▲ 130 points
back
57 comments
>Next, a short but powerful laser pulse is fired into the water jet. This bumps electrons out of the dissolved salts, essentially boosting the conductivity of the water. A second laser can read back what state the water is in, providing the “on” and “off” options of an existing transistor.
Sounds like an optical computer much more than a water computer (compared to e.g. microfluidics), and it's much less surprising that an optical switch would switch fast.
So I've been out of Chemistry for the best part of a decade but I think the novel part of this is exploiting the fact that water based solvents are insanely efficient at propagating charge.
Here is some reading I think nay be relevant[0][1][2] but tldr thanks to the miracle of hydrogen bonding, protons/electrons can "transfer" through water by an extremely fast chain reaction of proton association/dissociation - thereby circumventing diffusion as a transfer mechanism.
Laser is a useful interface because lasers can be controlled digitally and provide a convenient way of ionising the solution, but this seems like quite a cool hack exploiting the properties of water to me.
[0] https://en.m.wikipedia.org/wiki/Grotthuss_mechanism
[1] https://www.sciencedirect.com/topics/engineering/grotthuss-m...
So the water is the copper and the laser is the electrons?
This is really going to mess up the hydraulic analogy.
It's Water carrying Water, all the way down
It’s like someone turned https://xkcd.com/2571/ into reality.
Water is the conductor, yes. Electrons are still the electrons -- excited by or measured by laser
The question is: how are the laser an photodiodes driven? Also water? I bet no. Just transistors... so, why not use the transistors switching the laser directly, without the whole laser/water/photodiode path?
I do not think this is going anywhere.
I agree it's still time to be skeptical, but it is interesting.
Now if they can make a terrahertz adder out of this... then I'll consider getting excited
Is there a reason why you couldn't potentially use this same effect to switch the lasers?
> I do not think this is going anywhere.
It is... someone published a paper on this, and will get to keep his/her tenure in this horrible publish-or-perish world of academia.
Foundational research into principles that leads nowhere is not the problem. The pool of cool ideas like this is unknowably large, and it would be sad to not see papers like this from time time. [Edit: as long as it is a high-quality study of course.]
The true problem of the publish-or-perish world is the moral hazard that it causes. In booming fields, it is very seductive to abuse statistics or to outright fake data. These are not always caught by peer review!
> Because the laser pulse is so fast, the water can switch states in a matter of picoseconds, which are trillionths of a second. This translates to potential computer speeds in the terahertz (THz) range – that’s 1,000 GHz, which is far faster than any existing semiconductor can switch.
Picoseconds means under 1000GHz. 1ps would be 1000GHz.
Transistors can switch in picoseconds. The fastest basic bulk silicon ones hundreds of GHz as far as I know. More exotic ones I'm sure could be made a lot faster.
So exactly how much faster is this water based switching device than a transistor? Hopefully the author was not under the impression that transistor switching speed was on the order of CPU clock frequency.
According to the article, they have demonstrated the generation of a signal with the frequency of 3 THz, and the switching time is under 200 fs (exponential edge with a time constant of 70 fs).
Yes... and how is the laser driven? Of course, fast transistors...
There are transistors circuits that go into THz.
That is completely false.
The shortest laser pulses are much shorter than any pulses that can be generated with electronic devices, and they are generated by various optical means.
There are no transistor circuits that go into THz, but only into hundreds of GHz, and at those high frequencies they are not pulse circuits, but amplifiers or oscillators.
A review of potential future transistors that will go into low THz:
R u sure?
https://ieeexplore.ieee.org/document/8240463
For that matter, the laser controlled by other means will not be switched at total discretion, but at a fixed rate. Not different than an oscillator.
Iodine in water ionized on the picosecond scale, by laser pulse.
Would be mildly surprised if that didn't happen that fast. I guess the interesting twist is that the medium is flowing liquid.
How long does this state last? If one were to, say, write a series of bits into the upstream side of the flow, how many such bits could you recover downstream? The potential I see here is possibly a delay line for THz band signals, not so much the basis for logic for a 1000GHz CPU. But delay lines are darned useful things.
It’s been a while since I did circuits. What are delay lines used for?
Early computers used delay lines as memory
In RF, delay can be used to attenuate signals. So filtering of signals.
Storage, synchronization and signal recovery off the top of my head.
Also sometimes you may want to apply skew/deskew to signals, finely tune the arrival time of two or more signals, etc.
So it's not water that makes it fast, it's laser.
This is equivalent to saying, "So it's not the semiconductor that makes it fast, it's the electrical current."
The medium that is changing conductivity is water in this case, and it is able to do it much much faster.
That said, the system involves a laser that adds to the ability to switch quickly. "Because the laser pulse is so fast"
Wet, fast, excited laser.
Lasers aren't wet, water is!
Is the transistor driving the laser very fast...
https://youtu.be/ZfKGC3LFyFA?t=353
Predicted in Rollerball
How can you make logic gates with this?
Agreed: how is this a "circuit"? How does the output drive the input of another one of these?
It looks like this is just "shine a laser at a material to change it, and then use another laser to detect the difference."
You can describe any transistor with the same words as "shine a laser at a material to change it, and then use another laser to detect the difference":
"apply a voltage at a device to change it, and then use another voltage to detect the difference (i.e. to detect if a current passes)".
This is an optical switch, like a transistor is an electrical switch. An optical switch that is not very efficient, because the transmission is varied only between 100% and 80%, but one that is very fast.
It can be used in THz range mixers and oscillators.
https://en.wikipedia.org/wiki/Optical_computing
> To replace electronic components with optical ones, an equivalent optical transistor is required. This is achieved using materials with a non-linear refractive index.
If we do end up creating a processor using this, I wonder if it will need liquid cooling.
Presumably they can use cold water in the jets which itself chills the space. Nice fringe benefit If It Scales.
You need to cool the laser presumably, I wouldn’t use water for that, however.
Like with any switches, by connecting them in series and in parallel.
Nevertheless, it is unlikely to use such optical switches for logic gates. The most likely application for them is in mixers and oscillators.
But that's the problem: how does the "output" of one of these switches drive the "input" of the next one?
That doesn't involve a laser like in the article, and is probably much slower. I don't think you can get Terahertz speeds by just using Newtonian physics on water.
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2860314/ Microfluidic Photonic Integrated Circuits?
>> Because the laser pulse is so fast, the water can switch states in a matter of picoseconds, which are trillionths of a second. This translates to potential computer speeds in the terahertz (THz) range – that’s 1,000 GHz,
This in not comparable to transistors in a computer. Transistors use charge (electrons) to switch current (electrons) in order to build boolean gates and logic circuits. Those lasers are probably not optically controlled, and even if they were and could be made into optical logic circuits, their speed will be limited by their size and the speed of light.
Lasers are electric and the water is used to carry a current, so you presumably power the lasers using the current in the water to make gates.
This makes me think of the water pipe puzzles from bioshock, but I guess it isnt supposed to.
Which is more of a limit of clock speed, signal propagation or transistor switch time? I know heat dissipation is a large factor.
Yes, I'd say it's switching losses driving power dissipation, mostly, but also the amplification factor (often written as 60mV/decade). Bipolar transistors are still better at that.
I haven't read the article yet, but I can imagine a chip with watercooling microfluidic channels, possibly also used to transmit informations (kind of like hormones in the bloodstream). If it is useful as a computing medium why not?
Would it make a good waveguide too?
Compactness and timing. I don’t know why picosecond switching is a big deal, here? A 3Ghz part with a transistor depth of 40 would require a top end switching rate of 8 picoseconds — and that’s really going to make PD unhappy, as it’d leave no timing margin. At best, this tech is 2–4x faster than the device in your pocket; it’s probably slower, knowing the timing rates of some of the analog-on-digital parts being shipped, right now.
Does it evaporate?