Stellarator vs Tokamak: Why Wendelstein 7-X Is Twisted

Both machines are trying to do the same thing: hold a gas at over 100 million degrees away from any wall, using magnetic fields, for long enough that fusion releases more energy than it took to get there.

They differ in how they produce one specific component of that field — and almost every other difference follows from it.

Why a simple magnetic ring does not work

The obvious approach is a doughnut-shaped chamber wrapped in coils, producing a magnetic field that runs the long way around the ring. Charged particles spiral along field lines, so they circulate rather than hitting the wall.

It fails, for a geometric reason. In a toroidal coil arrangement, the coils are closer together on the inside of the ring than the outside, so the field is stronger on the inboard side. That gradient makes positive and negative charges drift in opposite vertical directions. Charge separates, an electric field builds between the top and bottom of the plasma, and the resulting force pushes the entire plasma outward into the wall.

The fix is to give the field lines a twist, so that a line spirals around the tube as it goes around the ring. A particle following it spends part of its orbit near the top and part near the bottom, and the drift cancels out instead of accumulating. This twist is called rotational transform.

Every magnetic confinement design is, at bottom, an answer to one question: where does the twist come from?

The tokamak answer: drive a current through the plasma

A tokamak generates the twist by driving an enormous electric current — millions of amperes — through the plasma itself. That current produces its own magnetic field looping around the plasma column, and the combination twists the field lines.

This is elegant. The plasma is doing the work, so the external coils can be simple planar rings, cheap to design and manufacture. It is why the tokamak became the mainstream approach, and why ITER is one.

It also creates two hard problems.

Tokamaks are naturally pulsed. The plasma current is driven by a transformer: the central solenoid ramps its own current, inducing current in the plasma. A transformer cannot ramp forever, so the pulse eventually ends. A power station that stops every few minutes is awkward.

The current can go unstable. A plasma carrying megaamps stores a lot of energy in its own magnetic field. If confinement is lost abruptly — a disruption — that energy dumps into the vessel within milliseconds. At reactor scale, disruptions are a serious engineering threat, and much of tokamak research is devoted to avoiding, predicting, or surviving them.

The stellarator answer: build the twist into the coils

A stellarator produces the entire twist externally. No plasma current is required.

That removes both tokamak problems at once. Nothing needs to be induced, so the machine can in principle run continuously. And with no large plasma current, the disruption mechanism largely does not exist.

The cost is the coils. Producing the right field with external magnets alone demands shapes that are not circles, not planar, and not intuitive — every coil a different three-dimensional curve, and the plasma cross-section changing shape as it travels around the machine, from a bean to a triangle and back.

Early stellarators in the 1950s and 60s performed badly. Their fields, designed by hand, confined particles poorly, and the tokamak overtook them.

What changed: optimisation

The revival came from computation. Rather than guessing a coil shape and calculating the resulting confinement, you can specify the properties you want — good particle confinement, stability, acceptable heat loads — and let a numerical optimiser search the space of possible field geometries for a configuration that satisfies them.

Wendelstein 7-X, at the Max Planck Institute for Plasma Physics in Greifswald, is the result of that approach at scale. Fifty non-planar superconducting coils, each a distinct shape, each manufactured to a tolerance of well under a millimetre, because the optimisation only works if the built machine matches the computed one.

Construction took about nineteen years. The first plasma was produced in December 2015.

What it was built to prove

W7-X is explicitly not a power plant, and cannot be one: it uses hydrogen rather than a deuterium-tritium fuel mix, so it does not produce significant fusion energy. It was built to answer a narrower question — whether an optimised stellarator actually confines as well as the optimisation predicted.

The results have broadly supported that. The machine has demonstrated the reduced transport its design targeted, and later campaigns with an actively cooled divertor have pushed toward the long, high-power pulses that are the whole point of the concept.

The open question is no longer whether stellarators work. It is whether they can be built affordably. A tokamak’s coils are simple and repetitive; a stellarator’s are fifty bespoke objects. Several private ventures are now betting that better computation, better magnets, and modern manufacturing shift that balance — which is a rather different argument from the one that sidelined stellarators sixty years ago.