Researchers Re-Examine Second Law of Thermodynamics 125
Many readers have written to tell us that researchers are examining the possibility of using Brownian ratchets to help combat the problem of heat dissipation in miniaturized electronics. "Currently, devices are engineered to operate near thermal equilibrium, in accordance with the Second Law of Thermodynamics which states that heat tends to transfer from a hotter unit to a cooler one. However, using the concept of Brownian ratchets, which are systems that convert non-equilibrium energy to do useful work, the researchers hope to allow computers to operate at low power levels, and harness power dissipated by other functions. 'The main quest we have is to see if by departing from near-equilibrium operation, we can perform computation more efficiently,' Ghosh told iTnews. 'We aren't breaking the Second Law — that's not what we are claiming,' he said. 'We are simply re-examining its implications, as much of the established understanding of power dissipation is based on near-equilibrium operation.'"
Re:Hmmmm, help me out here. (Score:4, Interesting)
maxwell's demon (Score:4, Interesting)
maxwell's demon [wikipedia.org]
old, well-tread, philosophically and scientifically fruitless territory here
Very weak on details (Score:4, Interesting)
What a crappy article. Subtracting the techno-babble, it sounds like they want to attach a thermocouple [wikipedia.org] or heat engine [wikipedia.org] to their chips, which has already been tried many times and found to be not worth the effort. Maybe they think they have a better method, but I sure couldn't tell from RTFA.
But Brownian Ratchets don't work? (Score:1, Interesting)
I'm not an expert and wikipedia isn't a great source but in the article on Brownian Ratchets it mentions that any machine small enough to move based on the Brownian motion of nearby matter would be subject to Brownian motion itself. So are they saying they have a way of making brownian ratchets work or are they just assuming they can use something that most people believe doesn't work?
Re:Hmmmm, help me out here. (Score:4, Interesting)
Well the idea of using the heat energy to do something is all well and good, but they would need something that actually needs to be done...Otherwise it would seem to be more efficient to simply strive for greater efficiency, and try to reduce the amount of waste heat.
Obligatory Wikipedia reference (Score:5, Interesting)
You are correct.
As described by Feynman, a Brownian Ratchetis a theoretical machine that can extract energy form a system in equilibrium. It is a kind of Maxwell's demon [wikipedia.org].
Feynman explains why such a machine will not work without a potential energy gradient and is in fact a perpetual motion machine.
TFA seems to indicate that they intend to operate from a system not in equilibrium, which is allowed by the Thermodynamics Police. But it isn't very clear from the summary.
Reading between the lines (Score:4, Interesting)
Subtracting the techno-babble, it sounds like they want to attach a thermocouple [wikipedia.org] or heat engine [wikipedia.org] to their chips...
Almost. Reading between the lines, it appears that they want to attach thermocouples or heat engines *IN* their chips rather then to them. They appear to be talking about the heat in the individual transistors within chips, rather than the entire chip. From the article, it sounded like they were trying to reduce the heat from each individual transistor and use that heat in different ways.
Can it be done? I have no clue. Can 50,000 nano sized thermocouples be more more efficient than 1 small one? Again, no clue.
Re:Hmmmm, help me out here. (Score:5, Interesting)
Just read the guy's home page for explanation (Score:2, Interesting)
http://people.virginia.edu/~ag7rq/ [virginia.edu]
follow the link to "Second Law? You must be kidding..."
"FYI -- there is some sensational press out there that makes it sound like we're planning to break/have already broken the 2nd law of thermodynamics. This is, of course, absurd -- but I think it's imperative we set the record straight before everyone starts jumping all over us.
The context.... a colleague and I received funding to study non-equilibrium switching invoking a concept called 'Brownian Ratchets' that has been well studied in nonequilibrium statistical physics over the years. The potential benefactor of this study is the chip industry, in a very broad way, as it is worried about rapidly increasing thermal budgets (chips are becoming very hot). We're simply trying to examine the physics of Brownian ratchets in a device context. A popular model for heat dissipation in binary switching (proposed by Victor Zhirnov and co-workers) looks at a two well one barrier geometry, with a gate controlling the barrier and a drain controlling the overall directionality. Each such raising and lowering of a barrier at the end dissipates energy irreversibly (during the reset step where one erases information), leading to a kTln2 dissipation per operation (kT is the thermal energy). And this analysis is usually done by assuming that you wait after you raise or lower a barrier and then let the electrons move and reach equilibrium with the surroundings. The analysis is thus based on equilibrium Boltzmann statistics -- since the electron was at equilibrium before a computation and reaches equilibrium after. What is not clear is what happens during the non-equilibrium transition phase, or if you switch before the equilibrium is reached. The aim of the study is not to attempt to deviate from cherished physical principles, but on the contrary to see what these cherished principles posit for such a situation. A ratchet is known to be able to rectify non-equilibrium noise to produce directed motion by transducing spatial asymmetries in the system (this is well recognized in nonequilibrium statistical mechanics and has been mulled over for years). The physics is well studied, but the context is perhaps new... we are interested in seeing if rectifying such non-equilibrium noise (as a ratchet does) can perhaps shave off some of the power dissipation limit associated with a drain bias in the regular example.
This is, of course, still at a toy model -- we need to worry about how to deal with compatibility of input and output, for example. Simply put, we don't know if this will bear fruit for the big picture of low-power device operation, but it's worth investigating.
That's about it... but then, cooling laptops as hot as the sun through the power of thinking or by breaking the 2nd law sounds fancier ... doesn't it? "
He's talking about studying the transient state of electrons switching in a semiconductor barrier, and how it may be useful in reduce semiconductor heating.
Re:Hmmmm, help me out here. (Score:3, Interesting)
"Brownian Ratchet" incessantly, and I know what those are: a theoretical molecular machine
I thought they were widely observed in microbial locomotion systems?
It's not perpetual motion machine. It's a fan. (Score:4, Interesting)
One of my friends got her degree in Linke's lab: http://www.uoregon.edu/~linke/res_ratchet.html [uoregon.edu] .
If the front page at Linke's lab is related to whatever inspired the article: I bet they're trying to make a microscopic fan (with an external power source) as a linear motor, not a perpetual motion machine. They're not trying to scavenge the power from the heat. They're trying to move the hot molecules around.
Such a fan could be in the form of a structure of electrodes on the top of the chip which moves the coolant by creating intermittent sloped potential wells, using the brownian motion from the heat to accomplish part of the motion of the surrounding coolant.
You'd still be providing the energy to move the molecules when you create and then dissipate the potential wells. You make a "traench with a sloped bottom", the molecules fall into it and slide to one end, you raise the bottom of the hole, lifting them, and they scatter, with some of them ending up over the NEXT trench location next time. No free lunch - you provided the energy to move them by lifting them out of the potential well when you demolished it.
I suspect that they are using brownian ratchets for the motors, rather than trying to move the molecules directly, because they found a way to implement the former efficiently.
But I'd like to see how it works and what makes it better than creating a similar array of stepwise-moving potential wells ala charge-coupled devices. More efficient? Fewer drivers? Sloped potential wells easy to make using triangular or other interesting electrode shapes? Larger structures that can be fabricated at current semiconductor feature sizes?
Re:Hmmmm, help me out here. (Score:4, Interesting)
Well, it depends a little on exactly what they're doing. There was an argument that you couldn't use air cooling to go below ambient temperature. Let us say you've N chips, with some method of transferring the heat from all of them to a common point. The common point can now be air cooled to ambient temperature, which means that the N chips must be cooled below that. You generate heat by doing so (2nd Law) but so long as that generated heat is outside the airstream you're using, it won't affect the ambient temperature, except in a closed system, where this must necessarily break down almost immediately.
It would follow that if you could transfer heat from surrounding areas to a more concentrated region, you can get enough heat to do interesting things with. But it has to be concentrated, or you won't have enough to be able to do anything. You won't be able to do anything useful with a thermocouple, as you don't have any inherent cold regions and making one will cost more energy than the thermocouple could provide. So what else can you do with heat? Heat causes expansion - a really bad idea for any material using variable materials in layers to produce tracks - but there are possibilities for nanoscale mechanical systems. Not many, though, and nothing I can think of that would be useful.
Let us say you have N compute devices, but for some reason (due to prior threading, perhaps) the ones in use are highly concentrated together. The heat could be used to trigger a re-distribution of workload. Seems unlikely to be fast enough, but it's one possibility.
Option 2 would seem to be based on electron tunneling. This phenomena is deliberately used to create jumps between lines that you can't build physically on a 2D circuit except by using lots of very slow logic. Electron tunneling is partially a function of the medium. If you could therefore alter the medium sufficiently, you basically have a very slow but serviceable switch. This is only useful if there's anything so long-term that an extremely high latency switching mechanism would be useful.
Option 3 is where data is retained in the absence of power (for some time - doesn't matter how long) but you need it to act like volatile memory. Maybe you could use heat to zero the state of such memory. Again, it's very slow, so you'd need something that needed so much zeroing that doing the same operation electronically would be slower. This is possible because although heat has a very high latency, it diffuses well and therefore provides a massively parallel method.
Option 4 is to find the researchers and tie them by their feet to the top of the mast of a Tall Ship and leave them there until they do something worthwhile. I favour option 4.
Re:Reading between the lines (Score:3, Interesting)
I'd like to add that, just because it wasn't worth the effort, doesn't mean that it isn't worth the effort. For various reasons, energy usage grows much faster than performance in current transistor designs. There are some experimental designs which reduce this significantly, but they're still in the early lab / press release stage, and nowhere near being ready for production. Just shrinking the process isn't giving the kinds of benefits it used to, because it increases leakage, which increases waste heat, which means that shrinking the process requires a lot of clever tricks to do well.
If you can extract useful work from the temperature gradient caused by an operating chip, then you can dramatically reduce the amount of energy required for a given performance level, which is very important at the moment. Whether they actually can, remains to be seen.