Wibbly-wobbly magnetic fusion stuff: The return of the stellarator

ARS Technica

Artistically shaped magnets may make stellarators easier to manage than ITER.

Design of the W7-X from the outside. Having many ports means the scientists can make a lot of different measurements at the same time.

The arrangement of magnets required to create a well-confined plasma.

The shape of the magnetic field.

Fusion powers the Sun, where hydrogen ions are forced together by the high pressure and temperature. The nuclei join to create helium and release a lot of energy in the process. Doing the same thing on Earth means creating the same conditions that drive hydrogen nuclei together, which is easier said than done. Humans are very clever, but achieving fusion in a magnetic bottle will probably be one of our cleverer tricks. Making that bottle is difficult, and Ars recently had the chance to visit the people and facilities behind one of our most significant attempts at it.

For most people, magnetic bottles for fusion bring to mind the tokamak, a donut-shaped device that confines the plasma in a ring. But actually, the tokamak is just one approach; there’s a more complicated version that is helical in shape. Somewhere in between the two is the stellarator. Here, the required magnetic field is a bit easier to create than for a helix, but it’s still far more complicated than for a tokamak.

At the Max Planck Institute for Plasma Physics (MPIPP) in Greifswald, located on the Baltic coast in Germany, the latest iteration of the stellarator design is preparing to restart after its first trial run. The researchers putting it all together are pretty excited by the prospect—frankly every engineer and scientist would be excited by the prospect of turning on a new piece of hardware. But it’s even more so the case at MPIPP since the new gear happens to be something they designed and built. The stellarator is something special: the realization of a design that is more than 50 years in the making.

Self-organized confinement

The heliac, the stellarator, and the tokamak are all trying to achieve the same thing: confine a plasma tightly in a magnetic bottle, tightly enough to push protons in close to each other. They all use a more-or-less donut shape, but that more-or-less involves some really important differences. That difference makes the stellarator a pretty special science and engineering challenge. To highlight that challenge, we can start with the simpler and more familiar tokamak.

The tokamak begins with a donut-shaped vacuum vessel. The magnetic field is applied by a series of flat coils that are wrapped around the tube of the donut (as in the diagram). This, along with a few other magnets, creates a magnetic field that runs in parallel lines around the interior of the donut. When a plasma is injected, its charged particles corkscrew around the field lines. At first sight this looks like it should confine the plasma in a series of tubes.

This doesn’t happen, though. As Professor Thomas Klinger, head of the stellarator project at the MPIPP says, “The vacuum magnetic field has no confinement properties because it’s a purely toroidal field. And a purely toroid field does not confine a plasma at all; that was already realised by Fermi in 1951.”

The problem is that the charged particles can drift from magnetic field line to magnetic field line. Since the magnetic field doesn’t have the same strength across the cross-section of the torus, particle drift to the outside is much more energetically favorable. So the plasma simply expands outward and hits the wall.

To obtain high plasma temperatures in a tokamak, this drift has to be stopped. To do this, a large current has to flow through the plasma. “You have to twist the magnetic field lines, which is done by the current,” says Klinger. The current generates a second magnetic field, which distorts the applied field so that the field lines run in a twisted spiral.

A charged particle in the very short-term can still be thought of as corkscrewing around a single field line. But, because the field line spirals around, it is better to think of a series of nested surfaces (like a matryoshka doll), with the particles in the plasma confined on these surfaces. One consequence of this design is that, while particles still hop between field lines, they can now drift from low magnetic field to high magnetic field, and vice versa—an outward flow is no longer favorable. So, on average, the rate at which particles escape confinement is much smaller.

Strong confinement means that the plasma has to support a large current to generate the right magnetic field shape. For the international thermonuclear experimental reactor (ITER), the plasma will generate several million amps of current. Unfortunately, the current through the plasma, the plasma density, and temperature don’t end up the same everywhere, and these differences have the potential to destabilize the current.

In particular, if the current is not evenly distributed across the plasma, the lovely nested surfaces that confine the plasma may be destroyed. This process can rapidly spiral out of control, dumping all the current in the plasma to the vessel walls in an event called a disruption. A disruption is not something to be taken lightly, as Klinger notes. “A grown-up tokamak like JET [joint European tokamak] or our ASDEX upgrade [axially symmetric diverter experiment] starts to jump in the case of a disruption,” he says. “These are big machines; imagine such a big machine starts jumping.”

So while the tokamak can use a self-organizing magnetic field to confine the plasma, that field is subject to various instabilities. To avoid these building into problems, the tokamak has to operate in pulsed mode (though those pulses may be hours in duration), and it requires a lot of sensors, control systems, and feedback to minimize the instabilities.

To get this right, you need a good physical model of the plasma physics. Researchers use the model to look for the telltale signs that indicate the beginning of an instability. “My modeling is mostly related to how do we control these instabilities. How do we affect these instabilities so that they either do not occur or that, when they occur, we suppress them or ameliorate their presence,” says Dr. Egbert Westerhof from the Dutch Institute for Fundamental Energy Research (DIFFER).

In the tokamak, this sort of modeling is simplified by the symmetry of the device, which reduces a 3D problem to 2D. The results from these physics-based models are then used to create empirical models that do not really contain detailed physics, but they can quickly provide predictive results within some limited range of plasma properties.

This simplicity has helped produce models that can calculate the tokamak’s behavior faster than the tokamak can misbehave, a necessity for a successful control system. This hasn’t really happened with the stellarator designs. “They are really far [ahead of us] in tokamaks because they have these models that work really well. They have been tested. And now they can actually predict the temperature and density profiles faster than real time, which is incredible. But we don’t have these models yet,” explains Dr. Josefine Proll, an assistant professor at Technical University Eindhoven.

One of the artistically shaped magnets required to generate the stellarator's magnetic field.

The crazy shape of the magnet is clearly visible.

The plasma vessel has a shape that mimics that of the magnetic field.

The W7-X under construction. The torus is about 16m in diameter.

The W7-X is open at the moment and covered with scaffolding for access.

For comparison, this is a drawing of an ITER magnet in one of the lifting tools.

Compare the simplicity of the arrangement of magnets in ITER with that of the W7-X.

The ITER torus.

The ITER torus.

The ITER torus.

A training section of the W7-X plasma vessel.

Inside the training mock-up. You have to learn where you can and can't place your feet.

My guide and I puzzled over this bit. We think it is an electrode.

Training to move around the W7-X's vessel.

The vacuum chamber is big. These, I think, are two of the pipes that lead to high speed vacuum pumps. But, to obtain really good vacuum, you have to cover everything with foil and heat the entire vessel to more than 100 degrees Celsius.

This is an access port for instrumentation. Laser positioning is used to ensure everything is mounted correctly.

The W7-X is full of ports, most of them packed with instrumentation.

When you work with charged particles, everything is done with coils. This coil tunes a high-power source of microwaves

Each microwave source produces 800kW of power at about 30-percent efficiency. All that excess heat has to be dumped somewhere.

About 10MW of microwave power from 10 different sources are combined and enter the stellarator through this box.

Angela Merkel pushed the button last year. Every big facility needs a big red button.


Externally organized confinement

The stellarator has little to no current in the plasma. This is because the externally applied magnetic field has all the properties required to confine the plasma. So, although the vacuum vessel is still basically a toroid, the magnets that loop around the tube are not planar. Instead, they have the shape needed to generate a twisted magnetic field. “If you shape your field in a clever way then you can make it so that the drifts basically cancel out, at least for those that would leave the plasma,” says Proll.

Theoretically, that is. In practice, well, we’re still working on it. To give a magnetic field precisely the right shape requires extensive calculation at many different scales, and all of it must happen in a 3D space.

So, computer code that simulates the plasma over the entire volume of a stellarator had to be developed, and that had to wait for computers that were powerful enough to perform the calculations. “These machines, these supercomputers of the ’80s, made it possible to crank through the equations, to solve the equations simultaneously, and then it was found out, okay, the stellarator needs optimization,” says Klinger.

Calling it optimization kind of undersells the problem, though. Scientists had to decide what parameters of the system need to be optimized and in what range. To make that decision more difficult, no single computer model can encompass the vast range of physics that needed to be included. To get an accurate picture of the plasma in a stellarator, you need separate models that calculate the applied magnetic field and the plasma’s fluid-like behavior, called a magnetohydrodynamic model. Then, to test the magnetic field confinement against particle drift and particle collisions, you need models that track individual particles along field lines and other models that deal with diffusion. All of these models needed to be created and then verified against experimental data before optimization was even possible.

Listing image by Max Planck Institute for Plasma Physics