A new quasi-linear method for predicting the characteristics of the micro turbulent eddies observed in tokamaks with 2D visualization measurements has been developed. Since the fluctuation spectrum of gyro-kinetic turbulence is sharply peaked at a particular poloidal and radial wavenumber, the mode at the peak of the spectrum is the most important linear mode. This is the basis for quasi-linear transport flux calculations but it can also be applied to visualize the turbulent eddies. Using the same quasi-linear elements as are used by the TGLF transport model, a movie of the dominant 2D eddies can be constructed in real space and time and compared to data. This method makes a direct connection between the quasi-linear fluxes and the eddy-scale turbulence properties validating the theory at a greater level of detail.

The relationship between previously predicted alpha particle density flattening by alpha-driven Alfvén eigenmodes (AEs) in a stiff transport model, and the corresponding redistribution of alpha heating, is now better understood. Evaluation with GYRO of the quasilinear flux due to local AEs, as a function of velocity for an ITER baseline case, found that energy flux in the driving alpha particle channel exceeds the value expected for purely convective transport by a factor of 1.2 to 1.7 across the radial domain. For strong drive, a broad resonance is clearly resolved in energy space. Very close to marginality, the resonance becomes narrow and required velocity-space resolution becomes prohibitive even for determining mode frequency. For well-resolved cases, the ratio of true energy transport to purely convective energy transport remains essentially unchanged with drive strength, even as the energy-space resonance is observed to broaden. The convective factor for ITG/TEM micoturbulent transport of alpha particles is much lower at about 0.03. Since the ITG/TEM transport dominates at the edge, the net birth alpha heating loss is one hundred times the net birth alpha particle loss.

An analytic theory for the neoclassical transport of a trace heavy impurity in a rotating plasma in the Pfirsch-Schluter regime has been derived. Unlike previous theories which rely on a fluid or Braginskii approximation, a unique kinetic approach for the both main ions and impurities is used, allowing for an extension of previous results for larger, more experimentally-relevant Mach numbers and general geometry. Treatment of ion-impurity collisions with the exact linearized Fokker-Planck operator, rather than a model operator, also reveals some corrections for the impurity transport. The method allows for solution of the complete diffusive and convective components of the ambipolar particle flux, which are essential for computing the profile shape in transport calculations. Consistent with previous theories, an enhancement of the diffusion coefficient is found for impurity Mach numbers of order one due to the increased effective toroidal curvature. However, a new result at large Mach number reveals a reduction due to the rotation-driven poloidal asymmetry in the density. The new formulae provide convenient forms for code validation and have been shown to agree to high accuracy with NEO, which is presently the only neoclassical code to fully include sonic toroidal rotation effects.

During the month of March, Professor Howard Wilson and student Amelia Dowsett are visiting GA from the University of York to work with Phil Snyder on extension of the ELITE stability code to lower toroidal mode numbers.

In a collaboration with ASIPP-Hefei, a parallel, Graphical Processing Unit (GPU)-optimized version of the EFIT equilibrium reconstruction code has been successfully installed and tested on a local Linux workstation with a GPU board. This GPU-optimized version (P-EFIT) of the GA-developed EFIT code was developed at ASIPP-Hefei to support EAST tokamak operation. P-EFIT is based on the EFIT framework but takes advantage of massively parallel GPU cores to significantly accelerate the computation. P-EFIT is developed using the CUDA (Compute Unified Device Architecture) parallel computing architecture to allow integration of Central Processing Units (CPUs) and GPUs using C programming. The parallel processing is implemented with the Single-Instruction Multiple-Thread (SIMT) architecture. Initial magnetic reconstruction testing results for a DIII-D discharge on a 65×65 spatial grid indicate that P-EFIT could accurately reproduce the EFIT reconstruction algorithms at a fraction of the computational cost. This P-EFIT version is limited to magnetic reconstructions only and has limited current profile representation capability. Ongoing work includes an extension to allow more current profile representations, higher resolution spatial grids, more experimental measurements and physics constraints, and more flexible user interfaces.

**Disclaimer**

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