NIMROD simulations of a shell-pellet injection induced thermal quench (TQ) have been carried out, scanning over several parameters including pellet speed, payload delivery location, and shell mass. Using a nominal carbon shell mass taken from DIII-D experiments (3.01×10^20 atoms), simulations assuming 100% ablation and 25% ablation of this mass before payload delivery were conducted. For 100% shell ablation, NIMROD predicts considerable destruction of outer flux surfaces before payload delivery, whereas 25% ablation has “ideal-like” behavior with flux surface break-up occurring from the inside-out once the payload is released. These simulations may bound the actual ablated material in DIII-D experiments, suggesting the experiments fall within the range necessary for ideal shell pellet behavior. With the larger ablated shell mass, some intermediate flux surfaces separate the core from the edge, as long as the payload delivery is sufficiently close to the axis. As expected, a less centered payload has less ideal-like behavior.

Members of GA's exascale computing team (J. Candy, F. Halpern, and M. Kostuk) attended the GPU Hackathon in Santa Fe, NM, to optimize CGYRO and ALMA code performance and scalability on GPU clusters. In the case of CGYRO, scripts for debugging and profiling GPU and MPI calls were developed and different techniques for manual binding of GPU to MPI tasks were established. Using these new capabilities, detailed CGYRO benchmarks were carried out on P4, P100, and V100 based GPU clusters. These also provided a clear assessment of the advantages of GPUDirect networking technology. The same profiling techniques were employed on the ALMA multigrid solver components, exposing possibilities for algorithmic improvements such as reduced communications, memory usage, and compute sharing between CPUs and GPUs. Several of these optimizations were then implemented in ALMA, leading to a factor of two speedup. The work was carried in close collaboration with Drs. P. Luszczek and D. Appelhans, GPU computing experts from ORNL and IBM who acted as GA's team mentors.

Chris Holland, Eric Bass, Phil Snyder, Orso Meneghini and Joseph McClenaghan attended the 2019 US-EU Transport Task Force meeting in Austin TX. Bass presented a talk on validation of energetic particle transport models, and McClenaghan presented a talk on transport in scenarios with high poloidal beta.

The paper, “Shattered Pellet Injection Simulations with NIMROD” by C. Kim, et al., has been accepted for publication in the Special Issue of Phys. Plasmas for invited APS presentations. The paper describes resistive MHD simulations using NIMROD for disruptions mitigated by Shattered Pellet Injection (SPI) as well as initial validation comparisons with DIII-D experiments. The 3D simulations show that parallel flows transport impurities and dictate the dynamics of the thermal quench.

A paper entitled “Axisymmetric benchmarks of impurity dynamics in extended-magnetohydrodynamic simulations,” by Brendan Lyons, Charlson Kim, Yueqiang Liu, Nate. Ferraro, Stephen Jardin, Joseph McClenaghan, Paul Parks and Lang Lao, was accepted for publication in the special issue of Plasma Physics and Controlled Fusion, associated with the 6th Annual Theory and Simulation of Disruptions Workshop held at Princeton NJ in 2018. The paper describes a verification benchmark of the coupled impurity and MHD models implemented in the M3D-C1 and NIMROD extended-magnetohydrodynamics (xMHD) codes for simulations of impurity-induced disruption mitigation, and will provide increased confidence in the ability of both codes to perform sophisticated disruption-mitigation simulations.

The 5th week-long OMFIT developers code-camp was held March 4-8 at General Atomics. Several important objectives were accomplished as a result of the efforts of many participants. For the first time, international researchers (Myungwon Lee and Jisung Kang from NFRI in Daejeon, South Korea) attended an OMFIT code camp, and generated the first kinetic equilibrium reconstructions of KSTAR in OMFIT coupling EFIT, OMFITprofiles, and TRANSP. Significant progress was also made on a broad range of physics modules: the TRANSP module now produces postprocessed quantities to facilitate DIII-D between-shot control-room analyses; the IPS-FASTRAN transport, the GINGRED automated SOL mesher, and the Ion Cyclotron Emission (ICE) modules were refined for official release; a new module has been developed to interface with the SOFT code for a synthetic synchrotron diagnostic of runaway electrons; the NIMROD module has been refactored for improved simplicity and generality; the 3D magnetics module was updated to bridge the gap to existing DIII-D IDL routines. Significant advancements were made with the SOLPS module in OMFIT which now supports SOLPS-ITER. Progress was also made towards further integrating OMFIT with the ITER-IMAS database via the OMAS library. Finally, numerous bug fixes and improvements were also made to the OMFIT framework itself. In particular, the work of transitioning from Python 2 to Python 3 by the end of 2019 has started, which is an important advancement to secure the longevity of the OMFIT project and ensure that OMFIT users and developers will remain at the forefront of fusion science for years to come.

Nonlinear gyrokinetic simulations comparing hydrogen, deuterium, and tritium plasmas have been used to explore the influence of kinetic electron effects on the scaling of turbulent transport with hydrogen isotope mass. These effects are neglected in the standard gyroBohm scaling. For ITG-driven turbulence, the energy flux increases as the ion mass increases. While this trend is qualitatively consistent with the standard gyroBohm scaling, the scaling factor is not. Relative to the deuterium transport, the scaling is stronger than the gyroBohm scaling for hydrogen, while for tritium the scaling is weaker than the gyroBohm scaling. The critical temperature gradient is, however, independent of mass ratio. For linear cases, strong collisionality can decouple the electron and ion dynamics, recovering the intrinsic mass scaling of the linear growth rate with the inverse square of the ion mass, but only for unrealistic collision frequency.