UIUC Researchers Create Light Emitting Transistor

thesilverbail writes "Researchers at the University of Illinois have demonstrated light emission from the base layer of a Bipolar Transistor. This discovery could be the beginning of an era in which photons are directed around a chip in much the same fashion as electrons have been maneuvered on conventional chips. It's reminiscent of the exciting days of the Miracle month November, 1947, when the transistor was first invented."

Damn

[cicatrix1]

If I would have known at the time I could have marketed the light emitting transistor after my discovery back in 7th grade industrial technology class.

Re:Damn

[Micro$will]

Sure, but assuming you've done something like I've done (try to switch several amps through a 2N2222 rated at .8 amps) isn't exactly considered "Solid State".

Re:Damn

[Patrik_AKA_RedX]

Among the years I've invented several new technologies in class:

The one-time self-deassembling capacitor. (put a high enough voltage on a elko to build one)

The flash diode. (build a bridge rectifier and replace one diode with a zenerdiode)

The automatic copper trace remover. (short a 7Ah lead accu through a trace on a circuit board. Capable of removing quite wide tracks)


Guess I should take a patent on those. They must have millions of possible uses.

Re:Damn

[jfw25]

The flash diode. (build a bridge rectifier and replace one diode with a zenerdiode)

Also referred to as the "Noise Emitting Diode".

Cool! Lights!

[YouHaveSnail]

From Univac to the Connection Machine, computers have always had lights that show what they're doing. I'd love to see a PowerPC or Pentium processor built with this tech and a little window on the chip package so you could look inside and see what the chip is doing! Cheesy, I know, but it'd be a fun and retro nod to our computing heritage. I'm not sure where you'd put the heat sink, though...

Re:Cool! Lights!

[deglr6328]

Can you see in the deep 'vacuum' ultraviolet range? Cuz that's what you'd need to be able to do in order to resolve the 90 nanometer transistors in a Pentium 4! :) Though maybe you could see large scale patterns in circuts lighting up for certain tasks....that would be neat.

Re:Cool! Lights!

[wiggling]

When first out of college I worked at a place that made 16-bit minicomputers. (No, you never heard of them.) They owned an H-P diagnostic device that plugged into the address bus and used the low-order 8 bits to drive the X axis of an oscilliscope and the high-order 8 bits for the Y axis.

You also hooked a sensor onto the memory read strobe (IIRC you could do choose between data fetch and instruction fetch on this architecture), and you could watch real-time memory accesses graphically -- it would light a

This doesn't seem to mean much

[Tau Zero]

First, there's no real magic here; the news release [uiuc.edu] says that the transistor is "made from indium gallium phosphide and gallium arsenide."

What this means is that someone has taken the same materials which emit light as part of a single-junction device (a diode) and have also made them do so as part of a bi-junction device. While this looks like it might be a good way to integrate light emission with the control circuitry, it's not going to do anything to make them easier to integrate into large devices (silicon works for this because its oxide, SiO2, is a pretty good insulator while gallium doesn't do anything so convenient).

I will admit that it's clever, and someone may find some unobvious way of turning it into a useful device (massively parallel optical interconnects?), but there's just no way that this is going to be slapped onto the next Intel or AMD die. It especially will not replace aluminum or copper interconnects between parts of one processor.

This isn't what you think

[Cyclone66]

It's an extension of the idea of the Light emitting Diode but it doesn't mean we'll have optical computers any time soon. The transistor still runs on current, it only outputs light! That means it can not drive another LET which means fully optical based devices aren't possible.

What this DOES mean however, is smaller devices in the realm of electronic/optical interaction. The basic functionality of this transistor is similar to a small combination of transistors and diodes except that this does it in one device. The device is rather large but it'll probably shrink rather quickly as it's still in the early stages of development.

Re:This isn't what you think

[grayrest]

Actually, it outputs both current and light. The light just happens to be a byproduct of the electron-hole recombination in the material. I do believe that the two outputs will be either the same or opposite depending on transistor type (pnp or npn) so you really don't get anything out of the transistor that you weren't getting already, you just get it in a different format. I'm sure, however, that someone more creative than myself will find some cool use for this.

You're extremely unlikely to see this type of technology in a microprocessor, at least in it's current BJT form. Most if not all microprocessors use MOS-type transistors because they're much easier to fabricate. BJTs are usually used in electronic amplifiers, though either can be used in either scenario depending on how the circuit is set up.

Re:This isn't what you think

[jralls]

The main reason that FETs are used instead of BJTs in logic devices is that FETs are voltage devices while BJTs are current devices. Logic is much easer to implement with voltage. Amplification, on the other hand, works better with current devices. Hence the BiCMOS processes which are common in the mixed-signal world.

As for these light-emitting BJTs, unless the light is emitted instead of heat from the resistance losses (which I'd think unlikely), the gain and efficiency of the transistor will be reduce

That's nothing

[TheOnlyCoolTim]

This one time I made some smoke emitting transistors.

Tim

Re:That's nothing

[Anonymous Coward]

This one time I made some smoke emitting transistors.

Once the smoke gets out, they don't work anymore.

Re:That's nothing

[eclectro]

Yes there is the Noise Emitting Diode (NED) too;

Noise Emitting Diode (NED)
When connected across a 1000 volt supply it makes a loud noise (once). The NED was discovered by Igor Pravaganda whom you'll recall worked many years trying to filter AC with electrolytics. He'll always be remembered as the father of the confetti generator.

I bought it as IR LED...

[SharpFang]

but besides Infrared, it emits visible light, noise and smell.

Electrolytic transistors work well as Boom generators, just reverse the polarisation.

End of interchip communincation problem era!

[stj]

This is great. It's not to do computations. Really, silicon is good and fast enough for that right now (of course it would be real cool if possible, but that won't work). However, couple this effect with phototransistors which already are in the domain of high frequency (at least theoretical results suggest so - 10 GHz here [aps.org]) and you end of with the dream of board engineers - forget wiring chips with metal. Interconnect them with fiber! Certainly with germanium it's gonna bit a bit difficult, but it's worth it: all you need to wire to chips is power, end of distance limit between chips, and forget problems of spilling something on the board! Even with current germanium version it'll work perfect for connectors between boards.

Re:End of interchip communincation problem era!

[SW6]

[...]all you need to wire to chips is power, end of distance limit between chips, and forget problems of spilling something on the board! Even with current germanium version it'll work perfect for connectors between boards.

You haven't thought this through, have you? One of the problems with chips is that they can't get signals between points on the chip fast enough because the speed of light limits it. So how is light going to make it any faster when it travels at, erm, the speed of light?

Re:End of interchip communincation problem era!

[stj]

You are right, however you are wrong ;-) Well, speed of light is not the only problem. One of the major complications in electrical wiring is that at high frequencies even a couple inches turns into long distance telecommunications and you have to construct actual transmission lines with matching impedances. Using optics eliminates this problem. Now if you talk about speed of light, at the distances of inches that's far less relevant than you think. Typically much more problem is with the transmission rate

Re:End of interchip communincation problem era!

[stj]

Actually they happen to be about the same. Copper wire speeds vary between 0.4 and 0.95c, depending on the desing of the cable. Typical assumption is 0.66c. Fiber optics propogation speeds vary, too. However the most common assumption is 0.66c, because SiO2 (silica glass) has just that factor (approximately).

just higher switching speeds

[ndevice]

Since switching speed is also a function of the size of the device (capacitance), this transistor probably wouldn't produce any benefit over the LED version.

The article seems to be saying that they could get higher switching speeds out of this transistor, but we still have inefficiencies here: either the amplified output is used, or the optical output is used. Why would it be necessary to use both? And if the benefits come from not having to drive interconnects between the transistor and LED, we already have integrated transistor/led packages on the same substrate. It might be a better idea to make the emitter current the light emmiting part because then you'd be able to save on component costs.

While this is kind on impressive...

[blankinthefill]

I don't think it heralds in any type of light passed processing age. There is still the unresolved problem of creating a sensor that is sesitive to pick up a small amount of light, let alone making such a sensor small amd efficient. Untill that occurs, then this is just another advance that we probably can't use for what we think we can.

Nothing to get excited about

[JGski]

Since this was done on GaAs/InP, it pretty much assures that the massive adoption date is anywhere from Today+20years to Today+infinity. In case you hadn't heard, GaAs, and III-V in general, has been the "next big thing" in semductors for nearly 30 years now. It's pretty much a niche technology only used when Si absolutely can't do the job at any price because the price to do III-V is usually so high in capital and tech issues that Si, even when price inflated 1-2 orders of magnitude, is *still* overall cheaper.

Things that would get my attention:

implemented in SiGe instead of III-V materials

bidirectional transduction in O->E not just O->E at usable efficiencies

demonstration of integration into "conventional" manufacturing processes

Otherwise, it's an interesting academic exercise that might lead to the above points, which is "A Good Thing".

my use

[OwlofCreamCheese]

I say we use this to make TVS with resolutions of ten trillion by ten trillion. then you can like... take a microscope to your tv and see the bacteria on the actors. I bet james earl jones has herpes.

reminiscent of the Miracle month

[frovingslosh]

It's reminiscent of the exciting days of the Miracle month November, 1947, when the transistor was first invented.

Oh yea, it's just like that, except that in this case we'll never hear about it again and in five years you'll do searches on it and not be able to find anything on it except the original articles. I could easily list 100 things this exciting or more that I've seen that just seem to vanish. Interesting news, but I think it's way over optomistic to equate it to the discovery of the transistor.

Re:reminiscent of the Miracle month

[dustmote]

Ah, but what was the *most* interesting one? That's the one I'd like to see more research on. :)

Big deal

[ScottForbes]

I invented a light-emitting resistor in a UIUC circuit design course over ten years ago. It only worked once, though, and it burned my fingers.

Where this may be useful...

[hobit]

The idea that this is somehow enabling optical computing is incorrect as these things can't take light as input. One could use a photodetector of course, but it doesn't seem like a big deal, one could use LEDs to do this today...

Debugging hardware could become simpler however. You could "see" (or have a computer watch) exactly what each transistor is doing. This could greatly enhance the ability to test real chips. Of course the transistors are usually buried under the metal layers, so I guess it depends on how bright the transistors are or if the designers are willing to leave (lots) of gaps in the metal layer for at least the "important" transistors.

Mark