A computational approach to spectroscopy of molecular systems : modeling, prediction, and design

Abstract

This thesis describes a series of approaches for modeling spectroscopy of molecular systems in aqueous environments, focusing on proton transfer, water dynamics, and hydrogen bonding interactions. The spectroscopy motivating this work ranges from nuclear to vibrational to electronic, spanning from 106 to 1015Hz. The work in this thesis focuses on connecting these spectroscopic measurements directly to the underlying molecular structure through a variety of computational methods. After a discussion of the properties of hydrogen bonds and strongly hydrogen bonded systems, I present a phenomenological approach for modeling linear and nonlinear infrared spectroscopy in condensed phase chemical systems, focusing on applications to strongly hydrogen bonded complexes. In this I also derive and demonstrate the application of a Langevin-like Brownian oscillator model for the bath in computational spectroscopy, utilizing the language of open quantum systems along with the semiclassical approximation for the linear and nonlinear response functions to numerically calculate nonlinear spectra . With this we can examine phenomena previously difficult with other methods, including non-Gaussian dynamics, correlated motions, highly anharmonic potentials, proton transfer, and complex system-bath relationships. Next I describe a design problem reliant on water dynamics and hydrogen bonding: improving and tuning the water enhancement properties of MRI contrast agents. This work focuses specifically on a new ligand architecture with promising modular, tunable synthetic properties. Motivated by the fundamental equations governing relaxivity enhancement, this work proposes systems for which it is possible to improve and tune the molecular rotational timescale, fast water motion, and coordinating water geometry to overcome fundamental limitations in currently available contrast agents. Lastly, this thesis discusses a method of feature selection that works to identify key molecular variables important in influencing the absorption profiles of fluorescent proteins, utilizing machine learning on spectral clusters built from an ensemble of ground state dynamics trajectories. Using the example of green fluorescent protein, I show that this new feature selection protocol can highlight important interactions in the native structure that can help inform rational design of fluorescent proteins.