Yesterday I attended a public lecture in Aachen held by Professor Eli Yablonovitch from UCLA on the subject of photonic crystals. I went there with only a general idea of what to expect (something about optics in silicon and optical processors) but was completely blown away by the possibilities of the topic and how much progress had already been made.
Let me first say a few words about Yablonovitch: The guy made his PhD in 1972 and has been a professor at Harvard and UCLA since, but also found time to co-found his own company, called Luxtera. The list of his academic awards is lengthy and some of his better-known papers have achieved over 1,000 citations (the serious scientists substitute for comparing penis length). In short, one of academia's success stories. But not terribly impressing to me, simply because it's all second-hand praise. I wanted to see what he'd talk about. I was not disappointed.
His talk began with a short introduction on photonic crystals – basically they're the optical version of semiconductors. They have optical bands with small band gaps in separating them. This means that light of certain wavelengths cannot propagate in these crystals. The crystals can be doped to create states in the band gap and thus become controllable, just like electronic semiconductors. In practice, photonic crystals are often a sort of silicon grid, i.e., a silicon block with a lot of holes drilled into it in a certain pattern. Silicon is a material we can manipulate in awesome precision and which we know a very great deal about, so that's fine.
Now he swerved towards the applications side. Luxtera has been cooperating with Intel and ordering prototypes at a Freescale fab in Austin. Chip fabs are expensive as hell and very difficult to get access to for small companies or research groups, so he must have convinced some people that there's money in his work. Serious people. Serious money. Anyway, using that capability, Luxtera has been working on several optical build elements for chips. Waveguides, filters, switches, detectors. One thing they don't have is a light source on the chip – silicon lasers still suck. Therefore, an external laser guided in an optical fiber is the preferred option. But since fibers have a core diameter of 8 µm and the silicon channels conducting the light on-chip have dimensions of ~500 nm, some serious connection work is required. The solution is one of Yablonovitch's proudest discoveries, apparently: A spot of silicon with riffles applied in a particular pattern creates an interference effect that couples the laser beam into the waveguide at 60% efficiency – very smooth.
There's a number of other things photonics promise to be really good at – filtering, for example. However, purely photonic chips are still far off, so Luxtera seems to be betting on hybrid designs, where electronics and photonics cooperate. By going this easier route, Luxtera hopes to enter the market in a few years with chips for 10 Gigabit ethernet. They want to compete against electronic solutions by offering several times lower power consumption. A modest goal, but one for which solid success is quite possible and in case of failure will not discredit the entire field. I have to say I approve (not that anyone cares).
So, some fun numbers on what incredible things can already be done in the lab with photonic crystals:
1.) Low loss optical fibers
Normally, fibers have a core of optically dense (high refraction index n) material, surrounding by thinner mantle. With optical semiconductors, this can be inverted by enclosing a core of air inside a photonic crystal mantle. The result are loss rates of 1,7 dB/km (i.e. a signal can travel five kilometers and still have 10% of its strength). Theoretically, values down to 0.001 dB/km are possible, although we all know about "theoretically possible."
2.) Highly precise filters
Some of the on-chip filters made with photonic crystals have extremely high quality, which means they only pass a very narrow frequency range and block everything else. Numerically, this is expressed via the Q factor: Divide the main frequency which is allowed to pass by the width of the frequency interval which can reasonably pass the filter. Photonic filters can get Q = 1.8e6, meaning if you filter for 1 GHz, you will filter +/- 270 Hz. Enormous precision!
3.) Highly sensitive switches
This is connected to the filters. If you want a switch, take a filter and make up some way of changing its acceptance frequency. Immediately, you can switch the filter to "pass" or "don't pass." Due to the high Q, changing the refraction index of the crystal by as much as 0.001 is sufficient to switch. This is easily accomplished in numerous ways.
After he'd wined and dined us with the possibilities of photonics (most of which already work in the lab), Prof. Yablonovitch wanted to get a little "futuristic." He speculated on what would come [i]after[/i] photonics, the next big thing. In his opinion, it would be (new word!) plasmonics. Electronics use electrons to send messages, photonics use photons, plasmonics use… yes, plasmons. What are plasmons? Simply put, they are shock waves in the electron cloud. Imagine electrons inside a solid like wind in the atmosphere. Normal electricity is like a steady wind, sometimes rising and falling but normally a uniform stream. Plasmons are like the shock waves created by an explosion. They are compression waves inside the electron stream. Normally, they die off way too fast to be useful. But if you make the circuits really small, they hang around long enough. They also have some very interesting properties, most noticeably that you can create "lenses" for them which a have a refraction index of 100 or more. For photons, 5 is pretty much the limit (the lenses in glasses have 1.5 or slightly more). With such lenses, it's possible to focus plasmons incredibly sharply and in the focus spot manipulate them in various ways. By doing this, it is amongst other things possible to build a real optical transistor. Until then, electronic transistor are needed to control the photonic circuits.
The lecture ended there, and some discussion ensued. The lecture was part of a workshop taking place, so there were several highly qualified scientists sitting there shooting questions. All of my detail questions were answered in the course of this, so I didn't actually ask anything myself. I left the lecture thoughtful but enthusiastic. I tend towards caution when people laud their own company's accomplishments, but the potential was obvious, and it was much closer than I had hoped. It's a matter of years, not decades, until photonics will meet the market.