Fusion's Fuel Math Was Wrong All Along

We Have Been Miscounting How Much Fusion Fuel We Need

Every fusion power plant design assumes it knows how much fuel to store — and this paper says that assumption has been wrong for years.

Fusion runs on two forms of hydrogen: deuterium, which you can pull from seawater, and tritium, which barely exists in nature, is radioactive, and has to be bred inside the reactor itself. Every serious design for a commercial plant includes a loop called Direct Internal Recycling — DIR — that catches unburned tritium from the exhaust and feeds it straight back into the plasma. The models behind that loop have been built on a key assumption: that the gas injected into the plasma is roughly half deuterium and half tritium. A team building on work from researchers including Meschini and Moscheni has now shown that assumption is off by about a factor of ten. Here is the camping trip version. You budget fuel for your camp stove. But you forgot that keeping the campsite liveable — managing heat, managing waste — burns ten times more fuel on background jobs you never counted. The fusion equivalent is called gas puffing. To protect the reactor walls from overheating, operators constantly spray extra fuel gas around the edge of the plasma to keep it in a cool, diffuse state called detachment. Across every present-day tokamak and next-step stellarator in the team's database, that puffing rate exceeds the core fuel need by roughly ten to one. Once you account for puffing, the DIR loop fills up with the wrong ingredient. If most of the puffed gas is deuterium — which is cheaper and safer to handle — the tritium fraction in the loop drops, and fusion power falls by around 10%. Breeding ratios, storage requirements, and pump sizes all need to be recalculated. The catch: this is a modelling and database study, not an experiment. The fixes — puff less, seed more impurity gas to compensate, redesign the loop — are identified but not yet tested at reactor scale. I simplified the loop chemistry here; the actual tritium accounting is debated even among specialists.

Electrons Gone Rogue Are Blasting Chunks Out of Reactor Walls

A controlled explosion inside a fusion reactor sounds like a catastrophe — at the DIII-D tokamak in California, it is an experiment.

In a working plasma, electrons bump into each other constantly, like pedestrians jostling on a busy pavement. That friction keeps them in check. But during a disruption — a sudden, uncontrolled collapse of the plasma — the crowd thins out. Some electrons stop colliding with anything and accelerate freely, approaching the speed of light. These runaway electrons can carry enormous currents, and when they slam into the reactor wall they do not just heat it: they blow chunks off. A team from the DIII-D disruption group at General Atomics ran controlled versions of this scenario. They fired precisely characterised beams of runaway electrons at small graphite test pieces mounted on the tokamak wall, filmed the resulting debris field with high-speed cameras tracking roughly a thousand particles per frame, and then tried to reproduce what happened in a computer model. The model they used borrows a constitutive framework called Johnson-Holmquist — originally developed for ceramic armour in the automotive and defence industries — and couples it to a finite-element structural solver and a particle-hydrodynamics code in LS-DYNA. Two experiments were run, each depositing about 10 kilojoules of energy: one in 2023 (1.2 mm crater depth, 140 cubic millimetres of material blown off) and one in 2024 (0.9 mm crater, only 30 cubic millimetres lost, because the same energy was spread over a wider area). The model reproduced both cases. Most debris flew at tens of metres per second; about 10% reached 200–400 metres per second. Why does this matter? ITER — the international experiment under construction in southern France — must survive disruptions without catastrophic wall damage. Knowing how materials fracture lets engineers design walls that fail more gracefully. The catch: two events is a thin validation set. More energies, more materials, and more beam geometries are needed before anyone relies on this model for ITER wall design.

A Deliberate Magnetic Wobble Makes the Plasma Denser, Not Leakier

You'd think deliberately warping the magnetic cage around a fusion plasma would cause it to leak — under the right conditions, it does the opposite.

The magnetic field that confines a fusion plasma is not perfectly smooth. Engineers at facilities like DIII-D in San Diego have long used deliberately applied wobbles — called resonant magnetic perturbations, or RMPs — to tame violent plasma eruptions called ELMs. Think of ELMs as the plasma equivalent of a sneeze: sudden, violent, and damaging to the walls. RMPs suppress them, but the standard deal was that you paid a price: the wobbles also pushed particles out of the plasma, making it thinner and harder to sustain. You took the tradeoff. A team at General Atomics running experiments on DIII-D has now shown that the tradeoff is not fixed. When the plasma rotates in a specific direction — counter to the main plasma current — applying an RMP does the reverse of leaking particles: it pumps them in. Electron density rose by up to 15% in their experiments. The overall plasma pressure improved by about 13%. Think of a garden tap where, if you turn the handle a particular way, pressure in the pipe goes up rather than down. It works for one direction of flow and not for others. The team used modelling tools called TM1 and GPEC to trace the cause. When the plasma rotates counter-current, the RMP reverses the sign of a particle-loss process called neoclassical transport — instead of pushing particles out, it pulls them toward the core. The catch: this only works over a specific window of rotation direction and collisionality. ITER is designed to operate with co-current rotation, so translating this result to the international experiment will require more work. The team is candid about this. But the finding does suggest the physics of RMPs is richer — and more controllable — than the standard picture assumed.

Put these three stories side by side and a picture emerges — not of a technology racing toward success, but of a field doing the slower, harder work of getting the basics right. The tritium paper says the supply chain for fusion fuel has been planned around a number that is wrong by an order of magnitude. That is not a minor correction; it cascades through every design choice for a commercial plant. The runaway electron work says we are finally building the predictive tools to understand how plasma damage actually unfolds — frame by frame, particle by particle — rather than just hoping to prevent it. And the RMP result says a physics effect everyone treated as a fixed cost turns out to be a variable, depending on how you run the machine. Taken together: fusion is moving from 'will it work in principle' to 'here are the specific engineering decisions that determine whether it works in practice.' That transition is progress. It is also where the hard work lives.

On the tritium front, watch for responses from the groups designing DIR loops for DEMO-class reactors — this paper will force a reckoning with fuel cycle models that have been largely settled for a decade. On ELM control, the key open question is whether the RMP pump-in effect can be reproduced or adapted in co-rotating plasmas, which is the regime ITER will operate in. The ITER team is expected to present updated ELM-control scenarios at the IAEA Fusion Energy Conference later this year — that would be the first credible place to see a community response.

• Wikipedia: Tritium as fusion fuel
• Wikipedia: Edge-localized modes (ELMs)
• Wikipedia: Tokamak disruptions