Accuracy of Several Wave Function and Density Functional Theory Methods for Description of Noncovalent Interaction of Saturated and Unsaturated Hydrocarbon Dimers.

The proper description of noncovalent complexes is a notoriously difficult problem, especially for complexes dominated by the dispersion energy. Accurate and reliable results can be obtained using computationally demanding methods such as the coupled clusters with iterative treatment of single and double excitations and perturbative triples correction (CCSD(T)), close to the complete basis set (CBS) limit. The sizes of the noncovalent complexes of interest, however, often exceed the computational capability of available computer facilities and software. Computationally efficient yet accurate and reliable theoretical methods are highly desired. In this work, we assembled a small test set of noncovalent complexes of un/saturated a/cyclic hydrocarbon (HC) dimers in order to inspect the accuracy and reliability of several routinely used low-order scaling wave function (WFT) and density functional theory (DFT) methods. The test set comprises dispersion dominated complexes of two different monomer types, saturated and unsaturated. The unsaturated systems are relatively well populated in one of the most popular training data sets for noncovalent complexes, the S22 set of Jurečka et al. The opposite is true for saturated systems, for which rather poor performance of "approximate" methods has been observed. From the results shown is this work, it is clear that unsaturated, e.g., π···π stacked, covalent complexes are described more accurately on average. With the exception of a few "balanced methods", such as MP2C, MP2.5, SCS-/SCS(MI)-CCSD, or DFT-D3 with the TPSS and PBE functionals, a simultaneous description of saturated and unsaturated HCs introduces serious errors (i.e., more than 1 kcal/mol).