Argonne-led AI “Adviser” Accelerates Robotic Design of Advanced Electronic Materials
A research team led by scientists at the U.S. Department of Energy’s Argonne National Laboratory has developed an innovative artificial intelligence (AI) “adviser” designed to monitor and optimize machine learning algorithms during autonomous scientific experiments. The new system is intended to accelerate the discovery of advanced electronic materials by guiding researchers and automated laboratories toward the most promising experimental paths. By continuously analyzing the performance of AI models in real time, the adviser identifies patterns, evaluates progress, and recommends adjustments that help scientists refine their strategies more efficiently.
To demonstrate the capabilities of the adviser, the research team integrated it with Polybot, an AI-driven robotic laboratory platform located at the Center for Nanoscale Materials, a user facility supported by the DOE Office of Science at Argonne. Polybot is designed to autonomously synthesize, characterize, and optimize materials with minimal human intervention. In this study, researchers used the system to investigate a class of electronic materials known as mixed ion-electron conducting polymers (MIECPs). These materials are considered highly promising for next-generation technologies, including wearable electronics, bioelectronics, and advanced energy-storage systems, because they can efficiently transport both ions and electrons.
Autonomous experimentation platforms like Polybot can rapidly explore complex material systems, but they often require large datasets to effectively train and adapt machine learning models. This requirement can slow progress when data is limited or experiments are costly. The newly developed AI adviser addresses this challenge by monitoring the performance of various machine learning algorithms in real time. As experiments proceed, the adviser analyzes results, identifies emerging trends, and communicates actionable insights to human researchers. Scientists can then use this information to modify experimental plans and guide the robotic system toward more productive directions.
When integrated into Polybot’s automated workflow, the adviser played a crucial role in optimizing the experimental process. Instead of testing thousands of possible conditions, the system focused only on the most promising ones. Out of more than 4,300 potential combinations of material processing conditions, the adviser helped narrow the investigation to just 64 experiments. This dramatic reduction significantly accelerated the research while maintaining high-quality results.
During the experimental campaign, the adviser detected that improvements in device performance were slowing when a particular AI optimization algorithm was used. Based on this observation, it recommended switching to a different machine learning optimizer. The research team followed the recommendation, and the change led to a substantial increase in device performance. This ability to dynamically evaluate and adjust algorithms demonstrates how AI advisers can improve decision-making in autonomous laboratories.
The adviser also identified deposition speed as a critical factor influencing material performance. Recognizing the importance of this parameter, researchers expanded their investigation to explore deposition conditions more thoroughly. This deeper analysis ultimately produced additional performance improvements in the materials being studied.
To better understand the relationship between material structure and device behavior, the team conducted detailed characterization on the ten most representative samples produced during the experiments. Some of these measurements were performed at the Advanced Light Source at Lawrence Berkeley National Laboratory. Through these studies, scientists discovered that two structural features strongly influenced performance: wider spacing between polymer layers and thinner fiber structures. These characteristics appeared to improve the materials’ ability to conduct both ions and electrons.
The researchers also found that the polymer material can crystallize into two distinct structural forms, each affecting device behavior differently. These insights provide valuable guidance for designing improved MIECP materials with enhanced performance for electronic and energy-related applications.
The findings from this work were published in the journal Nature Chemical Engineering. In addition to Argonne scientists, the research team included collaborators from the University of Chicago, Lawrence Berkeley National Laboratory, University of Southern Mississippi, and the University of Central Florida. Together, the collaboration highlights how advanced AI tools and autonomous laboratories can dramatically accelerate materials discovery and innovation.
Source link: https://www.businesswire.com/