KAIST said on Wednesday that a research team led by professor Ji-min Kwon (권지민) of the School of Electrical Engineering and the Department of AI Systems developed a technology that automatically screens 2D semiconductors using only optical microscope images and links the process to transistor fabrication. The work was conducted jointly with UNIST, Hanbat National University, Hanyang University and Washington University in St. Louis in the United States.

A 2D semiconductor is an ultrathin semiconductor only a few atomic layers thick. It is called a “dream semiconductor” because it can realize semiconductors that are smaller and use less electricity than conventional silicon semiconductors. In the future, 2D semiconductors are expected to be used in a range of technologies, including AI semiconductors, smartphones, data centres, wearable devices, foldable or stretchable electronics and ultraminiature medical sensors.

However, 2D semiconductors made by a solution process vary in the location, size and thickness of flakes, or small semiconductor pieces, requiring researchers to find desired samples one by one under a microscope. Researchers then had to design electrodes directly for the corresponding positions, requiring significant time and effort. It was also practically difficult to analyse more than thousands of devices at once.

The team used molybdenum disulfide (MoS2), a representative 2D semiconductor material. It used the property that RGB red, green and blue brightness values seen under a microscope vary by thickness, enabling a computer to automatically find the desired semiconductor and automatically design electrodes.

Verification with atomic force microscopy showed the team could distinguish even subtle thickness differences of 3 to 8 layers. Using the method, it succeeded in automatically screening suitable samples among more than 120,000 flakes and producing and analysing 1,615 transistors.

The team also confirmed the relationship between thickness and electrical performance. It statistically identified a characteristic in which current flows better as the semiconductor becomes thicker, but switching performance worsens. It revealed this through large-scale data analysis, a characteristic that was difficult to confirm previously because only a small number of samples could be analysed.

Kwon said, “Previously, researchers had to find the desired semiconductor directly under a microscope, but this study has made it possible to automate this process.” She added, “Going forward, it will lay the foundation for predicting a semiconductor’s electrical performance using only microscope photos and for developing better next-generation semiconductors much faster.”

The study had Kwon, Hak-soon Jung (정학순) and Yong-woo Lee (이용우) as co-corresponding authors, and Sang-hyun Lee (이상현), a researcher at UNIST, as the first author. The results were published on April 3 in the international materials science journal Advanced Functional Materials and were selected as an inside back cover paper in the fields of 2D materials and electronic devices.