Edward G. Hohenstein

Psi4: an open-source ab initio electronic structure program

The Psi4 program is a new approach to modern quantum chemistry, encompassing Hartree–Fock and density‐functional theory to configuration interaction and coupled cluster. The program is written entirely in C++ and relies on a new infrastructure that has been designed to permit high‐efficiency computations of both standard and emerging electronic structure methods on conventional and high‐performance parallel computer architectures. Psi4 offers flexible user input built on the Python scripting language that enables both new and experienced users to make full use of the programs capabilities, and even to implement new functionality with moderate effort. To maximize its impact and usefulness, Psi4 is available through an open‐source license to the entire scientific community.

An Assessment of Theoretical Methods for Nonbonded Interactions: Comparison to Complete Basis Set Limit Coupled-Cluster Potential Energy Curves for the Benzene Dimer, the Methane Dimer, Benzene−Methane, and Benzene−H2S†

Large, correlation-consistent basis sets have been used to very closely approximate the coupled-cluster singles, doubles, and perturbative triples [CCSD(T)] complete basis set potential energy curves of several prototype nonbonded complexes, the sandwich, T-shaped, and parallel-displaced benzene dimers, the methane-benzene complex, the H2S-benzene complex, and the methane dimer. These benchmark potential energy curves are used to assess the performance of several methods for nonbonded interactions, including various spin-component-scaled second-order perturbation theory (SCS-MP2) methods, the spin-component-scaled coupled-cluster singles and doubles method (SCS-CCSD), density functional theory empirically corrected for dispersion (DFT-D), and the meta-generalized-gradient approximation functionals M05-2X and M06-2X. These approaches generally provide good results for the test set, with the SCS methods being somewhat more robust. M05-2X underbinds for the test cases considered, while the performances of DFT-D and M06-2X are similar. Density fitting, dual basis, and local correlation approximations all introduce only small errors in the interaction energies but can speed up the computations significantly, particulary when used in combination.

Assessment of the Performance of DFT and DFT-D Methods for Describing Distance Dependence of Hydrogen-Bonded Interactions.

Noncovalent interactions such as hydrogen bonds, van der Waals forces, and π-π interactions play important roles influencing the structure, stability, and dynamic properties of biomolecules including DNA and RNA base pairs. In an effort to better understand the fundamental physics of hydrogen bonding (H-bonding), we investigate the distance dependence of interaction energies in the prototype bimolecular complexes of formic acid, formamide, and formamidine. Potential energy curves along the H-bonding dissociation coordinate are examined both by establishing reference CCSD(T) interaction energies extrapolated to the complete basis set limit and by assessing the performance of the density functional methods B3LYP, PBE, PBE0, B970, PB86, M05-2X, and M06-2X and empirical dispersion corrected methods B3LYP-D3, PBE-D3, PBE0-D3, B970-D2, BP86-D3, and ωB97X-D, with basis sets 6-311++G(3df,3pd), aug-cc-pVDZ, and aug-cc-pVTZ. Although H-bonding interactions are dominated by electrostatics, it is necessary to properly account for dispersion interactions to obtain accurate energetics. In order to quantitatively probe the nature of hydrogen bonding interactions as a function of distance, we decompose the interaction energy curves into physically meaningful components with symmetry-adapted perturbation theory (SAPT). The SAPT results confirm that the contribution of dispersion and induction are significant at and near equilibrium, although electrostatics dominate. Among the DFT/DFT-D techniques, the best overall results are obtained utilizing counterpoise-corrected ωB97X-D with the aug-cc-pVDZ basis set.

Basis set consistent revision of the S22 test set of noncovalent interaction energies.

The S22 test set of interaction energies for small model complexes [Phys. Chem. Chem. Phys. 8, 1985 (2006)] has been very valuable for benchmarking new and existing methods for noncovalent interactions. However, the basis sets utilized to compute the CCSD(T) interaction energies for some of the dimers are insufficient to obtain converged results. Here we consistently extrapolate all CCSD(T)/complete basis set (CBS) interaction energies using larger basis sets for the CCSD(T) component of the computation. The revised values, which we designate S22A, represent the most accurate results to date for this set of dimers. The new values appear to be within a few hundredths of 1 kcal mol(-1) of the true CCSD(T)/CBS limit at the given geometries, but the former S22 values are off by as much as 0.6 kcal mol(-1) compared to the revised values. Because some of the most promising methods for noncovalent interactions are already achieving this level of agreement (or better) compared to the S22 data, more accurate benchmark values would clearly be helpful. The MP2, SCS-MP2, SCS-CCSD, SCS(MI)-MP2, and B2PLYP-D methods have been tested against the more accurate benchmark set. The B2PLYP-D method outperforms all other methods tested here, with a mean average deviation of only 0.12 kcal mol(-1). However, the consistent, slight underestimation of the interaction energies computed by the SCS-CCSD method (an overall mean absolute deviation and mean deviation of 0.24 and -0.23 kcal mol(-1), respectively) suggests that the SCS-CCSD method has the potential to become even more accurate with a reoptimization of its parameters for noncovalent interactions.

Wavefunction methods for noncovalent interactions

Noncovalent interactions remain poorly understood despite their importance to supramolecular chemistry, biochemistry, and materials science. They are an ideal target for theoretical study, where interactions of interest can be probed directly, free from competing secondary interactions. However, the most popular tools of computational chemistry are not particularly reliable for noncovalent interactions. Here we review recent works in wavefunction‐based quantum chemistry techniques aimed at greater accuracy and faster computations for these systems. We describe recent developments in high‐accuracy benchmarks, a variety of recent wavefunction methods with promise for noncovalent interactions, various approximations to speed up these methods, and recent advances in wavefunction‐based symmetry‐adapted perturbation theory, which provides not only interaction energies but also their decomposition into physically meaningful components. Together, these advances are currently extending robust, accurate computations of noncovalent interactions from systems with around one dozen heavy atoms up to systems with several dozens of heavy atoms.

Density fitting and Cholesky decomposition approximations in symmetry-adapted perturbation theory: Implementation and application to probe the nature of π-π interactions in linear acenes

Density fitting (DF) approximations have been used to increase the efficiency of several quantum mechanical methods. In this work, we apply DF and a related approach, Cholesky decomposition (CD), to wave function-based symmetry-adapted perturbation theory (SAPT). We also test the one-center approximation to the Cholesky decomposition. The DF and CD approximations lead to a dramatic improvement in the overall computational cost of SAPT, while introducing negligible errors. For typical target accuracies, the Cholesky basis needed is noticeably smaller than the DF basis (although the cost of constructing the Cholesky vectors is slightly greater than that of constructing the three-index DF integrals). The SAPT program developed in this work is applied to the interactions between acenes previously studied by Grimme [Angew. Chem., Int. Ed. 47, 3430 (2008)], expanding the cases studied by adding the pentacene dimer. The SAPT decomposition of the acene interactions provides a more realistic picture of the interac...

Improvement of the coupled-cluster singles and doubles method via scaling same-and opposite-spin components of the double excitation correlation energy

There has been much interest in cost-free improvements to second-order Møller-Plesset perturbation theory (MP2) via scaling the same- and opposite-spin components of the correlation energy (spin-component scaled MP2). By scaling the same- and opposite-spin components of the double excitation correlation energy from the coupled-cluster of single and double excitations (CCSD) method, similar improvements can be achieved. Optimized for a set of 48 reaction energies, scaling factors were determined to be 1.13 and 1.27 for the same- and opposite-spin components, respectively. Preliminary results suggest that the spin-component scaled CCSD (SCS-CCSD) method will outperform all MP2 type methods considered for describing intermolecular interactions. Potential energy curves computed with the SCS-CCSD method for the sandwich benzene dimer and methane dimer reproduce the benchmark CCSD(T) potential curves with errors of only a few hundredths of 1 kcal mol(-1) for the minima. The performance of the SCS-CCSD method suggests that it is a reliable, lower cost alternative to the CCSD(T) method.

Density fitting of intramonomer correlation effects in symmetry-adapted perturbation theory

Symmetry-adapted perturbation theory (SAPT) offers insight into the nature of intermolecular interactions. In addition, accurate energies can be obtained from the wave function-based variant of SAPT provided that intramonomer electron correlation effects are included. We apply density-fitting (DF) approximations to the intramonomer correlation corrections in SAPT. The introduction of this approximation leads to an improvement in the computational cost of SAPT by reducing the scaling of certain SAPT terms, reducing the amount of disk I/O, and avoiding the explicit computation of certain types of MO integrals. We have implemented all the intramonomer correlation corrections to SAPT through second-order under the DF approximation. Additionally, leading third-order terms are also implemented. The accuracy of this truncation of SAPT is tested against the S22 test set of Hobza and co-workers [Phys. Chem. Chem. Phys. 8, 1985 (2006)]. When the intramonomer corrections to dispersion are included in SAPT, a mean absolute deviation of 0.3-0.4 kcal mol(-1) is observed for the S22 test set when using an aug-cc-pVDZ basis. The computations on the adenine-thymine complexes in the S22 test set with an aug-cc-pVDZ basis represent the largest SAPT computations to date that include this degree of intramonomer correlation. Computations of this size can now be performed routinely with our newly developed DF-SAPT program.

Origin of the surprising enhancement of electrostatic energies by electron-donating substituents in substituted sandwich benzene dimers.

A recent study of substituted face-to-face benzene dimers by Lewis and co-workers [J. Am. Chem. Soc. 2011, 133, 3854-3862] indicated a surprising enhancement of electrostatic interactions for both electron-withdrawing and electron-donating substituents. Here we demonstrate that charge penetration (an attractive electrostatic interaction arising from the overlap of the electron densities on the two monomers) is the cause of this counterintuitive effect. These charge penetration effects are significant at typical π-π interaction distances, and they are not easily described by multipole models. A simple measure of a substituents electron-donating or electron-withdrawing character, such as the Hammett parameter σ(m), is unlikely to capture subtle charge penetration effects. Indeed, correlation of the relative total energies or relative electrostatic energies with ∑σ(m) breaks down for multiply substituted face-to-face benzene dimers.

Tensor hypercontraction density fitting. I. Quartic scaling second- and third-order Møller-Plesset perturbation theory

Many approximations have been developed to help deal with the O(N(4)) growth of the electron repulsion integral (ERI) tensor, where N is the number of one-electron basis functions used to represent the electronic wavefunction. Of these, the density fitting (DF) approximation is currently the most widely used despite the fact that it is often incapable of altering the underlying scaling of computational effort with respect to molecular size. We present a method for exploiting sparsity in three-center overlap integrals through tensor decomposition to obtain a low-rank approximation to density fitting (tensor hypercontraction density fitting or THC-DF). This new approximation reduces the 4th-order ERI tensor to a product of five matrices, simultaneously reducing the storage requirement as well as increasing the flexibility to regroup terms and reduce scaling behavior. As an example, we demonstrate such a scaling reduction for second- and third-order perturbation theory (MP2 and MP3), showing that both can be carried out in O(N(4)) operations. This should be compared to the usual scaling behavior of O(N(5)) and O(N(6)) for MP2 and MP3, respectively. The THC-DF technique can also be applied to other methods in electronic structure theory, such as coupled-cluster and configuration interaction, promising significant gains in computational efficiency and storage reduction.