Future Optics: Exciting possibilities for silicon photonics emerge - Interview with Michal Lipson

Today, we have networks of light on silicon chips - it's unbelievable how far we have come from barely being able to propagate light.

1609lfw Fo H

OSA: How did you get interested in nanophotonics?

Michal Lipson: My father was a physics professor studying cosmology. He always felt that academic physicists are very fortunate to make a living by simply discovering. His enthusiasm led my twin sister, who is now a space physicist at Boston University, and me to feel from a very young age that we would pursue careers in physics. It would almost be a sin not to! I majored in physics as an undergraduate at the University of Sao Paulo, where I grew up. Later, when I emigrated to Israel, I decided I wanted to pursue a topic more technology-oriented with possible future applications. At the time, quantum heterostructures were a hot topic, and the Technion—where I did my PhD—was extremely active in that area. I moved toward optics during my studies when I got interested in how micro-cavities control the interaction of light with electronic transitions.

About that time, the first paper was published on a one-dimensional cavity, a photonic crystal waveguide defined by etching a series of holes to confine light in a region just a few hundred nanometers across. It was the smallest cavity ever made, and I really gravitated toward this topic.

OSA: What are the advantages of working on a nanoscale?

ML: Tailoring nanostructures can confine light in very small regions. In silicon, light is much more sensitive to nanoscale variations in refractive index than, for example, in glass. The field of silicon photonics can enable interesting new phenomena by localizing light within a few hundred nanometers.

In 2004, my group made a modulator by confining light very tightly in silicon. The main challenge with silicon is that it is fundamentally passive. Light propagates well in silicon, but it lacks a second-order nonlinearity or an electro-optic coefficient, and it does not emit or amplify light. However, the index of refraction is sensitive to carrier concentration, as in any semiconductor. By creating nanocavities in silicon, we were able to amplify this small sensitivity to make silicon an active material so it could modulate light. In 2004, we reported making the first gigahertz silicon modulators, showing that silicon photonics could produce active devices.

OSA: What materials are you working on?

ML: I work on silicon-based materials, including silicon nitride (SiN), silicon carbide, and silicon. I also integrate materials onto silicon, such as two-dimensional films of graphene, which we deposit on SiN. Graphene is important because it has extremely high electron mobility, while silicon and SiN do not. So, the graphene layer is used to electrically control light propagating in the silicon layer. Leveraging the mature technology developed to fabricate silicon microelectronics makes fabrication of photonic structures much easier in silicon than in II-VI, III-V, or more exotic semiconductors.

OSA: How has silicon photonics progressed since then?

ML: One of the big questions in silicon photonics was how to integrate a laser on silicon, a very difficult problem because silicon has an indirect bandgap. Around 2007, John Bowers' group at the University of California at Santa Barbara made a big advance in this area by bonding III-V gain chips with silicon, which led to further advances in integrating lasers with silicon photonics.

Nowadays, a big push is larger-scale integration in silicon photonics for use in data center applications, for example. My group, as well as several other groups, are working on a new class of interconnects that achieve very high bandwidth by multiplexing different modes in addition to different wavelengths. We can do modal division multiplexing very easily on silicon photonics because each mode has a very different effective index, much more so in silicon than in other materials.

Silicon photonics enabled several new areas. One of them is the field of micro-frequency combs: generating parametric oscillation in silicon-based nanocavities. The basic idea is to use four-wave mixing and optical parametric amplification within an oscillator, taking advantage of the very high quality factor in silicon photonics. Sending continuous-wave laser light with very little power into a tiny cavity engineered to have the desired dispersion can generate a whole octave-spanning micro-comb. We are collaborating with Alexander Gaeta, who is also here at Columbia and is an expert in nonlinear optics, on this work.

OSA: What can micro-combs do?

ML: They are very compact and have lots of potential applications, including precision timing and spectroscopy. DARPA has several programs to push the field forward.

For spectroscopy, the mid-infrared (mid-IR) is an important interest because of the limited optical sources that are currently available. For example, pumping micro-combs at 2 μm could generate light in the important 3–4 μm spectral range. That's very exciting for spectroscopy because essentially, all molecules have some signature in the mid-IR. With a good mid-IR source, you could identify molecular species without needing chemical labels by measuring directly the absorption in a gas or liquid. For timing, if the micro-comb can generate an octave—or even two-thirds of an octave—self-referencing the wavelengths within the comb could enable an absolute reference for metrology.

OSA: What's taken for granted now that was revolutionary 20 years ago?

ML: In 2000, we could barely propagate light in silicon, as we had enormous losses and everything in silicon photonics was almost completely passive. The only way to redirect light was to change the temperature. Today, we have networks of light on silicon chips—it's unbelievable how far we have come from barely being able to propagate light.

OSA: What's likely to emerge in the next 10–30 years?

ML: Silicon photonics is already being introduced in some markets now.

The ability to lithographically define optical structures opens up new sorts of active structures. You can imagine a new silicon "tabletop" with filters, amplifiers, and other active devices that can be miniaturized and massively integrated on a chip.

These silicon photonics tabletops could be used everywhere that optical systems are being used. For example, almost every neuroscience laboratory has a very complex optical imaging system. Silicon photonics could be used to integrate these complex systems onto a small chip.

Another example is quantum optics, where silicon photonics are being introduced to replace the bulk optics used today, and to enable new functions impossible with current discrete components. All kinds of different applications are emerging in computing—the prospects are extremely exciting.


MICHAL LIPSON is the Eugene Higgins professor of electrical engineering at Columbia University. She is a pioneer in silicon photonics and invented key building blocks, including the gigahertz silicon modulator. She was a 2010 MacArthur Fellow, and is a fellow of The Optical Society and IEEE.

More in Optics