Use cases

Research running on Falcon

How Cardiff researchers are using Falcon's CPU and GPU computing in practice.

Case study · Chemistry

Using AI and Advanced Computing to Accelerate Catalyst Discovery

Researchers at Cardiff University are using high-performance computing, AI and electron microscopy to understand catalysts better. Catalysts are central to tackling decarbonisation, using carbon dioxide and building a circular economy. The work runs on Falcon, Cardiff’s HPC and AI cluster, which gives the team both the CPU and GPU computing it needs.

The challenge

Catalysts cut the energy and waste of chemical processes, and more than 90% of those processes use at least one. The Cardiff Catalysis Institute studies them down to the atomic scale. Modern microscopy can produce terabytes of data, which has to be analysed alongside demanding simulations of atomic structures and reactions.

How advanced computing helps

Falcon’s CPU nodes process the experimental data and run the simulations, and its GPUs run the machine learning. AI can analyse microscopy images quickly enough to give feedback while an experiment is still running, and machine learning can make molecular simulation more efficient.

Results

The team has published a neural-network pipeline for automatically analysing catalyst images, as well as work combining modelling and machine learning to understand catalytic reactions. Their research has appeared in Nature, ACS Catalysis, Digital Discovery and the Journal of Chemical Theory and Computation.

People

Research team

Prof Sir Richard Catlow, Dr Alberto Roldan Martinez, Prof David Willock, Dr Andrew Logsdail, Dr Stefano Leoni and Dr Thomas Slater, from the School of Chemistry and the Cardiff Catalysis Institute.

This case study is summarised from an article by Supercomputing Wales. Read the full article, including publications, on the Supercomputing Wales website.