SiC-TGAP: A machine learning interatomic potential for radiation damage simulations in 3C-SiC

TL;DR

This work presents TGAP, a Gaussian Approximation Potential for 3C-SiC trained on crystalline, liquid, and amorphous configurations, including 21 defect types, to enable accurate radiation-damage simulations. By employing two-body and turboSOAP descriptors alongside the NLH repulsive potential, TGAP reproduces bulk thermomechanical properties, melting behavior, and disordered-phase characteristics with close alignment to DFT and experimental data, while improving upon classical potentials. TGAP delivers defect formation energies, interstitial stability, and vacancy migration barriers that closely track ab initio results, and it captures carbon decomposition and graphene-like clustering in liquid SiC—crucial features for modeling primary damage and damage accumulation. Overall, TGAP provides a robust, efficient tool for atomistic radiation-damage studies in cubic SiC, enabling more reliable predictions of cascade evolution and defect populations in high-temperature, high-dose environments.

Abstract

Silicon carbide (SiC) has long been a subject of study for its application in harsh environments. Existing empirical interatomic potentials for 3C-SiC show significant discrepancies in predicting the properties that are crucial in describing the evolution of defects generated in collision cascades. We present a Gaussian approximation potential model for 3C-SiC (TGAP) trained by two-body and the turboSOAP many-body descriptors. The dataset covers crystalline, liquid and amorphous phases. To accurately capture defect dynamics, twenty-one defect types have been included in the dataset. TGAP captures the experimentally observed decomposition of carbon atoms in the liquid phase at atmospheric pressure, while also accurately reproducing the radial distribution function of the high-temperature homogeneous liquid phase across a range of densities. Moreover, it predicts the melting point in very good agreement with density functional theory and experiments. The potential is equipped with the Nordlund-Lehtola-Hobler repulsive potential to capture the high repulsion of recoils in the collision cascades. TGAP provides an accurate tool for atomistic simulation of radiation damage in cubic SiC.

Figures (13)

• Figure 1: Energy-volume curve of 3C-SiC calculated with the TGAP, PBE-DFT and empirical models.
• Figure 2: Relative elastic modulus as a function of temperature obtained from MD simulations with TGAP, compared to experimental values. The MD simulations were performed on a 1000-atom cell, with a 4% lattice deformation applied. The black solid line represents $E = 460 - 0.04\,T\exp(-\frac{962}{T})$, an empirical fit to eight sets of experimentally measured elastic moduli presented in Ref. handbook. Two sets of experimental data points from Ref. handbook are shown with black markers. CVD stands for chemical vapor deposition, and PS for pressureless sintered. The dashed red line represents the $E = 425 - 0.036\,T\exp(-\frac{1237}{T})$, fitted to the TGAP-MD results.
• Figure 3: Phonon dispersion bands of 3C-SiC along the $\Gamma \to \mathrm{X} \to \Gamma \to \mathrm{L}$ pathway through reciprocal space, as calculated with TGAP, DFT, and empirical Tersoff-ZBL, Vashishta, MEAM meam, EDIP edip, and GW-ZBL models. The calculations were carried out with the Phonopy code, where a 128-atom primitive supercell generated with corresponding equilibrium lattice constant for each model was used.
• Figure 4: Linear thermal expansion and heat capacity of 3C-SiC calculated with QHA method as implemented in Phonopy. Phonopy calculations were performed with up to 5% lattice distortion. The experimental values are taken from Ref. handbook.
• Figure 5: Potential energy scan for isolated Si--Si, Si--C, and C--C dimers. DFT reference represents our calculations. DMOL data shows all-electron DFT calculations provided in Ref. NLH. TGAP has been trained with the $V_\mathrm{NLH}$ repulsive potential which is fitted to DMOL data.