Fluid Logic Chips

Doc Ruby writes "Colorado researchers 'have constructed microfluidic gates that use the relative flow resistance of liquid to carry out the basic logic operations NOT, AND, OR, XOR, NOR and NAND. The researchers have also combined a pair of gates into a half adder, which carries out half the operation of addition.' All CPUs processing binary logic are made of these types of gates, but usually execute as flows of electrons in wires, not fluids in tubes. Will this advance revolutionize chemistry and computing the way electric gates revolutionized electronics and computing? Will 'fluid programmers' give new meaning to "flowchart"?"

Comments

How fast?

[cbrocious]

How fast could this ever be? Neat, but I dunno how this could ever be put to a practical use. Cool hack none the less.

Re:How fast?

[ikewillis]

I think you mean the speed of electrons. Electrons can't travel the speed of light (in a vacuum)

Re:How fast?

[wass]

Practical uses? Well, for starters, it's microfluidics. So if we're lucky, we'll finally be able to get one of these babies into a package small enough to fit in a watch. You've always wanted a digital watch, right?

It's amusing, but in 1967 this Fluidic Amplifier was billed as "the simplest device known for setting up digital circuit applications."

Re:How fast?

[forkboy]

Electrons dont move through copper at the speed of light. However, they do move faster than sound does through a fluid.

Re:How fast?

[grumpygrodyguy]

Will this advance revolutionize chemistry and computing the way electric gates revolutionized electronics and computing? Will 'fluid programmers' give new meaning to "flowchart"?"

How fast could this ever be? Neat, but I dunno how this could ever be put to a practical use. Cool hack none the less.

In all likelyhood this will never be used as a replacement for silicon. It's much more likely that stuff like this will be used in bioinformatics & pharmacuetical circles in order to perform massively parallel tests on different molecular combinations.

If there are over 1,000,000 molecular permutations of a particular family of drugs(or DNA). Perhaps this kind of computer could rapidly cycle through all such combinations. Maybe the testing reaction could be performed with a liquid-mechanical ALU of sorts. Then the results could be stored in a liquid memory bank where they could be reviewed. Perhaps indicator dyes, or electrical dyes could be used to signal positive/negative results. *shrug*

Re:How fast?

[Detritus]

I mean the speed of electromagnetic radiation, not electrons. Which is in the neighborhood of 0.6c in coaxial cable.

Re:How fast?

[hunterx11]

You've always wanted a digital watch, right?

That and little green pieces of paper.

Re:How fast?

[wass]

Speed of light in vacuum, c, will always be faster than the speed of propogation of any particle, mode, or disturbance**.

The speed of light in a material is slower than in a vacuum, by a factor of the index of refraction (usually frequency dependent). Interestingly, it IS possible for particles to travel faster than this apparent speed of light, and in doing so they emit Cerenkov Radation, which is how many high-energy physics particle detectors (eg SNO) detect individual particles.

** For the nitpickers who will inevitably respond to that generalization, it is occasionally possible (in theory, at least) to set up a mode in some carefully-devised system where the speed of propogation of this mode is faster than c, but this mode cannot carry information. Simple example is a linear array of equally-spaced pendula, each with the same fundamental frequency, and with a spring connecting the weights at the bottom to the two pendula on either side. If a mode is set up where all pendula are oscillating at their fundamental frequency, all of them at exactly the same phase, (springs always remain at their unstretched length) then the phase velocity of this mode is infinite. However, there can be no 'information' or disturbance transmitted down the system. In reality, thermal and quantum disturbances would disrupt this mode and it would eventually become something much more complicated. These disturbances would be transmitted at a finite velocity, less than c.

Re:How fast?

[polecat_redux]

that is still a lot faster then we can move an object bigger then an atom.

True, but you wouldn't necessarily need to move a specific liquid molecule through an entire path/circuit. For example, say you have a tube filled with water, and you were to apply pressure on one end, almost instantly, water would be expelled from the other end. The "distance" that any single molecule of water along the path would need to travel depends only on how much water you want to get out of the other end.

For lack of a better analogy, it would be like poking someone with a stick rather than throwing a rock at them - travel time is mostly eliminated.

Re:How fast?

[Doc Ruby]

It *is* silicon - it just routes tiny, single-file fluid molecules through empty channels, rather than really tiny clouds of electrons through conductive channels.

Re:How fast?

[Jerf]

almost instantly, water would be expelled from the other end.

You can quantify that better. It basically travels at the speed of sound in the medium, because it uses the same forces that sound does.

This is also the solution to the relativity paradox, "What if I take an infinitely rigid rod and tap it on one end, causing the other end to instantly vibrate, with the tap exceeding the speed of light?" The answer is that in this universe, no such infinitely rigid rod is possible; the maximum speed possible is still the speed of light.

This also implies that fluidic computing will always be slower than electronics, because the fundmental speed is orders of magnitudes slower. Which doesn't mean it is useless, I'm just killing two birds with one stone here, showing why this is no threat to electronics :-)

Re:How fast?

[imgod2u]

Yes, however, but the same argument, electrons don't neccessarily need to flow too far inside an electric circuit. Merely the shift in electric field is neccessary to indicate an on (5V let's say) or off(0V) state. Very little current actually needs to flow. Changes in the electric field propogates at the speed of light, so modern CPU's do, indeed, operate at the speed of light.

Re:How fast?

[wass]

You, sir, are a moron. They didn't invent a new mechanism for transmitting a signal. Electrons _have_ to be used.

This seems to be a common misperception on slashdot.

Electrons are certainly used, of course, in digital logic circuits. For example representing bits as charge stored in a capacitor, or by mediating the quantum statistics of a transistor for switching (by controlling the charge on the gate of a MOSFET).

However - when a signal is sent down a wire (eg, from a microprocessor, over the data bus, to a peripheral) that signal is NOT being sent through the electron drift. [Although electrons will drift in presence of an electric field, the drift velocity is INCREDIBLY small, look it up.]

If the microprocessor wants to flip a bit from a 0 to a 1, the wire is originally at one potential, and the microprocessor will change the potential. This disturbance isn't instantaneous along the wire, that would violate relativity. The microprocessor basically creates an electromagnetic disturbance that travels down the wire to the peripheral.

Now let's look at this 'disturbance' more closely. Electrons at point A are being ultimately effected by electrons at point B. This effect is mediated through electron interactions, and one knows that that the electromagnetic force is the mediator between electrons. And from Quantum Field Theory one knows that photons are the quanta of the electromagnetic force.

So what this in effect means is that whenever electrons are interacting, photons are being transmitted somewhere during that exchange. Thus, the parent was correct that it's the electromagnetic wave, as opposed to the physical motion of the electrons themselves. that plays the role in limiting digital logic speed.

Re:How fast?

[MillionthMonkey]

The drift velocity for DC current is around a few mm/sec, but the individual root-mean-square electron velocities in a wire are comparable to c. (Regardless of current.) Although they keep scattering off the copper atoms and off each other and change direction all the time, so they don't travel very far. But the signal propagation velocity through a wire depends much more on the rms velocity than the aggregate current drift velocity.

People are generally surprised that drift velocity of electrical current is so slow. When I was in middle school there were two water fountains in the cafeteria, separated by a hundred feet or so. A common trick was to wait until someone was drinking at one fountain, then turn the handle on the other fountain really fast- on/off/on/off/on/off- and they would instantly get sprayed in the face. Even though the bulk of the water itself drifted slowly through the pipes as you drank it, the individual water molecules were bouncing around in all directions at high speeds and the signal conduction through the water between the fountains was very fast.

Re:How fast?

[spacecowboy420]

Best."Oh, no he di'n".evaar

Re:How fast?

[ddimas]

This also implies that fluidic computing will always be slower than electronics, because the fundmental speed is orders of magnitudes slower. Which doesn't mean it is useless, I'm just killing two birds with one stone here, showing why this is no threat to electronics :-)

This looks to be more useful for a fast and accurate "tricorder" device. Looking at it from a chemical point of view you can "preprogram" a whole lot of tests onto this thing and just fly. I for one (having done more than my share of chemical analysis) would love to see this device in a lab. I would save literally years of set up time.

Re:How fast?

[pVoid]

Perhaps this kind of computer could rapidly cycle through all such combinations.

You're completely missing the point. This machine is just a fluid implementation of a binary machine. In that respect, it has absolutely no logical difference from a silicon based digital machine.

I think you might be drawing some sort of paralel between the fact that Quantum computers can do many things in paralel, and the fact that this doesn't use electricity as a means of implementing classical digitial logic.

Quantum computers aren't just a different implementation of digital logic, in fact, there is nothing even remotely similar to the logic circuits you're used to in Q Computing.

If there's anything I can think of, these machines will be completely immune to EM interference. But aside from that, they will be regular computers that could eventually implement a RISC or CISC arch, and run linux or windows binaries.

Re:How fast?

[wass]

it uses the same forces that sound does.

Actually, if you look at the microscopic physics, they both use the same forces. It's primarily electromagnetic forces, although some quantum degeneracy statistics plays a role too, that prevent your hand from going through a door when you knock on it. However, in fluidics (and sound) phonons are being transmitted through the medium, just like photons are transmitted through the wires in electronic systems. However, the sound waves derive mostly from the usually harmonic potentials keeping molecules spaced apart at their average distances. EM waves derive from charges (ie, electrons) moving and reacting with each other.

This also implies that fluidic computing will always be slower than electronics

Practically yes, but to be pedantic - not necessarily always. Maxmimum signal speed in fluidics would by governed by phonons, and in electronics by photons.

In reality the phonon modes, which are usually pretty dispersive (ie propogation speed depends on frequency), have slower propagation speeds than photons (also usually dispersive but usually not as much) in most matter.

But to say 'always' isn't necessarily true, there's no reason a priori to assume in some random material photonic excitations are necessarily faster than phononic excitations.

Redundant Systems and Fluid Dynamics

[mfh]

Will this advance revolutionize chemistry and computing the way electric gates revolutionized electronics and computing?

Not really, because it's basically a copy of the old way except utilizing fluid dynamics. The way electric gates revolutionized electronics was special because there was nothing like it before. What this will do, is enable better redundant designs for deep space probes. Also, a liquid computer likely doesn't get as hot or it won't be as much of a problem if it does.

Will 'fluid programmers' give new meaning to "flowchart"?"

No, we'll just fill all the systems with coffee and call ourselves The Happy Folk.

Re:Redundant Systems and Fluid Dynamics

[deglr6328]

Why would deep space probes use microfluidic logic processors? They may, on the other hand, be very useful for carrying out microchemical analytical techniques with a limited amount of reagent for things like life detection and geochemistry experiments on future planetary(Mars probably) rovers though.

will...

[mr_burns]

will kevin costner star in a dramatization of the discovery as a bad actor with gills? "WaterLogicWorld".

So in the future

[Camel Pilot]

Will we have computers with a logo that says

"Guinness inside"

What about microscopic steam-based logic gates?

[MillionthMonkey]

That would be totally retro. And it would allow AMD to enter the business.

Old logic, new gates

[k_yarina]

Fluidics has been around for a long time.. http://en.wikipedia.org/wiki/Fluidic_logic

In Eastern Germany

[retostamm]

in the 1970's there was a lot of research on Pneumatic Computing. I read a book about that a while back (can't remember the title).

Essentially it worked the same way, plus they had a little "Transistor" where a big airstream would be disturbed if a small control airstream is on.

Obvious advantages of that technology:
- You only need to be able to cut sheetmetal and weld it together
- Not affected by X-Rays unless you melt it (think MAD/Nukes)
- Probably no cooling problems (not sure about this)

Of course, it'd be also very slow. And big.

chemistry and computing?

[k98sven]

Will this advance revolutionize chemistry and computing the way electric gates revolutionized electronics and computing?

I'm not sure if this is a typo.. but I see no real use for this in computing.. unless you want computers which (at best) work like conventional ones except much, much, much, slower.

However, in chemistry.. it may very well become a big thing. One possible use I can think of is for building automated little microlaboratories, controlling the mixage and flow of different chemicals.

This, in general, is a hot research topic in chemistry.. Already in biotech a lot of things similar to this are being put to practical use (Chip assays is an example).

Basically, it's the revolution of miniaturization which is (finally..) coming to chemistry.

Advanced? This is 50's technologies

[nurb432]

The application of fluidics has been around for ages.. even before tubes and 'electronic logic' we had fluidics.. both analog and digital.

Sure its still cool, but dont call it 'advanced'..

Geesh..

Frozen hack...

[sarcastro73]

Whoever thought that supercooling a processor would completely prevent ANDing two bits?

Teaching tool?

[supz]

Couldn't this be used as a great tool for teaching? You should show people exactly what is happening inside a processor. It's always so difficult to get people to picture something they cannot see, and this would make a great visual example

How would you cool such computing devices?

[deragon]

How would you cool such computing devices? Surround the tubes with coils and have electricity flowing through them? ;)

Cool Running

[overshoot]

Hate to break the news, but unless someone finds a way to use superfluids (lossless flow fluids, like liquid helium) for these, they're gonna take power to run.

At a rough guess from scaling theory, they're gonna take several orders of magnitude more energy/bit than electronic gates.

Re:It's not like this is new logic...

[jcr]

When it comes down to it, every programming language gets reduced to assembly level code in order to actually runs.

Close, but what the hardware executes is machine language, not assembly language.

-jcr

Re:It's not like this is new logic...

[NanoGator]

"This is a new way to do binary logic mechanically, but until they get this to the speed of copper chips they're not going to be useful for much."

Would they survive an EM burst?

Re:Sounds great but...

[k98sven]

The width of these channels is 100 micrometers.

The flows here are created by the capillary forces which dominate at that size.

No gravity required.

There are still some bugs in the water gates...

[hpa]

Okay, this definitely calls for a link to... Crunchly!

Mandatory reading for the larval geek...

Fluid computer in Cold War?

[isdale]

I recall there was a ballistics computer built back during the cold war. The idea was that it was immune to radiation effects (EMPulse). I cant find a reference to it in a quick web search.

Corning fluidics (from about 1972)

[klubar]

Many years ago (about 1972), Corning and others made "fluidics" devices that used air to implement simple nand gates. They were looking for applications, such as explosive environments (fireworks factories, cotton processing) that relays wouldn't work well in. The devices had simple sensors and could implement logic by combining nand gates. There were a couple of competitors that made fluidic devices. The Corning were small black cans about 2" high and 1/2 around; the air supply was connected on the top and there were 4-inputs and one output on the bottom.

Cute, but they went no where. I put together a neat high school science fair project with them and got to the county level.

Nice to see the concept recycled.

Zounds

[omarius]

Certinly puts a new spin on 'memory leak.'

I swim the body fluidic.

[Doc Ruby]

That could depend on the operations. In the electronic paradigm, fast CPUs process data in parallel, integrated across much slower networks, their messages processed by routers on a much higher symbolic level than processed in the CPUs. A possible fluidics architecture might process chemical reactions which code their results in their products, which are flags for the fluidic processor valve. So networks of partial results can be processed by these CPUs. There are many computational chemistry applications which could be complementary to this kind of processor, with fluids merely the medium which they chemistry conveniently produces, and these chips are suited to process. There's nothing uniquely informational about electrons; they're just the tiny tool we had mastered when we started applying the mechanics of info theory. Now we can harness our latent fluidics techniques, crossbred with our electronic techniques, for a hybrid that can use the most tractable properties of both.

Additionally, humans are more chemical than electronic. Even our neurology, often metaphorically "electric", is really an ion pump. All electronics require lots of adapters to couple with our senses, either chemical, optical or mechanical (including sound). These fluidics are in the same domain as our own primary physical existence. So integrating them with our biology might be more direct. Implants, sensors, medicine, all the much more personal tech applications might be more available to microfluidics than they've been to alien electronics. Surf's up!

Fluidic state machine: automatic transmissions

[MasterC]

Using fluid for logic isn't new. The best example I can think of is the automatic transmission. The valve body is a maze of pathways that essentially act as a state machine of the transmissions that chooses the appropriate bands and gears and such.

One link I found (go down to "Valve Body"):

http://www.familycar.com/transmission.htm

The modern processor is an electrical state machine and the valve body is a fluidic state machine.

The only real development is the physical implementation of the logic but considering that currently they can't link gates it's not of that much practical use since you can't form a state machine (or anything more complex than a gate)...at least I'm not aware of a way to make one layer of logic a state machine...

Cool nonetheless.

Uh, minor nit.

[chadjg]

Uhh, according to some random website

The compressibility of water is .0000034 lb/ft2. For a change in pressure of one pound per square inch, water is only compressed by about 3 one-millionths of its original volume.

However, at very large volumes, this small compression can become significant. For example, if water were truly incompressable, the oceans would stand about 100 feet higher than they do today. Compression of water in the ocean basins reduces global sea level by 100 feet.

I know the macro and micro worlds are quite different, but water does compress, and pipes and hoses do stretch, therefore there must be some delay in the propagation of the pressure wave, right?

Speaking from some minor experience with fire hoses and associated equipment, slamming valves on and off with a relatively incompressible fluid raises holy hell with fittings, pumps and hoses. It's called a "water hammer," and the effects can be costly. I'm not quite sure if this would be a problem in a microsopic array.

Im amazed how....

[Creepy Crawler]

Dumb people here are....

First off, electronics can be made EMP-proof, and a few ways at that.

1: Surround COMPLETELY the device with shielding, ala TEMPEST. Ground the shielding.

2: Use EMP detecters around the object to detect incoming pulse. When slight spike is recorded, cut electricity and ground the chips. These "hardened" chips have a cutoff which crucial parts are grounded/ungrounded on accepting a signal.

3: Use electron tubes for the base part of the system. Connect e-tubes to tempest surrounded internal computer. Use self-grounding chips for best survivalibility.

We already know the future.

[jd]

Fluid links are extensively used by the Time Lords as an essential component in the TARDIS flight controls.

The only reason we have Fluid logic chips today is that The Doctor defected from UNIT after the BBC cancelled the series, selling his advanced knowledge on the subject to these researchers.

Re:It's not like this is new logic...

[jfengel]

Yes, they would survive an EM burst.

I used to work in a fluidics laboratory as an intern [hey, I actually know something for once that every Slashdotter doesn't know!] and one of the purposes they were developing this stuff for was because of its ability to survive an EM burst. They were talking about using it in fighter planes for exactly that reason.

This was two decades ago, I'm ashamed to admit [I mean, I can't believe I've gotten so old that I remember two decades ago], and the things this lab built were way, way larger than the stuff being talked about in the article.

Missing the point, perhaps?

[Skinwalker]

As someone who does academic research in microfluidics, I should probably comment on this and some of the perceptions of it.

This stuff will not WILL NOT ever replace electronic circuitry. I don't think anyone who works with microfluidic applications would seriously claim this. There is just too large of a speed differential between fluids and semiconductors. Do you want your computer making decisions on the millisecond time scale (fluidics) or the subnanosecond time scale (silicon)? This work is a little misguided, and somewhat misleading as it tries to mimic electronic circuitry. Fluidic logic was rightly given up as a techy-backwater thirty years ago. There is tremendous potential in this field, however, when people start to think out of the box of usual engineering.

There are some really cool fluid physics that take over at these length scales. You can't have turbulence (the Reynold's number is far too low) so neighboring streams are totally laminar, and stay separate until they mix by slow diffusion. Buoyant forces dominate over convective forces (the Grashoff number is low), so you can do biology and chemistry experiments in these systems that were previously only practical in microgravity. For a tiny fraction of the cost, mind you... most microfluidicists use a channel-making process that employs photolithography, so you can use the economy of scale to do a f^Hckton of experiments for pennies on the dollar. Better than hoping your precious bugs survive the next shuttle flight.

This stuff is already having a serious impact in biotech and big pharma. The Human Genome Project wouldn't have been possible without technology that used these physics to shunt little packets of fluid around. Synthetic chemists use it to make thousands of variations on whatever drug they're working on.

Do some googling if you're interested... the field is booming right now.

Oh, and these guys are almost certainly using computers to drive their input pumps. Cheating, sorta...

NCR at the1963 World's Fair

[Baldrson]

From the Boundary Institute's CV list:

Thomas Etter's background is in mathematics and philosophy. He has worked in various ways with computers, holding several early patents on integrated circuits, one of which was demonstrated by National Cash Register Inc. at the 1963 World's Fair.

Etter's integrated circuit demonstration was of a fluidic device.