It was recalled that the migration of these elements yielded Si self-interstitial diffusivities, that exceeded those which were obtained from dopant marker experiments by 6 orders of magnitude at 800C. It was noted that both types of experiment could be reconciled by assuming the existence of a non-annihilating interstitial trap that was related to C. Selected metal diffusion data were re-analyzed in this context. Non-annihilating immobile traps and a second-order reaction which involved interstitial C and C-C pairs were considered. Quantitative point defect parameters were obtained at 1115C for an assumed trap concentration of 5 x 1016/cm3. Excellent fits to Zn concentration versus depth profiles were obtained in all cases, regardless of the trap concentration. The estimated equilibrium concentration of Si self-interstitials varied inversely with the trap concentration, while the product of the self-interstitial diffusivity and the self-interstitial concentration remained almost constant. It was concluded that agreement, or disagreement, of metal diffusivity results with the Si self-diffusion coefficient could not be used to exclude or prove the occurrence of trap-limited diffusion. The published values of point-defect parameters which had been obtained by neglecting traps were suggested to represent lower bounds on the self-interstitial diffusivity, and upper bounds on the equilibrium concentration of self-interstitials. The temporal evolution of the Zn profile in the presence of traps was considered, and the proper inclusion of traps in the analysis of metal diffusion data led to a decrease in the estimated activation energy for self-interstitial diffusion.

H.J.Gossmann, P.A.Stolk, D.J.Eaglesham, C.S.Rafferty, J.M.Poate: Applied Physics Letters, 1995, 67[21], 3135-7