Cornell professor Michal Lipson heads a research group of a dozen EEs pursuing nanoscale photonics. The team has created silicon photonic chips that confine light in very small cavities that can slow down, enhance and otherwise manipulate the optical properties of light. These pioneering nanostructures enhance the light-matter interaction in silicon by orders of magnitude, Lipson claims, enabling a material's optical properties to be controlled externally either electrically or with another optical signal.
Lipson aims to provide the basic building blocks for electro-optical and all-optical circuits that include all necessary passive and active photonic components on a single CMOS chip.
The daughter of two U.S. physicists, Lipson speaks English with a charming accent that reflects her upbringing in Brazil and Israel. She spoke recently with EE Times contributor R. Colin Johnson.
EE Times: Your group is working on several all-optical chip projects. Is there a rollout plan?
Michal Lipson: We are gradually progressing from individual centimeter-size photonic devices to integrated submicron-scale on-chip devices. Several photonic devices on-chip geared toward different applications are being developed in our lab, ranging from optical interconnects to all-optical networks to on-chip sensing. Several of our designs are silicon-based, such as our all-optical modulator for all-optical networks. All of our photonic devices are based on the principle of strong optical confinement. This confinement enables functionalities to be demonstrated in submicron-sized devices. For example, we have made a silicon biosensor that can sense a single nanoscale particle, because of the optical confinement. And we have made the first silicon all-optical modulator. We also have other optical efforts ongoing that are not based on silicon.
EET: Your all-optical ring modulator was the first silicon circuit to switch optical signals. It works by changing the index of refraction of a nanoscale optical cavity using optical confinement.
Lipson: Yes, we use optical confinement, which is based on confining photons in an optical cavity that is smaller than [the photon's] wavelength. [See www.eet.com/ article/showArticle.jhtml?articleId=51201471.] For modulation, the problem with silicon is that it does not have enough optical nonlinearities. The way we beat this problem is to confine light so strongly that we effectively enhance the very small nonlinearities that silicon does have. The optical properties of the material remain the same, but their effect on the transmission is amplified.
EET: What hurdles must be cleared to enable silicon optical chips?
Lipson: The most important goal of this field is to demonstrate integrated systems that have significant advantages over standard electronics for example, lower power dissipation, higher reliability and low latency. These aspects limit the scaling of current electronic chips and the performance of electronic technologies [such as high-data-rate interconnects]. The devices demonstrated today by ourselves and other groups at MIT, Caltech and Columbia indicate that optics could have significant advantages, especially for high-data-rate transmission systems. However, demonstrations of complete optical systems-on-chip still need to be done.
Our group and others, such as Stanford's and IBM's, are working on low-power systems using optics on-chip. There are also technological issues that need to be overcome at the devices level. These include, for example, overcoming the losses from free carriers in the material. Also, thermal stability is another problem that needs to be tackled, especially when light confinement is concerned. When you confine light to such a small area, the transmission properties become very sensitive to temperature. We and other research groups, such as the University of Rochester's, are trying to come up with ways of minimizing this effect.
EET: One advantage of commercial silicon chips is the wide temperature range they can tolerate.
Lipson: Yes, you are right. But there are several approaches that could solve this problem. For instance, one way is to create a structure in the silicon that reacts to temperature in a complementary manner so that it cancels out the overall temperature effect.
EET: Would that be a structure in the silicon architecture?
Lipson: Yes, it would. And all of what we are doing is compatible with standard CMOS processing techniques.
EET: Intel's reversed-biased devices for controlling carrier concentration in waveguides also are compatible with CMOS.
Lipson: Intel's is a very important breakthrough. Research groups worldwide for many many years have had this goal of achieving a silicon laser. The fact that Intel finally got it to lase is very encouraging. To get its silicon optical cavity to lase, Intel's laser uses another light source to pump it. One could use such a laser as Intel's, for instance, as nonlinear devices, such as a wavelength converter on an all-optical chip for example in all-optical networks, where Intel's laser could have a big impact.
EET: Is your group trying to make a silicon laser too?
Lipson: We are currently not working on making a silicon laser, and there are several reasons for it. I know it's an important hope, but I don't think the lack of a silicon laser is a stumbling block for the field. Think of it as an optical battery you don't expect an electronic battery to be on an electronic chip, and it's the same way here. We don't expect our optical battery to be on the all-optical chip, and we shouldn't expect it to be. A single source outside the all-optical chip distributes the power better.
EET: Which optical silicon functions will appear in chips first?
Lipson: The first will definitely be passive silicon optical components, such as waveguides, filters, bends and so on, because they have been worked on the most extensively. For almost two decades they have been researched in the laboratory, and now chips are being made that show very low losses with very efficient designs.
EET: Have passive silicon optical chips matured enough to be made commercially?
Lipson: Yes, definitely. Then the next step will be to introduce active devices that interconnect, so we can externally control the flow of the light where is the light going. And here is where modulators like ours will be introduced.
EET: And switches too?
Lipson: Yes, yes switches and modulators. We might need amplifiers as well. And the next step will be for all-optical circuits, like optical networks, where you will definitely want to do some logic steps, for example, in decoding routers.
EET: How far off are all-optical chips that perform logic functions?
Lipson: Now that we have demonstrated optical switches the building blocks it is really just a matter of engineering solutions.
EET: Can you give us a timetable?
Lipson: Today, networks from 1 to 10 km long are more economical if made all optical. So it will really depend on distance from that last kilometer on down. For instance, for the distance from rack to rack, it is predicted that in about two to five years it will become more economical to make rack-to-rack communications optical. These of course will not be silicon at first, but they will be optical.
EET: So, you think nonsilicon optical devices will work their way down to the backplane, with more-economical silicon replacements coming up right behind them?
Lipson: Well, for the shorter distance from board to board or from chip to chip, that will still be electrical for at least the next 10 years. And what really determines if and when optics will be introduced or not is how much the existing copper can satisfy user demand for more bandwidth, and how low the cost of optics can go. If we can make the optics on silicon, then it would make a stronger case for replacing copper with optics.
EET: What's the next step for you?
Lipson: We are working on demonstrating high-performance devices, at the submicron scale, as building blocks for on-chip optical integrated systems. We have several ongoing efforts in the lab, all based on strong optical confinement. We are developing new types of optical light-confinement structures that are extremely small so that we have better control over the optical properties. With such high confinements we will be able to achieve optical elements on-chip with unprecedented bandwidth, low power and high speed.