Abstract
Charge density is an important film characteristic that determines the quality of film surface passivation. In this study, a prediction model of charge density was constructed by using a neural network and generic algorithm. For a systematic modeling, plasma enhanced chemical vapor deposition process of silicon nitride films was characterized by means of a statistical experimental design. Effects of parameters under various radio frequency (rf) powers were examined under various plasma conditions. An increase in charge density was observed with increasing rf power over the entire range of SiH4 flow rate as well as at lower N2 flow rates or pressures. The effect of SiH4 flow rate on charge density was opposite to that for N2 flow rate given rf powers. For the variations in N2 or SiH4 flow rate, charge density seemed to be strongly correlated to N/Si ratio. Under the variations in pressure or N2 flow rate, maintaining a lower pressure was beneficial in achieving a higher charge density. A pronounced rf power effect at smaller grain size was expected.
