The adverse effects of disruptions in tokamaks can be mitigated by rapid density increases using cryogenic liquid jets. However, a gas jet would be simpler to use and recent experiments on massive gas jet injection into DIII-D achieved the goal of quenching the poloidal magnetic field without generating runaway electrons. We have now developed a new theoretical model for the penetration of such a gas jet. Since the surface of the jet ablates like a pellet an ablation recoil pressure will be impressed along the surface of the jet, causing a Bernoulli-like deceleration. Penetration is assisted by radial compression of the jet density which helps to keep the dynamic pressure rho*V^2 from falling too rapidly during propagation. The equations describing the (z, t) dependence of the jet radius, density, pressure, and velocity were derived using the long slender jet approximation as previously done in the liquid jet penetration model. A fairly straightforward solution was found when jet cooling of the background plasma is neglected; however such cooling may be necessary for good penetration.

A key prediction of the GA stability-based model of edge localized modes (ELMs) that the ELM depth is determined by the radial penetration of the computed unstable edge kink mode into the core has now been tested quantitatively for the first time. In the past, this model for ELMs and constraints on the H-mode pedestal has successfully described the qualitative behavior of ELMs in DIII-D, JT-60U, and Alcator C-Mod over a wide range of conditions. A direct quantitative comparison has been undertaken for DIII-D shot 97887. The radial eigenmode structure of the n=10 peeling-ballooning mode which is the most unstable mode just prior to the first ELM was calculated with the ELITE code and compared to the observed ELM depth - that is the radial region over which the ELM leads to significant direct loss of energy and particles determined by statistical analysis of Thomson data across the first two large ELMs. The calculated and observed radial widths are in good agreement. Also, both extend inward well inside the pedestal. The promising agreement in this first studied case encourages further more detailed studies.

GYRO (the GA global electromagnetic gyrokinetic solver) has been ported to the NERSC SP (seaborg) and T3E (mcurie) machines and finite-beta testing is underway. So far, low-resolution runs at up to 50% of the MHD beta limit, show no sign of beta-related numerical instability or convergence problems. Time stepping error, which is monitored dynamically, is well below acceptable limits. We emphasize that (i) full electron dynamics on the bounce/transit timescale are retained; (ii) both ExB and magnetic flutter nonlinearities are computed with 4th-order time accuracy; (iii) arbitrary profiles can be simulated (although we are using a switchable flux-tube mode of operation that allows precise comparison with the only other nonlinear, fully-toroidal electromagnetic solver: GS2). Our preliminary findings indicate that transport increases weakly with beta up to about half the critical beta, beyond which wildly intermittent fluctuations in the thermal diffusivity are observed.