Accurate prediction of structural and mechanical properties on amorphous materials enabled through machine-learning potentials: a case study of silicon nitride

Aug 6, 2026

Upload by Violet

Amorphous silicon nitride (a-SiN) is widely used for its excellent mechanical and electrical performance, yet the link between microscopic structure and macroscopic behavior remains incompletely understood. Ab initio approaches offer high fidelity but are limited by simulation cell size, which strongly affects amorphous-system results. This study addresses the limitation by training a machine-learning interatomic potential on ab initio data, then running molecular dynamics on much larger models. The simulations reproduce experimental elastic properties, including elastic isotropy, demonstrating machine-learning potentials as effective tools for structural and mechanical prediction in complex amorphous materials.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages12

Reading Time30 min

AccessFree to download

Introduction

Motivation: amorphous SiN and modeling challenges

Approach: machine-learning potentials and larger-scale molecular dynamics

Goal: reproduce experimental elastic/structural properties

arXiv :2408 .05782v1 [ cond-mat .mtrl-sci ] 11 Aug 2024 Accurate prediction of structural and mechanical properties on amorphous materials enabled through machine-learning potentials: a case study of silicon nitride Ganesh Kumar Nayaka,1 , Prashanth Srinivasanb , Juraj Todta , Rostislav Daniela , Paolo Nicolinic and David Holeca a Department of Materials Science, Montanuniversität Leoben, Franz-Josef-Strasse 18, 8700 Leoben, Austria b Department of Materials Design, Institute for Materials Science, University of Stuttgart, Pfaffenwaldring 55, 70569 Stuttgart, Germany c Institute of Physics (FZU), Czech Academy of Sciences, Na Slovance 2, 18200 Prague, Czechia | ARTICLE INFO | | AB STRACT | | --- | --- | --- | | Keywords:
Coatings
First-principles calculations Molecular Dynamics Machine-learning force field Amorphous Silicon Nitride Mechanical Properties | | Amorphous silicon nitride (a-SiN) is a material which has found wide application due to its excellent mechanical and electrical properties. Despite the significant effort devoted in understanding how the microscopic structure influences the material performance, many aspects still remain elusive. If on the one hand ab initio calculations respresent the technique of election to study such a system, they present severe limitations in terms of the size of the system that can be simulated. Such an aspect plays a determinant role, particularly when amorphous structure are to be investigated, as often results depend dramatically on the size of the system. Here, we overcome this limitation by training a machinelearning (ML) interatomic model to ab initio data. We show that molecular dynamics simulations using the ML model on much larger systems can reproduce experimental measurements of elastic properties, including elastic isotropy. Our study demonstrates the broader impact of machine-learning potentials for predicting structural and mechanical properties, even for complex amorphous structures. | 1. Introduction Silicon nitride is a ceramic material of great technological interest with diverse applications owing to its good mechanical and electrical properties [1–9] . Si3N4 can be synthesized using various processes such as sputtering, chemical vapor deposition, and glow-discharge decomposition [10] . Sintered Si3N4 components exhibit high density, high melting temperature, low mechanical stress, high thermal strength, strong resistance against thermal shock, and fracture toughness, and are used in many engineering applications [11, 12]. For example, they are used for making engine components and cutting tools due to their superior mechanical properties at high temperatures [3, 4] . SiN􀁸 in the nanocomposite TiN/SiN􀁸 acts as a protective coating due to its high hardness and excellent wear-resistance [5–8] . Incorporating Si3N4 as a second phase has been proposed to improve the tribological performance of a material, while keeping the other wear properties intact [8, 9] . Additionally, Si3N4 in the amorphous form also has several technological benefits. Thin films of a-Si3N4 exhibit a high dielectric constant, a high-energy barrier for impurity diffusion, high resistance against radiation, and show oxidation resistance up to 1500◦ C, making them ideal candidates for several microelectronic applications [13–15] and as agate dielectric in thin-film transistors [16] . Thick films of [nayak@mch.rwth-aachen.de](nayak@mch.rwth-aachen.de) (G.K. Nayak); [david.holec@unileoben.ac.at](david.holec@unileoben.ac.at) (D. Holec) [http://cms.unileoben.ac.at](http://cms.unileoben.ac.at) (D. Holec) ORCID(s): 0000-0002-6996-8998 (G.K. Nayak); 0000-0002-3516-1061 (D. Holec) 1Current affiliation: Materials Chemistry, RWTH Aachen University, Kopernikusstraße 10, 52074 Aachen, Germany Si3N4 are promising candidates for non-linear optical applications [17, 18] . The above properties together with the biocompatibility of a-SiN, also make it an exceptional candidate for

FAQs

Why are ab initio calculations difficult for studying amorphous silicon nitride?

They are constrained by small system sizes, and in amorphous materials the results can depend strongly on the modeled cell size.

How does the study enable accurate predictions of structural and mechanical properties?

It trains a machine-learning interatomic model on ab initio data, then uses molecular dynamics with this larger-scale ML potential.

What key mechanical property results are reproduced for larger amorphous systems?

The molecular dynamics simulations reproduce experimental elastic properties, including elastic isotropy.

Aug 6, 2026

Upload by Violet