The NIMROD extended MHD code was applied to simulate disruption mitigation scenarios for ITER baseline and hybrid plasmas using shattered pellet injection (SPI) with various choices of the impurity content. In particular, the ratio of the deuterium to neon number densities was varied within the initial pellet. With certain assumptions on the plasma model (single fluid, equal temperature for all particle species, fixed plasma resistivity and thermal conductivity coefficients), full thermal quench (TQ) is predicted for ITER 15MA/5.3T baseline plasmas within 8 ms, with a total of 0.5 kPa-m^3 pure neon impurity. Most of the thermal energy is dissipated via line radiation. The shattered pellets are fully ablated during the TQ. Doubling the neon fraction reduces the thermal quench time to about 5 ms, with partial ablation (~64%) of the initial pellet. Adding deuterium into the neon pellet further reduces the TQ time. For example, SPI with 0.5 kPa-m^3 of neon and 10 times larger amount of D_2 reduces the TQ time to about 4.5 ms. Modeling sensitivity scans show that the simulated time traces of the TQ and energy radiation are not sensitive to the assumed plasma resistivity or viscosity, though a high kinematic plasma viscosity does help to damp the MHD perturbations during the TQ.

Field-line integration simulations predict that a misalignment of the DIII-D toroidal field (see highlight for Sep 22, 2017 at https://fusion.gat.com/global/theory/weekly/0917) , if the plasma response is ignored, would lead to several 3D effects, including an n=1 toroidal modulation of the 3D strike point location and the appearance of a finger in the footprint of the separatrix manifolds on the divertor. A benchmarking of the predicted misalignment footprints has been carried out between several codes: namely TRIP3D, MAFOT, and VMEC. The comparison between TRIP3D and MAFOT, given the similarity of the algorithms and the vacuum fields only, is a check of the misalignment models implemented in the two codes, and results in excellent agreement. The comparison with VMEC, on the other hand, yields interesting results regarding the role of the ideal plasma response contribution for strike point splitting and finger formation. Preliminary analyses suggest that the plasma response partially screens the strike point splitting while the n=1 modulation of the strike point location remains similar to the vacuum models.

To accelerate prediction of the turbulent transport, the neural network (NN) accelerated model of the Trapped Gyro Landau-Fluid (TGLF-NN) code has been extended to ITER. Preparation for ITER operation relies on our ability to efficiently predict the plasma confinement with high physics fidelity. High-fidelity turbulent transport models such as TGLF remain one of the major bottlenecks in this process. A large database of ITER TGLF simulations was assembled for nuclear phase (D+T, He ash, and Ne impurity). The original NN implementation has been re-written to leverage state-of-the-art machine learning libraries (e.g. Tensorflow). Tuning of the NN model parameters (topology and training parameters) was carried out on Graphical Processing Unit (GPU) enabled clusters, both using Gaussian process based optimization, and random sampling of the configuration space. The resulting TGLF-NN model is capable of reproducing well the original energy, particle and momentum fluxes of TGLF for ITER conditions. Porting and testing of the TGLF-NN model in the TGYRO transport code is under way.

A more sophisticated investigation of the possibility of using the proposed 1 MW DIII-D helicon antenna to launch a fast wave into a post thermal quench discharge to mitigate runaway electrons was carried out. The analysis used the recently derived formula for the perpendicular absorption coefficient due to electron-ion collisions in the ray tracing code GENRAY. Using a DIII-D EFIT equilibrium for shot #122976, assuming a flat density profile with zero impurities, and for 500 MHz, two antenna locations were considered, namely at the midplane and at 45 deg above midplane. For both launch positions, the poloidal projection of the ray trajectory is downward then goes upwards and loops around a point near the magnetic axis before heading out to the boundary: in the midplane and 45 degree cases the distance of closest approaches of the trajectory to the magnetic axis are  min = 0.0225 and 0.0926, respectively. The corresponding path length of the rays projected on the poloidal plane are 1.76 m and 2.12 m. For Te = 25 eV the power remaining at  min is 14.2% for the midplane launch and 17.9% for the 45 degree launch. However, at Te = 10 eV the corresponding values drop drastically to 0.11 % and 0.25 %. Lowered collisional damping in post thermal-quench plasmas, however, makes fast-wave mitigation of RE challenging. Nevertheless, the density affects the ray path strongly and this will be explored in future work.