Different than a vacuum by far, the issues are mostly on the inside because there are gas and ions on that side too. The more or less standard HV feedthrough design does NOT work well here, for a number of reasons. The one I'm thinking of when I say standard is the one you get from say Lesker, where there is a wire or rod attached to the end of the ceramic, that then "floats" inside the large hole in the ceramic and eventually enters the tank. There are (at least) two reasons that doesn't work in a plasma set of conditions. One is Paschen's law, and that extra length back to the "top" of the insulator may be the longest path in the system -- so the gas/arc happens lengthwise down the central rod, and fries the ceramic near the bolt -- or fries the ceramic where it joins the flange with an arc from the top to the inside at the flange. Another issue is that with that long rod, it attracts all the ions, neutralizes them, and gets hot -- just what isn't wanted. So one has to eliminate both that long path, and any exposed extra conductor, a modification that turns out to be really hard to make on a normal feedthrough design.
I have learned a few things empirically, that I will share here, but I'm looking for more input from others who have fought this particular dragon.
One is that to insulate the central rod, it's not enough to insulate it "pretty well", it has to be really good, and able to stand off the full voltage, as otherwise the outside will attract ions, and they hit hard enough to reduce any oxide chemically --. Emprically, pyrex is pretty conductive by these standards, and so far, only quartz has been good enough for that. But near the end where it meets the fusor grid, it gets hit by high speed ions anyway, and tends to become metallic silicon...leading to quick failure there (pyrex fails immediately, shatters, ugh). So the only thing I've found that will do for that last inch or so is alumina, which when reduced evaporates off, and stays insulating for a lot longer (until some sputtered grid metal gets on it).
For the outermost part, I have found that the conductivity of pyrex is a good thing indeed, achieving much the same effect JohnF was mentioning on another thread of spreading out the voltage differential nicely along the length. So when that outermost piece is grounded in the middle (via the tubing coupler that lets it into the tank) we have no lengthwise arcing on the outside side, and on the inside side, it stays at near ground too, and doesn't attract hot D+ ions so bad, and therefore lives a long time -- and even if it becomes a little conductive on the end, it's insulated by the inner quartz tube well enough to not cause trouble.
Ok, those tricks, described elsewhere on this forum, work fine at up to my current 53kv limit, and the fat copper rod I use helps with heating issues, being able to carry heat out of the tank nicely, even though it's nearly a foot long. The problem is, now I want to go up to the hundreds kV, and there is no source of quartz tubing thick enough for the middle insulator I can find, they just don't make almost solid rod with a tiny hole in the middle (thick wall), and in general, due to tolerances, you can't just get telescoping sizes.
So, a new design is needed -- I already know there's nothing I can buy that solves all this, and my friends at CERN and ITER haven't solved this one either (private communications and thanks Joe S for hooking me up with those experts).
I calculate that a roughly 1" diameter piece of teflon will do the insulation fine...but...it won't like exposure to vacuum one bit, and especially not hot ions, and the stuff flows as Jerry points out elsewhere. So I began with the idea of using a pyrex outer, with teflon insert, but now I have to make the glass/metal seal on the in-tank end. Since pyrex is conductive (compared to quartz or teflon) I then have the problem of the in tank end no longer sitting at ground and it will attract ions with the usual destruction. McMaster doesn't sell alumina tubing large enough to slide a piece over that....it's extremely hard to make a quartz/metal seal or a quartz/pyrex seal -- the expansion tempcos are too far off. I suppose I could try the same geometry I'm using now, but using teflon instead of quartz for the inner insulator, but -- it would be in vacuum even if I use the alumina extension, and it would get hot, which would make it give off various nasty gases on the inside of the tank, not good. Probably not so hot as to ruin it, just everything else that got hit with the fluorine it gives off in that case -- out on the air side, no troubles in other words.
So I'm stuck between what I think I need, and what I can find materials to actually do here, and any thoughts are welcome. This is just a plain bizarre kind of application which no one is designing hardware to handle. It kills commercial feedthroughs right quick, well under 100 hours of running, sometimes inside 20, and at those prices, well, I can't do this. We are running at a pressure point where a spark doesn't take the shortest path -- on the left side of Paschen's law curve, and I don't think many other applications ever do that -- if any. Were this a vacuum, I'd be fat, but of course a fusor doesn't run in a vacuum, it needs fuel!
FWIW, I am running at 1.6 e-2 millibar indicated, or thereabouts. At 1.9 and above, I get fewer neutrons and the current tries to rise without limit, below that too much it just goes out and I can only run currents more or less equal to what my ion source provides, it's not self sustaining in the 6" diameter sidearm. If I put another grid out in the big part of the tank to act as an ion source, it will light off down to about 1.4 e-2 mbar at 40kv or so, but no lower and actually, I'd like to go lots lower than that if I can find a way (working on that, separate project). The reason is -- higher Q when the mean free path gets long so I'm mainly getting fast on fast collisions and boy does it ever go up quick -- double the output going from 1.9 e-2 mbar to 1.6 with the same input volts and ma.