Intel makes microchips, not radios. But if the company’s newest manufacturing plans pan out, the distinction between the two products may disappear. According to chief technology officer Patrick Gelsinger, Intel is quickly learning how to build tiny radio transceivers from the same material it uses in microchips: silicon. Research progress inspired Gelsinger to announce in February an audacious plan to put a silicon-based radio on the corner of every microchip the company sells, within as little as five years, at no extra cost to customers.

The announcement puts Intel at the forefront of industry efforts to build all-in-one chips that could replace the jumble of costly parts in cell phones and other wireless gadgets. Silicon integrated circuits already dominate when it comes to the digital “back end” of cell phones and other wireless communications devices, where signals are decoded for conversion into sound. But the analog “front-end” components, which capture and amplify radio signals and convert them to digital bits, are typically found on a separate, radio frequency section of the wireless circuit board. This section houses both large, three-dimensional parts such as capacitors and oscillators and transistorized components like amplifiers, which run at such high speeds that they have traditionally been made only using faster, more expensive “compound semiconductors” like gallium arsenide.

But Intel and other chip makers would prefer to stick all these functions onto a single silicon chip, which could be patterned using well-established photolithographic techniques and would cost about one-tenth as much as chips using compound semiconductors. For engineers and computer scientists looking to a future where computing power is ubiquitous and wireless, the potential cost and space savings of putting all of a radio’s parts onto the same chip that holds the computing components has a powerful appeal. “We can safely say that any intelligent device needs both a processor and some form of wireless connectivity,” says Turner Whitted, manager of the Hardware Devices Research group at Microsoft Research in Redmond, WA. “It makes sense to combine these functions to the greatest extent possible.”

There are still big technical barriers to mass-producing silicon radios, such as reducing three-dimensional parts to the micrometer scale with the needed precision and uniformity. But Gelsinger made his announcement on the strength of recent work in the labs of Steve Pawlowski, director of Intel Labs’ Communications and Interconnect Technology Group in Portland, OR, and Valluri Rao, who heads the company’s Analytical and Microsystems Technologies division. Rao’s staff is using silicon to build tiny structures that duplicate the functions of traditional capacitors, oscillators and other components. Pawlowski and his colleagues, meanwhile, are testing silicon circuitry that performs the amplifying, mixing and filtering functions typically handled by separate, more expensive front-end chips. By working out these core technologies, “We’re going to be able to dramatically reduce the size and cost of the components required in radio circuits,” Gelsinger says.

This is an ambitious agenda, considering that many of these components only work because of their macroscopic size and their three-dimensionality. Oscillators, which help to tune in and amplify radio signals, are one example. These crucial components are often made from quartz crystals that resonate electrically when a voltage is applied, with a resonant frequency partly determined by their dimensions-typically up to a centimeter square. Variable capacitors, used to filter out all frequencies except that of the signal, are another example; they usually consist of interleaved metal plates that hold a varying charge depending on the amount of space between them. Putting space between the plates requires depth, but a typical integrated circuit consists only of thin layers of semiconducting silicon and conducting substrates.

That’s why Rao and his fellow researchers have started thinking out of the plane, turning to microelectromechanical (MEMS) manufacturing techniques developed over the last decade in numerous labs at universities and startup firms. These labs have created a wide assortment of tiny structures like beams, bridges and springs from strips of silicon only a micrometer wide. But no one has yet fashioned such structures into a fully functional radio, or built them on a piece of silicon that contains all of the signal-processing circuitry needed to handle today’s digital cell phone transmissions.

Rao and his group believe this can be done without any revolutionary changes in manufacturing techniques. “You can actually build a mechanical device like a variable capacitor using the lithography we have available,” he says. Conventional lithography uses light to carve tiny features on silicon chips; MEMS builders go a step further, excavating around these features to make such suspended three-dimensional structures as cantilevers.

Rao’s group is using these lithography-based MEMS techniques to build prototype silicon capacitors in which the upper plates are suspended by tiny silicon springs. Applying a voltage makes the plates move up or down, changing the capacitance. It turns out that such structures leak far less charge to surrounding materials than conventional capacitors, Rao says. His researchers are also experimenting with oscillators made from free-hanging cantilevers. “Imagine a tuning fork with one prong, but so small that its resonant frequency is measured in gigahertz,” Rao explains. “That lets you start doing things at radio frequencies.”

While researchers at Intel and elsewhere have managed to build small clusters of tiny capacitors and resonators, building hundreds or thousands of identical MEMS structures using lithography is still a problem. “We’re looking at our lithographic process from the point of view of getting very uniform behavior over a wafer, so that we can build these things at high volume,” Rao says.

Pawlowski’s group, meanwhile, is demonstrating digital circuitry on silicon for components such as mixers and analog-to-digital converters, achieving what he calls “pretty good signal gain” even at the high frequencies usually handled by gallium arsenide chips. And while they’re at it, researchers in his group are designing signal-processing circuitry with the brains to switch between competing wireless-communications protocols. “If somebody is in a Starbucks and they have a connection on their laptop to a wireless local-area network, and they walk out, they need a second radio to keep the connection open,” notes Pawlowski. “The promise of this architecture is that it could run multiple protocols without having to have multiple, separate radios.”

Laptops, cell phones and other devices that let you roam seamlessly between wireless networks are only one of the industry niches where silicon radios could eventually dominate. Armies of small, low-power, constantly connected devices could eventually infiltrate the appliances and structures all around us. “For example, tiny sensors that communicate through different methods could go inside every window and every ventilation duct to monitor environmental conditions and improve energy efficiency,” says computer scientist Gaetano Borriello, who leads an Intel-sponsored ubiquitous-computing laboratory at the University of Washington in Seattle. “Eventually, they could even go inside of people. What we’re doing is expanding the range of possibilities.”