New paper tracks five years of progress towards digital twins for fusion, and the gaps that still need to be closed.
A digital twin is a virtual model that mirrors the structure, context and behaviour of a physical system. It can predict how that system will behave and inform decisions in the real world. A digital twin of Earth’s oceans, for example, could show how sea level rise would affect particular locations under different conditions. A digital twin of a city could help urban planners decide where to intervene first to get more people onto public transport.
Digital twins prove their worth in genuinely complex settings, which is exactly why they’re important for fusion, as the field itself grows more complex.
“Many aspects of fusion plasma can now be modelled with a high level of reliability and predictive capability,” says Frank Jenko, Director at the Max Planck Institute for Plasma Physics and Head of Digital Solutions for Fusion Office at EUROfusion. “So, the natural next step will be to combine as many of these elements as possible – and that’s the point at which we start to talk about digital twins.”
The digital twin concept has circulated in engineering since the early 2000s, and fusion researchers worldwide have been exploring it for years. In 2021, EUROfusion launched E-TASC (Theory and Advanced Simulation Coordination), an initiative to coordinate that effort across Europe. Now, in a new paper, Jenko and his colleagues report on the progress made over the past five years.
Physics of plasma
Some of the clearest progress reported in the paper concerns simulations of plasma inside a tokamak. Researchers have made notable strides in modelling the plasma core and its key features: turbulence, the instabilities that can disrupt magnetic confinement, and the behaviour of fast particles produced by the fusion reactions themselves.
“In the core we’re in relatively good shape. We’ve done a lot of code validation successfully, that’s one of the major advancements,” Jenko adds.
Until now, though, this progress has largely been made by different teams working in isolation. That siloed approach is starting to change, and the paper documents real steps in that direction: simulations that couple turbulence with fast-particle-driven waves, and codes that track how instabilities evolve deep into the nonlinear regime rather than stopping at their onset.
But while core-related research has advanced steadily, there’s still significant work to do on the plasma edge, where the plasma is closer to the reactor walls.
“One of the key open issues is what happens in the edge region,” Jenko says. “This is harder to model for various reasons. The physics is, by nature, more complex.”
Part of that complexity comes from how materials affect plasma. Tungsten, for example, is used in reactor walls, and over the many years a power plant should operate, it’s unavoidable that some of it will come loose and enter the plasma. Modelling how these stray particles affect the plasma is difficult – and so is capturing the turbulence that governs how heat and particles actually reach the wall.
Still, the paper reports progress: turbulence codes that now reproduce detachment (a regime in which the plasma cools sharply before it reaches the wall), and reactor-scale simulations of erosion and dust formation. But more work is needed.
Simulating stellarators
The paper also notes progress in simulating stellarators. Long overshadowed by tokamaks, stellarators gained considerable credibility from the success of Germany’s Wendelstein 7-X. Advances in simulation now let researchers design and assess stellarators with a confidence that wasn’t possible before.
“Stellarators aren’t quite as mature yet, simply because there isn’t the same amount of data available, in theory or in experiment,” Jenko says. “But we’re hopeful that we can advance these stellarator simulations to a level where they can also lead to power plant concepts.”
Jenko doesn’t see this as an either/or competition between stellarators and tokamaks. He sees two parallel paths that digital-twin modelling can help mature side by side.
Looking ahead
When it comes to next steps, future work will inevitably involve stitching together components that have been modelled in isolation. To do so, researchers will continue to explore frameworks that incorporate different pieces of information together. Bluemira, for example, is an integrated design tool for future fusion reactors. It incorporates different modules, some of which rely on other codes, to carry out a range of conceptual fusion reactor design activities.
Researchers will continue to use this and other frameworks to build systematic simulations rather than isolated models. Teams will also need to validate the integrated simulations against existing machines before trusting them with machines that don’t exist yet, such as ITER or a future power plant.
“Ultimately, the development of digital twins is really a game changer,” Jenko says. “We’ve taken steps in that direction, and the natural next step will be to put the pieces together into a coherent whole.”
A digital twin of a full fusion power plant isn’t here yet. But its building blocks – models of the plasma core, the edge, the wall, and the machine around them – are closer to reality than ever before.
“When it comes to fusion, we want to be as efficient as possible. This means we should use every opportunity that presents itself to us, including creating digital twins,” Jenko concludes. “Five years ago, this was just an aspiration. Now it’s a research and development programme, and it’s moving fast.”