MIT Researchers Make Advance Toward Photonic Circuits

MrSeb writes with this excerpt from an article in Extreme Tech: "Light-emitting diodes are a cornerstone of consumer tech. They make thin-and-light TVs and smartphones possible, provide efficient household, handheld, and automobile illumination, and, of course, without LEDs your router would not have blinkenlights. Thanks to some engineers from MIT, though, a new diode looks set to steal the humble LED's thunder. Dubbed a diode for light, and crafted using standard silicon chip fabrication techniques, this is a key discovery that will pave the path to photonic (as opposed to electronic) pathways on computer chips and circuit boards. The diode for light — which is made from a thin layer of garnet — is transparent in one direction, but opaque in the other. Garnet is usually hard to deposit on a silicon wafer, but the MIT researchers found a way to do it."

Comments

New we need a real one way mirror.

[wheeda]

New we need a real one way mirror.

What's it got to do with LEDs

[Zouden]

The summary (taken from the first article) implies that these new diodes are going to supersede LEDs, but they have completely different purposes. LEDs make lights, these things don't.

Re:What's it got to do with LEDs

[zlives]

umm they both have light and diodes... so that's 2 out of 3... right

Re:What's it got to do with LEDs

[snookerhog]

furthermore, they may want to rethink the name if "diode for light" is going to become DFL

Re:What's it got to do with LEDs

[The Askylist]

May I suggest Light Absorbing Diode? LADs could make even the Kindle cool.

Re:What's it got to do with LEDs

[froggymana]

furthermore, they may want to rethink the name if "diode for light" is going to become DFL

Hopefully it's dfl on their list of names.

This might actually make it

[bigredradio]

From the article -

The whole system could be made using standard microchip manufacturing machinery, Ross says. “It simplifies making an all-optical chip,” she says. The design of the circuit can be produced “just like an integrated-circuit person can design a whole microprocessor. Now, you can do an integrated optical circuit.” That could make it much easier to commercialize than a system based on different materials, Ross says. “A silicon platform is what you want to use,” she says, because “there’s a huge infrastructure for silicon processing. Everyone knows how to process silicon. That means they can set about developing the chip without having to worry about new fabrication techniques.”

It is good to see someone is coming up with an innovation that can "actually" be introduced. Seems like I read about new innovations every day on slashdot that never get off the ground because completely new manufacturing processes need to be created. Hopefully this will actually make it because it requires fewer changes by manufacturers (which can be significant barriers to innovation).

Re:This might actually make it

[kelemvor4]

manufacturers (which can be significant barriers to innovation).

They certainly can be!

light transistor

[greywire]

Does this mean a light transistor is coming soon?

I am wondering if there is any material that acts as a mirror and can be switched from reflective to transparent electronically? I assume there is not or you wouldn't have devices like MEMS displays. I'm thinking if you had such a material it would be essentially a light transistor.

Re:light transistor

[Anonymous Coward]

Well we currently have devices that can become opaque or transparent depending on signal.

Also, couldn't metamaterials technically be used in this way?
That EM blackhole that was created could be turned in to a one-way channel for light instead of just an infinite spiral to absorption
Reflective could be a mirror that punts it back through another channel that comes out at the same input, shift mirror to allow it through.
Of course, that involves mechanical stuff (even if incredibly tiny), but the use of photonics could possibly decrease the heat and space requirements for boards, so a significantly larger number could probably be used.

Re:light transistor

[Calos]

Light transistors? Not exactly as you describe, that I'm aware... But things that fit a more general description of "light transistor" do exist. They're basically optical ring resonators, but add some standard silicon doping and bias it, you can change the refractive index to stop the resonance. I'm not exactly up on my photonics, but I think in theory you could couple the ring resonator to another pathway which would ultimately go to the display, with the applied voltage controlling whether or not the source signal is coupled through to the display.

Not exactly what you have in mind, probably not efficient or manufacturable.

Re:light transistor

[wooferhound]

I am wondering if there is any material that acts as a mirror and can be switched from reflective to transparent electronically? I assume there is not or you wouldn't have devices like MEMS displays

Maybe a DLP mirror system ?

Re:light transistor

[greywire]

Yeah, thats the MEMS I was referring to. I was wonder if there was a purely solid state way of doing it.

Re:light transistor

[greywire]

I'd like to be able to reflect something back and forth, essentially keeping it in a holding pattern until such time as I wanted to release it through one side. The length of time of course would be very very short, and the switching time from transparent to reflective would need to be fast.

Re:light transistor

[amliebsch]

Liquid crystals respond to *electric* charge. An optical transistor should not require electrical charges at all. In other words, something like a liquid crystal that was activated or deactivated by light. Theoretically you could do this by slapping a small PV cell on to an existing crystal panel, but the goal here is to be able to fab everything out of silicon like we do with electronic integrated circuits.

What I can't wait for is the electricity-emitting light diode.

Re:light transistor

[_0xd0ad]

Liquid crystals to turn on/off an optical pathway.

...slowly

Re:light transistor

[greywire]

Yes, of course.

However I don't think you can control when the light is released. You don't turn the mirrors on and off, the light simply escapes when its reached a high enough intensity.

I have lots of questions

[Hentes]

Could someone with expertise in the topic tell me:

• How does this replace a LED? It's not a light source.
• How does this allow creating optical circuits? You need more than just diodes for a logic gate.
• If I make a sphere out of this, will it act like a black hole?

Re:I have lots of questions

[Wookie Monster]

Making a sphere out of this should be no different than making any type of black colored sphere. The light is converted to heat and the sphere warms up. If the heat is allowed to slowly escape, then equilibrium is reached. Otherwise, the sphere melts and potentially destroys its heat absorbing properties. Or the liquid sphere radiates heat away. In a black hole, heat cannot escape at all.

Re:I have lots of questions

[kelemvor4]

Could someone with expertise in the topic tell me:

• How does this replace a LED? It's not a light source.
• How does this allow creating optical circuits? You need more than just diodes for a logic gate.
• If I make a sphere out of this, will it act like a black hole?

From TFA:

Basically, it’s now possible, with regular chip-fab tools, to create an integrated silicon circuit with optical, rather than electronic, interconnects — both internally, and between other chips.

TFA makes it confusing by bringing up blinking lights used to convey various status to humans via light, which does not appear to be related.

In the near term, though, garnet-on-silicon chips are likely to be used in networking — first in backbone routers, which are physically huge and very power hungry because of the current size of optical switching hardware, and then hopefully at home and in the office (100Gbps home networks!)

It's all in the poorly written article.

Re:I have lots of questions

[kelemvor4]

This article is just a rewording of the original article: http://www.mit.edu/newsoffice/2011/optical-computing-diode-1123.html
Fortunately, the kids at ExtremeTech were good enough to at least link the original which isn't nearly as confusing.

Re:I have lots of questions

[Hentes]

So what did you want to achieve by quoting the parts of the article I was asking about?

Re:I have lots of questions

[Anonymous Coward]

It's not a Light Emitting Diode, it's basically a Light Diode. Light can only pass through in one direction. The way this is normally done is by rotating the polarization of light through two polarizers. Any light of the wrong polarization cannot return back to the light source. This is usually used to protect laser sources and their modulators from return loss reflections in fiber optic systems that make use of polarization maintaining fibers.

Anyway, a 'true' optical transistor can be fashioned out of this if coupled with an optically controlled gain medium. If you have optical transistors, you can create optical NOR or NAND gates, and all basic logic functions can be created solely from these gates.

Just skimming the actual Nature paper itself, it appears they've basically created an optical isolator on a planar optical waveguide circuit.

To answer your last question: no.

You'd be better off with black construction paper for that.

Re:I have lots of questions

[Anonymous Coward]

If I make a sphere out of this, will it act like a black hole?

No. Black hole studies typically utilize Dark Emitting Arsenic Diodes, (DEADs), not Light Emitting Diodes, (LEDs.)
DEADs don't really emit darkness, but rather just absorb light. The more current that's applied, the more light they absorb.
Originally developed by the Department of Defense for use as a type of energy absorbing material coating, (e.g. a laser weapon shield).
As laser research produced more and more powerful lasers, DEADS were tuned to absorb a more powerful beam but of a narrower frequency range.
It got to a point where DEADs were developed that could completed shield against a lasers of a certain frequency, but were completely ineffective against any other frequency. The project was finally officially cut for expense overruns.

Interesting concept

[colin_faber]

Assuming you don't `leak' light like you do with electricity thought a traditional transistor gate when scaling way down this technology could provide a method of continued packing of high speed transistor like elements. The next interesting question would be how hot the chips would actually run, considering we're really no longer resisting, resulting in high temperatures.

Re:Interesting concept

[Semptimilius]

Well, much of the leaking in traditional electronic transistors is due to quantum mechanical effects, which would still apply to photonic devices. (With differences arising from such things as spin.) Some people are using evanescent fields from thin fibre lines to actually couple the signals in the line to other devices.

Re:Interesting concept

[marcosdumay]

You surely leak light by the same mechanisms you leak electrons (tunneling), it is just that with light we are used to it, but still don't have a useful workaround. Optical elements also disperse light in a way that is quite similar to a conductor dispersing elecrons, and they also absorb the light (what have no equivalent for electrons).

The biggest advantage of an optical device is that it can act on several signals at once, in a SIMD way.

The MIT article is much better...

[divisionbyzero]

It doesn't go off and start talking about LEDs and WDM which just confuses the issue.

http://www.mit.edu/newsoffice/2011/optical-computing-diode-1123.html

Re:Still too low

[Anonymous Coward]

They did... they just don't interact with the real world

Switching speed.

[fish_in_the_c]

Just because you can use it to store data doesn't mean it will make a good processor any time soon.
how long does it take to switch states. That's why flash ram is not used in primary memory.

Re:Hasn't it been supposedly "paved" several times

[rubycodez]

Things don't work that way in the real world, kid. Have a lollipop.

I'm confused (happens a lot)

[The Living Fractal]

So if I make a 'photonic circuit'... doesn't some kind of electrical impulse have to modulate the light pathways, i.e. turning them on and off to allow light through, in order to create logic structure? i.e... doesn't this still require transistors? Or does this material actually *change* like a transistor gate when subjected to light, and change back when the light is taken away... Now that I could see making logic gates with.

Re:I'm confused (happens a lot)

[HiThere]

Not if it's a real photonic circuit. Of course, we don't have many of those right now. A few telescopes, but not the ones with photon amplifiers. A few microscopes, but not many of them, either. Mainly the only current "photonic circuits" are experimental. (To be photonic rather than just optical you need to have light intentionally modifying the signal that other light carries. Usually this only happens by accident, and degrades the signal.)

Most of the things that we have that approach photonics are actually electronic/optic hybrids.

(I may be overstating the current presence of photonic circuits. I'm giving interpretation that favor that to some things I read that could be interpreted either way.)

Another overhyped materials science article

[Animats]

We get at least one of these overhyped materials science articles each month. This time, someone has figured out how to deposit a garnet layer in a wafer fab. This is blown up into "photonic computing real soon now". It's not.

There's a lot of work in progress (PowerPoint) on optical on-chip interconnects. This is not "photonic computing". It's clusters of CPUs with a network of optical interconnects, all on one IC. The CPUs are still made of transistors. IBM has a very active research program in this area. But it's a long way from working. There are optical switching elements that work experimentally, but nothing ready for volume manufacturing yet. The optical interconnects themselves aren't considered to be the big problem.

So far, most of the proposed approaches involve un-buffered circuit switched networks. An optical connection is set up from CPU 1 to CPU 2 by electrical means, and then data is blasted across it. Circuit setup time is long compared to the data rate. So this is for long messages within a cluster, not cache synchronization. Think (inevitably) Beowulf cluster on a chip, not thousand-CPU shared memory microprocessors. The technology may also be useful as a network optical switch.

Short version: when this all works, servers get more densely packed.

Re:Hasn't it been supposedly "paved" several times

[timeOday]

Sadly, progress is not made in academia. It's made in the market. Until there is a photonic device on the market they are simply blowing smoke... Wake me up when they have actually got something on the market.

This mentality is a big part of our national decline. Nobody wants to make the investments or do the hard work. They just want to swoop in when the technology is ripe for commercialization and reap all the profit from others' years of investment. Individuals and big companies act this out in different ways, but it boils down to, "just wake me up when I can get it on sale at Walmart."

Great for permanent data storage, I guess...

[CCarrot]

This could be great for long term, non volatile data storage, I suppose, but unless they develop an efficient method of changing the state (i.e., which direction is opaque), I can't see this being much use for processing in general.

On the plus side, maybe we'll finally get to see those data cubes/crystals that popular SF books are always referring to...:-)

Re:Hasn't it been supposedly "paved" several times

[ragzz]

I agree with "Nobody wants to make the investments or do the hard work....just wake me up when I can get it on sale at Walmart." Remember that even 'normal transistors' were ideas conceived and realized in 'research labs' and definitely not the market. And there are companies which (search for infinera) are doing things like 'photonic integrated circuits' (the signal processing is in electronics and transmission is optical); and they're better off than blowing smoke.

Thermodynamics, anyone?

[ResidentSourcerer]

My reading of the summary says that they claim that is opaque in one direction and transparent in the other.

Consider the following thought experiment:
Assume that this material works the same way for far-infrared as it does for visible light.

In a well insulated cylinder place a sheet of this material across the interior of the cylinder so that light can pass to the left, but not to the right.

Now the material in the right end of the cylinder is emmitting infra-red radiation. It can pass through the material. This warms up the left side, and cools off the right side.

Now we have a temperature difference, so we can run a heat engine. Voila! perpetual motion.

Now if the material doesn't do this with room temperature radiation, then we have to put it in a red hot environment. Will the material still work at those temps?