videos of polywell phase space

[happyjack27]

Am I on crack? Fair enough if so. Just thinking out loud.

that all makes sense. i'm not talking net power either. i mean there really is absolutely no fusion at this range. if i want wb + fusion i will ultimately have to scale the velocities up to simulate what would happen in a larger machine.

for density calculation i'd still have to sort or bin, and i'd rather sort 'cause it's more accurate. and i'd rather sort on the gpu because it's faster and it eliminates the pci bus memory transfer bottleneck.


for statistics i'd have to add it up per particle, that's the biggest trouble. but in that i'd just do a parallel reduction and that's really no biggie. so i'm going to do that first, for things that don't require density, which means unfortunately not the total system fusion rate.

it all depends on electron confinement time per unit energy input anyways, so the electron loss rate and electron loss energy are the most important stats to get out, imho. and once i have those i can do a real quantitative comparison of diffierent configs and parameters, rather than just visualizations.

that makes it a priority.

[ladajo]

fair enough. You should get the basics of confinement worked out first.
e- leads to D-T leads to D-D leads to PB&J.
No Well, no Fusion.

[krenshala]

You mention it displaying the fusion rates; does it track/display the collision rates (yet)? Even if they weren't sufficient for fusion, knowing the collision rates of the various particles would be interesting, even if not actually useful (and I'm sure they are useful ).

[happyjack27]

You mention it displaying the fusion rates; does it track/display the collision rates (yet)? Even if they weren't sufficient for fusion, knowing the collision rates of the various particles would be interesting, even if not actually useful (and I'm sure they are useful ).

i don't know what's meant by "collision rates". i'm not modeling the residual strong force, and the coloumb force is 1/r^2 so in the sim the particles cannot "collide".

and each particle representes billions of real particles anyways.

for fusion rates i would model it by cross section * density squared.

[krenshala]

You mention it displaying the fusion rates; does it track/display the collision rates (yet)? Even if they weren't sufficient for fusion, knowing the collision rates of the various particles would be interesting, even if not actually useful (and I'm sure they are useful ).

i don't know what's meant by "collision rates". i'm not modeling the residual strong force, and the coloumb force is 1/r^2 so in the sim the particles cannot "collide".

and each particle representes billions of real particles anyways.

for fusion rates i would model it by cross section * density squared.

Well, you need collisions of some kind for actual fusion (at least, that is my understanding ). I was just thinking that tracking how many "collisions" you get in the sim, and correlating that with the fusion cross-section would be additional useful information.

I fully understand if that isn't something you want to tackle on the programming side right now, however.

[happyjack27]

wb-6, 15cm, lowered the mag field to get closer to beta=1. also started the electrons uniformly w/in half the magrid radius, simulating introduction via ionization. momentum space view, followed by position space + radial momentum. looks to me like closser to beta=1 and better confinement. but again no numbers. porbably after the holidays.

http://www.youtube.com/watch?v=XN9yr_5u6N4

[happyjack27]

more doedec video.

this starts out as x=volume elements from center, y = absolute momentum.

then i rotate it to show you x = absolute momentum, y = potential energy

then i rotate it again to show x = volume from center, y = potential energy

so that's all orthogonal projections of those 3 variables.

then i quickly rotate back through the other 2 projections.

http://www.youtube.com/watch?v=Mi9N29Tah3A


this is a video of the oscillating potential well.

x = volume elements from center, y = electric potential

http://www.youtube.com/watch?v=o__NmJV-LRA

[happyjack27]

you'll find this funny: apparently there's a stuck key on my home computer. so when i vnc in to it... well i can't unstick the key remotely of course, so i'm kind of screwed for the day. (i tried to disable keyboard inputs remotely, but it wouldn't let me check the box.)

[ladajo]

Happyjack,
Have you done anything further since the New Year?
Did you ever get in touch with Joel Rogers about your sims verses his PIC runs?
Just curious. Hate to see you give up on some hard work that was appreciated.

[rjaypeters]

Thought of hiring your time to Doc Nebel and team?

would love to. if they're interested, i'm available.

[speculate]In the best of all possible worlds*, this has been accomplished. Because of an iron-clad NDA, we won't be hearing from happyjack27 for a while.[/speculate]

*in which, Pangloss assures us, we certainly live.

[happyjack27]

i'm done simulating for the foreseeable future. i don't think there's anything more i can learn from it.

my main conclusions are:

1. it works.

2. injection methods are important for improving efficiency. magrid is as good at keeping out as it is at keeping in, so good to start with neutrals to bypass them right away. (in such a high-field vacuum, they should ionize right away.)

3. plain old higher-order polyhedra seems the way to go as far as magrid config is concerned. you don't need to go too high to get the payoff, though, and as you go higher expect it not to focus as tightly to the center, so you'll have a bigger wiffleball. which could be a good thing because that means more fusion area, but means you'll need a larger machine not just for the mechanics but to keep the well depth good. though due to scaling of field strength and what not, with a big enough machine, reaching optimal velocity for fusion is fairly easy, even for p-b11. so long story short before going to higher-order polyhedra make sure the size justifies it.

4. the ions will thermalize overtime. you'll get lower energy" ions staying in the center and higher energy ions oscillating through the whole machine, and everything in between. on average ions are "hotter" in the center and colder on the edge, as they oscillate through KE-PE exchange like a bunch of pendulums swinging at different frequencies.

5. electrons in the center are cold and dense. they get faster as you go out from the center. they'll escape at high energy out the cusps in generally the same direction - good way to recycle energy there.

6. electrons will get higher angular momentum as they go towards the outside, whereas ions will have very low angular momentum except in the very center as they pass by the center. this is because the ions are governed by columb forces of the electrons and the magrid bias, while electron motion is dominated by the magnetic fields, as expected by theory.

that probably about sums it up.

for any one who wants to continue, the source-code is available on source forge. the project is called "em-body". (requires a cuda gpu)

[MSimon]

4. the ions will thermalize overtime. you'll get lower energy" ions staying in the center and higher energy ions oscillating through the whole machine, and everything in between. on average ions are "hotter" in the center and colder on the edge, as they oscillate through KE-PE exchange like a bunch of pendulums swinging at different frequencies.

I'm wondering why you didn't see any self organizing effects such as bunching (ala the klystron).

[happyjack27]

4. the ions will thermalize overtime. you'll get lower energy" ions staying in the center and higher energy ions oscillating through the whole machine, and everything in between. on average ions are "hotter" in the center and colder on the edge, as they oscillate through KE-PE exchange like a bunch of pendulums swinging at different frequencies.

I'm wondering why you didn't see any self organizing effects such as bunching (ala the klystron).

not sure what you mean. any effects that come from QED (including "de-ionization") (with the exception of nuclear cross-section) will not show as the sim uses only classical electrodynamics. but at this scale and energy level, i'd imagine they can be safely ignored. (save some grid losses due to tunneling.)

anything that follows from classical electrodynamics (which is most everything.) should show up.

[MSimon]

4. the ions will thermalize overtime. you'll get lower energy" ions staying in the center and higher energy ions oscillating through the whole machine, and everything in between. on average ions are "hotter" in the center and colder on the edge, as they oscillate through KE-PE exchange like a bunch of pendulums swinging at different frequencies.

I'm wondering why you didn't see any self organizing effects such as bunching (ala the klystron).

not sure what you mean. any effects that come from QED (including "de-ionization") (with the exception of nuclear cross-section) will not show as the sim uses only classical electrodynamics. but at this scale and energy level, i'd imagine they can be safely ignored. (save some grid losses due to tunneling.)

anything that follows from classical electrodynamics (which is most everything.) should show up.

Self organizing (the beams - bunching) should show up in the simulation. It has shown up in the Joel Rogers simulations. I would not expect "various" frequencies of transit after sufficient operational time. A few hundred usec to a msec or so. By that time the particles should be bunched in oscillating packets. And the packets from all 6 directions should be in syncronism. i.e. they meet at the center.

The oscillation is why POPS should help.

[happyjack27]

...

anything that follows from classical electrodynamics (which is most everything.) should show up.

Self organizing (the beams - bunching) should show up in the simulation. It has shown up in the Joel Rogers simulations. I would not expect "various" frequencies of transit after sufficient operational time. A few hundred usec to a msec or so. By that time the particles should be bunched in oscillating packets. And the packets from all 6 directions should be in syncronism. i.e. they meet at the center.

The oscillation is why POPS should help.

probably not enough sim time for that, then. i run the sim very slow so that it gets good temporal resolution.

that or if the sim time isn't fine enough i suppose frequency-based effect might not materialize as the finite sampling rate would cause distortion of the very high frequencies. though i think i ran it slow enough for that.

some of the sims (w/no charge or negligble grid charge, i believe) showed oscillation of the p-well. this could be "bunching". but i'm kinda skeptical of that. not sure i was using good enough time res on those. (though part of my doubt was _because_ the well appeared to have a multi-frequency oscillation pattern.)

http://www.youtube.com/watch?v=5nmM9gnNmKc
http://www.youtube.com/watch?v=Md68rUeXD_8

(if the variation of particle density across the x axis looks off i might be showing the x axis as x^2 or x^3 to normalize density w/volume (since volume grows w/r^3 from the center))


i suppose it's also possible the effect doesn't happen in real life. but i think there just wasn't enough sim time. sim time to real time ratio is enormous . if i recall correctly something on the order of a picosecond per hour. running it for a day i still barely got it to the points for ions to be evenly spread out. for oscillations to form (if they do) would probably take considerably longer.