Toward Simulation of Silicon-Nanoparticle Shock-Synthesis Using Machine-Learned Interatomic Potentials

Silicon nanoparticles (Si-NP) exhibit superior optoelectronic and thermal properties that make them promising for advanced electronics, but scalable, controllable synthesis remains a challenge. Traditional low-pressure synthesis methods such as chemical vapor deposition offer precise control but are limited in scalability. Recent work has shown that detonation generated shockwaves in Si-containing precursors can be used to produce large quantities of Si-NPs on extremely rapid timescales1. This process could be replicated in a tunable fashion through laser driven shocks, as has been demonstrated both computationally and experimentally for nanocarbon synthesis2, 3. However, understanding of the underlying phenomena necessary to achieve this tunability remains a limiting factor. Simulations can fill this knowledge gap by providing an atomic-level description of the underlying phenomena but require interatomic models capable of “quantum accuracy” on large spatiotemporal scales. To meet these challenges, we are developing machine-learned interatomic models (ML-IAM) for Si-based materials using the ChIMES framework. In this presentation, we discuss the development and application of a ChIMES ML-IAM for Si designed to enable accurate description of phase transformations in Si. Considerations for efficient training set construction and model validation will also be discussed.

[1] M. J. Langenderfer, et al. April, 2020

[2] M. R. Armstrong, et al. Jan. 2020

[3] R. K. Lindsey, et al. Mar. 2022