X-ray and electron scattering experiments are in renaissance, and researchers are now able to use the latest high-brilliance light sources at particle accelerator facilities to follow photochemical dynamics far from equilibrium in real time, observe rare stochastic events like reaction barrier crossing, and record single-shot spectra of unstable compounds with ultrashort and ultrabright pulses of light.
Connecting the experimental observables to the operational chemistry and physics with computational simulations is the key to understanding and explaining physical phenomena; ultimately it is also the key to harnessing them for practical applications. The challenge lies in making the simulations capable of capturing the complexity of the experiments while simultaneously fast, accurate, affordable, and generally-applicable enough to appeal to users.
It turns out that this is a tall order - but it isn't impossible!
Using deep neural networks (DNNs; a kind of machine learning model inspired by the architecture of the brain), we can reduce the time that these simulations take from hours/days to a fraction of a second. We can democratise data analysis, unlock new kinds of high-throughput experiments, and help beamline users to better plan their beamtime allocations with ’on-the-fly’, ’limited-expertise-required’ machine-aided analyses. DNNs can also help us analyse the experimental observables directly from the detector(s), and computer vision- and decision-capable DNNs can even assert autonomous control over the instrumentation in the laboratory.
The research I do here at the University of York is all about making this a reality.
Conor came to the University of York as a Lecturer in Machine Learning in 2022. He was an undergraduate (MChem; 2011-2015) and postgraduate (PhD; 2015-2019) at the University, and he spent five years (2014-2019) working in Dr Derek Wann's gas electron diffraction laboratory.
From 2014-2019, he recomissioned a vintage gas electron diffractometer from the University of Reading and brought together theory and experiment to better plan, propose, and analyse gas electron diffraction experiments.
Afterwards, he left for Newcastle University to take up an EPSRC Doctoral Prize Fellowship (2019-2022) with Professor Tom Penfold. From 2019-2022, he developed machine-learning models for predicting K- and L-edge X-ray absorption and emission spectra of first-row transition metals.
Coming soon - check back for updates.