Decision trees for optimizing the minimum detectable concentration of radioxenon detectors

Area of Science:

• Nuclear Physics
• Data Science

Background:

• Radiation detection systems are crucial for monitoring radioactive isotopes.
• Distinguishing radioxenon decays from radon-progeny backgrounds is challenging.
• Traditional methods often involve physical radon removal, which is not assumed here.

Purpose of the Study:

• To apply machine learning for optimizing radiation detection system design.
• To evaluate the effectiveness of detector design choices in identifying radioxenon decays amidst radon interference.

Main Methods:

• A decision tree-based algorithm was developed for greedy optimization.
• The algorithm partitions energy depositions using a minimum detectable concentration metric.
• The method was applied to simulated radioxenon decay detection with radon-progeny backgrounds.

Main Results:

• The study found that high-resolution readout and spatial segmentation provided limited improvement.
• Simpler detector designs were found to be nearly as effective as complex ones.
• The machine learning approach successfully optimized the partitioning of energy depositions.

Conclusions:

• Sophisticated detector designs offer marginal benefits for radioxenon detection against radon backgrounds.
• Machine learning provides an effective framework for optimizing radiation detector design parameters.
• The findings suggest focusing on algorithmic improvements over hardware complexity for specific detection challenges.