Vacuum Technology is Black Magic

We've continued plugging away at the optical excitation experiment discussed in the Week in the Lab series last year, and have finally managed to get a decent metastable signal out of the thing. The signals are at pressures that are considerably higher than I would like (and quite a bit higher than the turbopump is happy with), but recent results from a colleague at Argonne National Lab suggest that this may well be due to the fact that one of our lasers is operating at much lower power than would be ideal. Still, it's data, and data are always good.

As always with experimental work, getting preliminary data triggered a bunch more work, including an interesting side trip into the wonderful, annoying world of vacuum technology. We fixed the problem on Thursday, but as is entirely typical of vacuum issues, I don't feel like I've learned anything in the process.

(Details below the fold.)

i-644d558aa4f6cc75a1ebf37f88ed4c09-sm_gas_flow.jpgThe basic arrangement of the system looks (schematically) like the figure at left. We have a bottle of krypton gas ("Kr") with a gas regulator ("X"), which feeds gas into a copper tube, then through a metering valve ("M") to control the rate of gas flow, and a cut-off valve ("C") which lets us seal the gas bottle off from the rest of the system (the first metering valve we used could only limit the flow, not cut it off completely). It then flows through some more copper and glass tubing into the main chamber. There's also a thermocouple gauge ("G") to monitor the pressure, though it doesn't go down all that low (there's an ion gauge on the main chamber that will read down to 10-9 torr or below).

We had this system all set up, and pumped down very nicely (to 10-7 torr, for those who care, which is very good given that there are three compression fittings in the system). Then we decided to upgrade the metering valve, and swapped in a better one from the other vacuum system in the lab. After that, the base pressure was almost a factor of ten higher-- still good, but not as good. We had a little adventure with the connections on the valve, though, so it wasn't too surprising.

This week, we did a second upgrade of the metering valve, to put a brand-new valve in the line (the old one had a tendency to stick and then suddenly release, letting way too much gas in), and suddenly, the pressure wasn't worth a damn. With both valves open, it wouldn't get below 10-5 torr, which is Not Very Good. And there was no obvious reason for it.

A little poking around suggested a leak between the gas bottle and the metering valve, which was about two feet of copper tubing with Swagelock fittings on both ends. Closing the cut-off valve brought the pressure all the way down to 10-7 torr, while closing the metering valve saw the pressure drop to the high 10-7 range, and then creep back up toward 10-5, consistent with the valve not closing all the way. I closed things off, removed that tube, and reconnected it, and it made no difference. I tightened down all the connections on that tube, but didn't manage to make the vacuum better. I did manage to crank one of the connectors down hard enough that it cut the end of the tube off, but never saw any improvement in the vacuum. Putting a brand new connector on didn't help, either.

Finally, I got disgusted with the whole thing, and told my student to just replace the whole thing-- take a fresh piece of copper tubing, put two new connectors on the ends, and try that. And damn if it didn't work. We had shut down the turbo by that point, but the mechanical pump by itself took the pressure down below where the thermocouple gauge would read, when it hadn't gone below 50 millitorr with the old tubing.

This ate the entire afternoon, and I still have no idea what the problem was, or why replacing the copper tube fixed it. The tube was moved around quite a bit during the various upgrades, but there were no visible kinks or damage, and the connections on both ends were replaced. There must be some pinhole leak somewhere in the middle of the tube, but I can't see where or how.

Such are the joys of experimental atomic physics. On the bright side, my student says everything's working nicely, so we should be able to get some good data tomorrow. Unless something else decides to stop working for no apparent reason...

Tags

More like this

Here's a picture of the gas-handling line leading to the discharge region seen in the plasma post: How many valves can you count in that picture? If you said "seven," give yourself a pat on the back. Here's the same picture with the valves numbered for your convenience: ("Seven! Seven valves! Ha…
I spent the bulk of yesterday afternoon doing vacuum system work, specifically working on the system to feed gas into the atomic beam source. My feelings about this can be inferred from the Facebook status message I set at the time: "Chad Orzel abhors a vacuum." The apparatus I'm building uses…
Over in the reader request thread, Richard asks for experimental details: I'd be interested in (probably a series) of posts on how people practically actually do cold atoms experiments because I don't really know. I needed to take some new publicity photos of the lab anyway, so this is a good…
A great many physics experiments need to be conducted at low pressures, in order to avoid sample contamination, thermal effects, or dissipative forces produced by interaction with air. Some experiments don't require all that much in terms of vacuum, while others require pressures so low that they'…

Gosh, things don't change at all. Back when I was serving a failed apprenticeship in atomic physics at least we had something called a leak detector, which was essentially a single-purpose mass spec and a tank of helium. Watch the fun when it runs out that the block of steel from which part of the apparatus was machined is, uh, porous.

But at least you're not working in a lab with so much stray RF running around that your pump laser output goes unstable...

Just cover the whole thing up with a 1/4" layer of Torr-Seal. During a previous incarnation as an experimentalist, I did that once with a chamber about a foot tall, 4" radius. Looked like the Michelin man, but I got the data, finished my report, and went theory all the way and never looked back!