Toward understanding the depletion of two-level systems in ultrastable glasses

TL;DR

This paper links the depletion of two-level systems in ultrastable glasses to the reduction of quasi-localized modes, providing a formalized model and numerical estimates that align with experimental observations.

Contribution

It formalizes the relationship between TLS depletion and QLM reduction using the soft-potential model and provides numerical estimates consistent with experiments.

Findings

TLS density decreases by over 1000 times in ultrastable glasses

Both QLM density and symmetric double-well fraction decrease significantly

Estimated TLS density matches experimental values in amorphous silicon

Abstract

The density of Two-level systems (TLS) controls the low-temperature thermal properties in glasses and has been found to be almost depleted in ultrastable glasses. While this depletion of TLS is thought to have a close relationship with the dramatic decrease of quasi-localized modes (QLMs), it has yet to be clearly formalized. In this work, we argue, based on the \textit{soft-potential} model, that TLS correspond to QLMs with typical frequency $\omega_0$. The density $n_0$ of TLS is proportional to both the density of QLMs $D_L(\omega_0)$, and the fraction of symmetric double-wells $f(\omega_0)$ at $\omega_0$, i.e., $n_0 \propto D_L(\omega_0)f(\omega_0)$. We numerically estimate $\omega_0$ and $n_0$ in computer glasses at different levels of stabilities, and find that $\omega_0$ is about $5\%$ to $10\%$ of the Debye frequency. $n_0$ in ultrastable glasses is over $1000$ times smaller than that in poorly prepared glasses, with both $D_L(\omega_0)$ and $f(\omega_0)$ decreasing significantly. Remarkably, the order of magnitude of estimations for $n_0$ agrees with that found in experiments in amorphous silicon. Our study paves the way to understanding the depletion of TLS through the rarefaction of QLMs.