Luxtera fabricates silicon-photonics photodetectors on standard SOI-CMOS wafers
March 16, 2007, Carlsbad, CA--Adding to its silicon-photonics technology portfolio, Luxtera has developed a germanium-enabled integrated photodetector fabricated on a mainstream SOI (silicon on insulator) CMOS wafer using standard processes. Strategically adding pure germanium around optical waveguides is what has allowed Luxtera to create its integrated long-wavelength photodetection capability. Along with silicon-photonic lasers and modulators, photodetectors are key to creating a complete silicon-photonics communications system.
This technology is well ahead of others in the market, according to Luxtera. The monolithic, fully integrated devices to consume far less power and realize orders of magnitude cost reductions that can make photodetectors virtually free, especially when large numbers of them are created on the same wafer.
To achieve photonic to electronic signal conversion, Luxtera applies small amounts of pure germanium along optical waveguides on a CMOS die and connects it to the chip level metallic interconnect by using a step that is common to the CMOS transistors and germanium photodetectors. This integration capability has the potential to make discrete photodetectors obsolete in the future, just as very-large-scale integration (VLSI) made discrete transistors practically obsolete in electronic design, notes Marek Tlalka, vice president of marketing for Luxtera.
Conventionally, commercial photodetectors are implemented in integrated-photonic systems as discrete components, meaning that customers are faced with the cost of purchasing individual photodetectors for each photonic receiver. Luxtera's SOI-CMOS technology allows a very large number of photodetectors to be grown directly on a die during wafer manufacturing, and to be economically tested at the wafer scale. At one low cost, thousands of photodetectors can now be grown at the same time; now, designers can use large numbers of photodetectors freely instead of sparingly because the cost of the final product will no longer be affected by the number of photodetectors used. In applications in which larger numbers of photodetectors are required, such as chip-to-chip and intrachip optical connectivity, this invention brings individual photodetector costs from greater than a dollar to less than a penny.
By placing the photodetector immediately adjacent to the receiver electronics on a common CMOS die, electronic noise is reduced, enabling the photodetector to detect much lower power signals. This improves receiver performance by a factor of four, according to Luxtera; therefore, less expensive low-power lasers can be used on the transmit side, reducing the transceiver cost.
"By quickly productizing this technology, Luxtera should be in a position to lower the cost of optics for future high-volume deployment," said Jag Bolaria, senior analyst with The Linley Group.
"This technology is revolutionary for the industry--it brings us much closer to delivering high performance optical transceivers at cost points associated with legacy copper," said Alex Dickinson, CEO of Luxtera.
Luxtera is currently demonstrating this technology on complete 10 Gbit/s 130 nm SOI-CMOS receivers using the Freescale Semiconductor (Austin, TX) foundry fabrication process. The demonstration receiver consists of a fully integrated die with germanium photodetectors and all required supporting electronic and photonic logic, including optical interface to fiber (fiber to the chip), transimpedance amplifier (TIA), and limiting amplifier (LA). The first commercial application is expected to target Luxtera's next-generation transceivers for optical connectivity in communications and consumer markets. Other potential applications include areas where large numbers of photodetectors are needed and include optical sensors in infrared cameras, interactive gaming, and medical imaging.