Hybrid methods for molecular spectroscopy and reactivity
File(s)
Author(s)
Fare, Clyde
Type
Thesis
Abstract
In this thesis a variety of different studies are undertaken coupled together through ap-
plication of the hybrid method ONIOM to compute reaction energies, barrier heights and
vibrational modes of molecules. ONIOM is a means to combine multiple levels of theory such
that different regions of a molecular system are treated with different levels of accuracy
enabling computation on large molecular systems with complex environments at an accuracy
that would be impossible using a single level of theory. The price of ONIOM is a requirement
for computational chemists to specify regions of interest, and in the case of biomolecules
where forcefields are typically used to treat the environment to choose suitably parameterised
forcefield. My initial investigations focused on means to automate this process hence allow
ONIOM to be used within high throughput studies to compute properties of populations of
molecules. Toward this end I present an API allowing construction and execution of ONIOM
calculations including parameterisation for biomolecular simulation which both improves
providence in computational chemical research and enables high throughput calculations.
Following this I investigate ONIOM applied to graphitic fragments first examining a 1,3
dipolar cycloaddition then in order to address the question of how to choose model regions
within these extended delocalised systems I examine whether systematic fragmentation and
use of very cheap computational metrics can allow automatic model selection within a test
hydrogenation reaction. I then turn to an alternative means of computing the effect of the
environment: machine learning and perform a small machine study on solvation of a series
of 6095 constitutional isomers examining whether both the implicit solvation method PCM
and an explicit model of solvation making use of ONIOM can be learned use the machine
learning algorithm Kernel Ridge Regression.
Finally in the last part of my thesis I perform an experimental time resolved spectroscopic
study of the photoactive protein EOSFP examining the photo induced changes that occur both
during and after exposure to the 405nm light which drives photoconversion in this protein
and make use ONIOM to compute vibrational modes that aid assignment of vibrational
spectra.
plication of the hybrid method ONIOM to compute reaction energies, barrier heights and
vibrational modes of molecules. ONIOM is a means to combine multiple levels of theory such
that different regions of a molecular system are treated with different levels of accuracy
enabling computation on large molecular systems with complex environments at an accuracy
that would be impossible using a single level of theory. The price of ONIOM is a requirement
for computational chemists to specify regions of interest, and in the case of biomolecules
where forcefields are typically used to treat the environment to choose suitably parameterised
forcefield. My initial investigations focused on means to automate this process hence allow
ONIOM to be used within high throughput studies to compute properties of populations of
molecules. Toward this end I present an API allowing construction and execution of ONIOM
calculations including parameterisation for biomolecular simulation which both improves
providence in computational chemical research and enables high throughput calculations.
Following this I investigate ONIOM applied to graphitic fragments first examining a 1,3
dipolar cycloaddition then in order to address the question of how to choose model regions
within these extended delocalised systems I examine whether systematic fragmentation and
use of very cheap computational metrics can allow automatic model selection within a test
hydrogenation reaction. I then turn to an alternative means of computing the effect of the
environment: machine learning and perform a small machine study on solvation of a series
of 6095 constitutional isomers examining whether both the implicit solvation method PCM
and an explicit model of solvation making use of ONIOM can be learned use the machine
learning algorithm Kernel Ridge Regression.
Finally in the last part of my thesis I perform an experimental time resolved spectroscopic
study of the photoactive protein EOSFP examining the photo induced changes that occur both
during and after exposure to the 405nm light which drives photoconversion in this protein
and make use ONIOM to compute vibrational modes that aid assignment of vibrational
spectra.
Version
Open Access
Date Issued
2017-12
Date Awarded
2018-09
Copyright Statement
Attribution NoDerivatives 4.0 International Licence (CC BY-ND)
Advisor
Bearpark, Michael
van Thor, Jasper
Sponsor
Engineering and Physical Sciences Research Council
Publisher Department
Chemistry
Publisher Institution
Imperial College London
Qualification Level
Doctoral
Qualification Name
Doctor of Philosophy (PhD)