janrinok writes:
"A report from Vanderbilt Institute for Nanoscale Science and Engineering brings news of a breakthrough in miniaturization that could have a major impact on hardware in the future."
From the report:
An ultra-fast and ultra-small optical switch has been invented that could advance the day when photons replace electrons in the innards of consumer products ranging from cell phones to automobiles.
The new optical device can turn on and off trillions of times per second. It consists of individual switches that are only one five-hundredth the width of a human hair (200 nanometers) in diameter. This size is much smaller than the current generation of optical switches and it easily breaks one of the major technical barriers to the spread of electronic devices that detect and control light: miniaturizing the size of ultrafast optical switches.
The ultrafast switch is made out of an artificial material engineered to have properties that are not found in nature. In this case, the 'metamaterial' consists of nanoscale particles of vanadium dioxide (VO2) a crystalline solid that can rapidly switch back and forth between an opaque, metallic phase and a transparent, semiconducting phase which are deposited on a glass substrate and coated with a 'nanomesh' of tiny gold nanoparticles.
(Score: 3, Funny) by GungnirSniper on Wednesday March 19 2014, @08:37AM
By the time this is in production, Microsoft Windows will need 25GHz and 1TB of RAM just to boot.
Tips for better submissions to help our site grow. [soylentnews.org]
(Score: 5, Funny) by mrbluze on Wednesday March 19 2014, @08:43AM
Nonsense, 640TB should be enough for anyone.
Do it yourself, 'cause no one else will do it yourself.
(Score: 3, Funny) by MrGuy on Wednesday March 19 2014, @12:48PM
..by next Thursday, then?
(Score: 5, Interesting) by Covalent on Wednesday March 19 2014, @09:43AM
It was my understanding that tunneling effects were the limiting factor for electronic chips and that these really start to be a problem at around 10nm. Perhaps I am just ignorant of how these switches are going to be used, but 200nm seems an order of magnitude too large. Impressive to be sure, but small enough to be a game changer?
You can't rationally argue somebody out of a position they didn't rationally get into.
(Score: 5, Informative) by VLM on Wednesday March 19 2014, @12:50PM
I had the same confusion in the opposite direction. So its an unholy expensive cutting edge technology to make "near UV" LEDs that at least temporarily at great cost and difficulty and relative ineffiency radiate 400 nm light. So LOL at the idea of running lower than 400 nm or so.
A little more realistically "most optical telecom" work is doing in near IR around 1300-1500 nm because glass has an optical window of low loss around there and besides, even if you're not using glass fiber, why not piggyback on decades of tech developed in near IR. So most likely running at 1300 nm or so.
I've done a modest about of microwave RF work with waveguides and the like and I'm mystified how you optically switch a 1300 nm wave using a switch thats only 200 nm in size. How does this "work" exactly? It could be that the explanation was blocked by the journalist filter.
Maybe its like a mirror. You can reflect 2 meter wavelength RF signals perfectly well using a thin cookie sheet.
None the less there are some severe optical coupling issues. Trying to wedge 256 signals each 400 nm wide in a small area is non-trivial. Could play sequential processing games, I suppose.
(Score: 2, Informative) by joshuajon on Wednesday March 19 2014, @04:22PM
You're getting hung up on the wavelength. You'd need to compare the 200nm to the wavewidth.
(Score: 5, Informative) by green is the enemy on Wednesday March 19 2014, @02:14PM
The main promising feature of optical computing is less energy loss in the transmission of a signal across the chip, because there is no line capacitance to charge. If optical transistors can be made to switch with comparable amount of energy as electronic transistors, they would probably hold an advantage. At ~200 nm feature size the density issue may be solvable with 3D stacking.
Amazingly, the article actually includes data on the energy to switch their optical transistors: "100 femtojoules per bit." In comparison, typical electronic transistor currently requires about 0.01 fJ (excluding line losses) according to one reference. Alternatively we may estimate the switching energy per transistor from a typical computing device, lets say the NVIDIA GK110 GPU: energy per transistor per clock cycle ~= 250 W / (7080e6 transistors * 980 MHz) = 0.036 fJ. The actual transistor switching energy is several times higher since each transistor doesn't switch every single clock cycle. Still, the optical transistors have a long way to go to compete with electronics.
(Score: 2) by bd on Wednesday March 19 2014, @03:17PM
What is far more interesting would be the amount of transmission modulation these optical switches produce.
Optical computing was a very popular research field. The problem was, you basically ran into problems if you wanted to use the output of an optical switch as the input of another one. Basically, all the optical switching materials discovered were lousy "transistors".
With this device being nearly the size of the wavelength of the light they use, I ask myself if it will probably do even less?
I think it is quite telling that nothing in the article mentions that.
(Score: 2) by gringer on Wednesday March 19 2014, @09:51AM
So when is this opaque/transparent metamaterial going to be used in screens?
Ask me about Sequencing DNA in front of Linus Torvalds [youtube.com]
(Score: 4, Funny) by dublet on Wednesday March 19 2014, @10:33AM
see that coming..
"If anyone needs me, I'm in the angry dome. [dublet.org]"
(Score: 1) by RamiK on Wednesday March 19 2014, @11:39AM
Even without all the unresolved technical problems (noise, heat... Yes, photon have their equivalence too), there's still no guarantee the manufacturing process will ever prove profitable.
compiling...
(Score: 2) by MrGuy on Wednesday March 19 2014, @11:46AM
"Prove profitable" feels overrated to me as the standard to apply. Every advance in semi-conductor technology has been incredibly expensive at first (clean room manufacturing by itself is phenomenally expensive to implement). But costs have dropped rapidly as scale has increased. I'm not too worried about a process that's feasible at lab scale being able to be engineered to an acceptable cost level. Industrial engineers are smart folks.
The real the performance advantage is sufficiently worthy to generate demand over and above what exists today. And if we're truly talking several orders of magnitude of speed improvements, I'm guessing it will.
(Score: 1) by RamiK on Wednesday March 19 2014, @04:03PM
"Every advance in semi-conductor" is selection bias since you don't know how many dead technology branches were explored and found unprofitable to mass-produce.
I don't know the future. I just know that demand for high performance computing is declining much like the demand for muscle cars. So just because you have smart engineers doesn't mean the market will be there buying there product.
compiling...
(Score: 0) by Anonymous Coward on Wednesday March 19 2014, @07:25PM
Hasn't Moore done enough?
(Score: 1) by Stuntbutt on Wednesday March 19 2014, @06:10PM
Current microelectronics materials are predominantly silicon (Si). Silicon is cheap to gather, expensive to process. Vanadium Dioxide, I am willing to speculate is not cheap to farm or produce.
While electronics might have hit a "wall" for speed (power/density) concerns, solutions to the problem won't have ubiquitous appeal unless ($price==right).
(Score: 2, Informative) by mrchew1982 on Thursday March 20 2014, @04:40AM
most of the cost in producing silicon is purifying it from its oxidized state (silicon dioxide, commonly found in sand) to a near pure silicone, which has to be done at very high temperatures and extremely controlled atmospheres.
Vanadium dioxide is probably the naturally occurring state for vanadium, and thus the only difficulty would be relative abundance and separating it from the other minerals. Levigation is commonly used to separate sand, i don't know about vanadium.