Abstract
This thesis presents the development and validation of a novel method for identifying illicit materials based on active photon interrogation coupled with photoneutron spectrometry (API-PS). The approach relies on inducing photo-nuclear reactions in concealed materials using a 23 MV LINAC and analyzing the energy spectra of the emitted neutrons with liquid organic scintillators and dedicated unfolding algorithms. A full experimental test bench was designed, including shielding, synchronization electronics, calibration procedures, and custom software. We successfully separated neutrons from gamma rays in harsh, pulsed, mixed-field environments, and unfolded neutron spectra from benchmark targets (graphite, glucose, melamine) to identify the distinct nuclear signatures of carbon, oxygen, and nitrogen. These signatures were then retrieved in realistic inspection scenarios involving suitcases and crates containing nitrogen-rich materials such as melamine. Finally, we introduced DeepNSI, a deep learning framework trained on experimental and simulated spectra, capable of identifying light elements and reconstructing neutron energy distributions. This proof-of-concept demonstrates that neutron spectrometry can be used not only to detect but also to characterize threat materials in complex inspection environments. The method opens new prospects for security screening, but also for applications in medical dosimetry, nuclear waste management, and potentially in environmental sciences such as lithium exploration.
Key words
Active Photon Interrogation, Photoneutron Spectrometry, Illicit Material Detection, Organic Scintillators, Neutron-Gamma Discrimination, Deep Learning for Nuclear Applications
Full text
Full text: https://theses.fr/2025UPASP081