While lying on the beach during a vacation on the Spanish coast in 1999, physicist Jamal Ramdani had an epiphany. As the sand complied to the contours of his body, Ramdani, a researcher at Motorola Labs in Tempe, AZ, suddenly envisioned a solution to a puzzle that had perplexed the semiconductor industry for 30 years: how to combine cheap silicon with high-speed, light-emitting but far more expensive semiconducting materials like gallium arsenide, all on a single wafer.
Because the materials are physically mismatched, layering one on top of the other to produce a chip with optimal electronic and optical properties has been virtually impossible. It may have been the sand on that Spanish beach, which is made of the same mineral from which silicon wafers are derived, that provided Ramdani with the pivotal hint. In any case, Ramdani recalls, “I came back to Phoenix, borrowed a machine for growing compound semiconductors, and in two or three shots, we had gallium arsenide sitting on silicon.”
The benefits of having the functionality of gallium arsenide-particularly its abilities to handle wireless communications and emit light-on an inexpensive silicon chip were not lost on Motorola executives. High-performance chips made out of gallium arsenide and other so-called compound semiconductors are widely used in everything from cell phones to switches in optical communications networks. At the very least, Ramdani’s invention could mean replacing these costly chips with far less expensive gallium-arsenide-on-silicon ones. In the two years since Ramdani’s breakthrough, Motorola has filed over 300 patents on the technology; last fall, the company used Ramdani’s method to build prototype chips for boosting signals in cell phones. To commercialize the new material, Motorola has started up a wholly owned subsidiary-Thoughtbeam, in Austin, TX-promising the new materials will find their way into electronic and optical devices within the next two years.
The impact of Motorola’s chip technology could go far beyond cheaper cell phones or optical devices. Today, if you want a fast, inexpensive microprocessor, you need a silicon chip; if you want a chip to handle optical functions or high-frequency radio signals, you need compound semiconductors like gallium arsenide or indium phosphide. As a result, equipment like cell phones and communications network switches requires multiple semiconductor devices. Eventually, predict some experts, the Motorola technology could make it possible to integrate the functions of gallium arsenide and silicon on a single chip, using each of the materials for what it does best. The result would be a superchip. Instead of having multiple chips in a DVD player doing different tasks-generating light to read the disc, fielding input from viewers, decoding digital data into images and sound-a single chip could handle it all.
The semiconductor industry has been dreaming of such a superchip for decades-and a number of researchers are actively pursuing that dream. For instance, Eugene Fitzgerald, a materials scientist at MIT, has been working on the problem for over a decade and has published descriptions of his own technique for growing gallium arsenide on silicon. He and many other skeptics question whether the Motorola technology will prove to be a grand slam. “Every few years, there is a so-called solution, but upon closer examination, you see that it isn’t one at all,” says Fitzgerald.
Others, however, are so impressed with the potential of Ramdani’s breakthrough that they believe the technology could fundamentally change the dynamics of the chip-making business, finally bridging the materials divide between silicon and compound semiconductors that has become a fundamental fact in the industry. According to Steve Cullen, director and principal analyst of semiconductor research services at Cahners In-Stat Group, the Motorola advance could “go down in history as a major turning point for the semiconductor industry.”
Silicon is the material of choice for the vast majority of chips used in microprocessing applications; it’s easy to handle, and manufacturers have learned to carve into it the tiny circuitry that makes possible today’s fast and inexpensive computers. But for all of its celebrity, silicon cannot match the wireless and optical capabilities of more expensive semiconductors like gallium arsenide and indium phosphide.
These materials are called “compound” semiconductors because their crystals-unlike silicon’s-are composed of more than one element. This more complex composition often endows them with desirable physical traits. For instance, because electrons travel faster in many compound semiconductors, the materials can process higher-frequency radio signals and hence larger amounts of data, which is just what you need if you want, say, handheld wireless devices that can receive seamless streams of video.
And unlike silicon, many of these compound semiconductors can emit beams of light when fed just a little bit of electrical current. That makes possible the solid-state lasers that can read the small bits of information densely packed on a CD or DVD. High-speed optical communication networks also rely on compound semiconductors for converting optical information into electronic information, and vice versa, at the thousands of places where optical fibers meet electronic switches and computers.
However, the same complexity that makes compound semiconductors so useful also makes them brittle, hard to synthesize, tough to integrate with other materials-and very expensive. At the moment, a 15-centimeter wafer of gallium arsenide costs about $300, while a 20-centimeter silicon wafer can cost about one-tenth as much. Ramdani’s breakthrough involves a way to deposit a patina of gallium arsenide atop a wafer of standard silicon. The top layer of gallium arsenide provides all the unique capabilities of that material, but putting it on a silicon substrate makes it much easier to handle and cheaper to manufacture.
At first blush, the procedure sounds about as simple as slathering peanut butter on top of a slice of bread. But in practice, it’s much trickier. The fundamental problem, says Fitzgerald, is that the underlying crystalline structures of silicon and gallium arsenide are so different that layering one on top of the other is like stacking grapefruits on a bed of oranges. “You get misfits and extra spaces,” says Fitzgerald. These defects in the crystal tend to snag electrons, disrupting the functions of the semiconducting devices.
So far, the mismatch problem has defeated just about anyone who has ever attempted to make gallium-arsenide-on-silicon wafers. That helps explain why many researchers at companies like IBM and the semiconductor technology startup AmberWave Systems of Salem, NH, are pursuing an alternative approach to more versatile and powerful semiconductors: tweaking silicon so that it behaves more like its fancier cousins. That way, they get the cost benefits of using the 50-year-old infrastructure of silicon manufacturing technology and yet still approach the performance of compound semiconductors.
AmberWave Systems, which was cofounded by MIT’s Fitzgerald, has developed a form of “strained” silicon crystal in which electrons move faster than in regular silicon. The material makes faster transistors possible, and that means, for example, higher-frequency radio signal processors. The researchers grow a layer of a silicon-germanium alloy on top of a silicon wafer and then top the alloy with a thin layer of silicon. Because the distances between atoms in the silicon-germanium crystal are longer than in silicon, the silicon atoms in the top layer must stretch in order to match the spaces between the atoms in the silicon-germanium below. When the silicon atoms are farther apart, electrons move about more freely, hence faster.
In fact, this little bit of crystal engineering has yielded samples in which electrons move up to 80 percent faster than they do in ordinary silicon wafers. Within the next year, AmberWave hopes to see devices made from this material hit the market-microprocessors, for instance, or signal-boosting chips in cell phones.
Tweaking silicon could make it faster, but for optical capabilities you still need compound semiconductors. While a number of researchers are trying to grow compound semiconductors on silicon, Motorola believes it has a head start in the race to commercialize the technology-thanks to both Ramdani and the company’s well-established manufacturing and marketing infrastructure.
The history of Ramdani’s breakthrough actually begins at least a year before his fateful Spanish vacation. Ramdani was part of a Motorola research group trying to make silicon faster when he made an accidental discovery that would lead to the gallium-arsenide-on-silicon project. At that time, he and his colleagues were focusing on the thin, glasslike layer of silicon dioxide that forms on top of silicon when it is exposed to oxygen during chip processing. This layer, known as a dielectric, is a vital chip component because it enables one transistor to control the electrical state of another while preventing electrons from leaking between them.
But as transistors get smaller and this layer gets thinner, it becomes more prone to leaking electrons. To solve this problem, Ramdani and his colleagues Ravi Droopad and Jimmy Yu were experimenting with an alternative to silicon dioxide-strontium titanate-that could improve the performance of silicon-based chips. Yet as the Motorola researchers deposited a bee’s breath’s worth of strontium titanate on a silicon surface, an intervening layer of silicon dioxide formed. It was like inadvertently covering a window with a layer of black paint when all you wanted to do was slightly tint it.
And then, Ramdani visited that Spanish beach. While relaxing on the sand, he realized that the silicon dioxide layer, along with the strontium titanate, might serve a far larger purpose than he had originally imagined: intermediate layers that, when sandwiched between silicon and gallium arsenide, could reconcile the crystalline mismatch between the two semiconductors. That’s because the distances between the atoms in strontium titanate, when on top of the silicon dioxide layer that forms underneath it, are longer than those in silicon but shorter than those in gallium arsenide. In fact, it’s the silicon dioxide that causes the atoms in strontium titanate to relax completely and assume a configuration more in line with that of the gallium arsenide atoms above. Within days of returning from his vacation, Ramdani and his team of engineers succeeded in growing gallium arsenide on silicon using these intervening layers (see “Superchip Ingredients”).
As the Motorola researchers refine their technology over the coming years and learn to grow other compound semiconductors on top of silicon, the potential applications of the material ought to continue expanding. As Ramdani and his colleagues see it, the same type of inner layer that they use to marry gallium arsenide to silicon could be used to grow indium phosphide or any number of other high-performance compound semiconductors on the same inexpensive silicon substrate. Each of these compound semiconductors has its own personality-its own speed and light-emitting properties.
Such technology could also lead to new kinds of devices or applications that previously have not been cost effective. Cheap sources of high-performance chips, for example, could make it easier for designers to add wireless communications to household appliances and connect them to the Internet. Visions of washing machines directly communicating with service centers when they go on the fritz or refrigerators that call in food orders to the supermarket could become cheaper to realize, if no more desirable. More affordable light-emitting and light-detecting chips could change the economics of fiber-optic links for directly connecting home computers, video cameras and other domestic gadgetry to the Internet.
Beyond that, chip manufacturers like Motorola and AmberWave Systems share the same longer-term techno-dream-an all-in-one wafer. In this vision, compound semiconductors are not just layered on top of a silicon substrate, but the different semiconductors are integrated together on the chip. “If we can grow a thin film of gallium arsenide on top of silicon wafers, then maybe we can selectively grow islands of gallium arsenide on silicon,” says Charles Huang, cofounder and chief technical officer of Anadigics, a chip manufacturing firm based in Warren, NJ.
Each island would have its own function-say, sending and receiving messages wirelessly, or transmitting data optically to the outside world. Yet the majority of the silicon would be available to do the actual computing or storing of data. Such multitalented chips would, for instance, be able to shuttle data around a microprocessor optically. In a computer, data currently moves electronically both within chips and between chips-between, say, a microprocessor and a memory chip-through tiny wires that slow everything down. “The wires are the real bottleneck in computers,” says Ramdani. If each silicon chip came with its own onboard lasers made of compound semiconductors for moving data around, such chips would both operate more quickly on their own and be able to trade larger amounts of data with other chips more quickly.
It’s too soon to rule out the possibility that some lurking glitch will send Motorola’s growing investment in its new technology into the vast heap of good tries gone bad. Certainly there are a number of skeptics out there still not convinced that the company’s superchip will ever live up to its hype. Nevertheless, in the three years since Ramdani’s original epiphany, Motorola has increasingly committed itself to making sure the technology fulfills its many promises, throwing its substantial financial and technical weight behind it.
Indeed, Ramdani’s excitement over the breakthrough is far from waning. “The way I see it, this technology is going to revolutionize the semiconductor industry,” he says. “It will allow us to do things that, 20 years ago, we could only dream of doing.”
Pumping Up Silicon
A sample of companies pushing the limits of semiconductor materials