Fusor operation pictures, comments on grid designs

[Doug Coulter]

Oh yeah, we do, though perhaps not perfectly. I work on things such that my (or anyone else here) yearly dose is no more than about doubled from the background, which is low here (and half what it was from nuclear tests when I was growing up). At least for gammas. So in other words, my net exposure to things that make a geiger counter count is about the same as Joe Sousa up in the Boston area who isn't doing this at all, but I'll have a better calibration on that soon as he's sending me one of his pancake tubes to build up here so we can compare directly.
In other words, we're still under (except in bursts) running lower exposures than some pretty ridiculously low limits they use for other workers. One could argue that bursts are worse than continuous low level background, of course -- that would be my guess at the limit anyway. Obviously if you took a lifetimes worth of cosmic background dose in a minute, it would kill you right off, where there's not much convincing evidence that it's a major cause of death when spread out over a lifetime.

This of course doesn't take neutrons into account at all. Still working that one, but in general, we don't run hard and long day after day, which kind of minimizes that. There is this very high thinking to running ratio we use, which seems the quickest way to get to improvements anyway (much like software, planning is faster than coding).

Typically, I will make a few runs on a day I run (perhaps 20 minutes total), but those days are not terribly frequent. Much more time is spent building apparatus, testing and measuring my metrology gear, things like that.

Every couple weeks theres enough "new" built up to be worth another run, more or less. That changes over time, but so far, as we've made say a 10x improvement, we've run that much less or at that much lower power, to compensate. So far that plan has worked. As we move to pulse mode, there's no particular reason to test with more pulses than it takes to get out of our measurement noise, so that's a potential boon. Other progress has made it possible (finally, that was one heck of a slog) to run the thing remotely, and I have a nearby cave I can stuff it into if that becomes an issue. So no big worries about that here, yet. But we pay attention, certainly.

And we're only doing the DD reaction because it's easy and easy to measure, the neutrons are a good signature and proof it's working. As we advance I plan to move to other reactions that make fewer neutrons, as I'm not in the fuel breeding business and have little use for medical radioisotopes. Not that it wouldn't be fun to try DT for a little bit....

[Joe Jarski]

Doug, what are your thoughts on the titanium grid vs tungsten or tungsten alloys? I'm trying to find a cheaper alternative (for now) to something like tungsten-rhenium that can be spot welded without becoming brittle like pure tungsten.

BTW, I found out that my TIG welder has a really cool spot welding feature that does a great job of welding thin material and wires together in an argon atmosphere.

[Doug Coulter]

Short answer -- they all work. W can take heat better than Ti, but not a whole lot. They all get brittle if they don't start out that way, but it's not like you're going to let your kitty cat play with them -- it just doesn't matter. I've had "issues" with spot welding W, because it sticks to electrodes, so I came up with another way for type C thermocouples (W/Re) that uses a tig and a jig. My main grid right now uses W rods (.040 from a tig supply house) and graphite ends, drilled for the rods, and a piece of type C wire "rubber band" or tie wrap to hold it together. It can be fairly delicate and still work. If you're going to do spheres, it's a lot harder...I'd go for Ti wire (McMaster or when you visit, I'll give you some) for starters. Pretty ductile before it gets run a lot, and usually not too brittle after that, as it is hot, and the way I shut down is to turn off HV and open up the pumping while it's still hot, so the hydrogen leaves it while it is still hot. Works fine.

[chrismb]

Rhenium should make a very interesting grid.

I posted a comment on it over in fusor.net, but no-one followed up with any further comments.

The reason it is so interesting is that, as far as I understand it, rhenium is the only material in the universe that is fully ohmically conductive both in pure and in oxidized form.

One might therefore, naively, presume that this would be a superior material for grids for that reason, that it never forms a dielectric layer under oxidation. As it happens, Wisconsin has run pure rhenium grids and, indeed, neutrons rates were highest for that material [I seem to recall].

The other thing about rhenium is that it is rarer than platinum, yet cheaper. I figure there must be some commodity futures worth looking at for that stuff. Stock up on it now...

I can't say if it would become more or less brittle that W, but I will presume the characteristics I describe are interesting enough to give it a shot, if you can.

[Doug Coulter]

I only have experience with the alloys of Re and W (Type C wire) which is pretty expensive stuff, but is ductile and easy to form. Hard to spot weld, as are all those super hi temp alloys. It retards the nasty water cycle with tungsten, and increases the ohmic resistance a bunch, which really isn't much of an issue here as we're in the milliamps. Nanmac sells the wire, minimum order, a couple bucks a foot. You get pairs, one 5% the other 16% Re. Makes great filaments for electron emitters, for which I coat it with yttria to help it emit at lower temps.

Any of this stuff works well for grids....NiCr hasn't worked out well for me, for reasons I don't understand -- it lives, just not many neutrons.

[Joe Jarski]

I've recently been thinking over a little diversion from where I was going to start with my fusor, but didn't want to spend a fortune until I proved out the (not really) "new" idea. The fabrication will be somewhat delicate compared to what I'm used to doing, so using Ti will give me plenty of practice material, but still be an improvement on SS in most respects. If it works out then I can go to some of the refractory metals. I like the "self bake out" to reduce hydrogen embrittlement.

Chris, I spent some time over there looking at the different materials that people were using and saw the post. I didn't realize that the oxides were conductive. It sounds like rhenium would be a good material from both a fabrication and properties standpoint. I'm surprised that no one has used it yet.

[Doug Coulter]

I've used the W/Re alloy and it's my #2 grid now, the one I use as an ion source. You can see it in the pic above.
I don't know that pure Re is available/affordable at this point. Oxides are not going to be an issue in the conditions here. The hot hydrogen will reduce them faster than they can form. Nor will a thin film of oxide insulate at these voltages, it'll get blasted right off should it form.

[chrismb]

I found the plot...

See the bottom plot, column 3;

http://fti.neep.wisc.edu/presentations/ ... r_tofe.pdf

[Doug Coulter]

Thanks Chris, that's actually quite encouraging, on analysis.

Don't know where they found that pure Re (someone else's money, no doubt), but heck, those numbers aren't as good as what I am getting here now at 50kv and a much simpler and smaller setup (with pure W) -- and with far less input power, even in the less-good static mode. Hah! That tiny difference could be grid-to-grid precision variations alone, I get tons of difference vs grid-build accuracy here. Going from 80 mils error to 30 in the build made the output go up on the close order of 10x for me in a more perfect geometry (where it is perfect, in other words, away from the cylinder ends). Next one will be inside 10 mil error. In other words, still horrible by optics standards but also better than before.

We're going to kick some butt and take some names, guys. Or do already if my numbers are correct. I think they might be close enough to almost start bragging...

I doubt the grid material matters that much if it lives under the conditions -- I'm more worried about finding lower secondary electron emission at this point. With one exception, that's what I see here-- material isn't too important; I never tried that hard with the NiCr (changed too many things at a time for that observation to be declared universal -- that was the spiral grid above that failed for other reasons -- crummy geometry for focus for one).

Based on what I see and their numbers, their Q is lower than mine, Richard's, JonR's, Tyler's, by quite a lot (factor of a few at least). I see that same kind of poisser on what I call "failures" here, not that you can assume all that much from that alone -- it's only semi-related as a diagnostic. It could be that making the thing too big is bad, dunno -- working on testing that here, actually -- already tested "too small" fairly well (but not to completion). What they are seeing is what I see when I try for too much current, and hit space charge defocusing, here -- neutrons stop going up fast with current and the curve rounds over (same as theirs). A lot of things have to be right together to get to good "luminosity" at focus, and they've not hit on a sweet spot as I have, is my take on that data.

The "error surface" in howevermany dimensions has a lot of local minima and maxima, so you can't just sweep in one parameter (or a couple) and find the global best spot (same problem as training a neural network), there are all these intermediate peaks in the function where it gets worse in all directions from there or doesn't change, and any sweep looking for "better" or a gradient will end at one of the local peaks -- but the real best might be a good ways off, with a big dip in between. This isn't a simple linear situation at all. For example, I hit the limit at 9.8 ma on mine (tried up to 40 ma) because it just defocuses -- more stuff going through the grid, but much less per beam area at the intersection as the beam(s) spreads out. In fact, the equations tend to indicate it should degrade somewhat before that, so I'd guess there's a little fudge from the electrons, or some other cause for it to even be as good as it is.

Consider the case of an optical telescope, far from focus. All you see is a blur, uniformly bright, no matter what it's pointed at. Changing focus slightly doesn't change anything you can observe if you're way off. In fact, until you get close, you can't even tell if you're moving in the right direction. Only quite near focus (with lenses that can even achieve that when properly adjusted) can you tell, and get it right.

Instead, start way off, and with a lens that is a funhouse mirror (imagine the image from an optical glass lens made from triangular shapes, or some polygons) -- you can fiddle endlessly and never get a decent image. With luck you can get some concentration, but you'll never image to a point.

Now on top of that -- you can't see the actual image, just the artifacts around where there was scatter (the poisser). That's the challenge we all face.

It could be possible to make what amounts to a Fresnel lens out of oddball shapes -- man figured that one out -- a long time after we figured out how to make normal lenses. I suspect history will at least rhyme, here.

In our case, the effective curvature of the effective lens is a function of field, which is a function of spacing of the grid wires. So, if you want a regular lens, you make those circles. If you're willing to accept the idea of a cylinder lens (in any case, remember it's a bunch of lenses, and a bunch of colliding beams), you make the elements straight rods, producing a cylinder lens between each pair of them. Either way, the object is to create the most possible particles per second going through a finite and small place so they can hit -- and the probability of hitting goes way up the smaller that space becomes (much more than linear) -- to the point where if that space was one nucleus wave-function wide -- it becomes 100%, more or less.

You can't get more by just running more pressure, as that implies collisions that defocus things before you get there. More current for a given ion velocity just means Coloumb defocusing. They are trying to overcome basic physics laws with brute force. It's unlikely to be the answer unless they get to a heck of a lot more brute force than is likely in that setup (think the lasers at NIF, with luck).

Well, enough for now, but that's my own philosophy (post a couple beers). So far, if these and my numbers are even close, it looks pretty good for my approach so far.

[fusordoug]

Here are some possible error surfaces in two dimensions. Remember we have a lot more of them to deal with.
Focal length a function of element spacing and net charge (including the stuff there, which is a variable itself).
Gas pressure, volts, current. Ratio of inner to outer dimensions. Ratio of ions to free electrons. Neutrals that
scatter things if there are many of them. And all this is potentially (and in my case hopefully, since I want to control that)
a function of time. These are from Timothy Masters book on neural network training and the issues with simple gradient descent
as a way to optimize things. This is really the same problem....Only in some cases the math that would even produce these plots doesn't yet exist,
the perturbations of one on another aren't convergent either.