Carnegie Mellon researchers have demonstrated atomically thin transistors with channels shorter than 5 nanometers. Their approach tackles one of semiconductor scaling’s most persistent problems: keeping current from flowing when a transistor is supposed to be off.
For decades, shrinking transistors has helped chipmakers pack more computing power into smaller spaces. But a transistor can only get so small before the physics that makes it work begins to work against it.
One of the biggest problems is keeping the transistor switched off.
A transistor controls current through a semiconductor channel using an electric field generated by its gate. As that channel becomes shorter, the source and drain move closer together, making it increasingly difficult for the gate to control the flow of electrons. Current can begin leaking through even when the device is supposed to be off.
That leakage wastes power and generates heat, two problems that become especially troublesome when billions of transistors are operating together.
Researchers at Carnegie Mellon University have demonstrated a way to push transistor dimensions below 5 nanometers while maintaining strong electrical control. Their design combines an atomically thin semiconductor with an air gap that changes how electric fields behave around the channel.
The work, published in Nature Communications and announced October 15, also includes devices fabricated across a four-inch wafer, an important step beyond demonstrating the concept in a single laboratory transistor.
Why Smaller Transistors Become Harder to Switch Off
A field-effect transistor works by using voltage at the gate to control whether current can flow between the source and drain.
In a conventional MOSFET, the gate creates an electric field that changes the conductivity of the semiconductor channel. Apply the appropriate voltage and the transistor turns on. Remove or change that voltage and the channel should stop conducting.
As transistors shrink, however, the source and drain begin to interfere with the gate’s control.
The problem is known as a short-channel effect.
When the channel becomes extremely short, the electric field from the drain can influence the energy barrier that is supposed to prevent electrons from crossing the channel. The gate no longer has exclusive control over whether current flows.
The result is increased leakage current, weaker switching behavior and potentially greater power consumption.
Engineers have spent years developing transistor architectures to manage these effects. FinFETs improved gate control by wrapping the gate around multiple sides of a narrow silicon fin. Gate-all-around transistors take that approach further, surrounding the channel with the gate.
But geometry alone cannot solve every scaling problem.
As channel lengths approach only a few nanometers, the thickness of the semiconductor itself becomes increasingly important. A thick channel makes it harder for the gate to control the entire region where current can flow.
That has encouraged researchers to investigate two-dimensional semiconductors.
An Atomically Thin Alternative to Silicon
The Carnegie Mellon team used molybdenum disulfide (MoS₂), a semiconductor that can be made just one atomic layer thick.
Unlike silicon, which has a three-dimensional crystal structure, MoS₂ belongs to a family of layered materials known as transition-metal dichalcogenides. Individual layers are held together by relatively weak van der Waals forces, making it possible to produce extremely thin semiconductor channels.
A monolayer of MoS₂ is only about 0.7 nanometers thick.
That thinness is useful because the gate’s electric field can influence the channel more effectively. With less semiconductor material to control, engineers can potentially suppress some of the electrostatic problems that become severe in very short silicon channels.
MoS₂ also retains a semiconductor bandgap at monolayer thickness, allowing it to switch between conducting and nonconducting states.
But using a thinner semiconductor doesn’t automatically produce a better transistor.
The surrounding device structure still determines how effectively the gate controls the channel. Contacts, insulating materials and the physical arrangement of the electrodes all influence switching behavior.
That is where Carnegie Mellon’s air-gap design comes in.
The Air Gap That Changes the Electric Field
Air isn’t usually the first material engineers think of when designing a transistor. In this case, however, leaving space beneath the semiconductor is part of the solution.
The Carnegie Mellon researchers developed a structure that incorporates an air gap below the MoS₂ channel. Because air has a much lower dielectric constant than many of the insulating materials used in semiconductor devices, the gap changes how electric fields are distributed around the transistor.
The purpose is to strengthen the gate’s electrostatic control relative to the unwanted influence of the source and drain.
In an extremely short transistor, that balance is critical. The gate must be able to establish a barrier that prevents electrons from crossing the channel when the device is off. If the drain’s electric field weakens that barrier too much, current leaks through.
By combining the air-gap geometry with an atomically thin channel, the researchers were able to maintain switching behavior at physical channel lengths below 5 nanometers.
The approach illustrates an important distinction in semiconductor scaling. Engineers aren’t simply trying to make every dimension smaller. They’re trying to preserve the electrical behavior of the transistor as those dimensions shrink.
Sometimes that means adding a carefully positioned empty space rather than another material.
From a Single Device to a Four-Inch Wafer
Many experimental transistor designs produce impressive results in individual devices but become much harder to reproduce when fabrication moves to larger substrates.
That makes manufacturing consistency an important part of evaluating a new architecture.
The Carnegie Mellon team fabricated its MoS₂ transistors across four-inch wafers, demonstrating that the process could be applied beyond an isolated device.
This does not mean the researchers have developed a commercial manufacturing process. Producing working transistors on a wafer is different from integrating billions of them into a complete processor.
A practical logic technology also requires consistent threshold voltages, low contact resistance, reliable gate dielectrics, high fabrication yield and compatibility with the many other steps involved in semiconductor manufacturing.
Nevertheless, wafer-scale fabrication addresses a question that frequently limits emerging transistor research: whether a promising device structure can be reproduced using a process rather than constructed individually.
For two-dimensional semiconductors, that question is particularly important. Researchers have demonstrated excellent electrical characteristics in carefully prepared flakes, but large-scale electronics require uniform materials and repeatable device performance across an entire wafer.
The Carnegie Mellon work brings the air-gap architecture into that broader manufacturing discussion.
The Next Scaling Challenge Is More Than Channel Length
A transistor with a physical channel shorter than 5 nanometers is an impressive demonstration, but channel length is only one measure of device performance.
Engineers also need to know how much current the transistor can deliver when switched on, how much leaks when switched off and how quickly the gate can change its state.
Contact resistance becomes particularly important in two-dimensional devices. Even when the semiconductor channel has favorable properties, resistance where the metal electrodes meet the material can limit the current available to a circuit.
There are also questions about how the air-gap structure behaves during fabrication and operation.
Any geometry that introduces a small unsupported region must be evaluated for mechanical stability, process variability and compatibility with subsequent manufacturing steps. The complete device must also survive electrical stress and repeated switching without significant degradation.
And a short channel does not necessarily mean a proportionally smaller complete transistor. Contacts, gate structures and spacing requirements all contribute to the footprint of a practical device.
These considerations will determine whether sub-5-nanometer MoS₂ transistors can eventually offer useful advantages over increasingly sophisticated silicon architectures.
Making Room for the Next Generation of Transistors
Silicon remains the foundation of modern computing, and the semiconductor industry continues to find new ways to extend its performance through transistor architecture, materials and advanced packaging.
Two-dimensional semiconductors offer another path, particularly as conventional channels become increasingly difficult to control at atomic dimensions.
The Carnegie Mellon research shows how the properties of an atomically thin semiconductor can be combined with device geometry to address one of the fundamental problems of transistor scaling.
The air gap doesn’t make leakage disappear, and the demonstration is still far from a finished logic chip. But it shows that controlling the electric field around a transistor can be just as important as changing the material carrying the current.
As transistors approach dimensions measured in only a few atoms, the next improvement may come from reconsidering not just what goes into the device, but what engineers deliberately leave out.
Sources
Carnegie Mellon University, October 15, 2026 Beyond Silicon: Researchers Advance Ultra-Short-Channel Transistors
Original university announcement covering the MoS₂ transistor architecture, sub-5-nanometer channels and wafer-scale fabrication.
