Abstract Accurate evaluation of the thermal conductivity of a material can be a challenging task from both experimental and theoretical points of view. In particular for the nanostructured materials, the experimental measurement of thermal conductivity is associated with diverse sources of uncertainty. As a viable alternative to experiment, the combination of density functional theory (DFT) simulations and the solution of Boltzmann transport equation is currently considered as the most trusted approach to examine thermal conductivity. The main bottleneck of the aforementioned method is to acquire the anharmonic interatomic force constants using the computationally demanding DFT calculations. In this work we propose a substantially accelerated approach for the evaluation of anharmonic interatomic force constants via employing machine-learning interatomic potentials (MLIPs) trained over short ab initio molecular dynamics trajectories. The remarkable accuracy of the proposed accelerated method is confirmed by comparing the estimated thermal conductivities of several bulk and two-dimensional materials with those computed by the full-DFT approach. The MLIP-based method proposed in this study can be employed as a standard tool, which would substantially accelerate and facilitate the estimation of lattice thermal conductivity in comparison with the commonly used full-DFT solution.

中文翻译：

---

通过机器学习原子间势加速热导率的第一性原理估计：MTP/ShengBTE 解决方案

摘要 从实验和理论的角度来看，准确评估材料的热导率是一项具有挑战性的任务。特别是对于纳米结构材料，热导率的实验测量与各种不确定性来源有关。作为实验的可行替代方案，密度泛函理论 (DFT) 模拟和玻尔兹曼输运方程解的组合目前被认为是检查热导率的最可靠方法。上述方法的主要瓶颈是使用计算要求高的 DFT 计算来获取非谐原子间力常数。在这项工作中，我们提出了一种显着加速的方法，通过采用在短的 ab initio 分子动力学轨迹上训练的机器学习原子间势 (MLIP) 来评估非谐原子间力常数。通过比较几种块状和二维材料的估计热导率与全 DFT 方法计算的热导率，证实了所提出的加速方法的显着准确性。本研究中提出的基于 MLIP 的方法可以用作标准工具，与常用的全 DFT 解决方案相比，这将大大加速和促进晶格热导率的估计。通过比较几种块状和二维材料的估计热导率与全 DFT 方法计算的热导率，证实了所提出的加速方法的显着准确性。本研究中提出的基于 MLIP 的方法可以用作标准工具，与常用的全 DFT 解决方案相比，这将大大加速和促进晶格热导率的估计。通过比较几种块状和二维材料的估计热导率与全 DFT 方法计算的热导率，证实了所提出的加速方法的显着准确性。本研究中提出的基于 MLIP 的方法可以用作标准工具，与常用的全 DFT 解决方案相比，这将大大加速和促进晶格热导率的估计。