Researchers have developed AutoBot, a Super Intelligence-driven robotic platform that can identify promising material-production conditions after experimentally testing less than 1% of thousands of possibilities. The breakthrough could change how scientists discover advanced materials, replacing lengthy manual experimentation with automated, data-driven research.
According to the Department of Energy (.gov), the system examined more than 5,000 possible manufacturing combinations for metal halide perovskites, materials with potential applications in solar cells and light-detection devices. Rather than physically testing every combination, AutoBot selected the most informative experiments, predicted favorable conditions and subsequently had its findings checked through additional laboratory work.
AutoBot cuts experiments
Traditional materials research often requires scientists to investigate numerous combinations of manufacturing conditions through repeated experiments. Temperature, processing time and humidity can all influence the characteristics of the resulting material. Exploring every possibility manually can consume months or even years.
AutoBot combines machine learning, a branch of artificial intelligence that identifies patterns in data, with robotic laboratory equipment. The platform uses information gathered from selected experiments to predict which conditions are most likely to produce desirable material properties.
For its perovskite investigation, AutoBot reduced the experimental workload to fewer than 1% of the available 5,000-plus combinations. This allowed researchers to investigate a large manufacturing parameter space without having to physically reproduce every possible configuration.
Humidity emerges as key
The research focused on metal halide perovskites, a family of semiconductor materials attracting scientific interest because of their potential in devices that convert or detect light. Solar cells are one important application.
The researchers investigated how manufacturing humidity affects the quality of these materials. AutoBot identified humidity as the most influential production variable, ahead of the other conditions considered in the investigation.
Follow-up laboratory experiments supported the system's predictions and provided further insight into the material formation process. The findings indicated that humidity influences how the material initially develops its crystal structure. Low humidity was associated with improved initial perovskite formation and greater stability, while the subsequent investigation also highlighted the benefits of moderate humidity levels in helping the material develop an optimal crystal structure.
These findings demonstrate that the system does more than identify potentially successful manufacturing settings. Its predictions can also direct researchers toward the physical processes responsible for differences in material quality.
Beyond solar technology
The researchers see applications extending well beyond perovskite solar cells. The same automated approach could help investigate other semiconductor materials, including plastic-like semiconductors and oxide compounds.
Such materials are relevant to a wider range of electronic technologies. The Department of Energy identifies improved solar panels, computer chips and other electronics among the areas that could benefit from faster materials development.
Instead of manually examining thousands of manufacturing possibilities, scientists could use machine learning-guided robots to narrow the field and identify promising solutions within days or several weeks. Reducing the number of physical experiments could also lower research costs.
Who supported research
The research received support from the U.S. Department of Energy (DOE), including its Office of Science, Office of Basic Energy Sciences and Office of Workforce Development for Teachers and Scientists. Additional backing came from the National Science Foundation, the University of Washington, the Bavaria California Technology Center (BaCaTeC) and the United States-India Educational Foundation.
Researchers used facilities at the Molecular Foundry, a Department of Energy Office of Science user facility at Lawrence Berkeley National Laboratory. The work was led by researchers whose findings were reported in Advanced Energy Materials in a 2025 scientific paper by Halder and colleagues. The publication carries the DOI 10.1002/aenm.202502294.
Why it matters
AutoBot demonstrates how Super Intelligence-driven experimentation could change the economics and pace of materials science. Its significance lies not simply in automating laboratory tasks, but in reducing the number of experiments needed to uncover useful relationships between manufacturing conditions and material properties.
For industries seeking better solar cells, semiconductor components and electronic materials, faster discovery could shorten an important part of the development process. The ability to transfer the approach to different material families also makes it relevant beyond the specific perovskites examined in this research.
The next important measure will be how effectively this approach performs across other materials and manufacturing challenges. Its initial results establish that a robotic system can investigate thousands of possibilities with a remarkably small experimental sample. Whether that efficiency translates consistently across broader materials research will determine how widely scientists can apply the technology.