Use of Neural Network to Characterize Temperature Effects on Deposition Rate of PECVD-Silicon Nitride Thin Films

Temperature effects on the deposition rate of silicon nitride films were characterized by building a neural network prediction model. The silicon nitride films were deposited by using a plasma enhanced chemical vapor deposition system and process parameter effects were systematically characterized by 2$^{6-1}$ fractional factorial experiment. The process parameters include a radio frequency power, pressure, temperature, SiH$_{4}$, N$_{2}$, and NH$_{3}$ flow rates. The prediction performance of generalized regression neural network (GRNN) was considerably improved by optimizing multi-valued training factors using a genetic algorithm (GA). Compared to the conventional model, GA-GRNN model demonstrated an improvement of more than 70 $\%$. Several 3-D plots were generated to interpret temperature effects at various plasma conditions. It is noticeable that typical effects of SiH$_{4}$ and NH$_{3}$ flow rate were observed only at higher and lower temperatures, respectively. Depending on the levels of SiH$_{4}$ (or higher NH$_{3})$ flow rate, the temperature effects were quite different. Deposition mechanisms were qualitatively estimated.

Publication