AI 中文总结
该研究从质量输运变分原理出发，提出统计框架描述复杂固体热扩散，表明微观涨落支配非阿累尼乌斯扩散，动力学项可机器学习处理，还阐述了固溶体空位扩散在不同条件下与阿累尼乌斯行为的关系。
AI 中文摘要
从质量输运的变分原理出发，我们提出了一个广泛适用的统计框架来描述复杂固体中的热扩散。我们表明微观热力学和动力学涨落支配非阿累尼乌斯扩散，其动力学项由机器学习可处理的相对扩散贡献来描述。对于固溶体中的空位扩散，当有序效应开始变得重要时可能会偏离阿累尼乌斯行为，但在较高温度下，由于构型熵和竞争能量涨落，近似阿累尼乌斯行为可能会出现。
英文摘要
Starting from a variational principle for mass transport, we present a broadly applicable statistical framework describing thermal diffusion in complex solids. We show that microscopic thermodynamic and kinetic fluctuations govern non-Arrhenius diffusion, with kinetic terms described by machine-learnable relative diffusion contributions. For vacancy diffusion in solid solutions, deviation from Arrheniusness may occur when ordering effects start to become important but at higher temperatures, approximately Arrhenius behavior can occur, aided by configurational entropy and competing energy fluctuations.
Comments14 pages, 3 figures, 24 pages supplementary material
