Our research programme has two connected directions: automated chemical exploration and scientific machine learning. Both begin with chemical questions; electronic-structure theory, search algorithms, data-driven models and scientific software are combined as the problem requires.
Automated chemical exploration
Chemical discovery is often limited by the structures and mechanisms a researcher thinks to test. We are interested in computational strategies that search more broadly: generating molecular structures, navigating reaction pathways and identifying chemically meaningful regions of large search spaces.
PyAR is our principal software programme for reproducible structure and reaction search. It supports systematic generation and exploration across reactions, molecular aggregates and atomic clusters.
Explore PyAR and automated structure and reaction search →
Scientific machine learning
We use machine learning when it makes a chemical problem more tractable: to accelerate an energy model, guide a search or identify relationships worth testing. We favour interpretable, chemically grounded models and make validation, uncertainty and domain limits part of the scientific argument. Prediction alone is not enough.
Application areas
Reaction discovery, astrochemistry and prebiotic chemistry
Some problems begin by asking which molecules are possible; others ask how those molecules form. HydroMol maps a defined space of small hydrocarbons, while our automated reaction studies explore networks involving HCN, HNC, ammonia, formaldehyde and related species. Astrochemistry and prebiotic chemistry provide demanding tests of both approaches.
Explore automated reaction discovery and astrochemical space →
Molecular aggregates, nanoclusters and nanoalloys
Atoms and molecules can assemble into many competing structures whose stability and properties depend on size, composition, bonding and collective interactions. We use automated cluster building, quantum chemistry and machine-learned potentials to explore these landscapes—from hydrogen-bonded molecular aggregates to metallic clusters and multicomponent nanoalloys.
Explore molecular aggregates, nanoclusters and nanoalloys →
Molecular discovery and functional materials
We combine molecular generation, quantum chemistry and machine learning to move from large candidate spaces towards molecules selected for particular functions. Current applications include strained hydrocarbons for energy storage, TADF emitters for OLEDs, photocages, photoswitches and photosensitisers.
Explore molecular discovery and functional materials →
The same two directions run through these application areas: automated exploration widens what can be considered, while scientific machine learning helps rank, accelerate or interpret where appropriate.