Cooling breakthrough for computers, car electronics

Researchers at Purdue University have had a breakthrough that may completely change how engineers design cooling systems in everything from computers to electric and hybrid cars.

Using special computer chips from Delphi Electronics, Suresh Garimella, the R. Eugene and Susie E. Goodson Professor of Mechanical Engineering at Purdue University, and doctoral student Tannaz Harirchia, have developed and tested new mathematical formulas concerning the properties of boiling liquids in "microchannels."

Purdue University doctoral student Tannaz Harirchia holding up the computer chip she and professor Suresh Garimella (right) used to develop new formulas concerning 'microchannels.'

(Credit: Purdue University)

It's no secret that engineers, particularly chipmakers and computer manufacturers, have been striving for years to design cooling systems with highly efficient heat-transfer rates.

Microchannels are tiny channels through which fluid is directed in some types of high-power electronic cooling systems. Purdue University researchers have been working on the idea of microchannel heat sinks, as well as liquid-filled chips, for some time. IBM has also been developing a liquid-based chip-cooling concept.

Garimella and Harirchia have now determined that "allowing a liquid to boil in cooling systems dramatically increases how much heat can be removed, compared to simply heating a liquid to below its boiling point," according to their report.

"Boiling occurs differently in tiny channels than it does in ordinary size tubing used in conventional cooling systems," Garimella said in a statement.

Details of the breakthrough will be presented on October 8 in Belgium at Therminic 2009, an annual conference on thermal research and technology for microelectronics.

Having come up with a new way to assess fluid boiling in microchannels, Garimella and Harirchia now plan to concentrate on developing heat-transfer models engineers can use when designing cooling systems for high-power electronics.

Conducted in conjunction with Delphi Electronics, the research was funded by Indiana's 21st Century Research and Technology Fund, and Purdue-based National Science Foundation Cooling Technologies Research Center, a consortium of corporate, academic, and government laboratories.

[jaguar717]

Sounds like the heat pipes that are already in aftermarket CPU and GPU cooling systems. Are they just scaling them way down in size so that surface area per volume skyrockets?

[texaslabrat]

They are scaling them down to the point that they can be included inside the IC packaging rather than as a "bolt on" piece that is in conductive contact with the outside of the packaging. Since you eliminate the R-value of the packaging itself, the heat can be drawn away from the actual IC at a faster rate with less parasitic energy loss in the cooling system, all else being equal. Or so goes the theory.

[arkile]

I need this for my xbox 360 so it doesn't get red lights lol

[tech_crazy]

Unless I am missing something fundamental, this just seems like old wine in a spanking new bottle. It is common physics knowledge about something called 'latent heat of vaporization". In simple terms, 1 gm of water requires (or uses) 1 calorie of energy to raise its temperature by 1degree C. This is its "specific heat". However, when water boils into steam, it requires 540 calories to boil (convert the water into steam). Thus boiling water can take away 540 cal of heat per gram instead of 1 cal/gm of water at less than boiling temperature. Wouldn't it thus be obvious to let water (or any other liquid) boil in the channel/tube/sink ? Heck, even ice has a latent heat of 80 cal/gm. What exactly was pathbreaking in the Purdue system?

[zyxxy]

The fluid is boiling in microscopic channels that are etched into the chip package itself. IBM has some great pictures of the actual channels in their design, but I cannot find the link right now.

Fluid vaporization in this confined space behaves differently than bulk fluid vaporization in a container. They have characterized that behavior.

Space constrained processing behaves differently than bulk. Consider the catalytic converter on your car. The exhaust to channeled through a constrained space that is coated with a catalyst. If you just lined the exhaust pipes with catalyst, it would not work the same. This is analogous to that, but dealing with heat transfer.

[Seaspray0]

@zyxxy. But that still doesn't explain how you are going to condense the boiled liquid back into liquid state. That requires releasing the heat somewhere and that will still take a heatsink. The heat transfer will be improved, though.

[texaslabrat]

@Seaspray:

There is a collection architecture on the other side of the chip which consolidates the microchannels into "macro" channels (one analogy is header pipes attached to the exhaust manifold, and then coming together to form a single, larger, exhaust pipe) and then on to a condensor/heatsink to complete the cycle. That part doesn't change substantially in concept from typical heatpipe architecture that you may be familiar with in consumer PC applications.

[jragosta]

"Garimella and Harirchia have now determined that "allowing a liquid to boil in cooling systems dramatically increases how much heat can be removed, compared to simply heating a liquid to below its boiling point," according to their report."

Sorry, pals, but that's Freshman Chemistry. Look at water, it takes 1 calorie to heat 1 g of water by 1 degree C. It takes 539 calories to boil 1 gram of water which is already at 100 degrees C. There is absolutely nothing new about that, nor did they discover it.

There may be something new about the microchannels, but the statement made above is complete BS.

[zyxxy]

I am certain that the researchers know what they are doing, and the reporter dumbed it down. What is new, is the process behavior in the microchannels. Understanding that part of the process lets you refine the channel design to exploit that behavior as much as possible.

Also, they are not using water, although your analogy still holds as far as latent heat is concerned. They are using something with a lower boiling temperature. Likely a fluorocarbon if some type.

[texaslabrat]

While it seems as though we'll have to wait until Oct 8 to get the full details...it seems as though the breakthrough is mathematically modeling the boiling/condendsation process for a heat pipe cooler in the very small scale. For those who posters who don't understand the breakthrough due to lack of advanced science education...let me just say that as you scale things down physical processes we take for granted don't always behave the same. Things like surface tension, Reynolds numbers, and the like make fluids flow very differently in small spaces than in a big beaker (or the relatively large spaces of common heat pipe coolers we use these days). To jragosta: did your freshman chemistry class ever demonstrate capillary action? Ok, now boil a liquid in that small tube and report back the exact mechanics of what happens in closed-form mathematical fashion that can be used to design a cooling system. Your gross oversimplification of the issue just shows your lack of understanding of the subject matter.

As for the supposed redundancy in "inventing" heat-pipe technology/phase change cooling...you have to remember that they are working at the chip-size-scale (actually flowing liquid THROUGH the chip packaging), and most efforts thus far have focused on simply just a working liquid that's moving around (like your car's radiator system) to remove heat since there was no working model on how to shrink a heat-pipe (with its associated phase change mechanisms) system to that scale. THAT's the breakthrough..and THAT's what's so awesome about the announcement.

I look forward to reading the details once they are presented...stuff like this takes me back to my engineering lab days ;)

[tech_crazy]

Dude, get off your high horse. I am into VLSI for close to 2 decades and have been following up on microchannels etc. . And that includes in-chip channels not just below-chip ones.

It seems you lack basic science education. Capillary action is in an open capillary. Do you even know that the chip microchannels are a closed system? And go easy on the intimidation with all the physics terms, we have known those for a couple of decades. Report back when (and if) you get actual details of the Purdue thingy. Until then, just chill.

[texaslabrat]

Dude, you're the one with the oblivious comments. I'm just trying to shed some light in terms that folks might understand (ie the "strange" behavior of liquid that occurs when you scale things down). Since some folks (like yourself) can't seem to grasp the gist of the breakthrough here, I was trying to introduce the scale. I am well aware of the closed system, thanks...as for the "physics terms"...I'm sorry. Did you want me to use less syllables? You can always google search if you don't understand them. If "we've known about them for a couple of decades", why is it you posted such an unenlightened comment above? Seems to me a great number of people on this board don't understand the physics of fluid flow, and need to be educated. So, perhaps you should take some of your own advice and "chill" and read up on the background of this stuff rather than chastising me for trying to erase a bit of ignorance.

[jaguar717]

He's taking issue with stuff like "due to lack of advanced science education", which reads like it was written by a career student with no real world experience appealing to authority.

Never confuse schooling with education. I hold an engineering degree and econ major from a top 10 university, and wouldn't presume for a second that the equations and terminology I picked up are an "education". All the outside reading has supplied the real knowledge.

Complicated language doesn't elevate you to some superior level, it's just posturing. As I'm sure Feynman would agree, all the true scientific knowledge is simple.

[texaslabrat]

Well, I took issue with the oversimplification and outright ridiculing of the researchers in question by people who, quite frankly, don't have a friggin' clue with which to even begin to engage in a conversation on the subject. If someone is going to be offended that I *indirectly* called them out for making an idiotic statement...so be it. And if my statement reads like a "career student"...well, lets just say some of the statements I was responding to read like someone who has had *no schooling*. Clear-cut cases of "a little knowledge is dangerous" where people who learned about heat of vaporization in high school think they know enough to ridicule researchers such as the ones detailed in this article. I refuse to apologize for coming down like a ton of bricks on such ignorance (though I thought I was quite diplomatic compared to my first draft of that post).

And while I'm sure you mean well, Jaguar...I really don't need your advice on the subject of education vs schooling. I have both covered in spades, thanks. Especially in the case pertaining to this particular article where I have deep theoretical and practical experience from university classrooms and research projects for NASA and DoD. And I'm not sure what you meant about "complicated language"...LOL unless you don't think anybody has ever heard of surface tension (you may be right, given some of the responses here) or Reynolds numbers (behold the power of Google!). Last I checked, this was a technology-based website were apropos technical terms are not verboten. If people can't be bothered to look up one or two terms that they might not be familiar with, I have no sympathy for them whatsoever.

In short, if people don't want to be called out for being morons, they should keep their moronic comments to themselves. Being an apologist for them doesn't help them, Jaguar...it only encourages them to make fools of themselves again later ;)

[texaslabrat]

On a lighter note...Ph.D students never looked like that when I was going to school..guess I should have picked Purdue LOL.

[mechengineer1400]

Phase change heat absorbtion is and has been the principle behind systems as diverse as mechanical refrigeration, heat pipe energy recovery and the ammonia heat exchangers which keep the Alaska pipeline pylons frozen, and is practically as old as mechanical engineering itself. For someone to just discver this they must have been living under a rock. Also, the microchannel concept being studied has existed for some time - check out U.S. patent 7,115,987

[texaslabrat]

No one is claiming the concept of the coolers is new...what is new is the fact that the processes can now be mathematically modeled in such a way that predictive designs can be engineered and optimized. Think of it as the difference of using CFD versus just making a bunch of iterative models for a wind tunnel in order to design an airplane. Yes, you can do the latter..but it's WAY more expensive and takes WAY longer to get right than if you can model things with math/computers and then make a proof-of-concept physical model to validate the virtual one.

[zyxxy]

@texaslabrat: Nicely put! It is all about characterization and modeling. Engineering is more efficient in the presence of knowledge, so understanding channel behavior over a range of sizes, pressures, temperatures, materials, etc., enables more efficient thermal design based on knowledge rather than trial and error.

[alegr]

...and all new is old...
Cooling of the vacuum tubes with boiling water was introduced in the beginning of the 20th century, because it was more effective than water cooling. For steady boiling, the surface was textured, which prevented forming of stable steam film.