[Paper Review] Validation of Compton Scattering Monte Carlo Simulation Models

G. Weidenspointner, Matej Batič+5 more|arXiv (Cornell University)|Jan 8, 2014

Electron and X-Ray Spectroscopy Techniques13 references3 citations

TL;DR

This study validates Compton scattering Monte Carlo models using experimental data, evaluating their accuracy in simulating total and differential cross sections. The Klein-Nishina model in Geant4 underperforms due to neglecting electron binding effects, while EPDL- and Penelope-based models show superior agreement with data, achieving 82–85% compatibility with experimental results at 1% significance level.

ABSTRACT

Several models for the Monte Carlo simulation of Compton scattering on electrons are quantitatively evaluated with respect to a large collection of experimental data retrieved from the literature. Some of these models are currently implemented in general purpose Monte Carlo systems; some have been implemented and evaluated for possible use in Monte Carlo particle transport for the first time in this study. Here we present first and preliminary results concerning total and differential Compton scattering cross sections.

Motivation & Objective

• To objectively assess the physical accuracy of Compton scattering simulation models in general-purpose Monte Carlo systems.
• To identify discrepancies between theoretical models and experimental data, particularly regarding electron binding effects.
• To evaluate the performance of models implemented in Geant4 9.6, including standard, low-energy, and alternative approaches.
• To quantify model compatibility with experimental data using statistical hypothesis testing.
• To provide a foundation for future validation of additional Compton scattering features such as polarization and shell effects.

Proposed method

• Collected and compiled a large dataset of experimental Compton scattering cross sections from the literature for validation.
• Applied a χ² goodness-of-fit test with significance level α = 0.01 to compare model predictions against individual experimental data points.
• Defined model efficiency as the fraction of test cases passing the χ² test (p ≥ 0.01), with total efficiency being the mean across all test cases.
• Evaluated models for total cross sections (one test case per element) and differential cross sections (one test case per energy-angle combination).
• Compared multiple models: Klein-Nishina (standard Geant4), EPDL-based, Penelope-based, Brusa, Biggs, and Hubbell scattering functions.
• Performed unit tests using alternative scattering functions and tabulated data from EPDL and other sources to ensure consistency.

Experimental results

Research questions

• RQ1 How well do standard Compton scattering models in Geant4 reproduce experimental total and differential Compton scattering cross sections?
• RQ2 To what extent do models that account for electron binding effects (e.g., EPDL, Penelope) outperform the free-electron Klein-Nishina model?
• RQ3 What is the statistical compatibility of different scattering function implementations with experimental data at α = 0.01?
• RQ4 Are there systematic discrepancies between experimental datasets or model predictions, particularly at low scattering angles or low energies?
• RQ5 How do code-level design issues such as duplication affect model reliability and validation outcomes?

Key findings

• All tested models for total Compton scattering cross sections achieved 100% efficiency, indicating full compatibility with experimental data at α = 0.01.
• The Klein-Nishina-based model in Geant4 showed significantly lower performance, with an efficiency of 54% ± 3% for differential cross sections, indicating systematic overestimation, especially at low angles.
• EPDL-based and Penelope-based models achieved 82% ± 2% efficiency, demonstrating strong agreement with experimental data.
• Models based on Brusa, Biggs, and Hubbell scattering functions achieved 84–85% ± 2% efficiency, indicating high compatibility with experimental measurements.
• Visual comparisons (Fig. 3) show that Klein-Nishina overestimates differential cross sections, while EPDL and Penelope models closely follow experimental trends.
• Discrepancies between experimental datasets were observed, suggesting potential outliers or systematic biases requiring further critical appraisal.