Quantitative 3D non-linear simulations of shattered pellet injection in ASDEX Upgrade using JOREK

This paper presents quantitative 3D non-linear JOREK simulations of shattered pellet injection in ASDEX Upgrade, demonstrating that incorporating a simplified parallel heat-flux limiting treatment resolves previous discrepancies with experimental data to enable reliable predictions of thermal quench duration and radiation fraction for ITER disruption mitigation.

The Gist

The Big Picture: Taming the Cosmic Fireball

Imagine a tokamak (like the ASDEX Upgrade machine in Germany) as a giant, super-hot donut-shaped cage holding a swirling storm of plasma. This plasma is hotter than the center of the sun. The goal is to keep it stable so we can eventually use it for clean energy.

But sometimes, the plasma gets too unstable and decides to "crash." This is called a disruption. It's like a car hitting a wall at 100 mph. If this happens in a future massive power plant (ITER), the crash could melt the walls of the machine or create dangerous beams of high-speed particles.

To stop this crash, scientists use a technique called Shattered Pellet Injection (SPI). Think of it like a fire extinguisher for a nuclear fire. Instead of shooting one big bullet of material, they fire a cannon that shatters a pellet into thousands of tiny snowflakes (fragments) right before the crash. These snowflakes fly into the plasma, cool it down, and spread the energy out safely so the machine doesn't break.

The Problem: The "Too Fast" Computer Model

For a while, the scientists had a super-computer model (using a code called JOREK) to simulate how these snowflakes would behave. But there was a glitch.

When they ran the simulation, the computer said the plasma would cool down way too fast. It was like the simulation predicted the fire extinguisher would freeze the room in a split second, whereas in real life experiments, the cooling happened more slowly and gently.

Why was the computer wrong?
The computer was using a rulebook for how heat moves through the plasma that was too optimistic. It assumed heat could zip along magnetic field lines as fast as a bullet (unlimited speed). In reality, the plasma is so dense and chaotic that heat hits a "speed limit." It's like trying to run through a crowded hallway; you can't sprint, you have to shuffle. The old model didn't account for this crowd, so it thought heat was leaving the system too quickly.

The Fix: Putting a Speed Limit on Heat

The authors of this paper realized they needed to put a speed limit on the heat in their computer model. They took the "heat flow" setting and turned it down by a factor of 10.

The Analogy:
Imagine you are pouring water (heat) from a bucket into a sink (the plasma).

• Old Model: You opened the tap fully. The water rushed out, the sink overflowed instantly, and the simulation said, "Disruption! Crash!"
• New Model: You put a flow restrictor on the tap. The water still flows, but it trickles out at a controlled pace. The sink fills up gradually, just like in real life.

By slowing down the heat flow in the simulation, the results suddenly matched the real-world experiments perfectly. The "cooling front" (the wave of coldness) moved at the right speed, and the amount of energy radiated away matched what the scientists saw on their cameras and sensors.

What They Learned: Size and Ingredients Matter

Once they fixed the "speed limit" bug, they used the now-accurate model to test two important variables:

1. The "Flavor" (Neon Content):
   They tested pellets with different amounts of neon gas mixed in.

   • Result: A little bit of neon (trace amounts) creates a long, slow cooling phase. A lot of neon (10%) creates a faster, more intense cooling phase. Both matched real experiments, giving them confidence they can tune the "flavor" to get the exact cooling speed needed for the big ITER machine.

2. The "Snowflake Size" (Fragment Size):
   They tested pellets that shattered into 53 big chunks versus 1,105 tiny chunks.

   • Result: The tiny chunks melt (ablate) very fast because they have a huge total surface area, but they might get blown away before they can do their job deep inside the plasma. The bigger chunks take longer to melt but penetrate deeper.
   • The Catch: The simulation showed that for the tiniest fragments, the real-world experiments performed even worse than the model predicted. The scientists suspect a phenomenon called the "Rocket Effect" (where the gas shooting off the back of the fragment pushes it backward, like a tiny rocket). Their current model doesn't include this rocket push yet, which is why the tiny fragments didn't penetrate as deep in the real world as the computer thought they would.

Why This Matters for the Future

This paper is a huge step forward because it moves the science from "qualitative" (looking good in a general sense) to "quantitative" (getting the exact numbers right).

• Before: "Hey, the simulation looks kind of like the experiment."
• Now: "The simulation predicts the cooling time to be 3.2 milliseconds, and the experiment was 3.1 milliseconds. We are spot on."

This accuracy is critical for ITER, the massive international fusion reactor under construction. If they get the pellet injection wrong, the reactor could be damaged. By proving that their computer model can now predict exactly how the plasma will react, they are giving engineers the confidence to design the perfect "fire extinguisher" system to keep the future of fusion energy safe and stable.

In short: They fixed the computer's "heat speed limit," making the simulation match reality. Now, they can use this super-accurate model to design the best possible safety system for the world's biggest fusion experiment.

Technical Summary

1. Problem Statement

Major disruptions in tokamaks pose severe risks to plasma-facing components and machine integrity. Shattered Pellet Injection (SPI) is the baseline mitigation strategy for the ITER tokamak, designed to rapidly radiate thermal energy and suppress runaway electrons. However, optimizing SPI requires navigating a vast parameter space (fragment size, material composition, injection timing).

A critical challenge identified in previous modeling efforts (specifically using the JOREK code) was a significant discrepancy between simulations and experiments in the ASDEX Upgrade (AUG) tokamak. Previous 3D non-linear Magnetohydrodynamic (MHD) simulations using the standard Spitzer–Härm formula for parallel thermal diffusivity predicted a pre-thermal quench (pre-TQ) phase that was too short compared to experimental observations. The standard model assumed a collision-dominated plasma, likely overestimating parallel heat flux and leading to artificially rapid thermal energy loss and exaggerated MHD instabilities.

2. Methodology

The authors revisited SPI simulations of AUG H-mode discharges (specifically shot #40355 and #40673) using the JOREK code. The study employed a two-temperature reduced MHD model with the following key methodological adjustments:

• Flux-Limited Heat Transport: Instead of the standard Spitzer–Härm diffusivity ($\chi_{\parallel, SH}$), the authors introduced a simplified treatment of parallel heat-flux limiting. They reduced the parallel thermal diffusivity by a constant factor of 10 ($\chi_{\parallel} = 0.1 \cdot \chi_{\parallel, SH}$). This adjustment mimics the saturation of heat flux observed in kinetic simulations and experiments, where ambipolar electric fields limit energy loss to values far below the free-streaming limit.
• Simulation Parameters:
  • Equilibrium: Reconstructed from AUG discharge #40355 ($B_t = 1.8$ T, $I_p = 0.8$ MA).
  • Injection Scenarios: Simulations covered varying neon fractions (0.12% trace to 10%) and fragment sizes (53 vs. 1105 fragments).
  • Physics Models: Used the Neutral Gas Shielding (NGS) model for ablation and OpenADAS for impurity atomic processes. No background impurities or initial magnetic perturbations were included to isolate the effects of the injected material.
• Comparison Strategy: Simulation results (thermal energy, plasma current, radiation fraction) were directly compared against experimental data from the 2022 AUG campaign, with time-shifting applied to align fragment arrival times.

3. Key Contributions

• Resolution of Simulation-Experiment Discrepancies: The paper successfully bridges the gap between previous JOREK simulations and experimental data by identifying parallel heat-flux limiting as the missing physical ingredient.
• Quantitative Validation: The work moves beyond qualitative comparisons to achieve quantitative agreement regarding pre-TQ duration and radiation fractions.
• Mechanistic Insight: The study elucidates the physical mechanisms by which overestimated heat transport distorts disruption dynamics, specifically linking excessive parallel heat flux to rapid plasmoid heating, charge separation, and the amplification of resistive MHD instabilities (e.g., $2/1$ tearing modes).
• Parameter Space Re-evaluation: The authors re-assessed the impact of neon concentration and fragment size using the corrected physics model, providing new insights for ITER optimization.

4. Key Results

A. Impact of Reduced Parallel Thermal Diffusivity

• Pre-TQ Duration: Reducing $\chi_{\parallel}$ by a factor of 10 extended the pre-TQ duration, bringing it into quantitative agreement with experimental observations (e.g., ~1 ms for 10% neon, ~3–4 ms for 0.12% neon).
• MHD Stability: The standard Spitzer–Härm model caused an artificially strong MHD response. The reduced diffusivity resulted in:
  • Weaker growth of $n=1$ and $n=2$ magnetic harmonics.
  • Preservation of core flux surfaces for longer.
  • A more gradual cooling front that tracks the fragment trajectory rather than propagating ahead of it.
• Radiation Fraction: The corrected model yielded a significantly higher radiation fraction during the initial cooling stage, matching experimental diagnostics. The standard model underestimated radiation because it assumed energy was lost too quickly via conduction/convection before radiation could dominate.

B. Influence of Neon Content

• Two-Stage Cooling: The simulations confirmed a two-stage cooling process: an initial phase dominated by conduction/convection (due to magnetic stochasticity) followed by a radiation-dominated phase.
• Neon Fraction Effects: Lower neon fractions (0.12%) resulted in longer pre-TQ durations and slower TQ processes, leading to higher electron density assimilation. This is beneficial for runaway electron suppression due to increased collisionality.

C. Influence of Fragment Size

• Ablation vs. Penetration: Smaller fragments (1105 count) have a larger total surface area, leading to faster initial ablation. However, larger fragments (53 count) exhibited slightly higher total assimilation by the end of the TQ because their longer pre-TQ duration allowed more time for material deposition.
• Discrepancy in Trace Neon: In experiments, smaller fragments with trace neon showed significantly lower assimilation than larger ones, a trend not fully captured by the current model. The authors attribute this to the rocket effect (momentum transfer from asymmetric ablation), which is not yet implemented in JOREK but is crucial for small fragment penetration.

5. Significance

This work represents a critical step toward predictive modeling for ITER's Disruption Mitigation System (DMS).

• Reliability: By incorporating a simplified flux-limiting model, the JOREK code can now reliably predict key disruption metrics (thermal quench duration, radiation fraction) without degrading agreement in other areas.
• Optimization: The validated model provides a robust tool for optimizing SPI parameters (fragment size, composition) to minimize thermal loads, electromagnetic forces, and runaway electron risks in ITER.
• Future Directions: The study highlights the necessity of implementing more sophisticated physics, such as a realistic flux-limited conduction model and the rocket effect, to further refine the modeling of small fragment penetration and improve the accuracy of disruption mitigation strategies.