The electronic properties of silicon nitride

Abstract

The nature of bonding in silicon nitride is treated using simple bond-orbital models. A nitrogen pπ lone-pair valence band maximum is found. This is due to the planar nitrogen site, which is in turn due to the repulsions between non-bonded second-neighbour silicon atoms. The conduction minimum is found to have a relatively low effective mass and be formed of Si 3s states. The density of states (DOS) for β-Si₃N₄ is calculated for two Si–N–Si bond angles, 120° and 107°. The DOS at the former angle, which corresponds to a planar nitrogen, shows an upper pπ valence band which has merged into the lower bonding bands at 107°. The effects of non-bonded silicon–silicon repulsions on the planarity of the nitrogen site and likely structure of amorphous silicon nitride are discussed. Unlike vitreous SiO₂, commercial silicon nitride contains an appreciable proportion of impurities which may determine electronic transport and low-energy optical properties. The local electronic structure of ≡Si–Si, ≡Si–H, =N–H and ≡Si–O–Si≡ impurity configurations are calculated, and the first two are found to give rise to states in the pseudogap—unlike in a-Si, hydrogenation will not remove all dangling-bond states from the gap of silicon nitride. The possible structures of coordination defects in silicon nitride are calculated by simple bond-orbital methods and by use of molecular analogies. The low Lewis basicity of (SiH₃)₃N shows that overcoordination of nitrogen is difficult, and so defects in silicon nitride are expected to have a positive effective correlation energy. This result rules out a model of charge storage in MNOS devices based on nitride charged defects.