In another instance of AI making itself genuinely useful, researchers at the University of Toronto (UoT) have identified a better catalyst for the production of green hydrogen using AI – saving themselves years in experimentation.
Search made faster
Led by PhD student Jehad Abed, the team developed an AI program and trained it on more than 36,000 metal oxide combinations. Its purpose was to identify an alloy to serve as an effective catalyst in hydrogen production. The AI ran simulations on these combinations to determine which one would be the most efficient, stable, and durable throughout the production process.
The AI could suggest any of billions of metal oxide combinations. Working out an alloy that would satisfy the researchers’ desired conditions could have taken them years, but as Abed noted, the AI arrived at a strong candidate in a matter of days.
Ultimately, the program identified an alloy consisting of specific proportions of ruthenium, chromium, and titanium. It turned out to be 20 times more stable and long-lasting than the team’s benchmark metal.
AI helps find cheaper way to make green hydrogen
A closer look at the results
Hydrogen is produced by passing electricity between two pieces of metal suspended in water; hydrogen and oxygen gases are released during this process.
While you can use renewable energy sources to supply the electricity, you still need a lot of it – along with expensive metals – to produce hydrogen. Introducing a catalyst in the mix can make the process more efficient and cost-effective.
Once the AI identified the ideal catalyst alloy, the researchers used ultra-bright X-rays to assess its performance through the course of a reaction. They experimented at Canadian Light Source, a research facility at the University of Saskatchewan equipped with a synchrotron that’s said to be “billions of times brighter than the sun.“
This enabled them to observe “how the [AI-identified catalyst’s] atomic arrangements would change and respond to the amount of electricity that we put in,” and conclude that it was a strong choice for the production of green hydrogen.
Will it scale?
The next step, of course, is to test the alloy under real world conditions and see if it delivers. This could make green hydrogen a more feasible choice as a clean fuel. It’s one of many recent advances in hydrogen production from around the world.
For example, scientists at Australia’s RMIT came up with a highly efficient, low-cost green hydrogen generation process using seawater and a novel catalyst last year. There’s also the recently invented proton-exchange-membrane method for producing hydrogen cheaply to keep an eye on. And we reported just last week that Swiss scientists devised a cheap and efficient way to store hydrogen for months without losing it into the atmosphere.
The UoT researchers’ program follows DeepMind’s innovation from last year, in which its GNoME AI tool helped discover 2.2 million new crystals that could find use in next-gen superconductors and EV batteries. Both breakthroughs underscore the groundbreaking capability and blinding speed with which AI can help us find useful materials for the tech of the future. Some other developments of note include a neural network that gives ships a 5-minute warning before rogue waves appear out of the sea, and smarter breast cancer screening.
A paper on the UoT research has been published in the Journal of the American Chemical Society.
Source: Canadian Light Source
