Physics-based method for generating probability table using random-matrix approach

This paper presents a physics-based method for generating probability tables by calculating fluctuating cross sections using a Gaussian Orthogonal Ensemble $S$-matrix model, validating the approach with $^{238}$U and $^{239}$Pu to show qualitative agreement with conventional Breit-Wigner formalism while accounting for statistical uncertainties and $S$-matrix unitarity.

The Gist

Imagine you are trying to predict how a crowd of people will move through a hallway. If the hallway is empty and wide, people walk smoothly and predictably. But if the hallway is packed with obstacles (like furniture or other people), the flow becomes chaotic. Some people get stuck, some speed up, and the path becomes unpredictable.

In the world of nuclear physics, neutrons are the people, and atomic nuclei (like Uranium or Plutonium) are the crowded hallways. When neutrons hit these nuclei, they don't just bounce off smoothly; they get caught in a chaotic dance of "resonances" (temporary traps).

This paper introduces a new, more reliable way to map out this chaotic dance, specifically for the middle ground where the chaos is too messy to track individual steps but too wild to be perfectly smooth.

Here is the breakdown of their work using simple analogies:

1. The Problem: The "Unresolved" Zone

Physicists have two main ways to describe how neutrons interact with nuclei:

• The Low-Energy Zone (Resolved): Here, the "obstacles" are far apart. You can see each one clearly, like individual trees in a forest. You can measure them one by one.
• The High-Energy Zone (Smooth): Here, the obstacles are so close together that they blur into a solid wall. You can't see individuals, so you just measure the average thickness of the wall.
• The Middle Zone (The Unresolved Resonance Region): This is the messy middle. The obstacles are overlapping. You can't see them individually, but the wall isn't smooth yet; it's bumpy and fluctuating.

Currently, to predict how neutrons behave in this messy middle zone, scientists use a method called SLBW (Single-Level Breit-Wigner). Think of this like trying to predict traffic by assuming every car drives at the exact same speed and never crashes. It's a useful simplification, but it has a flaw: sometimes, the math says cars are driving backwards (negative numbers), which is impossible in real life. This breaks the "rules of the road" (a concept physicists call unitarity).

2. The Solution: The "Random Matrix" Approach

The authors developed a new method using something called the GOE-S-matrix model.

• The Analogy: Imagine you want to predict the outcome of a massive, chaotic game of pinball. Instead of trying to calculate the path of every single ball (which is too hard), you use a giant, computer-generated "Random Matrix."
• How it works: This matrix is like a bag of marbles with specific rules. You pull out random numbers (representing the chaotic energy levels inside the nucleus) that follow a strict statistical pattern known as the Gaussian Orthogonal Ensemble (GOE).
• The Magic: By using this random-matrix approach, the authors can calculate the "bumpy" cross-sections (how likely a neutron is to hit or be absorbed) without ever needing to assume specific distributions for the chaos. Crucially, this method guarantees the rules of the road are followed. It never produces impossible "negative" results. It respects unitarity, meaning the total probability of everything happening always adds up to 100%.

3. The Process: Building a "Probability Table"

In nuclear reactors, engineers need a "cheat sheet" called a Probability Table. Since they can't know exactly where every neutron will go, this table tells them: "At this energy level, there is a 10% chance the neutron hits a big bump, a 50% chance it hits a medium bump, and a 40% chance it hits a small bump."

The authors did the following:

1. Simulated the Chaos: They used their new Random Matrix method to simulate millions of "ladders" (different possible scenarios of how the resonances could be arranged).
2. Found the Sweet Spot: They tested different sizes of their simulation (changing the number of "levels" or "ladders"). They found that using a specific, moderate size (25 levels) and focusing on the center of the energy range gave them the most accurate results without taking too much computer power.
3. Checked the Results: They compared their new tables against the old "SLBW" method.
   • The Result: The new tables looked very similar to the old ones in the big picture.
   • The Improvement: The new method didn't have the "negative number" glitches. It also handled the "lumped" channels (like capture and fission) more realistically, treating them as complex multi-channel processes rather than simple single-lane roads.

4. The Conclusion

The authors successfully built a new, physics-based engine to generate these probability tables.

• Why it matters: It's more solid theoretically because it doesn't rely on shaky assumptions about how the chaos is distributed.
• The Trade-off: It requires a bit more computer power to run the random matrix simulations, but the authors found a "Goldilocks" setting (25 levels) that is accurate enough without being too slow.
• The Bottom Line: They have proven that you can generate these essential nuclear data tables using a rigorous random-matrix approach that respects the fundamental laws of physics (unitarity), offering a cleaner, more reliable alternative to the traditional methods.

In short, they replaced a "best guess" map of a chaotic city with a mathematically guaranteed map that never tells you a street goes backwards.

Technical Summary

Technical Summary: Physics-based Method for Generating Probability Tables Using a Random-Matrix Approach

Problem Statement
In the Unresolved Resonance Region (URR) of neutron-induced reactions, typically occurring in the keV energy range for actinides, individual resonances overlap and become experimentally indistinguishable. While the Resolved Resonance Region (RRR) and the fast energy region are well-described by R-matrix theory and Hauser-Feshbach theory respectively, the URR presents a challenge. Conventional methods for generating probability tables in this region—essential for self-shielding calculations in reactor analyses—rely on simplified assumptions. Specifically, the Single-Level Breit-Wigner (SLBW) formalism combined with statistical distributions (Wigner for level spacing and $\chi^2$ for widths) suffers from deficiencies: it fails to properly account for interference between resonances, requires prior knowledge of statistical distributions, and, crucially, lacks unitarity, which can lead to unphysical negative cross sections. The paper addresses the need for a method to generate probability tables in the URR based on a solid theoretical foundation that avoids these simplifications.

Methodology
The authors develop a new method for generating probability tables using the Gaussian Orthogonal Ensemble (GOE) incorporated into the scattering matrix (S-matrix), referred to as the GOE-S-matrix model.

• Theoretical Framework: Unlike conventional approaches that assume statistical distributions for resonances, the GOE-S-matrix model derives cross sections from a time-reversal invariant real symmetric Hamiltonian ($H^{(GOE)}$) where elements are randomly sampled from a Gaussian distribution. The S-matrix is calculated analytically or via Monte Carlo techniques, incorporating transmission coefficients and a phase shift to map the model onto experimental cross sections.
• Input Parameters: The model utilizes transmission coefficients calculated via the optical model (CoH3 code) for elastic scattering, giant dipole resonance models for capture, and the Hill-Wheeler model for fission. It treats inelastic scattering channels explicitly based on excited states and partial waves.
• Probability Table Generation: The calculated fluctuating cross sections are converted into probability tables using the same binning and averaging procedures as the NJOY nuclear data processing code. This involves defining boundary cross sections and calculating the probability ($P_j$) and average cross section ($\bar{\sigma}_j$) for each bin.
• Validation Strategy: The study uses $^{238}\text{U}$ and $^{239}\text{Pu}$ as target nuclei. The authors determine optimal model parameters (specifically the number of levels $n$ and the energy range $E_\lambda$) by analyzing the convergence of average cross sections. They compare their results against:
  1. Evaluated nuclear data libraries (JENDL-5, ENDF/B-VIII.0, JEFF-3.3).
  2. Probability tables generated using the conventional SLBW formalism at 0 K to isolate model-based differences from temperature effects.
• Uncertainty Quantification: The statistical uncertainty of the probability tables is evaluated using the Root Mean Square Percentage Error (RMSPE) of the product of probability and average cross sections as a function of the number of ladders ($L$).

Key Contributions and Results

1. Parameter Optimization: The study identifies that restricting the GOE eigenvalue range to the "quarter $E_\lambda$ range" ($E_\lambda = -\lambda/2$ to $\lambda/2$), where the level density is approximately constant, yields better agreement with reference data than using the full semicircular range. For $^{238}\text{U}$, this reduces deviations in capture cross sections from >20% to <1%. A matrix dimension of $n=25$ is determined to be sufficient for convergence, balancing accuracy with computational cost.
2. Statistical Convergence: The RMSPE analysis demonstrates that statistical uncertainty decreases as the number of ladders ($L$) increases. For both target nuclei, the RMSPE drops below 5% at $L=1,000$ and approaches 1% at $L=10,000$.
3. Comparison with SLBW:
   • Qualitative Agreement: The probability tables generated by the GOE-S-matrix model are qualitatively comparable to those from the SLBW formalism.
   • Unitarity: A critical distinction is the preservation of unitarity in the GOE-S-matrix model. The SLBW approach, lacking unitarity, can produce negative cross sections (which are artificially replaced by $1 \mu\text{b}$ in NJOY), leading to unphysically low probabilities in certain bins. The GOE-S-matrix model avoids this entirely.
   • Fluctuations: The GOE-S-matrix model exhibits stronger fluctuations in probability, particularly at low and high cross-section values, due to event-to-event variations in resonance amplitudes.
   • Channel Treatment: The model successfully treats inelastic scattering on a channel-by-channel basis, whereas the SLBW approach often cannot calculate inelastic cross sections for nuclei like $^{239}\text{Pu}$ due to missing partial width data in evaluated libraries.

Significance and Claims
The paper claims to have established a reliable, physics-based method for generating probability tables in the URR that is grounded in random-matrix theory rather than simplified resonance formulas. The primary significance lies in the model's ability to guarantee unitarity and provide a more realistic treatment of lumped channels (capture and fission) and individual inelastic scattering channels without relying on pre-assumed statistical distributions for level spacings or widths.

The authors modestly conclude that while the method produces results comparable to the conventional SLBW formalism, the preservation of unitarity and the handling of channel-specific physics represent a theoretical improvement. They note that local differences exist, particularly in the magnitude of probability spikes, and identify future work as incorporating temperature effects and refining the determination of boundary cross sections to reduce distribution fluctuations. The study does not claim immediate replacement of existing libraries but presents a robust theoretical alternative for generating the underlying probability data.