Optimization of saturation current density of PECVD SiN coated phosphorus diffused emitters using neural network modeling

Emitter surface passivation by low temperature plasma enhanced chemical vapor deposition (PECVD) silicon nitride is investigated and optimized in this paper. We have found that the saturation current density of a 90±10 μ/sq phosphorus diffused emitter with N_s ≈3 x 10¹⁹ and X_j ≈0.3 µm can be lowered by a factor of eight by appropriate PECVD silicon nitride deposition and photoassisted anneal. PECVD silicon nitride deposition alone reduces the emitter saturation density (J_oe) by about a factor of two to three, and a subsequent photoanneal at temperatures ≥350°C reduces J_oe by another factor of three. In spite of the larger flat band shift for direct PECVD silicon nitride coating, the silicon nitride induced surface passivation is found to be about a factor of two inferior to the thermal oxide plus PECVD silicon nitride passivation due to higher interface state density at the SiN/SiO₂ interface compared to SiO₂/Si interface. A combination of statistical experimental design and neural network modeling is used to show quantitatively that lower radio frequency power, higher substrate temperature, and higher reactor pressure during the PECVD deposition can reduce the J_oe of the silicon nitride coated emitter.