Gravel, Plasma Exhaust, and a Real-Time Microwave Autopilot

Testing Rough Boron Pebbles as a Fusion Heat Shield in a Real Tokamak

What if the heat shield inside a fusion reactor was made of something closer to compressed gravel than a polished tile?

Picture a rod of sintered boron pebbles — roughly the texture of compacted gravel — protruding into the most hellish spot in a fusion device. That is exactly what a team at General Atomics tested in the DIII-D tokamak in San Diego. The divertor is the section of a tokamak that collects exhaust heat and particles from the plasma. Think of it as the drain in a bathtub that has to survive direct contact with plasma running at 100 million degrees. Current designs use solid tiles, but those degrade and become radioactive waste. The idea here: use boron pebble aggregates — replaceable, chemically useful (boron is already added to fusion plasmas to reduce impurities) — as the contact surface. In the experiment, a 1 cm rod of sintered boron pebbles was exposed to heat fluxes up to 80 megawatts per square metre. For reference, the surface of the sun radiates about 63 MW/m² — so this exceeds that. The pebbles eroded rapidly, shedding boron dust into the plasma at a surface recession rate of roughly 1 centimetre per second. About half the boron never came back; it floated through the plasma as fine dust. Here is the catch: this was five experimental shots, on tiny test rods, in a machine far smaller than a power plant. The erosion rate is high — you would chew through material fast at reactor scale. What the team found encouraging is that the core plasma performance was not significantly disrupted by all that boron dust flying around. That is the minimum bar you need to clear before scaling up. A small but real step.

Fusion's Own Helium Exhaust Helps Calm the Turbulence Inside the Plasma

The helium ash produced by fusion reactions might actually help the reactor hold onto its own heat — a self-reinforcing loop nobody wants to count on until it is confirmed.

Imagine you are trying to keep a pot of water at a precise rolling boil. Turbulence inside — bubbles jostling randomly — keeps pulling heat away in uncontrolled bursts. Now imagine the steam itself creates a gentle spinning current that damps the bubble chaos. That is, in rough terms, what a team working on MIT's ARC reactor design found in their simulations. When deuterium and tritium fuse, they produce helium nuclei called alpha particles moving at very high speeds. Previous models often treated these as passengers that slow down and blend in. This team used a simulation code called CGYRO and found something more interesting: fast-moving alpha particles drive long-wavelength waves in the plasma — called toroidal Alfvén eigenmodes, or TAEs, essentially slow ocean swells rolling through the fuel. Those swells pump energy into large rotating currents called zonal flows. Zonal flows are very good at breaking up the small, chaotic eddies that leak heat out of the plasma. In the inner half of the plasma volume, turbulent heat loss dropped significantly when fast alphas were included in the simulation. Why does it matter? Turbulence is one of the main reasons fusion plasma is so hard to keep hot. If the fusion reactions themselves help suppress that turbulence, the harder the reactor works, the better it contains its heat — a beneficial feedback loop. The catch: this is a simulation of one specific operating point of one reactor design. Real plasmas have complications that simulations miss, and this effect needs to be seen in an actual experiment before anyone designs a reactor around it. Honest answer: nobody knows yet how large the effect is in practice.

An Algorithm Steers Microwave Beams Inside a Fusion Reactor in Real Time

When a microwave source broke mid-experiment inside a live fusion plasma, the new control algorithm simply redistributed the remaining beams and kept going.

Keeping a fusion plasma stable is a bit like conducting an orchestra where some musicians keep trying to go rogue. One of the main tools for stopping dangerous magnetic instabilities — the kind that can tear the plasma apart in milliseconds — is firing precisely aimed microwave beams at specific spots inside the reactor to heat and calm those regions before they spiral. The problem: the right target spot keeps shifting as the plasma evolves, and recalculating it from scratch takes too long. A team at General Atomics built a system called ECHO — ECH Optimization — to solve this on the DIII-D tokamak. The core trick is a neural network trained to imitate a physics code called TORBEAM, which traces how microwave beams travel through plasma. The original code takes about 75 milliseconds per calculation — too slow for real-time use. The trained neural network does the same job fast enough to be paired with a genetic search algorithm: try thousands of antenna-angle combinations rapidly, keep the best one. The whole system runs inside the reactor's control computer and updates the microwave antenna mirrors during a live shot. Here is the moment that stuck with me: when one of the gyrotrons — the microwave sources — broke mid-shot during a test, ECHO automatically redistributed the remaining beams and kept hitting the target profile. That kind of fault tolerance matters enormously for a machine that will need to run for hours at a time. The catch: DIII-D has a modest number of heating beams compared to a full reactor. Scaling this to a machine with more sources, stronger fields, and faster-evolving plasmas is the real engineering test ahead. But validated real-time control in an actual tokamak is a genuinely useful result.

Look at these three papers together and a pattern emerges that the hype cycle tends to miss. The basic physics of fusion — that it works, that it releases energy — has been demonstrated. What the field is actually wrestling with right now is operational maturity: materials that survive brutal heat loads without turning into a liability, plasma effects that you can count on rather than fight, and control systems robust enough to handle real hardware failures mid-experiment. The boron pebble test is a materials question. The alpha particle simulation is a plasma behaviour question. The ECH algorithm is a control engineering question. None of them is glamorous in the way 'we hit ignition' is glamorous. But all three have the same practical shape: we need to know this before a fusion power plant runs reliably for thirty years. That is where the serious work is right now — and honestly, it is more interesting than the headline milestones.

The DIII-D boron pebble work will need follow-up at higher shot counts and with the pebble material fully integrated into the divertor rather than just protruding as a test rod — watch for a campaign result later this year. On the alpha particle side, the open question is whether any current experiment has the right conditions to measure the TAE-zonal-flow turbulence suppression directly; ITER's early deuterium-tritium campaigns would be the natural test, though those are still years away. The question I would most like answered: does the real plasma match the simulation, even roughly?

• Wikipedia: Divertor — what it is and why it matters
• Wikipedia: Electron cyclotron resonance heating
• Wikipedia: Alfvén wave — the wave type driving the alpha-particle effect