum map, without changing the exact polar projection. We choose the principal phase arg⁡λ∈(−π,π]\arg\lambda\in(-\pi,\pi], continuous near λ=1\lambda=1; frequencies differing by 2π/Δt2\pi/\Delta t cannot be distinguished at the update times. The case λ=0\lambda=0 has no finite correlation-matched decay rate.

Sums and integrals of correlations

Under correlation matching, the complex combination Cνdiag(n)+iCνoff(n)=λnC_\nu^{\mathrm{diag}}(n)+iC_\nu^{\mathrm{off}}(n)=\lambda^n equals the cABP exponential at every update time. For 0<∣λ∣<10<|\lambda|<1, the corresponding discrete sum and continuous integral are

Δt[12+∑n=1∞λn]=Δt21+λ1−λ.(97)\Delta t\left[\frac{1}{2}+\sum_{n=1}^{\infty}\lambda^n\right]=\frac{\Delta t}{2}\frac{1+\lambda}{1-\lambda}. \tag*{(97)}

and

∫0∞exp⁡[(−D~rcorr+iω~corr)t] dt=1D~rcorr−iω~corr.(98)\int_{0}^{\infty}\exp[(-\widetilde{D}_{r}^{\mathrm{corr}}+i\widetilde{\omega}^{\mathrm{corr}})t]\,\mathrm{d}t=\frac{1}{\widetilde{D}_{r}^{\mathrm{corr}}-i\widetilde{\omega}^{\mathrm{corr}}}. \tag*{(98)}

Multiplying their real and imaginary parts by v~02/2\widetilde{v}_{0}^{2}/2 gives the respective even and odd coefficients. The half weight at zero lag is the normalization in Eq. (37).

Expanding the sum at fixed correlation-matched rates yields

Δt[12+∑n=1∞λn]=1D~rcorr−iω~corr+Δt212(D~rcorr−iω~corr)+O(Δt4).(99)\Delta t\left[\frac{1}{2}+\sum_{n=1}^{\infty}\lambda^{n}\right]=\frac{1}{\widetilde{D}_{r}^{\mathrm{corr}}-i\widetilde{\omega}^{\mathrm{corr}}}+\frac{\Delta t^{2}}{12}\left(\widetilde{D}_{r}^{\mathrm{corr}}-i\widetilde{\omega}^{\mathrm{corr}}\right)+O(\Delta t^{4}). \tag*{(99)}

Thus, with the cABP coefficients evaluated using the correlation map,

Deff−D~eff=v~02D~rcorrΔt224+O(Δt4),(100)D_{\mathrm{eff}}-\widetilde{D}_{\mathrm{eff}}=\frac{\widetilde{v}_{0}^{2}\widetilde{D}_{r}^{\mathrm{corr}}\Delta t^{2}}{24}+O(\Delta t^{4}), \tag*{(100)}

and

Dodd−D~odd=−v~02ω~corrΔt224+O(Δt4).(101)D_{\mathrm{odd}}-\widetilde{D}_{\mathrm{odd}}=-\frac{\widetilde{v}_{0}^{2}\widetilde{\omega}^{\mathrm{corr}}\Delta t^{2}}{24}+O(\Delta t^{4}). \tag*{(101)}

This is the correction between the discrete, half-weighted sum and the continuous integral. The expansion requires Δt(D~rcorr)2+(ω~corr)2≪1\Delta t\sqrt{(\widetilde{D}_{r}^{\mathrm{corr}})^{2}+(\widetilde{\omega}^{\mathrm{corr}})^{2}}\ll1; Eqs. (97) and (98) remain exact outside this regime.

For the leading map, direct substitution into Eq. (49) instead gives

D~eff−Deff=ℓ24Δt2(1−qp)−(1−qp2−qc2)(1−qp)2+qc2=D0.(102)\widetilde{D}_{\mathrm{eff}}-D_{\mathrm{eff}}=\frac{\ell^{2}}{4\Delta t}\frac{2(1-q_{p})-(1-q_{p}^{2}-q_{c}^{2})}{(1-q_{p})^{2}+q_{c}^{2}}=D_{0}. \tag*{(102)}

The odd coefficients coincide by the same substitution, proving Eq. (56). These finite-step differences between models do not alter the equality of the lattice hydrodynamic and discrete Green–Kubo coefficients.

References

  1. [1]Pearson K. 1905a The problem of the random walk. Nature 72, 294. (10.1038/072294b0)
  2. [2]Pearson K. 1905b The problem of the random walk. Nature 72, 342. (10.1038/072342a0)
  3. [3]Ross R. 1905a The possibility of reducing mosquitoes. Nature 72, 151. (10.1038/072151a0)DOI
  4. [4]Ross R. 1905b The logical basis of the sanitary policy of mosquito reduction. Science 22, 689–699. (10.1126/science.22.570.689)pubmed.ncbi.nlm.nih.gov/17729473
  5. [5]Lord Rayleigh. 1905 The problem of the random walk. Nature 72, 318. (10.1038/072318a0)
  6. [6]Einstein A. 1905 Über die von der molekularkinetischen Theorie der Wärme geforderte Bewegung von in ruhenden Flüssigkeiten suspendierten Teilchen. Ann. Phys. 322, 549–560. (10.1002/andp.19053220806)DOI
  7. [7]von Smoluchowski M. 1906 Zur kinetischen Theorie der Brownschen Molekularbewegung und der Suspensionen. Ann. Phys. 326, 756–780. (10.1002/andp.19063261405)
  8. [8]von Smoluchowski M. 1916 Drei Vorträge über Diffusion, Brownsche Molekularbewegung und Koagulation von Kolloidteilchen. Phys. Z. 17, 557–571, 585–599.
  9. [9]Gudowska-Nowak E, Lindenberg K, Metzler R. 2017 Preface: Marian Smoluchowski’s 1916 paper—a century of inspiration. J. Phys. A: Math. Theor. 50, 380301. (10.1088/1751-8121/aa8529)
  10. [10]Alexander S, Orbach R. 1982 Density of states on fractals: “fractons”. J. Phys. Lett. 43, 625–631. (10.1051/jphyslet:019820043017062500)DOI
  11. [11]Gefen Y, Aharony A, Alexander S. 1983 Anomalous diffusion on percolating clusters. Phys. Rev. Lett. 50, 77–80. (10.1103/PhysRevLett.50.77)DOI
  12. [12]Kesten H. 1986 Subdiffusive behavior of random walk on a random cluster. Ann. Inst. Henri Poincaré Probab. Stat. 22, 425–487.
  13. [13]Stauffer D, Aharony A. 2018 Introduction to percolation theory. London: Taylor & Francis 2nd edition. (10.1201/9781315274386)
  14. [14]Barlow MT. 2004 Random walks on supercritical percolation clusters. Ann. Probab. 32, 3024–3084. (10.1214/009117904000000748)DOI
  15. [15]Lauber Bonomo O, Shitrit I, Reuveni S, Redner S. 2026 Pushy random walk: A minimal model for transport in deformable media. Phys. Rev. Lett. 137, 037101. (10.1103/7hjs-rx8d)DOI
  16. [16]Fisher DS. 1984 Random walks in random environments. Phys. Rev. A 30, 960–964. (10.1103/PhysRevA.30.960)DOI
  17. [17]Madras N, Slade G. 1996 The self-avoiding walk. Boston, MA: Birkhäuser. (10.1007/978-1-4612-4132-4)DOI
  18. [18]Henry BI, Langlands TAM, Wearne SL. 2006 Anomalous diffusion with linear reaction dynamics: From continuous time random walks to fractional reaction-diffusion equations. Phys. Rev. E 74, 031116. (10.1103/PhysRevE.74.031116)pubmed.ncbi.nlm.nih.gov/17025603
  19. [19]Seki K, Wojcik M, Tachiya M. 2003 Fractional reaction-diffusion equation. J. Chem. Phys. 119, 2165–2170. (10.1063/1.1587126)DOI
  20. [20]Perez LJ, Hidalgo JJ, Dentz M. 2019 Reactive random walk particle tracking and its equivalence with the advection-diffusion-reaction equation. Water Resour. Res. 55, 847–855. (10.1029/2018WR023560)
  21. [21]Sleeman BD, Levine HA. 1997 A system of reaction diffusion equations arising in the theory of reinforced random walks. SIAM J. Appl. Math. 57, 683–730. (10.1137/S0036139995291106)
  22. [22]Fama EF. 1995 Random walks in stock market prices. Financ. Anal. J. 51, 75–80. (10.2469/faj.v51.n1.1861)
  23. [23]Oshanin G, Lindenberg K, Wio HS, Burlatsky S. 2009 Efficient search by optimized intermittent random walks. J. Phys. A: Math. Theor. 42, 434008. (10.1088/1751-8113/42/43/434008)
  24. [24]Serfozo MR. 1981 Optimal control of random walks, birth and death processes, and queues. Adv. Appl. Probab. 13, 61–83. (10.2307/1426467)
  25. [25]Oshanin G, Wio HS, Lindenberg K, Burlatsky SF. 2007 Intermittent random walks for an optimal search strategy: One-dimensional case. J. Phys.: Condens. Matter 19, 065142. (10.1088/0953-8984/19/6/065142)
  26. [26]Berg HC. 1983 Random walks in biology. Princeton, NJ: Princeton University Press.
  27. [27]Grebenkov D, Metzler R, Oshanin G, editors. 2024 Target search problems. Cham: Springer. (10.1007/978-3-031-67802-8)
  28. [28]te Vrugt M, Liebchen B, Cates ME. 2026 Colloquium: What do we mean by ‘active matter’? Rev. Mod. Phys. 98, 031001. (10.1103/wd4f-q7kv)
  29. [29]de Gennes P-G. 1992 Soft Matter (Nobel Lecture). Angew. Chem. Int. Ed. Engl. 31, 842–845. (10.1002/anie.199208421)DOI
  30. [30]Romanczuk P, Bär M, Ebeling W, Lindner B, Schimansky-Geier L. 2012 Active Brownian particles: From individual to collective stochastic dynamics. Eur. Phys. J. Spec. Top. 202, 1–162. (10.1140/epjst/e2012-01529-y)
  31. [31]Prawar Dadhichi L, Maitra A, Ramaswamy S. 2018 Origins and diagnostics of the nonequilibrium character of active systems. J. Stat. Mech.: Theory Exp. 2018, 123201. (10.1088/1742-5468/aae852)
  32. [32]Astumian RD, Hänggi P. 2002 Brownian motors. Phys. Today 55, 33–39. (10.1063/1.1535005)DOI
  33. [33]Reimann P, Hänggi P. 2002 Introduction to the physics of Brownian motors. Appl. Phys. A 75, 169–178. (10.1007/s003390201331)
  34. [34]Fletcher DA, Theriot JA. 2004 An introduction to cell motility for the physical scientist. Phys. Biol. 1, T1–T10. (10.1088/1478-3975/1/1/T01)
  35. [35]Koch DL, Subramanian G. 2011 Collective hydrodynamics of swimming microorganisms: Living fluids. Annu. Rev. Fluid Mech. 43, 637–659. (10.1146/annurev-fluid-121108-145434)
  36. [36]Bechinger C, Di Leonardo R, Löwen H, Reichhardt C, Volpe G, Volpe G. 2016 Active particles in complex and crowded environments. Rev. Mod. Phys. 88, 045006. (10.1103/RevModPhys.88.045006)
  37. [37]Walther A, Müller AHE. 2008 Janus particles. Soft Matter 4, 663–668. (10.1039/b718131k)DOI
  38. [38]Gomez-Solano JR, Blokhuis A, Bechinger C. 2016 Dynamics of self-propelled Janus particles in viscoelastic fluids. Phys. Rev. Lett. 116, 138301. (10.1103/PhysRevLett.116.138301)
  39. [39]O’Byrne J, Kafri Y, Tailleur J, van Wijland F. 2022 Time irreversibility in active matter, from micro to macro. Nat. Rev. Phys. 4, 167–183. (10.1038/s42254-021-00046-2)
  40. [40]Vilk O, Aghion E, Avgar T, Beta C, Nagel O, Sabri A, Sarfati R, Schwartz DK, Weiss M, Krapf D, Nathan R, Metzler R, Assaf M. 2022 Unravelling the origins of anomalous diffusion: From molecules to migrating storks. Phys. Rev. Res. 4, 033055. (10.1103/PhysRevResearch.4.033055)
  41. [41]Löwen H. 2016 Chirality in microswimmer motion: From circle swimmers to active turbulence. Eur. Phys. J. Spec. Top. 225, 2319–2331. (10.1140/epjst/e2016-60054-6)DOI
  42. [42]Liebchen B, Levis D. 2022 Chiral active matter. Europhys. Lett. 139, 67001. (10.1209/0295-5075/ac8f69)DOI
  43. [43]Frymier PD, Ford RM, Berg HC, Cummings PT. 1995 Three-dimensional tracking of motile bacteria near a solid planar surface. Proc. Natl. Acad. Sci. USA 92, 6195–6199. (10.1073/pnas.92.13.6195)pubmed.ncbi.nlm.nih.gov/7597100
  44. [44]Lauga E, DiLuzio WR, Whitesides GM, Stone HA. 2006 Swimming in circles: Motion of bacteria near solid boundaries. Biophys. J. 90, 400–412. (10.1529/biophysj.105.069401)DOI
  45. [45]Petroff AP, Wu X-L, Libchaber A. 2015 Fast-moving bacteria self-organize into active two-dimensional crystals of rotating cells. Phys. Rev. Lett. 114, 158102. (10.1103/PhysRevLett.114.158102)pubmed.ncbi.nlm.nih.gov/25933342
  46. [46]Li H, Chaté H, Sano M, Shi X-Q, Zhang HP. 2024 Robust edge flows in swarming bacterial colonies. Phys. Rev. X 14, 041006. (10.1103/PhysRevX.14.041006)
  47. [47]Tan TH, Mietke A, Li J, Chen Y, Higinbotham H, Foster PJ, Gokhale S, Dunkel J, Fakhri N. 2022 Odd dynamics of living chiral crystals. Nature 607, 287–293. (10.1038/s41586-022-04889-6)DOI
  48. [48]Kümmel F, ten Hagen B, Wittkowski R, Buttinoni I, Eichhorn R, Volpe G, Löwen H, Bechinger C. 2013 Circular motion of asymmetric self-propelling particles. Phys. Rev. Lett. 110, 198302. (10.1103/PhysRevLett.110.198302)
  49. [49]Soni V, Bilišing ES, Magkiriadou S, Sacanna S, Bartolo D, Shelley MJ, Irvine WTM. 2019 The odd free surface flows of a colloidal chiral fluid. Nat. Phys. 15, 1188–1194. (10.1038/s41567-019-0603-8)
  50. [50]Massana-Cid H, Levis D, Hernández Hernández RJ, Pagonabarraga I, Tierno P. 2021 Arrested phase separation in chiral fluids of colloidal spinners. Phys. Rev. Res. 3, L042021. (10.1103/PhysRevResearch.3.L042021)
  51. [51]Cao X, Das D, Windbacher N, Ginot F, Krüger M, Bechinger C. 2023 Memory-induced Magnus effect. Nat. Phys. 19, 1904–1909. (10.1038/s41567-023-02213-1)arxiv.org/abs/2303.07416
  52. [52]Schimansky-Geier L, Mieth M, Rosé H, Malchow H. 1995 Structure formation by active Brownian particles. Phys. Lett. A 207, 140–146. (10.1016/0375-9601(95)00700-D)
  53. [53]Bonilla LL. 2019 Active Ornstein-Uhlenbeck particles. Phys. Rev. E 100, 022601. (10.1103/PhysRevE.100.022601)DOI
  54. [54]Pattanayak A, Shee A, Chaudhuri D, Chaudhuri A. 2024 Impact of torque on active Brownian particle: exact moments in two and three dimensions. New J. Phys. 26, 083024. (10.1088/1367-2630/ad6a32)DOI
  55. [55]Caprini L, Petrini A, Marini Bettolo Marconi U. 2026 Modeling chiral active particles: from circular motion to odd interactions. J. Stat. Mech.: Theory Exp. 2026, 024001. (10.1088/1742-5468/ae3d28)arxiv.org/abs/2510.15711
  56. [56]Taylor GI. 1922 Diffusion by continuous movements. Proc. Lond. Math. Soc. s2-20, 196–212. (10.1112/plms/s2-20.1.196)DOI
  57. [57]Fürth R. 1920 Die Brownsche Bewegung bei Berücksichtigung einer Persistenz der Bewegungsrichtung. Mit Anwendungen auf die Bewegung lebender Infusorien. Z. Phys. 2, 244–256. (10.1007/BF01328731)
  58. [58]Montroll EW. 1950 Markoff chains and excluded volume effect in polymer chains. J. Chem. Phys. 18, 734–743. (10.1063/1.1747735)DOI
  59. [59]Gilbert T, Sanders DP. 2010 Diffusion coefficients for multi-step persistent random walks on lattices. J. Phys. A: Math. Theor. 43, 035001. (10.1088/1751-8113/43/3/035001)DOI
  60. [60]Gilbert T, Sanders DP. 2012 Diffusion coefficients for periodically induced multi-step persistent walks on regular lattices. J. Phys. A: Math. Theor. 45, 255003. (10.1088/1751-8113/45/25/255003)DOI
  61. [61]Gilbert T, Nguyen HC, Sanders DP. 2011 Diffusive properties of persistent walks on cubic lattices with application to periodic Lorentz gases. J. Phys. A: Math. Theor. 44, 065001. (10.1088/1751-8113/44/6/065001)DOI
  62. [62]Gilbert T, Sanders DP. 2009 Persistence effects in deterministic diffusion. Phys. Rev. E 80, 041121. (10.1103/PhysRevE.80.041121)DOI
  63. [63]Bender EA, Richmond LB. 1984 Correlated random walks. Ann. Probab. 12, 274–278. (10.1214/aop/1176993392)DOI
  64. [64]Renshaw E, Henderson R. 1981 The correlated random walk. J. Appl. Probab. 18, 403–414. (10.2307/3213286)DOI
  65. [65]Chen A, Renshaw E. 1994 The general correlated random walk. J. Appl. Probab. 31, 869–884. (10.2307/3215313)DOI
  66. [66]Larralde H. 1997 Transport properties of a two-dimensional “chiral” persistent random walk. Phys. Rev. E 56, 5004–5008. (10.1103/PhysRevE.56.5004)DOI
  67. [67]Sadjadi Z, Shaebani MR, Rieger H, Santen L. 2015 Persistent-random-walk approach to anomalous transport of self-propelled particles. Phys. Rev. E 91, 062715. (10.1103/PhysRevE.91.062715)DOI
  68. [68]Jose S, Mandal D, Barma M, Ramola K. 2022 Active random walks in one and two dimensions. Phys. Rev. E 105, 064103. (10.1103/PhysRevE.105.064103)DOI
  69. [69]Osat S, Meyberg E, Metson J, Speck T. 2026 Topological chiral random walker. arXiv:2602.12020. (10.48550/arXiv.2602.12020)arxiv.org/abs/2602.12020
  70. [70]Wójcik J, Kalz E. 2026 The chiral random walk: A quantum-inspired framework for odd diffusion. New J. Phys. 28, 114603. (10.1088/1367-2630/aeaf77)arxiv.org/abs/2602.09920
  71. [71]Aharonov Y, Davidovich L, Zagury N. 1993 Quantum random walks. Phys. Rev. A 48, 1687–1690. (10.1103/Phys- RevA.48.1687)pubmed.ncbi.nlm.nih.gov/9909772
  72. [72]Lemaitre E, Sokolov IM, Metzler R, Chechkin AV. 2023 Non-Gaussian displacement distributions in models of heterogeneous active particle dynamics. New J. Phys. 25, 013010. (10.1088/1367-2630/acb005)DOI
  73. [73]Pattanayak A, Roy S, Chaudhuri A. 2026 Inertia–chirality interplay in active Brownian motion: Exact dynamics and phase maps. New J. Phys. 28, 054402. (10.1088/1367-2630/ae5f40)DOI
  74. [74]Zhou Y, Ge M, Wang T. 2025 Experimental modeling of chiral active robots and a minimal model of non-Gaussian displacements. Phys. Rev. E 111, 015404. (10.1103/PhysRevE.111.015404)DOI
  75. [75]Lauber Bonomo O, Reuveni S. 2023 Loss of percolation transition in the presence of simple tracer-media interactions. Phys. Rev. Res. 5, L042015. (10.1103/PhysRevResearch.5.L042015)
  76. [76]Moore F, Russo J, Liverpoool TB, Royall CP. 2023 Active Brownian particles in random and porous environments. J. Chem. Phys. 158, 104907. (10.1063/5.0131340)DOI
  77. [77]Zeitz M, Wolff K, Stark H. 2017 Active Brownian particles moving in a random Lorentz gas. Eur. Phys. J. E 40, 23. (10.1140/epje/i2017-11510-0)
  78. [78]Iyaniwura SA, Qiu M, Peng Z. 2026 Mean first passage time of chiral active Brownian particles. Soft Matter 22, 4489–4504. (10.1039/d6sm00448b)
  79. [79]Baouche Y, Le Goff M, Kurztahler C, Franosch T. 2025 First-passage-time statistics of active Brownian particles: A perturbative approach. Phys. Rev. E 111, 054113. (10.1103/PhysRevE.111.054113)
  80. [80]Pal PS, Park J-M, Pal A, Park H, Lee JS. 2024 Active motion can be beneficial for target search with resetting in a thermal environment. Phys. Rev. E 110, 054124. (10.1103/PhysRevE.110.054124)DOI
  81. [81]Uhlig J, Wójcik J, Metzler R, Kalz E. 2026 Chirality enhances diffusion in disordered environments. J. Chem. Phys. 165, 124118. (10.1063/5.0352851)DOI
  82. [82]Cates ME, Tailleur J. 2013 When are active Brownian particles and run-and-tumble particles equivalent? Con- sequences for motility-induced phase separation. Europhys. Lett. 101, 20010. (10.1209/0295-5075/101/20010)DOI
  83. [83]te Vrugt M, Wittkowski R. 2020 Relations between angular and Cartesian orientational expansions. AIP Adv. 10, 035106. (10.1063/1.5141367)DOI
  84. [84]Kalz E, Sharma A, Metzler R. 2024 Field theory of active chiral hard disks: A first-principles approach to steric interactions. J. Phys. A: Math. Theor. 57, 265002. (10.1088/1751-8121/ad5089)DOI
  85. [85]Frisch U, Hasslacher B, Pomeau Y. 1986 Lattice-gas automata for the Navier-Stokes equation. Phys. Rev. Lett. 56, 1505–1508. (10.1103/PhysRevLett.56.1505)pubmed.ncbi.nlm.nih.gov/10032689
  86. [86]Frisch U, d’Humieres D, Hasslacher B, Lallemand P, Pomeau Y, Rivet J-P. 1987 Lattice gas hydrodynamics in two and three dimensions. Complex Syst. 1, 649–707.
  87. [87]Bureković S, De Luca F, Cates ME, Nardini C. 2026 Active Cahn-Hilliard theory for nonequilibrium phase separation: Quantitative macroscopic predictions and a microscopic derivation. Phys. Rev. Res. 8, 033141. (10.1103/n155-s73b)DOI
  88. [88]Granek O, Kafri Y, Kardar M, Ro S, Tailleur J, Solon A. 2024 Colloquium: Inclusions, boundaries, and disorder in scalar active matter. Rev. Mod. Phys. 96, 031003. (10.1103/RevModPhys.96.031003)DOI
  89. [89]Escalante-Velázquez CA. 2026 Benchmarking wall models for confined active Brownian particles near planar and channel walls. Eur. Phys. J. E 49, 100. (10.1140/epje/s10189-026-00641-4)
  90. [90]Kalz E, Vuijk HD, Sommer J-JU, Metzler R, Sharma A. 2024 Oscillatory force autocorrelations in equilibrium odd-diffusive systems. Phys. Rev. Lett. 132, 057102. (10.1103/PhysRevLett.132.057102)DOI
  91. [91]Hargus C, Epstein JM, Mandadapu KK. 2021 Odd diffusivity of chiral random motion. Phys. Rev. Lett. 127, 178001. (10.1103/PhysRevLett.127.178001)DOI
  92. [92]Sherf N, Uhlig J, Wójcik J, Metzler R, Roichman Y, Kalz E. 2026 aCRW codes v1.2.0. Zenodo. (10.5281/zen- odo.24060170)
  93. [93]Risken H. 1996 The Fokker-Planck equation: Methods of solution and applications. Berlin: Springer 2nd edition. (10.1007/978-3-642-61544-3)
  94. [94]Othmer HG, Dunbar SR, Alt W. 1988 Models of dispersal in biological systems. J. Math. Biol. 26, 263–298. (10.1007/BF00277392)pubmed.ncbi.nlm.nih.gov/3411255
  95. [95]Tailleur J, Cates ME. 2008 Statistical mechanics of interacting run-and-tumble bacteria. Phys. Rev. Lett. 100, 218103. (10.1103/PhysRevLett.100.218103)DOI
  96. [96]Goldstein S. 1951 On diffusion by discontinuous movements, and on the telegraph equation. Q. J. Mech. Appl. Math. 4, 129–156. (10.1093/qjmam/4.2.129)DOI
  97. [97]Kiechl T, Altshuler A, Lüders A, Roichman Y, Franosch T. 2026 Free chiral self-propelled robots compared to active Brownian circle swimmers. Phys. Rev. E 113, 045409. (10.1103/274x-4vgn)DOI
  98. [98]Dauchot O, Démery V. 2019 Dynamics of a self-propelled particle in a harmonic trap. Phys. Rev. Lett. 122, 068002. (10.1103/PhysRevLett.122.068002)DOI
  99. [99]Khadem SMJ, Siboni NH, Klapp SHL. 2021 Transport and phase separation of active Brownian particles in fluctuating environments. Phys. Rev. E 104, 064615. (10.1103/PhysRevE.104.064615)DOI
  100. [100]Grebenkov DS, Metzler R, Oshanin G. 2026 Fastest first-passage time for multiple searchers with finite speed. Phys. Rev. Lett. 137, 127102. (10.1103/7q39-h9dv)
  101. [101]Lawley SD. 2026 Universality and ambiguity in extremes of anomalous diffusion. Phys. Rev. E 113, 064135. (10.1103/z398-p9lf)

Paper details

vlnc:W_4rp7hmut
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

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