How Materials Research Benefits from Artificial Intelligence

Research will continue to depend on work in the lab. But AI can greatly accelerate processes. © Shutterstock/Motortion Films

A researcher’s knowledge and experience are still key factors for recognizing and evaluating the potential of new materials. But Big Data analyses, text mining and machine learning can simplify research processes in many ways. AI can scour thousands and thousands of technical articles for certain topics and terms, exposing the target connections more quickly. Artificial intelligence also can develop materials virtually before the physical experimentation takes place in a lab. This raises possibilities that sound like fantasies. For example, Prof. Pascal Friederich of the Karlsruhe Institute of Technology (KIT) says that AI can predict “the characteristics of several million materials” based on the results of about 1,000 simulations. In the process, he says, AI took different formulations into account, along with the production conditions that have a major impact on material properties. Friederich says the goal is to produce the most promising candidates in the lab to validate the predictions. AI’s ability to suggest additional experiments based on automated experiments is another strength, he adds.

Huge Time-savings with AI  

Artificial intelligence clearly has what it takes to reduce the effort and time required for materials research and development. That’s an appealing prospect for departments and institutes that do research and development – especially in sectors with strong competitive pressures that shorten the development timeframes. Google subsidiary Deepmind got people’s attention on the question of AI’s potential in late 2023. It announced that it had identified two million new materials in one year with AI’s help. The company drew on a database containing the descriptions of about 50,000 materials from researchers around the world. It believes that 400,000 of the new compounds could be produced in short order under lab conditions. They could prove helpful with the production of faster computer chips, more efficient solar panels or higher-performance batteries.

A team from Microsoft has advanced further along these lines working with one of the U.S. Department of Energy’s research institutes. Using several AI models over a nine-month period, the staff analyzed 32 million substances with the goal of identifying the most promising material to improve battery performance. By comparison, the development of the lithium-ion battery took about 20 years. The cooperating research institute is currently testing a recipe proposed by AI that consists of lithium, sodium and other elements. If successful, it could reduce the lithium content in batteries by up to 70 percent. Microsoft Vice President Jason Zander was euphoric as he described how quickly the highly promising battery material was identified with AI’s help. “We are at the dawn of a new era of scientific discovery.”