Catch up on stories from the past week (and beyond) at the Slashdot story archive

 


Forgot your password?
Close
typodupeerror
×
Science Hardware

Turning Microchips Into Lasers 12

Posted by michael from the I-will-not-blind-people-with-my-pentium-7 dept.
An AC sends news of this New York times article on trying to use photons instead of electrons to make much faster chips. "Not available in stores near you".
This discussion has been archived. No new comments can be posted.

Turning Microchips Into Lasers More

Turning Microchips Into Lasers

Comments Filter:
  • all-optical repeaters are already possible, using rare earth elements, I believe erbium is a favourite.
  • > you could create several streams of photons in several directions at once from a multi-facited crystal

    You must be careful with your trig and margins for error. This design may fail in proximity to a massive object where the curve of space affects the path of the photons. You cannot increase the information density too much as a beam may strike it's neighbours receptor by mistake.


    --
  • > Say electrons move at 10^5 m/s (somebody help me out here)

    Electrons in silicon and metal circuits only move at a few mm/s but the change in electrical potential moves along very rapidly (like waves in water).

    A transistor in a microchip will switch when the fast moving wave moves a few extra electons into the semiconductor junction changing it's electrical properties.

    With light, each photon will be a part of the signal itself, so you need to find a crystal which needs a certain number of photons entering on one axis in a given time to prevent photons from passing through another axis (or some similar mechanism). This should present no heat problem if excess photons simply pass through unimpeded - so you have a sort of photonic capacitance, a light equivalent to the FET.

    Of course things are really much more complicated.
    --

  • This has been discussed before, when various companies have come out with optical switches. The trouble is not ampification (Erbium-Doped Fiber Amplifiers, EDFA's, are a stock part at many laser suppliers) or raw switching (see recent slashdot articles), the problem is packet switching. In order to read the header on a packet and decide where that packet should go, you need to convert back to electrical. There's where you take your performance hit. What's really needed for hugely fast networks is all-optical switching logic. And while I can see how you might make some incremental improvements when you bring the electronics closer to the lasers, I don't see a revolution in tech here.

    This story is about being able to integrate light and electronics. There are lots of different types of laser and detecting diodes, but to integrate them directly onto a single package, you currently need to fuse different chips onto a single substrate. This process is not the easiest or cheapest way to go. If you can integrate the photonics and the electronics onto a single chip, it would shortcut a bunch of the multi component issues.

    The article also tries to hint toward all-optical computing. As far as I can see, this work makes no real steps toward all optical computers. Claims five or ten years ago about optical computing have proved disappointments. No one can get it to work well. As far as I can see, this only is a possibility for better integration between electronics and optics, not a fundamental component of an optical computer. I could be wrong, of course, so if anyone out there knows, I'd be eager to hear how to build one with these.



    BMagneton
  • Actually we already have those. Erbium doped fiber amplifiers have been used for a while to combat the attenuation of fibers. They are simply a segment of fiber that has been doped with erbium and are then pumped, this then acts as an optical amplifier. I believe semiconductor lasers used to be used for this, but that was probably about 10 yeas ago.

    The difficulty with optical routers is trying to process and split the data optically.

  • Wait a minute : do someone know anything about the order of magnitude it takes for an object to curve photon trajectory like this ?
    I know that the trajectory must only be slighlty bended to do what you're meaning here, but earth or even sol are nowhere near to have the gravity field of a neutron star. And if space is bent, should the cristal be bent as well, so you problem is moot ?
    As well, since we are nearing science-fiction : what about micro-optical fibers ? Too much idea for a saturday morning anyway...
  • "We're still workin' on it, but we'll make you think we'll be done real soon so you keep giving us money."
  • I was under the impression that you could create several streams of photons in several directions at once from a multi-facited crystal, thereby increasing efficientcy. Another problem in nanotechnology is how to make the machines as small as possible. Currently, we are restricted to the width of the atom that is being used to transfer data. With lasers, we can use one crystal to "beam" the data to another in a string of photons. While the width would still be resricted by atoms, the ability to "layer" the information would be exponentially better. A network of crystals could send a info in a wafer pattern that could not be done in smaller electronic devices w/o having an insulator between wires (silicon). Add the speed of light to that, and you get a great increase in efficentcy with a milionth of the space.
  • ...processing using photonics - most of it dealt with simply producing light from silicon chips. Not too sure how you would process using photons. And what sort of perfomance gains would you be looking at?

    [implausible_math]Say electrons move at 10^5 m/s (somebody help me out here) and photons (in Si) move at 10^8 m/s. In a circuit 10^-2 m long, electrons would take 10^-7 s (100 ns scale) to give 10^7 ops per second while light would take 0.1 ns to give theoretically (10^10 ops per sec - 10 gigaflops).[/implausible_math]

    All my assumptions were b.s., but is a thousand-fold increase theoretically "in the ballpark"? Anybody?

    [doh]Photons are 10^3 times as fast as electrons, naturally I came up with a 10^3 performance boost![/doh]

    OTOH, I'm not sure I'd want a glow-in-the-dark CPU...
  • As I understand it, it's not much a matter of particle propagation speed that a matter of wave frequencies, as pointed out in previous comment. (please forget still pending debate of particle vs wave behaviour).

    The problem with electrical wave is that the higher the frequency, the less they tend to stay in the wire you're using to transmit them. Stating it otherwise : the higher the frequency is, the more likely you wire is turning into an antenna. And the signal carrying your bits vanishes into the air.

    With light, the problem is totally different, because once you get it into a optical fiber, it does not try to get out ( into normal conditions).

    As for frequencies, here is an example : a typical visible wavelenght is around 300nm, which gives you about 100000Gz. Even if that signal is modulated a 1/1000 of the main frequency, you have a frequency 100 time greater than the current 1.5 gz boasted by who you know.

    The only problem left to solve is to make diodes, transistor, that work with light...

  • I have posted the text from the article, so you don't have to log-in/register.


     December 7, 2000

    WHAT'S NEXT

    Researchers Trying to Find Ways to Turn
    Microchips Into Lasers

    By IAN AUSTEN

    WHATEVER their many talents, silicon microchips aren't very bright. That is, they can't produce significant amounts of light, certainly not enough to act as tiny lasers. But because photons -- the basic units of light -- make electrons seem relatively sluggish, a silicon chip that could spit out laser light is the stuff of dreams for engineers looking for ever faster computer chips.

    Light-producing silicon chips could even revolutionize consumer gadgets. A single silicon chip that could both calculate and glow, for example, could slim down laptop computers to the thickness of a credit card.

    Even a real-life variation of the science- fiction film "Fantastic Voyage" might be possible, but instead of a miniature Raquel Welch traveling through a patient's veins in a tiny submarine, this adventure would feature silicon microchips using their own laser light to guide a path through the human body and diagnose medical problems.

    Professor Lorenzo Pavesi of the University of Trento in Italy, says that the potential uses for such "thinking lasers" would be limited only by the imagination.

    Research led by Dr. Pavesi has overcome at least one hurdle between current reality and flights of fancy. While Dr. Pavesi's researchers have not created a silicon laser, they have managed to coax light from special silicon crystals, enough light that they may have taken a step toward integrating optical systems into silicon.

    Lionel Kimerling, the director of the Microphotonics Research Center at M.I.T., is developing a different approach, which keeps the light source separate from chips, but he does not dismiss the value of the Italian research. "We're just at the birth of integrating photonics," he said. "There is tremendous power from integration."

    Like many research groups trying to force light from silicon, Dr. Pavesi's team began working with conventional bulk silicon crystals. They found two problems. Not only was bulk crystal a particularly poor light producer, but its feeble light output diminished over time.

    Dr. Pavesi then tried silicon nanocrystals, very tiny objects made by manipulating individual atoms, molecules and groups of molecules. His group zapped a wafer laced with nanocrystals with an intense blue laser. Out came more light, now colored red, than went in. In fact, the light amplification was on about the same level as is achieved with more costly gallium arsenide crystals now used in lasers, Dr. Pavesi said.

    Because of their size, Dr. Pavesi said, silicon nanocrystals can be packed very densely. A large light gain, he said, requires a lot of nanocrystals.

    Dr. Pavesi said that the light's color could be changed by simply altering the size of the nanocrystals in the silicon sandwich or adding other chemicals.

    There are at least two very big problems to overcome before Dr. Pavesi's glowing nanocrystals become part of the first silicon laser. The photons that come out of Dr. Pavesi's chip do not produce the intense light of a laser beam. It will be necessary, Dr. Pavesi said, to surround the nanocrystals with two tiny mirrors facing each other.The light produced from the nanocrystals would bounce off the mirrors, building power before leaving the chip.

    The far more difficult problem is finding a way to make the chip generate light on its own instead of boosting light beamed against it from the outside. With regular lasers, electricity provides the energy that becomes light. Unfortunately for silicon laser researchers, silicon is the material of choice for microchips partly because it is a very good electrical insulator. "There are some tricks that can be worked on," Dr. Pavesi said.

    Dr. Kimerling said that some researchers claimed to have sent electricity through silicon, but said that their evidence was not convincing. Using silicon nanocrystals, he said, will not reduce the problem's difficulty. But Dr. Kimerling said that if Dr. Pavesi was able to change the color of light produced by his nanocrystals simply by altering their size, they could be useful in fiber optic communications -- even if they were simply amplifying external laser light. Fiber optic systems can increase the amount of data they carry by running several laser light colors simultaneously.

    Citing the history of lasers made with more light-friendly materials, Dr. Pavesi is reluctant to offer even a vague estimate of when consumers might see the first, ultra- slim, one-chip laptops based around silicon nanocrystal technology.

    "In technological development," he said, "you can have one bottleneck and that takes 10 years to solve. If we are lucky it will be 5 to 10 years to market. On the contrary, it may happen that due to engineering difficulties a silicon laser will never reach the market. It's very difficult to predict."

  • Optical-to-optical amplifiers would be useful now. (Score:3)

    by re-geeked ( 113937 ) on Friday December 08, 2000 @07:35AM (#573492)
    Even if this research doesn't create silicon lasers, the use of it to amplify fiber-optic signals would be of great benefit in today's hot new networking tech: all-optical routers.

    Today, if you want to amplify a light signal (they attenuate, too, at least when traveling through fiber) you need to convert the signal to electrical, then retransmit as optical. This is a big speed hit.

    The article does mention this briefly, but this practical short-term application gets lost in the breathless predictions of a glowing future.

    Does anyone have any idea whether the response of the silicon would be fast enough to improve on the current situation?

Slashdot Top Deals

"If you are afraid of loneliness, don't marry." -- Chekhov

Close