Machine Learning Framework Speeds Catalyst Discovery for Chemical Manufacturing

Scientists at Brookhaven National Laboratory have developed a multi-layer machine learning (ML) framework they say could significantly reduce the time and computational cost of identifying high-performing catalysts for chemical manufacturing and energy production applications.

The framework, published in Chem Catalysis and developed by researchers in Brookhaven's Catalysis Reactivity and Structure group, addresses key limitations of existing single-layer models — including high database generation costs, low data quality and lack of embedded chemical domain knowledge. 

Rather than predicting catalyst performance in a single step, the multi-layer approach breaks evaluation into a series of sequential binary classifications that mirror how chemists assess catalyst behavior experimentally, according to the researchers.

"Simpler one-layer models overlook the domain expertise needed to reliably predict a good catalyst," said Brookhaven chemist Ping Liu in a press statement. "We developed a multi-layer binary machine learning approach that targets complex reaction networks for real catalysis, which has never been considered before in this kind of model."

The team validated the framework using an existing commercial process, CO₂-to-methanol hydrogenation, with copper-based catalysts as the benchmark. Simulations, performed at low computational cost, generated training data that captured competition between multiple reaction pathways that simpler models typically omit, the researchers reported. 

According to the research, the framework successfully identified catalyst designs more active and selective than conventional copper catalysts and revealed that transitions between competing reaction pathways — rather than individual reaction steps alone — are the critical control points for both activity and selectivity.

"Highly active and selective catalysts save energy and costs," Liu emphasized. "An active catalyst means it doesn't require high pressure or high temperatures to speed up a reaction, and a selective catalyst means it doesn't require purification to get the product you want."

According to the press statement, the framework is adaptable to other catalytic processes beyond CO₂ hydrogenation and could improve catalyst discovery workflows for industry partners. The research was supported by the Department of Energy’s Office of Science.