Year
2019
Abstract
The detection of shielded special nuclear material (SNM) presents a challenge to passive inspection systems; however,active photon interrogation can be used to detect shielded SNM due to the ability of photons to penetrate low-Zmaterial traditionally used as neutron shielding. The high energy interrogation photons can penetrate shielding andinduce neutron emissions in SNM via photonuclear reactions. The neutron emissions can be detected, but the intensephoton flash from the source outnumbers the emitted neutrons by orders of magnitude. Pile-up will be likely, resultingin the loss of neutron information. This effect must be mitigated to overcome a substantial active background, and toensure timely inspections. We use the recently-licensed linear accelerator (linac) facility at the University of Michiganto evaluate the fast neutron detection capability of stilbene and to test a neural network-based algorithm that canseparate piled-up detector signals to recover pulse information. The Varian M9 linac installed in the facility has anendpoint energy of 9 MeV, which is sufficient to induce photonuclear reactions in materials such as uranium and lead.Due to photonuclear reaction rates, the irradiation will result in a relatively small amount of fast neutron emissions ina high-photon field. To provide a high-confidence benchmark for our experiment, we use a helium-3 detector that isgamma-blind and filters incoming thermal neutrons to be sensitive to only fast neutrons. By comparing detection ratesand pile-up rejection rates while varying distance from the irradiation target, we determine which detectorconfiguration(s) are ideal for future active interrogation experiments. We also test the neural network algorithm forrecovery of neutron information from pile-up events. This work benefits safeguards techniques, especiallymeasurements in high-count-rate environments such as fresh fuel assay using americium-lithium sources and photonactive interrogation using pulsed sources.