New semiconductor materials that potentially increase the efficiency of solar cells and other gadgets are being sought after by scientists. However, the speed at which scientists can manually monitor crucial material characteristics is a barrier to innovation. Things might move more quickly, thanks to a completely autonomous robotic system created by MIT researchers. The study was published in Science Advances.
Image credit: Pixel B/Shutterstock.com
Their technology makes use of a robotic probe to evaluate photoconductance, an important electrical characteristic that determines how electrically receptive a material is to light.
The researchers include human specialists’ knowledge of the materials science area into the machine-learning model that directs the robot’s decision-making. This allows the robot to choose the optimal locations to touch a material with the probe to obtain the most information about its photoconductance, while a unique planning method determines the quickest path between contact spots.
During a 24-hour test, the completely autonomous robotic probe acquired over 125 unique measurements each hour, demonstrating more precision and dependability than existing artificial intelligence-based technologies.
This technology might stimulate the development of more efficient solar panels by drastically accelerating the rate at which scientists can analyze crucial features of novel semiconductor materials.
I find this paper to be incredibly exciting because it provides a pathway for autonomous, contact-based characterization methods. Not every important property of a material can be measured in a contactless way. If you need to make contact with your sample, you want it to be fast and you want to maximize the amount of information that you gain.
Tonio Buonassisi, Study Senior Author and Professor, Massachusetts Institute of Technology
His co-authors include graduate student Alexander (Aleks) Siemenn as the main author, postdoctoral researchers Basita Das and Kangyu Ji, and graduate student Fang Sheng.
Making Contact
Since 2018, researchers in Buonassisi’s laboratory have been developing a fully autonomous materials discovery lab. Their recent focus has been on finding new perovskites, a type of semiconductor used in photovoltaics like solar panels.
In previous work, they established methods for rapidly synthesizing and printing novel combinations of perovskite materials. They also developed imaging-based methods for determining critical material characteristics.
However, the most precise way to determine photoconductance is to place a probe on the material, shine a light on it, and then measure the electrical reaction.
To allow our experimental laboratory to operate as quickly and accurately as possible, we had to come up with a solution that would produce the best measurements while minimizing the time it takes to run the whole procedure.
Alexander (Aleks) Siemenn, Study Lead Author and Graduate Student, Massachusetts Institute of Technology
This requires the combination of machine learning, robotics, and material science into a single autonomous system.
To begin, the robot’s onboard camera captures an image of a slide with perovskite material written on it.
The image is then segmented using computer vision and input into a neural network model that was specifically created to combine domain expertise from chemists and materials scientists.
“These robots can improve the repeatability and precision of our operations, but it is important to still have a human in the loop. If we don’t have a good way to implement the rich knowledge from these chemical experts into our robots, we are not going to be able to discover new materials,” added Siemenn.
The model applies this domain knowledge to find the best locations for the probe to touch depending on the sample’s shape and material composition. These contact points are sent into a path planner, which calculates the most efficient route for the probe to reach all spots.
This machine-learning approach’s versatility is especially crucial given the printed samples’ different shapes, which range from circular droplets to jellybean-like structures.
“It is almost like measuring snowflakes — it is difficult to get two that are identical,” said Buonassisi.
After determining the shortest path, the path planner sends signals to the robot’s motors, which maneuver the probe and take measurements at each contact point in quick succession.
The neural network model’s self-supervised nature is critical to the method’s speed. The approach identifies ideal contact points immediately on a sample image, eliminating the requirement for labeled training data.
The researchers also expedited the system by improving the path planning process. They discovered that adding a little bit of noise, or unpredictability, to the algorithm improved its ability to locate the shortest path.
“As we progress in this age of autonomous labs, you really do need all three of these expertise — hardware building, software, and an understanding of materials science — coming together into the same team to be able to innovate quickly. And that is part of the secret sauce here,” added Buonassisi.
Rich Data, Rapid Results
After building the system from the ground up, the researchers tested each component. Their findings revealed that the neural network model identified more contact locations with less calculation time than seven other AI-based approaches. In addition, the path planning algorithm consistently produced shorter path plans than previous approaches.
When they put all the components together for a 24-hour completely autonomous experiment, the robotic system took over 3,000 distinct photoconductance measurements at a pace of more than 125 per hour.
Furthermore, the degree of detail supplied by this exact measuring technique allowed the researchers to discover both hotspots with greater photoconductance and regions of material deterioration.
“Being able to gather such rich data that can be captured at such fast rates, without the need for human guidance, starts to open up doors to be able to discover and develop new high-performance semiconductors, especially for sustainability applications like solar panels,” stated Siemenn.
The researchers intend to continue developing this robotic system as they try to construct a completely autonomous lab for materials discovery.
This research is funded in part by First Solar, Eni through the MIT Energy Initiative, MathWorks, the University of Toronto’s Acceleration Consortium, the United States Department of Energy, and the National Science Foundation.
Journal Reference:
Siemenn, A., et al. (2025) A self-supervised robotic system for autonomous contact-based spatial mapping of semiconductor properties. Science Advances. doi.org/10.1126/sciadv.adw7071.
