Liang Yu Hsuan, Ye Hong-Zhou, Berkelbach Timothy C
Department of Chemistry, Columbia University, New York, New York 10027, United States.
J Phys Chem Lett. 2023 Nov 23;14(46):10435-10441. doi: 10.1021/acs.jpclett.3c02411. Epub 2023 Nov 13.
Attaining kJ/mol accuracy in cohesive energy for molecular crystals is a persistent challenge in computational materials science. In this study, we evaluate second-order Møller-Plesset perturbation theory (MP2) and its spin-component scaled models for calculating cohesive energies for 23 molecular crystals (X23 data set). Using periodic boundary conditions and Brillouin zone sampling, we converge results to the thermodynamic and complete basis set limits, achieving an accuracy of about 2 kJ/mol (0.5 kcal/mol), which is rarely achieved in previous MP2 calculations for molecular crystals. When compared to experimental data, our results have a mean absolute error of 12.9 kJ/mol, comparable to Density Functional Theory with the PBE functional and TS dispersion correction. By separately scaling the opposite-spin and same-spin correlation energy components, using predetermined parameters, we reduce the mean absolute error to 9.5 kJ/mol. Further fine-tuning of these scaling parameters specifically for the X23 data set brings the mean absolute error down to 7.5 kJ/mol.
在计算材料科学中，实现分子晶体结合能的千焦每摩尔精度是一个长期存在的挑战。在本研究中，我们评估了二阶Møller-Plesset微扰理论（MP2）及其自旋分量缩放模型，用于计算23种分子晶体（X23数据集）的结合能。使用周期性边界条件和布里渊区采样，我们将结果收敛到热力学和完全基组极限，实现了约2千焦每摩尔（0.5千卡每摩尔）的精度，这在以前的分子晶体MP2计算中很少实现。与实验数据相比，我们的结果平均绝对误差为12.9千焦每摩尔，与采用PBE泛函和TS色散校正的密度泛函理论相当。通过使用预定参数分别缩放相反自旋和相同自旋相关能分量，我们将平均绝对误差降低到9.5千焦每摩尔。针对X23数据集对这些缩放参数进行进一步微调，使平均绝对误差降至7.5千焦每摩尔。
