A GEANT4-Based Simulation of Directional Neutron Detectors Using Liquid Scintillators and Boron Carbide Moderators

TL;DR

This paper presents a GEANT4 simulation of a compact, directional neutron detector using liquid scintillators, boron carbide moderators, and machine learning for source direction classification, demonstrating high accuracy and robustness.

Contribution

The study introduces a novel simulation framework for a directional neutron detector with optimized geometry and machine learning-based directionality assessment.

Findings

Detection efficiency between 10% and 30%.

Machine learning classifier achieved 100% accuracy.

Detector effective even near edges.

Abstract

We present a simulation-based study of a compact directional neutron detector composed of liquid scintillator, Cesium Iodide, with boron carbide (B4C) moderation, and silicon photomultipliers (SiPMs). Using GEANT4, we explored multiple detector geometries and material configurations, finding neutron detection efficiencies ranging from approximately 10% to 30%. To evaluate directionality, spatial energy distributions were analyzed and used to train a machine learning classifier, which achieved 100% accuracy in identifying neutron source directions along four cardinal axes. The model remained effective for sources near detector edges, demonstrating robustness. These results establish the feasibility of the proposed detector for applications in nuclear safety, environmental monitoring, and scientific research applications, with future work focused on experimental validation.