Coupled cluster theory through quasiperturbative triple excitations [CCSD(T)] was used with large correlation consistent basis sets to obtain optimized structures, harmonic vibrational frequencies and atomization energies for 37 molecules from the G2/97 test set. In some cases, it proved possible to include the triple excitations iteratively via CCSDT. Use of various correlation consistent basis set sequences facilitated estimation of frozen core energies in the complete basis set limit. Tight d functions were added for all second row atoms in order to improve the basis set convergence properties. Core/valence correlation corrections were obtained from all electron CCSD(T)/cc-pCVQZ calculations. Scalar relativistic contributions to the atomization energy were obtained from configuration interaction mass-velocity/one-electron Darwin calculations and CCSD(T) Douglas-Kroll-Hess calculations. By combining results from the present work with previously reported findings, a total of 114 comparisons with reliable experimental data for molecular atomization energies were possible. A statistical analysis of the level of agreement with experiment was performed, leading to a mean absolute deviation of 0.8 kcal/mol and maximum absolute error of -4.4 kcal/mol. This represents the most thorough study to date of the reliability of a composite approach to computational thermochemistry based on coupled cluster theory. The approach avoids the use of additivity approximations to estimate the complete basis set limit and does not include empirical corrections to the electronic energy. Results from three parameterized methods (G2, G3, and CBS-Q) for the same set of molecules are compared to the coupled cluster results. (C) 2001 American Institute of Physics.
通过准微扰三重激发的耦合簇理论[CCSD(T)]与大的相关一致性基组一起被用于获取G2/97测试集中37个分子的优化结构、谐振动频率和原子化能。在某些情况下,通过CCSDT迭代地包含三重激发被证明是可能的。各种相关一致性基组序列的使用有助于在完全基组极限下估计冻结核能。为所有第二周期原子添加了紧的d函数以改善基组收敛性质。从全电子CCSD(T)/cc - pCVQZ计算中获得了核/价相关校正。从组态相互作用质速/单电子达尔文计算以及CCSD(T)道格拉斯 - 克罗尔 - 赫斯计算中获得了对原子化能的标量相对论贡献。通过将本工作的结果与先前报道的发现相结合,与分子原子化能的可靠实验数据总共进行了114次比较。对与实验的一致性水平进行了统计分析,得出平均绝对偏差为0.8千卡/摩尔,最大绝对误差为 - 4.4千卡/摩尔。这代表了迄今为止对基于耦合簇理论的计算热化学复合方法可靠性的最全面研究。该方法避免使用加和近似来估计完全基组极限,并且不包括对电子能的经验校正。将针对同一组分子的三种参数化方法(G2、G3和CBS - Q)的结果与耦合簇结果进行了比较。(C)2001美国物理学会