The systems that make fusion possible
From magnetic coils and special materials to plasma sources, heating, diagnostics and simulation, here are the core technologies behind our open magnetic traps.
Magnetic System
The key system in both tokamaks and open magnetic traps is the system that generates the magnetic field used to prevent fusion plasma from coming into contact with the walls of the vacuum chamber.
When an electric current flows through a conductor, it generates a magnetic field around it. If the conductor is wound into a coil, the magnetic fields produced by each turn add together and reinforce one another, creating a stronger and more directed magnetic field. The greater the number of turns and the higher the current, the stronger the resulting magnetic field.
Plasma Sources
Starting a fusion device requires more than simply energizing the magnets and creating a vacuum. Charged particles must first be introduced into the chamber to form the initial plasma. This is accomplished using dedicated plasma injectors or radio-frequency antennas.
Plasma Guns
Plasma injectors can be thought of as “electric guns” that fire compact packets of already-ionized gas into the vacuum chamber. Once captured by the magnetic trap, these particles act as a seed plasma. Electric fields and heating systems then increase the number of charged particles and gradually heat the material to millions of degrees.
Unlike a conventional gas source, a plasma gun introduces matter in an already ionized state, simplifying discharge initiation and helping to establish a stable plasma more rapidly.
Helicon Antennas
As an alternative to plasma guns, helicon antennas are increasingly used to generate the initial plasma. Instead of injecting a ready-made plasma packet, they create plasma directly inside the vacuum chamber using radio-frequency waves.
A helicon antenna emits a special type of electromagnetic wave that propagates along magnetic field lines. These waves efficiently transfer energy to low-density gas, stripping electrons from atoms and converting the gas into dense plasma. In essence, the antenna acts as a “contactless ionizer,” generating plasma without electrodes or powerful electrical discharges.
External Heating Systems
Once an initial plasma has been created in the magnetic trap, it must be heated to temperatures sufficient for fusion reactions.
There are two primary methods:
1. Electromagnetic heating using radiation at frequencies resonant with plasma electrons or ions. Resonant absorption transfers energy efficiently to the corresponding plasma species. Electron heating is carried out using wave-guides transporting beams emitted from powerful gyrotron sources. Ion heating is carried out using an antenna located relatively close to the plasma.
2. High-energy particle injection. Because magnetic fields prevent charged particles from being injected across field lines, specialized Neutral Beam Injectors (NBIs) have been developed. These systems first generate beams of positively or negatively charged ions, accelerate them using electric fields, and then neutralize them through interactions with gas or photon-based neutralization systems. These neutral beams can penetrate deep into the plasma, become ionized, and contribute not only to plasma heating but also to fueling.
Numerical Tools
Diagnostics, no matter how numerous or sophisticated, give only an incomplete picture of the phenomena at play within the plasma. Therefore, numerical simulation is an essential tool in plasma physics. It enables not only interpretation of experimental observations but also investigation of promising plasma concepts before building physical experiments.
Unlike ordinary gases or liquids, plasma is a complex system of charged particles interacting both with externally applied electromagnetic fields and with one another through self-generated electric and magnetic fields. This makes plasma simulation computationally demanding and scientifically challenging.
Plasma simulation is done following different kinds of modeling, depending on the accuracy that is required. The models go from 0D global descriptions to precise kinetic descriptions, which evolve in time the 6D (3D in space, 3D in velocity space) distribution of the charged particles of the plasma. In between, 3D fluid models are also very useful, because the kinetic simulations can be very expensive in terms of computing resources. The simulations can be used to model the physics of plasma equilibrium and stability, they are indispensable to test new ideas and guide the systems envisioned for heating, fueling, confining or stabilizing the plasma.
Artificial Intelligence and Machine learning
Recent advances in machine learning and artificial intelligence have found numerous applications in processing experimental data, developing codes for plasma simulations, and even for real-time control of laboratory experiments.
Processing Diagnostic Data
Modern fusion facilities generate enormous amounts of data from dozens of diagnostic systems. Machine-learning methods help automatically identify patterns, filter noise, reconstruct missing data, and rapidly infer plasma parameters from complex diagnostic signals.
This significantly accelerates experimental analysis and enables near-real-time results. Neural-network models can also combine information from multiple diagnostics to provide a more complete picture of plasma conditions.
Development and Acceleration of Simulation Codes
Machine learning is used not only for experimental analysis but also for accelerating plasma simulations. Neural-network models can serve as fast approximations of computationally intensive physics calculations, reducing simulation times from hours to seconds.
AI also helps automate parameter optimization, analyze large simulation datasets, and identify physical patterns in modeling results. In the future, these techniques may become an important complement to traditional MHD, PIC, and hybrid simulation codes used in fusion reactor design.
Real-Time Plasma Control
Artificial intelligence is increasingly being applied to real-time plasma control. By analyzing diagnostic signals, AI algorithms can predict instabilities, disruptions, or departures from operational limits.
Based on these predictions, control systems can automatically adjust magnetic-coil currents, plasma heating, or fuel injection. This approach improves plasma stability and enables safe operation closer to machine performance limits.