In the technology world, few things have been given more premature obituaries than Moore’s Law. But for traditional semiconductor transistor technology, an end will eventually be reached—the question is simply when that will be. Researchers at Australia’s University of New South Wales haven’t written an obituary for Moore’s Law, but they may have better envisioned its end than have any other scientists thus far. Their recently revealed single-atom transistor could foreshadow the smallest possible process technology that can be achieved using the basic semiconductor approach that has powered computer development in recent decades.
The Usual Disclaimer
The Data Center Journal regularly reviews new developments in science and technology that could have a potential impact on IT and the data center. And every such development is accompanied by a warning against overenthusiasm: yes, the potential implications of the development are usually exciting, but many of these developments actually have no future. (How much research, although conceivably revolutionary, is never heard of again after initial reports of some success or breakthrough?) But that’s par for the course in research: it’s a gamble that a particular avenue of inquiry will yield some results that are beneficial both technologically and financially. Many of these avenues are dead ends, but some aren’t—and that’s what makes keeping an eye on progress so interesting.
Reaching the End of an Impressive Road
Shrinking a semiconductor process technology yields a number of benefits: it reduces power consumption for a given performance level, it enables faster processors (components are closer together, meaning signals have less transit time between them) and it allows more circuitry to be packed into a smaller area. The fundamental limit, given the underlying approach of modern semiconductor manufacturing, is the size of the atoms that compose the transistors on a chip; you can’t get any smaller than that. (Yes, there are smaller particles, but that’s another matter—no pun intended.)
Thus, the size of the atoms in a semiconductor device represents the limit of semiconductor technology, barring any unforeseen development. And researchers at the University of New South Wales have realized this limit: a single-atom transistor in which an atom is positioned “perfectly”—that is, without the location uncertainty that has plagued previous prototypes of this kind. Martin Fuechsle, research fellow and lead author of the paper describing the device, states, “We have, with atomic precision, positioned this atom within our device.” (For a good, brief layman’s overview of the device, see the video “Tiny Transistor.”)
In addition to envisioning the limits of traditional semiconductor manufacturing, this device takes another potential step toward realization of a quantum computer, which is the hoped-for next generation of computing technology. (And for a nice layman’s overview of quantum computing, see “Quantum Computing.”) Even quantum computing has its theoretical limits (“Scientists Find Fundamental Maximum Limit for Processor Speeds”), but the progress of computer processing in a new quantum paradigm would likely kick off a “new Moore’s Law”—or would just breathe new life into Moore’s Law as it stands now. Sounds exciting! Now, for the reality check.
But What’s the ‘But’?
The single-atom transistor implemented by the researchers in Australia has at least one major drawback: the precise positioning of the atom, to be maintained, requires the device be held at a temperature of –391°F (that’s –235°C, or just 38°C from absolute zero)—colder even than liquid nitrogen, which has a boiling point of –321°F. The need to maintain a device at significantly lower than room temperature (in particular, beyond what you can get in a typical refrigerator) puts the device beyond the reach of consumer applications for now. And if you think cooling is a problem in the data center now, imagine what it would be if processor chips had to be kept really cold!
This is not to say that this single-atom transistor has no potential consumer or commercial use, but it suffers from the same problems that are stifling many avenues of research, such as practical superconductivity and even quantum computing: the need to maintain extremely low temperatures for proper functioning. The two possible solutions are to implement prototypes that can operate at more reasonable temperatures and to make extremely low temperatures more ubiquitous (e.g., a cylinder of liquid helium in every household—okay, that’s a little facetious, but it illustrates the problem with this latter route).
If it turns out that low temperatures are an inherent requirement of, say, quantum computing, then these futuristic processing models may always be limited to a few very expensive, very impractical models in the halls of research.
Moore’s Law: Not So Easy to Pin Down
Depending on whom in the press you ask, the single-atom transistor is either an outworking of Moore’s Law or a deathblow to Moore’s Law. Even the precise definition of Moore’s Law varies depending on who is describing it, but it basically says that the number of transistors on a manufactured semiconductor chip doubles every two years; that’s a compound annual growth rate (CAGR) of about 40%.
But a single-atom transistor means we can pack a ton more transistors on a chip, blowing away Moore’s Law, right? Not so fast. This development has potential, but it is not a realization of such a chip. For one, chips produced by Intel, GlobalFoundries and other fabs operate at room temperature, which by itself makes a comparison moot. If the economics of a process technology make it impractical, it arguably has no standing in comparison with those that consumers and companies are relying on for actual products.
This is not to say that the single-atom transistor will have no future effect on progress; it may simply arrive in practical form just in time to keep Moore’s Law going a little longer. But then again, temperature limitations may prevent it from ever reaching a home desktop or even a data center. Either way, however, this new transistor represents a possible way in which Moore’s Law might live another day; as it stands presently, it really has no impact on Moore’s Law because it isn’t really comparable to traditional semiconductor manufacturing and use. It must, like other new developments, make its slow way from prototype to practicality.
The Takeaway
The new single-atom transistor developed at the University of New South Wales is an admirable feat with two possible interpretations: it could represent a new age of computing, being an important step toward a practical quantum computer, or it could simply be a glimpse of the end of progress in semiconductor technology. Almost everyone agrees that the progress envisioned by Moore’s Law will eventually end; the question is simply when that will happen. For all practical purposes, the single-atom transistor may represent that end: the point at which software companies will have to think about de-bloating their products because the hardware can no longer support their endless expansion.
A balanced view is required as much here as anywhere. Overwhelming exuberance over technological innovation is as ridiculous as complete dismissal of progress. Nevertheless, eventually physical limits are reached and progress stops: it’s not unreasonable to expect that one day, Intel won’t offer a fundamentally faster computer compared with the previous year’s model. Computer technology has seen astounding progress over the past 50 or so years (or a little more or less, depending on your perspective), but there’s no reason to believe it will continue forever.
In the meantime, it’s still fun to keep an eye on new developments that hope to keep the Moore’s Law train moving a little further down the track.
Author contact
Photo courtesy of jma.work.
About Jeff Clark
Jeff Clark is editor for the Data Center Journal. He holds a bachelor’s degree in physics from the University of Richmond as well as master’s and doctorate degrees in electrical engineering from Virginia Tech.
One Comment