γ-ray Reconstruction Using Neural Networks

Measuring γ-rays from nuclear decay is important for distinguishing different models describing the photon strength function. Because γ-rays often scatter from one detector into another within an array, one must reconstruct the actual energy of the initial γ-ray from the energies deposited in multiple detectors. Current γ-ray emission data is reconstructed using modern day addback-clustering algorithms, however due to scattering and cross-talk in detectors, the algorithms can only reconstruct a fraction of the events accurately. This research investigates the potential use of neural networks to reconstruct gamma energy and multiplicity using simulated detector data originating from DICEBOX. These results are then compared to the performance of the current clustering algorithms. If the neural network outperforms the clustering algorithms and is able to identify previously unknown patterns in the data, then it may replace the clustering algorithms to reconstruct γ-ray emissions. Excited states of 58Fe made in DICEBOX in conjunction with GEANT4 to create simulated events were used to train and test various different networks.