ACE File Verification Project - Stage 1 Results

IAEA Coordinator: A.Trkov

Scope

Stage-1 is intended to test the capability of different data processing codes to produce library files in ACE format that reflect the cross sections in the original ENDF file and that run with the MCNP Monte Carlo code and produce equivalent results.

Source data files at this stage are the U-235 and U-238 evaluations from the ENDF/B-VIII.b4 library. This version of the files was chosen because they contain up-to-date cross sections, but without format extensions of the final ENDF/B-VIII.0 library, which were not processable even with the NJOY code at the start of the project.


Participants

NJOY

The NJOY data processing system from the Los Alamos National Laboratory (LANL) was for a long time the only code publically available for generating files in ACE format for continuous-energy Monte Carlo calculations.

Cross sections of U-238 after resonance reconstruction and Doppler broadening processed with NJOY agree with the equivalent PENDF generated with PREPRO, as seen from the cross section plots. Small differences of the order up to 0.3% can be seen in the capture cross section. Only the comparison for U-238 is shown, since this seems to be more restrictive in terms of code performance. The results for U-235 are very similar.

Cross sections of U-238 extracted from the ACE file generated with NJOY agree with the PENDF generated with PREPRO, as seen from the cross section plots.

The scattering in the capture cross section in the resonance range is consistent with the observed differences in the PENDF files processed by PREPRO and NJOY. It is believed that this discrepancy in the reconstructed cross sections has no effect on the results of transport calculations.

Elastic angular distributions are compared at energies 40 keV, 70 keV, 100 keV and 1 MeV, where most of the differences in the scattering moments were observed. The comparison is shown in the angular cross section plots. Differences are observed in the data retrieved from the ACE file, particularly at the lowest energies. Anisotropy at these energies is still relatively small and a 0.4% difference is probably unimportant, but there is also a difference of almost 1% at forward angles at 1 MeV.

Double differential cross sections are compared at the incident neutron energy of 2 MeV, which lies between two panels in the source ENDF files and requires interpolation in incident energy. A forward angle of 20 degrees and a backward angle of 160 degrees are considered. The reactions included in the comparison are: fission, first discrete inelastic level, continuum inelastic and the total neutron emission. The comparison is shown in the double-differential cross section plots. No significant differences are observed in the data retrieved from the ACE file.


LEGEND/NJOY

The case is essentially a standard NJOY except that angular distributions in the source ENDF file are converted into tabular form with the LEGEND module of PREPRO. The main purpose of the exercise is to check the magnitude of the impact of the inprecise angular distributions entered into the ACE file by NJOY.

The cross sections and the double-differential cross sections are identical to those of the pure NJOY case, so the plots are not given explicitly. Likewise, the angular distributions are the same as in the reference PENDF file, so the plots are not given explicitly.


PREPRO/ACEMAKER

The ACEMAKER code is a module developed by D.L. Aldama, Cuba, that starts from the files processed with PREPRO to assemble the ACE file.

The comparison of the cross sections and elastic scattering moments is shown in the cross section plots.

The comparison of angular distributions is shown in the angular cross section plots. No differences are observed in the data retrieved from the ACE file compared to PREPRO, since they are processed in exactly the same way.

A comparison of the double-differential cross sections is shown in the double-differential cross section plots. No significant differences are observed in the data retrieved from the ACE file.


FRENDY

The FRENDY data processing code package is under development at JAEA in Japan. It is complete as far as the current scope of testing is concerned.

The comparison of the cross sections and elastic scattering moments is shown in the cross section plots.

The comparison of angular distributions is shown in the angular cross section plots. Differences are observed in the data retrieved from the ACE file, particularly in the forward angles at 100 keV, where it exceeds 2%. A corrective action might be needed.

A comparison of the double-differential cross sections is shown in the double-differential cross section plots. No significant differences are observed in the data retrieved from the ACE file.


FUDGE

The FUDGE code is an open source code maintained by the Lawrence Livermore National Laboratory. It is the primary code with the capability to process the evaluated data in the GNDS format.

The comparison of the cross sections and elastic scattering moments is shown in the cross section plots. A small problem is seen in the Doppler-broadening of non-threshold reactions at very low energies. No other significant differences are observed in the data retrieved from the ACE file.

The comparison of angular distributions is shown in the angular cross section plots. Differences are observed in the data retrieved from the ACE file, particularly in the forward angles at 100 keV, where it exceeds 2%. A corrective action might be needed.

A comparison of the double-differential cross sections is shown in the double-differential cross section plots. The difference in the neutron emission cross section from the inelastic continuum reaction arose in the preliminary submission of the ACE file generated by FUDGE because the interpolation flag between the incident energy panels were not transferred into the ACE file. This inconsistency was already corrected in the FUDGE code.


GAIA/NJOY

The GAIA code system is under development at IRSN, France. At present only the modules for the resonance reconstruction and Doppler broadening are applicable for the exercise. The final file assembly is done with the NJOY code.

The originally submitted ACE files did not perform very well. The problem was traced to a large number of points added by GAIA below threshold after Doppler broadening. Although the effective Doppler-broadened cross sections can be non-zero below the nominal threshold defined by the Q-value due to the relative motion of the target nucleus, the range given by GAIA extended too far below threshold. Correcting the PENDF file of GAIA with the FIXUP module of PREPRO-2015 (threshold correction is not available in PREPRO-2017 and later), the re-assembled ACE file performance was restored.

The comparison of the cross sections and elastic scattering moments is shown in the cross section plots.

A comparison of the double-differential cross sections is not applicable, since the ACE file assembly is done with the NJOY code.


GALILEE/NJOY

The GALILEE code system is under development at CEA, France. At present only the modules for the resonance reconstruction and Doppler broadening are applicable for the exercise. The final file assembly is done with the NJOY code.

The comparison of the cross sections and elastic scattering moments is shown in the cross section plots.

A comparison of the double-differential cross sections is not applicable, since the ACE file assembly is done with the NJOY code.


GRUCON

The GRUCON data processing system has been developed at IPPE, Russian federation. It is complete and comparable in scope to NJOY. The GRUCON package is available from the IAEA.

The comparison of the cross sections and elastic scattering moments is shown in the cross section plots.

The comparison of angular distributions is shown in the angular cross section plots. The angular cross sections are in perfect agreement with PREPRO. The apparent difference in the first interval might be an artefact of the plotting procedure.

A comparison of the double-differential cross sections is shown in the double-differential cross section plots. No significant differences are observed in the data retrieved from the ACE file.


RULER

The RULER data processing code package is under development at the CNDC in China. It is complete as far as the current scope of testing is concerned.

The comparison of the cross sections and elastic scattering moments is shown in the cross section plots.

The comparison of angular distributions is shown in the angular cross section plots. Differences are observed in the data retrieved from the ACE file, which are similar to those observed with NJOY processing.

A comparison of the double-differential cross sections is shown in the double-differential cross section plots. No significant differences are observed in the data retrieved from the ACE file.


NECP-Atlas

The NECP-Atlas is being developed at the Xi'an Jiaotong University with Prof. Tiejun Zu as the primary contact person. Further details are available on the code web site.

The comparison of the cross sections and elastic scattering moments is shown in the cross section plots. The cross sections are in perfect agreement with the reference. Some differences are seen in the elastic scattering Legendre coefficients in the Lab coordinate system, reconstructed from the angular distributions when the anisotropy is small. These originate from insufficient reconstruction tolerance when tabulating angular distributions from Legendre coefficients (see below). Physically, this is not expected to impact transport calculations, since it is a relative difference when the anisotropy is still small.

The comparison of angular distributions is shown in the angular cross section plots. Differences are observed in the data retrieved from the ACE file, which are similar to those observed with NJOY processing.

A comparison of the double-differential cross sections is shown in the double-differential cross section plots. No significant differences are observed in the data retrieved from the ACE file.


Results of criticality calculations

Since this is a purely numerical exercise without taking self-shielding into account, comparison to the benchmark multiplication factor is not applicable. Arbitrarily, the NJOY results were taken as reference. The plot of the differences in k_eff shows that most of the differences lie within 20 pcm, except for a few isolated cases.


NOTE: This is work in progress. The results are preliminary and the authors may submit improved files that could reduce the observed differences.