The continuum gyrokinetic solver within the plasma physics code Gkeyll was used to model turbulence in open-field line devices and the scrape-off layer of fusion devices. Since neutral interactions are important in these scenarios, simplified models of neutral interactions, including electron-impact ionization and charge exchange, have been implemented in the code. The neutrals are modeled as a kinetic Boltzmann species using the continuum Vlasov solver within Gkeyll and coupled to the gyrokinetic plasma species. Tests of a simple SOL with neutral interactions in one spatial dimension were carried out with gyrokinetic plasma species (with two velocity-space dimensions) coupled to Vlasov neutral species (with three velocity-space dimensions) and demonstrated reasonable agreement with analytic two-point model predictions. Results of tests in three spatial dimensions will be forthcoming.

Dispersive shell pellet (DSP) injection simulations with the 3D extended MHD code NIMROD have been carried out for DIII-D to understand whether the injection of two DSPs from opposite toroidal locations can retain the key advantages of DSP for disruption mitigation, such as a high radiation fraction, while reducing the toroidal peaking factor of the thermal quench radiation. In the simulations, a single DSP traveling at 400 m/s is compared with two identical DSPs separated by 180 degrees, with one at 400 m/s and one at 420 m/s. A major concern is whether the delivery of the payload material from the faster pellet will effectively halt the ablation of the slower pellet so as to prevent payload delivery. Unsurprisingly, the solo DSP has the earliest payload delivery time at 1.03 ms due to the cooling effects of the extra ablated shell material in the dual DSP case, but this effect is not dramatic, with the dual DSPs delivering their payload just 0.06 ms and 0.18 ms later than the single DSP. The more surprising result is that, in the dual DSP simulation, the slower 400 m/s pellet releases its payload earlier than the faster 420 m/s pellet. This occurs due to the onset of MHD instabilities triggered by the ablated shell material that shifts the hot core away from the faster pellet and toward the slower, increasing its ablation rate. If we regard the faster pellet as dominant in driving the instability, this result is qualitatively consistent with earlier massive gas injection (MGI) modeling showing that the 1/1 mode tends to orient itself such that the heat flux is away from the injection location. Studies to quantify the effects of different relative pellet speeds on thermal quench characteristics are ongoing.

Quantitative prediction of the runaway electron (RE) avalanche production process is critical for future large tokamak experiments such as ITER. The MARS-Q code has been updated to enable toroidal modeling of this process, by coupling the Rosenbluth-Putvinski avalanche model with the axisymmetric (n=0) MHD solver and by taking the effect of plasma shaping into account. The avalanche model for the RE current and the MHD equations are simultaneously time advanced within MARS-Q. The plasma shaping effect enters into the avalanche model mainly via the spatial distribution of the electric field, which comes from the MHD solution. Initial-value simulations were carried out for two DIII-D discharges with different post-thermal quench plasma shapes, one near circular and one with high elongation. The discharges also had different plasma conditions, in particular different initial current. For the circular-shaped plasma, the Rosenbluth-Putvinski model somewhat underestimates the RE plateau current, as compared with that measured in DIII-D experiments. For the elongated plasma, simulations find strong runaway current avalanche production. Systematic scans of the plasma boundary shape, at fixed pre-disruption plasma current, found that plasma elongation helps to reduce the RE avalanche production. Plasma triangularity (either positive or negative), on the other hand, has only a minor effect. The avalanche process involves two competing mechanisms associated with the electric field. On one hand, stronger electric field produces more avalanche multiplication and thus more RE current. On the other hand, a fast growing RE current quickly reduces the fraction of the conduction current coupled with the electric field. As a result, the runaway plateau current is not always the largest with the strongest initial electric field. The results from this study help to better understand the RE avalanche dynamics and point to future directions to further improve the RE modeling for tokamak plasmas.