We present a perspective on the use of diffuse basis functions for electronic structure calculations by density functional theory and wave function theory. We especially emphasize minimally augmented basis sets and calendar basis sets. We base our conclusions on our previous experience with commonly computed quantities, such as bond energies, barrier heights, electron affinities, noncovalent (van der Waals and hydrogen bond) interaction energies, and ionization potentials, on Stephens et al.'s results for optical rotation and on our own new calculations (presented here) of polarizabilities and of potential energy curves of van der Waals complexes. We emphasize the benefits of partial augmentation of the higher-zeta basis sets in preference to full augmentation at a lower ζ level. Benefits and limitations of the use of fully, partially, and minimally augmented basis sets are reviewed for different electronic structure methods and molecular properties. We have found that minimal augmentation is almost always enough for density functional theory (DFT) when applied to ionization potentials, electron affinities, atomization energies, barrier heights, and hydrogen-bond energies. For electric dipole polarizabilities, we find that augmentation beyond minimal has an average effect of 8% at the polarized triple-ζ level and 5% at the polarized quadruple-ζ level. The effects are larger for potential energy curves of van der Waals complexes. The effects are also larger for wave function theory (WFT). Even for WFT though, full augmentation is not needed for most purposes, and a level of augmentation between minimal and full is optimal for most problems. The calendar basis sets named after the months provide a convergent sequence of partially augmented basis sets that can be used for such calculations. The jun-cc-pV(T+d)Z basis set is very useful for MP2-F12 calculations of barrier heights and hydrogen bond strengths.

No abstract

Coupled-cluster (CC) methods have been extensively used as the high-level approach in quantum electronic structure theory to predict various properties of molecules when experimental results are unavailable. It is often assumed that CC methods, if they include at least up to connected-triple-excitation quasiperturbative corrections to a full treatment of single and double excitations (in particular, CCSD(T)), and a very large basis set, are more accurate than Kohn-Sham (KS) density functional theory (DFT). In the present work, we tested and compared the performance of standard CC and KS methods on bond energy calculations of 20 3d transition metal-containing diatomic molecules against the most reliable experimental data available, as collected in a database called 3dMLBE20. It is found that, although the CCSD(T) and higher levels CC methods have mean unsigned deviations from experiment that are smaller than most exchange-correlation functionals for metal-ligand bond energies of transition metals, the improvement is less than one standard deviation of the mean unsigned deviation. Furthermore, on average, almost half of the 42 exchange-correlation functionals that we tested are closer to experiment than CCSD(T) with the same extended basis set for the same molecule. The results show that, when both relativistic and core-valence correlation effects are considered, even the very high-level (expensive) CC method with single, double, triple, and perturbative quadruple cluster operators, namely, CCSDT(2)Q, averaged over 20 bond energies, gives a mean unsigned deviation (MUD(20) = 4.7 kcal/mol when one correlates only valence, 3p, and 3s electrons of transition metals and only valence electrons of ligands, or 4.6 kcal/mol when one correlates all core electrons except for 1s shells of transition metals, S, and Cl); and that is similar to some good xc functionals (e.g., B97-1 (MUD(20) = 4.5 kcal/mol) and PW6B95 (MUD(20) = 4.9 kcal/mol)) when the same basis set is used. We found that, for both coupled cluster calculations and KS calculations, the T1 diagnostics correlate the errors better than either the M diagnostics or the B1 DFT-based diagnostics. The potential use of practical standard CC methods as a benchmark theory is further confounded by the finding that CC and DFT methods usually have different signs of the error. We conclude that the available experimental data do not provide a justification for using conventional single-reference CC theory calculations to validate or test xc functionals for systems involving 3d transition metals.

The OH radical is the most important radical in combustion and in the atmosphere, and methanol is a fuel and antifreeze additive, model biofuel, and trace atmospheric constituent. These reagents are also present in interstellar space. Here we calculate the rate constants, branching ratios, and kinetic isotope effects (KIEs) of the hydrogen abstraction reaction of methanol by OH radical in a broad temperature range of 30-2000 K, covering interstellar space, the atmosphere, and combustion by using the competitive canonical unified statistical (CCUS) model in both the low-pressure and high-pressure limits and, for comparison, the pre-equilibrium model. Coupled cluster CCSD(T)-F12a theory and multi-reference CASPT2 theory were used to carry out benchmark calculations of the stationary points on the potential energy surface to select the most appropriate density functional method for direct dynamics calculations of rate constants. We find a significant effect of the anharmonicity of high-frequency modes of transition states on the low-temperature rate constant, and we show how tunneling leads to an unusual negative temperature dependence of the rate constants in the range 200 K > T > 100 K. The calculations also demonstrate the importance of the extent of stabilization of the pre-reactive complex. The capture rate for the formation of the complex is the dominant dynamical bottleneck for T < 100 K, and it leads to weak temperature dependence of the rate below 100 K in the high-pressure-limit of the CCUS model. We also report the pressure dependence of branching ratios (which are hard to measure so theory is essential) and the KIEs, and we report an unusual nonmonotonic variation of the KIE in the high-pressure limit at low temperatures.

We introduce a new database called TSG48 containing 48 transition state geometrical data (in particular, internuclear distances in transition state structures) for 16 main group reactions. The 16 reactions are the 12 reactions in the previously published DBH24 database (which includes hydrogen transfer reactions, heavy-atom transfer reactions, nucleophilic substitution reactions, and association reactions plus one unimolecular isomerization) plus four H-transfer reactions in which a hydrogen atom is abstracted by the methyl or hydroperoxyl radical from the two different positions in methanol. The data in TSG48 include data for four reactions that have previously been treated at a very high level in the literature. These data are used to test and validate methods that are affordable for the entire test suite, and the most accurate of these methods is found to be the multilevel BMC-CCSD method. The data that constitute the TSG48 database are therefore taken to consist of these very high level calculations for the four reactions where they are available and BMC-CCSD calculations for the other 12 reactions. The TSG48 database is used to assess the performance of the eight Minnesota density functionals from the M05-M08 families and 26 other high-performance and popular density functionals for locating transition state geometries. For comparison, the MP2 and QCISD wave function methods have also been tested for transition state geometries. The MC3BB and MC3MPW doubly hybrid functionals and the M08-HX and M06-2X hybrid meta-GGAs are found to have the best performance of all of the density functionals tested. M08-HX is the most highly recommended functional due to the excellent performance for all five subsets of TSG48, as well as having a lower cost when compared to doubly hybrid functionals. The mean absolute errors in transition state internuclear distances associated with breaking and forming bonds as calculated by the B2PLYP, MP2, and B3LYP methods are respectively about 2, 3, and 5 times larger than those calculated by MC3BB and M08-HX.