# Anisotropic Crystallization Kinetics and Interfacial Dynamics of Phase-Change Material Sb$\_2$S$\_3$ from Machine Learning Force Field Simulations

Source: [Anisotropic Crystallization Kinetics and Interfacial Dynamics of Phase-Change Material Sb$\_2$S$\_3$ from Machine Learning Force Field Simulations](<https://www.sota2.com/research/paper/6a0e9172f984739ed67389aa>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-05-22T05:00:05.162Z

Authors: Souvik Chakraborty, Wen-Qing Li, Yun Liu

Published: 2026-05-20T06:29:29.000Z

## Abstract

The phase-change material antimony sulfide (Sb$\_2$S$\_3$) relies on rapid and reversible phase transitions between crystalline and amorphous states, which are critical for their performance in data storage and photonics applications. In this work, a machine learning force field is developed based on the moment tensor potential approach, allowing us to understand the atomistic origin of the structural evolution and crystallization kinetics in Sb$\_2$S$\_3$ for the first time, by enabling large-scale molecular dynamics simulations (up to 7680 atoms for 40 ns). Sb$\_2$S$\_3$ shows anisotropic growth rates with the \[100\] facet exhibiting the fastest growth due to the strong Sb-S covalent bonding along its quasi-1D ribbon-like structure of its crystalline phase. The activation energy for crystal growth is found to be 0.55-0.57 eV, whereas that for diffusion is around 1.16-1.56 eV. The lower activation energy for crystal growth indicates that its heterogeneous crystallization is interface controlled rather than diffusion limited, unlike GST and GeTe with atomic attachment at the solid-liquid interface being energetically favoured over long range atomic transport. These findings provide key insights into the structural, thermodynamic, and kinetic properties of Sb$\_2$S$\_3$, paving the way for optimizing its functionality including switching speed, reliability, and energy efficiency.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2605.20785>)

[PDF](<https://arxiv.org/pdf/2605.20785.pdf>)