Engineers are developing universal, high-speed technology to model, understand and respond to forces

Researchers have been studying industrial production of ammonia for a century. But they struggled to find ways to improve low-yield, inefficient operations.

Atmospheric nitrogen, with the help of iron ore, combines with hydrogen to produce ammonia. That process produces a lot of ammonia—160 million tons worldwide each year. Most of it is used in agriculture, mainly as a nitrogen fertilizer. It is also used in many industries, including refrigeration for food and beverage production. We all know it as a household cleaner.

A research team led by Qi An, an assistant professor of materials science and engineering at Iowa State University, has developed artificial intelligence technology that could find ways to improve researchers’ understanding of chemical reactions. involved in the production of ammonia and other complex chemical reactions.

“Our improved HDRL-FP project has the potential to significantly contribute to the development of this process, which can reduce production costs and CO₂ emissions, and help create smaller and more widespread plants,” the researchers wrote in a paper recently published online. newspaper Nature Communication. “Thus, the design highlights its effectiveness and ability to predict complex chemical pathways.”

HDRL-FP is a Higher Level Intensive Course in First Principles. An and his colleagues and co-authors—Tian Lan and Huan Wang of Salesforce AI Research in California—say the technology is full of potential.

“Exploring reaction mechanisms is important for understanding chemical processes, optimizing reaction conditions, and developing more efficient materials,” they wrote.

Of rewards and atoms

An said there are two keys to the researchers’ software technology: a type of machine learning called reinforcement learning and linking the simulation method to the positions of the atoms involved.

Reinforcement learning is like training a dog by using rewards to encourage behavior. In reinforcement learning, computers learn from their actions as they search for appropriate rewards. In this case, the rewards are about finding the best, most efficient, cheapest way.

This method, when used with image processing features and multi-generational strategies, can quickly and automatically determine the optimal response path from thousands of possible paths, An said. . That accurately identifies the active reaction mechanisms among the very noisy data in real chemical reactions.

Researchers have also developed the technology to be useful for general studies of catalytic reactions. Lessons begin with the positions of the atoms mapped out in the energy field. It is sufficient—researchers do not have to start with a more detailed representation of the response environment, including states, actions or rewards for a particular response.

An and his colleagues worked on the project for nearly two years. It started when An transferred to Iowa State and is supported by his university start-up funds.

He said the system’s readings for the ammonia-producing reaction are considered indicative.

“This allows us to know how to respond,” said An. “We are able to see key reaction steps in ammonia synthesis.”

They wrote, documenting the success of researchers in that reaction “to make the investigation of complex chemical reactions self-evident,” providing a promising approach for future research. comes with discoveries.