Facebook Inc.’s artificial intelligence research unit said today it has teamed up with Carnegie Mellon University’s Department of Chemical Engineering on a collaborative project to use AI to accelerate quantum mechanical simulations.

The idea is to discover new “electrocatalysts” that can provide a more efficient and scalable method of storing and using renewable energy. Facebook AI research scientist Larry Zitnick said in a blog post the project aims to tackle the intermittent nature of renewable energy sources such as wind and solar energy.

The problem is that although both energy sources are vital part of the modern power grid, at present they cannot be relied upon because, of course, “the sun doesn’t always shine and the wind doesn’t always blow.” In order to increase our reliance on such energy sources, he said, we need to find a better way to store this energy once it has been generated, so that it can be made available days, weeks or months later, when it’s most needed.

The obvious answer is to store this energy in lithium-ion batteries, but Zitnick said that would be unrealistic because of the prohibitive cost: There simply aren’t enough batteries available to do it on a global scale.

Instead, Facebook and Carnegie Mellon’s Department of Chemical Engineering have hit upon the idea of using chemical reactions to convert excess solar and wind energy into other fuels that can be stored more easily, such as hydrogen or ethanol. That’s already possible, but existing methods for doing so require the use of rare and expensive electrocatalysts such as platinum. Zitnick said the Open Catalyst Project is an effort to discover new, lower-cost catalysts that can drive these reactions instead.

It’s a big challenge. The discovery of new catalysts is an arduous process. They need to be made up of at least three known metals, and there are 40 in total that we know about, which means more than 10,000 possible combinations of elements. Worse still, each combination must be thoroughly tested by adjusting the ratios or configurations of each element to determine its efficiency as an electrocatalyst, which adds up to billions of possibilities.

Most researchers in this field use quantum mechanical simulation tools such as Density Functional Theory for this kind of research, but even the most advanced of them are limited to around 40,000 simulations per year, which isn’t nearly enough, Zitnick said.

The Open Catalyst Project aims to accelerate this research and help scientists to screen billions of possible catalysts each year, using AI to accurately predict atomic interactions at an order of magnitude faster than today’s existing, compute-heavy simulations.

“Calculations that take modern laboratories days could, with the help of AI, take seconds,” Zitnick said. “Leveraging artificial intelligence to approximate DFT computation is a necessity if we’re to explore the full field of possible catalysts.”

To that end, Facebook and Carnegie Mellon’s Department of Chemical Engineering said they’re releasing on GitHub the open-source Open Catalyst 2020 data set and a set of baseline machine learning models to accelerate this research. Zitnick said he hopes the project’s open and collaborative nature will inspire other researchers to join in the effort to discover new electrocatalysts for renewable energy. He said it’s a much larger and more important initiative than any single discipline or institution.

“If successful, this research has the potential to significantly accelerate the global shift towards renewable energy, removing the high costs associated with current electrocatalysts, providing a scalable alternative to expensive storage technologies like batteries, and supplying clean and sustainable power the world over,” Zitnick said.

Photo: Oimheidi/Pixabay