Robotic probe quickly measures key properties of new materials

Cambridge, MA – Scientists are on a relentless quest to uncover novel semiconductor materials that promise to revolutionize the efficiency of solar cells and various electronic devices. However, the pace of these groundbreaking discoveries has often been hampered by the time-consuming manual measurement of critical material properties.

Breaking this bottleneck, researchers at MIT have unveiled a fully autonomous robotic system designed to dramatically accelerate this process. Their innovative system deploys a robotic probe to swiftly and precisely measure photoconductance – a vital electrical property indicating a material’s responsiveness to light.

What sets this system apart is its sophisticated integration of human-expert knowledge directly into the machine-learning model guiding the robot’s decisions. This unique approach allows the robot to intelligently identify optimal contact points on a material, ensuring maximum information gain regarding its photoconductance. Complementing this, a specialized planning procedure calculates the most efficient path for the probe to move between these points, minimizing measurement time.

During a rigorous 24-hour test, the autonomous robotic probe showcased remarkable performance, conducting over 125 unique measurements per hour. This unparalleled speed was matched by superior precision and reliability when compared to other existing artificial intelligence-based methods.

This dramatic increase in the speed of characterizing new semiconductor materials holds immense promise for the rapid development of next-generation solar panels that could generate significantly more electricity, contributing to a more sustainable future.

“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,” stated Tonio Buonassisi, professor of mechanical engineering and a senior author of the seminal paper detailing the autonomous system. His co-authors include lead author Alexander (Aleks) Siemenn, graduate student; postdocs Basita Das and Kangyu Ji; and graduate student Fang Sheng. The research was published in Science Advances.

The journey towards a fully autonomous materials discovery laboratory began in Buonassisi’s lab in 2018, with a recent focus on novel perovskites – a class of semiconductor materials crucial for photovoltaics. While prior work involved rapid synthesis and imaging-based characterization, photoconductance demanded direct probe contact for accurate measurement.

“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,” explained Siemenn. This challenge necessitated a seamless fusion of machine learning, robotics, and advanced material science.

The system operates by first capturing an image of the perovskite material via its onboard camera. This image is then segmented by computer vision and fed into a neural network model, ingeniously designed to integrate expert domain knowledge from chemists and materials scientists. This human-in-the-loop AI ensures optimal probe contact points are determined based on the sample’s unique shape and composition, even for irregular forms like ‘snowflakes’. These points are then processed by a path planner, which calculates the most efficient route for the robotic probe to take measurements in rapid succession. The self-supervised nature of the neural network and the optimized path planning algorithm (incorporating a touch of controlled randomness for efficiency) are key contributors to the system’s remarkable speed.

“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,” Professor Buonassisi added.

Comprehensive testing of each component revealed that the neural network model outperformed seven other AI-based methods in identifying superior contact points with reduced computation time. Similarly, the path planning algorithm consistently delivered shorter, more efficient routes.

During a full 24-hour autonomous experiment, the integrated robotic system executed over 3,000 unique photoconductance measurements. The rich, precise data gathered at such high rates enabled researchers to pinpoint ‘hotspots’ of higher photoconductance and identify areas exhibiting material degradation, offering invaluable insights for material optimization.

“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,” Siemenn concluded.

This groundbreaking work received support from First Solar, Eni through the MIT Energy Initiative, MathWorks, the University of Toronto’s Acceleration Consortium, the U.S. Department of Energy, and the U.S. National Science Foundation.