Phiar Claims Detector Breakthrough
Phiar (pronounced: Fire!) says it’s hit on a way of making detectors out of silicon, and that could be a big deal. It would mean that detectors could be integrated directly with standard CMOS electronics -- an important step towards putting optics and electronics on the same chip. This in turn could slash costs, while also boosting the performance of all kinds of semiconductor devices.
As an early-stage startup, Phiar hasn't pinned down the best way of applying its technology. One use would be to eliminate some components from optical transcievers, namely a separate detector and electronic amplifier chip. It might also be used for high-speed chip-to-chip interconnections, for example, for optical clock distribution -- detectors could be placed at various points on a chip to pick up high-speed timing signals.
Phiar also points out that its technology could be used to build much faster detectors than those available today. Its detectors should be able pick up signals as fast as 1 Tbit/s, the company claims.
Founded in July 2001 by university professor Garret Moddel from the University of Colorado in Boulder, Phiar has recently been granted several key patents. And that means it's prepared to lift the covers a little on the underlying technology.
All optical detectors today are based on the particle nature of light, using a semiconductor to absorb incoming photons, says Moddel. Phiar's approach is different -- it relies on the wavelike nature of light. Simply put, it uses an antenna to capture the variations in the electromagnetic field and convert them into alternating current. A rectifying diode is then used to smooth out the carrier frequency of the light, leaving just the data signal.
It sounds straightforward until you realise that the carrier frequency -- the frequency of 1550nm light -- lies in the region of several hundred terahertz -- aproximately 20,000 times faster than the fastest silicon electronic chips used in today's networks.
There's another technical difficulty -- the size of the antenna must match the wavelength of light impinging on it. That means Phiar requires an antenna roughly 1 micron wide. In fact, it probably requires an array of these antennas to capture a decent amount of light from a singlemode fiber, for example.
Both these issues have prevented anyone from commercializing optical antenna technology in the past, says Moddel. But thanks to advances in standard silicon processing technology, both problems can now be overcome -- with a little ingenuity. "We aren't asking for anything that state-of-the-art chip makers can't already do," he adds.
The patents that were granted last month relate to the rectifying diode -- a device that only passes current in one direction. Standard semiconductor diodes are way too slow to keep up with terahertz frequencies. A faster device exists, called a metal-insulator-metal (MIM) tunneling diode, which transports electrons across the diode by a quantum mechanical process called tunneling. Tunneling is very fast, compared to the ordinary transport of electrons inside semiconductors.
Old MIM diodes weren't very responsive however, so Phiar has come up with a variation that has a much higher sensitivity. Instead of a single insulator material, the company uses a second type of insulating material to create a "quantum well" -- a thin sandwich of one material inside the other. It's only in the past few years that semiconductor manufacturers have become adept at deposting very thin layers of insulator, Moddel notes.
More patents, still pending, relate to a "traveling wave" structure, which sends the super-high-speed electronic signals from the antenna to the diode without slowing them down.
With the technology foundations laid, Phiar's next step is to find a partner to help it take the fabrication process out of the lab and into a commercial environment. It also needs to decide which applications to target first -- something a partner could help with.
"We're a small company, so we can't be making our own chips," Moddel notes. "We are seeking two types of partnerships. One would be with a foundry, the other would be working with a manufacturer where we put our technology on their chip under license."
"We're very focused on what we've got to do to prove that we're not just a pretty face," he adds.
So far, Phiar has convinced at least one VC that its ideas have merit. Menlo Ventures provided seed funding and furnished a further $3 million in August 2002.
— Pauline Rigby, Senior Editor, Light Reading