Gd-containing scintillators for thermal neutron detection via graph-based particle discrimination

Area of Science:

• Nuclear instrumentation
• Materials science
• Data science

Background:

• Neutron-gamma discrimination (NGD) is crucial for efficient neutron detection but challenging with gadolinium-containing scintillators.
• Existing methods struggle with high signal-to-noise ratios due to noise and data complexity.
• Pulse-shape based digital-signal processing offers a path to mitigate these issues.

Purpose of the Study:

• To develop advanced digital classifiers for enhanced NGD in gadolinium scintillators.
• To reduce data dimensionality and mitigate noise and outliers for accurate classification.
• To achieve high thermal neutron detection efficiency and NGD ratio simultaneously.

Main Methods:

• Two graph-embedded non-negative matrix factorization (NMF) digital classifiers were developed.
• An event graph (network) was introduced to uncover low-dimensional structures in high-dimensional experimental data.
• Energy-independent normalized waveform features and mixed-source reference events were utilized.

Main Results:

• The proposed algorithms achieved high thermal neutron detection efficiency (77%-80%) with a high NGD ratio (∼10⁹).
• Excellent discrimination between neutron events and ambient background events was demonstrated.
• A smooth or sparse low-dimensional representation of detector signals was obtained from a Ce:Gd₃Al₂Ga₃O₁₂ scintillator.

Conclusions:

• Graph-embedded NMF algorithms effectively address NGD challenges in gadolinium scintillators.
• The methods enable simultaneous high detection efficiency and superior neutron-gamma discrimination.
• These algorithms have broad applicability in various radiation detection systems and applications.