Preview

PDF, English - main document Download (29MB) | Terms of use

Abstract

Quantum chemical simulations of molecular properties are crucial to obtain in-depth insight into a multitude of chemical and biological phenomena. In particular for investigating light-driven systems, modeling of electronic excitations by computational means is indispensable for supporting, complementing, and extending experimental findings. The complexity in terms of electronic structure, intermolecular interactions, and dynamics of the involved molecular systems, however, pushes the limits of computational feasibility. Hybrid quantum-classical environment schemes tackle this complexity by splitting the system into a quantum region and its environment. Thus, they retain the high-level quantum chemical description for the part of interest without neglecting the pivotal effects of the environment.

In this thesis, I develop methods for modeling molecular properties in complex environments. The first half of the thesis is dedicated to new combined approaches of the polarizable embedding (PE) model and the algebraic-diagrammatic construction (ADC) scheme for the polarization propagator for computational spectroscopy simulations. I derive and implement two PE-ADC coupling schemes: The first scheme – pt-PE-ADC – uses a self-consistent PE reference state with a canonical ADC procedure and is suited for computation of electronic excitation energies including a posteriori perturbative corrections. The second scheme – LR-PE-ADC – includes direct coupling to the polarizable environment in an iterative manner, making it suitable for excited electronic states and higher-order response properties. Furthermore, I derive working equations to evaluate analytic nuclear gradients using PE-ADC. To advance the availability of the PE model in general, I implement a standalone, open-source, and hybrid Python/C++ library, called CPPE, and interface it to several freely available quantum chemical host programs. The PE-ADC schemes are implemented with adcc, a toolkit for development of ADC-based methods and combinable with several Python-driven host programs. The simple and clean design of both libraries allows for extension of existing workflows and rapid prototyping. Moreover, I implement response properties using ADC and the intermediate state representation (ISR) in a new Python library, called respondo. The synergy of all three libraries enables the user to implement new features in a straightforward manner, while maintaining usability and efficiency for practical calculations. I test the individual approaches in several benchmark calculations and case studies. For example, I find that excitation energy errors using pt-PE-ADC for microsolvated p-nitroaniline are much smaller than the intrinsic error of ADC itself. Furthermore, I investigate the charge transfer (CT) state involved in the photoprotection mechanism of the flavoprotein dodecin. In addition, I conduct the first computations of higher-order response properties with ADC and a polarizable model. In these studies, I observe that LR-PE-ADC greatly improves the accuracy of the property compared to simpler coupling schemes. I further show that corrections for electron spill-out artifacts and the physically sound evaluation of PE-ADC intensities are decisive when benchmarking against supersystem calculations. With my theoretical derivations and open-source implementations, I provide, to the best of my knowledge, the most complete and unique feature set of polarizable models combined with ADC to date.

The second half of the thesis first contains a general performance improvement of PE models. I implement a PE scheme where the classical electric fields in the environment are evaluated using the fast multipole method (FMM) instead of direct summation. Consequently, the electric field evaluations as rate-limiting step of the classical part exhibit an asymptotic linear scaling in the PE-FMM scheme, making it suitable for efficient simulations of environments with over a million polarizable sites. Next, I show algorithm details for numerically stable solution of response equations in the ADC/ISR framework, and I analyze convergence behavior of different solver algorithms. These algorithms are beneficial for efficient evaluation of PE-ADC response properties, too. I present derivations and numerical case studies of complex excited state polarizabilities which extend the ADC/ISR framework beyond ground state response properties. Then, I investigate the distortion of molecules under external forces. I develop a new electronic structure method to apply hydrostatic pressure in standard quantum chemical simulations via Gaussian potentials, called GOSTSHYP. This implicit embedding scheme directly exerts pressure on a molecule via compression of the electron density, such that it becomes possible to treat atoms and molecules and to run geometry optimizations and dynamics simulations at a pre-defined pressure. This feature set is not found in any other comparable method. I use steered molecular dynamics (SMD) simulations with quantum chemical strain analysis tools to elucidate the rupture process of rubredoxin. I prove that the extremely low rupture force does not result from hydrogen bond networks to the protein as assumed so far in the literature, but that its origin is likely more intricate. Finally, I present the design of novel photocages based on fluorene derivatives. Using an efficient computational screening protocol, I propose cyclopenta-dithiophene as scaffold, leading to the next generation of fluorene-based photocages with desirable absorption and uncaging properties.