IBM lights up silicon chips to handle ‘Big Data’

19 Dec 2012

Big Blue says breakthrough shows feasibility of silicon nanophotonics for chip manufacturing.

IBM has announced what it is calling “a major advance in the ability to use light instead of electrical signals to transmit information for future computing”.

The breakthrough technology – dubbed silicon nanophotonics – allows the integration of different optical components side-by-side with electrical circuits on a single silicon chip using sub-100nm semiconductor technology.

The company says that its 90nm silicon integrated nanophotonics technology can integrate a photodetector and modulator fabricated side-by-side with silicon transistors. Silicon nanophotonics circuits and silicon transistors can be interconnected with nine levels of metal wires.

IBM says that silicon nanophotonics provides a super highway for large volumes of data to move at rapid speeds between computer chips in servers, large data centers, and supercomputers, thus alleviating the limitations of congested data traffic and high-cost traditional interconnects.

John Kelly, Senior VP and Director of IBM Research, commented, “This breakthrough is a result of more than a decade of pioneering research at IBM. This allows us to move silicon nanophotonics technology into a real-world manufacturing environment that will have impact across a range of applications.”

Silicon nanophotonics technology provides answers to “Big Data” challenges by seamlessly connecting various parts of large systems, whether few centimeters or few kilometers apart from each other, and move terabytes of data via pulses of light through optical fibers.

Building on its initial proof of concept in 2010, IBM says it has solved the key challenges of transferring the silicon nanophotonics technology into the commercial foundry.

By adding a few processing modules into a high-performance 90nm CMOS fabrication line, a variety of silicon nanophotonics components such as wavelength division multiplexers (WDM), modulators, and detectors are integrated side-by-side with a CMOS electrical circuitry.

As a result, single-chip optical communications transceivers can be manufactured in a conventional semiconductor foundry, providing significant cost reduction over traditional approaches.

IBM’s CMOS nanophotonics technology demonstrates transceivers to exceed the data rate of 25Gbit/s per channel. In addition, the technology is capable of feeding a number of parallel optical data streams into a single fiber by utilizing compact on-chip wavelength-division multiplexing devices.

Analysis

Considering that a semiconductor company as significant as IBM is now talking about the commercialization of silicon photonics, optics.org interviewed Professor Michael Hochberg, Director of silicon photonics foundry OpSIS, who also specializes researching the subject at both the University of Delaware and the National University of Singapore.

"The thing that is novel about IBM’s latest work is that it is in a sub-100nm electronics process; a 90nm process rather than a 130nm process, which Luxtera has been using. 90 nm is certainly better than 130nm because the transistors have narrower gates, and processes run faster and at a lower power, on the whole. But both of these wavelengths are really trailing edge nodes, in a world where advanced CMOS is at 22nm.

"This is nice work – it’s great to develop a wholly integrated process, it’s very exciting to see IBM working on this integrated approach, and to see the seriousness of their commitment to silicon photonics. Developing a process like this is neither cheap nor easy.

"It’s a good achievement for IBM to have developed monolithically integrated CMOS, transistors, germanium, with a decent modulator and a decent photodetector. They are showing some supporting circuits as well, which is also great to see. And it’s a good thing for multiple foundries to be able to produce integrated silicon photonics-electronic circuits.

"It is very difficult and expensive to get photonics systems to work monolithically in very advanced processing beyond the 90 / 130nm node. The economics pushes you away from doing it. However because the photonics doesn’t need super advanced lithography and super resolution, the economics of it pushes developers towards a two-chip solution where the photonics is done in one chip and the electronics in another, then the two are bonded together using low-capacitance interconnects.

"I really see multi-chip integration approach as the direction which a lot of the field will be heading in over the next few years. Because that approach enables the leveraging of really advanced CMOS and advanced bipolar features. It allows development of the best possible photonics process without compromising the electronics or vice versa.

"IBM’s achievement is exciting because of the scale of the company and its track record of integrating R&D into products and commercializing them. So IBM’s seriousness about this is a very good sign for silicon photonics as a field.

Emerging silicon photonics marketplace

"What we are seeing is one major semiconductor company after another making major investments that are going from research into meaningful products in silicon photonics. Some of them are doing it by developing their own products, such as Intel and IBM, some are doing it by leveraging outsourced R&D in various ways - some of Opsis’ users are doing it in that way; some of them are doing it by allying themselves with smaller companies, such as ST teaming up with Luxtera; some of them are getting into it through acquisition, such as Cisco buying Lightwire.

"So what we are seeing is players up and down the semiconductor industry hierarchy from FABs to design house to consumers of high-end silicon, there is a wide diversity of strategies being implemented, but they are all coming to the conclusion that they have to be in this space.

"I don’t believe that it is inevitable that it will become a commoditized marketplace in the near term. I think what’s going to happen is that silicon photonics is going to end up looking a lot like analog electronics. Anyone who wants to build high performance digital circuits and have fast I/Os ends up having to figure out how to engage on the analog side.

"What’s going to occur is that we will see the same thing in silicon photonics. The bottom line is that the people who want to do the most advanced electronic circuits and get the most data in and out are going to be forced to engage with silicon photonics. There will be a variety of business strategies. But it’s not going to be commoditized, at least not for a long while."

About the Author

Matthew Peach is a contributing editor to optics.org