DOI: 10.1002/cphc.201200469 Assessing the Accuracy of SAPT(DFT) Interaction Energies by Comparison with Experimentally Derived Noble Gas Potentials and Molecular Crystal Lattice Energies Andrew J. Bordner* [a] 1. Introduction Non-covalent molecular interactions are important for a wide range of chemical phenomena. In particular, non-covalent in- teractions are responsible for the structures and properties of molecular crystals and biological macromolecules. Computa- tional modeling of these systems can guide experiments. How- ever, because of their size, they are not directly amenable to high-level quantum chemistry methods. Instead, they are usu- ally modeled using approximate energy functions called force fields. An appealing approach to deriving such force fields is to fit them to quantum chemistry calculations for the molecule of interest, in the case of crystals, or a small molecule analog of subunits (amino acids or nucleic acids), in the case of biopoly- mers. A highly accurate, yet computationally feasible quantum chemistry method is needed in order to perform these calcula- tions. The density functional version of symmetry-adapted per- turbation theory, or SAPT(DFT), [1–3] is a promising candidate for such a method. The accuracy of SAPT(DFT), or indeed any quantum chemis- try method, in calculating non-covalent interactions can be as- sessed using two approaches. One approach is to compare SAPT(DFT) interaction energies with supramolecular calcula- tions using benchmark wavefunction methods, typically CCSD(T) extrapolated to the complete basis set (CBS) limit. CCSD(T) is often chosen for comparison because it is arguably the most accurate yet computationally feasible method for small systems. One such study [4] found good agreement be- tween CCSD(T) and SAPT(DFT) interaction energy calculations for the benzene dimer, which is dominated by dispersion inter- actions. SAPT(DFT) was also compared with CCSD(T)/CBS re- sults for argon, krypton and benzene dimers. [5] Another study [6] compared interactions energies for the S22 set of dimers con- formations using the closely related DFT-SAPT method [7–10] with CCSD(T)/CBS benchmark values. The overall RMSD with the aug-cc-pVTZ basis set was 2.3 kJ mol 1 . Importantly, the errors for hydrogen-bonded (2.2 kJ mol 1 ) and dispersion- bonded (2.3 kJ mol 1 ) complexes were of comparable magni- tude, indicating that DFT-SAPT could accurately reproduce both types of interactions. The other main approach for assess- ing accuracy is by comparison with experimental data. Podesz- wa and Szalewicz [5] found close agreement between the SAPT- (DFT)-calculated potentials for argon and krypton dimers and empirical potentials. In two other studies by the same group, 6-dimensional potentials were fit to SAPT(DFT) calculations for benzene [4] and water [11] dimers and compared with second virial coefficients and other experimentally determined quanti- ties. In addition SAPT(DFT)-derived force fields have been used to predict crystal structures and properties for a number of or- ganic compounds including RDX, [12, 13] HMX, [14] C 6 Br 2 CIFH 2 , [15] and FOX-7. [16] In this study, we followed the latter approach and per- formed a comprehensive comparison of SAPT(DFT) calculations with experimental data. We examine the interaction potentials of six different noble gas dimers as well as crystal lattice ener- gies for five different small molecules in comparison with SAPT(DFT) and wavefunction methods [MP2 and CCSD(T)]. These two physical properties were chosen since they permit a relatively direct comparison with experimental values, unlike for example, liquid properties that require thermodynamic averaging. Because noble gas atoms interact through weak dis- persion forces, those results provide an indication of how well different methods account for these interactions. It is well known that dispersion interactions are poorly reproduced by standard DFT functionals [17–19] and wavefunction methods with The density functional version of symmetry-adapted perturba- tion theory, SAPT(DFT), is a computationally efficient method for calculating intermolecular interaction energies. We evaluate its accuracy by comparison with experimentally determined noble gas interaction potentials and sublimation enthalpies, most of which have not been previously calculated using this method. In order to compare the results with wavefunction methods, we also calculate these quantities using MP2 and, for noble gas dimers, using CCSD(T). For the crystal lattice energy calculations, we include corrections to the dispersion, electro- static, and induction energies that account for the finite inter- action distance cutoff and higher-order induction contribu- tions. Overall, the energy values extrapolated to the complete basis set limit show that SAPT(DFT) achieves significantly better agreement with experiment than MP2. [a] Dr. A. J. Bordner Mayo Clinic 13400 East Shea Boulevard Scottsdale, AZ 85259 (USA) E-mail : Bordner.Andrew@mayo.edu ChemPhysChem 2012, 13, 3981 – 3988  2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim 3981