Intel claims opto-electronic breakthrough

A breakthrough in opto-electronics detectors by Intel promises to address the shortage of network bandwidth which some believe could lead to the collapse of the internet. The system could also improve inter-chip and on-chip communications.

A paper published in Nature Photonics details the performance of an Avalanche Photo Detector (APD) based on silicon rather than the normally used indium phosphide.

Intel's research team created a sensor for light detection and amplification using silicon and Complementary Metal-Oxide Semiconductor processing to give a "gain-bandwidth product" of 340GHz, which they claim is the "best result ever measured for this key APD performance metric".

'Gain' is an electronic system's ability to increase the signal power of the device, and Intel claims that its research could "lower the cost of optical links running at data rates of 40Gbit/s or higher" and "proves for the first time that a silicon photonics device can exceed the performance of a device made with traditional, more expensive optical materials such as indium phosphide".

"The gain [multiplied by] the bandwidth is a fixed number," said Dr Mario Paniccia, of Intel's Photonics Technology Lab, explaining the significance of a 'gain-bandwidth product'.

"Commercial APDs made with indium phosphide have a typical gain bandwidth of 120GHz. For use in 10Gbit/s long-haul communications links, typical APDs would cost $200 to $300 [£134 to £202].

"Because of the high gain bandwidth we're achieving here in a low-cost device, we can produce a device that gives better performance at 10Gbit/s, and can also operate better at 20Gbit/s and 40Gbit/s and higher."

Dr Paniccia added that, depending on the application, Intel could adapt the device to focus on speed, distance or power savings.

"We can reduce the laser power, but still get the same gain as current APDs, " he said. "The potential applications of the new APD range from communications interconnects to quantum cryptography, biochips and eventually chip-to-chip and on-chip interconnects."

Questioned about commercialisation of the product, Paniccia admitted that Intel still has to optimise the device for performance and increase its sensitivity. "We are working to drive this aggressively into our platforms and also looking at the reliability of these systems," he said.

Dr Paniccia explained that optics would evolve in the next five to seven years, but that it will be difficult to replace copper at distances of under six inches for chip-to-chip interconnects.

"Predicting something five years ahead - an eternity in this industry - would be difficult," he said. "On-chip interconnects are still quite a way out. Ten plus years."