Two-dimensional semiconductors, such as molybdenum disulfide (MoS₂), consist of layers just a few atoms thick and offer remarkable properties for miniaturized electronics. These materials hold the potential to surpass traditional silicon by enabling smaller, more energy-efficient chips critical for AI, wearable devices, flexible electronics, and medical sensors. However, variability in flake thickness, size, and position has made manual identification under microscopes a painstaking bottleneck, limiting large-scale device fabrication and analysis.

The KAIST team exploited the subtle variations in RGB brightness values of MoS₂ flakes under optical microscopy, which correspond to their thicknesses. Leveraging this, they designed an algorithm to automatically classify flakes and precisely design electrodes for transistor fabrication. Atomic Force Microscopy (AFM) validation confirmed that the system accurately distinguished layer thickness differences even within a narrow range of three to eight layers.

Using this approach, the researchers scanned over 120,000 MoS₂ flakes, fabricating and characterizing 1,615 transistors—a scale previously unfeasible through manual methods. This massive dataset enabled the team to statistically reveal a critical and previously elusive relationship: thicker MoS₂ layers facilitate higher current flow but exhibit diminished on/off switching capabilities. Such insights are vital for optimizing device performance yet had remained inaccessible due to limited sample sizes.

Beyond streamlining fabrication, this work introduces a data-driven paradigm to 2D semiconductor research, transforming the process from experience-based experimentation to systematic, large-scale analysis. This shift not only accelerates discovery but also opens pathways for artificial intelligence to aid in the design of novel semiconductor materials with tailored properties.

Published in Advanced Functional Materials, this study marks a significant advance in semiconductor research methodologies. By integrating automated optical recognition with device fabrication, KAIST’s innovation paves the way for rapid, high-throughput screening and accelerates the commercialization prospects of next-generation AI chips and ultra-low-power devices.

As silicon technology approaches its physical limits due to power loss and heat generation, automated approaches like this are essential to unlocking the potential of 2D materials. The collaboration between KAIST, UNIST, Hanbat National University, Hanyang University, and Washington University in St. Louis demonstrates the powerful synergy of interdisciplinary efforts driving the future of electronics.

This achievement heralds a new era where data-driven insights guide materials research, expediting developments that could reshape the landscape of electronics, healthcare sensors, and flexible technologies for years to come.

Subject of Research: Two-dimensional semiconductors and automated device fabrication

Article Title: KAIST Automates the Search for “Dream Semiconductor” 2D Semiconductors

News Publication Date: 8-Jul-2026

Web References: http://dx.doi.org/10.1002/adfm.202532204

References: Statistically Resolving Thickness-Dependent Electrical Characteristics in Multilayer-MoS₂ Transistors, Advanced Functional Materials, DOI: 10.1002/adfm.202532204

Image Credits: KAIST

Keywords

2D semiconductors, molybdenum disulfide, MoS₂, automated screening, transistor fabrication, AI semiconductors, data-driven materials research, optical microscopy, nanotechnology