🔬 physics

AXUV synthetic diagnostic for ASDEX Upgrade and its application for SPI simulations

This paper presents a synthetic diagnostic tool based on the Cherab-Raysect framework to model AXUV diode measurements on the ASDEX Upgrade tokamak, enabling the validation of shattered pellet injection simulations by comparing synthetic signals with experimental data and revealing both qualitative agreements and specific discrepancies related to impurity content.

Original authors: Ferenc Lengyel, Weikang Tang, Matthias Hölzl, Matthias Bernert, Matěj Tomeš, Peter Halldestam, Paul Heinrich, Gergely Papp, Stefan Jachmich, Umar Sheikh, Mathias Dibon, Pascal de Marné, Jörg Hobirk, T

Published 2026-08-25

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Ferenc Lengyel, Weikang Tang, Matthias Hölzl, Matthias Bernert, Matěj Tomeš, Peter Halldestam, Paul Heinrich, Gergely Papp, Stefan Jachmich, Umar Sheikh, Mathias Dibon, Pascal de Marné, Jörg Hobirk, Thomas Eberl, Gergő I. Pokol, for the ASDEX Upgrade Team, the EUROfusion Tokamak Exploitation Team

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Inside the massive, doughnut-shaped machines known as tokamaks, superheated gas called plasma is held in place by powerful magnetic fields. This plasma is the fuel for future fusion power plants, which aim to replicate the energy source of the sun. However, these machines are fragile. If the plasma loses its stability, it can collapse in a fraction of a second, an event known as a disruption. Such a collapse releases a torrent of energy that can damage the machine's interior walls, threatening the viability of fusion as a clean energy source. To prevent this, scientists have developed a safety system called shattered pellet injection. Imagine firing a frozen pellet of gas into the heart of the plasma at high speed. The pellet shatters into thousands of tiny fragments, which then spread through the plasma like a cloud. This cloud cools the plasma down rapidly and evenly, draining its energy in a controlled way before it can cause damage.

The challenge lies in understanding exactly how this cooling cloud behaves. Scientists need to know where the fragments go, how fast they move, and how they radiate heat away. To see this, researchers at the ASDEX Upgrade tokamak in Germany installed a specialized set of cameras equipped with sensors called AXUV diodes. These sensors are incredibly fast, capable of detecting flashes of light in millionths of a second, far quicker than standard cameras. They are designed to see the specific types of light emitted by the gas inside the pellet as it breaks apart and cools the plasma. However, these sensors are not perfect. Over time, the intense radiation inside the machine dulls their sensitivity, and their ability to see different colors of light changes unpredictably. This makes it difficult to turn the raw signals from the cameras into a clear picture of what is happening inside the machine, especially when the gas inside the pellet is a complex mixture.

To solve this puzzle, a team of researchers developed a new digital tool, a "synthetic diagnostic," which acts as a virtual mirror to the real experiment. Instead of trying to interpret the messy, degraded signals directly, they built a computer model that simulates exactly how the cameras should see the plasma. They fed this model detailed data from high-fidelity computer simulations of the shattered pellet injection process. The model accounts for the specific shape of the cameras, the way light bounces off the machine's walls, and the changing sensitivity of the sensors. By running the simulation through this virtual camera, the team could generate a predicted signal that could be directly compared to the real signals recorded during actual experiments. This approach allowed them to bypass the uncertainties of the physical sensors and focus on the physics of the event itself.

The researchers tested this tool against two different types of pellet injections. In the first case, they used a pellet containing a significant amount of neon gas mixed with hydrogen. The results were strikingly clear. The synthetic signals generated by the model matched the experimental data with remarkable precision. The virtual camera showed the cooling cloud moving inward toward the center of the plasma at the same speed and with the same intensity as the real camera saw it. The model successfully tracked the radiation front as it spread, confirming that the computer simulations of the pellet's behavior were accurate for this high-neon scenario. This agreement gave the scientists confidence that their understanding of how a heavy, neon-rich pellet cools the plasma was correct.

The second test, however, revealed where the current models fall short. This time, the pellet contained only a tiny trace of neon. In the real experiment, the cooling cloud did not behave as the simulation predicted. The actual camera saw fine, vertical structures of light drifting downward toward the bottom of the machine, a motion that the computer model completely missed. The researchers realized that the simulation had failed to account for a physical phenomenon known as the "rocket effect." As the pellet fragments vaporize, they shoot gas out the back, propelling the fragments forward and causing the cloud to drift, much like a rocket. Because the simulation ignored this drift, it could not reproduce the downward motion seen in the real data. Furthermore, the real machine contained background impurities that radiated light, adding a layer of complexity that the simplified simulation did not include.

This work demonstrates that while the shattered pellet injection method is a powerful tool for protecting fusion reactors, the physics governing the injection of light pellets is more complex than previously thought. The new synthetic diagnostic tool proved its worth by not only validating the simulations for heavy pellets but also by pinpointing exactly where the models for light pellets were missing key physics. The researchers concluded that to accurately predict the behavior of future fusion reactors, especially those using lighter pellets, their computer models must be updated to include the rocket effect and the influence of background impurities. By bridging the gap between messy real-world data and clean computer simulations, this tool provides a clearer path toward mastering the safety mechanisms needed for the next generation of fusion energy.

Technical Summary: AXUV Synthetic Diagnostic for ASDEX Upgrade and SPI Simulations

Problem Statement
The paper addresses the challenges in interpreting Absolute eXtended UltraViolet (AXUV) diode measurements during Shattered Pellet Injection (SPI) experiments on the ASDEX Upgrade (AUG) tokamak. While AXUV diodes offer high time resolution ( $\sim\mu$ s) suitable for studying fast phenomena like SPI, their utility for determining absolute radiated power is compromised by two main factors:

Spectral Responsivity Degradation: The diodes' spectral sensitivity degrades over time due to exposure to extreme UV and vacuum UV radiation, and this degradation is non-uniform across the spectral range.
Time-Varying Spectra: In mixed Ne/D $_2$ SPI experiments, the plasma emission spectrum evolves rapidly (e.g., shifting from bremsstrahlung to line radiation as the plasma cools). Since the diodes have non-uniform spectral responsivity, the effective response ( $\eta_{eff}$ ) changes dynamically during the discharge.

These uncertainties, particularly when neon is injected (which radiates strongly in the diodes' sensitive range), make absolute power reconstruction and tomography unreliable. Consequently, the authors sought a method to bridge the gap between first-principles plasma simulations and experimental AXUV data without relying on absolute calibration.

Methodology
The authors developed a 3D synthetic diagnostic tool within the Cherab-Raysect optical modelling framework to simulate AXUV camera signals based on plasma simulation data.

Framework: The tool utilizes the Raysect ray-tracing library, adapted to handle the wide spectral range (1 eV to 1.24 keV) required for AXUV diodes, which extends beyond Raysect's typical visible-light focus.
Geometry: The diagnostic models 192 lines of sight (LOS) across four AXUV cameras (two vertical and two horizontal types) in two toroidal sectors (Sector 16, where SPI occurs, and Sector 5, $\sim110^\circ$ away). It incorporates the full 3D CAD geometry of the plasma-facing components (PFCs) to account for spectrally dependent reflections, assuming tungsten surfaces with a roughness parameter of 0.29.
Simulation Input: The synthetic diagnostic processes output from JOREK, a non-linear extended magnetohydrodynamics (MHD) code. Plasma parameters (electron density, temperature, and impurity charge states) from JOREK are super-sampled onto a uniform 2D poloidal grid (2 cm $\times$ 2 cm resolution) and convolved with pre-calculated diode sensitivity matrices.
Spectral Handling: To manage computational cost while maintaining resolution, the spectrum is divided into three regions (1–10 eV, 10–100 eV, 100–1240 eV) with 100 bins each (300 total bins). The model includes bremsstrahlung and neon line radiation (OPEN-ADAS), but excludes background impurities in the simulations to isolate the pellet radiation effects.
Verification: The tool was verified against analytically calculated etendue values (mean relative difference of 2.3%) and cross-checked against the Cherab PowerPipeline Monte Carlo method (average relative error of 0.017%).

Key Contributions

Development of a 3D Synthetic Diagnostic: A robust tool capable of generating synthetic AXUV signals from 3D MHD simulations, accounting for complex geometry, spectral responsivity degradation, and PFC reflections.
Methodological Shift: The paper argues against attempting absolute radiated power reconstruction or tomography for mixed Ne/D $_2$ SPI due to spectral uncertainties. Instead, it proposes comparing qualitative features and time-evolution patterns between synthetic and experimental signals.
Application to JOREK Simulations: The diagnostic is applied to two specific AUG discharges from the 2022 SPI campaign:
- #40673: High neon content (10 mol% Ne).
- #41007: Low neon content (0.17 mol% Ne).

Results
The synthetic signals were compared with experimental AXUV data from the 2022 SPI campaign:

High Neon Content (#40673):
- Agreement: There was particularly good agreement in the time evolution of the radiation front. The inward radial penetration of the radiation cloud in the simulation matched the experiment in both velocity and amplitude up to $t \approx 2.3285$ s.
- Features: Both synthetic and experimental data showed fine radiation structures moving from the main cloud toward the high-field side and spreading along outer flux surfaces.
- Discrepancies: Minor differences were observed in the timing and intensity of vertical spreading in the horizontal camera data. The synthetic data showed a single abrupt spreading event, while experiments showed multiple smaller events. The authors attribute this partly to the stochastic nature of pellet fragmentation (simulations use the Parks model which may overestimate fragment sizes) and the absence of background impurities in the simulation.
- Toroidal Spread: At Sector 5 ( $\sim110^\circ$ away), the simulation showed delayed radiation onset compared to experiments, likely due to the simulation not accounting for equilibrium toroidal flow.
Low Neon Content (#41007):
- Discrepancies: The match was significantly poorer. The experiment showed fine vertical radiation structures drifting downward shortly after fragment arrival, which were absent in the simulation.
- Cause: The authors attribute this to the rocket effect and plasmoid drift, which are prominent in low-neon pellets but were not included in the JOREK simulation used.
- Background Radiation: In the experiment, background impurity radiation (e.g., tungsten) was comparable in magnitude to the pellet radiation, whereas the simulation (lacking intrinsic impurities) showed only reflection-based background.
- Conclusion: Despite the drift discrepancies, both simulation and experiment showed the radiation spreading poloidally by the time thermal energy dropped to 10% ( $t_{0.1W}$ ).

Significance and Claims
The paper claims that the developed synthetic diagnostic is a capable tool for validating and interpreting SPI simulations against experimental AXUV data, specifically by identifying missing physics in the models rather than relying on absolute power calibration.

Physics Validation: The tool successfully detected the absence of specific physics models. For the low-neon case, the mismatch in radiation drift patterns directly indicated the need to include the rocket effect and plasmoid drift in future simulations.
Robustness: The study demonstrates that even with order-unity uncertainties in channel sensitivities and degradation, qualitative comparisons of signal features (spatial structure and time evolution) remain valid and informative.
Future Recommendations: Based on the analysis, the authors recommend that future SPI simulations should include:
1. Rocket effect and plasmoid drift models (especially for trace-neon scenarios).
2. Background impurity radiation.
3. Equilibrium toroidal flows.

The paper concludes that this approach allows researchers to assess which physics models are necessary to reproduce experimental observations, thereby supporting the development of disruption mitigation strategies for ITER.

Technical Summary: AXUV Synthetic Diagnostic for ASDEX Upgrade and SPI Simulations

More like this