There are some things silicon doesn’t do well. It neither absorbs nor emits light efficiently. So, silicon photonics systems, which connect racks of servers in data centers via high-data rate optical fiber links, are dependent on photodiode receivers made of germanium or other materials to turn optical signals into electronic ones. Saif Islam, professor of electrical and computer engineering at University of California Davis, and colleagues have come up with a way for silicon photodiodes to do the job, potentially driving down the cost of optical computer-to-computer communications.
The difficulty for silicon photodiodes, explains Islam, is the trade-off between speed and efficiency. When a photon is absorbed in a photodiode, it becomes an electron and a hole (a positive charge), which must then travel to the device’s anode and cathode to produce current. The thicker the absorption region, the more likely a photon is to be absorbed, increasing efficiency. But thinner regions have a speed advantage: “When you make it very thin, your electrons and holes take little time to reach the electrical terminals.” And that speeds up the photodiode’s switching speed and the system’s data rate.
Islam and his colleagues came up with a silicon structure that makes photodiodes both fast and efficient by being both thin and good at capturing light. The structure is an array of tapered holes in the silicon that have the effect of steering the light into the plane of the silicon. “So basically, we’re bending light 90 degrees,” he says.
Ordinarily, most light would pass through the photodiode, with only a few percent of photons captured and turned into electricity. But by steering photons into the specially-structured photodiode, the light spends more time interacting with it instead of falling straight through it, leading to higher efficiency. Because the light is traveling the photodiode’s length, not its thickness, the device can be made thin and therefore speedy.
Using data from these photodiodes, Islam calculates that it could help a single silicon transceiver operate at 60 gigabits per second, well above the 25 Gb/s available now. He expects that his team will have their efficient-but-speedy transceiver constructed later this year. Being able to make the whole system as one integrated silicon circuit should help drive down the cost of data connections. Right now they’re around US $10 per gigabit per second, and the industry goal is below $1/Gb/s, he says.
His team is also experimenting with the kind of photodetector used in lidar systems for self-driving cars. Called avalanche photodetectors, these are designed for picking up the relatively few photons that bounce back from a laser’s target and amplify those into a detectable electronic system. These are usually made from non-silicon materials, but he’s confident the nanostructured silicon device can do the job. “Our device could offer speed and sensitivity in low-light conditions,” he says.