It used to be merely a dream — to be able to produce an unlimited amount of clean energy in the same way as the sun itself, using only abundantly available hydrogen gas as source material.

But after nearly a century of effort, the dream seems finally about to become real. August 2020 marked the beginning of the crucial construction phase of the gigantic ITER nuclear fusion reactor in southern France, near Marseille.

In the reactor hall, the high-vacuum chamber is being assembled that will later harbor the ring-shaped reactor core. A research reactor of the most promising tokamak type, ITER is meant to be the ultimate touchstone that will show if and how humanity will be able to harness the power of nuclear fusion.

Similar to much smaller research reactors already in use, ITER will be the magnetic vessel for ultra-hot hydrogen plasma, designed to keep the plasma hot enough to fuel enough fusion reactions to produce net energy. At its core, the plasma will reach the unimaginably high temperature of 100 million degrees Celsius. In comparison, the outer rim of the plasma will be much colder, with temperatures of a few thousand degrees Celsius.

“These strong temperature gradients are the source of free energy that triggers a number of instabilities, which in turn spontaneously develop into turbulences,” explains Laurent Villard, an adjunct professor of plasma physics theory and numerical simulation at The Federal Institute of Technology in Lausanne (EPFL).

“Understanding, predicting, and ultimately controlling these turbulences will be crucial for the successful operation of ITER.” To investigate these turbulent processes, Villard and his colleague Paolo Ricci, an associate professor at EPFL, together with their co-workers use the Piz Daint supercomputer to perform plasma simulations. The goal is to determine how the heat and particles will be transported within the plasma and how they will impact the reactor wall. 

Charged particles make for complex turbulence 

Turbulence in a plasma is essentially different and more complex than turbulence in water or air because the particles of a plasma are electrically charged. Their movement creates electromagnetic fields, which in turn affect particle motion, which again affects local charges and therefore the electromagnetic field, and so on.

This results in a plethora of various waves and instabilities. In particular, the particles’ motion also slightly affects the strong background magnetic field originating from the gigantic reactor magnets holding the plasma in place. “All these interactions make solving the system of mathematical equations particularly challenging,” says Villard.

He and his team use gyrokinetic modeling to represent the physics of the ultra-hot, very low-density core of the plasma. Contrary to fluid dynamics, gyrokinetics is used to represent the flow and properties of particles that have a long mean free path, meaning they travel a long average distance before they collide with another particle. In these situations, the particles’ velocity and, consequently, their kinetic energy is extremely high, so the use of kinetic theory is essential for a complete and accurate description. 

Accelerated simulations 

For their numerical simulations, Villard and his team collaborated with experts at the Swiss National Supercomputing Centre (CSCS) to refactor their ORB5 code and adapt it to run on GPUs instead of CPUs. “This accelerated the performance by a factor of 8, reducing previous weeks of simulation time to mere days,” says Villard.

The improvement now allows the scientists to perform large scale turbulence simulations that were not practical before. In addition, the team enhanced the capabilities of their code by introducing a sort of virtual antenna that provides the possibility to introduce external perturbations into the system. This means that simulated added effects, such as sheared flows, can be investigated for their ability to suppress plasma turbulence.

With their new simulations, the scientists have discovered that turbulence of the plasma core is affected by instabilities occurring at the very periphery of the plasma, even if those instabilities are very small-scale. They are also examining the effect of large-scale instabilities within the plasma.

In preliminary work, they have identified a few mechanisms of how such large-scale instabilities interact with turbulences. “These are processes that can change the whole reactor behavior,” explains Villard. This coincides with another study in collaboration with the Max Planck Institute for Plasma Physics that is investigating how turbulence interacts with the fast particles created by the fusion reaction and with large-scale instabilities that are created in the process. 

The outer rim of the plasma 

Complementary to Villard’s investigation of the plasma core, his colleague Paolo Ricci uses enhanced fluid dynamics to simulate the activity at the edge of the plasma and the interaction between the plasma and the reactor wall. “At this interface between plasma and wall, we face the steepest temperature gradient in the system,” explains Ricci.

The inner wall of ITER will be built using materials such as tungsten and beryllium that can withstand large heat fluxes. Still, the confinement of the plasma will never be perfect. In operation, the heat flux will have to be reduced to an acceptable level, meaning around 5 megawatts per square meter, similar to what spacecrafts have to survive when they re-enter the atmosphere. “To achieve this, we need to understand what happens at the interaction between the plasma and the wall,” says Ricci. 

That’s where his models based on the Global Braginskii Solver (GBS) code come into play. Apart from fluid dynamics, they include representations for the electric and magnetic fields as well as for the reactor wall, including its inherent material impurities. This method enabled Ricci and his team to understand the dependence of the plasma heat flux on factors like machine size, plasma density, and applied heating power, and then estimate the plasma heat flux in ITER during its start-up phase.

In the future, Ricci plans to port the GBS code to GPUs as well, which will enable more complete simulations of the ITER plasma cycles. Their findings, he says, will help to operate the fusion reactor in the most effective way, using its full power safely and reliably.

Snowflake should make things cooler

In part, however, the scientists’ work is already focusing on a next fusion reactor, DEMO, the device that is meant to one day actually bring fusion electricity to the grid. Using Piz Daint, Ricci and his group were the first to simulate plasma turbulence in the so-called snowflake configuration.

This novel setup aims to distribute the immense heat load more evenly in a larger area by employing more complicated magnetic fields than ITER. The snowflake configuration was already tested experimentally using TCV, a versatile research fusion experiment at EPFL. Now, Ricci’s snowflake simulation has proven to be in agreement with the TCV experiment, and thus suitable for further investigation.

In the long term, Ricci and Villard plan to couple their distinct representations of edge and core plasma—an ambitious and challenging enterprise since the structures of the two currently used codes, ORB5 and GBS, are fundamentally different, as are the physical processes they describe.

In the meantime, construction of the ITER reactor is advancing, and with it the chance to witness the exciting possibilities of nuclear fusion. The first ITER plasma is expected by the end of 2025, the first net energy near 2050.