KAIST Team Shows One SnSe2 Contact Works for Both CMOS Polarities

A single contact material, tin diselenide, raised maximum drive current by more than 1,000 times in p-type monolayer tungsten diselenide transistors compared with conventional nickel electrodes, according to a Korea Advanced Institute of Science and Technology study published in Advanced Materials.

The same SnSe2 contact, paired with n-type monolayer molybdenum disulfide, produced steep switching and an on/off current ratio above one billion. The team, led by KAIST Professor Joonki Suh, calls the structure a universal van der Waals tunneling injector. The paper, "A Universal van der Waals Tunneling Injector for Monolayer CMOS," appeared online on August 12, 2026. KAIST PhD candidate Hanbin Cho was first author. Suh and Professor Kyungmin Ko of Yonsei University served as co-corresponding authors.

Collaborators included Yonsei University, the Beijing Computational Science Research Center in China, the Korea Institute of Science and Technology, Hanyang University, the Ulsan National Institute of Science and Technology, and Samsung Electronics.

Why does one contact for two polarities matter?

Monolayer semiconductor channels can be less than a nanometer thick. Depositing a conventional metal electrode directly onto such a crystal can disrupt the atomic structure and leave an energy barrier that blocks electrons or holes. Because n-type and p-type channels carry current through different carriers, engineers have usually optimized separate contacts and deposition conditions for each.

SnSe2 takes a different route. It joins the channels through weak van der Waals attraction rather than strong chemical bonds, avoiding the damage that metal deposition causes. Electron microscopy showed an atomically sharp SnSe2/MoS2 interface with uniform interlayer spacing of approximately 2.15 angstroms. Surface-potential mapping indicated the SnSe2 contact was more uniform than a directly deposited nickel contact.

The mechanism differs at each junction. In WSe2, SnSe2 forms a type-III "broken-gap" band alignment, letting holes cross the barrier by band-to-band tunneling. In MoS2, an applied gate field narrows the electron-injection barrier so electrons tunnel through it. The shared material therefore does not force both polarities into identical electronic arrangements.

What did the devices actually demonstrate?

• p-type WSe2 transistors: more than 1,000-fold increase in maximum drive current versus nickel contacts.
• n-type MoS2 transistors: steep switching with an on/off ratio above 10^9.
• A CMOS inverter combining both channel types responded reliably to repeated inputs.

The inverter is the study's system-level proof. It shows the universal contact concept works in complementary operation, the basic digital-circuit scheme that uses both channel polarities to invert an input signal.

"This study demonstrates that efficient charge injection, one of the most challenging bottlenecks in monolayer two-dimensional semiconductors, can be addressed using a single material platform," Suh said.

What the study does not show

The release reported no complete processor, no multilayer three-dimensional chip, and no commercial AI accelerator. It did not establish energy savings for such products. The improved devices used ultrathin channels, not bulk materials, and the study did not test complete AI chips or quantify performance in multilayer devices.

Those caveats matter. Two-dimensional semistors attract interest because atomic thickness permits vertical stacking of multiple device layers, fitting more transistors into the same chip footprint. A single-polarity contact strategy could simplify fabrication in such circuits. But the researchers themselves identified direct growth, large-area processing, and device integration as areas still requiring advances before practical implementation of low-power two-dimensional CMOS circuits.

The evidence supports a narrower claim than the headline framing suggests: SnSe2 served as the common injector, its two junctions used distinct tunneling pathways, and both device types showed strong electrical performance. The work does not yet show that replacing metal contacts throughout an AI chip will make it smaller or more energy-efficient.

What the study establishes is a peer-reviewed demonstration that one van der Waals material can contact both major transistor polarities while addressing the interface barriers that have constrained monolayer devices. Whether it scales to manufacturable stacked chips will depend on progress in large-area growth and integration, which remains future work rather than a demonstrated outcome.