Analysis of Fission Matrix Databases using Temperature Profiles obtained from High-Fidelity Multiphysics Simulations

This paper demonstrates that utilizing temperature profiles derived from high-fidelity Multiphysics simulations, rather than uniform profiles, to construct Fission Matrix databases significantly improves the accuracy of multiplication factor and fission source distribution predictions for Molten Salt Fast reactors.

The Gist

The Big Picture: Predicting a Nuclear Reactor's Mood

Imagine you are trying to predict how a nuclear reactor behaves. The reactor is like a giant, complex engine where fuel is constantly heating up and cooling down. To keep it safe and efficient, engineers need to know exactly how the heat is distributed inside the core.

The problem is that the most accurate way to calculate this (called the Monte Carlo method) is like trying to count every single grain of sand on a beach by picking them up one by one. It's incredibly accurate, but it takes so much time and computer power that you can't do it quickly enough to react to changes in real-time.

This paper introduces a "shortcut" called the Fission Matrix (FM) method. Think of this as a cheat sheet or a lookup table. Instead of doing the hard math every time, you pre-calculate a bunch of scenarios and store them in a database. When you need an answer, you just look up the closest match and interpolate (guess) the result. It's fast and cheap.

The Problem: The "Uniform" Assumption Was Wrong

In the past, scientists made these cheat sheets (called Fission Matrix Databases or FMDBs) assuming the reactor fuel was the same temperature everywhere, like a loaf of bread baked evenly in an oven.

However, real reactors (specifically the Molten Salt Fast Reactor, or MSFR) are more like a stirred pot of soup. The liquid fuel flows in, gets hot, swirls around, and cools down in different spots. The temperature isn't uniform; it has hot spots and cool spots.

The authors realized that using a "uniform" cheat sheet for a "swirly soup" reactor was giving them slightly wrong answers. They wanted to see if making a cheat sheet based on the actual swirling temperature patterns would make the predictions better.

The Solution: The "Cardinal" Super-Computer Chef

To get the real temperature patterns, the team used a sophisticated software tool called Cardinal.

• The Analogy: Imagine Cardinal is a master chef who runs a kitchen simulation.
  • First, it simulates the neutronics (the nuclear reactions) to see where the heat is being generated.
  • Then, it passes that heat data to a fluid dynamics simulator (NekRS) to see how the liquid fuel moves and cools.
  • The chef sends the new temperature data back to the nuclear simulator, which updates the heat generation.
  • They repeat this loop thousands of times until the "soup" settles into a realistic, steady state.

This process gave them a highly accurate map of the temperature inside the reactor, complete with all the swirls and hot spots.

The Experiment: Two Cheat Sheets

The team created two different "cheat sheets" (databases) to test their theory:

1. The Uniform Sheet: Made using the old-school assumption that the fuel is the same temperature everywhere.
2. The Cardinal Sheet: Made using the complex, realistic temperature maps generated by the Cardinal simulation.

They then used both sheets to predict the reactor's behavior in two different test scenarios (changing the inlet temperature and the flow rate).

The Results: Accuracy Wins

When they compared the predictions against the "gold standard" (the slow, super-accurate Monte Carlo method), the results were clear:

• The Uniform Sheet was okay, but it had noticeable errors. It was like trying to navigate a city using a map that assumes all streets are straight lines, ignoring the curves and traffic.
• The Cardinal Sheet was much closer to the truth. Because the cheat sheet was built using realistic temperature patterns, the predictions for the reactor's power and safety were significantly more accurate.

The Metaphor:
Imagine you are trying to guess the weather in a city.

• Method A (Uniform): You assume the temperature is exactly 70°F everywhere in the city.
• Method B (Cardinal): You use a weather map that shows it's 60°F by the river, 80°F in the city center, and 65°F in the hills.
• The Result: If you need to know if it's going to rain in the city center, Method B gives you the right answer, while Method A might get it wrong because it ignored the local details.

Conclusion

The main takeaway is simple: If you want a fast and accurate prediction of a nuclear reactor, your "cheat sheet" needs to look like the real thing.

By using high-fidelity simulations to create databases that match the complex, swirling temperature profiles of a liquid-fuel reactor, the team proved they can get much better results without waiting days for a supercomputer to finish the calculation. This is a big step toward making nuclear reactors safer and easier to design.

Technical Summary

1. Problem Statement

Monte Carlo (MC) methods are the gold standard for neutronic calculations due to their use of continuous-energy cross-sections and exact geometry modeling. However, they suffer from high computational costs when low uncertainty is required, making them inefficient for transient calculations or scenarios requiring temperature feedback (coupled neutronics/thermal-hydraulics).

The Fission Matrix (FM) method offers a faster alternative by utilizing pre-calculated databases (FMDBs) generated from MC simulations. However, the accuracy of the FM method depends heavily on the temperature profiles used to generate these databases. Previous studies utilized simplified uniform or non-uniform temperature profiles that did not accurately reflect the complex thermal-hydraulic behavior of Molten Salt Fast Reactors (MSFR), specifically regarding internal recirculation and stagnation zones. This discrepancy can lead to errors in the multiplication factor ($k_{eff}$) and fission source distribution.

2. Methodology

The authors developed a workflow to generate high-fidelity temperature profiles and evaluate their impact on FMDB accuracy.

• Multiphysics Coupling (Cardinal):

  • The study utilized Cardinal, an open-source application wrapping OpenMC (neutronics) and NekRS (thermal-hydraulics) within the MOOSE framework.
  • OpenMC Model: A simplified cylindrical MSFR geometry using liquid fuel (LiF-ThF4-233UF4) and a fertile blanket. It employs a $k$-eigenvalue calculation with the ENDF/B-VIII.0 library.
  • NekRS Model: Solves fluid mass, momentum, and energy conservation using a RANS $k-\tau$ turbulence model. It receives volumetric heat sources from OpenMC and calculates fluid temperature and velocity distributions.
  • Coupling Scheme: A Picard iteration "in time" loop was used. OpenMC calculates the heat source $\rightarrow$ NekRS calculates temperature/velocity $\rightarrow$ Cardinal updates fuel density and temperature $\rightarrow$ OpenMC recalculates. This continues until steady-state convergence.

• Fission Matrix Database (FMDB) Generation:

  • Serpent 2.2 was used to generate FMDBs based on specific temperature profiles.
  • Two distinct sets of FMDBs were created:
    1. Cardinal-based FMDB: Generated using complex, non-uniform temperature profiles derived from the Cardinal multiphysics simulations (varying inlet temperatures from 700 K to 1300 K).
    2. Uniform-based FMDB: Generated using simplified uniform temperature profiles (800 K to 1200 K in 100 K steps).
  • The databases used a $13 \times 13 \times 13$ Cartesian discretization.

• Interpolation and Solving:

  • A C++ script solved the FM equations. For a given target temperature profile, the fission matrix elements ($a_{ij}$) were linearly interpolated between the nearest FMDB entries based on cell temperatures.
  • The fundamental eigenvector ($F$) and eigenvalue ($k_{eff}$) were solved to determine the fission source distribution and reactivity.

3. Key Contributions

• High-Fidelity Profile Integration: The paper demonstrates the integration of high-fidelity, time-averaged multiphysics temperature profiles (capturing complex MSFR flow dynamics) into FMDB generation, moving beyond simplified uniform assumptions.
• Quantitative Comparison: It provides a rigorous comparison between FM results derived from "realistic" (Cardinal) profiles versus "simplified" (Uniform) profiles against a high-fidelity Serpent reference.
• Error Characterization: The study quantifies how the mismatch between the database generation profile and the target application profile affects $k_{eff}$ and the spatial distribution of the fission source.

4. Results

The study tested two specific temperature profiles (Test Profile 1: 898 K inlet; Test Profile 2: 1700 kg/s mass flow rate) against both FMDB sets.

• Multiplication Factor ($k_{eff}$):

  • Using Cardinal-based FMDBs resulted in significantly lower errors compared to the Serpent reference.
    • Test Profile 1: Difference of -3.1 pcm (vs. -23.9 pcm for Uniform).
    • Test Profile 2: Difference of 7.4 pcm (vs. -14.7 pcm for Uniform).
  • The Cardinal-based approach reduced the bias in $k_{eff}$ by approximately 20 pcm in the tested scenarios.

• Fission Source Distribution:

  • RMS Difference: Both methods showed similar RMS differences (~1.8%), indicating general agreement in the overall shape.
  • Maximum Difference: The Cardinal-based FMDBs showed lower maximum deviations (17.4% - 18.5%) compared to the Uniform-based FMDBs (22.8% - 24.6%).
  • Spatial Bias: Analysis of the difference maps (Figures 6 and 7) revealed that the Uniform-based FMDB introduced a systematic spatial error: negative differences at the bottom of the core and positive differences at the top. This was attributed to the inability of uniform profiles to capture the specific axial temperature gradients present in the MSFR.

5. Significance

• Accuracy Improvement: The study confirms that generating FMDBs using temperature profiles that closely mimic the actual reactor operating conditions (obtained via high-fidelity multiphysics) significantly improves the accuracy of the Fission Matrix method.
• Operational Relevance: For MSFRs, where internal recirculation creates complex, non-uniform temperature fields, relying on uniform temperature databases introduces systematic errors in reactivity and power distribution predictions.
• Future Application: The methodology validates the use of Cardinal-coupled simulations to create robust FMDBs, enabling faster, lower-cost neutronic calculations that retain high fidelity for transient analysis and temperature feedback studies in liquid-fueled reactors.

In conclusion, the paper establishes that the fidelity of the temperature profile used to generate the Fission Matrix Database is a critical factor in minimizing errors in $k_{eff}$ and fission source distribution, advocating for the use of high-fidelity multiphysics data over simplified assumptions.