Dopant diffusion and segregation in semiconductor heterostructures: Part III, diffusion of Si into GaAs

We have mentioned previously that in the third part of the present series of papers, a variety of n-doping associated phenomena will be treated. Instead, we have decided that this paper, in which the subject treated is diffusion of Si into GaAs, shall be the third paper of the series. This choice is arrived at because this subject is a most relevent heterostructure problem, and also because of space and timing considerations. The main n-type dopant Si in GaAs is amphoteric which may be incorporated as shallow donor species Si-Ga(+) and as shallow acceptor species Si-As(-). The solubility of Si-As(-) is much lower than that of Si-Ga(+) except at very high Si concentration levels. Hence, a severe electrical self-compensation occurs at very high Si concentrations. In this study we have modeled the Si distribution process in GaAs by assuming that the diffusing species is Si-Ga(+) which will convert into Si-As(-) in accordance with their solubilities and that the point defect species governing the diffusion of Si-Ga(+) are triply-negatively-charged Ga vacancies V-Ga(3-). The outstanding features of the Si indiffusion profiles near the Si/GaAs interface have been quantitatively explained for the first time. Deposited on the GaAs crystal surface, the Si source material is a polycrystalline Si layer which may be undoped or nf-doped using As or P. Without the use of an As vapor phase in the ambient, the As- and P-doped source materials effectively render the GaAs crystals into an As-rich composition, which leads to a much more efficient Si indiffusion process than for the case of using undoped source materials which maintains the GaAs crystals in a relatively As-poor condition. The source material and the GaAs crystal together form a heterostructure with its junction influencing the electron distribution in the region, which, in turn, affects the Si indiffusion process prominently.