Screening catalyst surfaces at 73,000 CPU-hour scale

Assessing whether a catalyst surface holds up under reducing conditions requires modelling many candidate configurations at the atomic scale. Each configuration is an expensive DFT calculation, the space of plausible surfaces is large, and stability is not obvious from structure alone — so the screening has to be systematic rather than intuition-led.

• Built and screened 40+ aMOC surface configurations for computational catalysis research.

• Ran approximately 73,000 CPU-hours of density functional theory calculations using ASE and VASP.

• Assessed structural stability of candidate surfaces under reducing conditions.

• Connected atomic-scale modelling to potential hydrogen fuel cell applications.

• Separately built Sankey diagrams mapping high-performance polymer supply chains (see Explorations).

1. 01

   Constructed candidate surface configurations systematically with ASE so the screen covers the space rather than sampling it arbitrarily.

2. 02

   Ran DFT via VASP at a scale where job orchestration, convergence behaviour, and failure handling matter as much as the physics.

3. 03

   Evaluated structural stability under reducing conditions — the environment the material would actually face.

4. 04

   Tied results back to the application question: whether these surfaces are viable for hydrogen fuel cell contexts.

• Compute at this scale is unforgiving: a misconfigured calculation is not a fast failure, it's thousands of wasted CPU-hours.

• Convergence is not guaranteed. A meaningful share of the work is diagnosing why a calculation didn't settle rather than interpreting one that did.

• 40+ configurations only produce a usable comparison if the setup is consistent across every one of them.

• The result is a stability assessment, not a product — long-horizon work where the payoff is a credible answer, including a negative one.

Comfort with long-horizon, compute-heavy technical work where the answer isn't guaranteed and the discipline is in the setup. This is the layer beneath the product — the part that has to be right before anything gets built on top of it.