Effect of topological disorder on structural, mechanical, and electronic properties of amorphous silicon nitride: An atomistic study

Abstract

ABSTRACT:We present a first principles study of the effect of atomic variability on the structure, mechanical and electronic properties of amorphous silicon nitride. Using a combination of molecular dynamics and density functional theory calculations we predict an ensemble of statistically independent, well-relaxed and stress-free amorphous silicon nitride structures. We analyze the short, intermediate, and long-range order of the structures generated using radial distribution functions, ring statistics, bond angle distributions, and translational invariance parameters. Though energetically very similar, these structures span a wide range of densities (2.75-3.25 g/cm 3 ) and bulk moduli (115GPa to 220 GPa) in good agreement with the fabrication-dependent experimental range. Chemical bonds and atomic defects are identified via a combination of bond distance cutoff and maximally localized Wannier function analysis. A significant number of the amorphous structures generated (~30%) are defect-free providing an ideal reference to characterize the formation energy of the various point defects and their defect energy levels. An analysis of the Kohn-Sham density of states and energetics of the structures reveals that defects in amorphous dielectrics have a distribution of associated properties (e.g. formation energies and electronic energy levels) due to variations in local atomic structure; this should be taken into consideration in physics-based continuum models of these materials.

Model form uncertainty versus intrinsic atomic variability in amorphous silicon oxides and nitrides

“…[18] and [10] respectively. The samples were equilibrated at a temperature higher than the predetermined melting point and ensembles of statistically independent samples were annealed to room temperature using empirical force fields.…”

“…A detailed description of structural generation characterization for both a-SiO 2 and a-Si 3 N 4 can be found in Refs. [18] and [10], respectively. We note that while all the a-SiO 2 structures where obtained under nominally identical conditions the a-Si 3 N 4 are annealed at different densities to mimic different processing conditions.…”

“…The use of the average energy as a reference to compute defect energies is a necessary evil due to the lack of one-to-one correspondence between defective and defect-free structures. Also of note is that we do not distinguish between defect types and compute the overall distribution of defect-pair formation energies; a detailed defect analysis was included in our prior publications [10,18]. We quantify model form uncertainties by comparing LDA and GGA predictions on a structure-by-structure basis.…”

“…[10] and references therein. The second kind has been particularly challenging to capture in simulations.…”

“…Computational techniques have become a useful and cost-effective way to predict the materials properties of a wide range of materials; in particular, multiscale approaches have been recently used to predict ensembles of low energy a-SiO 2 [6,18] and a-Si 3 N 4 [10,19] structures and their properties. At the heart of these predictions is density functional theory (DFT) [20], a quantum mechanical based technique that provides a good balance between accuracy and computational efficiency for a large variety of ground state properties of condensed matter systems [9].…”