MERCURY vs. TART Comparisons to Verify Thermal Scattering

Recently the results from many Monte Carlo codes were compared for a series of theoretical pin-cells; the results are documented in ref. [3]; details are also provided here in Appendix A and B. The purpose of this earlier code comparison was primarily to determine how accurately our codes model both bound and free atom neutron thermal scattering. Prior to this study many people assumed that our Monte Carlo transport codes were all now so accurate that they would all produce more or less the same answers, say for example K-eff to within 0.1%. The results demonstrated that in reality we see a rather large spread in the results for even simple scalar parameters, such as K-eff, where we found differences in excess of 2%, far exceeding many people's expectations. The differences between code results were traced to four major factors, (1) Differences between the sets of nuclear data used. (2) The accuracy of nuclear data processing codes. (3) The accuracy of the models used in our Monte Carlo transport codes. (4) Code user selected input options. Naturally at Livermore we would like to insure that we minimize the effects of these factors. In this report we compare the results using two of our Monte Carlo transport codes: MERCURY [2] and TART [2], with the following constraints designed to address the four points listed above, (1) Both codes used exactly the same nuclear data, namely the TART 2005 data. (2) Each code used its own nuclear data processing code. Even though these two data processing codes are independent, they have been extensively tested to insure the processed output results closely agree. (3) Both used the same nuclear physics models. This required that some physics be turned off in each code, namely, (a) Unresolved resonance energy region self-shielding was turned off in TART, since this is not currently available in MERCURY. (b) Delayed neutrons were treated as prompt in TART, since this is not currently available in MERCURY. (c) Classical, rather than relativistic, kinematics were used in MERCURY, since relativistic kinematics is not currently available in TART. (4) Both codes used the same input options; both modeled the geometry and materials in exactly the same manner; both used the same number of neutron histories to produce results to within comparable statistical accuracy. These constraints were used to test the accuracy of both our nuclear data processing codes and our Monte Carlo transport codes, while avoiding differences due to the use of different sets of nuclear data, since both of our codes used the same TART 2005 data.