Modelling of alpha particle slowing down, confinement and redistribution by sawteeth in TFTR using the FPPT code

Results from recent deuterium-tritium (DT) experiments on TFTR to measure the energy spectra and radial profiles of well trapped confined alpha particle distributions using the pellet charge exchange (PCX) diagnostic in quiescent plasmas are compared with a numerical modelling using the Fokker-Planck post-TRANSP (FPPT) processor code and show a classical slowing down behaviour of the alphas (Fisher, R.K., et al., Phys. Rev. Lett. 75 (1995) 846). However, in the presence of sawtooth oscillations PCX experimental data indicate a significant broadening of the trapped alpha radial distributions (Petrov, M.P., et al., Nucl. Fusion 35 (1995) 1437). Conventional models consistent with measured sawtooth effects on passing particles do not provide satisfactory simulations of the trapped particle mixing measured by the PCX diagnostic. A mechanism is proposed for fast particle mixing during the sawtooth crash to explain the trapped alpha particle radial profile broadening after the crash. The model is based on the fast particle orbit averaged toroidal drift in a perturbed helical electric field with an adjustable absolute value (similar to that in Kolesnichenko, Ya.I., et al., Nucl. Fusion 36 (1996) 159). Such a drift of the fast particles results in a change of their energy and a redistribution id phase space. To show the sensitivity of trapped particles to sawteeth, the redistribution in toroidal momentum Pphi (or in minor radius) was calculated in two ways. The first is based on the assumption that Pphi redistribution is stochastic with a large diffusion coefficient and was taken to be flat. The second way is to apply Kolesnichenko's inversion formula (Kolesnichenko, Ya.I., et al., Nucl. Fusion 32 (1992) 449). Both methods are compared with PCX data. The distribution function in a pre-sawtooth plasma and its evolution in a post-sawtooth crash plasma are simulated using the FPPT code. It is shown that FPPT calculated alpha particle distributions are consistent with TRANSP Monte Carlo calculations. Compariso