Processing of JEFF nuclear data libraries for the SCALE Code System and testing with criticality benchmark experiments

. In the last years, a new version of the Joint Evaluated Fission and Fusion File (JEFF) data library, namely JEFF-3.3, has been released with relevant updates in the neutron reaction, thermal neutron scattering and covariance sub-libraries. In the frame of the EU H2020 SANDA project, several e ﬀ orts have been made to enable the use of JEFF nuclear data libraries with the extensively tested and veriﬁed SCALE Code System. With this purpose, AMPX processing code has been applied to enable such application, allowing to provide insight into the interaction between the code and the new versions of JEFF data ﬁle. This paper provides an overview about the processing of JEFF-3.3 nuclear data library with AMPX for its application within the SCALE package. The AMPX-formatted cross-section library has been widely veriﬁed and tested using a comprehensive set of criticality benchmarks from ICSBEP, by comparing both with results provided by other processing and neutron transport codes and experimental. Processing of JEFF-3.3 covariances is also addressed along with their corresponding veriﬁcation using covariances processed with NJOY. This work paves the way towards a successful future interaction between JEFF libraries and SCALE.


Introduction
Nuclear data processing is the procedure devoted to the conversion of evaluated nuclear data into libraries for specific final applications such as neutron transport or inventory calculations. Computational codes are specifically dedicated to nuclear data processing. AMPX [1] is the modular processing code of SCALE Code System that takes basic cross section data in Evaluated Nuclear Data File (ENDF) format to provide both multigroup (MG) or continuous energy (CE) libraries for their use by the neutron transport codes included within SCALE [2].
The OECD Nuclear Energy Agency (NEA) Data Bank coordinates the Joint Evaluated Fission and Fusion (JEFF) nuclear data library project. In the last years, a new version of the JEFF library, namely JEFF-3.3 [3], has been released with relevant updates in the neutron reaction and the thermal neutron scattering sub-libraries. The library is publicly released through the NEA in ENDF-6 format [4]. Thus, users should perform the nuclear data processing to produce a nuclear data set in an adequate format for the final application. Therefore, the use of JEFF nuclear data libraries within SCALE system is not straightforward so that the processing of the nuclear data library must be undertaken with AMPX.
Past efforts set the first milestones for the usage of JEFF libraries within SCALE [5]. As a continuation of that work, within the EU H2020 SANDA (Supplying Accurate Nuclear Data for energy and non-energy Applica- * e-mail: antonio.jcarrascosa@upm.es tions) project, AMPX is being used for processing the JEFF-3.3 neutron library.
This paper deals with the processing of JEFF-3.3 neutron data libraries into a CE library for its use with SCALE transport codes such as KENO-VI. Main aspects concerning the processing of CE and covariance libraries with AMPX are depicted. The CE library performance is also evaluated for a set of criticality benchmarks. This allows to identify the application domain of the generated library and those issues that require further development activities.

Processing with AMPX
AMPX is the modular processing code of SCALE Code System, developed at Oak Ridge National Laboratory (ORNL). In this work, the CE library is generated using AMPX code available with SCALE6.3β11. This version incorporates relevant updates regarding the generation of probability tables for the unresolved resonance region (URR), affecting to intermediate and fast spectrum systems [6].
This section presents a brief summary about the processing of both CE libraries and covariance matrices.

Continuous energy libraries
The generation of a CE library with AMPX is performed through a multi-step procedure based on the usage of different modules. For each available isotope and starting from the ENDF-6 file, POLIDENT is firstly used to reconstruct point-wise CE cross sections at 0 K, with a default reconstruction tolerance of 0.1%. Then, BROADEN performs the Doppler-broadening for those temperatures required by the user. The first stage is completed by TGEL, which ensures the consistency between partial and total reactions.
Separately, PURM generates probability tables for the URR (if present), setting the number of probability bins to 20. Y12 is applied to generate two-dimensional kinematics data for neutron scattering producing doubledifferential data and JAMAICAN converts the data into marginal probability distribution in exit energy. Finally, PLATINUM creates the final CE library by merging the data produced in previous steps.
Y12 and JAMAICAN are also used for processing the Thermal Scattering Libraries (TSL), combining the thermal moderator data with the proper evaluation in the higher energy range (>10 eV).

Covariance libraries
This work also deals with the processing of JEFF-3.3 covariance libraries. AMPX is applied to generate COVERX-format covariances for average number of neutrons per fission (MF31), resonance parameters (MF32) and neutron cross-sections (MF33) and prompt fission spectrum (MF35).
PUFF is the module devoted to generating covariance libraries according to the group-averaged cross section data on the user-defined energy structure. Files produced by PUFF are then merged into a library that contains cross-reaction and cross-material covariance matrices (if present). Corrections are finally applied to the library by means of the COGNAC module.
In this work, two sets of covariance libraries are generated using a weighting function generically optimized for fast reactor analyses. The first covariance library is created for general purposes using a 33-energy group structure. Moreover, a 7-group structure covariance matrix is also created in the frame of the OECD/NEA WPEC Subgroup-46 [7].

Processing JEFF-3.cross section library
The latest release of the JEFF project, JEFF-3.3, is a thorough update of the neutron, decay data, fission yields, dpa and neutron activation libraries with TSLs for 20 compounds. It also includes new evaluations for the major nuclides 235 U, 238 U and 239 Pu along with important updates for many other isotopes in terms of neutron cross sections. It is worth mentioning that JEFF-3.3 improvements targeted the needs for advanced reactors developments programs, including upgrades for both sodium and lead. This work addresses the processing of JEFF-3.3 neutron data for 562 isotopes and 20 TSLs. The procedure detailed in Sect. 2.1 is successfully applied for the processing of all the isotopes. Nonetheless, several issues are identified through the processing itself and via the infinite dilution testing phase. The latter consists of very simple criticality calculations of an infinite dilution containing fissile material ( 238 U) in a solution, allowing to test all the nuclides included in the library.
The main outcomes of the testing phase are detailed below: • Negative cross section values are found when reconstructing cross section from resonance parameters for 19 isotopes.
• The total resonance width given in the file differs from calculated for the same set of isotopes.
• The lower limit of the URR does not include some unresolved resonance parameters for 239 U.
• Regarding TSL, data for 1 H in CaH 2 , Ca in CaH 2 and Mg in Mg metal are not considered since they are not identified in SCALE.
The infinite dilution calculation also reveals relevant issues concerning TSLs since it is observed that SCALE is not currently able to manage the following compounds: 16 O in Al 2 O 3 and 16 O in D 2 O. This is due to the lack of available material identification for them. However, this can be solved in future iterations using the COMPOZ module to update the standard composition library. Additionally, certain metastable isotopes are not manageable by SCALE so that they do not pass this phase: 106m Ag, 62m Co, 152m Eu, 94m Nb, and 135m Xe.
The rest of the isotopes are successfully processed and can be used in transport calculations. The performance of the library as a whole is evaluated in the next section.

Benchmarking
In order to test the new cross section library, a comprehensive set of experiments from the International Handbook of Evaluated Criticality Safety Benchmark Experiments (ICSBEP) [8] has been selected and evaluated. This analysis includes a comparison between KENO-VI and MCNP6.1, that use AMPX and NJOY-processed cross section data respectively. It is worth mentioning that both processing routes rely on the same set of parameters. Results are presented in terms of C/E since experimental values are also considered and provided insight into the behaviour of the library itself.
A set of ICSBEP benchmarks is created aiming to cover a variety of fuel, moderators, reflectors, spectra and geometries. This set consists of 120 benchmarks, divided into 43 highly enriched uranium (HEU) cases, 10 intermediate-enriched uranium (IEU) cases, 14 lowenriched uranium (LEU) cases, 8 mixed uranium and plutonium (MIX) cases, 28 plutonium (PU) cases and 17 233 U systems (U233). Of these, 71 corresponds to fast neutron spectra (FAST), 6 as for intermediate spectrum (INTER) and 43 for thermal spectrum (THERM). This set is mostly composed by cases included within ICSBEP database along with updated inputs provided by OECD/NEA, ensuring the consistency between KENO-VI and MCNP inputs. The latter has been widely used in previous works [9].
Multiplication factor calculations for the set of benchmarks using both KENO-VI and MCNP are presented from Fig. 1 to Fig. 4, ensuring a 1σ statistical error lower than 10 pcm.
For HEU category (Fig. 1) a good agreement can be observed between the values provided by KENO-VI and MCNP. Nonetheless, a dramatic deviation of around 900 pcm appears for the HMF009-001 benchmark. Further analyses reveals that this difference can be explained by the presence of 9 Be. In fact, this behaviour is systematically found in subsequent cases and this also affects to 9  This issue lies in the description of the (n,2n) reaction in the 9 Be ENDF-6 file provided by JEFF-3.3 evaluation. It is described by means of its partial reactions (i.e., MT875+ reaction channels), but the total reaction is not included. AMPX properly deals with these channels but an additional patch should be included to construct the (n,2n) description. This issue also affects to HCI003-007, for which a deviation of around 300 pcm is found.
The rest of cases presents discrepancies below 100 pcm except for HMF-003-009, HMF-003-011, HMF-011-001, HMI-006-003 and HMI-006-004. Deviations between 100 and 200 pcm are found for these cases suggesting that models shall be reviewed and updated. Fig. 2 depicts results for IEU, LEU and MIX benchmarks. A very good agreement is obtained for IEU benchmarks since differences are lower than 30 pcm in each case. Nonetheless, only benchmarks with fast spectra are included in this case so that more configurations with different physical forms and spectra may be added to the study for a wider comparison.
Regarding LEU category, results are consistent between both codes considering that deviations are not larger than 60 pcm in all cases. This is also observed for MIX benchmarks, where a very good agreement is also obtained (deviations <60 pcm). However, it is worth mentioning that both MCT002-001 and -002 present differences of around 100 pcm because simplified models are used in SCALE while MCNP results are obtained with detailed models. Results for PU benchmarks are presented in Fig 3. In general, both codes predict reasonably similar multiplication factors. The effect of the presence of 9 Be is again observed for PMF018-001, PMF019-001 and PMF021-001, showing differences larger than 2000 pcm. Apart from that, PMF005-001 presents deviations of around 150 pcm, even after updating the KENO-VI model. This benchmark may suggest that additional verification exercises are recommended for W isotopes.
This test is performed based on infinite dilution cases along with verification calculations for PMF-005-001. Firstly, infinite dilution tests show remarkable discrepancies between KENO-VI and MCNP for several W isotopes: 182 W, 184 W and 186 W . Concerning PMF-005-001, deviation between both codes is initially around 150 pcm but it is reduced up to 30 pcm when these nuclides are removed from the calculations. This behaviour is also confirmed for other cases such as HMF-003-009, that also contains these isotopes. Thus, further analyses are mandatory to solve this issue. Finally, U233 cases (Fig. 4) are considered covering a wide range of physical forms. Deviations are consistent between KENO-VI and MCNP besides the fact that of 9 Be is involved in both UMF005-001 and -002. On the other hand, UMF004-001 and -002, for which differences of around 200 pcm are found between both codes, are also affected by W isotopes.  In general, the AMPX-formatted JEFF-3.3 library shows a reasonably good performance. Results provided in this work are accompanied by extensive verification and validation activities carried out in the frame of the JEFF project. This library has been used to assess temperature trends observed for the IRPhEP KRITZ (KRITZ-LWR-RESR-004) benchmarks [10]. This allowed to test the library at room and elevated temperatures, showing a good performance compared to benchmark results.
Additionally, reactor physics calculations have been also performed for the SEFOR fast reactor, evaluating the associated Doppler reactivity effect [11].
Benchmarking activities successfully support the performance of the processed JEFF-3.3 library, establishing a reference processing route for future releases. In fact, preliminary recent works have been carried out for the JEFF-4T1 testing library [12], paving the way towards an optimized interaction between the future JEFF-4 library and AMPX processing code.

Processing and testing COVERX JEFF-3.3 covariances
As aforementioned, JEFF-3.3 neutron library contains 562 isotopes, of which 447 include covariances so that the procedure detailed in Sect. 2.2 is applied for all of them. Two different energy group structures are used for collapsing both cross sections and covariances in PUFF. For verification purposes, the processing is also carried out with NJOY, specifically using the ERRORR module, allowing to check the consistency between both codes. In general, AMPX properly deals with JEFF-3.3 covariances and a brief comparison for relevant isotopesreactions is presented in this work. Fig. 5 and 6 shows the relative standard deviations for 239 Pu (n,f) and 238 U (n,γ), respectively. They include results using both AMPX and NJOY and for the mentioned energy group structures. It can be seen a very good agreement between both codes in all cases and also for the 238 U (n,γ) correlation matrix (Fig. 7). NJOY BOXER -33g AMPX COVERX -33g NJOY BOXER -7g AMPX COVERX -7g Figure 6. Relative standard deviation for 238 U (n,γ) collapsed into 7-and 33-energy groups using AMPX and NJOY.
This agreement can be observed for the rest of relevant isotopes/reactions but a more extensive analysis would be worthwhile following the methodology proposed in [13]. This would allow to test in a more comprehensive way. Nonetheless, the COVERX JEFF-3.3 covariance library has been already tested in terms of uncertainty propagation and compared to NJOY-based covariances [14]. A very good agreement was observed between these two different methodologies, ensuring the adequate performance of the covariance matrix.

Conclusions and future work
JEFF-3.3 neutron cross section library, including TSLs, has been successfully processed with AMPX using the most recent SCALE release. The CE library has been created for its use with SCALE neutron transport tools.
This work is a step further on the interaction between AMPX and JEFF libraries, identifying remaining issues towards a more efficient procedure. During the testing and  benchmarking phases, relevant improvements have been highlighted concerning the treatment of 9 Be. On the other hand, several W isotopes require a more specific analysis in order to adjust potential inconsistencies. Nonetheless, the CE library performs adequately for the set of 120 benchmarks and according to all the verification activities already carried out. Associated covariances have been also generated using different energy group structures and verified against NJOY-processed files.
Finally, it is worth mentioning that these libraries (along with JEFF-3.1.1 CE library) have been recently released through the NEA/CPS aiming to expand the user's group of JEFF nuclear data libraries in the frame of the widely used SCALE Code System. This will also contribute to a more extensive verification and validation capabilities of the current JEFF-3.3, targeting future releases: JEFF-4.