The amazing technological progress of the past century (or so) has led to a vigilant search for the next technology that will create the kind of revolution that sprang from the first silicon transistor built at Bell Labs in the late 1940s. A research project at the University of Pennsylvania has produced optical switches using nanowires—in essence, these switches are the photonic (i.e., light-based) analog to semiconductor switches, which rely on electronic signals.
Ever in Search of the Next Great Thing
In the realm of computing, the current media favorite is likely quantum computing. Quantum computing has a certain exotic flavor that tends to leave other alternatives to traditional semiconductor-based computing in the proverbial dust. But quantum computing faces a number of fundamental problems that may or may not prove to be intractable (“Is Quantum Computing real?”). In any event, the potential for quantum computing beyond a few niche applications is questionable (although any number of developments—theoretical or practical—could render such speculation moot).
But disentangling ourselves from the subject of quantum computing reveals that a number of other, albeit less exotic, avenues of research also present interesting possibilities for the future of computing. As Moore’s Law continues toward the end of its impressive road—when the size of semiconductor components becomes so small that electrons begin “leaking” from their channels—researchers are pursuing alternatives that, they hope, will one day be as revolutionary (or nearly so) as the first transistor. One possibility is photonics: the use of photons as the means of carrying signals instead of electrons.
Researchers at the University of Pennsylvania have taken a step toward showing that photonic computation may be a possibility. The project has led to the creation of “the first all-optical photonic switch out of cadmium sulfide nanowires. Moreover, they combined these photonic switches into a logic gate, a fundamental component of computer chips that process information,” notes Penn News (“Penn Researchers Make First All-optical Nanowire Switch”). Of particular interest is the formation of a NAND gate using these photonic switches; NAND gates can be used to construct any logic operation and thus are in some sense the most basic element of digital computer processing.
According to lead researcher Ritesh Agarwal, “We see a future where ‘consumer electronics’ become ‘consumer photonics,’” and “this study shows that is possible.” But what’s the real story?
The Need for More Work
The inevitable reaction of many at the announcement of a new research development is projection of the technology into the consumer space. After all, what good is it if it doesn’t produce something consumers can consume? This isn’t to assign overexuberance to Agarwal—he may simply have been pointing out that his group’s development potentially has consumer applications. But just as when a development in the world of quantum computing is reported, some restraint is necessary.
Universities regularly churn out useless research (not everyone with a doctorate can possibly have done earth-shattering work). Occasionally, developments at universities reach semi-technical and popular-level media outlets. The question, however, is whether the development will amount to anything useful. In the case of the nanotube-based optical switch implemented at the University of Pennsylvania, the technology may be similar to the first transistor: a proof of concept that eventually leads to a revolution that permeates the consumer space, or just a curiosity that cannot or will not be applied to many (if any) practical purposes.
In some cases, the limits are technical. The laws of physics may prohibit practical application of a technology in a manner that rivals existing technologies. If photonics cannot do significantly better—that is, provide faster processing capability at a similar cost—then it probably won’t be pursued very far. The limits can also be economic. If the research, implementation or use of a technology is prohibitive (or even just more expensive) relative to competing technologies, a new product or idea may never reach its potential.
To some extent, chance is also a factor. A development with significant potential may simply be passed by, or preconceptions about what’s possible may prevent enthusiasm for further research. Thus, any fledgling technology has a number of hurdles to jump before it can even be considered for commercial products.
The Appeal of Photonic Devices
The usefulness of photonics relative to electronics is apparent in the comparison of copper cabling versus fiber. Fiber cables, owing to the much higher-frequency transmissions, can carry much more information than copper cables, which rely on electronic signals. Furthermore, fiber enables transmission over longer distance for a given amount of signal loss. But the bulkiness of optical modules limits the practicality of fiber for many applications. Research at Intel, for instance, is seeking to make fiber interconnects more practical at smaller scales, as one of the major limiting factors of computers today is bandwidth between devices (such as CPU cores).
This area of research, called silicon photonics, seeks to embed optical components in silicon chips, integrating electronics and photonics in the same device. Two years ago, Intel announced “a research prototype representing the world’s first silicon-based optical data connection with integrated lasers. The link can move data over longer distances and many times faster than today's copper technology; up to 50 gigabits of data per second” (“Intel Milestone Confirms Light Beams Can Replace Electronic Signals for Future Computers”).
Application of photonics to connect devices (fiber optics) and to make connections between components within a device (silicon photonics) leads to the next logical step: using photonics as the elements of the devices themselves—such as processors. This is where the University of Pennsylvania research makes headway. If a photonic switch, and thus photonic gate, can be mass produced—and here’s the key—densely, relatively inexpensively and at nanoscales, then in combination with fiber optics and other photonics-based connectivity, it could be the basis for a photonic computer.
Obviously, myriad technical dilemmas—up to and including how to manufacture products economically—must be pursued. In that sense, the photonic switch is similar to the first transistor. The state of the art didn’t jump from a single (rather ugly) transistor to 22nm FinFET process technologies overnight, and the same will be true with photonics, assuming a photonic computer is truly possible.
As The Data Center Journal has noted numerous times in the past, sobriety is called for when evaluating new research developments. Nevertheless, although the future of photonics in computers is uncertain, it does present an avenue of research that seems for the present to be more promising for widespread applicability than, say, quantum computing. The creation of a photonic NAND gate by University of Pennsylvania researchers is an interesting stride toward a photonic computer. But perhaps what such research directions really need most is an end to Moore’s Law for traditional silicon. When billions of dollars are invested in developing newer types of computing devices instead of just smaller silicon process technologies, then we may find more quickly which directions hold real promise.
Photo courtesy of MikeBlogs