Saturday, March 24, 2007

Polywell - Adding Details

Tom Ligon, a researcher who worked witth Dr. Bussard on the Polywell machines has added a few points to Polywell - As I Currently Understand It. Let me note to start that he thinks the basic description is excellent (with minor corrections). You can read what Tom has to say at General Fusion Theory. I'm going to put it here in my own words to make sure my understanding is correct. If I make a mistake I'm sure Tom will correct it when he gets back from vacation in a couple of weeks or so. Keep your eye on this space.

First point of minor correction. The magnets can be all North poles facing in or all South poles facing in. I picked one just to make the explanation simpler.

The next part is trickier. Here is where the electromagnets come in. We replace the ordinary magnets with coils of wire. You put a current through them and you have an electromagnet. The coils will be shaped to match the face of the solid it conforms to. If the solid is a cube the coils will be square. If the solid is an octahedron the coils would be triangular.

Now here is a part that wasn't clear to me earlier. Once you have the coils all wrapped up nice and tidy like you cover them with a metal sheath which then becomes the + grid.

Here is a nice picture of what it looks like in the WB-6 experimental version:

Fusion Reactor

Note that the coils do not conform to the shape of the solid picked. That is definitely an experimental error and will be corrected in the next version. This is what the Bussard folks call the MaGrid which stands of course for Magnetic Grid. Cooling is definitely going to be a problem because the cooling properties of a high vacuum are not very good. It is why we use vacuum bottles to keep things hot.

Tom points out that the Deuterium should be injected inside the grid so that the atoms are quickly ionized. That gives you a number of engineering problems I'm not going to go into here. Let us get the theory down first. You can't engineer what you don't understand.

Tom says that the plasma physics is probably the hardest thing to understand in the machine. I'd agree with that. You have electric fields, magnetic fields, coils, and charged particles. This is probably one of the toughest areas of physics because of the interactions. Every thing affects everything all at once.

Let's start this next stage with the magnetic field. Here is a nice drawing of what it would look like before any particles are circulating:
Fusion Reactor Magnetics Used by permission of Tom Ligon, copyright 2007
Used by permission of Tom Ligon, copyright 2007

Next lets look at it with electrons in the mix:
Fusion Reactor Magnetics Used by permission of Tom Ligon, copyright 2007
Used by permission of Tom Ligon, copyright 2007

So the electrons are injected into the center of the machine and last a long time. Any electrons that leak out are attracted back in by the grid. As long as the magnetic fields keep the electrons from hitting the grid they can just keep cycling along inside the machine. Long electron life time is one of the keys to net energy production.

Tom claims that in operation the circulating currents inside the machine are very high. Maybe millions of amps. This current of course, like any current, generates a magnetic field. This new magnetic field pushes back on the field around the coils making the field stronger near the coils. Now I'm having trouble visualizing a spherical current that creates north poles on all sides of a shere. Of course if the circulating currents were mirrors of the currents in the coils that could happen. However, I don't see how that happens naturally without creating forces that drive the electrons out of the center of the machine. We will put this down to details to be explained later.

OK some how you have electrons madly circulating around the center of the machine. When I get a better explanation of how this happens I will give it to you. These electrons form a potential well between the electrons in the center and the grid. The potential difference is about 80% of the grid to shell voltage. This is good because it means the electrons are less likely to have enough energy to reach the grounded outer shell where they are lost.

Now you start injecting fuel into the machine.The fuel either comes in as ions from an ion gun or it comes in without a charge and some of it is ionized by collisions with the madly spinning electrons. The fuel is affected by the same forces as the electrons but a little differently because it is going much slower. About 64 times slower in the cae of Deuterium fuel (a hydrogen with one neutron). Now these positively charged Deuterium ions are attracted to the virtual elctrode (the electron cloud) in the center of the machine. So they come rushing in. If they come rushing in fast enough and hit each other just about dead on they join together and make a He3 nucleus (two protons and a neutron) and give off a high energy neutron.

Ions that miss will go rushing through the center and then head for one of the grids. When the voltage field they traveled through equals the energy they had at the center of the machine the ions have given up their energy to the grids (which repel the ions), they then go heading back to the center of the machine where they have another chance at hitting another ion at high enough speed and close enough to cause a fusion.

Hopefully this happens often enough so that there is more energy coming out than going in by a lot.

As you can see there are a lot of loose ends to be tied down on this. There is lots more to understand what is actually happening. Fortunately there are a lot of amateurs working on this and progress is being made with electrostatic machines.

More details on these experiments are going to have to come out before any serious engineering could start.

There is more than enough information here for a serious experimenter to build a test machine to see if this path to fusion holds promise. You might not even need to try fusing Deuterium to do diagnostics on virtual cathode formation and other details on what is going on in the plasma.

Since diagnostics, and not power, are the object, the electron guns could be made oversize to compensate for electron losses. If you could cool the grid with something like Flourinert or liquid nitrogen and crank up the electron injection really high despite the losses a lot of useful information could be found.

So I see two steps at the very beginning:

1. Verify virtual cathode production in a magneto/electric field of minimum size.

2. Scale it up to a test reactor capable of operating the grid at around 30KV (max) at useful currents with Dueterium fuel. It should be possible to test how reaction rates scale with voltage on the grid and current through the coils. You can also monitor electron gun currents vs reaction rates (neutron production) to get some idea of the losses.

The next step after you have some good data (you might need 4 or 5 proto types) is to scale it up to a machine with 1 KW energy production, follow that with 100 KW, then 10 MW. Start the scale up when you are at least 70% confident of the next design. For experimental purposes lots of things can be fixed along the way. In fact depend on the fact that problems that are insignificant at 1 KW will be serious obstacles at 10 MW. That is the key to development. Fixing problems as they come up so as to meet targets for cost, development time, specifications, etc. Solvitur Ambulando.

WB-6 coils were .15 meter across and looked like this under construction:

WB-6 under construction

They were run at around .1 Tesla for .25 milliseconds.

The final article in this series is:

Polywell - Making The Well


Anonymous said...

It's possible I'm missing something obvious, but which parts form the usable, current-producing potential?

M. Simon said...

I don't understand your question.

The object of the device is not current production. It is fusion.

Anonymous said...

As I understand it, the Bussard reactor produces a potential (and current) as a direct product of fusion. I realize the models being described are just test versions, but once break-even fusion is achieved, where do the spare electrons come from?

Or maybe that's the obvious point I'm missing? The Bussard reactor is different from the Polywell models?

M. Simon said...


The spare electrons are injected by an electron gun.

The Bussard reactor = Polywell.

Anonymous said...

So how does one get power out of a Bussard reactor? I'm still confused.

M. Simon said...

The holy grail is to collect the energetic alphas in a 2 million volt (or so) grid.

Direct conversion so to speak.

The alternative is to figure some way to get the alphas to boil water.

Neutrons are easier. However, neutron fluxes are not a good idea for a number of reasons. One is that if you are using superconducting magnet coils they cause loss of superconductivity.

This idea is a long way from practical application. It needs lots of work.

Anonymous said...

Ah, I understand, thanks. I misunderstood the direct conversion mechanism. I wish you the best in this work, it's very exciting.

bbot said...

Neutron flux causes loss of superconductivity because they transmutate the superconductor's atoms, mostly by smashing them apart. Neutron radiation is tremendously destructive, which is why nice, neutron free reactions are so helpful. In the case of p-B11, you also get some high-energy alpha particles of a precise energy, which can be decelerated gracefully and used to produce power. This effect has been used in some radioisotope batteries, and is a fairly mature technology.

Anonymous said...

I don't think your description of the magnets is right. The polyhedral geometry of the WB6 is a truncated cube, not a regular cube. It's true that the coils don't quite fit the shape of the polyhedron, but if they did, they would be diamonds (i.e. squares with corners at top, bottom, left, and right), not squares like the face of a cube. You need someplace for the field lines that go in through the centers of the coils to come back out; that's the corners. The circular coils are a fair approximation of the correct truncated-cube design.

M. Simon said...

Anon. 30 March,

The field lines are going to close no matter what.

The question is what will the field gradient be. Which will determine what the density of ion current will be. Which will in turn determine how the ion current generated magnetic field will interact with the coil generated magnetic fields.

It may be that the truncated polyhedron is optimum.

I might add that there is also the question of field lines intersecting the adjacent magnet structure. Which is why the spacing Bussard talks about is important.

All the interactions get complicated which is why what is going on is so hard to visualise.

Anonymous said...

Related to the comment above:

You cannot have a magnetic field that is "all North pole" in all directions: that would violate Maxwell's equations. If the magnetic field lines are envisioned as "coming out" of the faces, then they are "going back" into the corners. So I find the term "quasi-spherical magnetic field" to be rather confusing.

With respect to collecting the power: During steady-state operation, nuclei should be dropping into the potential well, and the fusion products should come boiling out. The outward pressure from these exiting alpha particles will combat the in-falling motion of the p and B-11, as well as the electrons. (It's not mechanical pressure: the electromagnetic effect of fast-passing nuclei should be like being hit by photons.) This needs to be taken into account: without the outward flow of the alpha particles, there's no way to extract the energy of the process.

M. Simon said...


I understand English is not your first language. However, if you see something that seems incredible re-read it to make sure you are not jumping to conclusions.

The magnets can be all North poles facing in or all South poles facing in.

This is true without recourse to magnetic monopoles.

As to your second point. There is no computer today fast enough to model all this in a reasonable amount of time. So build one and take measurements.

Anonymous said...

M. Simon,

Even if you use 6 magnets oriented to "point" outward, the field will not be anything like spherical. However much magnetic flux comes out of the faces will have to snake back in at the corners: This is the implication of:
div B = 0
in Maxwell's equations.

For that reason, using the term "quasi-spherical", as I think even Bussard does in his talk to Google, is profoundly wrong. There will be a symmetry, but that corresponding to a cube, not at all to a sphere. A sphere can be turned in any direction, and to any degree, without a change. Whereas everytime you rotate this cube by 90 degrees, the field at the upper face will convert from outward to inward for as the corner passes, and then back again. Not even roughly spherical.

M. Simon said...


You are correct except that the rotation of electrons in the magnetic field produces a magnetic field.

This magneticfield "circularizes" the magnetic field from the coils.

Plasma physics is not easy because everything affects everything.

Anonymous said...

M. Simon,

That always has the effect of countering the local magnetic in which the electron is orbiting: The current loop formed by the electron's orbit produces a field (within the circular region bounded by the orbit) which is opposite to the field which is forcing the electron to move that way. Makes sense: If it worked the other way, the magnetic field would be getting stronger and stronger, which has a perpetual-motion-machine feel to it.

As a local effect, it weakens the field wherever it finds it. Therefore: Since the basic symmetry is cubical and not spherical, that will not be changed by this effect.

Further, both the fields produced by the device and by the individual electrons satisfy
div B = 0 ,
so the sum of the fields do as well, because this equation is linear.

If you imagine the magnetic field as a flow of liquid, it would mean that overall the flow keeps continuing: If it's coming out of a region, it has to go back into that region. A spherically symmetric flow that satisfies that does not exist; the best you can get is an axially symmetric flow, with the B-field confined to the plane transverse to the axis; and along the axial direction, you could confine it to a region. But the field will still be confined to the transverse plane in that configuration.

Yes, I took a class in plasma physics decades ago. An old buddy of mine from grad school is a professor of plasma physics in East Lansing; although his interest is in atmospheric physics rather than fusion.

M. Simon said...


The effect is a diamagnetic one. The field lines get compressed.

True it is not spherical. It is described as quasi-spherical.

Look at Indrek's latest simulation at:

IEC Fusion Newsgroup

You can see electron circulation and also what looks like well formation.

Anonymous said...

The animation I find clearer than the static illustration, but it shows up another curious point: To form the well, the electrons have to stick around - at least mostly.

In the animation, they come in from all sides, but they don't get out very far. If an electron heads straight out one face, it shouldn't get any fight from the magnetic fields; and since a static magnetic field does not affect kinetic energy, that electron should be heading out with all the energy it started with, which was quite a lot. So why do the electrons in the animation stop and turn around? Just as the well is attracting the positive nuclei, it will be actively pushing away the electrons.

Now, maybe there is an ongoing balance between the incoming electrons from the beam and the escaping electrons; but this is not reflected in the animation.

Is there any explanation for this?

rubycodez said...

some misconceptions here: the "quasi-spherical" phrase Bussard used refers to electron cloud. Magnetic mirroring (tendency of charged particles with spiraling along magnetic lines to "bounce back" from regions where field strength higher) is used to confine electrons. Particles with angle of velocity to magnetic field greater than a critical value will be reflected.