Carbon Nanotube Transistors: The Material That Could Take Over When Silicon Runs Out of Room

The transistor is the foundational unit of modern computing. Intel's 4004 processor from 1971 contained 2,300 transistors. A modern Apple M4 chip contains approximately 28 billion. The doubling of transistor density roughly every two years — Moore's Law — has driven five decades of exponential improvement in computing performance and cost. And it is now, unmistakably, running into physical limits.

At the 2nm node (a number that is more a marketing label than a literal measurement, but corresponds to gate lengths approaching 10–15nm with current FinFET and Gate-All-Around architectures), silicon MOSFETs face quantum mechanical constraints that cannot be engineered around indefinitely. Electrons tunnel through barriers. Thermal noise at room temperature becomes a fundamental limit. Dopant atoms must be placed with atomic precision.

Carbon nanotubes (CNTs) offer a materials path forward — one that has been tantalising researchers since their electrical properties were characterised in the early 1990s.

## Why Carbon Nanotubes Are Interesting for Transistors

A single-walled carbon nanotube (SWCNT) is a graphene sheet rolled into a cylinder with a diameter of roughly 1–2nm. The electrical properties depend critically on the **chirality** — the angle at which the graphene lattice is rolled. Roughly one-third of SWCNTs are metallic (conducting like a wire); two-thirds are semiconducting.

Semiconducting SWCNTs have several theoretical advantages over silicon as a transistor channel material:

- **Electron mobility** of 100,000 cm²/V·s versus ~1,400 for silicon — roughly 70x higher in ideal conditions
- **Ballistic transport**: electrons can traverse short CNT channels without scattering, meaning the channel can carry current without the resistive losses of silicon
- **Atomically thin body**: the 1nm diameter means the gate electrode can electrostatically control the channel with near-perfect efficiency at short gate lengths, reducing short-channel effects that plague silicon at sub-5nm dimensions
- **Low power**: the high electrostatic control enables steep subthreshold slopes, meaning the transistor switches from off to on with less voltage — directly reducing dynamic power consumption

In simulation and in carefully fabricated individual CNT transistors, these advantages are real. The question has always been whether they can be realised in a manufacturable array.

## The Chirality Problem

Here is the fundamental difficulty: carbon nanotube synthesis produces a mixture of metallic and semiconducting species. A single metallic CNT in a transistor channel creates a conduction path that never turns off — the transistor is effectively broken. For logic circuits, you need semiconducting purity approaching 99.9999% or better.

Sorting SWCNTs by electronic type has been a research focus for over two decades. The leading approaches include:

- **Density gradient ultracentrifugation**: separates by buoyancy differences between species in a density gradient medium
- **Gel chromatography**: selective adsorption of semiconducting SWCNTs to gel columns
- **Aqueous two-phase extraction**: partition between polymer phases based on CNT surface chemistry
- **DNA wrapping**: specific single-stranded DNA sequences selectively wrap particular CNT chiralities, enabling purification by chromatography

By 2024, research groups at MIT, Stanford, and IBM had demonstrated CNT solutions with 99.99%+ semiconducting purity. IBM's 2024 demonstration chips used sorted, deposited CNT arrays as the channel material in ring oscillators and logic gates, achieving switching speeds competitive with the best silicon node equivalents in the same geometry.

## The Alignment and Density Challenge

Sorted CNTs must be deposited onto a wafer substrate in **aligned, high-density arrays** — all pointing in the same direction, with sufficient density (ideally >100 CNTs/µm) to carry the required current. Aligned deposition techniques include:

- **Dielectrophoresis**: AC electric fields orient CNTs in suspension during deposition
- **Langmuir-Blodgett assembly**: CNT monolayers formed at a water-air interface and transferred to substrate
- **Chemical epitaxy**: quartz substrate surface chemistry aligns CNTs during growth by chemical vapour deposition

Each approach achieves alignment but struggles with density and uniformity at wafer scale. IBM's 2024 demonstration chips had CNT densities of roughly 200 CNTs/µm — sufficient for proof-of-concept but below the 500+ CNTs/µm that would be needed for competitive logic circuits.

## Comparison with GaN and SiC

Wide-bandgap semiconductors — gallium nitride (GaN) and silicon carbide (SiC) — are the most commercially mature alternatives to silicon for specific applications. GaN excels in power electronics and RF (5G base stations, EV power converters). SiC dominates in high-voltage power switching (EV inverters, industrial drives). Neither is positioned as a replacement for logic transistors.

The CNT value proposition is specifically for **low-power logic** — the kind of computing done in processors, AI accelerators, and memory controllers — not for power electronics. In this space, CNT transistors compete with silicon at sub-3nm nodes, not with GaN or SiC.

## Timeline to Commercial CNT Logic

The consensus among researchers in 2026 is that CNT logic is 8–15 years from commercial chip production, assuming current rates of progress. The bottlenecks are wafer-scale uniformity, defect density (a single defective CNT in a circuit cell causes a fault), and integration with existing CMOS back-end-of-line processes.

Intel, TSMC, and Samsung are all running internal CNT research programmes. Academic spinouts (Carbonics in the US, and several groups in China) are building pilot fabrication lines. The trajectory is real — but the distance between a demonstration chip and a volume-manufactured logic node remains substantial.

Silicon will not run out of room tomorrow. But the roadmap has a visible end, and carbon nanotubes are the most technically credible candidate to extend it.

Carbon Nanotube Transistors: The Material That Could Take Over When Silicon Runs Out of Room

// COMMENTS

ON THIS PAGE