Colour is everywhere, in autumn leaves, glowing screens, and solar panels converting sunlight into electricity. The molecules behind these phenomena, known as chromophores, are well-studied experimentally, yet their behaviour at the atomic level remains poorly understood, particularly when they cluster together or solidify into disordered materials.
This PhD project addressed that gap by developing a simulation framework that combines artificial intelligence with atomistic modelling. Machine-learned potentials allow simulations to run at quantum-level accuracy for far larger and more realistic systems than was previously feasible, making it possible to follow molecular motion across the time and length scales that matter most for real materials.
Neutron scattering was the experimental anchor for this work. Neutrons are uniquely sensitive to hydrogen-rich molecules like chromophores and can capture both where atoms sit and how they move. A major achievement of the project was learning to predict what a neutron instrument would actually measure from a simulation – turning theoretical calculations into something directly testable at a facility and giving experimentalists a richer lens through which to interpret their data.
The methods were put to work on three fronts: modelling the slow relaxation of glass-forming materials, computing how chromophores interact with light, and reproducing neutron scattering spectra from first principles. The project also produced calorine, a software package that makes these workflows accessible to researchers who may have little prior experience with simulations.
The long-term value of this work extends beyond chromophores. By making simulation-to-experiment pipelines easier to use and more accurate, it lays groundwork for faster development of light-harvesting and energy storage technologies, and potentially changes how neutron facilities approach data analysis.