Calculation of Upper Subcritical Limits for Nuclear Criticality in a Repository

The purpose of this document is to present the methodology to be used for development of the Subcritical Limit (SL) for post closure conditions for the Yucca Mountain repository. The SL is a value based on a set of benchmark criticality multiplier, keff results that are outputs of the MCNP calculation method. This SL accounts for calculational biases and associated uncertainties resulting from the use of MCNP as the method of assessing keff. The context for an SL estimate include the range of applicability (based on the set of MCNP results) and the type of SL required for the application at hand. This document will include illustrative calculations for each of three approaches. The data sets used for the example calculations are identified in Section 5.1. These represent three waste categories, and SLs for each of these sets of experiments will be computed in this document. Future MCNP data sets will be analyzed using the methods discussed here. The treatment of the biases evaluated on sets of keff results via MCNP is statistical in nature. This document does not address additional non-statistical contributions to the bias margin, acknowledging that regulatory requirements may impose additional administrative penalties. Potentially, there are other biases or margins that should be accounted for when assessing criticality (keff). Only aspects of the bias as determined using the stated assumptions and benchmark critical data sets will be included in the methods and sample calculations in this document. The set of benchmark experiments used in the validation of the computational system should be representative of the composition, configuration, and nuclear characteristics for the application at hand. In this work, a range of critical experiments will be the basis of establishing the SL for three categories of waste types that will be in the repository. The ultimate purpose of this document is to present methods that will effectively characterize the MCNP computations with respect to bias, as applicable to the repository setting. Combining varied sets of critical experiments into a single source of benchmark criticals provides wider ranges of applicability and, potentially, additional variability contribution to the treatment for the uncertainty of the bias. This will allow the estimation of the bias characteristics that will be useful in establishing the SL. If extrapolation is required, there may be need for ad lioc analyses to evaluate the bias characteristics, or at a minimum to recalculate the SLY based on the new range for the trending variable. This may also require extending the data set of critical experiments