Argonne-led AI “Adviser” Accelerates Robotic Design of Advanced Electronic Materials

LEMONT, Ill.--(BUSINESS WIRE)--A research team led by the U.S. Department of Energy’s (DOE) Argonne National Laboratory developed an innovative AI “adviser” that monitors and optimizes the performance of machine learning algorithms as autonomous experiments progress, enabling faster discovery of advanced electronic materials.

The researchers applied the adviser to Polybot, Argonne’s AI-guided robotic laboratory, to accelerate the investigation of electronic materials called mixed ion-electron conducting polymers (MIECPs), materials promising for wearable electronics and energy storage. Polybot is in the Center for Nanoscale Materials, a DOE Office of Science user facility at Argonne.

Autonomous platforms typically require large datasets to adapt effectively. The adviser mitigates data scarcity by evaluating algorithm performance in real time, extracting actionable patterns, and communicating those insights to human scientists who refine experimental plans. Integrated with Polybot’s autonomous synthesize-characterize-optimize workflow, the adviser guided adaptive choices that reduced the study to just 64 experiments out of more than 4,300 possible processing-condition combinations.

During the campaign, the adviser observed diminishing performance improvements from one AI optimizer and suggested switching to another AI algorithm for subsequent experiments. The scientists implemented the recommendation, and device performance improved significantly.

The adviser also flagged deposition speed as a key driver of performance, prompting a broader investigation of that parameter that led to further gains.

Researchers performed in-depth characterization of the 10 most representative material samples—including measurements conducted at the Advanced Light Source, another DOE Office of Science user facility at LBNL—to link device behavior to material structure. Two structural features played a crucial role in better performance: wider spaces between layers and thinner fibers. The team also discovered that the material crystallizes into two distinct structures. These significant findings can be leveraged to design higher-performing MIECPs.

The study was published in Nature Chemical Engineering. In addition to Argonne, the research team included the University of Chicago, Lawrence Berkeley National Laboratory (LBNL), University of Southern Mississippi and the University of Central Florida.

A full version of the press release is available on Argonne’s website.