betaFeedback

WorksPhysics and Astronomy

Finite-Larmor-radius vortex interaction from δf to full-f gyrofluid dynamics

Abstract

We study finite-size electrostatic gyrofluid vortex pairs in a two-dimensional isothermal model, focusing on finite-Larmor-radius effects and the transition from local perturbative to finite-amplitude full-distribution dynamics. An effective response kernel based on Padé ion gyroaveraging regularizes the short-range Euler interaction on the Larmor-radius scale while preserving its logarithmic far field, yielding a finite- core prediction for the orbital frequency. Local-model scans show that increasing the ion temperature relative to the electron temperature enhances gyrospinning and filamentation and drives a separation-dependent transition from merger to long-lived co-rotation. Long-wavelength simulations reproduce the predicted finite-Larmor- radius-induced dip in orbital angular frequency, corresponding to a pronounced slowdown of the vortex-pair rotation. For the full-distribution regime, we introduce a quasi-neutrality-consistent initialization that reconstructs the ion gyrocenter density from prescribed electron density and electrostatic potential. It recovers the local limit at small amplitude, whereas finite-amplitude simulations reveal additional mean-density polarization vorticity, delayed or arrested merger, and a loss of density- vorticity sign symmetry already in the cold-ion full-f system. An independent sign scan of density and vorticity then shows how this nonlinear NOB asymmetry combines with the local warm-ion FLR symmetry breaking identified in earlier δf vortex studies; for warm ions, Padé-based arbitrary-wavelength NOB and NOB+LWL polarization additionally produce distinct morphologies and trajectories. These results demonstrate that nonlinear density-dependent polarization and FLR physics form separate but coupled mechanisms that can qualitatively alter elementary vortex interactions in two-dimensional gyrofluid dynamics.

Introduction

Isolated vortices are among the most persistent coherent structures of nonlinear fluid motion and provide a particularly transparent route to the elementary dynamics underlying two-dimensional turbulence. In an ideal two-dimensional fluid, kinetic energy and enstrophy are simultaneously conserved, producing the familiar dual-cascade phenomenology in which energy is transferred towards larger scales while enstrophy is transported towards smaller scales [1, 2]. The dynamics of this transfer is not exhausted by spectral arguments: it is realized through the formation, deformation, merger and filamentation of coherent vortices. Numerical and experimental studies have linked the inverse cascade to the elongation and thinning of small vortices by large-scale strain, while filamentary structures are central to the forward transfer of enstrophy [3, 4]. At the level of a vortex pair, mutual rotation, co-advection, merger and the generation of vorticity sheets already reproduce many of the basic nonlinear ingredients of a turbulent vortex population [5, 6, 7, 8]. Such deliberately reduced configurations are therefore valuable not because an isolated vortex pair is expected to occur unmodified in a magnetized plasma, but because they isolate the interaction law that is otherwise embedded in a much more complicated turbulent state.

Magnetized plasmas constitute an important physical realization of quasi-two-dimensional vortex dynamics. For low-frequency fluctuations in a strong magnetic field, the leading perpendicular motion is the E×B\mathbf{E} \times\mathbf{B} drift, vE=E×B/B2\mathbf{v}_{E} = \mathbf{E} \times\mathbf{B}/B^{2}, with E=−∇ϕ\mathbf{E} = -\nabla\phi. In a straight and approximately homogeneous guide field the electrostatic potential consequently plays the role of a stream function, while the corresponding perpendicular flow vorticity is proportional to ∇⊥2ϕ\nabla_{\perp}^{2}\phi. This close correspondence between drift dynamics and two-dimensional fluid motion has long motivated reduced descriptions of plasma turbulence and coherent structures [9, 10]. In magnetically confined plasmas the same nonlinear E×B\mathbf{E} \times\mathbf{B} advection contributes directly to turbulent cross-field particle and heat transport, while self-generated zonal and mean flows can in turn shear and regulate the turbulence [11, 12, 13].

The neutral-fluid analogy becomes incomplete when perpendicular scales approach the ion gyroradius. A charged ion does not sample the electrostatic potential only at its instantaneous gyrocenter position, but experiences an orbit-averaged field over its Larmor motion. At k⊥ρi∼1k_{\perp}\rho_{i} \sim1, finite-Larmor-radius (FLR) effects therefore modify both the gyrocenter advection and the polarization relation connecting the particle and gyrocenter densities to the electric potential. Gyrokinetic theory provides the systematic low-frequency reduction in which the fast gyrophase is removed while the relevant orbit-width, kinetic and polarization effects are retained [14, 15, 16]. Gyrofluid models proceed by taking velocity-space moments of the gyrokinetic distribution and closing the resulting hierarchy, retaining gyroaveraging and polarization physics at a substantially reduced computational cost [17, 18, 19, 20, 21, 22]. In this description the evolved densities are gyrocenter rather than particle densities, and the electrostatic field is obtained from quasi-neutrality together with the polarization response. The underlying role of polarization in guiding-center and gyrocenter reductions has a long theoretical history [23, 24]. This distinction is essential here: in the cold-ion, long-wavelength limit the gyrofluid polarization reduces to a Poisson equation and the vorticity dynamics approaches the two-dimensional Euler equation, whereas at finite ion temperature the gyroaveraging operators alter the connection between density, potential and vorticity.

FLR effects on coherent plasma structures have been observed in several related settings. Knorr et al. already showed that finite-Larmor-radius corrections can qualitatively alter a vortex-merger calculation [17]. In interchange-driven filament dynamics, FLR corrections break poloidal symmetry and generate sheared or spinning flows around the structures [25]. For finite-amplitude blobs, ion-temperature effects can alter coherence, morphology and propagation, while nonlinear polarization introduces amplitude-dependent inertia and blob–hole asymmetries [26, 27, 28]. More generally, rotational dynamics of plasma blobs can be important for their stability and propagation [29]. These results establish that ion gyro-orbit physics cannot in general be treated as a perturbative correction to otherwise Euler-like motion: it can modify the effective velocity field, the redistribution of vorticity, and the lifetime of coherent structures.

The direct conceptual predecessor of the present work is the gyrofluid vortex-interaction study [30]. That work reduced the problem to two-dimensional, isothermal, electrostatic δf\delta f gyrofluid dynamics and showed explicitly how the classical vortex-pair picture is reorganized by FLR polarization. In the cold-ion limit, co-rotating vortices undergo the expected parity-related Euler-like merger, while counter-rotating vortices co-advect in the familiar dipolar fashion. For warm ions, however, the evolution is no longer determined by the initial vorticity profile alone. The relative distribution of electron and ion gyrocenter densities fixes the FLR contribution to polarization, so two states with identical initial vorticity can evolve differently. Spatial asymmetry between density and vorticity contours generates gyrospinning: spiral arms form around the vortex cores, vorticity is wound into thin sheets, and the resulting structures can either accelerate merger or drive separation depending on the underlying gyrocenter-density configuration [30]. The same mechanism also modifies counter-rotating co-advection and was shown to persist in many-vortex drift-wave turbulence modelled with the Hasegawa–Wakatani system, where FLR effects enhance vorticity filamentation and alter the small-scale structure of the flow [30, 31].

That observation is particularly important because vorticity sheets are not merely a visual by-product of vortex deformation. In ordinary two-dimensional turbulence, vortex thinning and filamentation are closely connected with inter-scale transfer [3, 4]; in drift-wave turbulence, nonlinear self-organization and dissipation likewise depend sensitively on the generation of fine vorticity structure [32]. The FLR-induced sheets in gyrofluid vortices therefore provide a plausible microscopic route by which ion-gyroradius physics can influence the spectral and transport properties of a turbulent plasma. Previous works isolated this mechanism, rather than embedding it immediately in a fully realistic edge model, precisely because the fate of a vortex pair gives a clean measure of how polarization changes nonlinear advection [30]. The present paper follows the same philosophy but asks how far that picture survives once two assumptions of the local model are relaxed or examined more closely: the form of the effective interaction at finite ρi\rho_i, and the small-amplitude δf\delta f approximation itself.

The first question is analytical. Classical point-vortex and finite-core theories express the orbital motion of a symmetric vortex pair through the Green function that connects vorticity to the stream function [5, 6, 7]. In a gyrofluid plasma, the electric potential plays the role of that stream function, but it is obtained through a scale-dependent polarization response containing gyroaveraging operators rather than through a simple Laplacian inversion. This suggests a natural reformulation: FLR effects may be viewed as a modification of the interaction kernel itself. Such a representation makes it possible to separate the long-range Euler-like part of the interaction from the short-range gyro-orbit correction and to derive an explicit prediction for the orbital frequency of finite-core vortices. It also provides a more quantitative basis for interpreting the dependence on ion temperature and vortex separation than a purely morphological comparison of merger snapshots.

The second question concerns fluctuation amplitude. We denote the electron gyrocenter density by nen_e, the ion gyrocenter density by NiN_i, and the electrostatic potential by ϕ\phi. The vortex problem in Ref. [30] is formulated in a local δf\delta f model, in which the dynamical fields (e.g. species gyrocenter density n~s\widetilde{n}_s) are small perturbations about a prescribed stationary background NbgN_{\mathrm{bg}}. This approximation is extremely powerful for turbulence and stability problems satisfying ∣n~s∣/Nbg≪1|\widetilde{n}_s|/N_{\mathrm{bg}} \ll1, because the background density can be removed from the nonlinear part of the polarization operator and the field inversion remains comparatively simple. The approximation is not uniformly valid, however, for structures whose density variation is comparable with the reference level. Such amplitudes occur naturally for coherent filaments and blobs in the plasma edge and scrape-off layer, motivating global or full-ff gyrofluid descriptions in which the complete moments, rather than only perturbations, are evolved [26, 27, 28, 29, 33]. In an unapproximated full-ff polarization equation the complete, spatially varying ion gyrocenter density enters the polarization response; in the terminology used below this is the non-Oberbeck–Boussinesq (NOB) formulation. Spatial variations of NiN_i therefore modify the local polarization inertia and can produce dynamics that have no counterpart in a constant-background δf\delta f inversion. The passage from δf\delta f to full-ff is consequently not just a change in normalization or bookkeeping; it changes the field equation that generates the E×B\mathbf{E} \times\mathbf{B} flow.

A further distinction is required between amplitude effects and wavelength effects in the polarization equation. The Oberbeck–Boussinesq (OB) approximation replaces the spatially varying density multiplying the polarization response by a fixed reference value. Thus the unapproximated full-ff polarization used here is NOB; applying OB to a full-ff model would be an additional approximation. Independently, one may retain the polarization response at arbitrary perpendicular wavelength or expand it in the long-wavelength limit (LWL), k⊥ρi≪1k_\perp\rho_i \ll1. Historically, full-ff gyrofluid polarization was commonly restricted to the LWL because of the short-wavelength closure problem. Held, Wiesenberger and Kendl developed Padé-based full-ff closures that retain linear polarization effects at arbitrary wavelength while recovering both the OB and LWL limits [34], and Held and Wiesenberger later demonstrated explicitly that NOB effects and arbitrary-wavelength effects can alter different aspects of finite-amplitude dynamics [35]. In the terminology of the present paper, NOB therefore denotes full-ff polarization without the OB replacement, while NOB+LWL denotes the same full-ff/NOB density dependence combined with the additional long-wavelength approximation. This distinction is especially relevant for vortex interaction because the FLR mechanism of Ref. [30] acts through thin, high-wavenumber vorticity sheets. Recent full-ff, full-kk implementations have extended this framework to quasi-two-dimensional drift-wave and interchange turbulence [36], and three-dimensional full-ff gyrofluid turbulence calculations demonstrate the continued development of global gyrofluid modelling beyond simplified two-dimensional benchmarks [37].

A controlled full-ff continuation of the vortex-pair problem nevertheless introduces an additional difficulty that is absent in the local calculation: initialization. In the δf\delta f model, electron and ion gyrocenter perturbations can be chosen so that the linear polarization equation generates a prescribed Gaussian vorticity. In full-ff, prescribing nen_e, NiN_i and ϕ\phi in a way that is directly consistent with the corresponding δf\delta f vortex is nontrivial because the polarization response is nonlinear and variable-coefficient; the three fields therefore cannot in general be prescribed independently while maintaining quasi-neutrality. A meaningful comparison with the local vortex problem requires a construction that preserves the intended electron-density and potential structure and reconstructs the compatible ion gyrocenter density. The GREENY framework provides both δf\delta f and full-ff formulations, together with NOB arbitrary-wavelength and NOB+LWL polarization options, within one numerical implementation [38]. It therefore permits the same elementary vortex configuration to be followed continuously from the small-amplitude regime into a genuinely finite-amplitude polarization problem.

The literature thus contains two well-developed but only partially connected lines: local gyrofluid studies that isolate how FLR polarization changes elementary vortex interaction [17, 30], and global/full-ff studies that establish the importance of nonlinear polarization, finite inertia and arbitrary-wavelength response for large-amplitude blobs and turbulence [26, 27, 28, 33, 34, 35, 36, 37]. What has remained largely unexplored is the direct continuation of the same finite-core, freely interacting vortex pair from one regime into the other. Addressing this gap allows the new physics to be attributed cleanly: one can first identify the FLR modification of the local interaction, verify the δf\delta f limit, and then determine which additional effects arise only when the complete finite-amplitude density participates dynamically in polarization.

The purpose of the present paper is to make this connection systematically. We first formulate an effective FLR-modified interaction kernel for the local gyrofluid polarization response and derive an orbital-frequency prediction for Gaussian finite-core vortices. We then revisit the δf\delta f co-rotating vortex problem over a broader range of initial separation and ion temperature, including the separation-dependent transition from complete merger to long-lived orbital motion and a comparison of measured orbital frequencies with the analytical prediction. Finally, we construct quasi-neutrality-consistent full-ff initial states, verify recovery of the local limit at small amplitude, and determine how nonlinear mean-density polarization modifies the same interaction when the localized density variation becomes comparable with the mean-density level. By varying the density amplitude n0n_0 and vorticity amplitude δ\delta independently, we then separate the local warm-ion density–vorticity symmetry breaking of Ref. [30] from the additional finite-amplitude sign asymmetry introduced by the full-ff/NOB polarization operator. The analysis is deliberately restricted to freely evolving electrostatic vortices in a homogeneous guide field. Curvature drive, parallel dynamics, sheath coupling and electromagnetic vortex interaction are excluded so that FLR effects, fluctuation amplitude and the polarization closure can be separated as cleanly as possible. For the summary below, τi=Ti/Te\tau_i = T_i/T_e is the ion-to-electron temperature ratio, n0n_0 is the amplitude of the localized electron-density structure, and NmeanN_{\mathrm{mean}} denotes the uniform mean-density level used to construct the full-ff initial state; NmeanN_{\mathrm{mean}} is not a static background field evolved by the full-ff model.

Main results. The paper establishes five connected results:

  • Finite-core analytical interaction law. A reduced gyrofluid polarization response yields an FLR-modified interaction kernel and an explicit finite-core prediction for the orbital angular frequency of Gaussian vortices.

  • δf\delta f merger–orbit transition. In positive-vorticity parameter scans, increasing τi\tau_i produces a transition from complete merger to a long-lived co-rotating state. The location of this transition in τi\tau_i depends on the initial vortex separation and is therefore not assigned a universal critical temperature ratio; across the scans, the transition coincides with enhanced small-scale vorticity in the FLR-generated sheets.

  • Recovery and breakdown of the local limit. A generalized full-ff initialization reconstructs the ion gyrocenter density NiN_i for prescribed electron gyrocenter density nen_e and electrostatic potential ϕ\phi, recovers the δf\delta f dynamics for sufficiently small localized density amplitudes, and exposes a distinct mean-density polarization contribution when the full-ff density variation becomes comparable with NmeanN_{\mathrm{mean}}.

  • Finite amplitude breaks the local density–vorticity sign symmetry. An independent scan over n0=±0.5n_0 = \pm0.5 and δ=±0.01\delta= \pm0.01 shows that full-ff/NOB polarization distinguishes opposite density departures even at fixed initial vorticity already for cold

ions. At finite τi\tau_i, this density-dependent asymmetry combines with the local FLR symmetry breaking known from δf\delta f vortex dynamics.

  • Arbitrary-wavelength polarization matters once the two mechanisms interact. For warm ions, the competition between NOB density dependence and FLR-generated fine structure makes the evolution sensitive to whether arbitrary-wavelength NOB or NOB+LWL polarization is retained, producing qualitatively different trajectories and shedding behavior.

Gyrofluid model, initialization and diagnostics

Electrostatic isothermal gyrofluid equations

We use the two-dimensional isothermal gyrofluid model implemented in GREENY [38], following the full-ff polarization structure of global gyrofluid formulations [26, 33]. The geometry is a slab perpendicular to a strong, homogeneous guide field. Electron FLR effects are neglected and the magnetic field magnitude is normalized to unity. In the electrostatic limit, and in the absence of external drives, curvature terms, parallel dynamics and sources, the normalized full-ff density equations reduce to

∂ne∂t=−[ϕ,ne]⊥,(1)\frac{\partial n_e}{\partial t} = -[\phi,n_e]_\perp, \tag*{(1)}
∂Ni∂t=−[ψi,Ni]⊥,(2)\frac{\partial N_i}{\partial t} = -[\psi_i,N_i]_\perp, \tag*{(2)}

with the perpendicular Poisson bracket

[f,g]⊥=∂f∂x∂g∂y−∂f∂y∂g∂x.(3)[f,g]_\perp= \frac{\partial f}{\partial x}\frac{\partial g}{\partial y} - \frac{\partial f}{\partial y}\frac{\partial g}{\partial x}. \tag*{(3)}

Here nen_e is the electron gyrocenter density, NiN_i the ion gyrocenter density and ϕ\phi the electrostatic potential. Throughout this work, gyro-Bohm normalization is employed: densities are normalized to a fixed normalization density N0N_0, perpendicular spatial scales to the drift scale ρS=cS/Ωi\rho_S = c_S/\Omega_i, time to L⊥/cSL_\perp/c_S, and the electrostatic potential to Te/eT_e/e, where cS=Te/mic_S = \sqrt{T_e/m_i} is the sound speed, Ωi=eB0/mi\Omega_i = eB_0/m_i the ion gyrofrequency, and L⊥L_\perp the perpendicular reference length used in the time normalization. All quantities appearing below are therefore dimensionless, and normalized variables are written without hats. With the normalized ion mass parameter equal to unity, the gyroaveraged ion potential is

ψi=Γ1ϕ−12∣∇⊥Γ1ϕ∣2.(4)\psi_i = \Gamma_1\phi- \frac{1}{2}\lvert\nabla_\perp\Gamma_1\phi\rvert^2. \tag*{(4)}

The Padé FLR operators are

Γ0=11−τi∇⊥2,Γ1=11−τi2∇⊥2,(5)\Gamma_0 = \frac{1}{1-\tau_i\nabla_\perp^2}, \qquad\Gamma_1 = \frac{1}{1-\frac{\tau_i}{2}\nabla_\perp^2}, \tag*{(5)}

where τi=Ti/Te\tau_i = T_i/T_e controls the ion gyroradius in the adopted gyro-Bohm normalization.

The full arbitrary-wavelength NOB polarization equation is

∇⊥⋅[Γ0(NiΓ0∇⊥ϕ)]=ne−Γ1Ni.(6)\nabla_\perp\cdot\left[\sqrt{\Gamma_0}\left(N_i\sqrt{\Gamma_0}\nabla_\perp\phi\right)\right] = n_e-\Gamma_1N_i. \tag*{(6)}

The corresponding NOB long-wavelength-limit (LWL) polarization is

∇⊥⋅(Ni∇⊥ϕ)=ne−Γ1Ni.(7)\nabla_\perp\cdot\left(N_i\nabla_\perp\phi\right) = n_e-\Gamma_1N_i. \tag*{(7)}

Thus, in full-ff the ion gyrocenter density enters inside the polarization operator itself. For numerical stabilization, GREENY adds the fourth-order hyperviscous sink Λf=−ν∇⊥4f\Lambda_f=-\nu\nabla_\perp^4 f to each evolved gyrocenter-density field ff; the value of ν\nu used in each simulation set is stated below.

For the local δf\delta f limit, the normalized prescribed static background is unity and we write

ne≈1+n~e,Ni≈1+N~i,∣n~e∣,∣N~i∣≪1.(8)n_e \approx1+\widetilde{n}_e,\qquad N_i \approx1+\widetilde{N}_i,\qquad|\widetilde{n}_e|,|\widetilde{N}_i| \ll1. \tag*{(8)}

Here n~e\widetilde{n}_e and N~i\widetilde{N}_i denote, exclusively, the electron and ion gyrocenter fluctuations about that static δf\delta f background. We retain the nonlinear perpendicular advection of these fluctuations. The density evolution equations become

∂n~e∂t=−[ϕ,n~e]⊥,(9)\frac{\partial\widetilde{n}_e}{\partial t}=-[\phi,\widetilde{n}_e]_\perp, \tag*{(9)}
∂N~i∂t=−[Γ1ϕ,N~i]⊥,(10)\frac{\partial\widetilde{N}_i}{\partial t}=-[\Gamma_1\phi,\widetilde{N}_i]_\perp, \tag*{(10)}

closed by the constant-background polarization equation

1τi(Γ0−1)ϕ=n~e−Γ1N~i.(11)\frac{1}{\tau_i}(\Gamma_0-1)\phi=\widetilde{n}_e-\Gamma_1\widetilde{N}_i. \tag*{(11)}

Using (Γ0−1)/τi=Γ0∇⊥2(\Gamma_0-1)/\tau_i=\Gamma_0\nabla_\perp^2, the constant-density linearization of equation (6) reduces directly to equation (11), fixing the sign convention consistently between the full-ff and local models. In the cold-ion limit, τi→0\tau_i\to0, this reduces to a Poisson equation for ϕ\phi, and the electrostatic vorticity Ω=∇⊥2ϕ\Omega=\nabla_\perp^2\phi obeys the two-dimensional Euler form ∂tΩ=−[ϕ,Ω]⊥\partial_t\Omega=-[\phi,\Omega]_\perp.

Density–vorticity symmetry from δf\delta f to full-ff

The symmetry breaking discussed in Ref. [30] can be recovered directly from the local equations and then distinguished from the finite-amplitude asymmetry of the full-ff polarization used here. These are related but not identical mechanisms. The local warm-ion effect shows that a prescribed vorticity does not uniquely specify the gyrofluid state, whereas the full-ff/NOB effect makes a complete finite-amplitude blob–hole inversion itself nonlinear.

In the cold-ion local limit, the vorticity satisfies

DϕΩ=0,Dϕ≡∂t+[ϕ,⋅]⊥,Ω=∇⊥2ϕ.(12)D_{\phi}\Omega= 0,\qquad D_{\phi} \equiv\partial_t + [\phi,\cdot]_{\perp},\qquad\Omega= \nabla_{\perp}^{2}\phi. \tag*{(12)}

The density perturbations do not enter this equation, so two cold-ion initial states with the same Ω\Omega have the same vorticity evolution. For warm ions, a long-wavelength expansion of the local gyrofluid equations gives, to leading FLR order [30],

Dϕ(Ω+τi2∇⊥2n~e)=τi2[Ω,n~e]⊥+O(τi2∇⊥4).(13)D_{\phi}\left(\Omega+ \frac{\tau_i}{2}\nabla_{\perp}^{2}\widetilde{n}_e\right) = \frac{\tau_i}{2}[\Omega,\widetilde{n}_e]_{\perp} + O\left(\tau_i^2\nabla_{\perp}^{4}\right). \tag*{(13)}

The relevant comparison in Ref. [30] keeps the initial vorticity fixed while changing the density/gyrocenter configuration. In its simplest sign test,

Ω↦Ω,n~e↦−n~e,(14)\Omega\mapsto\Omega,\qquad\widetilde{n}_e \mapsto-\widetilde{n}_e, \tag*{(14)}

with N~i\widetilde{N}_i adjusted through equation (11) so that the prescribed Ω\Omega remains unchanged. The two initial generalized vorticities are then

W+=Ω+τi2∇⊥2n~e,W−=Ω−τi2∇⊥2n~e,(15)W_{+} = \Omega+ \frac{\tau_i}{2}\nabla_{\perp}^{2}\widetilde{n}_e,\qquad W_{-} = \Omega- \frac{\tau_i}{2}\nabla_{\perp}^{2}\widetilde{n}_e, \tag*{(15)}

and the FLR source changes sign,

τi2[Ω,−n~e]⊥=−τi2[Ω,n~e]⊥.(16)\frac{\tau_i}{2}[\Omega,-\widetilde{n}_e]_{\perp} = -\frac{\tau_i}{2}[\Omega,\widetilde{n}_e]_{\perp}. \tag*{(16)}

Thus equal initial vorticity does not imply equal warm-ion evolution: the relative electron–ion gyrocenter-density configuration determines the sign of the FLR polarization response and can select accelerated merger or separation. This is the precise local symmetry breaking identified by Kendl. It should not be confused with a complete inversion of all perturbations. If (n~e,N~i,ϕ)(\widetilde{n}_e,\widetilde{N}_i,\phi) are all reversed, then the polarization relation is reversed consistently and one obtains the corresponding globally sign-reversed state (with the usual reversal of the flow direction). The non-equivalence in equation (14) instead arises because Ω\Omega is deliberately held fixed while the density configuration is changed.

Full-ff polarization introduces a second and independent loss of sign symmetry. Define the full-ff polarization charge

Q≡ne−Γ1Ni=P[Ni,ϕ],P[Ni,ϕ]≡∇⊥⋅[Γ0(NiΓ0∇⊥ϕ)].(17)Q \equiv n_e - \Gamma_1 N_i = \mathcal{P}[N_i,\phi], \qquad\mathcal{P}[N_i,\phi] \equiv\nabla_\perp\cdot\left[\sqrt{\Gamma_0}\left(N_i\sqrt{\Gamma_0}\nabla_\perp\phi\right)\right]. \tag*{(17)}

Writing Ni=Nmean+ΔNiN_i = N_{\mathrm{mean}} + \Delta N_i, its arbitrary-wavelength polarization separates exactly into

Q=NmeanΓ0∇⊥2ϕ+∇⊥⋅[Γ0(ΔNiΓ0∇⊥ϕ)]≡QOB+QNOB.(18)\begin{aligned} Q = N_{\mathrm{mean}}\Gamma_0\nabla_\perp^2\phi+ \nabla_\perp\cdot\left[\sqrt{\Gamma_0}\left(\Delta N_i\sqrt{\Gamma_0}\nabla_\perp\phi\right)\right] \\ &\equiv Q_{\mathrm{OB}} + Q_{\mathrm{NOB}}. \tag*{(18)} \end{aligned}

The full-ff comparison is a genuine finite-amplitude blob–hole inversion about the same uniform mean. The natural candidate transformation is

(ΔNi, ne−Nmean, ϕ)↦(−ΔNi, −(ne−Nmean), −ϕ).(19)(\Delta N_i,\, n_e - N_{\mathrm{mean}},\, \phi) \mapsto(-\Delta N_i,\,-(n_e - N_{\mathrm{mean}}),\,-\phi). \tag*{(19)}

Here both density departures change sign and ϕ↦−ϕ\phi\mapsto-\phi is the potential that a linear polarization equation would produce. The density side Q=ne−Γ1NiQ = n_e - \Gamma_1N_i is odd under this transformation. On the polarization side, however, the OB term is odd but the NOB term is even, because it is bilinear in ΔNi\Delta N_i and ∇⊥ϕ\nabla_\perp\phi. Hence

QOB+QNOB↦−QOB+QNOB≠−Q.(20)Q_{\mathrm{OB}} + Q_{\mathrm{NOB}} \mapsto-Q_{\mathrm{OB}} + Q_{\mathrm{NOB}} \ne-Q. \tag*{(20)}

The transformed fields therefore do not satisfy the full nonlinear polarization equation: the potential of the hole is not simply the negative of the potential of the corresponding blob. This is the full-ff/NOB blob–hole symmetry-breaking mechanism. Density enhancements and depletions carry different polarization inertia even before warm-ion gyroaveraging is considered.

The same result is especially transparent in the LWL, where

Q=∇⊥⋅(Ni∇⊥ϕ)=NmeanΩ+ΔNiΩ+∇⊥ΔNi⋅∇⊥ϕ.(21)Q = \nabla_\perp\cdot(N_i\nabla_\perp\phi) = N_{\mathrm{mean}}\Omega+ \Delta N_i\Omega+ \nabla_\perp\Delta N_i \cdot\nabla_\perp\phi. \tag*{(21)}

The constant-density term is odd, whereas both NOB terms are even under equation (19). They are of second order for ∣ΔNi∣/Nmean≪1|\Delta N_i|/N_{\mathrm{mean}} \ll1, which explains why the local sign symmetry is recovered with the δf\delta f/OB limit, but become comparable to the OB term at finite amplitude. Locally, if density gradients across a vortex core are neglected, Q≃NiΩQ \simeq N_i\Omega and therefore Ω≃Q/Ni\Omega\simeq Q/N_i: equal and opposite polarization charges embedded in a density enhancement and depletion need not generate equal and opposite vorticities.

Finally, full-ff and FLR effects act together rather than replacing one another. Using equations (1), (2) and (17), the exact polarization-charge evolution may be written

DϕQ=Γ1[ψi,Ni]⊥−[ϕ,Γ1Ni]⊥,(22)D_\phi Q = \Gamma_1[\psi_i,N_i]_\perp- [\phi,\Gamma_1N_i]_\perp, \tag*{(22)}

or, after inserting ψi=Γ1ϕ+ψ2i\psi_i = \Gamma_1\phi+ \psi_{2i} with ψ2i=−∣∇⊥Γ1ϕ∣2/2\psi_{2i} = -\lvert\nabla_\perp\Gamma_1\phi\rvert^2/2,

DϕQ=Γ1[Γ1ϕ,Ni]⊥−[ϕ,Γ1Ni]⊥+Γ1[ψ2i,Ni]⊥(23)D_\phi Q = \Gamma_1 [\Gamma_1\phi, N_i]_\perp- [\phi, \Gamma_1N_i]_\perp+ \Gamma_1[\psi_{2i},N_i]_\perp \tag*{(23)}

Equations (18) and (23) display the combined mechanism. Under the candidate blob–hole transformation equation (19), the two gyroaveraging-mismatch brackets in the first line of equation (23) are even: both ϕ\phi and the non-constant part of NiN_i change sign. In contrast, ψ2i\psi_{2i} is quadratic and therefore even in ϕ\phi, so its bracket with the sign-reversed density departure is odd. The full evolution consequently contains contributions of different parity in addition to the already non-odd NOB field inversion. No single transformation ϕhole=−ϕblob\phi_{\mathrm{hole}}=-\phi_{\mathrm{blob}} can map the finite-amplitude warm-ion system onto itself.

In the local small-amplitude limit, NiN_i is replaced by NmeanN_{\mathrm{mean}} inside the polarization operator, the even NOB contribution in equation (18) is discarded, and the higher-order ψ2i\psi_{2i} coupling is absent from the retained δf\delta f equations. In the further cold-ion limit, Γ1→1\Gamma_1\to1 and equation (12) is recovered. At finite amplitude and finite τi\tau_i, neither reduction applies. The NOB blob–hole asymmetry of the field inversion and the FLR sensitivity to the relative density–vorticity configuration coexist and jointly determine whether the vortices merge, orbit or separate. This separation of mechanisms is tested directly in § 5.3 by independently reversing the density amplitude and the prescribed vorticity sign.

δf\delta f co-rotating vortex initialization

To connect directly to the established gyrofluid vortex benchmark [30], two equal Gaussian electron-density fluctuations of width aa are initialized on the prescribed static δf\delta f background at a center-to-center separation R0R_0,

n~e,0=n0[g1(x,y)+g2(x,y)],(24)\widetilde{n}_{e,0}=n_0\left[g_1(x,y)+g_2(x,y)\right], \tag*{(24)}

where n0n_0 is the fluctuation amplitude in the δf\delta f model, and

gj=exp⁡[−(x−x0(j))2+(y−y0)2a2],x0(1,2)=Lx2∓R02.(25)g_j=\exp\left[-\frac{(x-x_0^{(j)})^2+(y-y_0)^2}{a^2}\right], \qquad x_0^{(1,2)}=\frac{L_x}{2}\mp\frac{R_0}{2}. \tag*{(25)}

Following the vortex-interaction construction of Ref. [30], the ion gyrocenter fluctuation is chosen as

N~i,0=Γ1−1(1−δn0Γ0)n~e,0,(26)\widetilde{N}_{i,0}=\Gamma_1^{-1}\left(1-\frac{\delta}{n_0}\Gamma_0\right)\widetilde{n}_{e,0}, \tag*{(26)}

where the sign of δ\delta specifies the initial electron–ion gyrocenter asymmetry. Inserting equation (26) into equation (11) gives the desired Gaussian vorticity pair,

Ω0=∇⊥2ϕ0=δn0n~e,0.(27)\Omega_0 = \nabla_\perp^2 \phi_0 = \frac{\delta}{n_0}\widetilde{n}_{e,0}. \tag*{(27)}

For n0≥0n_0 \ge0, the sign of δ\delta therefore also fixes the sign of the initialized vorticity. At finite τi\tau_i, δ\delta additionally selects the relative electron–ion gyrocenter-density configuration that drives the FLR gyrospinning response: the positive-δ\delta branch of the benchmark in Ref. [30] produces an attractive FLR contribution and promotes merger of the co-rotating pair, whereas reversing δ\delta reverses the initialized gyrocenter asymmetry and selects the branch in which the FLR contribution drives the two cores apart. Thus δ\delta is not merely an amplitude parameter; it selects the sign and the warm-ion interaction branch of the initialized vortex pair.

Quasi-neutrality-consistent full-ff initialization

In full-ff, prescribing nen_e, NiN_i and ϕ\phi in a δf\delta f-consistent way is nontrivial because of the more complex nonlinear polarization response in equation (6). We therefore prescribe the physically desired electron density and electrostatic potential and reconstruct the ion gyrocenter density numerically. To make the comparison with the local initialization explicit, we write the initial electron density as

ne,0=Nmean+Δne,0,Δne,0=n0[g1(x,y)+g2(x,y)].(28)n_{e,0} = N_{\mathrm{mean}} + \Delta n_{e,0}, \qquad\Delta n_{e,0} = n_0 [g_1(x,y) + g_2(x,y)]. \tag*{(28)}

where NmeanN_{\mathrm{mean}} is the uniform mean-density level of the initial full-ff state away from the localized vortex cores and Δne,0\Delta n_{e,0} is the finite-amplitude density excess relative to that level. Importantly, NmeanN_{\mathrm{mean}} is not a prescribed static background field of the full-ff dynamics: nen_e and NiN_i remain the complete evolved densities. The ratio n0/Nmeann_0/N_{\mathrm{mean}} is used only to characterize the amplitude of the localized structure relative to this mean-density level. Introducing ϕG≡Γ0ϕ\phi_G \equiv\sqrt{\Gamma_0}\phi, we rewrite the full-ff polarization equation by decomposing the complete densities into the constant mean-density level and their finite-amplitude departures, Ni=Nmean+(Ni−Nmean)N_i = N_{\mathrm{mean}} + (N_i - N_{\mathrm{mean}}) and ne=Nmean+(ne−Nmean)n_e = N_{\mathrm{mean}} + (n_e - N_{\mathrm{mean}}). This decomposition is exact and remains valid for finite-amplitude departures. Since the gyroaveraging operators act as the identity on the spatially constant NmeanN_{\mathrm{mean}}, the constant terms on the right-hand side cancel, while the mean-density contribution inside the polarization flux yields Nmean∇⊥2ϕGN_{\mathrm{mean}}\nabla_\perp^2\phi_G. The polarization equation can therefore be written as

∇⊥⋅[(Ni−Nmean)∇⊥ϕG]+Γ0−1Γ1(Ni−Nmean)=Γ0−1(ne−Nmean)−Nmean∇⊥2ϕG.(29)\nabla_\perp\cdot[(N_i - N_{\mathrm{mean}})\nabla_\perp\phi_G] + \sqrt{\Gamma_0}^{-1}\Gamma_1(N_i - N_{\mathrm{mean}}) = \sqrt{\Gamma_0}^{-1}(n_e - N_{\mathrm{mean}}) - N_{\mathrm{mean}}\nabla_\perp^2\phi_G. \tag*{(29)}

Thus, the inversion is formulated in terms of the finite-amplitude density departures from NmeanN_{\mathrm{mean}}, while the polarization associated with the mean-density level is retained explicitly. The formulation therefore naturally accommodates ∣ne−Nmean∣/Nmean=O(1)\lvert n_e-N_{\mathrm{mean}}\rvert/N_{\mathrm{mean}}=O(1), as in the finite-amplitude cases considered below.

We impose homogeneous Dirichlet conditions for the ion-density departure from the mean level,

(Ni−Nmean)∣∂A=0.(30)\left.(N_i-N_{\mathrm{mean}})\right|_{\partial A}=0. \tag*{(30)}

The localized Gaussian fields used here decay sufficiently strongly near the domain boundary that the spectral application of the gyro-operators remains compatible with the periodic continuation to numerical accuracy.

Equation (29) is discretized on a uniform Cartesian grid with a conservative finite-volume treatment of the nonlinear flux. Face-centered gradients of ϕG\phi_G define the flux direction and a donor-cell value of Ni−NmeanN_i-N_{\mathrm{mean}} is used at each face. The FLR and polarization operators are chosen consistently such that the corresponding square-root approximation Γ0≃Γ1\sqrt{\Gamma_0}\simeq\Gamma_1 is fulfilled within the adopted second-order Padé representation [34]. Consequently, Γ0−1Γ1\sqrt{\Gamma_0}^{-1}\Gamma_1 reduces to the identity contribution in the sparse operator. The resulting linear system, where pp indexes the set I\mathcal{I} of interior grid cells,

Mx=b,xp=(Ni−Nmean)p,p∈I,(31)\mathbf{M}\mathbf{x}=\mathbf{b}, \qquad x_p=(N_i-N_{\mathrm{mean}})_p,\quad p\in\mathcal{I}, \tag*{(31)}

contains at most five non-zero entries per interior row and is solved with BiCGSTAB preconditioned by an incomplete LU factorization. A Laplacian regularization ϵ∇⊥2(Ni−Nmean)\epsilon\nabla_\perp^2(N_i-N_{\mathrm{mean}}) is available for poorly coupled closed-contour configurations, but was not required for the vortex simulations discussed below. The standard inversion tolerance was 10−810^{-8} with a maximum of 500 iterations.

The initialization is verified by manufactured solutions and by recovery of the δf\delta f limit. For the numerical test we prescribe

ϕG(x,y)=Asin⁡2(kxx)sin⁡2(kyy),(32)\phi_G(x,y)=A\sin^2(k_xx)\sin^2(k_yy), \tag*{(32)}
Ni,exact(x,y)=Nmean+Bsin⁡2(kxx)sin⁡2(kyy),(33)N_{i,\mathrm{exact}}(x,y)=N_{\mathrm{mean}}+B\sin^2(k_xx)\sin^2(k_yy), \tag*{(33)}

construct the corresponding nen_e analytically from equation (29), and compare the recovered NiN_i with the exact field. The first-order donor-cell discretization yields the expected approximately first-order convergence in grid resolution; the errors also reproduce the expected scaling with effective mode number and potential amplitude. The three verification measures are collected in Figure 1.

Resolution scaling, effective-wavenumber scaling, and potential-amplitude scaling

Figure 1. Verification of the generalized full-ff initialization inversion. The error norms show approximately first-order convergence with grid resolution, cubic scaling with effective wavenumber, and linear scaling with potential amplitude, as expected for the employed discretization.

Numerical diagnostics

The vortex positions are obtained from a local center-of-mass tracker, but the tracked electron-density field must be distinguished between the two models. In the δf\delta f calculations the weight is the evolved fluctuation n~e\widetilde{n}_e about the prescribed static background. In the full-ff calculations there is no corresponding static background or fluctuation variable; instead we use the finite-amplitude density excess

Δne(r,t)=ne(r,t)−Nmean,(34)\Delta n_e(\mathbf{r},t)=n_e(\mathbf{r},t)-N_{\mathrm{mean}}, \tag*{(34)}

relative to the mean-density level introduced for the matched initial state. This subtraction is a diagnostic decomposition of the complete full-ff density and must not be identified with the δf\delta f fluctuation n~e\widetilde{n}_e. Denoting the appropriate tracking field by

ntrk={n~e,δf,Δne,full-f,(35)n_{\mathrm{trk}}= \begin{cases} \widetilde{n}_e, & \delta f,\\ \Delta n_e, & \text{full-}f, \end{cases} \tag*{(35)}

the center of vortex jj at output time tit_i is computed inside a disk of radius ρ\rho around its previous position,

ci(j)=∑∣r−ci−1(j)∣≤ρr ntrk(r,ti)∑∣r−ci−1(j)∣≤ρntrk(r,ti).(36)\mathbf{c}_i^{(j)}= \frac{\sum_{|\mathbf{r}-\mathbf{c}_{i-1}^{(j)}|\le\rho}\mathbf{r}\,n_{\mathrm{trk}}(\mathbf{r},t_i)} {\sum_{|\mathbf{r}-\mathbf{c}_{i-1}^{(j)}|\le\rho}n_{\mathrm{trk}}(\mathbf{r},t_i)}. \tag*{(36)}

In the tracking procedure used here, the local radius is typically chosen in the range ρ=5\rho=5–1515 grid cells, with smaller values used when strong deformation or breakup requires tighter localization. The separation diagnostic is

R(t)=∣c(2)(t)−c(1)(t)∣.(37)R(t)=\left|\mathbf{c}^{(2)}(t)-\mathbf{c}^{(1)}(t)\right|. \tag*{(37)}

Early-time accelerations are extracted from a quadratic fit to R(t)R(t), while the orbital angular frequency is obtained from a linear fit to the pair

angle

θ(t)=arctan⁡[c2(2)(t)−c2(1)(t)c1(2)(t)−c1(1)(t)](38)\theta(t)=\arctan\left[\frac{c_{2}^{(2)}(t)-c_{2}^{(1)}(t)}{c_{1}^{(2)}(t)-c_{1}^{(1)}(t)}\right] \tag*{(38)}

with a continuous branch chosen over the initial constant-frequency fit interval.

Analytical vortex-interaction model

From interaction energy to orbital frequency

For a two-dimensional incompressible flow with stream function ψ\psi and vorticity Ω=∇2ψ\Omega=\nabla^{2}\psi, the vortex-pair interaction can be expressed through a Hamiltonian functional [5, 6, 7]

H[Ω]=−12∫d2r∫d2r′ Ω(r)G(r−r′)Ω(r′)(39)H[\Omega]=-\frac{1}{2}\int\mathrm{d}^{2}r\int\mathrm{d}^{2}r'\,\Omega(\mathbf{r})G(\mathbf{r}-\mathbf{r}')\Omega(\mathbf{r}') \tag*{(39)}

For two identical vortices of circulation Γ\Gamma and a symmetric pair Hamiltonian H(R)H(R), the non-canonical point-vortex equations give

ω(R)=2ΓRdHdR.(40)\omega(R)=\frac{2}{\Gamma R}\frac{\mathrm{d}H}{\mathrm{d}R}. \tag*{(40)}

To obtain a form applicable to a general linear inversion kernel, we define the inverse-polarization kernel by

ψ^(k)=−G^(k)Ω^(k),(41)\hat{\psi}(\mathbf{k})=-\hat{G}(\mathbf{k})\hat{\Omega}(\mathbf{k}), \tag*{(41)}

so that G^(k)=1/k2\hat{G}(k)=1/k^{2} in the Euler limit. For two identical axisymmetric vortices, Fourier transformation and angular integration yield a Bessel representation of the pair interaction. A direct interaction-energy calculation for two extended cores contains the product ∣Ω^0(k)∣2\lvert\hat{\Omega}_{0}(k)\rvert^{2}. For Gaussian cores this gives a different finite-core factor from the standard velocity-at-the-center result because the interaction energy averages over both extended distributions. The numerical diagnostic used here instead follows the center of one vortex in the velocity field generated by the other. We therefore use the corresponding single-spectrum form,

ωeff(R)=1πΓR∫0∞dk k2G(k)Ω^0(k)J1(kR),(42)\omega_{\mathrm{eff}}(R)=\frac{1}{\pi\Gamma R}\int_{0}^{\infty}\mathrm{d}k\,k^{2}G(k)\hat{\Omega}_{0}(k)J_{1}(kR), \tag*{(42)}

where J1J_{1} is the Bessel function of the first kind. For the Euler kernel this choice recovers the standard Gaussian finite-core law ω=Γ(1−e−R2/a2)/(πR2)\omega=\Gamma(1-e^{-R^{2}/a^{2}})/(\pi R^{2}) [5]. For a Gaussian vortex

Ω0(r)=Ω0exp⁡(−r2/a2),Γ=πa2Ω0,(43)\Omega_{0}(r)=\Omega_{0}\exp(-r^{2}/a^{2}),\qquad\Gamma=\pi a^{2}\Omega_{0}, \tag*{(43)}

its normalized spectral form contributes the finite-core factor exp⁡(−a2k2/4)\exp(-a^{2}k^{2}/4).

Reduced gyrofluid response kernel

A closed analytical kernel can be obtained from a deliberately reduced electrostatic δf\delta f polarization response. Starting from equation (11), we retain ion gyroaveraging but replace the potential-side polarization response (Γ0−1)/τi(\Gamma_0-1)/\tau_i by its long-wavelength form ∇⊥2\nabla_\perp^2, thereby neglecting its finite-k⊥ρik_\perp\rho_i gyroscreening. In Fourier space this gives

−k2ϕ^(k)=n^e(k)−11+τi2k2N^i(k).(44)-k^2\hat{\phi}(\boldsymbol{k})=\hat{n}_e(\boldsymbol{k})-\frac{1}{1+\frac{\tau_i}{2}k^2}\hat{N}_i(\boldsymbol{k}). \tag*{(44)}

Introducing the generalized source

N^=(1+τi2k2)n^e−N^i,(45)\hat{N}=\left(1+\frac{\tau_i}{2}k^2\right)\hat{n}_e-\hat{N}_i, \tag*{(45)}

we obtain

Ω^=N^1+τi2k2(46)\hat{\Omega}=\frac{\hat{N}}{1+\frac{\tau_i}{2}k^2} \tag*{(46)}

and the effective algebraic interaction kernel

G^τi(k)=1k2(1+τi2k2).(47)\hat{G}_{\tau_i}(k)=\frac{1}{k^2\left(1+\frac{\tau_i}{2}k^2\right)}. \tag*{(47)}

Using partial fractions, the corresponding real-space kernel is

Gτi(r)=−12πln⁡r−12πK0(2τi r)+C,(48)G_{\tau_i}(r)=-\frac{1}{2\pi}\ln r-\frac{1}{2\pi}K_0\left(\sqrt{\frac{2}{\tau_i}}\,r\right)+C, \tag*{(48)}

with K0K_0 the modified Bessel function of the second kind [39 Chapter 10]. Its radial derivative is

dGτidr=−12πr+12π2τiK1(2τi r).(49)\frac{\mathrm{d}G_{\tau_i}}{\mathrm{d}r}=-\frac{1}{2\pi r}+\frac{1}{2\pi}\sqrt{\frac{2}{\tau_i}}K_1\left(\sqrt{\frac{2}{\tau_i}}\,r\right). \tag*{(49)}

The additional Bessel contribution regularizes the r→0r\to0 Euler singularity on the characteristic scale

ℓFLR=τi2 ρs,(50)\ell_{\mathrm{FLR}}=\sqrt{\frac{\tau_i}{2}}\,\rho_s, \tag*{(50)}

while the far field remains logarithmic and therefore Euler-like, as illustrated in Figure 2. For two point vortices, equations (40) and (49) give

Interaction kernel and radial derivative for the reduced $\delta f$ gyrofluid polarization response

Figure 2. Interaction kernel Gτi(r)G_{\tau_i}(r) and radial derivative ∂rGτi\partial_rG_{\tau_i} for the reduced δf\delta f gyrofluid polarization response. In the cold-ion limit the interaction approaches the logarithmic Euler point-vortex potential. Finite τi\tau_i removes the short-range singularity on the scale ℓFLR∼τi/2ρS\ell_{\mathrm{FLR}} \sim\sqrt{\tau_i/2}\rho_S while preserving the asymptotic far-field behavior.

ωpoint(R)=ΓπR2[1−xK1(x)],x=R2τi.(51)\omega_{\mathrm{point}}(R)=\frac{\Gamma}{\pi R^2}\left[1-xK_1(x)\right],\qquad x=R\sqrt{\frac{2}{\tau_i}}. \tag*{(51)}

For the finite Gaussian cores used in the simulations, equations (42) and (47) instead give

ωeff(R;a,τi)=ΓπR∫0∞J1(kR)e−a2k2/41+τi2k2 dk(52)\omega_{\mathrm{eff}}(R;a,\tau_i)=\frac{\Gamma}{\pi R}\int_0^\infty\frac{J_1(kR)e^{-a^2k^2/4}}{1+\frac{\tau_i}{2}k^2}\,\mathrm{d}k \tag*{(52)}

which reduces to the finite-core Euler result for τi→0\tau_i \to0 and to the point-pair FLR expression for a→0a \to0. Expanding the polarization response to first non-trivial order at long wavelength gives

ωLWL(R;a,τi)=ΓπR∫0∞dk J1(kR)e−a2k2/4(1−τi2k2).(53)\omega_{\mathrm{LWL}}(R; a, \tau_i) = \frac{\Gamma}{\pi R} \int_0^\infty\mathrm{d}k\, J_1(kR)e^{-a^2k^2/4}\left(1-\frac{\tau_i}{2}k^2\right). \tag*{(53)}

Numerical evaluation predicts an FLR-dependent dip in the orbital angular frequency. This prediction is tested below.

δf\delta f vortex interaction

The local-model results are presented first in detail. For a direct side-by-side comparison with the finite-amplitude full-ff dynamics, the principal findings of both descriptions are summarized at the beginning of section 5.

Warm-ion parameter scan and arrested merger

The δf\delta f implementation first reproduces the established cold-ion co-rotation and the warm-ion gyrospinning, merger enhancement or repulsion reported in [30]. Unless stated otherwise, these δf\delta f vortex scans use hyperviscosity ν=10−6\nu= 10^{-6}. We then extend the co-rotating positive-vorticity case over a wider range of separation and τi\tau_i. Representative vorticity fields across the scan are shown in Figure 3, while the corresponding separation histories are summarized in Figure 4. Cold-ion pairs proceed toward complete coalescence. At finite ion temperature the vortices spin up and develop thin surrounding vorticity sheets. For a given initial separation, sufficiently strong finite-ion-temperature effects can instead prevent complete merger, leaving a long-lived co-rotating pair at approximately constant separation; the required τi\tau_i shifts with the initial separation.

Evolved vorticity for positive initial vorticity amplitude at different initial separations and ion-to-electron temperature ratios

Figure 3. δf\delta f evolved vorticity for positive initial vorticity amplitude at the common time t=9000L⊥/cSt = 9000 L_{\perp}/c_{S}, displayed for different initial separations (rows) and ion-to-electron temperature ratios τi\tau_i (columns). Cold-ion cases merge fully, whereas the largest-τi\tau_i cases form a long-lived co-rotating pair at approximately constant separation.

Time evolution of the normalized separation of two positive-vorticity delta-f vortices

Figure 4. Time evolution of the normalized separation of two positive-vorticity δf\delta f vortices for different τi\tau_i, including the early-time evolution. Larger τi\tau_i initially decreases the separation more rapidly, but can also cause the vortices to interlock and remain at an approximately constant distance. The scan shows a transition between merging and long-lived co-rotation, but its location in τi\tau_i depends on the initial vortex separation and should not be interpreted as a universal threshold.

The extracted arrested-orbit distances increase monotonically with ion temperature over the investigated orbiting cases (Table 1). The present data do not justify a unique scaling law for this distance or a universal critical value of τi\tau_i: the location of the merger–orbit crossover shifts with the initial vortex separation. The transition should therefore be interpreted as a numerically observed, separation-dependent change of regime rather than as a fitted critical threshold or exponent.

τi\tau_iorbiting separation Rorb/aR_{\mathrm{orb}}/a
43.2
53.4
7.53.6
103.8

Table 1. Measured asymptotic separation of the long-lived co-rotating δf\delta f vortex pairs.

The early-time cold-ion acceleration shows two additional empirical trends in the investigated parameter range: an approximate R−6R^{-6} dependence on initial separation and an approximately quadratic dependence on the initial vorticity amplitude. At finite τi\tau_i the amplitude dependence becomes substantially more complicated. We therefore do not use these empirical fits as a theoretical scaling law; their main role is to demonstrate that warm-ion interaction cannot be represented by a simple multiplicative correction to the cold-fluid pair dynamics.

Small-scale vorticity and the merger transition

The change from merger to arrested co-rotation is accompanied by a redistribution of vorticity toward finer spatial scales. We quantify this through the total enstrophy, the cumulative dissipative enstrophy loss and the fraction of enstrophy above kc=0.75kmax⁡k_c = 0.75k_{\max}. The total enstrophy is defined as

Z(t)=12∫AdA Ω2.(54)Z(t) = \frac{1}{2} \int_A \mathrm{d}A\,\Omega^2. \tag*{(54)}

To quantify the enstrophy removed by hyperviscous dissipation, we use the cumulative dissipative enstrophy

Zcum(t)=∫0tdt′ ν∫AdA(∇⊥2Ω)2.(55)Z_{\mathrm{cum}}(t) = \int_{0}^{t} dt' \, \nu\int_{A} dA \left(\nabla_{\perp}^{2}\Omega\right)^{2}. \tag*{(55)}

The fraction of enstrophy contained at high perpendicular wavenumbers is defined as

fhigh=∫kc∞dk ∣Ω^(k)∣2∫0∞dk ∣Ω^(k)∣2,kc=0.75 kmax⁡.(56)f_{\mathrm{high}} = \frac{ \int_{k_c}^{\infty} dk \, \lvert\widehat{\Omega}(k)\rvert^{2}}{ \int_{0}^{\infty} dk \, \lvert\widehat{\Omega}(k)\rvert^{2}}, \qquad k_c = 0.75\,k_{\max}. \tag*{(56)}

As shown in Figure 5, larger τi\tau_i generates stronger early FLR spin-up and a larger high-kk fraction. Within a fixed-separation scan, the marked increase in fine-scale enstrophy coincides with entry into the arrested-orbit regime; the corresponding value of τi\tau_i is separation dependent. On the merging side of the crossover, subsequent density-core interaction continues to generate vorticity during merger. On the long-lived co-rotating side, the interaction is increasingly dominated by the encounter of the enveloping vorticity sheets, while later enstrophy production becomes weak. This supports the interpretation that sufficiently developed FLR-generated sheets inhibit further mixing of the density cores and thereby arrest complete coalescence.

Time evolution of total enstrophy, cumulative dissipative enstrophy loss, and fraction of enstrophy above $k_c = 0.75k_{\max}$

Figure 5. Time evolution of total enstrophy, cumulative dissipative enstrophy loss and the fraction of enstrophy above kc=0.75kmax⁡k_c = 0.75k_{\max} for positive-vorticity δf\delta f vortex pairs at different τi\tau_i. The enhanced high-wavenumber content at larger ion temperature accompanies the transition to long-lived co-rotation.

Orbital frequency compared with the analytical prediction

The finite-core theory of § 3 predicts an FLR-dependent modification of the orbital frequency. The comparison shown in Figure 6 uses direct δf\delta f long-wavelength-limit (LWL) simulations in a (32ρS)2(32\rho_S)^2 domain with 102421024^2 grid points, Gaussian vortex width a=2ρSa = 2\rho_S and initial vorticity amplitude Ω0=0.01\Omega_0 = 0.01. The vortex-center frequency is extracted directly from these LWL runs, without any spectral filtering or post-processing modification of the polarization response. The comparison therefore provides a direct numerical test of the long-wavelength analytical prediction for the same parameter set.

Orbital angular frequency as a function of initial separation for several ion-temperature ratios

Figure 6. Orbital angular frequency as a function of initial separation RR for several τi\tau_i, compared with the long-wavelength finite-core prediction equation (53). The simulation data were obtained using the δf\delta f model with LWL polarization and (Lx,Ly)=(32ρS,32ρS)(L_x,L_y)=(32\rho_S,32\rho_S), a numerical resolution of 102421024^2, a=2ρSa=2\rho_S and Ω0=0.01\Omega_0=0.01.

Within this qualification, the simulated orbital frequencies follow the theoretical curves well over the scanned initial separations and reproduce the predicted τi\tau_i-dependent dip. The agreement supports the main physical premise of the analytical model: FLR physics can be interpreted, at the level of the reduced vortex-pair problem, as a modification of the response kernel that connects the generalized source to the advecting electrostatic potential.

From δf\delta f to finite-amplitude full-ff interaction

Key findings: δf\delta f versus full-ff. The principal numerical results can be read most directly by comparing the two descriptions point by point:

  • Polarization regime. δf\delta f: the background is fixed in the local polarization response and the approximation is controlled for small fluctuations. full-$f: the complete

gyrocenter density enters the nonlinear polarization operator, so the complete density dependence participates dynamically once the localized amplitude reaches n0/Nmean=O(1)n_0/N_{\mathrm{mean}} = \mathcal{O}(1).

  • Small-amplitude limit. δf\delta f: provides the reference local dynamics. full-ff: the generalized initialization converges toward the δf\delta f separation histories as n0/Nmean→0n_0/N_{\mathrm{mean}} \to0; in the investigated cases the trajectories are already close for n0/Nmean=0.01n_0/N_{\mathrm{mean}} = 0.01, while warm-ion deviations remain visible at n0/Nmean=0.1n_0/N_{\mathrm{mean}} = 0.1.

  • FLR-mediated interaction. δf\delta f: FLR modifies the effective interaction kernel, drives gyrospinning and generates thin enveloping vorticity sheets. Positive-vorticity pairs change from complete merger to long-lived co-rotation as ion temperature is increased, but the crossover value of τi\tau_i shifts with the initial vortex separation. Across the orbiting cases in the present scan, Rorb/aR_{\mathrm{orb}}/a increases with τi\tau_i. full-ff: the same FLR dynamics acts together with an additional finite-amplitude mean-density polarization contribution, which can delay or arrest the ordinary merger and qualitatively alter the subsequent interaction.

  • Density–vorticity sign symmetry. δf\delta f: at τi=0\tau_i = 0, DϕΩ=0D_\phi\Omega= 0 and reversal of the density perturbation at fixed Ω0\Omega_0 is a symmetry of the vorticity dynamics; at finite τi\tau_i, the FLR source ∝τi[Ω,n‾e]⊥\propto\tau_i[\Omega,\overline{n}_e]_\perp changes sign and breaks this equivalence, as established in Ref. [30]. full-ff: the NOB contribution is bilinear in the density departure and electric field, so the same density reversal is already non-equivalent at τi=0\tau_i = 0. Warm ions therefore combine two distinct parity-breaking mechanisms: nonlinear NOB polarization and local FLR density–vorticity coupling.

  • Small-scale structure and wavelength response. δf\delta f: increasing τi\tau_i enhances the high-kk enstrophy fraction, and the onset of strongly developed FLR sheets coincides with the merger–orbit transition; direct LWL simulations reproduce the predicted orbital-frequency dip. full-ff: mean-density polarization adds shielded opposite-sign vorticity. For τi=1\tau_i = 1, arbitrary-wavelength NOB polarization retains finer sheets and produces an extended orbit followed by rapid repulsion, whereas NOB+LWL develops coarser structures and vortex shedding for positive initial vorticity; corresponding shedding is not observed in the arbitrary-wavelength NOB case over the simulated interval.

  • Physical interpretation. δf\delta f: finite ion temperature changes how one vortex advects the other through an FLR-modified response kernel. full-ff: finite amplitude changes the vorticity field that mediates the interaction itself through the complete density-dependent polarization response; the global dynamics therefore cannot, in general, be represented by a simple correction to vortex speed or inertia.

Recovery of the small-amplitude limit

A necessary validation of the generalized initialization is that the full-ff evolution approaches the δf\delta f result as the localized density amplitude is reduced relative to NmeanN_{\mathrm{mean}}. This is shown in Figure 7. At n0/Nmean=0.1n_0/N_{\mathrm{mean}} = 0.1, the cold-ion trajectories are already close while warm-ion cases still develop significant deviations at later time. At n0/Nmean=0.01n_0/N_{\mathrm{mean}} = 0.01, with the vorticity amplitude reduced by the same factor, the δf\delta f and full-ff separation curves are substantially more similar. The full-ff evolution is slightly delayed, consistent with the additional inertia associated with the finite mean-density contribution to polarization.

Separation histories for co-rotating positive-vorticity pairs for $n_0/N_{\mathrm{mean}}=0.1$, $\Omega_0=0.01$

(a) n0/Nmean=0.1n_0/N_{\mathrm{mean}} = 0.1, Ω0=0.01\Omega_0 = 0.01.

Separation histories for co-rotating positive-vorticity pairs for $n_0/N_{\mathrm{mean}}=0.01$, $\Omega_0=0.001$

(b) n0/Nmean=0.01n_0/N_{\mathrm{mean}} = 0.01, Ω0=0.001\Omega_0 = 0.001.

Figure 7. Separation histories for co-rotating positive-vorticity pairs in δf\delta f and full-ff, demonstrating recovery of the local limit as the localized density amplitude decreases relative to NmeanN_{\mathrm{mean}}. Curves with the Nmean∇⊥2ϕGN_{\mathrm{mean}}\nabla_{\perp}^{2}\phi_G contribution removed isolate the leading-order term responsible for the local δf\delta f limit.

The same comparison also identifies the term responsible for the local limit. If the mean-density contribution proportional to Nmean∇⊥2ϕGN_{\mathrm{mean}}\nabla_{\perp}^{2}\phi_G is removed from the initial full-ff polarization balance, the corresponding vortices remain nearly stationary in the small-amplitude cases. This is consistent with an amplitude ordering in which the mean-density term is O(ϵ)\mathcal{O}(\epsilon) while the additional density-dependent polarization terms are O(ϵ2)\mathcal{O}(\epsilon^{2}). Thus the constant-background polarization equation of the δf\delta f model emerges as the leading-order contribution of the full-ff equation, even though NmeanN_{\mathrm{mean}} itself is not a static full-ff background field.

Finite-amplitude mean-density polarization

We next consider a finite-amplitude pair with n0=Nmean=1n_0 = N_{\mathrm{mean}} = 1, a=6ρSa = 6\rho_S, R0=4aR_0 = 4a, ∣Ω0∣=0.01|\Omega_0| = 0.01, domain (128ρS)2(128\rho_S)^2 and 5122512^2 grid points. In this regime the localized density excess is comparable to the mean-density level and a full-ff treatment is required. The hyperviscosity is ν=10−5\nu= 10^{-5}, a factor of ten larger than in the corresponding δf\delta f studies, because substantially finer scales develop in the finite-amplitude calculation.

For cold ions, the generalized full-ff initialization produces a strong departure from the local dynamics, as shown in Figure 8. Opposite-sign vorticity is generated around the original cores, the positive–negative symmetry breaks early, and merger is delayed. Positive vortices merge only around t≃15000 L⊥/cSt \simeq15000\,L_{\perp}/c_S, while the negative pair remains in the merger process at the end of the run, t=20000 L⊥/cSt = 20000\,L_{\perp}/c_S. The origin of the delay can be isolated by subtracting the mean-density component −Nmean∇⊥2ϕG-N_{\mathrm{mean}}\nabla_{\perp}^{2}\phi_G from the initial vorticity balance, as shown in Figure 9. What remains is the additional vorticity generated by the density-dependent part of the full-ff polarization response. It forms a shielded structure with opposite-sign surrounding vorticity and generates an initially repulsive contribution. The full finite-amplitude dynamics can therefore be interpreted as a competition between the ordinary vortex interaction and a nonlinear mean-density/density-variation polarization response absent from the local model.

Finite-amplitude full-$f$ cold-ion co-rotating vortex evolution for positive and negative initial vorticity

Figure 8. Finite-amplitude full-ff cold-ion (τi=0\tau_i = 0) co-rotating vortex evolution for positive and negative initial vorticity. Opposite-sign vorticity is generated around the original cores and the positive–negative symmetry breaks early. The positive-vorticity pair merges at approximately t=15000L⊥/cSt = 15000 L_\perp/c_S, whereas the negative-vorticity pair is still merging at t=20000L⊥/cSt = 20000 L_\perp/c_S. Parameters: (Lx,Ly)=(128ρS,128ρS)(L_x, L_y) = (128\rho_S, 128\rho_S), numerical resolution 5122512^2, Ω0=δ=±0.01\Omega_0 = \delta= \pm0.01, n0=Nmean=1n_0 = N_{\mathrm{mean}} = 1, a=6ρSa = 6\rho_S, R0=4aR_0 = 4a, and ν=10−5\nu= 10^{-5}. The plotted nx,nyn_x, n_y axes denote grid indices.

Finite-amplitude cold-ion initialization with the mean-density component subtracted

Figure 9. Same finite-amplitude cold-ion initialization as Figure 8, with the mean-density component −Nmean∇⊥2ϕG-N_{\mathrm{mean}}\nabla_{\perp}^{2}\phi_G subtracted initially. The remaining contribution is the additional vorticity generated by the density-dependent part of the full-ff polarization response. The resulting shielded-vorticity structure exhibits positive–negative symmetry and drives early mutual repulsion.

Density–vorticity symmetry breaking at finite amplitude

The generalized full-ff initialization allows the density and vorticity signs to be varied independently. With

G(x,y)≡g1(x,y)+g2(x,y),Δne,0=n0G,Ω0=δG,(57)G(x,y) \equiv g_1(x,y) + g_2(x,y), \qquad\Delta n_{e,0} = n_0G, \qquad\Omega_0 = \delta G, \tag*{(57)}

we define two elementary sign operations,

Sn:(n0,δ)↦(−n0,δ),(Δne,0,Ω0)↦(−Δne,0,Ω0),(58)S_n : (n_0,\delta) \mapsto(-n_0,\delta), \qquad(\Delta n_{e,0},\Omega_0) \mapsto(-\Delta n_{e,0},\Omega_0), \tag*{(58)}
SΩ:(n0,δ)↦(n0,−δ),(Δne,0,Ω0)↦(Δne,0,−Ω0).(59)S_\Omega: (n_0,\delta) \mapsto(n_0,-\delta), \qquad(\Delta n_{e,0},\Omega_0) \mapsto(\Delta n_{e,0},-\Omega_0). \tag*{(59)}

Their composition SnΩ=SnSΩS_{n\Omega} = S_nS_\Omega reverses both the localized density departure and the circulation. For the scan shown in Figure 10, it is convenient to denote

Independent density–vorticity sign scan for finite-amplitude full-$f$ vortex pairs

Figure 10. Independent density–vorticity sign scan for finite-amplitude full-ff vortex pairs. The four rows correspond to the states B+B_{+}, B−B_{-}, H+H_{+} and H−H_{-} defined in equation (60). Comparing BσB_{\sigma} with HσH_{\sigma} applies SnS_{n} and reverses the localized density departure at fixed Ω0\Omega_{0}; comparing opposite subscripts applies SΩS_{\Omega} and reverses the circulation at fixed density character. Columns show the cold-ion case (τi=0\tau_{i}=0), warm-ion arbitrary-wavelength NOB polarization (τi=1\tau_{i}=1), and warm-ion NOB+LWL polarization (τi=1\tau_{i}=1). The cold-ion column already violates the local SnS_{n} symmetry through finite-amplitude NOB polarization, whereas the warm-ion columns combine this effect with the FLR symmetry breaking of the local gyrofluid dynamics. The ×10−2\times10^{-2} annotations belong to the vorticity color scales.

Bσ:(n0,δ)=(+0.5,σ 0.01),Hσ:(n0,δ)=(−0.5,σ 0.01),σ=±1,(60)B_\sigma: (n_0,\delta) = (+0.5,\sigma\,0.01), \qquad H_\sigma: (n_0,\delta) = (-0.5,\sigma\,0.01), \qquad\sigma= \pm1, \tag*{(60)}

where BB and HH denote density blobs and holes, while the subscript gives the sign of the initial vorticity. Thus

Sn:Bσ↔Hσ,SΩ:B+↔B−, H+↔H−,(61)S_n : B_\sigma\leftrightarrow H_\sigma, \qquad S_\Omega: B_+ \leftrightarrow B_-, \ H_+ \leftrightarrow H_-, \tag*{(61)}

provides two independent tests of density and circulation parity.

In the cold-ion local model, DϕΩ=0D_\phi\Omega= 0 [equation (12)], so SnS_n is invisible to the vorticity dynamics: BσB_\sigma and HσH_\sigma have identical Ω0\Omega_0 and therefore the same Euler-like evolution. At finite τi\tau_i, this local density symmetry is broken at leading FLR order by equation (13); under SnS_n,

W=Ω+τi2∇⊥2n~e↦Ω−τi2∇⊥2n~e,τi2[Ω,n~e]⊥↦−τi2[Ω,n~e]⊥,(62)W = \Omega+ \frac{\tau_i}{2}\nabla_\perp^2\tilde n_e \mapsto\Omega- \frac{\tau_i}{2}\nabla_\perp^2\tilde n_e, \qquad\frac{\tau_i}{2}[\Omega,\tilde n_e]_\perp\mapsto-\frac{\tau_i}{2}[\Omega,\tilde n_e]_\perp, \tag*{(62)}

which is the Kendl symmetry breaking of the local warm-ion system [30].

Full-ff polarization adds a second mechanism. In the LWL form of the NOB field equation,

Q=NmeanΩ+ΔNiΩ+∇⊥ΔNi⋅∇⊥ϕ(63)Q = N_{\mathrm{mean}}\Omega+ \Delta N_i\Omega+ \nabla_\perp\Delta N_i\cdot\nabla_\perp\phi \tag*{(63)}

only the first term survives the local OB reduction. The nonlinear contribution QNOBQ_{\mathrm{NOB}} depends explicitly on the reconstructed density departure. Consequently, at fixed prescribed Ω0\Omega_0, the transformation SnS_n requires a different NiN_i reconstruction and need not preserve the field inversion. Likewise, under the simultaneous finite-amplitude sign reversal (ΔNi,ϕ)↦(−ΔNi,−ϕ)(\Delta N_i,\phi) \mapsto(-\Delta N_i,-\phi), the OB term is odd but QNOBQ_{\mathrm{NOB}} is even [equations (20) and (21)]. Hence the full-ff/NOB system has no exact blob–hole parity even at τi=0\tau_i = 0.

This separation is visible directly in Figure 10. In the cold-ion column, the pairs

B+≁H+,B−≁H−,(64)B_{+} \not\sim H_{+}, \qquad B_{-} \not\sim H_{-}, \tag*{(64)}

show that reversing the density departure at fixed initial vorticity already produces different finite-amplitude vorticity fields; this difference cannot originate from ion-FLR gyroaveraging. At τi=1\tau_i = 1, the same NOB asymmetry is superposed with the local FLR mechanism of equation (62). The arbitrary-wavelength NOB column therefore contains both symmetry-breaking contributions, whereas the NOB+LWL column removes part of the short-wavelength polarization response while retaining the nonlinear density dependence. The resulting morphologies confirm that the two mechanisms coexist rather than replace one another: full-ff polarization breaks density parity through NiN_i in the field inversion, while finite-ρi\rho_i dynamics remains sensitive to the relative density–vorticity configuration.

The scan may therefore be summarized schematically as

τi=0, δf:Sn symmetry,τi>0, δf:Sn broken by FLR,τi=0, full-f:Sn broken by NOB polarization,τi>0, full-f:NOB and FLR symmetry breaking coexist.(65)\begin{aligned} \tau_i = 0,\ \delta f &: \quad S_n\ \text{symmetry},\\ \tau_i > 0,\ \delta f &: \quad S_n\ \text{broken by FLR},\\ \tau_i = 0,\ \text{full-}f &: \quad S_n\ \text{broken by NOB polarization},\\ \tau_i > 0,\ \text{full-}f &: \quad\text{NOB and FLR symmetry breaking coexist}. \tag*{(65)} \end{aligned}

Thus the initial vorticity field Ω0(r)\Omega_0(\boldsymbol{r}) alone is insufficient to determine a finite-amplitude gyrofluid vortex trajectory: the complete density configuration enters both through the nonlinear polarization inertia and, for warm ions, through the FLR-modified advection.

Arbitrary-wavelength (second order Padé) polarization effects

For τi=1\tau_i = 1 the additional mean-density polarization contribution competes with FLR gyrospinning. The full arbitrary-wavelength NOB and NOB+LWL polarizations can then produce markedly different finite-amplitude evolutions, particularly for positive initial vorticity; the corresponding arbitrary-wavelength NOB evolution is shown in Figure 11. In the NOB case, finely structured enveloping vorticity sheets are retained and the vortices orbit at nearly constant separation up to approximately t=1750L⊥/cSt = 1750 L_\perp/c_S. The separation subsequently decreases slightly before a rapid mutual repulsion at t≃3750L⊥/cSt \simeq3750 L_\perp/c_S. The initial orbit indicates that the attractive component associated with FLR spin-up can temporarily balance the repulsive mean-density polarization contribution.

Finite-amplitude full-$f$ warm-ion co-rotating vortex evolution

Figure 11. Finite-amplitude full-ff warm-ion (τi=1\tau_i = 1) co-rotating vortex evolution with arbitrary-wavelength NOB polarization for positive and negative initial vorticity. Parameters: (Lx,Ly)=(128ρS,128ρS)(L_x, L_y) = (128\rho_S, 128\rho_S), numerical resolution 5122512^2, Ω0=δ=±0.01\Omega_0 = \delta= \pm0.01, n0=Nmean=1n_0 = N_{\mathrm{mean}} = 1, a=6ρSa = 6\rho_S, R0=4aR_0 = 4a, and ν=10−5\nu= 10^{-5}. The plotted nx,nyn_x, n_y axes denote grid indices.

The NOB+LWL case, shown in Figure 12, suppresses part of the short-scale polarization structure. In the simulations this results in coarser evolved structures and pronounced vortex-shedding behavior reminiscent of a Kármán-type street; no corresponding shedding is observed in the arbitrary-wavelength NOB evolution over the simulated interval. For negative initial vorticity, by contrast, the NOB and NOB+LWL trajectories are more similar, consistent with the repelling FLR branch already dominating the interaction. The strong sign dependence emphasizes that the finite-amplitude correction is not a uniform change in inertia or interaction strength: it changes the spatial vorticity structure that then interacts nonlinearly with the FLR-generated sheets. The corresponding separation histories are collected in Figure 13.

Same warm-ion finite-amplitude co-rotating setup as Figure 11, evolved with NOB+LWL polarization

Figure 12. Same warm-ion finite-amplitude co-rotating setup as Figure 11, evolved with NOB+LWL polarization. For the positive-vorticity branch, the LWL model develops coarser structures and pronounced vortex-shedding behavior, which is not observed in the arbitrary-wavelength NOB evolution over the simulated interval; the negative-vorticity NOB and NOB+LWL evolutions are more similar.

Separation histories for the generalized finite-amplitude full-$f$ co-rotating simulations

Figure 13. Separation histories for the generalized finite-amplitude full-ff co-rotating simulations. The warm-ion NOB case exhibits an extended nearly constant-separation orbit followed by rapid repulsion, illustrating the competition between FLR-induced attraction and the repulsive mean-density polarization component.

Discussion

The δf\delta f and full-ff calculations expose two distinct ways in which gyrofluid polarization changes an otherwise familiar two-dimensional vortex-pair problem. Within the local model, finite ion temperature modifies the response kernel and generates gyrospinning, thin vorticity sheets and a separation-dependent orbital response. The analytical model isolates this effect by replacing the Laplacian Green function with equation (47). Although the closed form was obtained from a reduced polarization relation, the good direct comparison with the LWL orbital frequencies indicates that the kernel picture captures a meaningful part of the interaction physics. The observed dip and the arrested-merger regime should nevertheless not be conflated: the former is a prediction of the reduced LWL interaction model, whereas the latter is a nonlinear morphological transition in the full arbitrary-wavelength δf\delta f simulations.

The full-ff extension introduces an additional mechanism. Because the complete NiN_i multiplies the polarization response, the finite mean-density level and the localized density variation both participate in the field inversion. The mean-density contribution together with the nonlinear density-dependent terms produces vorticity outside the prescribed Gaussian cores. In the cold-ion case this additional contribution is directly visible as a shielded, initially repulsive structure. At finite τi\tau_i it competes with the FLR-generated vorticity sheets. This explains why the difference between local and global dynamics becomes qualitative once n0/Nmean=O(1)n_0/N_{\mathrm{mean}} = O(1), while the same generalized initialization correctly approaches the δf\delta f result as n0/Nmean→0n_0/N_{\mathrm{mean}} \to0.

The independent sign scan in Figure 10 makes the associated symmetry structure explicit. In the cold-ion local model, density does not enter the Euler vorticity equation, whereas the warm-ion δf\delta f system loses this density–vorticity equivalence through FLR polarization, as identified by Kendl [30]. In full-ff, a second symmetry breaking appears already at τi=0\tau_i = 0 because the spatially varying NiN_i multiplies the polarization response. At finite τi\tau_i the two mechanisms coexist rather than replacing one another: the nonlinear NOB field inversion distinguishes density enhancements from depletions, while gyroaveraging remains sensitive to the relative density–vorticity configuration. The absence of a simple sign-reversed partner in the warm-ion scan is therefore a direct consequence of these two parity-breaking contributions acting simultaneously.

The comparison between arbitrary-wavelength NOB and NOB+LWL polarization also suggests caution when using a long-wavelength closure in strongly filamenting vortex dynamics. In the present finite-amplitude positive-vorticity cases, suppressing the shorter-scale polarization response changes the morphology, stability and shedding of the vortices. The result does not imply that an LWL treatment is generally invalid; rather, it identifies a regime in which the fine vorticity sheets generated by FLR and finite-amplitude polarization participate directly in the subsequent interaction and therefore cannot be assumed dynamically passive.

Conclusions

We have used a deliberately elementary vortex-pair configuration to separate two effects that are often entangled in more complete plasma-turbulence simulations: FLR modification of the gyrofluid response and finite-amplitude modification of the polarization itself. The main results are:

R1. FLR physics can be expressed as an effective interaction kernel. The reduced δf\delta f polarization relation gives a finite-core orbital-frequency law in which the short-range Euler interaction is regularized on the FLR scale while the far field remains Euler-like. Direct δf\delta f LWL simulations reproduce the predicted τi\tau_i-dependent dip in the orbital frequency.

R2. Increasing ion temperature produces a distinct, separation-dependent merger–orbit transition. Positive-vorticity δf\delta f pairs change from complete merger to long-lived co-rotating states as τi\tau_i is increased, but the crossover value depends on the initial vortex separation and is not a universal critical temperature ratio. Within each scan, the onset of long-lived co-rotation coincides with a pronounced increase of enstrophy in the fine, FLR-generated vorticity sheets, indicating that these sheets become an active part of the interaction rather than a passive by-product of gyrospinning.

R3. The local δf\delta f model is recovered as a controlled small-amplitude limit of the generalized full-ff initialization. The mean-density contribution to the full-ff polarization is the leading O(ϵ)\mathcal{O}(\epsilon) term and yields the constant-background δf\delta f polarization at leading order, while the additional density-dependent nonlinear polarization terms enter at higher order. Accordingly, the full-ff and δf\delta f trajectories approach one another as the matched localized density amplitude is reduced.

R4. Finite amplitude introduces a second, genuinely full-ff symmetry breaking. Writing Δne,0=n0G\Delta n_{e,0}=n_0G and Ω0=δG\Omega_0=\delta G makes density reversal (n0,δ)↦(−n0,δ)(n_0,\delta)\mapsto(-n_0,\delta) and circulation reversal (n0,δ)↦(n0,−δ)(n_0,\delta)\mapsto(n_0,-\delta) independent operations. In the cold-ion local model the first is a symmetry because DϕΩ=0D_\phi\Omega=0; in full-ff it is broken already at τi=0\tau_i=0 by the nonlinear NOB terms ΔNiΩ+∇⊥ΔNi⋅∇⊥ϕ\Delta N_i\Omega+\nabla_\perp\Delta N_i\cdot\nabla_\perp\phi. Warm ions then superpose the local FLR symmetry breaking ∝τi[Ω,n~e]⊥\propto\tau_i[\Omega,\widetilde n_e]_\perp on this finite-amplitude NOB asymmetry.

R5. Arbitrary-wavelength response controls the morphology of the symmetry-broken warm-ion state. When the NOB and FLR mechanisms act simultaneously, retaining arbitrary-wavelength polarization preserves finer vorticity sheets and can produce trajectories that differ qualitatively from NOB+LWL evolution, including different repulsion and shedding behavior.

The central implication is that the passage from δf\delta f to full-ff cannot in general be represented by a simple correction to vortex speed or inertia. Once the localized full-ff density variation is comparable to the mean-density level, the nonlinear polarization equation changes the spatial vorticity field that mediates the interaction and removes the simple density–vorticity sign equivalence of the local cold-ion problem. In the controlled two-vortex system studied here, the combined NOB and FLR symmetry breaking is already sufficient to determine whether vortices merge, remain in a long-lived orbit, separate or shed secondary structures. These configurations therefore provide sensitive benchmark problems for future full-ff gyrofluid developments and a natural starting point for testing the same mechanisms in driven vortex ensembles and edge-turbulence models.

Data availability statement

The GREENY source code used for the simulations is available at https://git.uibk.ac.at/c7441315/greeny. The data that support the findings of this study are available from the corresponding author upon reasonable request.

Acknowledgements

F.F.L. acknowledges financial support by KKKÖ of ÖAW. M.H. acknowledges support by the UiT Aurora Centre Program, UiT The Arctic University of Norway (2020). This research was funded in whole or in part by the Austrian Science Fund (FWF) [P 34241-N]. The authors thank Lukas Einkemmer for insightful discussions and valuable advice concerning the numerical inversion of the polarization equation.

Author contributions

M. Rinner: Data curation (lead); Conceptualization (equal); Validation (equal); Software (equal); Visualization (lead); Writing – original draft (equal); Methodology (equal); Formal analysis (lead). F. F. Locker: Supervision (equal); Conceptualization (equal); Writing – original draft (equal); Software (equal); Methodology (equal); Visualization (support); Formal analysis (support). M. Held: Supervision (equal); Conceptualization (equal). A. Kendl: Supervision (equal); Resources (lead); Project administration (lead).

References

  1. [1]R. H. Kraichan and D. Montgomery. Two-dimensional turbulence. Reports on Progress in Physics, 43(5):547, 1980. doi: 10.1088/0034-4885/43/5/001.
  2. [2]Guido Boffetta and Robert E. Ecke. Two-dimensional turbulence. Annual Review of Fluid Mechanics, 44:427–451, 2012. doi: 10.1146/annurev-fluid-120710-101240.DOI
  3. [3]Shiyi Chen, Robert E. Ecke, Gregory L. Eyink, Michael Rivera, Minping Wan, and Zuoli Xiao. Physical mechanism of the two-dimensional inverse energy cascade. Physical Review Letters, 96(8):084502, 2006. doi: 10.1103/PhysRevLett.96.084502.DOI
  4. [4]C.-H. Bruneau, P. Fischer, and H. Kellay. The structures responsible for the inverse energy and the forward enstrophy cascades in two-dimensional turbulence. EPL (Europhysics Letters), 78(3):34002, 2007. doi: 10.1209/0295-5075/78/34002.DOI
  5. [5]P. G. Saffman. Vortex Dynamics. Cambridge Monographs on Mechanics. Cambridge University Press, Cambridge, 1993.
  6. [6]Paul K. Newton. The N-Vortex Problem: Analytical Techniques, volume 145 of Applied Mathematical Sciences. Springer, New York, 2001. doi: 10.1007/978-1-4684-9290-3.DOI
  7. [7]Hassan Aref. Point vortex dynamics: A classical mathematics playground. Journal of Mathematical Physics, 48(6):065401, 2007. doi: 10.1063/1.2425103.DOI
  8. [8]T. Leweke, S. Le Dizès, and C. H. K. Williamson. Dynamics and instabilities of vortex pairs. Annual Review of Fluid Mechanics, 48:507–541, 2016. doi: 10.1146/annurev-fluid-122414-034558.DOI
  9. [9]W. Horton and A. Hasegawa. Quasi-two-dimensional dynamics of plasmas and fluids. Chaos, 4(2):227–251, 1994. doi: 10.1063/1.166049.DOI
  10. [10]W. Horton. Drift waves and transport. Reviews of Modern Physics, 71(3):735–778, 1999. doi: 10.1103/RevModPhys.71.735.DOI
  11. [11]B. D. Scott, A. Kendl, and T. Ribeiro. Nonlinear dynamics in the tokamak edge. Contributions to Plasma Physics, 50(3–5):228–241, 2010. doi: 10.1002/ctpp.201010039.DOI
  12. [12]A. Kendl and B. D. Scott. Magnetic shear damping of dissipative drift wave turbulence. Physical Review Letters, 90(3):035006, 2003. doi: 10.1103/PhysRevLett.90.035006.DOI
  13. [13]Z. Lin, T. S. Hahm, W. W. Lee, W. M. Tang, and R. B. White. Turbulent transport reduction by zonal flows: Massively parallel simulations. Science, 281(5384):1835–1837, 1998. doi: 10.1126/science.281.5384.1835.DOI
  14. [14]T. S. Hahm. Nonlinear gyrokinetic equations for tokamak microturbulence. The Physics of Fluids, 31(9):2670–2673, 1988. doi: 10.1063/1.866544.
  15. [15]A. J. Brizard and T. S. Hahm. Foundations of nonlinear gyrokinetic theory. Reviews of Modern Physics, 79(2):421–468, 2007. doi: 10.1103/RevModPhys.79.421.DOI
  16. [16]John A. Krommes. The gyrokinetic description of microturbulence in magnetized plasmas. Annual Review of Fluid Mechanics, 44:175–201, 2012. doi: 10.1146/annurev-fluid-120710-101223.DOI
  17. [17]G. Knorr, F. R. Hansen, J. P. Lynov, H. L. Pécseli, and J. J. Rasmussen. Finite larmor radius effects to arbitrary order. Physica Scripta, 38(6):829–834, 1988. doi: 10.1088/0031-8949/38/6/014.DOI
  18. [18]W. Dorland and G. W. Hammett. Gyrofluid turbulence models with kinetic effects. Physics of Fluids B: Plasma Physics, 5(3):812–835, 1993. doi: 10.1063/1.860934.DOI
  19. [19]M. A. Beer and G. W. Hammett. Toroidal gyrofluid equations for simulations of tokamak turbulence. Physics of Plasmas, 3(11):4046–4064, 1996. doi: 10.1063/1.871538.DOI
  20. [20]Bruce D. Scott. Computation of electromagnetic turbulence and anomalous transport mechanisms in tokamak plasmas. Plasma Physics and Controlled Fusion, 45(12A):A385, 2003. doi: 10.1088/0741-3335/45/12A/025.
  21. [21]B. D. Scott. Free-energy conservation in local gyrofluid models. Physics of Plasmas, 12(10):102307, 2005. doi: 10.1063/1.2064968.DOI
  22. [22]B. Scott. Derivation via free energy conservation constraints of gyrofluid equations with finite- gyroradius electromagnetic nonlinearities. Physics of Plasmas, 17(10):102306, 2010. doi: 10.1063/1.3484219.
  23. [23]D. Pfirsch. New variational formulation of maxwell–vlasov and guiding center theories: Local charge and energy conservation laws. Zeitschrift für Naturforschung A, 39(1):1–8, 1984. doi: 10.1515/zna-1984-0102.
  24. [24]Alain J. Brizard. Beyond linear gyrocenter polarization in gyrokinetic theory. Physics of Plasmas, 20(9):092309, 2013. doi: 10.1063/1.4823716.DOI
  25. [25]Jens Madsen, Odd E. Garcia, Jeppe Stærk Larsen, Volker Naulin, Anders Henry Nielsen, and Jens Juul Rasmussen. The influence of finite larmor radius effects on the radial interchange motions of plasma filaments. Physics of Plasmas, 18(11):112504, 2011. doi: 10.1063/1.3658033.
  26. [26]M. Wiesenberger, J. Madsen, and A. Kendl. Radial convection of finite ion temperature, high amplitude plasma blobs. Physics of Plasmas, 21(9):092301, 2014. doi: 10.1063/1.4894220.DOI
  27. [27]Alexander Kendl. Inertial blob-hole symmetry breaking in magnetised plasma filaments. Plasma Physics and Controlled Fusion, 57(4):045012, 2015. doi: 10.1088/0741-3335/57/4/045012.DOI
  28. [28]M. Held, M. Wiesenberger, J. Madsen, and A. Kendl. The influence of temperature dynamics and dynamic finite ion larmor radius effects on seeded high amplitude plasma blobs. Nuclear Fusion, 56(12):126005, 2016. doi: 10.1088/0029-5515/56/12/126005.DOI
  29. [29]D. A. D’Ippolito, J. R. Myra, D. A. Russell, and G. Q. Yu. Rotational stability of plasma blobs. Physics of Plasmas, 11(10):4603–4609, 2004. doi: 10.1063/1.1785791.
  30. [30]Alexander Kendl. Gyrofluid vortex interaction. Plasma Physics and Controlled Fusion, 60(2):025017, 2018. doi: 10.1088/1361-6587/aa9f94.DOI
  31. [31]Masahiro Wakatani and Akira Hasegawa. A collisional drift wave description of plasma edge turbulence. The Physics of Fluids, 27(3):611–618, 1984. doi: 10.1063/1.864660.DOI
  32. [32]P. Manz, G. Birkenmeier, M. Ramisch, and U. Stroth. A link between nonlinear self-organization and dissipation in drift-wave turbulence. Physics of Plasmas, 19(8):082318, 2012. doi: 10.1063/1.4748143.DOI
  33. [33]Jens Madsen. Full-f gyrofluid model. Physics of Plasmas, 20(7):072301, 2013. doi: 10.1063/1.4813241.DOI
  34. [34]M. Held, M. Wiesenberger, and A. Kendl. Padé-based arbitrary wavelength polarization closures for full-f gyro-kinetic and -fluid models. Nuclear Fusion, 60(6):066014, 2020. doi: 10.1088/1741-4326/ab7a68.DOI
  35. [35]M. Held and M. Wiesenberger. Beyond the oberbeck–boussinesq and long wavelength approximation. Nuclear Fusion, 63(2):026008, 2023. doi: 10.1088/1741-4326/aca9e0.DOI
  36. [36]A. Kendl. TIFF: Gyrofluid turbulence in full-f and full-k. Computer Physics Communications, 294:108953, 2024. doi: 10.1016/j.cpc.2023.108953.DOI
  37. [37]M. Wiesenberger and M. Held. Effects of plasma resistivity in three-dimensional full-f gyro-fluid turbulence simulations. Plasma Physics and Controlled Fusion, 66(6):065003, 2024. doi: 10.1088/1361-6587/ad3670.DOI
  38. [38]F. F. Locker, M. Held, T. M. Stocker-Waldhuber, A. Stürz, M. Rinner, and A. Kendl. Greeny: A full-f 2d gyrofluid reconnection code. Computer Physics Communications, 316:109807, 2025. doi: 10.1016/j.cpc.2025.109807.DOI
  39. [39]DLMF. NIST Digital Library of Mathematical Functions. https://dlmf.nist.gov/, Release 1.2.7 of 2026-06-15, 2026. URL https://dlmf.nist.gov/. F. W. J. Olver, A. B. Olde Daalhuis, D. W. Lozier, B. I. Schneider, R. F. Boisvert, C. W. Clark, B. R. Miller, B. V. Saunders, H. S. Cohl, and M. A. McClain, eds.

Paper details

vlnc:W_66dy24r6
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 8

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