A major flaw in group constant calculation (July 12, 2009)
A major problem in the calculation of homogenized multi-group constants was discovered in July 2009. The problem originates from the fact that the basic collision flux estimator used with delta-tracking yields a zero value in void regions, which results in the under-estimation of flux integrals. Since multi-group cross sections are calculated by dividing integral reaction rates by the integral flux, the result is a systematic over-prediction of the values.
The problem affects only homogenized multi-group constants when low-density material regions are modeled as void. In most LWR cases these regions are limited to the gas gap between fuel pellet and cladding, and the errors are in the order of a few percent. The problem becomes more severe in CANDU and especially gas-cooled reactor geometries, in which the results may be off by 30%.
The problem was fixed in code update 1.1.3 (July 12, 2009). The reason why this major flaw was not discovered earlier is that the standard set of MCNP validation calculations were carried out using a tallyx subroutine that also discarded flux calculation in void regions. Comparison to deterministic codes was limited to LWR cases, and the few-percent discrepancies were considered originating from the cross section data and the differences in the transport methods.