betaFeedback

WorksPhysics and Astronomy

Odd Darcy Law: Odd Viscosity and Edge Currents in Porous Media

Abstract

We derive and numerically validate an odd Darcy law for chiral active fluids in porous media, in which an antisymmetric permeability tensor induces a flow transverse to the applied pressure gradient. We identify two distinct mechanisms leading to this odd response: odd viscosity in the fluid or the nonlinear coupling of edge currents to inertia. We then show how the transverse transport depends on porosity, boundary friction, and chirality.

Introduction

Rome’s aqueducts and lead-pipe networks attest to the sophistication of ancient hydraulic engineering, but a decisive step toward modern watersupply systems came much later in the less renowned French city of Dijon. In the 1850s, while designing the city’s water network, Henri Darcy investigated flow through sand filters and formulated the phenomenological relation now known as Darcy’s law [1]:

U=−1ηK⋅∇⟨p⟩,(1)U = -\frac{1}{\eta} K \cdot\nabla\langle p \rangle, \tag*{(1)}

which relates the volume-averaged (superficial) velocity UU to the macroscopic pressure gradient ∇⟨p⟩\nabla\langle p \rangle. Here, η\eta is the fluid viscosity, while the permeability tensor KK encodes the effects of pore geometry and connectivity on fluid transport. It may be determined empirically from bulk-flow measurements or predicted from pore-scale models using, for example, homogenization [2, 3], closures based on volume or ensemble averaging [4, 5], or approximate hydrodynamic cell models [6, 7]. Today, Darcy’s law underpins models of groundwater flow, oil and gas extraction, soil drainage, filtration, dissolution, and geothermal reservoirs [8, 9, 10, 11]. In biology, where crowded and porous environments are ubiquitous, it describes interstitial fluid transport in articular cartilage, tumors, the lymphatic interstitium, brain parenchyma, and deformable tissues [12, 13, 14, 15, 16].

Many biological media are also active: cells, molecular motors, and other internally driven components continuously generate stresses and flows [17, 18, 19, 20, 21, 22]. Extending Darcy-type descriptions to active media is therefore essential for understanding biological transport, especially because active matter responds strongly to obstacles via boundary accumulation, anchoring-dependent sorting, or active surface stresses [23, 24, 25, 26, 27, 28]. Recent progress includes studies of active self-propelled particles [29] or nematics in Hele-Shaw cells [30], corrugated channels [31] and porous media [32], the latter revealing a nonmonotonic dependence of permeability on activity. Much less is known about transport in chiral systems, ranging from circle-swimming particles and suspensions of spinning colloids to living assemblies of rotating cells [33, 34, 35, 36, 37, 38]. At the hydrodynamic level, chirality permits nondissipative antisymmetric transport coefficients [39, 40, 41, 42, 43], most notably odd viscosity [44, 45, 46, 47, 48, 49, 50]. Boundaries then play a central role [34, 51, 52, 53, 54, 55, 56, 57]: in 2D incompressible fluids, odd viscosity enters the linear response only through stress-dependent boundary conditions [58, 59], while torque densities generated by internal spin or transverse chiral forces drive edge currents [60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70].

Porous media, where fluids repeatedly interact with internal boundaries, provide a natural setting for converting odd stresses into macroscopic transport. Hall-like transport has been observed in active spinner fluids flowing through fixed obstacle arrays [71]. Related transverse responses have also been derived for parity-odd fluids confined in Hele-Shaw geometries [72]. These observations raise a broader question. In porous media, how do porosity, odd viscosity, and edge currents affect the transport in a chiral active fluid? We show that Darcy’s law remains valid, but with a permeability tensor that acquires an antisymmetric component. We derive this odd permeability quantitatively and identify two distinct origins of the transverse response: odd viscosity and edge currents at finite inertia.

Chiral hydrodynamics

Incompressible 2D chiral fluids are described on mesoscopic scales by [39, 73]:

ρ(∂tu+u⋅∇)u=∇⋅Σ,∇⋅u=0,(2)\rho(\partial_t \mathbf{u}+\mathbf{u}\cdot\nabla)\mathbf{u}=\nabla\cdot\boldsymbol{\Sigma}, \qquad\nabla\cdot\mathbf{u}=0, \tag*{(2)}

with ρ\rho the mass density and u(r,t)\mathbf{u}(\mathbf{r},t) the local velocity field driven by the stress:

Σij=−pδij+τεij+η(∂iuj+∂jui)+ηo(εjk∂iuk+εik∂kuj),(3)\Sigma_{ij}=-p\delta_{ij}+\tau\varepsilon_{ij}+\eta(\partial_i u_j+\partial_j u_i)+\eta_o(\varepsilon_{jk}\partial_i u_k+\varepsilon_{ik}\partial_k u_j), \tag*{(3)}

Here, repeated indices are summed, with εxy=−εyx=1\varepsilon_{xy}=-\varepsilon_{yx}=1 and εxx=εyy=0\varepsilon_{xx}=\varepsilon_{yy}=0. The pressure pp enforces incompressibility, while an imposed pressure difference across opposite boundaries drives a flow. η\eta and ηo\eta_o denote the even and odd viscosities, respectively, the latter requiring broken parity and time-reversal symmetries. The homogeneous torque density τ(ρ)\tau(\rho) contributes an antisymmetric stress characteristic of chiral systems, arising from effective transverse interparticle forces or the elimination of internal spin [39, 63]. As the transverse counterpart of pressure, it exerts tangential forces at boundaries that drive edge currents.

We model the porous medium as a collection of fixed obstacles, representing walls or large colloids in spinner suspensions, cells in biological tissues, or micropillars in microfluidic arrays. At their surfaces ∂B\partial B, we impose impermeability and Navier-slip boundary conditions.

u(r)⋅n^(r)=0,\mathbf{u}(\mathbf{r}) \cdot\hat{\mathbf{n}}(\mathbf{r}) = 0,
t^(r)⋅Σ(r)⋅n^(r)=ξu(r)⋅t^(r),r∈∂B.(4)\hat{\mathbf{t}}(\mathbf{r}) \cdot\boldsymbol{\Sigma}(\mathbf{r}) \cdot\hat{\mathbf{n}}(\mathbf{r}) = \xi\mathbf{u}(\mathbf{r}) \cdot\hat{\mathbf{t}}(\mathbf{r}), \qquad\mathbf{r} \in\partial B. \tag*{(4)}

where n^\hat{\mathbf{n}} points from the solid into the fluid, t^=−ε⋅n^\hat{\mathbf{t}} = -\boldsymbol{\varepsilon} \cdot\hat{\mathbf{n}} is the corresponding tangent, and ξ\xi is the wall-friction coefficient. The limits ξ→0\xi\to0 and ξ→∞\xi\to\infty recover perfect slip and no slip (u(r)⋅t^(r)=0\mathbf{u}(\mathbf{r}) \cdot\hat{\mathbf{t}}(\mathbf{r}) = 0), respectively.

For constant ηo\eta_o in 2D, incompressibility reduces the odd-viscous force density to

∇⋅Σodd=ηo∇ω,ω=εij∂iuj,(5)\nabla\cdot\boldsymbol{\Sigma}_{\mathrm{odd}} = \eta_o \nabla\omega, \qquad\omega= \varepsilon_{ij}\partial_i u_j, \tag*{(5)}

which can be absorbed into the effective pressure peff=p−ηoωp_{\mathrm{eff}} = p - \eta_o\omega. Odd viscosity therefore affects the flow only through the boundary conditions. It is invisible under no-slip, where the boundary velocity is fixed, but enters under partial slip through the tangential traction. Finite slip also permits the edge currents suppressed by no-slip boundaries, yielding richer and more realistic dynamics.

Odd viscosity dominated Stokes regime—We first consider steady Stokes flow, ∇⋅Σ=0\nabla\cdot\boldsymbol{\Sigma} = 0, without edge currents (τ=0\tau= 0), such that chirality enters exclusively through the odd viscosity. To model an extended porous medium without external boundary effects, we periodically repeat a unit cell containing obstacles. In Figure 1, we solve the resulting steady-state equations using a Taylor-Hood finite-element discretization (see End Matter), with the flow driven by a horizontal pressure drop and the porous medium represented by obstacles where the boundary conditions in Eq. (4) apply. Panel (a) shows an achiral fluid (ηo=0\eta_o = 0) driven through an isotropic medium by a pressure drop Δp\Delta p between the boundaries. The resulting mean velocity, U≃Ue^xU \simeq U\hat{\mathbf{e}}_x, is aligned with the forcing. In panel (b), the fluid remains passive, but anisotropy of the medium deflects the mean flow away from the forcing direction. Panel (c) exhibits a comparable transverse response, but now in an isotropic medium, where it is generated solely by odd viscosity in the fluid (ηo≠0\eta_o \ne0). Although their macroscopic signatures may appear similar, geometric anisotropy of the porous medium and odd viscosity of the fluid are fundamentally different mechanisms for producing transverse flow. The anisotropy-induced deflection depends on the orientation of the medium and vanishes when the pressure gradient is aligned with a principal axis. By contrast, the transverse response in the isotropic medium reflects the fluid’s chirality, here manifested through odd viscosity. As we show later, it can also arise from other chiral effects, such as torque-driven edge currents coupled to inertia. We first focus on the odd-viscous mechanism.

Passive and chiral fluids driven through porous media

Figure 1. Passive and chiral fluids driven along e^x\hat{\mathbf{e}}_x in porous media. Passive flow (ηo=0\eta_o = 0) in isotropic (a) and anisotropic (b) media. (c) Chiral flow (ηo/η=−1.5\eta_o/\eta= -1.5) in an isotropic medium. The insets show the mean flow direction. In all cases ξ=0\xi= 0.

Odd viscosity induced odd permeability—We consider a periodic medium containing nn nonoverlapping circular obstacles of radius aa in an area AA, with porosity φ=1−nπa2/A\varphi= 1 - n\pi a^2/A. The pressure is decomposed as

p(r)=p(r)+⟨∇p⟩⋅r.(6)p(\mathbf{r}) = p(\mathbf{r}) + \langle\nabla p \rangle\cdot\mathbf{r}. \tag*{(6)}

where p(r)p(\mathbf{r}) is the periodic part of pp, while ⟨∇p⟩⋅r\langle\nabla p \rangle\cdot\mathbf{r} represents its macroscopic affine component (e.g., ⟨∇p⟩=−(Δp/L)e^x\langle\nabla p \rangle= -(\Delta p/L)\hat{\mathbf{e}}_x for an inlet-to-outlet pressure drop Δp≡p(0)−p(L)\Delta p \equiv p(0) - p(L) across a length LL). This isolates the boundary-induced pressure drop in the affine term, allowing periodic boundary conditions on pp while representing the driving as a uniform body force.

To better understand how the transverse response emerges, we adapt a theory developed for a passive porous medium to our case [7, 74]. To this end, we adapt Kuwabara’s circular-cell approximation, replacing the complex porous medium by a single circular obstacle of radius aa surrounded by fluid up to an effective radius R=a/1−φR = a/\sqrt{1-\varphi} that preserves the porosity. At r=ar = a, we impose Eq. (4) while at an effective cell boundary r=Rr = R, with e^r=(cos⁡θ,sin⁡θ)\hat{\mathbf{e}}_r = (\cos\theta,\sin\theta), we impose:

u(R,θ)⋅e^r=U⋅e^r,ω(R,θ)=εij∂iuj∣r=R=0.(7)\mathbf{u}(R,\theta) \cdot\hat{\mathbf{e}}_r = U \cdot\hat{\mathbf{e}}_r, \qquad\omega(R,\theta) = \left.\varepsilon_{ij}\partial_i u_j\right|_{r=R} = 0. \tag*{(7)}

The first condition fixes the cell flux through the area-averaged velocity UU. The second is the Kuwabara closure [7], which effectively accounts for neighboring obstacles in a many-body system by imposing zero vorticity at the cell boundary (see Figure 4). This approximation is most accurate for well-separated obstacles in locally isotropic environments. Deferring the derivation to the End Matter, the resulting area-averaged velocity obeys Darcy’s law:

Figure 4. Kuwabara construction with cells around each obstacle corresponding approximately to ω=0\omega= 0. (a) velocity and vorticity fields. The exact points at which ω=0\omega= 0 are also shown, but these are not the ones considered by Kuwabara’s construction. (b) Corresponding Kuwabara cell. Here, ∣α∣=1|\alpha| = 1 and β=0\beta= 0.

Comparison of theory and simulations for periodic porous media

Figure 2. Comparison of theory and simulations for periodic porous media. (a,b) Even and odd permeabilities versus porosity for several lattices and periodic arrangements of randomly positioned disks averaged over 3 realizations, with ∣α∣=1|\alpha| = 1 and β=0\beta= 0. (c) Hall angle θo=arctan⁡(Ko/K)\theta_o = \arctan(K_o/K) versus porosity and odd viscosity for the hexagonal lattice at β=0\beta= 0. (d,e) Even and odd permeabilities versus α\alpha and the slip parameter β\beta for the hexagonal lattice at φ=0.9\varphi= 0.9. (f) Values of α\alpha maximizing ∣θo∣|\theta_o| and ∣Ko∣|K_o|, compared to the hexagonal lattice. (g) Normalized response for all simulated geometries, porosities, and slip coefficients. The gray disk is the passive-response bound Eq. (16). Opposite signs of α\alpha are folded through ∣Ko∣|K_o|. Here Ac=πa2/(1−φ)A_c = \pi a^2/(1-\varphi).

U=−1ηK⋅∇⟨p⟩,(8)\boldsymbol{U}=-\frac{1}{\eta}\boldsymbol{K}\cdot\nabla\langle p\rangle, \tag*{(8)}

where the permeability tensor reads

K=(Kns+cc2+α2ΔKαc2+α2ΔK−αc2+α2ΔKKns+cc2+α2ΔK),(9)\boldsymbol{K}=\begin{pmatrix} K_{\mathrm{ns}}+\frac{c}{c^2+\alpha^2}\Delta K & \frac{\alpha}{c^2+\alpha^2}\Delta K \\ -\frac{\alpha}{c^2+\alpha^2}\Delta K & K_{\mathrm{ns}}+\frac{c}{c^2+\alpha^2}\Delta K \end{pmatrix}, \tag*{(9)}

with α=ηo/η\alpha=\eta_o/\eta, β=ξa/η\beta=\xi a/\eta, c=1+β/2c=1+\beta/2, and ΔK=Kps−Kns\Delta K=K_{\mathrm{ps}}-K_{\mathrm{ns}}. Here, Kps(φ)≡Kxx(φ;α=0,β=0)K_{\mathrm{ps}}(\varphi)\equiv K_{xx}(\varphi;\alpha=0,\beta=0) and Kns(φ)≡Kxx(φ;α=0,β→∞)K_{\mathrm{ns}}(\varphi)\equiv K_{xx}(\varphi;\alpha=0,\beta\to\infty) are the perfect-slip and no-slip permeabilities of the corresponding passive fluid. Their dependence on porosity φ\varphi is given explicitly in the End Matter. The transverse response is encoded by the off-diagonal permeability, which deflects the flow perpendicular to the pressure gradient. Despite this deflection, the permeability remains isotropic, since it is built only from rotation-invariant tensors

K=K1−Koϵ,K=Kns+cΔKc2+α2,Ko=−αΔKc2+α2,(10)\boldsymbol{K}=K\mathbf{1}-K_o\boldsymbol{\epsilon},\qquad K=K_{\mathrm{ns}}+\frac{c\Delta K}{c^2+\alpha^2},\qquad K_o=-\frac{\alpha\Delta K}{c^2+\alpha^2}, \tag*{(10)}

where 1\mathbf{1} is the identity tensor and KK and KoK_o denote the conventional and odd permeabilities, respectively.

(12) can be recast compactly in complex notation as

U=KηG,K≡K+iKo=Kns+ΔK1+β/2+iα,(11)U=\frac{\mathcal{K}}{\eta}\mathcal{G},\qquad\mathcal{K}\equiv K+iK_o=K_{\mathrm{ns}}+\frac{\Delta K}{1+\beta/2+i\alpha}, \tag*{(11)}

with U=Ux+iUyU=U_x+iU_y and G=−∂x⟨p⟩−i∂y⟨p⟩\mathcal{G}=-\partial_x\langle p\rangle-i\partial_y\langle p\rangle. The permeability K\mathcal{K} acquires an imaginary part only when α=ηo/η≠0\alpha=\eta_o/\eta\ne0, thereby producing a flow component transverse to the pressure gradient. Conversely, for α=0\alpha=0, K\mathcal{K} is real, and the transverse response vanishes. The same occurs in the no-slip limit β→∞\beta\to\infty, where the odd-viscous correction is negligible and K→Kns\mathcal{K}\to K_{\mathrm{ns}}.

Fig. 2 tests the response across several porous geometries. Panels (a) and (b) compare the predicted and simulated permeabilities for periodic lattices and a periodic random array. Agreement is excellent at high porosity, where obstacles are well separated and many-body effects are weak, but worsens as porosity decreases and local packing and lattice anisotropy become important. The hexagonal lattice therefore agrees most closely with the rotationally symmetric cell model, whereas the Kagome lattice exhibits larger deviations despite its global isotropy. Capturing these geometry-specific effects would require solving the Stokes problem in the corresponding unit cell [6]. Panel (c) shows the Hall angle θo=arctan⁡(Ko/K)\theta_o=\arctan(K_o/K), @page 3/21 starts_mid=1 ends_mid=1

Comparison of theory and simulations for periodic porous media

Figure 2 (PDF p. 3). Comparison of theory and simulations for periodic porous media. (a,b) Even and odd permeabilities versus porosity for several lattices and periodic arrangements of randomly positioned disks averaged over 3 realizations, with ∣α∣=1|\alpha| = 1 and β=0\beta= 0. (c) Hall angle θo=arctan⁡(Ko/K)\theta_o = \arctan(K_o/K) versus porosity and odd viscosity for the hexagonal lattice at β=0\beta= 0. (d,e) Even and odd permeabilities versus α\alpha and the slip parameter β\beta for the hexagonal lattice at φ=0.9\varphi= 0.9. (f) Values of α\alpha maximizing ∣θo∣|\theta_o| and ∣Ko∣|K_o|, compared to the hexagonal lattice. (g) Normalized response for all simulated geometries, porosities, and slip coefficients. The gray disk is the passive-response bound Eq. (16). Opposite signs of α\alpha are folded through ∣Ko∣|K_o|. Here Ac=πa2/(1−φ)A_c = \pi a^2/(1-\varphi).

is most accurate for well-separated obstacles in locally isotropic environments. Deferring the derivation to the End Matter, the resulting area-averaged velocity obeys Darcy’s law:

U=−1ηK⋅∇⟨p⟩,(12)\boldsymbol{U}=-\frac{1}{\eta}\boldsymbol{K}\cdot\nabla\langle p\rangle, \tag*{(12)}

where the permeability tensor reads

K=(Kns+cc2+α2ΔKαc2+α2ΔK−αc2+α2ΔKKns+cc2+α2ΔK),(13)\boldsymbol{K}=\begin{pmatrix} K_{\mathrm{ns}}+\frac{c}{c^2+\alpha^2}\Delta K & \frac{\alpha}{c^2+\alpha^2}\Delta K \\ -\frac{\alpha}{c^2+\alpha^2}\Delta K & K_{\mathrm{ns}}+\frac{c}{c^2+\alpha^2}\Delta K \end{pmatrix}, \tag*{(13)}

with α=ηo/η\alpha=\eta_o/\eta, β=ξa/η\beta=\xi a/\eta, c=1+β/2c=1+\beta/2, and ΔK=Kps−Kns\Delta K=K_{\mathrm{ps}}-K_{\mathrm{ns}}. Here, Kps(φ)≡Kxx(φ;α=0,β=0)K_{\mathrm{ps}}(\varphi)\equiv K_{xx}(\varphi;\alpha=0,\beta=0) and Kns(φ)≡Kxx(φ;α=0,β→∞)K_{\mathrm{ns}}(\varphi)\equiv K_{xx}(\varphi;\alpha=0,\beta\to\infty) are the perfect-slip and no-slip permeabilities of the corresponding passive fluid. Their dependence on porosity φ\varphi is given explicitly in the End Matter. The transverse response is encoded by the off-diagonal permeability, which deflects the flow perpendicular to the pressure gradient. Despite this deflection, the permeability remains isotropic, since it is built only from rotation-invariant tensors

K=K1−Koϵ,K=Kns+cΔKc2+α2,Ko=−αΔKc2+α2,(14)\boldsymbol{K}=K\mathbf{1}-K_o\boldsymbol{\epsilon},\qquad K=K_{\mathrm{ns}}+\frac{c\Delta K}{c^2+\alpha^2},\qquad K_o=-\frac{\alpha\Delta K}{c^2+\alpha^2}, \tag*{(14)}

where 1\mathbf{1} is the identity tensor and KK and KoK_o denote the conventional and odd permeabilities, respectively.

(12) can be recast compactly in complex notation as

U=KηG,K≡K+iKo=Kns+ΔK1+β/2+iα,(15)U=\frac{\mathcal{K}}{\eta}\mathcal{G},\qquad\mathcal{K}\equiv K+iK_o=K_{\mathrm{ns}}+\frac{\Delta K}{1+\beta/2+i\alpha}, \tag*{(15)}

with U=Ux+iUyU=U_x+iU_y and G=−∂x⟨p⟩−i∂y⟨p⟩\mathcal{G}=-\partial_x\langle p\rangle-i\partial_y\langle p\rangle. The permeability K\mathcal{K} acquires an imaginary part only when α=ηo/η≠0\alpha=\eta_o/\eta\ne0, thereby producing a flow component transverse to the pressure gradient. Conversely, for α=0\alpha=0, K\mathcal{K} is real, and the transverse response vanishes. The same occurs in the no-slip limit β→∞\beta\to\infty, where the odd-viscous correction is negligible and K→Kns\mathcal{K}\to K_{\mathrm{ns}}.

Fig. 2 tests the response across several porous geometries. Panels (a) and (b) compare the predicted and simulated permeabilities for periodic lattices and a periodic random array. Agreement is excellent at high porosity, where obstacles are well separated and many-body effects are weak, but worsens as porosity decreases and local packing and lattice anisotropy become important. The hexagonal lattice therefore agrees most closely with the rotationally symmetric cell model, whereas the Kagome lattice exhibits larger deviations despite its global isotropy. Capturing these geometry-specific effects would require solving the Stokes problem in the corresponding unit cell [6]. Panel (c) shows the Hall angle θo=arctan⁡(Ko/K)\theta_o=\arctan(K_o/K), which quantifies the transverse deflection of the mean flow. While the cell theory predicts a monotonic dependence on porosity, the simulations reveal a maximum at intermediate porosity, showing that geometry and nontrivial correlations can be used to optimize the transverse response. Panels (d) and (e) show the dependence on the odd-viscosity parameter α\alpha and slip friction β\beta. At fixed β\beta, KK decreases monotonically with ∣α∣|\alpha|. Its dependence on β\beta is more unexpected. For ∣α∣≤1|\alpha| \le1, KK decreases monotonically, while for ∣α∣>1|\alpha| > 1 it first increases, reaches a maximum at β∗=2(∣α∣−1)\beta_* = 2(|\alpha| - 1), and then approaches KnsK_{\mathrm{ns}}. Thus, increasing wall friction can enhance longitudinal transport when odd viscosity is sufficiently strong. The odd permeability ∣Ko∣|K_{\mathrm{o}}| is always suppressed by increasing β\beta and is maximal at ∣α∗∣=1+β/2|\alpha_*| = 1 + \beta/2. Panel (f) confirms this relation and shows that the value of α\alpha maximizing the Hall angle is different and depends strongly on porosity. Similar optima arise in other chiral systems [75, 76, 77].

Because odd viscosity does no work, energy balance bounds the odd permeability independently of the porous medium geometry (see SM) [78]:

Ko2≤(K−Kns)(K(α=0)−K).(16)K_{\mathrm{o}}^2 \le(K-K_{\mathrm{ns}})(K(\alpha=0)-K). \tag*{(16)}

Two passive measurements thus bound the transverse response: ∣Ko∣≤(K(α=0)−Kns)/2|K_{\mathrm{o}}| \le(K(\alpha= 0) - K_{\mathrm{ns}})/2. The inequality follows by neglecting a nonnegative dissipation term associated with flow carrying no net flux. This term vanishes in Kuwabara’s rotationally symmetric cell, but is generally finite in more complex geometries, moving the response inside the disk in Fig. 2(g).

External damping, which captures momentum loss to a substrate and is particularly relevant in biological systems, leaves our results largely unchanged, as shown in the Supplemental Material (SM) [78]. Three-dimensional flows are qualitatively different. In 3D, odd viscosity requires anisotropy and remains dynamically relevant even under no-slip conditions. As shown in the SM [78], it produces an odd permeability with a surprising sign dependence on boundary conditions.

Inertia and torque-driven odd permeability—We now restore the torque density τ\tau in Eq. (3). Since τ=τ(ρ)\tau= \tau(\rho) and incompressibility fixes ρ\rho, τ\tau is constant in the fluid bulk. Its only variation is the jump across the fluid-solid interface, so the associated force density ϵ⋅∇τ\boldsymbol{\epsilon} \cdot\nabla\tau has support only on ∂B\partial B [79]. Equivalently, the antisymmetric stress acts as a tangential boundary traction, and the boundary condition becomes

t^(r)⋅Στ=0(r)⋅n^(r)=ξu(r)⋅t^(r)+τ(ρ).(17)\hat{\mathbf{t}}(\mathbf{r}) \cdot\boldsymbol{\Sigma}^{\tau=0}(\mathbf{r}) \cdot\hat{\mathbf{n}}(\mathbf{r}) = \xi\mathbf{u}(\mathbf{r}) \cdot\hat{\mathbf{t}}(\mathbf{r}) + \tau(\rho). \tag*{(17)}

where τ\tau has been removed from the bulk stress and retained only as a boundary forcing which drives edge currents [79]. In an isotropic medium, however, their net contributions to a global flow vanish. At the Stokes level, linearity also decouples them from the pressure-driven flow. The Darcy law therefore remains unchanged.

Once inertia is retained, however, the conclusion changes since the nonlinear convective term couples the boundary-driven flow to the edge currents. We therefore consider the stationary momentum balance

ρ(u⋅∇)u=∇⋅Στ=0,(18)\rho(\mathbf{u}\cdot\nabla)\mathbf{u} = \nabla\cdot\boldsymbol{\Sigma}^{\tau=0}, \tag*{(18)}

subject to impermeability and the driven boundary condition in Eq. (17). We assume that the pressure gradient and are weak enough such that the stationary solution of Eq. (18) is stable and below the onset of turbulence. We focus on perfect slip, β=0\beta= 0, and vanishing odd viscosity, with the general case discussed in the SM [78]. Figure 3 shows simulations of Eq. (18). Panel (a) displays the full flow around an obstacle. Torque-induced circulation weakly breaks the top-bottom symmetry of the pressure-driven flow. Panel (b) isolates the inertial correction by subtracting the solution without the convective term. The resulting upward flow demonstrates that inertia and edge currents alone generate a transverse response, even without odd viscosity.

Figure 3. Odd permeability induced by inertia and torque density. (a) Full flow and (b) inertial contribution alone at φ=0.97\varphi= 0.97. (c) Comparison of measured KoK_{\mathrm{o}} with theory. Here, α=β=0\alpha= \beta= 0, ρ∣τ∣Ac/η2=1\rho|\tau|A_{\mathrm{c}}/\eta^2 = 1 and ρ∣G∣Ac3/η2=1\rho|G|A_{\mathrm{c}}^3/\eta^2 = 1.

To quantify this effect, we solve the cell problem perturbatively to first order in inertia, as detailed in the SM [78]. For β=α=0\beta= \alpha= 0, the resulting permeabilities are

K≃KStokes1+Λ2,Ko≃KStokesΛ1+Λ2,Λ∝ρτa2η2,(19)K \simeq\frac{K_{\mathrm{Stokes}}}{1+\Lambda^2}, \qquad K_{\mathrm{o}} \simeq\frac{K_{\mathrm{Stokes}}\Lambda}{1+\Lambda^2}, \qquad\Lambda\propto\frac{\rho\tau a^2}{\eta^2}, \tag*{(19)}

where KStokesK_{\mathrm{Stokes}} is the passive-fluid permeability at β=α=0\beta=\alpha=0 found in Eq. (9) in the Stokes regime. The odd permeability is approximately proportional to Λ\Lambda, which depends nontrivially on porosity, changes sign with τ\tau, and vanishes with the torque density and its associated edge currents. The theoretical predictions are in good agreement with simulations of Eq. (18) for various lattices as shown in Figure 3(c). The nonmonotonic dependence of KoK_o on porosity results from two competing effects. Increasing porosity reduces viscous resistance but simultaneously decreases the boundary supporting the edge currents. The transverse response is therefore suppressed in both the dense and dilute limits and reaches a maximum at ϕ≃0.9\phi\simeq0.9 within our cell theory. We also measured in the SM how our first-order approximation degrades as the torque density is increased [78].

**Conclusion—**We have shown that Darcy’s law extends to chiral active fluids through an antisymmetric permeability, and identified two distinct mechanisms underlying this transverse response: odd viscosity, which acts already in the Stokes regime, and torque-driven edge currents, which couple to pressure-driven flow through inertia. Our theory quantitatively predicts these effects and their dependence on porosity and boundary friction.

These findings open several questions about transport in chiral porous media. Relaxing incompressibility may allow the torque density τ\tau to generate odd permeability already in the Stokes regime. More fundamentally, torque-driven edge currents in porous media remain poorly understood, even in the absence of pressure driving. The effects of weak anisotropy, local porosity variations, and tortuosity also warrant further study, particularly in biological environments. At the opposite extreme, recent studies of turbulence in bulk odd fluids [80, 81, 82] provide a starting point for exploring the largely uncharted turbulent regime of chiral porous media.

References

End Matter

*Numerical methods—*We solve the steady incompressible momentum equations using a finite-element method. Multi-obstacle periodic systems are meshed with Gmsh [83]. The velocity and pressure are discretized using P2-P1 Taylor-Hood elements [84]. The macroscopic pressure gradient is represented by a constant body force, while the pressure null mode is removed through a zero-mean constraint [84]. Impermeability is imposed as an essential constraint, whereas the slip friction, odd-viscous traction, and imposed torque traction enter through the boundary terms of the weak formulation of the Stokes equation [84]. The finite-element integrals in the linear Stokes problem yield a system of equations solved via SciPy [85]. When a finite inertia result is wanted, this Stokes solution provides the initial iterate for a Newton method [86]. The area-averaged velocity UU is calculated by integrating the finite-element velocity over the fluid region and dividing by the full system’s area.

*Cell theory derivation—*The Stokes equation in the fluid domain is:

−∇(p−ηoω)+η∇2u+G=0.(20)-\nabla(p-\eta_o\omega)+\eta\nabla^2\boldsymbol{u}+\boldsymbol{G}=0. \tag*{(20)}

where G=−∇⟨p⟩\boldsymbol{G}=-\nabla\langle p\rangle and with boundary conditions

u(r)⋅n^(r)=0,t^(r)⋅Σ(r)⋅n^(r)=ξu(r)⋅t^(r),(21)\boldsymbol{u}(\boldsymbol{r})\cdot\hat{\boldsymbol{n}}(\boldsymbol{r})=0,\qquad\hat{\boldsymbol{t}}(\boldsymbol{r})\cdot\boldsymbol{\Sigma}(\boldsymbol{r})\cdot\hat{\boldsymbol{n}}(\boldsymbol{r})=\xi\boldsymbol{u}(\boldsymbol{r})\cdot\hat{\boldsymbol{t}}(\boldsymbol{r}), \tag*{(21)}

on ∂B\partial B where the reduced stress tensor (using pp rather than pop_o), with τ=0\tau=0 reads:

Σij=−pδij+η(∂iuj+∂jui)+ηo(εjk∂iuk+εik∂kuj).(22)\Sigma_{ij}=-p\delta_{ij}+\eta(\partial_i u_j+\partial_j u_i)+\eta_o(\varepsilon_{jk}\partial_i u_k+\varepsilon_{ik}\partial_k u_j). \tag*{(22)}

In general, such a flow cannot be solved in closed form, even for a circular obstacle in a square box, because the geometry lacks continuous rotational invariance. We therefore study a simpler setup. We suppose that the obstacles are identical, nonoverlapping, and sufficiently evenly spaced, as in Figure 4. Each obstacle is then assigned a “cell” whose outer boundary has approximately zero vorticity (see the construction marked “Kuwabara” in Figure 4(b)). Instead of solving the dynamics of the fluid for the full system, we consider one obstacle of radius aa surrounded by an effective circular boundary of radius RR, representing the cell described above [7].

The porosity φ\varphi of this cell is φ=1−a2/R2\varphi= 1-a^2/R^2, which can be mapped to the porosity of the original medium. Since a uniform drive excites only the m=1m = 1 angular harmonic in this rotationally invariant system, the streamfunction can be written as

ψ(r,θ)=fx(r)sin⁡θ−fy(r)cos⁡θ,(23)\psi(r,\theta) = f_x(r)\sin\theta- f_y(r)\cos\theta, \tag*{(23)}

with

ur=fxrcos⁡θ+fyrsin⁡θ,uθ=−fx′sin⁡θ+fy′cos⁡θ.(24)u_r = \frac{f_x}{r}\cos\theta+ \frac{f_y}{r}\sin\theta,\qquad u_\theta= -f_x'\sin\theta+ f_y'\cos\theta. \tag*{(24)}

We package these two real radial amplitudes, and likewise the area-averaged Darcy velocity U=(Ux,Uy)U = (U_x,U_y) and pressure-gradient drive G=(Gx,Gy)G = (G_x,G_y), into the complex quantities

f=fx+ify,U=Ux+iUy,G=Gx+iGy.(25)f = f_x + if_y,\qquad U = U_x + iU_y,\qquad G = G_x + iG_y. \tag*{(25)}

This is a useful construction to compactly express the boundary conditions and the solution.

Taking the curl of the Stokes equation gives the biharmonic equation ∇4ψ=0\nabla^4\psi= 0. Both real harmonics obey the same radial equation, so their complex combination has the general solution

f(r)=Ar+Br+Crlog⁡(rR)+Dr3,(26)f(r) = Ar + \frac{B}{r} + Cr\log\left(\frac{r}{R}\right) + Dr^3, \tag*{(26)}

where AA, BB, CC, DD are complex constants.

We now impose the four cell conditions

f(a)=0,(27)f(a) = 0, \tag*{(27)}
η[−f′′(a)+f′(a)a]+2iηof′(a)a+ξf′(a)=0,(28)\eta\left[-f''(a) + \frac{f'(a)}{a}\right] + 2i\eta_o\frac{f'(a)}{a} + \xi f'(a) = 0, \tag*{(28)}
f(R)=UR,(29)f(R) = \mathcal{U}R, \tag*{(29)}
f′′(R)+f′(R)R−f(R)R2=0.(30)f''(R) + \frac{f'(R)}{R} - \frac{f(R)}{R^2} = 0. \tag*{(30)}

Eq. (27) is impermeability at the obstacle, and Eq. (28) is the Navier slip condition. At the artificial outer circle r=Rr = R, Eq. (29) imposes the radial velocity

ur(R,θ)=Uxcos⁡θ+Uysin⁡θ=U⋅e^r,(31)u_r(R,\theta) = U_x\cos\theta+ U_y\sin\theta= U\cdot\hat{e}_r, \tag*{(31)}

with UU conveniently corresponding to the area-averaged velocity. Indeed, incompressibility and impermeability at the obstacle give

1πR2∫a<r<Ru dr=1πR2∫r=Rr(u⋅n^) ds=U.(32)\frac{1}{\pi R^2}\int_{a<r<R} \boldsymbol{u}\,dr = \frac{1}{\pi R^2}\int_{r=R} r(\boldsymbol{u}\cdot\hat{\boldsymbol{n}})\,ds = U. \tag*{(32)}

Eq. (30) is the Kuwabara closure [7], which imposes vanishing vorticity at r=Rr = R to approximate the cancellation of the rotation generated by the central obstacle by that generated in neighboring cells of the original periodic problem, as illustrated in Figure 4.

Eqs. (27)–(30) form the linear system

(aa−1alog⁡(a/R)a3LALBLCLDRR−10R3002R−18R)(ABCD)=(00UR0)(33)\begin{pmatrix} a & a^{-1} & a\log(a/R) & a^3 \\ L_A & L_B & L_C & L_D \\ R & R^{-1} & 0 & R^3 \\ 0 & 0 & 2R^{-1} & 8R \end{pmatrix} \begin{pmatrix} A\\B\\C\\D \end{pmatrix} = \begin{pmatrix} 0\\0\\\mathcal{U}R\\0 \tag*{(33)} \end{pmatrix}

with

LA=(η+2iηo)/a+ξ,LB=−(3η+2iηo)/a3−ξ/a2,L_A = (\eta+ 2i\eta_o)/a + \xi,\qquad L_B = -(3\eta+ 2i\eta_o)/a^3 - \xi/a^2,
LC=ηalog⁡(aR)+2iηoa[log⁡(aR)+1]+ξ[log⁡(aR)+1],L_C = \frac{\eta}{a}\log\left(\frac{a}{R}\right) + \frac{2i\eta_o}{a}\left[\log\left(\frac{a}{R}\right) + 1\right] + \xi\left[\log\left(\frac{a}{R}\right) + 1\right],
LD=(−3η+6iηo)a+3ξa2.(34)L_D = (-3\eta+ 6i\eta_o)a + 3\xi a^2. \tag*{(34)}

Inverting the system (33) yields the solution to the flow. By integrating the traction around the obstacle, the force on it is found. It must be balanced against the cell pressure force πR2G\pi R^2G, which yields Darcy’s law

U=KηG,K=K+iKo,K=πR2K∗,(35)\mathcal{U} = \frac{\mathcal{K}}{\eta}G,\qquad\mathcal{K} = K + iK_o,\qquad\mathcal{K} = \pi R^2\mathcal{K}^*, \tag*{(35)}

with

K∗(φ;α,β)=Kns∗(φ)+Kps∗(φ)−Kns∗(φ)1+β/2+iα,(36)\mathcal{K}^*(\varphi;\alpha,\beta) = K_{\mathrm{ns}}^*(\varphi) + \frac{K_{\mathrm{ps}}^*(\varphi)-K_{\mathrm{ns}}^*(\varphi)}{1+\beta/2+i\alpha}, \tag*{(36)}

where α=ηo/η\alpha= \eta_o/\eta and β=ξa/η\beta= \xi a/\eta and Kps∗K_{\mathrm{ps}}^* and Kns∗K_{\mathrm{ns}}^* are the perfect-slip and no-slip permeabilities in the absence of odd viscosity:

Kps∗=−2log⁡(1−φ)−1+(1−φ)216π,(37)K_{\mathrm{ps}}^* = \frac{-2\log(1-\varphi)-1+(1-\varphi)^2}{16\pi}, \tag*{(37)}
Kns∗=−2log⁡(1−φ)−3+4(1−φ)−(1−φ)216π.(38)K_{\mathrm{ns}}^* = \frac{-2\log(1-\varphi)-3+4(1-\varphi)-(1-\varphi)^2}{16\pi}. \tag*{(38)}

Energy bound on odd-viscous permeability

We derive Eq. (16) for a pressure gradient G=−∇⟨p⟩=Gex\mathbf{G}=-\nabla\langle p\rangle=G\mathbf{e}_x and the corresponding flow ux\mathbf{u}_x. For two velocity fields u\mathbf{u} and v\mathbf{v}, we define their “mutual dissipation”

D(v,u)=2η∫ΩfE(v):E(u)dA+∫∂Bξvtutds,Eij(u)=∂iuj+∂jui2.(39)\mathcal{D}(\mathbf{v},\mathbf{u})=2\eta\int_{\Omega_f}\mathbf{E}(\mathbf{v}):\mathbf{E}(\mathbf{u})\mathrm{d}A+\int_{\partial B}\xi v_tu_t\mathrm{d}s,\qquad E_{ij}(\mathbf{u})=\frac{\partial_i u_j+\partial_j u_i}{2}. \tag*{(39)}

where E\mathbf{E} is the strain-rate tensor, ut=u⋅t^u_t=\mathbf{u}\cdot\hat{\mathbf{t}}, and Ωf\Omega_f is the fluid region of a cell of total area AA. In particular, D(u,u)≥0\mathcal{D}(\mathbf{u},\mathbf{u})\ge0 is the power dissipated by bulk viscosity and wall friction.

Let uiu_i and ui0≡ui(α=0)u_i^0\equiv u_i(\alpha=0) denote the flows driven by GeiG\mathbf{e}_i with and without odd viscosity, at the same wall friction. Using the no-slip reference uins=ui0(β→∞)u_i^{\mathrm{ns}}=u_i^0(\beta\to\infty), define

wi0=ui0−uins,wi=ui−uins,i=x,y.(40)w_i^0=u_i^0-u_i^{\mathrm{ns}},\qquad w_i=u_i-u_i^{\mathrm{ns}},\qquad i=x,y. \tag*{(40)}

The no-slip reference is independent of α\alpha in 2D. We study wxw_x, using wx0w_x^0 and wy0w_y^0 as passive auxiliary fields.

In the steady state, the pressure work is converted into bulk and wall dissipation such that D(ux,ux)=AG2K/η\mathcal{D}(\mathbf{u}_x,\mathbf{u}_x)=AG^2K/\eta. Multiplying the Stokes equation by a no-slip reference and integrating by parts removes the odd boundary traction and gives D(uins,wx)=0\mathcal{D}(\mathbf{u}_i^{\mathrm{ns}},\mathbf{w}_x)=0. The slip correction therefore dissipates exactly the extra pressure work relative to no slip:

D(wx,wx)=AG2η(K−Kns).(41)\mathcal{D}(\mathbf{w}_x,\mathbf{w}_x)=\frac{AG^2}{\eta}(K-K_{\mathrm{ns}}). \tag*{(41)}

To identify how much of this power carries mean flow, we use the passive solutions. Multiplying their Stokes equations by an incompressible, periodic, impermeable field v\mathbf{v} and integrating by parts gives

D(ui0,v)=G∫ΩfvidA.(42)\mathcal{D}(\mathbf{u}_i^0,\mathbf{v})=G\int_{\Omega_f}v_i\mathrm{d}A. \tag*{(42)}

Thus the mutual dissipation with a passive flow measures the flux along its driving direction. By isotropy, wx0w_x^0 and wy0w_y^0 carry the same extra mean speed G(K(α=0)−Kns)/ηG(K(\alpha=0)-K_{\mathrm{ns}})/\eta, along xx and yy, respectively. We normalize the extra longitudinal and transverse fluxes by the passive slip-induced flux:

X=K−KnsK(α=0)−Kns,Y=KoK(α=0)−Kns.(43)X=\frac{K-K_{\mathrm{ns}}}{K(\alpha=0)-K_{\mathrm{ns}}},\qquad Y=\frac{K_{\mathrm{o}}}{K(\alpha=0)-K_{\mathrm{ns}}}. \tag*{(43)}

Writing

wx=Xwx0+Ywy0+w⊥,(44)w_x = X w_x^0 + Y w_y^0 + w_\perp, \tag*{(44)}

the remaining motion w⊥w_\perp has zero mean flux. Eq. (42) and the no-slip subtraction give D(wx0,w⊥)=0\mathcal{D}(w_x^0,w_\perp)=0, while isotropy gives D(wx0,wy0)=0\mathcal{D}(w_x^0,w_y^0)=0. Eq. (41) then yields:

K−Kns=(K(α=0)−Kns)(X2+Y2)+ηAG2D(w⊥,w⊥).(45)K-K_{\mathrm{ns}} = \left(K(\alpha=0)-K_{\mathrm{ns}}\right)(X^2+Y^2)+\frac{\eta}{AG^2}\mathcal{D}(w_\perp,w_\perp). \tag*{(45)}

The first term is the cost of carrying the extra mean flow while the last is dissipation without net transport. Since the latter is nonnegative, by completing the square, we obtain the wanted bond:

(X−12)2+Y2≤14,Ko2≤(K−Kns)[K(α=0)−K].(46)\left(X-\frac{1}{2}\right)^2+Y^2\leq\frac{1}{4},\qquad K_o^2\leq(K-K_{\mathrm{ns}})[K(\alpha=0)-K]. \tag*{(46)}

Hence Kns≤K≤K(α=0)K_{\mathrm{ns}}\leq K\leq K(\alpha=0) and ∣Ko∣≤(K(α=0)−Kns)/2|K_o|\leq(K(\alpha=0)-K_{\mathrm{ns}})/2.

Circular-cell theory saturates the bound. Indeed, Eq. (36), with c=1+β/2c=1+\beta/2, gives

X+iY=11+iα/c,(X−12)2+Y2=14.(47)X+iY=\frac{1}{1+i\alpha/c},\qquad\left(X-\frac{1}{2}\right)^2+Y^2=\frac{1}{4}. \tag*{(47)}

More complex geometries can generate additional motion with zero mean flux, placing the response inside the disk in Fig. 2(g). The distance from its boundary measures this extra dissipation.

A fixed substrate damping γ≥0\gamma\geq0 simply adds the nonnegative power γ∫Ωf∣u∣2 dA\gamma\int_{\Omega_f}|u|^2\,\mathrm{d}A to Eq. (39) and the argument remains unchanged. By contrast, in the 3D geometry considered below, using the full passive flows instead of the slip corrections gives the weaker bound Ko2≤K⊥[K(α=0)−K⊥]K_o^2\leq K_\perp[K(\alpha=0)-K_\perp], where K⊥=Kxx=KyyK_\perp=K_{xx}=K_{yy}. The no-slip subtraction is unavailable because odd viscosity modifies the no-slip flow. In the torque-driven inertial case also considered below, the imposed torque density supplies additional power, so this pressure-work bound no longer applies.

Permeabilities with external damping

We repeat the calculation done in the End Matter, but now adding damping. The momentum equation is

−∇(p−ηoω)+η∇2u−γu+G=0,∇⋅u=0,(48)-\nabla(p-\eta_o\omega)+\eta\nabla^2u-\gamma u+G=0,\qquad\nabla\cdot u=0, \tag*{(48)}

where G=−∇⟨p⟩G=-\nabla\langle p\rangle and the external damping γ\gamma has units of viscosity per area and defines the screening length ℓγ=η/γ\ell_\gamma=\sqrt{\eta/\gamma}. We use

δ=Rℓγ.(49)\delta=\frac{R}{\ell_\gamma}. \tag*{(49)}

When δ≪1\delta\ll1, the cell is small compared with the screening length and the undamped result should be recovered (although we must be careful with the Stokes paradox at φ→1\varphi\to1). Taking the curl of Eq. (48) removes the pressure and the constant drive and gives

(∇2−ℓγ−2)∇2ψ=0,(50)(\nabla^2-\ell_\gamma^{-2})\nabla^2\psi=0, \tag*{(50)}

with ψ\psi the streamfunction. Using the first-harmonic convention of Eq. (23), the complex radial solution is

f(r)=Ar+Br+CI1(r/ℓγ)+DK1(r/ℓγ),(51)f(r)=Ar+\frac{B}{r}+CI_1(r/\ell_\gamma)+DK_1(r/\ell_\gamma), \tag*{(51)}

where I1I_1 and K1K_1 are modified Bessel functions. Define

L1g=g′′+g′r−gr2,Tα,βg=−g′′(a)+1+β+2iαag′(a).(52)\mathcal{L}_1g=g^{\prime\prime}+\frac{g^\prime}{r}-\frac{g}{r^2},\qquad\mathcal{T}_{\alpha,\beta}g=-g^{\prime\prime}(a)+\frac{1+\beta+2i\alpha}{a}g^\prime(a). \tag*{(52)}

The four boundary conditions are:

f(a)=0,Tα,βf=0,f(R)=UR,L1f(R)=0.(53)f(a)=0,\qquad\mathcal{T}_{\alpha,\beta}f=0,\qquad f(R)=UR,\qquad\mathcal{L}_1f(R)=0. \tag*{(53)}

With

ϕ(r)=(r,r−1,I1(r/ℓγ),K1(r/ℓγ)),(54)\phi(r) = \left(r, r^{-1}, I_1(r/\ell_\gamma), K_1(r/\ell_\gamma)\right), \tag*{(54)}

the prescribed-U\mathcal{U} problem is the 4×44 \times4 system

\begin{pmatrix} \phi(a) \\ \mathcal{T}_{\alpha,\beta}\phi(a) \\ \phi(R) \\ \mathcal{L}_1\phi(R) \end{pmatrix} \begin{pmatrix} A \\ B \\ C \\ D \end{pmatrix} = \begin{pmatrix} 0 \\ 0 \\ \mathcal{U}R \\ 0 \end{pmatrix}. \tag*{(55)}\end{pmatrix}

This determines AA, BB, CC, DD.

Let F∂BF_{\partial B} be the integral of the complete traction exerted by the fluid on the obstacle, including its pressure, even-viscous, and odd-viscous parts. Force balance yields:

πR2G=F∂B+γ∫a<r<Ru dr=F∂B+γπR2U,(56)\pi R^2 G = F_{\partial B} + \gamma\int_{a<r<R} \mathbf{u}\, \mathrm{d}r = F_{\partial B} + \gamma\pi R^2 \mathcal{U}, \tag*{(56)}

where the last equality follows from Eq. (32). Everything on the right-hand side is known from the prescribed-U\mathcal{U} solution, so Eq. (56) directly determines GG:

U=KδηG,Kδ∗≡Kδ∗+iKo,δ∗=Kns,δ∗+Kps,δ∗−Kns,δ∗1+qδ(β/2+iα),Kδ∗=KδπR2,(57)\mathcal{U} = \frac{K_\delta}{\eta}\mathcal{G}, \qquad K_\delta^* \equiv K_\delta^* + iK_{o,\delta}^* = K_{\mathrm{ns},\delta}^* + \frac{K_{\mathrm{ps},\delta}^* - K_{\mathrm{ns},\delta}^*}{1 + q_\delta(\beta/2 + i\alpha)}, \qquad K_\delta^* = \frac{K_\delta}{\pi R^2}, \tag*{(57)}

where the endpoint permeabilities and qδq_\delta have a compact explicit form. We introduce

Hδ(x)=I1(δx)K1(δ)−K1(δx)I1(δ)(58)H_\delta(x) = I_1(\delta x)K_1(\delta) - K_1(\delta x)I_1(\delta) \tag*{(58)}

and its logarithmic derivative at the obstacle (a prime denotes differentiation with respect to the argument),

hδ≡χHδ′(χ)Hδ(χ)=χδI0(χδ)K1(δ)+K0(χδ)I1(δ)I1(χδ)K1(δ)−K1(χδ)I1(δ)−1,χ=aR=1−ϕ.(59)h_\delta\equiv\chi\frac{H_\delta'(\chi)}{H_\delta(\chi)} = \chi\delta\frac{I_0(\chi\delta)K_1(\delta) + K_0(\chi\delta)I_1(\delta)}{I_1(\chi\delta)K_1(\delta) - K_1(\chi\delta)I_1(\delta)} - 1, \qquad\chi= \frac{a}{R} = \sqrt{1-\phi}. \tag*{(59)}

Only the two combinations

Aδ=2−φ+φhδ,Bδ=φ+(2−φ)hδ(60)A_{\delta}=2-\varphi+\varphi h_{\delta},\qquad B_{\delta}=\varphi+(2-\varphi)h_{\delta} \tag*{(60)}

are needed. Direct solution of the velocity problem followed by the force balance gives

Kns,δ∗=Aδπδ2Bδ,Kps,δ∗=2Aδ−φ(χδ)2πδ2[2Bδ−(2−φ)(χδ)2],qδ=2Bδ2Bδ−(2−φ)(χδ)2.(61)K^{*}_{\mathrm{ns},\delta}=\frac{A_{\delta}}{\pi\delta^{2}B_{\delta}},\qquad K^{*}_{\mathrm{ps},\delta}=\frac{2A_{\delta}-\varphi(\chi\delta)^{2}}{\pi\delta^{2}\left[2B_{\delta}-(2-\varphi)(\chi\delta)^{2}\right]},\qquad q_{\delta}=\frac{2B_{\delta}}{2B_{\delta}-(2-\varphi)(\chi\delta)^{2}}. \tag*{(61)}

We note that χδ=a/ℓγ\chi\delta=a/\ell_{\gamma}. Eq. (57) reduces continuously to Eq. (36) in the undamped limit δ→0\delta\to0 except at φ→1\varphi\to1. For every fixed γ>0\gamma>0, momentum is screened over the finite length ℓγ\ell_{\gamma}. As φ→1\varphi\to1 (or a→0a\to0 at fixed RR), the boundary on which odd viscosity affects the traction disappears, the perfect and no-slip passive permeability converge to the same value Kps,δ∗,Kns,δ∗→1/(πδ2)K^{*}_{\mathrm{ps},\delta},K^{*}_{\mathrm{ns},\delta}\to1/(\pi\delta^{2}), and hence Ko,δ∗→0K^{*}_{\mathrm{o},\delta}\to0. The remaining uniform flow is simply U=G/γU=G/\gamma. However, when γ=0\gamma=0, the momentum is unscreened, and the two-dimensional small-obstacle limit is singular with Kps∗−Kns∗=φ2/(8π)K^{*}_{\mathrm{ps}}-K^{*}_{\mathrm{ns}}=\varphi^{2}/(8\pi), which remains finite as φ→1\varphi\to1, even though the even permeability diverges. Thus the limits γ→0\gamma\to0 and φ→1\varphi\to1 do not commute.

We test our theory in Figure S1 using the same parameters as in the main text, except for a finite damping ℓγ≃0.26\ell_{\gamma}\simeq0.26, and find agreement as good as in the main text. Apart from the behavior of KoK_{\mathrm{o}} at φ→0\varphi\to0, another difference from the undamped case is that now the α\alpha at which ∣Ko∣\lvert K_{\mathrm{o}}\rvert is maximal depends on φ\varphi. Panel (g) also shows that damping preserves the passive-response bound when the reference permeabilities are evaluated at the same γ\gamma.

Comparison of theory and simulations for periodic porous media with damping

Figure S1. Comparison of theory and simulations for periodic porous media with damping, ℓγ≃0.26\ell_\gamma\simeq0.26. Panels (a)–(f) show the same quantities as Fig. 2. (g) Normalized response using the passive references of each geometry at the same damping. The dashed semicircle is the circular-cell equality and the gray disk is the passive-response bound.

Three-dimensional odd permeability

Setup

We now consider the three-dimensional extension of the cell construction. For simplicity, we do not include a torque density. In 3D, odd viscosity requires a distinguished axis, which we choose to be e^z\hat{\mathbf{e}}_{z}. Writing ϵijk\epsilon_{ijk} for the three-dimensional Levi-Civita symbol, we define εij≡ϵij3\varepsilon_{ij}\equiv\epsilon_{ij3}, so that ε⋅v=−e^z×v\varepsilon\cdot v=-\hat{\mathbf{e}}_{z}\times v, and introduce α=ηo/η\alpha=\eta_{\mathrm{o}}/\eta. After the same affine-pressure decomposition as in the two-dimensional case, the Stokes equation considered here is

−∇pm+η∇2u−ηoe^z×∇2u+G=0,∇⋅u=0.(62)-\nabla p_{m}+\eta\nabla^{2}u-\eta_{\mathrm{o}}\hat{\mathbf{e}}_{z}\times\nabla^{2}u+G=0,\qquad\nabla\cdot u=0. \tag*{(62)}

The bulk equation does not uniquely determine the local viscosity tensor in 3D. To evaluate the force on the obstacle, we use the tensor structure of Refs. [48, 87]:

Σij=−pδij+2ηEij+2ηo(δikεjℓ+εikδjℓ)Ekℓ,Eij=12(∂iuj+∂jui).(63)\Sigma_{ij}=-p\delta_{ij}+2\eta E_{ij}+2\eta_{\mathrm{o}}\left(\delta_{ik}\varepsilon_{j\ell}+\varepsilon_{ik}\delta_{j\ell}\right)E_{k\ell},\qquad E_{ij}=\frac{1}{2}\left(\partial_{i}u_{j}+\partial_{j}u_{i}\right). \tag*{(63)}

The sign of ηo\eta_{\mathrm{o}} is chosen consistently with the two-dimensional stress in Eq. (3); it is opposite to the convention in those references. The factor of two in the 3D tensor normalization is retained: for an incompressible planar flow, its in-plane odd stress is twice that of Eq. (3) at the same ηo\eta_{\mathrm{o}}. Its divergence gives Eq. (62), with pm=p−ηoωzp_{m}=p-\eta_{\mathrm{o}}\omega_{z} and ω=∇×u\omega=\nabla\times u. The identity

e^z×∇2u=∂zω−∇ωz(64)\hat{\mathbf{e}}_{z}\times\nabla^{2}u=\partial_{z}\omega-\nabla\omega_{z} \tag*{(64)}

then allows us to rewrite Eq. (62) as

−∇q+η∇2u−ηo∂zω+G=0,q=pm−ηoωz=p−2ηoωz.(65)-\nabla q+\eta\nabla^{2}u-\eta_{\mathrm{o}}\partial_{z}\omega+G=0,\qquad q=p_{m}-\eta_{\mathrm{o}}\omega_{z}=p-2\eta_{\mathrm{o}}\omega_{z}. \tag*{(65)}

In contrast with the two-dimensional case, odd viscosity modifies the bulk velocity directly, in addition to its contribution to the boundary traction.

Cell theory

We consider a solid spherical obstacle of radius aa surrounded by an artificial spherical boundary of radius RR. We introduce

χ=aR,φ=1−χ3,β=ξaη.(66)\chi=\frac{a}{R},\qquad\varphi=1-\chi^{3},\qquad\beta=\frac{\xi a}{\eta}. \tag*{(66)}

At the solid obstacle, we impose impermeability and Navier slip. Defining the tangential projector Pt=1−e^re^r\mathbf{P}_t = 1-\hat{\mathbf{e}}_r\hat{\mathbf{e}}_r and the tangential velocity ut=Pt⋅u\mathbf{u}_t = \mathbf{P}_t \cdot\mathbf{u}, the boundary conditions read

u(a)⋅e^r=0,Pt⋅Σ(a)⋅e^r=ξut(a).(67)\mathbf{u}(a) \cdot\hat{\mathbf{e}}_r = 0,\qquad\mathbf{P}_t \cdot\boldsymbol{\Sigma}(a) \cdot\hat{\mathbf{e}}_r = \xi\mathbf{u}_t(a). \tag*{(67)}

At the artificial outer boundary r=Rr=R we impose the radial velocity:

u(R)⋅e^r=U⋅e^r,(68)\mathbf{u}(R) \cdot\hat{\mathbf{e}}_r = \mathbf{U} \cdot\hat{\mathbf{e}}_r, \tag*{(68)}

with U\mathbf{U} again the superficial velocity. We also impose the three-dimensional extension of the Kuwabara closure

e^r×ω(R)=0.(69)\hat{\mathbf{e}}_r \times\boldsymbol{\omega}(R) = 0. \tag*{(69)}

The second condition sets the tangential vorticity to zero. For the passive axisymmetric flow, the vorticity is purely azimuthal, and Eq. (69) reduces to the usual Kuwabara condition [7]. The passive construction does not constrain the radial vorticity, which vanishes identically. We therefore do not additionally impose e^r⋅ω(R)=0\hat{\mathbf{e}}_r \cdot\boldsymbol{\omega}(R) = 0, because odd viscosity generates a toroidal mode with a non-zero radial vorticity and this extra condition would overconstrain the cell problem.

We first recall the passive solution. Choosing the polar axis along U\mathbf{U}, the streamfunction is written as

ψ0(r,θ)=UR2F(s)sin⁡2θ,s=r/R.(70)\psi_0(r,\theta) = U R^2 F(s) \sin^2\theta,\qquad s=r/R. \tag*{(70)}

The two inner conditions at s=χs=\chi, together with the two outer conditions at s=1s=1, become

F(χ)=0,χF′′(χ)−(β+2)F′(χ)=0,F(1)=12,F′′(1)=1.(71)F(\chi)=0,\qquad\chi F^{\prime\prime}(\chi)-(\beta+2)F^{\prime}(\chi)=0,\qquad F(1)=\frac{1}{2},\qquad F^{\prime\prime}(1)=1. \tag*{(71)}

Introducing

Dβ(χ)=β(5−9χ+5χ3−χ6)+3(5−6χ+χ6),(72)D_\beta(\chi)=\beta\left(5-9\chi+5\chi^3-\chi^6\right)+3\left(5-6\chi+\chi^6\right), \tag*{(72)}

we obtain

F(s)=5(βχ3+2β+6)s2−15χ(β+2)s+χ3[5β+(6−2β)χ3]s−1−3χ(β+2)s44Dβ(χ).(73)F(s)=\frac{5(\beta\chi^3+2\beta+6)s^2-15\chi(\beta+2)s+\chi^3\left[5\beta+(6-2\beta)\chi^3\right]s^{-1}-3\chi(\beta+2)s^4}{4D_\beta(\chi)}. \tag*{(73)}

Integrating the passive traction over the obstacle and balancing it with the pressure force gives the passive spherical Kuwabara permeability:

Kα=0(β)=2a2Dβ(χ)45χ3(β+2)=Kns(χ)+Kps(χ)−Kns(χ)1+β/2.(74)K_{\alpha=0}(\beta)=\frac{2a^2D_\beta(\chi)}{45\chi^3(\beta+2)}=K_{\mathrm{ns}}(\chi)+\frac{K_{\mathrm{ps}}(\chi)-K_{\mathrm{ns}}(\chi)}{1+\beta/2}. \tag*{(74)}

where the 3D perfect-slip and no-slip permeabilities are

Kps(χ)=a215χ3(5−6χ+χ6),Kns(χ)=2a245χ3(5−9χ+5χ3−χ6).(75)K_{\mathrm{ps}}(\chi)=\frac{a^2}{15\chi^3}\left(5-6\chi+\chi^6\right),\qquad K_{\mathrm{ns}}(\chi)=\frac{2a^2}{45\chi^3}\left(5-9\chi+5\chi^3-\chi^6\right). \tag*{(75)}

For β→∞\beta\to\infty, Eq. (74) reduces to the usual no-slip spherical Kuwabara permeability [7].

To include odd viscosity, we expand the velocity and modified pressure as

u=u(0)+αu(1)+α2u(2)+O(α3),q=q(0)+αq(1)+α2q(2)+O(α3).(76)\mathbf{u}=\mathbf{u}^{(0)}+\alpha\mathbf{u}^{(1)}+\alpha^2\mathbf{u}^{(2)}+O(\alpha^3),\qquad q=q^{(0)}+\alpha q^{(1)}+\alpha^2q^{(2)}+O(\alpha^3). \tag*{(76)}

The distinguished axis breaks rotational invariance, and the perturbation therefore excites higher spherical-harmonic modes. Although O(α)O(\alpha) already yields an odd permeability, we retain terms through O(α2)O(\alpha^2), where odd viscosity alone makes the permeability anisotropic.

Eq. (65) gives the hierarchy

−∇q(0)+η∇2u(0)+G=0,−∇q(n)+η∇2u(n)=+η∂zω(n−1),n=1,2.(77)-\nabla q^{(0)}+\eta\nabla^2\mathbf{u}^{(0)}+\mathbf{G}=0,\qquad-\nabla q^{(n)}+\eta\nabla^2\mathbf{u}^{(n)}=+\eta\partial_z\boldsymbol{\omega}^{(n-1)},\qquad n=1,2. \tag*{(77)}

The correction fields satisfy homogeneous impermeability and outer conditions. At the obstacle, their tangential boundary conditions are driven by the odd traction generated by the preceding order. We solve this hierarchy using spherical harmonics, separating the case where U\mathbf{U} is parallel or orthogonal to e^z\hat{\mathbf{e}}_z. The passive flow is an ℓ=1\ell=1 poloidal mode. For a longitudinal flow, the first-order correction is an ℓ=2\ell= 2 toroidal mode, while for a transverse flow, finite slip also generates an ℓ=1\ell= 1 poloidal component. The second-order correction contains ℓ=1\ell= 1 and ℓ=3\ell= 3 modes. However, only the ℓ=1\ell= 1 mode contributes to the force applied onto the obstacle by the fluid.

For example, when U=UezU = U\mathbf{e}_z, the first-order velocity can be written as

u(1)=−Uh(s)e^r×∇SP2(cos⁡θ),(78)\mathbf{u}^{(1)}=-U h(s)\mathbf{\hat{e}}_r\times\nabla_S P_2(\cos\theta), \tag*{(78)}

where

h(s)=Aχ,β(s2+32s−3)−Hχ,β6s−1,(79)h(s)=A_{\chi,\beta}\left(s^2+\frac{3}{2}s^{-3}\right)-\frac{H_{\chi,\beta}}{6}s^{-1}, \tag*{(79)}

with

Hχ,β=15χ(β+2)2Dβ(χ),Aχ,β=5χ3[β2+6β−8+(12−2β)χ3]2Dβ(χ)[β(2χ5+3)+12−2χ5].(80)H_{\chi,\beta}=\frac{15\chi(\beta+2)}{2D_\beta(\chi)},\qquad A_{\chi,\beta}=\frac{5\chi^3\left[\beta^2+6\beta-8+(12-2\beta)\chi^3\right]}{2D_\beta(\chi)\left[\beta(2\chi^5+3)+12-2\chi^5\right]}. \tag*{(80)}

Here, P2P_2 is the second Legendre polynomial and ∇S\nabla_S is the surface gradient on the unit sphere. The solution satisfies the Navier condition at s=χs=\chi and [d(sh)/ds]s=1=0[d(sh)/ds]_{s=1}=0, but is nonzero in the fluid domain. Odd viscosity therefore generates a transverse bulk flow even in the no-slip β→∞\beta\to\infty limit.

For compactness, we introduce

Cχ,β=2(1−χ)(1+χ+χ2)3(β+2)(81)C_{\chi,\beta}=\frac{2(1-\chi)(1+\chi+\chi^2)}{3(\beta+2)} \tag*{(81)}

and

X1=15χ(β+6)(1−χ)4(1+χ+χ2)(χ3+3χ2+6χ+5)4Dβ(χ)2,(82)X_1=\frac{15\chi(\beta+6)(1-\chi)^4(1+\chi+\chi^2)(\chi^3+3\chi^2+6\chi+5)}{4D_\beta(\chi)^2}, \tag*{(82)}
X2=−75(β+6)(1−χ)3(1+χ+χ2)34Dβ(χ)2,(83)X_2=-\frac{75(\beta+6)(1-\chi)^3(1+\chi+\chi^2)^3}{4D_\beta(\chi)^2}, \tag*{(83)}
X3=45(β+6)(1−χ)2(1+χ+χ2)(4χ5+4χ4+4χ3−χ2−χ+5)4χDβ(χ)2,(84)X_3=\frac{45(\beta+6)(1-\chi)^2(1+\chi+\chi^2)(4\chi^5+4\chi^4+4\chi^3-\chi^2-\chi+5)}{4\chi D_\beta(\chi)^2}, \tag*{(84)}

together with

Tχ,β=920(3Aχ,βχ+12Aχ,βχ−4−Hχ,βχ−2).(85)T_{\chi,\beta}=\frac{9}{20}\left(3A_{\chi,\beta}\chi+12A_{\chi,\beta}\chi^{-4}-H_{\chi,\beta}\chi^{-2}\right). \tag*{(85)}

The three dimensionless resistance coefficients required below are

co=12−5(β+6)(1−χ)2(1+χ+χ2)22(β+2)Dβ(χ),(86)c_o=\frac{1}{2}-\frac{5(\beta+6)(1-\chi)^2(1+\chi+\chi^2)^2}{2(\beta+2)D_\beta(\chi)}, \tag*{(86)}
c∥=15+3Cχ,β5[Hχ,βχ−3Aχ,β(χ2+4χ−3)],(87)c_{\parallel}=\frac{1}{5}+\frac{3C_{\chi,\beta}}{5}\left[\frac{H_{\chi,\beta}}{\chi}-3A_{\chi,\beta}\left(\chi^2+4\chi^{-3}\right)\right], \tag*{(87)}
c⊥=320+χDβ(χ)45(β+2)(X3−2X2χ−2X1χ2)+Cχ,β(−χTχ,β+2X1χ+X22).(88)c_{\perp}=\frac{3}{20}+\frac{\chi D_\beta(\chi)}{45(\beta+2)}\left(X_3-\frac{2X_2}{\chi}-\frac{2X_1}{\chi^2}\right)+C_{\chi,\beta}\left(-\chi T_{\chi,\beta}+\frac{2X_1}{\chi}+\frac{X_2}{2}\right). \tag*{(88)}

Solving the radial equations in the longitudinal and transverse sectors, integrating the traction, and inverting the resulting resistance tensor yields, after some algebra,

K(α,β)=Kα=0(β)(1−(c⊥+co2)α2−coα0coα1−(c⊥+co2)α20001−c∥α2)+O(α3).(89)K(\alpha,\beta)=K_{\alpha=0}(\beta) \begin{pmatrix} 1-(c_{\perp}+c_o^2)\alpha^2 & -c_o\alpha& 0\\ c_o\alpha& 1-(c_{\perp}+c_o^2)\alpha^2 & 0\\ 0 & 0 & 1-c_{\parallel}\alpha^2 \end{pmatrix} +O(\alpha^3). \tag*{(89)}

For α≠0\alpha\ne0, an odd permeability emerges in the xyxy plane. We retain the convention Ko=Kyx=−KxyK_o=K_{yx}=-K_{xy} used in the two-dimensional Darcy law. Odd viscosity also induces an anisotropic even response, with Kxx=Kyy≠KzzK_{xx}=K_{yy}\ne K_{zz}, even for an isotropic medium.

We can check our results in several limits. For α=0\alpha=0, Eq. (89) reduces exactly to the passive isotropic result in Eq. (74). In the isolated-sphere limit φ→1\varphi\to1, it gives the small-α\alpha translational mobility of a Navier-slip sphere. Taking subsequently β→∞\beta\to\infty where:

co→12,c⊥→320,c∥→15,(90)c_o \to\frac{1}{2}, \qquad c_{\perp} \to\frac{3}{20}, \qquad c_{\parallel} \to\frac{1}{5}, \tag*{(90)}

recovers, after reversing the sign of their odd-viscosity parameter, the no-slip result of Refs. [48, 87]. The full no-slip limit of the permeability is

K(α,β→∞)=Kα=0(β→∞)(1−25α2−α20α21−25α20001−15α2)+O(α3).(91)\mathbf{K}(\alpha,\beta\to\infty)=\mathbf{K}_{\alpha=0}(\beta\to\infty)\begin{pmatrix} 1-\frac{2}{5}\alpha^2 & -\frac{\alpha}{2} & 0 \\ \frac{\alpha}{2} & 1-\frac{2}{5}\alpha^2 & 0 \\ 0 & 0 & 1-\frac{1}{5}\alpha^2 \end{pmatrix}+\mathcal{O}(\alpha^3). \tag*{(91)}

Hence, as expected, in 3D, odd viscosity enters the permeability, even in the no-slip case.

Periodic-cell simulations

We compare the spherical-cell theory with finite-element solutions of Eq. (62) for a stationary sphere in a cubic periodic cell of side LL, with ϕ=1−4πa3/(3L3)\phi=1-4\pi a^3/(3L^3) and the boundary condition (S29). Both permeabilities are normalized by Vc=L2V_c=L^2, matching the total volume VcV_c of the spherical and periodic cells. Figure S2,Figure S2) illustrates the flow at large odd viscosity: a transverse response persists under no slip, while the correction u(α)−u(0)\boldsymbol{u}(\alpha)-\boldsymbol{u}(0) also contains a reduction of the longitudinal flow. For the +x+x drive shown here, the mean flow is deflected towards −y-y. This Hall response is associated with the antisymmetric xyxy block, Kxy=−KyxK_{xy}=-K_{yx}, while Kxz=Kzx=0K_{xz}=K_{zx}=0. The horizontal section shows the resulting sideways deflection; the xzxz projection shows local bending around the sphere but no mean deflection along zz, and omits the transverse yy velocity.

Figure S2. Three-dimensional flow and permeability in a simple-cubic periodic array of spheres, with odd axis e^z\hat{e}_z. (a,b) Full flow u(α)\mathbf{u}(\alpha) and correction Δu=u(α)−u(0)\Delta\mathbf{u}=\mathbf{u}(\alpha)-\mathbf{u}(0) at φ=0.99\varphi=0.99, α=−2\alpha=-2, and no slip, driven by G=Ge^x\mathbf{G}=G\hat{\mathbf{e}}_x. Colors show ∣u∣/umax⁡|\mathbf{u}|/u_{\max} and ∣Δu∣/Δumax⁡|\Delta\mathbf{u}|/\Delta u_{\max} on the central sections z=L/2z=L/2 and y=L/2y=L/2, respectively, where umax⁡=max⁡r∣u(r)∣u_{\max}=\max_{\mathbf{r}}|\mathbf{u}(\mathbf{r})| and Δumax⁡=max⁡r∣Δu(r)∣\Delta u_{\max}=\max_{\mathbf{r}}|\Delta\mathbf{u}(\mathbf{r})| are evaluated over the sampled fluid unit cell. Black curves and arrows follow the velocity projected into each section. The transverse response to this drive is Kyx=−KxyK_{yx}=-K_{xy}. (c)–(e) Symbols: periodic simulations; lines: spherical-cell theory, Eq. (89). (c) Porosity scan at α=−0.1\alpha=-0.1 and no slip. (d) Odd-viscosity scan at φ=0.99\varphi=0.99 and no slip; the perturbative theory is continued over the full range to show its limitations. (e) Friction scan at φ=0.99\varphi=0.99 and α=−0.1\alpha=-0.1; β=0\beta=0 is perfect slip, and the detached β=∞\beta=\infty point is an exact no-slip simulation.

At small ∣α∣\lvert\alpha\rvert and high porosity, the theory captures the diagonal and transverse permeabilities [Figure S2,Figure S2)]. Deviations grow at lower porosity, especially for KxyK_{xy}, reflecting the approximate spherical outer boundary. The wider α\alpha scan also shows the departure from the small-α\alpha expansion. Finally, increasing the boundary friction at fixed negative α\alpha changes KxyK_{xy} from negative at perfect slip to positive at no slip [Figure S2]. Both the periodic simulations and the cell theory predict this sign reversal, so the direction of transverse transport can be reversed by changing slip alone.

Weak inertia and an imposed torque density

In this section, we derive the permeabilities of a porous medium driven by a pressure gradient combined with a torque density while retaining odd viscosity and finite slip. At the Stokes level, the torque and pressure drivings remain additive because the equations are linear. Once convection, and hence inertia, is included, their cross-advection modifies the permeabilities.

Cell problem and torque convention

We consider the dynamics of an incompressible fluid governed by

Iρ(u⋅∇)u=∇⋅Σ(+)+G,∇⋅u=0.(92)\mathcal{I}\rho(\boldsymbol{u}\cdot\nabla)\boldsymbol{u}=\nabla\cdot\boldsymbol{\Sigma}^{(+)}+\boldsymbol{G},\qquad\nabla\cdot\boldsymbol{u}=0. \tag*{(92)}

The bookkeeping parameter I\mathcal{I} controls the inertial expansion and may be set to unity at the end of the calculation. The (+)(+) notation is explained below. The reduced stress tensor is

Σij(+)[u,p]=−pδij+η(∂iuj+∂jui)+ηo(εjk∂iuk+εik∂kuj).(93)\Sigma_{ij}^{(+)}[\boldsymbol{u},p]=-p\delta_{ij}+\eta(\partial_i u_j+\partial_j u_i)+\eta_o(\varepsilon_{jk}\partial_i u_k+\varepsilon_{ik}\partial_k u_j). \tag*{(93)}

The obstacle boundary conditions are

u(r)⋅n^(r)=0,t^(r)⋅Σ(+)(r)⋅n^(r)=ξu(r)⋅t^(r)+τ,r∈∂B.(94)\boldsymbol{u}(\boldsymbol{r})\cdot\hat{\boldsymbol{n}}(\boldsymbol{r})=0,\qquad\hat{\boldsymbol{t}}(\boldsymbol{r})\cdot\boldsymbol{\Sigma}^{(+)}(\boldsymbol{r})\cdot\hat{\boldsymbol{n}}(\boldsymbol{r})=\xi\boldsymbol{u}(\boldsymbol{r})\cdot\hat{\boldsymbol{t}}(\boldsymbol{r})+\boldsymbol{\tau},\qquad\boldsymbol{r}\in\partial B. \tag*{(94)}

Before introducing inertia, consider the case without pressure driving, G=0\boldsymbol{G}=0. The torque produces a finite circulating Stokes field, but a system invariant under a nontrivial rotation R\boldsymbol{R} cannot acquire a mean velocity from τ\boldsymbol{\tau} or ηo\eta_o:

⟨uτ⟩=R⋅⟨uτ⟩⟹⟨uτ⟩=0.(95)\langle\boldsymbol{u}_{\tau}\rangle=\boldsymbol{R}\cdot\langle\boldsymbol{u}_{\tau}\rangle\quad\Longrightarrow\quad\langle\boldsymbol{u}_{\tau}\rangle=0. \tag*{(95)}

This applies to the exact square, hexagonal, and Kagome lattices. A finite random realization need not possess this rotational symmetry and may therefore exhibit a realization-dependent torque-induced current, although its isotropic ensemble average must vanish.

Stokes solution

Take the pressure gradient along xx, G=Gxe^xG = G_x\hat{\mathbf{e}}_x. The complete solution in the Stokes regime is the sum of two physical solutions evaluated at the same values of ηo\eta_o and ξ\xi:

  1. (uG(+),pG(+))(\mathbf{u}^{(+)}_G, p^{(+) }_G) is driven by a unit xx-directed force and satisfies the homogeneous passive slip condition.

  1. (uτ(+),pτ(+))(\mathbf{u}^{(+)}_\tau, p^{(+)}_\tau) has zero body force and unit excess tangential traction.

Both fields satisfy impermeability and periodicity. Linearity of the Stokes problem gives

u(0)=GxuG(+)+τuτ(+).(96)\mathbf{u}^{(0)} = G_x\mathbf{u}^{(+)}_G + \tau\mathbf{u}^{(+)}_\tau. \tag*{(96)}

We then work perturbatively in O(I)O(\mathcal{I}) to capture the effects of the inertia. Hence, the nonlinear solutions are expanded as:

u=u(0)+Iu(1)+O(I2),p=p(0)+Ip(1)+O(I2).(97)\mathbf{u} = \mathbf{u}^{(0)} + \mathcal{I}\mathbf{u}^{(1)} + O(\mathcal{I}^2), \qquad p = p^{(0)} + \mathcal{I}p^{(1)} + O(\mathcal{I}^2). \tag*{(97)}

Substitution into Eq. (S54) gives, at first order,

∇⋅Σ(+)[u(1),p(1)]=ρ(u(0)⋅∇)u(0),∇⋅u(1)=0.(98)\nabla\cdot\boldsymbol{\Sigma}^{(+)}[\mathbf{u}^{(1)},p^{(1)}] = \rho(\mathbf{u}^{(0)}\cdot\nabla)\mathbf{u}^{(0)}, \qquad\nabla\cdot\mathbf{u}^{(1)} = 0. \tag*{(98)}

The first-order boundary conditions are homogeneous:

u(1)⋅n^=0,t^⋅Σ(+)[u(1),p(1)]⋅n^=ξu(1)⋅t^.(99)\mathbf{u}^{(1)}\cdot\hat{\mathbf{n}} = 0, \qquad\hat{\mathbf{t}}\cdot\boldsymbol{\Sigma}^{(+)}[\mathbf{u}^{(1)},p^{(1)}]\cdot\hat{\mathbf{n}} = \xi\mathbf{u}^{(1)}\cdot\hat{\mathbf{t}}. \tag*{(99)}

Pressure-torque cross-advection

Using Eq. (S58), the known first-order forcing is

(u(0)⋅∇)u(0)=Gx2cGG+Gxτc×+τ2cττ,(\mathbf{u}^{(0)}\cdot\nabla)\mathbf{u}^{(0)} = G_x^2\mathbf{c}_{GG} + G_x\tau\mathbf{c}_{\times} + \tau^2\mathbf{c}_{\tau\tau},
cGG=(uG(+)⋅∇)uG(+),cττ=(uτ(+)⋅∇)uτ(+),c×=(uG(+)⋅∇)uτ(+)+(uτ(+)⋅∇)uG(+).(100)\mathbf{c}_{GG} = (\mathbf{u}^{(+)}_G\cdot\nabla)\mathbf{u}^{(+)}_G, \qquad\mathbf{c}_{\tau\tau} = (\mathbf{u}^{(+)}_\tau\cdot\nabla)\mathbf{u}^{(+)}_\tau, \qquad\mathbf{c}_{\times} = (\mathbf{u}^{(+)}_G\cdot\nabla)\mathbf{u}^{(+)}_\tau+ (\mathbf{u}^{(+)}_\tau\cdot\nabla)\mathbf{u}^{(+)}_G. \tag*{(100)}

The pressure-only term is quadratic in GG, and the torque-only term is independent of GG, so neither contributes to the linear permeability. The term c×\mathbf{c}_{\times} is the unique first-order forcing that is linear in both GxG_x and τ\tau.

To isolate this contribution, define (u×,p×)(\mathbf{u}_{\times},p_{\times}) by

∇⋅Σ(+)[u×,p×]=c×,∇⋅u×=0.(101)\nabla\cdot\boldsymbol{\Sigma}^{(+)}[\mathbf{u}_{\times},p_{\times}] = \mathbf{c}_{\times}, \qquad\nabla\cdot\mathbf{u}_{\times} = 0. \tag*{(101)}

This problem satisfies the homogeneous conditions in Eq. (S61) and periodicity on the external cell faces. The factor ρ\rho is not included in the definition of u×\mathbf{u}_{\times} and is restored when the physical first-order field is reconstructed:

u(1)∣Gxτ=ρGxτu×.(102)\left.\mathbf{u}^{(1)}\right|_{G_x\tau} = \rho G_x\tau\mathbf{u}_{\times}. \tag*{(102)}

Darcy law and generalized reciprocity

Using the Lorentz reciprocal theorem adapted to a system with odd viscosity [46], we can find the desired result without having to explicitly find u×\mathbf{u}_{\times}.

We define the “adjoint” reduced stress by changing the sign of the odd viscous term:

Σij(−)[v,q]=−qδij+η(∂ivj+∂jvi)−ηo(ϵjk∂ivk+ϵik∂kvj).(103)\Sigma^{(-)}_{ij}[\mathbf{v},q] = -q\delta_{ij} + \eta(\partial_i v_j + \partial_j v_i) - \eta_o(\epsilon_{jk}\partial_i v_k + \epsilon_{ik}\partial_k v_j). \tag*{(103)}

For any two incompressible fields u\mathbf{u} and v\mathbf{v}, the constitutive tensors obey

(∂jvi)Σij(+)[u,p]=(∂jui)Σij(−)[v,q].(104)(\partial_j v_i)\Sigma^{(+)}_{ij}[\mathbf{u},p] = (\partial_j u_i)\Sigma^{(-)}_{ij}[\mathbf{v},q]. \tag*{(104)}

The pressure contributions vanish by incompressibility, the even-viscous contributions are symmetric under interchange of the two fields, and the odd-viscous contributions are antisymmetric under that interchange. Integration by parts therefore gives the generalized Lorentz relation

∫Ωfv⋅∇⋅Σ(+)[u,p] dr−∫Ωfu⋅∇⋅Σ(−)[v,q] dr=−∫∂Ωf[v⋅Σ(+)[u,p]⋅n^−u⋅Σ(−)[v,q]⋅n^] ds,(105)\int_{\Omega_f} \boldsymbol{v} \cdot\nabla\cdot\boldsymbol{\Sigma}^{(+)}[\boldsymbol{u},p]\,\mathrm{d}r-\int_{\Omega_f} \boldsymbol{u} \cdot\nabla\cdot\boldsymbol{\Sigma}^{(-)}[\boldsymbol{v},q]\,\mathrm{d}r=-\int_{\partial\Omega_f}\left[\boldsymbol{v}\cdot\boldsymbol{\Sigma}^{(+)}[\boldsymbol{u},p]\cdot\hat{\boldsymbol{n}}-\boldsymbol{u}\cdot\boldsymbol{\Sigma}^{(-)}[\boldsymbol{v},q]\cdot\hat{\boldsymbol{n}}\right]\,\mathrm{d}s, \tag*{(105)}

where Ωf\Omega_f is the fluid domain and, in this boundary integral, n^\hat{\boldsymbol{n}} points into the fluid domain. On an obstacle, this agrees with the solid-to-fluid normal defined in the main text; the minus sign accounts for this inward-normal convention.

Let (wj(−),πj(−))(\boldsymbol{w}^{(-)}_j,\pi^{(-)}_j) be the adjoint passive Stokes field driven by a unit pressure-gradient body force in direction jj:

−∇⋅Σ(−)[wj(−),πj(−)]=e^j,∇⋅wj(−)=0,-\nabla\cdot\boldsymbol{\Sigma}^{(-)}[\boldsymbol{w}^{(-)}_j,\pi^{(-)}_j]=\hat{\boldsymbol{e}}_j,\qquad\nabla\cdot\boldsymbol{w}^{(-)}_j=0,
wj(−)⋅n^=0,t^⋅Σ(−)[wj(−),πj(−)]⋅n^=ξwj(−)⋅t^.(106)\boldsymbol{w}^{(-)}_j\cdot\hat{\boldsymbol{n}}=0,\qquad\hat{\boldsymbol{t}}\cdot\boldsymbol{\Sigma}^{(-)}[\boldsymbol{w}^{(-)}_j,\pi^{(-)}_j]\cdot\hat{\boldsymbol{n}}=\xi\boldsymbol{w}^{(-)}_j\cdot\hat{\boldsymbol{t}}. \tag*{(106)}

On every obstacle boundary, the integrands in the boundary contributions to Eq. (105) vanish pointwise. On the external boundary of a periodic cell, opposite faces also cancel pairwise because the fields and reduced stresses are periodic while their outward normals have opposite signs. Using Eqs. (S63) and (S68) therefore gives

∫Ωfwj(−)⋅c× dr=−∫Ωfu×,j dr.(107)\int_{\Omega_f}\boldsymbol{w}^{(-)}_j\cdot\boldsymbol{c}_{\times}\,\mathrm{d}r=-\int_{\Omega_f}u_{\times,j}\,\mathrm{d}r. \tag*{(107)}

Define

Mj≡1A∫Ωfu×,j dr=−1A∫Ωfwj(−)⋅c× dr,(108)M_j\equiv\frac{1}{A}\int_{\Omega_f}u_{\times,j}\,\mathrm{d}r=-\frac{1}{A}\int_{\Omega_f}\boldsymbol{w}^{(-)}_j\cdot\boldsymbol{c}_{\times}\,\mathrm{d}r, \tag*{(108)}

with AA the total periodic-cell area, including the fluid and solid regions. Then using Eqs. (S64) and (S70), we find:

U=KStokesηGx+IρGxτM+O(IGx2,I2),M≡Mx+iMy,KStokes≡KStokes+iKo,Stokes,(109)U=\frac{\mathcal{K}_{\mathrm{Stokes}}}{\eta}G_x+\mathcal{I}\rho G_x\tau M+O(\mathcal{I}G_x^2,\mathcal{I}^2),\qquad M\equiv M_x+iM_y,\qquad\mathcal{K}_{\mathrm{Stokes}}\equiv K_{\mathrm{Stokes}}+iK_{o,\mathrm{Stokes}}, \tag*{(109)}

where KStokes\mathcal{K}_{\mathrm{Stokes}} is the permeability found in the Stokes regime (I=0\mathcal{I}=0). Comparing the term linear in GxG_x with U=(K/η)GxU=(K/\eta)G_x gives

K=KStokes+IρητM+O(I2),(110)K=K_{\mathrm{Stokes}}+\mathcal{I}\rho\eta\tau M+O(\mathcal{I}^2), \tag*{(110)}

where pressure-only terms proportional to Gx2G_x^2 belong to the nonlinear Darcy-Forchheimer response and are excluded from the linear permeability [88]. In general, inertial effects are subtle [89] and here, we only focus on the small contribution from the torque density.

Inertial cell theory

Up to this point, Eq. (108) is exact for a periodic cell. We now approximate the three zero-inertia fields entering its volume integral by their circular-cell counterparts using Kuwabara construction again. The annular domain contains one obstacle of radius aa and an artificial outer boundary of radius RR, so

ϕs≡1−ϕ=a2R2,Ac=πR2,s≡rR,χ≡ϕs.(111)\phi_s\equiv1-\phi=\frac{a^2}{R^2},\qquad A_c=\pi R^2,\qquad s\equiv\frac{r}{R},\qquad\chi\equiv\sqrt{\phi_s}. \tag*{(111)}

The obstacle is located at s=χs=\chi.

Pressure-driven Stokes flow

In the End Matter, we solved the pressure-driven Stokes flow in the Kuwabara cell. We recall the steps here in convenient reduced units. Introduce

z≡β+2iα,α=ηoη,β=ξaη,Ls≡log⁡ϕs.(112)z\equiv\beta+2i\alpha,\qquad\alpha=\frac{\eta_o}{\eta},\qquad\beta=\frac{\xi a}{\eta},\qquad\mathcal{L}_s\equiv\log\phi_s. \tag*{(112)}

For the physical field driven by a unit xx-directed pressure-gradient body force, write

fG(r)=UStokesRFz(s),UStokes≡1πR2∫a<r<R(uG,x(+)+iuG,y(+)) dr=KStokesη.(113)f_G(\mathbf{r})=\mathcal{U}_{\mathrm{Stokes}}RF_z(s),\qquad\mathcal{U}_{\mathrm{Stokes}}\equiv\frac{1}{\pi R^2}\int_{a<r<R}\left(u_{G,x}^{(+)}+iu_{G,y}^{(+)}\right)\,\mathrm{d}r=\frac{\mathcal{K}_{\mathrm{Stokes}}}{\eta}. \tag*{(113)}

The four circular-cell conditions are

Fz(χ)=0,−Fz′′(χ)+1+zχFz′(χ)=0,Fz(1)=1,Fz′′(1)+Fz′(1)−Fz(1)=0.(114)F_z(\chi)=0,\qquad-F_z''(\chi)+\frac{1+z}{\chi}F_z'(\chi)=0,\qquad F_z(1)=1,\qquad F_z''(1)+F_z'(1)-F_z(1)=0. \tag*{(114)}

Their solution was already found to be

Fz(s)=Nz(s)Dz.(115)F_z(s)=\frac{N_z(s)}{D_z}. \tag*{(115)}

with

Nz(s)=2[z(1−ϕs+Ls)+2Ls]s+ϕs[z(ϕs−2)−2ϕs]1s−4(z+2)slog⁡s+(z+2)s3,(116)N_z(s)=2\left[z(1-\phi_s+\mathcal{L}_s)+2\mathcal{L}_s\right]s+\phi_s\left[z(\phi_s-2)-2\phi_s\right]\frac{1}{s}-4(z+2)s\log s+(z+2)s^3, \tag*{(116)}

and

Dz=2(1−ϕs2+2Ls)+z(ϕs2−4ϕs+3+2Ls).(117)D_z=2(1-\phi_s^2+2\mathcal{L}_s)+z(\phi_s^2-4\phi_s+3+2\mathcal{L}_s). \tag*{(117)}

The zero-inertia complex permeability is the previously found

KStokes∗≡KStokesπR2=−Dz16π(z+2)=Kns∗+Kps∗−Kns∗1+β/2+iα.(118)\mathcal{K}_{\mathrm{Stokes}}^*\equiv\frac{\mathcal{K}_{\mathrm{Stokes}}}{\pi R^2}=-\frac{D_z}{16\pi(z+2)}=\mathcal{K}_{\mathrm{ns}}^*+\frac{\mathcal{K}_{\mathrm{ps}}^*-\mathcal{K}_{\mathrm{ns}}^*}{1+\beta/2+i\alpha}. \tag*{(118)}

For the reciprocal integral, it is useful to eliminate DzD_z between FzF_z and U\mathcal{U}:

S^z(s)≡ηfG(Rs)R3=−Nz(s)16(z+2).(119)\widehat{S}_z(s)\equiv\frac{\eta f_G(Rs)}{R^3}=-\frac{N_z(s)}{16(z+2)}. \tag*{(119)}

The adjoint operator has odd viscosity −ηo-\eta_o. Rotational invariance and the reality of all other coefficients therefore give the complex radial amplitudes

fwx(r)=fG(r)‾,fwy(r)=ifG(r)‾.(120)f_{wx}(r)=\overline{f_G(r)},\qquad f_{wy}(r)=i\overline{f_G(r)}. \tag*{(120)}
Torque-driven Stokes flow

The torque-only Stokes solution is axisymmetric and azimuthal:

uτ(+)=hβ(r)e^θ,hβ(r)=Ar+Br.(121)u_\tau^{(+)}=h_\beta(r)\hat{\mathbf{e}}_\theta,\qquad h_\beta(r)=Ar+\frac{B}{r}. \tag*{(121)}

Odd viscosity contributes only to the normal traction of this axisymmetric field and does not enter its tangential boundary condition. For the tangent and normal convention used here,

t^⋅Σ(+)⋅n^=−2ηB/a2,uτ(+)⋅t^=hβ(a).(122)\hat{\mathbf{t}}\cdot\boldsymbol{\Sigma}^{(+)}\cdot\hat{\mathbf{n}}=-2\eta B/a^2,\qquad\mathbf{u}_\tau^{(+)}\cdot\hat{\mathbf{t}}=h_\beta(a). \tag*{(122)}

For the m=0m=0 mode, the Kuwabara closure is imposed on the circulation rather than the vorticity. The circulation at the artificial outer boundary is set to zero to reflect the cancellation of the identical rotational flows generated by neighboring obstacles:

∮r=Ruτ(+)⋅dℓ=0⇒A=−B/R2.(123)\oint_{r=R}\mathbf{u}_\tau^{(+)}\cdot\mathrm{d}\boldsymbol{\ell}=0\Rightarrow A=-B/R^2. \tag*{(123)}

The unit excess-traction condition in Eq. (94) then gives

−2ηBa2=1+ξBa(1−ϕs),B=−a2η[2+β(1−ϕs)].(124)-\frac{2\eta B}{a^2}=1+\frac{\xi B}{a}(1-\phi_s),\qquad B=-\frac{a^2}{\eta[2+\beta(1-\phi_s)]}. \tag*{(124)}

The velocity field in this cell geometry is therefore

hβ(r)=−a2η[2+β(1−ϕs)](1r−rR2).(125)h_\beta(r)=-\frac{a^2}{\eta[2+\beta(1-\phi_s)]}\left(\frac{1}{r}-\frac{r}{R^2}\right). \tag*{(125)}
Periodic reciprocal integral

Let fG=fG,x+ifG,yf_G=f_{G,x}+if_{G,y} and define

A×(r)=hβ(2fG′r−fGr2),B×(r)=fGhβ′r−hβfG′r+hβfGr2.(126)\mathcal{A}_{\times}(r)=h_{\beta}\left(\frac{2f'_G}{r}-\frac{f_G}{r^2}\right),\qquad\mathcal{B}_{\times}(r)=\frac{f_Gh'_{\beta}}{r}-\frac{h_{\beta}f'_G}{r}+\frac{h_{\beta}f_G}{r^2}. \tag*{(126)}

Writing A×=Ax+iAy\mathcal{A}_{\times}=A_x+iA_y and B×=Bx+iBy\mathcal{B}_{\times}=B_x+iB_y, the pressure-torque cross-advection is

c×,r=Axsin⁡θ−Aycos⁡θ,c×,θ=Bxcos⁡θ+Bysin⁡θ.(127)c_{\times,r}=A_x\sin\theta-A_y\cos\theta,\qquad c_{\times,\theta}=B_x\cos\theta+B_y\sin\theta. \tag*{(127)}

Substitution of the adjoint amplitudes in Eq. (120) into Eq. (108), followed by the angular integration, gives the complex response

M=−iR2∫aRr(fGA×r+fG′B×) dr.(128)\mathcal{M}=-\frac{i}{R^2}\int_a^R r\left(\frac{f_G\mathcal{A}_{\times}}{r}+f'_G\mathcal{B}_{\times}\right)\,\mathrm{d}r. \tag*{(128)}

Define

dβ≡2+β(1−ϕs),H(s)≡1s−s.(129)d_{\beta}\equiv2+\beta(1-\phi_s),\qquad H(s)\equiv\frac{1}{s}-s. \tag*{(129)}

Using Eqs. (119) and (S87), Eq. (128) becomes

M=iϕsR4η3dβJz,Jz=∫χ1[S^zS^z′H′−H(S^z′)2+3HsS^zS^z′−Hs2S^z2] ds.(130)\mathcal{M}=i\frac{\phi_sR^4}{\eta^3d_{\beta}}\mathcal{J}_z,\qquad\mathcal{J}_z=\int_{\chi}^{1}\left[\widehat{S}_z\widehat{S}'_zH'-H(\widehat{S}'_z)^2+\frac{3H}{s}\widehat{S}_z\widehat{S}'_z-\frac{H}{s^2}\widehat{S}_z^2\right]\,\mathrm{d}s. \tag*{(130)}

The products in Eq. (130) are ordinary complex products, not modulus squares. Since NzN_z is a linear combination of ss, s−1s^{-1}, slog⁡ss\log s, and s3s^3, the integral is elementary:

Jz=−P(ϕs,z)96(z+2)2,P(ϕs,z)=P0(ϕs)+zP1(ϕs)+z2P2(ϕs).(131)\mathcal{J}_z=-\frac{P(\phi_s,z)}{96(z+2)^2},\qquad P(\phi_s,z)=P_0(\phi_s)+zP_1(\phi_s)+z^2P_2(\phi_s). \tag*{(131)}

The three real functions are

P0(ϕs)=6(log⁡ϕs)2−6log⁡ϕs−16+12ϕs−3ϕs2−24ϕs2log⁡ϕs+4ϕs3+3ϕs4P_0(\phi_s)=6(\log\phi_s)^2-6\log\phi_s-16+12\phi_s-3\phi_s^2-24\phi_s^2\log\phi_s+4\phi_s^3+3\phi_s^4
P1(ϕs)=6(log⁡ϕs)2−30ϕslog⁡ϕs−3ϕs4+13ϕs3−12ϕs2+27ϕs−25P_1(\phi_s)=6(\log\phi_s)^2-30\phi_s\log\phi_s-3\phi_s^4+13\phi_s^3-12\phi_s^2+27\phi_s-25
4P2(ϕs)=6(log⁡ϕs)2+24ϕs2log⁡ϕs−60ϕslog⁡ϕs+6log⁡ϕs+3ϕs4−26ϕs3+51ϕs2−6ϕs−22.(132)4P_2(\phi_s)=6(\log\phi_s)^2+24\phi_s^2\log\phi_s-60\phi_s\log\phi_s+6\log\phi_s+3\phi_s^4-26\phi_s^3+51\phi_s^2-6\phi_s-22. \tag*{(132)}

Therefore,

M=−iϕsR496η3dβP(ϕs,z)(z+2)2.(133)\mathcal{M}=-i\frac{\phi_sR^4}{96\eta^3d_{\beta}}\frac{P(\phi_s,z)}{(z+2)^2}. \tag*{(133)}

The normalized complex permeability is, to first order in inertia,

K∗=KStokes∗+δKI∗+O(I2),(134)\mathcal{K}^*=\mathcal{K}^*_{\mathrm{Stokes}}+\delta\mathcal{K}^*_{\mathrm{I}}+O(\mathcal{I}^2), \tag*{(134)}

where

δKI∗≡IρητπR2M=−iCϕs,βRϕs,z,(135)\delta\mathcal{K}^*_{\mathrm{I}}\equiv\frac{\mathcal{I}\rho\eta\tau}{\pi R^2}\mathcal{M}=-iC_{\phi_s,\beta}R_{\phi_s,z}, \tag*{(135)}

with

Cϕs,β≡ϕsΘ96π2dβ,Rϕs,z≡P(ϕs,z)(z+2)2,Θ≡IρτAcη2.(136)C_{\phi_s,\beta}\equiv\frac{\phi_s\Theta}{96\pi^2d_{\beta}},\qquad R_{\phi_s,z}\equiv\frac{P(\phi_s,z)}{(z+2)^2},\qquad\Theta\equiv\frac{\mathcal{I}\rho\tau A_c}{\eta^2}. \tag*{(136)}

Thus, the controlled first-order even and odd permeabilities are

K∗=Re⁡KStokes∗+Cϕs,βIm⁡Rϕs,z+O(I2),Ko∗=Im⁡KStokes∗−Cϕs,βRe⁡Rϕs,z+O(I2).(137)\mathcal{K}^*=\operatorname{Re}\mathcal{K}^*_{\mathrm{Stokes}}+C_{\phi_s,\beta}\operatorname{Im}R_{\phi_s,z}+O(\mathcal{I}^2),\qquad\mathcal{K}^*_o=\operatorname{Im}\mathcal{K}^*_{\mathrm{Stokes}}-C_{\phi_s,\beta}\operatorname{Re}R_{\phi_s,z}+O(\mathcal{I}^2). \tag*{(137)}
Resummed resistance closure

It is possible to ameliorate the perturbative result of Eq. (137) by considering the resistance, rather than the permeability, and expanding it at first order:

1K∗=1KStokes∗−δKI∗(KStokes∗)2+O(I2).(138)\frac{1}{K^{*}}=\frac{1}{K^{*}_{\mathrm{Stokes}}}-\frac{\delta K^{*}_{I}}{(K^{*}_{\mathrm{Stokes}})^2}+O(\mathcal{I}^{2}). \tag*{(138)}

We can then obtain back the permeability by inverting this result at this order:

Kres∗≡[1KStokes∗−δKI∗(KStokes∗)2]−1=KStokes∗1−Γφs,z,Γφs,z≡δKI∗KStokes∗=iφsΘ6πdβP(φs,z)(z+2)Dz.(139)K^{*}_{\mathrm{res}}\equiv\left[\frac{1}{K^{*}_{\mathrm{Stokes}}}-\frac{\delta K^{*}_{I}}{(K^{*}_{\mathrm{Stokes}})^2}\right]^{-1}=\frac{K^{*}_{\mathrm{Stokes}}}{1-\Gamma_{\varphi_s,z}},\qquad\Gamma_{\varphi_s,z}\equiv\frac{\delta K^{*}_{I}}{K^{*}_{\mathrm{Stokes}}}=i\frac{\varphi_s\Theta}{6\pi d_{\beta}}\frac{\mathcal{P}(\varphi_s,z)}{(z+2)D_z}. \tag*{(139)}

By construction, the resummed expression in Eq. (139) and the direct expansion in Eq. (137) agree through O(I)O(\mathcal{I}). The resummed form is an extrapolation beyond controlled first order, but it remains better behaved at larger inertia and yields better numerical agreement.

Hence, the permeabilities are

KStokes∗=KStokes∗+iKo,Stokes∗,Γφs,z=ΓR+iΓI,(140)K^{*}_{\mathrm{Stokes}}=K^{*}_{\mathrm{Stokes}}+iK^{*}_{o,\mathrm{Stokes}},\qquad\Gamma_{\varphi_s,z}=\Gamma_R+i\Gamma_I, \tag*{(140)}

the resummed components are

Kres∗=KStokes∗(1−ΓR)−Ko,Stokes∗ΓI(1−ΓR)2+ΓI2,Ko,res∗=Ko,Stokes∗(1−ΓR)+KStokes∗ΓI(1−ΓR)2+ΓI2.(141)K^{*}_{\mathrm{res}}=\frac{K^{*}_{\mathrm{Stokes}}(1-\Gamma_R)-K^{*}_{o,\mathrm{Stokes}}\Gamma_I}{(1-\Gamma_R)^2+\Gamma_I^2},\qquad K^{*}_{o,\mathrm{res}}=\frac{K^{*}_{o,\mathrm{Stokes}}(1-\Gamma_R)+K^{*}_{\mathrm{Stokes}}\Gamma_I}{(1-\Gamma_R)^2+\Gamma_I^2}. \tag*{(141)}

For α=β=0\alpha=\beta=0, which is the case studied in the Letter,

Kres∗=KStokes∗1−iΛ,Λ=φsP0(φs)48π[1−φs2+2log⁡(φs)]Θ.(142)K^{*}_{\mathrm{res}}=\frac{K^{*}_{\mathrm{Stokes}}}{1-i\Lambda},\qquad\Lambda=\frac{\varphi_s P_0(\varphi_s)}{48\pi[1-\varphi_s^2+2\log(\varphi_s)]}\Theta. \tag*{(142)}

Figure S3 shows how the theory deteriorates as the torque increases and how the resummed theory improves upon the naive first-order result. We recall that the simulations are done by enforcing a stationary chiral Navier–Stokes equation. At large Reynolds numbers, the resulting permeability may therefore differ from that obtained in time-dependent simulations, where turbulence may develop.

Figure S3. (a) Finite-drive permeability estimates, normalized by the circular-cell KStokesK_{\mathrm{Stokes}}, and (b) relative error against theory. At fixed τ\tau, the simulations use Gx=±GG_x=\pm G and extract (K,Ko)=η[U(+G)−U(−G)]/(2G)(K,K_o)=\eta[U(+G)-U(-G)]/(2G). The horizontal axis is Re⁡τ=ρUτ,rmsAc/η\operatorname{Re}_{\tau}=\rho U_{\tau,\mathrm{rms}}\sqrt{A_c}/\eta, with Uτ,rms=[Ac−1∫Ac∣uτ∣2 dr]1/2U_{\tau,\mathrm{rms}}=[A_c^{-1}\int_{A_c}|\mathbf{u}_{\tau}|^2\,d\mathbf{r}]^{1/2} and u‾τ=[u(+G)+u(−G)]/2\overline{\mathbf{u}}_{\tau}=[\mathbf{u}(+G)+\mathbf{u}(-G)]/2, an estimate of the torque-only flow. The lattice is hexagonal, φ=0.90\varphi=0.90, α=β=0\alpha=\beta=0, and ρGAc3/2/η2≃0.0403\rho G A_c^{3/2}/\eta^2\simeq0.0403.

On the use of the reciprocal relation

The reciprocal-theorem expression for the permeability is derived under periodic boundary conditions, which ensure that the boundary integral vanishes. Our cell geometry is not periodic, so this contribution cannot be discarded a priori. Nevertheless, direct evaluation of the boundary terms shows that it vanishes in this case.

References

  1. [1]Darcy, H. (1856). Les Fontaines publiques de la ville de Dijon.
  2. [2]Torquato, S. (2002). Random Heterogeneous Materials. In Interdisciplinary Applied Mathematics (Vol. 16).
  3. [3]U. Hornung, ed., Homogenization and Porous Media, Interdisciplinary Applied Mathematics, Vol. 6 (Springer, New York, NY, 1997).DOI
  4. [4]Sahimi, M. (2011). Flow and Transport in Porous Media and Fractured Rock: From Classical Methods to Modern Approaches.
  5. [5]S. Whitaker, The Method of Volume Averaging, Theory and Applications of Transport in Porous Media, Vol. 13 (Springer Netherlands, Dordrecht, 1999).DOI
  6. [6]A. S. Sangani and A. Acrivos, Slow flow past periodic arrays of cylinders with application to heat transfer, International Journal of Multiphase Flow 8, 193 (1982).DOI
  7. [7]S. Kuwabara, The forces experienced by randomly distributed parallel circular cylinders or spheres in a viscous flow at small reynolds numbers, Journal of the physical society of Japan 14, 527 (1959).DOI
  8. [8]Nield, D. A., & Bejan, A. (2006). Convection in porous media.
  9. [9]Goflier, F., Zarcone, C., Bazin, B., Lenormand, R., Lasseux, D., & Quintard, M. (2002). On the ability of a darcy-scale model to capture wormhole formation during the dissolution of a porous medium. Journal of fluid Mechanics, 457(213).
  10. [10]Freeze, R. A., & Cherry, J. A. (1979). Groundwater.
  11. [11]Dake, L. P. (1978). Fundamentals of Reservoir Engineering (Vol. 8).
  12. [12]V. C. Mow, S. C. Kuei, W. M. Lai, and C. G. Armstrong, Biphasic creep and stress relaxation of articular cartilage in compression: Theory and experiments, Journal of Biomechanical Engineering 102, 73 (1980).pubmed.ncbi.nlm.nih.gov/7382457
  13. [13]C. Pozrikidis and D. A. Farrow, A model of fluid flow in solid tumors, Annals of Biomedical Engineering 31, 181 (2003).pubmed.ncbi.nlm.nih.gov/12627826
  14. [14]T. Roose and M. A. Swartz, Multiscale modeling of lymphatic drainage from tissues using homogenization theory, Journal of Biomechanics 45, 107 (2012).DOI
  15. [15]H. E. Schreder, J. Liu, D. H. Kelley, J. H. Thomas, and K. A. S. Boster, A hydraulic resistance model for interstitial fluid flow in the brain, Journal of the Royal Society Interface 19, 20210812 (2022).DOI
  16. [16]A. H.-D. Cheng, Poroelasticity, 1st ed., Theory and Applications of Transport in Porous Media, Vol. 27 (Springer International Publishing, Cham, 2016).DOI
  17. [17]M. C. Marchetti, J.-F. Joanny, S. Ramaswamy, T. B. Liverpool, J. Prost, M. Rao, and R. A. Simha, Hydrodynamics of soft active matter, Reviews of modern physics 85, 1143 (2013).DOI
  18. [18]C. Bechinger, R. Di Leonardo, H. Löwen, C. Reichhardt, G. Volpe, and G. Volpe, Active particles in complex and crowded environments, Reviews of modern physics 88, 045006 (2016).DOI
  19. [19]R. Alert and X. Trepat, Physical models of collective cell migration, Annual Review of Condensed Matter Physics 11, 77 (2020).DOI
  20. [20]D. Needleman and Z. Dogic, Active matter at the interface between materials science and cell biology, Nature Reviews Materials 2, 17048 (2017).DOI
  21. [21]J. Prost, F. Jülicher, and J.-F. Joanny, Active gel physics, Nature Physics 11, 111 (2015).DOI
  22. [22]J. Elgeti, R. G. Winkler, and G. Gompper, Physics of microswimmers—single particle motion and collective behavior: a review, Reports on progress in physics 78, 056601 (2015).DOI
  23. [23]R. Alonso-Matilla, B. Chakrabarti, and D. Saintillan, Transport and dispersion of active particles in periodic porous media, Physical Review Fluids 4, 043101 (2019).DOI
  24. [24]K. J. Modica, Y. Xi, and S. C. Takatori, Porous Media Microstructure Determines the Diffusion of Active Matter: Experiments and Simulations, Frontiers in Physics 10 (2022).DOI
  25. [25]C. Reichhardt and C. J. Olson Reichhardt, Active matter transport and jamming on disordered landscapes, Phys- ical Review E 90, 012701 (2014).pubmed.ncbi.nlm.nih.gov/25122329
  26. [26]T. G. J. Chandler and S. E. Spagnolie, Active nematic response to a deformable body or boundary: Elastic deformations and anchoring-induced flow, Phys. Rev. Res. 7, 023263 (2025).DOI
  27. [27]J. Shea, G. Jung, and F. Schmid, How boundary interactions dominate emergent driving of inertial passive probes in active matter, Journal of Physics A: Mathematical and Theoretical 57, 235006 (2024).DOI
  28. [28]S. Bhattacharyya and J. M. Yeomans, Active sorting to boundaries in active nematic–passive isotropic fluid mixtures, Soft Matter 21, 7641 (2025).DOI
  29. [29]P. Das, M. Residori, A. Voigt, S. Mandal, and C. Kurzthaler, Vorticity-induced surfing and trapping in porous media, arXiv:2511.02471 (2025).arXiv
  30. [30]F. Mackay, J. Toner, A. Morozov, and D. Marenduzzo, Darcy’s Law without Friction in Active Nematic Rheology, Physical Review Letters 124, 187801 (2020).DOI
  31. [31]J. P. Vaidya, T. N. Shendruk, and S. P. Thampi, Active nematics in corrugated channels, Soft Matter 20, 8230 (2024).DOI
  32. [32]R. R. Keogh, T. Kozhukhov, K. Thijssen, and T. N. Shendruk, Active Darcy’s Law, Physical Review Letters 132, 188301 (2024).DOI
  33. [33]Mecke, J., Gao, Y., Ramírez Medina, C. A., Aarts, D. G., Gompper, G., & Ripoll, M. (2023). Simultaneous emergence of active turbulence and odd viscosity in a colloidal chiral active system. Communications Physics, 6, 324.
  34. [34]C. W. Chan, D. Wu, K. Qiao, K. L. Fong, Z. Yang, Y. Han, and R. Zhang, Chiral active particles are sensitive reporters to environmental geometry, Nature communications 15, 1406 (2024).DOI
  35. [35]J. Katuri, N. Kaur, W. Uspal, A. Cornelius, D. Quashie Jr, and J. Ali, Control of colloidal cohesive states in active chiral fluids, Communications Physics 7, 291 (2024).DOI
  36. [36]P. Chen, S. Weady, S. Atis, T. Matsuzawa, M. J. Shelley, and W. T. M. Irvine, Self-propulsion, flocking and chiral active phases from particles spinning at intermediate Reynolds numbers, Nature Physics 21, 146 (2025).DOI
  37. [37]T. H. Tan, A. Mietke, J. Li, Y. Chen, H. Higinbotham, P. J. Foster, S. Gokhale, J. Dunkel, and N. Fakhri, Odd dynamics of living chiral crystals, Nature 607, 287 (2022).DOI
  38. [38]Chao, Y.-C., Gokhale, S., Lin, L., Hastewell, A., Bacanu, A., Chen, Y., Li, J., Liu, J., Lee, H., Dunkel, J., & al., . et . (2026). Selective excitation of work-generating cycles in non-reciprocal living solids. Nature Physics, 1.
  39. [39]R. Maire, A. Petrini, U. Marini Bettolo Marconi, and L. Caprini, Kinetic theory of chiral active disks: Odd transport and torque density, The Journal of Chemical Physics 165, 034501 (2026).DOI
  40. [40]Lier, R., Patus, M., & Chapman-enskog. (2026). Chapman-enskog expansion for chirally colliding disks. arXiv:2602.21367
  41. [41]E. Eren, M. Fruchart, and V. Vitelli, A collisional model of odd fluids: from boltzmann equation to chiral hydrodynamics, arXiv:2508.12944 (2025).arXiv
  42. [42]C. Hargus, J. M. Epstein, and K. K. Mandadapu, Odd diffusivity of chiral random motion, Physical Review Letters 127, 178001 (2021).DOI
  43. [43]T. Markovich and T. C. Lubensky, Chiral active fluids: Insights from the total momentum, Physical Review E 112, 035409 (2025).DOI
  44. [44]M. Fruchart, C. Scheibner, and V. Vitelli, Odd viscosity and odd elasticity, Annual Review of Condensed Matter Physics 14, 471 (2023).DOI
  45. [45]Y. Hosaka, S. Komura, and D. Andelman, Nonreciprocal response of a two-dimensional fluid with odd viscosity, Physical Review E 103, 042610 (2021).DOI
  46. [46]Y. Hosaka, R. Golestanian, and A. Vilfan, Lorentz reciprocal theorem in fluids with odd viscosity, Physical Review Letters 131, 178303 (2023).DOI
  47. [47]R. Lier, C. Duclut, S. Bo, J. Armas, F. Jülicher, and P. Suŕowka, Lift force in odd compressible fluids, Physical Review E 108, L023101 (2023).DOI
  48. [48]R. Lier, Odd viscous flow past a sphere at low but non-zero reynolds numbers, Journal of Fluid Mechanics 998, A40 (2024).DOI
  49. [49]A. Daddi-Moussa-Ider, Y. Hosaka, and S. Komura, Analytical response functions for a compressible thin fluid layer with odd viscosity, New Journal of Physics (2026).arxiv.org/abs/2602.18136
  50. [50]Hosaka, Y., Chatzitofi, M., Golestanian, R., & Vilfan, A. (2024). Chirotactic response of microswimmers in fluids with odd viscosity. Physical Review Research, 6, L032044.
  51. [51]F.-H. Meng, J.-L. Liu, and Y.-L. He, Transport of the obstacle arrays driven by chiral active particles with temperature difference, Journal of Physics A: Mathematical and Theoretical 53, 095005 (2020).DOI
  52. [52]E. Irani, Z. Mokhtari, and A. Zippelius, Dynamics of bacteria scanning a porous environment, Physical Review Letters 128, 144501 (2022).DOI
  53. [53]D. M. Van Roon, G. Volpe, M. M. Telo da Gama, and N. A. Araújo, The role of disorder in the motion of chiral active particles in the presence of obstacles, Soft Matter 18, 6899 (2022).DOI
  54. [54]F. Siebers, A. Jayaram, P. Blümler, and T. Speck, Exploiting compositional disorder in collectives of light-driven circle walkers, Science Advances 9, eadf5443 (2023).DOI
  55. [55]J. Mecke, Y. Gao, and M. Ripoll, Obstacle-tuned transition from chaotic to coherent vortex flows and odd diffusion in chiral active fluids, Physics of Fluids 37 (2025).arxiv.org/abs/2508.15210
  56. [56]J. Mecke, Y. Gao, and M. Ripoll, Confinement-induced pumping of chiral active fluids, arXiv:2609.12611 (2026).arXiv
  57. [57]J. Uhlig, J. Wójcik, R. Metzler, and E. Kalz, Chirality enhances diffusion in disordered environments, The Journal of Chemical Physics 165, 124118 (2026).DOI
  58. [58]S. Ganeshan and A. G. Abanov, Odd viscosity in two-dimensional incompressible fluids, Physical review fluids 2, 094101 (2017).DOI
  59. [59]R. Lier, Slip-induced odd viscous flow past a cylinder, Physical Review Fluids 9, 094101 (2024).DOI
  60. [60]J. Metzger, C. Hargus, J. Tailleur, and F. van Wijland, Equation of state for the edge flow of chiral colloidal fluids, Physical Review Letters (2026), accepted for publication.arxiv.org/abs/2604.18708
  61. [61]V. Yashunsky, D. Pearce, C. Blanch-Mercader, F. Ascione, P. Silberzan, and L. Giomi, Chiral edge current in nematic cell monolayers, Physical Review X 12, 041017 (2022).DOI
  62. [62]Y. Kuroda, E. Meyberg, G. Gardi, T. Speck, and S. Osat, Designing topological edge currents in chiral active matter, arXiv:2606.31840 (2026).arXiv
  63. [63]U. Marini Bettolo Marconi, A. Petrini, R. Maire, and L. Caprini, Emergent hydrodynamics of chiral active fluids: vortices, bubbles and odd diffusion, New Journal of Physics 28, 064401 (2026).arxiv.org/abs/2601.19432
  64. [64]Caporusso, C. B., Gonnella, G., & Levis, D. (n.d.). Phase co-.
  65. [65]L. Caprini, M. Musacchio, U. M. B. Marconi, B. Liebchen, and H. Löwen, Active thermodynamics of inertial chiral active gases: Equation of state and edge currents, Proceedings of the National Academy of Sciences 123, e2532158123 (2026).DOI
  66. [66]B. Wang, P. Pietzonka, and F. Jülicher, Edge currents shape condensates in chiral active matter, Physical Review Research 8, 043093 (2026).DOI
  67. [67]Z.-F. Huang, M. t. Vrugt, R. Wittkowski, and H. Löwen, Anomalous grain dynamics and grain locomotion of odd crystals, Proc. Natl. Acad. Sci. U.S.A. 122, e2511350122 (2025).DOI
  68. [68]A. Petrini, R. Maire, U. M. B. Marconi, and L. Caprini, Curvature-driven wall accumulation in chiral active particles, Soft Matter 22, 6543 (2026).DOI
  69. [69]S. Bureković, S. Kole, A. Maitra, and C. Nardini, Bulk and microphase separation in chiral active systems, arXiv:2607.12045 (2026).arXiv
  70. [70]F. V. Reyes, Steady base states in a two-dimensional chiral fluid. the chiral stokes cavity, arXiv:2607.24279 (2026).arXiv
  71. [71]X. Lou, Q. Yang, Y. Ding, P. Liu, K. Chen, X. Zhou, F. Ye, R. Podgornik, and M. Yang, Odd viscosity-induced hall-like transport of an active chiral fluid, Proceedings of the National Academy of Sciences 119, e2201279119 (2022).DOI
  72. [72]D. Reynolds, G. M. Monteiro, and S. Ganeshan, Hele-shaw flow for parity odd three-dimensional fluids, Physical Review Fluids 7, 114201 (2022).DOI
  73. [73]D. Banerjee, A. Souslov, A. G. Abanov, and V. Vitelli, Odd viscosity in chiral active fluids, Nature communications 8, 1573 (2017).DOI
  74. [74]D. Lasseux and F. J. Valdés-Parada, Darcy’s law survival from no-slip to perfect-slip flow in porous media, Journal of Fluid Mechanics 997, A65 (2024).DOI
  75. [75]Y. Hosaka, S. Komura, and D. Andelman, Hydrodynamic lift of a two-dimensional liquid domain with odd viscosity, Physical Review E 104, 064613 (2021).DOI
  76. [76]Y. Hosaka, R. Golestanian, and A. Daddi-Moussa-Ider, Hydrodynamics of an odd active surfer in a chiral fluid, New Journal of Physics 25, 083046 (2023).DOI
  77. [77]I. Abdoli, H. D. Vuijk, J.-U. Sommer, J. M. Brader, and A. Sharma, Nondiffusive fluxes in a brownian system with lorentz force, Physical review E 101, 012120 (2020).DOI
  78. [78]See the Supplemental Material for the energy bound on odd-viscous permeability, external damping, three-dimensional transport, and the weak-inertia derivation and validation, including Refs. [7, 46, 48, 87–89].. (n.d.). See the Supplemental Material for the energy bound on odd-viscous permeability, external damping, three-dimensional transport, and the weak-inertia derivation and validation, including Refs. [7, 46, 48, 87–89]..
  79. [79]R. Maire and I. Pagonabarraga, Kinetic and hydrodynamic theories of chiral intruder dynamics in nonequilibrium baths, arXiv:2607.23223 (2026).arXiv
  80. [80]X. M. de Wit, M. Fruchart, T. Khain, F. Toschi, and V. Vitelli, Pattern formation by turbulent cascades, Nature 627, 515 (2024).DOI
  81. [81]X. M. de Wit, L. Touzo, S. Galtier, M. Fruchart, F. Toschi, and V. Vitelli, Wave turbulence theory of odd fluids and solids: kinetic equations and solutions, arXiv:2606.14583 (2026).arXiv
  82. [82]X. M. de Wit, L. Touzo, S. Galtier, M. Fruchart, F. Toschi, and V. Vitelli, Odd wave turbulence, arXiv:2504.15403 (2026).arXiv
  83. [83]Geuzaine, C., & Remacle, J.-F. (2009). Gmsh: a three-dimensional finite element mesh generator with built-in pre- and post-processing facilities. International Journal for Numerical Methods in Engineering, 79.
  84. [84]D. Boffi, F. Brezzi, M. Fortin, et al., Mixed finite element methods and applications, Vol. 44 (Springer, 2013).DOI
  85. [85]P. Virtanen, R. Gommers, T. E. Oliphant, M. Haberland, T. Reddy, D. Cournapeau, E. Burovski, P. Peterson, W. Weckesser, J. Bright, et al., SciPy 1.0: fundamental algorithms for scientific computing in python, Nature methods 17, 261 (2020).DOI
  86. [86]C. T. Kelley, Solving nonlinear equations with Newton’s method (Society for Industrial and Applied Mathematics, 2003).DOI
  87. [87]J. G. Everts and B. Cichocki, Dissipative effects in odd viscous stokes flow around a single sphere, Physical Review Letters 132, 218303 (2024).DOI
  88. [88]D. Lasseux and F. J. Valdés-Parada, On the developments of darcy’s law to include inertial and slip effects, Comptes Rendus. Mécanique 345, 660–669 (2017).DOI
  89. [89]Lasseux, D., Abbassian Arani, A. A., & Ahmadi, A. (2011). On the stationary macroscopic inertial effects for one phase flow in ordered and disordered porous media. Physics of Fluids, 23, 073103.

Paper details

vlnc:W_ngah4ua7
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 9

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