Study of Soft Errors in Dynamic Random Access Memories for Neutron Dosimetry

Binary information on a dynamic random access memory (DRAM) can be changed when a charged particle passes by without damaging the device itself (Soft Error). Soft error rates (SER) have been studied in 256K DRAMs employing alpha particles from *“!Am, and thermal neutrons through the action of alpha particles and tritons, generated by the (n,Q) reaction in °LiF converters. The energy dependence of SER has been studied by degrading 5.5 MeV alpha particles after passage through melinex films. The dependence of SER on the operating bias of DRAM, having been studied, a reduced value of +2.8 V has been used throughout the experimental work, leading to higher sensitivity to soft error. For the study of thermal neutron sensitivity a paraffin howitzer was used (*’Cf and Am-Be neutron sources). After optimising the thickness of ‘LiF converters for SER production, a calibration curve of soft errors against dose equivalent has been obtained. A dose equivalent of 25 wSvh"? is equivalent to 1 error per hour. It is concluded that a higher density (1M/4M) DRAM has the potential of measuring a dose as low as 1 uSvh"’. Experimental soft errors have been modelled by considering charged particle environment, path length distributions of the charged particles through the DRAM sensitive volume and the critical charge of the device. For the charged particle environment, its spectral distribution has been simulated by a Monte Carlo method. SER’s have then been computed for sensitive volumes in the shape of rectangular parallelepipeds (RPP) and hemispheres (HEM), integral path length distributions in these having been computed first. In general, the integral distributions for an RPP lies below that of HEM(i.e.lower SER for RPP), except for very large chord lengths. In order to match the experimental and simulated SER’s for RPP, threshold energy ranges have been estimated for collection depths 1,2,3 um, these being 450-950, 750-1200, and 925-1325 keV, respectively. For HEM’s, the ranges are 675-1050, 825-1225, 950-1350 keV. Further, the matching exercise has led us to conclude that tritons make no contribution to soft error production in the present study. It is also proposed that a DRAM sensor could replace the proportional/Scintillation counter in a Bonner sphere, and that by combining a single moderating assembly with a number of sensors, an automatic semi-portable survey/ personnel neutron monitor could be assembled.