Gyro Fluid System in Plasma Research

• Gyro Fluid System is a family of gyroaveraged reduced models that simplify gyrokinetic dynamics into fluid-like moments with adjustable velocity-space resolution.
• It employs spectral methods such as Hermite and Laguerre expansions to capture key phenomena including drift-wave turbulence, reconnection, and pedestal stability.
• GFS frameworks facilitate rapid eigenmode analysis and transport studies, providing actionable insights for optimizing plasma confinement and stability.

Gyro Fluid System (GFS) denotes a class of reduced descriptions in which gyrokinetic dynamics are represented by gyrocenter fields or fluid-like moments, with closure through quasi-neutrality and, in electromagnetic formulations, Ampère’s law. In contemporary plasma research, GFS formulations are used for low-frequency magnetized-plasma turbulence, coherent-vortex dynamics, magnetic reconnection, and linear pedestal stability analysis; in a distinct relativistic usage, “gyrohydrodynamics” has also been presented as a Gyro Fluid System for charged spinful fluids under strong vorticity (Yang et al., 16 Sep 2025, Cao et al., 2022). Taken together, these usages suggest that GFS is best understood as a family of gyro-averaged reduced models rather than a single universal equation set.

1. Model class and scope

In the NSTX pedestal literature, GFS is defined as a flexible reduced model consisting of linear moment equations designed to approximate gyrokinetic physics while allowing adjustable velocity-space resolution. Its stated features include the absence of bounce-averaging, retention of the mirror force in the gyrokinetic equations, a spectral representation using Hermite polynomials in parallel velocity and Laguerre polynomials in perpendicular energy, a two-parameter real closure for Hermite moments, and a Hermite representation of the field-line coordinate $z$ with parameters $n_z$ and $w_z$. In this setting, GFS is intended as a fast linear eigenmode solver for instability identification, growth-rate and frequency prediction, quasi-linear transport studies, and pedestal stability constraints for EPED-like modeling (Yang et al., 16 Sep 2025).

Other implementations broaden the same reduced-model idea. TIFF introduces an isothermal gyrofluid code for quasi-two-dimensional interchange and drift-wave turbulence with full-$f$ dynamics and full-$k$ polarization, meaning arbitrary fluctuation amplitudes and arbitrary polarization wavelengths. A full-$F$ gyro-moment approach for LAPD expands the ion distribution function on a Hermite–Laguerre basis and evolves the expansion coefficients as fluid-like gyro-moments, yielding an arbitrary number of coupled moment equations. A three-dimensional diverted-tokamak model evolves total gyro-center fields in a full-$F$, isothermal, electromagnetic setting. These formulations differ in geometry, closure, and target application, but all replace full phase-space kinetics by gyro-averaged fields or moments (Kendl, 2023, Frei et al., 2023, Wiesenberger et al., 2023).

This diversity also clarifies a common ambiguity. GFS is not restricted to a single closure philosophy such as $\delta f$, full-$f$, full-$k$, long-wavelength, or fixed-moment truncation. The literature instead presents a spectrum of models whose shared structure is gyroaveraging, polarization closure, and reduced evolution equations.

2. Governing equations and closure structure

A representative quasi-two-dimensional GFS evolves gyrocenter densities $n_z$0 rather than particle densities $n_z$1. For species $n_z$2,

$n_z$3

with $n_z$4, $n_z$5, and $n_z$6, $n_z$7 the gyro-operators. In Fourier space,

$n_z$8

Quasi-neutrality closes the system through

$n_z$9

and the gyrocenter densities satisfy advection equations

$w_z$0

with $w_z$1 and $w_z$2. In the cold-ion limit $w_z$3, the polarization equation reduces to

$w_z$4

and subtraction of the density equations yields the two-dimensional Euler vorticity equation

$w_z$5

At finite ion temperature, finite-Larmor-radius effects modify vorticity transport through terms such as

$w_z$6

or, at lower order,

$w_z$7

With $w_z$8 and $w_z$9, the generalized vorticity

$f$0

approximately obeys

$f$1

This makes explicit that FLR effects couple vorticity advection to density and pressure asymmetries rather than to the vorticity field alone (Kendl, 2017).

A multi-species electromagnetic gyro-fluid model in flux-tube geometry retains the same basic ingredients. For fluctuating gyro-center densities $f$2,

$f$3

while parallel momentum and electromagnetic dynamics satisfy

$f$4

with closure by the polarization equation

$f$5

and Ampère’s law

$f$6

In the Padé approximation,

$f$7

The effective polarization mass is

$f$8

and in the cold-ion limit the polarization equation becomes

$f$9

This form isolates mass dependence in the inertial response (Meyer et al., 2016).

Full-$k$0, electromagnetic GFS variants extend the closure to evolving total fields. In a diverted-tokamak formulation, the evolved variables are gyro-center density $k$1, gyro-center parallel velocity $k$2, electrostatic potential $k$3, and parallel magnetic vector potential $k$4, with field equations

$k$5

and

$k$6

summed over species. In GREENY, a two-dimensional full-$k$7 reconnection model closes through the non-Oberbeck–Boussinesq arbitrary-wavelength polarisation equation

$k$8

and Ampère’s law

$k$9

These examples show how GFS formulations interpolate between electrostatic vorticity models and electromagnetic moment systems (Wiesenberger et al., 2023, Locker et al., 14 Feb 2025).

3. Representations, closures, and numerical realizations

Different GFS implementations are distinguished less by the presence of gyroaveraging than by the representation used for velocity space, polarization, and field inversion.

Implementation	Core representation	Application
GFS linear eigenmode solver	Hermite in parallel velocity, Laguerre in perpendicular energy, Hermite in $F$0	NSTX H-mode pedestal stability
TIFF	Isothermal full-$F$1, full-$F$2 gyrofluid model with generalized Poisson solve	Quasi-2D interchange and drift-wave turbulence
GREENY	Full-$F$3 and $F$4 2D electromagnetic gyrofluid system	Collisionless magnetic reconnection
Full-$F$5 gyro-moment hierarchy	Hermite–Laguerre expansion of ion distribution, Braginskii electrons	LAPD turbulence

In the full-$F$6 gyro-moment approach, the ion distribution is expanded as

$F$7

with moment coefficients defined by projection. Closure is obtained by truncation,

$F$8

The LAPD implementation uses a fourth-order Runge–Kutta scheme, Arakawa discretization for Poisson brackets, centered finite differences for other derivatives, and a staggered grid in $F$9 for odd-$F$0 moments and parallel velocities (Frei et al., 2023).

TIFF and GREENY emphasize the elliptic field solve. TIFF implements PCG, SOR, and a dynamically corrected Fourier method for the generalized Poisson problem

$F$1

where the coefficient $F$2 evolves with the density. Its DCF method builds on Teague’s method and uses information from the previous time step. GREENY uses a Karniadakis multistep scheme, finite-difference or spectral gradients and Laplacians, Arakawa brackets, Teague, PCG, and SOR solvers for the polarization equation, and SOR or FFTW-based spectral inversion for Ampère’s law. For GREENY’s manufactured-solution tests, the reported scaling is $F$3 and $F$4 for Ampère’s law, while the polarization equation shows $F$5 for SOR and $F$6 for PCG and the spectral Teague method (Kendl, 2023, Locker et al., 14 Feb 2025).

The choice between full-$F$7, $F$8, full-$F$9, long-wavelength, and Oberbeck–Boussinesq variants has direct numerical consequences. TIFF states that its model reduces to Hasegawa–Wakatani in the limits of small turbulence amplitudes, cold ions without FLR effects, and homogeneous magnetic field. GREENY explicitly distinguishes Full-$\delta f$0, Full-$\delta f$1+OB, Full-$\delta f$2+LWL, Full-$\delta f$3+OB+LWL, and $\delta f$4 closures through the treatment of the polarization density $\delta f$5 and the polarization contribution to the gyrofluid potential $\delta f$6. This suggests that “GFS closure” is application-dependent rather than unique.

4. Nonlinear dynamics, transport, and reconnection

The nonlinear dynamics of GFS depart sharply from ordinary two-dimensional Euler flow once FLR physics is retained. In gyrofluid vortex interaction, the fate of co-rotating eddies is decided between accelerated merging or explosion by the asymmetry of initial density distributions. For $\delta f$7, a positive initial amplitude choice $\delta f$8 yields accelerated merging, with the first minimum in vortex separation occurring around $\delta f$9 rather than $f$0 in the cold system. The same vorticity sign realized with the opposite density asymmetry, $f$1 and $f$2, leads instead to explosive separation. The underlying mechanism is the FLR term $f$3, interpreted as gyroinduced vortex spiraling. In warm gyrofluid turbulence, this produces filamentation into thin vorticity sheets and enhances vorticity amplitude across the spectrum, while density fluctuations grow especially at intermediate scales $f$4 (Kendl, 2017).

Mass dependence produces another characteristic GFS effect. In local electromagnetic multi-species gyro-fluid computations, transport decreases with increasing effective plasma mass for protium, deuterium, and tritium mixtures. This trend appears in both cold and warm ion models and in both electrostatic and electromagnetic runs. Warm ions show the stronger isotope dependence. Over $f$5, density and potential autocorrelation times increase by about $f$6, perpendicular correlation lengths increase by about $f$7, and zonal flow activity changes only weakly with isotope mass, with about a $f$8 increase for warm ions. No significant isotope dependence of GAM intensity is found in these isothermal simulations. The stated interpretation is that isotope-improved confinement is dominated by mass-dependent polarization and FLR gyro-dynamics rather than by stronger zonal flows (Meyer et al., 2016).

Three-dimensional full-$f$9, isothermal, electromagnetic simulations in a COMPASS-sized diverted tokamak identify two transport regimes separated by a critical resistivity

$k$0

Beyond this value, mass and energy confinement reduce with increasing resistivity. Relative fluctuation amplitudes increase from below $k$1 in the core to $k$2 in the edge and up to $k$3 in the scrape-off layer. The electron force balance and energy conservation show relative errors on the order of $k$4, while particle conservation and ion momentum balance show errors on the order of $k$5. The turbulence remains field aligned, with density filaments and parallel-current tubes visible in three-dimensional visualization (Wiesenberger et al., 2023).

Reconnection-oriented GFS models exhibit a related sensitivity to closure and dissipation. In GREENY’s Harris-sheet studies, higher $k$6 gives a more X-shaped reconnection region, while lower $k$7 gives a more elongated or Y-shaped structure. In the arbitrary-wavelength Full-$k$8 case, the reconnection region exhibits fine $k$9-scale structure. With warm ions $n_z$00 and ion-current-driven reconnection, ion FLR effects can decouple ion and electron potentials and form plasmoids. A hyperviscosity scan over $n_z$01 to $n_z$02 shows that the main reconnection peak is not strongly altered for moderate $n_z$03, but the transient phase and onset can change; for $n_z$04 the system becomes linearly stabilized, whereas too small $n_z$05 leads to granular noise. Recommended values are roughly $n_z$06, case-dependent (Locker et al., 14 Feb 2025).

5. Linear eigenmode GFS and pedestal stability

In pedestal applications, GFS is used as a reduced linear gyrokinetic eigenmode solver benchmarked against CGYRO. For NSTX H-mode pedestal conditions, the validation database contains 864 local linear stability cases from discharge 139047 at $n_z$07 ms, classified as 706 KBM cases, 89 MTM cases, and 69 TEM cases, spanning roughly $n_z$08 to $n_z$09. The RMS error for a quantity $n_z$10 is defined as

$n_z$11

and branch matching uses the closest eigenvalue among the first $n_z$12 GFS modes: $n_z$13 The branch-mismatch fraction is

$n_z$14

For fixed $n_z$15, Bayesian optimization searches $n_z$16 using Gaussian process regression, 2 random initial points, 40 iterations, Expected Improvement, and an even-$n_z$17 constraint. The optimization is performed on a balanced subsample of 40 KBMs, 30 TEMs, and 40 MTMs. Core-optimized settings $n_z$18 yield a growth-rate RMS error of $n_z$19 and a frequency RMS error of $n_z$20. With optimized pedestal settings $n_z$21, the reported performance is a growth-rate RMS error of $n_z$22, a frequency RMS error of $n_z$23, and a mode mismatch of $n_z$24. Mode-by-mode, the optimized solver gives KBM errors of $n_z$25 in growth rate and $n_z$26 in frequency with $n_z$27 mismatch, TEM errors of $n_z$28 and $n_z$29 with $n_z$30 mismatch, and MTM errors of $n_z$31 and $n_z$32 with $n_z$33 mismatch. The $n_z$34 case is about $n_z$35 slower than $n_z$36, but an estimate in the paper states that CGYRO-like GFS resolution would be 18,956 times slower than the optimized pedestal-resolution setup (Yang et al., 16 Sep 2025).

The same reduced-model strategy is then embedded in pedestal prediction. A later study describes GFS as a reduced electromagnetic gyrofluid stability code developed to provide a fast but physics-faithful estimate of kinetic ballooning mode stability in tokamak pedestals. It solves the fluid moments of the linear gyrokinetic equation using Hermite polynomials in the parallel velocity or field-aligned coordinate, Laguerre polynomials in perpendicular velocity, and the full electromagnetic perturbation set $n_z$37, $n_z$38, and $n_z$39, with pitch-angle scattering and like-species collisions. The stated workflow is to build a bootstrap-consistent equilibrium, fix pedestal density and geometry, increase pedestal electron temperature until the local KBM becomes unstable, scan the temperature gradient via $n_z$40, identify the critical $n_z$41 from a switch from electrostatic to electromagnetic behavior and a sharp growth-rate increase, and iterate until the computed critical $n_z$42 matches the equilibrium $n_z$43 within $n_z$44. In the limit $n_z$45, the KBM threshold approaches the ideal ballooning threshold, but at finite $n_z$46 the KBM is more restrictive. The benchmarked calculations use $n_z$47, and the paper reports agreement with CGYRO for about 800 NSTX-U KBM cases. It further reports experimentally consistent pedestal-width scalings: for DIII-D,

$n_z$48

with $n_z$49–0.09, and for NSTX-U,

$n_z$50

with $n_z$51. The combined GFS+ELITE framework, labeled EPED3 in the study, is reported to improve agreement with experiment relative to EPED1 for the presented DIII-D dataset (Tzanis et al., 15 Sep 2025).

These pedestal studies also correct a frequent oversimplification. GFS is not used there merely as a cheaper stand-in for ideal MHD; it is explicitly designed to retain finite-$n_z$52 effects, ion drift resonances, trapped-particle physics, and electromagnetic coupling while remaining much cheaper than full gyrokinetics.

6. Terminological breadth and related theoretical frameworks

Terminology is not uniform across disciplines. In relativistic many-body theory, gyrohydrodynamics has been introduced as a relativistic quasi-hydrodynamic framework for a charged spinful fluid under strong vorticity and is explicitly identified with the Gyro Fluid System. Its power counting treats thermal vorticity $n_z$53 as $n_z$54, the difference $n_z$55 as $n_z$56, and spin density as leading in derivatives but suppressed by $n_z$57. The leading-order energy-momentum tensor is anisotropic,

$n_z$58

and the complete first-order constitutive relations involve 17 transport coefficients: three bulk viscosities, four shear viscosities, three rotational viscosities, four cross viscosities, and three conductivities. Hall-like coefficients do not contribute to entropy production (Cao et al., 2022).

Adjacent theoretical literatures use different names while addressing structurally related questions. An action-principle formulation for generalized fluid motion including gyroviscosity derives nondissipative momentum transport from gyroscopic motion of fluid elements and gives broad conservation-law statements for energy, linear momentum, and angular momentum. A differential-geometric theory of systems with gyroscopic forces defines a mechanical system with gyroscopic forces as a 4-tuple

$n_z$59

with $n_z$60 a closed 2-form, modified symplectic form

$n_z$61

and a generalized cyclic integral criterion

$n_z$62

These works do not supply the modern plasma GFS nomenclature, but they provide the geometric and variational language for gyroviscous transport and gyroscopic forcing (Lingam et al., 2014, Kharlamov, 2014).

A plausible implication is that the phrase “Gyro Fluid System” names a methodological principle more than a single canonical model: start from a system with fast gyromotion, retain gyroaveraging and polarization, reduce the kinetic description to moments or effective fields, and choose closure, geometry, and solver architecture to match the target regime. In plasma physics that target may be drift-wave turbulence, pedestal stability, or reconnection; in relativistic hydrodynamics it may be spin transport under strong vorticity.