The neutral gas shielding model for ablation of frozen hydrogenic pellets is extended to include the effects of (1) an initial Maxwellian distribution of incident electron energies; (2) a cold plasma shield outside the neutral shield and extended along the magnetic field; (3) energetic neutral beam ions and alpha particles; and (4) self-limiting electron ablation in the collisionless plasma limit. Including the full electron distribution increases ablation, but adding the cold ionized shield reduces ablation; the net effect is a modest reduction in pellet penetration compared with the monoenergetic electron neutral shielding model with no plasma shield. Unlike electrons, fast ions can enter the neutral shield directly without passing through the cold ionized shield because their gyro-orbits are typically larger than the diameter of the cold plasma tube. Fast alpha particles should not enhance the ablation rate unless their population exceeds that expected from local classical thermalization. Fast beam ions, however, may enhance ablation in the plasma periphery if their population is high enough. Self-limiting ablation in the collisionless limit leads to a temporary distortion of the original plasma electron Maxwellian distribution function through preferential depopulation of the higher energy electrons.

A method of analysing plasma performance over large regions of density and temperature space with time-dependent multi-dimensional transport codes is presented. Contour plots of global steady-state plasma parameters are generated and then used to show: the relationships between driven and ignited operation; regions of thermal stability; the effects of the shift of the magnetic axis with increasing beta on beam penetration and thermal conduction losses from toroidal field ripple; and optimal neutral beam heating during start-up in tokamak reactor applications.

The measured penetration depth lambda of deuterium pellets injected into the Joint European Torus (JET) confirms some features of the neutral gas shielding (NGS) model, but not others. The scaling of lambda with plasma and pellet parameters agrees with the NGS model, as in earlier ASDEX studies. Pellet velocity was varied over the range 0.46-1.35 km/s in the JET experiments to test specifically the scaling of lambda with velocity. This scaling also agrees with the NGS model. However, the penetration is deeper in JET than in ASDEX when it is corrected for the expected machine size dependence. Furthermore, the penetration depths measured in JET are greater (by nearly a factor of two) than those predicted by local ablation calculations using the NGS model with an incident Maxwellian distribution of electrons. Plasma shielding used in previous modelling of the JET penetration data can account for the additional shielding, but it also removes the observed velocity dependence. The implications of both the scaling observations and the penetration depths for improvements in ablation theory and in the models are discussed

A centrifuge injector that repetitively fires 1.3 mm deuterium pellets (1 torr • L per pellet) at a rate of 32 pellets per second was used to build up and maintain a Doublet III 2.4 MW neutral-beam-heated limiter discharge at a line-averaged density of 1 X 10 14 cm" 3 . When compared to a conventional gas-fuelled plasma at similar density, the pelletfuelled plasma was characterized by a factor-of-three reduction in edge neutral density and limiter recycling, a centrally peaked profile, a 70% increase in global energy confinement, and a tenfold increase in the fusion reaction rate.

Initial hydrogen pellet injection experiments have been performed in plasmas characterized by a low particle re-cycle fraction resulting from the action of a poloidal magnetic divertor. As in past experiments, the interaction of the fuel with the plasma is observed to be adiabatic. Moreover, the pellet mass is accounted for in the core plasma density increase, indicating only a small loss of fuel while the pellet transits the divertor scrape-off plasma. The effect of edge re-cycling on the density was studied by comparing divertor (low-re-cycle) and limiter (high-re-cycle) plasmas; the distinction between the two cases is clearest in the edge plasma region where the density decay rates differ most. Particle transport subsequent to pellet injection is less ambiguous for divertor cases, and, by comparing the density profile relaxation and the electron temperature recovery with an empirical transport model that closely approximates the pre-injection plasma conditions, it is concluded that the plasma confinement properties do not deteriorate as a result of pellet injection. The principal difference between central and edge fuelling is demonstrated by a peaking of the density profile and an extended decay time for the density perturbation.