Development and Application of Hybrid Methods

Abstract

This thesis focuses on the development and application of hybrid methods for modeling excited state properties of large systems. The rapid growth of technology and computer processing power allows the wide application of accurate quantum mechanical (QM) methods to different areas of scientific research. However, more accurate computational methods are usually associated with higher computational cost that limits their application to small to medium sized systems. In many cases, modeling the environment at QM level is not possible, and yet completely neglecting the effect of the environment may not be a reasonable assumption. One popular approach for reducing computational cost is to use hybrid methods, which combine two or multiple computational methods to treat a single system. An area where major development is still necessary for hybrid methods is modeling excited states. For a realistic description of these phenomena, a combination of two different QM methods (QM/QM) may even be required. The challenging part of such a hybrid approach is to properly describe the mutual polarization among layers. Currently, in the field, there is no accepted standard computational protocol for a multi-layered embedding scheme that provides reliable modeling of excited state properties in complex environments. In order to design a multi-layered, fully mutually-polarizable embedding (MPE) scheme, implementing and thoroughly benchmarking the methods that act as its components is crucial. Our research aims to benchmark and implement several key methods that serve as components of an MPE, and to further expand the area of hybrid methods development for accurately modeling excited state properties. One focus of our research is the development and application of hybrid methods that use coupled cluster with single and double excitations (CCSD) and its excited state version, equation-ofmotion CCSD (EOM-CCSD), as high level. Methods based on CC theory provide consistently accurate results in gas phase. However, due to their steep computational scaling, works that focus on CCSD as part of hybrid schemes are limited, and even less has been done for excited state CCSD compared to density functional theory (DFT)/time dependent (TD)-DFT. We first benchmark EOM-CCSD combined with the polarizable continuum model (PCM) of solvation, for computing electronic excitation energies of solvated molecules. We address the PCM shortcoming of neglecting specific interactions, such as H bonding, through micro-solvated clusters inside the PCM cavity. We then combine EOM-CCSD with a classical polarizable molecular mechanics force field (MMPol) based on the induced dipole model to describe such specific interactions more efficiently. Our work also extends to QM/QM hybrid methods. We consider embedding for the ONIOM (our own n-layered integrated molecular orbital and molecular mechanics) hybrid method, and investigate the importance of including a polarizable embedding based on the induced dipole model. In addition, to simulate large regions of the UV/vis spectrum, we develop a multi-state extrapolation scheme based on the ONIOM extrapolation formula. This method overcomes the state-matching difficulty among sub-calculations in standard ONIOM, and is able to efficiently extrapolate several bands of the spectra at once. In summary, our research expands the field of hybrid methods development for excited states, by proposing and testing several approaches for the simulation of large chromophores in complex environments.