Speaker
Description
Monte Carlo (MC) simulations are a useful tool in benchmarking a range of calibration services, from confirming ionization chamber calibration factors to aiding the development of novel calibration systems. Different MC programs can be more tuned towards different applications and user populations. However, due to the difference in the underlying physics and transport mechanisms, different MC programs can produce different results. Modelling a similar geometry in various MC programs can increase confidence in the results and encourage their utility.
Here, the Vinten 671 ionization chamber (VIC) was modelled using four MC programs: EGSnrc, EGS++, Penelope, and TOPAS. VICs with well-characterized response relationships are deployed at several national measurement institutes (NMIs). Modelling the VIC can serve as a benchmarking tool for MC programs to determine the validity of the simulation output in the energy range of the radioisotope studied. Calibration coefficients for many radionuclides have been measured, modeled, and compared between institutes.
VIC models were constructed based on a CT scan of the VIC at the National Institute of Standards and Technology (NIST) and parameters in the literature [1]. Radioisotopes were modeled as distributed aqueous sources in 5 mL borosilicate flame-sealed ampoules; 12 radionuclides with various decay emissions were assessed as well as 14 monoenergetic photon sources. To compare the response of the nuclides, the energy deposited in the simulated VIC nitrogen gas volume was observed and used to calculate the calibration coefficient. These values were compared to experimental values from the literature and the percent error was assessed. The error per nuclide was added in quadrature (the 'sum of square errors', SSE) as a comparison between models.
Of the nuclides assessed, 10 agreed with published values within 2%; 5 agreed within 1%. The largest disagreement was observed for Pb-212, which had a 1-4% error for all models. The best agreement was observed for Ba-133, which had a 0.05-0.2% error for all models. Overall, all models were within an SSE of 5%. Over the photon energy range from 0.2 to 2.02 MV, the relative response of all models agreed from 0.2 to 5%.
This study provides the geometry and decay data used so that others may develop and benchmark their own implementations of these VIC models.
[1] Townson et al., 2018. ARI 134.
