WorksPhysics and Astronomy
Explosive instability, quiet plasma: phase-space cascades, coherent vortices, and transport suppression in a dipole pair plasma
Abstract
Turbulence in a magnetised dipole pair plasma can transfer electrostatic energy into coherent flows while carrying entropy to small scales. This occurs via a dual cascade in a four-dimensional phase space, whereby advection by the fluctuating electric field nonlinearly phase-mixes the transit-averaged particle motion. Below a critical pressure gradient, zonal bands suppress the kinetic mode, leaving a transport floor. When the fluid mode is unstable, an explosive burst collapses into a zonal condensate on which two vortices "breathe" by reversible energy exchange, without causing net transport.
Magnetic confinement of dilute electron–positron plasmas is being pursued by the APEX levitated-dipole and EPOS stellarator programmes [1, 2]. These plasmas offer a simple setting in which to study turbulence under conditions of charge and mass symmetry. When their equilibrium densities and temperatures are also equal, electrostatic drift modes are stable in a straight magnetic field [3]. The curvature needed for dipole confinement nevertheless makes the magnetic drift charge dependent, so that density and temperature gradients drive interchange instabilities [4, 5, 6]. Above a threshold in the gradients (strong drive) this is the fluid interchange; below it (weak drive), a kinetic interchange, the entropy mode [7, 8], destabilised by particles resonant with the magnetic drift [4]. Linear thresholds are only a first step towards predicting confinement. What matters is how much transport survives once the fluctuations saturate.
In two-dimensional fluids, an inverse energy cascade builds large vortices while enstrophy is transferred to small scales [9]. A kinetic plasma can also develop fine structure in the particle distribution, allowing coherent flows to coexist with a forward entropy cascade [10, 11]. In a dipole, particles of different pitch angle follow different orbits through the varying magnetic field. Nonlinear advection can therefore mix the distribution in pitch angle even when perpendicular mixing by Larmor motion [12, 13] is negligible.
We predict and demonstrate a four-dimensional phase-space cascade in which electrostatic energy collects at large scales while entropy develops fine pitch-angle structure. The resulting zonal states exhibit a finite transport floor at weak drive, but transition to vortex breathing at strong drive: the periodic growth and decay of coherent vortices through energy exchange with the zonal flow, accompanied by negligible mean transport.
Kinetic equations and invariants.— We consider the ordering , where is the Lar-mor radius, the Debye length, and an equilibrium length. The unstable scales, , are much larger than . This reverses the usual ordering of magnetic confinement, in which and the plasma is quasineutral, , at the unstable Lar-mor radius scales.
We use flux coordinates with and the arc length along the field line, and write , with and a common Maxwellian . Both the transit and bounce frequencies, the leading is independent of [4] [3], where and the pitch-angle variable . The potential sampled by an orbit is
The limits are bounce points for trapped particles and one field-line period for passing particles. To leading order in , the kinetic and field equations are [4, 14]
Here is the advection, is the orbit-averaged binormal magnetic drift, and is the gradient drive at fixed . The dipole has no radial magnetic drift.
Let , the average over the flux tube of volume , and define the field energy . The two nonlinear invariants are
Here is the free energy, the perturbed entropy of the particles weighted by plus the field energy [15], and is the electrostatic energy seen by orbit-averaged particles. The second term in Eq. (5), the orbit-variance term, is non-negative [16] and vanishes for a flute-like potential, constant along the field line. This is the electrostatic energy of the linear theory of trapped-particle modes [16, 17]; to our knowledge its role as a nonlinear invariant of the orbit-averaged problem has not been examined. The entropy of the orbit-averaged perturbation is
where . We call entropy throughout. The bracket therefore conserves , but the driven dynamics need not.
The pair constrains transfer between perpendicular Fourier modes, as energy and enstrophy do in two-dimensional fluids [9]. At , the least entropy a mode carries per unit of obeys , and being spectral densities of and , and the state attaining this scaling is flute-like [proof in the End Matter, Eqs. (10)–(19)]. Electrostatic energy must therefore collect at large scales whenever entropy moves to small scales [10, 11] [{xref}Eq. (A11)`]. That constraint fixes the end state, not the route to it: it holds whether the energy arrives through a constant-flux cascade, as in two-dimensional fluids, or in single nonlocal steps. A flute-like potential has the same orbit average for every pitch angle, so the endpoint of the inverse transfer is also the state in which particles of every pitch angle are advected together and nonlinear pitch-angle mixing (described below) switches off.
Dipole cascade and phase mixing.—We use local, electrostatic GENE simulations [18] in a circular-coil dipole [5], on the flux surface whose outboard equatorial radius is coil radii, so that varies by a factor of two along the line. Both drives have , with the magnetic-drift frequency of a thermal particle and the specific volume of the flux tube (with ), and the ratio , with and , fixed as 0 or 1. The fluid-interchange threshold is in the -pinch limit [4]: at the density gradient alone drives the kinetic interchange across a broad band of ( and are the radial and binormal coordinates of the box, along and ); at the drive is above the threshold and the fluid interchange grows at a scale-independent rate, , fastest for streamers (). Times are in with ; fluxes are quoted in the Debye mixing-length units and , energies in . Both runs use and a box wide at the equator; fourth-order hyperdiffusion acts in and along the field line and there is no collision operator. Resolution and validation will be detailed in a future work [19].
The nonlinear transfer , from the free-energy diagnostics of GENE [20, 21], defines a cumulative flux
A negative is transfer to larger scales, a positive one to smaller. At the two fluxes take opposite signs as soon as the run becomes nonlinear (Figure 1): across the long-wavelength band and at every . This happens before any zonal flow exists: over the averaging window the zonal share of the potential power is negligible, and the store is still filling [Figure 2(c)]. The condensate then grows out of the turbulence as narrow bands, many Debye lengths wide but far below the box scale, with vortices persisting between them [Figure 3(c)]; late in the run they hold most of the equatorial potential power and a large part of , in Figure 2(c), still rising, and the transport has not switched off (Figure 2).
Figure 1. (a) The flux of the electrostatic energy of Eq. (5), and (b) the flux of the entropy of Eq. (6), through the perpendicular wavenumber , Eq. (7), averaged over the nonlinear-onset phase before zonal flows have developed; each is normalised to its own peak. For comparison the black curves use both terms of Eq. (5) instead of just its field term; the vertical dotted line marks the largest complete shell of the box.
Figure 2. (a) Heat flux and (b) particle flux , every output sample (pale) with a short running mean (black). (c) The free energy of Eq. (4), the electrostatic energy of Eq. (5) with both its terms and its zonal and non-zonal parts, , .
Figure 3. (a,c) Electrostatic potential for during the nonlinear onset and late in the zonal-band state. (b,d) Same for just before the flux peak and late in the drain; the dashed lines are contours of , i.e. streamlines.
At there is no turbulent cascade since there is no inertial range to cascade through: the run passes from linear growth to a box-scale zonal flow in one burst, without developed turbulence (Figure 3).
The simulations retain the parallel motion, so their nonlinearity conserves and but not the orbit-variance term of Eq. (5) instantaneously. However, the full conservation is indeed observed, over timescales longer than the fast parallel dynamics.
Equation (1) explains the nonlinear phase mixing. A deeply trapped particle samples the potential near the magnetic-field minimum; a passing particle samples the whole field line. Tracers launched from one point in the turbulence, each advected by its own , spread to a separation of many Debye lengths [Figure 4(a) and Movie 1 [22]; compare Figure 1 of Ref. [23]]. Here, the orbit average plays a role analogous to the velocity-dependent gyroaveraging in finite-Larmor-radius turbulence [12, 13].
Figure 4. (a) Tracers launched at one (star), stacked by pitch angle and each advected by its own ; grey curves are their projections on the floor. (b) Orbit-averaged potential power (blue) and the entropy per unit density of states, each normalised to its maximum, over the onset; the dashed line marks the trapped–passing boundary . (c) over its maximum.
Pitch-angle structure produces parallel structure through Eq. (3), coupling and within the four-dimensional phase space . At , the potential power is dominated by nearly flute-like large scales, whereas has order-one variation [Figure 4(b)]. Smaller scales retain order-one pitch dependence and separate the tracers. To measure transfer in pitch angle, we expand the orbit-averaged distribution in polynomials in of degree , orthogonal under the orbit measure , so that plays the role of a pitch-angle wavenumber. The flux , the nonlinear transfer of out of degrees up to , is positive [Figure 4(c)] and entropy moves to finer pitch-angle structure.
Coherent vortices and breathing.—At streamers break into small vortices that fill the box (Figure 3). At streamers () of every wavelength grow at the same linear rate, while the box-scale streamer dominates in saturation amplitude. A secondary instability collapses the streamer in one burst into box-scale zonal bands modulated by two vortices, one on each jet: above the fluid threshold, here , the instability does not rely on the drift resonance that phase-mixes the kinetic interchange, so free energy accumulates undissipated and must drain afterwards. The collapse is the two-dimensional secondary instability of the streamer, whose growth rate through the rise of the zonal energy follows the streamer amplitude as [24]. The burst leaves the zonal flow with most of the electrostatic energy . Nothing in the collisionless model damps the flute-like part of this condensate: there is no geodesic curvature, gyroaveraging is negligible at , and the velocity-space hyperdiffusion conserves particles, so it cannot change . Collisions would damp the flow at a small fraction of the collision frequency, so the condensate is long-lived rather than permanent. In electron-ion plasmas, zonal flows can be driven unstable (a tertiary instability) by a resonance involving nonlinear phase mixing [24, 25, 26]; here, the orbit average could provide such a mechanism through the trapped particles, but none is observed: the condensate persists to the end of the run.
In the condensate each vortex gains and loses electrostatic energy through exchange with the zonal flow, and the nonlinear conservation law gives
where is the part of and the rest. Each vortex, grows and shrinks periodically: it is advected there at a fraction of , being the peak zonal flow speed, tilted and untilted by the shear on either side, and the tilting is the exchange of Eq. (8). The two vortices move in opposite directions, so in the frame of either the other passes at twice the advection frequency; the tilt of the flow between them, and with it the exchange, follows their relative phase, and the breathing runs at that doubled frequency (Figure 5 and Movie 2 [22]).
Figure 5. (a) The zonal and non-zonal parts of , Eq. (5) with both its terms, each over its window mean. (b) The two sides of Eq. (8) for the field term , the first term of Eq. (5).
The large oscillating fluxes of Figure 2(a,b) are caused by these vortices. Each vortex carries its own density and pressure perturbation outward through one half of its circulation and back through the other, so both fluxes reverse every half turn and average to zero. The free energy injection , Eq. (9), is zero on average. This is a dynamic analogue of the Dimits state [27], in which zonal flows quench transport: here the eddies survive, the shear returning the energy it takes.
The zonal shear is a few times the peak linear growth rate at , but tens of times that rate at [Figs. 7(c) and 8(c)]. The more explosive instability builds the stronger flow and leaves less residual transport.
Transport and the free-energy balance.— Define outward-positive, species-summed radial fluxes and . Here is the total heat flux. The total free-energy equation is [15, 28]
Figure 6. (a) and ; (b) and ; both normalised by the mean of . (c) , and the breathing exchange, Eq. (8), divided by the non-zonal electrostatic energy .
Magnetic drift and nonlinear advection conserve ; is its irreversible removal.
Figure 6(a,b) splits the rates of change of and into reversible parts, and , from the gradient drive and the linear terms, and irreversible parts, and , from dissipation, with and . At , over the onset, injection exceeds dissipation and the difference fills the store of energy. When the zonal bands take over the particle flux is smaller but steady, injection and dissipation balance [Fig. 6(c)] and the store stops filling. This is the transport floor of the kinetic interchange, carried by a residual bath of coherent fluctuations.
At , electrostatic energy feeds entropy while the sinks remove free energy [Fig. 6(a,b)]. Mean injection is negligible, so , while the reversible zonal–vortex exchange is much faster [Fig. 6(c) and 5]. The state therefore changes little over a breathing cycle but decays over many cycles. Its long-time outcome remains open: dissipation may vanish, leaving a transport-free state, or interchange activity may resume if the zonal shear falls below the stabilising level.
Confinement implications.— For weak drive, the heat flux is small, an order of magnitude below [Fig. 2(a,b)], suggesting the simplifying assumption . Steady balance then gives . For and positive mean dissipation, particles move inward for and outward for , the reversal that Helander and Connor [29] found for the quasilinear particle flux near marginal stability. The run lies on the predicted side: its particle flux is outward throughout [Figs. 2(b) and 7(a)].
At the burst moves particles inwards and heat outward Figure 8(a). An inward pinch has precedent in electron–ion plasmas [30, 31, 32, 33], in LDX [34] and RT-1 [35].
Figure 8. As Figure 7, for the run; (a) on a symmetric logarithmic axis, with (dashed); in (c). There are two vortices because the box-scale zonal flow is one jet in each direction, and exchanging the species with , and a half-box shift, a symmetry of the pair-plasma equations kept by the simulated state, carries each jet and its vortex onto the other.
Figure 7. The run. (a) The terms of Eqs. (9) times , and their residual. (b) The zonal fraction of the equatorial potential power. (c) At the end of the run, the -averaged flow (grey) and the zonal shearing rate (black) over the peak linear growth rate (dotted, ).
These observations motivate the conjecture that evolving profiles could relax towards . An inward pinch steepens the density profile and lowers ; outward transport flattens it and raises . The proposed endpoint, , has uniform Maxwellian entropy per particle [36]. The value forms one boundary of Helander’s point-dipole ground-state region [37]; the full condition additionally constrains the absolute gradients. It is also the ray on which the fluid- and kinetic-interchange thresholds of the point dipole meet [4, 29].
The more explosive instability leads to the quieter plasma end state, and the particle fluxes of the two runs point towards from either side. The strong-drive condensate reaches the box scale, but its size in a device remains undetermined. Global simulations with physical collisions could test its persistence, residual transport, and coupling to the equilibrium. These results concern magnetically confined pair plasmas with , a regime also found in pulsar magnetospheres [38, 39]. The APEX experiment could test these predictions through the evolution of its density and temperature profiles and through coherent potential fluctuations at twice the vortex advection frequency.
Acknowledgments
Acknowledgements.—We thank J. Ball, R. Nies, A. A. Schekochihin, and W. Clarke for helpful discussions. Part of this work used the CINECA PITAGORA supercomputer under project GEMST26. This work has been funded by the EPSRC Fusion Grant 2022/27 [grant number EP/W006839/1]. To obtain further information on the data and models underlying this paper please contact [email protected]. For the purpose of open access, the author(s) has applied a Creative Commons Attribution (CC BY) licence (where permitted by UKRI, ‘Open Government Licence’ or ‘Creative Commons Attribution No-derivatives (CC BY-ND) licence’ may be stated instead) to any Author Accepted Manuscript version arising.
References
End Matter
Appendix A: Least entropy per unit of electrostatic energy
We derive bounds on the least entropy a perpendicular Fourier mode must carry per unit of electrostatic energy, restricted to the one-well case of a dipole geometry. States are pairs , independent of , with fixed by Eq. (3).
Phase-space measure and orbit averages.— Let denote the phase-space integral per unit volume, so that Eq. (6) reads . In the variables of Eq. (1), and summed over the sign of , the measure is
The last factor is proportional to the time a particle spends in , and with a single well the accessible part of the line at each is one orbit, so integrating any over at fixed gives , the orbit average of Eq. (1) applied to . Two consequences of this result allow us to find a useful expression for . Firstly, the phase-space integral of any quantity equals that of its orbit average, , so that for independent of , such as or , and for any ,
the second because . Secondly, for independent of and , . With and the second identity of Eq. (11), the orbit-variance term of Eq. (5) becomes .
Electrostatic energy in terms of the orbit-averaged perturbation.— To express through , take the -moment of Eq. (3): multiplying by , averaging over the periodic box and integrating by parts gives . Expanding and using the first identity of Eq. (11) to move the orbit average off the potential,
Lower bound on the entropy.— The Cauchy–Schwarz inequality for the measure , applied to and , bounds the square of Eq. (12) and gives
Two dimensionless numbers describe the potential: its mean square perpendicular wavenumber in units of , and its orbit variance as a fraction of its mean square,
The orbit-variance term of Eq. (5) is non-negative, so , with only for a flute-like potential. With and , Eq. (5) reads , which is positive unless is constant. Hence also : at the orbit average vanishes at every pitch angle, so the product in Eq. (12), and with it , would vanish. Equation (A4) then becomes
Equality holds in the first step only if with a single constant for both species and all of phase space, and in the second only if .
The bound for a single perpendicular mode.— A single Fourier mode obeys the same bound. Equation (3) is linear, so a single pair , with conjugate to , is a state on its own. In the local flux tube, whose metric and depend on alone, and are sums over of their parts and , so Eq. (15) holds for each mode with and replaced by and . For such a state the averages over in Eq. (14) reduce, because , to line averages, , and , where is the amplitude and varies along the line with the metric. Writing , Eq. (14) becomes
Here is a weighted average of along the line, and so lies between and .
Attainment of the bound.— Equality throughout Eq. (15) requires , so that is constant along the line, and , so that ; Eq. (3) at each then fixes to , which a single constant satisfies only if is constant along the line. On such a line the bound is attained, exactly: the analogue of the entropy-to-energy ratio of two-dimensional flow, with in the place of the Larmor radius of gyrokinetics [10]. In the dipole varies along the line and no state attains equality throughout Eq. (15). Equation (15) with still bounds the minimum of from below, and one trial state bounds it from above. The pitch-angle-independent perturbation in the mode , with a constant amplitude, has the potential by Eq. (3), so with . Then and the second identity of Eq. (11) bound the entropy of Eq. (6) from above, while the field term of Eq. (5) bounds the electrostatic energy from below,
and their ratio is an average of with the positive weight , so the trial perturbation gives
Since Eqs. (15) and (16) give for every state, for every mode and at any ,
the minimum being over all states of the mode, , with . It is attained by , equality in the first step of Eq. (15), with the least eigenvalue of . Along the simulated flux tube varies by a factor of order unity, so the minimum is . Equation (15) also gives , so a state with has : its potential departs from a flute-like one by in root mean square.
Consequence for the cascade. — The bracket in Eq. (2) conserves and separately, and Eq. (19) says where it can put . Let be the minimum in Eq. (19), a property of the mode alone lying between and , so that for every state. Summing over modes gives : for any threshold , the electrostatic energy at scales with can never exceed , while entropy above the bound is not fixed by , and the bound places no constraint on the direction in which it moves. This inequality applied to a three-scale transfer gives the kinetic analogue of Fjørtoft’s argument [9, 23]. Take three groups of modes with , all of initially at , and let the nonlinearity send electrostatic energies and into the groups and . Conserving and , with in each group, gives
where is the entropy initially above the bound, zero for a state at it. Doubling multiplies by four at every , so it multiplies by exactly four where is constant along the line, ; in the dipole it multiplies only the bounds and of Eq. (19) by four. With and , Eq. (20) reads : a state at the bound sends at most a quarter as much electrostatic energy an octave towards smaller scales as an octave towards larger. In the dipole Eq. (20) holds as written, with the ratio in place of , and
the octave estimate is exact only for constant . Equation (20) restricts which transfers are possible rather than predicting which occur: once is comparable with it allows either direction; Figure 1 shows the one taken. Two-dimensional fluid turbulence, with enstrophy exactly times energy, is the case of equality [9].
Appendix B: Time histories of the two runs
Figures 7 and 8 show both runs over their full length.
References
- [4]geometry, J. Plasma Phys. 84, 905840201 (2018), arXiv:1711.08021.DOI
- [5]D. Kennedy, A. Mishchenko, P. Xanthopoulos, P. Helander, A. Bañón Navarro, and T. Görler, Linear gyrokinetics of electron–positron plasmas in closed field-line systems, J. Plasma Phys. 86, 905860208 (2020).DOI
- [6]A. Mishchenko, D. Kennedy, P. Helander, A. Könies, G. G. Plunk, P. Xanthopoulos, and A. Zocco, Gyrokinetic applications in electron–positron and non-neutral plasmas, J. Plasma Phys. 89, 935890403 (2023).DOI
- [7]A. N. Sinakov, P. J. Catto, and R. J. Hastie, Kinetic stability of electrostatic plasma modes in a dipolar magnetic field, Phys. Plasmas 8, 4414 (2001).
- [8]P. Ricci, B. N. Rogers, W. Dorland, and M. Barnes, Gyrokinetic linear theory of the entropy mode in a Z pinch, Phys. Plasmas 13, 062102 (2006).DOI
- [9]R. Fjørtoft, On the changes in the spectral distribution of kinetic energy for two-dimensional non-divergent flow, Tellus 5, 225 (1953).
- [10]G. G. Plunk, S. C. Cowley, A. A. Schekochihin, and T. Tatsuno, Two-dimensional gyrokinetic turbulence, J. Fluid Mech. 664, 407 (2010).DOI
- [11]G. G. Plunk and T. Tatsuno, Energy transfer and dual cascade in kinetic magnetized plasma turbulence, Phys. Rev. Lett. 106, 165003 (2011).pubmed.ncbi.nlm.nih.gov/21599375
- [12]W. Dorland and G. W. Hammett, Gyrofluid turbulence models with kinetic effects, Phys. Fluids B 5, 812 (1993).
- [13]A. A. Schekochihin, S. C. Cowley, W. Dorland, G. W. Hammett, G. G. Howes, G. G. Plunk, E. Quataert, and T. Tatsuno, Gyrokinetic turbulence: a nonlinear route to dissipation through phase space, Plasma Phys. Control. Fusion 50, 124024 (2008).DOI
- [14]I. G. Abel, G. G. Plunk, E. Wang, M. Barnes, S. C. Cowley, W. Dorland, and A. A. Schekochihin, Multiscale gyrokinetics for rotating tokamak plasmas: fluctuations, transport and energy flows, Reports on Progress in Physics 76, 116201 (2013).DOI
- [15]J. A. Krommes and G. Hu, The role of dissipation in the theory and simulations of homogeneous plasma turbulence, and resolution of the entropy paradox, Phys. Plasmas 1, 3211 (1994).DOI
- [16]P. Helander, J. H. E. Proll, and G. G. Plunk, Collisionless mirror instabilities in stellarators. I. Analytical theory of trapped-particle modes, Phys. Plasmas 20, 122505 (2013).
- [17]J. H. E. Proll, P. Helander, J. W. Connor, and G. G. Plunk, Resilience of quasi-isodynamic stellarators against trapped-particle instabilities, Phys. Rev. Lett. 108, 245002 (2012).pubmed.ncbi.nlm.nih.gov/23004281
- [18]F. Jenko, W. Dorland, M. Kotschenreuther, and B. N. Rogers, Electron temperature gradient driven turbulence, Phys. Plasmas 7, 1904 (2000).DOI
- [19]D. Kennedy and G. G. Plunk, Electron–positron plasma turbulence in closed-field-line geometry (2026), unpublished manuscript.
- [20]A. Bañón Navarro, P. Morel, M. Albrecht-Marc, F. Merz, T. Görler, F. Jenko, and D. Carati, Free energy cascade in gyrokinetic turbulence, Phys. Rev. Lett. 106, 055001 (2011).pubmed.ncbi.nlm.nih.gov/21405402
- [21]A. Bañón Navarro, B. Teaca, and F. Jenko, The anisotropic redistribution of free energy for gyrokinetic plasma turbulence in a Z-pinch, Phys. Plasmas 23, 042301 (2016).arxiv.org/abs/1512.06058
- [22]See Supplemental Material at [URL will be inserted by publisher] for Movies 1 and 2.
- [23]G. G. Plunk, T. Tatsuno, and W. Dorland, Considering fluctuation energy as a measure of gyrokinetic turbulence, New J. Phys. 14, 103030 (2012).DOI
- [24]B. N. Rogers, W. Dorland, and M. Kotschenreuther, Generation and stability of zonal flows in ion-temperature-gradient mode turbulence, Phys. Rev. Lett. 85, 5336 (2000).
- [25]P. G. Ivanov, A. A. Schekochihin, W. Dorland, A. R. Field, and F. I. Parra, Zonally dominated dynamics and Dimits threshold in curvature-driven ITG turbulence, J. Plasma Phys. 86, 855860502 (2020).DOI
- [26]A. Hallenbert and G. G. Plunk, Predicting the Dimits shift through reduced mode tertiary instability analysis in a strongly driven gyrokinetic fluid limit, Journal of Plasma Physics 87, 905870508 (2021).DOI
- [27]A. M. Dimits, G. Bateman, M. A. Beer, B. I. Cohen, W. Dorland, G. W. Hammett, C. Kim, J. E. Kinsey, M. Kotschenreuther, A. H. Kritz, L. L. Lao, J. Mandrekas, W. M. Nevins, S. E. Parker, A. J. Redd, D. E. Shumaker, R. Sydora, and J. Weiland, Comparisons and physics basis of tokamak transport models and turbulence simulations, Phys. Plasmas 7, 969 (2000).
- [28]H. Sugama, M. Okamoto, W. Horton, and M. Wakatani, Transport processes and entropy production in toroidal plasmas with gyrokinetic electromagnetic turbulence, Phys. Plasmas 3, 2379 (1996).DOI
- [29]P. Helander and J. W. Connor, Gyrokinetic stability theory of electrostatic plasmas, J. Plasma Phys. 82, 905820301 (2016), arXiv:1606.04656.DOI
- [30]V. V. Yankov, The pinch effect explains turbulent transport in tokamaks, JETP Lett. 60, 171 (1994).
- [31]M. B. Isichenko, A. V. Gruzinov, and P. H. Diamond, Invariant measure and turbulent pinch in tokamaks, Phys. Rev. Lett. 74, 4436 (1995).pubmed.ncbi.nlm.nih.gov/10058506
- [32]M. B. Isichenko, A. V. Gruzinov, P. H. Diamond, and P. N. Yushmanov, Anomalous pinch effect and energy exchange in tokamaks, Phys. Plasmas 3, 1916 (1996).DOI
- [33]S. Kobayashi, B. N. Rogers, and W. Dorland, Particle pinch in gyrokinetic simulations of closed field-line systems, Phys. Rev. Lett. 105, 235004 (2010).
- [34]A. C. Boxer, R. Bergmann, J. L. Ellsworth, D. T. Garnier, J. Kesner, M. E. Mauel, and P. Woskow, Turbulent inward pinch of plasma confined by a levitated dipole magnet, Nat. Phys. 6, 207 (2010).
- [35]H. Saitoh, Z. Yoshida, J. Morikawa, Y. Yano, T. Mizushima, Y. Ogawa, M. Furukawa, Y. Kawai, K. Harima, Y. Kawazura, Y. Kaneko, K. Tadachi, S. Emoto, M. Kobayashi, T. Sugiura, and G. Vogel, High-β plasma formation and observation of peaked density profile in RT-1, Nucl. Fusion 51, 063034 (2011).DOI
- [36]L. D. Landau and E. M. Lifshitz, Statistical Physics, Part 1, 3rd ed., Course of Theoretical Physics, Vol. 5 (Pergamon Press, Oxford, 1980) secs. 43 and 45.
- [37]P. Helander, Available energy and ground states of collisionless plasmas, J. Plasma Phys. 83, 715830401 (2017), arXiv:1706.05219.DOI
- [38]P. Goldreich and W. H. Julian, Pulsar electrodynamics, Astrophys. J. 157, 869 (1969).DOI
- [39]A. Philippov and M. Kramer, Pulsar magnetospheres and their radiation, Annu. Rev. Astron. Astrophys. 60, 495 (2022).DOI