A carbon-nanotube transistor result can describe a single switching device, a radio-frequency measurement, a working logic circuit or a fabrication process. Each answers a different question. Read the device structure, test conditions and demonstrated scale before comparing a headline with a silicon processor or predicting a manufacturing transition.
Start with the device being measured
A field-effect transistor uses a gate to control current through a channel between source and drain. In a carbon-nanotube field-effect transistor, semiconducting nanotubes form that channel. The contacts, gate dielectric, tube arrangement and surrounding interconnect all affect the measured result. A complete device specification therefore describes the surrounding structure as well as the nanotube.
Ask whether the sample uses one tube or an array, how channel length and width are defined, and which voltage and temperature conditions apply. Current normalized per unit width can be useful, but it depends on what the width represents. Comparisons should also account for contact resistance, parasitic capacitance and whether a reported value is measured directly or extracted from a model.
On narrow screens, focus the table and use the arrow keys or swipe to see every column.
| Evidence level | What it can establish | What still needs evidence |
|---|---|---|
| Individual transistor | Switching or gain under stated conditions | Variation, integration and useful circuit behavior. |
| Logic or analog circuit | Several devices working together for a task | Scale, reproducibility, power and reliability. |
| Wafer-level process | Fabrication across a stated area or process flow | Yield distribution, long-term behavior and production economics. |
| Manufactured product | A qualified implementation for a defined use | Its measured advantages for the relevant workload. |
Read dated research examples
A 2019 Nature paper reported a microprocessor built with more than 14,000 carbon-nanotube field-effect transistors. Its 16-bit implementation executed 32-bit RISC-V instructions with 16-bit data and address paths. That establishes integrated programmable logic; comparison with a desktop processor would require a shared workload and measurement method.
A 2020 Nature Electronics paper reported carbon-nanotube transistor fabrication in commercial silicon manufacturing facilities, including processing on 200 mm wafers. That addresses an important integration question: whether a process can be implemented in such facilities. It does not by itself establish a high-volume shipping processor, its cost or its lifetime reliability.
A 2025 Nature Electronics radio-frequency study reported extrinsic cutoff and maximum oscillation frequencies of 551 GHz and 1,024 GHz for 35 nm Y-gate devices. It also described a 30 GHz amplifier with 21.4 dB gain using 50 nm devices. These are specific RF device and circuit results, not a claim that a general-purpose CPU runs at a terahertz.
A January 2026 Nature Communications study investigated gamma-ray treatment to improve the nanotube–dielectric interface and reduce off-state leakage. It reported an on/off current ratio around 100,000 for the studied devices. This establishes results for a processing technique and device behavior. Product-level yield, lifetime and application benefits require their own measurements.
This selection illustrates different levels of research evidence. Keeping the publication date, device geometry and demonstrated circuit beside the headline makes progress easier to assess without treating unlike metrics as substitutes.
Keep switching, gain and computing performance distinct
For RF transistors, cutoff frequency and maximum oscillation frequency characterize small-signal behavior under defined conditions. A digital processor’s useful work depends on a complete circuit, clocking, memory, interconnect, power constraints and software. A large number attached to one device cannot be substituted for instructions per second or application throughput.
For a circuit demonstration, ask what inputs were applied, what outputs were verified, how long it operated and how power was measured. For a claimed improvement, identify the baseline and whether both implementations were measured under comparable conditions. An especially good device and a typical device population can tell very different stories.
Look for the evidence needed to scale
- Semiconducting material purity and control of unwanted conducting paths.
- Tube placement, alignment and density across the intended circuit area.
- Contact and threshold variation, with distributions rather than only best results.
- Compatibility with interconnect, dielectrics and thermal processing.
- Circuit yield, aging, environmental stability and repeatability across lots.
- A useful system-level benefit after fabrication and integration costs are included.
These are evaluation questions, not a declaration that every research process has the same unresolved problem. Read the methods and supplementary information to see which were tested. A credible technical summary states what was demonstrated, under what conditions, and which conclusions remain outside the experiment.