This is how artificial intelligence is already helping us overcome one of the biggest challenges of nuclear fusion

The construction of ITER is on the right track. The agenda planned by EUROfusion, the international consortium that is responsible for the development of this experimental nuclear fusion reactor, is meeting deadlines little by little. And, if nothing goes wrong, in 2025 the assembly of this extremely complex machine will conclude and they will begin the first tests with plasma.

There is no doubt that this will be an important milestone, but there will still be several challenges ahead that will need to be overcome to enable the arrival of commercial nuclear fusion, which is estimated, according to EUROfusion, by the 60’s. The IFMIF-DONES project will be in charge of developing the materials to be used in the internal lining of the DEMO vacuum chamber. And there is no doubt that it is a great challenge.

But there is another major challenge that we cannot ignore: it is imperative to understand how plasma behaves, which is at a temperature close to 150 million degrees Celsius, to stabilize it. It is necessary to solve this problem in order to be able to sustain the nuclear fusion reaction over time, and in this area the threat comes from the turbulence that originates naturally in the outermost layer of the plasma, which is precisely the one that is most near the walls of the vacuum chamber.

The turbulences that originate in the ‘crust’ of this extremely high temperature gas are in a way similar to the deflagrations emitted by our Sun, but the fact that the plasma is confined by a magnetic field housed inside a chamber forces technicians to prevent it from coming into direct contact with the container walls. Otherwise, if you touch them will degrade them and the nuclear fusion reaction will not be able to be sustained.

Researchers are currently working on various strategies to solve this problem, and one of the most promising is trying to take advantage of the stabilizing effect they exert on plasma. ionized helium-4 nuclei (We talk about it in some depth in the article that I link here). This is the line being followed by a research group at MIT, and precisely another scientific team from this institution has made a very important discovery in this same area.

Deep learning helps us understand how plasma behaves

Even if the IFMIF-DONES project fulfills its mission and manages to find the inner lining materials that will allow the vacuum chamber to withstand the impact of high-energy neutrons, a challenge will remain: preventing the plasma heat fluxes in the form of turbulence. Damage the walls of this chamber. To achieve this, it is necessary to know precisely how plasma behaves, which has led a group of MIT researchers to develop a mathematical model that seeks to predict it.

The curious thing is that deep learning is playing an absolutely essential role in the elaboration of this turbulence model, which serves to put it to the test and assess your predictive ability. In fact, in two of the articles they have recently published (we leave the links right at the end of this text), these researchers explain in detail the strategy they are using to ensure that deep learning allows them to infer new knowledge about the behavior of the student. plasma. Its purpose is to refine your turbulence model as much as possible to precisely stabilize the plasma.

This statement by Abhilash Mathews, one of the MIT researchers, describes his approach very well: “A successful theory must be able to predict what you are going to observe, such as temperature, density, electric potential. or the flow. In fact, it is the relationship that exists between all these variables that fundamentally defines a theory of turbulence. Our work essentially analyzes the dynamic relationship that exists between two of these variables: the electric field of the turbulence and the pressure of the electrons.

His words clearly show how difficult it is to predict the behavior of the plasma at very high temperatures confined inside the nuclear fusion reactor. In fact, according to these researchers, the turbulence to which this gas is subjected they are much more complex than those that we can observe in other fluids, such as air or water. Still, the advances these and other scientists are making invite us to view the future of nuclear fusion with reasonable optimism. There is still much to do, but the outlook is hopeful.

Cover image | Steve Jurvetson

More information | Physical Review E | American Institute of Physics