Do TECs work in vacuum?

[azhari]

We're trying to use TECs to cool things down in a vacuum, but are having problems. In air they work fine and cool the parts down properly. But as soon as we put the parts in vacuum the temperature starts going up until both sides are at the same temp. We are cooling the hot side with water cooling (vacuum sealed of course), so the warm side is at room temperature. The parts that are being cooled don't generate any heat, we just need to have them cold (silicon detectors have lower noise when cold, same reason your cpu oc's better when cold).

We haven't put any ASII on there since under vacuum it would just boil off. We are planning to use AS epoxy, but even without any goop, there should be some thermal transfer.

Anyone have any ideas why these TECs don't work in vacuum? These are standard 80W TECs with a pulsed regulated power supply which allows us to control the temperature (I would love to have one of these at home, but at over $800 each, I'll pass).

[Aesik]

What are you using to pull your vacuum?

[Diggrr]

I can't see why tec's wouldn't work in a vacuum, do you have them bolted tightly into a sandwich assembly? The reason I'm asking, is that any air gaps between the tec and the object to be cooled would become an insulating vacuum along with the rest of the chamber. You could try lapping the tec's to make sure they're flat...They are ceramic and hard as all getout, but persistence would pay. Be carefull not lap the insulation off of the wires.
Also make sure that the bolting plates in the sandwich are sufficiently thick enough. Bowing could be the cause of the air gaps on the tec's surface. 3/8 or thicker plates would be recommended.

Hope this helps you out.

*edit* you could also try using Arctic Silver epoxy. Place it on the ceramic, and assemble the sandwich. Tighten after the epoxy has set.

[Voodoo Rufus]

That's odd. They should work. Very wierd. They aren't dependent on air for operation.

:beer:

[azhari]

Thanks for the suggestions guys and I'll answer your questions in the order received:

Aesik: We use Turbo pumps to get down to about 10-100 nano torrs of pressure. In case you're not familiar with the units, that is one hell of a low pressure :)

Diggrr: I thought of the TEC's surface not being perfectly flat. That would mean that under vacuum there isn't any air to act as a thermal transfer medium. But we do tighten things pretty good. However, your second point could be the issue. The plates are not very thick and it is quite possible that they are bowing under pressure. In fact I'll check them out for that in the morning (gotta run some calculations to see what thickness they need to be).

Voodoo: That's exactly what I thought. The only thing that vacuum could affect is if we pump down while the TEC is being powered at about 100 militorr the TEC could spark. But once we get way down in pressure there should be no sparking. My colleague admitted that he pumped on it while biased, so we checked them out in air and they worked fine.

Thanks for all the suggestions guys.

[docJ]

What kind of thermal couples are you using and how are they attached to the surfaces you are trying to measure ?

[azhari]

They are 40 mm square, 80W TECs. Pretty standard. They are sandwiched between a copper water cooler and a BIG aluminum piece. So there is no sagging. Right now we're not using any goop between the pieces, so there might be a lot of air if the TECs surface is not flat (bowed in). Gonna have to play some more with it to figure out what's going on. I was just wondering if TECs have problems running in vacuum.

[docJ]

Let me restate " thermal couples " to temp probes. Are you sure about good physical contact with the probes ?

[Patchmaster]

I'm far from an expert with this stuff, but I recall reading that TECs need to be under a LOT of pressure to work properly even in atmosphere. If the the TEC isn't practically flowing into the mating surfaces you're going to trap a lot of air -- or more to the point in your case: nothing -- in the microscopic nooks and crannies on the surfaces. As I'm sure many of us know, "nothing" is the ultimate insulator. So be sure you have the TECs torqued to manufacturer specs.

I'd also suggest using some AS2 and running it in atmosphere for 100 hours or so to allow the AS2 to go through its phase changes. Even if the silicone boils off when you put it in vaccuum, you should still have the silver particles left to fill in all those little gaps and provide some level of heat conductance.

[azhari]

Just an update. Today I finally decided to go down and play with the setup. So far it had been some students working with it. I tightened the bolts holding the TECs and made sure they worked in air (the temp probes are actually bolted on also). Then I turned them off and pumped down. Lo and behold, they started cooling just fine in vacuum. Without any thermal paste. Next step is to put the AS Epoxy in there and see how things work. I think the TECs weren't making good contact (as Patchmaster suggested). So the problem seems to be resolved.

Thanks everyone for your help.

[GigaFrog]

The reason why they were not making good contact in vacuum is that any air buble, even microscopic, will expand in vacuum. If you go real down in vacuum, then the air bulb that got trapped at atmospheric pressure will expand under vacuum and lift the plates from the tec, degrading the thermal transfert.
I am not sure any thermal transfert compound will help, as it will expand in the same way.
The best would be to mount the setup under vacuum, but that may not be easy.

[donny_paycheck]

You could try coating either side with vacuum pump oil or diffusion pump oil.....something with a low vapor pressure that wouldn't hit it's boiling point at high vacuum. Vacuum grease might work too. Sounds like you've got it figured out, but this is a suggestion.

[azhari]

What we typically do with epoxy for vacuum use is that after mixing the two components we pump on it slowly so that all the air bubbles that are trapped in the mixture boil out. Then when we reach a point that the boiling is down to a minimum, this is where the epoxy itself starts to evaporate, then we take out and apply it to the surfaces very gently. I've done this with other types of epoxy so it shouldn't be too different in this case. This way there will be minimal amount of air trapped in the epoxy. This way it comes out even better at conducting heat since there will be no microscopic air bubbles left in there.

The real problem was to cool things down without any epoxy at first to make sure things worked under vacuum. Even without epoxy there should be some cooling. But the students were seeing no cooling whatsoever.

Of course this has given me some ideas about cooling with liquid nitrogen :D If I put the comp in vacuum then I can cool the cpu with a copper pipe carrying the LN. No worries about condensation this way. I've done this for a hydrogen gas target that we use in our experiments. The hydrogen is in one chamber that gets surrounded by LN that an autofill device fills from the outside of the vacuum. Unfortunately, I don't think I can convince the lab director to do this unless there is a very good reason to do it :(

[GigaFrog]

If I put the comp in vacuum then I can cool the cpu with a copper pipe carrying the LN. No worries about condensation this way

That is a very interesting idea!
Why not filling the comp with inert gas (like nitrous or something like that). You won't have to deal with keeping the pressure low (you will have to air-thightened the comp to keep the inert gas in). GPU, chipset and HDD will need cooling, as no airflow will be available.

Sounds crazy, but interesting.

We use inert gas in microelectronics, so that we reduce corrosion on connections, without having to deal with vacuum.

[azhari]

The problem with the inert gas is that it could still liquify at LN temps. Also, unless the chamber is completely sealed you can get air in through diffusion. Essentially, even under positive pressure air molecules can get in. With a vacuum pump the air molecules are less likely to get to the cold surfaces.

The nice thing is that we have some pretty big chambers to put the comp in. In fact we have chambers that can fit 10 people in them :D The only way a project like this would be approved is if someone is willing to pay $1000 a day to use the equipment. That isn't much compared to those guys who used that expensive 3M fluid with LN to cool things. I figure with a couple of thousand dollars we could set up a real nice vacuumed LN cooled computer. With continuous LN. Should be able to oc pretty good until the cpu breaks apart due to shrinkage.

[UserName]

as long as the chamber the CPU is in is airtight you have no condensation issue