地核温压条件下铁的深度学习分子动力学模拟

Deep learning molecular dynamics simulation of iron under the temperature and pressure of the Earth's core

铁(Fe)是地核中的主要元素,研究铁在地核温压条件下的状态方程及热力学性质对理解地核至关重要。为高效地构建出准确和稳健的势能面模型,本文使用第一性原理数据为初始数据,利用深度势能生成器对地核内温压下构型进行采样,通过深度神经网络训练出深度学习势,完成对大体系超胞的深度学习分子动力学高精度模拟。所构建的地核温压下铁的深度势能,与第一性原理计算结果相吻合。利用该势能,可实现内核温压下超过10000原子体系的高精度分子动力学模拟,计算出密堆六方(hcp)和体心立方(bcc)结构铁的密度和状态方程。本文利用深度势能方法提高了计算效率、保证了计算精度同时可利用于大规模体系、多晶体系,为更接近地球内部条件物质状态、性质的模拟打下基础。

Iron is the main element in the Earth's core.Studying the state equation and thermodynamic properties of iron is crucial to the understanding of the core.In order to efficiently construct an accurate and robust potential energy surface model,this paper uses the first-principles data as the initial data,samples the configuration of the core under the tempera-ture and pressure of the Earth's core using a deep potential energy generator,trains a deep-learning potential through the deep neural network,and completes a deep learning molecular dynamics high-precision simulation of the large system su-percell.The constructed depth potential energy of iron at the temperature and pressure of the core is in agreement with the first-nature principle calculations.Using this well trained potential,high-precision molecular dynamics simulations are re-alized for systems with more than 10000 atoms at the temperature and pressure of the Earth's core,and densities and state equations of iron with hexagonal close packed(hcp)and body-centered cubic(bcc)structures are calculated,all of which are consistent with previous research results.In summary,the depth potential method,which can be used for large-scale and polycrystalline systems,improves computational efficiency,ensures computational accuracy,and provides a ba-sis for simulating the state and properties of matter under conditions of the Earth's interior.