A new numerical scheme for core transport calculations using the GLF23 model has been successfully tested on the PTOR transport code to simulate the evolution of a DIII-D NCS discharge. The scheme is fully implicit allowing much larger time steps than previous methods. The method involves a new way to map the GLF23 fluxes onto a conventional diffusion and convection type transport system of equations. The simulation evolved both ion and electron temperatures, and toroidal and ExB velocity. This is the first time the ExB velocity evolution has been computed by solving the perpendicular momentum balance equation.

The bootstrap current modules, eqd_jbs and toq_jbs, were upgraded to allow analytic and spline inputs for density and zeff profile variation. The eqd_jbs code uses EFIT equilibrium data, and toq_jbs those of TOQ. The graphic library, PGPLOT, was used to produce the graphic output. The source codes are now platform independent and run on HP and alpha workstations. The new modules will greatly improve the modeling of the bootstrap current in DIII-D Advanced Tokamak discharges with steep internal or edge density gradients.

Test runs for the DCON ideal MHD stability code developed by A. Glasser (LANL) were performed in the control room between shots for the recent wall stabilization experiments. Preliminary analysis indicates that the code predictions were generally in reasonable qualitative agreement with observed MHD behavior; for example, the code predicted an internal instability at the time a beta collapse was observed. Confirmation of the predictions will still need to be made using post control room kinetic EFIT equilibria but this test indicates that DCON can provide a reasonable guide between shots. Work is now underway to link DCON directly with the MDSPLUS data base so that it can be used as a routine and semi automated control room tool.

The 1/R variation in the tokamak magnetic field strength induces an excess magnetic-curvature drift current inside the ablated and ionized pellet substance, causing acceleration of pellet clouds toward the large-R side of the tokamak. The acceleration persists until the cloud comes into pressure equilibrium with the background plasma by a combination of two effects (1) parallel expansion and (2) cross-field drift up the pressure gradient, as in the case of pellets launched from the inner wall of the tokamak. An extension of a recent cloud drift model (Parks et al APS 1999) includes both effects self-consistently. The extended model predicts that inside launch will be successful for a tokamak with nominal ITER-FEAT parameters: the ablated fuel from a 6mm diameter pellet with a velocity of only 800 m/s will drift all the way to the magnetic axis and stop.

The 12-processor Linux PC cluster has been put into production mode for between-shot EFITs during this week's operation. As a result, the EFITs become available usually within two minutes upon the completion of a shot at 25 millisecond resolution. A new interface has also been created so that the physics operators and session leaders can directly control the EFIT setups including the time resolution and the input parameter file being used.

A stand-alone code XNRATE has been written to compute the neutron rates from thermal-thermal, beam-thermal, and beam-beam interactions using the fusion cross sections from Bosch and Hale (Nucl. Fusion, 1992). It has been recently parallelized with MPI and tested on the LUNA PC Linux cluster, and dramatic improvements in the CPU time have been achieved. With one processor, the CPU time is approximately 100 secs and with 16 processors the CPU time is reduced to under 12 secs. This complements the other physics modules already parallelized including GLF23, FREYA, and TORAY. Future plans are to port XNRATE and the other parallel codes to ONETWO and begin testing it on the Luna system.

Theoretical examples were used to demonstrate that, with appropriate feedback algorithm, including high negative values of gain, control of RWM using sensor loops outside of the resistive vessel should be as effective as using sensor loops inside. It is speculated that the advantages of inside placed sensor loops reported previously in other studies resulted either from not fully optimizing the feedback system with respect to the gain, or from an incomplete specification of the feedback algorithm.