Automated 2D Semiconductor Screening Could Speed Low-Power AI Chip Development

Researchers from KAIST and several partner institutions have developed an automated system for screening two-dimensional (2D) semiconductors and fabricating transistors, moving away from manual, labor-intensive processes. By utilizing optical microscope images to identify material thickness and automate electrode design, the team successfully analyzed over 1,600 devices to establish a clear link between material thickness and electrical performance. This breakthrough is expected to accelerate the commercialization of ultra-low-power AI semiconductors by transitioning 2D material research into a data-driven, high-throughput discipline.

A collaborative research team led by Professor Jimin Kwon at KAIST has introduced a technology that automates the identification of 2D semiconductors, specifically molybdenum disulfide (MoS₂), from optical microscope images. Traditionally, researchers had to manually locate individual semiconductor flakes and design electrodes for each, a process that severely limited the scale of device analysis. The new system leverages RGB brightness values to distinguish subtle thickness differences—ranging from three to eight layers—which was verified using atomic force microscopy (AFM). This automation allowed the team to screen more than 120,000 semiconductor flakes and successfully fabricate 1,615 transistors for large-scale evaluation.

The study, published in Advanced Functional Materials, provides critical statistical insights into the relationship between the physical dimensions of 2D materials and their electrical properties. The researchers clarified for the first time that while thicker MoS₂ layers allow current to flow more easily, they simultaneously suffer from a decreased ability to switch electricity on and off effectively. These findings were only possible through the analysis of thousands of devices, a feat previously unattainable through manual methods. Such data-driven insights are vital for the semiconductor industry as it seeks to overcome the physical limits of conventional silicon, which faces increasing challenges with power loss and heat generation during miniaturization.

The implications for the Semiconductors & Chips sector are significant, as 2D semiconductors are often referred to as "dream semiconductors" for their potential to enable ultra-thin, low-power electronics. This automated fabrication and screening process paves the way for the mass production of next-generation components used in AI semiconductors, data centers, and wearable devices. By transforming 2D semiconductor research from a process based on human experience into a high-speed, data-centric model, the technology sets the stage for future AI-driven semiconductor design. This shift is expected to accelerate the timeline for commercializing foldable electronics, high-performance medical sensors, and more efficient mobile hardware.