Slashdot Mirror
Light Emitting Silicon Steps It Up
h4mm3r writes "STMicroelectronics plans to announce a breakthrough on Monday in light-emitting silicon that could lead to a new generation of more powerful computing processors and more efficient automobile components as well as
potentially higher-speed optical data-transmission systems. (gotta register, free yadda yadda)"
Saab's newest 9-3 sport sedan has all of its electronics talking to each other over a fibre optic line - supposedly allowing more crap to be added later with greater bandwidth for cooler new features and stuff.
I don't really see how it will have an effect on me, but I think it is a cool idea in general.
Saab is owned by GM, soi I don't know if it is a trend that all of GM is heading towards, or if Saab is somehow special.
There are some odd things afoot now, in the Villa Straylight.
LIght Emitting Silicon? - LIES
;)
Any technology distinguishable from magic, is insufficiently advanced.
Light-Emitting Silicon Shines Much Brighter in New Invention
Why can't slashdot become a partner to NYT?
If you don't want to give google false page hits there's always majcher
Being able to integrate light emitting semiconductor devices together with electronic circuits is one of the holy grails of the semiconductor industry. Not only would the benefit come to increasing the speed of processors, optoelectronic devices would benefit greatly from this technology through integration. The cost savings and increased functionality would be incredible. Can't wait for it to become a reality.
"It's comin' back around again..." -RATM
I'm more curious about whether this could make photo-sensor diodes (read: solar cells) more efficient as well. That could lead to widely used poly-silicon being a reasonable alternative to Gallium Arsenide as far as power (whereas now, it's used solely due to cost).
120V 20A will make almost any chip glow!
Great! Now strippers can do their own light shows! (Okay, it's silicon vs silicone, but shhhhh, it's funnier that way)
One of the things this article really doesn't elaborate on is just how difficult the road has been to make efficient light emitting silicon. I had a professor as an undergraduate at the University of Rochester who spent a significant amount of time trying to get it to work. The article doesn't go into the technology, but I'm guessing they're using porous silicon. Porous-Si has small nanometer scale pores in (etched via electochemistry). The pores effectively alter the band gap of the silicon, increasing it to that of the compound light-emitting semiconductors such as GaAs. While this technique works well at generating light, the problem is getting it to generate light efficiently. Hence the exotic rare-earth materials such as erbium. I'm impressed that STMicroelectronics was able to increase the light output 100-fold. Extravagent claims such as these make me want to take a wait-and-see attitude. The process might be so difficult that it wont be practicle on high-performance chips for some time. Also, the processing techniques of light emitting silicon is different than for standard logic. I'd like to see how well these two processes can be merged.
"Case dismissed!"
---
Does this mean that part of the heat from the CPU will be light in the future? No more "monitor glow", more like "computer glow"... perhaps if different parts had different colours, e.g. floating point = green, integer = blue, cachemiss = red. Then you would know what part of your code to optimize without running a profiler: "It's all green and f**king slow, make your inner loops fixed point, dumbass!" :)
Here's something else (only slightly different):E G20021028S001 4
http://siliconstrategies.com/story/O
Basically, LEDs use Gallium and some other material because Silicon is horribly inefficient at photo-applications (its a electron band-gap thing, ask a physicist), but because its so cheap and GaAs is very not cheap, they still use polycrystalline Si for large solar cells.
Unfortunately, Leds are just too dim when silicon is used, so Gallium and whatever else (depends on wavelength) is still necessary there. By getting efficient light emitting Silicon, a whole pantload of money gets saved by avoiding Gallium.
End note: Why is Si cheaper than Ga? Refinement is more complex for Ga, Si is much more plentiful, and it hard to make large wafers of GaAs. Plus GaAs oxide (don't know the formula) is liquid at room temperature, so the only demand is photo applications (and stressed Si) because making IC with just GaAs means you can't use a liquid GaAs-oxide as a mask/gate/whatever.
This sounds like what they're going to be doing at first... The article points out that current opto-isolators need to be made with external components, whereas these would be made as a monolithic device. Still, opto-isolators are fairly cheap. I wonder how STMicroelectronics plans on selling these for cheaper. Eventually, I think the long term goals for this technology (if it proves to be really useful) is for use in high-performance logic chips. The problem with clocking large scale chips (such as CPUs) is that the clock signal has to arrive at all the gates at the exact same time. This is actually a very big challenge because resistance*capacitance slows things down. Trying to propagate a signal all the way across a chip to a large number of gates means that you need large driver transistors to supply the large current necessary. With optical clocking, you eliminate the RC time delay. You simply need to generate a pulsed optical signal and then make conduits across the chip to channel it to all the gates.
Of course, I'm guessing that is not as easy as it seems, which is why STMicroelectronics is making simple devices like opto-isolators. It could be several years before optical clocking is perfected.
No more need to buy cases with neon lights in them... just grab the latest motherboard from your favorite mobo manufacturer and voila :)
But in all seriousness, after I saw the article a while back (on slashdot) with something about optical traces on a motherboard in about 5 years from now, it had me very intrigued. I mean if you can shave a few nanoseconds from every bus cycle that's gotta be worth 10% increase in performance eventually. Especially on a clawhammer/sledgehammer where you've eliminated the north bridge part of the chipset.
For those of you who don't like to register...
Google News (Beta) Link
The next remark is false. The previous remark is true.
well now, let us see.. 100 times ZERO. Maybe that "new math" we all learned can help us with this.
Please forgive this post, I am a bitter III/V (read GaAs et al) guy
The AC that replied to you pointed out the possible benefits for solar cells, but...
The reason the Sun designer described it as "the holy grail" is timing circuitry on CPU's. What's the figure, something like 75-80% of a CPU is dedicated to timing circuitry? Think about what happens when you replace all that timing circuitry with a light pulse, and just pick it up wherever you need it. Eliminate all the wiring currently used to distribute the timing, and you get lower power, tons more silicon to devote to other things, and probably the potential for speed gains.
Wow, sorry but I need to correct you on a bunch of points.
Gallium Arsenide (not gallium) is used to make a variety of LED and semiconductor lasers. Silicon is unattractive for light-emitting applications because it has an indirect bandgap, making emission of photons much less efficient than in direct bandgap materials.
Making large wafer of GaAs is not so much a processing issue as a cost issue (i.e. how much would one wafer end up having to sell for, and would anyone at all even think of dropping that much money on one). HOWEVER, neither GaAs nor its native oxide(s) are liquid or even water-soluble at room temperature. You were perhaps thinking of Germanium. The problem with GaAs oxides is that they do not form into such nice layers as SiO2, and that they do not effectively passivate the GaAs surface such that MOSFETs cannot be fabricated. GaAs (and InP) and still widely used (in your cellphone for example), but in different ways than silicon and not nearly as widely as silicon.
Solar Cells == Photovoltaic Cells
:)
Solar Cells != Photo-Diodes
While both are PN junctions, insofar as the construction is concerned, photovoltaic cells actually produce a voltage, while photodiodes behave...differently.
When you have a photovoltaic cell, you need only connect a load to it to use EM energy for whatever work you need done.
With diodes, you have one PN junction (meaning P-type material on one side, and N-type material on the other.). To forward-bias the diode, you attach your positive voltage source to the P region, and your negative voltage source to the N region. This causes your electrons (called current carriers) to be pushed across the PN junction toward your positive voltage supply.
If you reverse-bias the diode, your current carriers will be drawn away from the PN junction, and almost no current, called leakage current, can cross.
All PN junctions are sensitive to light in that light striking silicon will produce current carriers(disclaimer: I'm only telling half the story...it can get confusion if you start considering "electron holes"...but if generation of free electrons bothers you, feel free.), wherever they strike. If they're particularly near the PN junction, they will serve to cause an increase in the leakage current, the external measurement of which is how the information is retrieved.
A rudimentary photodiode is simply a PN junction with a glass window.
There are all sorts of things you can do with semiconductors, doped or not. I keep seeing discussion proclaiming the downfall of semiconductors, but I wouldn't count on, say, quantum computing, to be able to function without supporting circuitry for the next twenty to thirty years. I hope to retire about then.
What's this Submit thingy do?