Starting from the amorphous ground state: linking landscape thermodynamics to slow dynamics and crossover
Abstract
A microscopic understanding of low-temperature thermodynamics and its relation to dynamical features such as a fragile-to-strong crossover (FSC) remains a central challenge in glass physics. Using swap Monte Carlo combined with a full potential-energy-landscape (PEL) analysis of a non-network-forming model, we obtain equilibrium data deep into the glassy regime and identify a finite system size that simultaneously reproduces bulk behaviour for and allows complete sampling of the PEL down to its lowest-energy amorphous states. This enables the direct computation of the configurational entropy over the full temperature range of the finite system without relying on liquid-state thermodynamic integration. We find a pronounced depletion of low-energy states relative to the Gaussian regime of the PEL, which governs the low-temperature curvature of the configurational entropy. Numerically, the apparent activation energy of the diffusivity closely follows the temperature dependence of the mean inherent structure energy and exhibits a gradual crossover towards Arrhenius-like behaviour. This correlation is consistent with a trap-model description of the PEL, in which the FSC emerges naturally as a consequence of the depletion of low-energy states and thus of the lower bound of the PEL. We further argue, as illustrated analytically for a simple binomial model of the PEL, that the observability of a FSC depends on whether the depletion regime is reached within the accessible temperature window.
Keywords
Cite
@article{arxiv.2307.10143,
title = {Starting from the amorphous ground state: linking landscape thermodynamics to slow dynamics and crossover},
author = {Anshul D. S. Parmar and Simon G. Kellers and Andreas Heuer},
journal= {arXiv preprint arXiv:2307.10143},
year = {2026}
}
Comments
14 pages with 9 figures and additional details are provided in supplementary material. Accepted in The Journal of Chemical Physics (2026)