CMOS integrated nano-photonics -- coming soon?
Building on earlier proof-of-concept work, IBM researchers presented the results of their first silicon integrated nano-photonics. That’s a mouthful but a little easier to digest than the full title of the paper: A 90nm CMOS Integrated Nano-Photonics Technology for 25Gbps WDM Optical Communications Applications. In a nutshell, this was the first time that optical components were integrated on the same chip as standard silicon CMOS transistors. The big step here was producing the devices in on of IBM’s commercial foundries on a 90 nm generation production line.
I reviewed the pre-conference program and presented my picks of the most interesting papers here and here. I did not mention the integrated nano-photonics paper since I concentrated more on near-term technology. So why did it attract attention, and why did I overlook it?
Don’t get me wrong. The paper represents a major milestone in a challenging development program. Furthermore, IBM develops a lot of semiconductor technology that ends up as industry standard – especially when it comes to interconnect. Remember the switch from aluminum wiring on ICs to copper? IBM was the first in production. Optical communication is faster than metal wiring even copper. That’s why everyone wants an optical fiber connection for the Internet. The semiconductor industry craves it at the chip level too as metal wiring represents a data bottleneck in communications between chips in systems such as data exchanged between microprocessors and RAM. You can bet that if the industry picks up on optical interconnect, IBM will be paving the way.
But I think there is still a long way to go before the industry is ready to switch to optical interconnect at the board level. The all-optical backplane idea has been around for a long time, and there is a lot of development between 90 nm and the current production state-of-the-art at 22 nm.
There is also the possibility that integration of optical components and silicon transistors for very large scale may actually become unnecessary in the time frame one might expect the IBM nano-photonics to roll out. That’s because there is a lot of work on silicon channel replacement. Future chips may actually be switching to the III-V materials used in the optical components for the standard digital logic devices as well. Integrating optical components and logic gates will require a lot less effort at that time.
In short, I don’t expect to see this technology in our gadgets in the foreseeable future, but the tech blogs like gee whiz –especially beautiful art and graphics. IBM always obliges in this regard. As you can see, I have taken full advantage above.