Talk-Polywell.org

So I give this about 0% chance of going anywhere.

Well there ya' go. Someone call the Skunk Works and tell them to kill the project.

Lets not forget that they are also using neutral beam injection to further stabilize the plasma.

hanelyp wrote:Plasma stable over 1 second isn't the same as 1 second confinement time. In the simulations I'm running (OOPIC) of a slightly simplified version of the skunkworks configuration the confinement is amazingly stable, absorbing an order of magnitude more plasma than beta=1 before appearing to blow out. Keep feeding the plasma in at the right rate and indefinite steady state is plausible. But the confinement becomes leaky via cross field diffusion, mostly near the pole cusps, and unless the latest revision I've made fixes it it gets confinement time in microseconds in simulation. For reference, in my setup beta=1 is around 4e19 ions/m^3, proper fusion temperatures for deuterium.

It would be interesting to see what a sinusoidal current with DC bias in the central coil does to the plasma density distribution in your sim. Frequency slowly swept up to the maximum practical (inductive reactance limit), with neighboring coil currents held steady for the first sweep and ramped up or down with frequency for other cases.

EDIT: Should have included DC bias and bias-relative sinusoid amplitude. Get a 3D plot with peak plasma density as z and frequency and amplitude as x and y for each bias, or use color for density as 4th D and freqency, amplitude, bias as x, y, z.

Hanleyp:

"...But the confinement becomes leaky via cross field diffusion, mostly near the pole cusps, and unless the latest revision I've made fixes it it gets confinement time in microseconds in simulation...."

One observation on the LM field arrangement (at least in the patent) is that it is all-but to a 'unit cell', in that the reactors look like they would 'bolt together' end-on-end. One might think this would reduce or even eliminate net losses at the relevant (joined) centre-line point cusps. With even more imagination, maybe they could be fitted together in a ring to completely eliminate the pole cusp losses.

With that in mind, one might simulate a 'best possible case' scenario if one could somehow set the pole cusp losses to zero in the simulation of the unit cell. The obvious method would be to take everything moving out-on-the-left back in-on-the-right & vv.

Is this possible in the simulations that you run, and if so what happens to confinement time and the triple product?

RERT wrote:With that in mind, one might simulate a 'best possible case' scenario if one could somehow set the pole cusp losses to zero in the simulation of the unit cell. The obvious method would be to take everything moving out-on-the-left back in-on-the-right & vv.

I think the easiest way to simulate an "endless chain" of cells with the software I'm using is to put a dielectric with 100% reflection across the point cusp. I can already tell by visual inspection of the setup I'm currently running that the vast majority of loss is through those cusps. (ignoring supports, not simulated) Ions and electrons are getting nowhere near a loss surface anywhere else.

If anyone else is interested I could share my OOPIC input file.

choff wrote:My guess is that while the electrons can circulate outside the coils circumference the ions stay within. There's only two line cusps and two point cusps. One of the point cusps would be where electrons are injected, the other where they leave. So they don't have to worry about wiffleball formation in order to keep electrons from leaving through corners, they can operate at 1KW.

In the simulations I'm running the ions have no difficulty following electrons wherever they go, though at much less velocity. A polywell style excess of electrons could change that.

Getting electrons in through a point cusp doesn't seem to work well unless you can get a source smaller than the gyroradius (not practical with the PIC software I'm using) or sufficient intensity to exceed beta=1 for electron injection and brute force the way in. The problem with brute force injection is it tends to then brute force the way out on the other side.

Joel Rodgers called it as 100 times less output than polywell, LM claimed a 10 fold output breakthrough without stating what the original output would have been. My understanding is that there's an electrode at one end for injecting electrons and one at the other for removing those electrons with enough energy to leave the system. If they're working in ac mode then they might be injecting the electrons when the magrid field is at the low end of wave strength, the other source would be the neutral beams themselves that ionize upon entering. Being it's Skunkworks they might also be withholding a key piece of information, not showing photo's of a key component.

Looks like the plasma is flowing into the supports for the magnets. Wouldn't this cause the plasma to lose energy like the fusor plasma impacting the grid?

It is interesting to note that the patent applications by Lockheed Martin only mention deuterium-tritium as a fuel. Other companies have listed advanced fuels also in their applications.

A strange thing about the patent applications is the amount of text spent on describing the controlling computer system. It is hard to see the point in listing obsolete buses like AGP, ISA, MCA and VLB, that have disappeared from the market many years ago.

ohiovr wrote:Looks like the plasma is flowing into the supports for the magnets. Wouldn't this cause the plasma to lose energy like the fusor plasma impacting the grid?

Very much so. Unless something can largely prevent plasma impacts on the supports the reactor will have trouble breaking even.

Maybe L-M is seeking "jet mode" out of one of the ends instead of a more uniform confinement.

That might make either high-Isp/low-thrust (QED-DFP) or low-Isp/high-thrust (QED-ARC) a little easier.

hanelyp wrote: I think the easiest way to simulate an "endless chain" of cells with the software I'm using is to put a dielectric with 100% reflection across the point cusp. I can already tell by visual inspection of the setup I'm currently running that the vast majority of loss is through those cusps. (ignoring supports, not simulated) Ions and electrons are getting nowhere near a loss surface anywhere else.

If anyone else is interested I could share my OOPIC input file.

Thanks for the offer, but being completely unfamiliar with OOPIC, and having a day-job and a bunch of other projects already in hand, I think I'd be deluding myself if I took you up on it! Thanks again...

hanelyp wrote:

ohiovr wrote:Looks like the plasma is flowing into the supports for the magnets. Wouldn't this cause the plasma to lose energy like the fusor plasma impacting the grid?

Very much so. Unless something can largely prevent plasma impacts on the supports the reactor will have trouble breaking even.

I don't recall which thread, but I suggested earlier that a current flowing in (or out) through the supports would protect them with a magnetic field in the same way the walls are protected by the parallel magnetic field lines from the coils. Even if they don't *need* a current into the coils through the supports, they could run one in anyway to protect them. If they send current both in and out through the supports, they can even arrange for that field to die off extremely quickly with distance.

Being very naive in this field, is there something obvious which makes this approach a non-starter?

hanelyp wrote:

ohiovr wrote:Looks like the plasma is flowing into the supports for the magnets. Wouldn't this cause the plasma to lose energy like the fusor plasma impacting the grid?

Very much so. Unless something can largely prevent plasma impacts on the supports the reactor will have trouble breaking even.

I think the key is how much plasma is in the vicinity of the supports. Assuming that my view that the Lockheed concept is similar to the Polywell, implies that the plasma, especifically the electrons outside the central core is much less than in that core. The electron rich plasma that is flowing around the end magnets is made up of escaped electrons just as in the Polywell. If confinement time in terms of number of passes is 10,000 or more under Wiffleball mode, the internal density of electrons is ~ 10,000 times greater than the electron plasma outside the magnets. Thus the relative numbers of electrons that have an opportunity to hit the supports is 10,000 times less relative to the core exposure. It is tolorable. Shielding the supports with B fields of electrostatic shields would help further provided it does not compromise other aspects. The nubs in WB6 was thought to be tolerable- at least in the experimental machine. Further research with WB7 showed further efforts to minimize exposed surfaces near the magrid was desirable and obtainable (?). Wall standoffs seem to be the answer. Admittedly though, the electron dynamics outside the magrid are different with a positively charged magrid verses a grounded neutral magrid.

Note the image of the two magnet test setup. The neutral beam and electron injector protrude well within the magrid (internal plasma volume). They have to be highly exposed surfaces. That they appear to be away from the cusp might help in some ways- though it may hurt in others. It might be useful for testing certain concepts, but I suspect they would be intolerable in any machine that hopes to produce net power.

Dan Tibbets

RERT wrote:I don't recall which thread, but I suggested earlier that a current flowing in (or out) through the supports would protect them with a magnetic field in the same way the walls are protected by the parallel magnetic field lines from the coils.

Running a 2D simulation that works fabulous, at least if you run enough current. The 3D case I suspect has problems with the interface between current carrying standoff and toroid magnet.

Squeezing the magnetic field between inner and outer coils, excluding plasma by mirror effect, may be worth looking at.

Electrostatic shielding against electrons would work well if ions can be kept at a distance by some other means.