Polarity-controllable FETs offer a different approach to transistor design: rather than fixing a device as n-type or p-type through fabrication, they use an additional gate to select which carrier type conducts. Research on two-dimensional (2D) semiconductors such as tungsten diselenide (WSe2) suggests that this electrostatic approach could support compact devices while reducing reliance on precise dopant placement. The results are promising, but they also highlight the process and integration challenges that must be addressed before the technology can move beyond research demonstrations.
Why consider polarity-controllable FETs?
Conventional CMOS logic relies on complementary n-type and p-type transistors. As device dimensions shrink, controlling abrupt doping profiles becomes more difficult. Variations in the number of dopant atoms in a channel can also shift transistor threshold voltage, contributing to device variability.
A polarity-controllable device aims to avoid setting carrier type through channel doping. It instead exploits the ambipolar behavior of a Schottky-barrier field-effect transistor (SB-FET). In an ambipolar transistor, the gate voltage determines whether electrons or holes are injected into the channel. Although this behavior is generally suppressed in standard CMOS devices, it is central to polarity control.
The device uses two independently controlled gates. A conventional control gate (CG) switches the transistor between on and off states. A polarity gate (PG) changes the Schottky barriers at the source and drain, selecting electron or hole conduction. In principle, the same device can therefore operate as either an n-type or p-type FET, with its polarity selected during operation rather than fixed during fabrication.
What the WSe2 experiment demonstrated
Experimental devices were fabricated using mechanically exfoliated, multilayer WSe2, approximately 7.5 nm thick. Buried metal lines formed the polarity gates, while the silicon substrate served as the control gate. The metal contacts were selected to support injection of both carrier types.
With the two gates operated together, the device showed ambipolar behavior. When the gates were controlled independently, changing the PG bias selected the transistor polarity, while the CG controlled switching. The same device showed p-type behavior at negative PG bias and n-type behavior at positive PG bias. Reported on/off current ratios were 106 for p-type operation and 107 for n-type operation.
These measurements establish polarity switching in the demonstrated WSe2 device. They should be distinguished from the performance projections discussed below, which came from quantum-transport simulations of scaled devices.
Materials and manufacturing remain central challenges
Early work on 2D materials often uses mechanical exfoliation, which can produce high-quality flakes but does not provide a scalable, uniform manufacturing method. Flake dimensions, thickness, folds and cracks can vary, making it difficult to establish consistent device behavior across a production-relevant population.
Research into larger-area material production includes chemical vapor deposition (CVD), van der Waals epitaxy and metal-organic CVD. These approaches have demonstrated growth of several transition-metal dichalcogenides, including MoS2, WSe2 and WS2. The source reports wafer-scale growth of monolayer MoS2 and WS2 using MOCVD, with a device yield of 99%, while noting that the process took several hours per wafer. This illustrates both the progress in film uniformity and the need to consider throughput.
Transfer processes are another part of the integration problem. A 2D film may be grown on a substrate suited to material formation but unsuitable for standard electronics. Moving the film to a desired substrate could separate growth from device fabrication. However, some high-quality CVD processes require temperatures above 800°C, whereas monolithic 3D integration on top of CMOS calls for processes below 350°C. High-temperature annealing associated with doping can create an additional obstacle. Electrostatic polarity control is attractive partly because it may reduce the need for such doping steps.
Scaling potential in quantum simulations
Self-consistent ballistic quantum-transport simulations were used to examine how polarity-controllable devices might behave at very short gate lengths. The analysis modeled source and drain contacts without dopant implantation and assumed mid-gap Schottky barriers to produce symmetric n-type and p-type behavior. Results therefore describe modeled structures and assumptions, not measured performance in manufactured ultra-scaled devices.
For bilayer WSe2, the simulations produced on-state current around 300 µA/µm and on/off ratios above 106 down to a 4 nm gate length. The study also examined a modeled 2D material with a bandgap of approximately 0.8 eV. Its lower Schottky barrier was associated with higher simulated on-current, around 1.5 mA/µm, and on/off ratios above 105 down to a 5 nm gate length for both polarities.
The comparison points to bandgap as an important design consideration. A lower bandgap can reduce the contact barrier and improve carrier injection in this device concept. But material selection involves balancing current, off-state control and the feasibility of producing suitable films; simulation alone does not resolve those manufacturing questions.
Circuit-level implications
The device projections were also used in a benchmarking study of a 32-bit ripple-carry adder. The proposed double-independent-gate FET logic uses compact three-input XOR and majority gates, taking advantage of the independently controlled terminals. Its adder architecture also computes carry and inverted-carry signals in parallel.
In that modeled comparison, a 2D MX2-based implementation achieved an energy-delay product (EDP) 7.2 times lower than the high-performance CMOS reference. The result reflects both the device assumptions and the selected logic architecture. It is a circuit-level benchmark, not evidence of a production-ready processor or a guaranteed advantage in a particular control system.
Practical implications
For engineers and technology decision-makers, polarity-controllable FETs are a research direction worth tracking, particularly where device scaling, variability and future integration options are strategic concerns. The work demonstrates polarity selection in WSe2 and identifies promising simulated device and circuit behavior. At the same time, scalable film growth, transfer, process temperature, uniformity and validation of scaled-device performance remain significant requirements.
The broader lesson is that transistor architecture, material growth and circuit design must be evaluated together. Electrostatic polarity control could simplify the definition of n-type and p-type behavior and enable logic structures that use additional gate functionality. Realizing that potential will depend on translating promising device demonstrations and simulations into repeatable, integrable processes.