DeepScience — Nuclear Fusion

DeepScience · Nuclear Fusion · Daily Digest

Boron dust, argon gas, and magnets that crack themselves

Three concrete steps this week toward a fusion reactor that doesn't tear itself apart.

            June 15, 2026
          

Three papers landed this week that each attack a different version of the same problem: fusion plasmas are brutally violent, and the machines that hold them have to survive. I spent the morning working through the physics so you don't have to. Let me walk you through what each one found, what it costs, and what still has to happen.

Today's stories

              01 / 03
            

Sprinkling boron powder into a plasma makes it far calmer

What if the key to taming a 100-million-degree plasma was dusting it with a few milligrams of boron powder per second?

A team working on the DIII-D tokamak at General Atomics in San Diego tried exactly that. Their target was a recurring problem called ELMs — edge-localized modes. Think of ELMs like the violent splatter when water hits a red-hot pan: sudden, repetitive bursts of plasma that slam the reactor wall with a pulse of heat. In a commercial reactor, those bursts would gradually erode the wall. Taming them is one of fusion's most stubborn engineering challenges. The team used a device called an Impurity Power Dropper to feed boron powder into the plasma at different rates across five experimental runs. At a modest dose of 4.5 milligrams per second, ELM frequency dropped by 76 percent compared to a reference plasma with no injection. At nearly double that dose — 9.7 mg/s — the plasma went ELM-free entirely for stretches of about 300 milliseconds at a time. The reason this works is subtle. The boron doesn't just coat things — it selectively stirs up low-frequency turbulence at the plasma's outer edge, which bleeds off pressure gradually rather than letting it build up and snap. The team also observed something not previously reported: the two stability limits that normally converge to trigger an ELM actually pull apart under boron injection, opening a safe channel between them. The catch, honestly, is that 300 milliseconds is still short. When the ELM-free phase eventually broke down, the plasma lost about 15 percent of its stored energy in one go. And this was five discharges on a single experimental day — not a sustained campaign. The authors are at an early but real step: a mechanism has been identified, not yet a reliable control strategy.

Glossary

ELM (edge-localized mode) — A sudden burst of heat and particles from the edge of a tokamak plasma, like a pressure valve snapping open — potentially damaging to reactor walls.

H-mode — A high-confinement operating state for a tokamak, where a sharp pressure gradient builds at the plasma edge — the same gradient that makes ELMs possible.

pedestal — The steep pressure cliff at the edge of an H-mode plasma, like the raised lip of a bowl; it improves overall confinement but also stores the energy that ELMs release.

Source: Impurity-driven turbulence opens a pathway to ELM-free operation and enhanced pedestal stability in tokamaks

              02 / 03
            

A model of a new compact reactor hits near-gigawatt power while keeping walls cool

Can you push a fusion reactor to nearly a gigawatt of fusion power while keeping the exhaust surfaces cool enough to survive?

That is the question a team of modelers tackled for ARC, the compact high-field tokamak being developed by Commonwealth Fusion Systems (CFS), the MIT spin-off behind the SPARC project. ARC is designed to run at very high magnetic field in a small package — which is attractive, but it also means the plasma is running hot and dense, and the exhaust heat has to go somewhere. The team ran a large suite of computer simulations using several coupled physics codes — ASTRA for the core plasma, TGLF for turbulent transport, and an extended Lengyel model for the exhaust region — to ask: can ARC produce near-gigawatt fusion power while also keeping the divertor (the reactor's exhaust system, roughly analogous to a car's radiator) cool enough not to melt? The answer, with the right impurity seeding strategy, appears to be yes. Injecting small amounts of argon gas into the exhaust region — which radiates heat away before it concentrates on surfaces — maintained wall temperatures below 2 electron volts (a standard target for survivability) while allowing fusion power in the 750–1000 MW range. Neon injection also worked, but proved less stable: too much neon crept into the plasma core, cooling it and risking a collapse out of high-confinement mode. This is all still modeling. ARC has not been built. The codes used are well-established, but they rely on approximations — in particular, the pedestal model uses a neural network trained on other machines. The team is clear this is a sensitivity study, not a prediction. What it tells you is that the heat-exhaust problem does not, on paper, rule out high-performance operation. That is worth knowing.

Glossary

divertor — A specially shaped region at the bottom of a tokamak that channels exhaust heat and particles away from the main plasma-facing walls — like the drain in a sink.

detachment — A desirable operating condition where the plasma exhaust cools and loses contact with the wall before hitting it, dramatically reducing heat load on materials.

impurity seeding — Deliberately injecting a small amount of a heavy gas (like argon) into the plasma edge to radiate heat away as light rather than dumping it as kinetic energy onto wall surfaces.

H-mode — High-confinement mode — a state where the plasma is well-insulated from the wall, enabling higher pressure and more fusion reactions.

Source: Core-edge integrated modeling of ARC: on the effect of impurity transport and detachment conditions

              03 / 03
            

There is a hard limit on how strong fusion magnets can be, and engineers found it

At some point, making a fusion magnet stronger stops helping — the structure needed to hold it together gets so thick that it squeezes the reactor out of existence.

One of the central ideas in modern fusion design is that stronger magnetic fields let you build a smaller, cheaper reactor. Higher field means you can squeeze the plasma harder in a smaller volume. That logic drives CFS, Tokamak Energy, and several national programs to push toward very high magnetic fields using new superconducting tapes. But a team from the French fusion systems group working with the D0FUS tokamak design code has just quantified something engineers had sensed but not fully mapped: there is a ceiling. Their analysis focuses on the mechanical structure of the magnets themselves. A toroidal field coil — the hoop-shaped magnet that circles the plasma like a barrel stave — experiences enormous inward magnetic forces. At high fields, the steel structure holding the coil together has to be so thick that it eats into the space available for the plasma. Think of it like a bicycle wheel: the stronger the tension in the spokes, the thicker the hub has to be, until the hub takes up the whole wheel. The team found that with standard steel (316L stainless) in a conventional wedged coil arrangement, no viable power-plant-scale design (2 GW fusion power, Q=40) exists beyond a peak field of 20 Tesla. That is a concrete number. Three fixes help approximately equally: switching to a stronger steel alloy (called CHSN01), changing the coil architecture from wedging to bucking, or reducing the demands on the central solenoid. Combining all three opens up compact designs with a major radius below 4 metres. This was a purely analytical study — no hardware was built or tested. The models were validated against six existing machines, which builds confidence, but real magnets always surprise engineers in ways models miss.

Glossary

toroidal field coil — The large electromagnets that wrap around a tokamak like barrel staves, creating the main magnetic field that keeps the plasma from touching the walls.

central solenoid — A tall electromagnet running through the centre of a tokamak, used to drive electrical current through the plasma and heat it up — like a transformer's primary coil.

radial build — The total thickness of all layers — magnets, shielding, structure — stacked from the plasma outward; the thicker this is, the bigger (and costlier) the machine must be.

Q factor — The ratio of fusion energy produced to energy put in to heat the plasma; Q=10 means the plasma produces ten times more energy than was injected. Commercial viability generally requires Q well above 1.

Source: Impact of mechanical constraints on tokamak design and implications for high field power plants

The bigger picture

What these three papers share is a preoccupation with limits — not the exciting limits of how much energy fusion can produce, but the boring, physical limits that could stop a working reactor from surviving its own operation. ELMs hammer the wall. Exhaust heat melts it. Magnet structure crowds out the plasma. None of these is a new problem, but this week brought three genuinely independent attempts to quantify and mitigate them: a powder that calms the plasma's edge instabilities, a seeding strategy that keeps exhaust surfaces survivable near a gigawatt, and a design map showing exactly where the magnet arms race hits a wall. The picture this paints is of an engineering discipline that has moved past 'can we make fusion happen at all' and is now deep in 'can we build something that runs for decades without destroying itself.' That is real progress. It is also a lot harder than it sounds.

What to watch next

The boron powder result from DIII-D will be worth following — the team needs to demonstrate ELM-free periods lasting seconds, not hundreds of milliseconds, before this becomes a credible control strategy. On the ARC side, Commonwealth Fusion Systems has said SPARC (its proof-of-concept predecessor to ARC) is targeting first plasma in the late 2020s; any update on SPARC's construction timeline will tell us whether the modeling results this week are feeding into a real design cycle. The open question I'd want answered: can argon seeding be sustained in pulsed operation without gradually poisoning the core plasma?