Parameterizations for shielding electron accelerators based on Monte Carlo studies

Numerous recipes for designing lateral slab neutron shielding for electron accelerators are available and each generally produces rather similar results for shield thicknesses of about 2 m of concrete and for electron beams with energy in the 1 to 10 GeV region. For thinner or much thicker shielding the results tend to diverge and the standard recipes require modification. Likewise for geometries other than lateral to the beam direction further corrections are required so that calculated results are less reliable and hence additional and costly conservatism is needed. With the adoption of Monte Carlo (MC) methods of transporting particles a much more powerful way of calculating radiation dose rates outside shielding becomes available. This method is not constrained by geometry, although deep penetration problems need special statistical treatment, and is an excellent approach to solving any radiation transport problem providing the method has been properly checked against measurements and is free from the well known errors common to such computer methods. This present paper utilizes the results of MC calculations based on a nuclear fragmentation model named DINREG using the MC transport code GEANT and models them with the normal two parameter shielding expressions. Because the parameters can change with electron beam energy, angle to the electron beam direction and target material, the parameters are expressed as functions of some of these variables to provide a universal equations for shielding electron beams which can used rather simply for deep penetration problems in simple geometry without the time consuming computations needed in the original MC programs. A particular problem with using simple parameterizations based on the uncollided flux is that approximations based on spherical geometry might not apply to the more common cylindrical cases used for accelerator shielding. This source of error has been discussed at length by Stevenson and others. To study this source of error, a run in both spherical and cylindrical geometry were compared