World’s Fastest Transistor

World’s Fastest Transistor and could soon be used to make silicon-free chips     

Advances in materials and architecture could lead to silicon-free chip manufacturing thanks to a new type of transistor. The new transistor runs 40% faster and uses less power. The 2D transistor’s design looks like a series of interwoven bridges. 2D bismuth transistors are less brittle and more flexible than transistors made using conventional silicon, the scientists said. Researchers in China say they have created a new silicon-free transistor which could significantly boost performance while reducing energy consumption. The team says this development represents a new direction for transistor research. At Peking University, a group of Chinese scientists may have just turned the computing industry up on its head. With a slender sheet of lab-grown bismuth and an architecture unlike anything inside today’s silicon chips, they’ve built what they call the world’s fastest and most efficient transistor. Not only does it outperform the best processors made by Intel and TSMC, but it also uses less energy doing so. And most important of all, there’s no trace of silicon involved. This two-dimensional, silicon-free transistor represents a radical rethinking of what chips can be and how they can be made. The scientists said that the new transistor could be integrated into chips which could one day perform up to 40% faster than the best existing silicon processors made by US companies like Intel. “If chip innovations based on existing materials are considered a ‘short cut’, then our development of 2D material-based transistors is akin to ‘changing lanes’,” Hailin Peng, professor of chemistry at Peking University and the study’s lead author, said. Despite that dramatic increase in power, the researchers claim that such chips would also draw 10% less power. 

Since the early 1990s, transistors, the tiny switches which drive everything from smartphones to supercomputers, have largely depended on silicon and a design known as the Fin Field-Effect Transistor, or FinFET. These resemble tiny skyscrapers, standing vertically on chips to allow better control of current flow at Nano scale dimensions. But FinFETs are running out of space. As chips shrink to just a few nanometres, engineers are hitting hard physical limits. Below 3 nanometres, performance gains become harder and power consumption rises. Something has to give. So, Peng’s team decided not to shrink the old design any further. They scrapped it altogether. The efficiency and performance gains are possible thanks to the chip's unique architecture, the scientists said, specifically the new two-dimensional silicon-free transistor they created. This transistor is a gate-all-around field-effect transistor (GAAFET). Unlike previous leading transistor designs like the fin field-effect transistor (FinFET), a GAAFET transistor wraps sources with a gate on all four sides, instead of just three. Their new transistor uses a gate-all-around field-effect transistor (GAAFET) structure. Instead of wrapping the gate around three sides of the transistor’s channel like FinFETs do, a GAAFET encircles it on all four. This offers better control of the current and drastically cuts down on wasted energy. But that’s not new. The real innovation lies in what the transistor is made of. At its most basic level, a transistor is a semiconductor device found in every computer chip. Each transistor has a source, a gate and a drain, which allow the transistor to function as a switch. Rather than silicon, the Peking University team built their transistor using bismuth oxyselenide (Bi₂O₂Se) for the channel, and bismuth selenite oxide (Bi₂SeO₅) as the gate material. These materials are part of a class known as two-dimensional semiconductors, atomically thin sheets with exceptional electrical properties. Bismuth oxyselenide, in particular, offers something silicon struggles with at ultra-small sizes: speed.

The gate is how a transistor controls the flow of current between the source and drain terminals and can act as both a switch and amplifier. Wrapping this gate around all sides of a source (or sources, as some transistors contain multiple), instead of just three as in conventional transistors, leads to potential improvements in both performance and efficiency. Electrons move through it faster, even when packed into tiny spaces. It also has a higher dielectric constant, meaning it can hold and control electric charge more efficiently. So it makes it, for faster switching, reduced energy loss, and, very importantly, a lower chance of overheating. “This reduces electron scattering and current loss, allowing electrons to flow with almost no resistance, akin to water moving through a smooth pipe,” Peng explained. The interface between these materials is also smoother than that of common semiconductor-oxide combinations used in industry today. This means fewer defects and less electrical noise. All of this adds up to stunning results. According to the team, their transistor can run 40% faster than today’s most advanced 3-nanometer silicon chips, and it does so while using 10% less energy.

This is because a fully wrapped source provides better electrostatic control (as there is less energy loss to static electricity discharges) and the potential for higher drive currents and faster switching times. There’s a geopolitical current flowing beneath this research. Due to on-going US-led export restrictions, Chinese firms have been blocked from buying the latest silicon chip-making equipment. The most advanced lithography machines, which can manufacture 3-nanometer chips, are made by a handful of companies in the West. By creating a transistor that doesn’t rely on silicon, and which can be fabricated using existing tools in China, Peng’s team may have found a way around those sanctions. “While this path is born out of necessity due to current sanctions, it also forces researchers to find solutions from fresh perspectives,” Peng said. Despite the impressive performance in the lab, big questions remain. Can these transistors be manufactured at scale? Will they survive the heat and stress of real-world computing? And how long will it take for the technology to reach consumer devices?

Peng’s team says they’re already working on scaling up production. Early prototypes of logic units built with the transistor showed ultra-low operating voltages and high voltage gain, two promising signs for integration into actual circuits. The fact that they used existing fabrication platforms also hints that the barrier to mass production may not be as high as with other experimental technologies. While the GAAFET architecture isn’t itself new, the PKU team's use of bismuth oxyselenide as the semiconductor was, as well as the fact they used it to create an "atomically-thin" two-dimensional transistor. “This work demonstrates that 2D GAAFETs do exhibit comparable performance and energy efficiency to commercial silicon-based transistors,” the researchers wrote, “making them a promising candidate for the next technology node.” But turning laboratory breakthroughs into commercial chips typically takes years, sometimes decades. It’s one thing to design a single transistor. It’s quite another to integrate billions of them onto a reliable, manufacturable chip.

2D bismuth transistors are less brittle and more flexible than traditional silicon, the scientists added in the study. Bismuth provides better carrier mobility, the speed at which electrons can move through it when an electrical field is applied. It also has a high dielectric constant, a measure of a material's ability to store electrical energy, which contributes to the transistor’s increased efficiency. Still, if successful, this breakthrough could give China a new technological foundation. And more broadly, it signals that the semiconductor race may no longer be just about who can make smaller silicon but who can think beyond it. Should this transistor be fitted into a chip that does prove faster than US-made chips by Intel and other companies, it could also allow China to sidestep current restrictions on buying advanced chips and tap into US chip-making by shifting to a wholly different manufacturing process.