Department of Chemistry, University of Zurich , Winterthurerstrasse 190, 8057 Zurich, Switzerland.
School of Pharmaceutical Science and Technology, University of Tianjin , 92 Weijin Road, Nankai District, Tianjin 3000072, P. R. China.
J Chem Theory Comput. 2017 Jun 13;13(6):2650-2666. doi: 10.1021/acs.jctc.7b00220. Epub 2017 May 24.
The implementation of 300 combinations of generalized gradient approximation/local density approximation exchange-correlation dispersion-corrected spin-component-scaled double-hybrid (DSD) density functional theory (DFT) methods has been carried out and the performance assessed against several DFT and post-Hartree-Fock methods, enabling further advancements toward the long-standing challenge of accurate prediction of interaction energies and associated properties. The resulting framework is flexible and has been further extended to include the resolution of identity (RI) approximation for solving the critical four-center two-electron repulsion integrals in the basis of the Kohn-Sham orbitals for cost effectiveness. To evaluate the performance of this set of new cost-effective methods, denoted as RI-DSD-DFTs, seven validation data sets were designed to cover a broad range of non-covalent interactions with characteristic stabilizing contributions. Inclusion of the perturbative treatment of correlation effects is shown to significantly improve the description of weak interactions. The set of DSD-DFTs provide interaction energies with root-mean-square deviations and mean absolute errors within 0.5 kcal/mol. The cost-effective RI-DSD-DFT counterparts deviate by less than 0.18 kcal/mol on average with only 2% of the computational cost.
已经实现了 300 种广义梯度近似/局域密度近似交换相关色散校正自旋分量缩放双杂交(DSD)密度泛函理论(DFT)方法的组合,并对几种 DFT 和 Hartree-Fock 后方法进行了评估,从而进一步推进了准确预测相互作用能和相关性质的长期挑战。所得到的框架是灵活的,并进一步扩展为包括分辨率身份(RI)近似,以解决基于 Kohn-Sham 轨道的关键四中心二电子排斥积分的成本效益。为了评估这组新的具有成本效益的方法的性能,称为 RI-DSD-DFTs,设计了七个验证数据集,以涵盖具有特征稳定贡献的广泛非共价相互作用。包括相关效应的微扰处理被证明可以显著改善弱相互作用的描述。DSD-DFT 组提供的相互作用能的均方根偏差和平均绝对误差在 0.5 kcal/mol 以内。具有成本效益的 RI-DSD-DFT 对应物的平均偏差小于 0.18 kcal/mol,仅消耗 2%的计算成本。