[extropy-chat] Photonics

Dan Clemmensen dgc at cox.net
Sun Apr 3 14:38:21 UTC 2005

Mike Lorrey wrote:

>A few years ago we were reading articles on this list of an
>IBM-Northwestern U project demonstrating quantum well photonic
>circuits, in which the researchers were stating that their circuits
>would be 1000 times faster and 1000 times smaller than electronic
>circuits. What happened to those claims?
I just did a Google toe find those articles.

Yes, I did recall them correctly. The articles you refer to conflated 
three distinct technologies. In retrospect, this was a deliberate 
strategy to get in on the dot.com financial boom. Here is what I found: 
In 1998, the technology for manipulating photons was still mostly at the 
level of discrete components. This included WDM multiplexers and 
demultiplexers, EDFAs, and lasers. at the time, any approach using 
lithographic techniques to manipulate photons promised to reduce the 
cost of photon manipulation by orders of magnitude.

Yes, the researchers were finding ways to use quantum coupling for 
multiplexing and demultiplexing.

However, the inputs and outputs were still  "photon wires," and the 
operating wavelength was 850nm, so the "photon wires" were in fact at 
least 850nm in diameter. The best that could possibly be hoped for was 
to reduce the mux and de-mux to zero, which would still leave the 
"photon wires" as the dimensional constraint. Even with 32 discrete 
signals per wire, this approach was still "bigger" than the dimensions 
of a purely electronic circuit.

In the event, quantum coupling of photonic wires never actually became 
viable, The closest we ever got to a cost-effective replacement for 
discrete optical components was the AWG (array waveguide.) AWGs are huge 
by comparison to electronic circuits, for obvious reasons.

Quantum wells are still being investigated, but only in the electronic 
domain, not (as far as I can tell) in the optical domain.

On a personal note, after spending nearly 30 years on the electronic 
side of telecommunications, I shifted to core routing in 2000 and spent 
four years pursuing this dream. From 2000 to 2004, I worked with two 
core router startups. We were attempting to take advantage of just such 
technologies as you describe, but at one remove: if photonics makes 
long-haul circuits cheaper, then surely the industry will be in 
desperate need of better/cheaper/faster scalable core routers.

In the event, the less-radical discrete techniques proved more than 
adequate to decrease the cost of long-haul bandwidth to the point that 
it is now a "free"commodity, Telecoms costs are now driven by the last 
mile rather than by the long-haul cost. The last mile, in turn, is 
driven by regulatory constraints.

During the period from 1995 to 2005, the technology has increased the 
capacity of an individual long-haul fiber from about 200Mbps to about 
6.4Tbps. This is a factor of 16000, which is a whole heck of a lot 
faster than Moore's law. The world now has a gross overcapacity of 
long-haul fiber. Instead of digging new trenches or laying new undersea 
cable, the long-haul providers can add new capacity by "merely" 
deploying new multiplexers to increase the capacity of existing fiber.

Here is the conflation: the following statements are true:
  1)new techniques involving quantum wells can reduce the cost of 
photonic manipulations by orders of magnitude
  2) the quantum resonators are electronic.
  3) quantum-well electronics has the potential to supersede coventional 
electronics and reduce the power and increase the speed by a factor of 1000.
  4) lithographic photonics has the potential to supersede conventional 
photonics and reduce the power and increase the speed by a factor
of 1000.

Each statement was true (or at least defensible) in 1998. However, The 
statements have essentially nothing to do with each other. The photonic 
components are still two or three orders of magnitude larger than the 
electronic components.

More information about the extropy-chat mailing list