Key Takeaways

• MIT researchers have developed a fully autonomous robotic system to measure photoconductivity in semiconductor materials, enhancing efficiency in material property characterization.
• The system combines machine learning and robotics, enabling rapid and precise measurements, with over 3,000 data points collected in a 24-hour span.
• This innovation could significantly aid the development of advanced solar cells and other electronic devices, fostering faster discovery of new materials.

Innovative Robotic System for Semiconductor Research

MIT scientists are advancing the search for new semiconductor materials to enhance the efficiency of solar cells and electronic devices. However, the slow manual measurement process limits innovation. A new autonomous robotic system developed by these researchers aims to expedite this process significantly.

The robotic probe focuses on photoconductivity, which measures how a material responds to light. By integrating expert knowledge from material scientists into a machine-learning model, the robot can optimize data collection by identifying the most informative contact points on a material. Additionally, a specialized planning algorithm allows the probe to move quickly between these points.

During a 24-hour test, the system managed over 3,000 unique measurements, averaging more than 125 per hour. This efficiency and accuracy surpasses other AI-based methods used for similar tasks. According to Tonio Buonassisi, a principal researcher, this innovation represents a significant step toward autonomous characterization techniques, especially for properties requiring direct contact measurement.

The research team, based in the Accelerated Materials Laboratory for Sustainability, is particularly focused on developing new perovskite materials for solar panels. Previously, they created methods to rapidly synthesize and analyze combinations of perovskites. However, accurately measuring photoconductivity necessitates direct contact with the material, which previously posed challenges.

To optimize the measuring process, the robotic system uses an onboard camera to capture images of printed perovskite samples. These images are then analyzed using computer vision and a neural network designed to incorporate insights from chemists. The model pinpoints ideal contact areas based on the material’s characteristics and directs the robotic probe accordingly.

Siemenn, the lead author of the research, emphasizes the importance of human expertise throughout this process, stating that integrating knowledge from materials scientists is crucial for effective material discovery.

The unique shapes of the printed samples—ranging from circular drops to jellybean-like forms—make adaptability a key feature of the machine-learning model. The system’s path planner enhances efficiency further by adding randomness to the algorithm, enabling quicker route calculations.

Testing confirmed that this model outperformed seven other AI methods in identifying contact points and produced more efficient path-planning results. In a single autonomous session, the robotic system generated a substantial amount of data, allowing researchers to locate areas with higher photoconductivity and identify any material degradation.

Siemenn remarked on the potential for this data-rich approach, facilitating the discovery of high-performance semiconductors, particularly for sustainability applications like solar energy. The research team aims to refine this robotic system further as they work towards a fully autonomous laboratory dedicated to materials discovery.

The content above is a summary. For more details, see the source article.