An Examination of the Mechanisms of Si Diffusion in GaAs
- PDF / 276,394 Bytes
- 5 Pages / 420.48 x 639 pts Page_size
- 72 Downloads / 235 Views
AN EXAMINATION OF THE MECHANISMS OF Si DIFFUSION INGaAs SHAOFENG YU, ULRICH M.GOSELE AND TEH Y.TAN School of Engineering, Duke University Durham, NC 27706 ABSTRACT An examination of the three available quantitative models of Si diffusion in GaAs has led to the conclusion that the Fermi-level effect mechanism plays the most essential role. In some experimental results a point defect concentration transient is involved which should be incoorporated in future models. INTRODUCTION Being the main n-type dopant used in device fabrications in GaAs based materials, one would think that a thorough understanding of the mecahnisms governing Si diffusion has long been established. The contrary, however, is true. At low concentrations, Si atoms occupy the Ga sites to constitute shallow donors, SiGa+, and it is known that the Si diffusivity increases with an increase of the As pressure [1,2]. This means SiGa+ diffuse via the Ga vacancy VGa. At high concentrations, Si does not produce an equal amount of electrons, because now some Si atoms occupy the As sites to constitute shallow acceptors, SiAs, which leads to a self-compensation [3,4]. Measured complete Si diffusion profiles were not available until quite recently [5-10]. Analyses of these profiles have led to the three quantitative Si diffusion models discussed in the following. QUANTITATIVE Si DIFFUSION MODELS With a high surface concentration, the measured Si diffusion profiles is concentration dependent. At the diffusion front, the profile declines steeply [5-10]. To explain this kind of feature, Greiner and Gibbons [5,6] proposed a SiGa-SiAs pair diffusion model, the GG-model. In this model it is assumed that the concentration of the SiGa-SiAs pairs increases as the Si concentration is increased, since there will then be more SiAsavailable. It is further assumed that the pairs diffuse fast by hopping alternately into VGa and -LAs while SiGa+ and SiAs- do not move. The concentrations of YVGa and V_As are assumed to be at their thermal equilibrium values. The GG-model has yielded satisfactory fits for the main concentration dependent part of some experimental Si profiles obtained under rapid thermal annealing (RTA) conditions [ 5-7].
Mat. Res. Soc. Symp. Proc. Vol. 163. ©1990 Materials Research Society
672
For the RTA results, however, a gradient in the concentrations of the point defects, induced by the annealing process, may not be ignorable. Noting this point, Kavanagh et al. [8] proposed a point defect gradient model, the K-model, which fitted also satisfactorily the main concentration dependent part of some RTA results. In this model it is assumed that vacancies ., not distinguished between those of the Ga and those of the As sublattices, are injected from the surface into the GaAs interior starting at the onset of the RTA process. Invoking an increasing V surface concentration as the RTA time increases, which in principle can result from the transient V. injection process, a K concentration gradient produces in the near-surface region of the GaAs interior which in
Data Loading...
