Fusion Helps Itself, Better Magnets, and a Leaky Salt

Fusion's own exhaust helps the plasma hold together better

The helium exhaust from a fusion reaction may be quietly fixing the plasma's biggest heat-loss problem — without any help from engineers.

Imagine you're trying to keep a bath hot. Tiny swirling eddies on the surface constantly pull heat out into the air. This is roughly what kills a fusion plasma: ion-scale turbulence — swirling currents about the size of a proton's orbit — bleeds heat out of the plasma faster than the reaction can replace it. Fixing this is one of the hardest problems in the field. A team using the GENE simulation code, applied to both the ITER and SPARC reference designs, found something unexpected: the fusion reaction itself may help fix this. When deuterium and tritium fuse, they release helium nuclei called alpha particles. These alpha particles are fast and energetic. They slightly excite oscillations in the plasma's magnetic structure — waves called toroidal Alfvén eigenmodes, or TAEs. Those waves then amplify large-scale circulation currents inside the plasma, called zonal flows. Those flows, in turn, act like a strong crosswind that breaks up the small turbulent eddies before they can carry heat away. The predicted payoff is significant: the simulations show SPARC getting 25% more usable heating power from its alpha particles when this feedback loop is included (uncertainty ±3%), and ITER getting 18% more (±5%). The plasma core also develops steeper temperature gradients — a sign of better heat confinement. The catch is real: this is entirely simulation, using GENE coupled to a transport solver called Tango. Neither SPARC nor ITER has yet burned deuterium-tritium fuel. The prediction is physically well-motivated, but plasmas regularly surprise us. File this one under 'very promising, needs experimental confirmation.' The first burning-plasma experiments at SPARC, expected later this decade, will be the first real test.

Engineers wind the first new-shape superconducting magnet for fusion-class coils

Wrapping a stiff, expensive tape into a precise spiral without snapping it sounds like a craft problem — but it is one of the real bottlenecks between us and working fusion magnets.

The magnets that make modern fusion reactors possible are wound from a material called ReBCO tape — a high-temperature superconductor that carries enormous electrical currents with almost no resistance, at temperatures cold enough to need a cryocooler but nowhere near as brutal as older superconductors. The problem is that ReBCO tape is stiff, brittle under the wrong kind of bending, and expensive. Every new magnet geometry has to be proved out physically before you trust it. A team at CERN built the first magnet ever wound in a geometry called canted-cosine-theta — think of it as a spiral staircase of tape, carefully angled so the cumulative current produces a very specific field shape — using this HTS ReBCO tape. The design is a sextupole, a magnet type that corrects focusing errors in particle beams, intended for the proposed FCC-ee collider. But the winding technique and the tape qualification are directly relevant to fusion: SPARC, Commonwealth Fusion's prototype reactor, uses the same ReBCO tape in its own high-field coils. The prototype — about the length of a large wine bottle — was cooled to 46.5 K (roughly −227°C) and ramped up to 300 A of current. That is 15% above its designed operating current, and it held steady without quenching, meaning without the superconductor suddenly losing its special properties and dumping all that energy as heat. Two tape manufacturers were qualified in the process. The catch: this is a single prototype, tested once, at one measurement plane along its length. Making dozens of identical coils that survive years of thermal cycling is a different, harder problem. But you have to prove the recipe once before you can bake at scale.

Our tritium leakage estimates were off because we ignored the sides of the container

Published measurements of how fast tritium leaks through a key reactor material disagree by a factor of up to 100 — and a team at MIT just found a big part of the reason why.

Tritium — the rare hydrogen variant that fusion reactors will consume by the kilogram — does not exist in meaningful quantities on Earth. Reactors will have to breed it themselves, typically inside a liquid blanket surrounding the plasma. FLiBe, a molten salt made of fluorine, lithium, and beryllium, is a leading candidate for that blanket. But tritium is small and slippery: it migrates through metal container walls the way moisture seeps through a poorly sealed window frame. To build a safe reactor, engineers need to know exactly how fast it moves. A team at MIT's Plasma Science and Fusion Center ran an experiment called HYPERION, pushing hydrogen isotopes through a FLiBe sample, and then used a finite-element code called FESTIM to model the results and infer FLiBe's permeability — how easily tritium passes through it. The uncomfortable finding: the standard approach of treating the problem as one-dimensional, imagining tritium flowing straight through like water in a pipe, is wrong. A significant fraction was leaking sideways through the nickel walls of the containment structure — pathways that one-dimensional models simply cannot see. When you account for that, your inferred permeability shifts substantially. The scatter in published FLiBe measurements — already spanning a factor of 10 to 100 — is partly explained by this: different labs built different containers and each measured a different combination of 'through the salt' and 'around the sides.' The catch: the MIT team bracketed the answer between two limiting assumptions about the container boundary; they have not pinned down a single definitive permeability number yet. But knowing why previous measurements disagree is already a meaningful result — you can't fix a measurement problem you haven't diagnosed.

These three papers share a quiet common theme: we are getting better at finding the mistakes in our own models. The alpha-particle turbulence result says burning plasmas may be kinder to themselves than we thought — a self-correcting feedback loop our simulations had been missing, and one that could meaningfully improve performance predictions for both SPARC and ITER. The HTS magnet result says we can physically build the coil geometries that next-generation reactors need, one new winding pattern at a time. And the FLiBe result says our tritium transport estimates have been systematically off because we were solving a three-dimensional problem with a one-dimensional tool. None of these is a rescue or a crisis. They are corrections. Some are encouraging — the plasma helps itself more than we knew. Some are sobering — our tritium models were wrong in ways that matter for reactor safety and fuel supply. That pattern of correction, honestly reported, is what real progress looks like.

The alpha-particle turbulence prediction will only be tested when a machine actually burns deuterium-tritium fuel at relevant power — Commonwealth Fusion's SPARC is the nearest candidate, with DT operations currently targeted for the late 2020s. On the tritium side, blanket qualification experiments are ramping up at facilities in Europe and Asia over the next two to three years; watch for new FLiBe and PbLi permeability measurements that will either confirm or complicate the MIT picture. The open question I'd most want answered: how much does the alpha-particle turbulence suppression hold up as plasma conditions vary during a real pulse?

• Quanta Magazine: Why Fusion Is Hard (turbulence explainer)
• Wikipedia: Tritium breeding in fusion reactors
• Wikipedia: High-temperature superconductivity