Chip-level advances that may change computing

1 2 Page 2
Page 2 of 2

Next, Tabula plans to target the chips found in everyday mainstream electronics, such as digital cameras, video games and maybe even full computers. The company's current eight-level design is in production, and Tabula is well on the way to creating a 12-layer version, with a 20-level chip on the drawing board. "There's no limit to the number of levels we can integrate," Tieg says.

From soot to circuits: Graphene

For the past 45 years, almost like clockwork, the number of transistors on a state-of-the-art silicon computer chip has roughly doubled every two years, making Moore's Law as reliable as the law of gravity. As the active elements on a chip have gotten smaller and cheaper to make, more of them could be shoehorned into devices of increasing complexity, ability and power -- all at roughly the same cost as the previous generation of products.

This embarrassment of digital riches may at last be heading toward a dead end. Scientists trying to stuff ever more transistors onto a silicon chip are having trouble reliably making active elements smaller than the current best of 14 nm -- roughly double the size of a hemoglobin molecule in blood or about one-thousandth the size of a grain of talcum powder.

A substance called graphene could breathe new life into Moore's Law by augmenting silicon technology. Made from nothing more glamorous than soot, graphene is an atom-thick layer of carbon atoms arranged in a hexagonal pattern. Under an electron microscope, graphene looks like a cross between chicken wire and a honeycomb.

"It not only looks strange but has incredible properties," says Walt de Heer at his nanoscience lab at the Georgia Institute of Technology. "Graphene is a wonderful material to make electronics out of," he says. "It's fast, doesn't use a lot of power and can be made with very small features. It outperforms silicon and does things silicon can't. It could be the future of electronics."

Semiconductor researchers have been experimenting with graphene since the 1970s but have had problems making ultrathin layers of the honeycomb. University of Manchester researchers Andre Geim and Konstantin Novoselov successfully produced graphene layers in 2004 (this and other advances in graphene research earned them the 2010 Nobel Prize in physics), and the field has advanced rapidly since then.

Earlier this year, de Heer's group fabricated graphene wires -- an essential first step in making microchips -- that were about 10 nm wide by using epitaxy to deposit a sheet of pure graphene onto a silicon chip. (Epitaxy is the process of growing a thin crystalline layer on the surface of another crystal so that the layer mimics the structure of the substrate.)

Eventually, electronic structures as small as 1 nm and much faster than silicon are possible, de Heer says. "If it pans out, graphene could yield a terahertz processor," he predicts -- roughly 20 times faster than today's best silicon chip.

Next year, the Georgia Tech group hopes to finish work on a prototype graphene integrated circuit to use as a test bed for exploring the material's unique properties and refining the technology for creating circuits.

Meanwhile, researchers at IBM have produced experimental graphene-based transistors and integrated circuits using standard semiconductor manufacturing techniques. IBM's Guha points to these as the first steps toward graphene being used on an industrial scale.

"This area has great potential," he says. "It has applications in military and wireless technology and the possibilities for integration with silicon. What is needed now is a lot of hard work to demonstrate the ability to build amplifier circuits and to create large areas of high-quality graphene active circuits integrated in them."

While the first graphene products could appear in the 2013 time frame, don't expect to see super-fast laptops powered by graphene chips anytime soon. Because of their expense, they are likely to show up initially for specialty uses where cost doesn't matter as much as top speed and low power use.

Similarly, integrated circuits that seem rudimentary today were once expensive specialty items used in military and space applications where cost wasn't the main consideration. "The history of this area," says NIST's Seiler, "is that these things start out expensive and rare and become inexpensive and everywhere."

HP Labs' Williams adds, "It's like creating a new way of making chips that could be a lot faster. Graphene has a lot of potential and could be in everyday items in 10 years."

Standard semiconductor processing involves a series of intricate steps that need to be carried out in an expensive clean room that's free of electronics-destroying dust and contaminants. But Xerox is working on a cheaper and easier way to make electronics by printing circuits on a plastic sheet. The process uses equipment that might cost hundreds of thousands of dollars, not the billions needed for traditional chip-making plants like the one Intel recently broke ground for in Chandler, Ariz.

"Conventional electronics are fast, small and expensive," observes Jennifer Ernst, formerly director of business development at Xerox's PARC research lab in Palo Alto, Calif. By printing them directly on plastic, however, PARC is making electronics that are "slow, big and cheap," says Ernst, now a vice president at Thin Film Electronics.

PARC's design prints circuits directly on the base material in a process that's often only slightly more involved than printing a mailing label. It requires some special materials, like silver ink, but these devices can be printed on flexible polyethylene sheets rather than on brittle silicon. In fact, the results probably shouldn't even be called chips anymore.

By adapting a variety of printing techniques, including ink-jetting, stamping and silk screening, PARC has made amplifiers, batteries and switches for a fraction of what it costs to manufacture them the traditional way. The company recently succeeded in making a 20-bit memory and controller circuit this way, and will start selling it next year. It's a drop in the digital bucket compared with megabit flash and DRAM chips, but it's a start.

Another interesting printed-circuit project is the blast detection sensor tape that PARC is developing for the U.S. Defense Advanced Research Projects Agency (DARPA). It's made by printing circuits on a flexible tape that can be pressed onto a soldier's helmet. With a flexible film battery on the back, the sensors measure the pressure (up to 100 psi), acceleration (up to 1,000 Gs), sound levels (up to 175 decibels) and light (up to 400 lux) experienced in battlefield conditions.

After a week on the front line, the soldier tears the tape off the helmet and sends it to a lab, where the data is downloaded and analyzed so that doctors can see if the soldier is in danger of a debilitating brain injury. "It replaces a $7 sensor, costs less than $1 and performs just as well," says Ernst.

On the downside, printed circuits will likely never catch up to silicon in terms of speed or the ability to put billions of transistors on something the size of a fingernail. But there are lots of places where cost counts for more than speed. As early as 2012, printed devices should start showing up in toys and games that incorporate rudimentary computing, like synthetic voices, as well as in car seat sensors for controlling the deployment of air bags in an accident. (Printed circuits are slow compared to traditional silicon electronics, but still fast enough to for air bag deployment.)

Further out -- around 2015, Ernst estimates -- printed circuits could end up in some very interesting places, such as flexible e-book readers that can be rolled up when not in use or clothing made of a solar-cell fabric that can charge a music player or cell phone. Market analysis firm IDTechEx forecasts that sales of these flexible printed circuits will grow from $1 billion in 2010 to $45 billion in 2016 and show up in a variety of devices.

IBM's Guha also sees a bright future for printed circuits. "Anytime you remove a clean room from making electronics, it becomes much cheaper," he says. "Cheap and dirty is good enough for many uses, provided that the circuits can be made with acceptable quality."

For more breakthrough technologies, be sure to check back next week, when we'll look at innovations in access, power and computer control.

This story, "Chip-level advances that may change computing" was originally published by Computerworld.

Join the Network World communities on Facebook and LinkedIn to comment on topics that are top of mind.

Copyright © 2011 IDG Communications, Inc.

1 2 Page 2
Page 2 of 2
SD-WAN buyers guide: Key questions to ask vendors (and yourself)