“What is particularly significant about this study, which combines experiment and theory, is that in addition to identifying several new compounds for solar fuel applications, we were also able to learn something new about the underlying electronic structure of the materials themselves,” Neaton said in a Caltech press release.

To discover these new photoanodes, the team combined computational and experimental approaches. A Materials Project database was mined for potentially useful compounds. Hundreds of theoretical calculations were performed using computational resources at the National Energy Research Scientific Computing Center (NERSC), together with software and expertise from the Molecular Foundry. Once the best candidates for photoanode activity were identified, it was time to test those materials in the laboratory.

The materials were simultaneously tested for anode activity under different conditions using high-throughput experimentation. This was the first time these kinds of experiments had been run this way, according to Gregoire.

“The key advance made by the team was to combine the best capabilities enabled by theory and supercomputers with novel high throughput experiments to generate scientific knowledge at an unprecedented rate,” Gregoire said in the press release.

Eoin Treacy's view

There has been great deal of commentary in the media about the advances in artificial intelligence and how it is represents a threat to employment across a number of fields. A broader perspective to the easy application of massive computing power is the scale that can be brought to experiments through computer simulation and data analysis. Artificial intelligence represents a major facilitator for technological innovation. Coupled with rapid prototyping and CRISPR the potential for unprecedented change in a range of sectors, stretching from materials to healthcare, is looking increasingly like the base case.