论文部分内容阅读
本文介绍一个近期发展的基于神经元网络的校正密度泛函(DFT)计算标准生成焓(?fHo)的方法,即X1se方法.首先,该方法通过B3LYP/6-311+G(d,p)优化构型和计算零点能,采用B3LYP/6-311+G(3df,2p)计算单点能量,由此获得DFT计算的?fHo;接着,将DFT计算获得的生成焓和零点能,以及分子的各种描述子输入已训练好的神经网络模型,无需额外的计算量,就可获得校正后的生成焓.与本研究组早期发展的X1和X1s方法相比,X1se新增加了环境描述子,因而进一步提高了该系列方法的应用范围和计算精度.
This paper presents a recently developed method for calculating the standard enthalpy of formation (? FHo) based on neural networks based on the density functional theory (DFT) of the X1se method.Firstly, this method uses B3LYP / 6-311 + G (d, p) Optimize the configuration and calculate the zero-point energy, calculate the single-point energy using B3LYP / 6-311 + G (3df, 2p), and then obtain the? FHo calculated by the DFT; then calculate the enthalpy of formation and the zero-point energy obtained by the DFT calculation, , All kinds of descriptors are input into the trained neural network model, and the corrected enthalpy of formation can be obtained without additional computation.Compared with the earlier developed X1 and X1s methods, X1se newly added the environment descriptor , Thus further improving the scope of application of the series of methods and calculation accuracy.