Boron Pebbles, Self-Calming Plasma, and a Real-Time Aim Bot

Tiny Boron Balls Survive Fusion's Most Brutal Heat Test

Imagine dropping boron marbles into the hottest exhaust pipe ever built — that is roughly what the DIII-D team just did.

The divertor is the exhaust pipe of a tokamak. It is where the spent plasma slams into a solid surface, dumping enormous amounts of heat. It is also, quietly, one of the most unsolved problems in fusion engineering. A team at General Atomics mounted a rod of tiny boron spheres — each about the size of a peppercorn, sintered together with a carbon binder — inside DIII-D's lower divertor and ran plasma over it. The heat load reached 80 megawatts per square metre. To put that in terms you can feel: one megawatt per square metre would melt steel within a second. Think of it as holding a piece of blackboard chalk against a sustained blowtorch. You expect it to erode. It did — at roughly one centimetre per second when the plasma swept directly over the sample. Why bother? Boron is a favourite wall-coating material in fusion because it chemically mops up oxygen and other impurities that would cool the plasma. Right now, tokamaks coat their walls by vaporising boron powder in separate conditioning sessions. A renewable pebble bed that replenishes itself passively is an appealing alternative. And crucially, the team found that despite a cloud of boron dust spraying into the plasma, the core plasma kept performing normally. No sudden energy loss. No disruption. The catch: this was a first-look test — five plasma shots, two samples. About half the released boron went missing as fine dust scattered into the vacuum chamber. Where that dust settles over hundreds of shots, and whether it accumulates somewhere damaging, is genuinely unknown. A promising opening result, not a solution.

Fusion's Own Helium Ash Might Calm Its Worst Turbulence

The helium ash from a fusion reaction might actually calm the very turbulence that has been threatening to derail the whole process.

Turbulence inside a tokamak is a relentless leak. Heat that should stay in the plasma and keep the fusion going instead bleeds outward through chaotic, swirling motions, like body heat escaping through a badly insulated jacket. Managing that leak is one of the central engineering challenges of fusion. Now, a team using the CGYRO simulation code for a design called ARC — one of the most credible compact high-field tokamak concepts in development — has found evidence that burning plasmas might naturally suppress some of that leak themselves. Here is the mechanism. The fusion reaction produces fast helium nuclei, called alpha particles — the main ash of the reaction. These energetic particles stir up long, slow waves in the plasma. Think of a boat moving through a still lake: its wake generates a slow rolling swell that, somewhat counterintuitively, actually calms the smaller surface ripples around it. Those slow waves, in the simulation, enhance what physicists call zonal flows — organised, ring-shaped currents in the plasma that act like windbreak hedges, breaking up the chaotic turbulent mixing. The result: turbulent heat loss in the inner half of the plasma is significantly reduced when fast alpha particles are modelled properly, compared to a simulation that pretends they have already slowed down to background temperature. The team's integrated model gave an energy gain factor of about 22 for a slightly reduced-current ARC scenario — roughly ten times break-even. The catch: these are simulations, not an experiment. ARC is still a design on paper. And the suppression only works where fast alphas are dense — in the inner core. The outer edge remains turbulent territory, and honestly, nobody has confirmed this in a real burning plasma yet.

An AI Learns to Steer Fusion's Microwave Heater in Real Time

Steering a microwave beam inside a shifting magnetic bottle in real time — while hardware fails mid-shot — turned out to be a solvable problem.

One of the ways fusion machines heat their plasma is by firing microwaves at it — a technique called electron cyclotron heating, or ECH. The microwaves are injected by devices called gyrotrons, and they enter the plasma through steerable mirrors mounted on the machine's wall. The goal is to deposit that heat at exactly the right radial location inside the plasma. Aim too far in or out, and you waste the energy or, worse, you trigger instabilities. The problem: the plasma shifts and changes shape during a shot, so the optimal mirror angle ten milliseconds ago may already be wrong. A team at General Atomics built and deployed an algorithm called ECHO — ECH Optimization — directly on the DIII-D tokamak. It works like a smart navigation app for your microwave beam. Instead of re-running a full physics calculation from scratch every time the plasma changes — which takes too long — ECHO uses a neural network that has already learned a compressed, fast version of a complex beam-tracing code called TORBEAM. A genetic algorithm then searches that fast model for the best mirror angle in real time. The whole loop runs during a live plasma shot. The team demonstrated it working in actual experiments. It achieved target heating profiles even when one of the gyrotrons failed mid-shot — the algorithm noticed the gap and redistributed the remaining power automatically. The catch: the neural network was trained specifically on DIII-D's geometry and plasma conditions. Adapting it to a different machine — say, ITER — would require retraining from scratch on that machine's data. The method is proven; the portability is not yet.

Look at what these three papers share: they are all about managing energy at the boundary between the plasma and the physical machine. The boron pebble test asks whether you can build a renewable material buffer against the plasma exhaust. The alpha particle simulation suggests that burning plasmas may be naturally more cooperative than feared — the fuel itself might help manage turbulence. The ECHO algorithm shows that real-time computation is becoming a first-class tool for keeping the plasma well-behaved during a shot. None of these are theoretical gestures. Two involve measurements on a real machine. One is a simulation of a specific near-term design with named parameters and a named energy gain. What I find worth noting: the wall problem and the turbulence problem are no longer being treated as separate research tracks. They are converging on a shared question — how do you keep a burning plasma from destroying the thing containing it, for long enough to matter?

The DIII-D boron pebble team will almost certainly run follow-up shots with different protrusion geometries and more systematic dust inventory measurements — that missing 50% of boron is the number to track. On the broader calendar, ITER's tokamak assembly is the event that gives all wall-material and plasma-control research its deadline pressure; any material science result feeds directly into ITER's wall strategy decisions. The open question I would want answered next: does that fugitive boron dust accumulate somewhere in the vacuum vessel in ways that become a contamination problem over hundreds of shots, or does it benignly redistribute itself as a wall coating?

• Wikipedia: Divertor — what it is and why it matters
• Wikipedia: Gyrotron — how fusion microwave heaters work
• Wikipedia: Alpha particle — the helium ash of fusion