A Neural Network Approach for Online Reconstruction of Bremsstrahlung Spectra Produced by Electron Accelerator

The characterization of bremsstrahlung spectra generated by electron accelerators is becoming increasingly crucial, particularly in radiation processing applications such as sterilization of medical devices or food irradiation. The growing transition from isotopic to electric irradiators presents new challenges related to the control of beam properties. In this context, the technology resource center Aerial in collaboration with CEA/IRESNE and IPHC is looking to develop a tool and methodology enabling the online characterization of the bremsstrahlung spectra generated in its feerix[1] installation. This step is very important for their irradiation operations to ensure precise dose deposition in the sample and to precisely estimate the activation product when the photon energy exceeds the photonuclear reaction threshold. Information on the energy spectrum is also a key input for Monte Carlo simulations, which are increasingly used in radiation processing. However, conventional direct and indirect spectrometry methods are limited in meeting the challenges of Aerial’s high energy and high-power irradiation platform. In this study, we propose a new theoretical approach based on neural networks to solve an ill-posed inverse problem, enabling the reconstruction of bremsstrahlung spectra from depth-dose measurements. This approach is motivated by the limitations of previously discussed regularization methods and existing neural network approaches. We focus here on an analytical approach for generating realistic training and validation datasets, consisting of Bremsstrahlung spectra and their corresponding dose distributions in any medium. This neural network approach will also be compared with other methods reported in the literature.