The transistors at the heart of every computer, today numbering in the billions on a single chip, have generally been based on the concept John Bardeen, Walter Brattain and William Shockley first turned into a prototype at the Bell Labs in 1947. Physicists have now demonstrated a radically simpler transistor design, first patented by Austrian physicist Julius Edgar Lilienfeld in 1925 but never turned into a practical device until now. This simpler version could push computers to become faster and to consume less power.
In every transistor, an electrode, called the gate, governs whether current can run along a semiconductor strip, thereby defining an on or off state essential to a computer’s binary function. Traditionally, the semiconductor strip is structured like a sandwich, with one type of material between two layers of another type. In the “off” position the sandwich acts as an electrical insulator, but the gate can turn it into a conductor, typically by creating an electric field. In chip fabrication the sandwich is obtained from a strip of silicon “doped” with other elements. For example, the middle section can be created by adding in atoms that tend to hog extra electrons; the side sections get atoms that tend to give electrons away. Each section separately could conduct electricity, but electrons will refuse to move across the middle section unless the gate is turned on.
The boundaries between consecutive sections are called junctions. As transistor size shrinks, it is becoming a challenge to produce sharp boundaries where doping concentrations change abruptly over distances of just nanometers, says Jean-Pierre Colinge of the Tyndall National Institute in Ireland.
A solution, then, is to eliminate those boundaries. Following Lilienfeld’s idea, Colinge and his team have built a transistor with one type of doping only and thus no junctions. The new device is a one-micron-long nanorod of heavily doped silicon, with the gate crossing over its middle section. An electric field from the gate turns the transistor off by depleting that middle section of its electrons, preventing the flow of current through the rod. The team describes its result in the March Nature Nanotechnology (Scientific American is part of Nature Publishing Group).
An effective depletion of electrons requires the rod to be just 10 nanometers thick, a feat that has only recently become possible in large-scale manufacturing. “The device should be able to be integrated in silicon chips quite readily,” because it is compatible with existing fabrication processes, Colinge says. The junctionless design is more effective at switching currents on and off, he says, which means it could work at lower voltages, producing less waste heat, and at faster speeds. (After increasing for decades, computer clock speeds have essentially been stuck at about three gigahertz for the past several years.)
Thomas Theis, director of physical sciences at the IBM Watson Research Center in Yorktown Heights, N.Y., says that the junctionless transistor could be useful if the authors can make it much shorter than the current one micron, to better match existing components. Colinge says that shrinking it down to 10 nanometers should be feasible, and his team is working on getting there. Since the paper’s publication, Colinge says, several semiconductor companies have shown interest in the transistor, perhaps getting ready to give new meaning to a future that has no boundaries.
Already a Digital subscriber? Sign-in Now
If your institution has site license access, enter here.
See what we're tweeting about
10 Comments
Add CommentYour description of a transistor only fits bipolar transistors. Field effect transistors use a depletion area, much like how you describe this old/new design. I'd expect better from a technical publication.
Reply | Report Abuse | Link to thisSomebody is having their leg pulled here. The described Field Effect Transistor became the predominant type used in computers over 30 years ago. I think this writer needs to be a little more knowledgeable about this field if he is going to write about it.
Reply | Report Abuse | Link to thisI was thinking the same thing. Only Bipolar Junction Transistors (BJTs) and Junction Field Effect Transistors (JFETs) have junctions. These types are mostly used in analog domains. The most widely used transistors (in the digital domain at least) are MOSFETs, which don't have junctions. What am I missing?
Reply | Report Abuse | Link to thisThe researcher mentioned has investigated about nanowire transistors...
Reply | Report Abuse | Link to thisttp://www.tyndall.ie/research/ultimate-silicon-devices/Research/physics.html
BJTs, MOSFETs both have junctions... BJTs are, in a basic form, p-n-p or n-p-n.... MOSFETs are the same, the drain and source electrodes are connected to a very tiny (in comparison to the gate) p or n material while the gate is the opposing type, n or p.
Reply | Report Abuse | Link to thisThe description is poor.... They are mixing the explainations of two different types of transistors.
It's an insulated gate FET
Reply | Report Abuse | Link to thisMicrosoft just filed the patent on the "page-turn-gesture".
Reply | Report Abuse | Link to thisYeah, they "invented" it, right after they did that "gravity" thing.
I agree completely!
Reply | Report Abuse | Link to thisIt sounded like the old model was a FET and the new a BJT (but without junctions?), whereas it's actually just a FET where the source/drain are doped identically to the bulk film, and the gate electrode is counter-doped (and the size of the gate is really important, apparently)...
The author should have specified we were talking about FET's (and not transistors in general) at the beginning of the article...
So, is this something new and paradigm changing that could result in a supercomputer in every home, or, is it an incremental step based on an older design? I'm a layperson who wouldn't know my n from my p if I dropped it on my foot.
Reply | Report Abuse | Link to thisIt's been a while since I've looked at my solid state text, so correct me if I'm wrong, but the physics of the BJT and FET are basically the same when it comes to carrier transport. Just like there's drift and diffusion in a PN junction, there's drift and diffusion through the depletion region of a FET. Besides, depletion regions only form when you have a junction, so common sense tells you FETs have junctions (and PN diodes have depletion regions).
Reply | Report Abuse | Link to thisActually, I'm a little confused about how a nanorod device would NOT have a junction. If they're growing a heavily doped silicon nanorod in a silicon matrix (through MOCVD or something), then there will be a junction between the rod and its surrounding matrix. They will have found a way to make a very well-defined junction as opposed to the ambiguous junction made using ion implantation, but there will still be a junction.