59th Annual Report on Research 2014

This project concerns the development of new massively parallel coupled-cluster (CC) implementations to accurately predict electronic spin resonance (ESR) tensors in large radicals. Free radicals play a crucial role in many biological and industrial processes, particularly in petroleum industry. ESR spectroscopy is a technique that can uniquely study the nature of free radicals alone and also in their cooperative effect during chemical reactions. One of the difficulties with ESR spectra is that they are far more complex to interpret than those from other widely used spectroscopic techniques (e.g. infrared spectroscopy). Therefore, being able to simulate ESR spectra theoretically is an invaluable and indispensable component in their analysis. However, in order to be useful for such analysis, the employed theory must be absolutely predictive in the sense that it can be reliably used as a complement to experiments. Density functional theory (DFT) models have been widely applied to ESR simulations for large systems, but they have been proven to be unreliable and fraught with many difficulties. On the other hand, more accurate wavefunction-based methods with predictive quality remain limited to small benchmark systems. In particular, CC methods are known to provide the most accurate results for nearly all the electronic structure problems including ESR spectra. In fact, one of the CC models, CC with single, double and perturbative triple excitations [CCSD(T)], is considered as the gold standard for electronic structure calculations. However, the high computational cost of the CC methods has precluded their routine application to large systems. To overcome this serious limitation, we are developing massively parallel CC implementations to predict ESR properties. Those implementations follow a new domain specific software paradigm developed by the Bartlett research group for its public domain code ACES III. This paradigm includes the super instruction processor (SIP) and the super instruction architecture language (SIAL) to run and develop efficient parallel codes. In that framework, we are further extending the massively parallel CC implementations of ACES III to systematically predict the three main ESR tensors in large radicals, namely, the A-, g-, and d-tensors.

During the second reporting period of this grant, we published the first article in our ongoing series of publications reporting our massively parallel CC implementations to predict ESR tensors; this whole series of papers is being published in the Journal of Chemical Physics. This first article reported our completed implementation of CC capabilities in ACES III to predict A-tensors and their application to calculate A-tensor’s isotropic hyperfine coupling constants in large radicals. These calculations were performed on 38 neutral, cationic, and anionic radicals that include the B, O, Be, F, H, C, Cl, S, N, P, and Zn nuclei. Some of the large radicals featured in this article include diethylaminyl radical, benzyl radical, 1,3,2-benzodithiazolyl radical, phenylaminyl radical, cyclo-hexyl radical, aniline cation radical, 4-nitroaniline cation radical, p-phenylenediamine cation radical, N,N-dimethylamino-p-phenylenediamine cation radical , N,N-Dimethyl-4-nitroaniline cation radical, 1-adamantyl radical , and Zn-porphycene anion radical. The published parallel calculations were conducted at the Hartree-Fock (HF), second-order many-body perturbation theory [MBPT(2)], CC with single and double excitations (CCSD), and CCSD(T) levels using various large atomic basis sets. HF results consistently overestimated isotropic hyperfine coupling constants. However, inclusion of electron correlation effects in the simplest way via MBPT(2) provided significant improvements in the predictions, but not without occasional failures. In contrast, CCSD and CCSD(T) results were consistently in very good agreement with experimental results. These reported calculations for the above radicals with up to 35 atoms and with up to 925 basis functions are among the largest applications of the CC theory in general and constitute the largest CC prediction of ESR spectra to date. After finishing the implementation of A-tensor capabilities, we completely implemented a new general linear-response-CC (LR-CC) parallel module to evaluate any type of first-order (zero-response) and second-order (linear-response) properties (i.e. properties that are from the first- and second-order derivatives of the energy with respect to the strength parameters, respectively). The implementation of this LR-CC parallel module with the SIAL language in ACES III is the most challenging and time-consuming task in this project and has been successfully accomplished. The developed LR-CC parallel module is the crucial component in this project because it allows the calculation of the two remaining ESR tensors: the g- and D-tensor. However, this LR-CC parallel module is general enough to calculate any type of properties and not only ESR tensors. The first application of the LR-CC parallel module is focusing on the calculation of dipole polarizabilities in large molecules at the CCSD level. Molecules under study include quinoline, isoquinoline, benzonitrile, nitrobenzene, acenaphthene, benzanthracene, fluorene, and thiophene oligomers. The predicted CCSD polarizabilities compare well with available results from experiments and alternative theoretical methods. This current application of the LR-CC parallel module to dipole polarizabilities constitutes the largest application of the CC theory to this type of property to date. This dipole polarizabilities’ study will be completed and sent for publication very soon. In addition, the LR-CC parallel module is prepared for the calculation of the two remaining ESR tensors: the g- and D-tensors, in large radicals. Calculations of these properties and their publication in the remaining articles in our ongoing series of publications on ESR tensors will be completed during the third, extended year of this project. Finally, to highlight the capabilities of the massively parallel implementations of closed- and open-shell CC methods in ACES III, the photoelectron spectrum of the Sobolewski and Domcke’s chlorophyll-imidazole-benzoquinone model complex was calculated with the equation of motion CC (EOM–CCSD) method. The calculated spectrum results augmented other recently published results with the approximate resolution of identity CC singles and doubles (RI-CC2) method. In addition, the structure of this supramolecule was calculated at the MBPT(2) level with two different basis sets and compared with its counterpart at the DFT level. This study has been published in Molecular Physics.