Could Nanowire Transistors Rescue Moore’s Law from Obsolescence?

Researchers in France suggest they may have found a way to extend Moore's Law through the 2020s after all.

  • Share
  • Read Later
Pgiam / Getty Images

A year ago today I wrote a story titled “The Collapse of Moore’s Law: Physicist Says It’s Already Happening” highlighting theoretical physicist Michio Kaku’s argument that Moore’s Law — the notion that the number of transistors on a computer chip will double every two years — is on the verge of collapse. Kaku put the terminal timeframe for Intel co-founder Gordon Moore’s exponential forecast at roughly 10 years out, thus somewhere in the early 2020s.

For every claim foretelling Moore’s Law’s demise, you can find another pitching its redemption. Like this one noticed by IEEE Spectrum, starring, of all things, nanowires, referring to wiring at the nanoscopic scale — one nanometer is equal to one-billionth of a meter.

Moore’s Law, which has pretty reliably accounted for periodic increases in computer power without chips exploding in size or radically increasing in power consumption, is coming up hard against basic spatial limits and physical laws (that is, actual laws). In order to keep increasingly powerful chips reasonably small and power-sipping, engineers have to cram incredible numbers of transistors into ever-smaller spaces (transistors, now well into the billions per chip, are how a CPU controls current). But you eventually run out of room, and even if you could prolong the shrinking game, you eventually spring current leaks you can’t insulate around — that, and once you’re down to atom-sized transistors, it’s pretty much everyone out of the pool, time to go home.

Well, unless we come up with some seemingly unorthodox workarounds. Take Intel’s tri-gate technology (the company’s so-called “3D processors”), which offers a stop-gap measure by stacking “gates” (electric circuits) to give electrons a threefold increase in space, reducing leakage, increasing efficiency and reducing power consumption dramatically. Even then, as Kaku notes, “there is an ultimate limit set by the laws of thermal dynamics and set by the laws of quantum mechanics as to how much computing power you can do with silicon.”

But what if we could fundamentally change how the transistors themselves were designed, such that we could enjoy all the benefits of them being transistors but with far better ability to control messy current?

Such a solution exists in the form of something called a nanowire field-effect transistor (FET), also known as a “gate-all-around” transistor due to its manipulation by a single, surrounding gate (gates act as “switches” for current). Individual nanowires are too small to effectively carry current, so scientists came up with the idea of jamming a bunch together in something like a wiring mesh. The problem? Size. Even at the nanoscale, you have to come in with extremely small gates if you’re going to produce nanowire FETs capable of operating within Moore’s Law’s strictures by the early 2020s.

Enter researchers in Lille and Toulouse, France, who report they’ve created a nanowire transistor capable of satiating Moore’s Law’s relentlessly exponential appetite. According to IEEE Spectrum:

It consists of an array of 225 doped-silicon nanowires, each 30 nm wide and 200 nm tall, vertically linking the two platinum contact planes that form the source and drain of the transistor. Besides their narrowness, what’s new is the gate: A single 14-nm-thick chromium layer surrounds each nanowire midway up its length.

That single gate is where the magic happens, according to Guilhem Larrieu with the Toulouse-based Laboratory for Analysis and Architecture of Systems: “The advantage of an all-around gate allows the creation of shorter gates, without loss of control on the current through the channel. We demonstrated the first vertical nanowire transistor with such a short gate.”

Other upsides of this approach, according to the research team’s paper (published in the journal Nanoscale) include “better immunity to short channel effects, reduction of device-to-device variability, and nanometer gate length patterning without the need for high-resolution lithography.”

No, it’s not a total game-changer in that you’d simply be extending and not saving Moore’s Law (that’s also assuming the idea can travel all the way from the research to the manufacturing stage), but it’s a reminder that putting precise timeframes around the demise of remarkably prescient observations, such as the one Gordon Moore made all the way back in 1965, is as much an article of faith as the “law” itself.