Impact of thermal and intermediate energy neutrons on the semiconductor memories for the CERN accelerators

A wide quantity of SRAM memories are employed along the Large Hadron Collider (LHC), the main CERN accelerator, and they are subjected to high levels of ionizing radiations which compromise the reliability of these devices. The Single Event Effect (SEE) qualification for components to be used in the complex high-energy accelerator at CERN relies on the characterization of two cross sections: 200-MeV protons and thermal neutrons. However, due to cost and time constraints, it is not always possible to characterize the SEE response of components to thermal neutrons, which is often regarded as negligible for components without borophosphosilicate glass (BPSG). Nevertheless, as recent studies show, the sensitivity of deep sub-micron technologies to thermal neutrons has increased owing to the presence of Boron 10 as a dopant and contact contaminant. The very large thermal neutron fluxes relative to high-energy hadron fluxes in some of the heavily shielded accelerator areas imply that even comparatively small thermal neutron sensitivities could dominate the overall Single Event Upset (SEU) rate. For instance, in some locations that host electronic devices, the thermal neutrons fluence can be up to 15 times larger than that of the high-energy hadron. For this reason, in this work I explore the option of measuring the thermal neutron sensitivity through high-energy mixed field irradiations, in conditions with different ratios between the thermal and high-energy hadron fluxes. I studied SEU and SEL SRAM cross sections using the Cern High energy AcceleRator Mixed-field (CHARM) facility, where a wide variety of accelerator environments can be reproduced by combining different test positions and shielding configurations. The mixed-field radiation environment simulated in the past with the FLUKA Monte Carlo tool was investigated in order to select a location with a large thermal neutron fluence compared with the equivalent high-energy hadron fluence. From this perspective, I selected a location with a strong contribution from thermal and intermediate energy neutrons and I characterized it, by combining FLUKA Monte Carlo simulations and the Radiation Monitor (RadMon) measurements. In order to vary the amount of thermal neutrons in the selected position, I designed and built a box of boron carbide, a material that has a high capture cross section for thermal neutrons. After preliminary FLUKA simulations to outline the spectra outgoing the boron carbide absorber, I carried out the RadMon tests-analysis with the purpose of experimentally verifying the results. The former evidences that a large portion of the neutron spectra is fully absorbed below 1 eV and partially cut until 0.01 MeV, while hardly affecting the high-energy flux. Once the neutron test-position was calibrated, I tested and studied the upset and latch-up sensitivities of different SRAM memories to thermal neutrons and high-energy hadron. This investigation was made with differential measurements using the boron carbide box as a thermal neutrons absorber. One of the tested components was the ESA SEU Monitor, an SRAM-based radiation detector employed to prove the effectiveness of the differential approach and to assess the beam spatial uniformity. Moreover, to benchmark the cross sections results retrieved on the neutron-dominated position at CHARM, I tested the same memories in two facilities in Grenoble (France): a 14 MeV mono-energetic neutrons source and a reactor providing a thermal neutron spectrum for testing electronics. Since 14 MeV neutron beams are typically more accessible and cost-efficient than several hundred MeV protons at cyclotron facilities, this work also evaluates their possible use for deriving the saturation cross section, representative of the high-energy hadron response of candidate components. Furthermore I worked on an Americium-Beryllium neutron source at CERN to assess its possible use for SEU testing when CHARM is not available. After the calibration of the facility carried out with FLUKA simulations and ESA Monitor experimental measurements, I proved its potential employment for future tests. Finally, I compared the mono-energetic neutron beam test approach results against those at CHARM, obtaining a highly satisfactory agreement between the memories cross sections measured in these facilities. In this way, CHARM is shown to successfully reproduce the conditions to obtain an SEE qualification compatible with that at standard mono-energetic facilities.