Year: 2023
Type: dissertation
Abstract: This work focuses on using classical machine learning (ML) models to study the quantum dynamics \nof excitation energy transfer (EET) within strongly coupled open quantum systems relevant to light \nharvesting complexes (LHCs). Direct evidence for long-lived quantum coherence has been found to \nplay an important role in EET processes during the first step of photosynthesis in certain LHCs where \nexcitation energy is transmitted from the antenna pigments to the reaction center in which \nphotochemical reactions are initiated [1–3]. The numerically exact method used to simulate the \ndynamics in this work is the hierarchical equations of motion (HEOM) adapted by Ishizaki and \nFleming to suit the quantum biological regime [4–6]. In the case of an open quantum system, such as \na photosynthetic pigment-protein complex, evolving over time we can generate a set of time \ndependent observables that depict the coherent movement of electronic excitations through the \nsystem by solving a suitable set of quantum dynamic equations such as the HEOM. \nWe have focused on solving two problems, the first being the inverse problem. That is, the objective \nis to determine whether a trained ML model can perform Hamiltonian tomography by using the time \ndependence of the observables as inputs. We demonstrate the capability of the convolutional neural \nnetwork (CNN) to solve the inverse problem. That is, the trained CNN can accurately describe the \nsystem under study by predicting the parameters of the system Hamiltonian when given the \naforementioned time dependent data. The models developed can predict Hamiltonian parameters \nsuch as excited state energies and inter-site couplings of a system up to 99.28% accuracy. \nThe second use of the same data set of observables involves time-series analysis. Although various \nanalytical solutions for the dynamics of open quantum systems such as the HEOM have been \ndeveloped, these often require immense computational resources. We demonstrate that models such \nas SARIMA, CatBoost, Prophet, convolutional and recurrent neural networks can predict the \nlong-time dynamics provided that the initial short-time dynamics is given. Our results suggest that \nSARIMA can serve as a computationally inexpensive yet accurate way to predict long-time \ndynamics.
Author: Naicker, Kimara.
Institution: University of KwaZulu-Natal
Language: English
Subfield: Atomic and Molecular Physics, and Optics
Field: Physics and Astronomy
Domain: Physical Sciences
Open Access status: green
Funder: National Research Foundation
Award: G1113979091
OpenAlex ID: W6979973648
