Lab-made 2D transistor switches below 60 mV decade⁻¹

Inside a laboratory at The Hong Kong Polytechnic University, researchers stacked ultrathin layers of bismuth and indium selenide to make a transistor behave differently from the silicon devices that dominate integrated circuits. The resulting 2D tunneling field-effect transistor switched at room temperature with a subthreshold swing below 60 mV decade⁻¹, the room-temperature limit that constrains conventional MOSFETs.

That limit comes from how a conventional transistor moves electrical charges: they pass over an energy barrier, a process known as thermionic emission. The PolyU design uses quantum tunneling instead, allowing charges to pass through the barrier. By controlling the nanoscale structure of alternating bismuth and indium selenide layers, the team created an energy-band alignment intended to make that tunneling efficient. In its two-dimensional form, bismuth becomes a semiconductor rather than behaving as a semimetal, according to the researchers.

The device operated with a gate-voltage range of 160 mV, far below the 800 mV required by the advanced MOSFETs cited in the study. Its subthreshold swing remained below the thermionic limit across six orders of magnitude of current switching. The transistor also delivered up to several microamps per micrometer and a high ON/OFF current ratio—enough current, the team says, to drive multiple downstream logic gates and limit circuit delay.

The work was led by Prof. Jianhua Hao and Dr. Zehan Wu, with researchers from the National University of Singapore, The Hong Kong University of Science and Technology, Peking University, and the Singapore University of Technology and Design. The device was fabricated on standard centimeter-scale silicon substrates, and the team used pulsed laser deposition to demonstrate a possible route toward wafer-scale production of 2D materials.

So, concretely, the result offers chip designers a laboratory-tested transistor architecture that could reduce switching voltage and energy in future integrated circuits, including specialized AI hardware. It does not yet show a mass-produced processor: the reported achievement is a device experiment, and the manufacturing route still has to move from demonstrated viability to dependable chip production.