In this paper we present molecular dynamics simulations of silicon nitride, both in bulk and as an interface to crystalline silicon. We investigate, in particular, the bonding structure of the silicon nitride and analyze the simulations to search for defective geometries which have been identified as potential charge carrier traps when silicon nitride forms an interface with silicon semiconductors. The simulations reveal how the bonding patterns in silicon nitride are dependent upon the stoichiometry of the system. Furthermore we demonstrate how having an "interphase", where the nitrogen content in silicon gradually reduces towards pure silicon across a boundary region, as opposed to an interface where there is an abrupt drop in nitrogen concentration at the boundary, can result in significantly different numbers of certain important carrier trap

The effects of an interface gradient in nitrogen concentration on a number of important properties of amorphous hydrogenated silicon nitride / crystalline silicon (a-SiNx:H/c-Si) interfaces in the context of solar cell devices are investigated using molecular dynamics simulations. We simulate interfaces with a gradient of nitrogen which goes from SiN1.2 to Si over widths from 2 9 nm, in the presence of 10 at % hydrogen, to recreate the conditions present when SiNx layers are deposited onto c-Si by plasma enhanced vapour deposition. We examine how changing the width of the nitrogen gradient can affect a number of atomic level structural properties which influence the optical and electrical performance of solar cells. We examine the trajectories of our simulations to search for certain geometries which have previously been identified as being important at this interface. Siliconsilicon and silicon hydrogen bonds, which are help to determine the refractive index of the interface are shown to increase with increasing N gradient width. The fixed charge in the interface is also shown to increase with the width of the gradient. The results demonstrate how altering the width of the N layer can affect the efficiency of a-SiNx:H as both an anti-reflective coating and a passivation layer, and we suggest an optimal gradient width of, in the region of, 2 nm