Since the 1950s, scientists have sought to harness the power of nuclear fusion. When two atomic nuclei collide at high speeds, they fuse into a single nucleus that has less mass than the sum of two original nuclei and in return releases energy. By harnessing energy sloughed off during this rection and efficiently perpetuating it on a large enough scale, humanity could one day build fusion reactors that would offer a source of electricity that is both carbon-free and that uses a less volatile and less dangerous process than today’s nuclear-fission-based power plants.
Jun 18, 2025
Home Top News
User Research
Physics
Energy
Simulation
See all news
Researchers at the Max Planck Institute for Plasma Physics (IPP) are currently focused on one of the largest remaining obstacles to making fusion practical: effectively confining the ultra-hot plasma needed to sustain a large-scale fusion reaction inside donut-shaped devices called tokamaks.
“The quest for fusion energy is a massive collaborative effort and the challenges involved go well beyond problem of disruptions stemming from loss of plasma confinement,” said Dr. Matthias Hölzl, Group Leader of the Non-Linear Magnetohydrodynamics (MHD) group at IPP. “That said, disruptions present one of the leading concerns for the feasibility of the tokamak concept as a whole. Understanding how to avoid and mitigate disruptions is essential if we aim to build reactors that could actually power our lives.”
Hölzl and his collaborators are using high-performance computing (HPC) resources at the High-Performance Computing Center Stuttgart (HLRS) to focus on a particular phenomenon that makes plasma disruptions so problematic: highly energetic runaway electrons that may arise because of the loss of plasma confinement and cause large, localized heat loads on the walls.
Creating a large-scale fusion reaction is no small feat: it is essentially trying to mimic the sun. Researchers confine and heat plasma material in a tokamak until it gets hot enough for fusion reactions. “For fusion to occur, the fuel needs to heat to around 100 million degrees Celsius,” said Hannes Bergström, doctoral student at IPP and first author of a refereed journal article that describes the newest findings. “At these extreme temperatures, electrons are no longer bound to atomic nuclei, so we say that matter is in a plasma state. We use magnetic fields to confine plasma, in part by running a current through it, but the entire system is a delicate balance, and instabilities can appear that may rapidly drop the temperature and disrupt the reaction.”
In these extreme environments, these liberated electrons can accelerate almost to the speed of light, becoming “relativistic” particles in the process. Via collisions, these electrons can create an avalanche with more and more of these highly energetic particles. Such a beam of high-speed particles can pose a risk to the reactor wall.
Due to the risk of damaging machine prototypes through experiment, plasma physicists rely on simulations to help understand how runaway electrons begin and how they could be averted or mitigated. To simulate plasma accurately, researchers use MHD, which treats plasma as a continuous object rather than a collection of individual particles. But the runaway electrons are not described accurately enough by such methods, and a so-called hybrid fluid-kinetic approach is needed. This approach captures the mutual interactions between the background plasma and the relativistic particles. The team was the first to develop such a model.
To get an accurate simulation of these complicated processes, researchers require HPC resources to efficiently run calculations needed to account for how the various properties during a reaction influence one another. “We have to simulate the evolution of temperature, particle transport, runaway electron generation, as well as electric and magnetic fields, and all of these things influence one another,” said Bergström. “In addition, we have to model the entire tokamak realistically in 3D and account for differences in time scales—disruption to plasma can happen over seconds, but runaway electrons can start to be lost in a matter of microseconds.”
The team uses its JOREK MHD code to model fusion plasma dynamics, but until recently, had been unable to use JOREK to accurately model runaway electrons from first principles, relying instead on less-accurate fluid models to include them in the simulation. As part of his PhD work, Bergström developed a model to include in JOREK that would both accurately simulate runaway electrons’ behavior and couple those reactions to the general plasma dynamics in a JOREK simulation. Using the Hawk supercomputer at HLRS, the team was able to verify this additional model’s accuracy, creating a simulation of a runaway electron beam hijacking the plasma current during a disruption, then converting back to being carried by the cooler plasma after the loss of the runaway electrons to the walls. “With this new model, we can use JOREK to study the interaction between runaway electrons and the rest of the plasma with accuracy that previously just wasn’t possible,” Hölzl said. “We benchmarked it to make sure the model behaves as we would expect, but now we need to simulate more realistic scenarios with larger plasma volumes and longer time scales, where earlier models might not be able to tell us what we should expect to see.”
With this new model integrated into JOREK, the researchers have set their sights on further developing the code to take advantage of the newest generation of HPC systems at HLRS and the other GCS centers—namely, they are porting their application to run on GPUs and AMD’s accelerated processing units installed in HLRS’s latest system, Hunter. “We expect this work to yield a substantial speed up and allow us to cross longer time scales—this is crucial if we want to study aspects of the runaway electron process that are happening across widely different timescales,” Hölzl said.
With access to these accelerated systems, Hölzl hopes to continue pushing the frontier of fusion simulations from emulating conditions in a tokamak built for fundamental fusion research to a device like ITER—a large experimental fusion reactor being built in France that aims to create the largest-ever sustained fusion reaction when it goes into operation in the early 2030s. Hölzl noted that while accurately simulating a reactor on the scale of ITER with the full complexity of the new model is still beyond current capabilities, the team’s work is laying the groundwork for next-generation simulations to provide insight into next-generation tokamaks.
“Input from our work will help make future large fusion devices more reliable and more efficient,” he said. “The size of future devices does increase the computational cost of these simulations by orders of magnitude. That means we must focus on model development, validation, and optimization. The availability of powerful computational resources remains essential.”
— Eric Gedenk
Bergström H, Liu SJ, Bandaru V, et al. 2025. Introduction of a 3D global non-linear full-f particle-in-cell model for runaway electrons in JOREK. Plasma Phys Control Fusion 67: 035004.
Funding for HLRS's Hawk and Hunter supercomputers was provided by the Baden-Württemberg Ministry for Science, Research, and the Arts and the German Federal Ministry of Research, Technology and Space through the Gauss Centre for Supercomputing (GCS).
