Magnets, Runaway Electrons, and a 30,000x Speed Trick

Superconducting Magnet Joints That Stay Reliable Under Pressure

The connector between two layers of a superconducting magnet can silently become ten thousand times worse at conducting electricity — and nobody had a reliable fix.

Think of a superconducting magnet as a very tightly wound spool of special tape, layer upon layer. Between each layer, there is a tiny electrical contact. In normal operation, that contact has a specific resistance — a deliberate, engineered resistance, not zero, because you actually want some resistance there to protect the magnet if something goes wrong. This is called 'resistive insulation', and it is one of the tricks that makes modern compact fusion magnets safer. The problem a team at MIT's Plasma Science and Fusion Center has been studying: when that contact is pressed and released thousands of times — as happens every time a magnet charges and discharges — the resistance can drop by up to four orders of magnitude. That is a factor of ten thousand. Imagine a dimmer switch in your living room that, after a year of use, suddenly stops dimming and just goes to full brightness. You wanted control. You lost it. The team tested several surface treatments on the tape at 4.2 Kelvin (colder than outer space) and cycled the pressure 30,000 times. The winner: coating the REBCO tape with a layer of tin-lead solder just two to three micrometres thick — roughly one-thirtieth the width of a human hair. That coating kept the contact resistance stable across all 30,000 cycles. They then validated this in an actual six-layer test coil running at fields above 11 Tesla, and the short-sample results held. The catch: this was a test coil, not a reactor-scale magnet. Scaling this manufacturing process to the kilometres of tape in a real fusion device is still an open engineering challenge, and the solder coating process has to be done reel-to-reel without degrading the tape's ability to carry current. So far, it hasn't. But 'so far' in a controlled lab is different from a production line.

Why Injecting Gas Safely Stops a Fusion Plasma from Exploding

When a tokamak plasma suddenly collapses, it can fire a beam of electrons at near-light speed straight into the reactor wall — and for years, we didn't fully understand how to stop it cleanly.

Here is a scenario fusion engineers worry about constantly. A tokamak plasma loses its stability and collapses in a fraction of a second — a 'disruption'. The magnetic energy left over can accelerate a small population of electrons to nearly the speed of light. These 'runaway electrons' then slam into the reactor wall like a particle beam, punching holes in the material. In a big reactor like ITER, that is potentially hundreds of megajoules of damage. One mitigation strategy is to inject a puff of gas or frozen pellets — often deuterium — into the plasma right as it collapses. This is called 'shattered pellet injection'. It works, but until now the precise reason why it works, and more importantly why it only works in a specific quantity window, was not well understood. A team modelling parameters close to the DIII-D and AUG tokamaks, using the JOREK nonlinear simulation code paired with first-principles kinetic modelling, found the key mechanism. When enough neutrals are injected, free electrons recombine with ions — the plasma partially un-ionises. This temporarily raises the electrical resistance of the plasma by about ten times above its normal value. That spike in resistance preferentially amplifies tearing instabilities at the outer edge of the plasma rather than deeper inside it. Edge tearing makes the outer magnetic field lines go chaotic in a controlled way, which disperses the runaway beam over a wider area instead of focusing it on one spot. Benign termination. The catch: the simulations use uniform, time-independent resistivity — a simplification. Real plasma resistance is spatially uneven and evolves rapidly. The team acknowledges this is a stepping stone, not a full engineering prescription. The window of 'right amount of neutral injection' is narrow, and understanding its exact boundaries in reactor conditions is still work in progress.

A 30,000-Times Faster Way to Simulate Mirror Fusion Machines

Simulating a plasma trapped in a magnetic mirror from cold start to steady state would have taken longer than your career — until a team found a shortcut that is 30,000 times faster.

Magnetic mirrors are a different breed of fusion machine from tokamaks. Instead of bending plasma into a doughnut, a mirror machine traps plasma between two strong magnetic field zones, like squeezing a garden hose at both ends to stop water escaping. They have been mostly out of fashion for decades, but small HTS-magnet machines like WHAM at the University of Wisconsin are reviving interest — partly because they are cheaper and faster to build. The problem: to simulate how plasma behaves in these machines from first principles, you need to track billions of particles across a time range that spans nine orders of magnitude. At one end, a particle bounces back and forth in microseconds. At the other end, collisions slowly leak particles out over tenths of a second. Simulating that full range with standard methods would require computational time measured in geological eras, not grant cycles. A team working on the WHAM configuration developed two combined tricks. The first, called pseudo orbit-averaging, alternates between simulating full particle motion and a mathematically averaged version of it — like occasionally fast-forwarding between key scenes in a film rather than watching every frame. The second trick artificially slows down the fastest timescales by a factor of ten, so the simulation can take bigger steps without missing important physics. Together, these deliver a claimed 30,000-times speedup. The resulting equilibrium — how the plasma settles into a stable state — matches analytic theory derived decades ago, which is a meaningful validation. The catch: this is a one-dimensional model along a single field line, not a full three-dimensional simulation. Real mirrors have three-dimensional instabilities that this approach cannot yet capture. Think of it as correctly solving a simplified version of the problem — which is nonetheless a prerequisite for solving the real one.

Look at these three stories together and a pattern emerges: fusion's engineering problems are increasingly becoming precision manufacturing and simulation problems, not just physics problems. The REBCO coil work says: we can build the magnets, but we need to control the interfaces between layers at the micrometre scale, under cryogenic cycling, reliably enough for an industrial process. The runaway electron work says: we understand disruption mitigation well enough to explain why a specific intervention works — but the operating window is narrow and we need to map its edges before ITER switches on. The mirror simulation work says: we now have tools to compute plasma equilibria in a new class of machine, which matters because the mirror revival is real and HTS magnets are what make it viable. None of these is a shortcut to the finish line. All three are the kind of incremental, unfashionable engineering rigour that actually gets there.

The WHAM mirror experiment at Wisconsin is actively taking data — watch for first plasma results that can be compared against the gyrokinetic equilibria described in today's paper. On the disruption side, ITER is still several years from first plasma, but the shattered pellet injection system is being designed now; any updates from the ITER Organization on pellet injection validation experiments are worth tracking. Open question I'd want answered: does that 30,000x speedup hold when you extend the mirror simulation to two or three dimensions, where drift instabilities live?

• Wikipedia: High-temperature superconductivity and REBCO
• Wikipedia: Magnetic mirror fusion concept
• Wikipedia: Tokamak disruptions