betaFeedback

WorksPhysics and Astronomy

Odd Capillarity

Abstract

Capillarity governs interfacial relaxation and hydrodynamic motion in binary fluids through parity-even mechanical stresses. Here, we introduce odd capillarity, a parity-breaking interfacial mechanical response, by incorporating a symmetric, nonvariational, parity-odd capillary stress into the Cahn–Hilliard–Navier–Stokes (CHNS) equations. This stress generates forces along interfaces, transverse to ordinary capillary forces, and performs mechanical work that is not balanced by changes in the composition free energy. Direct numerical simulations reveal that odd capillarity arrests domain coarsening while sustaining turbulent fluid motion without a conventional inertial energy cascade, as energy injection is predominantly balanced by viscous dissipation. In asymmetric mixtures, odd capillarity driven deformation and breakup interrupt droplet growth, producing dynamically arrested emulsions with composition hyperuniformity and suppressed large-scale vorticity fluctuations. These findings establish odd capillarity as a distinct interfacial mechanism for controlling phase separation and sustaining nonequilibrium fluid motion.

Introduction: Capillarity converts the geometry of a fluid interface into mechanical stress [1, 2]. This fundamental coupling governs a broad range of interfacial phenomena, from phase separation and wetting to droplet coalescence and breakup, and capillary waves [3, 4, 5, 6, 7, 8]. By favoring the reduction of interfacial free energy, capillary forces drive interfacial relaxation and generate flows in the surrounding fluid, which in turn advect and deform the interfaces. This basic interplay is captured by the Cahn–Hilliard–Navier–Stokes (CHNS) equations, or Model H, in which the capillary stress couples composition gradients to fluid momentum [9, 10, 11, 12, 13]. Derived from the free energy, this stress preserves spatial parity symmetry.

The mechanical response of a material can change fundamentally when spatial reflection symmetry (parity) is broken. In chiral fluids, broken parity permits odd viscosity, a nondissipative transverse stress response [14, 15]. In active solids and mechanical metamaterials, non-reciprocal interactions can give rise to odd elasticity, where elastic stresses are not derivable from a scalar energy and can perform work over cyclic deformations [16, 17]. These developments [18] have primarily concerned bulk mechanical responses, whereas parity-odd capillary stresses at fluid interfaces remain comparatively unexplored. Recent theories of chiral phase separation have demonstrated persistent interfacial currents and their influence on domain morphology [19, 20, 21, 22]. These theories, however, describe dry systems, where chirality drives tangential currents of particles or composition along interfaces without momentum-conserving hydrody-

namics. This raises a fundamental question: If capillarity itself acquires a parity-odd component, how does it reshape interfacial dynamics through its coupling to fluid motion?

In this Letter, we introduce odd capillarity, a parity-breaking interfacial mechanical response, and develop a minimal hydrodynamic model incorporating a nonvariational, parity-odd capillary stress within the CHNS equations. This stress generates a force transverse to the even capillary force without introducing an intrinsic torque density. Unlike the conventional capillary coupling, the odd stress can perform mechanical work that is not balanced by changes in the binary-fluid free energy. By direct numerical simulations (DNSs) we explore the possibility of odd capillarity to arrest domain coarsening and study its impact on turbulent fluid motion. We perform a scale-by-scale kinetic-energy analysis to understand the impact of odd capillarity and analyse the vorticity field with respect to hyperuniformity. Our findings establish odd capillarity as a distinct mechanism through which fluid interfaces can organize and sustain nonequilibrium motion.

Model: We consider an incompressible binary fluid in two dimensions, described by a conserved composition field ϕ(x,t)\phi(\mathbf{x},t), a velocity field u(x,t)\mathbf{u}(\mathbf{x},t) and a pressure field p(x,t)p(\mathbf{x},t). The composition distinguishes the two coexisting phases, while its spatial gradients identify the diffuse interfaces separating them. We formulate a minimal hydrodynamic model by extending the CHNS equations [9, 10, 11, 12, 13], retaining the usual conservation laws and introducing parity breaking solely through an additional interfacial mechanical stress. The governing equations read

∂tϕ+u⋅∇ϕ=M∇2μ,(1)\partial_t\phi+\mathbf{u}\cdot\nabla\phi=M\nabla^2\mu, \tag*{(1)}
∂tu+u⋅∇u=−∇p+ν∇2u−ϕ∇μ+∇⋅Σodd,(2)\partial_t\mathbf{u}+\mathbf{u}\cdot\nabla\mathbf{u}=-\nabla p+\nu\nabla^2\mathbf{u}-\phi\nabla\mu+\nabla\cdot\mathbf{\Sigma}^{\mathrm{odd}}, \tag*{(2)}

with ∇⋅u=0\nabla\cdot\mathbf{u} = 0. Here MM is the mobility, and ν\nu the kinematic viscosity. We assume equal densities and viscosities in both phases, with the density set to unity. The chemical potential μ=δF/δϕ\mu= \delta F/\delta\phi follows from the standard Ginzburg–Landau free energy

F[ϕ]=∫Ω[3σ16ϵ(ϕ2−1)2+3σϵ4∣∇ϕ∣2] dx,(3)F[\phi] = \int_{\Omega} \left[\frac{3\sigma}{16\epsilon}(\phi^2 - 1)^2 + \frac{3\sigma\epsilon}{4}|\nabla\phi|^2\right]\,\mathrm{d}x, \tag*{(3)}

where σ\sigma is the equilibrium surface tension and ϵ\epsilon sets the diffuse-interface thickness. Thus, the phase-separation thermodynamics remains unchanged and the additional nonequilibrium ingredient ∇⋅Σodd\nabla\cdot\boldsymbol{\Sigma}^{\mathrm{odd}} enters only through the stress.

In ordinary CHNS equations, the capillary force −ϕ∇μ-\phi\nabla\mu derives from the same free energy and can be expressed, up to an isotropic pressure contribution, as the divergence of the Korteweg stress,

Σijeven=−32σϵ[∂iϕ∂jϕ−12∣∇ϕ∣2δij].(4)\Sigma_{ij}^{\mathrm{even}} = -\frac{3}{2}\sigma\epsilon\left[\partial_i\phi\partial_j\phi- \frac{1}{2}|\nabla\phi|^2\delta_{ij}\right]. \tag*{(4)}

This stress is symmetric and parity even. Its mechanical work is compensated by changes in the binary-fluid free energy, so that capillary forces redistribute energy between interfacial and hydrodynamic degrees of freedom.

To construct an odd interfacial response, we exploit the geometry intrinsic to a diffuse interface. The composition gradient ∇ϕ\nabla\phi defines its local normal direction, while a rotation by 90∘90^\circ gives the corresponding tangent. These two directions allow an interfacial stress to distinguish longitudinal and transverse mechanical responses without introducing an independent orientational field. We therefore seek a stress that is quadratic in composition gradients, symmetric under interchange of tensor indices, and odd under spatial reflection. A minimal choice satisfying these requirements is

Σijodd=32σ~ϵ[ϵik∂kϕ∂jϕ+ϵjk∂kϕ∂iϕ],(5)\Sigma_{ij}^{\mathrm{odd}} = \frac{3}{2}\tilde{\sigma}\epsilon\left[\epsilon_{ik}\partial_k\phi\partial_j\phi+ \epsilon_{jk}\partial_k\phi\partial_i\phi\right], \tag*{(5)}

where ϵij\epsilon_{ij} is the two-dimensional Levi–Civita tensor, with ϵ12=1\epsilon_{12} = 1. The coefficient σ~\tilde{\sigma} determines the strength and handedness of odd capillarity and has the dimensions of surface tension. For σ~=0\tilde{\sigma} = 0, the model reduces to the classical CHNS equations.

Several distinguishing mechanical properties follow from this construction. The stress is symmetric and traceless, and therefore introduces no intrinsic antisymmetric torque density. It vanishes in homogeneous phases and is concentrated in the interfacial regions where ∇ϕ≠0\nabla\phi\ne0. Most importantly, its divergence generates a force perpendicular to the passive capillary force:

fodd=2χe^z×feven,χ=σ~/σ,(6)\mathbf{f}_{\mathrm{odd}} = 2\chi\hat{\mathbf{e}}_z \times\mathbf{f}_{\mathrm{even}}, \qquad\chi= \tilde{\sigma}/\sigma, \tag*{(6)}

where feven=∇⋅Σeven\mathbf{f}_{\mathrm{even}} = \nabla\cdot\boldsymbol{\Sigma}^{\mathrm{even}} and fodd=∇⋅Σodd\mathbf{f}_{\mathrm{odd}} = \nabla\cdot\boldsymbol{\Sigma}^{\mathrm{odd}}. The dimensionless chirality parameter χ\chi measures the relative strength of odd capillarity, with χ=0\chi= 0 recovering the achiral limit and its sign determining the handedness of the transverse response. Thus, the same composition gradients that generate ordinary capillary forcing also support an independent, parity-breaking mechanical response. For a planar interface parallel to the xx-axis, with ϕ=ϕ(y)\phi= \phi(y), the odd stress produces a tangential force density fxodd=∂yΣxyoddf_x^{\mathrm{odd}} = \partial_y\Sigma_{xy}^{\mathrm{odd}}, forming a localized force-dipole layer characterized by ∫−∞∞fxodd dy=0\int_{-\infty}^{\infty} f_x^{\mathrm{odd}}\,\mathrm{d}y = 0, ∫−∞∞yfxodd dy=−σ~\int_{-\infty}^{\infty} y f_x^{\mathrm{odd}}\,\mathrm{d}y = -\tilde{\sigma}, where the second equality follows from the equilibrium diffuse-interface profile [see Appendix § 1 for the derivations]. The resulting force-dipole layer drives tangential shear across the interface, while the symmetry of the underlying stress ensures the absence of intrinsic torque density.

The proposed coupling is distinct from other nonequilibrium mechanical responses. In Active Model H [23], an independent active stress modifies the parity-even interfacial coupling and can oppose ordinary surface tension. Its chiral extension has been developed to describe layered chiral active matter, where it yields odd elastic responses [24]. In this Letter, we instead consider a parity-odd capillary stress acting at the interfaces of a phase-separating binary fluid. This also makes it distinct from odd viscosity, which is a parity-breaking constitutive response to velocity gradients [15].

The energetic consequences of the additional stress distinguish odd capillarity from passive interfacial relaxation. We define the mean kinetic energy e=⟨∣u∣2⟩x/2e = \langle|\mathbf{u}|^2\rangle_x/2, the mean free energy f=F/∣Ω∣f = \mathcal{F}/|\Omega|, and the total mean energy E=e+f\mathcal{E} = e + f, where ⟨⋅⟩x\langle\cdot\rangle_x denotes a spatial average. For periodic boundary conditions, the corresponding energy balances are

dedt=Peven+Podd−Dν,(7)\frac{\mathrm{d}e}{\mathrm{d}t} = P_{\mathrm{even}} + P_{\mathrm{odd}} - D_\nu, \tag*{(7)}
dfdt=−Peven−Dμ,(8)\frac{\mathrm{d}f}{\mathrm{d}t} = -P_{\mathrm{even}} - D_\mu, \tag*{(8)}
dEdt=Podd−Dν−Dμ,(9)\frac{\mathrm{d}\mathcal{E}}{\mathrm{d}t} = P_{\mathrm{odd}} - D_\nu- D_\mu, \tag*{(9)}

where Peven=⟨u⋅feven⟩xP_{\mathrm{even}} = \langle\mathbf{u}\cdot\mathbf{f}_{\mathrm{even}}\rangle_x and Podd=⟨u⋅fodd⟩xP_{\mathrm{odd}} = \langle\mathbf{u}\cdot\mathbf{f}_{\mathrm{odd}}\rangle_x are the powers transferred to the fluid by even and odd capillary forces, respectively. The viscous and diffusive dissipation rates are Dν=−ν⟨u⋅∇2u⟩xD_\nu= -\nu\langle\mathbf{u}\cdot\nabla^2\mathbf{u}\rangle_x and Dμ=−M⟨∣∇μ∣2⟩xD_\mu= -M\langle|\nabla\mu|^2\rangle_x, respectively, with Dν,Dμ≥0D_\nu,D_\mu\ge0 [see Appendix § 2 for detailed derivations]. Unlike the passive capillary contribution, whose mechanical work is balanced by changes in the composition free energy, the odd stress produces an independent power term PoddP_{\mathrm{odd}}, which can have either sign. The model therefore permits sustained mechanical energy injection through stresses localized at fluid interfaces, even though the composition dynamics retains its conventional Cahn–Hilliard form.

Phase ordering and coarsening arrest: We first investigate how odd capillarity modifies the phase-ordering dynamics of a binary fluid. We perform DNS of Eqs. (1) and (2) using a Fourier pseudospectral method [13] in a two-dimensional periodic square of side 2π2\pi, with 102421024^2 collocation points [see Appendix § 4 for numerical details]. Starting from a nearly homogeneous mixture, ϕ(x,0)=ϕ0+δϕ(x)\phi(\boldsymbol{x},0)=\phi_0+\delta\phi(\boldsymbol{x}), where ϕ0\phi_0 is the mean composition and δϕ\delta\phi is a small-amplitude random perturbation, we examine the evolution of the characteristic domain size. We first consider a symmetric mixture, ϕ0=0\phi_0=0, and define the coarsening length scale as

L(t)=2π∑kS(k,t)∑kkS(k,t),(10)L(t)=2\pi\frac{\sum_k S(k,t)}{\sum_k kS(k,t)}, \tag*{(10)}

where S(k,t)=∑k≤∣k′∣<k+Δk∣δϕ^(k′,t)∣2S(k,t)=\sum_{k\leq|\boldsymbol{k}'|<k+\Delta k}|\widehat{\delta\phi}(\boldsymbol{k}',t)|^2 is the shell-integrated composition spectrum, with δϕ=ϕ−ϕ0\delta\phi=\phi-\phi_0. Here, ϕ^(k′,t)\widehat{\phi}(\boldsymbol{k}',t) denotes the spatial Fourier coefficient of ϕ\phi, k′\boldsymbol{k}' is the wavevector, kk is the wavenumber defining the shell, and Δk\Delta k is its width. Figure 1(a) shows the temporal evolution of L(t)L(t) for different values of chirality parameter χ\chi, while Figure 1(b)–(Figure 1e) display representative pseudocolor plots of ϕ\phi. In the achiral limit (χ=0\chi=0), the mixture undergoes spinodal decomposition, forming interconnected domains that grow through diffusive transport and hydrodynamic motion [Fig. 1(b)]. In the observed scaling regime, L(t)∼tL(t)\sim t, consistent with the classical viscous hydrodynamic growth law of binary-fluid mixtures [3, 25]. Introducing odd capillarity significantly alters these dynamics. For χ=0.5\chi=0.5 and 11, coarsening continues but becomes progressively slower, with the growth rate decreasing as χ\chi increases. The corresponding interfaces exhibit pronounced deformations and irregular domain morphologies [Fig. 1(c)]. Upon increasing χ\chi to 22, the characteristic length ceases to grow systematically

Figure 1. Odd capillarity slows and arrests domain coarsening. (a) Evolution of the characteristic coarsening length L(t)L(t) for different chirality parameter χ=σ~/σ\chi= \tilde{\sigma}/\sigma, with σ=1\sigma= 1. For χ=0\chi= 0 (black solid line), the ordinary CHNS model exhibits approximately linear hydrodynamic coarsening, L(t)∼tL(t) \sim t (black dotted line). Increasing χ\chi progressively suppresses domain growth, with coarsening arrest observed for χ=2\chi= 2, 44, and 1010. The arrested length decreases with increasing χ\chi. (b)–(e) Representative pseudocolor plots of the composition field ϕ\phi: (b) χ=0\chi= 0 at t=4t = 4; (c)–(e) χ=1\chi= 1, 22, and 44, respectively, at t=20t = 20, illustrating the increasingly fine domain structures.

and fluctuates around a statistically stationary value, indicating coarsening arrest. At stronger odd capillarity (χ=4,10\chi=4,10), the arrested state persists with a further reduction in the characteristic domain size [Figs. 1(d) and 1(e)]. In the arrested state, the interfaces keep evolving through diffusive growth, reconnection, neck thinning, and breakup, and spiral-like structures form. The state is dynamically sustained rather than frozen. The free-energy balance [Eq. 7b] provides an energetic interpretation of this behaviour. In a non-equilibrium steady state, ⟨Peven⟩t=−⟨Dμ⟩t<0\langle P_{\mathrm{even}}\rangle_t=-\langle D_\mu\rangle_t<0. Thus, on average, the flow must continuously do work against capillarity and regenerate interfacial free energy at the rate it is dissipated by diffusion. In contrast, for conventional CHNS equations in the Stokes limit, Peven=Dν≥0P_{\mathrm{even}}=D_\nu\geq0, and the free energy decreases monotonically. Odd capillarity therefore sustains a dynamically arrested state in which interfaces continually evolve despite the absence of systematic domain growth. Coarsening arrest in hydrodynamically coupled binary fluids has previously been observed under externally driven turbulent flows [26, 27, 28], through contractile interfacial stresses in Active Model H [25, 29, 30], and through bulk active flows [31, 32]. Arrested phase separation has also been reported in active-spinner fluids and non-reciprocal binary-fluid turbulence [33, 34]. The described mechanism differs from these examples as odd capillarity suppresses coarsening through self-generated tangential interfacial flows and the resulting hydrodynamic deformation of domains.

Odd-capillarity-driven turbulence: The persistent interfacial deformations and coarsening arrest raise a fundamental question about the origin of the accompanying fluid motion. Can odd capillarity itself sustain complex flows in the absence of appreciable inertia or external forcing? To address this question, we examine the kinetic-energy spectrum and the mechanisms of energy transfer across scales. The shell-integrated kinetic-energy spectrum, E(k,t)=∑k≤∣k′∣<k+Δk∣u^(k′,t)∣2E(k,t) = \sum_{k \leq|\mathbf{k}'| < k+\Delta k} |\hat{\mathbf{u}}(\mathbf{k}',t)|^2, satisfies the scale-by-scale energy balance [[35, 36] [37]

∂tE(k,t)=Tu(k,t)+Teven(k,t)+Todd(k,t)−Dν(k,t),(11)\partial_t E(k,t) = T^u(k,t) + T^{\mathrm{even}}(k,t) + T^{\mathrm{odd}}(k,t) - D_\nu(k,t), \tag*{(11)}

where TuT^u, TevenT^{\mathrm{even}}, and ToddT^{\mathrm{odd}} denote the contributions from nonlinear advection (inertia), even capillary forcing, and odd-capillary forcing, respectively, and DνD_\nu is the viscous dissipation spectrum [see Appendix § 3 for the definitions]. Hereafter, E(k)=⟨E(k,t)⟩tE(k) = \langle E(k,t)\rangle_t, Tα(k)=⟨Tα(k,t)⟩tT^\alpha(k) = \langle T^\alpha(k,t)\rangle_t, and Dν(k)=⟨Dν(k,t)⟩tD_\nu(k) = \langle D_\nu(k,t)\rangle_t denote their statistically stationary time averages, with α=u\alpha= u, even, odd. In the dynamically arrested regime, the kinetic energy fluctuates around a statistically stationary mean [Figure 2(b)], while its spectrum extends over a broad range of wavenumbers [Figure 2(c)], with E(k)∼k−0.5E(k) \sim k^{-0.5} for k<2π/Lck < 2\pi/L_c and

Figure 2. Odd-capillarity-driven turbulence. (a) Representative pseudocolor plot of the scalar vorticity field ω=(∇×u)⋅e^z\omega= (\nabla\times\mathbf{u}) \cdot\hat{\mathbf{e}}_z, normalized by its root-mean-square value ωrms\omega_{\mathrm{rms}}, for χ=4\chi= 4 at t=400t = 400. (b) Time evolution of the spatially averaged kinetic energy e(t)e(t) for χ=2\chi= 2, 44, and 1010. (c) Time-averaged kinetic-energy spectra E(k)E(k), normalized by ⟨e(t)⟩t\langle e(t)\rangle_t, for the same values of χ\chi. The dotted and dashed lines indicate the reference scalings k−1/2k^{-1/2} and k−2k^{-2}, respectively. The spectra are averaged over 400 snapshots in the statistically stationary state. (d) Spectral contributions Tu(k)T^u(k), Teven(k)T^{\mathrm{even}}(k), Todd(k)T^{\mathrm{odd}}(k), and Dν(k)D_\nu(k) to the kinetic-energy balance [Eq. (11)] for χ=4\chi= 4, plotted on a logarithmic wavenumber axis. The inertial transfer is negligible, while odd-capillary energy injection is predominantly balanced by viscous dissipation; the even capillary contribution is approximately two orders of magnitude smaller. (e) Mean odd-capillary power ⟨Podd⟩t\langle P_{\mathrm{odd}}\rangle_t, viscous dissipation ⟨Dν⟩t\langle D_\nu\rangle_t, and even capillary power −⟨Peven⟩t-\langle P_{\mathrm{even}}\rangle_t versus χ\chi. The dotted line indicates the scaling ⟨Podd⟩t∼χ1.45\langle P_{\mathrm{odd}}\rangle_t \sim\chi^{1.45}. The ratio ∣⟨Peven⟩t∣/⟨Podd⟩t∼10−2|\langle P_{\mathrm{even}}\rangle_t|/\langle P_{\mathrm{odd}}\rangle_t \sim10^{-2}. (f) Probability density functions of the local odd-capillary power podd(x,t)=u⋅foddp_{\mathrm{odd}}(\mathbf{x},t) = \mathbf{u} \cdot\mathbf{f}_{\mathrm{odd}}, normalized by its root-mean-square value, for χ=2\chi= 2, 44, and 1010.

E(k)∼k−2E(k) \sim k^{-2} for k>2π/Lck > 2\pi/L_c, where LcL_c is the characteristic arrested domain size. The scale-by-scale energy balance reveals that Todd(k)T^{\mathrm{odd}}(k) is predominantly positive, demonstrating that odd capillarity injects mechanical energy directly into the fluid across a range of scales [Figure 2(d)]. The injected energy is balanced primarily by viscous dissipation Dν(k)D_\nu(k), while the nonlinear advective transfer Tu(k)T^u(k) is negligible. The even capillary contribution is comparatively weak and changes sign with wavenumber, while the nonlinear advective transfer Tu(k)T^u(k) is negligible. In the non-equilibrium stationary state, ⟨∂tE(k,t)⟩t=0\langle\partial_t E(k,t)\rangle_t = 0, so Eq. 9 gives Todd(k)+Teven(k)≃Dν(k)T^{\mathrm{odd}}(k) + T^{\mathrm{even}}(k) \simeq D_\nu(k). Since the even capillary contribution is comparatively small, Todd(k)≃Dν(k)T^{\mathrm{odd}}(k) \simeq D_\nu(k), indicating that energy injected by odd capillarity is dissipated predominantly at the same scales by viscosity, without appreciable transfer through an inertial cascade.

Since the advective transfer only redistributes energy among scales, ∑kTu(k)=0\sum_k T^u(k) = 0 for incompressible periodic flow, summing Eq. 9 over all wavenumbers recovers the global kinetic-energy balance Eq. 7a, with e(t)=∑kE(k,t)e(t) = \sum_k E(k,t), Podd(t)=∑kTodd(k,t)P_{\mathrm{odd}}(t) = \sum_k T^{\mathrm{odd}}(k,t), Peven(t)=∑kTeven(k,t)P_{\mathrm{even}}(t) = \sum_k T^{\mathrm{even}}(k,t), and Dν(t)=∑kDν(k,t)D_\nu(t) = \sum_k D_\nu(k,t). Figure 2(e) shows the time-averaged terms in the non-equilibrium steady state. We find ⟨Podd⟩t>0\langle P_{\mathrm{odd}}\rangle_t > 0 and ⟨Peven⟩t<0\langle P_{\mathrm{even}}\rangle_t < 0, which indicates that odd capillarity injects kinetic energy globally, while the even capillary force converts a small fraction of it into composition free energy, −⟨Peven⟩t/⟨Podd⟩t≈10−2-\langle P_{\mathrm{even}}\rangle_t/\langle P_{\mathrm{odd}}\rangle_t \approx10^{-2}. This fraction balances diffusive dissipation, ⟨Peven⟩t=−⟨Dμ⟩t<0\langle P_{\mathrm{even}}\rangle_t = -\langle D_\mu\rangle_t < 0, as required by the free-energy balance [Eq. 7b]. The remainder is dissipated by viscosity, so that ⟨Podd⟩t=⟨Dν⟩t+⟨Dμ⟩t≃⟨Dν⟩t\langle P_{\mathrm{odd}}\rangle_t = \langle D_\nu\rangle_t + \langle D_\mu\rangle_t \simeq\langle D_\nu\rangle_t. The mean odd capillary power scales as ⟨Podd⟩t∼χ1.45\langle P_{\mathrm{odd}}\rangle_t \sim\chi^{1.45}. Although odd capillarity injects energy on average, the local power density podd(x,t)=u⋅foddp_{\mathrm{odd}}(\mathbf{x}, t) = \mathbf{u} \cdot\mathbf{f}_{\mathrm{odd}} fluctuates between positive and negative values. Its probability density P(podd/podd,rms)\mathcal{P}(p_{\mathrm{odd}}/p_{\mathrm{odd,rms}}) exhibits tails of both signs, with a stronger positive tail [Figure 2(f)]. The distributions broaden with increasing χ\chi, indicating increasingly pronounced relative fluctuations in local energy transfer. Thus, odd capillarity acts as a net energy source while locally both injecting and extracting kinetic energy. In contrast to externally sustained interface-induced turbulence in viscous binary fluids [38], the present turbulent dynamics are powered directly by parity-odd stresses at evolving fluid interfaces.

Arrested droplet growth and hyperuniformity: Having established coarsening arrest and sustained fluid motion in symmetric mixtures, we now examine phase separation in asymmetric mixtures with ϕ0=0.1,0.2,0.3,0.4,\phi_0 = 0.1, 0.2, 0.3, 0.4, and 0.50.5. Figure 3(a) shows a representative composition field for ϕ0=0.4\phi_0 = 0.4 and χ=4\chi= 4, revealing a dynamically evolving droplet-rich state. Unlike passive binary-fluid mixtures, where droplets grow through Ostwald ripening and coalescence, odd capillarity interrupts their continued growth. Although smaller droplets shrink and disappear while larger ones grow through Ostwald ripening, the latter undergo interfacial instabilities that lead to deformation and breakup. The associated velocity fields [Figure 3(b) and (c)] reveal strong tangential flows near the droplet interfaces, which also hinder coalescence between neighboring droplets. Thus, the competition between diffusive growth, coalescence, and odd capillarity driven breakup maintains a dynamically arrested droplet-rich state.

Figure 3. Arrested droplet growth and hyperuniformity. All results are for χ=4\chi= 4. (a) Representative pseudocolor plot of the composition field ϕ\phi for ϕ0=0.4\phi_0 = 0.4 at t=300t = 300. (b) Corresponding pseudocolor plot of the velocity magnitude ∣u∣|\mathbf{u}|, overlaid with velocity vectors. (c) Magnified view of an individual droplet in (b), highlighting the interfacial flow and velocity vectors. (d) Time evolution of the spatially averaged kinetic energy e(t)e(t) for ϕ0=0\phi_0 = 0, 0.10.1, 0.20.2, 0.30.3, 0.40.4, and 0.50.5. (e) Time-averaged vorticity spectral density Ψω(k)\Psi_\omega(k) for different ϕ0\phi_0, showing suppressed long-wavelength fluctuations and the emergence of hyperuniformity. The reference slopes indicate the scalings k1/2k^{1/2}, kk, and k4k^4. (f) Time-averaged composition spectral density Ψϕ(k)\Psi_\phi(k) for the same values of ϕ0\phi_0. The reference slopes k0k^0 and kk illustrate the transition from nonhyperuniform to hyperuniform composition fluctuations. The color legend in (e) applies to (d)–(f).

Figure 3(d) shows that the mean kinetic energy e(t)e(t) approaches a statistically stationary state for different ϕ0\phi_0. Its temporal fluctuations become progressively weaker with increasing ϕ0\phi_0, accompanied by increasingly localized flows associated with individual droplets. To characterize the resulting large-scale vorticity fluctuations, we examine its hyperuniformity, which is characterized by anomalously suppressed large scale fluctuations [39]. We consider the time-averaged spectral density of the vorticity field, ω=(∇×u)⋅e^z\omega= (\nabla\times\mathbf{u}) \cdot\hat{\mathbf{e}}_z, defined as Ψω(k)=⟨1Nk∑k≤∣k′∣<k+Δk∣ω^(k′,t)∣2⟩t\Psi_\omega(k) = \left\langle\frac{1}{N_k} \sum_{k \leq|\mathbf{k}'| < k+\Delta k} |\hat{\omega}(\mathbf{k}',t)|^2 \right\rangle_t, where NkN_k is the number of Fourier modes in the shell k≤∣k′∣<k+Δkk \leq|\mathbf{k}'| < k+\Delta k. In two dimensions, Ψω(k)∝kE(k)\Psi_\omega(k) \propto k E(k) [40], which relates the vorticity spectral density to the kinetic-energy spectrum. For symmetric mixtures (ϕ0=0\phi_0 = 0), we find Ψω(k)∼k0.5\Psi_\omega(k) \sim k^{0.5} at small wavenumbers [Fig. 3(e)], consistent with E(k)∼k−0.5E(k) \sim k^{-0.5} discussed earlier [Fig. 2(c)]. With increasing ϕ0\phi_0, this scaling gives way to two distinct regimes, Ψω(k)∼k\Psi_\omega(k) \sim k at smaller wavenumbers and Ψω(k)∼k4\Psi_\omega(k) \sim k^4 at relatively higher wavenumbers. These positive exponents indicate hyperuniform vorticity fluctuations, with increasingly strong suppression at large scales.

The composition field also exhibits a marked change in its large-scale fluctuations. Figure 3(f) shows the time-averaged spectral density Ψϕ(k)\Psi_\phi(k), defined analogously to Ψω(k)\Psi_\omega(k). For symmetric mixtures, Ψϕ(k)∼k0\Psi_\phi(k) \sim k^0 at small wavenumbers, indicating nonhyperuniform composition fluctuations. In contrast, the droplet-rich states exhibit Ψϕ(k)∼k\Psi_\phi(k) \sim k, revealing the emergence of composition hyperuniformity. As vortical flows become increasingly localized around individual droplets, their spatial arrangement influences the large-scale distribution of vorticity, providing a physical connection between the hyperuniformity of the composition and vorticity fields. Hydrodynamically interrupted droplet growth has also been reported in Active Model H with contractile interfacial stresses [41]. However, such activity can suppress composition hyperuniformity and produce anti-hyperuniform vorticity fluctuations, Ψω(k)∼k−2/3\Psi_\omega(k) \sim k^{-2/3} [25]. In contrast, odd capillarity arrests droplet growth while promoting composition hyperuniformity and suppressing large-scale vorticity fluctuations.

Conclusions and outlook: We have shown that capillarity admits a parity-odd counterpart, odd capillarity, through which fluid interfaces generate transverse forces and sustain non-equilibrium motion. We introduced this response within the CHNS equations by supplementing the ordinary Korteweg stress with a symmetric, parity-odd capillary stress. The resulting localized force-dipole structure drives interfacial shear flows that, in turn, deform and reorganize the interfaces themselves. This feedback fundamentally changes phase separation, progressively slowing and ultimately arresting coarsening.

We have identified a sustained energy-transfer cycle underlying coarsening arrest. Odd-capillarity stresses inject kinetic energy into the fluid, most of which is dissipated by viscosity, while a smaller fraction is transferred through ordinary capillary stresses to replenish the composition free energy lost through diffusion. This energy transfer sustains interfacial rearrangements and prevents continued domain growth. We have further demonstrated that the arrested state exhibits odd-capillary-driven turbulence, characterized by persistent kinetic-energy fluctuations and broadband kinetic energy spectra.

Our scale-by-scale analysis reveals that energy injection is predominantly balanced by viscous dissipation at the same scales, with negligible inertial transfer. In asymmetric mixtures, the same interfacial feedback interrupts droplet growth: although Ostwald ripening persists, growing droplets deform and break under odd-capillarity forcing, while tangential interfacial flows hinder coalescence. The resulting dynamically arrested droplet-rich states exhibit composition hyperuniformity and strongly suppressed large-scale vorticity fluctuations.

Odd capillarity opens new possibilities for revisiting classical capillary phenomena in systems with broken parity. Unlike chiral surface waves arising from bulk odd viscosity [42] or odd slip at chiral active boundaries [43], parity breaking here enters directly through the capillary stress. A natural next step is to develop a systematic sharp-interface theory to establish the mechanisms of interfacial deformation and instability. Such a theory would provide a foundation for exploring whether odd capillarity can destabilize interfaces even in the absence of inertia, how it modifies capillary waves and Rayleigh–Plateau breakup [44], and how it interacts with external forces. In particular, the motion of rising or falling droplets under gravity presents an intriguing problem, complementing studies of deformable droplets with bulk odd viscosity [45]. Deriving odd capillarity from microscopic models of chiral active constituents remains an open challenge. Coarse-graining approaches that have yielded interfacial tension and odd surface flows [19] and odd transport coefficients [46] for chiral active particles offer natural routes. Experiments provide different routes to generating chiral flows near fluid interfaces. In a colloidal fluid driven by external torques, the particles develop persistent edge flows, while the free surface exhibits waves and instabilities [47]. Another possibility involves motile constituents interacting directly with fluid interfaces. Bacteria adsorbed at oil–water interfaces can drive droplet rotation, with angular speed decreasing with droplet size [48]. Bacteria trapped at fluid interfaces also exhibit circular swimming trajectories [49] and generate localized hydrodynamic force dipoles [50]. These observations motivate investigating whether microscopic chirality can produce the parity-odd interfacial stress proposed here, providing a possible experimental route to odd capillarity. These observations suggest possible experimental routes to odd capillarity, although whether microscopic chirality generates the proposed parity-odd interfacial stress remains to be established. Incorporating substrate friction, which is relevant to many experimental systems, would extend our model to settings with momentum loss. More broadly, from ordinary surface tension in passive fluids to negative mechanical tension in active fluids [41], capillarity has acquired new roles in interfacial dynamics. Odd capillarity introduces a distinct possibility: fluid interfaces that retain their ordinary surface tension while generating handed, tangential flows that inject mechanical energy and organize the surrounding fluid.

Acknowledgments: We acknowledges the support by the German Research Foundation (DFG) through the project “Analysing structure–property relations in equilibrium and non-equilibrium hyperuniform systems” (project number VO 899/32-1). We further acknowledge computing resources at JSC under grant ”MORPH” and at ZIH under grant ”WIR”.

Tangential force-dipole structure of odd capillarity

To characterize the interfacial forcing generated by the odd stress, consider a planar interface parallel to the xx-axis, with ϕ=ϕ0(y)\phi= \phi_0(y). For ϵxy=1\epsilon_{xy}=1, the odd stress reduces to

Σodd=32σ~ϵ[ϕ0′(y)]2(0110).(12)\boldsymbol{\Sigma}^{\mathrm{odd}} = \frac{3}{2}\widetilde{\sigma}\epsilon[\phi_0'(y)]^2 \begin{pmatrix} 0 & 1 \\ 1 & 0 \end{pmatrix}. \tag*{(12)}

Its divergence yields a purely tangential force density,

fxodd=32σ~ϵddy[ϕ0′(y)]2,fyodd=0.(13)f_x^{\mathrm{odd}} = \frac{3}{2}\widetilde{\sigma}\epsilon\frac{d}{dy}[\phi_0'(y)]^2,\qquad f_y^{\mathrm{odd}}=0. \tag*{(13)}

For the equilibrium diffuse-interface profile ϕ0(y)=tanh⁡[y/(2ϵ)]\phi_0(y)=\tanh[y/(2\epsilon)], this becomes

fxodd(y)=−3σ~4ϵ2sech⁡4(y2ϵ)tanh⁡(y2ϵ).(14)f_x^{\mathrm{odd}}(y)=-\frac{3\widetilde{\sigma}}{4\epsilon^2}\operatorname{sech}^4\left(\frac{y}{2\epsilon}\right)\tanh\left(\frac{y}{2\epsilon}\right). \tag*{(14)}

The force density is antisymmetric with respect to the interface. Its zeroth moment vanishes, whereas its first normal moment remains finite:

∫−∞∞fxodd dy=0,∫−∞∞yfxodd dy=−σ~.(15)\int_{-\infty}^{\infty}f_x^{\mathrm{odd}}\,dy=0,\qquad\int_{-\infty}^{\infty}y f_x^{\mathrm{odd}}\,dy=-\widetilde{\sigma}. \tag*{(15)}

The latter follows by integration by parts and the identity ∫−∞∞[ϕ0′(y)]2 dy=2/(3ϵ)\int_{-\infty}^{\infty}[\phi_0'(y)]^2\,dy=2/(3\epsilon). Thus, the odd stress produces a localized tangential force-dipole layer with zero resultant force and a nonzero first moment.

Derivation of the energy balances

We derive the kinetic-, free-, and total-energy balances [Eqs. (7)] for incompressible flow with periodic boundary conditions. The derivation also applies to boundary conditions for which the corresponding boundary terms vanish. We define the spatial average as

⟨g⟩x=1∣Ω∣∫Ωg(x) dx.(16)\langle g\rangle_x=\frac{1}{|\Omega|}\int_{\Omega}g(x)\,dx. \tag*{(16)}

The mean kinetic energy, mean free energy, and total mean energy are

e=12⟨∣u∣2⟩x,f=F[ϕ]∣Ω∣,E=e+f,(17)e=\frac{1}{2}\langle|\mathbf{u}|^2\rangle_x,\qquad f=\frac{\mathcal{F}[\phi]}{|\Omega|},\qquad\mathcal{E}=e+f, \tag*{(17)}

where the density is constant and set to ρ=1\rho=1. The chemical potential is defined by μ=δF/δϕ\mu=\delta\mathcal{F}/\delta\phi.

Free-energy balance. Since μ=δF/δϕ\mu=\delta\mathcal{F}/\delta\phi, the time derivative of the mean free energy is

dfdt=⟨μ ∂tϕ⟩x.(18)\frac{df}{dt}=\langle\mu\,\partial_t\phi\rangle_x. \tag*{(18)}

Substituting the Cahn–Hilliard equation,

∂tϕ=−u⋅∇ϕ+M∇2μ,(19)\partial_t\phi=-\mathbf{u}\cdot\nabla\phi+M\nabla^2\mu, \tag*{(19)}

gives

dfdt=−⟨μ u⋅∇ϕ⟩x+M⟨μ∇2μ⟩x=⟨ϕ u⋅∇μ⟩x−M⟨∣∇μ∣2⟩x,(20)\frac{df}{dt}=-\langle\mu\,\mathbf{u}\cdot\nabla\phi\rangle_x+M\langle\mu\nabla^2\mu\rangle_x =\langle\phi\,\mathbf{u}\cdot\nabla\mu\rangle_x-M\langle|\nabla\mu|^2\rangle_x, \tag*{(20)}

where we have used incompressibility, periodicity, and integration by parts. The ordinary capillary force appearing in the Navier–Stokes equation is −ϕ∇μ-\phi\nabla\mu. Its mean mechanical power is

Peven=−⟨ϕ u⋅∇μ⟩x.(21)P_{\mathrm{even}}=-\langle\phi\,\mathbf{u}\cdot\nabla\mu\rangle_x. \tag*{(21)}

Equivalently, since −ϕ∇μ-\phi\nabla\mu and feven=∇⋅Σeven\mathbf{f}_{\mathrm{even}}=\nabla\cdot\boldsymbol{\Sigma}^{\mathrm{even}} differ only by a gradient, whose mechanical work vanishes for incompressible periodic flow,

Peven=⟨u⋅feven⟩x.(22)P_{\mathrm{even}}=\langle\mathbf{u}\cdot\mathbf{f}_{\mathrm{even}}\rangle_x. \tag*{(22)}

Defining the nonnegative diffusive dissipation rate as

Dμ=M⟨∣∇μ∣2⟩x,(23)D_\mu=M\langle|\nabla\mu|^2\rangle_x, \tag*{(23)}

we obtain

dfdt=−Peven−Dμ.(24)\frac{df}{dt}=-P_{\mathrm{even}}-D_\mu. \tag*{(24)}

Kinetic-energy balance. Multiplying the Navier–Stokes equation [Eq. (2)] by u\mathbf{u} and taking the spatial average gives

dedt=−⟨u⋅∇p⟩x+ν⟨u⋅∇2u⟩x+⟨u⋅feven⟩x+⟨u⋅fodd⟩x.(25)\frac{de}{dt}=-\langle\mathbf{u}\cdot\nabla p\rangle_x+\nu\langle\mathbf{u}\cdot\nabla^2\mathbf{u}\rangle_x +\langle\mathbf{u}\cdot\mathbf{f}_{\mathrm{even}}\rangle_x+\langle\mathbf{u}\cdot\mathbf{f}_{\mathrm{odd}}\rangle_x. \tag*{(25)}

where the nonlinear advective contribution vanishes because

⟨u⋅(u⋅∇)u⟩x=⟨u⋅∇∣u∣22⟩x=⟨∇⋅(∣u∣22u)⟩x=0.(26)\langle\mathbf{u}\cdot(\mathbf{u}\cdot\nabla)\mathbf{u}\rangle_x =\left\langle\mathbf{u}\cdot\nabla\frac{|\mathbf{u}|^2}{2}\right\rangle_x =\left\langle\nabla\cdot\left(\frac{|\mathbf{u}|^2}{2}\mathbf{u}\right)\right\rangle_x=0. \tag*{(26)}

Likewise, the pressure term vanishes:

⟨u⋅∇p⟩x=⟨∇⋅(pu)⟩x−⟨p∇⋅u⟩x=0.(27)\langle\mathbf{u}\cdot\nabla p\rangle_x=\langle\nabla\cdot(p\mathbf{u})\rangle_x-\langle p\nabla\cdot\mathbf{u}\rangle_x=0. \tag*{(27)}

For constant viscosity, integration by parts gives

ν⟨u⋅∇2u⟩x=−ν⟨∣∇u∣2⟩x=−2ν⟨S:S⟩x,(28)\nu\langle\mathbf{u}\cdot\nabla^2\mathbf{u}\rangle_x=-\nu\langle|\nabla\mathbf{u}|^2\rangle_x=-2\nu\langle\mathbf{S}:\mathbf{S}\rangle_x, \tag*{(28)}

where S=(∇u+∇uT)/2\mathbf{S}=(\nabla\mathbf{u}+\nabla\mathbf{u}^{\mathsf{T}})/2. We therefore define the positive viscous dissipation rate as

Dν=−ν⟨u⋅∇2u⟩x=2ν⟨S:S⟩x≥0.(29)\mathcal{D}_{\nu}=-\nu\langle\mathbf{u}\cdot\nabla^2\mathbf{u}\rangle_x=2\nu\langle\mathbf{S}:\mathbf{S}\rangle_x\ge0. \tag*{(29)}

The mechanical power supplied by the odd capillary force is

Podd=⟨u⋅fodd⟩x,fodd=∇⋅Σodd.(30)P_{\mathrm{odd}}=\langle\mathbf{u}\cdot\mathbf{f}_{\mathrm{odd}}\rangle_x,\qquad\mathbf{f}_{\mathrm{odd}}=\nabla\cdot\boldsymbol{\Sigma}^{\mathrm{odd}}. \tag*{(30)}

Collecting the contributions, the kinetic-energy balance becomes

dedt=Peven+Podd−Dν.(31)\frac{de}{dt}=P_{\mathrm{even}}+P_{\mathrm{odd}}-\mathcal{D}_{\nu}. \tag*{(31)}

Total-energy balance. Adding Eqs. (B9) and (B16), the ordinary capillary power cancels exactly between the kinetic- and free-energy balances. Thus,

dEdt=Podd−Dν−Dμ.(32)\frac{d\mathcal{E}}{dt}=P_{\mathrm{odd}}-\mathcal{D}_{\nu}-\mathcal{D}_{\mu}. \tag*{(32)}

This recovers the total-energy balance in Eq. (7c). The ordinary capillary force mediates energy exchange between the fluid motion and the composition field, whereas the odd capillary stress contributes an independent, sign-indefinite power to the total-energy budget. For σe=0\sigma_{\mathrm{e}}=0, the odd power vanishes, and the total mean energy satisfies

dEdt=−Dν−Dμ≤0,(33)\frac{d\mathcal{E}}{dt}=-\mathcal{D}_{\nu}-\mathcal{D}_{\mu}\le0, \tag*{(33)}

recovering the dissipative structure of passive Cahn–Hilliard–Navier–Stokes dynamics.

Spectral kinetic-energy balance

We derive the scale-by-scale kinetic-energy balance used in the main text, following the spectral-transfer framework for Cahn–Hilliard–Navier–Stokes turbulence [30, 35, 38]. For constant density ρ=1\rho=1, the incompressible Navier–Stokes equation is

∂tu+(u⋅∇)u=−∇p+ν∇2u+feven+fodd,(34)\partial_t\mathbf{u}+(\mathbf{u}\cdot\nabla)\mathbf{u}=-\nabla p+\nu\nabla^2\mathbf{u}+\mathbf{f}_{\mathrm{even}}+\mathbf{f}_{\mathrm{odd}}, \tag*{(34)}

where feven=∇⋅Σeven\mathbf{f}_{\mathrm{even}}=\nabla\cdot\boldsymbol{\Sigma}^{\mathrm{even}} and fodd=∇⋅Σodd\mathbf{f}_{\mathrm{odd}}=\nabla\cdot\boldsymbol{\Sigma}^{\mathrm{odd}}. The ordinary capillary force is equivalent to −ϕ∇μ-\phi\nabla\mu up to a gradient, which does not contribute to the kinetic-energy balance of incompressible periodic flow.

For an N×NN\times N periodic grid, we define the normalized discrete Fourier coefficients as

u^(k,t)=1N2∑i,j=0N−1u(xij,t)e−ik⋅xij.(35)\widehat{\mathbf{u}}(\mathbf{k},t)=\frac{1}{N^2}\sum_{i,j=0}^{N-1}\mathbf{u}(\mathbf{x}_{ij},t)e^{-i\mathbf{k}\cdot\mathbf{x}_{ij}}. \tag*{(35)}

The same convention applies to all other fields. Here k=(2π/L)(nx,ny)\mathbf{k}=(2\pi/L)(n_x,n_y) is the physical wavevector for a square domain of side LL. The Fourier-transformed momentum equation can be written as

∂tu^(k,t)=−P(k)⋅(u⋅∇)u^(k,t)−ν∣k∣2u^(k,t)+P(k)⋅[f^even(k,t)+f^odd(k,t)],(36)\partial_t\widehat{\mathbf{u}}(\mathbf{k},t)=-\mathbf{P}(\mathbf{k})\cdot\widehat{(\mathbf{u}\cdot\nabla)\mathbf{u}}(\mathbf{k},t)-\nu|\mathbf{k}|^2\widehat{\mathbf{u}}(\mathbf{k},t)+\mathbf{P}(\mathbf{k})\cdot\left[\widehat{\mathbf{f}}_{\mathrm{even}}(\mathbf{k},t)+\widehat{\mathbf{f}}_{\mathrm{odd}}(\mathbf{k},t)\right], \tag*{(36)}

where

Pij(k)=δij−kikj∣k∣2(37)P_{ij}(\mathbf{k})=\delta_{ij}-\frac{k_i k_j}{|\mathbf{k}|^2} \tag*{(37)}

is the incompressible projection operator for k≠0\mathbf{k}\ne0. It removes the pressure contribution and ensures that k⋅u^(k,t)=0\mathbf{k}\cdot\widehat{\mathbf{u}}(\mathbf{k},t)=0.

The shell-integrated kinetic-energy spectrum is

E(k,t)=12∑k′∈Sk∣u^(k′,t)∣2,(38)E(k,t)=\frac{1}{2}\sum_{\mathbf{k}'\in S_k}|\widehat{\mathbf{u}}(\mathbf{k}',t)|^2, \tag*{(38)}

where Sk={k′:k≤∣k′∣<k+Δk}S_k=\{\mathbf{k}': k\le|\mathbf{k}'|<k+\Delta k\} denotes a wavenumber shell of width Δk\Delta k. Multiplying Eq. (36) by u^∗\widehat{\mathbf{u}}^*, taking the real part, and summing over the shell gives

∂tE(k,t)=Tu(k,t)+Teven(k,t)+Todd(k,t)−Dν(k,t).(39)\partial_t E(k,t)=T^{u}(k,t)+T^{\mathrm{even}}(k,t)+T^{\mathrm{odd}}(k,t)-D_{\nu}(k,t). \tag*{(39)}

The nonlinear advective transfer is

Tu(k,t)=−∑k′∈SkRe⁡[u^∗(k′,t)⋅P(k′)⋅(u⋅∇)u^(k′,t)].(40)T^{u}(k,t)=-\sum_{\mathbf{k}'\in S_k}\operatorname{Re}\left[\widehat{\mathbf{u}}^*(\mathbf{k}',t)\cdot\mathbf{P}(\mathbf{k}')\cdot\widehat{(\mathbf{u}\cdot\nabla)\mathbf{u}}(\mathbf{k}',t)\right]. \tag*{(40)}

The ordinary and odd capillary contributions are

Teven(k,t)=∑k′∈SkRe⁡[u^∗(k′,t)⋅P(k′)⋅f^even(k′,t)],(41)T^{\mathrm{even}}(k,t)=\sum_{\mathbf{k}'\in S_k}\operatorname{Re}\left[\widehat{\mathbf{u}}^*(\mathbf{k}',t)\cdot\mathbf{P}(\mathbf{k}')\cdot\widehat{\mathbf{f}}_{\mathrm{even}}(\mathbf{k}',t)\right], \tag*{(41)}
Todd(k,t)=∑k′∈SkRe⁡[u^∗(k′,t)⋅P(k′)⋅f^odd(k′,t)].(42)T^{\mathrm{odd}}(k,t)=\sum_{\mathbf{k}'\in S_k}\operatorname{Re}\left[\widehat{\mathbf{u}}^*(\mathbf{k}',t)\cdot\mathbf{P}(\mathbf{k}')\cdot\widehat{\mathbf{f}}_{\mathrm{odd}}(\mathbf{k}',t)\right]. \tag*{(42)}

The viscous dissipation spectrum is

Dν(k,t)=ν∑k′∈Sk∣k′∣2∣u^(k′,t)∣2.(43)D_{\nu}(k,t)=\nu\sum_{\mathbf{k}'\in S_k}|\mathbf{k}'|^2|\widehat{\mathbf{u}}(\mathbf{k}',t)|^2. \tag*{(43)}

Direct Numerical Simulations

We solve the coupled Cahn–Hilliard–Navier–Stokes equations using a Fourier pseudospectral method in a two-dimensional periodic square domain Ω=[0,2π]2\Omega=[0,2\pi]^2, following the numerical framework of Ref. [25]. To eliminate the pressure, we rewrite the incompressible momentum equation in the vorticity–streamfunction formulation,

∂tω+u⋅∇ω=ν∇2ω+[∇×(feven+fodd)]⋅e^z,(44)\partial_t\omega+\mathbf{u}\cdot\nabla\omega=\nu\nabla^2\omega+\left[\nabla\times(\mathbf{f}_{\mathrm{even}}+\mathbf{f}_{\mathrm{odd}})\right]\cdot\widehat{\mathbf{e}}_z, \tag*{(44)}

where ω=(∇×u)⋅e^z\omega= (\nabla\times\mathbf{u}) \cdot\hat{\mathbf{e}}_z is the scalar vorticity. The streamfunction ψ\psi satisfies

ω=−∇2ψ,u=(∂yψ,−∂xψ),(45)\omega= -\nabla^2\psi,\qquad\mathbf{u} = (\partial_y\psi,-\partial_x\psi), \tag*{(45)}

which automatically enforces ∇⋅u=0\nabla\cdot\mathbf{u} = 0.

The composition, vorticity, and streamfunction fields are represented by truncated Fourier series,

q(x,t)=∑kq^(k,t)exp⁡(ik⋅x),q∈{ϕ,ω,ψ},(46)q(\mathbf{x},t) = \sum_{\mathbf{k}} \widehat{q}(\mathbf{k},t)\exp(i\mathbf{k}\cdot\mathbf{x}),\qquad q \in\{\phi,\omega,\psi\}, \tag*{(46)}

where k=(kx,ky)\mathbf{k} = (k_x,k_y) is the Fourier wavevector, and the sum extends over all resolved Fourier modes. The evolution equations in Fourier space become

∂tω^=−u⋅∇ω^−νk2ω^+i(kxf^y−kyf^x),(47)\partial_t\widehat{\omega} = -\widehat{\mathbf{u}\cdot\nabla\omega} - \nu k^2\widehat{\omega} + i\left(k_x\widehat{f}_y-k_y\widehat{f}_x\right), \tag*{(47)}
∂tϕ^=−u⋅∇ϕ^−Mk2μ^,(48)\partial_t\widehat{\phi} = -\widehat{\mathbf{u}\cdot\nabla\phi} - M k^2\widehat{\mu}, \tag*{(48)}

where k=∣k∣k = \lvert\mathbf{k}\rvert and f=feven+fodd\mathbf{f} = \mathbf{f}_{\mathrm{even}} + \mathbf{f}_{\mathrm{odd}}. The chemical potential is

μ^=3σ4ϵ(ϕ3^−ϕ^)+3σϵ2k2ϕ^.(49)\widehat{\mu} = \frac{3\sigma}{4\epsilon}\left(\widehat{\phi^3}-\widehat{\phi}\right) + \frac{3\sigma\epsilon}{2}k^2\widehat{\phi}. \tag*{(49)}

The velocity is reconstructed from ψ^=ω^/k2\widehat{\psi} = \widehat{\omega}/k^2 for k≠0k \ne0, with the zero mode of the streamfunction set to zero. Spatial derivatives are evaluated spectrally, while nonlinear products and capillary forcing are computed in physical space and transformed back to Fourier space. The coupled equations are advanced using the second-order exponential time-differencing Runge–Kutta method (ET-DRK2) [51], which treats the linear dissipative operators exactly and the remaining nonlinear contributions explicitly. We implement the Fourier pseudospectral solver in CUDA C, with GPU-optimized routines for the numerical operations. All simulations are performed on NVIDIA GPUs, including the H100 architecture.

Throughout our simulations, we fix the parameters N=1024N = 1024, ϵ=0.02\epsilon= 0.02, M=ϵ2/2M = \epsilon^2/2, ν=1\nu= 1, and σ=1\sigma= 1. We define the Reynolds number as Re=UrmsLc/νRe = U_{\mathrm{rms}}L_c/\nu, where Urms=2⟨e(t)⟩tU_{\mathrm{rms}} = \sqrt{2\langle e(t)\rangle_t} and LcL_c is the characteristic coarsening length in the statistically stationary state. For χ=2\chi= 2, 4, and 10, we obtain Re=0.20Re = 0.20, 0.23, and 0.28, respectively, confirming that the arrested states occur at low Reynolds numbers.

References

References

  1. [1]P.-G. Gennes, F. Brochard-Wyart, D. Quéré, et al., Capillarity and wetting phenomena: drops, bubbles, pearls, waves (Springer, 2004).
  2. [2]Y. Pomeau and E. Villermaux, Physics Today 59, 39 (2006).
  3. [3]E. D. Siggia, Physical review A 20, 595 (1979).
  4. [4]A. A. Hyman, C. A. Weber, and F. Jülicher, Annual Review of Cell and Developmental Biology 30, 39 (2014).
  5. [5]D. Bonn, J. Eggers, J. Indekeu, J. Meunier, and E. Rolley, Reviews of modern physics 81, 739 (2009).
  6. [6]J. Eggers, Reviews of modern physics 69, 865 (1997).
  7. [7]J. Eggers, J. R. Lister, and H. A. Stone, Journal of Fluid Mechanics 401, 293 (1999).
  8. [8]M. J. Lighthill and J. Lighthill, Waves in fluids (Cambridge university press, 2001).
  9. [9]P. C. Hohenberg and B. I. Halperin, Reviews of Modern Physics 49, 435 (1977).
  10. [10]D. Jacqmin, Journal of computational physics 155, 96 (1999).
  11. [11]H. Abels, Archive for rational mechanics and analysis 194, 463 (2009).
  12. [12]M. E. Cates and E. Tjhung, Journal of Fluid Mechanics 836, P1 (2018).
  13. [13]N. B. Padhan and R. Pandit, Journal of Fluid Mechanics 1010, P1 (2025).
  14. [14]J. Avron, Journal of statistical physics 92, 543 (1998).
  15. [15]D. Banerjee, A. Souslov, A. G. Abanov, and V. Vitelli, Nature communications 8, 1573 (2017).
  16. [16]C. Scheibner, A. Souslov, D. Banerjee, P. Surowka, W. T. Irvine, and V. Vitelli, Nature Physics 16, 475 (2020).
  17. [17]L. Braverman, C. Scheibner, B. VanSaders, and V. Vitelli, Physical review letters 127, 268001 (2021).
  18. [18]M. Fruchart, C. Scheibner, and V. Vitelli, Annual Review of Condensed Matter Physics 14, 471 (2023).
  19. [19]L. Langford and A. K. Omar, Physical Review Letters 134, 068301 (2025).
  20. [20]S. Burekovic, S. Kole, A. Maitra, and C. Nardini, arXiv preprint arXiv:2607.12045 (2026).arxiv.org/abs/2607.12045
  21. [21]B. Wang, P. Pietzonka, and F. Jülicher, arXiv preprint arXiv:2603.20064 (2026).arxiv.org/abs/2603.20064
  22. [22]K. Blom and U. Thiele, arXiv preprint arXiv:2607.27305 (2026).arxiv.org/abs/2607.27305
  23. [23]A. Tiribocchi, R. Wittkowski, D. Marenduzzo, and M. E. Cates, Physical review letters 115, 188302 (2015).
  24. [24]S. J. Kole, G. P. Alexander, S. Ramaswamy, and A. Maitra, Phys. Rev. Lett. 126, 248001 (2021).
  25. [25]N. B. Padhan and A. Voigt, Journal of Physics: Condensed Matter 37, 105101 (2025).
  26. [26]S. Berti, G. Boffetta, M. Cencini, and A. Vulpiani, Phys. Rev. Lett. 95, 224501 (2005).
  27. [27]P. Perlekar, R. Benzi, H. J. H. Clercx, D. R. Nelson, and F. Toschi, Phys. Rev. Lett. 112, 014502 (2014).
  28. [28]P. Perlekar, N. Pal, and R. Pandit, Scientific Reports 7, 44589 (2017).
  29. [29]A. Tiribocchi, R. Wittkowski, D. Marenduzzo, and M. E. Cates, Phys. Rev. Lett. 115, 188302 (2015).
  30. [30]N. B. Padhan, K. V. Kiran, and R. Pandit, Soft Matter 20, 3620 (2024).
  31. [31]F. Caballero and M. C. Marchetti, Phys. Rev. Lett. 129, 268002 (2022).
  32. [32]R. Backofen and A. Voigt, arXiv preprint arXiv:2601.15001 (2026).arxiv.org/abs/2601.15001
  33. [33]B. Maji, N. B. Padhan, and R. Pandit, Communications Physics 8, 488 (2025).
  34. [34]B. Maji, N. B. Padhan, A. Voigt, and R. Pandit, arXiv preprint arXiv:2602.06928 (2026).arxiv.org/abs/2602.06928
  35. [35]P. Perlekar, Journal of Fluid Mechanics 873, 459 (2019).
  36. [36][36*] A. Alexakis and L. Biferale, Physics Reports 767, 1 [37] M. K. Verma, Energy transfers in fluid flows: multiscale and spectral perspectives (Cambridge University Press, 2019).
  37. [38]N. B. Padhan, D. Vincenzi, and R. Pandit, Phys. Rev. Fluids 9, L122401 (2024).
  38. [39]S. Torquato, Physics Reports 745, 1 (2018).
  39. [40]R. Backofen, A. Y. Altawil, M. Salvalaglio, and A. Voigt, Proceedings of the National Academy of Sciences 121, e2320719121 (2024).
  40. [41]R. Singh and M. E. Cates, Phys. Rev. Lett. 123, 148005 (2019).
  41. [42]A. G. Abanov, T. Can, and S. Ganeshan, SciPost Phys. 010 (2018).
  42. [43]Y. Hosaka and A. Vilfan, arXiv preprint arXiv:2609.17005 (2026).DOI
  43. [44]B. Barker, J. B. Bell, and A. L. Garcia, Proceedings of the National Academy of Sciences 120, e2306088120 (2023).
  44. [45]H. França and M. Jalaal, Phys. Rev. Res. 8, 033140 (2026).
  45. [46]R. Maire, A. Petrini, U. Marini Bettolo Marconi, and L. Caprini, The Journal of Chemical Physics 165 (2026).
  46. [47]V. Soni, E. S. Billigin, S. Magkiriadou, S. Saccanna, D. Bartolo, M. J. Shelley, and W. T. Irvine, Nature physics 15, 1188 (2019).
  47. [48]N. K. Dewangan and J. C. Conrad, Soft Matter 15, 9368 (2019).
  48. [49]J. Deng, M. Molaei, N. G. Chisholm, and K. J. Stebe, Langmuir 36, 6888 (2020).
  49. [50]J. Deng, M. Molaei, N. G. Chisholm, and K. J. Stebe, Journal of Fluid Mechanics 976, A18 (2023).
  50. [51]S. M. Cox and P. C. Matthews, Journal of Computational Physics 176, 430 (2002).

Paper details

vlnc:W_v8rjr8hu
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 9

downloads · 2

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