Proving FRC fusion stability at scale, with Senior Scientist Roelof Groenewald

For decades, Field-Reversed Configurations (FRCs) have been one of the most attractive opportunities in fusion because of the large power output, but one of the biggest doubts was whether they scale to power plant conditions. Conventional wisdom says the higher the temperature, the less stable plasma would become.

But now according to TAE’s new paper, that roadblock has been cleared.

In this episode, TAE computational physicist Roelof Groenewald shares how he and his team overturned this long-standing assumption. Largely conducted using an award of computer time at the National Energy Research Scientific Computing Center through the ALCC program, their research shows that, somewhat paradoxically, FRC plasmas actually become more stable as they heat up. Think of a spinning top or bicycle or basketball spinning on your finger that becomes more stable the faster it spins. This breakthrough, along with TAE’s other enabling technologies, has enormous implications for the future of clean energy.

Why This Matters 

• Lower costs, higher efficiency: Unlike other fusion approaches that require massive superconducting magnets, FRCs generate much of their own magnetic field, making them more compact and economical. 
• Proven stability at scale: Real-world experiments and advanced simulations now confirm FRCs can operate stably at commercial size. 
• Accelerating timelines: With stability confirmed, TAE can confidently design its first prototype power plant—Da Vinci—aimed at delivering grid-ready fusion power. 
• Smarter R&D: Using some of the world’s fastest supercomputers, TAE can now simulate and optimize future machines virtually, cutting development costs and speeding progress. 

The next challenge is building the next generation of high-energy particle beams that can sustain fusion continuously, just like the sun. If successful, TAE will have achieved what many thought impossible: a practical, scalable path to fusion power with FRC topology.

JIM MCNIEL: For years, physicists, plasma physicists have been under the impression that the hotter that a plasma gets in a field-reversed configuration reactor, the less stable it becomes. That this was due to a stacking or a building up of fast ions. I mean, this is above some people’s pay grade in terms of understanding it, but the general outcome was that an FRC could not become big enough to become a viable commercial power generator. 

I’m joined today by Roelof Groenewald, who is a computational physicist at TAE and is responsible for being the co-author of a paper that basically turns this whole theory upside down, proves that the simulations were incorrect and in practice, the science, the actual physical operation of the machine was what you needed to follow.

 Roelof, welcome to Good Clean Energy.

ROELOF GROENEWALD: Thank you so much, Jim.

MCNIEL: Perhaps we should start with a basic background of what an FRC is. Can you explain what that is and what role plasma plays inside of it?

GROENEWALD: Absolutely. So the analogy that is commonly used at TAE is to think of an FRC plasma as a spinning smoke ring, if you’ve seen those that people blow. I think for listeners that are maybe more in tune with different fusion approaches that people generally use. A good analogy is to start thinking from the donut kind of tokamak picture where you have fields that go the long way around the donut. The FRC is kind of the other extreme of that when instead of the lines going all the long way around the donut, the magnetic fields go the short way around the donut through the hole in the middle of the donut.

MCNIEL:Yeah, I kind of think of it as a tokamak is like a hula hoop. The plasma spins like a hula hoop around your hips, and it never is exactly parallel to the ground. And an FRC is like a football spinning through air, but it’s held stationary inside of the linear machine.

GROENEWALD: Yeah. That’s a nice way to think of it. 

MCNIEL: And what we’re talking about is the super hot plasma. It’s the fourth state of matter, right? There’s solids, liquids, gases, and then there’s plasma.

GROENEWALD: Exactly. Where the electrons are stripped off the atoms and the whole thing is a soup of charged particles.

MCNIEL: And we need the plasma because what happens in a plasma.

GROENEWALD: Well, what we really need is for extremely hot particles to collide with each other so that they will fuse. And the only way that we can get enough energy at the point where there’s enough energy in the particles for that to happen, they can’t hold onto their electrons anymore. The benefit is that once you have these charged particles, they tend to stick to magnetic field lines so you can confine these particles in a region of space if you’re setting up magnetic field lines to kind of hold them in a bit. So it kind of is a benefit that nature gives us where in the state where particles will fuse, they are also charged, and that allows us to confine them with a magnetic field.

MCNIEL: Okay. Yeah, so let me unpack that for just a second. So we’re talking about super hot temperatures. So to get to boron fusion, we’re talking about 1.5 billion degrees C so that’s obviously really hot. 

GROENEWALD: Significantly hotter than the sun. 

MCNIEL: Yeah, like a thousand times hotter than the sun’s core. And to keep that in control, there’s no material in the universe that can suffer that kind of heat so we hold it in a magnetic cage. 

GROENEWALD: Exactly. 

MCNIEL: So what was the prevailing logic or wisdom or theory about FRCs and why FRCs were not a viable solution for fusion?

GROENEWALD: So the special thing about FRCs compared to most other magnetic confinement approaches is that in an FRC, the magnetic field that we use to confine the particles are generated by the plasma itself. That means the current that creates the magnetic field is the plasma current itself. And specifically in an FRC, what that results is, as you mentioned before, we have this linear machine that has external magnetic coils, and inside that we set up this kind of football-shaped plasma that has a field-reversed region. That’s the field-reversed part of the FRC. That aspect you can think of, as I’m sure all your listeners have played with magnets where if you hold one magnet and you hold another magnet up with the opposite polarity or I guess the same polarity, the magnets will tend to want to spin around.

MCNIEL: They repel each other. 

GROENEWALD: Yeah and they want to flip around so north and south poles are aligned. If you push two north poles together, they want to flip around. And that flipping motion, that tilt. It’s the same thing that occurs in FRCs because we have this magnetic dipole, as we call it, that’s oppositely aligned to the external field. It wants to flip around. It wants to tilt.

MCNIEL: So the field we’re talking about is moving from one end of the football to the other, and then it reverses. What causes that reversal of the field?

GROENEWALD: That is the plasma current that we were describing before, so that the current flowing inside the plasma itself creates such a strong magnetic field that it can overcome the external field and set up this region of reversed field.

MCNIEL: Right. And so the distinct difference between an FRC and say a tokamak is the tokamak, they’re building these super powerful magnets that are like 20 Tesla  magnets, and they need that because their plasma doesn’t create its own magnetic field. 

GROENEWALD: Yes, exactly. So what plasma physicists think of this, we call it plasma beta. 

MCNIEL: So beta is the ratio of pressure of the plasma against the external pressure of the magnets. It’s high in an FRC because the FRC creates its own magnetic field and it’s low in a tokamak because the tokamak doesn’t. Is that true?

GROENEWALD: That is exactly right. And that’s what makes FRCs special and that’s what makes FRCs conceivably able to run with advanced fuels like proton and boron fuels as you mentioned. The p-B11 fuel or Deuterium-Helium3.

MCNIEL: And beta’s important because the power output is beta squared, right?

GROENEWALD: Exactly. Yeah. You can kind of think of beta as almost a measure of the economical efficiency of a reactor.

MCNIEL: Okay, so with a higher beta, you’re going to get a higher Q. You’re going to get greater energy out than in.

GROENEWALD: At a fixed magnetic field, yes. So if you’re comparing the same magnetic field.

MCNIEL: Okay, so well then why wouldn’t everybody just go build an FRC? Because you don’t have to pay for super expensive, high temperature superconducting magnets or make them two stories tall. It makes a lot more sense. Why hasn’t the entire industry chased an FRC?

GROENEWALD: Well, that’s the critical thing, right? To have fusion happen in the first place, you need to have a plasma that’s stable, that’s going to stick around. And lots of people have done various analysis on different machines and how stable they are. And in particular, for FRCs, there was analysis done in the Seventies with sort of very simple models suggesting that FRCs would be very unstable, specifically due to this tilt motion that we described before.

MCNIEL: What was causing that tilt motion? It was the buildup of fast ions?

GROENEWALD: In that specific case, it’s simply this misaligned magnetic dipoles as we were describing with the magnets that you hold north pole to north pole, and they want to flip around. Those FRCs didn’t have fast ions in them, and they were purely thermal FRCs.

MCNIEL: Okay. Well, anybody who’s ever held two magnets close together knows it’s almost impossible to hold them steady, negative to negative. So it would kind of make sense that the plasma would tilt, wouldn’t it?

GROENEWALD: Yes. From that sort of picture, it would. The benefit is that the reality is a plasma isn’t this rigid object like a magnet that we’re holding. The plasma is  a collection of many particles and those particles are moving around. 

The analysis that was done at that point, that is this sort of basic magneto hydrodynamic analysis treats the plasma as a kind of rigid body. And in that analysis, the FRC is very unstable. But then when you extend the theory to include these effects, we take into account that particles are kind of individually moving around and the plasma isn’t a rigid body, and then the situation isn’t as bad as Rosenbluth found. Instead, the results show that there’s a strong dependence on the radius of the plasma. That at small radii, the FRCs would be stable, but as the radius gets very big, it would become unstable again. And there was this kind of trend towards Rosenbluth’s results as the radius increased.

MCNIEL: So most physicists would look at that and say, well, if I want to build a gigawatt power plant, that’s not going to work because I’m going to have to be at a certain scale and the plasma is no longer going to be stable based on Rosenbluth, who is like the father of plasma physics.

GROENEWALD: Yeah, absolutely. Rosenbluth was one of the fathers of plasma physics pushing for fusion research. Analysis done in the 90s showed that if you want to have ignition in an FRC, your radius would need to be one to two meters. 

MCNIEL: So it’s out of the question you can get there with an FRC. that’s just never going to work according to Rosenbluth, right?

GROENEWALD: Exactly.

MCNIEL: Okay, so how did you prove this to be wrong?

GROENEWALD: So also in the 90s, it was found from simplified models that if you had a large number of particles that were very energetic, whose orbits span the entire size of the FRC, and that would create a stabilizing effect against this tilt — this you can almost think of as a top that has a heavy top. If you spin that, it can sort of balance on a point. It’s a similar effect to that, where as long as you have enough particles that move through the entire FRC, they would stabilize it. We call such particles “fast ions.” Fast electrons will also work, but fast ions are easier to do and that is what TAE does. And so this effect was theorized to stabilize the FRC against tilt. But you have to have enough of these fast ions,

MCNIEL: Okay. Let’s unpack a little bit of this so people understand. One of the things you can help us understand is that there’s a direct correlation between fast and hot. You want to talk about that?

GROENEWALD: Absolutely. So when we talk about hot, if we talk about temperature of something, we are really talking about the average kinetic energy that particles would have of that temperature. And of course the more kinetic energy a particle has, the faster it moves. So we think fast and think hot.

MCNIEL: So temperature and energy are kind of interchangeable here.

GROENEWALD: Yes.

MCNIEL: In simple terms. So we’re talking about particles that are not moving at walking speed, they’re moving at thousands of miles an hour.

GROENEWALD: Absolutely. 

MCNIEL: Or tens of thousands of meters a second or some insane number like that.

GROENEWALD: Tens of thousands of kilometers per second.

MCNIEL: Yeah, so basically a percentage, a measurable percentage of the speed of light. We’re talking about fast. If you’re going to get to a billion five degrees Celsius, these things are moving really quick and they’re being held into like a soap bubble. You know, it’s this invisible cage of magnetic fields, and so they’re rotating around and then reversing direction, firing back down through the middle, and going around and round and around.

So your paper basically challenged the Rosenbluth finding and said, here’s why it’s not correct. So what did you find to prove that you’re right because you’ve done it in real machines and then you also got the simulation to match up with reality. How did that come about?

GROENEWALD: Yes. So the theory and the expectation was there that fast ions, these energetic particles would stabilize tilt. But as I mentioned the critical thing is that there has to be enough of them. At the same time, there’s another theory that showed if you have a lot of these fast particles, they could drive modes in the system themselves. So plasmas are very lively beasts and there’s many different types of excitations or many types of collective modes that come up in these systems. And so the kind of issue that computational physicists specifically had was we were kind of between this rock and a hard place where on one hand we have to have enough fast ions in order to stabilize that tilt mode we talked about. But then simultaneously, this other aspect of the system said that if you have a lot of these particles, they will drive modes that we call energetic particle modes or fast ion driven modes. 

And up until this, this recent work that we did, attempts to simulate the stabilization of tilt by having enough fast ions always ran into this problem where those systems would be very unstable to beam-driven modes. So it was never possible to show the stabilization because while you don’t get the FRC disrupting due to tilt, you do get it disrupting due to a beam-driven mode.

MCNIEL: And the beams, they’re basically the fuel injectors of the fusion device. They’re introducing really fast particles. The particles are neutral because they have to pass through a magnetic wall, and then when they encounter the particles inside of the magnetic cage, they assume the charge of that particle that they encounter. They become charged. And so if you think about a top, I like to think about a basketball on your finger because a top, you pull a string and it spins and it’s stable, but as it loses energy, it starts to wobble and it falls down. What we have to have is a top that’s constantly being spun faster, like a basketball on your finger.

And in this case, we have the finger holding it up, which is kind of the center of the magnetic field and the fingers on the outside are beams. Eight beams that are spinning the ball faster and faster and faster and faster. And the faster it spins, the hotter it gets, and you proved the more stable it becomes.

GROENEWALD: Yes. We showed that you could have a large radius and still have a stable evolution, which at this point computational theory is very much a following experiment because TAEhas demonstrated this 10 years ago. 

MCNIEL: But the math didn’t match up with reality.

GROENEWALD: Yes, exactly. The sort of specific way that we thought about the fast ions was the critical bit that we figured out in this paper. So I talked about these beam-driven modes that you can get if you have lots of these energetic particles, they will drive modes for themselves. But the kind of theory that said there’s a limit on how many of these you can have, that part essentially was incomplete. That it doesn’t, it’s not really a limit on how many, it’s a limit on what type. Or exactly how they behave. 

And we showed that if we capture in our modeling self consistently how the FRC, this magnetic configuration, responds to kind of increasing numbers of these particles being added to the system, it will relax in such a way that it pushes out the fast ions, the energetic ones that would drive the mode but keeps the ones that don’t. So we could find that you can simultaneously have large amounts of this fast ion current that gives you stability against tilt while simultaneously not driving the system to be unstable to these beam-driven modes.

MCNIEL: So originally the thought was that the fast ions would create a wave that  would destabilize the plasma. Is that correct?

GROENEWALD: That’s exactly correct.

MCNIEL: But you discovered that there are these little micro-perturbances or discrepancies that take place and kind of breaks up that wave.

GROENEWALD: That is right. Yes. Here, theory is kind of following experiment, because experimentally these, we call them microbursts, have been observed going back to [TAE’s fourth-generation machine,] C-2U, almost like 10 years ago.

MCNIEL: So I guess I think about this from a layman’s standpoint: If you have a really big earthquake near a body of water, you create a huge wave or a tsunami. But in this case, we have these little micro-perturbances which are maybe like little micro counter-waves that break up the huge wave into smaller waves. And so you don’t necessarily flood the town or tilt the plasma. Is that reasonable? 

GROENEWALD: Yeah, exactly. That’s a good way to think about it.

MCNIEL: Okay, so the paper demonstrated that we can have stable plasma at scale, which we need for a power plant. But how did you correct the simulation? You guys have been working with a language called WarpX.  Can you talk about that for a minute?

GROENEWALD: Yes, the code we’re using is WarpX, which is this excellent code developed by a group at Lawrence Berkeley National Lab primarily. It’s an open-source code. It’s a community code, so people are contributing from all over the world. But the code was originally built for this exascale project

MCNIEL: For the Frontier computer.

GROENEWALD: For all of the exascale computers that the DOE is building, of which Frontier was the first one. So this code allows us to run simulations a lot faster than people were able to before. And this kind of puts us in a position that we could look at what we call transport timescales in the system. So if you think about that reactor, there’s a fusion reactor, there’s lots of different processes going on, and those processes happen on kind of individual scales, depending on the physics that drives them. So historically people were able to only look at short time scales, things on the tens of microsecond dynamics in the system. But now because we’re using this code, that is so much faster than what people had access to before. We can run simulations past the millisecond of simulation time, a hundred times longer than we would before. 

MCNIEL: So the outcome of this is with just a neutral-beam-driven plasma. We can prove stable plasma and we can prove it gets more stable the hotter and more energetic it becomes. And you were talking about density or number of particles. That’s really important because what we really want is a plasma full of atoms that are going to get hot enough and fast enough to merge. 

GROENEWALD: Yes, exactly. This is, as you mentioned, scaling. What we’re interested in is being able to run simulations. Now that we have the ability to reproduce what we’re seeing experimentally on the current machine and verifying it by comparing sort of specific observables from the simulation to the experiments such as those microbiomes we described. That gives us confidence that the simulations are reproducing what the experiment is doing. And then allows us to confidently say, let’s simulate what a hypothetical next machine would look like, and then what the machine after that would look like. And we can see how does the stability and the temperature, the confinement properties change as we kind of successfully step towards bigger, bigger reactors that eventually gets to the kind of reactor that would form a fusion power plant. 

MCNIEL: Right. So now that you have more confidence in the accuracy of your simulation, which has been proven by real experiments, you could start to trust it to try out new things. Like for instance, the current Norm configuration is running 8 neutral beams. We don’t know if the first commercial plant is gonna need 12 beams or 4 beams, or 24 beams, right?

GROENEWALD: Exactly, and we don’t know what the optimal energy for those beams would be. So Norm injects fast ions at 15 kiloelectron volts and some at 30 kiloelectron volts. Maybe for a reactor we would want to have much higher energy. 

These are the kinds of questions that we can explore with the simulation. Now, do we want more current rather than more energy? Do we want a different angle on the injection of these beams, things of that sort.

MCNIEL: Right, so we can run those experiments in the simulation and test them and get the answers that we need before we start cutting metal.

GROENEWALD: Exactly. It’s a lot cheaper for us to run a simulation than it is for the engineers to design and build a reactor and commission it and test it.

MCNIEL: Well, how long does it take to do this? So let’s say you start with 4 beams and then you want to do 6, and then you want to do 8. Is this like a day at a time, a week at a time, a month at a time? What does it take to iterate on those models?

GROENEWALD: Currently it’s a couple days, so running a single simulation would take five days if we want to run it for these very long scales, right? So up to a millisecond or so on. But this is where we really benefit from these giant computers that the DOE has built and that they kindly make available through grant proposals to the public where we can run many of these simulations at the same time. So as we talked about Frontier earlier, it has thousands of computers of nodes. If we only need 4 at a time, then maybe we can run on a hundred and run 25 simulations at the same time. And that’s kind of the power of the iteration.

MCNIEL: It’s amazing and we have access to this computational power at Oak Ridge and Frontier and Lawrence…

GROENEWALD: Lawrence Berkeley, the NERSC facility, the National Energy Research and Scientific center we have time at, so we run on ProMat supercomputer, and we have time on the Aurora supercomputer that just recently,

MCNIEL: That’s amazing. These computers, I mean, they’re simulating our solar system. Or they’re simulating I think actually our Milky Way galaxy.

GROENEWALD: They do amazing things.

MCNIEL: They’re doing insane things. So what kind of feedback have you had on the paper so far from your community?

GROENEWALD: Of course, people are excited about this. I’ve had some people reach out that have worked on these topics 10 years ago and have kind of encountered these beam-driven modes. They’ve been very excited to see the next step in this understanding forming. Internally at TAE we’re fairly excited about just being past this contradiction that we’ve had to deal with. 

MCNIEL: Yeah, we had this obstacle, this opposition that now you’ve been able to resolve. And so there should be no longer any debate about how big an FRC can get and how energetic it can become. 

GROENEWALD: Exactly.

MCNIEL: Yeah. Well, that leads us to commercialization. So I guess you and the rest of the computational team are going to help the engineers figure out the best configuration for Da Vinci, which is going to be TAE’s first commercial power plant or prototype. It’s going to be producing electrons. Anything else about this paper that we should share with our listeners that you’re excited about?

GROENEWALD: I think the follow-up is what’s exciting for me most now is once we’ve been able to kind of unblock this discrepancy that was holding us up and we understand now, we’ve kind of been able to go past and really do an in-depth model validation study, which is where things get very exciting.

Because then the theoretical people, the computational group and the experimental group have to come together and look at their individual data coming from simulations and experiments and say, do we really see the same things going on? 

MCNIEL: So it seems to me now that TAE has achieved this great breakthrough where we’ve removed the formation sections of the machine, so we’re no longer creating smoke rings of plasma and colliding them together. We’re just taking a seed plasma and heating it up with neutral beams.

We’ve cut the hardware in half. We’ve reduced complexity. It’s a much simpler device. So from what I understand, the real tall order to get to commercial fusion is we have to increase beam power. And you said we’re at 30 or 40,000 thousand volts of energy. But we know that we’re probably going to have to get up as maybe high as 800, 850,000 volts. Is that the main objective now is to get to higher performance particle beams?

GROENEWALD: It is certainly a large part of the effort. These beams they’re kind of a scientific problem still. No one in the world has been able to build a kind of very high energy beams that operate for a very long time. Because we want to approach steady state, right? So we also want beams that aren’t charged and then fire for a few milliseconds. We want them to be operating the whole time. So yes, that’s a very big push on the TAE side. 

MCNIEL: We’re talking about a steady-state FRC, which kind of more mimics the way the sun operates, right? The sun’s not a pulse reactor. It’s a steady state reactor. We want to mimic the sun. We want to put a star in a jar. So the beams have to be operating continuously at these high-energy levels to continue to put energy or fuel into the plasma, so the plasma can stay hot enough to let fusion happen.

GROENEWALD: Fortunately, TAE has a world-class beam team as we call them that’s working on these next generation neutral beams. 

MCNIEL: Roelof, thank you very much for the work you’ve done and congratulations on a very exciting paper. Thank you for helping us to understand it a little bit better.