Improved Linear Congruential Random Number Generators

Abstract

The random number generator in MCNP and most other Monte Carlo codes for particle transport (e.g, RCP, RACER, MORSE, KENO, VIM, MCPP) is based on algorithms called linear congruential random number generators(LCGs). However, all LCGs have a defect that cannot be removed by adjusting their parameters. The problem lies in the "crystalline" nature of LCGs. This article presents the improved linear congruential random number generators (ILCGs), which preserve the advantage of rapidity, and the statistical tests indicate that the statistical qualities have been obviously improved.