Dissimilar thermal transport properties in $\kappa$-Ga$_2$O$_3$ and $\beta$-Ga$_2$O$_3$ revealed by machine-learning homogeneous nonequilibrium molecular dynamics simulations

Abstract

The lattice thermal conductivity (LTC) of Ga$_2$O$_3$ is an important property due to the challenge in the thermal management of high-power devices. We develop machine-learned neuroevolution potentials for single-crystalline $\beta$-Ga$_2$O$_3$ and $\kappa$-Ga$_2$O$_3$, and apply them to perform homogeneous nonequilibrium molecular dynamics simulations to predict their LTCs. The LTC of $\beta$-Ga$_2$O$_3$ was determined to be 10.3 $\pm$ 0.2 W/(m K), 19.9 $\pm$ 0.2 W/(m K), and 12.6 $\pm$ 0.2 W/(m K) along [100], [010], and [001], respectively, aligning with previous experimental measurements. For the first time, we predict the LTC of $\kappa$-Ga$_2$O$_3$ along [100], [010], and [001] to be 4.5 $\pm$ 0.0 W/(m K), 3.9 $\pm$ 0.0 W/(m K), and 4.0 $\pm$ 0.1 W/(m K), respectively, showing a nearly isotropic thermal transport property. The reduced LTC of $\kappa$-Ga$_2$O$_3$ versus $\beta$-Ga$_2$O$_3$ stems from its restricted low-frequency phonons up to 5 THz. Furthermore, we find that the $\beta$ phase exhibits a typical temperature dependence slightly stronger than $\sim T^{-1}$, whereas the $\kappa$ phase shows a weaker temperature dependence, ranging from $\sim T^{-0.5}$ to $\sim T^{-0.7}$.