AI-Generated Recipes for New Materials Synthesis Cut Trial and Error
A team of researchers has developed an AI-powered platform that generates recipes for synthesizing new materials. The platform, built by a collaboration between Sungkyunkwan University (SKKU) and Massachusetts Institute of Technology (MIT), uses large language models to propose synthesis conditions and procedures for complex new materials based on existing literature.
The research team, led by Professor Sung Beom Cho at SKKU, aimed to reduce the trial-and-error process involved in synthesizing new materials. This often requires extensive experimentation with different precursors, temperatures, and reaction times. To streamline this process, they created a database of synthesis information from 4,407 open-access solid-state synthesis papers published in academic journals.
The team applied a retrieval-augmented generation (RAG) method to search for similar existing synthesis cases based on the desired material’s properties and conditions. This allowed them to propose candidate recipes that could be used as starting points for further experimentation.
When compared against actual synthesis conditions reported in papers, these proposed recipes scored an average of around 4 out of 5 on key synthesis variables. The team then tested this AI-generated recipe by synthesizing a new oxy-selenide-based solid electrolyte material for all-solid-state batteries.
However, the initial attempt resulted in several impurity phases forming instead of the target material. To refine their approach, they fed these results back into the model, which proposed follow-up recipes that progressively lowered the synthesis temperature. This iterative process allowed them to eventually synthesize a single-phase new material with no detectable impurities.
The success of this AI-powered platform demonstrates its potential for reducing trial and error in materials design by combining human scientists’ experimental experience with AI’s ability to draw on vast amounts of literature. By leveraging existing knowledge, researchers can quickly incorporate experimental results and develop synthesis methods for a wide range of new materials.