Theoretical study of charge trapping levels in silicon nitride using the LDA-1/2 self-energy correction scheme for excited states

Silicon nitride, with a permittivity mid-way between SiO2 and common high-k materials such as HfO2, is widely used in microelectronics as an insulating layer on top of oxides where it serves as an impurity barrier with the positive side effect of increasing the dielectric constant of the insulator when it is SiO2. It is also employed as charge storage in nonvolatile memory devices thanks to its high concentration of charge traps. However, in the case of memories, it is still unclear which defects are responsible for charge trapping and what is the impact of defect concentration on the structural and electronic properties of SiNx. Indeed, for the amorphous phase the band gap was measured in the range 5.1–5.5 eV, with long tails in the density of states penetrating the gap region. It is still not clear which defects are responsible for the tails. On the other hand, the K-center defects have been associated with charge trapping, though its origin is assigned to one Si back bond. To investigate the contribution of defect states to the band edge tails and band gap states, we adopted the β phase of stoichiometric silicon nitride (β-Si3N4) as our model material and calculated its electronic properties employing ab initio DFT/LDA simulations with self-energy correction to improve the location of defect states in the SiNx band gap through the correction of the band gap underestimation typical of DFT/LDA. We considered some important defects in SiNx, as the Si anti-site and the N vacancy with H saturation, in two defect concentrations. The location of our calculated defect levels in the band gap correlates well with the available experimental data, offering a structural explanation to the measured band edge tails and charge trapping characteristics.