Author ORCID Identifier

https://orcid.org/0000-0002-2659-552X

Date Available

5-6-2025

Year of Publication

2025

Document Type

Doctoral Dissertation

Degree Name

Doctor of Philosophy (PhD)

College

Arts and Sciences

Department/School/Program

Chemistry

Faculty

Dr. Chad M. Risko

Faculty

Dr. Kenneth Graham

Abstract

Chemistry is entering a new paradigm of automation and data-driven discovery. Automated discovery is grounded in well-curated “big data.” As generative and predictive models fueled by simulation data see growing success, emerging robotic automation enables the generation of unprecedented volumes of experimental data. Automation-powered, data-driven approaches hold tremendous potential for groundbreaking insights and innovations, particularly in the study and discovery of electroactive π-conjugated molecules. Realizing this potential, however, requires democratizing chemical data and the automation needed to generate and use it. There is a need to expand access to the tools for findable, accessible, interoperable, and reusable (FAIR) data management and experimental automation. This dissertation contends that efficient discovery in the realm of electroactive π-conjugated molecules requires a coalition of automation and data-driven design with chemists and chemical intuition; this necessitates both large-scale FAIR data and intuitive man-machine interfaces. This dissertation investigates the automation of big-data generation, management, and analysis in the context of studying small electroactive π-conjugated molecules. First, this work examines the philosophical and historical foundations underpinning chemical data ontologies upon which automation and data-driven approaches depend. It advocates for interdisciplinary collaboration between philosophers and chemists to create more realistic, intuitive, and FAIR-compliant data structures. Then, this dissertation explores data generation and management in practice by producing computational data for over 40,000 electroactive molecules via automated high-throughput quantum chemical calculations and building a management infrastructure for the resulting data. It next demonstrates the insights gained through analyzing big data with a study of dihedral angle rotations in π-conjugated systems. The results demonstrate the ability of data-empowered machine learning (ML) to inexpensively automate the estimation of experiment-aligned estimations for mesoscale properties. Likewise, it discusses how big data can be utilized for informing the selection of similarity measures, a key metric in many automated discovery applications. This work finally transitions to the automated generation of experimental data. It overviews a software developed for translating experimental protocols to robotic actions, validating the system by reproducing well-reported electrochemical experiments. Overall, this dissertation offers a path through effective organization, generation, management, and use of chemical data towards the automated study and discovery of electroactive π-conjugated molecules.

Digital Object Identifier (DOI)

https://doi.org/10.13023/etd.2025.100

Funding Information

This work was generously supported by the National Science Foundation (NSF) under Cooperative Agreement Numbers 2019574 and 1849213 and the Office of Naval Research under Award Numbers N00014-22-1-2179 and N00014-18-1-2448. This work was also supported by the University of Kentucky Lyman T. Johnson Fellowship and the PEO Scholars Award.

Download

Share

COinS