The growth of Si doped LEC GaAs shows a significant deviation in the Si incorporation from the prediction of a classic segregation model. As a rule, it is expected that for a given impurity segregation coefficient, the dopant incorporation throughout the crystal may be calculated with fairly good accuracy. In fact, Te-doped GaAs LEC doping profiles show a close approximation to this classic model. However, in Si doped GaAs LEC growth, the doping distribution in the crystals deviates significantly from the segregation model. In some cases, this deviation is of several orders of magnitude. The degree of deviation was found to be dependent on the growth conditions. This work was performed to understand the source of the deviation from the model and to allow for an accurate account of the Si incorporation. In this work, Si doped GaAs crystals were grown by low pressure LEC. These crystals were targeted for doping in the 1016 to low 1018 cm-3 range. The actual dopant incorporation was significantly lower than the segregation model predicted, and the axial doping variation was much too flat. Moreover, in spite of an intentional Si doping, a significant number of crystals were semi-insulating. Analysis of the experimental data has shown that this doping behavior cannot be explained by the assumption of an unknown compensating impurity. A different model has been developed that is consistent with all of our observations and that provides an accurate account of the Si incorporation. Understanding the Si incorporating anomaly has allowed us to make successful process changes and improvements. This model explains the observed doping anomaly by taking account of the diffusion of Si between the GaAs melt and the B2O3 encapsulant, as well as some permanent Si trapping in the B2O3. When the interaction of the Si with B2O3 is taken into account, the Si incorporation follows the segregation model and the anomaly vanishes.
Read full abstract