"we've modelled it..."
Unless they've made something, anything, even one single gate, this story is a load of vapour-rub.
Boffins have made optical transistors that can reach 4 TERAHERTZ
Aluminum-doped zinc oxide is the key to building faster, optical chips, according to researchers at Purdue University, Indiana. They've modelled an all-optical, CMOS-compatible transistor capable of 4THz speeds, potentially more than 1,000 times faster than silicon transistors. The all-optical bit means that the data stream …
-
-
This post has been deleted by its author
-
-
Friday 7th August 2015 15:31 GMT Dave 32
They might be on to something here. I did my MS/EE project paper on optical computing, close to 30 years ago, and most of the limitations he speaks of were known back then. The problem has been that no one has found a way to get around those limitations, at least until now. That idea of controlling a photo-reactive material with UV light, while it switches IR light is interesting, and makes a lot of sense. All they would need to do now is to run the IR light coming through the device through a frequency doubling crystal (e.g., something like KDP) twice to get the IR up to UV frequencies, and then use that to control the switches. And, no, it won't be easy. Based on the progress I've seen over the past 30 years, I'd estimate that his two decade number is quite optimistic. Still, it could happen.
Probably the next thing that needs to happen, though, is for someone to come up with a better/faster/cheaper way of producing the Al doped ZnO. CVD is a bit of a royal pain. Something like the Czochralski process, to produce large, single crystals of ZnO would be interesting, and, based on a quick literature search, it appears that it may be possible, given some rather exotic constraints of crucible/oven conditions.
Of course, massively parallel, ultra high speed cross-bar switches, using the optical switches could also be quite interesting, and may be just the thing that's needed to allow massively parallel conventional processors to be networked. Imagine, if you will, a few thousand (or 10s of thousands, or 100s of thousands) of processors being cross-coupled to many TB of RAM, I/O devices, etc., all of which can be dynamically reconfigured at 4 THz speeds.
Dave
-
-
Saturday 8th August 2015 14:54 GMT Preston Munchensonton
Re: Hey Dave - I think you could get a very high paying job with this knowledge!
Would you mind taking stock options until the company becomes profitable (~20+ years)?
I would love that opportunity. Unfortunately, it's being funded by the people of the US state of Indiana and the US Federal government. I doubt the timeframe would require nearly as long in the private sector.
-
-
Friday 7th August 2015 16:04 GMT Speltier
That is at least 3 and maybe 5 or 6 PhD thesis of work right there. And after that, some poor process engineering team will have to make the optical ivory idea actually yield in a production environment...
However it is interesting as a fast switch for certain niche military and financial tasks where cost is less of an object.
-
Sunday 9th August 2015 10:17 GMT John Smith 19
" All they would need to do now is to run the IR light coming through the device through a frequency doubling crystal (e.g., something like KDP) twice to get the IR up to UV frequencies, and then use that to control the switches. "
IIRC this process has a threshold intensity level and substantial losses. Good thermal management will be important.
"Probably the next thing that needs to happen, though, is for someone to come up with a better/faster/cheaper way of producing the Al doped ZnO. "
Perhaps easier than you realize.
You might be looking at something called "Atomic Layer Epitaxy." Originally developed in the 708's to improve the quality of EL displays it uses a sequence of chemicals to progressively build single atomic layer structures through electrochemistry. Basically electroplating but with the reagents in an inert carrier gas, rather than in water.
In principal you're looking at a layer of Zinc, annodizing it IE making it the +ve eletrode to build up an oxide layer, then the ALE to add the Aluminium.
The trouble is that however you slice it photons of different wavelengths are not like signals at different voltages or currents.
Such a device is loosely like a MOS transistor with the gate voltage being the input wavelength and the switched signal the output current.
Hardware logic design tells us that any logic function can be realized if you can do NOT.
But someone has to actually build this and characterize it before this is anywhere close to starting to sound interesting.
-
-
Friday 7th August 2015 15:57 GMT Tom 7
Road to no-where?
I worked on some mixed bipolar and GaAs stuff 25 years ago. It seemed clear then that the optical stuff was always going to be on the periphery and not the core - the wavelength of the light being used being a major problem. The transistors involved here are pretty huge compared to modern devices. Make the light shorter wavelength and you've got serious problems of photons tunneling through several neighbouring devices - at these scales everything's see-through.
As the man said - might be useful doing some switching/encoding in the data centre but not really.
-
Friday 7th August 2015 18:17 GMT Andus McCoatover
...and in the years to come...
...our grandkids will be cracking open an ancient El Reg. printout, and have a giraffe at this. Teraherz? Oh, so...50's...
Much like when we crack open an ancient copy of "Metronome Monthly"/"101 things to do with a 555 timer"/Practical Electronics.
Whatever it's called nowadays.
Howls of derisive laughter.
-
Saturday 8th August 2015 04:36 GMT John Savard
Just two things needed
They just need to do two things to be well on the road to making a faster microprocessor with this, actually.
First, actually produce a real switch of the type they believe should work this fast.
Second, also produce a switch which works as fast, but which has a carrier of 450nm and a control of 1300nm - the other way around to the one in the story.
Then optical logic circuitry would be entirely possible.
-
Sunday 9th August 2015 03:35 GMT John Tserkezis
"Boffins have made optical transistors that can reach 4 TERAHERTZ"
And my mother would still claim she can do maths in her head faster.
She is so anti-tech, she has a cell phone, but never takes it anywhere. In fact, every month or so when she forgets to recharge it, she gets me to turn it on again. So it can sit on the dresser for another month not being used. I'm looking forward to brain implants, maybe we can port Crysis to run on her without overclocking.
Sigh.
-
Biting the hand that feeds IT
