Page 9 - Electrostatic quasi-neutral formulation of global cross-separatrix particle simulation in field-reversed configuration geometry

Physics of Plasmas ARTICLE
FIG. 6. The density profile used in the simulation is shpowffiffiffinffiffiffi in the top while the inverse scale length normalized by ion gyro-radius (q 􏰂 mi Ti ) is shown in the bot-
i eB0 tom. The dashed gray line represents the separatrix.
FIG. 7. The electron ae, calculated by (16) for the simulation parameters, is shown. Electrons are fully-kinetic-like (ae 􏰈 1) where the magnetic field is low but are drift-kinetic- like (ae < 1) where the magnetic field is larger.
scitation.org/journal/php
19–21
As the instability saturates, there is an inverse spectral cascade from the higher to lower toroidal wavenumbers. This can be seen as the change from the shorter wavelength mode structure in Fig. 8 to the larger eddies in Fig. 9. Compared with the previous gyrokinetic turbu- lence simulations,8 there is less inward spread from SOL to core; how- ever, it is difficult to discern whether this is due to the correct representation of particle trajectories or the change in the nature of the instability due to the inclusion of the non-adiabatic electron response. After saturation, the amplitude of fluctuations in the core are about two orders of magnitude below the SOL at longer wavelengths, similar to the gyrokinetic turbulence simulations.
Simulations using more realistic profiles (including equilibrium electric fields) reconstructed from experimental shots are ongoing and will be reported on, in detail in future work.
VII. DISCUSSION
To address the coupled nature of the core and edge regions of field-reversed configuration plasmas, a formulation for electrostatic particle simulation across the separatrix is presented. A field-aligned simulation mesh and corresponding mesh operations have been devel- oped and verified. A formulation for efficient particle pushing, free from gyrokinetic validity and based on the blended particle model, is also established. A corresponding quasi-neutrality equation for self- consistent fields is also detailed. Using about 2.5 􏰅 106 cells and 19 􏰅 106 particles, a total wall-time of roughly 2500 node-hours was required for this numerically converged blended FRC simulation run, comparable to previous gyrokinetic FRC simulation runs. For mag- netic geometry with large variations in magnetic field magnitude, this model is an improvement in simulation time over the fully kinetic par- ticle model and an improvement in model validity over the gyrokinetic model. The efficient use of the blended model for turbulent transport simulations will aided by the upcoming exascale computing platforms. All of these new features have been implemented in ANC and bench- marked in simple cylindrical geometry. Nonlinear simulations in FRC geometry have also been performed.
mode.
when the mode frequency is roughly equal to the electron magnetic gradient drift frequency, suggesting that this mode is, at least, partially driven by the electron grad-B motion.
Initial analysis of the non-adiabatic electron response peaks
corresponding to hkfqii > 10), and is shown in Fig. 8. During the ini- tial linear growth stage, the instability for mode numbers n 1⁄4 1⁄260; 70􏰆 has real frequency xr =Xc;p 1⁄4 1⁄20:52; 0:56􏰆 and linear growth-rate c=Xc;p 1⁄4 1⁄20:11; 0:11􏰆. The mode rotates in the direction of the ion dia- magnetic drift, and the frequency is comparable to the so-called unconventional short wavelength ion temperature gradient (SWITG)
Phys. Plasmas 27, 082504 (2020); doi: 10.1063/5.0012439 Published under license by AIP Publishing
27, 082504-9