NUMERICAL INVESTIGATION OF PULSED CHEMICAL VAPOR DEPOSITION OF ALUMINUM NITRIDE TO REDUCE PARTICLE FORMATION

摘要

Particles of aluminum nitride (A1N) have been observed to form during epitaxial growth of A1N films by metal organic chemical vapor deposition (MOCVD). Particle formation is undesirable because particles do not contribute to the film growth, and are detrimental to the hydraulic system of the reactor. It is believed that particle formation is triggered by adducts that are formed when the group-Ill precursor, namely tri-methyl-aluminum (TMA1), and the group-V precursor, namely ammonia (NH3), come in direct contact in the gas-phase. Thus, one way to eliminate particle formation is to prevent the group-Ill and the group-V precursors from coming in direct contact at all in the gas-phase. In this article, pulsing of TMA1 and NH3 is numerically investigated as a means to reduce A1N particle formation. The investigations are conducted using computational fluid dynamics (CFD) analysis with the inclusion of detailed chemical reaction mechanisms both in the gas-phase and at the surface. The CFD code is first validated for steady-state (non-pulsed) MOCVD of A1N against published data. Subsequently, it is exercised for pulsed MOCVD with various pulse widths, precursor gas flow rates, wafer temperature, and reactor pressure. It is found that in order to significantly reduce particle formation, the group-Ill and group-V precursors need to be separated by a carrier gas pulse, and the carrier gas pulse should be at least 5-6 times as long as the precursor gas pulses. The studies also reveal that with the same time-averaged precursor gas flow rates as steady injection (non-pulsed) conditions, pulsed MOCVD can result in higher film growth rates because the precursors are incorporated into the film, rather than being wasted as particles. The improvement in growth rate was noted for both horizontal and vertical reactors, and was found to be most pronounced for intermediate wafer temperature and intermediate reactor pressure.