Calculations of the free-boundary plasma response to applied 3D fields in DIII-D have recently been carried out using M3D-C1. This contrasts with previous M3D-C1 calculations, in which a perfectly conducting boundary was used at roughly the location of the DIII-D first wall. In the free-boundary calculations, the first wall is modeled as a resistive structure, with the conducting boundary moved far outside of the vacuum vessel. This capability will allow a more accurate model of the low-frequency plasma response, in which eddy currents do not play a significant role. Additionally, because the fields due to the plasma can now extend beyond the location of the first wall, this will allow a comparison with the M3D-C1 results with magnetic probe data and a more accurate accounting of the electromagnetic torque on the plasma (since the JxB torque on the plasma is equal and opposite to Jcoil x Bplasma, which is zero in the conducting wall case). Finally, the resistive wall capability will allow closer comparison with other codes, such as IPEC and MARS, which are typically run with a free boundary condition. Preliminary results from free-boundary calculations with M3D-C1 generally show somewhat improved screening of the pitch-resonant field components, and also reduced kink excitation in some cases.

A neoclassical model for studying poloidal asymmetries in the density that are induced by anisotropic RF minority ions, based on a generalization of the Hinton-Wong strong toroidal rotation theory, has been developed. Even in the diamagnetic limit, the anisotropic minority causes a nonuniform redistribution of the particles around a flux surface due to the mirror force, analogous to the centrifugal effect in sonic rotation. As a result of quasi-neutrality, a poloidally-varying electostatic potential is generated to balance the density asymmetry. The effect of this potential on the neoclassical transport of heavy impurities is studied with NEO. The poloidal variation in the potential is computed using a form for the density variation of a species with an anisotropic temperature derived from parallel force balance and assuming a bi-Mawellian distribution. In contrast with toroidal rotation effects (for which the density is localized at the outboard midplane), the poloidal distribution induced by ICRH produces a reverse in-out symmetry. NEO simulations find that, unlike for toroidal rotation, small temperature anisotropies reduce rather than enhance the impurity diffusivity, which may have implications for studies of high-Z transport in ICRH plasmas.

C. Holland (UCSD), J. Candy (GA), and N. Howard (ORISE) have received a 2014 ASCR Leadership Computing Challenge (ALCC) award titled “Validation Studies of Gyrokinetic Simulations to Understand the Coupling of Ion and Electron Scale Turbulence in Tokamak Plasmas”. This award provides 50,000,000 hours at NERSC and 90,000,000 at ALCF to perform physically comprehensive gyrokinetic simulations of coupled ITG and ETG turbulence in the DIII-D and Alcator C-mod tokamaks using the GYRO code. Initial studies will focus upon the well-studied L-mode shortfall issue in DIII-D, and extend ongoing analysis of Alcator C-mod L-mode discharges that have a well-documented under-prediction of the electron heat flux by conventional ion-scale gyrokinetic simulations. These studies build on initial simulations by N. Howard which demonstrated for the first time that ITG and ETG turbulence can co-exist at experimentally significant levels in gyrokinetic simulations using experimentally measured inputs, and that inclusion of the ETG scale fluctuations both drives an increase in the ion-scale turbulence and transport, and resolves the previously observed under prediction for the case considered. These results were presented at both the Sherwood theory and Transport Task Force meetings.

A highlight was published by Stephen Benka in the May issue of Physics Today on the recent solution of the mystery of magnetic plasma confinement in transport barriers by Gary Staebler and the DIII-D team. The article can be accessed at http://scitation.aip.org/content/aip/magazine/physicstoday/news/10.1063/PT.5.7071;jsessionid=7e668jkjepa7c.x-aip-live-06

The neural network-based (NN) model developed previously (see Highlight from January 31 2014 at Theory Weekly Highlights for January 2014) has been extended to simulate particle and momentum fluxes, in addition to the ion and electron heat transport. A FORTRAN implementation of the NN model was also subsequently integrated into the ONETWO transport code. The NN predictions were benchmarked against over 4000 DIII-D experimental discharge time-slices across different regimes, and were able to capture the experimental behavior inside of rho < 0.95, with average error for all transport channels of less than 20%. The numerical efficiency of this method, requiring only a few CPU microseconds per data point, makes it an ideal candidate for real time plasma control and scenario preparation for experiments. With the NN model embedded in ONETWO, time-dependent DIII-D transport simulations are now being validated against the experiments. The NN model is proving to be numerically non-stiff, which results in robust convergence, and stable transport solutions can be found with a few minutes of computation.

Gary Staebler was elected vice-chair of the US-TTF workshop. He plans to promote the use of theoretical predictions in the planning of experiments and the validation of the outcomes.

A simulation of localized above-midplane massive gas injection into DIII-D was performed with NIMROD to mimic the MEDUSA gas-injection valve and compare to below-midplane (CERBERUS-like) injection. The MEDUSA simulation reproduced the finding from the CERBERUS simulation that a very localized gas source model produces a radiation toroidal peaking factor in line with experimental measurements (~1.2-1.4). In both simulations, the injected impurity ions spread helically along field lines due to the parallel pressure gradient introduced by the impurities. The spreading was more pronounced in the core of the plasma (inside q=2) than closer to the separatrix. The spreading in each case was toroidally asymmetric, with the preferred toroidal direction reversing for the upper versus lower valve injection. In each case the direction corresponding to poloidal propagation toward the high-field side was favored. Certain free parameters, such as particle diffusion, will continue to be adjusted to get the best experimental match for the individual CERBERUS and MEDUSA simulations, after which two-valve simulations will be run.

**Disclaimer**

These highlights are reports of research work in progress and are accordingly subject to change or modification