Roger L. Martz

Initial MCNP6 Release Overview

MCNP6 is simply and accurately described as the merger of MCNP5 and MCNPX capabilities, but it is much more than the sum of those two computer codes. MCNP6 is the result of five years of effort by the MCNP5 and MCNPX code development teams. These groups of people, residing in Los Alamos National Laboratory’s (LANL) X Computational Physics Division, Monte Carlo Codes Group (XCP-3), and Decision Applications Division, Radiation Transport and Applications Team (D-5), respectively, have combined their code development efforts to produce the next evolution of MCNP. While maintenance and bug fixes will continue for MCNP5 1.60 and MCNPX 2.7.0 for upcoming years, new code development capabilities only will be developed and released in MCNP6. In fact, the initial release of MCNP6 contains 16 new features not previously found in either code. These new features include the abilities to import unstructured mesh geometries from the finite element code Abaqus, to transport photons down to 1.0 eV, to transport electrons down to 10.0 eV, to model complete atomic relaxation emissions, and to generate or read mesh geometries for use with the LANL discrete ordinates code Partisn. The first release of MCNP6, MCNP6 Beta 2, is now available through the Radiation Safety Information Computational Center, and the first production release is expected in calendar year 2012. High confidence in the MCNP6 code is based on its performance with the verification and validation test suites, comparisons to its predecessor codes, the regression test suite, its code development process, and the underlying high-quality nuclear and atomic databases.

MCNP6 UNSTRUCTURED MESH INITIAL VALIDATION AND PERFORMANCE RESULTS

Los Alamos National Laboratory Monte Carlo N-Particle transport code (MCNP) Version 6 (MCNP6) has been extended to include a new capability that permits tracking of neutrons and photons on an unstructured mesh that is embedded as a mesh universe within its constructive solid geometry capability. The mesh geometry is created through Abaqus/CAE using its solid modeling capabilities. MCNP transport results are calculated for mesh elements using a path length estimator while element-to-element tracking is performed on the mesh. The results from MCNP6 can be exported to Abaqus/CAE for visualization or other physics analysis. Three geometrically simple benchmark experiments were analyzed: Godiva, Osaka nickel sphere, and fusion neutron source vanadium cube. Computer run time is proportional to the number of mesh elements, element order, and element type specified in the input. Good agreement of our MCNP6 results with the measured neutron leakage for the nickel sphere and the measured neutron and gamma spectra from the vanadium assembly was observed.

MCNP Variance Reduction Developments in the 21st Century

Improvements incorporated into MCNP variance reduction methodology and code releases since 2000 are discussed. Some of the improvements are modifications or generalizations of older techniques, and some are entirely new. In particular, pulse-height-tally variance reduction is now possible in MCNP, and the dxtran technique has been generalized to allow an arbitrary nesting of dxtran spheres. A new precollision, next-event estimator is discussed along with flux-at-a-point image tallies. Additionally, the event log analyzer is a tool designed to help the user understand what causes the variance in the user’s particular MCNP calculation.

The MCNP6 Book On Unstructured Mesh Geometry: User's Guide for MCNP 6.2.1

Hybrid geometries with unstructured mesh embedded in a constructive solid geometry universe representation is a recently added (less than 10 years old) feature to MCNP6 and was present in its first production release as well as several prior beta versions. This capability requires new input and produces its own output that is available in a special file for post-processing. This user’s guide provides an overview of the capability and then discusses the input, output, and some of the preand post-processing capability. The MCNP Book On Unstructured Mesh Geometry: User’s Guide

Evaluation of the Pool Critical Assembly Benchmark with Explicitly-Modeled Geometry using MCNP6

Abstract Despite being one of the most widely used benchmarks for qualifying light water reactor (LWR) radiation transport methods and data, no benchmark calculation of the Oak Ridge National Laboratory (ORNL) pool critical assembly (PCA) pressure vessel wall benchmark facility (PVWBF) using MCNP6 with explicitly modeled core geometry exists. As such, this paper provides results for such an analysis. First, a criticality calculation is used to construct the fixed source term. Next, ADVANTG-generated variance reduction parameters are used within the final MCNP6 fixed source calculations. These calculations provide unadjusted dosimetry results using three sets of dosimetry reaction cross sections of varying ages (those packaged with MCNP6, from the IRDF-2002 multigroup library, and from the ACE-formatted IRDFF v1.05 library). These results are then compared to two different sets of measured reaction rates. The comparison agrees in an overall sense within 2% and on a specific reaction and dosimetry location basis within 5%. Except for the neptunium dosimetry, the individual foil raw calculation-to-experiment comparisons usually agree within 10% but are typically greater than unity. Finally, in the course of developing these calculations, geometry that has previously not been completely specified is provided herein for the convenience of future analysts.

Evaluation of the Kobayashi Analytical Benchmark Using MCNP6’s Unstructured Mesh Capabilities

Abstract This paper provides results for calculations performed using MCNP6’s unstructured mesh (UM) capabilities based on the three problems described in the Kobayashi benchmark suite. These calculations are performed to provide a comprehensive and consistent basis for the verification and validation of MCNP6’s constructive solid geometry (CSG) and UM neutron transport capabilities relative to a well-known analytic benchmark. First, preexisting MCNP5 CSG models are updated and reexecuted to form a basis of comparison with UM for both the consistency of the numeric results and speed of execution. Next, a series of UM calculations is performed using first- and second-order tetrahedral and hexahedral elements with mesh generated using Abaqus. In addition, a different first-order tetrahedral mesh is generated with Attila4MC in order to investigate the effect on the results. When executed, the results for both CSG and UM agree among themselves and with the benchmark quantities within reasonable statistical fluctuations (at worst, the results agree within 2σ or 10% but generally within 1σ or 5%) and recognizing from historical work that improved agreement is possible with additional variance-reduction effort. As expected, for the simple geometries herein, we find the CSG calculations completing approximately ten times faster than the comparable fastest UM calculations. We find minor speed differences (~1%) between multigroup and continuous-energy nuclear data and significant speed differences (factor ~100) between different element types. As such, the timing results support the recommendation that users run with the simplest UM element type that adequately represents the problem geometry, ideally first-order hexahedra, and with the most convenient nuclear data energy treatment.

Evaluation of Pulsed Sphere Time-of-Flight and Neutron Attenuation Experimental Benchmarks Using MCNP6’s Unstructured Mesh Capabilities

Abstract This paper extends the verification and validation of MCNP6’s unstructured mesh (UM) features for neutron transport capabilities by comparing code and experimental results for two different sets of experiments. The first set of experiments comprises time-of-flight spectrum measurements of spheres pulsed by 14-MeV neutrons performed by Lawrence Livermore National Laboratory in the early 1970s. The second set of experiments comprises spontaneous fission neutron attenuation measurements in relatively simple geometries with varying shield thicknesses performed by Ueki et al. in the early 1990s. First, traditional constructive solid geometry (CSG) models are analyzed to ensure agreement with experimental values and to form a basis of comparison with UM results. For the pulsed sphere experiments, a series of UM calculations is performed using first-order tetrahedral elements with various levels of mesh refinement. For the Ueki experiments, purely CSG, purely UM, and hybrid CSG/UM calculations are performed using first- and second-order tetrahedral and hexahedral elements. In the purely UM cases, two different meshing algorithms are used to specify the first-order tetrahedral mesh. The pulsed sphere calculated and experimental time-of-flight spectra agree with p-values >0.999 when compared using χ2 goodness-of-fit tests. Furthermore, the UM results show discrepancies with the experimental values comparable to the CSG cases. The Ueki neutron attenuation calculated values using track-length and point detector tallies agree with the experimental values within 1σ with a single exception that agrees well within 2σ. As such, we conclude that the results for the CSG and UM calculations agree among themselves and with the experimental quantities when considering the associated statistical uncertainties.

A Notable Comparison of Computational Geometries in MCNP6 Calculations

Abstract MCNP6 has been extended to include a new capability that permits tracking of neutrons and photons on an unstructured mesh (UM) embedded as a mesh universe within its constructive solid geometry capability. Our mesh geometry was created through Abaqus/CAE using its solid modeling capabilities. Monte Carlo transport results are calculated for mesh elements using a path length estimator while particles track from element face to element face on the mesh. This paper presents some performance comparisons for the initialization and calculation phases of two well-known benchmark problems using both the legacy and the UM tracking capabilities. For detailed geometries, UM initialization is always faster. For very detailed geometries where the models are comparable, the UM capability is faster than the legacy geometry capability.

Verification of Unstructured Mesh Capabilities in MCNP6 for Reactor Physics Problems

New unstructured mesh capabilities in MCNP6 (developmental version during summer 2012) show potential for conducting multi-physics analyses by coupling MCNP to a finite element solver such as Abaqus/CAE[2]. Before these new capabilities can be utilized, the ability of MCNP to accurately estimate eigenvalues and pin powers using an unstructured mesh must first be verified. Previous work to verify the unstructured mesh capabilities in MCNP was accomplished using the Godiva sphere [1], and this work attempts to build on that. To accomplish this, a criticality benchmark and a fuel assembly benchmark were used for calculations in MCNP using both the Constructive Solid Geometry (CSG) native to MCNP and the unstructured mesh geometry generated using Abaqus/CAE. The Big Ten criticality benchmark [3] was modeled due to its geometry being similar to that of a reactor fuel pin. The C5G7 3-D Mixed Oxide (MOX) Fuel Assembly Benchmark [4] was modeled to test the unstructured mesh capabilities on a reactor-type problem.